BIO-PLY was prepared from eight healthy horses (4 geldings and 4 mares between the ages of 6 and 19 years). The Institutional Animal Care and Use Committee of North Carolina State University (16-189-O) approved the use of these horses for BIO-PLY preparation. Whole blood was collected via jugular venipuncture into syringes containing acid citrate dextrose (ACD). Erythrocytes were removed by sedimentation and the layer above the erythrocytes or leukocyte-rich platelet-rich plasma was centrifuged at 250g for 15 minutes to remove leukocytes. The supernatant or platelet-rich plasma (PRP) was then centrifuged at 1500g for 15 minutes to pellet the platelets. The platelet-poor plasma (PPP) or supernatant above the platelet pellet was saved and the platelet pellet was re-suspended in PPP to generate a 50× PRP. The importance of this plasma component for antimicrobial activity has been previously described (Drago et al., 2014; López et al., 2014; Gilbertie et al., 2020). Leukocyte, erythrocytes, and platelet concentrations were determined as previously described (Gilbertie et al., 2020). The 50× PRP contained greater than 1,000,000 platelet/µL, less than 100 WBC/µL and <10 RBC/µL. PRP lysate (PRP-L) was generated from the 50x PRP by five freeze/thaw cycles. The platelet debris was removed from PRP-L by centrifugation at 20,000g for 20 minutes. Removal of anionic components was performed by incubation with a washed, loose anion exchange resin (UNOsphere Q resin, Bio-Rad Laboratories, Hercules, CA, USA) and subsequent separation of unbound components (cationic and neutral) using a bottle-top filter (0.22µm PES, Nalgene). Fractionation by molecular weight was performed with a 10kDa molecular weight cutoff filter (Amicon^® Ultra 15 mL Centrifugal Filters, 10kDa, MilliporeSigma, Burlington, MA). The filtrate containing proteins and peptides <10kDa in size was collected, aliquoted into 5mL aliquots, and stored at -80°C until used in this study. In vitro anti-biofilm efficacy was confirmed prior to use in this study (Gilbertie et al., 2020).

Skeletally mature horses (n=12; 7 mares, 4 geldings, 1 stallion; ages 2-14 years) with normal physical examinations, bloodwork, and tarsocrural radiographs were randomly allocated into treatment (AMK+BIO-PLY) or control (AMK) groups (IACUC protocol #16-194 of NC State University). A graphical abstract of the study design is presented in Figure 1 . Horses were quarantined for 14 days and stall acclimated for 5-7 days before entrance into the study. On day 0, horses underwent standing sedation with intravenous detomidine (0.005-0.01mg/kg) and butorphanol (0.005-0.01 mg/kg) to administer epidural analgesia and perform the tarsocrural joint inoculation. A sacrocaudal epidural injection of buprenorphine (0.005 mg/kg) and detomidine (0.01 mg/kg) was performed using a 20-gauge 3.5-inch spinal needle for hindlimb pain control (Fischer et al., 2009). S. aureus (ATCC 25923) was prepared by diluting a 0.5 McFarland (1x10⁸ CFU/mL) to yield an inoculum of 1x10⁶ CFU/mL in sterile saline. The final inoculum suspension was then serially diluted, and plate counted to verify the bacterial concentration. Each horse received a 1mL S. aureus inoculum (1x10⁶ CFU) via intra-articular injection into one randomly assigned tarsocrural joint using a standard dorsomedial approach and a 21-gauge 1.5-inch hypodermic needle. Prior to withdrawal of the needle from the joint, an additional 1mL of sterile saline was injected into the joint to limit S. aureus extravasation. Starting 24 hours post-infection, horses were treated with BIO-PLY (5mL) and 500mg amikacin (treatment or AMK+BIO-PLY group; n=6) or 500mg of amikacin and sterile saline (5mL) (control or AMK group; n=6) daily for 7 days. All horses received systemic antimicrobials by intravenous administration of potassium penicillin (22,000 U/kg every 6 hours) in combination with gentamicin (6.6 mg/kg every 24 hours) for 10 days post-infection as well as a tapering course of phenylbutazone (4.4 mg/kg every 12 hours for days 1-3, 2.2 mg/kg every 12 hours for days 4-10 and 2.2 mg/kg every 24 hours for the duration of the study). The minimum inhibitory concentration of S. aureus ATCC 25923 for these antimicrobials as measured by antimicrobial susceptibility testing using the Sensititre Complete Automated AST System and the equine (Equine EQUIN1F Vet AST Plate) was measured to be <4 ug/mL for amikacin, ≤ 0.06 ug/mL for penicillin; and ≤1 ug/mL for gentamicin (Gilbertie et al., 2019).

Horses were examined twice daily by a veterinarian for comfort and any clinical signs of systemic infection or sepsis (change in general demeanor, fever, etc.). Horses were evaluated and scored daily for pain with each category scored on a scale of 0-3 (0 = most normal, 3 = most abnormal): lameness, tarsocrural joint swelling, limb edema, pain to palpation of the joint, and heat at the site of infection.

Starting on day 0, whole blood, plasma and serum samples were collected. After day 7, blood was collected weekly until the end of the study. Whole blood and serum were analyzed for changes in white blood cell populations and biochemistry values using the Clinical Pathology Laboratory at NC State University College of Veterinary Medicine. SF was aspirated prior to treatment. Portions of the SF samples collected daily were submitted for analysis of total protein, total nucleated cell count and differential cell count. A portion of that sample was processed as a cytospin as well for further cellular analyses. Aliquots of clarified SF were also saved for biomarker/biochemical analysis at the conclusion of the study. All samples were stored at -80°C until analysis.

Horses had pre-infection radiographs performed followed by end-term radiographs at 21 days post-infection. Ultrasonography was performed at day 0 (pre-infection), day 1 (post-infection, pre-treatment), and days 7, 14, and 21 (post-treatment). Tarsocrural joints were imaged with grayscale using the Aplio 500 system (Canon Medical Systems, CA, USA) with a linear 12 MHz broadband transducer. Transverse and longitudinal grayscale images of the dorsomedial and plantarolateral recesses of the tarsocrural joint ( Supplementary Figure 1 ) were acquired with the limb weight bearing. All images were anonymized, randomized prior to analyses, assessed independently by two blinded radiologists using established criteria for infectious arthritis (Beccati et al., 2015), and scores were averaged. Degree of distension, synovial thickening, fibrinous loculation, and vascularity as visualized with power Doppler were scored on a scale of 0-3 (0 = most normal, 3 = most abnormal). Character of synovial effusion and presence of hyperechoic foci were scored on a scale of 0-1 (0 = anechoic or absent, 1 = echogenic or present). Radiographs were evaluated using a previously published scoring system for osteoarthritis on a scale of 0-4 (0 = normal, 4 = severe change) evaluated on the basis of: bone proliferation at the joint capsule attachment; subchondral bone lysis; subchondral bone sclerosis; and osteophyte formation (Frisbie et al., 2013).

At the conclusion of the study, horses were euthanized with an overdose of pentobarbital sodium administered intravenously following sedation according to AVMA Guidelines for the Euthanasia of Animals: 2020 Edition. The infected tarsocrural joint was aseptically prepared and dissection of the tarsocrural joint was performed aseptically from the dorsomedial aspect. The joint was evaluated for gross morphology and photographed as a reference. Synovial membrane samples were collected from four different sites within the joint (dorsomedial, dorsolateral, plantaromedial, and plantarolateral; Supplementary Figure 1 ) for histological and microbiological analysis. Four osteochondral samples were taken from both the lateral and medial trochlea of the talus from the dorsal and plantar aspects for histological analysis.

Synovial and osteochondral tissues harvested at necropsy were fixed in neutral-buffered 10% formalin, paraffin embedded, microtome sectioned at 5μm thickness and staining for histologic examination using bright field microscopy and scoring using a modified OARSI system (McIlwraith et al., 2010). Osteochondral tissues were demineralized in an aerated 10% formic acid solution prior to trimming into cassettes. Synovial tissue specimens were stained with Hematoxylin & Eosin (H&E) to assess inflammatory lesions and immunohistochemical (IHC) stains including CD3 for T lymphocytes (DAKO), CD20 for B lymphocytes (Thermo Scientific), CD204 for dendritic cells and macrophages (TransGenic Inc.), and MAC387 (i.e., S100A9, MRP-14, calprotectin) for myeloid leukocytes (i.e., neutrophils and monocytes/macrophages) (DAKO) to further characterize inflammatory cell infiltrates. QuPath (Bankhead et al., 2017) was used for quantitative analyses of positive cell counts identified with each IHC marker from two sections representing the synovial superficial intimal layers including adherent exudate (if applicable) and deeper subintimal regions. MAC387+ cells were further evaluated for morphologic features including cell size and nuclear morphology (i.e., polysegmentation) to determine relative numbers of polymorphonuclear granulocytes (i.e., neutrophils) and monocytes/macrophages within synovial sections. Osteochondral sections were stained with H&E and Safranin O (SafO). Assessment and scoring of histology specimens were performed by a blinded observer. MetaMorph^® image analysis software (Molecular Devices Corporation, San Jose, CA, USA) was used for colorimetric quantification to quantify percentage SafO staining within total areas (mm²) of non-mineralized articular cartilage within osteochondral sections.

Historically, it has been difficult to isolate and culture bacteria from the SF of patients with infectious arthritis (Gallo et al., 2008; Taylor et al., 2010). Therefore, we enzymatically digested SF to disperse S. aureus biofilm aggregates and quantify bacterial numbers in infected SF ( Supplementary Figure 2 ). In brief, 10mL of SF was treated with 200µg/mL proteinase K (QIAGEN, Hilden, Germany) for 1 hour before centrifugation at 300g for 10 min to remove host cells. The supernatant containing SF and dispersed bacteria from SF biofilms was then serially diluted and spot plated in triplicate at each serial dilution for colony counts. At end-term, synovial tissue was removed aseptically at necropsy and weighed in grams. Tissue was then gently homogenized and enzymatically digested for 1 hour in media containing 200µg/mL proteinase K and 1.5mg/mL collagenase type II (ThermoFisher Scientific). Tissue homogenates were then centrifuged at 300g for 15 min to remove host cells and debris and the supernatant containing dispersed bacteria was then serially diluted and spot plated in triplicate at each serial dilution for colony counts. A 1mL aliquot of either the SF or tissue homogenate was also enriched in Thioglycolate broth. The lower limit of detection for this method of bacterial enumeration is 16.7 CFU/mL or 3.4 CFU/g. We reported 0 CFU/mL or CFU/g tissue if all three spots had no visible colonies and the enrichment broth yielded no growth.

Venous blood samples were used to determine concentrations of D-Dimer (Shahi et al., 2017) and serum amyloid A (Jacobsen et al., 2006) using the MILLIPLEX MAP Human Cardiovascular Disease (CVD) Magnetic Bead Panel (MilliporeSigma, MA, USA) that the manufacturer predicted to have cross reactivity with equine samples. Concentrations of the predominate inflammatory cytokines found in SF were quantified with the MILLIPLEX MAP Equine Cytokine/Chemokine Magnetic Bead Panel (MilliporeSigma, MA, USA).

All statistical analyses were performed using Stata 14.1MP, StataCorp, College Station TX, with two-sided tests of hypotheses and a p-value < 0.05 as the criterion for statistical significance. Descriptive analyses include computation of means (with 95% confidence intervals [95%CI]), standard deviations, medians, interquartile ranges (IQR) of continuous variables and tabulation of categorical variables. Tests of normal distribution were performed to determine extent of skewness. Frequency counts and percentages were used for summarizing categorical variables (e.g., sex, signalment and others). As part of the exploratory analysis, principal component analysis was performed to examine intrinsic clusters and obvious outliers within the observations. Discriminant analysis was conducted to establish a set of biomarker concentrations based on variable selection to distinguish between groups. All scores (imaging and histology) were compared between control and treatment groups using Wilcoxon rank-sum tests. Continuous data from ultrasonography, biomechanical testing and cytokine analyses was compared between control and treatment groups using t-tests or non-parametric tests based on normality. Inference statistical analysis was based on a generalized linear mixed model. Sex and other covariates were included as fixed effects in the model where indicated. Systemic and SF cell parameters and biomarkers, and bacterial load were analyzed using multivariate statistical methods. Post-hoc pairwise comparisons were used to assess the influence of group.

At day 1 post-infection, all horses had macroscopically visible SF biofilm aggregates within their experimentally infected joints ( Figure 2A ). S. aureus concentrations were measured within their SF ( Figure 2B ) after dispersal of aggregates ( Supplementary Figure 2 ). Over the 7 days of treatment, SF from both AMK and AMK+BIO-PLY treated joints showed a time-dependent decrease in SF bacterial counts. By day 3, AMK+BIO-PLY treated joints had significantly fewer bacteria within the SF (2.52 ± 0.82 log CFU/mL) compared to horses treated with AMK alone (3.82 ± 0.40 log CFU/mL) (p<0.005; Figure 2B ). At end-term, AMK+BIO-PLY treated horses had significantly fewer bacteria within the synovial tissue (2.06 ± 4.86 CFU/g tissue) compared to horses treated with AMK alone (80.95 ± 58.77 CFU/g tissue) (p<0.009; Figure 2C ). Notably, 5 out of 6 AMK+BIO-PLY treated synovial tissues showed complete bacterial clearance (based on the detection limits of our assay) and the 1 out of 6 that did not show complete bacterial clearance showed a drastically reduced bacterial load.

AMK treated limbs showed persistent swelling above and immediately at the level of the fetlock ( Figure 2D ). This swelling was not apparent at day 7 in the AMK+BIO-PLY treated limbs. The pain score, based on lameness, tarsocrural swelling, limb edema, pain to palpation of the joint, and heat at the site of infection decreased for all horses over time ( Figure 2E ). Notably, by day 4, AMK+BIO-PLY treated horses had significantly lower pain scores compared to AMK treated horses (p<0.008).

Using ultrasound to monitor the progression of infectious arthritis (Gaigneux et al., 2017), AMK+BIO-PLY treated horses had an improved ultrasonographic appearance compared to AMK treated horses as evident by changes in joint capsule and synovial thickening as well as the presence of intra-articular fibrin in both the dorsomedial ( Figures 3A, E ). and plantarolateral recesses ( Figures 3B, F ). Over the 21 days, AMK+BIO-PLY treated horses had improved (=lower) ultrasonographic scores compared to AMK horses at days 7 (p<0.007), 14 (p<0.006), and 21 (p<0.02) for the dorsomedial recess ( Figures 3C, G ) and at day 21 (p<0.01) for the plantolateral recess ( Figures 3D, H ). Reduced joint distension ( Supplementary Figure 3A ; p<0.03), reduced synovial thickening ( Supplementary Figure 3B ; p<0.004) and fewer fibrinous loculations within the joint space ( Supplementary Figure 3C ; p<0.03) were evident. No differences in joint effusion ( Supplementary Figure 3D ), intra-articular hyperechoic foci ( Supplementary Figure 3E ) or vascularity ( Supplementary Figure 3F ) were found. Overall, these ultrasonographic changes were consistent with a resolving infection.

No differences in SF total nucleated cell count ( Figure 4A ) or SF total protein content ( Figure 4B ) was found between the groups. The SF percentage of neutrophils ( Figure 4C ) decreased in AMK+BIO-PLY treated horses concomitant with increased percentages of mononuclear cells ( Figure 4D ); these values were significantly different from the AMK group at day 6 and 7 (p<0.02). No differences in systemic leukocyte counts were observed between the two groups throughout the study ( Figure 4E ). Fibrinogen ( Figure 4F ; a protein for blood coagulation) and circulating D-dimer ( Figure 4G ; product of fibrin breakdown) associated with coagulopathy and infection (Shahi et al., 2017; Klim et al., 2018) were decreased with AMK+BIO-PLY treatment compared to the AMK control horses. Serum amyloid A ( Figure 4H ; an acute phase protein) associated with infection and inflammation (Jacobsen et al., 2006) was not different between treatment groups. No alterations in other systemic parameters (i.e., circulating inflammatory cell counts) were observed between treatment groups ( Supplementary Figure 4 ). While many markers of inflammation remained similar between the two groups during treatments, synovial neutrophil and systemic D-dimer content were significantly decreased after treatment with AMK+BIO-PLY.

Infected joints respond to microbial challenge by inducing the production of cytokines and chemokines that activate resident macrophages and serve as neutrophil chemoattractants (Davis, 2005). Cytokine levels increased with the onset of infection ( Figure 5 ) and the overall increase was moderated by AMK+BIO-PLY treatment, where day 7 levels associated with AMK+BIO-PLY treatment approached the initial day 0 levels. This was not achieved with AMK alone. Specifically, IL-1β ( Figure 5A ) and IL-6 ( Figure 5B ), IL-8 ( Figure 5D ), IL-18 ( Figure 5G ), MCP-1 ( Figure 5E ), IL-4 ( Figure 5H ), and IL-5 ( Figure 5I ) initially increased in all joints but were significantly lower in AMK+BIO-PLY-treated joints compared to AMK treated joints by day 7 (p<0.05). IFNγ ( Figure 5L ) had an initial increase at day 2 in AMK+BIO-PLY horses (p<0.003) but decreased by day 3 to the same level as AMK treated horses. AMK+BIO-PLY treatment did not alter SF concentrations of TNF-α ( Figure 5C ), G-CSF ( Figure 5F ), IL-10 ( Figure 5J ), IL-2 ( Figure 5K ), FGF2 ( Supplementary 5A ), eotaxin ( Supplementary Figure 5B ), GM-CSF ( Supplementary Figure 5C ), IL-1α ( Supplementary Figure 5D ), fractalkine ( Supplementary Figure 5E ), IL-13 ( Supplementary Figure 5F ), IL-17A ( Supplementary Figure 5G ), IL-12p70 ( Supplementary Figure 5H ), IP-10 ( Supplementary Figure 5I ), or GRO ( Supplementary Figure 5J ). Taken together, these cytokine profiles suggest an immune-modulating response for AMK+BIO-PLY in the resolution of infection.

The cardinal histologic features of joint infections are synovial inflammation with surface erosions and fibrinosuppurative exudates, subintimal infiltration by neutrophils and mononuclear cells, subintimal neovascularization, and fibrosis (Della Valle et al., 1999). Following euthanasia of horses at day 23-24, a series of postmortem analyses were performed on target tissues from four sites within the tarsocrural joint ( Supplementary Figure 1 ). Joints treated with AMK+BIO-PLY had lower (i.e., improved) total scores for synovial lesions based upon the scoring system for horses recommended by the Osteoarthritis Research Society International (OARSI) (McIlwraith et al., 2010) (p<0.04) ( Figure 6A ). When individual components of the grading scheme were assessed, AMK+BIO-PLY treated joints had significantly lower (i.e., less severe) scores for both vascularity and subintimal fibrosis (p<0.04) ( Figure 6B ). Cellular infiltration and subintimal edema trended lower for AMK+BIO-PLY treated joints but were not statistically significant. Because the OARSI scoring system is defined for horses with non-septic osteoarthritis, tissue sections were evaluated for characteristics of septic joints exemplified by fibrin and inflammatory cells within adherent surface exudates as well as the subintima. AMK treated synovium had more extensive intimal ulceration with thick, adherent, poorly organized mats of fibrin with numerous, predominantly neutrophilic, inflammatory cell pockets. In contrast, AMK+BIO-PLY treated joints had less extensive intimal ulcers with scant, better defined, organizing layers of fibrin that contained few, predominantly mononuclear, inflammatory cells ( Figure 6C , upper and middle panels). AMK treated subintima showed expansion by poorly organized, thick bands of immature fibrovascular tissue with relative increases in inflammatory cells infiltrating deeper layers of collagenous tissue. Within AMK+BIO-PLY treated joints, the subintimal fibrovascular tissue was more organized and compacted beneath a thin regenerating intimal membrane composed of flattened spindle-shaped intimal cells ( Figure 6C lower panels), corresponding to less severe synovial hypertrophy. Thus, by day 23-24, AMK+BIO-PLY treated joints, based on the improved OARSI scores and histology, showed comparatively less severe fibrinosuppurative inflammation and synovial hypertrophy associated with infectious arthritis.

Histochemical and IHC stains were used to analyze synovial inflammation and inflammatory cell populations within superficial intimal and deep subintimal regions of synovial membranes ( Figure 7 ) from four representative locations of the tarsocrural joint ( Supplementary Figure 1 ). AMK treated synovium had higher numbers and percentages of Mac387+ myeloid leukocytes (i.e., neutrophils and monocytes/macrophages) than AMK+BIO-PLY treated synovium ( Figure 7A ). Myeloid leukocytes were concentrated within the superficial intima and adherent fibrinous exudates (p<0.03) but infiltrated all synovial layers. In AMK treated synovium, 90-95% of Mac387+ cells were neutrophils (i.e., ~12μm diameter cells with polysegmented nuclei), whereas in AMK+BIO-PLY treated synovium, nearly all Mac387+ cells were mononuclear (i.e., monocyte/macrophage). The Mac387 IHC results correlated well with differential cell counts and cytologic slide preparations performed on synovial fluid at day 7 ( Figure 7A , insets). Increased numbers and percentage of B lymphocytes (CD20+) were identified within superficial synovial layers of AMK+BIO-PLY treated synovium compared to AMK treated controls (p<0.04), although there was variability among individual horses most pronounced in AMK+BIO-PLY treatment group ( Figure 7B ). Numbers and percentages of CD204+ cells (i.e., dendritic cells and macrophages) trended higher in AMK+BIO-PLY treated synovial samples, although numbers of CD204+ cells identified within exudate, synovial intimal membranes and subintimal stroma varied among horses representing both treatment groups ( Figure 7C ). No differences in numbers, percentage, or distribution of CD3+ T lymphocytes were identified between the AMK and AMK+BIO-PLY treated groups ( Figure 7D ).

Osteochondral explants representing four different locations within the joint were evaluated using the OARSI grading scheme for equine cartilage and subchondral bone (McIlwraith et al., 2010). AMK+BIO-PLY treated osteochondral explants had improved total OARSI scores compared to AMK treated explants ( Figure 8A ; p<0.007). Specifically, when individual components of the grading scheme were compared between groups, scores for focal cell loss (p<0.02) and numbers of chondrocyte clones (i.e., chondrones) (p<0.005) were decreased in AMK+BIO-PLY treated horses as compared to AMK alone ( Figures 8B, C ). Morphometric analyses applied to evaluate the percentage of SafO staining per total area of non-mineralized articular cartilage showed a decreased percentage area of SafO staining (p<0.03) in AMK treated joints compared to AMK+BIO-PLY treated horses ( Figures 8D, E ), indicating that BIO-PLY treatment mitigated inflammation-associated loss of proteoglycan content within the cartilage extracellular matrix. Together these findings suggest that AMK+BIO-PLY mitigates the severity of catabolic tissue response.