A commercially available S. aureus Xen29 (PerkinElmer Inc., Waltham, MA, United States) was used for in vivo cutaneous wound challenge studies (Table 1). S. aureus Xen29 from glycerol stocks stored at –80°C were streaked for isolation onto Brain-Heart Infusion (BHI)(Research Products International Corp., Mt. Prospect, IL, United States) agar plates containing 50 mg/mL of kanamycin monosulfate (Research Products International Corp., Mt. Prospect, IL, United States) and incubated for 16 h at 37°C. Following incubation, plates were imaged using a LAS-1000 luminescent image analyzer (FUJIFILM Corporation) to confirm bioluminescence. One colony was selected to inoculate a flask containing 50 mL of BHI containing 50 mg/mL of kanamycin monosulfate. The culture was incubated at 37°C for 16 h, shaking at 225 RPM. The bacterial suspension was subjected to centrifugation at 2,200 × g for 15 min at room temp. The supernatant was discarded, and the pellet was resuspended in 20 mL of sterile-filtered 1 × PBS, followed by centrifugation at 2,200 × g for 15 min at room temp. The supernatant was again discarded, and the pellet was suspended in 5 mL of sterile-filtered 1 × PBS with 10% glycerol. Serial dilutions of the bacterial suspension were diluted into sterile-filtered PBS with 10% glycerol and the optical density was determined at 600 nm (OD₆₀₀). According to the manufacturer’s instructions, an optical density of 0.5 units is approximately equivalent to 1.4 × 10⁸ CFU/mL.

Bacillus spp. strains were cryopreserved in Tryptic Soy Broth (TSB) containing 20% glycerol. Each strain was grown on Tryptic Soy Agar (TSA) overnight at 30°C to obtain isolated colonies. Bacillus spp. spores used for in vivo challenge were prepared by inoculating an isolated bacterial colony into a 20 mL culture tube containing 5.0 mL of TSB and incubated on a rotatory shaker at 200 RPM at 30°C for 5 days, after which the vegetative cells were killed by heating at 80°C for 20 min, and the CFU/mL were determined by serial dilution and plating onto TSA. To prepare for injection, spores were dispensed into 1.0 mL aliquots with 10% glycerol.

Experimental protocols were reviewed, approved, and performed under regulatory supervision of Auburn University’s Institutional Biosafety Committee (IBC) and Institutional Animal Care and Use Committee (IACUC), under IACUC protocols 2011-2006 and 2016-2988. For this study, 22 female C57BL/6J mice at 6–8 weeks old were anesthetized by 1–3% isoflurane gas and hair was shaven with electronic clippers and depilated using Veet^® (Reckitt Benckiser Group, Berkshire, United Kingdom). Treatment groups (n = 3 mice per Bacillus strain) were injected subcutaneously between the shoulder blades with a 20 mL suspension comprised of 10 mL 1.0 × 10⁸ CFU/mL S. aureus Xen29 and 10 mL of 1 × 10⁸ CFU/mL B. velezensis spores with metabolites (total of 1.0 × 10⁶ CFUs) for strains AB01, AP143, AP183, AP191, and AP218. Internal controls were administered by subcutaneous injection of 1.0 × 10⁶ CFUs of S. aureus Xen29 in a volume of 10 mL on the lower back. Untreated control mice (n = 3) were given subcutaneous injections of 1.0 × 10⁶ CFUs of S. aureus Xen29 between the shoulder blades and the lower back. In a subsequent experiment, a group of mice was administered subcutaneous injections of AP183 spores (1 × 10⁶ CFU/mL) with associated metabolites between the shoulder blades and lower back and another group was injected subcutaneously with metabolites only between the shoulder blades and lower back of each animal. At the end point of the experiment at 6 days post-inoculation, all mice were sacrificed by CO₂ asphyxiation, followed by cervical dislocation, and wound samples were harvested for histological analysis and homogenized for CFU determination.

All in vivo imaging was performed using an IVIS Lumina XRMS (Caliper Life Sciences). Mice were anesthetized by 1–3% isoflurane gas and placed in the imaging chamber of the IVIS Lumina XRMS. Bioluminescence exposures were 5 min with medium binning, f-2 aperture, and a 12.5 cm × 12.5 cm field of view. All mice were imaged once per day for a period of 7 days. At the endpoint, mice were sacrificed according to IACUC protocol guidelines and wound sites were excised and stored in 1 × PBS. Wound samples were split in half, with one half embedded for histological analysis and the other half homogenized for DNA isolation (see DNA isolation section below) to enable 16S rRNA gene amplicon sequencing and serial dilutions were plated to determine final CFU/mg of tissue.

To determine the optimal formulation for administration of AP183 to wound sites, three formulations were prepared and tested in vivo for inhibition of S. aureus proliferation. AP183 spores (1.0 × 10⁶ CFUs), supernatant containing AP183 metabolites without spores, and AP183 spores with supernatant (1.0 × 10⁶ CFUs) were co-administered with 1.0 × 10⁶ CFUs S. aureus Xen29. Mice (n = 3 per group) in a volume of 10 mL were imaged for bioluminescence once a day for 6 days. At the endpoint, mice were sacrificed by CO₂ asphyxiation and cervical dislocation. Wound tissue was resected, homogenized, and diluted followed by plating on TSA and incubated for 18 h at 37°C. S. aureus colonies were counted by bioluminescent imaging to determine CFU counts. Plates were incubated for 18 h at 37°C and bioluminescence imaging was performed on each plate to allow counting of luminescent colonies and ultimately determine CFUs present in each wound. The other half of each wound was prepared for histological Gram and hematoxylin and eosin (H&E) staining.

Wound tissue from two mice from each treatment or control group was embedded in clear frozen section compound (VWR International, West Chester, PA, United States) medium and fixed by submersing in a metal container with 2-methylbutane (Alfa Aesar, Haverhill, MA, United States) on dry ice/liquid nitrogen. Sectioning was performed using a Microm HM 525 (Thermo Fisher Scientific Inc.) with a section thickness of 0.6 mM per slice. Alternating slices were stained with either the following Gram staining or H&E staining procedures. Microscopy was performed using a Zeiss Axioskop with Plan-NEOFLUAR objective lenses and Nikon Sight DS-Fi2 digital camera.

Gram staining of slides was performed by submersion in crystal violet (EMD Millipore, Billerica, MA, United States) for 15 s, followed by gentle rinsing with tap water, submersion in Gram’s Iodine (EMD Millipore, Billerica, MA, United States) for 15 s, tap water rinse, submersion in tartrazine (EMD Millipore, Billerica, MA, United States) solution for 15 s, and a final tap water rinse. Slides rinsed with 95% alcohol and allowed to dry, then dripped with xylene (EMD Millipore, Billerica, MA, United States) and allowed to dry before a coverslip with Cytoseal^TM (Thermo Fisher Scientific Inc.) was placed over each slide.

H&E staining was performed by submersing slides in hematoxylin (EMD Millipore, Billerica, MA, United States) and placed on a rocker for 15 min, then rinsed in tap water with gentle shaking for 15 min. Slides were submersed in 95% alcohol for 30 s and submersed into eosin (VWR International, West Chester, PA, United States) for 45 s. Slides were then submerged into 95% alcohol for 1 min, followed by submersion in 100% alcohol for 1 min. Slides were then submersed in xylene for 1 min and allowed to air dry. A coverslip with Cytoseal^TM (Thermo Fisher Scientific Inc.) was then placed over each slide prior to microscopic examination.

A single colony of B. velezensis AP183 was inoculated into 3 mL of TSB and cultivated overnight at 37°C with shaking at 250 rpm. Small (∼3 mm) sections of a Shiley^TM tracheostomy tube inner cannula (Medtronic, Minneapolis, MN, United States) were sterilized in 70% ethanol and dried prior to placing into a culture tube containing 3 mL of TSB. Then 100 μL of the overnight culture of B. velezensis AP183 was added to the tube and incubated for 72 h at 37°C. The supernatant was then removed, and planktonic cells were removed by washing the inner cannula sections three times in 1 × phosphate-buffered saline. After washing the inner cannula sections, 100 μL of an overnight TSB culture of S. aureus NE1260F (Bose et al., 2013) was added to each tube and incubated at 37°C with shaking at 250 RPM for 24 h. The experiment was conducted in triplicate, and a negative control was included that only had 100 μL of S. aureus culture added to the inner cannula sections, which were incubated at 37°C with shaking at 250 RPM for 24 h. The next day each inner cannula section was washed three times in 1 × phosphate-buffered saline and DNA was extracted from each section (see DNA isolation section below).

In this experiment the competition between B. velezensis AP183 and S. aureus was evaluated when log-phase cultures were co-inoculated in a 96-well plate containing a defined M9 salts minimal medium containing different carbon sources. Two of the carbon sources (glucose and sucrose) could be utilized for growth by both cultures, whereas pectin could only be utilized by B. velezensis AP183 (Hassan et al., 2019). In this experiment a thin film of the carbon source being evaluated (pectin, glucose, or sucrose) was prepared in 96 well plates by evaporative drying from a saturated solution. Each solution was prepared by mixing the respective carbon source with sterilized DI water for 24 h under magnetic stirring. The required amount of solution was then pipetted into the plate wells and the plate dried at 30°C for 24 h. The amount of carbon source added ranged from 1 to 20% (w/v) with four replicates per each concentration. B. velezensis AP183 and S. aureus NE1260F were inoculated into separate 2 mL TSB cultures and incubated overnight at 37°C with shaking at 250 rpm. The cultures were normalized to OD₆₀₀ = 1.0 and then 1 mL was removed, subjected to centrifugation at 10,000 × g and resuspended in 1 mL of M9 minimal salts medium. The B. velezensis AP183 and S. aureus NE1260F cells were then diluted 1:100 into the same tube of M9 medium, and 20 mL of this stock was added to each well of the 96-well plate containing different carbon sources (for a total 1:1000 dilution, approximately 10⁶ CFU of each culture in each well) with a total volume of 200 mL per well. The 96-well plate was incubated for 72 h at 37°C with shaking at 250 rpm and then each well was serially diluted and plated on TSA to determine CFU/mL for each of the two cultures, which were discriminated based on their unique colony morphologies. Due to the expected inhibition of S. aureus viability by B. velezensis AP183 when grown on TSA at high densities, only plates with lower numbers of CFUs (<100) were selected for calculation of CFU/mL.

The bacterial strains used in this study included the mastitis isolate S. aureus RF122 and other bacterial isolates isolated from clinical mastitis cases (Table 1). All Bacillus and Staphylococcus cultures were cryopreserved in TSB containing 20% glycerol and isolated colonies obtained on TSA. The antimicrobial activity of B. velezensis AP183 against bacteria isolated from mastitis cases (Table 1) was assayed using a soft agar overlay technique. The assay was performed in triplicate and inhibitory activity was quantified by measuring the zone of inhibition in mm.

To generate a S. aureus RF122 culture for intramammary inoculation, a single colony was inoculated into 3 mL of TSB and cultivated at 37°C for 16 h followed by transfer of 1 mL of the bacterial culture to 100 mL of TSB and incubated overnight at 37°C with shaking. After overnight incubation, the culture was normalized to an OD₆₀₀ of 0.5 and then diluted to an approximate concentration of 10² CFU/mL and 10 mL of this diluted culture was used for the intramammary inoculation of bovine quarters.

For cultivation of B. velezensis AP183, an isolated colony was inoculated into 3 mL of TSB and grown at 30°C for 18 h followed by transferring 1 mL of the bacterial culture to 100 mL of TSB. After overnight incubation at 30°C with shaking, the final approximate concentration of the bacterial suspension was 10⁸ CFU/mL and 10 mL of this dose was used directly for the high dose B. velezensis AP183 treatment containing 10⁹ CFU total. For the low dose treatment, the overnight culture was diluted to 10² CFU/mL and 10 ml was used for intramammary inoculation that contained 10³ CFU total. To produce spores of B. velezensis AP183, a TSB culture was used to inoculate 10 spore preparation agar plates using a cotton swab and incubated for 10 days at 30°C. The spores were harvested, and the spore suspension was heat treated at 80°C for 20 min to kill vegetative cells. The concentration of the spore suspension was determined by serially diluting the spore suspension in sterile water and incubating in TSA plates for overnight at 30°C. The spore suspension was diluted to 10⁴ CFU/mL and were preserved at 4°C until further use. On the experimental day (D₀), the final concentration of the spore suspension was adjusted to 10² CFU/mL with sterile water and used for intramammary inoculation.

The experimental protocol was approved by Auburn University Institutional Animal Care & Use Committee (IACUC Protocol Number: 2017-3120). This study was conducted on four first-lactation Holstein dairy cows, age 2 years, at the Auburn University College of Veterinary Medicine. Animal health status was monitored clinically, and milk quality was monitored by bacteriological culture and somatic cell count analysis of individual quarter milk samples.

Prior to evaluating interactions between B. velezensis AP183 and S. aureus within mammary tissue, experiments were conducted to assess the impact of B. velezensis AP183 alone on mammary inflammatory responses.

In the first experiment a high dose inoculum of B. velezensis AP183 (10 mL of a 10⁸ CFU/mL bacterial suspension) was used for an intramammary inoculation in four animals, with half of the 16 healthy mammary glands (a fore and a rear quarter from each cow) inoculated and the other uninoculated quarters considered as negative controls. Surface cleaning was performed with a dry paper towel and a commercial teat dip containing 1% povidone iodine disinfectant for 30 s, followed by disinfecting the teat ends using isopropyl alcohol swabs, milk samples were collected every day from 1 day prior to inoculation (D_–1), inoculation day (D₀) and at the post inoculation days from one to fifteen (i.e., D₁ to D₁₅) in 15 mL conical tubes. Milk samples were immediately placed on ice and then transferred to the laboratory for analysis of microbiological counts on TSA, SCC, fat, protein, lactose, and milk solids-not-fat (SNF) content.

In a second experiment, a lower dose inoculum of Bacillus AP183 (10 mL of a 10² CFU/mL) was administered intramammary in six healthy bovines, as described above. Milk samples were collected on inoculation day (D₀) and at the post inoculation days from D₁ to D₄, for determination of Bacillus CFU counts, SCC, and other raw components. The procedure for quarters disinfection, milk collection, transportation and analysis were the same as described above for the higher dose study.

To investigate the impact of B. velezensis AP183 on the microbiology and mammary inflammatory response during S. aureus-induced clinical mastitis, 10 mL of a 10² CFU/mL S. aureus RF122 culture were administered intramammary in all quarters of four animals, 1 day prior to low dose B. velezensis inoculation (D_–1). At time D₀, half of the quarters were subjected to intramammary inoculation with 10 mL of a 10² CFU/mL B. velezensis AP183 suspension. Milk samples were collected every day starting the day prior to S. aureus inoculation (D_–2) and continuing every day to 9 days post-inoculation (i.e., D₉). The sample handling protocol was the same procedure described above for milk sample collection, transport and analysis. In addition, the milk samples collected from this S. aureus-induced mastitis experiment were aliquoted into microcentrifuge tubes and stored at -80^oC until processed for microbiome analysis for samples D₀, D₂, D₅, and D₈.

Milk samples from the high dose B. velezensis AP183 study were analyzed for bacillusin A using liquid chromatography-mass spectrometry (LC-MS) at the Auburn University Mass Spectrometry Lab, using purified bacillusin A diluted into milk known to be without exposure to the macrodiolide. The milk samples and standards were mixed with ice cold acetone (100 μL milk and 400 μL acetone), vortexed for 30 s, frozen for 15 min, subjected to centrifugation at 10,000 × g for 5 min, and decanted. Exposure to light was minimized during all sample and standard handling steps. The liquid was evaporated for 4 h in a Thermo Savant DNA 120 speed vac concentrator without any heating. The milk standards and samples were reconstituted in 50 μL of 70% water and 30% acetonitrile, vortexed, and incubated for 5 min in a water bath ultrasonicator, and subjected to centrifugation at 10,000 × g for 5 min. The supernatant was analyzed using an ultra-performance LC system (ACQUITY, Waters Corp., United States) coupled with a quadrupole time-of-flight mass spectrometer (Q-Tof Premier, Waters) with electrospray ionization (ESI) in negative mode using Masslynx software (V4.1). Injection of 5 mL of the sample or standard was made on a C18 column (ACQUITY UPLC^® BEH C18, 1.7 μm, 1 × 50 mm, Waters) with a 200 μL/min flow rate of mobile phase solution A (0.1% formic acid in 95% water and 5% acetonitrile) and solution B (0.1% formic acid in 95% acetonitrile and 5% H₂O) beginning at 35% B, held for 1 min followed by a linear ramp to 50% B in 9 min, then to 100% B at 11 min, held 2 min, and back to 35%B with 6 min of re-equilibration. The MS spectral range was 120–1300 m/z with a scan time of 0.1 s and 0.05 interscan delay. The 597.3 ion was chosen for MSMS with maximum sensitivity, collision energy of 30 eV, LM resolution of 4.7, and 0.3 s scan and 0.05 s interscan delay. The capillary voltage was set at 2.8 kV, the sample cone voltage was 30 V, and the extraction cone was 4.0 V. The source and desolvation temperature were maintained at 95 and 400°C, respectively, with the desolvation gas flow at 600 L/h. The lock mass was used to correct instrument accuracy with a 0.1 μM solution of HP 1221 (Agilent part number G1969-85003).

Cows were pastured with the Auburn University Large Animal Teaching Hospital dairy herd and fed the same ration as the remainder of the herd. Milking was carried out twice daily in a milking parlor. The California mastitis test was performed on each quarter at the time of morning milking and milk was collected for laboratory submission to quantify somatic cell counts, percent lactose, percent protein, percent fat, and percent SNF and bacterial culture. A daily physical examination was conducted immediately following milking each morning with the cows restrained in self-locking stanchions in the feeding area.

Milk samples were analyzed for the determination of SCC by flow cytometry utilizing a Somacount FCM (Bentley Instruments Chaska, MN, United States) and other milk components including lactose, protein, fat, SNF were analyzed using a FTS Fourier Transform Spectrometer (Bentley Instruments) at the Mid-South Dairy Records Laboratory (Springfield, MO, United States). Microbiological enumeration was performed by culture plate count method by serial dilution and plating for CFUs on TSA as described above.

Following endpoint procedures, wounds for each respective treatment or control group were excised using a sterile scalpel for homogenization and extraction of genomic DNA as described previously (Nasrin, 2015). In the case of biofilm formed on inner cannula sections, the sections were removed sterilely and placed into DNA isolation tubes. Genomic DNA was extracted from wound homogenates and biofilm samples using an ultraclean microbial DNA isolation kit following the manufacturer’s instruction (MO BIO laboratories Inc., Carlsbad, CA United States). For milk samples, genomic DNA was isolated using a milk bacterial DNA isolation kit (Norgen Biotek, Thorold, ON, Canada). Briefly, the milk samples were subjected to centrifugation at 20,000 × g for 2 min using a Microfuge 22R (Beckman Coulter Life Sciences, Indianapolis, IN, United States). The pellet was resuspended and used for DNA isolation according to the manufacturer’s protocol. The purity of the DNA from each sample was determined using a TECAN infinite M1000 PRO and the DNA samples were quantified using Qubit dsDNA HS Assay kit and read with a Qubit 2.0 fluorometer (Invitrogen, Life Technologies) according to manufacturer’s instructions and stored at –20°C until further use.

Each genomic DNA sample was used as a template for 16S rRNA gene sequencing using the “universal bacteria” 16S rRNA gene bacterial primer set 515F/926R that targets the hypervariable V4 and V5 regions and a high-fidelity polymerase to generate amplicons for sequencing using a MiSeq sequencer (Illumina, San Diego, CA, United States), with a 2 × 300 paired-end v2 sequencing kit and a 30% phiX control, at the University of Illinois at Chicago’s Sequencing Core. Changes in the relative abundance of bacterial taxa were assessed using the Python-based QIIME bioinformatics pipeline v.1.9.1 at the family and genera phylogenetic levels (Kuczynski et al., 2012). The specific order of processing was as follows: library generation (files available upon request), barcode trimming with the script split_libraries.py, selection of operational taxonomic units (OTUs) with the script pick_otus.py, selection of a representative set with the script pick_rep_set.py, assignment of taxonomy with the script assign_taxonomy.py, generation of an OTU table with the script make_otu_table.py, summarizing taxa with the script summarize_taxa.py, testing evolutionary distance by building an alignment with the script align_seqs.py, hard sequence filtration of the alignment using a Lane mask with the script filter_alignment.py, generation of a phylogenetic tree with FastTree with the script make_phylogeny.py, rarefaction analyses with the script multiple_rarefactions.py, assessment of alpha diversity with the script alpha_diversity.py, assessment of beta diversity with the script jackknifed_beta_diversity.py, and distance statistics were created with the script dissimilarity_mtx_stats.py. All scripts referenced above are available on the QIIME website and are also available upon request.

The R Software Package (R Foundation for Statistical Computing, Vienna, Austria) and Microsoft Excel were used to perform the statistical analyses of microbiome data. Student’s t-test, a one-way analysis of variance and paired t-tests were performed for significance determination (statistical significance required a P-value < 0.05). Statistical analyses for other data were performed using Microsoft Excel. Microsoft Excel was used to determine standard deviation and statistical significance using two-tailed paired and unpaired t-test (statistical significance required a P-value < 0.05). Data composed of pre and post treatment samples within the same group (Treatment T₀ vs. T₈) was analyzed using paired t-test and to compare the effect of treatment on microbiota (Treatment vs. Control at T₈), an unpaired t-test was performed.

The initial animal model consisted of creating four distinct wounds on the dorsal skin of the mouse and pipetting S. aureus Xen29 onto each wound, followed by application of B. velezensis strains to designated treatment wounds. This procedure proved inconsistent due to mice licking the bacterial suspensions from the wounds (data not shown). This prompted modification of the protocol wherein bacterial cultures were injected subcutaneously into two regions of the dorsal skin. To ensure that treatment was applied in the same site as the infection, B. velezensis spores and S. aureus Xen29 were co-administered by mixing in equal parts immediately prior to injection. This procedure ensured that the numbers of injected cells remained consistent, and mouse communal grooming habits did not interfere with the results.

In the first disease challenge using co-inoculation subcutaneously, spores of five B. velezensis (strains AP143, AP191, AP183, AP218, and AB01) were selected for evaluation of their respective ability to inhibit S. aureus Xen29 colonization in a mouse cutaneous wound model. Each mouse in the treatment group was simultaneously challenged with S. aureus Xen29 in two independent cutaneous wounds on each mouse back. The S. aureus Xen29 suspension and/or spores were pipetted onto wounds of two mice per treatment group. The final concentrations of S. aureus Xen29 and B. velezensis spores per wound were approximately 1.0 × 10⁷ CFU respectively. The results in these small number of animals indicated that co-administration of S. aureus Xen29 with spores of B. velezensis AP183 resulted in the greatest inhibition of S. aureus Xen29 growth and bioluminescence as compared to the other B. velezensis strains AP191, AP218, AB01, or AP143, with B. velezensis AP191 showing a moderate degree of S. aureus inhibition (data not shown).

A subsequent in vivo challenge of mice (n = 5 in each treatment and control group) with wounds treated with a B. velezensis AP183 culture resulted in a significant decrease (P < 0.05) of bioluminescence from S. aureus Xen29 (Figures 1A,B). Furthermore, we observed a significant reduction (P < 0.05) in the number of cultured CFUs in homogenates corresponding to AP183 and supernatant treated wounds (Figure 1C). Injection sites treated with B. velezensis AP191 spores and supernatant were less consistent, but still resulted in decreased bioluminescent signal overall as compared to control wounds. Bioluminescent signal in untreated control wounds peaked between days 2 and 3 with a substantial decrease in reported signal by the end of the study. However, one control mouse inoculated with S. aureus Xen29 but no Bacillus spores succumbed to systemic infection 7 days post-inoculation despite a substantial decrease in bioluminescent signal from wound sites, suggesting that in this control mouse that the S. aureus Xen29 had spread beyond the wound site several days after it was subcutaneously inoculated. Because B. velezensis AP183 was observed to provide the greatest degree of S. aureus inhibition compared to that observed for B. velezensis AP191 and other strains, and there was greater knowledge concerning the bioactive metabolite (i.e., bacillusin A) expressed by this strain that was responsible for S. aureus growth inhibition, subsequent biofilm and in vivo experiments in dairy cows focused on this strain.

Histological Gram staining revealed an abundance of Gram-positive cocci on sections of excised control wounds as compared to wounds treated with AP183 spores and/or metabolites (Figures 2A,B). However, Gram-positive cocci were also observed in treated wounds. H&E-stained sections of excised tissue appear necrotic and inflamed in contrast to AP183 treated wounds (Figures 2C,D).

The previous experiments in a mouse wound model indicated that B. velezensis AP183 provided the strongest inhibition of S. aureus proliferation and systemic infection, among the tested B. velezensis strains. In those experiments the S. aureus inoculation was performed with a bacterial suspension, and co-inoculated with a B. velezensis culture. Because S. aureus commonly forms biofilms on medical device substrates that can lead to systemic infections, an experiment was conducted to evaluate if a B. velezensis AP183 biofilm formed on a tracheostomy tube inner cannula substrate could impact subsequent S. aureus colonization. Due to difficulties in assessing accurate CFU counts from a biofilm, DNA was isolated from biofilm samples in which B. velezensis AP183 had formed a biofilm prior to inoculation with S. aureus NE1260F or from biofilm formed by S. aureus NE1260F only. In the treatment group that was inoculated previously with B. velezensis AP183, a significant decrease was observed in S. aureus relative abundance on the inner cannula sections (P = 0.01) compared to control samples. In the treatment group the S. aureus relative abundance was 3.5%, whereas the relative abundance of Bacillus was 96.5% (Figure 3A). In contrast, the control group was dominated by 99.4% S. aureus (Figure 3A).

The previous experiment showed that a B. velezensis AP183 biofilm was able to resist colonization by S. aureus. A subsequent experiment was designed to evaluate the relative abundance of these two bacteria when co-inoculated into M9 minimal medium containing different carbon sources. Whereas both B. velezensis AP183 and S. aureus can grow in the presence of glucose or sucrose, only B. velezensis AP183 can use pectin as a sole carbon source. When these two cultures were inoculated into 96-well plates containing each of these carbon sources, it was observed that the relative abundance of S. aureus NE1260F increased when the concentration of glucose or sucrose reached very high levels, but no corresponding increase in S. aureus NE1260F relative abundance was observed when pectin was the sole carbon source and this difference was significant at a 10% (w/v) concentration of these carbon sources (Figure 3B; P < 0.05). While no significant difference was observed when lactose was included as a sugar (P = 0.09) the mean relative abundance of S. aureus NE1260F in the presence of lactose remained less than 25% even at the highest lactose concentrations (Figure 3B). This experiment revealed that the competition between these bacteria is influenced by the available growth substrate. Subsequent experiments were conducted to evaluate the ability of B. velezensis AP183 to antagonize a S. aureus biofilm that was formed within a microfluidic device, which similarly showed a strong inhibitory effect of B. velezensis AP183 on S. aureus viability and resulted in biofilm disruption (data not shown, manuscript in preparation). Collectively, these experiments indicated that B. velezensis AP183 can compete with S. aureus under different environmental conditions, including in microbial biofilms on a medically relevant substrate, and that this competition is influenced by the growth substrate.

The antibacterial activity of B. velezensis AP183 against mastitis-inducing pathogens was evaluated by recording the respective zones of inhibition against each isolate. B. velezensis AP183 was observed to produce antibacterial metabolites that inhibited the growth of all eight S. aureus mastitis isolates, one S. uberis mastitis isolate, and one B. cereus mastitis isolate (Supplementary Table 1). No inhibitory effects were observed for B. velezensis AP183 against a Gram-negative bacterial mastitis isolate, K. pneumoniae (Supplementary Table 1).

To evaluate the ability of B. velezensis AP183 to induce an immune response within mammary tissue, and the timing for inflammation resolution, a high dose (approximately 10⁹ CFU) of B. velezensis AP183 was inoculated intramammary into healthy bovine mammary glands to evaluate the in vivo effects compared to healthy quarters that served as a negative control. This high-dose inoculum induced a significant increase in SCC (approximately 7 × 10⁶ cells/mL) by 24 h post-inoculation which was significantly higher than the SCC observed for the negative control quarters (at 24 h, P = 1.5 × 10^–6; P < 0.01 for all post-inoculation time points) (Figure 4). All cows developed localized clinical mastitis with visible abnormal mammary secretions from inoculated quarters without any signs of systemic disease. All quarters infused with the challenge dose of B. velezensis AP183 showed visibly evident caseous material in the milk. The California mastitis test indicated an increase in somatic cells for the Bacillus-infused quarters and for some non-infused quarters. No signs of systemic infection of elevated temperature were noted in any of the infused cows. All clinical cases resolved within 5 days without antimicrobial treatment and CMT results regressed without any further observation of mastitis symptoms. The somatic cell counts decreased to a normal level (<1.42 × 10³ cells/mL) by 10 days post-inoculation. This result as well as results from previous studies (Klostermann et al., 2008; Frola et al., 2012) suggest that, while B. velezensis could induce significant inflammation, the inflammatory response resolved within a short period of time. This also suggested that inoculation of mammary tissue with a lower dose of B. velezensis AP183 would be less inflammatory. Further analysis of the milk obtained from this preliminary experiment indicated that milk from quarters receiving a high inoculum dose had greater protein and fat content compared with control quarters at all time points with significant differences observed at D₂ and D₆ for fat (P < 0.05 at both time points) and D₂ and D5 for protein (P < 0.05 at both time points). Interestingly, the lactose and SNF values were less in the treatment group with significant differences observed for lactose levels at Days 1 through 8 (P < 0.05 for each time point) and for SNF levels at D₁, D₄, and D₅ (P < 0.05 at each time point) (Supplementary Figure 1). The reduced lactose and SNF levels in the treatment group with high SCC in this study is in accordance with the study of dos Reis et al. (2013) but the high protein and fat content of milk from inoculated, high SCC quarters that were observed in this study did not agree with observations made in some other studies (Schultz, 1977; dos Reis et al., 2013).

The milk samples from quarters treated with the high dosage of B. velezensis AP183 at D₅ did not have detectable bacillusin A based on LC-MS analysis. The detection of purified bacillusin A indicated that the limit of detection sensitivity was 3 ppm. A possible reason for the poor sensitivity for bacillusin A detection could include its instability (Ravu et al., 2015), or possibly during extraction from milk samples the bacillusin A was degraded or inefficiently recovered. It is unknown whether B. velezensis AP183 germinates and expresses bacillusin A within mammary tissues. Further research to provide better recovery of intact bacillusin A from milk samples and to assess bacillusin A biosynthetic gene cluster expression would be helpful in assessing any antibiotic residues present in milk, especially when a lower spore inoculum is used.

A subsequent experiment was conducted to determine the effect of a lower dosage of B. velezensis AP183 on the mammary inflammatory response. A lower dose inoculum of 10³ CFU B. velezensis AP183 did not induce clinical mastitis in any of the inoculated cows. Milk sampled from B. velezensis AP183-inoculated quarters was visibly abnormal, and a slight increase of SCC was observed for those samples, although the difference was not significant. In this experiment, the mean SSC was higher in treatment group compared to the control from day D₁ to D₄; however, the SCC for milk from B. velezensis AP183-inoculated quarters was below the regulatory threshold of bulk-tank SCC which is 750,000 in the United States (Supplementary Figure 2A). The Bacillus CFU counts observed from inoculated quarters increased slightly at D₂ compared to control (Supplementary Figure 2B). In addition, protein concentration was increased in milk from inoculated quarters, while lactose concentration was reduced in the treatment group compared to control. No changes were observed in fat in the treatment group compared to control, though SNF showed a different pattern with a reduction of SNF observed in the treatment group at D₁ while it increased the control group (Supplementary Figure 3). The significance of this brief change in SNF at D₁ in terms of milk quality is unknown. While the experiments evaluating the mammary inflammatory response at the two dosages of B. velezensis AP183 were performed with different animals, at different times and with different sample sizes, the procedures for inoculation, milk collection, testing and analyzing were the same for these experiments. A comparison of the mammary tissue inflammatory response at low vs. high dose of B. velezensis AP183 over the time course post-inoculation revealed significant differences in SCC between low dose and high dose at each treatment day from D₁ through D₄ (D₁, P = 6.8 × 10^–6; D₂, P = 0.0005; D₃, P = 0.04; D₄, P = 0.03), in each case a vigorous inflammatory response was observed for the high dosage of B. velezensis AP183 whereas the low dosage never exceeded the regulatory threshold of bulk-tank SCC in the United States (Figure 4).

Based on the observation of a significant inflammatory response observed when B. velezensis AP183 was inoculated by intramammary infusion at a high dose, we sought to evaluate the ability of this strain to antagonize S. aureus within mammary tissues when inoculated at a lower dosage, post-induction of clinical mastitis. All quarters challenged with 10³ CFU of S. aureus RF122 were observed to have increased SCC with clinical mastitis by 24 h post-inoculation, with decreasing SCC observed at later time points (Figure 5A). Interestingly, after D₁ a greater SCC reduction was observed in the B. velezensis AP183 treatment group as compared to control (Figure 5A), indicating that the inoculated B. velezensis AP183 reduced mastitis severity. At 7 days post-inoculation with B. velezensis AP183, SCC levels were decreased to approximately day 0 levels. Furthermore, at D₃ S. aureus counts were significantly higher (P = 0.04) in control samples as compared to samples treated with B. velezensis AP183 (Figure 5B). The levels of fat, protein, lactose and milk solids-not-fat (MSNF) were similar between the control and treatment groups in this lower dose inoculum experiment (Supplementary Figure 4), with no significant differences observed between control and B. velezensis AP183-treated quarters (P > 0.05).

In the previously described study in which S. aureus RF122 was inoculated into all quarters, followed by a low dosage inoculum of B. velezensis AP183 into half of these quarters, the milk bacterial community composition was observed to have differences between the control (only S aureus RF122) and treatment (S. aureus RF122 and B. velezensis AP183) groups. The microbiota of milk from cows affected by mastitis was comprised mainly of four bacterial phyla, including taxa affiliated with the Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidetes (Figure 6A). At D₀, the milk microbiota was dominated by members of the Firmicutes (mean relative abundance of 74.3% at D₀) followed by Proteobacteria taxa (mean relative abundance of 17.5% at D₀). Interestingly, the quarters treated with B. velezensis AP183 were observed to have a decreased Firmicutes relative abundance over time post-inoculation, changing from 70.4% at D₀ to 48.0% at D₂ to 33.1% at D₈ (Figure 6A). A similar trend for Firmicutes relative abundance was observed for the control group which was observed to be 78.3% at D₀ to 31.9% at D₂ to 42.6% at D₈ (Figure 6A). It is important to note that the increased Firmicutes relative abundance in the control group at the final time point D₈ was due at least in part to the higher relative abundance of the genus Staphylococcus. In contrast, the relative abundance of Proteobacteria taxa were observed to increase in both the control and B. velezensis AP183-treated samples at these same time points, with observations of Proteobacteria abundance for control samples changing from 14.6% at D₀ to 53.4% at D₂ to 40.1% at D₈, and similarly the Proteobacteria relative abundance for the B. velezensis AP183 treated samples were at 20.4% at D₀ and increased to 41.1% at D₂ to 40.1% at D₈ at the final time point (Figure 6A). The relative abundance of Actinobacteria and Bacteroidetes taxa were observed to be at comparable levels in control and treatment groups.

In addition to phylum-level differences, differences in the relative abundance of bacterial genera were also observed in milk microbiome samples. Milk samples were dominated by Staphylococcus, Achromobacter, Delftia, Pseudomonas, Agrobacterium, Acinetobacter, Hydrogenophaga, Corynebacterium, and Bacillus spp. (Figure 6B). In the quarters treated with B. velezensis AP183, a significant decrease was observed in the relative abundance of the genus Staphylococcus in milk samples over time (D₀ vs. D₈, P = 0.003; D₂ vs. D₈, P = 0.030), with the treatment group showing Staphylococcus relative abundance changing from 61.4% at D₀ to 39.3% at D₂, and to 7.4% at D₈. In contrast, in the control group the Staphylococcus relative abundance also decreased significantly over time (P = 0.002) from 69.8% at D₀ to 20.2% at D₈, but no significant difference was observed from 22.3% at D₂ to 20.2% at D₈. Most importantly, the results indicated that B. velezensis AP183 treatment had a significant effect in decreasing Staphylococcus relative abundance relative to the control group at D₈ (P = 0.021). Further, the relative abundance of the genus Bacillus was higher in the treatment group (2.9%) at D₈ compared to control group (1.5%) as would be expected, although this result was not significant. Besides Staphylococcus, the only other bacterial genus in which the relative abundance exhibited a significant difference between early time points (D₀ or D₂) and D₈ in control and treatment groups was the genus Hydrogenophaga, which in control samples went from 2.3% at D₀ to 1.4% at D₈, whereas in treatment samples the Hydrogenophaga relative abundance was observed to be at 0.4% at D₀ and was at 3.5% at D₈.

Shannon and Simpson diversity indices were not different between the control and treatment groups, but the sample diversity was slightly higher at day 12 in treatment group based on species richness and chao 1 (Table 2), although no statistically significant differences were observed (P > 0.05). Interestingly, there were distinct differences between the control and treatment groups observed in the principal component analysis based on weighted Unifrac distances (Figure 7). The treatment group clustered together except for two outliers. While the distribution of data in the control group was dispersed and there were apparent differences from the treatment samples. Principal component 1 explained a large percentage of the variation (86%) for Staphylococcus spp. that demonstrated a high correlation of the relative abundance of the Staphylococcus with the first principal component.