Medium 200 (M200PRF500), low serum growth supplement (LSGS; S00310), 0.025% trypsin-EDTA (R001100), VEGF (PHC9344), Pierce Protease Inhibitor Mini Tablets (A32953), HisPur™ Cobalt Resin (89964), Detoxi-Gel™ (20339), penicillin-streptomycin (15140122), amphotericin B (15290018), Qubit™ Protein Broad Range kit (A50668) and human umbilical vein endothelial cells (HUVECs; C0035C; RRID: CVCL K312) were purchased from Thermo Fisher Scientific/Fisher Scientific. ToxinSensor™ LAL Endotoxin Assay Kit (L00350) was purchased from Genscript. HisLink™ Protein Purification Resin (V8823) and CellTiter 96^® AQueous One Solution (G3582) were purchased from Promega. Mitomycin C (M4287) was purchased from Sigma Aldrich. Proteome Profiler™ Human Angiogenesis Antibody Array (ARY007) was purchased from R&D Systems. IRDye 800 CW Streptavidin (926-32230) was purchased from LI-COR. Growth factor reduced Matrigel (GFR-Matrigel; 356231) was purchased from Corning. Axitinib (HY-10065) was purchased from MedChemExpress. 0.1 mm glass beads (P000929LYSK0A.0) and soft tissue homogenizing kits (P000933-LYSK0-A) were purchased from Bertin Corp. 4-well culture inserts (80466) and angiogenesis slides (81506) were purchased from ibidi. MycoAlert™ Plus Mycoplasma Detection Kit (LT07-701) was purchased from Lonza.

N-terminal His₆-tagged β-toxin was previously cloned into E. coli TOP10 using a pTrcHis TOPO vector (Herrera et al., 2016). The plasmid was maintained with 100 μg mL^–1 carbenicillin in all growths. Cells were grown in 1 L terrific broth (24 g yeast extract, 12 g tryptone, 4 mL glycerol, 100 mL of supplement [0.17 M KH₂PO₄, 0.72 M K₂HPO₄]) at 37°C to an OD₆₀₀ 0.4–0.8 followed by induction with 1 mM IPTG overnight at 30°C. Pelleted cells were resuspended in 25 mL resuspension buffer (50 mM NaH₂PO₄, 500 mM NaCl, 20 mM imidazole, pH 8) and three Pierce Protease Inhibitor Mini Tablets. 10 mL aliquots were divided into bead lysing tubes containing 7 g of 0.1 mm glass beads and homogenized using a PreCellys Cryolys Evolution (bertin Instruments) with the following settings: 9,900 rpm, 6–30 s cycles with 60 s rests, 4°C. Lysate was centrifuged (40 min, 50,000 x g, 4°C) and clarified with a 0.45 μm filter. Affinity chromatography with HisPur™ Cobalt Resin followed by an imidazole-gradient elution (50 mM NaH₂PO₄, 500 mM NaCl, 250 mM imidazole, pH 8) was used to separate β-toxin. Protein-containing fractions, assessed by SDS-PAGE, were dialyzed against 4 L PBS, pH 7.4, overnight at 4°C. Protein concentration was determined by Qubit™ using the Qubit™ Protein Broad Range kit. Typical yield was 4–5 mg per liter of growth. Purity was assessed by Coomassie stain of SDS-PAGE gels and was at least 95% by visual observation. All proteins were cleaned of endotoxin via Detoxi-Gel™ resin and endotoxin levels were assessed using the ToxinSensor™ LAL Endotoxin Assay Kit. Proteins were used when the final endotoxin concentration in experiments was ≤ 0.025 ng mL^–1 (Kulhankova et al., 2018).

Immortalized human aortic endothelial cells (iHAECs) are a recently established cell line shown to retain phenotypic and functional characteristics of primary cells, serving as a large-vessel model system in which to address questions relevant to vascular biology (Kulhankova et al., 2018). Human umbilical vein endothelial cells (HUVECs) were obtained from Thermo Fisher.

Cells were grown at 37°C, 5% CO₂ in phenol red–free, endothelial cell basal medium (Medium 200) supplemented with low-serum growth supplement (LSGS, final concentrations of: FBS 2%, hydrocortisone 1 μg mL^–1, human epidermal growth factor 10 ng mL^–1, basic fibroblast growth factor, 3 ng mL^–1, heparin 10 μg mL^–1). Cells were maintained on 1% gelatin-coated plates unless otherwise stated. Cells were passaged at least twice before use in experiments. iHAECs were used at passages between 4 and 10 from a single clone. Primary HUVECs were used between 4 and 12 passages. Mycoplasma-testing was conducted every 6 months using MycoAlert Plus Mycoplasma™ Detection Kit.

An MTS assay was used to determine cell viability. Cells were seeded at 7,000 cells/well into gelatin-coated 96-well plates and grown overnight to near confluency. Media was removed and replaced with 100 μL of media containing increasing concentrations of β-toxin, VEGF, axitinib, or mitomycin C followed by overnight incubation. 20 μL of CellTiter 96^® AQueous One Solution was added to each well followed by a 1 h incubation at 37°C, 5% CO₂. A plate reader was used to read absorbance at 490 nm. Three biological replicates were conducted with each containing three technical replicates. The data were normalized so that untreated cells were considered 100% activity by dividing the absorbance of treated cells by the absorbance of untreated cells.

Gelatin-coated 96-well tissue culture plates were seeded at 7,000 cells/well and grown to near confluence. Fresh media containing β-toxin at 50 μg mL^–1 was added, and plates were incubated for 24 h at 37°C, 5% CO₂. The conditioned media was removed and stored at -80°C until analyzed. Three biological replicates were conducted with each containing three technical replicates for every treatment. The relative expression of 55 angiogenesis-related proteins was determined from the conditioned media of various experiments using a Proteome Profiler™ Human Angiogenesis Antibody Array with fluorescent analysis according to the manufacturer’s instructions modified for fluorescent analysis. A total of 120 μL of conditioned media was incubated with a cocktail of biotinylated detection antibodies (“sample-antibody mixture”) for 1 h at room temperature. During this incubation, the membrane containing the capture antibodies was blocked using kit-supplied blocking buffer at room temperature. After the hour incubation, the sample-antibody mixture was added to the washed membrane and incubated overnight at 4°C. After three 10-min washes at room temperature, the membrane was incubated with IRDye 800 CW Streptavidin (1:2,000 dilution) for 30 min at room temperature in the dark. After three 10-min washes at room temperature, the fluorescent signal was detected using an Azure c600 (Azure Biosystems; 120 μm resolution, auto intensity). The signal produced at each spot is proportional to the amount of analyte bound and the mean pixel intensity of the duplicate spots on the membrane was calculated and averaged using Image Studio Software (LI-COR; RRID: SCR_013715). Fold-changes over untreated controls were calculated for each detected protein. An untreated control was always performed. All treatments were matched. After an extensive literature search and cross-referencing of the GTExPortal and Expression Atlas databases, seven analytes unlikely to be produced by endothelial cells were removed from final analysis (angiopoietin-1, angiostatin/plasminogen, EG-VEGF, FGF-4, leptin, platelet factor 4, serpin B5). None of these analytes were produced by iHAECs.

4-chamber silicone inserts were placed into 12-well uncoated tissue culture treated plates. Each chamber was seeded with 3.08 × 10⁴ cells and plates were incubated at 37°C, 5% CO₂ for 4 h. The media was removed and replaced by media containing β-toxin (50 μg mL^–1) and incubated overnight. The inserts were removed, and conditioned media placed aside. The wells were washed with DPBS, and the conditioned media returned to the wells with additional media containing effectors so that the final volume was 1.5 mL per well. Experiments were also conducted where β-toxin (50 μg mL^–1), VEGF (10 ng mL^–1), axitinib (10 μM), or mitomycin C (2 μg mL^–1) was added after insert removal. The plates were incubated overnight in a Leica DMi8 equipped with a Tokai Hit stage-top incubator set to 37°C, 5% CO₂. Images were captured every 30 min for 24 h using a HC PL FLUOTAR 4x/0.13 objective lens. Five independent experiments were conducted for each treatment condition. Images were automatically analyzed via ImageJ (RRID: SCR_002285). The edges were found (Process → Find Edges) and the image was smoothed 10 times (Process → Smooth). A MinError (I) threshold was then applied [Image → Adjust → Auto Local Threshold: MinError (I)] to automatically detect cells. The particle count was then quantified with a particle size of 1,000-infinity (Analyze → Analyze Particles [size: 1,000—infinity]) (Venter and Niesler, 2019).

For cell proliferation assays, cells were seeded at 7,000 cells/well into gelatin-coated 96-well plates and immediately treated with either β-toxin (50 μg mL^–1), VEGF (10 ng mL^–1), axitinib (10 μM), or mitomycin C (2 μg mL^–1). Plates were incubated overnight in a Leica DMi8 equipped with a Tokai Hit stage-top incubator (Tokai Hit Co., Ltd.) set to 37°C, 5% CO₂. Merged images were captured every 30 min for 5 h, then every 5 h using a HC PL FLUOTAR 4x/0.13 objective lens. Three independent experiments were conducted for each treatment condition. Cells were automatically counted using ImageJ by modifying the protocol outlined by Venter and Niesler (2019). Images were converted to grayscale and the edges found (Process → Find Edges). An Isodata threshold was then applied (Image → Adjust → Auto Threshold: Isodata) to automatically detect cells. The particle count was then quantified after determining the appropriate particle size to decrease background (Analyze → Analyze Particles [size: 0.003–0.2]).

Wells in angiogenesis μ-slides were coated with 10 μL of GFR-Matrigel and allowed to polymerize for 1 h in a humidified chamber at 37°C, 5% CO₂. Cells were seeded at 10,000 cells/well in media containing β-toxin (50 μg mL^–1) or axitinib (30 μM). The μ-slides were incubated overnight in a Leica DMi8 equipped with a Tokai Hit stage-top incubator set to 37°C, 5% CO₂. Images were captured every hour for 12 h using a HC PL FLUOTAR 4 x/0.13 objective lens. Images were analyzed via the ImageJ Angiogenesis Analyzer plugin (Carpentier et al., 2020). A minimum of six independent experiments with five technical replicates were conducted.

Mixed-sex New Zealand white rabbits, 2–3 kg, were purchased from Charles River Laboratories (Massachusetts) and maintained at Charmany Instructional Facility at the School of Veterinary Medicine (SVM) of the University of Wisconsin (UW)-Madison. All rabbits were individually caged and given access to food and water ad libitum. Rabbits were given a period of at least 4 days to acclimate and deemed generally healthy by a veterinarian before experimental procedures.

Aortic ring explants were prepared following the thin-layer method (Zhu and Nicosia, 2002) with modifications to the matrix embedding process as described below. The thoracic and abdominal aortas were excised immediately after euthanasia. In a petri dish containing PBS, excess fascia and connective tissue were removed, then 1–1.5 mm² cross-sections were cut with a scalpel. 300 μL phenol red-free GFR-Matrigel was added to wells in 24-well plates and rings immediately embedded. After 10 min polymerization at 37°C, 5% CO₂, 500 μL supplemented Medium 200 was added and plates were incubated at 37°C, 5% CO₂ up to 14 days. Medium 200 contained LSGS, 100 U mL^–1 penicillin-streptomycin, 2.5 μg mL^–1 amphotericin B, and relevant treatments. Media (± treatments) was changed every 3–5 days. Merged images were captured every other day using the Leica DMi8 equipped with a Tokai Hit stage-top incubator set to 37°C, 5% CO₂ using a HC PL FLUOTAR 4x/0.13 objective lens. Growth was assessed using ImageJ (Stiffey-Wilusz et al., 2001). A total of three rabbits were used with a minimum of three rings per condition.

Statistical analyses were performed using GraphPad Prism software (RRID: SCR_002798). For each experiment, the precision measures and number of technical and biological replicates are indicated in figure legends. The number of cells, number of measurements and timing of experiments can be found in the Method Details for each experimental setup. For the proteomics analysis, emphasis was placed on proteins with mean fold change outside of the 0.5–1.5-fold change previously described using this same array (Barron et al., 2017). Cell proliferation, tube formation, and wound healing were analyzed by two-way repeated measures ANOVA with α = 0.05. For MTS assays unpaired two-tailed t-tests were conducted. Statistical significance was given as ^∗p < 0.0332, ^∗∗p < 0.0021, ^∗∗∗p < 0.002, ^∗∗∗∗p < 0.0001.

SMase activity results in the production of bioactive sphingolipids, recognized as signaling molecules that modulate angiogenesis (Van Brocklyn and Williams, 2012; Mehra et al., 2014). Hence, we first sought to establish whether S. aureus β-toxin alters the secretion profile of angiogenesis-related proteins in immortalized human aortic endothelial cells (iHAECs). For this purpose, we used a human angiogenesis proteome array to profile 48 proteins in iHAEC supernatants from subconfluent monolayers treated for 24 h under proangiogenic (growth medium ± VEGF), antiangiogenic (+ axitinib), or toxin conditions. In complete medium (basal medium with growth supplements), the most highly detected proteins secreted by endothelial cells were serpin E1, endothelial growth factor (EGF), thrombospondin-1, and endothelin-1, which promote either matrix degradation, growth, or proliferation and migration (Supplementary Figure 1A). These were followed largely by proteins that promote endothelial cell proliferation and migration (artemin, insulin-like growth factor binding protein (IGFBP)-2, IGFBP-3, pentraxin 3), capillary formation (angiopoietin-2), and tissue inhibitor of metalloproteinases (TIMP)-1. Hence, as expected, iHAECs in growth medium are under angiogenic-inducing conditions. For the subsequent studies, the secretion profile of iHAECs treated with VEGF (10 ng mL^–1), axitinib (30 μM), or β-toxin (50 μg mL^–1) were expressed as fold change from growth medium control, with a cut-off of ≥ 1.5-fold or ≤ 0.5-fold as thresholds for 50% increases or decreases in protein levels, respectively (Barron et al., 2017).

VEGF promotes angiogenesis and is produced by iHAECs during growth in a monolayer, albeit at low levels (Supplementary Figure 1A). Thus, VEGF treatment was used to maximally induce angiogenesis under experimental conditions tested in our study. VEGF-treated cells exhibited a similar profile to that of growth medium control, with the exception of IL-8 which showed increased production (1.5-fold; + 50%) (Figure 1). Furthermore, in the presence of complete medium, VEGF did not further induce proliferation, confirming optimal proangiogenic conditions (Supplementary Figures 2A,B). Axitinib is an antiangiogenic molecule that inhibits VEGF receptor-1, -2, and -3 signaling in endothelial cells (Kelly and Rixe, 2009) and inhibits both cell proliferation and cell migration (Siedlecki et al., 2017). At working concentrations, axitinib significantly inhibited metabolic activity and proliferation (Supplementary Figures 2C,D) while preserving integrity of the monolayer as visually established. Consistent with its antiangiogenic activity, axitinib induced an overall inhibitory profile with < 0.5-fold decreases in production of granulocyte-macrophage colony-stimulating factor (GM-CSF), A Disintegrin and Metalloproteinase with Thrombospondin Motifs (ADAMTS)-1, Hepatocyte Growth Factor (HGF), pentraxin 3, TIMP-1, and dipeptidyl peptidase IV (DPPIV) (Figure 1). β-toxin treatment at a sublethal concentration (Supplementary Figure 1B) resulted in an overall inhibitory profile that was centered around proteins important for cell proliferation and migration: IGFBP-3 (0.4-fold; −64%), TIMP-1 (0.2-fold; −75%), TIMP-4 (0.5-fold; −50%), thrombospondin-1 (0.4-fold; −63%), and endothelin-1 (0.3-fold; −71%) (Figure 1). The protein profile resulting from β-toxin treatment suggests that β-toxin is an antiangiogenic molecule that inhibits angiogenesis by targeting proliferation and migration.

As in vitro wound healing assays are a function of proliferation and migration, we used this approach to directly evaluate the ability of iHAECs to close a gap in the monolayer in the presence or absence of β-toxin (50 μg mL^–1). VEGF (10 ng mL^–1) and axitinib (10 μM) were used as inducing or inhibition controls, respectively. Of note, axitinib, at the concentration used to treat monolayers (30 μM) resulted in generalized iHAEC toxicity during wound healing (as observed by loss of integrity of the monoloayer) (Supplementary Figure 2D). Hence, we decreased the axitinib concentration to 10 μM for this assay which inhibited proliferation but did not result in cytotoxicity (Supplementary Figure 3).

Without treatment, iHAECs closed 80% of the gap by 24 h, as measured in time-lapse analyses (Figure 2). iHAECs stimulated with VEGF showed an increase in percent gap closure of 15%, while those treated with axitinib displayed significant inhibition with a 28% decrease in gap closure compared to untreated cells (Figures 2A,B). Similarly, iHAECs treated with β-toxin exhibited a 28.5% decrease in gap closure (Figures 2A,C). During infection, host cells can undergo prolonged exposure to β-toxin before vascular damage occurs, potentiating the inhibitory phenotype. To assess this, iHAECs were treated overnight with β-toxin, prior to gap formation, and thereafter as previously performed. Pretreatment with β-toxin significantly inhibited gap closure by 34.3% (Figures 2A,D), but overall provided no significant further inhibition (Figures 2B,C). These results were confirmed in human umbilical vein endothelial cells (HUVECs), where similar inhibition of gap closure was observed upon exposure to β-toxin with and without pretreatment (Supplementary Figure 4). These data indicate that β-toxin markedly inhibits re-endothelialization, contributing to defects in vascular repair.

Because differential effects were observed in the wound healing assay in response to β-toxin and experimental controls, we investigated the proteome profile in media collected at the end of the assays. This provided an opportunity to establish whether distinct profiles would ensue from cell populations that contain cells actively migrating and proliferating to close a gap vs. those in a monolayer. In the wound healing assay, serpin E1, EGF, and endothelin-1 remained unchanged compared to growth in a monolayer (Supplementary Figure 1A). Overall, these remained the most highly detected proteins secreted by iHAECs in vitro. Interestingly, compared to growth in a monolayer, iHAECs in the wound healing assay exhibited both increases and decreases in proteins involved in growth, proliferation, and migration, where placental growth factor (PlGF), TIMP-1, and thrombospondin-1 were increased by 1.6-fold (+ 57%), 2.1-fold (+ 108%), and 1.7-fold (+ 67%), respectively, while hepatocyte growth factor (HGF), and IGFBP-2 were decreased by 0.5-fold (−51%) and 0.4-fold (−57%), respectively (Supplementary Figure 1A).

iHAECs in the wound healing assay were more responsive to the effects of VEGF and axitinib. Cells treated with VEGF exhibited an induced angiogenesis profile compared to both untreated cells (Figure 3) and VEGF-treated cells grown in a monolayer (Supplementary Figure 5). Proteins that showed increases during wound healing spanned most categories, with those important in extracellular matrix degradation being the sole exception (Figure 3). These results are consistent with enhanced gap closure (Figure 2A) and reiterate the prominent effects of VEGF on regulation of angiogenic factors. While axitinib significantly inhibited gap closure (Figure 2A), it induced a mixed protein profile and a shift away from the inhibitory profile observed in iHAEC monolayers (Supplementary Figure 5). Instead, during wound healing, axitinib specifically targeted matrix metalloproteinase (MMP)-8 (0.5-fold; −47%), PDGF-AA (1.7-fold; + 34%), and TIMP-4 (0.5-fold; −41%) (Figure 3). Subtle decreases in a select group of proteins caught our attention as they stand opposite to the changes observed in VEGF-treated cells. These included the cytokines CXCL-16 and IL-1β, MMP-9, the growth factor HB-EGF, and proteins involved in regulating or aiding proliferation and migration (IGFBP-2, TIMP-1, activin A, serpin F1, and DPPIV).

Strikingly, iHAECs treated with β-toxin during wound healing exhibited a mostly neutral profile with subtle changes observed throughout. Most notable were decreased MMP-8 (important in matrix remodeling for angiogenesis) and increased endostatin (an angiogenesis inhibitor) (Figure 3). The most relevant change after β-toxin treatment was increased production of the growth factor HGF by 1.7-fold (+ 69%) (Figure 3). This profile is intriguing given that in that same context β-toxin significantly inhibited gap closure (Figure 2C). Pretreatment shifted the profile back to mostly inhibitory, where iHAECs exhibited large decreases in proliferation/migration proteins PD-ECGF (0.4-fold; −59%) and TIMP-1 (0.5-fold; −46%) (Figure 3). Therefore, subtle changes in the balance of angiogenic proteins may be sufficient to significantly impact wound healing as measured in vitro. Collectively, these results illustrate the context–dependent characteristics of the angiogenic profile of iHAECs in response to exogenous agents.

Having established that β-toxin inhibits wound healing, we addressed whether this effect was the result of impaired cell migration, cell proliferation, or both. For this purpose, we conducted wound healing assays in the presence of mitomycin C, an antiproliferative agent. Mitomycin C was used at 2 μg mL^–1 as this concentration reduced metabolic activity by 49.2%, and in combination with β-toxin did not further decrease metabolic activity (Figure 4A). Treatment with mitomycin C above 2 μg mL^–1 resulted in cell toxicity (data not shown). Mitomycin C inhibited wound healing by 17.8% compared to untreated cells, where the largest effect on cell proliferation and concomitant decreases in wound closure were measured past 15 h (Figures 4B,C). iHAECs concurrently treated with β-toxin (50 μg mL^–1) and mitomycin C exhibited a 40.9% decrease in wound closure compared to untreated cells and a 28.8% decrease compared to those treated only with mitomycin C (Figures 4B,C). The additive effect of mitomycin C and β-toxin in the wound healing assay is consistent with β-toxin inhibiting cell migration.

In the wound healing assay, initially cells exist in a monolayer and gap closure is largely driven by actively migrating cells of the gap cell front. Therefore, we sought to directly test β-toxin effects on cell proliferation in a population of actively proliferating cells. For this, iHAECs were cultured at the time of seeding in the presence or absence of β-toxin or mitomycin C and total cell counts were measured from images captured every 30 min for 5 h, then every 5 h over a 20-h period (Figure 4D). In this context, β-toxin significantly inhibited cell proliferation comparable to the inhibition induced by mitomycin C (Figure 4D). These data provide evidence that β-toxin can interfere with the angiogenic process by both inhibiting cell migration and cell proliferation.

Re-vascularization is the ultimate outcome of angiogenesis. In vitro, endothelial cell differentiation into capillaries or tubulogenesis can be addressed with a tube formation assay. For this assay, iHAECs and HUVECs were seeded on growth factor reduced (GFR)-Matrigel in medium containing serum to induce tube formation. Cells were treated with either β-toxin or axitinib (inhibitor control) at the time of seeding and the average tube length and loop count was measured over a 12-h time frame. Axitinib significantly decreased tube length and loop count over time (Figures 5A–C). Yet, β-toxin treatment produced mixed results in aortic vs. umbilical vein endothelial cells. In the presence of β-toxin, the average tube length and loop count remained unaffected in iHAECs (Figures 5D,E). With HUVECs, β-toxin had no effect on tube length but did significantly decrease loop count, suggesting a direct effect on network complexity during re-vascularization (Figures 5F,G). The differential responses of HUVECs and iHAECs to β-toxin likely reflects either the heterogeneity of endothelial cells or the altered physiology of immortalized cells (Ribatti et al., 2002).

Hence, it remained to be established if β-toxin inhibits angiogenesis by targeting capillary formation. To address this in a more physiologically relevant system, we utilized the rabbit aortic ring model of angiogenesis. In this model, thoracic and abdominal aortas were explanted from New Zealand white rabbits, cut into ∼1 mm sections, embedded into a thin layer of GFR-Matrigel, and cultured in complete medium to induce sprouting at the severed edge of the explant. We used equal numbers of thoracic and abdominal aortic rings per condition obtained from 3 individual rabbits. After embedding, aortic rings were cultured in the presence or absence of β-toxin or axitinib for 14 days. Untreated aortic ring explants (n = 26) formed sprouts within a week that continued to grow in density and complexity over time while rings treated with axitinib (n = 8) or β-toxin (n = 12) failed to sprout over the course of the experiment (Figure 6 and Supplementary Figures 6, 7). Thus, in the more physiologically relevant context of the aortic ring model, β-toxin completely inhibited angiogenesis.