To study the role of endothelial ADAM10 in S. aureus sepsis, we generated transgenic mice that enable endothelial cell–specific deletion of ADAM10 under control of the VE-cadherin promoter that drives expression of a tamoxifen-inducible (TAM-inducible) Cre recombinase (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI168450DS1) (43). ADAM10 staining of isolated aorta and mesenteric artery after TAM treatment of VE-Cad ADAM10loxP/loxP mice demonstrated a 53% and 37% reduction in Adam10 expression, respectively, compared with that observed in vessels from control VE-Cad ADAM10WT/WT mice (Figure 1A and Supplemental Figure 1B). Although WT control mice (VE-Cad ADAM10+/+) succumbed to sepsis induced by i.v. inoculation of 1 × 108 CFU S. aureus USA300/LAC, endothelial ADAM10-knockout mice (VE-Cad ADAM10–/–) were protected against lethal infection (Figure 1B). Similarly, VE-Cad ADAM10–/– mice were protected against lethality induced by i.v. delivery of purified active Hla, confirming the sufficiency of this toxin for ADAM10-dependent outcomes (Figure 1C). Evaluation of health scores (Supplemental Table 1 and Supplemental Figure 1C) and physical activity (Supplemental Videos 1 and 2) over the course of S. aureus infection reflected protection against the clinical disease observed in VE-Cad ADAM10–/– mice relative to controls, despite similar weight loss in both groups (Supplemental Figure 1D). S. aureus recovery from multiple tissues was comparable in control and VE-Cad ADAM10–/– mice 24 hours after infection (Supplemental Figure 1E), suggesting that the improved survival of VE-Cad ADAM10–/– mice was not simply related to early bacterial control.

Staphylococcus aureus targets endothelial ADAM10 to cause lethal sepsis. (A) ADAM10 (red) staining in the descending aorta and mesenteric artery from control or VE-Cad ADAM10–/– mice. DAPI (blue) denotes cell nuclei; collagen autofluorescence (green). Scale bars: 30 μm. (B) Survival following lethal S. aureus infection in VE-Cad ADAM10–/– (n = 37) or control (n = 18) female mice. Independent experiments were repeated 4 times, and data were pooled. (C) Survival following lethal purified Hla sepsis in VE-Cad ADAM10–/– (n = 9) or control (n = 10) female mice. Independent experiments were repeated twice, and data were pooled. (D) Serum IL-10 analysis 8 and 24 hours after infection in female VE-Cad ADAM10–/– mice or controls. Data represent 3 independent pooled experiments. Data are presented as the mean ± SEM. (E) Mouse platelet count enumerated 4 hours after lethal infection with S. aureus in male and female mice. Data are from 3 independent pooled experiments and are presented as the mean ± SD. *P ≤ 0.05 and **P ≤ 0.01, by unpaired, 2-tailed t test.

As human mortality from S. aureus sepsis correlates with host immunologic markers of inflammation and a sustained decrease in circulating platelet numbers (24–26, 44), we evaluated serum cytokine levels and quantified platelets in WT and VE-Cad ADAM10–/– mice that received a sublethal i.v. inoculum of S. aureus. Serum IL-10 was increased in WT mice 8 hours after infection, whereas VE-Cad ADAM10–/– mice had a minimal increase in IL-10 levels (Figure 1D), mirroring the response seen in humans who survive infection. Although serum IL-1β levels were elevated 8 hours after infection, we observed no significant difference in this response between VE-Cad ADAM10–/– and control mice (Supplemental Figure 1F). Similarly, both groups of mice exhibited comparable platelet counts 4 hours after infection (Figure 1E).

On the basis of these findings, we hypothesized that Hla-induced endothelial damage may incite immunothrombosis during S. aureus sepsis, coupling a functional alteration of endothelial integrity and platelet function. To inform an understanding of microvascular injury induced by S. aureus in situ, we used 2-photon microscopy to visualize platelet-endothelial interactions within the hepatic vasculature following infection of VE-Cad ADAM10–/– mice and matched controls with a nonlethal inoculum of S. aureus. WT mice developed more substantial platelet aggregates than did VE-Cad ADAM10–/– mice 6–8 hours after infection (Figure 2, A and B, and Supplemental Figure 2A). We observed a limited thrombus burden in both control and VE-Cad ADAM10–/– uninfected mice (Supplemental Figure 2B), which was statistically different from that observed in infected VE-Cad ADAM10–/– mice (average thrombi area of 104 μm2 vs. 635 μm2, SEM ± 97.93, P < 0.0001). Mice with platelet-specific ADAM10 deletion (PF4 ADAM10–/–) were also protected from platelet aggregate formation within the liver vasculature, whereas mice with myeloid-specific ADAM10 deletion (LysM ADAM10–/–) were not protected (Supplemental Figure 2, C and D). As neither PF4 ADAM10–/– nor LysM ADAM10–/– mice are protected from lethal S. aureus sepsis (12), these data together suggest that microvascular injury and platelet aggregation are intimately linked to impair tissue oxygenation and organ function (45, 46). We thus examined end-organ injury in VE-Cad ADAM10–/– mice. Although gross pathologic analysis of control livers revealed large areas of ischemia and necrosis (Figure 2C), VE-Cad ADAM10–/– mice had a markedly lesser degree of injury following infection that was also evident on histopathologic analysis of the liver (Figure 2D) and in quantification of serum alanine aminotransferase levels (Figure 2E). We did not detect differences in liver IL-1β or IL-10 levels after infection (Supplemental Figure 2, E and F), suggesting that local inflammation alone is not readily equated with the end-organ tissue damage observed.

ADAM10 alters endothelial cell–platelet interactions in response to S. aureus and contributes to sepsis-associated injury. (A) Representative 2-photon image of control or VE-Cad ADAM10–/– mouse livers 6 to 8 hours after nonlethal S. aureus sepsis. Vasculature (red, Qdots655), platelets (green, GPIbβ). Scale bars: 50 μm. Lower panels display images outlined by a dashed line box. Scale bars: 20 μm. White arrows denote thrombi. (B) Quantification of the total area of platelet accumulation within the vasculature in mouse liver as treated in A. Data represent 5–7 FOV per mouse in control (n = 7F) and VE-Cad ADAM10–/– (n = 6F) mice. Data are presented as the mean ± SEM. Data for individual FOV within each mouse are shown in Supplemental Figure 2A. (C) Representative images of the liver 24 hours after S. aureus infection in control or VE-Cad ADAM10–/– mice. (D) H&E-stained liver sections from control or VE-Cad ADAM10–/– mice 24 hours after infection, with the area of necrosis outlined. Scale bars: 100 μm. Images in C and D are representative of 5 mice per condition from 2 independent experiments. (E) Serum ALT in infected control (n = 11 males, 7 females) or VE-Cad ADAM10–/– (n = 7 males, 6 females) mice 24 hours after nonlethal S. aureus infection. Data represent 3 independent pooled experiments and are presented as the mean ± SEM. (F) Representative 2-photon images of control or VE-Cad ADAM10–/– mouse livers 2–4 hours after lethal S. aureus sepsis. Vasculature (red, Qdots655); vWF (green). Scale bars: 10 μm. White arrows denote vWF deposition. (G) Quantification of the total area of vWF accumulation within the vasculature in mouse liver as treated in F. Data represent 5–6 FOV per mouse in control (n = 6F) and VE-Cad ADAM10–/– (n = 4 females, 2 males) mice per group and represent the mean ± SEM. Data for individual FOV within each mouse are shown in Supplemental Figure 3F. *P ≤ 0.05 and **P ≤ 0.01, by nested t test for in vivo imaging (B and G) or unpaired, 2-tailed t test (E).

Our hypothesis that endothelial injury caused by Hla initiates a pathologic platelet response within the vasculature was consistent with findings in a recent study by Sun et al. demonstrating that treatment with ticagrelor, a platelet P2Y12 inhibitor, protected against S. aureus sepsis (28). Ticagrelor blocked Hla-mediated platelet cytotoxicity in an ADAM10-dependent manner in vitro, allowing for improved clearance of the pathogen. Consistent with these findings, WT mice treated with ticagrelor in our model system exhibited a significant reduction in platelet aggregation within the hepatic vasculature compared with vehicle-treated controls (Supplemental Figure 3, A and B). As ticagrelor has been shown to exhibit pleiotropic effects beyond platelet inhibition (47), we examined whether ticagrelor affected endothelial ADAM10 activation and endothelial dysfunction. Hla-induced microvascular injury leads to barrier disruption and vascular permeability (10), inducing the deposition of von Willebrand factor (vWF) on the endothelium (12). vWF is critical in hemostasis, mediating initial platelet tethering under fluid shear stress in the vasculature (48). Treatment of human pulmonary artery endothelial cells (HPAECs) with purified Hla led to increased ADAM10 activity (Supplemental Figure 3C) and LDH release indicative of cellular injury (Supplemental Figure 3D), which was abrogated by pretreatment of HPAECs with the specific ADAM10 inhibitor GI254023X. Pretreatment of HPAECs with ticagrelor did not modulate ADAM10 activation (Supplemental Figure 3C) or LDH release induced by Hla (Supplemental Figure 3D); similarly, ticagrelor did not prevent VE-cadherin cleavage or vWF extrusion in response to Hla (Supplemental Figure 3E), but these were ablated by pretreatment with GI254023X. Thus, we conclude that the effects of ticagrelor on S. aureus sepsis appeared to be focused on the modulation of platelet function, which we hypothesize is downstream of endothelial injury elicited by the pathogen.

To assess the temporal nature of endothelial injury and platelet aggregation in vivo, we used 2-photon microscopy to visualize vWF deposition within the hepatic vasculature during S. aureus sepsis in VE-Cad ADAM10–/– mice or matched controls. Within 4 hours of infection, WT mice developed increased vWF deposition compared with VE-Cad ADAM10–/– mice (Figure 2, F and G, and Supplemental Figure 3F) preceding the difference in platelet aggregation seen at 6–8 hours (Figure 2, A and B). Together, these findings suggest a sequence of events in the microvasculature in which Hla-mediated endothelial injury functioned as an inciting event, leading to the presentation of vWF on the injured endothelium as a platelet tethering site. This pathogenic sequence presents 2 distinct therapeutic opportunities to combat sepsis: protecting the endothelium from initial injury and preventing platelet aggregation that is catalyzed by this injury.

Endothelial injury is not unique to S. aureus disease, as it has been studied in the molecular pathogenesis of experimental sepsis in multiple models. The clinical manifestations of endovascular injury and progression of severe sepsis exhibit an unexpected homogeneity across distinct infectious pathogens, suggesting that pathogen-specific pathways that incite injury trigger a convergent biologic process in the host. We considered whether ADAM10 may be an upstream molecular pathway in sepsis pathogenesis, especially given the observation that an ADAM10 promoter polymorphism modulates sepsis severity (32). Three other bacterial cytotoxins have been linked to ADAM10 activation in vitro. We previously demonstrated that treatment of alveolar epithelial cells with S. pneumoniae pneumolysin (PLY), a cholesterol-dependent cytolysin, elicited ADAM10-mediated cleavage of epithelial cadherin (E-cadherin) (19). PLY is a PFT produced by nearly all clinical isolates and is required for virulence in invasive infection (49). Formation of the PLY pore on endothelial cells causes calcium influx, thereby activating phospholipase A2 (50) and stimulating vWF secretion (51). Reboud et al. demonstrated that Pseudomonas aeruginosa PFT exolysin (ExlA) and Serratia marcescens hemolysin (ShlA) trigger ADAM10 activation by promoting Ca2+ influx into the endothelial cell, leading to VE-cadherin cleavage (52). P. aeruginosa also disrupts endothelial barrier integrity through the secreted elastase LasB (53), whereas the type 3 secretion system (T3SS) effector ExoU exhibits phospholipase A2 activity, rapidly inducing membrane injury (54, 55) and triggering vWF release from endothelial cells in vitro (56). Among P. aeruginosa strains, expression of ExlA or the T3SS appear to be mutually exclusive mechanisms by which endothelial cells are injured (55, 57, 58). The ability of diverse bacterial cytotoxins to elicit endothelial injury, together with the central role of ADAM10 in endothelial homeostasis, led us to hypothesize that ADAM10 is a central mediator of sepsis-induced endothelial dysfunction. To address this, we initially focused on P. aeruginosa, S. pneumoniae, group B Streptococcus (GBS), and Candida albicans, which are leading causes of sepsis in pediatric and adult populations (59, 60). Like S. aureus, P. aeruginosa, and S. pneumoniae, both GBS and C. albicans harbor cytotoxins that cause membrane injury. GBS β-hemolysin/cytolysin is a cytolytic pigment that plays a key role in GBS pathogenesis, resulting in membrane injury in multiple cell types that govern barrier stability and the host immune response (61). Candidalysin, the PFT of C. albicans, compromises barrier integrity by uncontrolled Ca2+ influx leading to a proinflammatory cascade (62, 63). Upon infection of WT and VE-Cad ADAM10–/– mice with these pathogens, VE-Cad ADAM10–/– mice were protected against lethal sepsis caused by P. aeruginosa (Figure 3A) and S. pneumoniae (Figure 3B), but not sepsis due to GBS (Figure 3C) or C. albicans (Figure 3D). Clinical health scores during P. aeruginosa and S. pneumoniae infection mimicked those observed in VE-Cad ADAM10–/– mice during S. aureus sepsis (Supplemental Figure 4, A–D). We next assessed whether platelet aggregation within the hepatic vasculature was altered in an ADAM10-dependent manner following infection with these pathogens. In vivo imaging of VE-Cad ADAM10–/– mice infected with P. aeruginosa or S. pneumoniae revealed significantly reduced areas of platelet aggregation in the liver 6–8 hours after infection compared with control WT mice (Figure 4, A and B, and Supplemental Figure 4, E and F). In contrast, platelet thrombus formation was indistinguishable in control and VE-Cad ADAM10–/– mice after infection with GBS or C. albicans (Figure 4, A and B, and Supplemental Figure 4, G and H). To further investigate the role of endothelial ADAM10 as an inciting factor in microvascular injury in vivo following infection with P. aeruginosa or S. pneumoniae, we examined whether the use of the ADAM10 active site inhibitor GI254023X could prevent pathogen-induced platelet aggregation on the injured endothelium. C57Bl/6 mice were pretreated for a 3-day period with GI254023X via i.p. injection and then infected with either P. aeruginosa or S. pneumoniae and imaged as described above. We found that treatment with the ADAM10 inhibitor protected against platelet aggregation within the hepatic vasculature in both P. aeruginosa and S. pneumoniae sepsis in a manner similar to that observed in VE-Cad ADAM10–/– mice (Figure 5, A and B, and Supplemental Figure 5, A and B). Together, these studies provide the first in vivo evidence to our knowledge that the role of endothelial ADAM10 in the pathophysiology of sepsis extends beyond that of the known Hla-ADAM10 complex, exhibiting pathogen specificity.

Endothelial ADAM10 mediates lethal sepsis in a pathogen-specific manner. Survival curves for control and VE-Cad ADAM10–/– mice infected with lethal P. aeruginosa (A, n = 21 [13 males, 8 females], n = 20 [11 males, 8 females]); S. pneumoniae (B, n = 24 [8 males, 16 females], n = 23 [11 males, 12 females]; GBS (C, n = 23 [9 males, 14 females], n = 14 [7 males, 7 females]; or C. albicans (D, n = 12 males, 13 males ). **P ≤ 0.01, by log-rank (Mantel-Cox) test for survival curves.

Endothelial ADAM10 mediates platelet aggregation in a pathogen-specific manner. (A) Representative 2-photon images of control and VE-Cad ADAM10–/– mouse livers at 6–8 hours and (B) corresponding quantification of the total area of platelet accumulation within the vasculature of mouse livers after P. aeruginosa, S. pneumoniae, GBS, or C. albicans sepsis. Vasculature (red, Qdots655); platelets (green, GPIbβ). Scale bars: 20 μm. White arrows denote thrombi. Data represent 5–7 FOV per mouse in 5–6 male and female mice per group and indicate the mean ± SEM. Measurements for individual FOV within each mouse are displayed in Supplemental Figure 4, E–H. *P ≤ 0.05 and ***P ≤ 0.001, by nested t test.

ADAM10 mediates platelet aggregation. (A) Representative 2-photon images of C57Bl/6 mice 6–8 hours after P. aeruginosa or S. pneumoniae sepsis, pretreated for 3 days with vehicle or the ADAM10 inhibitor GI254023X. Vasculature (red, Qdots655), platelets (green, GPIbβ). Scale bar: 20 μm. White arrows denote thrombi. (B) Quantification of the total area of platelet accumulation within the vasculature in mouse livers as treated in A. Data represent 5–7 FOV per mouse in 6–7 male and female mice per group and are presented as the mean ± SEM. Measurements for individual FOV within each mouse are displayed in Supplemental Figure 5, A and B. **P ≤ 0.01, by nested t test for each treatment group.

Like S. aureus, many pathogens utilize PFTs or other membrane-injurious cytotoxins to damage and manipulate the endothelium (54, 64–66), raising the possibility that these virulence factors may couple intravascular infection with the observed dependence on ADAM10. PFTs and other membrane-active cytotoxins generally have 2 main effects during infection: the compromising of barrier integrity and disruption of the host immunological response (64, 67, 68). On the basis of our observation that microvascular occlusion in response to endothelial injury distinguished S. aureus, P. aeruginosa, and S. pneumoniae from pathogens that exhibited ADAM10 independence, we leveraged the 2-photon in vivo imaging model of platelet aggregation to examine the impact of membrane-active cytotoxins on disease pathogenesis. Upon infection of C57Bl/6 mice with WT S. aureus LAC or an isogenic Hla-deficient mutant (LACΔhla), we confirmed that platelet aggregation in response to S. aureus sepsis was dependent on Hla (Figure 6, A and D, and Supplemental Figure 5C). We then extended these studies to determine whether microvascular occlusion in P. aeruginosa and S. pneumoniae infection was dependent on cytotoxin expression. We used WT P. aeruginosa PA99 or its isogenic strain PA99ΔexoSTU, which is impaired in effector secretion without disruption of expression of the T3SS (69), and WT S. pneumoniae 6A or its isogenic PLY-deficient mutant (6AΔply). Compared with their WT strains, both toxin mutant strains elicited a reduction in platelet aggregation within the hepatic vasculature 6–8 hours after infection (Figure 6, B–D, and Supplemental Figure 5, D and E). Together, these data suggest that specific bacterial cytotoxins may either initiate or amplify ADAM10-dependent endothelial injury, resulting in platelet aggregation and microvascular thrombosis in the pathogenesis of sepsis.

Bacterial cytotoxins mediate sepsis-induced platelet aggregation in the liver. Representative 2-photon images of C57Bl/6 mice 6–8 hours after sepsis with (A) S. aureus LAC or LACΔhla, (B) P. aeruginosa PA99 or PA99ΔexoSTU, and (C) S. pneumoniae 6A or 6AΔply. Vasculature (red, Qdots655); platelets (green, GPIbβ). Scale bars: 20 μm. White arrows denote thrombi. (D) Quantification of the total area of platelet accumulation within the vasculature in mouse livers as treated in A–C. Data represent 5–7 FOV per mouse in 5–7 female mice per group and are presented as the mean ± SEM. Measurements for individual FOV within each mouse are displayed in Supplemental Figure 5, C–E. *P ≤ 0.05, **P ≤ 0.01, and ****P ≤ 0.0001, by nested t test for each treatment group.