MARTX toxin of Vibrio vulnificus induces RBC phosphatidylserine exposure that can contribute to thrombosis

V. vulnificus-infected patients suffer from hemolytic anemia and circulatory lesions, often accompanied by venous thrombosis. However, the pathophysiological mechanism of venous thrombosis associated with V. vulnificus infection remains largely unknown. Herein, V. vulnificus infection at the sub-hemolytic level induced shape change of human red blood cells (RBCs) accompanied by phosphatidylserine exposure, and microvesicle generation, leading to the procoagulant activation of RBCs and ultimately, acquisition of prothrombotic activity. Of note, V. vulnificus exposed to RBCs substantially upregulated the rtxA gene encoding multifunctional autoprocessing repeats-in-toxin (MARTX) toxin. Mutant studies showed that V. vulnificus-induced RBC procoagulant activity was due to the pore forming region of the MARTX toxin causing intracellular Ca2+ influx in RBCs. In a rat venous thrombosis model triggered by tissue factor and stasis, the V. vulnificus wild type increased thrombosis while the ΔrtxA mutant failed to increase thrombosis, confirming that V. vulnificus induces thrombosis through the procoagulant activation of RBCs via the mediation of the MARTX toxin.

Infection of V. vulnificus, an opportunistic human pathogen and a causative agent of foodborne disease, is characterized by rapid disease progression leading to hemolysis and fatal septicemia with high mortality rates within a few days 1,2 . The pathogenesis of V. vulnificus infection is also often accompanied by circulatory disorders such as venous thrombosis, which brings about tissue destruction and limb amputation [3][4][5] . These pathological features imply that V. vulnificus infection may induce intravascular coagulation 6 , which causes blood clotting and blockade of small blood vessels. Also, the pathological features suggest that the prevention of venous thrombosis is an attractive strategy to alleviate the symptoms of V. vulnificus infection. However, little is known about the virulence factor(s) of V. vulnificus responsible for the development of venous thrombosis and its molecular mechanisms.
Normally, thrombosis, such as arterial thrombosis, occurs via platelet activation under high blood flow 7 . However, recent studies demonstrated that thrombosis, especially in the vein, also occurs through the increased procoagulant activity of human RBCs under low blood flow 8,9 . Procoagulant activity is often increased along with the elevation of the intracellular calcium [Ca 2+ ] i level, and caspase-3 and scramblase activities by various external stimuli such as mechanical and chemical stress, resulting in the externalization of an anionic phospholipid, phosphatidylserine (PS) of RBCs 9,10 . Along with PS exposure, RBCs often shed the PS-bearing microvesicles (MVs; less than 1 μm in diameter) by exocytosis 11 and undergo progressive morphological changes from normal discocytes to echinocytes 12 . The PS exposure of RBCs provides sites for the assembly of prothrombinase and tenase complexes, which facilitates thrombin generation and enhances prothrombotic activity to prompt blood clotting and, ultimately, venous thrombosis 13 .
Among multiple virulence factors of V. vulnificus, the multifunctional autoprocessing repeats-in-toxin (MARTX) toxin, which is encoded by the rtxA gene, is the most crucial exotoxin responsible for the pathogenicity of V. vulnificus 14,15 . As a very large protein toxin, the MARTX toxin forms a pore on the host cell membrane using its repeats-containing regions and then delivers its own multiple cytotoxic or cytopathic effector domains into the cytosol of epithelial cells 16 . A mutant V. vulnificus defective in the MARTX toxin production was less cytotoxic and lethal when compared with the isogenic wild type (WT) 14,15 . However, it remains unelucidated whether the MARTX toxin also plays a role in venous thrombosis caused by V. vulnificus infection.
Here, we show that V. vulnificus induces shape changes, PS exposure, and MV generation in RBCs, along with the elevation of the [Ca 2+ ] i level, ultimately leading to the procoagulant and prothrombotic activity of RBCs. Especially, V. vulnificus infecting RBCs substantially upregulated rtxA, illuminating that the pore-forming activity of the MARTX toxin induced the cytopathic changes of RBCs. Finally, we conducted rat model studies in vivo to confirm the role of the MARTX toxin in V. vulnificus-associated thrombosis.

Morphological changes of RBCs infected with V. vulnificus at the sub-hemolytic level
To examine the possible changes of human red blood cells (RBCs) before hemolysis, an optimal titer and incubation time for subhemolytic infection of V. vulnificus were screened. The RBCs freshly isolated from healthy volunteers were infected with V. vulnificus MO6-24/O (Supplementary Table 1), and then the multiplicity of infection (MOI)-dependent and incubation time-dependent hemolysis was observed (Fig. 1a, b). When incubated for 1 h, V. vulnificus induced only minimal hemolysis (<20%) at an MOI lower than 10, but severe hemolysis (up to 80%) was observed at an MOI of 25 (Fig. 1a). Similarly, V. vulnificus lysed less than 20% of RBCs at an MOI of 5 or 10 and incubation for up to 1 h (Fig. 1b). Therefore, infections of RBCs with V. vulnificus at an MOI up to 10  were defined as the sub-hemolytic level infections in the following experiments. Interestingly, V. vulnificus (red arrowheads) infection at the subhemolytic level induced the morphological changes of the RBCs (discocyte) as analyzed with scanning electron microscopy (SEM) (Fig. 1c) and transmission electron microscopy (TEM) ( Supplementary  Fig. 1). As the MOI increased up to 10, echinocytes (white arrows) characterized by a shell-like appearance with evenly spaced thorny projections, followed by spherocytes (white arrowheads) with a smallsized round shape, were observed (Fig. 1c, d). The morphological changes of discocyte infected with V. vulnificus into echinocyte and then spherocyte were further confirmed with confocal microscopy (Fig. 1e). The morphological changes of RBCs infected with V. vulnificus into the echinocytes accompanied with numerous MVs generate spherocytes (Fig. 1f) (Fig. 2a, bottom). Confocal microscopy analysis demonstrated that the uninfected RBCs were not stained with Annexin V specific for the exposed PS (Fig. 2b). However, the outer leaflet of the echinocyte and spherocyte membranes were positively stained (Fig. 2b), confirming PS exposure of RBCs upon V. vulnificus infection. Notably, infection of RBCs with V. vulnificus increased MV generation and PS exposure not only in an MOI-dependent but also in an incubation time-dependent manner ( Supplementary Fig. 2a, b). The results confirmed that the morphological changes of RBCs infected with V. vulnificus into spherocytes at the sub-hemolytic level are accompanied by MV generation and PS exposure.

Procoagulant activity of RBCs
PS exposure of RBCs under mechanical and chemical stress is often accompanied by procoagulant activity 9 . Interestingly, RBCs infected with V. vulnificus also revealed procoagulant activity along with the elevation of the intracellular Ca 2+ [Ca 2+ ] i level (Fig. 2c), Ca 2+ -dependent caspase-3 activity (Fig. 2d), and then scramblase activity (Fig. 2e) in an MOI-dependent manner. This result indicated that the increased PS exposure of the RBCs, observed upon infection with V. vulnificus (Fig. 2a), is also accompanied by increased procoagulant activity. To confirm this relationship between PS exposure and procoagulant activity, RBCs were pretreated with specific inhibitors of caspase-3, Z-DEVD-FMK (VI), and Q-VD-OPh (Q) before being infected with V. vulnificus. When infected with V. vulnificus, the pretreated RBCs revealed significantly reduced PS exposure and procoagulant activity (Fig. 2f), confirming that PS exposure is caused by the procoagulant activity of RBCs.

Prothrombotic activity of RBCs
The increased procoagulant activity could result in functional alterations of RBCs, such as promoting their prothrombotic activity, including thrombin generation, adherence to endothelial cells, and self-aggregation 9,13,17,18 . When infected with V. vulnificus, RBCs increased thrombin generation in an MOI-dependent and an incubation time-dependent manner (Fig. 2g). When the exposed PS in RBCs was blocked with purified Annexin V, the procoagulant activity of V. vulnificus was attenuated significantly, suggesting that PS exposure plays a key role in V. vulnificus-induced procoagulant activity (Supplementary Fig. 3). RBCs were pre-infected with V. vulnificus at different MOIs, and then their adherence and aggregation to human umbilical vein endothelial cells (ECs) were analyzed by fluorescence microscopy. Notably, the RBCs pre-infected with more V. vulnificus (MOI of 5 to 10) revealed more adherence (white arrows) and aggregation (yellow arrowheads) to ECs (Fig. 2h). Indeed, V. vulnificus infection led RBCs to aggregate themselves in an MOI-dependent manner even in the absence of ECs (Fig. 2i). The combined results suggest that V. vulnificus infection at the sub-hemolytic level results in the morphological changes of RBCs, with increased procoagulant activity leading to increased prothrombotic activity.

MARTX toxin responsible for the functional changes of RBCs
Since V. vulnificus infection increased procoagulant activity (Fig. 2a-f and Supplementary Fig. 2a, b), it is reasonable to assume that the V. vulnificus virulence factor(s) responsible for procoagulant activity could be specifically induced upon exposure to RBCs. Indeed, the sequencing analyses of the whole transcriptome of V. vulnificus unexposed or exposed to RBCs identified a total of 121 differentially expressed genes (DEGs; with |log 2 fold change| ≥1.0, and P value ≤0.05) ( Fig. 3a and Supplementary Table 2). Most of the 36 genes upregulated are involved in the transport and metabolisms of various nutrients, including iron. Interestingly, the only upregulated exotoxin gene is rtxA (VVMO6_03947, log 2 fold change = 1.20, P value = 2.72E-43) encoding the MARTX toxin. To examine the effects of the MARTX toxin on procoagulant activity, effects of the rtxA deletion in V. vulnificus on the changes of PS exposure and MV generation of the RBCs were primarily characterized. The ΔrtxA mutant showed significantly reduced PS exposure and MV generation (Fig. 3b, c). Furthermore, the ΔrtxA mutant induced significantly less elevation of the [Ca 2+ ] i level, scramblase activity, and thrombin generation in the RBCs than the parental WT ( Fig. 3d-f). Notably, another rtxA mutant strain (rtxA::nptI), of which the rtxA gene was inactivated by insertional rather than deletional mutation, revealed similarly attenuated procoagulant activity of RBCs ( Supplementary Fig. 4a-c). Importantly, the attenuation was recovered in a revertant strain, indicating that the effects of the rtxA mutation on the cytopathic changes of RBCs were not due to the polar effects ( Supplementary  Fig. 4b, c). These results confirmed that the MARTX toxin plays a crucial role in the increase of the procoagulant activity of RBCs infected with V. vulnificus.

Pore-forming regions of MARTX toxin responsible for the functional changes of RBCs
To further identify functional domains of the multifunctional MARTX toxin responsible for the increased procoagulant activity of RBCs, an effector-free (EF) V. vulnificus mutant (EF-rtxA) was constructed (Fig. 3g). Of note, the EF-rtxA strain produces an EF-MARTX toxin that does not carry the effector domains but is still able to generate a pore on the intoxicated host cell membrane. The pore-forming activity of the EF-rtxA strain was proved by the bleb formation in human epithelial cells ( Supplementary Fig. 4d), as reported previously 19 . When infected with either the V. vulnificus WT or EF-rtxA strains, RBCs revealed no statistical differences in the changes of PS exposure, [Ca 2+ ] i level, and thrombin generation ( Fig. 3h-j). Altogether, the results indicate that the pore-forming regions, not the effector domains, of the MARTX toxin of V. vulnificus are responsible for the cytopathic changes of RBCs with their enhanced procoagulant activity.

Venous thrombosis of rat
Prior to confirming the effects of the MARTX toxin on the RBCs in vivo, the changes in PS exposure and thrombin generation in rat RBCs in response to V. vulnificus infection in vitro were examined as bridge studies. The rat RBCs infected with the sub-hemolytic level of V. vulnificus (MOI 1, 5, or 10) in vitro increased PS exposure and thrombin generation in an MOI-dependent manner (Fig. 4a, b), which are similar to the changes observed in human RBCs (Fig. 2a, g). Next, rats were intravenously (IV) infected with either the V. vulnificus WT or ΔrtxA for 1 h, and then thrombin generation from the whole blood cells ex vivo were examined. The thrombin generation level of the rats infected with  the ΔrtxA mutant was lower than that infected with the WT, even close to that uninfected (Fig. 4c), indicating that the MARTX toxin is the virulence factor responsible for thrombin generation from the whole blood cells of the rats. In order to observe the morphological change of RBCs following V. vulnificus infection in vivo, blood was collected 1 h after IV injection of WT and ΔrtxA mutant to rats. Representative image of RBCs was observed under scanning electron microscopy (Fig. 4d). Consistently with in vitro results (Fig. 1c, e), administration of WT induced morphological changes of RBCs into echinocytes (white arrows), implying that PS-bearing microvesicles were formed from RBCs in vivo following V. vulnificus infection. On the other hand, injection of the ΔrtxA mutant did not change the shape compared to the untreated control. Finally, effects of V. vulnificus infection on thrombus formation in vivo were examined using a rat venous thrombosis model (Fig. 4e). When rats were IV infected with the V. vulnificus WT, and then injected with the thromboplastin, thrombus formation increased in an incubation time-dependent and an infectious dose-dependent manner (Fig. 4f, g). In contrast, rats infected  with the deletion mutant (ΔrtxA) as well as the insertion mutant (rtxA::nptI) did not show any significant increase in thrombus formation when compared with the uninfected control. Furthermore, the revertant of the insertion mutant (rtxA::nptI) rescued the ability of V. vulnificus WT to increase thrombus formation (Fig. 4f, g), as consistent with in vitro data ( Supplementary Fig. 4b, c). These results support our conclusion that the MARTX toxin of V. vulnificus is mainly responsible for the procoagulant activity of RBCs and thrombus formation in vivo.

Discussion
Various pathogens can cause venous thrombosis 3,4,6,20,21 and severe hemolytic anemia 22,23 as reported in many clinical case reports. However, studies on the morphological change of human RBCs caused by these pathogens or its implication in venous thrombosis have not been reported. Here, we demonstrated that V. vulnificus infection at the subhemolytic level induces remarkable morphological changes in RBCs from normal discocytes to echinocytes and further to the spherocytes (Fig. 1c-f). Along with the morphological changes of RBCs, the were determined. c Rats were infected IV with either V. vulnificus WT or ΔrtxA mutant (10 9 CFU/Rat) for 1 h, and then whole blood cells were collected and used to determine thrombin generation ex vivo (n = 5 in control, n = 5 in WT, n = 7 in ΔrtxA mutant, two-tailed Student's t-test). d Rats were infected IV with either V. vulnificus WT or ΔrtxA mutant (10 8 CFU/Rat) for 1 h, and then whole blood cells were collected and removed platelet-rich plasma and buffy coat to observe RBCs morphology using SEM. The representative images by SEM images among five independent experiments were presented. The white arrow indicates echinocyte. The scale bar is presented. e A schematic diagram presents the isolation of thrombus formed in 16 mm of inferior vena cava (IVC) exposed after 1 h IV infection with V. vulnificus and subsequent injection with thromboplastin. f Rats were infected IV with V. vulnificus WT, ΔrtxA, rtxA::nptI mutant, or revertant strain (10 7 CFU/Rat) as described above and then incubated for different times as indicated, and then thrombus formed in the IVC were analyzed by stereomicroscopy. The scale bar is presented. g Rats were infected IV with V. vulnificus WT, ΔrtxA, rtxA::nptI mutant, or revertant strain for 1 h at various doses as indicated, and then weights of the formed thrombus in the IVC were determined (n = 5, two-tailed Student's t-test). The means ± SE were calculated from at least five independent experiments. WT wild type, ΔrtxA a mutant producing no MARTX toxin, rtxA::nptI a mutant producing no MARTX toxin, nptI aminoglycoside 3'-phosphotransferase gene, Rev a revertant producing the WT-MARTX toxin, SEM scanning electron microscopy, control, uninfected, MOI multiplicity of infection, and IVC inferior vena cava. procoagulant activity of RBCs enhances PS exposure and MV generation, ultimately promoting the risk of venous thrombosis. This was fully demonstrated by our in vivo study, where normal healthy rats infected with V. vulnificus showed increased thrombin generation of RBCs and thrombus formation while those infected with the rtxA mutants did not show any increase (Fig. 4c, f, g). Moreover, the revertant of the rtxA insertion mutant (rtxA::nptI) rescued the ability of V. vulnificus WT to increase thrombus formation (Fig. 4f, g), confirming that the MARTX toxin of V. vulnificus is mainly responsible for the procoagulant activity of RBCs and thrombus formation. We also elucidated the biochemical and molecular mechanism underlying the procoagulant activity and morphological changes of the RBCs infected with V. vulnificus. V. vulnificus infection at the subhemolytic level elevates the [Ca 2+ ] i level with increased caspase-3 and scramblase activities accompanied by PS exposure and MV generation of RBCs (Fig. 2a-e). These molecular alterations suggest that the elevation of the [Ca 2+ ] i level of the RBCs infected with V. vulnificus is key to the procoagulant activity and morphological changes of RBCs, which ultimately enhances the prothrombotic activity (Fig. 2g-i). Although the previous studies reported that the elevation of the [Ca 2+ ] i level in RBCs is closely linked with thrombotic complications and hemolytic anemia caused by various external stress 24 , the elevation of the [Ca 2+ ] i level in RBCs mediated by a pathogen infection has never been reported to the best of our knowledge.
While purified bacterial toxins listeriolysin O (LLO) and alphahemolysin (HlyA) were also observed to induce the elevation of the [Ca 2+ ] i level in RBCs in vitro 25,26 , their effects on the procoagulant and prothrombotic activity and subsequent thrombosis in vivo have not been addressed. Mutational analysis revealed that the MARTX toxin of V. vulnificus was responsible for the elevation of the [Ca 2+ ] i level, leading to the procoagulant activity of RBCs and thrombin generation (Fig. 3b-f). Notably, further functional dissection analysis revealed that the pore-forming activity of the EF-MARTX toxin, but not the multiple cytotoxic or cytopathic effector functions, is attributed to the elevation of the [Ca 2+ ] i level, which triggers PS exposure of RBCs and thrombin generation (Fig. 3g-j). Previous studies reported that the exposure of RBCs to exogenous and endogenous stimuli like calcium ionophore A23187 and lysophosphatidic acid (LPA) induces a morphological change from discocyte into echinocytes mediated through the elevation of [Ca 2+ ] i level 11,27 . These results suggest that the poreforming activity of the EF-MARTX toxin may be responsible for the morphological change of RBCs from normal discocyte into pathologic echinocyte through the elevation of the [Ca 2+ ] i level.
It has been previously reported that HlyU, a transcription regulator, upregulates the rtxA encoding MARTX toxin by directly binding to its promoter 28 . Interestingly, this study revealed that the expression of rtxA was significantly increased following the exposure of V. vulnificus to RBCs (Fig. 3a and Supplementary Table 2), supporting the previous report that direct contact of V. vulnificus to host cells is required to display MARTX cytotoxicity 29 . However, the transcriptomic analysis showed that, unlike rtxA, hlyU encoding HlyU was not upregulated in V. vulnificus infecting RBCs (Supplementary Table 2), indicating that other mechanism(s), yet unknown, could be involved in the upregulation of rtxA following exposure of V. vulnificus to RBCs. In addition to the rtxA gene, several genes related to iron uptake (Supplementary Table 2) were also significantly upregulated in V. vulnificus exposed to RBCs, along with the generation of numerous MVs. Recent reports suggest that MVs derived from RBCs retain residual hemoglobin as an important resource of iron 30 . Incidentally, iron is essential for the growth and lethality of V. vulnificus 31 . Our results on the upregulation of iron uptake-related genes of V. vulnificus combined with the numerous MVs generation from V. vulnificus-infected RBCs imply that increased availability of iron from the MVs may promote the growth and lethality of V. vulnificus, which may be crucial to the manifestation of its pathogenicity.
V. vulnificus infection may accompany circulatory disorders from blood clotting and blockade of the venous vessels leading to deep vein thrombosis 3,5,21,32,33 , highlighting the importance of thrombosis for the pathological complications of V. vulnificus infection. However, the mechanism underlying V. vulnificus infection-associated thrombosis has not been fully elucidated. Here, we investigated the role of RBCs and MARTX toxin in coagulation and thrombosis associated with V. vulnificus infection. A pivotal role of RBCs in hemostasis and thrombosis has been well explained in recent reviews 8,9,34 . Several mechanisms were suggested for the involvement of RBCs in the promotion of blood coagulation and thrombosis. Previous studies suggested that RBC aggregation may aggravate deep vein thrombosis by increasing the hydrodynamic resistance in the veins in the lower limbs 9,18 . However, it is yet to be clarified whether V. vulnificus-induced RBC aggregation alone is sufficient to trigger venous thrombosis. Attachment of RBCs to vascular endothelium also plays a role in hemostasis and thrombosis. Unlike normal RBCs, RBCs are prone to attach to endothelium under certain pathological conditions such as sickle cell disease 35 . It has been suggested that PS exposure in RBCs was linked to increased adhesion to endothelium 17,36 . Consistently with this notion, our in vitro results demonstrated that PS-exposing RBCs following V. vulnificus infection increased endothelial cell attachment (Fig. 2h) and RBC aggregation (Fig. 2i), indicating that these events may promote the vascular occlusions associated with venous thrombosis observed in the V. vulnificus-infected patients.
More attention has recently been paid to the role of PS exposure of RBCs in coagulation and thrombosis 8,9,34 . Under pathological states, negatively charged PS in the surface of the RBC membrane provides sites for the assembly of prothrombinase complex with factor Xa and factor Va, which facilitates thrombin generation from prothrombin to enhance blood clotting and ultimately promote venous thrombosis 13 . In accordance with this downstream component of coagulation, our in vitro results showed that the PS-exposing RBCs and MVs occurring upon V. vulnificus infection increase thrombin generation by exogenous addition of factor Xa, and factor Va (Fig. 2g), suggesting that PSexposing RBCs accelerate the downstream component of coagulation cascade to promote coagulation in vivo even though coagulation is not triggered by them. Previous studies reported that V. vulnificus infection secretes a metalloprotease to activate factor XII, that involves the first step of the intrinsic pathway [37][38][39] . In addition, there are several reports that infection with bacteria such as Staphylococcus aureus initiates the extrinsic coagulation pathway by inducing tissue factor production from monocytes and endothelium via inflammatory mediators such as TNF-alpha [40][41][42] . There is no report on the increase of tissue factor expression by V. vulnificus infection, but the levels of TNFalpha were significantly increased in V. vulnificus-infected patients [43][44][45] . It is possible that V. vulnificus infection may induce the expression of tissue factors on the surface of cells such as monocytes by the increased TNF-alpha, allowing the complex formation with factor VII, finally activating Factor X. Collectively, V. vulnificus infection may induce inflammation and activate the upstream coagulation factors such as tissue factor and factor XII in extrinsic/intrinsic pathway, which culminates in the formation of factors Xa/Va. These factors Xa/Va readily form the prothrombinase complex in the presence of PSexposing RBCs resulting in increased generation of thrombin from prothrombin, ultimately forming fibrin clot.
In summary, we discovered that V. vulnificus infection induces the morphological and functional changes of RBCs, ultimately leading to thrombosis (Fig. 5). This study is the first to demonstrate molecular mechanisms of bacterial infection leading to the thrombotic complications through inducing the procoagulant and prothrombotic activity of RBCs. We believe that our study provides an important insight into understanding pathogen-induced thrombosis at molecular levels, which could help to develop innovative strategies to control the intravascular coagulation caused by pathogens.

Strains, plasmids, culture conditions, and mutant construction
The strains and plasmids used in this study are listed in Table S1 14,46-50 . A clinical isolate of V. vulnificus MO6-24/O 46 and its three isogenic mutants were used in this study. The ΔrtxA mutant, which was constructed by deleting 89% of the rtxA structural gene and then inserting a nptI gene cassette conferring resistance to kanamycin into the deleted rtxA, produces no MARTX toxin 14 . The rtxA::nptI mutant with the insertion of nptI at the 5′-end of rtxA also produces no MARTX toxin 47 . The revertant, in which the rtxA structural gene was restored from the rtxA::nptI mutation, produces the WT-MARTX toxin 47 . To construct the EF-rtxA mutant, the pDS_EF-MARTX 50 , carrying rtxA with the deletion of 34% structural gene encoding effector domains of the MARTX toxin, was conjugally transferred into the V. vulnificus WT and a resulting mutant by the double homologous recombination was screened out. Desired mutation and reversion were validated by sequencing the rtxA gene. Growths of the V. vulnificus strains were obtained in a Luria-Bertani medium supplemented with 2% NaCl (w/v) (LBS) and appropriate antibiotics at 30°C and monitored by measuring A 600 of the cultures.

Preparation of human RBCs
With the approval from the Ethics Committee of the Health Service Center at Seoul National University (IRB No. 1909/001-007), all participants provided written informed consent before enrollment for the study. Human blood was obtained from healthy male donors (20-30 years old) using a vacutainer with acid citrate dextrose (ACD) and a 21-gauge needle (Becton Dickinson) on the day of each experiment. Platelet-rich plasma and buffy coat were removed by aspiration after centrifugation at 200×g for 15 min. Packed human RBCs were washed two times with phosphate-buffered saline (PBS; 1.06 mM KH 2 PO 4 , 154 mM NaCl, 2.96 mM, and Na 2 HPO 4 at pH 7.4) and once more with Ringer's solution (125 mM NaCl, 5 mM KCl, 1 mM MgSO 4 , 32 mM HEPES, and 5 mM glucose, pH 7.4). The washed RBCs were resuspended in Ringer's solution to a concentration of 1 × 10 8 cells/ml, and CaCl 2 was added to a final concentration of 1 mM prior to use for further experiments.

Detection of hemolytic activity
The RBCs resuspension infected with V. vulnificus at various MOIs and incubated for different incubation times was centrifuged at 10,000×g for 2 min. The extent of hemolysis of the RBCs was determined by measuring released hemoglobin spectrophotometrically at 540 nm (A 540 ). The complete lysis of RBCs by 1% Triton X-100 (Sigma-Aldrich) was expressed as 100% hemolysis.

Microscopy analyses
For transmission electron microscopy (TEM) analysis, the RBCs were infected with V. vulnificus for 1 h and then fixed with 2% glutaraldehyde solution in the refrigerator overnight. On the second day, the fixed RBCs were washed three times with PBS and then post-fixed with 1% osmium tetroxide for 30   Fluorescence data from 5,000 events were collected and analyzed using Cell Quest Pro software v.6.0. When required, the RBCs were pretreated with caspase-3 inhibitors (Z-DEVD-FMK and Q-VD-OPh (Calbiochem)) before being infected with V. vulnificus. A gate for PS exposure-positive events was set using the analyzed data from RBCs treated with 2.5 mM of CaCl 2 instead of 2.5 mM EDTA before the flow cytometry analysis. Data from 5000 events were collected and analyzed using Cell Quest Pro software v.6.0. For the detection of the intracellular Ca 2+ level, the RBCs were pretreated with 3 μM of Fluo-4 acetoxymethyl ester (fluo-4 AM) (Thermo Fisher Scientific) for 1 h at 37°C in the dark were infected with V. vulnificus and then the fluorescence of Fluo-4 AM was analyzed by flow cytometry. For the evaluation of caspase-3 activity, three hundred microliters of the RBCs infected with V. vulnificus were mixed with 1 μL of the in situ fluorescent marker for caspase-3 (FITC-DEVD-FMK (Calbiochem)) and then incubated for 30 min in a thermomixer (37°C, 1000 rpm, dark). Then, the RBCs were centrifuged (1000×g for 5 min), washed twice with Ringer's solution, and resuspended in 500 μL of Ringer's solution. The fluorescence from the RBC suspension was analyzed by flow cytometry. Data from 10,000 events were collected and analyzed. For the assessment of scramblase activity, 500 μL of the RBCs were infected with V. vulnificus in the presence of 0.5 μL of fluorescent phosphatidylcholine (PC) (C6-NBD-PC (Avanti Polar Lipids)) for 0, 10, and 30 min at 37°C. Then, 50 μL of aliquots were transferred to tubes containing 450 μL of ice-cold Ringer's solution either with or without 1% BSA and further incubated for 10 min. Because the PCs associated with the outer leaflet of the RBC membrane would be back-extracted by the BSA, the degree of PC translocation into the inner leaflet of the RBC membrane was measured by comparing the fluorescence intensity of the cells before (without 1% BSA) and after (with 1% BSA) backextraction. Data from 5000 events were collected and analyzed using Cell Quest Pro software v.6.0. Percentage of PC translocation was calculated by the following: PC translocation (%) = (fluorescence intensity after back-extraction/fluorescence intensity before backextraction) × 100.

Prothrombinase assay
The RBCs were infected with V. vulnificus, washed with Ringer's solution with 1 mM of CaCl 2 , and then incubated with 5 nM of factor Xa (Hematologic Technologies) and 10 nM of factor Va (Hematologic Technologies) in Tyrode buffer (134 mM NaCl, 10 mM HEPES (Sigma-Aldrich), 5 mM glucose, 2.9 mM KCl, 1 mM MgCl 2 , 12 mM NaHCO 3 , 0.34 mM Na 2 HPO 4 , 0.3% BSA, and 2 mM CaCl 2 at pH 7.4) for 3 min at 37°C. After the addition of 2 μM of purified human prothrombin (factor II) (Hematologic Technologies) to the suspension for 3 min, a 10 μl aliquot was transferred to a tube containing 490 μl of stop buffer (50 mM Tris-HCl, 120 mM NaCl, and 2 mM EDTA at pH 7.9). Thrombin generation was determined by measuring A 405 and 1 μmol thrombin generation per min is defined as a unit (U) per 10 6 RBCs in vitro and per μL whole blood cells ex vivo 11 . In experiments using purified Annexin V (BD Bioscience, 556416, final concentration 2 uM) 51 , human RBCs were pre-incubated with the Annexin V for 10 min, and then infected with V. vulnificus at 5, and 10 MOIs for 30 min. Thrombin generation was determined using the prothrombinase assay as mentioned above.

Adherence and aggregation of RBCs to EC and self-aggregation
Human umbilical vein endothelial cells were originally purchased from Lonza (C2517A). The RBCs pre-infected with V. vulnificus for 30 min were moved to human umbilical vein endothelial cells (ECs, 2 × 10 4 cells) pre-stained with Calcein green (Invitrogen) and then incubated at 37°C for 30 min. The RBCs pre-infected but not adhered to ECs were washed out using the EC growth medium (EBM-2 (Lonza)). The RBCs pre-infected and adhered to ECs were stained with the antiglycophorin-A-PE (BD Bioscience, 555570, 1:100 dilution) and then analyzed by fluorescence microscopy (Axiovert 200 M, Zeiss). To observe self-aggregation, the RBCs were infected with V. vulnificus, stained with the anti-glycophorin-A-PE (BD Bioscience, 555570, 1:100 dilution), and then analyzed by fluorescence microscopy.

Transcriptome analysis
For transcriptome analysis, the freshly prepared RBCs were infected with V. vulnificus MO6-24/O WT at an MOI of 10. After 30 min incubation at 37°C, total RNAs from the V. vulnificus were isolated 52 . Libraries of total RNAs were prepared with the TruSeq stranded mRNA library prep kit (Illumina) according to the manufacturer's protocol and then sequenced on the Illumina NovaSeq6000 platform. The sequencing reads were mapped to the complete reference genome of V. vulnificus MO6-24/O (GenBank accession numbers: CP002469 and CP002470, https://www.ncbi.nlm.nih.gov/), and the expression level of each gene was calculated as a relative log expression (RLE) value.
The RLE values were normalized and analyzed statistically using DESeq2 v.1.26.0 to identify differentially expressed genes (P < 0.05) when exposed to the RBCs.

Animal experiments
All the protocols used for animal experiments were approved by the Ethics Committee of the Animal Service Center at Seoul National University (SNU-170417- . For bridge studies, the rat RBCs were prepared from the whole blood (citrate-dextrose treated) withdrawn from the abdominal aorta of male Sprague-Dawley (SD) rats (280-350 g, 8 weeks old) anesthetized with urethane (1.25 g/kg, intraperitoneal injection) following the method used for the human RBC preparation, and then, PS exposure and thrombin generation were determined as mentioned above. For ex vivo studies, whole blood cells were collected from the abdominal aorta of the rats after 1 h IV infection with V. vulnificus and then, thrombin generation was determined using the prothrombinase assay as mentioned above. For in vivo RBC morphological studies, rats were infected IV with either V. vulnificus WT or ΔrtxA mutant for 1 h, and then whole blood cells were collected. RBCs were prepared after centrifugation at 5800×g for 1 min followed by removal of platelet-rich plasma and buffy coat. RBC morphology was analyzed using SEM as mentioned above. For in vivo thrombus formation studies, a venous thrombosis rat model was used, based upon the previous report that thrombus formed in this model was due to RBCs and fibrin clots with only a few platelets 53 . The rat abdomen was surgically opened, and the inferior vena cava (IVC) was exposed after 1 h IV infection with V. vulnificus. Both ends of the IVC, about 16 mm apart from each other, were tied loosely. However, all side branches were tied tightly with cotton threads. Thromboplastin (Instrumentation Laboratory) was infused into the IVC for 1 min to initiate thrombus formation. Upon tightening the cotton threads at the ends of the IVC, the abdominal cavity was provisionally closed to maintain the blood stasis for 15 min. The tied IVC segments were excised and opened longitudinally to isolate the thrombus, of which weights were immediately determined. The thrombus was photographed at a 1× magnification using a Stemi 305 stereomicroscope (Zeiss) equipped with an Axiocam 105 color camera (Zeiss).

Statistical analysis
The means and standard errors (SE) of means were calculated from at least five independent experiments. The data were subjected to a twoway analysis of variance (ANOVA) followed by Duncan's multiple range test or two-tailed Student's t-test to determine statistical significance when compared with the control using SigmaPlot v.12.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.