RTX-family toxin EhxA drives morphological remodeling and thrombogenesis in RBCs during enterohemorrhagic Escherichia coli infection

Abstract

Enterohemorrhagic Escherichia coli (EHEC) causes thrombotic microangiopathy, yet the red blood cell (RBC)–centered mechanism has remained unclear. We identify the RTX-family hemolysin EhxA as the driver of RBC-mediated thrombogenesis. Deletion of ehxA abolishes Ca²⁺ influx, phosphatidylserine (PS) exposure, progression from discocyte to echinocyte to spherocyte, thrombin generation, RBC-endothelium adhesion, and RBC aggregation. Genetic complementation restores these readouts to wild type, and purified EhxA in bacteria-free assays recapitulates them while localizing to intact RBC membranes. By contrast, Δstx2 mutants do not elicit these RBC phenotypes, distinguishing this pathway from Shiga toxin–dependent effects. Multiple regression quantifies the link between PS exposure, morphology, and procoagulant outputs. In rats, infection with wild type increased RBC remodeling and venous thrombosis, whereas infection with ΔehxA did not. Together, the data define an EhxA-Ca²⁺-PS pathway that drives RBC structural remodeling and procoagulant activation during EHEC infection and nominate RTX toxins as targets for preventing toxin-induced coagulopathies.

EhxA toxin uncovers RBC-driven thrombosis independent of Shiga toxin, redefining bacterial pathogenicity.

INTRODUCTION

Enterohemorrhagic Escherichia coli (EHEC) is a highly virulent foodborne pathogen that causes severe gastrointestinal infections and systemic complications, representing a major global health burden (1, 2). Among the various serotypes, E. coli O157:H7 is the most well characterized (3, 4). However, non-O157 serotypes also play an important role in disease outbreaks (5), particularly affecting vulnerable populations, including children, the elderly, and immunocompromised individuals (6, 7). A life-threatening complication of EHEC infection is hemolytic uremic syndrome (HUS), which develops in 5 to 15% of cases and rises to 20 to 30% during outbreaks, with a high mortality rate of 1.35%, resulting in 54 deaths among nearly 4000 cases (8, 9). HUS is a thrombotic microangiopathy characterized by microangiopathic hemolytic anemia and acute kidney injury, primarily caused by renal endothelial damage induced by Shiga toxin (Stx) (10, 11).

Stx, the hallmark virulence factor of EHEC, is synthesized in the intestines by EHEC during infection (12, 13) and after entering the bloodstream, Stx binds to globotriaosylceramide (Gb3) receptors on renal endothelial cells (ECs) (14, 15). Once internalized, Stx inhibits protein synthesis by cleaving ribosomal RNA, resulting in endothelial apoptosis and vascular dysfunction (14, 16). The resulting endothelial injury induces microvascular thrombosis, exacerbating renal damage (1, 17), which can further increase the risk of red blood cell (RBC) disintegration and hemolysis (18, 19). While the role of Stx in renal endothelial dysfunction is well established (20, 21), its direct impact on RBCs and its contribution to thrombogenesis remain unclear.

Recent studies suggest that RBCs, traditionally recognized primarily for oxygen transport, actively contribute to coagulation and thrombogenesis under pathophysiological conditions (22, 23). Factors such as high shear stress or inflammatory stimuli can increase phosphatidylserine (PS) exposure on the RBC membrane, inducing a procoagulant shift that enhances thrombin generation, promotes RBC adhesion to ECs, and facilitates thrombus formation (18, 24). Although PS exposure is established as a key regulator of RBC prothrombotic activity, its direct role in driving morphological remodeling of RBCs remains largely unexplored. Moreover, studies addressing RBC responses to chemical and pathological stimuli have not systematically examined the potential mechanistic link between PS exposure and morphological changes.

The role of bacterial infections in modulating RBC procoagulant activity (22, 25) is similarly underexplored. Our previous research demonstrated that the highly virulent pathogen Vibrio vulnificus induces distinct thrombogenic responses in RBCs via its RtxA toxin (26). Specifically, RtxA-mediated calcium-permeable pore formation leads to PS exposure and microvesicle (MV) generation, key factors promoting prothrombotic activity. Early morphological alterations in RBCs accompanying these processes ultimately resulted in hemolysis. However, the precise relationship between PS exposure and RBC morphological changes induced by bacterial toxins remains unclear. In addition, it remains to be determined whether these RBC-mediated thrombogenic mechanisms are pathogen-specific adaptations unique to V. vulnificus or represent a broader strategy among various bacterial pathogens.

To address these critical knowledge gaps, we used EHEC as a comparative model pathogen. EHEC is known to cause hemolytic anemia and severe thrombotic complications, particularly in HUS (8, 27). Through this model, we aim to determine whether bacterial-induced procoagulant activation and associated RBC morphological changes are driven specifically by Stx or if other virulence factors contribute. By elucidating the mechanistic connections between PS exposure and morphological remodeling in pathogen-induced thrombosis, this study advances understanding of the pathophysiology of EHEC-associated complications and identifies potential therapeutic targets. Furthermore, our findings may establish bacterial modulation of RBC function as a conserved mechanism in microbial pathogenesis, warranting expanded investigation across diverse bacterial species (28).

RESULTS

Calcium-dependent PS exposure and procoagulant activation in EHEC-infected RBCs

Freshly isolated RBCs from healthy volunteers were infected with EHEC to assess its impact on PS exposure and MV generation. To focus on early functional and morphological changes while avoiding overt lysis, subhemolytic conditions were defined as <10% total lysis and infection parameters were optimized accordingly. An exposure of 24 hours at a multiplicity of infection (MOI) of 100 or less kept hemolysis below this threshold (Fig. 1A) and enabled reliable measurement of PS exposure and MV generation, with time-dependent profiles shown in fig. S1A. EHEC-infected RBCs were analyzed for PS exposure and MV generation using flow cytometry. While uninfected RBCs exhibited a uniform size distribution, EHEC-infected RBCs displayed two distinct populations corresponding to RBCs and MVs (fig. S1B). Notably, both PS exposure and MV generation increased in an MOI-dependent manner following EHEC infection (Fig. 1, B and C). PS exposure showed a dramatic increase, rising from 0.97% in uninfected controls to 63.89% at an MOI of 100 (Fig. 1B, inset).

Fig. 1. EHEC infection promotes PS externalization and MV release in human RBCs via a calcium-dependent mechanism.

(A) RBC hemolysis after 24 hours infection with wild-type (WT) EHEC (EDL933) at the indicated MOIs (25 to 100), quantified by hemoglobin release (n = 5). (B and C) Flow cytometric quantification of PS exposure (B) and MV generation (C) across increasing EHEC doses (n = 5); a representative histogram for PS is shown in the right inset of (B). (D) Intracellular calcium ([Ca²⁺]_i) measured 24 hours after EHEC infection across MOIs (n = 5); right inset shows a representative histogram. (E) Schematic of Ca²⁺-dependent regulation of PS externalization. (F to H) Quantification of caspase-3 activation [(F), n = 5], scramblase activity [(G), n = 5], and flippase activity [(H), n = 6] in RBCs after EHEC infection at the indicated times (10 and 30 min). (I) Schematic illustrating how PS exposure supports assembly of coagulation factors, culminating in thrombin generation. (J) Thrombin generation in RBCs following infection with increasing MOIs of WT EHEC, quantified as described in Materials and Methods (n = 5). Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s post hoc test versus the control group and is annotated as ns (not significant), *P < 0.05, **P < 0.01, and ***P < 0.001.

Previous studies have established a strong correlation between PS exposure and MV generation, both primarily mediated by elevated intracellular calcium levels (29, 30). Consistent with these findings, EHEC infection induced an MOI-dependent increase in intracellular calcium (Fig. 1D). This calcium influx activated caspase-3, which in turn stimulated scramblase activity, leading to PS externalization. Simultaneously, calcium suppressed flippase activity, preventing the internalization of PS (Fig. 1E) (31, 32). The activation of caspase-3 was also MOI-dependent (Fig. 1F), paralleling the observed increase in intracellular calcium, while scramblase activity progressively increased with MOI (Fig. 1G). Notably, flippase suppression was prominent only at a low MOI of 25 (Fig. 1H). These results suggest that at low MOI, both scramblase activation and flippase suppression contribute to PS exposure. However, at higher MOIs, scramblase activity became the dominant driver of PS exposure, with minimal involvement of flippase suppression. PS exposure on the RBC membrane generates negatively charged surfaces that facilitate the assembly of coagulation factors, driving thrombin generation, a key component of the coagulation cascade (Fig. 1I). As previously reported (18, 24), PS exposure positively correlates with thrombin generation, an observation confirmed here in wild-type (WT) EHEC–infected RBCs (fig. S2). Infection of RBCs with WT EHEC significantly enhanced thrombin generation in a concentration-dependent manner (Fig. 1J). Collectively, these data support a calcium-dependent mechanism in which EHEC induces intracellular Ca²⁺ influx, drives PS externalization on RBCs, and thereby confers a procoagulant surface that augments thrombin generation.

Calcium-driven mechanisms underlying PS exposure and RBC morphological changes in EHEC infection

Following the identification of molecular mechanisms driving PS exposure and MV generation in EHEC-infected RBCs, we sought to further investigate the accompanying morphological changes and their underlying mechanisms. Previous studies using light microscopy have suggested general morphological changes in RBCs during E. coli infections (33). However, no research has thoroughly investigated these changes in the context of highly virulent EHEC. To fill this gap, we used advanced imaging techniques to analyze morphological alterations in human RBCs following EHEC infection. Transmission electron microscopy (TEM) images revealed that EHEC infection induces notable deviations from the typical biconcave discocyte structure (Fig. 2A). At an MOI of 50, discocytes transformed into echinocytes, distinguished by spiky, shell-like protrusions (blue arrowheads), a process likely facilitated by MV shedding. As the MOI increased to 100, this transformation progressed further, with increased MV generation contributing to the formation of spherocytes, characterized by their spherical shape (green arrowheads). These morphological alterations were corroborated by scanning electron microscopy (SEM) (Fig. 2B) and quantitatively assessed (Fig. 2C). The gradual transformation from discocytes to echinocytes, and eventually to spherocytes, underscores the direct impact of EHEC infection on RBC morphology. To further investigate the connection between PS exposure and RBC morphological changes, we performed simple linear regression (SLR) analysis (fig. S3). This analysis revealed a strong positive correlation between PS exposure and morphological changes, with a coefficient of determination (R²) of 0.8530 (P < 0.0001). These findings provide quantitative evidence that increased PS externalization is strongly correlated with RBC morphological remodeling during EHEC infection.

Fig. 2. Calcium-dependent mechanisms driving PS exposure and morphological changes in RBCs during EHEC infection.

Representative transmission electron microscopy (TEM) (A) and scanning electron microscopy (SEM) images (B) illustrating distinct morphological transitions of RBCs into echinocytes, and further into spherocytes, following EHEC infection at varying MOIs (25 to 100) for 24 hours Scale bars: (A), 2 μm (left inset, 1 μm); (B), 4 μm; and (D) and (F), 10 μm. White arrows indicate EHEC; blue arrowheads highlight echinocytes; green arrowheads denote spherocytes. (C) Quantification of RBC morphological changes from SEM images, expressed as the percentage of total RBCs by morphology (discocytes, echinocytes, and spherocytes) (n = 5). In (D) and (E), RBCs were pretreated with the calcium chelator EGTA as described in Materials and Methods, followed by infection with EHEC for 24 hours (D) Confocal microscopy images illustrating morphological changes in RBCs. (E) Flow cytometric quantification of PS exposure in EHEC-infected RBCs (n = 5), confirming the Ca²⁺ dependence of both morphological alterations and PS externalization. (F) Effect of PS blockade on RBC morphological alterations. RBCs were treated with purified Annexin V before EHEC infection at MOI 100 for 24 hours Morphological changes were assessed by confocal microscopy. Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s post hoc test versus the control group, and by two-tailed Student’s t test between the MOI 50 and MOI 50 + EGTA groups. Significance is annotated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

To delineate the factors driving these morphological changes, we first examined the role of calcium (Ca²⁺) in RBC morphological alterations. Treatment with EGTA, a calcium chelator, at an MOI of 100 preserved the normal biconcave disc-like structure, similar to that of the uninfected control group (Fig. 2D). Since PS externalization is closely linked to calcium signaling, we next assessed its relationship with Ca²⁺ levels. EGTA treatment significantly reduced PS exposure at an MOI of 50 compared to uninfected controls (Fig. 2E). Given that PS externalization provides a procoagulant surface for thrombin generation (Fig. 1I), we further investigated the connection between calcium signaling and EHEC-induced procoagulant pathway, focusing on its downstream effects. As shown in fig. S4 (A to C), EGTA treatment reduced MV generation and caspase-3 activity, most notably restoring thrombin generation. To confirm the critical role of PS exposure in morphological changes, we applied purified annexin V, a PS-blocking agent, to EHEC-infected RBCs at MOI 100. Annexin V treatment completely abrogated the EHEC-induced morphological changes, and the cells retained their discocyte shape, closely resembling the control group (Fig. 2F). This strongly implicates PS exposure, mediated by calcium influx, as the key driver of shape changes in RBCs during EHEC infection. Together, these results emphasize Ca²⁺ influx as a pivotal regulator orchestrating the simultaneous occurrence of PS exposure and morphological transformations in EHEC-infected RBCs.

Lack of Stx2 contribution to RBC procoagulant activity and morphology

To identify EHEC virulence factors responsible for RBC morphological remodeling and PS exposure, we profiled the bacterial transcriptome after infection to RBCs and compared it with uninfected controls. The volcano plot identified 653 DEGs (criteria: |log₂FC| ≥ 1.0; P ≤ 0.05), including 417 up-regulated genes enriched for nutrient transport and metabolic pathways (Fig. 3A). Virulence-associated transcripts were summarized in a heatmap (Fig. 3B). Among virulence-associated transcripts, the most increased exotoxin transcript was ehxA (enterohemolysin, EDL933_p0046; log₂FC = 1.37; P = 4.99 × 10⁻²) (table S1), whereas stx genes were not significantly induced (stx1A, log₂FC = 0.01 and P = 1.00; stx1B, log₂FC = −0.98 and P = 3.44 × 10⁻¹; stx2A, log₂FC = 0.57 and P = 3.51 × 10⁻¹; and stx2B, log₂FC = 0.33 and P = 9.13 × 10⁻¹) (table S1).

Fig. 3. EhxA, not Stx2, drives PS exposure, intracellular Ca²⁺ increase, and RBC remodeling during EHEC infection.

(A) RNA-seq of EHEC after 24 hours coincubation with human RBCs versus an uninfected control. Volcano plot depicting differentially expressed genes (DEGs; |log₂ fold change| ≥ 1.0, P ≤ 0.05), with up-regulated genes in red, down-regulated genes in blue, ehxA highlighted in green, and stx2A/stx2B highlighted in orange (total DEGs = 669). (B) Heatmap of EHEC virulence factor genes during RBC infection showing significant up-regulation of ehxA and down-regulation of stx2A/B relative to uninfected controls. (C) Schematic of Δstx2 construction in EDL933 (CRISPR-Cas9 deletion of the stx2A-stx2B locus). (D) PS exposure (left axis) and MV generation (right axis) in RBCs infected with WT or Δstx2 at MOIs 50 or 100 (n = 5). (E) Intracellular Ca²⁺ level ([Ca²⁺]_i; left axis) and thrombin generation (right axis) in RBCs infected with WT or Δstx2 across MOIs (n = 5). (F) Representative confocal images of RBC morphology after infection with WT or Δstx2 (MOIs 50 or 100), showing transitions to echinocytes (blue arrowheads) and spherocytes (green arrowheads); scale bar, 10 μm. (G and H) Bacteria-free exposure of RBCs to purified Stx2 (0 to 10 μg/ml) for 24 hours (n = 4). Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s post hoc test versus the control group, and by two-tailed Student’s t test for comparisons between wild-type–infected and mutant-infected groups at the same MOI. Significance is annotated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

Because lack of induction alone does not exclude a functional contribution and because Stxs are widely regarded as central to EHEC pathogenesis (34), we directly tested whether Stx contributes to the RBC phenotypes in this RBC experimental system. We focused on Stx2, which is generally more toxic than Stx1 in clinical and experimental contexts (12). A schematic of the stx2A-stx2B locus and the RT-primer design used in this study is shown in Fig. 3C. We first examined genetic requirement by constructing an isogenic Δstx2 mutant through deletion of the stx2A-stx2B locus in EDL933 and validating the edit by polymerase chain reaction (PCR) and whole-genome sequencing (WGS; Materials and Methods; fig. S5). When RBCs were infected for 24 hours at an MOI of 50 and 100, Δstx2 was indistinguishable from WT in PS exposure and MV generation, intracellular Ca²⁺ levels and thrombin generation, and RBC morphology by confocal imaging (Fig. 3, D to F).

To provide protein-level confirmation, human RBCs were exposed to purified Stx2 at concentrations up to 10 μg/ml. Purified Stx2 did not induce PS exposure but produced at most a small increase in MV at the highest tested concentration. This small MV increase is unlikely to be biologically meaningful, confirming that Stx2 does not drive these end points in our system (Fig. 3, G and H). Together with the transcriptomic screen that highlighted ehxA rather than stx genes, these results argue against a role for Stx in RBC remodeling and procoagulant activity in this RBC experimental context studied and motivate a subsequent focus on the pore-forming enterohemolysin EhxA.

Unveiling the critical role of EhxA in EHEC-induced PS exposure and morphological changes

To determine the pivotal role of the virulence factor EhxA in driving PS exposure and morphological changes in RBCs, an ehxA deletion mutant (ΔehxA::cat) was constructed from the WT EDL933 strain (Fig. 4A) and the edit was validated by PCR and WGS (Materials and Methods; fig. S6). Compared with WT, ΔehxA infection significantly reduced intracellular Ca²⁺ levels, PS exposure and MV generation across MOIs (Fig. 4B and fig. S7), indicating that EhxA is required for the Ca²⁺-dependent scrambling pathway. Consistent with these molecular readouts, confocal imaging revealed that WT progressively transformed RBCs from discocytes into echinocytes and spherocytes in an MOI-dependent manner, whereas ΔehxA failed to induce discernible morphological changes (Fig. 4C). Quantification confirmed strong suppression of echinocyte/spherocyte formation in the mutant, with morphology distributions approaching those of uninfected controls (Fig. 4D).

Fig. 4. Identification of EhxA as a key virulence factor underlying Ca²⁺ influx, PS externalization, RBC remodeling, and thrombin generation in EHEC-infected human RBCs.

(A) Construction of an ehxA deletion mutant (ΔehxA::cat) in EDL933 via CRISPR-Cas9. (B) Flow cytometric quantification of intracellular calcium ([Ca²⁺]_i level, MFI; left axis) and PS exposure (%; right axis) in RBCs infected for 24 hours with WT EHEC or ΔehxA at MOI 50 and 100 (n = 5). (C) Representative confocal images of RBC morphology after infection with WT or ΔehxA (MOIs 50 and 100; 24 hours). Blue arrowheads, echinocytes; green arrowheads, spherocytes; scale bar, 10 μm. (D) Quantification of RBC morphology from confocal images, expressed as the percentage of total RBCs by morphology (discocytes, echinocytes, and spherocytes) after infection with WT or ΔehxA (MOI 100; 24 hours) (n = 5). (E) Multiple linear regression (MLR) modeling the contribution of MOI, strain (WT versus ΔehxA), and PS exposure to abnormal RBC morphology; regression coefficients (β), R², and P values are provided in table S2. (F) Thrombin generation measured in RBCs after 24 hours infection with WT or ΔehxA at MOIs 50 and 100 (n = 5). (G) MLR modeling the contribution of MOI, strain, and PS exposure to thrombin generation; detailed statistics are provided in table S3. Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s or Tukey’s post hoc test, as appropriate, and by two-tailed Student’s t test for pairwise comparisons between wild-type–infected and mutant-infected groups at the same MOI. Significance is annotated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

To quantitatively confirm these morphological observations and the associated reduction in PS exposure caused by the ΔehxA mutation, multiple linear regression (MLR) analysis incorporating both WT and ΔehxA mutant strains was conducted. The analysis revealed that PS exposure remained significantly associated with RBC morphological alterations [β = 0.4398, 95% confidence interval (CI) = 0.03991 to 0.8397, and P = 0.0327] (Fig. 4E and table S2), while mutant status (ΔehxA) exhibited a pronounced negative impact (β = −39.32, 95% CI = −53.40 to −25.23, and P < 0.0001), reinforcing the essential role of EhxA in mediating RBC morphological changes through PS exposure. As a functional readout of RBC-dependent procoagulant activity, thrombin generation was significantly lower in ΔehxA-infected RBCs than in WT at matched MOIs (Fig. 4F). Using MLR that included both WT and ΔehxA datasets, PS exposure remained significantly associated with thrombin generation, and deletion of ehxA had a pronounced negative impact on thrombin generation (β = −0.8734 and P = 0.0001) (Fig. 4G and table S3), indicating that EhxA is essential for PS-mediated propagation of the procoagulant cascade. Collectively, MLR identifies PS exposure as the principal predictor of RBC morphology and thrombin, positioning PS at the core of the EhxA-driven cascade.

EhxA as a key mediator of prothrombotic activity in EHEC-infected RBCs

Beyond thrombin generation, we further investigated two functional consequences in RBCs. First, RBC adhesion to ECs was assessed by fluorescence microscopy. WT infection increased RBC adhesion to human umbilical vein ECs (HUVECs) in an MOI-dependent manner (yellow arrowheads) (Fig. 5A) and quantitative counts confirmed the rise at MOI 50 and 100. In contrast, RBCs infected with the ΔehxA mutant exhibited significantly reduced adhesion, approaching uninfected control levels (Fig. 5B), indicating that EhxA is essential for mediating EHEC-induced RBC-EC interactions. Second, WT infection promoted concentration-dependent RBC self-aggregation even in the absence of ECs (green arrows) (Fig. 5C), and this effect was robustly reduced in ΔehxA-infected RBCs (Fig. 5D).

Fig. 5. PS exposure–dependent prothrombotic RBC-EC adhesion and RBC self-aggregation mediated by EhxA.

(A) Representative fluorescence images of RBC adhesion to confluent HUVEC monolayers after 24 hours infection with WT EHEC or ΔehxA at MOIs 50 and 100. ECs (green), RBCs (red); yellow arrowheads denote adherent RBCs; scale bar, 100 μm. (B) Quantification of RBC adhesion to HUVECs, expressed as adhered RBCs per field (counts/field) across strains and MOIs (n = 4). (C) Representative fluorescence images showing RBC self-aggregation in the absence of ECs after 24 hours infection with WT or ΔehxA at MOIs 50 and 100; green arrows indicate aggregates; scale bar, 10 μm. (D) Quantification of RBC aggregation, expressed as the percentage of total RBCs that are self-aggregated (n = 4). (E) Multiple linear regression (MLR) relating PS exposure (%) to RBC adhesion (counts/field); regression coefficients (β), R², and P values are provided in table S4. (F) MLR relating PS exposure (%) to RBC aggregation (%); detailed statistics are provided in table S5. Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s post hoc test versus the control group, and by two-tailed Student’s t test for comparisons between wild-type– and mutant-infected groups at the same MOI. Significance is annotated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

With PS exposure linked to thrombin generation in EHEC-infected RBCs (Fig. 4G), the next step was to test whether PS quantitatively predicts additional procoagulant readouts. Using MLR that pooled WT and ΔehxA data, PS exposure remained a significant independent association with RBC adhesion to HUVECs (β = 0.3069, 95% CI = 0.1607 to 0.4532, and P = 0.0004), whereas ΔehxA status exerted a strong negative effect (β = −7.789, 95% CI = −13.26 to −2.319, and P = 0.0082) (table S4). The overall model was highly significant (P < 0.0001) and showed substantial fit (R² = 0.8684) (Fig. 5E). A parallel analysis for RBC self-aggregation showed the same pattern. PS exposure remained significantly associated (β = 0.2904, 95% CI = 0.0196 to 0.5613, and P = 0.0371). ΔehxA status had a pronounced negative effect (β = −12.37, 95% CI = −22.49 to −2.238, and P = 0.0198) (table S5). The model was significant overall (P < 0.0001) and provided good explanatory power (R² = 0.8106) (Fig. 5F). Together, these data demonstrate that EhxA drives not only thrombin generation but also EHEC-induced RBC adhesion and self-aggregation, and that both processes are quantitatively linked to PS exposure, placing PS at the core of the EhxA-dependent prothrombotic program.

Evidence from complementation and purified-protein assays for EhxA’s central role in RBC procoagulant activity and remodeling

To move beyond conclusions from the ΔehxA deletion alone, EhxA’s role in EHEC-infected RBCs was assessed in parallel by genetic restoration of ehxA and, in a separate set of experiments, by bacteria-free exposure to purified EhxA. Figure 6A outlines the complementation strategy in brief, where the ΔehxA::cat strain was transformed with an empty vector plasmid pJK1113 or an arabinose-inducible pHY2508, alongside a WT strain with empty vector as positive control, and the same RBC readouts were quantified. Reintroducing ehxA restored PS exposure, intracellular Ca²⁺ increase, thrombin generation, and the characteristic transitions from discocyte to echinocyte and then spherocyte to WT levels (Fig. 6, B and C, and fig. S8).

Fig. 6. Conclusive evidence for EhxA-dependent RBC procoagulant activity and remodeling from genetic complementation and purified EhxA.

(A) Plasmid/strain schematics: wild-type (WT) EDL933 carrying the empty vector pJK1113; ΔehxA::cat carrying the empty vector pJK1113; and ΔehxA::cat complemented with pHY2508 expressing ehxA under the arabinose-inducible P_BAD promoter. (B) Flow cytometric quantification of PS exposure (left axis) and MV generation (right axis) in RBCs infected for 24 hours with the indicated strains at MOIs 25 and 50 (n = 5). (C) Intracellular calcium ([Ca²⁺]_i level, MFI; left axis) and thrombin generation (right axis) in RBCs 24 hours after infection with WT carrying pJK1113, ΔehxA::cat carrying pJK1113, or ΔehxA::cat complemented with pHY2508 at MOIs 25 and 50 (n = 5). (D) Experimental outline for parallel approaches: genetic ehxA complementation (top) and bacteria-free exposure to purified EhxA protein (bottom), followed by measurement of [Ca²⁺]_i, PS exposure, and morphology over 24 hours Created in BioRender. Im, H. (2026) https://BioRender.com/mycfstr. (E) Concentration-response to purified EhxA (0 to 10 μg/ml; 24 hours) showing [Ca²⁺]_i level ([Ca²⁺]_i level, MFI; left axis) and PS exposure (%, right axis) (n = 4). (F) Confocal imaging of His₆-tagged EhxA on intact RBCs (bright field/FITC/merge): Anti–His-FITC signal localizes to the RBC surface; blue arrowheads, echinocytes; green arrowheads, spherocytes; scale bar, 10 μm. Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s post hoc test versus the control group and by one-way ANOVA with Tukey’s multiple comparisons test for analyses among three groups at the same MOI. Significance is annotated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

Furthermore, in a bacteria-free setting, purified EhxA induced a concentration-dependent increase in intracellular Ca²⁺, PS exposure, and MV generation, with parallel increases in thrombin generation (Fig. 6, D and E, and fig. S9). After exposure to His-tagged EhxA, anti–His–fluorescein isothiocyanate (FITC) signal colocalized with RBC outlines in bright-field images, indicating surface association across a morphological transition from discocytes to echinocytes to spherocytes (Fig. 6F). These imaging data support robust EhxA engagement of the RBC surface, regardless of RBC shape, and are consistent with the concentration-dependent patterns observed for Ca²⁺ influx, PS exposure, and procoagulant activity. Together, the complementation results and the purified-protein assays substantiate a direct role for EhxA in the RBC phenotypes and procoagulant activity.

EhxA-dependent RBC morphological remodeling and venous thrombosis during in vivo EHEC infection in rats

To confirm that rat RBCs exhibit similar responses to human RBCs upon EHEC infection and validate the relevance of the rat model for in vivo studies, we first conducted in vitro experiments. EHEC infection of rat RBCs resulted in approximately 10% hemolysis, a statistically significant (P < 0.05) but biologically modest effect (fig. S10). EHEC infection increased PS exposure and thrombin generation in rat RBCs in a concentration-dependent manner (Fig. 7, A and B), closely mirroring the responses observed in human RBCs. These findings demonstrate that rat RBCs exhibit comparable prothrombotic alterations to human RBCs during EHEC infection, thereby supporting the use of a rat model to investigate the in vivo effects of the virulence factor EhxA.

Fig. 7. EhxA-induced RBC morphological remodeling and prothrombotic activation in ex vivo and in vivo rat models.

(A) PS exposure of rat RBCs after 6 hours in vitro infection with EHEC at varying MOIs, analyzed by flow cytometry (n = 6). (B) Thrombin generation quantified by prothrombinase assay after 6-hour infection (n = 6). (C) Experimental schematic illustrating ex vivo and in vivo models. Rats were intravenously infected with WT EHEC (10⁷ or 10⁹ CFU per rat) or ΔehxA mutant (10⁹ CFU per rat) for 3 hours RBCs isolated from infected rats were examined for morphological changes using scanning electron microscopy (SEM; ex vivo). Separately, venous thrombosis was induced by thromboplastin injection 3 hours postinfection, and thrombi were analyzed 15 min later (in vivo). (D) Representative SEM images showing echinocyte (blue arrowheads) and spherocyte (green arrowheads) formation in RBCs from rats infected with WT EHEC in a dose-dependent manner. RBC morphology in ΔehxA-infected rats remained largely unchanged, maintaining discocyte shape. (E) Quantification of RBC morphological distribution from SEM images, expressed as percentages (Discocyte, Echinocyte, and Spherocyte). (F) Representative stereomicroscopy images depicting thrombi formation in the inferior vena cava (IVC) of rats following infection with WT EHEC or ΔehxA mutant. (G) Quantitative analysis of thrombus area (mm²) measured from stereomicroscopy images (n = 5). (H) Independent assessment of thrombus weight in rats infected with varying doses of WT EHEC or ΔehxA mutant (n = 5). Data are presented as mean ± SEM. Statistical significance was assessed by one-way ANOVA with Dunnett’s post hoc test versus the control group, and by two-tailed Student’s t test for comparisons between wild-type–infected and mutant-infected groups at the same MOI. Significance is annotated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

To further assess the effects of EHEC infection ex vivo, rats were intravenously infected with either the WT strain or the ΔehxA mutant for 3 hours, and RBC morphology was subsequently analyzed (Fig. 7C, top). Infection with the WT triggered a transformation of normal biconcave discocytes into echinocytes at 10⁷ colony-forming units (CFUs) per rat and spherocytes at 10⁹ CFUs per rat (Fig. 7D). In contrast, RBCs from rats infected with the ΔehxA mutant exhibited minimal morphological changes compared to the uninfected control group, while retaining their characteristic discocyte shape, as quantified in Fig. 7E. These findings align with prior observations in human RBCs, confirming EhxA as the principal virulence factor responsible for the morphological alterations induced by EHEC infection.

To evaluate the pathological consequences of the observed morphological and functional alterations in RBCs, including increased PS exposure and thrombin generation, the impact of EHEC infection on thrombus formation was investigated using a rat venous thrombosis model (Fig. 7C, bottom). Thrombus images from each group are shown in Fig. 7F, and the area of thrombosis was quantified using the ImageJ software. In addition, separate independent experiments were performed to determine thrombus weight. Together, these data show that infection with the WT EHEC strain resulted in a dose-dependent increase in both thrombus area and weight, as quantified in Fig. 7 (G and H, respectively). In stark contrast, rats infected with the ΔehxA mutant exhibited results similar to those of the uninfected control in both parameters. These results highlight the critical role of EhxA in mediating thrombotic complications during EHEC infection, further establishing its importance as a key driver of EHEC pathogenicity.

DISCUSSION

Our study identifies the RTX-family toxin EhxA as a previously unrecognized virulence factor in EHEC that forms calcium-permeable pores in the RBC membrane (35) and initiates a PS-centric procoagulant pathway. Calcium entry promotes PS externalization, accompanies RBC shape remodeling from discocytes to echinocytes and ultimately to spherocytes, and increases thrombin generation. Although Stx is the principal driver of renal endothelial injury leading to hemolytic anemia and microvascular thrombosis (36, 37), our findings challenge this paradigm by showing that in RBCs, EhxA rather than Stx2 drives the thrombogenic and morphological phenotypes. EhxA’s necessity was demonstrated by the ΔehxA mutant, which markedly reduced Ca²⁺ influx, PS exposure, MV release, morphological transitions, and thrombin generation. Sufficiency was established by genetic complementation, which restored all end points to WT levels, and by bacteria-free exposure to purified EhxA, which reproduced these phenotypes (Fig. 6). The EhxA-PS axis translated in vivo in rats, where EhxA provoked RBC remodeling and increased venous thrombosis, whereas the ΔehxA mutant did not (Fig. 7). Together, these results define a robust, RBC-intrinsic, PS-anchored mechanism of thrombogenesis in EHEC.

This paradigm shift not only reshapes our understanding of RBC membrane biology but also raises important questions about how pathogens may exploit RBC morphological remodeling to influence disease progression. These findings broaden the understanding of bacterial pathogenesis by revealing a direct RBC-mediated procoagulant mechanism that could operate independently of endothelial injury. Furthermore, to date, only V. vulnificus has been demonstrated to mediate RBC-driven thrombosis (26). This discovery suggests that RBCs may play a previously unrecognized role in bacterial pathogenesis, potentially representing a broader but underexplored microbial strategy to modulate host hemostasis and promote thrombotic complications. This finding not only deepens our understanding of microbial contributions to thrombogenesis but also highlights potential therapeutic targets to mitigate pathogen-induced coagulopathies.

In this study, we show that, upon infection to RBCs, E. coli O157:H7 (EDL933) up-regulates ehxA, which encodes the pore-forming toxin EhxA, and that EhxA-dependent Ca²⁺ entry in RBCs is tightly coupled to PS externalization, progressive remodeling from discocytes to echinocytes and spherocytes, and increased thrombin generation. The interdependence of these end points was established by Ca²⁺ chelation with EGTA, which abolished PS exposure, morphology change, and thrombin generation (Fig. 2, D and E, and fig. S4), by genetic loss of function (ΔehxA), which eliminated the same readouts (Fig. 4), and by PS blockade with annexin V, which prevented morphological remodeling (Fig. 2F). Evidence for a direct role of EhxA was provided in a bacteria-free system, where purified EhxA reproduced the Ca²⁺ rise, PS exposure, and shape transitions and showed clear association with RBC membranes (Fig. 6, D to F), indicating that bacterial adhesion is not required. Although several T3SS transcripts increased during RBC infection (table S1), the ensemble of results, comprising loss with ΔehxA (Fig. 4), restoration by complementation (Fig. 6, A to C), and reproduction with purified EhxA (Fig. 6, D to F), supports an EhxA-driven, PS-anchored mechanism as the principal driver of the observed phenotypes rather than T3SS-dependent effects. Collectively, these data provide an integrated causal framework grounded in genetic necessity, genetic rescue, and protein-level reconstitution for EHEC-induced, PS-anchored procoagulant remodeling of RBCs.

To further corroborate the mechanism at the transcriptional level, RNA sequencing (RNA-seq) during RBC infection showed selective up-regulation of ehxA among exotoxin genes, whereas Stx-related transcripts were unchanged (Fig. 3B and table S1). Consistent with this expression pattern, an isogenic Δstx2 strain was indistinguishable from WT across RBC readouts (Fig. 3, D and E), and purified Stx2 did not elicit PS exposure or MV generation (Fig. 3, G and H). In contrast, hlyE, a chromosomal pore-forming cytolysin outside the RTX family (38), was not induced, and other pore-forming hemolysins likewise showed no induction in our dataset (table S1). Together with functional tests, including loss of the phenotype in ΔehxA (Fig. 4), restoration by ehxA complementation (Fig. 6, A to C), and reproduction by purified EhxA under bacteria-free conditions (Fig. 6, D to F), these transcriptomic findings identify EhxA as the specific driver of the RBC procoagulant phenotype in our experimental model.

Beyond identifying EhxA as the trigger of RBC dysfunction, we provide a methodological advance by establishing, to our knowledge, a quantitative link between PS exposure and RBC morphology in bacteria-infected RBCs using MLR. Across WT and ΔehxA datasets, PS exposure emerged as the principal predictor of morphological remodeling, elevating PS from a correlative marker to a central determinant of the discocyte-echinocyte-spherocyte transition (R² = 0.8759 and P < 0.0001; Fig. 4E). In line with the established association between PS and thrombin generation (Fig. 1I), the same modeling framework further showed that PS exposure quantitatively predicts a broader panel of procoagulant readouts within EHEC infection, including thrombin generation (Fig. 4G), RBC adhesion to HUVECs (Fig. 5E), and RBC self-aggregation (Fig. 5F). Together, these analyses indicate that the overall procoagulant activity observed in EHEC is organized around a PS-anchored axis that integrates morphological remodeling with thrombin production and cell-cell interactions. This conceptual integration is summarized in Fig. 8, which outlines the experimental workflow and mechanistic scheme spanning Ca²⁺ entry, PS exposure, morphological remodeling, and thrombotic outcomes. Collectively, our findings expand current views of RBC pathophysiology by establishing PS as the mechanistic hub linking calcium entry to structural and functional activation and by positioning EhxA as a bacterial effector that leverages this hub to drive thrombogenesis.

Fig. 8. Schematic of the EhxA-driven Ca²⁺-PS axis that organizes RBC remodeling and thrombogenesis during EHEC infection.

The varying lengths of the three dark red arrows represent differences in blood flow and the shear stress exerted on RBCs. RTX-family toxins, including V. vulnificus RtxA and EHEC EhxA, share a conserved RTX family toxin pore-forming region, highlighting a conserved mechanism in Ca²⁺-mediated thrombogenesis (upper left). During EHEC infection, EhxA associates with the RBC membrane and forms a calcium-permeable pore, triggering intracellular calcium influx ([Ca²⁺]_i). Elevated [Ca²⁺]_i activates caspase-3 and scramblase while suppressing flippase, resulting in phosphatidylserine (PS) externalization (lower left). PS-centered signaling accompanies a morphological progression from discocytes to echinocytes and spherocytes (top). The PS-rich surface supports prothrombinase assembly, increasing thrombin generation, and thereby enhances RBC-endothelial adhesion and RBC self-aggregation, culminating in venous thrombosis (central schematic). This model highlights an EhxA-dependent Ca²⁺-PS pathway that drives procoagulant activation and RBC remodeling during EHEC infection, consistent with venous thrombosis observed in patients. Created in BioRender. Im, H. (2026) https://BioRender.com/d96bk8k.

Our findings extend the understanding of RTX-family toxins beyond EHEC, paralleling previous reports in V. vulnificus, where RtxA induces RBC-mediated thrombogenesis via calcium-permeable pores. In this study, we revealed that EhxA from EHEC also contributes to RBC morphological change and thrombogenesis through calcium-permeable pore formation, similar to V. vulnificus RtxA. This finding further supports the broader role of RTX toxins in host-pathogen interactions (39, 40). RTX genes are widely present in pathogenic Gram-negative bacteria, encoding exotoxins capable of pore formation, which are involved in host cell lysis, adhesion, and immune modulation (41, 42). Given these similarities, we propose that RTX toxins may have a conserved role in modulating RBC morphology and thrombogenic potential. Our data suggest that, like RtxA, EhxA from EHEC uses a pore-forming mechanism to influence RBC morphology and promote thrombogenesis. However, RTX toxins exhibit host specificity, with functional variations requiring direct validation in other pathogens, such as Actinobacillus pleuropneumoniae, which primarily affects the respiratory system (43, 44). This suggests RTX toxins may have a broader but species-specific role in pathogen-induced thrombogenesis, although further research is needed to confirm its conservation.

Several studies have shown that elevations of intracellular Ca²⁺ in RBCs under physical, chemical, or pathological stimuli promote PS exposure and can culminate in either eryptosis or hemolysis (45–47). Although PS externalization is shared by both processes, they serve distinct roles: Eryptosis primarily supports clearance of senescent RBCs, whereas hemolysis contributes to procoagulant activity and thrombogenesis (45–47). Building on this Ca²⁺-centered framework, it is important to note that RBCs, which are anucleate and lack mitochondria, engage a calcium-activated remodeling program that is distinct from the classical caspase-dependent apoptosis of nucleated cells. In RBCs, increases in Ca²⁺ drive PS externalization through scramblase activation and flippase inhibition and may also activate calpains, which can reshape the spectrin/actin cytoskeleton (48). In our model, EhxA-mediated Ca²⁺ influx links PS exposure to the characteristic progression of morphology from discocytes through echinocytes to spherocytes and augments procoagulant activity, as supported by the suppression of PS exposure, shape change, and thrombin generation with EGTA, annexin V, and ΔehxA (Figs. 2 and 4), the restoration of these readouts by genetic complementation of ehxA, and the recapitulation of key phenotypes with purified EhxA in bacteria-free assays (Fig. 6). We detect only modest caspase-3 activation (Fig. 1F), consistent with an eryptosis-like process rather than engagement of the classical apoptosis pathway typical of nucleated cells (49). Thus, while calpain may operate in parallel to facilitate cytoskeletal remodeling, the primary route to the procoagulant output under our experimental conditions is an EhxA-induced Ca²⁺ increase that drives PS exposure.

Our study points to a previously underrecognized, pathogen-triggered program of RBC disintegration mediated by RTX-family toxins. Exposure of RBCs to RTX toxins initiates morphological remodeling from discocytes to echinocytes and then to spherocytes and can ultimately culminate in hemolysis. These morphological transitions coincide with increased PS exposure, thereby amplifying the procoagulant activity of RBC membranes. To our knowledge, shape-resolved evidence of PS externalization specifically on echinocytes and spherocytes in bacteria-infected RBCs remains limited, and Annexin V–based fluorescence staining suggested PS exposure on both shape states in our previous work (26). This observation suggests that echinocytes and spherocytes may directly contribute to pathogen-driven procoagulant processes by providing PS-rich surfaces for coagulation complex assembly as well. Notably, hemolysis induced by chemical stimuli has been reported to generate PS-exposing membrane fragments that directly promote procoagulant activity (50, 51). Building on these observations, we propose that RTX toxins engage a calcium-dependent sequence comprising morphological remodeling, PS externalization, and, in later stages, hemolysis that contributes to thrombogenic activity (Fig. 8, top right). Future studies will be important to determine whether similar morphological cues and PS exposure also serve as signals for physiological RBC removal at end of life, thereby clarifying the broader implications for RBC biology and pathogen-induced coagulopathy.

Given the widespread presence of RTX-family toxins in various Gram-negative pathogens and their ability to modulate host cell function, targeting these virulence factors presents a promising therapeutic avenue. Strategies aimed at neutralizing pore-forming toxins, such as monoclonal antibodies, small-molecule inhibitors, or receptor decoys, could mitigate toxin-induced cellular dysfunction and thrombogenesis. In addition, calcium channel blockers or ion chelators may effectively prevent intracellular calcium dysregulation mediated by pore-forming toxins, thereby preserving RBC integrity and reducing pathogenic coagulopathies. Moreover, given the role of PS exposure in thrombus formation, therapeutic strategies specifically targeting inhibition of PS externalization or procoagulant MV release could provide broad protection against bacterial-induced thrombotic complications. Future studies should explore these targeted approaches across different pathogens to develop effective countermeasures against RTX-mediated bacterial pathogenesis and its associated systemic complications.

MATERIALS AND METHODS

Bacterial strains, plasmids, mutant construction, and complementation

The strains and plasmids used in this study are listed in table S6, and the oligonucleotides are listed in table S7. Escherichia coli O157:H7 strain EDL933, originally isolated from the 1982 U.S. outbreak of hemorrhagic colitis (52), was used as the WT strain. This strain harbors ehxA, which encodes enterohemolysin (EhxA), a member of the RTX (repeats-in-toxin) family implicated in host cell lysis and virulence (53).

A ΔehxA mutant was constructed using a CRISPR-Cas9 mutagenesis system adapted from Jiang et al. (54). The pTargetF plasmid was modified to express a single-guide RNA (sgRNA) targeting ehxA (N20 adjacent to a PAM, nGG), and the left/right homology arms together with a chloramphenicol (Cm) resistance (cat) cassette (amplified from pKD3) were assembled into a repair template by overlap PCR. The repair template and pTargetF-ehxA were co-electroporated into λ-Red–induced EDL933 carrying pCas, and transformants were recovered on LB agar supplemented with kanamycin (Km; 50 μg/ml) and spectinomycin (Spc; 50 μg/ml) at 30°C overnight. Plasmid curing was performed as described (54). The final ΔehxA mutant was verified by PCR (fig. S6) and WGS and used for downstream functional assays.

A Δstx2 mutant was generated by CRISPR-Cas9 without insertion of an antibiotic-resistance cassette. pTargetF was modified to express an sgRNA targeting the chromosomal stx2 locus and electroporated into λ-Red–induced EDL933 carrying pCas. Transformants were recovered on LB agar containing Spc (50 μg/ml) at 30°C overnight. Plasmid curing was performed as above, and allelic exchange was confirmed by PCR (fig. S5) and WGS.

For complementation, the ehxA CDS was PCR-amplified (primers in table S7) and cloned into the Nco I–Sph I sites of pJK1113 under the arabinose-inducible P_BAD promoter. The resulting plasmid (pHY2508; pJK1113::ehxA) was introduced into the ΔehxA strain by electroporation (54), and transformants were selected on LB agar containing Km (50 μg/ml). Successful transformation was confirmed by PCR (fig. S11). In complementation experiments only, strains were precultured at 37°C for 2 hours with 0.08% L-(+)-arabinose, and arabinose was maintained during RBC incubation. At 24 hours postinfection, RBC alterations were evaluated by quantifying PS exposure, MV generation, intracellular Ca²⁺ level, thrombin generation, and hemolysis using the procedures described below.

Selection antibiotics were used only for strain construction and complementation (Km, 50 μg/ml; Spc, 50 μg/ml; and Cm, 25 μg/ml), and all phenotypic assays, including RBC coincubation and purified-protein experiments, were performed in antibiotic-free media after washing and resuspension in antibiotic-free buffer.

WGS and mutant validation

Genomic DNA from WT EDL933 and isogenic mutants (ΔehxA, Δstx2) was sequenced on the PacBio Revio platform to generate HiFi reads. De novo assemblies were constructed with Flye and annotated using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP). Long-read alignments and whole-genome comparisons confirmed precise deletion of ehxA (EHEC-hlyA) in ΔehxA and stx2 in Δstx2, with clean junctions and no structural rearrangements at the edit boundaries. As an independent PCR-based validation, primers located outside the homology arms were used to amplify across each edit junction, yielding the expected amplicons only in the corresponding mutants (primer sequences in table S7; gels in figs. S5 and S6). Sequencing datasets are publicly available at NCBI: ΔehxA (BioProject PRJNA1321016, SRA SRR35313728, WGS JBQXGD000000000, and BioSample SAMN51193716) and Δstx2 (BioProject PRJNA1322117, SRA SRR35316131, WGS JBQYGB000000000, and BioSample SAMN51235973).

Bacterial growth curve measurement

Bacterial strains were grown overnight in LB broth at 37°C with shaking. Cultures were then diluted 1:3 with fresh LB broth, and the optical density at 600 nm (OD₆₀₀) was measured. Subsequently, cultures were normalized to the lowest OD₆₀₀ by diluting each culture with LB to match the minimal value. We then added 10 μl of each normalized inoculum to 990 μl of LB in individual wells of a 24-well microplate (final volume 1 ml per well). Bacterial growth was monitored hourly at OD₆₀₀ using a Spark microplate reader (Tecan, Männedorf, Switzerland) at 37°C with continuous orbital shaking and humidity control (humidity cassette, Tecan) to prevent evaporation. All experiments were performed with at least five independent biological replicates.

Ethical approval and blood sample collection

This study was conducted with the approval of the Ethics Committee of the Health Service Center at Hanyang University (HYUIRB-202012-001-9) and Seoul National University (IRB no. 2108/001-010). Whole-blood samples were obtained from healthy male volunteers aged 20 to 30 years following informed consent. Blood was collected on the day of each experiment using a 21-gauge needle and acid citrate dextrose–coated vacutainers to prevent coagulation. All procedures were performed in accordance with institutional guidelines and regulations.

Isolation and preparation of RBCs

Following whole-blood collection, samples were centrifuged at 200g for 15 min to separate platelet-rich plasma and the buffy coat, which were carefully removed. The remaining packed RBCs were washed twice with phosphate-buffered saline (PBS; 1.06 mM KH₂PO₄, 154 mM NaCl, and 2.96 mM Na₂HPO₄; pH 7.4), followed by a final wash with Ringer’s solution (125 mM NaCl, 5 mM KCl, 1 mM MgSO₄, 32 mM HEPES, and 5 mM glucose; pH 7.4). The processed RBCs were resuspended in Ringer’s solution at a final concentration of 1 × 10⁸ cells/ml. For subsequent experiments, CaCl₂ was added to the suspension to reach a final concentration of 1 mM.

Assessment of hemolytic activity

To evaluate hemolytic activity, RBC suspensions were exposed to EHEC at varying MOIs and incubated at 37°C with gentle shaking (1000 rpm) across multiple time points. Following incubation, the samples were centrifuged at 10,000g for 2 min, and hemoglobin release, indicative of hemolysis, was quantified by measuring absorbance at 540 nm (A₅₄₀) using a spectrophotometer. Complete RBC lysis induced by 1% Triton X-100 served as the reference point for 100% hemolysis. Hemolysis data for the Δstx2 mutant, purified Stx2, and ehxA complementation are shown in fig. S15 (A to C).

Detection of PS exposure and MV generation

Flow cytometry was used to assess PS exposure and MV generation in EHEC-infected RBCs. RBCs were infected with EHEC at varying MOIs and incubated at 37°C with gentle shaking (1000 rpm) for 24 hours before analysis. To determine whether the observed effects were specifically mediated by EhxA rather than Stx, RBCs were treated with purified Stx2 (List Biologicals, 162L) (10). A FACS Calibur system (Becton Dickinson) equipped with a 488-nm argon-ion laser was used for all measurements. PS exposure was detected using Annexin V–FITC (BD Bioscience, 556419, 1:20 dilution), and RBCs and MVs were identified using anti–glycophorin-A-PE (BD Bioscience, 555570, 1:20 dilution). RBCs and MVs were distinguished based on forward scatter height for size and side scatter height for granularity. Data were collected from 5000 events and analyzed using Cell Quest Pro software version 6.0.

Measurement of intracellular Ca²⁺ levels and caspase-3 activity

To measure intracellular calcium levels, RBCs were incubated with 3 μM Fluo-4 acetoxymethyl ester (Fluo-4 AM, Thermo Fisher Scientific, F14201) for 1 hour at 37°C in darkness, followed by infection with EHEC for 24 hours The fluorescence intensity of Fluo-4 AM was recorded via flow cytometry.

For caspase-3 activity analysis, RBCs were infected with EHEC at varying MOIs and incubated at 37°C with gentle shaking (1000 rpm) for 24 hours After infection, 300 μl of RBC suspension was collected and mixed with 1 μl of FITC-DEVD-FMK (Calbiochem, 9499), a fluorescent caspase-3 marker. The mixture was then incubated at 37°C with gentle shaking (1000 rpm) for 30 min in darkness. Following incubation, cells were centrifuged at 1000g for 5 min, washed twice with wash buffer (Biovision, 1210), and resuspended in 500 μl of Ringer’s solution. Fluorescence intensity was measured using flow cytometry, with data collected from 5000 events.

Measurement of phospholipid translocation

Phospholipid translocation was assessed using a protocol adapted from Noh et al. (55). RBCs at a density of 1 × 10⁸ cells/ml were infected with EHEC at varying concentrations and incubated at 37°C with gentle shaking (1000 rpm) for 24 hours. To evaluate scramblase activity, 1 μM C6-NBD-PC (Avanti Research, 810132P) was added to the suspension, while 1 μM C6-NBD-PS (Avanti Research, 810194P) was used for flippase activity. Samples were incubated at 37°C, and aliquots were collected at 10 and 30 min. After incubation, samples were chilled in ice-cold tris buffer for 10 min with or without 1% bovine serum albumin (BSA). The degree of phospholipid internalization was calculated by comparing the fluorescence intensity of cells before (no BSA) and after (with BSA) back-extraction. A total of 5000 events were recorded per sample and analyzed using Cell Quest Pro software.

Microscopy analyses

Transmission electron microscopy

RBCs were infected with EHEC for 24 hours and fixed overnight in 2% glutaraldehyde at 4°C. The following day, the samples were rinsed three times with PBS and postfixed with 1% osmium tetroxide for 30 min. After a brief rinse with PBS, RBCs were dehydrated in a graded ethanol series (30 to 90%), followed by three washes with 100% ethanol. The dehydrated cells were treated twice with propylene oxide for 10 min each and infiltrated with a 1:1 mixture of propylene oxide and Spurr’s resin for 2 hours. The samples were left in a desiccator overnight with Spurr’s resin alone. On the third day, they were further infiltrated with fresh resin for 2 hours, then polymerized in a 70°C oven overnight. Processed samples were examined using a JEM-1010 TEM (JEOL).

Scanning electron microscopy

For SEM analysis, RBCs were infected with EHEC for 24 hours, then fixed in 2% glutaraldehyde for 1 hour at 4°C. Fixed cells were centrifuged, washed three times with PBS, and postfixed in 1% osmium tetroxide for 30 min at room temperature. After two PBS rinses, the cells underwent sequential dehydration using ethanol (50 to 100%). Once dehydrated, samples were coated with gold and analyzed using a Merlin Compact FE-SEM (Zeiss).

Confocal microscopy

RBCs were diluted to 3 × 10⁶ cells/ml and seeded onto four-well chambered coverslips (Thermo Fisher Scientific, 155383). Cells were incubated at room temperature for 1 hour for attachment, and unattached RBCs were removed by washing with Ringer’s solution containing 1% BSA before infection with EHEC.

To investigate the role of calcium in EHEC-induced RBC morphological alterations, RBCs (1.5 × 10⁶ cells per well) were pretreated 1 hour before infection with either Ringer’s solution containing calcium (Ca²⁺ group) or Ringer’s solution without calcium supplemented with 300 mM EGTA (final concentration 3 mM; EGTA group). After pretreatment, cells were washed with Ringer’s solution containing 1% BSA and infected with EHEC (500 μl per well).

To further explore whether PS exposure directly affects RBC morphological alterations, purified annexin V (150 μl, final concentration 7 μM, BioLegend, 640902) was added to RBCs after washing with Ringer’s solution containing 1% BSA. Annexin V binds to PS and inhibits its externalization, preventing downstream effects. Cells were incubated with annexin V in the dark for 5 min, followed by EHEC infection (350 μl) for 24 hours at 37°C.

After infection, RBCs from all experimental conditions were stained with anti–glycophorin-A-PE (1:100 dilution) for 10 min, washed with Ringer’s solution, and visualized using a TCS SP8 confocal microscope (Leica) equipped with an argon laser. The excitation and emission wavelengths were set to 488 nm and 550 to 600 nm, respectively. Captured images were analyzed to evaluate RBC morphology, focusing on the roles of calcium and PS exposure in EHEC-induced morphological alterations.

Calcium chelation assay

To assess the role of calcium in EHEC-induced procoagulant activity, RBCs were preincubated with EGTA at a final concentration of 3 mM for 30 min at 37°C with gentle shaking (1000 rpm) in darkness. The cells were then infected with EHEC and incubated for 24 hours under the same conditions. Following incubation, the effects on PS exposure, caspase-3 activity, MV generation, and thrombin generation were analyzed as described in the respective sections.

Transcriptome analysis

Freshly prepared RBCs were exposed to EHEC EDL933 at an MOI of 100 and incubated at 37°C for 24 hours Total RNA was extracted from EHEC cells following the protocol described by Im et al. (56). RNA libraries were prepared using the TruSeq Stranded mRNA Library Prep Kit (Illumina) according to the manufacturer’s instructions and sequenced on the Illumina NovaSeq 6000 platform. The sequencing reads were aligned to the complete reference genome of EHEC EDL933, available under GenBank accession numbers CP008957 and CP008958 (https://ncbi.nlm.nih.gov/). Gene expression levels were calculated as relative log expression (RLE) values. The RLE values were normalized and analyzed statistically using DESeq2 v.1.26.0 to identify differentially expressed genes (P < 0.05) in response to RBC exposure.

RBC functional assay

Prothrombinase assay

RBCs were infected with EHEC for 24 hours at 37°C with gentle shaking (1000 rpm) and subsequently washed with Ringer’s solution containing 1 mM CaCl₂. The washed RBCs were incubated with 5 nM factor Xa and 10 nM factor Va (Hematologic Technologies, HCXA-0060 and HCVA-0110, respectively) in Tyrode’s buffer (134 mM NaCl, 10 mM HEPES, 5 mM glucose, 2.9 mM KCl, 1 mM MgCl₂, 12 mM NaHCO₃, 0.34 mM Na₂HPO₄, 0.3% BSA, and 2 mM CaCl₂; pH 7.4) for exactly 3 min at 37°C.

Purified human prothrombin (factor II, 2 μM; Hematologic Technologies, HCP-0010) was then added, and the suspension was incubated for an additional 3 min. A 10-μl aliquot was transferred to 490 μl of stop buffer (50 mM tris-HCl, 120 mM NaCl, and 2 mM EDTA; pH 7.9) to halt the reaction. Thrombin generation was quantified by measuring absorbance at 405 nm, as described previously (26). Thrombin activity was expressed in units (U), where 1 μmol thrombin generated per minute was defined as 1 U. This definition was applied both in vitro per 10⁶ RBCs and ex vivo per microliter of whole-blood cells.

Adherence of RBCs to ECs and RBC self-aggregation

HUVECs (Lonza, C2517A) were seeded at 2 × 10⁴ cells per well in a 24-well plate and prelabeled with Calcein Green (Invitrogen, C3100MP) before the experiment. RBCs were infected with EHEC for 24 hours at 37°C with gentle shaking (1000 rpm), followed by a washing step with Ringer’s solution, and then added to prestained HUVECs. Coincubation was performed at 37°C for an additional 30 min. Nonadherent RBCs were washed off using EC growth medium (EGM-2, Lonza). RBCs adhering to HUVECs were stained with anti–glycophorin-A-PE (1:100 dilution) and visualized using fluorescence microscopy (Axiovert 200 M, Zeiss). For self-aggregation analysis, EHEC-infected RBCs were stained with anti–glycophorin-A-PE (1:100 dilution) and analyzed by fluorescence microscopy to evaluate their aggregation behavior.

Purified EhxA preparation and treatment of RBCs

Recombinant enterohemolysin EhxA (EDL933) carrying a C-terminal His₆ tag (pET21a-EhxA) was produced under contract by Bionics (Seoul, Republic of Korea). For toxin-exposure assays, human RBCs were incubated with purified EhxA at 1, 2.5, 5, or 10 μg/ml for 24 hours at 37°C. Intracellular Ca²⁺, PS exposure, hemolysis, MV generation, and thrombin generation were quantified using the same procedures as for bacteria-infected RBCs.

Subcellular localization of EhxA toxin on RBC membranes

Human RBCs prepared as described in the “Microscopy analyses” section were incubated with purified EhxA toxin for 24 hours at 37°C under the same conditions. After incubation, cells were washed, fixed (2% glutaraldehyde), permeabilized (0.1% Triton X-100), and blocked (1% BSA). EhxA was detected using a monoclonal anti-6×His antibody (AD1.1.10) followed by an FITC-conjugated secondary antibody (5 μg/ml). Confocal microscopy was performed on a K1-Fluo laser-scanning confocal microscope (Nanoscope Systems, Daejeon, Republic of Korea). RBCs were imaged in bright field, and the localization of EhxA was visualized by FITC fluorescence.

Animal experiments

All animal experiments were conducted with the approval of the Ethics Committee of the Animal Service Center at Seoul National University (SNU-170417-27-5).

To validate the findings from human RBCs and assess the feasibility of using rat RBCs for further analysis, rat RBCs were isolated from whole-blood treated with citrate-dextrose, collected from the abdominal aorta of urethane-anesthetized male Sprague-Dawley rats (280 to 350 g). RBC preparation followed the protocol established for human RBCs, and hemolysis, PS exposure, and thrombin generation were assessed following a 6 hours in vitro infection with EHEC. This preliminary experiment was conducted to establish a basis for subsequent ex vivo and in vivo studies.

Ex vivo studies

For ex vivo analysis, whole blood was collected from the abdominal aorta 3 hours after intravenous infection with EHEC. Morphological changes in RBCs were analyzed using SEM, as previously described (26).

In vivo studies

Three hours after intravenous infection with EHEC, the abdomen of the rats was surgically opened to expose the inferior vena cava (IVC). Two sections of the IVC, approximately 16 mm apart, were loosely tied, while all side branches were tightly ligated using cotton threads. Thromboplastin (Instrumentation Laboratory, 0020002950) was infused into the IVC for 1 min to initiate thrombus formation. Both ends of the IVC were then tightly tied, and the abdominal cavity was temporarily closed to maintain blood stasis for 15 min. After this period, the IVC segments were excised, opened longitudinally, and the formed thrombus was either weighed or imaged for area analysis.

For thrombus area analysis, rats were imaged at 1× magnification using a Stemi 305 stereomicroscope equipped with an Axiocam 105 color camera (Zeiss). The pixel-to-mm ratio was calibrated to ensure that the measured area was expressed in mm², and the area of each thrombus was subsequently measured using ImageJ software. For thrombus weight measurement, independent experiments were conducted using a different group of rats. The thrombi were extracted immediately after excision and weighed.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 10 (GraphPad Software, CA). For multiple-group comparisons, data were analyzed by one-way analysis of variance (ANOVA) followed by Dunnett’s post hoc test. Comparisons between two groups at the same MOI, such as WT- and deletion mutant–infected RBCs, were performed using a two-tailed Student’s t test. In experiments involving the complementation strain, one-way ANOVA followed by Dunnett’s post hoc test was used for comparisons with the control, whereas one-way ANOVA followed by Tukey’s post hoc test was applied for comparisons among WT-, deletion mutant–, and complemented strain–infected groups at the same MOI. Data are presented as mean ± SEM. Significance is indicated as ns (not significant); *P < 0.05, **P < 0.01, and ***P < 0.001.

To evaluate the relationship between PS exposure and RBC morphological alterations, SLR analysis was performed. MLR analysis was conducted to assess the relative contributions of PS exposure, bacterial MOI, and mutant strain status to RBC morphological changes. Similarly, SLR and MLR were used to determine the association between PS exposure and thrombin generation. Regression coefficients (β), R² values, and P values were reported to quantify statistical significance. All statistical tests were two-tailed, and P values <0.05 were considered statistically significant.

Supplementary Materials

This PDF file includes:

Figs. S1 to S15

Tables S1 to S7

References