Targeting the tissue factor coagulation initiation complex prevents antiphospholipid antibody development

The antiphospholipid syndrome (APS) remains a clinically challenging acquired thrombophilia associated with autoimmune diseases and often severe morbidity and mortality.^1-3 APS is diagnosed based on circulating antiphospholipid antibodies (aPL) which are reactive with the phospholipid (PL) cardiolipin and coagulation regulatory proteins, foremost β2-glycoprotein I (β2GPI),^4,5 protein C (PC), and protein S.⁶ aPL also prolong clotting times in the lupus anticoagulant assay by interacting with procoagulant PLs, prothrombin, and β2GPI.⁷ The combination of persistent autoantibodies in systemic lupus erythematosus (SLE) defines a patient subgroup with high mortality.⁸ Although diagnosed by cardiolipin binding, aPL recognize the structural endolysosomal phospholipid lysobisphosphatidic acid (LBPA, or bis[monoacylglycerol]phosphate).⁹ The endothelial cell–protein C receptor (EPCR) exchanges bound phosphatidylcholine to LBPA after ALIX-dependent endosomal retrieval and presents LBPA on the cell surface for recognition by aPL.¹⁰ Lipid-reactive aPL, including those with protein cross-reactivity,^11,12 engage EPCR-LBPA to promote prothrombotic responses and autoimmunity in vivo.¹⁰ 

Coagulation initiation by tissue factor (TF) is central to the thrombotic complications in APS.¹³ In addition, complement activation contributes to thrombosis and fetal loss in APS¹⁴ and complement mutations are associated with catastrophic APS.¹⁵ Complement is directly involved in activating monocyte-expressed cryptic TF by protein disulfide isomerase–dependent thiol-disulfide exchange,¹⁶ which also regulates the APS autoantigen β2GPI.¹⁷ Complement C3 activation triggered by aPL not only renders TF prothrombotic¹⁸ but also contributes to proinflammatory signaling in monocytes and endothelial cells.^19,20 Mechanistically, aPL bind to EPCR which functionally associates with the coagulation initiation complex of TF coagulation factor (F) VIIa and FXa in protease-activated receptor (PAR) signaling.^21,22 Signaling and prothrombotic activity of this complex are inhibited by TF pathway inhibitor (TFPI)²³ and binding of aPL to EPCR-LBPA dissociates the cell surface TF-FVIIa-FXa-TFPI complex on monocytes. This liberates FXa for thrombin-dependent PAR1/PAR2 heterodimer signaling¹⁹ leading to procoagulant acidic sphingomyelinase activation.¹⁰ Furthermore, TF-FVIIa endosomal trafficking¹⁹ leads to proinflammatory toll-like receptor (TLR)–7 signaling.^24,25 

The engagement of EPCR-LBPA by aPL on dendritic cells (DCs) also triggers TLR7-dependent type I interferon (IFN) production that activates PL-reactive B1 cells in an autoamplification loop required for persistent aPL production in experimental APS and SLE.¹⁰ In addition to autoimmune diseases, coagulation activation by TF is a hallmark of viral infections²⁶ and a major contributor to thromboinflammation in severe COVID-19.^27-29 aPL develop transiently in various infectious diseases³⁰ and are prothrombotic when isolated from patients with COVID-19 who were hospitalized ³¹ through specific engagement of EPCR-LBPA.²⁰ Circulating PL-reactive B cells in patients with COVID-19 are phenotypically similar to circulating B1 cells in mouse models of APS and SLE but do not react with β2GPI.^10,20 Here, we show that the TF coagulation initiation complex not only participates in aPL-induced thrombosis but also promotes persistent aPL production. We furthermore characterize the effects of targeted TF inhibition on immune-cell activation and the prevention of aPL development in latent viral infection and autoimmune disease.

Detailed lists of reagents, suppliers, and primer sequences are given in the supplemental Tables 1 - 3, available on the Blood website. Human monoclonal aPL representative of patient reactivities HL5B (cardiolipin-reactive), rJGG9 (β2GPI-reactive), and HL7G (dual reactivity) were purified under endotoxin-free conditions and characterized extensively.¹⁰ Patient immunoglobulin fractions were purified by ammonium sulfate precipitation and MiniTrap G25 columns.²⁰ 

Mutant and genetically matched wild-type (WT) mice were bred under identical housing conditions in the central animal facility of the University Medical Center Mainz or at TSRI. Mice were from the following sources: MRL-lpr (MRL/MpJ-Fas^lpr/J, JAX); C57BL/6J TF cytoplasmic domain deleted (TF^ΔCT) mice, EPCR^C/S (Procr^C/S) and strain-matched EPCR^WT mice;^10,19 gp47phox^−/− and gp91phox^−/− (B6N.129S2-Ncf1^tm1Shl/J and B6.129S-Cybb^tm1Din/J, JAX). Experiments with age- and sex-matched mice were approved by the TSRI IACUC or Landesuntersuchungsamt Rheinland-Pfalz, Koblenz, Germany (no. 23-177-07/G20-1-067, no. 23-177-07/G22-1-025, no. 23-177-07/G18-1-003S, and no. 23-177-07/G19-1-089E2).

aPL-amplified thrombus development by 10 μg immunoglobulin G (IgG) purified from patients with COVID-19 was evaluated with and without inhibition by 1 μg/g bodyweight nematode anticoagulant protein c2 (NAPc2) subcutaneously by high-speed fluorescence video microscopy on an Olympus BX51WI, as described.²⁰ 

Experimental APS was induced by repeated subcutaneous immunizations with human β2GPI or by immunization with aPL or control IgG in complete Freund adjuvant into 1 footpad, followed 2 weeks later by a boost with incomplete Freund adjuvant in the contra-lateral footpad.^10,32 

TF^ΔCT and WT C57BL/6J control mice, as well as EPCR^C/S and strain-matched EPCR^WT mice were infected with 2 × 10⁵ PFU murine cytomegalovirus (mCMV) (Smith strain, ATCC-VR1399) intraperitoneally and received 10 subcutaneously doses of 0.5 μg/g NAPc2 in 100 μL phosphate-buffered saline (PBS) or vehicle control every other day. Cytokine levels were measured with the LEGENDplex Mouse Anti-Virus Response Panel (13-plex). Viral DNA load was determined by M55-specific and pthrp-specific quantitative polymerase chain reactions (PCRs) normalized to a log₁₀-titration of standard plasmid pDrive_gB_PTHrP_Tdy, as described.³³ 

The effect of NAPc2 on lupus-like autoimmune disease was evaluated in MRL-lpr mice³⁴ receiving 0.5 mg/kg rNAPc2 or saline control every other day for 3 weeks starting at an age of 11 weeks. Lymphadenopathy was scored at 3 sites (cervical, axillary, axillary or brachial, and inguinal (on a scale from 0-4 [none-very severe]). Skin lesions were scored at snout, ears, and neck (from 0-3 [none-severe]). Albumin in spot urine samples was detected with a mouse albumin enzyme-linked immunosorbent assay (ELISA) (Bethyl Laboratories) and normalized to creatinine content with the Creatinine Assay Kit (Creative BioMart).

Renal paraffin sections stained with periodic acid–Schiff reagent were scored for glomerulus pathology in 20 glomeruli per kidney.¹⁰ Interstitial or tubular pathology was assessed semiquantitatively in 10 randomly selected high-power fields. Glomerular and interstitial or tubular scores were averaged to obtain the renal pathology score. CD68, CD4, and B220 cell infiltration was quantified in stained sections and randomly selected high-power fields.

Monocytic MM1 cells received inhibitors (20 μg/mL anti-PAR1 ATAP/WEDE, 50 μg/mL anti-TF 5G9/10H10, and 200 nM NAPc2) 15 minutes before stimulation with 400 ng/mL aPL HL5B and HL7G, 1 μg/mL aPL rJGG9 and control IgG, 10 μg/mL patient immunoglobulin, or 10 ng/mL lipopolysaccharide (LPS) for 1 hour. Endosomal reactive oxygen species (ROS) was detected with 10 μM fluorescent probe H₂DCFDA by flow cytometry using a FACS Canto.¹⁹ Messenger RNA (mRNA) expression analysis by real-time PCR on the iCycler iQ thermal cycler (Bio-Rad) was normalized to glyceraldehyde-3-phosphate dehydrogenase.¹⁰ DC stimulation and IFNα measurements were performed as described.²⁴ 

RNA isolated from MM1 cells was sequenced at the Max Delbrück Center for Molecular Medicine, Berlin, Germany, and next-generation sequencing (NGS) data were mapped to GRCh38.p13 with STAR on the fly. Count matrices were generated and transcripts with expression <10 were removed and analyzed for differential expression with the DESEQ2 package (version 4.2.0). After log normalization, differentially expressed genes were volcano plotted with ggplot2. Top differentially expressed genes were clustered with the Euclidean distance algorithm and log2-fold changes between groups were visualized with Pheatmap, using an adjusted P value <.05 as threshold.

NGS sequencing of MRL-lpr DCs used spleens digested with collagenase II (Worthington) and DNase I (Sigma-Aldrich) for 30 minutes at 37°C followed by cell dissociation with 10 mM EDTA and enrichment on an Optiprep density gradient (Sigma-Aldrich). Dead cells were stained with 7-aminoactinomycin D for cell sorting on an ARIA II cell sorter (Becton Dickinson) and live DCs identified by major histocompatibility complex II and CD11c in the CD3/CD19/NK1.1 negative population were lysed in 350 μL RLT lysis buffer (Qiagen), 10 μL β-MeEtOH and stored at −80°C. RNA was sequenced by Novogene and mapped to reference genome Mus_musculus.GRCm38.96 with STAR2.7.2. After quality control, low-count gene removal (gene expression <10), DESEQ2 analysis and normalization, gene set enrichment analysis was performed with the package Clusterprofiler version 4.8.2.

Immunoglobulin fractions or mouse sera were analyzed using ELISA plates coated overnight with 2 μg/mL PL in ethanol at 4°C, blocked with 1% Tween 20, and incubated with 40 μg/mL of immunoglobulin fractions or mouse sera in PBS, 0.1% Tween 20. Bound antibody was detected with horseradish peroxidase (HRP)-coupled anti-IgG secondary antibody and tetramethylbenzidine substrate. Binding to β2GPI was analyzed in a commercially available ELISA plates (AESKU Diagnostics, Wendelsheim, Germany).

aPL producing B cells were analyzed in peritoneal lavage or EDTA blood by PL vesicles or β2GPI staining and gating for live B1 cells.²⁰ Changes in B1 cells in the bone marrow and peritoneal lavage were quantified with standardized multicolor flow cytometry using antibodies listed in the supplemental Table 4. Cells were washed with PBS or bovine serum albumin (0.5%), measured on an Attune Nxt flow cytometer (ThermoFisher), and data analyzed with FlowJo 7.2 software (TreeStar Inc).

Mean and standard deviations of biological replicates are shown. GraphPad Prism 9.5.0 was used for statistical analysis by 2-way analysis of variance, followed by Sidak multiple comparison test for multiple groups or 2-tailed unpaired Student t test for 2 groups.

TF-initiated thromboinflammation is a hallmark of severe COVID-19^27-29 and COVID-19 aPL induce thrombosis in mice.^20,31 We analyzed the role of TF with the specific TF inhibitor NAPc2 that locks the TF-FVIIa-FXa complex in a TFPI-like manner.^35,36 NAPc2 inhibits TF pathological functions in mouse and primate models^37-41 and coagulation activation in humans.^42-44 In an established model of aPL-enhanced thrombosis in the flow-restricted vena cava inferior,^10,18-20,45 NAPc2 markedly reduced thrombosis induced by prothrombotic immunoglobulin fractions isolated from patients with COVID-19 and known to engage EPCR-LBPA²⁰ (Figure 1A). Thus, COVID-19 aPL induced TF-dependent thrombosis.

To better understand the inhibitory function of NAPc2 on aPL–induced immune cell activation, we stimulated human monocytic cells with aPL derived from patients with COVID-19 or prototypic aPL monoclonal antibodies⁴⁶ in human serum (Figure 1B), because NAPc2 has very low affinity for bovine FX/Xa.⁴⁷ Except rJGG9 reactive only with β2GPI,^10,48 COVID-19 and monoclonal aPL HL5B and HL7G reactive with EPCR-LBPA induced primarily proinflammatory tumor necrosis factor (TNF) and TF. We had previously shown that mouse monocytes also upregulated IFN-regulated transcripts (GBP6 and IFN-regulated factor [IRF] 8) in response to aPL.¹⁰ A repeat of the experiment in plasma showed that lipid-reactive aPL and LPS now robustly induced IFN-regulated transcripts (Figure 1B), suggesting that a labile factor required for IFN responses was severely reduced in the preparation of human serum.

Under these experimental conditions, NAPc2 also markedly attenuated induction of TF and TNF, but not GBP6 and IRF8 by COVID-19 aPL (Figure 1C). Although NAPc2 inhibits TF-FVIIa in a TFPI-like manner requiring FX/Xa, NAPc2 does not block the active site of FXa and permits TF-FVIIa-FXa–mediated proteolytic activation of PARs⁴⁹ and other substrates of the nascent product FXa.⁵⁰ Importantly, NAPc2 also maintains the signaling conformation of TF-FVIIa-FXa required for TLR4/PAR2-dependent IFN responses involving IRF7 in DCs.²² Addition of the FXa inhibitor NAP5 together with NAPc2 prevented the induction of IFN-regulated genes by aPL (Figure 1C), confirming that FXa mediates the IFN responses in the NAPc2-inhibited complex.

To fully understand the inhibitory properties of NAPc2, we analyzed the genome-wide transcriptional response induced by 2 prototypic human monoclonal aPL—HL5B solely reactive with EPCR-LBPA and HL7G reactive with EPCR-LBPA and β2GPI. Principal component analysis of transcript changes in NGS data (Figure 2A) showed that aPL HL5B and HL7G induced a very similar response distinct from LPS stimulation. The aPL HL5B (Figure 2B) and HL7G (Figure 2C) induced an overlapping response of IFN-regulated transcripts, including IRF1 and IRF7, proinflammatory TNF and interleukin (IL) 1β, which were found in the top 100 transcripts induced by aPL HL5B vs control IgG (Figure 2D). These similarities of prototypic aPL indicated that the cross-reactivity of aPL HL7G with the plasma protein β2GPI did not influence the early immune cell activation after engagement of EPCR-LBPA.

Remarkably, NAPc2 altered, but did not revert to baseline, the transcriptional profiles induced by aPL HL5B and HL7G (Figure 2A,D). The comparison of NAPc2 effects on the transcriptional profiles of aPL HL5B (Figure 2E) and HL7G (Figure 2F) stimulated cells revealed a suppression of proinflammatory responses and of TF induction by NAPc2, in line with the targeted transcript analysis with COVID-19 aPL (Figure 1C). In contrast, IFN-regulated transcripts were still induced by aPL HL5B (Figure 2G) and HL7G (Figure 2H) in the presence of NAPc2 in comparison with IgG control. Thus, stabilization of the TF-FVIIa-FXa complex by NAPc2 on aPL-stimulated monocytes suppressed prothrombotic and proinflammatory effects but not the induction of IFN responses implicated in antimicrobial host defense.

We next asked whether NAPc2 might also interfere with the development of aPL. The proinflammatory activation of monocytes by aPL requires endosomal ROS production and TF-FVIIa–mediated translocation of gp91phox to the endosome.¹⁹ In line with the observed inhibition of proinflammatory responses by NAPc2 (Figure 2), NAPc2 was as potent as the inhibitory anti-TF 10H10^19,51 or cleavage blocking anti-PAR1¹⁹ in preventing aPL HL5B-induced ROS generation (Figure 3A). Although the prothrombotic⁴⁵ and proinflammatory⁴⁶ effects of aPL are dependent on the nicotinamide adenine dinucleotide phosphate (NADPH) oxidase, the role of the NADPH oxidase complex in the development of APS is unknown. In an established experimental model of APS,¹⁰ we found that aPL reactive with cardiolipin (Figure 3B) or β2GPI (Figure 3C) did not develop in mice lacking components of the NADPH oxidase complex, p47phox and gp91phox. In this model, cardiolipin-reactive aPL develop dependent on TLR7 and EPCR¹⁰ and β2GPI-reactive aPL development was also abolished in EPCR^C/S mice defective in LBPA-loading or Tlr7^−/−, but not Tlr9^−/− mice (supplemental Figure 1A-B), validating this prothrombotic APS model⁵² as highly relevant for autoimmune aPL development. Experimental APS can also be induced by immunization with β2GPI.^32,52 In this alternative model, aPL reactive with LBPA and cell surface–expressed EPCR-LBPA and capable of proinflammatory and procoagulant monocyte activation also did not develop in EPCR^C/S mice (supplemental Figure 2), indicating that experimental APS induction by immunization with aPL or β2GPI evokes similar autoimmune responses.

These data implicated TF-mediating endosomal translocation of the NADPH complex in aPL development. To further analyze aPL development in a pathological setting, we asked whether viral infections associated with activation of IFN responses could stimulate aPL production by B1 cells. Indeed, infection with mCMV caused a persistent production of aPL reactive with LBPA (Figure 3D) and β2GPI (Figure 3E) appearing during viral latency. These antibodies did not develop in EPCR^C/S mice that were defective in aPL signaling.¹⁰ The TF cytoplasmic domain couples to the NADPH oxidase complex^53,54 and is required for endosomal ROS production in aPL signaling.¹⁹ We, therefore, analyzed the development of aPL in mCMV-infected TF cytoplasmic domain deleted (TF^ΔCT) mice. TF^ΔCT mice did not develop autoantibodies (Figure 3D-E) or circulating PL-reactive B1 cells (Figure 3F). These data implicated TF and especially TF cytoplasmic domain signaling in the development of APS.

EPCR-LBPA recognition by aPL on DCs is also required for IFN signaling–dependent expansion of B1 cells with PL reactivity.¹⁰ Stimulation with aPL HL5B or COVID-19 aPL induced IFNα mRNA and protein secretion from DCs (supplemental Figure 3A-B). IFNα, but not GPB6 or IRF7 mRNA induction was markedly reduced in the presence of NAPc2, indicating that the TF-induced IFN responses were independent of autocrine IFNα signaling. We next asked whether the inhibition of the TF initiation complex by NAPc2 could prevent autoantibody production in latent mCMV infection. We treated mice during the acute viral infection from day 0 to 20 with NAPc2. Cytokine levels peaked at 42 hours and returned to baseline 8 days after infection (Figure 4A). NAPc2 treatment significantly reduced CXCL1, CCL2, CXCL10, IL12, IFNγ, and IFNα levels 42 hours after infection. NAPc2 treatment did not impair the control of acute infection as indicated by the viral genome load in lung, spleen, liver, and salivary gland which was indistinguishable between saline- and NAPc2-treated mice 21 days after infection (Figure 4B). Remarkably, treatment with NAPc2 attenuated the persistent production of aPL with cardiolipin reactivity (Figure 4C) and suppressed the appearance of circulating B cells reactive with PL vesicles (Figure 4D-E). These PL-reactive B cells circulating at day 10 after infection had the phenotypic makers CD5, CD20, CD27, and CD43 characteristic for B1 cells and were IgG class-switched (Figure 4D), a feature of B1 cells exiting the peritoneum.⁵⁵ We also detected PL-reactive and IgG⁺ B1 cells in the peritoneal cavity and the blood 21 days after infection (supplemental Figure 3C), and NAPc2 suppressed the expansion of these B1 cells not only in the blood, but also the peritoneal cavity (supplemental Figure 3D). In addition, short-term treatment with NAPc2 for 20 days suppressed LBPA- and β2GPI-reactive aPL 12 weeks after infection (Figure 4F). Thus, targeting the TF initiation complex with NAPc2 prevented the development of aPL in latent viral infection without apparent impairments of antiviral immunity.

Viral sensing and autoimmune signaling pathways overlap in disease development.⁵⁶ We, therefore, evaluated NAPc2 therapy in an established model of autoimmunity. MRL-lpr mice develop a TLR7-dependent⁵⁷ SLE-like syndrome and aPL¹⁰ at an age of 5 to 6 weeks.¹⁰ The severe SLE-like syndrome progresses and requires euthanization of mice before they reach an age of 15 to 16 weeks. We treated MRL-lpr mice at an age of 11 weeks with 0.5 mg/kg rNAPc2 or saline control every other day for 3 weeks. At the time of randomization to therapy, both groups had similar anticardiolipin aPL titers, but aPL titers rather declined in rNAPc2-treated mice, whereas they increased in mice receiving saline (Figure 5A). In a second cohort, MRL-lpr mice were similarly randomized for treatment and the reactivity of serum samples against cardiolipin or β2GPI markedly increased over time in saline-treated mice, whereas mice randomized to receive rNAPc2 therapy continued to produce very low autoantibody titers (Figure 5B).

The development of lymphadenopathy was also attenuated in NAPc2-treated animal vs those treated with saline (Figure 5C) and only 1 of 5 of the NAPc2-treated mice had visible skin lesions at the end of the experiment as compared with 4 of 5 in the control group. NAPc2 treatment had no effect on peripheral blood cell counts (supplemental Figure 4). To evaluate the immune modulatory effect of NAPc2, we isolated splenic DCs by flow cytometry for NGS sequencing. Treatment with NAPc2 altered the expression profile of DCs (Figure 5D). Biological pathways analysis showed a suppression of inflammatory responses in DCs isolated from NAPc2-treated mice (Figure 5D). We next addressed the effects of NAPc2 on B1 cells in the peritoneum. Although NAPc2 treatment did not influence CD19⁺/CD43⁺ B1 cell abundance (supplemental Figure 5A-B), NAPc2 increased the proportion of peritoneal B1 cells expressing CD5, which is a negative regulator of B-cell receptor signaling.⁵⁸ In contrast, NAPc2 did not alter B1 phenotypes in the bone marrow (supplemental Figure 5C), indicating that NAPc2 primarily altered immune pathology in the periphery.

Kidney pathology in the NAPc2 treatment cohorts revealed a significantly improved renal pathology score, a composite of glomerular and tubular injury, in rNAPc2- vs control-treated mice (Figure 5E). Treatment with rNAPc2 also significantly reduced kidney immune cell infiltration, based on scoring for macrophages (CD68), T cells (CD4), and B cells (B220) by immune-histochemistry (Figure 5E). In addition, renal function assessed by measuring albumin in the urine showed that sham-treated control mice developed more albuminuria in comparison to rNAPc2-treated mice (Figure 5F). These data showed that NAPc2 prevented not only the development of aPL but also disease severity in MRL-lpr mice.

Current clinically used antithrombotic therapies have limited effects on aPL titers. We confirmed this observation in a single center cohort without antithrombotic therapy and under heparin or vitamin K–antagonist therapy (Figure 6A). The measured aPL titers in each treatment group did not change significantly over the observation period of 1 year. We next treated MRL-lpr mice with either NAPc2 or low-molecular weight heparin, using a previously established effective dosing scheme.⁵⁹ Treatment with NAPc2, but not with heparin, suppressed anticardiolipin and anti-β2GPI titers (Figure 6B). At the end of the treatment experiment, circulating CD19⁺CD27⁺CD43⁺ B1 cells (Figure 6C) demonstrated binding of both autoantigens, labeled PL vesicles, and β2GPI (Figure 6D). As expected from the measured aPL titers in these mice, sham- and heparin-treated MRL-lpr mice showed a high percentage of PL and β2GPI-reactive circulating B1 cells which were not detected in NAPc2-treated mice (Figure 6D). PL vesicle staining was blocked by the pathogenic target of aPL, LBPA-loaded soluble EPCR (sEPCR-LBPA) as well as unlabeled β2GPI, but not soluble EPCR loaded with phosphatidylcholine (sEPCR-PC) (Figure 6D). Conversely, β2GPI staining was specifically prevented by sEPCR-LBPA as well as unlabeled PL vesicles, indicating that circulating B1 cells in this mouse model of autoimmunity had dual reactivity with EPCR-LBPA and β2GPI. In addition, these data showed that specific inhibition of TF with rNAPc2 but not anticoagulation with heparin effectively prevented the development of autoimmune antibodies.

In this study, we show that TF plays a crucial role not only in the pathogenic effects of aPL but also in the development of autoreactive B1 cells in latent viral infection and autoimmune disease. We demonstrate in preclinical proof-of-concept experiments that specific blockade of the TF coagulation initiation complex with the nematode-derived inhibitor, NAPc2 prevents the prothrombotic effects of aPL and attenuates the aPL–induced proinflammatory activation of monocytes. The stabilization of the TF initiation complex by NAPc2 not simply inhibits aPL signaling globally but rather preserves the induction of monocyte IFN responses relevant for antimicrobial host defense. This unique biased antagonistic mechanism of action can explain the previously demonstrated beneficial effects of NAPc2 in attenuating inflammation, whereas preserving antiviral immune defense in lethal Ebola infection in primates.³⁸ 

Infection with the single-stranded RNA severe acute respiratory syndrome coronavirus 2 is associated with the development of prothrombotic aPL and the appearance of PL-reactive B1 cell in the blood of patients with severe acute COVID-19.²⁰ We, here, demonstrate that infection with the DNA virus mCMV distinctly causes a delayed development of aPL that react with both autoantigens, EPCR-LBPA and β2GPI. The development of aPL in mCMV-infected mice is prevented in EPCR^C/S mice defective in LBPA lipid presentation¹⁰ or mice lacking the TF cytoplasmic domain, which supports aPL–induced endosomal ROS production.¹⁹ Consistently, functional inhibition of TF by NAPc2 not only attenuates aPL–induced endosomal ROS production in monocytes and IFNα secretion in DCs but also the development of aPL. These data provides initial evidence that TF signaling contributes to the development of autoimmune pathologies in persistent viral infections and suggests a potentially beneficial strategy for therapeutic intervention.

TF inhibition by NAPc2 also effectively prevents aPL development in a relevant preclinical model of SLE, implicating TF signaling as a common pathway linking autoimmune and viral pathologies. In both disease models, PL-reactive B1 cells are found in the blood and disappear after pharmacological intervention with NAPc2. Circulating B1 cells also react with the autoantigen β2GPI. Importantly, B1-cell binding of PL vesicles and β2GPI is specifically abolished by competition with EPCR-LBPA. In addition, unlabeled autoantigens crosscompete the B1-cell staining by the other, indicating that the majority of these circulating B1 cells have dual reactivity, previously demonstrated for human aPL.^10,46 

Therapeutic anticoagulation in APS aims to prevent thromboembolic events but typically fails to attenuate aPL antibody titers. Immunotherapy also has limited effects on aPL antibody titers in contrast to hydroxychloroquine⁶⁰ that interferes with NADPH-dependent endosomal proinflammatory aPL signaling.⁶¹ The challenges in antithrombotic therapy is furthermore highlighted in patients with triple-positive APS who experienced more arterial thromboembolic complications in a randomized trial with oral FXa inhibitors vs vitamin K–antagonist,⁶² emphasizing on the need for improved therapies. Here, we show that anticoagulation with heparin is ineffective in suppressing the development of circulating autoreactive B1 cells. In contrast, NAPc2 suppresses the appearance of aPL and PL-reactive B1 cells in blood and the peritoneal cavity, a major location for B1 cell in the body. Although NAPc2 has no effect on B1 cells in the bone marrow, NAPc2 increased CD5 expression in peritoneal B1 cells in SLE mice. This finding suggests that innate immune cell targeting by NAPc2 may counteract B1-cell activation and rebalance CD5-dependent regulation of autoimmune pathology.

We have previously shown that specific targeting of the autoantigen EPCR-LBPA attenuates the progression of SLE-like autoimmune pathology.¹⁰ Engagement of EPCR-LBPA on DCs sustains a paracrine amplification loop that expands PL-reactive B1 cells and ultimately persistent autoimmune antibody production. Specific targeting of TF in SLE-prone mice also attenuates systemic signs of autoimmune disease and suppresses a proinflammatory gene signature in isolated splenic DCs. Thus, the demonstrated suppression of B1 cell activation leading to aPL production with a clinically well-tolerated TF antagonist⁴⁴ provides a rationale for therapeutic intervention in patients with severe APS and may have broader therapeutic benefits in the prevention of end organ damage in autoimmune diseases.

The authors thank P. Wilgenbus, S. Reyda, and G. Carlino for technical assistance and managing the mouse colonies.

This study was supported by the German Research Foundation (DFG Project Number 318346496, SFB1292/2 TP02 to W.R., TP21 to S.M., TP13 to H.S., and TP11 to N.L.) and (MU4233/1-1 to N.M.-C.), the Humboldt Foundation of Germany (Humboldt Professorship to W.R.), the German Ministry for Education and Technology (Cluster4future curATime to W.R. and N.M.-C.), the German Center for Cardiovascular Research (shared expertise 81X2100161 to N.H. and W.R.), and National Institutes of Health, National Heart, Lung, and Blood Institute grant P01-HL139420 (S.K.). N.L. is a member of the DFG–funded Cluster of Excellence Immunosensation (EXC2151) at the University of Bonn.

Contribution: N.M.-C. designed, performed, and analyzed experiments and wrote the manuscript; K.G., T.S.N., M.M., S.M., H.S., K.S., J.R., N.L., and J.W.-M. designed, performed, and analyzed experiments; J.P., J.B., and N.H. provided bioinformatic analysis; S.K., L.T., and K.J.L. provided crucial reagents and expertise; and W.R. conceptualized and directed the study and wrote the manuscript.

Conflict-of-interest disclosure: W.R. is a consultant for ARCA Biopharma. W.R., N.M.-C., K.L., and J.W.-M. are listed as inventors on intellectual properties for the therapeutic use of NAPc2. The remaining authors declare no competing financial interests.

Correspondence: Wolfram Ruf, Center for Thrombosis and Hemostasis, University Medical Center of the Johannes Gutenberg University Mainz, Langenbeckstr 1, 55131 Mainz, Germany;.