A total of 41 female patients were included in this study: 21 diagnosed with APS and 20 in the control group. According to clinical characteristics, female patients with obstetric APS and thrombosis (n = 11) were divided as follows: refractory primary (RI+, n = 3), refractory secondary [with systemic lupus erythematosus (SLE)] (RII+, n = 2), and non-refractory primary (NR+, n = 6). Refractoriness was defined as obstetric manifestations as patients manifest a new episode of pregnancy morbidity despite optimal pharmacological treatment with heparin and ASA during pregnancy (Mekinian et al., 2017). Patients with vascular thrombosis (n = 10) were classified as primary (VTI, n = 6) and secondary with SLE (VTII, n = 4). The control group included patients who were negative for aPL and with clinical manifestations of thrombosis (n = 10), classified as primary (patients who were negative for aPL with thrombosis without autoimmune disease (VTI/aPL-, n = 5) or secondary with SLE (patients who were negative for aPL with thrombosis and SLE (VTII/aPL-, n = 5) and patients with previous uncomplicated pregnancies [normal human serum (NHS), n = 10]. Exclusion criteria for all study participants were other associated diseases such as infections, diabetes, cancer, or chronic disease other than systemic autoimmune diseases due to the inclusion of patients with secondary APS. Patients were recruited from the Recurrent Pregnancy Loss Program of Reproduction Group (University of Antioquia) and Anticoagulation Clinic (San Vicente Fundación Hospital), with the previous approval of the Ethics Committee from the Medical Research Institute-School of Medicine (University of Antioquia). Informed consent was obtained from each participant. None of the patients were pregnant or presented an acute thrombosis episode when the samples were obtained.

Reagents, materials, and antibodies were acquired from the following manufacturers: Limulus amebocyte lysate, NuncTM filter flask, BODIPY C11, MitoSOX, eFluor, and anti-CD31-FITC from Thermo Scientific (Waltham, MA, United States); Type I collagenase from Invitrogen (San Diego, CA, United States); basal endothelial cell culture medium and fetal bovine serum (FBS) from Promocell (Heidelberg, Germany); RPMI-1640, FBS, Opti-MEM, and PBS from Gibco (Grand Island, NY, United States); penicillin and gentamicin from Genfar (Bogotá, Colombia); human-β2GPI from Louisville APL Diagnostics (Louisville, KY, United States); Amphotericin B, PKH67, dichlorofluorescein diacetate (DCFH-DA), propidium iodide, lipopolysaccharide (LPS), ASA, HCQ, vascular endothelial growth factor (VEGF) 165, and NG-nitro-L-arginine methyl ester (L-NAME) from Sigma (St. Louis, MO, United States); anti-VCAM-1-PE and anti-E-selectin-Alexa Fluor 700, from R&D Systems by Bio-techne (Minneapolis, MN, United States); 4-amino-5-methylamino-2′, 7-difluoroflurescein diacetate (DAF-FM-DA) from Cayman Chemical (Ann Arbor, MI, United States); enoxaparin (ENX) from Procaps (Barranquilla, Colombia); and annexin V and 7-aminoactinomycin D (7-AAD) from BD Pharmigen (San Diego, CA, United States).

The Clinical and Laboratory Standards Institute recommendations were followed for plasma LA determination (Pengo et al., 2009). APTT-SP (Instrumentation Laboratory, Orangeburg, NY, United States) was used to demonstrate the antibody dependence on phospholipids. Dilute Russell’s viper venom time (dRVVT) screen and dRVVT confirm (Instrumentation Laboratory, Orangeburg, NY, United States) were used to detect LA. The presence of aCL and aβ2GPI was evaluated using commercial kits (BioSystems, Barcelona, Spain and Human, Wiesbaden, Germany, respectively) in serum and purified immunoglobulin-G (IgG). IgG from the serum pool of patients in each group was purified by affinity chromatography using protein G-Sepharose (General Electric Healthcare, NY, United States), as described (Alvarez et al., 2017). Endotoxins were detected in purified IgG using the Limulus amebocyte lysate assay.

Human umbilical cords were obtained from female patients with uncomplicated pregnancies who attended obstetric services at the Hospital San Vicente Fundación, Medellín, Colombia. Voluntary female patients signed informed consent. The Ethical Committee of the same Hospital approved the sample collection. Human umbilical vein endothelial cells (HUVECs) were isolated by mechanical and enzymatic digestion based on the modified protocol of Jaffe (Jaffe et al., 1973; Gil-Villa et al., 2020). Briefly, the umbilical vein was perfused with a pericranial needle with 5 ml of type I collagenase followed by 20 min of incubation at 37°C. The umbilical vein content was centrifuged, and the button of cells (including HUVECs) was cultured in a NuncTM filter flask with 10 ml of basal endothelial cell culture medium and 2% of FBS. Different umbilical cords from healthy patients were used to isolate HUVECs included in each experiment. HUVECs were used until passage three. In all the experiments, HUVECs were maintained in Opti-MEM in serum-free conditions to perform the stimuli with IgG and human-β2GPI.

THP-1 cells, derived from a 1-year-old infant with acute monocytic leukemia, were obtained from the American Type Culture Collection (CRL-1593, Manassas, VA, United States) and cultured in RPMI plus 10% FBS and antibiotic solution of 100 IU of penicillin, 50 μg of gentamicin, and 250 ng of amphotericin B at 37°C with 5% CO₂ and 60% relative humidity.

Aspirin at 10 mM, ENX at 50 IU/ml, and HCQ at 1 μg/ml were simultaneously added with IgG to detect the modulating effect on NO release.

The effect of IgG, ASA, ENX, HCQ, and β2GPI on cell viability was evaluated using 7-AAD. The stimuli were performed under the same conditions previously described to detect NO. Following 1-h incubation, the dead cells of 7-AAD+ were detected using an LSR Fortessa flow cytometer (BD). The obtained data were analyzed using the FlowJo^® v7.6.2 program and reported as% dead cells. As a positive control of dead cells, 4- and 8-mM hydrogen peroxide (Sigma, St. Louis, MO, United States) were used.

Data distribution was evaluated using the Shapiro–Wilk tests. Data are expressed as mean ± SD. A t-test, Mann–Whitney test and two-way ANOVAII were performed, and the comparison between means was determined using the Holm–Sidak post-test. Statistical analyses were performed using Prism6 (GraphPad Software, Inc., San Diego, CA, United States). For all the cases, the value of p < 0.05 was considered significant.

No differences were observed in the age of patients of different groups. The samples from patients with VTI, RI+, RII+, and NR+ were positive for aPL in the serum and purified IgG. Contrarily, the group with VTII alone was positive for LA and serum aCL. The RII+ group was positive for aPL; however, it had lower values than patients with refractory and NR+ APS (RI+ and NR+). Patients with RI+ had higher LA values compared with other groups. Two patients in this group had catastrophic APS. The IgG samples of controls (NHS, VTI/aPL-, and VTII/aPL-) were negative for aPL (Table 1). All the tested IgG samples were negative for endotoxins. For all the cases, secondary APS indicated the presence of SLE. All patients with refractory APS had pregnancy morbidity even after ASA and heparin treatment (Table 1). Different characteristics of thrombotic events were observed among patients with thrombosis from which the received treatment was derived (Table 1).

In all the experiments, LPS increased E-selectin and VCAM-1 expressions (Figures 1A,E). In β2GPI-stimulated endothelial cells, the VTI and RI+ group IgG increased E-selectin compared with NHS control IgG (Figures 1B–D). Also, in β2GPI-stimulated endothelial cells, RI+ IgG increased VCAM-1 in contrast with the NHS control IgG (Figures 1F,G).

Monocyte aggregates were not observed in the endothelium regarding basal adhesion and cells treated with VTI/aPL, VTII/aPL, and NHS controls (Figures 2A,C–E). As expected, LPS increased the adhesion of THP-1 monocytes to an endothelial monolayer (Figures 2B,G). Complementarily, RI+ IgG increased the adhesion of monocytes to the endothelium compared with VTI/aPL and VTII/aPL controls (Figures 2D–F,H). The number of aggregates was higher and statistically significant upon RI+ stimulation (Figures 2F,H).

Hydrogen peroxide and LPS increased the endothelial O₂^– production (Figures 3A,B,D). RII+ IgG increased endothelial O₂^– production compared with the NHS and NR+ IgG without β2GPI (Figures 3C,E). No IgG effect was observed in cells treated with β2GPI (Figure 3E). In addition, IgG from the patient groups included in this study and β2GPI alone did not affect the ROS production detected by DCF signal or lipoperoxidation by BODIPY C11 staining (Supplementary Figure 1). IgG from patients with positive or negative aPL and thrombosis (VTI/aPL-, VTII/aPL-, VTI, and VTII) did not affect the generation of oxidative stress in HUVECs (data not shown).

A higher number of 0.5- and 1-μm endothelial microparticles for CD31+/annexin V-, CD31+/annexin V+, CD31-/annexin V+, and CD31-/annexin V- were generated in cells stimulated with LPS compared with basal microparticle production (Supplementary Figures 2A–N). After the initial set up of an experimental protocol, including a potential experimental confounder generated by electronic noise or negligible count of particles in the binding buffer (Figures 4A,B) and using LPS as a positive control to increase the total microparticle count (Figures 4C,D), the effect of IgG from patients with APS was tested. In addition, the microparticle size was estimated using the polystyrene beads of 0.5, 1, and 2 μm (Figure 4E). Contrarily, NR+ IgG with β2GPI increased the generation of 0.5-μm CD31+/annexin V+ microparticles compared with IgG from the same group without β2GPI and NHS IgG, VTII/aPL-, and VTII controls with β2GPI (Figures 5B–F,K). IgG from this group with β2GPI increased the generation of 1-μm CD31+/annexin V- microparticles compared with IgG without β2GPI or NHS control IgG (Figures 5H–J,L). RII+ IgG increased the production of 1-μm endothelial dysfunction biomarkers of the total cell-derived endothelial microparticles. RII+ IgG increased the generation of microparticles (1 μm) in a β2GPI-dependent manner compared with the NHS control (Figures 4F–H).

Non-refractory primary obstetric APS and VTI IgG increased the procoagulant microparticles or Annexin V+ (CD31+/annexin V+ and CD31-/annexin V+) by 0.5 μm (Figure 5M). IgG from the patient groups included in this study did not induce 2-μm microparticles (data not shown). The dot-plot labeled with control indicates microparticles without antibodies, which were used to define the location of negative and positive MPs for Annexin V and CD31 (Figures 5A,G).

Vascular endothelial growth factor-induced NO synthesis in a dose-dependent manner, whereas L-NAME inhibited its production in all the VEGF-used doses (Figures 6A,B). β2GPI alone did not alter NO synthesis in contrast with the baseline control (Figure 6A). RI+ IgG decreased NO synthesis in HUVECs compared with NHS control IgG without β2GPI (Figure 6C). All IgG from patients with pregnancy morbidity and thrombosis (RI+, RII+, and NR+) reduced NO synthesis compared with the NHS IgG control in a β2GPI-dependent manner (Figure 6C). ASA did not induce a modulatory effect on reducing NO induced by IgG from patients with pregnancy morbidity and thrombosis (Figures 6D,E). ENX restored the reduced synthesis of NO generated by RI+ IgG with and without β2GPI (Figures 6F,G). HCQ did not induce a modulatory effect on reducing NO induced by IgG from patients with pregnancy morbidity and thrombosis without β2GPI (Figure 6I). HCQ restored the synthesis of NO reduced by IgG from RI+ with β2GPI (Figure 6H). IgG, ASA, EXN, and HCQ did not induce dead cells compared with baseline control (Supplementary Figure 3).

Refractory secondary obstetric APS and NR+ IgG increased the presence of endothelin-1 compared with NHS IgG without the addition of β2GPI (Figure 7A). RI+ IgG with β2GPI increased the production of endothelin-1 compared with RI+ IgG without β2GPI (Figure 7A). IgG from patients with VTI and VTII increased the VEGF in contrast with NHS IgG in the absence of β2GPI. IgG from patients with NR+ decreased the VEGF levels in contrast with IgG from the NHS in the absence of β2GPI. NR+ IgG with β2GPI increased VEGF levels compared with that without β2GPI (Figure 7B).

Different endothelial activation and dysfunction mechanisms induced by IgG in groups of patients included in our study are associated with the development of APS clinical manifestations. Our main findings include: (1) VTI IgG increased the expression of E-selectin, generated positive procoagulant microparticles, and increased VEGF generation (Figure 8A). (2) RI+ IgG increased the adhesion of monocytes and decreased NO, which were modulated by HCQ and ENX (Figure 8B). The results suggest that HCQ and ENX together become an effective alternative for these patients, considering the endothelial dysfunction modulation induced by the previous activation of these cells. (3) NR+ IgG decreased NO bioavailability, increased procoagulant microparticles, induced endothelin-1, and reduced VEGF production (Figure 8C). In these NR+ patients, heparin and ASA could modulate the clinical manifestations using the mechanisms that were not evaluated in this study. (4) IgG from RII+ decreased NO bioavailability, generated the production of microparticle and endothelin-1, and induced oxidative stress (mitochondrial O₂^– production) (Figure 8D). In these groups of patients, the antioxidant therapy, which was proposed by a few authors, was used for APS treatment.

Both the aPL that require or do not require β2GPI to increase their pathological effect on the endothelium could act synergistically in each group of patient to induce clinical manifestations. aPL are associated with different clinical manifestations of APS, but the mechanism that explains these associations or the presence of these autoantibodies or epitopes in distinct groups of patients with APS is unknown, especially in the refractory cases. aPL generate endothelial activation leading to cell dysfunction and favor clinical manifestations of thrombosis and gestational morbidity. These aPL recognize antigens, such as cardiolipin and β2GPI, and epitopes of these antigens that determine the pathological effect or clinical manifestation. An example of this is the antibodies anti-domain-I of β2GPI (aD1-β2GPI) in the glycine40-arginine 43 associated with thrombosis (de Laat et al., 2005). However, the presence of these aD1-β2GPI is a predictor of thrombosis and pregnancy morbidity; contrarily, the antibodies anti-domain-4/5 of β2GPI is not associated with these clinical manifestations (Chighizola et al., 2018). Additionally, patients with SLE, LA, and aβ2GPI IgA antibodies are also associated with developing thrombosis and aCL with preterm delivery (Saleh et al., 2020; Demir et al., 2021). Patients with different clinical manifestations, refractoriness, and other autoimmune diseases, such as SLE, were included to differentiate between the pathological mechanisms of aPL on endothelial activation and dysfunction that explain the aPL generation of different pathological effects and triggering thrombosis alone or thrombosis with pregnancy morbidity.

Immunoglobulin-G from the different groups of patients triggered different mechanisms to generate endothelial activation and dysfunction that explain APS clinical manifestations. RI+ IgG with β2GPI induced endothelial activation by expressing E-selectin, VCAM, and consequent monocyte adhesion to the endothelium. Both molecules increased by aPL indicate endothelial activation associated with thrombus formation (Gandhi et al., 2021). In our experimental model, VTI IgG did not induce endothelial dysfunction as evaluated with NO synthesis. RI+ IgG induced endothelial activation and dysfunction. RI+ IgG induced endothelial dysfunction due to the NO bioavailability reduction with and without β2GPI. The expression of adhesion molecules is mediated by the factor NF-kB that is inhibited by NO (Liao, 2013). NO produced by eNOS or NO-donors reduces endothelial activation by inhibiting the expression of adhesion molecules, leukocyte adhesion and traffic, platelet reactivity, and vascular proliferation and angiogenesis modulation (Liao, 2013). The expression of adhesion molecules allowing the binding of monocytes to the endothelium and the decreased NO were associated with thrombosis (Ghimire et al., 2017). Monocytes are the primary sources of tissue factor, a key in the extrinsic coagulation cascade. These cells also bind platelets to each other and participate in thrombus recanalization (Mukhopadhyay et al., 2019). On the other hand, NO reduction represents a prothrombotic microenvironment considering that NO inhibits platelet and leukocyte adhesion to the endothelium (Costa et al., 2019). These mechanisms of endothelial activation and dysfunction induced by RI+ IgG associated with thrombosis also lead to gestational morbidity (Possomato-Vieira and Khalil, 2016). Therefore, decreased NO was associated with increased vasoconstriction and hypertension, leading to an alteration in adequate uterine spiral artery remodeling (Possomato-Vieira and Khalil, 2016). In addition, similar monocyte-mediated activation of endothelial cells was described in preeclampsia, a pregnancy condition characterized by hypertension, endothelial dysfunction, and, in severe cases, thrombocytopenia.

In APS, the development of thrombosis is associated with reduced endothelial NO through the binding of aPL to the apolipoprotein E receptor 2 (ApoER2) by β2GPI, which leads to the protein phosphatase 2A (PP2A) activation that induces eNOS dephosphorylation (Sacharidou et al., 2018). Recently, the same proteins were detected in trophoblast with pathological functions, leading to pregnancy morbidity. The binding of aPL by ApoER2-stimulated PP2A, which reduced proliferation and trophoblastic migration associated with the development of preeclampsia for upregulation in hypoxia-inducible factor 1 and soluble endoglin (Chu et al., 2021). In APS, the development of pregnancy morbidity is related to endothelial activation and dysfunction that induce defective placental formation; however, the effect of aPL on trophoblast cells is vital (Abrahams et al., 2017; Quao et al., 2018). aPL reduce decidual endovascular trophoblast invasion (Sebire et al., 2002), leading to early recurrent miscarriages in APS (Abrahams et al., 2017; Jovanovic Krivokuca et al., 2017; Quao et al., 2018). Contrarily, aPL also induce pregnancy morbidity in the second and third trimester by inducing placental dysfunction that is evidenced by preeclampsia and/or intrauterine growth restriction (Antovic et al., 2018; Chu et al., 2021).

Hydroxychloroquine (HCQ) and ENX modulate the endothelial dysfunction induced by RI+ IgG and β2GPI. ENX only modulated endothelial dysfunction generated by RI + IgG without β2GPI. Miranda et al. (2019) showed that HCQ restores eNOS phosphorylation decreased by aPL and thrombus formation in vivo. These authors found that HCQ reverses the increased tissue factor and decreased thrombomodulin induced by aPL in human aortic endothelial cells (Miranda et al., 2019). However, the previous study did not include different groups of patients, and the effects of HCQ on NO bioavailability through the blockade of NO-donors are unknown. Our study found the existence of the modulating effect of HCQ in patients with refractory primary APS with pregnancy morbidity and thrombosis (RI+). This modulating effect of HCQ is also explained by the disintegration that this drug induces on the phospholipid/β2GPI/aβ2GPI complexes (Rand et al., 2008). HCQ disintegrates the β2GPI/aPL complexes, preventing downstream pathological effects; it also modulates proinflammatory cytokine production, such as tumor necrosis factor-alpha, and mitigates the increase of HUVEC permeability induced by the serum from patients with preeclampsia (Rahman et al., 2020). Likewise, heparin binds to β2GPI in domain V through an interaction with Lys284, Lys286, and Lys287, and decreases the ability to recognize aPL aβ2GPI by reducing their prothrombotic activity (Guerin et al., 2002). Similar to HCQ, heparin increases the phosphorylation of eNOS, thus increasing NO (Li et al., 2020).

Non-refractory primary obstetric APS IgG with β2GPI decreased NO bioavailability and VEGF production but increased endothelin-1 generation, which explains one of the triggering mechanisms of thrombosis and pregnancy morbidity and hypertension. Our results agree with previous findings in which pregnancy morbidity was associated with a decreased NO that induces endothelin-1 production (Saleh et al., 2016). Nevertheless, a VEGF reduction was related with enhancing this vasoconstrictor as a switch in the angiogenic mechanisms, associated with placental malformation, pregnancy morbidity, and other pathologies, such as cancer (Lankhorst et al., 2016). Despite this, IgG from patients with VTI and VTII increased the VEGF production as a possible thrombosis-inducing mechanism associated with intimal hyperplasia at the thrombus site (Williams et al., 2000).

On the other hand, RII+ IgG with β2GPI decreased NO bioavailability, but without this cofactor increased endothelin-1 and induced O₂^– production compared with the NR+ primary APS (NR+). Independent results of β2GPI indicate that this effect is generated by aCL without requiring this cofactor or other autoantibodies present in RII+ patients. Human monoclonal aCL injected into BALB/c mice was reported to induce O₂^– production. However, these antibodies showed reactivity to β2GPI (Delgado Alves et al., 2005). Hence, the role of aCL in this event was unclear, but Simoncini et al. (2005) showed that after eliminating aβ2GPI, aPL from patients with APS-induced ROS production in endothelial cells compared with the NHS IgG control, suggesting the relevance of β2GPI-independent aCL. A few studies reported that the serum of patients with SLE, positive for anti-double-stranded DNA antibodies, induced intracellular ROS production and increased NADPH oxidase activity, which catalyzes O₂^– production (Didion and Faraci, 2002; Toral et al., 2017). However, patients were not classified by clinical manifestation in these previous studies. According to the results in this study, β2GPI-independent autoantibodies from patients with RII+ trigger thrombosis and pregnancy morbidity through endothelial oxidative stress, indicating a potential antioxidant therapy in patients with refractoriness to conventional treatment. These results indicate that not only β2GPI-independent aCL generate events associated with endothelial activation in these patients. RII+ IgG triggers endothelial activation and dysfunction, characterized by ROS production: O₂^– production reduces NO bioavailability, impairing vascular relaxation, which increases in endothelin-1 and enhances vasoconstriction (Silva et al., 2012).

Endothelial dysfunction is characterized by the outburst of microparticles (Amabile et al., 2005; Deng et al., 2017; Leite et al., 2020). IgG from patients with APS, presenting thrombosis with or without pregnancy morbidity, increased endothelial microparticles in vitro (Pericleous et al., 2013). Endothelial microparticles reportedly increased only in the plasma of patients with thrombotic APS compared with healthy controls and patients with obstetric APS (Breen et al., 2015). VTI IgG with β2GPI produced 0.5-μm microparticles positive for phosphatidylserine, considered as procoagulants. These microparticles induce coagulation by exposing phosphatidylserine, which activates tissue factor, a receptor for factor VIIa, and initiates the coagulation cascade through an extrinsic pathway (Owens and Mackman, 2011). Additionally, NR+ IgG increased the number of 0.5-μm procoagulant microparticles (CD31+/annexinV+ and CD31-/annexinV+) and a 1-μm CD31+ microparticle. RII+ IgG generates total microparticles of 1 μm. Microparticles of endothelial cells were found in patients with recurrent miscarriage, indicating damage and endothelial activation (Carp et al., 2004).