Prevalence, specificity, and clinical association of anti-phospholipid antibodies in COVID-19 patients: are the antibodies really guilty?

Background. Critically ill patients with coronavirus disease 2019 (COVID-19) have a profound hypercoagulable state and often develop coagulopathy which leads to organ failure and death. Because of a prolonged activated partial-thromboplastin time (aPTT), a relationship with anti-phospholipid antibodies (aPL) has been proposed, but results are controversial. Functional assays for aPL (i.e., lupus anticoagulant) can be influenced by concomitant anticoagulation and/or high levels of C reactive protein. The presence of anti-cardiolipin (aCL), anti-beta2-glycoprotein I (anti-{beta}2GPI and anti-phosphatidylserine/prothrombin (aPS/PT) antibodies was not investigated systematically. Epitope specificity of anti-{beta}2GPI antibodies was not reported. Aim. To evaluate the prevalence and the clinical association of aPL in a large cohort of COVID-19 patients, and to characterize the epitope specificity of anti-{beta}2GPI antibodies. Methods. ELISA and chemiluminescence assays were used to test 122 sera of patients suffering from severe COVID-19. Of them, 16 displayed major thrombotic events. Results. Anti-{beta}2GPI IgG/IgA/IgM were the most frequent in 15.6/6.6/9.0% of patients, while aCL IgG/IgM were detected in 5.7/6.6% by ELISA. Comparable values were found by chemiluminescence. aPS/PT IgG/IgM were detectable in 2.5 and 9.8% by ELISA. No association between thrombosis and aPL was found. Reactivity against domain 1 and 4-5 of {beta}2GPI was limited to 3/58 (5.2%) tested sera for each domain and did not correlate with aCL/anti-{beta}2GPI nor with thrombosis. Conclusion. aPL show a low prevalence in COVID-19 patients and are not associated with major thrombotic events. aPL in COVID-19 patients are mainly directed against {beta}2GPI but display an epitope specificity different from antibodies in antiphospholipid syndrome.


INTRODUCTION
Critically ill patients with coronavirus disease 2019  have a profound hypercoagulable state and often develop thrombosis in veins, arteries and in the microcirculation [1]. As such, these patients may experience organ failure and eventually die [2]. Recent analysis of coagulation screening parameters were instrumental at identifying several abnormalities in these patients, including prominent elevation of fibrin/fibrinogen degradation products (i.e., D-dimer) and a prolonged activated partial-thromboplastin time (aPTT). While high levels of D-dimer are consistent with sustained activation of the clotting and fibrinolytic cascades, the combination of prolonged aPTT and both arterial and venous thrombosis was, however, surprising, and it is reminiscent of a clinical scenario known as antiphospholipid syndrome (APS) [3]. APS is a systemic autoimmune disorder characterized by venous and arterial thrombosis in the presence of anti-phospholipid antibodies (aPL) that can prolong phospholipid-dependent clotting time assays [3].
Looking at the causes for aPTT prolongation, recent studies have shown that lupus anticoagulant (LA) can be detected in a significant percentage of COVID-19 positive samples [4][5][6]. Since LA is often caused by aPL, these findings support the idea that aPL may play a role in COVID-19 [7] However, it is important to point out that LA is a very sensitive assay and its outcome can be influenced by several factors, most notably heparin administration [8] and a profound inflammatory state characterized by high levels of C reactive protein (CRP) [9][10][11]. Both of them are present in .
Another method to detect aPL that is in principle insensitive to anticoagulation and other confounding agents relies on the detection and quantification of autoantibodies using solid phase assays [3]. Using this method, the presence of aPL was recently reported in a handful of case reports and a small cohort of patients [4][5][6][7]. While encouraging, this data is limited and its interpretation remains controversial, with some investigators proposing an important role of aPL in COVID-19 patients [7] while others suggesting a very poor correlation between aPL and thrombotic All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 19, . . https://doi.org/10.1101 doi: medRxiv preprint 5 events [2]. A larger study, possibly multicenter, would be ideal to clarify these outstanding issues and draw conclusions regarding the prevalence, epitope specificity and clinical correlation between aPL and thrombotic manifestations in COVID-19 patients. To fill this gap in knowledge, we used a combination of ELISA and chemiluminescence assays to quantify the levels and identify the type of aPL present in 122 sera of patients suffering from severe or critical COVID-19. Our data indicates a low prevalence of aPL in COVID-19 patients. Yet, they are not associated with major thrombotic events. Importantly, aPL in COVID-19 patients are mainly reacting against β 2 -glycoprotein I (β 2 GPI) but display an epitope specificity different from antibodies found in APS patients.

Patients
A total of 122 patients were enrolled from two COVID-19 referral centers in Lombardia, which is one of the regions in Northern Italy heavily affected by the pandemic. Specifically, 58 patients were recruited in Milan and 64 patients were recruited in Brescia. All patients tested positive to SARS-CoV-2, as detected by . Mean age was 68.5 (+ SD 16.4) years; 77 were men and 45 women. All 122 patients were classified as severe or critical COVID-19 according to Chen et al. [12]. No diagnosis of previous autoimmune diseases was made; six patients had a thrombotic event (3 arterial and 3 venous) in the past clinical history. Eighty-seven patients suffering from APS were also tested for aCL and anti-β 2 GPI IgG, IgM [13]. The study was approved by the Ethics Committees in Milan (Ethic Committee Istituto Auxologico Italiano 3-04-2020) and Brescia (ASST Spedali Civili NP4187).

Detection of aPL
Anti-cardiolipin (aCL) and anti-β 2 GPI IgG, IgA and IgM were detected by chemiluminescence immunoassay (CIA; Quanta Flash, Inova, San Diego, CA, US) and by a home-made ELISA as described elsewhere [13,14]. Anti-β 2 GPI domain 1 IgG (anti-D1) were detected by CIA (Inova) All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Statistical analysis
Data were analyzed using R version 3.4.0. Descriptive statistics was used to summarize data.
Associations and differences between categorical or continuous variables were tested by using Fisher's exact test and non-parametric Mann-Whitney test, respectively. A p-value <0.05 was considered statistically significant.

RESULTS & DISCUSSION
Eighty-three per cent of COVID-19 patients enrolled in our study displayed elevated D-dimer and elevated levels of C reactive protein (CRP). Prolonged aPTT (>30 sec) was found in 57.6% of COVID-19 patients while PT INR values were above the cut-off in 24.8% of the cases. Most of the patients (120 out of 122) were on anticoagulation with low molecular weight heparin (70% of the patients was on therapeutic and the remaining on prophylactic heparin dosage). Despite anticoagulation, we observed sixteen thrombotic events (13.1%), of which 8 were in veins and 8 in arteries. These statistics are in agreement with previous reports [2,[16][17][18][19][20][21] and document abnormal coagulation activities in our patient population.
In the APS field, testing for LA is not recommended by the official guidelines when patients are on heparin, since the presence of heparin, even if neutralized, may lead to false positive results [8].
Likewise, high levels CRP, such as those found in our cohort of patients, have been shown to prolong aPTT independently from the presence of aPL [9][10][11]. On these bases, the presence of aPL were researched using solid phase assays, and not LA. First, we investigated the presence of aCL and anti-β 2 GPI, two very established classification criteria diagnostic tools in APS [3]. Testing was independently performed at two different sites, Milan and Brescia, using standardized All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted . . https://doi.org/10.1101/2020 methodologies harmonized between the two laboratories [22]. The prevalence of COVID-19 patients positive for IgG/IgA/IgM aCL and anti-β 2 GPI antibodies detected by ELISA and CIA is summarized in Table 1. The raw data of the ELISA assays are shown in Figure 1. We found IgG/IgM aCL in 5.7/6.6% of patients, whereas anti-β 2 GPI IgG/IgA/IgM were found in 15.6/6.6/9.0% of patients. Similar values were obtained for aCL antibodies using CIA assays (Table 1) [23], whereas a slightly lower sensitivity was obtained for anti-β 2 GPI antibodies. Taken together, our data shows a low prevalence of classification criteria aPL in COVID-19 patients. In this regard, our study confirms a recent study obtained with smaller cohorts of patients [4,21].
Importantly, our data also shows that aPL are slightly more reactive towards β 2 GPI-coated plates as compared to CL-coated ones and that, regardless of the nature of aPL, there is no association between aPL positivity and thrombotic events (p = 1).
A striking difference between the autoantibody profile found in COVID-19 patients as compared to the one found in APS concerned the titers of aPL. Medium/low aPL titers were consistently found in the patients with COVID-19. By contrast, medium/high titers are usually found in APS patients ( Figure 1). This difference suggests that aPL found in COVID-19 may be different from aPL found in APS and led us to further investigate the epitope specificity of anti-β 2 GPI antibodies. We focused on autoantibodies directed against the N-terminal domain 1 (anti-D1) or the C-terminal domains 4-5 (anti-D4-5) of the molecule [14] (Figure 2A). This is because anti-D1 antibodies are associated with an increased risk of thrombosis and pregnancy complications in APS patients [13,14,24]. By contrast, anti D4-5 antibodies, are associated neither with vascular nor obstetric APS manifestations [13,25]. Furthermore, anti D4-5 antibodies are also reported at high levels in the so called asymptomatic aPL carriers and are frequently found in non-APS (e.g. patients with leprosy, atopic dermatitis, atherosclerosis and in children born to mothers with systemic autoimmune diseases) [25]. Figure 2B shows that three out of 58 samples reacted with D1, while in Figure 2C, three samples tested positive for D4-5. None of the sera was positive for both domains and all displayed a weak reactivity. Although the number of the investigated sera is relatively small, this finding is All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted . . https://doi.org/10.1101/2020 8 quite different from the results found in APS in which almost all the sera positive for the whole β 2 GPI molecule also reacted with domain D1 at high titer [13,24]. Furthermore, at variance with APS patients, none of the anti-D1 positive patients displayed thrombotic events [24].
Approximately fifty-seven per cent of COVID-19 patients has prolonged aPTT. Yet, only a small proportion of COVID-19 patients carry aCL and anti-β 2 GPI antibodies. This suggests that other factors must be responsible for the prolonged aPTT phenomenon. Based on current literature, aPS/PT can be associated with a prolonged aPTT and with the presence of LA [15]. We found fifteen out of 122 sera positive for aPS/PT (12.3%), mostly of the IgM isotype (12 out 15) and at a low titer (Figure 3). There was no association between prolonged aPTT and the presence of aPS/PT antibodies nor with thrombotic events in our COVID-19 cohort. This indicates that aPS/PT are not responsible for the prolongation of aPTT nor are a predictor of adverse clinical outcomes.

Furthermore, in contrast to what we would have expected in APS [26], we found no associations
between the presence of aPS/PT, aCL and anti-β 2 GPI antibodies. This data is in line with the unusual epitope specificity of anti-β 2 GPI antibodies documented in Figure 2, supporting the hypothesis that aPL found in COVID-19 patients are different from aPL found in APS patients.
Whether COVID-19 aPL are similar to the one found in other infectious diseases such HCV, HBV and HIV [27] remains to be determined.
In conclusion, this study documents a low prevalence of aPL in COVID-19 patients and found no association between aPL positivity and major thrombotic events. This leads to the conclusion that aPL are not the main cause of prolonged aPTT nor can explain the pathogenesis of venous and arterial thrombosis in COVID-19 patients. While testing for aPL in COVID-19 patients cannot be recommended at this time, it is important to keep in mind that COVID-19 patients suffer of an acute form of systemic inflammation with complement activation [28], which may be responsible for endothelial perturbation. In this context, since β 2 GPI can accumulate on the activated endothelium at high density, even low titers of aPL may become pathogenic thus potentiating or even triggering thrombus formation, especially when anticoagulation is suspended. A comparable condition in All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted . . https://doi.org/10.1101/2020 9 which low titers of aPL can cause substantial damage is seen in obstetric APS, where high levels of β 2 GPI can be found in the placenta [29]. Hence, while transitory aPL are likely to be clinically irrelevant in COVID-19 patients as in other infections [27], detection of persistent aPL (12 weeks apart from the first detection) may represent a true risk factor. Accordingly, aPL detection may offer a more robust risk stratification for thrombosis and could eventually justify post-discharge anticoagulant prophylaxis.

CONFLICTS OF INTEREST
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted . . https://doi.org/10.1101/2020 M. Mahler is an employee at Inova Diagnostics, Inc. All the other authors declared no conflict of interest.

8.
Martinuzzo ME, Barrera LH, D 'Adamo MA, Otaso JC, Gimenez MI, Oyhamburu J. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 19, . . https://doi.org/10.1101 doi: medRxiv preprint  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted June 19, . . https://doi.org/10.1101 doi: medRxiv preprint  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

re .4
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.