SARS‐CoV‐2 infection results in upregulation of Plasminogen Activator Inhibitor‐1 and Neuroserpin in the lungs, and an increase in fibrinolysis inhibitors associated with disease severity

Abstract Severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) infection results in coagulation activation although it is usually not associated with consumption coagulopathy. D‐dimers are also commonly elevated despite systemic hypofibrinolysis. To understand these unusual features of coronavirus disease 2019 (COVID‐19) coagulopathy, 64 adult patients with SARS‐CoV‐2 infection (36 moderate and 28 severe) and 16 controls were studied. We evaluated the repertoire of plasma protease inhibitors (Serpins, Kunitz, Kazal, Cystatin‐like) targeting the fibrinolytic system: Plasminogen Activator Inhibitor‐1 (PAI‐1), Tissue Plasminogen Activator/Plasminogen Activator Inhibitor‐1 complex (t‐PA/PAI‐1), α‐2‐Antiplasmin, Plasmin‐α2‐Antiplasmin Complex, Thrombin‐activatable Fibrinolysis Inhibitor (TAFI)/TAFIa, Protease Nexin‐1 (PN‐1), and Neuroserpin (the main t‐PA inhibitor of the central nervous system). Inhibitors of the common (Antithrombin, Thrombin‐antithrombin complex, Protein Z [PZ]/PZ inhibitor, Heparin Cofactor II, and α2‐Macroglobulin), Protein C ([PC], Protein C inhibitor, and Protein S), contact (Kallistatin, Protease Nexin‐2/Amyloid Beta Precursor Protein, and α‐1‐Antitrypsin), and complement (C1‐Inhibitor) pathways, in addition to Factor XIII, Histidine‐rich glycoprotein (HRG) and Vaspin were also investigated by enzyme‐linked immunosorbent assay. The association of these markers with disease severity was evaluated by logistic regression. Pulmonary expression of PAI‐1 and Neuroserpin in the lungs from eight post‐mortem cases was assessed by immunohistochemistry. Results show that six patients (10%) developed thrombotic events, and mortality was 11%. There was no significant reduction in plasma anticoagulants, in keeping with a compensated state. However, an increase in fibrinolysis inhibitors (PAI‐1, Neuroserpin, PN‐1, PAP, and t‐PA/PAI‐1) was consistently observed, while HRG was reduced. Furthermore, these markers were associated with moderate and/or severe disease. Notably, immunostains demonstrated overexpression of PAI‐1 in epithelial cells, macrophages, and endothelial cells of fatal COVID‐19, while Neuroserpin was found in intraalveolar macrophages only. These results imply that the lungs in SARS‐CoV‐2 infection provide anti‐fibrinolytic activity resulting in a shift toward a local and systemic hypofibrinolytic state predisposing to (immuno)thrombosis, often in a background of compensated disseminated intravascular coagulation.

keeping with a compensated state. However, an increase in fibrinolysis inhibitors (PAI-1, Neuroserpin, PN-1, PAP, and t-PA/PAI-1) was consistently observed, while HRG was reduced. Furthermore, these markers were associated with moderate and/or severe disease. Notably, immunostains demonstrated overexpression of PAI-1 in epithelial cells, macrophages, and endothelial cells of fatal COVID-19, while Neuroserpin was found in intraalveolar macrophages only. These results imply that the lungs in SARS-CoV-2 infection provide anti-fibrinolytic activity resulting in a shift toward a local and systemic hypofibrinolytic state predisposing to (immuno)thrombosis, often in a background of compensated disseminated intravascular coagulation.

Study design and participants
Cross-sectional study of 64 adult (≥18 years) patients with moderate disease (referred to as "moderate disease group", n = 36) and severe or critical disease (referred to as "severe disease group", n = 28), based on World Health Organization guidelines [18], with clinical and/or radiologic indications for hospitalization at the Johns Hopkins Hospital (period April 2020-October 2020), as reported [5].

Autopsy
All autopsies were consented for by legal next of kin and performed by the Autopsy Service at the Johns Hopkins Hospital as described [19].

Statistical analysis
Descriptive statistics as described [5].

Demographics
Sixty-four adult patients (36 with moderate and 28 with severe disease) with PCR-confirmed SARS-CoV-2 were included in our cohort.
Sixteen individuals were selected as controls (

Routine biochemical, hematologic, and coagulation parameters
Several parameters were evaluated, with the mean ± SEM or median, indicated in Table 1 Table 1 shows that fibrinogen and platelets did not differ between moderate and severe disease groups, while PT (p = 0.0004), aPTT (p = 0.012), and D-dimer (p = 0.025) were discriminatory. The DIC score, based on the International Society on Thrombosis and Haemostasis criteria, was similar in severe versus moderate disease groups (p = NS) and averaged less than two in both groups.  Figure 1A). FXIa plays a major role in the downstream activation of FIX and serves as an amplification step by thrombin/polyphosphate.

Protease inhibitors of the contact pathway
FXIa is under the control of PN2/AβPP, a potent and specific inhibitor [22]. No difference in PN2/AβPP levels was observed between control (22.50 ± 0.4 ng/ml) and moderate disease groups, although PN2/AβPP was significantly elevated in severe cases versus controls (p = 0.01) ( Figure 1B). SARS-CoV-2 infection is associated with the NET formation and high levels of neutrophil elastase [5]. α1-antitrypsin (A1AT) is an abundant inhibitor of elastase as well as other enzymes [23]. A1AT showed an increase in moderate (p = 0.001) and severe (p = 0.009) disease versus controls (1.56 ± 0.1 mg/ml) ( Figure 1C). Complement activation is a well-described event in SARS-CoV-2 infection, and C1-INH is a main inhibitor of the classical and lectin complement pathways [24]. Figure 1D shows no statistical (p = NS) difference in C1-INH concentrations in moderate or severe disease versus normal controls (280 ± 50 μg/ml).

Protease inhibitors of the common and Protein C pathway
The contact and extrinsic pathways result in the generation of thrombin, platelet aggregation, clot formation, and inflammation. There are HCII is a thrombin inhibitor whose activity is increased 1000-fold by dermatan sulfate [23]. HCII concentration was higher in moderate and severe cases compared to controls (215 ± 45 μg/ml), but it did not reach statistical significance (p ≤ 0.05)( Figure 2C). PN-1 is an efficient inhibitor of thrombin released by activated platelets and a negative modulator of fibrinolysis [25]. The concentration of PN-1 in controls (3.6 ± 0.2 ng/ml) was similar in moderate and severe disease (p = NS), although a significant increase in severe versus moderate disease was observed (p = 0.038)( Figure 2D). A2M can trap many proteinases involved in coagulation and fibrinolysis including thrombin, FXa, plasmin, and kallikrein, among other enzymes, cytokines, and growth factors. A high concentration of A2M in control plasma (5.43 ± 0.5 mg/ml) remains stable in moderate and severe disease ( Figure 2E). FXa is a pro-inflammatory and pro-coagulant enzyme of the common pathway. FXa inhibition requires the presence of PZ as a cofactor for ZPI. There were no statistically significant differences in PZ between controls (1.78 ± 0.13 μg/ml) and the other two groups (p = NS)( Figure 2F). However, a significant increase in PZI was found when moderate and severe disease groups (p ≤ 0.0001, for both) were compared to control individuals (1.04 ± 0.1 μg/ml) ( Figure 2G). In the PC pathway, activated PC (aPC) functions as an anticoagulant by cleaving Factor Va and Factor VIIIa, in the presence of cofactor Protein S, being under the control of PCI [23]. Figure 2H shows that the plasma level of PC in normal individuals (4.06 ± 0.4 μg/ml) is similar to the other two groups, while there is an increase in PS (total) in these groups compared to controls (0.35 ± 0.1 ng/ml) ( Figure 2I). PCI levels in moderate or severe SARS-CoV-2 infection were similar to normal individuals (3.73 ± 0.7 μg/ml) ( Figure 2J).

Protease inhibitors of the fibrinolytic pathway
Fibrinolysis is stimulated by t-PA, an enzyme that activates the zymogen plasminogen into the enzyme plasmin, which degrades fibrin clots [26]. PAI-1 is the main inhibitor of t-PA in human plasma.
PAI-1 levels were higher in patients with moderate and severe disease (p = 0.01 and p = 0.009, respectively) ( Figure 3A in plasma [27]. A2AP in controls (88.35 ± 14 μg/ml) remained at the same level in moderate and severe disease groups ( Figure 3D).

Post-mortem studies
We studied four patients without COVID-19 who died of conditions not associated with pulmonary disease, and 4 other patients who died from COVID-19 with acute lung injury [19].  the fibrinolytic state in the lung, immunohistochemistry for PAI-1 and Neuroserpin was studied using polyclonal antibodies. Figure 5A shows the H&E of Control case #1 with preserved lung architecture. To illustrate the cellular sources for PAI-1 and Neuroserpin in detail, high-power views of the morphology and immunostain results for COVID-19 case #2 are presented in Figure 6. Figure 6A shows the H&E of the alveolar epithelium and abundant intra-alveolar macrophages. Figure 6B reveals marked staining for PAI-1 in cells morphologically consistent with epithelial cells as reported [29], but also in macrophages [30][31][32], with variable/weak staining in endothelial cells [33]. Figure 6C shows punctate Neuroserpin staining typically concentrated in vesicles located in close proximity to the plasma membrane of macrophages, as reported [34]. In contrast, epithelial cells and endothelial cells were negative for Neuroserpin staining. To compare the specificity of our immunostains for PAI-1 and Neuroserpin, with another polyclonal antibody, Figure 6D shows TF staining in epithelial cells only, but not in endothelial cells or macrophages, as reported [5]. In the absence of a primary antibody, staining was negative ( Figure 6D, inset). To confirm cell derivations, Figure 6E shows and Figure 6F reveals the staining of macrophages with anti-CD68. Anti-cytokeratin highlights alveolar epithelium (arrrows). All images (x200). Bar represents 100 μm.

SARS-CoV
PN-2/AβPP (FXIa inhibitor) in moderate and severe disease, likely secondary to release by activated platelets [38]. Altogether, these results are in keeping with adequate regulation of the common, contact, and complement pathways at several steps by their corresponding inhibitors/regulators. These findings are typical for a non-consumptive coagulopathy known as compensated DIC [39] and indicate that most of our patients are in this category. In these cases, a continuous or intermittent slow rate of initiation of intravascular coagulation occurs and control mechanisms (e.g., anticoagulants) may effectively prevent severe clinical manifestations, such as bleeding and hemorrhage [39],  [29] and in macrophages (arrowheads) [30][31][32], with variable/weak staining in endothelial cells (long arrow) [33] (x500). (C) Neuroserpin expression in macrophages (arrowheads) is typically associated with the plasma membrane as reported [34], but not in epithelial cells (arrows) (x500). (D) As a comparison for staining specificity, anti-tissue factor (TF) polyclonal antibody shows staining for TF in epithelial cells only (arrows), but not in macrophages (arrowheads)(x400), as reported [5]. The inset shows negative staining in the absence of a primary antibody. (E) Cytokeratin staining of epithelial cells with AE1/AE3 (arrows), but negative for macrophages (arrowheads) (x500). (F) CD68 confirms abundant intra-alveolar macrophages (arrowheads) and is negative for epithelial cells (arrows) (x500). Bar represents 50 μm. and severe cases with a significant increase in t-PAI/PAI-1 complex formation, indicative of t-PA neutralization. Additionally, PAI-1 and t-PA/PAI-1 were associated with disease severity based on our univariate analysis, suggesting that PAI-1 contributes to the pathogenesis of SARS-CoV-2 infection [40]. Of note, PAI-1 is produced by adipose tissue and is elevated in obese patients [30,41], a morbidity associated with poor outcomes in the infection [1][2][3][4]. Notably, our autopsy cases revealed overexpression of PAI-1 in epithelial cells [29] and macrophages [30][31][32], with variable/dim staining in endothelial cells [33], indicating a high anti-fibrinolytic activity in the lungs. This interpretation is supported by high PAI-1 antigen levels detected in bronchoalveolar lavage samples collected from critically ill patients with SARS-CoV-2 infection and suppressed fibrinolysis by gene expression analysis [15,36]. Of note, PAI-1 is also present in large amounts in platelets, which are often detected as platelet-rich thrombi in the pulmonary vessels of COVID-19 infection [5], and in neutrophils [42]; both cells play a proinflammatory/procoagulant role in the disease [1][2][3][4]43]. Furthermore, PAI-1 is the principal fibrinolytic inhibitor in the pathogenesis of acute respiratory distress syndrome, representing an independent risk factor for poor prognosis and mortality in acute lung injury (ALI) [44][45][46], a condition that may develop COVID-19 infection [1][2][3][4]. Taken together, anti-fibrinolytic activity through PAI-1 expression takes place in the lungs through different cell types and conceivably contributes to circulating PAI-1 levels leading to impaired systemic fibrinolysis.
Neuroserpin showed an increase in both moderate and severe cases of SARS-CoV-2 infection. Neuroserpin inhibits tPA, and to a lesser extent uPA and plasmin. Because Neuroserpin preferentially inhibits tPA, in contrast with PAI-1 inhibits tPA and uPA, and preferentially localizes to neurons, it has been proposed that this serpin is the selective inhibitor of tPA in neurons. The mechanism(s) of Neuroserpin increase in the plasma of our patients is not clear, since the CNS typically produces Neuroserpin [47]. However, macrophages are potentially an important source, since these are among the myeloid cells known to produce this inhibitor [34]. Accordingly, our immunostain showed a typical pattern of expression of Neuroserpin [34] in intraalveolar macrophages where it accumulates in the lungs. Neuroserpin function is complex, with both vascular and cellular effects, and it is considered an endogenous neuroprotectant in the course of cerebral ischemia through plasmin-dependent and independent mechanisms [47][48][49]. Interestingly, our results show that plasma Neuroserpin positively correlates with PAI-1 and t-PA/PAI-1, and is associated with severity, suggesting participation in disease pathogenesis. However, a definitive role for Neuroserpin in venous thrombosis, stroke, and nonischemic neurologic abnormalities of SARS-CoV-2 infection remains to be determined [49,50]. Finally, the results showed a modest but statistically significant increase in PN-1 in severe cases only, which may negatively modulate fibrinolysis by inhibiting t-PA [25].
Plasmin is the main enzyme that degrades the fibrin clot resulting in D-dimer formation and is under inhibitory control of A2AP [27].
High levels of A2AP are associated with ischemic stroke in humans, and ischemic events in mice. The levels of A2AP remained constant in moderate and severe SARS-CoV-2, with an increase in PAP complex formation consistent with an activated fibrinolytic system, an interpretation also supported by an increase in t-PA and D-dimers in our cohort [5]. Elevated PAP also indicates plasmin inhibition and potential failure of fibrin dissolution, contributing to hypofibrinolysis.
These findings are potentially important because clots are associated with trapped A2AP leading to fibrinolysis resistance, and circulating microclots are associated with long COVID-19 [51]. Fibrinolysis is also modulated by TAFI, a carboxypeptidase that cleaves terminal lysine in fibrin, preventing t-PA-mediated plasminogen activation. We did not detect consumption of TAFI/TAFIa, in contrast to variable or increased levels reported in critically ill/ICU patients [14,17,20,52], suggesting that TAFIa also functions as a negative modulator of fibrinolysis in advanced conditions.
Our results showed consumption of HRG, a cystatin-like inhibitor involved in the regulation of coagulation/fibrinolysis by promoting the assembly of t-PA with plasminogen leading to plasmin generation [28]. Low levels of HRG were associated with moderate and severe disease, suggesting a role in pathogenesis. HRG also negatively correlated with inhibitors of fibrinolysis (e.g. PAI-1). In mice, HRG prevents septic lethality through negative regulation of immunothrombosis [28].
It is conceivable that reduced HRG contributes to hypofibrinolysis, and/or promotes inflammation in SARS-CoV-2 infection. Other mechanisms may explain fibrinolysis shutdown in the disease. SARS-CoV-2 spike protein S1 binds to fibrinogen and induces structurally abnormal blood clots with heightened proinflammatory activity and fibrin(ogen) resistant to fibrinolysis [53]. This lytic impairment may result in persistent large microclots [54]. It is plausible that local expression and systemic hypofibrinolysis by numerous inhibitors described above, in addition to abnormalities in the fibrin(ogen) molecule, contribute to a hypofibrinolytic state typical of COVID-19 patients. This interpretation is supported by numerous TEG results showing impaired fibrinolysis in SARS-CoV-2 and its association with thrombosis [11,13,14,40].
Altogether, a picture emerges where two compartments contribute to hemostasis dysregulation in the disease. The lung ("first compartment") is the epicenter of immunothrombosis by numerous mechanisms, resulting in a heightened coagulation-inflammation cycle. Accordingly, intermittent/localized coagulation factor activation induced by pulmonary TF expression, and NETs formation (among others) overcomes inhibition by local anticoagulants whose expression and functions are downregulated or impaired due to endothelial cell damage (e.g., thrombomodulin). This results in abundant pulmonary fibrin deposition and platelet thrombi [1][2][3][4][5][6][7]. Increased D-dimers, Acquired or inherited thrombophilia may also contribute to thrombotic events in some cases [55][56][57]. In patients who are critically ill, marked activation of the coagulation cascade promotes a decompensated response resembling overt DIC seen in sepsis [8,9,55].
One important limitation of this study is that results and conclusions may not be generalizable to all critically ill/ICU patients in whom a decompensated state (overt DIC) may develop. Another limitation is that most patients were on prophylactic heparin, which may have prevented coagulation dysregulation to a certain extent. Although our results by ELISA are in excellent agreement with proteome analysis of SARS-CoV-2 plasma/serum for several inhibitors (e.g., Kallistatin, PZI, A2M, HCII, PS, A1AT, and HRG) [35,58,59], our analysis was limited to the concentration of inhibitors in the plasma; therefore, type II deficiencies (antigen levels are normal, but the activity levels are low) cannot be excluded. Despite these limitations, the marked increase in anti-fibrinolytic molecules in the lung and in the circulation suggests that going forward, targeting fibrinolysis may be a useful therapeutic strategy when combined with other modalities [60].

ACKNOWLEDGMENTS
We thank all physicians and medical staff involved in patient care, and all the technicians of the Core Lab and Department of Pathology for their support during the coronavirus disease 2019 (COVID- 19) pandemic. We are grateful to Ms. Samantha Olszewski for support.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.

AVAILABILITY STATEMENT
The data that support the findings of this study are available upon request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

ETHICS STATEMENT
All procedures were in accordance with the ethical standards of the respective local research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.