Tie2 activation protects against prothrombotic endothelial dysfunction in COVID-19

Profound endothelial dysfunction accompanies the microvascular thrombosis commonly observed in severe COVID-19. In the quiescent state, the endothelial surface is anticoagulant, a property maintained at least in part via constitutive signaling through the Tie2 receptor. During inflammation, the Tie2 antagonist angiopoietin-2 (Angpt-2) is released from activated endothelial cells and inhibits Tie2, promoting a prothrombotic phenotypic shift. We sought to assess whether severe COVID-19 is associated with procoagulant dysfunction of the endothelium and alterations in the Tie2-angiopoietin axis. Primary human endothelial cells treated with plasma from patients with severe COVID-19 upregulated the expression of thromboinflammatory genes, inhibited expression of antithrombotic genes, and promoted coagulation on the endothelial surface. Pharmacologic activation of Tie2 with the small molecule AKB-9778 reversed the prothrombotic state induced by COVID-19 plasma in primary endothelial cells. On lung autopsy specimens from COVID-19 patients, we found a prothrombotic endothelial signature as evidenced by increased von Willebrand Factor and loss of anticoagulant proteins. Assessment of circulating endothelial markers in a cohort of 98 patients with mild, moderate, or severe COVID-19 revealed profound endothelial dysfunction indicative of a prothrombotic state. Angpt-2 concentrations rose with increasing disease severity and highest levels were associated with worse survival. These data highlight the disruption of Tie2-angiopoietin signaling and procoagulant changes in endothelial cells in severe COVID-19. Moreover, our findings provide novel rationale for current trials of Tie2 activating therapy with AKB-9778 in severe COVID-19 disease.


Introduction
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus responsible for coronavirus disease (COVID- 19), can result in critical illness characterized by severe pulmonary manifestations 1 , in addition to several extrapulmonary manifestations 2 , and carries a significant morality rate. Critical COVID-19 illness is characterized by a prothrombotic coagulopathy, and higher D-dimer concentrations and activation of coagulation are associated with worse outcomes [3][4][5] . Fibrin deposition in the lung vasculature is a commonly identified histopathologic finding, and microvascular thrombosis may be a key driver of COVID-19 pathophysiology 6,7 .
Multiple lines of evidence, including measurement of circulating proteins and metabolites and analysis of histologic specimens, have demonstrated severe vascular inflammation and endothelial injury 8 . Therefore, a phenotypic switch of endothelial cells to a procoagulant state appears to be a critical disease mechanism. Endothelial dysfunction in COVID-19 may be mediated through circulating inflammatory cytokines 9,10 , auto-antibodies 11,12 , neutrophil extracellular traps (NETs) 13 , and potentially via direct viral infection 14 .
Procoagulant changes in endothelial cells can be characterized by loss of constitutive anticoagulant function and/or upregulation of thromboinflammatory mediators. Several studies have demonstrated that severe COVID-19 is associated with increased circulating levels of such markers, including procoagulant Von Willebrand factor (VWF) 15 and plasminogen activator inhibitor (PAI-1) 16 , vascular cell adhesion markers (VCAM, ICAM, E-selectin) [17][18][19] , and increased circulating antithrombotic endothelial surface proteins tissue factor pathway inhibitor (TFPI) 20 and thrombomodulin 21 , which are cleaved off the endothelial surface during inflammation 22 .
We have previously demonstrated via unbiased proteomics that endothelial dysfunction is strongly implicated in the coagulopathy of critical illness 23 . Specifically, dysregulation of the Tie2-angiopoietin pathway emerged as a central link between vascular inflammation and inappropriate coagulation 23 . The receptor tyrosine kinase Tie2 is highly enriched in vascular endothelium, and its ligand angiopoietin 1 (Angpt-1) promotes vascular stability and quiescence through Tie2 activation 24 . Moreover, activation of the Tie2 pathway can prevent the heightened thrombosis that occurs during septic endothelial injury 23 . Angiopoietin 2 (Angpt-2) is an Angpt-1 paralog that competitively inhibits Tie2. Angpt-2 expression is promoted by tissue hypoxia and other inflammatory pathways and is subsequently stored in endothelial Weibel-Palade bodies and released during endothelial activation 25 . Angpt-2 levels are even further elevated in the plasma of septic patients with disseminated intravascular coagulation (DIC) compared to those with sepsis but without DIC, and high Angpt-2 levels potentiate endothelial dysfunction of critical illness 23,24 . Loss of constitutive Tie2 signaling is mediated through Angpt-2 antagonism or by Tie2 cleavage from the endothelial surface 26,27 . Inhibition of Tie2 results in loss of barrier function and anti-inflammatory transcriptional machinery, most notably characterized by upregulation of Angpt-2 itself, leading to a positive feedback loop and further Tie2 suppression.
Since Angpt-2 contributes to the coagulopathy of sepsis 23 , we hypothesized that Angpt-2 may also regulate the host endothelial response in COVID-19 and, more importantly, represent a therapeutic target. We used a cell culture model to perturb theTie2-angiopoietin system and determine if increasing COVID-19 disease severity leads to procoagulant changes in the endothelium. We found that plasma from COVID-19 patients can directly promote a prothrombotic state in endothelial cells and that pharmacologic activation of Tie2 signaling rescues endothelial antithrombotic function. We established the presence of prothrombotic endothelial markers in lung tissue biopsies from COVID-19 patients. We next validated the relationship between COVID-19 severity and endothelial dysfunction by measuring Angpt-2 and other markers of thromboinflammatory endothelial activation in plasma from patients with mild, moderate and severe COVID-19. Markers of endothelial dysfunction and thrombosis are strongly correlated with COVID-19 disease severity and survival. Our findings indicate a role for Angpt-2-Tie2 system in endothelial dysfunction in COVID-19 pathogenesis and offer potential targets for therapeutic intervention.

COVID-19 Cohort
Plasma was collected from 98 patients with PCR-confirmed SARS-CoV-2 infection: 42 were admitted to an intensive care unit (ICU), 37 were hospitalized in a non-ICU medical floor, and 19 were outpatients. In addition, we studied 12 historical healthy controls. Patient characteristics are shown in Table 1 Plasma from patients with severe or moderate COVID-19 induced significant upregulation of tissue factor, E-selectin, and Angpt-2 gene expression and concurrently decreased expression of antithrombotic genes EPCR, TFPI, and thrombomodulin (Figure 1). Plasma from patients with severe COVID-19 did not significantly alter Tie2 or VE-PTP gene expression (Figure 1) Given the significant induction of Angpt-2 and its association with the coagulopathy of critical illness via inhibition of Tie2 23 , we tested whether pharmacologic activation of Tie2 could reverse thromboinflammatory gene expression triggered by COVID-19 plasma. Indeed, stimulation of Tie2 by recombinant Angpt-1 or the small molecule VE-PTP inhibitor AKB-9778 normalized much of the gene expression changes induced by plasma from COVID-19 patients (Figure 1).
We further evaluated the ability of humoral factors present in plasma from patients with COVID to promote procoagulant changes in the endothelium by measuring endothelial factor Xa and thrombin generation following exposure to COVID-19 plasma. These are the final steps immediately preceding fibrin formation and are promoted by inflamed, but not quiescent, endothelial cells. Plasma from both severe and moderate COVID-19 patients stimulated tenase and prothrombinase activity on the endothelial surface (Figure 2). Treatment with Tie2 activator Angpt-1 or AKB-9778 significantly attenuated the ability of plasma from severe COVID-19 to promote tenase and prothrombinase on endothelial cells. Furthermore, plasma from severe COVID-19 patients induced externalization of endothelial phosphatidylserine, an anionic phospholipid critical for assembly of coagulation enzyme complexes (Figure 2). Treatment with Angpt-1 or AKB-9778 also reduced PS externalization in response to severe COVID-19 plasma.
These results suggest that plasma from moderate and severe COVID-19 patients promotes a prothrombotic state in endothelial cells, and that activation of Tie2 can inhibit this response.

Increased protein markers of thrombosis in COVID-19 lung biopsies
We next sought to examine changes in expression in COVID-19 lung specimens of endothelial proteins associated with thrombotic risk. Core needle biopsies were obtained during limited autopsies performed shortly after expiration in five COVID-19 patients (see Supplemental Appendix). Histopathologic evaluation revealed lesions similar to those described in published autopsy series-including fibrin plugs within alveolar spaces and associated microvascular thrombi-suggesting that core needle sampling was representative of typical findings of COVID-19 32 . We compared these specimens to contemporaneously obtained normal lung tissue regions from tumor resections in non-COVID patients. Despite limited COVID-19 autopsy specimen availability, we observed increased levels of the prothrombotic endothelial protein von Willebrand factor (VWF) and decreased levels of the antithrombotic endothelial proteins EPCR and thrombomodulin (Figure 3).

disease severity and survival
We measured plasma concentration of proteins associated with thrombosis and procoagulant endothelial activation and correlated these values with the degree of COVID-19 severity ( Figure   4). Our thromboinflammtory panel included markers of general thrombotic activation, namely Ddimer and tissue factor; markers of procoagulant endothelial activation, VWF and P-selectin; and endothelial inflammation markers, E-selectin and soluble VEGFR-1. We also investigated levels of the anti-thrombotic endothelial proteins, endothelial protein C receptor (EPCR), TFPI, and thrombomodulin, which are cleaved from inflamed endothelium. Finally, given its role linking vascular inflammation and thrombosis, we investigated components of the Tie2-angiopoietin pathway, specifically Tie2-activating cytoprotective Angpt-1, Tie2-inhibiting proinflammatory Angpt-2, and soluble Tie2, which is also cleaved from endothelial cells during inflammatory states 26 .
There was significant correlation among markers of prothrombotic endothelial activation. Angpt-2 levels were highly correlated with D-dimer, E-selectin, thrombomodulin, P-selectin, and VWF (Supplemental Figure 1, top). D-dimer correlated with P-selectin, tissue factor, thrombomodulin, VWF, and TFPI. Levels of many thromboinflammatory proteins were elevated in patients who developed acute kidney injury or died during their index hospital admission (Supplemental Figure 2). The relatively small number of thrombotic events, with the majority being catheter-related thrombosis, did not make it possible for us to draw meaningful conclusions regarding the association of endothelial dysfunction and clotting events in this cohort.
We performed survival analysis for selected analytes, including ones that were significantly higher among patients who died during index hospitalization. Given there were 12 deaths among 42 ICU patients, we stratified ICU patients by tertiles and analyzed survival in the top tertile of analyte concentration versus the bottom tertile. Patients in the highest tertile of Angpt-2 (Log-rank P = 0.048), E-selectin (Log-rank P = 0.019) and P-selectin (Log-rank P = 0.044) had worse survival compared to patients in the bottom two tertiles ( Figure 5). These results suggest that Angpt-2 and other markers of endothelial dysfunction and thrombosis are strongly correlated with COVID-19 disease severity and implicate perturbation of the Tie2-angiopoietin pathway in this process.

Discussion
To interrogate the Ang2-Tie2 system in COVID-19-mediate endothelial dysfunction, we used an in vitro model in which primary endothelial cells were treated with COVID-19 plasma.
The effects of COVID-19 patient plasma on endothelial cells are notable in two related but distinct ways. First, there is a coordinated response such that pro-inflammatory and prothrombotic genes are upregulated whereas anti-inflammatory or anti-thrombotic genes are downregulated. Second, the effects of COVID humoral milieu extend to alterations of the endothelial cell surface that catalyze clot formation. Severe COVID-19 is associated with elevated levels of inflammatory cytokines, most notably TNFa, IL-6, and IL-8 1,9,10,33 , that scale with the degree of disease severity. These cytokines are known to promote transcription of prothrombotic genes such as tissue factor and selectins and to decrease platelet inhibitory nitric oxide 34,35 . Similar or even more profound cytokine elevations are associated with other critical illnesses including severe septic shock and cytokine release syndrome 9,36 , conditions also associated with profound endothelial injury and microvascular thrombosis. Loss of the constitutive endothelial anticoagulants, such as TFPI, EPCR, and thrombomodulin, is an important component of the prothrombotic endothelial transformation and may be an underappreciated driver of COVID-19 coagulopathy.
A primary function of the endothelium is to provide barrier defense to prevent excessive vessel permeability and an antithrombotic surface to promote blood circulation. Tie2 has a critical role in this process and remains activated throughout healthy adult vasculature via continuous secretion of Angpt1 from perivascular cells and platelets. As a receptor tyrosine kinase nearly exclusive to endothelial cells, Tie2 signaling promotes vascular quiescence by activating Kruppel-like factor 2 (KLF2) 37 and inhibiting inflammatory nuclear factor-kB (NF-kB) 38 , thus promoting expression of anticoagulant genes 39 and repressing tissue factor expression 23,40 . In our study, two unrelated approaches to Tie2 activation, direct Tie2 agonism with Angpt-1 and VE-PTP antagonism with AKB-9778, inhibit procoagulant changes in endothelial cells induced by plasma from COVID-19 patients. This result is consistent with the ability of Tie2 activation to suppress proinflammatory endothelial phenotypes in non-COVID infectious diseases such as gram-negative sepsis, anthrax, and malaria 23,41,42 . Our findings support previous work suggesting that preservation of Tie2 signaling is both necessary to prevent hypercoagulation and sufficient to normalize pathological thrombosis during systemic inflammation 23 .
VE-PTP interacts closely with Tie2 and is a natural brake on Tie2 activity. VE-PTP is highly expressed in the lung, the organ that also bears the highest concentration of Tie2. Like the other endogenous Tie2 antagonist, Angpt-2, VE-PTP is induced by pulmonary vascular stressors such as hypoxia 28 . AKB-9778 (Razuprotafib), has demonstrated beneficial activity in multiple animal models of vascular leak including LPS-induced acute lung injury 30 . Applied to cultured endothelial cells, AKB-9778 achieves ligand-independent Tie2 activation and activates Tie2 even when Angpt-1 is unable to do so during the high VE-PTP state of endothelial hypoxia.
In the present study, we demonstrate that AKB-9778 strongly suppresses the procoagulant 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 May 17, 2021. response in endothelial cells induced by COVID-19 plasma. AKB-9778 was more efficacious than Angpt-1, which may reflect the ability of VE-PTP inhibition to promote the weak agonist properties of Angpt-2 43 . Tie2 activation by AKB-9778 may therefore be an extremely effective means of dampening the thromboinflammtory state of the endothelium in COVID-19. In aggregate, our findings support clinical trials of AKB-9778 to improve pulmonary outcomes and mortality in moderate to severe COVID-19 (RESCUE, https://clinicaltrials.gov/ct2/show/NCT04511650). In addition to promoting endothelial barrier function to reduce the incidence or severity of acute respiratory disease syndrome, by lowering the procoagulant potential of the endothelium, AKB-9778 may also inhibit the microvascular thrombosis that is a hallmark of COVID-19 pathophysiology. The assays employed in this study also suggest a method to evaluate other endothelial-cell targeted therapies and assess individual-level prothrombotic risk.
While several studies have measured circulating proteins in COVID-19 patients, few to date have examined endothelial markers in lung histopathological specimens of SARS-CoV-2 infection. We attempted to examine expression of prothrombotic and antithrombotic endothelial proteins in situ, in lung tissue from COVID-19 autopsies. Our results demonstrate upregulation of VWF and loss of thrombomodulin and EPCR in lung tissue from COVID-19. It is therefore likely microvascular thromboses are driven at least in part by prothrombotic endothelial cell changes at the local tissue level. The profound downregulation of EPCR and thrombomodulin suggest that loss of constitutive antithrombotic function may promote clotting in COVID-19.
Increased circulating thrombomodulin levels in severe COVID-19 likely reflect its cleavage from the cell surface 22 . Loss of EPCR in lung specimens was unexpected because circulating EPCR was not significantly different between groups in our cohort, even comparing critical COVID-19 disease to healthy controls ( Figure 5). Nevertheless, the dramatic lack of EPCR antigen in COVID-19 lung specimens suggests mechanisms other than cleavage from the cell surface, such as downregulation of gene expression, may be responsible for loss of anticoagulant endothelial proteins. Indeed, plasma from patients with severe COVID-19 inhibited expression of EPCR by nearly 50% (Figure 1).
Our plasma vascular survey demonstrates that increasing severity of COVID-19 is associated with coagulopathy and shows a clear relationship between increasing levels of endothelial cell dysfunction and increasing clinical strata of disease severity 15,20,21 . Our data provide further support that extensive endothelial inflammation occurs in severe COVID. This study builds on 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 May 17, 2021. ; prior work by demonstrating that the degree of endothelial dysfunction scales with COVID-19 severity and with procoagulant biomarkers. Specifically, markers supporting a procoagulant endothelial phenotype including Angpt-2, P-selectin, thrombomodulin, and TFPI were all elevated in severe disease compared to moderate COVID-19 in our cohort. Our data are in agreement with several lines of evidence suggesting endothelial dysfunction promotes the coagulopathy of critical illness 23 and is a central driver of the pathobiology of severe COVID-19 8 .
Previous studies suggest that circulating Angpt-2, thrombomodulin, and VWF levels predict adverse outcomes in COVID-19 15,21,44 . We expand this list to include E-selectin and P-selectin as well. Correlation of Angpt-2, P-selectin, and VWF levels (Fig. S1) in our work and that of others suggest that Weibel-Palade body extrusion may be a common manifestation of COVID-19 endotheliopathy. Altogether, these findings suggest that a coordinated phenotypic shift of the endothelium contributes to adverse outcomes. Our current data add to mounting evidence that Angpt-2 is a highly sensitive and specific indicator of endothelial damage in acute illness, with COVID-19 being no exception to this trend. Extensive research has documented that Angpt-2 is a powerful biomarker for outcomes of acute respiratory distress syndrome and sepsis 27,45 , and our current findings extend prior findings implicating Angpt-2 in the coagulopathy of critical illness 23 . Indeed, data suggest that Angpt-2 levels are increased in severe COVID-19 and may predict adverse outcomes such as ICU admission 18 , acute kidney injury 46 , and survival 44 .
Our study has several limitations. Given the retrospective and non-consecutive nature of patient recruitment into the biorepositories, several biases may have inadvertently been introduced. It would have been useful to have readouts of gene expression and coagulation assays of individual plasmas correlated with markers detected in vivo or whether these functional assays provided additional predictive clinical information. Given the limited amount of plasma supply however, it was necessary to pool the individual plasmas. The amount of autopsy specimens were limited, which impaired our ability to perform comprehensive histopathological assessment of multiple endothelial markers. We cannot rule out that differences in specimen preparation affected antibody staining. That samples were not collected uniformly early upon hospital or ICU admission limits the applicability of our conclusions regarding endothelial biomarkers and clinical endpoints.
In conclusion, the present study supports the concept that moderate and severe COVID-19 are driven at least in part by procoagulant endothelial cell dysfunction, the degree of which 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.  Thrombotic events were reviewed and included deep vein thrombosis, pulmonary embolism, catheter-related thrombosis, or clinically documented thrombosis at another location.

Plasma biomarker measurements
Plasma was thawed on ice, then centrifuged for 5 min to pellet any debris. Concentrations of the following proteins were measured using a Luminex Human Premixed Multi-Analyte Kit (R&D Systems) for the following analytes: Angpt-1, Angpt-2, D-dimer, E-selectin, P-selectin, thrombomodulin, Tie2, Tissue factor, VWF, and VEGFR1, according to manufacturer's protocol.
Samples were diluted 1:2 and run on a MAGPIX system (Millipore Sigma) which was preprogrammed according to kit specifications. For EPCR, plasma samples were analyzed by ELISA (Diagnostica Stago) at 1:51 dilution according to manufacturer's protocol. For TFPI, 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.

Human lung tissue isolation and immunofluorescence microscopy
Core needle biopsies were obtained during limited autopsy of COVID-19 patients within 3 h of patient expiration. For controls, surgical specimens from patients undergoing lung tumor resections were analyzed by a pathologist, and lung tissue within the tumor-free margins was isolated. These controls were selected due to the similar nature in which the lung tissue was processed compared to standard autopsy where there is a longer delay before tissue is placed in fixative. Tissue was washed and fixed in freshly prepared paraformaldehyde 4% for 24 h and

Gene expression analysis.
Confluent HUVECs were treated with pooled patient plasma as described above and gene expression was determined using a 2-step Cell-to-Ct Taqman kit (Thermo Fisher Scientific). The following gene expression probes (Taqman, Thermo Fisher Scientific) were used: 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.

Phosphatidylserine Externalization
HUVECs were grown to confluence in glass chamber slides and incubated with pooled patient plasma as described above. Cells were washed with 10 mM HEPES buffer (pH 7.4) containing 140 mM NaCl, 2.5 mM CaCl2, and 2% FBS (annexin V binding buffer) and stained with annexin V-Alexa Fluor 488 (Thermo Fischer Scientific) at a 1:50 dilution and Zombie Red viability dye (BioLegend) at a 1:1000 dilution for 15 minutes at room temperature in annexin V binding buffer.
Cells were washed and fixed in paraformaldehyde 4% / annexin V binding buffer for 7 minutes. 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.

Image analysis
Fluorescent images were analyzed using Image J software (NIH). For phosphatidylserine externalization, annexin V staining was thresholded and total fluorescent area was normalized to the number of nuclei. For immunofluorescence microscopy of lung sections, fluorescence intensity was quantified per tissue area, after measuring and subtracting background signal from each image.

Statistics
Tests of normality were performed using the Anderson-Darling and D'Agonstino-Pearson method. Statistical significance for binary comparisons of continuous variables were assessed by unpaired two-tailed Student's t test unless the data did not demonstrate normality, in which case differences between groups were analyzed by Mann-Whitney U test. For comparison of continuous variables across multiple groups, all of which did not pass the test of normality, with the exception of Tie-2, the Kruskal-Wallis test with Dunn's post-hoc for multiple comparisons was performed. Correlation matrix analysis was performed using two-tailed nonparametric Spearman correlation. Survival analysis for ICU-patients was performed by segregating patients into the top tertile versus bottom two tertiles for each analyte. Kaplan-Meier analysis was performed, and survival was compared for the top tertile versus bottom two tertiles using the Mantel-Cox log-rank test. Comparison of categorical variables were performed using Fisher's exact test. All statistical analysis was performed using GraphPad Prism (version 9.0; GraphPad Software, San Diego, CA). P values of less than 0.05 were considered significant.

Study approval
This study measuring biomarkers and associated clinical data in patients enrolled in the BIDMC COVID-19 Data and Tissue Repository was approved by the BIDMC IRB (protocols 2020P000621). For all COVID-19 autopsy studies, patients were consented for limited autopsies by a pathologist during a witnessed phone call immediately after death of the patient.
Research using autopsy tissue was approved by the BIDMC institutional review board (IRB 2020P000525). A HIPPA waiver was granted to access the medical records of the patients undergoing autopsy (IRB 2020P00412). 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. 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.    tile scan image at 20X magnification ± SD. Significance was determined by 1-way ANOVA using Dunnett's post-test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  A) and loss of the antithrombotic factors thrombomodulin (THBD, B) and the endothelial protein C receptor (EPCR, C). Two images were obtained per tissue section and mean fluorescence intensity was analyzed for each tile-scanned image and normalized to background intensity. For graphs, mean is represented by the bar with each dot as a replicate, error bars indicate SD. Significance was determined by a 2-tailed Mann-Whitney test, **P < 0.01, ****P < 0.0001.  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.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (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 May 17, 2021. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.