key: cord-0979379-z108ra48
authors: Hultstrom, M.; Fromell, K.; Larsson, A.; Quaggin, S. E.; Betsholtz, C.; Frithiof, R.; Lipcsey, M.; Jeansson, M.
title: Elevated Angiopoietin-2 inhibits thrombomodulin-mediated anticoagulation in critically ill COVID-19 patients
date: 2021-01-15
journal: nan
DOI: 10.1101/2021.01.13.21249429
sha: 6f7bd601b828602ff4e6d212a2941ee49187e7cf
doc_id: 979379
cord_uid: z108ra48

Several studies suggest that hypercoagulation and endothelial dysfunction play central roles in severe forms of COVID-19. Here, we hypothesized that the high levels of the inflammatory cytokine Angiopoietin-2 (ANGPT2) reported in hospitalized COVID-19 patients might promote hypercoagulation through ANGPT2 binding to thrombomodulin with resulting inhibition of thrombin/thrombomodulin-mediated physiological anticoagulation. We therefore investigated plasma samples taken at two timepoints from 20 critically ill COVID-19 patients in intensive care regarding ANGPT2 levels and coagulation markers in comparison with 20 healthy blood donors. We found that ANGPT2 levels were increased in the COVID-19 patients in correlation with disease severity, hypercoagulation, and mortality. To test causality, we administered ANGPT2 to wildtype mice and found that it shortened bleeding time in a tail injury model. In further support of a role for ANGPT2 in physiological coagulation, bleeding time was increased in endothelial-specific Angpt2 knockout mice. Using in vitro assays, we found that ANGPT2 inhibited thrombomodulin-mediated anticoagulation and protein C activation in human donor plasma. Our data reveal a novel mechanism for ANGPT2 in hypercoagulation and suggest that Angiopoietin-2 inhibition may be tested in the treatment of hypercoagulation in severe COVID-19, as well as in certain other conditions, including sepsis.

SARS-CoV-2 infection may be paucisymptomatic, or lead to coronavirus disease-2019 (COVID- 19) , which has a wide range of symptoms and may cause severe illness, in particular in individuals with other cardiovascular risk factors (1). Thrombotic and thromboembolic disease has emerged as a major COVID-19 complication despite routine thrombosis prophylaxis now being standard of care (2) (3) (4) (5) . Microthrombosis has been suggested to contribute to both respiratory failure and neurological complications (6, 7) , and activation of the coagulation system indicates a poor prognosis among COVID-19 patients in intensive care (1, (7) (8) (9) .

Angiopoietin-2 (ANGPT2) is an inflammatory cytokine, the circulatory level of which correlates with adverse outcomes in several critical care syndromes, including acute respiratory disease syndrome (ARDS) and sepsis (reviewed in (10) ). An elevated plasma ANGPT2 level is a strong predictor of death in infection-mediated ARDS independent of the infectious agent (11) , and elevated plasma ANGPT2 is further associated with disseminated intravascular coagulation (DIC) in conjunction with sepsis (12) . In COVID-19, recent data show that ANGPT2 level is a good predictor of intensive care unit (ICU) admission (13) and correlates with the severity of disease (14, 15) . ANGPT2 exerts its effects through different molecular mechanisms, the most well-studied being inhibition of Tie2 receptor signaling. This causes destabilization of the endothelium in most vascular beds and promotes inflammation, vascular leakage, impairment of the endothelial glycocalyx, and activation of α5β1 integrin signaling (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) . The administration of Tie2 activating agents confers vascular protection and reduced mortality in experimental models of sepsis (12, 27, 28) . Moreover, the recent discovery that ANGPT2 binds thrombomodulin (29) suggest that ANGPT2 may have additional and direct effects on coagulation. Thrombomodulin is constitutively expressed on the luminal surface of endothelial cells, where it is an important member of the intrinsic anticoagulant pathway and also an anti-inflammatory agent (30) . Thrombomodulin inhibits the procoagulant functions of thrombin by binding and inhibiting its interaction with procoagulant substrates and instead promoting thrombin-catalyzed activation of protein C (APC) (31) . Endothelial-specific knockout of thrombomodulin in mice disrupts APC formation and causes lethal thrombus formation (32) , highlighting the potency of this pathway.

We hypothesized that the increased plasma levels of ANGPT2 observed in COVID-19 patients might have a direct effect on the coagulation system by inhibition of thrombomodulin-mediated activation of protein C. To investigate this, we measured plasma ANGPT2 and coagulation parameters in relation to clinical outcome in a cohort of critically ill COVID-19 patients and healthy blood donor controls. We further utilized experimental animals and in vitro assays to investigate if ANGPT2 could inhibit thrombomodulin-mediated anticoagulation and activation of protein C.

The study included a selected group of 20 COVID-19 patients with extended (>10 days) stays at the ICU at Uppsala University Hospital due to SARS-CoV-2 infection. Plasma samples were collected at two timepoints for each patient: 1-4 days after admission (early) and 10-14 days after admission (late). Age and gender matched healthy blood donors were used as controls. In the patient group, 15 patients subsequently recovered (recovering patients (Rec)) and 5 patients died (non-recovering patiens (Dec)). Patient demographic, comorbidities, clinical features at arrival, and outcomes are listed in Table 1 . Assessment of plasma ANGPT2 concentrations revealed that ANGPT2 was elevated (P<0.01) in patients already at the early timepoint after ICU admission compared to controls (Fig. 1A) . At the late timepoint ANGPT2 level was further increased (P<0.05) in the non-recovering patients compared to recovering patients (Fig. 1A) . Sequential organ failure assessment (SOFA) score was calculated for the same days as collection of plasma samples. The SOFA score represents six organ system, were each organ system is assigned a point value from 0 (normal) to 4 (high degree of dysfunction/failure) (33) . As expected, SOFA score was increased in all patients and were significantly (P<0.01) higher at the late timepoint in non-recovering patients compared to recovering patients (Fig. 1B) . ANGPT2 and SOFA score were significantly correlated (R=0.49, P=0.0016) (Fig. 1C) . To further investigate if ANGPT2 levels correlated with survival, we performed receiver operating characteristics (ROC) analysis on the late timepoint and found that a cutoff value of 7.4 ng/ml for plasma ANGPT2 significantly (P<0.05) predicted mortality in our cohort (Suppl. Fig1A, Fig. 1D ). ANGPT1 levels were not different between patients and controls ( Table 2) . Other clinical parameters, including PaO2/FiO2, C reactive protein, ferritin, and lactate are shown in Table 2 .

To investigate if ANGPT2 correlated with increased hypercoagulation, we assessed several markers of the coagulation system. Of note, all patients received prophylactic anticoagulation therapy with dalteparin sodium during the ICU stay (5) . Platelet counts were significantly (P<0.05) increased in patients at the late timepoint compared to the reference interval of healthy controls, as well as to the early timepoint (Table 2) . D-dimer levels were significantly increased (P<0.05) in all patients, however, without any difference between timepoints or outcome ( Table  2 ). Next, we measured levels of von Willebrand factor (VWF). All patients had significantly (P<0.01) increased VWF levels at the early timepoint, which further increased at the late timepoint, but without difference between recovering or non-recovering patients ( Fig. 2A) . ADAMTS13 is a metalloprotease that degrades large VWF multimers, thereby decreasing VWF's procoagulation properties (34) . We found that ADAMTS13 was significantly (P<0.01) decreased in the patients already at the early timepoint (Fig. 2B ). In addition, there was an even further reduction (P<0.01) of ADAMTS13 at the late timepoint in non-recovering patients (Fig.  2B ). ADAMTS13 and ANGPT2 levels were inversely correlated (R=0.457, P=0.0002) (Fig.  2C ). Receiver operating characteristics (ROC) analysis of ADAMTS13 identified a cutoff value of 424 ng/ml that significantly (P<0.05) predicted mortality in our cohort (Suppl. Fig1B, Fig.  2D ). All patients had significantly decreased levels of activated protein C (APC) compared to controls (P<0.0001) (Fig. 2E ). Thromboelastography (TEG) was performed on some patients during their ICU stay. TEG maximal amplitude (MA), representing thrombus strength, was significantly increased at the late timepoint (P<0.01, P<0.0001) with significant (P<0.05) further increase in non-recovering patients (Fig. 2F) . MA values ≥ 77 mm were only present at the late timepoint in non-recovering patients ( Table 2) . MA from TEG and ANGPT2 were significantly (P<0.05) correlated (Fig. 2G) . We did not find differences in other TEG parameters ( Table 2) .

To investigate if elevated ANGPT2 can directly affect coagulation in vivo, we performed tail bleeding experiments in mice. We injected intraperitoneally a recombinant His-tagged human ANGPT2 fragment corresponding to the thrombomodulin and Tie2 binding region and measured tail bleeding 15 minutes later. Control mice received recombinant His-tagged human IgG or albumin. Two different doses of ANGPT2 (25 and 250 µg/kg) were used. Injection of 250 µg/kg of ANGPT2 resulted in significantly (P<0.01) reduced bleeding time. A trend towards reduced bleeding time was observed also with 25 µg/kg of ANGPT2, but it was not statistically significant. Injection of ANGPT1 resulted in a bleeding time similar to that in controls (Fig. 3A) . The resulting plasma concentrations after ANGPT2 injection correlated roughly with the size of the injected dose Fig. 3B . To examine if ANGPT2 bound directly to thrombomodulin, we performed experiments on lung tissue after injection of 250 ug/kg ANGPT2, ANGPT1 or IgG. Thrombomodulin was immunoprecipitated, and the signals for His-tag and total thrombomodulin were evaluated. ANGPT2 showed significantly (P<0.01) more binding to thrombomodulin compared to IgG or ANGPT1 (Fig. 3C, D) . Interestingly, less thrombomodulin was immunoprecipitated from the lung tissue of ANGPT2-injected mice. To investigate this further, we performed Western blot analysis on lung lysates and probed for thrombomodulin and a loading control. These experiments revealed that ANGPT2-injected mice had significantly (P<0.05) lower thrombomodulin in lung lysates (Fig. 3E , F), suggesting that ANGPT2 directly or indirectly induce shedding of thrombomodulin from the endothelial surface. In contrast, mice with an induced endothelial-specific knockout of Angpt2 showed significantly (P<0.0001) increased bleeding time (Fig. 3G ). This change in tail bleeding time was independent of Tie2-dependent, because endothelial-specific Tie2 knockouts showed no differences in bleeding time (Fig. 3H ), in agreement with previously published data on Angpt1 knockout mice (29) .

To further study the effects of ANGPT2 on coagulation, we utilized thromboelastography (TEG) on human plasma supplemented with thrombomodulin and ANGPT2. As expected, thrombomodulin, through its negative regulation on coagulation, significantly (P<0.01) increased the time for coagulation to start (reaction time -TEG R) (Fig. 4A, B ). This effect was completely inhibited by ANGPT2 (Fig. 4A, B ). Trends towards a thrombomodulin-induced decreased thrombus strength (maximal amplitude -TEG MA) and its inhibition by ANGPT2 were also observed, although these effects were not statistically significant (Fig. 4C , D). It should be noted that plasma from all donors responded to thrombomodulin and ANGPT2 but to a variable degree. A TEG curve from one of the high responder donors can be seen in is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint 4E. Next, we investigated ANGPT2's effect on APC formation in human plasma in vitro. Soluble thrombomodulin was incubated with different doses of ANGPT2, ANGPT1, and control IgG to study thrombin/thrombomodulin mediated activation of protein C. Activated protein C was measured with chromogenic APC substrate. As expected, the addition of thrombomodulin significantly (P<0.0001) increased protein C activation (Fig. 4F) . Furthermore, these experiments showed that both ANGPT2 and ANGPT1 significantly (P<0.01) reduced thrombomodulin-mediated APC formation (Fig. 4F ).

Our comprehensive translational approach, comprising analysis of plasma and clinical features in critically ill COVID-19 patients together with mechanistic studies in mice and in vitro, suggests a novel role for ANGPT2 in hypercoagulation. We found that elevated ANGPT2 correlated with markers of coagulation in plasma in COVID-19 patients, with the highest levels in patients that subsequently died from the disease. Using mice, we further found a procoagulant effect of administered ANGPT2 and an anti-coagulant effect of genetic inactivation of the Angpt2 gene. In vitro experiments with human plasma showed that ANGPT2 inhibited thrombomodulin-mediated anticoagulation and activation of protein C. Taken together, our data suggest that elevated ANGPT2 might have an important pathogenic role in critically ill COVID-19 patients, and potentially also in other diseases with hypercoagulation. Our suggested function of ANGPT2 in hypercoagulation is summarized in Fig. 5 .

Circulating ANGPT2 levels correlated with severity of disease, hypercoagulation, and mortality in the studied cohort. Our results for ANGPT2 are in line with recently published data in COVID-19 patients (13) (14) (15) 35) . To investigate if ANGPT2 levels correlated with hypercoagulation in these patients, we assessed several markers of coagulopathy. Despite ongoing anticoagulant therapy, all patients displayed increasing levels of von Willebrand factor (VWF) over time during intensive care, however without difference between recovering and non-recovering patients (Fig. 2) . A recent study reported a correlation between VWF and mortality in ICU-treated COVID-19 patients (36) . ADAMTS13, a metalloprotease that degrades large VWF multimers and thereby decreases VWF's pro-hemostatic properties (34) , was significantly decreased in our COVID-19 patient cohort already at the early timepoint in ICU. In addition, there was an even further reduction of ADAMTS13 at the late timepoint in the non-recovering patients (Fig. 2 ). ADAMTS13 correlated with ANGPT2 and could independently predict mortality in our cohort (Fig. 2) . Previously, decreased concentrations of ADAMTS13 have been shown to correlate with mortality in COVID-19 and septic shock patients (15, 37) .

Clinical data for thromboelastography (TEG) were available for most patients in our study and showed that hypercoagulopathy was a general feature in these patients. This was seen as increased maximal amplitude (MA), but not as a decreased R-time. MA increased over time at ICU and correlated with mortality. MA also significantly correlated with ANGPT2 (Fig. 2) . The risk of developing venous thromboembolism in orthopedic trauma patients in known to . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint increase when MA ≥ 65 mm, and further doubles when MA ≥ 72 mm (38) . In this study, all patients had MA ≥ 65 mm, and at the late timepoint, 3 out of 4 non-recovering patients had MA ≥ 77 mm (Table 2) . Increased MA has previously been reported in critically ill COVID-19 patients (5, 39, 40) . Platelet count and D-dimer were not different between recovering and nonrecovering patients, but both were significantly increased compared to healthy controls (Table  2) . Recently, thrombin-antithrombin complexes and factor VIIIa were shown to be significant upregulated in critically ill ICU contained COVID-19 patients compared to hospitalized non-ICU COVID-19 patients (36) .

Recently, Daly et al demonstrated that ANGPT2 binds thrombomodulin (29) in addition to the already known binding partners Tie2 (41) and integrin β1α5 (26) . This raises the interesting possibility that ANGPT2 may directly inhibit thrombomodulin-mediated anticoagulation in vivo. Thrombin coupled to thrombomodulin converts protein C to activated protein C (APC), an endogenous protein that promotes fibrinolysis and inhibits thrombosis and inflammation (30) . Daly et al showed in vitro that ANGPT2 as well as ANGPT1 binding to thrombomodulin inhibit its binding to thrombin and subsequent activation of protein C (29) . An ANGPT2dependent inhibition of thrombomodulin has not previously been demonstrated in vivo. In the present study, we found a significant decrease of APC in all COVID-19 patients in comparison with to healthy controls (Fig. 2) . One limitation of our assay is that the measured APC levels were close to the detection limit, hence, possible differences between groups would not be detected. In contrast, a recent study did not find changes in APC in critically ill COVID-19 patients (36) . However, reduced levels of APC are found in a majority of patients in sepsis and are associated with increased risk of death (42) (43) (44) (45) . APC formation may be impaired because of down-regulation or shedding of thrombomodulin induced by inflammatory cytokines (46) , and, as we hypothesize herein, by ANGPT2 binding to thrombomodulin. Recent data show that circulating thrombomodulin was elevated in critically ill COVID-19 patients, suggesting shedding of thrombomodulin from the endothelium (15, 36) . Furthermore, Goshua et al, reported that soluble thrombomodulin correlates with mortality in critically ill COVID-19 patients (36) . In the current study, we noted that injection of ANGPT2 in mice resulted in the loss of thrombomodulin in lung tissue (Fig. 3) . Further studies are needed to investigate if this is a direct or indirect effect of ANGPT2.

To investigate if ANGPT2 can directly affect coagulation in vivo, we performed several experiments in mice. One simple but highly relevant experiment to evaluate coagulation is tail bleeding time (47) . In these experiments, recombinant ANGPT2 and ANGPT1 fragments corresponding to the thrombomodulin and TIE2 binding region were injected before the measurement of tail bleeding time. These experiments showed that ANGPT2, but not ANGPT1, could decrease bleeding time in vivo. Immunoprecipitation of thrombomodulin from lung tissue showed that it had bound ANGPT2, and to a smaller extent also ANGPT1 (Fig. 3) . In contrast, mice with endothelial-specific deletion of Angpt2 displayed longer bleeding times. Furthermore, endothelial-specific deletion of Tie2 did not change bleeding time, excluding a Tie2-dependent mechanism. Angpt1 knockout mice were also not affected, as shown in a previous study (29) . In contrast, Higgins et al, reported that heterozygous Tie2 knockout mice had an increased thrombotic response at the site of laser injury compared to controls (12) . As is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint Tie2 signaling regulates the transcription of Angpt2 (48) , it is intriguing to speculate that heterozygous Tie2 knockout mice had more Angpt2 protein stored in endothelial secretory vesicles , and that increased local release of Angpt2 occurred upon laser injury. Nonetheless, it is evident that more studies are needed concerning this pathway and its implications for disease.

TEG was also used to study ANGPT2 inhibition of thrombomodulin-mediated anticoagulation in freshly collected plasma from healthy donors. All donor plasma had a thrombomodulinmediated increase in reaction time that could be inhibited by ANGPT2 (Fig. 4) . The ability of ANGPT2 and ANGPT1 to inhibit thrombomodulin-mediated APC production was investigated in an APC assay with human plasma and a chromogenic APC substrate. In this assay, we found that both ANGPT2 and ANGPT1 could significantly reduce activation of protein C by approximately 50% (Fig. 4) , which is in agreement with the data reported by Daly et al (29) . Why is this effect elicited by ANGPT1 in vitro, but not in vivo? A possible answer is provided by Daly et al, who showed that in the presence of both thrombomodulin and TIE2, ANGPT1 would preferably bind TIE2, whereas ANGPT2 bound both thrombomodulin and TIE2 (29) . Currently, we do not know the local concentrations of ANGPT2 and ANGPT1 in vivo. Endothelial ANGPT2 levels vary extensively among vessel types and location and are upregulated in response to angiogenic and inflammatory activation. It is therefore possible that inhibition of thrombomodulin occurs only locally at sites of high ANGPT2 release. ANGPT1 on the other hand, is important for endothelial stabilization and anti-inflammatory properties through TIE2 signaling. In line with this, release of ANGPT1 in conjunction with platelet degranulation and resulting signaling through TIE2 have been shown to be important for endothelial closure after neutrophil extravasation (49) . Although ANGPT1 might inhibit thrombomodulin in vitro, this may not be a major concern in vivo, as most studies show unchanged or decreased ANGPT1 in disease, including the current study and a large sepsis study (12) .

Our data suggest that inhibition of ANGPT2 may be explored as a therapeutic approach in COVID-19 and other diseases with hypercoagulation. A compound known as trebananib (formerly AMG386) binds to both ANGPT2 and ANGPT1 and inhibit interaction with TIE2 (50). Trebananib has been tested in multiple clinical cancer trials and has been administered to a large number of patients (51) (52) (53) , and to our knowledge, coagulation disorders have not been reported as adverse effects in these trials. While trebananib might be an interesting compound to test in COVID-19 patients, its binding and inhibition of ANGPT1-TIE2 signaling may complicate matters as ANGPT1-TIE2 signaling in known to protect the vasculature and decreases inflammation (28) . Inhibiting ANGPT1 may therefore have adverse effects. There are currently several clinical trials registered with Angiopoietin-2 antibodies, many for the treatment of solid tumors. Critically ill COVID-19 patients may benefit from these antibodies. Another very interesting compound at the preclinical stage is ABTAA, a humanized Angiopoietin-2 Binding Tie2 Antibody (28) . ABTAA binds and clusters ANGPT2, converting it into a TIE2-activating molecule while decreasing free ANGPT2 (which antagonizes TIE signaling) at the same time. ABTAA treatment has shown promising results in experimental models of sepsis (28), however, coagulation was not evaluated. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint

We acknowledge certain limitations in our study. First, this study was neither designed nor powered to test the performance of parameters for outcome prediction. However, our findings are plausible, hypothesis-generating, and clearly deserve validation in a larger cohort of patients. Second, most of the COVID-19 patients were male, as this sex is overrepresented among COVID-19 patients in intensive care, and inference of our results to female COVID-19 patients should therefore be made with caution. Importantly, experiments in mice and analyses on donor blood had representation of both sexes.

In conclusion, we show that ANGPT2 levels in critically ill COVID-19 patients correlate with severity of disease, hypercoagulation, and mortality. In addition, we provide novel in vivo evidence for a direct role for ANGPT2 in coagulation through binding to and inhibition of thrombomodulin-mediated anticoagulation. These findings suggest that inhibition of ANGPT2 might not only benefit critically ill COVID-19 patients but also other patients with hypercoagulation.

This study was performed with patients at Uppsala University Hospital, Sweden. 20 patients with confirmed Sars-CoV-2 infection with an ICU duration longer than 10 days were included. Plasma samples were collected within 4 days after admission (early timepoint) and 10-14 days after admission (late timepoint). For each patient, clinical evaluation, and baseline characteristics (demography, pre COVID-19 treatments, clinical manifestations, cardiovascular risk factors, and body mass index) were retrieved from patient records. Healthy controls were asymptomatic adult blood donors, matched with patient cases for age and gender.

The study was approved by the Swedish Ethical Review Authority at the Ministry of Education and Research for patient samples (approved permit 2020-01623) and blood donor samples (approved permit 01/367), in accordance with the Swedish Ethical Review Act (SFS 2003:460). Informed consent was obtained from the patient, or next of kin if the patient was unable to give consent. The Declaration of Helsinki and its subsequent revisions were followed. STROBE guidelines were followed for reporting. All animal experiments were approved by the Uppsala Committee of Ethics of Animal Experiments (approved permit 5.8.18-04862-2020), in accordance with the Swedish Animal Protection Act (SFS 1988:534), and were conducted according to guidelines established by the Swedish Board of Agriculture.

All samples were collected in EDTA or 0.129 M trisodium citrate tubes (9NC BD Vacutainer, Becton Dickinson). Plasma was obtained after centrifugation at 3000 g for 10 min and stored at -80ºC until analysis. ELISA's were used to measure plasma protein concentrations for Angiopoietin-2 (DANG20, R&D Systems), Angiopoietin-1 (DANG10, R&D Systems), von Willebrand factor (ab108918, Abcam), and ADAMTS13 (ab234559, Abcam) according to the is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint manufacturer's instructions. Activated protein C was evaluated with a chromogenic APC substrate. A standard curve was created by a dilution series from pooled samples from five healthy controls. Samples (50 µl) from the dilution series, controls, and COVID-19 patients were incubated with 0.1 U of thrombin inhibitor hirudin (H0393, Sigma) for 10 minutes at 37ºC in a 96-well plate. After the addition of 50 µl chromogenic APC substrate (229021, Biophen CS-21(66)) the increase in absorbance was measured at 405 nm for 8 minutes (linear phase) at 37ºC in a temperature-controlled plate reader (Synergy HT, Biotek). The area under the curve was used to calculate APC concentrations expressed as arbitrary units (a.u.).

Floxed Angpt2 mice (54) and Tie2 mice (54) were crossed to tamoxifen inducible Cdh5-Cre ERT2 (55) mice to generate endothelial specific knockout of Angpt2 (Angpt2 iECKO ) and Tie2 (Tie2 iECKO ). Controls were littermate mice with wt/wt alleles for Angpt2 (WT) and Tie2 (WT). Mice were genotyped with primers for Angpt2 (for 5'-GGGAAACCTCAACACTCCAA and rev 5'-ACACCGGCCTCAAGACACAC, wt 222 bp, floxed 258 bp), Tie2 flox (for 5'-TCCTTGCCGCCAACTTGTAAAC and rev 5'-TTTCCTCCTCTCCTGACTACTCC, 604 bp), Tie2 wt (for 5'-TCCTTGCCGCCAACTTGTAAAC and rev 5'-AGCAAGCTGACTCCACAGAGAAC, 175 bp), and general Cre allele (for 5-ATGTCCAATTTACTGACCG and rev 5'-CGCCGCA TAACCAGTGAA, 673 bp). Knockout was induced with 3 doses of tamoxifen (2 mg) in peanut oil by oral gavage at 4 weeks of age. Mice for other experiment came from in house breeding on a C57BL6/J background. All experiments were performed in both female and male mice.

Mice with isoflurane anesthesia were subjected to surgical dissection of the tail (3 mm from the tip). The tail was prewarmed for 2 minutes before dissection and immediately after immersed in buffered saline prewarmed to 37°C. The time of bleeding was recorded. The tail bleeding assay were performed in 6-12-week-old Angpt2 iECKO Tie2 iECKO , and control mice (WT). In addition, the same experiment was performed in WT mice 15 minutes after receiving an i.p. injection of a recombinant human Angiopoietin-2 His tagged fragment (ab220589, Abcam) or a recombinant human Albumin His tagged fragment (ab217817, Abcam). The proteins were diluted in PBS and injected at 25 µg/kg and 250 µg/kg body weight. After the assay, heart puncture was performed to collect blood diluted 1:10 in citate-dextrose anticoagulant (C3821, Sigma), centrifuged and prepared as above. Lungs were harvested, snap frozen and stored at -80ºC for later protein analysis.

Plasma concentrations of injected recombinant human ANGPT2 was measured by ELISA as above (DANG20, R&D Systems). Immunoprecipitation experiments were performed to evaluate the binding of ANGPT2 and ANGPT1 to thrombomodulin after injection. Lung tissue was homogenized in RIPA buffer (89901, Pierce) with proteas and phosphatase inhibitor (A32959, Pierce). Lung lysates were immunoprecipitated with a rabbit anti-mouse thrombomodulin antibody (ab230010, Abcam) attached to protein G conjugated Dynabeads (10004D, Thermo Fisher Scientific). Immunoprecipitated proteins were separated on 4-20% . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint Mini Protean TGX gels (4561094, Biorad) and then transferred using Trans-blot turbo 0.2 µm PVDF membranes (1704156, Biorad). Blots were blocked with 5% BSA for 1 h and incubated overnight with mouse anti-6X His tag antibody (27E8, 2366, Cell Signaling). After washing and incubating with anti-mouse HRP conjugated secondary antibody (NA931, Sigma), proteins were visualized using ECL plus detection reagents (GERPN2232, Sigma). Blots were stripped with Re-Blot Plus Strong solution (2504, Millipore), blocked, and probed with antithrombomodulin antibody followed by anti-rabbit HRP conjugated secondary antibody (711-035-152, Jackson Immuno Research). Total thrombomodulin in lung lysates was done as above and correlated to loading control, anti-Gapdh HRP conjugated antibody (ab9482, Abcam). Band density was quantified with ImageJ (NIH).

Kaolin activated thromboelastography (TEG) was utilized to evaluate the effect of ANGPT2 on thrombomodulin dependent anticoagulation. Freshly drawn human citrate plasma, from both female and male donors, was incubated in kaolin tubes with 1000 ng/ml recombinant thrombomodulin (3947-PA-010, R&D Systems) and 100 ng/ml ANGPT2 (ab220589, Abcam) at room temperature for 20 minutes. Analysis was carried out on a TEG5000 (Nordic Biolabs) for 15 minutes + reaction time (R). The measurements were started by the addition of plasma to the TEG cups containing 20 µl 0.2 M CaCl2.Samples were run in duplicate and averaged. A sample without additives was run at the start and end of the experiment and averaged as control.

Effect of ANGPT2 on thrombomodulin dependent activation of protein C Pooled plasma from healthy controls was incubated with the same volume of 6 mM CaCl2 with 0.2 U/ml thrombin (T8885, Sigma), 1000 ng/ml recombinant thrombomodulin (3947-PA-010, R&D Systems), ANGPT2 (ab220589, Abcam), ANGPT1 (ab69492, Abcam), IgG (ab219660, Abcam) at 37ºC for 30 minutes at indicated concentrations in Fig. 4 . The reaction was terminated by addition of 0.2 U/ml hirudin (H0393, Sigma) at 37ºC for 10 minutes. The generation of APC was evaluated by adding chromogenic APC substrate as above.

Data is expressed as geometric mean ± geometric 95% confidence interval. To test for statistical differences, we utilized Student's t-test or ANOVA (>2 groups) where appropriate. ANOVA was followed by Bonferroni's pos hoc test. Data were tested for normal distribution and in the case of uneven distribution, data was log transformed before statistical analysis. Pearson correlation was used to measure dependence between two variables. Receiver operating characteristics (ROC) curve was used to determine the cut off value for ANGPT2 and ADAMTS13, to compare survival curves a log-rank test was performed. All statistical analysis was done in GraphPad Prism 8. All analyses were 2-sided and a P value of P<0.05 was considered statistically significant. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint

Geometric mean ± 95% CI for VWF (A), ADAMTS13 (B) for healthy controls (HC), recovered (Rec) and deceased (Dec) patients at shortly after admission (early) and 10-14 days after admission (late). (C) Correlation between ADAMTS13 and ANGPT2. Receiver operating characteristics (ROC) curve was used to determine a cutoff value of 424 ng/ml plasma ADAMTS13. (D) Survival graph with the determined ADAMTS13 cutoff of 424 ng/ml which could significantly predict mortality (D). Data as above for activated protein C (E), and maximal amplitude (MA) (F) from thromboelastography (TEG). (G) Correlation between MA from TEG and ANGPT2. Of note, not all patients had TEG data. Additional results from TEG analysis can be found in Table 2 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted January 15, 2021. ; https://doi.org/10.1101/2021.01.13.21249429 doi: medRxiv preprint

risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

COVID-19 and coagulation: bleeding and thrombotic manifestations of SARS-CoV-2 infection

Incidence of thrombotic complications in critically ill ICU patients with COVID-19

Facing COVID-19 in the ICU: vascular dysfunction, thrombosis, and dysregulated inflammation

Inadequate prophylactic effect of low-molecular weight heparin in critically ill COVID-19 patients

Nervous system involvement after infection with COVID-19 and other coronaviruses

Abnormal coagulation parameters are associated with poor prognosis in patients with novel coronavirus pneumonia

Clinical Characteristics of Coronavirus Disease 2019 in China

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Endothelial Activation: The Ang/Tie Axis in Sepsis

Circulating angiopoietin-2 and the risk of mortality in patients with acute respiratory distress syndrome: a systematic review and meta-analysis of 10 prospective cohort studies

Tie2 protects the vasculature against thrombus formation in systemic inflammation

Angiopoietin-2 as a marker of endothelial activation is a good predictor factor for intensive care unit admission of COVID-19 patients

Circulating markers of angiogenesis and endotheliopathy in COVID-19

Microvascular dysfunction in COVID-19: the MYSTIC study

Endothelial glycocalyx breakdown is mediated by angiopoietin-2

Gene control of tyrosine kinase TIE2 and vascular manifestations of infections

Tie2 expression and phosphorylation in angiogenic and quiescent adult tissues

Angiopoietin-1 protects the adult vasculature against plasma leakage

Leakage-resistant blood vessels in mice transgenically overexpressing angiopoietin-1

Angiopoietin-2 is critical for cytokine-induced vascular leakage

Angiopoietin-2 sensitizes endothelial cells to TNF-[alpha] and has a crucial role in the induction of inflammation

Angiopoietin 2 mediates microvascular and hemodynamic alterations in sepsis

Angiopoietin-2 may contribute to multiple organ dysfunction and death in sepsis*

Angiopoietin 2 is a partial agonist/antagonist of Tie2 signaling in the endothelium

Endothelial destabilization by angiopoietin-2 via integrin beta1 activation

Tie2 Activation Promotes Protection and Reconstitution of the Endothelial Glycocalyx in Human Sepsis

Amelioration of sepsis by TIE2 activation-induced vascular protection

Angiopoietins bind thrombomodulin and inhibit its function as a thrombin cofactor

Inflammation and the activated protein C anticoagulant pathway

Thrombomodulin in disseminated intravascular coagulation and other critical conditions-a multi-faceted anticoagulant protein with therapeutic potential

Endothelium-specific loss of murine thrombomodulin disrupts the protein C anticoagulant pathway and causes juvenile-onset thrombosis

The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine

Purification of human von Willebrand factor-cleaving protease and its identification as a new member of the metalloproteinase family

Circulating Markers of Angiogenesis and Endotheliopathy in COVID-19

Endotheliopathy in COVID-19-associated coagulopathy: evidence from a single-centre, cross-sectional study

The prognostic value of ADAMTS13 (a disintegrin and metalloprotease with thrombospondin type 1 repeats, member 13) deficiency in septic shock patients involves interleukin-6 and is not dependent on disseminated intravascular coagulation

Can Thrombelastography Predict Venous Thromboembolic Events in Patients With Severe Extremity Trauma?

COVID-19-Related Severe Hypercoagulability in Patients Admitted to Intensive Care Unit for Acute Respiratory Failure

Hypercoagulability of COVID-19 patients in intensive care unit: A report of thromboelastography findings and other parameters of hemostasis

Isolation of angiopoietin-1, a ligand for the TIE2 receptor, by secretion-trap expression cloning

Septic shock, multiple organ failure, and disseminated intravascular coagulation. Compared patterns of antithrombin III, protein C, and protein S deficiencies

Time course of hemostatic abnormalities in sepsis and its relation to outcome

Changes of the hemostatic network in critically ill patients--is there a difference between sepsis, trauma, and neurosurgery patients?

Epidemic meningococcemia and purpura fulminans with induced protein C deficiency

Release of thrombomodulin from endothelial cells by concerted action of TNF-alpha and neutrophils: in vivo and in vitro studies

Animal Models Subcommittee of the, I. Standardization Committee Of The, Towards a standardization of the murine tail bleeding model

Plastic roles of pericytes in the blood-retinal barrier

Platelets docking to VWF prevent leaks during leukocyte extravasation by stimulating Tie-2

Suppression of angiogenesis and tumor growth by selective inhibition of angiopoietin-2

Randomized, double-blind, placebo-controlled phase II study of AMG 386 combined with weekly paclitaxel in patients with recurrent ovarian cancer

Anti-angiopoietin therapy with trebananib for recurrent ovarian cancer (TRINOVA-1): a randomised, multicentre, double-blind

Safety, pharmacokinetics, and antitumor activity of AMG 386, a selective angiopoietin inhibitor, in adult patients with advanced solid tumors

A lymphatic defect causes ocular hypertension and glaucoma in mice

Inducible gene targeting in the neonatal vasculature and analysis of retinal angiogenesis in mice

We thank Jana Chmielniakova, Pia Peterson, and Cecilia Olsson at Uppsala University for technical assistance, as well as research nurses Joanna Wessbergh and Elin Söderman, and the biobank assistants Erik Danielsson and Philip Karlsson for their expertise in compiling patient samples. We thank Peetra Magnusson at Uppsala University for valuable comments on the manuscript. 

The authors have declared that no conflict of interest exists Data and materials availability: All data are presented within the paper