Testing the effect of PAR1 inhibitors on Plasmodium falciparum

Sequestration and cytoadherence of Background: Plasmodium falciparum -infected erythrocytes (IE) to microvascular endothelium alters endothelial barrier function and plays a role in the pathogenesis of severe malaria. Binding of IE is mediated by erythrocyte membrane protein 1 P. falciparum (PfEMP1) and the PfEMP1 variants that binds to endothelial protein C receptor (EPCR) have, in particular, been associated with the dysregulation of the coagulation/inflammation pathways in endothelial cells. This has prompted speculation about the role of protease-activated receptor-1 (PAR1) activation and signalling in causing endothelial activation and loss of barrier function in cerebral malaria. We used a co-culture of primary human brain microvascular Methods: endothelial cells (HBMEC) with material, recombinant P. falciparum PfEMP1 or lysates from IE, and measured barrier function by trans endothelial electrical resistance (TEER).  A selection of PAR1 inhibitors was tested for their ability to reverse the and thrombin induced P. falciparum decrease in barrier function. An initial screen in the presence of recombinant PfEMP1 identified Results: a few inhibitors that were able to reduce the rapid thrombin-induced barrier disruption even when activated protein C (aPC) was unable to do so. However, in the IE lysate co-culture system we identified a mechanism that slowly reduces barrier function and which is insensitive to PAR1 inhibitors. The selected PAR1 inhibitors were able to reverse the Conclusions: disruption of barrier function by thrombin but did not reverse the IE lysate induced disruption of barrier function, implicating a different PAR1-independent mechanism.  These findings have implications for the design of adjunct therapies to reduce brain swelling in cerebral malaria.


Introduction
The pathology of cerebral malaria (CM) is not fully understood but is associated with sequestration of P. falciparum-infected erythrocyctes (IE) and involves interactions between IE and host endothelial cells (EC) as well as host inflammation (for a review see (Wassmer & Grau, 2017)). In a clinical study in Malawi, brain swelling was found to be strongly associated with fatal outcome in CM (Seydel et al., 2015), which has subsequently been confirmed by studies in India (Mohanty et al., 2017). The aetiology of this swelling is probably multifactorial, involving cytopathic and vasogenic mechanisms (e.g. local and systemic inflammation; endothelial cell death; IE and erythrocyte accumulation). Multiple lines of work indicate involvement of the protein C (PC)-protease-activated receptor-1 (PAR1) axis and the consequences of the loss or inactivation of endothelial protein C receptor (EPCR) (for a review see (Mohan Rao et al., 2014)) (Kessler et al., 2017). PAR1 is a seven-transmembrane G protein-coupled receptor and is the prototypical thrombin receptor. It contains its own tethered ligand and exhibits biased antagonism. Cleavage events at two different sites lead to opposing effects: when cleaved at amino acid R41 by thrombin, PAR1 is pro-inflammatory and barrier disrupting whereas non-canonical cleavage of PAR1 at amino acid R46 by activated PC (aPC) has a barrier stabilising and anti-inflammatory effect, significantly reducing the inflammatory and barrier disruptive effect of thrombin on endothelial cells. Non-canonical cleavage of PAR1 by aPC is dependent on aPC binding to the active site of EPCR ( Figure 1). This dysregulation of the coagulation system does not often result in clinically evident thrombosis or bleeding in children and some researchers in this field have stressed the distinction between cytopathic and coagulant effects in this pathway (Mosnier et al., 2007). However, laboratory measurements of coagulation factors do show coagulation activation that is associated with fatal outcome in cerebral malaria (CM) (Moxon et al., 2015).
Two mechanisms for the action of cytoadherence on coagulation/ inflammation dysregulation have been proposed (Figure 1). The direct binding of the P. falciparum variant surface protein P. falciparum erythrocyte membrane protein 1 (PfEMP1) to EPCR blocks the conversion of PC to aPC, affecting the ability of this host control system to control PAR1 cleavage by thrombin and altering the PAR1 signalling pathway (Avril et al., 2019;Gillrie et al., 2015;Kessler et al., 2017;Mosnier & Lavstsen, 2016;Sampath et al., 2015). This mechanism is dependent on the presence of EPCR-binding parasite variants, and these PfEMP1 types are strongly associated with severe malaria (Mkumbaye et al., 2017;Storm et al., 2019). However, the same effect can also be achieved through the reduction of expression of EPCR on the endothelial surface, which is thought to be expressed at lower levels in the brain microvasculature. We (Moxon et al., 2013) have shown that this does occur and is associated with CM, and while linked to IE sequestration, this appears to be due to receptor shedding rather than steric inhibition and does not require cytoadherence via EPCR (Moxon et al., 2013). Either mechanism would disrupt EPCR function and its capacity to form a complex with aPC and modify PAR1 signalling; thus leading to an unmodified effect of Under normal circumstances, the presence of endothelial protein C receptor (EPCR)-activated protein C (aPC) on the endothelial surface is able to modulate the protease-activated receptor 1 (PAR1) response (by cleavage of R46) to thrombin along a cytoprotective pathway. In the presence of IE, the binding site for aPC can be occupied by P. falciparum erythrocyte membrane protein 1 (PfEMP1) or EPCR expression can be reduced, both limiting the availability of EPCR-aPC and resulting in cytopathic responses due to unmodified thrombin cleavage (R41) of PAR1. P. falciparum-infected erythrocytes (IE) are also able to release soluble factors that can influence endothelial integrity directly. thrombin with pro-inflammatory changes and loss of EC barrier function. Because the dysfunction of EPCR is at the level of the receptor, these cannot be mitigated through the addition of aPC (Figure 2A).
Our hypothesis is that, through both of these mechanisms, the Thrombin-PAR1 coagulation/inflammation axis plays a significant role in barrier loss in CM, and that adjunct treatments targeting this might alleviate mortality or post-CM neurological sequelae in CM. Since the activation of PC may be prevented by either of these mechanisms of EPCR disruption (steric inhibition or receptor cleavage) it is important to identify treatments that could be barrier stabilising to EPCR abrogation by either mechanism. We therefore investigated a range of PAR1 antagonists that are able to directly inhibit thrombin cleavage of the PAR1 extracellular domain without affecting other thrombin-dependent pathways. We tested their ability in preventing loss of barrier function in human endothelial cells in response to treatment with parasite material.

Culture of endothelial cells and Plasmodium falciparum
The human umbilical vein endothelial cell line EA.hy926 (ATCC) was cultured in DMEM medium supplemented with 10% Figure 2. Measuring effect of protease-activated receptor 1 (PAR1) inhibitors on endothelial barrier function of EA.hy926 cells using trans endothelial electrical resistance. (A) Cell Index traces for thrombin induced decrease in barrier function in EA.hy926 cells. Cell index was normalised at the time point immediately prior to the addition of 2 nM thrombin, indicated by the black triangle, and medium (black line) was set as baseline. The decrease in normalised cell index by thrombin (red) was set at 100% and the effect of 57 nM aPC (blue), 25 nM recombinant P. falciparum erythrocyte membrane protein 1 (rPfEMP1) (bright green) and 57 nM activated protein C (aPC) and 25 nM rPfEMP1 combined (dark green) on thrombin induced decrease in barrier function was determined. (B) The effect of inhibitors on thrombin induced decrease in barrier function. Inhibitors (concentrations in Table 1) were tested in the presence of 57 nM aPC and 25 nM rPfEMP1. The inhibitors in light brown do not have an effect, the inhibitors in light green have and intermediate effect and the inhibitors in purple reverse the effect of thrombin. (C) Graph of the data depicted in B with corresponding colours. Thrombin induced decrease in barrier function was set as 100% (red) and shown are the mean ± SD of 3 independent experiments for the 10 inhibitors. *denotes a P-value <0.05. foetal calf serum. Primary human brain microvascular endothelial cells (HBMEC, Cell Systems) were cultured in Endothelial Cell Growth Medium 2 containing 2% foetal calf serum, 5 ng/ml epidermal growth factor, 10 ng/ml basic fibroblast growth factor, 20 ng/ml insulin-like growth factor, 0.5 ng/ml vascular endothelial growth factor 165, 1 µg/ml ascorbic acid, 0.2 µg/ml hydrocortisone and 22.5 µg/ml heparin (EGM2, Promocell). The Plasmodium falciparum IT4 lab strains IT4var16 (ItG), IT4var14 (A4) and IT4var37 (4E12) were cultured according to our standard laboratory methods (Wu et al., 2018) in complete RPMI medium (RPMI 1640 with 25 mM HEPES, 11 mM glucose, 2 mM glutamine, 0.2% NaHCO 3 , 25 mg/l gentamicin and 10% Human Serum, pH 7.4) in normal group O red blood cells (RBC) at 3% haematocrit. rPfEMP1 preparation and PAR1 antagonists Recombinant EPCR-binding PfEMP1 (rPfEMP1) was produced as previously described (Turner et al., 2015). In short, HIS-tagged cysteine-rich interdomain region domain α1 (CIDRα1) protein derived from the IT4var20 PfEMP1 sequence was produced in the baculo-virus expression system, purified by nickel affinity chromatography and validated for binding to EPCR in ELISA. PAR1 antagonists, supplied by Eisai Co. Ltd. (Japan), and listed in Table 1, or commercially sourced from Sigma (Varopaxar), were dissolved at 10 mM in DMSO and stored in aliquots at -20°C.

Preparation of parasite lysates
Lysates of RBC and mid-trophozoite stage IE were prepared by freeze-thawing three times a suspension of 1 % haematocrit or a suspension of 50% parasitaemia at 1% haematocrit, respectively, in EGM2-min medium (EGM2 without hydrocortisone and heparin). Enrichment of the IE by gelatine flotation was used to obtain 50% parasitaemia.
Trans endothelial electrical resistance (TEER) analysis Barrier function was measured by TEER with the iCELLi-gence™ system (ACEA Biosciences). EA.hy926 cells and HBMEC (up to passage 9) were seeded in 8-well E plates (L8) and TEER was recorded at specific time intervals and expressed as cell index. For the initial inhibitor screen, EAhy926 cells were grown overnight and when confluent, the culture medium was exchanged for serum-free DMEM medium and combinations of 25 nM rPfEMP1, 57 nM aPC or PAR1 inhibitors (see Table 1 for concentrations) or vehicle were added 2 hours prior to treatment with 2 nM thrombin (see Figure 2 for details, n=3). For subsequent experiments with HBMEC, the overnight culture medium was first replaced with EGM2 medium and then with EGM2-min medium 2 hours prior to the addition of lysates in the absence or presence of selected PAR1 inhibitors (n=5). After ~20 hours, the additional effect of 2 nM thrombin was determined (n=2). Control experiments were performed with medium or inhibitor only. The cell index of medium only was set as baseline to correct for non-specific fluctuations of cell index during the experiment.
The decrease in barrier function by modulators was determined by normalising the cell index at the time point immediately prior to modulator addition. The maximum decrease in normalised cell index (NCI) by thrombin was determined for medium or lysate in the absence of inhibitor and set at 100%. The protective effect of inhibitor on the thrombin-induced disruption of barrier function calculated as follows: ∆NCI (thrombin + inhibitor)/∆NCI (thrombin − inhibitor) × 100.
For lysates, the normalised cell index was determined at 2 and 16 hours after adding the IT4var16 lysate. The decrease in normalised cell index was calculated and ∆NCI in absence of inhibitor was set at 100%. The effect of inhibitor on the lysate-induced decrease in barrier function was calculated as follows: ∆NCI (lysate + inhibitor)/∆NCI (lysate − inhibitor) × 100.

Measuring EPCR expression on endothelial cells
The effect of IT4var14 and IT4var37 IE or IE lysate on the EPCR expression on HBMEC was determined after 16 -20 hours of co-culture. A suspension of 50% parasitaemia at 1% haematocrit in EGM2-min medium was prepared and from this suspension lysate was also made, as described above. HBMEC were detached with Accutase ® , washed with cold PBS/1% BSA/ 2 mM EDTA, labelled with phycoerythrin conjugated rat anti-human EPCR antibody (Biolegend, 351904), washed and stained for cell viability with live/dead fixable yellow stain (Invitrogen). EPCR expression levels of viable cells were detected by flow cytometry; 10 ng/ml TNF was used as a positive control, since it decreases EPCR expression.

Statistical analysis
Statistical significance was calculated by ANOVA with Dunnett's post-test comparing the mean value (± SD) for each individual modulator addition (inhibitor, IE, lysate or TNF) with the mean value (± SD) of the condition without modulator (GraphPad Prism version 5). A P-value < 0.05 was deemed significant.

Results
Thrombin-induced barrier disruption and the effect of prior treatment with either PAR1 inhibitors or aPC in the presence of rPfEMP1 In the initial screen we investigated whether the effect of EPCR binding by PfEMP1 on thrombin-mediated barrier disruption of the human umbilical vein endothelial cell line EA.hy926 could be modified by PAR1 inhibitors. The loss of barrier function caused by the addition of thrombin was restored by the prior addition of aPC in the absence of rPfEMP1. However, aPC had no effect on preventing loss of barrier function when rPfEMP1 was bound to EPCR (see condition rPfEMP1-APC-thrombin in Figure 2) blocking aPC-mediated barrier strengthening. Thus, aPC might not be a useful treatment when EPCR-binding IE are present, such as in severe malaria. However, normal barrier function was obtained with the addition of inhibitors 1, 5 and 6 ( Figure 2), suggesting that they are able to act to stabilise EC barrier function despite the loss of protective aPC-mediated PAR1 cleavage. Raw data for Figure 2, in addition to that for Figure 3- Figure 6, are available as Underlying data (Storm et al., 2020).
Developing an endothelial cell barrier model using IE lysates Based on these findings we identified a subset of the Eisai compounds that were able to inhibit the action of thrombin on barrier function in the presence of rPfEMP1 and aPC. Next we used a co-culture system with primary HBMEC and IE. The work focussed on three Eisai compounds: E5555 (Atopaxar (Serebruany et al., 2009)), #04 (inhibitor 5) and #91 (inhibitor 1). E5555 was included as it has been through Phase 1 and Phase 2 clinical testing. In addition, we used the clinically licenced PAR1 inhibitor Vorapaxar (brand name Zontivity (Aralez Pharmaceuticals)), obtained as a non-clinical compound from Sigma-Aldrich.
The effect of PfEMP1 ligation to EPCR on barrier function has been described elsewhere (Gillrie et al., 2015;Kessler et al., 2017), but there has been little work on the impact of EPCR removal from the endothelial surface during cytoadherence. To address this, we focussed on EPCR abrogation in our co-culture model and first measured the EPCR expression on HBMEC after overnight co-culture with IE or IE-lysate ( Figure 3). Lysates had a more pronounced effect on EPCR expression to whole IE and were easier to prepare and more reproducible in TEER experiments. We therefore switched to a HBMEC/IE lysate co-culture system for subsequent experiments.
We observed that there were two 'behaviours' taking place in the TEER system ( Figure 4): A. A rapid response to thrombin in reducing EC barrier function transiently, which could be controlled under normal conditions using aPC ( Figure 4A).
B. A slower reduction in barrier function, activated by treatment with IE lysate and producing a cumulative effect over 30 hours ( Figure 4B) that was not seen in control wells with RBC lysate where barrier function remained constant throughout the incubation period. IE had less of an effect on barrier function than IE lysate. The fluctuations in cell index in the first 2 hours are consistent in each experiment, probably due to the addition of cells or cellular material.

Figure 3. Endothelial protein C receptor (EPCR) expression on human brain microvascular endothelial cells after overnight culturing
with P. falciparum-infected erythrocytes (IE) or IE lysate. EPCR expression was determined by flow cytometry after overnight culturing with IT4var14 or IT4var37 IE or IE lysate (n=2). EPCR expression on cells with medium only was set at 100%. Shown is mean ± SD of the combined IT4var14 and IT4var37 data (n=4) and the tumour necrosis factor (TNF) data (n=2) with * indicating a P-value <0.05.

Figure 5. Effect of inhibitors on thrombin-induced decrease in barrier function of human brain microvascular endothelial cells (HBMEC).
Cell index of HBMEC without (medium) and with P. falciparum-infected erythrocyte lysate in the absence and presence of 0.3 µM inhibitor was monitored for ~20 hrs prior to addition of 2 nM thrombin. The cell index was normalised at the time point immediately prior to the addition of thrombin and medium only was set as baseline. The maximum decrease in normalised cell index by thrombin was determined for medium or lysate in the absence of inhibitor and set at 100%. The inhibitors reverse the effect of thrombin. Shown are the mean ± SD of 2 independent experiments, with * indicating a P-value <0.05 compared to no inhibitor.

Figure 6. Effect of inhibitors on lysate-induced decrease in barrier function of human brain microvascular endothelial cells (HBMEC).
Barrier function was determined in the absence and presence of 0.3 µM inhibitor. Cell index was measured and normalised at the time point immediately prior to the addition of P. falciparum-infected erythrocyte lysate. The normalised cell index was determined at 2 and 16 hours after adding the lysate. The decrease in cell index was calculated and ∆normalised cell index in absence of inhibitor was set at 100%. Shown is the mean ± SD of 5 independent experiments with * indicating a P-value <0.05 compared to no inhibitor.
Behaviour A was as expected and confirms that aPC rescues thrombin induced barrier disruption in HBMEC as seen above in EAhy926 cells. Behaviour B has been reported previously by several investigators using a range of co-culture systems (Tripathi et al., 2007) (Gillrie et al., 2012, with either IE or parasite material affecting barrier integrity. Based on these findings we have tested the compounds for their ability to block both behaviours.

Effect of PAR1 inhibitors on thrombin-induced and lysateinduced reduction in EC barrier function
Compounds were tested for their ability to inhibit disruption of barrier function by thrombin and the slower effect caused by the IE lysate directly. All tested PAR1 inhibitors were able to significantly reverse thrombin-induced barrier disruption even in the presence of IE lysate ( Figure 5). However, three of the four inhibitors had no significant effect on the lysate-induced decrease in barrier function when used at 0.3 µM ( Figure 6) or at 0.6 µM (see Underlying data; Storm et al., 2020). Only E5555 significantly reverses the decrease in barrier function by IE lysate (-79.9 ± 16.5%, P = 0.026), albeit only by 20%. This still leaves a substantial EC barrier disruption by IE lysates, indicating that this phenomenon is independent of PAR1.

Discussion
There is considerable interest in developing adjunct therapies for CM based on controlling brain swelling, following the MRI finding of a strong association between this phenotype and death from CM (Mohanty et al., 2017;Seydel et al., 2015) and that brain swelling has also been associated with EPCRbinding IE phenotype (Kessler et al., 2017). As expected, the PAR1 inhibitors (including Vorapaxar) were able to block the effect of thrombin on transient reduction in barrier function, including in the presence of IE lysate. However, they had no significant effect on the slower reduction in barrier function caused by the lysate itself (thrombin independent).
Blocking of the thrombin-mediated pathway with the PAR1 antagonists might reduce brain swelling sufficiently to relieve mortality/morbidity in CM, but whether these types of inhibitors could be used safely in children/adults with multiple ring haemorrhages in their brains is a concern given the potential bleeding risk, owing to their inhibition of PAR1 in platelets (https://www.rxlist.com/zontivity-side-effects-drug-center.htm). Further consideration of risk/benefit calculations will be required to answer this.
The slower effect on brain barrier function by IE lysate requires further investigation. Other work has implicated a role for β-catenin in regulating brain barrier integrity (Gallego-Delgado et al., 2016), showing that IE-derived soluble mediators are able to cause a reduction in TEER of brain endothelial cells over several hours. Similarly, parasite kinins were also shown to reduced brain endothelial cell barrier function, as well as enhancing IE cytoadherence (Silva et al., 2019), and HRP2 has been implicated in modifying barrier function (Pal et al., 2016). Parasite histones have also been shown to cause barrier disruption on lung and brain endothelial cells and to be released in culture by IE and in vivo in patients with CM (Gillrie et al., 2012) and are increased in children with CM and associated with blood brain barrier breakdown and brain swelling (Moxon et al., 2019). For histones, there are some therapeutic avenues that could be explored. Non-anticoagulant heparins prevent histone-induced lethality in bacterial sepsis models in vivo (Wildhagen et al., 2014) and block the toxic effects of plasmodial histones in vitro. A non-anticoagulant heparin, Sevuparin, is being used in phase II trials for malaria for its potential to block parasite binding (Leitgeb et al., 2017;Saiwaew et al., 2017). Based on the work in this paper, an option would be to consider a combination therapy, based on stopping the thrombin-mediated effect using a PAR1 inhibitor combined with a non-anti-coagulant heparin to mitigate the histone effect. Interestingly, the two pathways are not completely separate, as aPC cleaves histones to regulate histone concentration in the vascular compartment (Gillrie et al., 2012;Xu et al., 2009) and histones can reduce aPC production (Kowalska et al., 2014).
In summary, we have shown that PAR1 inhibitors are able to reduce thrombin-induced barrier disruption even when the normal aPC control pathway has been disabled due to direct EPCR engagement with PfEMP1 or by reduced EPCR expression on EC induced by parasite lysates. The identification of a second pathway for barrier disruption induced by parasite lysate that is insensitive to PAR1 inhibitor treatment suggests that a more complex approach to adjunct therapy may be needed. This project contains the following underlying data:

Data availability
• Combined analysis data for fig 2_TEER_includes normalisation.csv (normalised data used to generate Figure 2).
• Combined analysis data for fig5_TEER_includes normalisation.csv (normalised data used to generate Figure 5).
• Combined analysis data for fig6_TEER_includes normalisation.csv (normalised data used to generate Figure 6).
Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).

Author information
Janet Storm and Yang Wu are joint first authors; Christopher A. Moxon and Alister G. Craig are joint last authors.
This work explores the possible mechanisms underlying cerebral malaria pathogenesis. It is therefore very relevant. The authors develop an model of the BBB and test the effect of in vitro Plasmodium and its effect on the ePCR/thrombin/PAR1 axis in barrier function of brain endothelial cells. falciparum The findings implicate that the parasite causes a more potent delayed barrier disruption that is not related to the classical ePCR/PAR1 model.

Minor points:
Include the number of IE per surface area added in every experiment or the equivalent for IE lysates.
Include the number of cells seeded per well (or surface area) used in xCelligence experiments.
Include the Cell index of cultures after the overnight growth before starting the experiments.
Indicate the stage of IE used in the experiments or if asynchronous cultures were used. If a single stage was used, explain how was it isolated.
"For subsequent experiments with HBMEC, the overnight culture medium was first replaced with and then with EGM2-min medium 2 hours prior to the addition of lysates in the EGM2 medium absence or presence of selected PAR1 inhibitors (n=5)." please indicate the time of incubation with EGM2 medium.
In Fig. 2, the effect of aPC and recombinant PfEMP1 alone on cells should be included as controls.
In Fig. 3, a control of uninfected RBC and their lysate should be included.
In Fig.4, it is not clear which are the lines corresponding to lighter and darker shades of green. Please use different colors.
In Figure 5, it would be interesting to see the original cell index graph showing the effect of In Figure 5, it would be interesting to see the original cell index graph showing the effect of thrombin on cells already incubated with the parasite lysate for 20 h.
In Figure 6 it is stated that "the normalised cell index was determined at 2 and 16 hours", but it is not clear which time point is shown in the graph or if it is the different between the values at these two time points. It would be informative to see the original cell index graphs as in figure 4 or to show the results at 2 h and 16 h separately in two independent graphs.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
No competing interests were disclosed.

Competing Interests:
Reviewer Expertise: Malaria, cerebral malaria, malaria anemia, malaria pathogenesis I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 19 May 2020 , Liverpool School of Tropical Medicine, Liverpool, UK Janet Storm We thank the reviewer for her constructive comments and have addressed these in our reply and in the manuscript.

Include the number of IE per surface area added in every experiment or the equivalent for IE lysates.
For IE and IE lysates, 300 µl was added per well and always at 50% parasitaemia at 1% haematocrit. This is 1.5 x 10 IE per well (0.64 cm surface area). This is now added to the methods. In addition, related to comment 4, the preparation of IE suspension has also been added to the methods section. . experiments Cell were seeded at 50,000 cells/cm , equal to 32,000 cells/well (surface area per well is 0.64 cm ). This is added to the methods.

Include the Cell index of cultures after the overnight growth before starting the experiments.
On average the cell index, after refreshing the medium and before adding lysates and/or inhibitors, is 6.9. This information is available in full in the raw data ( ). https://doi.org/10.6084/m9.figshare.11558889 Indicate the stage of IE used in the experiments or if asynchronous cultures were used. If a single stage was used, explain how was it isolated. The section in the methods describing the preparation of lysate has been altered and includes the preparation of IE suspension at 50% parasitaemia at 1% haematocrit. Lysates were made from these IE suspensions.

"For subsequent experiments with HBMEC, the overnight culture medium was first replaced with EGM2 medium and then with EGM2-min medium 2 hours prior to the addition of lysates in the absence or presence of selected PAR1 inhibitors (n=5)." please indicate the time of incubation with EGM2 medium.
Fresh EGM2 was normally added in the morning, resulting in 3-6 hrs incubation. This has been added to the methods.

, the effect of aPC and recombinant PfEMP1 alone on cells should be included as controls.
The TEER traces in figure 2A are a representative and are derived from an experiment that did not include aPC and rPfEMP1 on their own. We choose this representative experiment as it included the inhibitor data shown in figure 2B. From other experiments it was clear that neither aPC nor rPfEMP1 affected the normalised cell index more that 5%, this has been added to the results section.
In Fig. 3, a control of uninfected RBC and their lysate should be included. These controls are now added to figure 3, its legend and the text describing the result.
In Fig.4, it is not clear which are the lines corresponding to lighter and darker shades of green. Please use different colors. The colours in new figure 4C (was 4B) have been changed. IE lysate is now green and IE is cyan.

In Figure 5, it would be interesting to see the original cell index graph showing the effect of thrombin on cells already incubated with the parasite lysate for 20 h.
Cell index traces have been added to figure 5, as requested by the other reviewer too. It shows the slow recovery from thrombin after co-culture with IE lysate. For clarity, a similar cell index trace has been added to figure 4. See also our reply to the other reviewer.
In Figure 6 it is stated that "the normalised cell index was determined at 2 and 16 hours", but it is not clear which time point is shown in the graph or if it is the different between the values at these two time points. It would be informative to see the original cell index graphs as in figure 4 or to show the results at 2 h and 16 h separately in two independent graphs.
Representative cell index traces have also been added to figure 6 and indicate the 2 and 16 hour 2 2 Representative cell index traces have also been added to figure 6 and indicate the 2 and 16 hour time points. The difference between the normalised cell index at these time points was calculated and used to make the graph. These calculations are also provided in the raw data file "Combined analysis data for fig6_TEER_includes normalisation_2020-02-04.csv".
No competing interests were disclosed. Competing Interests:

Maria Bernabeu European Molecular Biology Laboratory, Barcelona, Spain
The study from Storm , describe how PAR1 inhibitors could counteract the loss of endothelial barrier et al. that occurs in cerebral malaria. The authors show that PAR1 inhibitors revert barrier disruption mediated by thrombin, but are not effective against long-term barrier breakdown caused by parasite lysates. The finding that PAR1 inhibitors are, even more effective than aPC, at rescuing thrombin-induced barrier opening in the presence of PfEMP1 EPCR binders is important for the field. The finding that other therapeutic approaches will be needed for barrier caused by parasitic molecules/toxins is also relevant. Overall, the study is well-designed and the conclusions are supported by the results. However, the study might benefit from additional improvements.

Methods:
The authors used the same parasitemia and hematocrit for long-term culture or to prepare the IEs lysate and they obtained significantly differences in barrier dysfunction. Did the authors observe merozoite egress after long term IE incubation? Are the molecules released on a mid-trophozoite lysate a good proxy of schizont molecules released after rupture? Is the concentration used "physiologically" present in the cerebral malaria microvasculature. Although I agree that the concentration used in this study might be necessary to observe a major barrier breakdown in in cultures, it might not be accurately mimic the pathogenic mechanisms on patients. A sentence vitro in the discussion reflecting the limitations of the model might improve the interpretation of the paper by readers outside the field.

Results:
Long term incubation with both IE or IE lysate showed 1) reduced surface expression of EPCR and reduced HBMEC cell index. However, the reduced cell index might not be a direct consequence of loss of EPCR in the surface. As the authors accurately describe in the discussion, loss of barrier in their experiments is more likely a consequence of the presence of parasite histones, HRP2 or other soluble mediators. However, the hypothesis exposed in the introduction might be misleading for the average reader. "Our hypothesis is that, through both of these

mechanisms, the Thrombin-PAR1 coagulation/inflammation axis plays a significant role in barrier mechanisms, the Thrombin-PAR1 coagulation/inflammation axis plays a significant role in barrier loss in CM, and that adjunct treatments targeting this might alleviate mortality or post-CM neurological sequelae in CM. Since the activation of PC may be prevented by either of these mechanisms of EPCR disruption (steric inhibition or receptor cleavage) it is important to identify
Since we treatments that could be barrier stabilising to EPCR abrogation by either mechanism." don't understand quite well the molecular mechanisms that induce loss of endothelial integrity and surface EPCR, the authors should be more careful while establishing associations.
I would recommend two include both 1) representative traces of cell index and bar plots in figure 3, 4, 5 and 6 in a similar way than in figure 2. Or include it in supplementary data. Was the experiment shown in figure 4 done on independent days? Although TEER measurements are highly variable depending on the day of the experiment, box plots and dot plots might better represent data distribution than bar graphs. Discussion: Avril , 2019 showed using the same methodology that schizont-stage parasites (EPCR et al. binding strains) prolonged thrombin-induced barrier disruption in resting and activated HBMEC, and HBMEC do not get back to baseline after IE incubation. Thrombin and IE colocalization is frequent in the microvasculature in cerebral malaria autopsies. As the effects seen in Avril are et al. likely to occur in patients this experiment is highly relevant. Is the experiment in Figure Figure 5 will clarify that.

Minor comments:
Abstract: "However, in the IE lysate co-culture system we identified a mechanism that slowly reduces barrier function and which is insensitive to PAR1 inhibitors." Please, rephrase to "However, PAR1 inhibitors did not rescued the barrier dysfunction after long-term co-culture with IE lysate". The molecular mechanisms has not been described in this manuscript.

Reproducibility of Methods:
Which amount of IE is used in the long-term co-culture in Figures 3, 4 and 5? The authors give the parasitemia, the hematocrit but as the volume is missing, the quantity or concentration of IE cannot be calculated. References:

Results
Kessler , 2017 is not well cited in the first paragraph as a review of aPC-PAR1 axis. Is this a et al.

Are the molecules released on a mid-trophozoite lysate a good proxy of schizont molecules released after rupture?
To mimic the release of schizont molecules, we wanted to use lysate of schizonts and we spent considerable time optimising culture methods to obtain parasites in schizogony. This proved to be rather difficult and therefore we made lysates of trophozoites in a late as possible stage. The cultures were regularly synchronised with sorbitol to keep them synchronous. In the methods section, preparation of IE and lysates has been altered to mid-late trophozoite stage. We did notice that lysates with a higher proportion of schizonts had a greater effect on the cell index, but for the experiments described in this manuscript it was not investigated further.
Is the concentration used "physiologically" present in the cerebral malaria microvasculature. Although I agree that the concentration used in this study might be necessary to observe a major barrier breakdown in in vitro cultures, it might not be accurately mimic the pathogenic mechanisms on patients. A sentence in the discussion reflecting the limitations of the model might improve the interpretation of the paper by readers outside the field. The reviewer is correct in pointing out that the conditions might not mimic the in vitro in vivo situation accurately, but the release of parasite molecules at sequestration sites could result in high local concentrations of these parasites factors. Measurements from patient plasma may only partially reflect these local concentrations, but hopefully new 3D models of cytoadherence may be able to resolve this issue. A comment has been added to the discussion to reflect this.
Long term incubation with both IE or IE lysate showed 1) reduced surface expression of EPCR and reduced HBMEC cell index. However, the reduced cell index might not be a direct consequence of loss of EPCR in the surface. As the authors accurately describe in the discussion, loss of barrier in their experiments is more likely a consequence of the presence of parasite histones, HRP2 or other soluble mediators. However, the hypothesis exposed in the introduction might be misleading for the average reader.

"Our hypothesis is that, through both of these mechanisms, the Thrombin-PAR1 coagulation/inflammation axis plays a significant role in barrier loss in CM, and that adjunct treatments targeting this might alleviate mortality or post-CM neurological sequelae in CM. Since the activation of PC may be prevented by either of these mechanisms of EPCR disruption (steric inhibition or receptor cleavage) it is important to identify treatments that could be barrier
Since we don't understand quite stabilising to EPCR abrogation by either mechanism." well the molecular mechanisms that induce loss of endothelial integrity and surface EPCR, the authors should be more careful while establishing associations. The reason to test the PAR1 inhibitors was to test our initial hypothesis that the Thrombin-PAR1 axis plays a significant role in barrier loss in CM. If true, the PAR1 inhibitors should affect the parasite-induced disruption of barrier function. Our results show that this is not the case and the implications of these data are discussed. We have changed the wording at the start of this paragraph to indicate the context of the hypothesis as a starting point. We agree with the reviewer that we do not understand all the mechanisms that contribute to endothelial dysfunction in severe malaria. figure 3, 4, 5 and 6 in a similar way than in figure 2. Or include it in supplementary data. Cell index traces have been added to figure 5 and 6. In figure 5B, the plot shows the effect of Vorapaxar on thrombin induced disruption of barrier function in the absence and presence of lysate. Figure 5A depicts the same traces but over a longer time frame, also showing the effect of lysate. Figure 5A depicts the same traces but over a longer time frame, also showing the effect of Vorapaxar on IE lysate induced decrease in cell index. Figure 6A then shows the effect of all four inhibitors on the lysate induced decrease in cell index. Note that the calculations are also provided in the raw data files "Combined analysis data for fig5_TEER_includes normalisation_2020-02-04.csv" and "Combined analysis data for fig6_TEER_includes normalisation_2020-02-04.csv".

I would recommend to include both 1) representative traces of cell index and bar plots in
Figure 4 already depicts TEER traces, but a panel (4B) has been added on request from both reviewers. It shows that HBMEC exposed to IE lysate recovers slowly from thrombin induced barrier disruption. Figure 3 shows the EPCR expression data, derived from flow cytometry.

Was the experiment shown in figure 4 done on independent days?
Yes, the data shown in figure 4 A, B and C (note that there is an additional panel) are from three different experiments.
Although TEER measurements are highly variable depending on the day of the experiment, box plots and dot plots might better represent data distribution than bar graphs. The bar graphs in figure 5 and 6 are showing mean ± SD, with relatively low SD values. We do not think depicting it as box plots or dot plots will change the interpretation of the data.

Avril
, 2019 showed using the same methodology that schizont-stage parasites et al.  Figure 5 will clarify that. We thank the reviewer for their comment and we observed the same delayed recovery from thrombin treatment after co-culture with IE lysate. We have added a representative cell index trace to show this in figure 5 (A and B) and added text to the results and discussion sections to acknowledge this.

(EPCR binding strains) prolonged thrombin-induced barrier disruption in resting and
A similar cell index trace was also added to figure 4 to show all three observations with HBMEC in one figure: A) Barrier function disruption by thrombin with a fast recovery. This can partly be rescued by the addition of aPC. B) Slow recovery from thrombin treatment after co-culture with IE lysate. C) Disruption of barrier function by IE lysate. "However, in the IE lysate co-culture system we identified a mechanism that slowly reduces barrier function and which is insensitive to PAR1 inhibitors." Please, rephrase to "However, PAR1 inhibitors did not rescued the barrier dysfunction after long-term co-culture with IE lysate". The molecular mechanisms has not been described in this manuscript. The reviewer is correct and we have changed this in the abstract. Figures 3, 4 and 5? The authors give the parasitemia, the hematocrit but as the volume is missing, the quantity or concentration of IE cannot be calculated. This has been added in the methods, it is 300 µl. The number of IE (or equivalent for lysates) has also been added, as requested by the second reviewer.

Which amount of IE is used in the long-term co-culture in
Why did the authors test the rescue after short-term EPCR-PfEMP1 steric inhibition on EA.hy.926 cells (HUVEC) (behavior A) and the long-term parasite effects on HBMEC (behavior B)? Is the rescue seen on Figure 2 also reproducible on HBMEC?
The initial screening experiments were done with a cell line to simplify the approach, and compounds that were identified from this screen were entered into more complex investigations using primary human brain endothelial cells. We recognise that the screen may miss some details but we used this as a cost-effective way to down-select the number of experiments needed. As others have reported on the steric effects of PfEMP1, we did not replicate this work on HBMEC.
A visual timeline of the experiments in each figure might improve the clarity. The experiments performed are quite complex and difficult to interpret and to compare with the current figure legends. A timeline has been added to figure 2, 4, 5 and 6.

Kessler et al., 2017 is not well cited in the first paragraph as a review of aPC-PAR1 axis. Is this a typo?
We had included it to give full coverage but the reviewer is correct and we have removed the reference at this point -it is cited later. The reference to has been added to the results and discussion section, specifically after Avril et al adding TEER traces to figure 4 and 5 describing a similar observation with delayed thrombin recovery after IE lysate exposure.
No competing interests were disclosed. Competing Interests: