Plasmodium falciparum disruption of pericyte angiopoietin-1 secretion contributes to barrier breakdown in a 3D brain microvessel model

Disruption of the vascular protective angiopoietin-Tie axis is common in cerebral malaria (CM) patients, with elevated angiopoietin-2 (Ang-2) and reduced angiopoietin-1 (Ang-1) blood concentrations. The role of pericytes in CM pathogenesis remains unexplored, despite being a major source of brain Ang-1 secretion and evidence of pericyte damage observed in CM postmortem samples. Here we engineered a 3D microfluidics-based microvessel model containing both human primary brain microvascular endothelial cells and pericytes. This model replicated pericyte vessel coverage and ultrastructural interactions present in the brain microvasculature. When exposed to P. falciparum-iRBC egress products, 3D brain microvessels presented decreased Ang-1 secretion, increased vascular permeability, and minor ultrastructural changes in pericyte morphology. Notably, P. falciparum-mediated barrier disruption was partially reversed after pre-treatment with recombinant Ang-1. Our approach suggests a novel mechanistic role of pericytes in CM pathogenesis and highlights the importance of the angiopoietin-Tie axis in vascular dysfunction caused by P. falciparum.


Introduction
Cerebral malaria (CM) is a severe neurological complication of Plasmodium falciparum infection, clinically characterized by coma and fatality rates averaging 15-20% [1] .Furthermore, approximately one third of CM survivors endure lifelong neurological sequelae such as hemiplegia, ataxia, epilepsy or speech disorders [2] .The majority of pediatric CM-related deaths present severe brain swelling due to vasogenic edema, likely as a consequence of blood-brain barrier dysfunction [3]   .A hallmark of CM is the sequestration of P. falciparum-infected red blood cells (iRBC) in the brain microvasculature [4] .While the molecular players responsible for P. falciparum-iRBC ligandreceptor interactions with cerebral endothelial cells have been extensively studied [5][6][7] , the consequences of iRBC sequestration to the brain microvasculature require further investigation.
Nevertheless, evidence suggests that the increase in vascular permeability is a result of a multifaceted process including a blockade of barrier-supportive endothelial receptors by iRBC, the release of toxic P. falciparum products from egressed iRBC, and a dysregulated host innate immune response [8,9] .
An important pathway commonly dysregulated in CM patients is the angiopoietin-Tie axis, crucial to maintaining endothelial pro-barrier, anti-inflammatory, and anti-apoptotic functions [10] .Binding of the vascular ligand angiopoietin-1 (Ang-1) to the endothelial receptor Tie-2 promotes vascular quiescence and the stabilization of endothelial tight junction proteins, such as Occludin and Zona Occludens-1 (ZO-1), necessary for maintaining the barrier function of the brain microvasculature [11-13]   .Conversely, angiopoietin-2 (Ang-2), which is rapidly released from small endothelial storage granules called Weibel-Palade bodies, generally acts as a Tie-2 antagonist leading to endothelial activation and vascular leakage [14] .Given the opposing roles of these two molecules in vascular function, increased levels of Ang-2 and decreased levels of Ang-1 have been largely associated with cerebral vascular pathogenesis in a multitude of diseases [15,16] .Similarly, a decrease in Ang-1, and an increase in Ang-2 and the Ang-2:Ang-1 ratio has been well documented in fatal cases of both pediatric and adult CM [17][18][19][20][21] .Despite its importance, the mechanisms leading to angiopoietin-Tie axis disruption in CM are poorly understood.Human pro-inflammatory cytokines and procoagulation proteins, such as TNF-α and thrombin, have been proposed as responsible for the increase of Ang-2 secretion by endothelial cells [22,23] , but the causative agent responsible for decreased secretion of Ang-1 in CM patients remains unknown.

Vascular mural cells, including smooth muscle cells and pericytes are the main cellular sources for
Ang-1 secretion in the brain microvasculature [10] .Beyond Ang-1 secretion, pericytes support the microvasculature and promote quiescence through additional roles.Structurally, pericytes wrap around cerebral microvessels and interact with the endothelium through reciprocal ultrastructural membrane protrusions, known as peg-and-socket junctions [24] , and secrete extracellular matrix components of the microvascular basal lamina that strengthen microvascular architecture.
Functionally, pericytes regulate brain blood flow by controlling vessel diameter through constriction [25] .Cerebral pericytes have been implicated in various vascular disorders of the central nervous system.Loss of pericytes from the cerebral vasculature in ischemic stroke and Alzheimer's disease is associated with blood-brain barrier disruption, extravasation of blood components into the brain parenchyma, and neuronal loss [26,27] .Similarly, histopathological studies in retinal samples of fatal pediatric CM cases have revealed pericyte damage in regions of P. falciparum-iRBC sequestration, suggesting a similar role of these cell types in severe malaria infections [28] .
Yet, the role of cerebral pericytes in CM pathogenesis and in the dysregulation of the angiopoietin-Tie pathway remains largely unknown.
Bioengineered vascular models are emerging as powerful tools to study disease mechanisms in vitro.They have been successfully used to recapitulate many aspects of in vivo vascular disease including dysfunction of the angiopoietin-Tie pathway [29] , and more recently, malaria pathogenesis, by modeling ligand-receptor interaction of iRBC in microvessels [30] or transcriptional changes on endothelial cells upon exposure to iRBC [31] .Here, we have further developed our existent 3D brain microvascular model [30] , with the addition of pericytes to create a device that recapitulates in vivo interactions between both cell types.We have used this advanced model to understand the molecular basis of angiopoietin-Tie dysregulation in CM.Our results indicate that iRBC and products released during parasite blood stage egress play a role in angiopoietin-Tie dysregulation, and propose brain pericytes as a new player in CM pathogenesis.

In vitro 3D brain microvessel model recapitulates in vivo cerebral endothelial-pericyte interactions
To investigate the role of cerebral pericytes in P. falciparum-mediated vascular pathogenesis, we have generated a 3D brain microvessel model that recapitulates the degree of pericyte coverage of the endothelium present in the brain microvasculature.The device is fabricated in a collagen type I hydrogel containing a microfluidic network with a pre-defined geometry, which is connected to both an inlet and outlet, enabling vessel perfusion [32] .Previous incorporation of pericytes in similar bioengineered 3D models involved cellular addition into the collagen hydrogel itself.However, this resulted in sparse interaction of pericytes with the endothelial microvessels [32] .Indeed, cerebral pericytes have been shown to cover approximately one third of the brain microvasculature, albeit with slight variations along the brain vascular hierarchical network [33,34] .In this model, primary human brain vascular pericytes (HBVP) were seeded directly into the channels along with primary human brain microvascular endothelial cells (HBMEC) at a 5:1 endothelial to pericyte ratio (Figure 1A).After 3 days in culture, the two cell types appear to reorganize into two different layers with luminal endothelial cells and mCherry-expressing pericytes sitting on the 100 μm endothelial channels (Figure 1B).Pericytes distributed homogenously around the vessel cross-section, with approximately a third of the microvessel surface area covered by pericytes (29.6% of the bottom surface and 27.3% of the top surface) (Figure 1C).Notably, endothelial VE-cadherin junctional labeling is continuous along the vessel luminal surface, suggesting that the presence of pericytes does not disrupt the endothelial layer (Figure 1D).Furthermore, endothelial cells in the microvessel network express other common endothelial markers, such as von Willebrand factor, in addition to adherens and tight junction markers, β-catenin and ZO-1 respectively (Figure 1E, Figure S1A, Supporting information).Labeling by the mural cell marker, α-smooth muscle actin (αSMA), shows that pericytes appear as elongated cells with thin processes stretching over multiple endothelial cells (Figure 1D).Pericyte cellular identity was confirmed by the expression of a combination of cerebral pericyte cellular markers including αSMA (Figure 1D), platelet derived growth factor receptor β (PDGFRβ) and nerve/glial antigen 2 (NG-2) (Figure 1E, Figure S1B, Supporting information) [35,36] .
PDGFRβ and NG-2 labeling of cerebral pericytes colocalized with laminin and collagen IV staining, suggesting that pericytes contribute to the secretion of extracellular matrix components that compose the basal lamina in vivo (Figure 1E) [37] .Volumetric electron microscopy has been previously used to characterize in vivo 3D ultrastructural interactions between pericytes and endothelial cells of the cerebral microvasculature [24] .To characterize the spatial and ultrastructural organization of pericytes and endothelial cells in the model, we utilized Serial Block-Face Scanning Electron Microscopy (SBF-SEM) [38] , a volume electron microscopy technique that provides nanoscale resolution of samples in a large field of view [39]   .We imaged a 100 μm segment of a microvessel branch with a 15x15x50 nm xyz resolution, followed by cellular segmentation and rendering.We characterized microvessel coverage of approximately 25 individual pericytes, distinguished from endothelial cells by the presence of a highly granulated cytoplasm [40] .Cross-sectional analysis and segmentation of the microvessel revealed an ovoid microvessel morphology, and confirmed the presence of pericytes at the abluminal side of the endothelial cells (Figure 2A).Pericytes acquired a classical bump-on-a-log morphology, with the soma being the thickest region of the cell, and their longitudinal axis aligned with the direction of flow (Figure 2B).Most of the endothelial-pericyte interface is composed of thin pericytic lamellae (100-300 nm thickness) that cover large areas of the endothelial surface, with multiple thin pericytic branches appearing mostly at the cell periphery, as described previously [24]   .A close up of regions of pericyte-endothelial cell interaction reveal that the membranes of both cell types are in close proximity, even though peg-and-socket junctions were not found (Figure 2C).
Pericytes constitute a heterogenous cell population depending on their location within the vascular branch.Considering that our model possesses 100 μm diameter microvessels the co-expression of αSMA, PDGFRβ and NG-2 staining, and the bump-on-a-log morphology(Figure 1D,E) is reminiscent of ensheathing pericytes, commonly found on cerebral arterioles rather than capillary pericytes that do not express αSMA [41,42] .Overall, the 3D in vitro brain microvessel model recapitulates in vivo interactions between cerebral endothelial cells and pericytes, both at the tissue and ultrastructural level, that can next be utilized to model the role of pericytes in malaria pathogenesis.

P. falciparum egress products increase vessel permeability but only cause minor changes in pericyte morphology
Multiple in vitro studies have shown that P. falciparum products released during malaria parasite egress disrupt the endothelial barrier [43][44][45][46][47] .To characterize the impact that these could have on pericyte function, we established a protocol to create media enriched for P. falciparum-iRBC egressed products, denoted from here on as iRBC-egress media.Briefly, tightly synchronized schizont stage-iRBCs were purified and allowed to egress in vascular growth media, followed by recovery of the supernatant fraction containing P. falciparum soluble products at an estimated concentration of 5x10 7 ruptured iRBC/mL (equivalent to simultaneous egress of approximately a 1% parasitemia) (Figure S2A,B, Supporting information).We measured the endothelial disruptive properties of iRBC-egress media by the xCELLigence system, which provides real time measurement of cell barrier integrity through impedance.The P. falciparum egress products caused a significant, dose-dependent decrease in cell index, a measurement of monolayer barrier function normalized to untreated cells, that reached a maximum disruption at 18 hours (Figure S2C, Supporting information).To elucidate if iRBC-egress media also induces barrier breakdown in the brain 3D microvessel model, we performed permeability assays in a simpler microvessel network which consists of a single 200 μm channel.After perfusion and incubation with iRBC-egress media for 18 hours, P. falciparum egress products accumulated on the lateral and bottom walls of the microvessel and appeared as dark spots in brightfield that were DAPI-positive by confocal microscopy, suggesting the presence of hemozoin, food vacuoles and parasite DNA remnants (Figure 3A,B).Microvessels perfused with iRBC-egress media presented a significant increase to 70 kDa FITC-dextran permeability across the entire vessel wall, corresponding to a 6-fold increment compared to vessels perfused with vascular cell media (4.08x10 -6 cm/s in media-only vs. 2.66x10 -5 cm/s in devices exposed to iRBC-egress media) (Figure 3B,C).
To determine whether increased permeability was associated with physical changes in pericytes, we assessed their cellular morphology in 3D microvessels exposed to iRBC-egress media for 18hours by SBF-SEM.Pericytes maintained close spatial proximity to the endothelium in the presence of P. falciparum-iRBC ghosts, similar to control devices exposed to vascular growth media (Figure 3D).To investigate potential changes in pericyte shape, we isolated and analyzed each 3D rendered pericyte (Figure 3E,F, Figure S3, Supporting information).Cellular features such as circularity and maximum width did not differ between conditions, however, iRBC-egress media treated pericytes presented fewer branches suggesting a response to iRBC-egress media (Figure 3E,F).Altogether, our results show that P. falciparum egress products cause an increase in permeability of the 3D brain microvessel model, accompanied by minor changes in pericyte morphology.
To validate whether changes in the angiopoietin-Tie axis occur physiologically following P. falciparum blood stage egress within the microvasculature, we perfused the 3D brain microvessels with schizont stage P. falciparum-iRBC for 30 minutes followed by a 10-minute wash.After an 18hour incubation period, supernatants were again recovered and analyzed by Luminex.Similar results were obtained, with a reduction in both Ang-1 and Ang-2 secretion and an increase in the Ang-2:Ang-1 ratio (Figure 4B).Altogether, these results suggest that P. falciparum-iRBC hamper Ang-1 secretion by pericytes, revealing a previously unappreciated role of pericytes in the dysregulation of the angiopoietin-Tie axis in CM. (rAng-1) to resting endothelium did not alter barrier properties.Conversely, an 18-hour pretreatment of rAng-1 prior endothelial cell exposure to iRBC-egress partially protected against parasite-mediated endothelial barrier disruption (Figure 5A-C).While both iRBC-egress media conditions follow parallel endothelial breakdown trajectories, rAng-1 addition dampened the maximum levels of disruption by iRBC-egress media, and was followed by a faster recovery to endothelial monolayer baseline levels (Figure 5B).Indeed, pre-treatment with rAng-1 conferred a 40% barrier breakdown protection after normalization to the untreated endothelial condition (Figure 5C).Nevertheless, shorter pre-treatment periods with rAng-1 (1-hour) did not protect against iRBCegress media disruption, suggesting that the establishment of endothelial protective pathways is not immediate (Figure S5, Supporting information).
Next, we tested the protective capacity of rAng-1 against iRBC-egress products in the 3D microvessel model (Figure 5D).While 3D microvessels presented a significant increase in permeability values to 70 kDa FITC-dextran upon exposure to iRBC-egress media, a much smaller non-significant increase was found when microvessels were pre-treated with rAng-1 for 18-hours (Figure 5E,F).Indeed, Ang-1 pre-treatment presented a 45% protection when median permeability values were compared (2.66x10 -5 cm/s in iRBC-egress media vs. 1.45x10 -5 cm/s iRBC-egress media with rAng-1 pre-treatment), similar to that seen in 2D endothelial culture.Taken together, pre-treatment with rAng-1 partially protects against microvascular permeability caused by P. falciparum egress products, both in 2D and 3D, highlighting the importance that a decrease in Ang-1 could have in CM pathogenesis.

Discussion
Disruption of the angiopoietin-Tie axis is a common feature in cerebral malaria patients, often associated with disease severity and fatality [18,20] .This dysregulation is also found in pediatric and adult severe malaria patients without cerebral complications [17,19] or in placental malaria [49] .
Central to this pathway are cerebral pericytes, an Ang-1 secreting cell type that is located at the interface between the vasculature and brain parenchyma, yet largely understudied in the context of cerebral malaria.Here, we developed a 3D brain microvessel microfluidic model that recapitulates the high pericyte coverage present in the brain microvasculature.Exposure of the 3D brain microvessels to P. falciparum-egress products caused an increase in vascular permeability accompanied by a decrease in Ang-1.Furthermore, we showed that pre-treatment with rAng-1 could, at least, partially protect against endothelial barrier breakdown.Taken together, these results suggest that the cessation of Ang-1 release upon exposure to iRBC-egress products further contributes to microvessel disruption.
The bioengineered 3D model provides important vascular mechanical cues, including estimated flow and extracellular matrix stiffnesses present in the human brain.We could reproduce the high pericyte coverage found in brain capillaries by co-seeding primary cerebral endothelial cells and pericytes through the microfluidic network.Simultaneous cell seeding, at a ratio of 5:1 (endothelial cells: pericytes), did not prevent the formation of an intact luminal endothelial monolayer that presents continuous labeling of adherens and tight junctional markers.SBF-SEM microscopy showed that both cell types sit at nanoscale distance, with the presence of thin lamellae extending from the pericyte soma covering large areas of the endothelium.Yet, despite close proximity, the recently well-described reciprocal peg-and-socket junctions were absent in the model [24] .Overall, the 3D brain microvessel model recapitulates pericyte-endothelial interactions found in human brain vasculature making it suitable to study the role of pericytes in CM pathogenesis.
P. falciparum products released during blood stage egress have been shown to possess barrier breakdown capabilities on 2D endothelial monolayers grown on plastic.Proposed barrier disruptive P. falciparum-iRBC egress products include merozoite proteins, histones, heme, hemozoin and P. falciparum histidine rich protein 2 [22,[43][44][45][46][47]50] . Howeer, their effects on pericytes remain unexplored.In this study, we confirmed that these products are disruptive in a microfluidics-based 3D brain microvessel model in the presence of pericytes.Pericytes play an important role in maintaining vascular homeostasis.For example, loss of pericytes from the cerebral vasculature leads to increased permeability in mouse models [51] .Furthermore, dysregulation of cerebral pericytes and detachment from the vasculature have been observed in other cerebral vasculature pathologies including stroke and Alzheimer's disease [26,27] .In our study, we did not observe gross differences on pericyte morphology or loss of vascular coverage upon treatment with iRBC-egress media.Nevertheless, pericytes from iRBC-egress media-treated 3D microvessels did display subtle morphological changes, having significantly less branches.The relevance of these minor ultrastructural modifications in vascular barrier dysfunction could imply molecular mechanisms of disease.For example, Luminex quantification upon stimulation with iRBC-egress media in the 3D brain microvessel model revealed a significant reduction in N-cadherin shedding.This could point towards reduced functional endothelial-pericyte interaction upon short-term incubation with P. falciparum-egress products.Future studies should explore whether exposure to P. falciparum for longer timepoints recapitulates phenotypes observed in post-mortem samples of cerebral malaria patients such as vacuolization [52] or pericyte loss [28] .Nevertheless, our results point towards a loss of integrity at the endothelial and pericyte interface.
Ang-1 and Ang-2 have been considered in the past as potential biomarkers of cerebral malaria [53] .
Ang-1 has excellent predictive power to distinguish malaria severity scores [19] or malaria from other central nervous system febrile diseases [54] .Nevertheless, the underlying mechanism of Ang-1 decrease in CM remains unknown, as previous in vitro studies overlooked the effect that iRBCs could exert on brain pericytes.To our knowledge, our study is the first one to report that P. falciparum products released upon iRBC egress halt Ang-1 secretion by pericytes.A complementary hypothesis for Ang-1 decrease in CM patients could be that severe thrombocytopenia, commonly associated with cerebral complications in P. falciparum infections [7]   , plays an additional role as platelets are another major source of serum Ang-1 [55] .In light of this, it is likely that both the effect of P. falciparum-iRBC on perivascular cells and thrombocytopenia could work in concert to further contribute to a widespread decrease of Ang-1 levels.As for the parasite component responsible for the cessation of Ang-1 secretion, the presence of accumulated light scattering, DAPI-positive iRBC-egress material suggest a role of parasitic DNA or hemozoin, yet further studies are necessary to identify which iRBC-egress derived component is responsible for altered signaling pathways in pericytes.Nevertheless, our results highlight brain pericytes as a new player in the development of CM through dysregulation in the secretion of Ang-1.
The use of recombinant Ang-1 as an adjunctive therapeutic against CM has long been of interest [48] because of its role in strengthening endothelial tight junctions, as well as, upregulating antiapoptotic and anti-inflammatory pathways [56] .Interestingly, we showed that pre-incubation of human microvessels with recombinant Ang-1 partially protected against iRBC-egress mediated barrier breakdown.Nevertheless, the failure to achieve complete protection upon rAng1 supplementation hints at the existence of other pathogenic mechanisms involved in cerebral malaria-induced barrier breakdown.While the concentration of recombinant Ang-1 utilized in this study was supraphysiological compared to peripheral blood samples, it is quite likely that peripheral blood concentrations underestimate local concentrations at the endothelial-pericyte interface.In our model, protection was only achieved after a pre-incubation of 18-hours, suggesting that downstream signaling through the Tie2 receptor, such as PI3K/Akt signaling, is required [10] .Future interventions targeting the angiopoietin-Tie axis should focus on approaches that reverse microvascular dysfunction after exposure to P. falciparum.As a proof of concept, recombinant Ang-1 has previously been used as a treatment in a rodent experimental cerebral malaria model after onset of disease, preserving blood-brain barrier integrity, and leading to a significantly increased survival rate [48] .Although we did not observe complete protection after short term exposure, our findings point towards a mechanistic role of Ang-1 in protecting against iRBC egressmediated endothelial barrier breakdown, further highlighting Ang-1 alteration as a functional contributor to CM pathogenesis.
Clinical studies have shown a lack of correlation between Ang-2 levels and P. falciparum-iRBC microvascular obstruction in the rectal microvasculature [57] or iRBC sequestration in brain postmortem samples [58] .Similarly, our study confirms that P. falciparum-iRBC sequestration in 3D brain microvessel models does not directly cause Ang-2 secretion by endothelial cells.This result is in concordance with a recent study that showed that Ang-2 is not secreted in response to P. falciparum lysate products, but rather occurs following endothelial activation with TNF [22] .Other studies have shown Ang-2 release as a result of exposure to thrombin or hypoxia [23,59] , conditions that have also been associated with CM pathogenesis.Altogether, our study is in agreement with previous reports suggesting that P. falciparum sequestration does not play a direct role in the secretion of Ang-2, which instead is more likely a consequence of the activation of host proinflammatory factors.
To note, our study has several limitations, including the low-throughput of the model or challenges in growing the devices for long periods of time, hence preventing the modelling of long-term disruptive mechanisms, such as pericyte loss.Furthermore, our devices currently represent larger microvessels of the brain given the 100 μm diameter of our model and α-SMA expression on pericytes.Therefore, future studies could exploit 2-photon laser ablation or bioengineered selfassembled models [60,61] to generate brain capillary-sized microvessels (5-10 μm), a major pathogenic site in CM.Nevertheless, this study establishes a key role of pericytes on microvessels that are similar in size to arterioles or post-capillary venules, regions of the brain vascular tree that are equally damaged in CM patients [28] .Finally, an additional limitation of our model is the absence of other cells that could play an important role in maintaining the angiopoietin-Tie axis.For example, Ang-1 secretion has been described in other brain and blood cell types, such as smooth muscle cells [62,63] , platelets and more recently, neurons and astrocytes [64] .Yet, one of the major advantages of in vitro bioengineered models is the opportunity to sequentially introduce different components that, independently or collectively, could play a role in a complex and multifaceted disease such as CM.In light of this, subsequent iterations of the model could introduce other cell types that produce Ang-1 or incorporate pro-inflammatory cytokines or thrombin, to concurrently model the functional consequence of increased secretion of Ang-2 by endothelial cells.
In summary, our study highlights the role that pericytes could have in the development of CM pathogenesis by showing for the first time that egress of P. falciparum-iRBC interrupt Ang-1 secretion by pericytes.In addition, our findings confirm that external addition of Ang-1 partially protects against the increase in vascular permeability mediated by P. falciparum egress products, providing further evidence that interventions involving the angiopoietin-Tie axis could be part of a multi-targeted adjunctive therapy against CM.
Parasites were grown in sealed top T75 flasks in a gas mixture of 90% N2, 5% CO2 and 1% O2.To maintain synchronous cultures, parasites were synchronized twice a week in their ring stage with 5% sorbitol and once a week in their trophozoite stage with 40 mg/mL gelaspan (Braun).

P. falciparum-iRBC egress media preparation:
Late-stage P. falciparum-iRBC (schizonts, parasitemia of 5-10%), synchronized in a 6-hour window, were purified by use of 40 mg/mL gelaspan gradient separation to a final purity of >60% parasitemia.The enriched P. falciparum-iRBC were then placed in complete RPMI media containing 1 μM compound 2, a reversible PKG inhibitor that inhibits iRBC egress (kindly donated by Michael Blackman, The Francis Crick Institute), at a concentration of 50 million P. falciparum-iRBC/mL.After 5 hours, compound 2 containing media was removed and the parasites were resuspended at a concentration of 100 million P. falciparum-iRBC/mL in vascular growth media, put into a sealed top T25 flask, gassed and left in the incubator overnight on a shaker set to 50 rpm to facilitate parasite egress.The resulting parasite egress efficiency was assessed by a hemocytometer count and blood smear, and was then concentrated to a working concentration of 50 million ruptured P. falciparum-iRBC/mL.The resulting iRBC-egress media is spun at 1000 rpm to remove cellular debris and flash frozen in liquid nitrogen until used.This concentration was chosen as this value is equivalent to a circulating parasitemia of 1%, concentrations frequently found in malaria patients.
Primary human brain microvascular endothelial cell and primary human brain vascular pericyte culture: Primary HBMEC (Cell Systems; ACBRI 376) were cultured according to the manufacturer's recommendations in complete endothelial growth media-2MV (Lonza) containing 5% fetal bovine serum.Cells were split using Trypsin/EDTA.When 90% confluent, HBMEC were seeded on 15 μg/mL Poly-L-lysine (P8920, Sigma) coated T75 flasks.Primary HBVP (ScienCell) were cultured according to the manufacturer's recommendations in basal pericyte media supplemented with 1% pericyte growth supplement, 1% penicillin/streptomycin and 2% fetal bovine serum (ScienCell).Cells were split using Trypsin/EDTA when 90% confluent, seeded on 15 μg/mL Poly-L-lysine coated T75 flasks, and maintained in a humidified incubator at 37 °C and 5% CO2.All experiments were conducted with HBMEC and HBVP with a passage number of 7-9.mCherry lentivirus transduction: HBVPs were grown in a T75 flask to a ~90% confluency and incubated with lentiviral particles containing a mCherry vector (kindly donated by Kristina Haase Lab, EMBL Barcelona) in serum free pericyte media at a multiplicity of infection of 10.After 24 hours, the lentivirus particles were removed by washing once every 24-hours for a 2-day period.mCherry-positive cells were then selected by fluorescence-activated cell sorting, expanded and froze down for future experiments.
3D brain microvessel model fabrication: Type 1 collagen was extracted from rat tails, dissolved in 0.1% acetic acid, lyophilized in a freezer dryer (Labconco Freezone 2.5 Plus) and resuspended at 15 mg/mL in 0.1% acetic acid for storage.The stock collagen solution-0.1% acetic acid solution was neutralized and diluted to 7.5 mg/mL on ice.A 13 by 13 channel grid or single channel pattern were negatively imprinted into the collagen hydrogel by soft lithography using a PDMS stamp, and inlet and outlets were created by insertion of stainless-steel dowel pins.The top pre-patterned collagen hydrogel contained in a plexiglass top jig is sealed to a flat collagen-layered bottom that sits on a coverslip and a plexiglass bottom jig.The assembly creates perfusable 120 μm diameter microvessels (grid design) or 200 μm diameter microvessels (single channel design), as described previously [32,65] .Prior to cell seeding, the channels were incubated with vascular growth media, consisting in EGM-2MV (Lonza) supplemented with 1x pericyte and 1x astrocyte growth factors of endothelial surface that was covered by pericytes was calculated by creating a mask of both the mCherry HBVP and the vWF-positive HBMEC.
Serial Block-Face Scanning Electron Microscopy: Microvessel devices were grown for 3 days, then treated with iRBC-egress media or vascular growth media for 18hours, and then fixed by adding 2% PFA and 2.5% glutaraldehyde (GA) to the device inlet for 30-minutes and washed twice with EGM-2MV at 37 °C.The collagen hydrogel was then carefully removed from the plexiglass jig and microvessel regions exposed to low shear stress were cut out and fixed with a secondary fixative solution (2% PFA, 2.5% GA, 0.25 mM CaCl2, 0.5 mM MgCl, 5% sucrose in a pH 7.4 0.1 M Cacodylate buffer) overnight at 4 °C, and then rinsed twice for 15-minutes with 0.1 M Cacodylate buffer.Samples were post-fixed in a reduced Osmium solution (1% OsO4, 1.5% K3FeCN6 in 0.065 M Cacodylate buffer) for 2-hours at 4 °C followed by six 10-minute washes in dH2O.Post-staining consisted of subsequent incubation steps of 1% thiocarbohydrazide in dH2O, 2% Osmium tetroxide in dH2O, and 1% Uranyl acetate in dH2O, aided by a PELCO Biowave Pro+ (Ted Pella) containing a SteadyTemp Pro and ColdSpot set to 20 °C at 7x 2 minute cycling on-off at 100 W under vacuum, with dH2O rinses in between steps once in a fume hood and twice in the microwave at 250 W for 40 s without vacuum.The stained samples were then dehydrated by serial additions of 30, 50, 80, and 100% ethanol solutions in the microwave at 250 W for 40 s without vacuum with ColdSpot set to 4 °C, infiltrated with EPON 812 hard epoxy resin in steps of 25, 50, 75, 90, and 3x 100% resin diluted in ethanol with each step in the microwave at 150 W for 3 minutes under vacuum and polymerized at 60 °C for 48 hours.The sample was trimmed (UC7, Leica Microsystems) using a 90° cryo-trimmer (Diatome) to generate a small block face.The resulting resin block was mounted on a pin stub using silver conductive epoxy resin (Ted Pella).The SBF-SEM acquisition was performed with a Zeiss Gemini2 equipped with a Gatan 3view microtome and a focal charge compensation device (Zeiss).The SEM was operated at 1.5 kV 300 pA, using a pixel size of 15 nm, a slice thickness of 50 nm and a dwell time of 1.6 μs.The SBF-SEM acquisition was performed using the software SBEMimage [66] .The acellular center of the lumen was not imaged to save imaging time.Following acquisition, the positional metadata of the sections and tiles obtained was converted with SBEM-specific import scripts (https://git.embl.de/schorb/rendermodules-addons)for subsequent assembly, stitching and alignment of large EM datasets [67] .Aligned images were then binned to 22 by 22 by 250 nm 3 in xyz and segmentation was then performed using the software package Amira (Thermo Fischer Scientific).Primary HBVP and HBMEC were segmented semi manually using the magic wand and brush tool to segment every fifth section followed by use of the interpolation tool to segment the entire volume.The resulting segmented microvessels were rendered using the generate surfaces module for 3D visualization.

P. falciparum-iRBC and iRBC-egress media perfusion to 3D brain microvessels: All experiments
were done in 3D microvessel devices grown for 3 days.Schizont-stage P. falciparum-iRBC were purified to >60% parasitemia by 40 mg/mL gelaspan and the resulting enriched population was diluted in vascular growth media to 50 million iRBC/mL (same concentration as the P. falciparum-iRBC-egress media).200 μL of P. falciparum-iRBC or P. falciparum-iRBC-egress media was added into the device inlet and perfused by gravity flow for 30-minutes with the outlet effluent being reintroduced to the inlet every 10 minutes.Microvessels treated with intact P. falciparum-iRBC (but not P. falciparum-iRBC-egress media), were then washed for 10 minutes with vascular growth media.Both conditions were incubated overnight for 18 hours.Supernatant was then removed from the outlet and pooled for Luminex analysis, and 3D brain microvessels were used for permeability studies or fixed for imaging with 3.7% PFA for 15-minutes (see section immunofluorescence microscopy).
Fluorescent dextran-based permeability assay: Single channel microvessel devices were grown for 3 days, then perfused and incubated for 18 hours with either vascular growth media or P.
falciparum-iRBC-egress media as described above.To determine resulting changes in permeability, they were perfused with a 70 kDa FITC-dextran solution as follows: the device was washed 1x with PBS and placed in a Zeiss LSM 980 Airyscan2 confocal microscope with a temperature and CO2 controlled imaging chamber (37 °C, 5% CO2).PBS was aspirated and 100 μL of 70 kDa FITC-dextran solution (100 μg/mL) in PBS was pulled from the inlet to the outlet at 6.5 μL/min flow rate using a syringe pump (Harvard Apparatus PHD 2000).Tilescan confocal images of the entire channel were taken every 30 seconds for 5 minutes once the channel had been filled with FITC dextran.In devices pre-treated with recombinant Ang-1 (923-AB-025, R&D Systems), pre-incubation occurred at 18-hours, starting on day 2.
Quantification of apparent permeability: Apparent permeability is determined as the flux of fluorescently labeled dextran across the microvessel wall into the surrounding collagen hydrogel.
All steps of permeability analysis were performed using ImageJ.By using the following equation: The apparent permeability (Papp) was calculated using the fluorescence intensity inside the vessel (IV) and in the surrounding collagen hydrogel (IT) at two time points t1 (0 minutes after the vessel is filled with dextran) and t2 (5 minutes after the vessel is filled).Δt is the change in time, Alateral tissue is the area of the collagen hydrogel being examined and P is the perimeter of the microvessel.
To account for P. falciparum-iRBC-egress material that appears dark even in the presence of 70 kDa FITC-dextran, regions of P. falciparum-iRBC-egress material were excluded from the calculation of the fluorescence intensity inside the vessel (IV) by using ImageJ median smoothing and the magic wand tool to select only the area of the channel without P. falciparum-iRBC-egress material.
Serial Block-Face Scanning Electron Microscopy pericyte morphology analysis: 3D-rendered pericyte meshes, segmented from SBF-SEM images of microvessels treated with either media alone or iRBC-egress media, were generated using Amira.Following segmentation, the meshes were exported as STL files and processed with a custom-developed Python script, designed to quantify aspects of pericyte morphology, including mesh circularity and maximum width [68] .Firstly, the 3D pericyte meshes were mapped onto a 2D plane through principal component analysis (PCA), effectively reducing the data's dimensionality while retaining its most significant morphological features.This mapping uses the two principal components that encapsulate the dimensions with the maximum variance in the 3D mesh, thus ensuring the conservation of structural information.
Subsequently, convex hull analysis [69] was applied to delineate the pericyte mesh boundaries, providing a clear representation of each structure's shape by which circularity and maximum width could be measured.The circularity of each mesh is calculated using the convex hull, by assessing the ratio of the area enclosed by the hull to the square of the hull perimeter, with the formula for circularity given as 4 × / 2 .This circularity metric provides insight into the roundness of the mesh, where a value closer to 1 indicates a shape that is more circular, and values closer to 0 reflect more elongated or irregular shapes.Additionally, the maximum width is defined by identifying the two farthest points along the convex hull boundary and then calculating the greatest perpendicular distance from this line (connecting the farthest points) to any point within the convex hull.This measurement represents the maximum width of the pericyte mesh by assuming its length is the axis with the greater size.In addition, the number of pericyte branches was calculated manually by counting the number of branches on each 3D rendered pericyte.

Figure 1 .
Figure 1.A) Schematic depiction of the fabrication pieces and device set-up (top-left), including a representation of resultant endothelial-pericyte interactions (top-right).Schematic depiction of the seeding method used to generate the 3D microvessel model (bottom).B) Immunofluorescence assay (IFA) maximum z-projection of the full network of cellularized channels labeled with Von-Willebrand factor for HBMEC (red), mCherry expressing HBVP (white) and DAPI (blue) (left).Inset highlighting pericyte coverage of the microvessels (right).Scale bars represent 500 μm and 100 μm (Inset).C) Quantification of the percentage of endothelial microvessel area covered by pericytes on either the top or bottom microvessel surface in a z cross-sectional view.Red dashed line represents estimated in vivo brain microvascular pericyte coverage.Data points represent quantification of 17 devices with statistical analysis determined by Mann-Whitney U test.Error bars represent mean +/-standard deviation.D) IFA maximum z-projection of VE-cadherin, αSMA, and DAPI labeling on the top and bottom cross-sectional surfaces of microvessels (left).Merge maximum z-projections with VE-cadherin (red), αSMA (white) and DAPI (blue) (right).E) IFA maximum z-projection of brain endothelial markers, ZO-1 and β-catenin, as well as, colocalization between brain pericyte markers, NG-2 and PDGFRβ (red), and extracellular matrix markers,

Figure 2 .
Figure 2. A) A cross-section of the microvessel imaged through the SBF-SEM volume (top) and the corresponding segmentation outlines of the endothelium (grey) and surrounding pericytes (coloured).Scale bars represent 20 μm.B) 3D rendering of the segmented microvessel (grey) and the surrounding pericytes viewed from above or below.Scale bars represent 10 μm.C) Zoomed images of the region of interest (ROI) in A. Insets show ROIs i and ii where pericytes are denoted with "P", endothelial cells by "EC" and pericyte lamellae by "P*".Scale bars represent 5 μm (left) and 500 nm in the insets (center and right).

Figure 3 .
Figure 3. A) Fluorescence confocal imaging of phalloidin (white) and DAPI (blue) of single channel 3D brain microvessels treated with either media only or P. falciparum-iRBC-egress media for 18-hours.Insets show ROIs of i and ii that highlight endothelial junctions and pericyte coverage

Figure 4 .
Figure 4. Ang-1 and Ang-2 concentrations measured from 3D brain microvessels treated with media only, iRBC-egress media A) or P. falciparum-iRBC B) for 18-hours (left and center).Ang-2:Ang-1 ratio in supernatants collected from 3D brain microvessels.Box and whisker plots display the median, 25th and 75th percentiles and the minimum and maximum data points.Data points represent supernatants pooled from 2-3 devices in 6-7 independent experiments and were analyzed by Mann-Whitney U test.Data points in which Ang-1 concentrations measured from 3D brain microvessels were below the limit of detection were removed to calculate the Ang-2:Ang-1 ratio.

Figure 5 .
Figure 5. A) Experimental outline of the rAng-1 pre-treatment permeability experiment measured in 2D HBMEC monolayers in the absence of HBVP.B) Representative recording data of xCELLigence measurements taken after an 18-hour +/-rAng-1 pre-treatment followed by +/-iRBC-egress media addition.All conditions were normalized to the media only control.C) Area under the curve analysis of 4 independent experiments run in triplicate with iRBC-egress media normalized as 100.Error bars represent mean +/-standard deviation.Statistical significance is measured by repeated measures one-way ANOVA test with Dunnett's multiple comparisons test.D) Experimental outline of the rAng-1 pre-treatment permeability experiment on 3D brain microvessels.E) Confocal image of the 70 kDa FITC-dextran flux into the surrounding collagen at 0 and 5 minutes after dextran perfusion.Images correspond to a representative rAng-1 pretreated microvessel exposed to either media only or iRBC egress media for 18 hours.Scale bars represent 200 μm.F) Apparent permeability of 70 kDa FITC-dextran in microvessels pretreated with or without rAng-1 followed by 18-hour media only or iRBC egress media treatment in 8-10 individual devices per condition.Box and whisker plots display the median, 25th and 75th percentiles and the minimum and maximum data points.Statistical significance was analyzed by Kruskal-Wallis test with Dunn's multiple comparisons test.

10 5
HBVPs were seeded on the basolateral side of Poly-L-lysine treated transwell inserts (12 well PET membrane inserts with a pore size of 3 μm) in 100 μL of media and left to attach for 4-6 hours.Transwell inserts were then flipped into a 12 well culture plate containing 1200 μL of pericyte media and allowed to grow for 2 days.Next, 10 5 HBMECs were seeded onto the apical side of the Poly-L-lysine treated transwell inserts in 300 μL of media and left to attach for 4-6 hours.Then media was removed from both sides of the insert and 1200 μL of pericyte media was added to the basolateral side and 800 μL of EGM-2MV media to the apical side.The cells were allowed to grow for 2 more days.On day 5, supernatants were taken from the basolateral side for angiopoietin-1 ELISA.Endothelial cell-only transwells underwent the same fabrication protocol minus the initial pericyte steps.Angiopoietin-1 ELISA: Ang-1 protein concentrations were measured from supernatants taken from the basolateral side of the transwell model.Supernatant concentrations were obtained by running 5X diluted samples on either the Human Angiopoietin-1 DuoSet ELISA kit (DY923, R&D Systems).Protein concentrations were interpolated from sigmoidal 4-parameter-fit standard curves generated from a 7-point standard curve of recombinant human proteins using GraphPad Prism.Quantification of secreted proteins by Luminex Assay: Secreted protein concentrations were measured from supernatants taken from the outlets of 3D brain microvessels with a 13 x 13 grid geometry.Supernatants from 2-3 devices exposed to the same condition were pooled, diluted 2X and assayed using a 10-plex Human Luminex Discovery Assay from R&D Systems on a Luminex 100/200.The 10-plex panel included: Ang-1, Ang-2, PDGF-BB, N-cadherin, TIMP-1, ANGPTL4, VEGF, Tie-2, IL-8, and CXCL-1.Protein concentrations were interpolated from a 11-point standard curve of known concentrations of recombinant human proteins provided by the vendor and reported as pg/mL using xPONENT 4.2.7 out of 10 analytes were in the standard curve quantification range for 100% of measurements recorded, with Timp-1, N-cadherin and IL-8 having 100% of measurements above the upper limit of quantification.Measuring changes in barrier integrity by xCELLigence: 96 well PET E-plates (300600910,Agilent) were coated with 15 μg/mL Poly-L-lysine and 5000 HBMECs in 200 μL EGM-2MV were added per well.After cell adherence, the plate was placed into the xCELLigence RTCA SP reader to begin baseline measurement of growth-related changes in cell index (an arbitrary measure of impedance) and the cells were incubated for 3 days until cell index reached a plateau (indicative of cell confluence).Media was changed every two days.For initial testing of the impact of iRBCegress media on barrier integrity, on the day of the experiment iRBC-egress media was added at concentrations of either 12.5, 25 or 50 million ruptured iRBC/mL.Changes in cell index were measured every 15-seconds for 8 hours and then every 15-minutes for 48-hours.To test the impact of Ang-1 on P. falciparum-iRBC-egress media mediated barrier breakdown, iRBC-egress media was added at 50x10 6 ruptured iRBC/mL after either a 1-or 18-hour pre-incubation with 1 μg/mL of rAng-1.Again, changes in cell index were measured every 15 seconds for 8 hours and then every 15-minutes for 48-hours.Quantification of partial protection was done by first normalizing each condition to the media only control.Then the area under the curve of the negative values, indicative of barrier breakdown, was calculated using GraphPad Prism (version 10.0.2) where the media only without rAng-1 condition represented an average value of 0 and the iRBC-egress media without rAng-1 condition an average value of 100.Statistical analysis:All statistics were obtained by use of GraphPad Prism (version 10.0.2).Nonparametric Mann-Whitney U tests were performed for most experiments, except to calculate Ang-1 protection in the 2D xCELLigence and microvessel experiments in which a repeated measures one-way ANOVA test with Dunnett's multiple comparisons test and Kruskal-Wallis test with Dunn's multiple comparisons test were performed respectively.A P-value<0.05 was considered statistically significant.

Figure S3 .
Figure S3.A) The analysis pipeline begins with the extraction of 3D segmented pericyte meshes.Shown here are four representative pericytes.B) Pericyte geometry is flattened into two principal dimensions using PCA by a Python-based image analysis pipeline.The 2D geometry's shape is then determined using convex hull analysis.To describe pericyte morphology, pericyte cell borders are analyzed for circularity, calculated as 4π*Area/Perimeter 2 , where a value of 1 indicates a perfect circle and values approaching 0 indicate increasingly elongated shapes, and maximum width.

Figure S4 .
Figure S4.A) Concentration of secreted angiopoietin-1 in supernatant obtained from either HBVP-HBMEC co-culture or HBMEC-only monolayers grown in a transwell model.Data points represent 3 independent experiments.Box and whisker plots display the median and the minimum and maximum data points.B) Concentrations of vascular factors PDGF-BB, N-cadherin, Timp-1, Angptl4, VEGF and Tie-2 measured by Luminex from 3D brain microvessels supernatants treated with media only or iRBC-egress media for 18-hours.C) Concentrations of released cytokines IL-8 and CXCL-1 measured by Luminex from 3D brain microvessels supernatants treated with media only or iRBC-egress media for 18-hours.Box and whisker plots display the median, 25th and 75th percentiles and the minimum and maximum data points in B and C. Data points represent supernatants pooled from 2-3 devices in 6-7 independent experiments.Statistical significance is analyzed by Mann-Whitney U test.

Figure S5 .
Figure S5.A) Experimental outline of short-term (1-hour) incubation with rAng-1 in a 2D HBMEC monolayer in the absence of HBVP.B) Representative recording data of xCELLigence measurements taken after +/-addition of rAng-1 followed by addition of media only or iRBCegress media.All conditions were normalized to the media only control.C) Area under the curve analysis.All conditions were normalized to P. falciparum-iRBC-egress media as 100.Data points represent 3 independent experiments run in triplicate.Statistical significance is analyzed by repeated measures one-way ANOVA test with Dunnett's multiple comparisons test.