Protocol for studying dynamics of parasite-host-mediated endothelial inflammation in 3D engineered brain microvessels

Summary

Bioengineered microvessels offer a controllable human system over cellular composition, vessel architecture, flow, and circulating components for studying vascular diseases. This protocol details the use of 3D brain microvessels to model in vivo-like parasite binding, maturation, and induced vascular inflammation during cerebral malaria progression. We outline steps for studying microvascular inflammation through immunofluorescence, ultrastructural imaging, RNA sequencing, and functional assays for permeability and leukocyte recruitment. This protocol can be adapted to investigate vascular dysfunction and blood-endothelium interactions in other diseases.

For complete details on the use and execution of this protocol, please refer to Howard et al.¹

Graphical abstract

Highlights

• •Steps to build 3D microvessels supporting parasite maturation in co-culture
• •Controllable model to study endothelial interactions with blood components
• •Instructions to study endothelial inflammatory processes

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

Bioengineered microvessels offer a controllable human system over cellular composition, vessel architecture, flow, and circulating components for studying vascular diseases. This protocol details the use of 3D brain microvessels to model in vivo-like parasite binding, maturation, and induced vascular inflammation during cerebral malaria progression. We outline steps for studying microvascular inflammation through immunofluorescence, ultrastructural imaging, RNA sequencing, and functional assays for permeability and leukocyte recruitment. This protocol can be adapted to investigate vascular dysfunction and blood-endothelium interactions in other diseases.

Before you begin

With an estimated 263 million malaria cases globally in 2023, Plasmodium falciparum infection remains the leading cause of malaria deaths.² During the blood stage of infection, P. falciparum malaria can progress to cerebral malaria (CM), a severe neurological complication with high mortality and long-term brain injury, primarily affecting children aged under five years.³ A hallmark of CM is the sequestration of P. falciparum-infected erythrocytes (IEs) in the brain microvasculature, accompanied by brain edema and inflammatory host responses.^4,5,6 In particular, high levels of parasite sequestration and elevated circulating cytokine concentrations are associated with increased malaria severity.^6,7,8 Post-mortem studies have revealed breakdown of the blood brain barrier (BBB), presence of brain hemorrhages, microvascular thrombosis, and accumulations of intravascular monocytes, indicating a complex interplay of vascular pathology in CM pathogenesis.^6,7 However, the dynamic parasite-host interactions that drive endothelial activation and cerebral microvascular dysfunction in CM progression remain incompletely understood, in large part due to the inaccessibility of the brain in life and the lack of animal models for P. falciparum-CM.

We have established a hydrogel-based, 3D human microvessel model employing a 13-by-13 grid patterned network to study parasite binding and inflammatory responses at the blood-brain endothelium interface.^1,9,10 Combining soft lithography and collagen injection molding, the fabrication process enables defined branching vessel geometries and arteriole- and venule-sized channels.¹¹ The perfusable microvessel molded in native extracellular matrix (ECM) is highly adaptable for luminal culture of various organ-specific endothelial cells, including primary human brain microvascular endothelial cells (HBMECs) used in this protocol and human pluripotent stem cell-derived endothelial cells, as well as incorporation of perivascular cells directly into the collagen matrix to establish microvessels with BBB properties.^11,12,13 Generation of 3D microvessel further permits controlled perfusion of CM-relevant stimuli and blood components, including uninfected red blood cells (RBCs), IEs, immune components (e.g., peripheral blood mononuclear cells (PBMCs)) and inflammatory cytokines (e.g., tumor necrosis factor alpha (TNF-α)).^1,14 Other perfusates, such as patient plasma-exposed IEs, can also be incorporated into experimental designs to study malaria immunity.¹⁵ Thus, this 3D microvessel system serves as a versatile platform for parametric investigations of vascular biology with broad applications in modeling organ-specific microvascular dysfunction and exploring cellular interactions under disease-specific biomolecular and biophysical perturbations.

Innovation

While traditional endothelial monolayers are used to study CM pathogenesis, the advent of tissue engineering has enabled disease modeling in a more physiological and controlled 3D environment.^{16,17,18,19,20} To facilitate mechanistic studies of a human-tropic pathogen, this protocol employs a 3D perfusable human microvessel model fabricated in collagen hydrogel to recreate in vivo-like brain microvascular branching and luminal structure. This protocol presents a comprehensive workflow for studying time-dependent brain endothelial responses to complex CM stimuli in a 3D microvessel model (Figure 1). The 3D microvessel system permits straightforward experimental design to perfuse CM-relevant stimuli and blood components independently, sequentially, or in combination to dissect the roles of host and parasite factors involved in systemic and localized endothelial activation (Figure 1A). With optimized microvessel-IE co-culture conditions, this protocol describes a robust approach that extends the flow-based IE cytoadhesion assay to a 3D microvessel-IE co-culture, supporting intravascular maturation and rupture of IEs, which mimic key events of the blood-stage infection cycle (Figure 1B). The microvessel-IE co-culture system supports a broad suite of downstream assays for spatiotemporal analysis of how disease stimuli drive brain endothelial inflammation and progression to CM-associated vascular pathology. The protocol integrates immunofluorescence and ultrastructural imaging, permeability assay, leukocyte perfusion, and bulk RNA sequencing to provide a multi-scale characterization of endothelial inflammation at phenotypic, functional, and transcriptomic levels. Perfusion-based studies can be tailored to specific questions on malaria biology, and the overall approach can be readily applied to study other vascular diseases and to evaluate therapeutic compounds aimed at restoring vascular function.

Figure 1.

Experimental timeline and schematic representation of the engineered 3D human brain microvessel-IE co-culture system

(A) Schematic of the 3D microvessel device showing a 13-by-13 patterned microvessel network embedded in the collagen hydrogel (left) and a schematic representation of a selection of perfusates at the microvessel inlet (right).

(B) Overview of key steps in co-culturing P. falciparum-IEs with 3D-engineered brain microvessels. The timeline highlights five experimental techniques (in box) described in this protocol. Below is an illustration of expected parasite intraerythrocytic maturation.

Institutional permissions

This protocol requires the use of whole blood from human subjects. Ethics approval from the local Institutional Review Board or relevant governing institution is required before proceeding with steps that involve human blood sample collection. All volunteers provided informed consent for the use of their blood in this study. Experimental procedures in this study were conducted under a University of Washington Institutional Review Board-approved protocol (IRB ID: STUDY00012120).

P. falciparum parasite culture

Timing: ∼10 days

Note: We provide summary points for culturing the clonal parasite line IT4var19, which was first selected for binding to primary HBMECs. More detailed information on general handling of asexual blood stage parasite culture using human uninfected RBCs in vitro can be referred to Lopez-Perez and Seidu and “Methods in Malaria Research, 6^th Edition (2013)”.^21,22

CRITICAL: The following steps are performed under sterile conditions using a cell culture hood. Disinfect all consumables used in parasite culture with 10% bleach before discarding.

• 1.Grow IEs in a suspension of anonymized human O+ type RBCs at 5% hematocrit (diluted from packed RBC) in complete parasite culture media.

Note: The blood stage of P. falciparum is characterized by parasite growth and replication inside human RBCs. Add uninfected RBCs to support asexual replication of the parasite in vitro. The IT4var19 parasite line is grown in complete P. falciparum serum media containing 10% pooled human serum to promote optimal surface display of the P. falciparum erythrocyte membrane protein 1 (PfEMP1) cytoadhesion ligand on IEs.²³

• 2.Perform gas equilibration of the media by gas exchange to establish a microaerophilic atmosphere (e.g., gas mixture of 5% O₂, 5% CO₂, and 90% N₂) in a non-vented cell culture flask.
• 3.Monitor parasitemia (the percentage of IEs) and assess parasite life cycle stages daily using a thin blood smear and parasite stain (e.g., Hemacolor or Giemsa stain).

Note: For routine cultivation, P. falciparum parasites are maintained at a low parasitemia (parasite density below 5%) to prevent the culture from collapsing due to nutrient depletion and waste products. Refer to the desired parasitemia and parasite multiplication rate of a specific strain.²¹

• 4.Replace parasite culture media daily and maintain parasitemia at less than 5%.
• 5.Determine var gene transcripts by quantitative RT-PCR using IT4 var strain-specific primer sets.

Note: To study the cytoadherence of CM parasite isolates in an in vitro system, use the P. falciparum laboratory line IT4var19, a DC8-EPCR-binding PfEMP1 variant (recommended). IT4var19 was derived from the IT4 strain and initially generated after repeated panning on immortalized human brain microvascular endothelial cells followed by limited clonal dilution.¹⁶ Cloned parasite lines should predominantly express a single var transcript.

Preparation of HBMEC culture

Timing: ∼2–5 days, depending on culture passages prior to microvessel seeding

Note: Refer to vendor’s instructions for detailed thawing, plating, and culture of primary human brain microvascular endothelial cells (HBMECs). Perform HBMEC culture in sterile conditions using a cell culture hood.

• 6.Prepare complete EGM-2MV growth media supplemented with 5% FBS (see materials and equipment section) for primary HBMEC culture.
• 7.Coat a T-75 culture flask with 5 mL of Attachment Factor and remove the solution after 30 seconds (s).
• 8.Thaw the cryopreserved cell vial in a 37°C water bath.
• 9.Dilute cells in 8 mL of complete EGM-2MV growth media and centrifuge at 500 x g for 4 min to remove residual DMSO from the supernatant.
• 10.Resuspend cells in 1 mL of complete EGM-2MV growth media.
• 11.Plate the 1 mL cell suspension into a T-75 culture flask containing 9–14 mL of EGM-2MV growth media.
• 12.Refresh EGM-2MV growth media at least 6 h after cell plating to remove residual DMSO and unattached cells.
• 13.Refresh EGM-2MV growth media every 2 days and subculture when the culture reaches 90% confluency.

Note: HBMECs are grown as monolayers until seeded into 3D microvessels up to passage 6 (P 6). While in monolayer culture, HBMECs should always be maintained at 80%–90% confluency (passage density). If plated too sparsely, HBMECs will not proliferate because they require intercellular contact for optimal growth. An HBMEC monolayer reaching 80%–90% confluency is optimal for seeding into microvessel devices. However, high monolayer confluency (95%–100%) will lead to contact inhibition in HBMEC, which can decrease the efficiency of cell attachment during seeding.

Preparation of microvessel fabrication

Timing: ∼1 week

Advanced tissue engineering technologies have enabled their application in the generation of perfusable, 3D endothelialized vascular networks in vitro. The microvessel system used in this protocol employs soft lithography and collagen molding to create a vascular network with defined geometry.^11,24 We describe a summary of the essential components required for microvessel fabrication: acrylic-based microfluidic culture jigs, polydimethylsiloxane (PDMS) stamps for casting collagen hydrogel, and collagen stock solution. Other vascular patterning approaches like templating and bioprinting require different setup processes.

Note: Example computer-aided design (CAD) files for the microvessel culture jigs (step 14) and the wafer consisting of the 13-by-13 microvessel network geometry (step 15) are provided in the supplemental zipped folder (Data S1).

• 14.
  Manufacture transparent acrylic-based microvessel culture jigs.

  • a.
    Design the top and bottom culture jigs using any CAD software. See Figure S1A for technical drawings of the microvessel culture jigs used in this protocol (designed in Fusion 360).

    Note: Ensure the jig design includes shallow molding wells for collagen casting, collagen injection ports, media reservoirs aligned with the microvessel channels on the PDMS stamp, threaded holes for screw placement and tweezer handling, and features compatible with imaging or pump systems.

  • b.
    Machine all pieces.

    Note: Acrylic jigs can be fabricated using CNC mills, which are often available from local machine shops. Ensure jig material is biocompatible.

• 15.
  Fabricate the positive micro-patterned PDMS stamps for casting the top collagen piece.

  • a.
    Design microvessel network geometry using any CAD software (See Figure S1B for a 13-by-13 grid network used in this protocol).

    Note: Design network geometry based on specific research or engineering goals. Consider design criteria such as throughput, biologically relevant vascular architecture (e.g., diameter, curvature, and vascular density), and compatibility with other fabrication techniques (e.g., photoablation or self-assembly to achieve a multi-scale vascular network) and assays (e.g., binding or angiogenesis).^1,25

  • b.
    Fabricate the micro-patterned master mold with a negative feature for casting the PDMS stamps.

    Note: The master mold can be fabricated via micromilling or 3D printing.

  • c.
    Mix and de-gas the PDMS mixture (mix 184 Silicone Elastomer and curing reagent in a 10:1 ratio in a plastic cup).
  • d.
    Cast and cure PDMS on the mold.
  • e.
    Peel off and cut into the positive patterned top stamps.
• 16.
  Fabricate the flat PDMS stamp.

  • a.
    Cast and cure PDMS in a clean Petri dish.
  • b.
    Peel off the flat PDMS stamp and cut it into appropriate sizes for casting the flat, thin bottom collagen piece.
• 17.
  Prepare type I collagen stock from rat tail tendons (Rockland) in 0.1 % acetic acid at a concentration of 15 mg/mL.

  • a.
    Prepare fresh 0.1% (v/v) acetic acid in molecular biology grade water and store at 4°C for 12–16 h.
  • b.
    Extract collagen fibers from rat tail tendons into ice-cold 70% isopropanol.
  • c.
    Wash collagen fibers in ice-cold 70% isopropanol three times followed by one wash in molecular biology grade water.
  • d.
    Dissolve collagen fibers in ice-cold 0.1% acetic acid (at least 60 mL per gram of collagen fibers) for at least 48 h until a viscous solution appears.
  • e.
    Centrifuge dissolved collagen fibers and collect the supernatant to remove debris and undissolved collagen fibers.
  • f.
    Freeze the collagen supernatant at −20°C for at least 48 h.
  • g.
    Lyophilize frozen collagen tubes on a freeze dryer for at least three days.
  • h.
    Weigh the lyophilized collagen ‘sponge’ and dissolve in freshly prepared, 4°C 0.1% acetic acid to the desired stock concentration (e.g., 15 mg/mL).

Note: Please refer to Rajan et al. for detailed steps on Type I collagen preparation, involving tendon extraction, dissolution in acetic acid, centrifugation, freeze-drying, and reconstitution to the desired stock concentration.²⁶ High concentration collagen I sourced from other species is also commercially available. Alternatively, decellularized matrices from specific tissues can be extracted and used as ECM sources in this protocol.²⁷

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
ICAM-1 monoclonal antibody (clone MEM-111), FITC, dilution 1:100	Thermo Fisher Scientific	Cat# MHCD5401, RRID:AB_10373958
APC conjugated, mouse monoclonal anti-VE-cadherin (clone 16B1), dilution 1:100	Thermo Fisher Scientific	Cat#17-1449-42, RRID: AB_467495
Mouse monoclonal anti-human CD14 Alexa Fluor 488 (clone 61D3), dilution 1:50-1:100	eBioscience	Cat# 53014942; RRID: AB_2744748
Mouse monoclonal PE conjugated anti-human CD45 (clone H130), dilution 1:50–1:100	BioLegend	Cat# 304008; RRID: AB_314396
Biological samples
O+ human red blood cell CPDA-1	BioIVT	Cat# HUMANRBCPDA-0111751
A+ human serum (heat-inactivated)	Interstate Blood Bank Inc	N/A
Human peripheral blood mononuclear cell (PBMC)	Healthy donor	N/A
Rat tail	Rockland	Cat# RT-T297
Chemicals, peptides, and recombinant proteins
RPMI 1640 medium (parasite culture)	Gibco	Cat# 23400-021
10 M sodium hydroxide (NaOH)	Sigma-Aldrich	Cat# 72068
Hypoxanthine	Sigma-Aldrich	Cat# H9377
D-(+)-glucose	Sigma-Aldrich	Cat# G7021
Sodium bicarbonate (7.5% solution)	Corning	Cat# 25-035-CI
Gentamicin (50 mg/mL)	Gibco	Cat# 15750078
EGM-2 MV microvascular endothelial cell growth medium SingleQuots supplements	Lonza	Cat# CC-4147
EBM-2 basal medium	Lonza	Cat# CC-3156
90% nitrogen, 5% oxygen, 5% carbon dioxide (gas blend)	Linde Gas and Equipment Inc.	Cat# BI NICDOXC1-K
100% methanol	Thermo Fisher Scientific	Cat# A412-4
Giemsa stain	Sigma-Aldrich	Cat# GS500
Hemacolor solution II	Sigma-Aldrich	Cat# 65044B
Hemacolor solution III	Sigma-Aldrich	Cat# 65044C
70% isopropanol	Thermo Fisher Scientific	Cat# 02-003-140
Glacial acetic acid	Thermo Fisher Scientific	Cat# A38-500
Poly(ethyleneimine) solution (for coating)	Supelco	Cat# P3143
Glutaraldehyde (for coating)	Sigma-Aldrich	Cat# 49629
Medium 199 (10X)	Gibco	Cat# 11825015
Dulbecco’s phosphate-buffered saline (DPBS)	Corning	Cat# 21-031-CV
Trypsin-EDTA (0.25%, sterile-filtered)	Sigma-Aldrich	Cat# T4049
Tumor necrosis factor alpha human	Sigma-Aldrich	Cat# H8916
Paraformaldehyde (PFA) solution, 4% in PBS	Thermo Fisher Scientific	Cat# J19943.K2
Aqueous glutaraldehyde EM grade 25%	Electron Microscopy Sciences	Cat# 16200
Hoechst 33342	Thermo Fisher Scientific	Cat# H1399
Dextran, fluorescein, 70,000 MW, anionic	Invitrogen	Cat# D1823
Lymphoprep solution	STEMCELL Technologies	Cat# 18061
RPMI 1640 medium, no glutamine, no phenol red (for PBMC isolation)	Gibco	Cat# 32-404-014
Triton X-100 for molecular biology	Sigma-Aldrich	Cat# T8787
Cytiva HyClone bovine serum albumin (BSA)	Thermo Fisher Scientific	Cat# SH3057402
Phalloidin	Invitrogen	Cat# A12380
Karnovsky’s fixative, ½ strength	Fred Hutch Imaging Core	2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer
Aqueous glutaraldehyde EM grade 50% (for Karnovsky’s fixative)	Electron Microscopy Sciences	Cat# 16310
Paraformaldehyde, reagent grade (for Karnovsky’s fixative)	Thermo Fisher Scientific	Cat# 04042-500
Sodium cacodylate, trihydrate	Electron Microscopy Sciences	Cat# 12310
2-mercaptoethanol	Thermo Fisher Scientific	Cat# ICN19483425
Critical commercial assays
PKH67 green fluorescent cell linker kit for general cell membrane labeling	Sigma-Aldrich	Cat# PKH67GL
RNeasy micro kit	QIAGEN	Cat# 74004
QIAshredder	QIAGEN	Cat# 79654
Qubit RNA IQ assay kits	Thermo Fisher Scientific	Cat# Q33221
RNA ScreenTape	Agilent Technologies	Cat# 5067-5576
RNA ScreenTape ladder	Agilent Technologies	Cat# 5067-5578
RNA ScreenTape sample buffer	Agilent Technologies	Cat# 5067-5577
XT DNA library prep kit	Nextera XT	Cat# FC-131-1024
SMART-Seq v.4 ultra-low input RNA kit	Takara Bio	Cat# 634890
Deposited data
Sequencing data	Howard et al.1	GEO: GSE227672
Experimental models: Cell lines
Primary human brain endothelial cell (HBMEC)	Cell Systems	Cat# ACBRI 376
Experimental models: Organisms/strains
Plasmodium falciparum strain: IT4 (line: IT4var19)	Avril et al.16	Provided by Dr. Joseph Smith’s lab at Seattle Children’s Research Institute
Oligonucleotides
Primer set for the IT4 var repertoire	Viebig et al.28	N/A
Software and algorithms
Fusion 360	Autodesk Fusion	N/A
NIS-Elements	Nikon Instruments Inc.	NIS Elements Ar (Advanced Research)
iDEP.95 (integrated, differential expression, and pathway analysis)	Ge et al.29	http://bioinformatics.sdstate.edu/idep95/
DEseq2	Bioconductor30	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
ShinyGO v.0.75 (Gene Ontology Enrichment Analysis)	Ge et al.31	http://bioinformatics.sdstate.edu/go75/
ImageJ (version 2.3.0)	Schneider et al.32	https://imagej.nih.gov/ij/
MATLAB	MathWorks	https://www.mathworks.com/products/matlab.html
GraphPad Prism	GraphPad	https://www.graphpad.com/
Excel	Microsoft	N/A
Other
Filter unit (pore size 0.8 μm)	Nalgene	Cat# 126-0080
Filter unit (pore size 0.45 and 0.2 μm)	Nalgene	Cat# Z370657 and Z358207
T-75 U-shaped canted neck cell culture flask with vent cap	Corning	Cat# 430641U
184 silicone elastomer kit	SYLGARD	Cat# 184 SIL ELAST KIT
Rat tails for type I collagen isolation	Rockland Immunochemicals, Inc.	Cat# RT-T297
Square cover glass	Thermo Fisher Scientific	Cat# 50-238-9165
Bleach	VWR	Cat# 89501-620
HyClone HyPure water, molecular biology grade	Cytiva	Cat# SH30538
30 mL freestanding specimen and centrifuge tubes	VWR	Cat# 89012-778
1 mL BD tuberculin syringe	BD	Cat# 00382903096596
Handheld corona surface treater (Plasma treater)	McMaster-Carr	Cat# 5693K14
Stainless steel dowel pin	McMaster-Carr	Cat# 93600A088
Stainless steel pan head Phillips screws (4-40 1/2″)	McMaster-Carr	Cat# 91735A106
QuadroMACS separator	Miltenyi Biotech	Cat# 130-090-976
LD columns	Miltenyi Biotech	Cat# 130-0492-901
ColorFrost microscope slides	Thermo Fisher Scientific	Cat# 12-550-17
Vacutainer whole blood tubes with ACD solution A (8.5 mL)	BD Biosciences	Cat# BDV364606
SepMate 50 mL tubes	STEMCELL Technologies	Cat# 85450
Razor blade	Red Devil	Cat# 3272BB
Critical point dryer	Tousimis	Cat# Autosamdri-815A
Cylinder SEM mount for JEOL JSM-840	Electron Microscopy Sciences	Cat# 75734
Gold/palladium sputter coater target	Electron Microscopy Sciences	Cat# 91017-AP
Sputter coater	Denton Vacuum	Desk IV
JSM-6610LV scanning electron microscope	JEOL Ltd.	JSM-6610LV
1.7 mL microcentrifuge tubes	Corning	Cat# MCT-175-C-S
Nanodrop spectrophotometer	Thermo Fisher Scientific	Cat# ND-2000
Qubit fluorometer	Invitrogen	Cat# Q33238
4200 TapeStation system	Agilent Technologies	Cat# G2991BA
Sequencing system	Illumina	NovaSeq 6000 System

Materials and equipment

1000 mL of complete P. falciparum serum culture media (parasite culture media)

Reagent	Final concentration	Amount
RPMI 1640 Medium (containing L-glutamine and 25 mM HEPES)	N/A	863.8 mL
0.1 M hypoxanthine in 1 M NaOH	0.05 g/L	4 mL
45% glucose	0.9 g/L	2 mL
Pooled A+ human sera	10%	100 mL
Sodium Bicarbonate (7.5%)	2.25 g/L	30 mL
Gentamicin (50 mg/mL)	10 mg/L	200 μL

Store at 4°C and use within one month. Alternatively, store at −20°C and use within 6 months.

Complete (or serum-free) EGM-2MV growth media (microvessel media)

• •Thaw the EGM-2 MV microvascular endothelial cell growth medium SingleQuots supplements (CC-4147) at 18°C–23°C.
• •Prepare complete EGM-2MV growth media supplemented with 5% fetal bovine serum (FBS) by adding all supplements to EBM-2 basal media (CC-3156).
• •To make serum-free EGM-2MV growth media, skip the addition of 5% FBS.

Store at 4°C and use within one month.

P. falciparum-3D brain microvessel co-culture media

• •Add an additional 5% pooled A+ human serum (v/v) to the complete EGM-2MV growth media supplemented with 5% FBS.

Note: This microvessel-IE co-culture media is optimized to promote parasite asexual maturation in 3D brain microvessels. See troubleshooting 1 for media formulation optimization.

• •Filter the complete co-culture media (5% FBS + 5% pooled malaria-naïve A+ human serum) through a sterilizing-grade membrane with a pore size of 0.22 μm.

CRITICAL: Skipping the co-culture media pre-filter step may result in serum-associated aggregates and clogging of the microvessel lumen during perfusion.

Store co-culture media at 4°C for up to 2 weeks and use fresh.

Tumor necrosis factor alpha-containing microvessel stimulation media

• •Reconstitute recombinant human TNF-α in sterile water to an initial concentration of 0.1 mg/mL.
• •Dilute to a stock concentration of 10 μg/mL in 0.1% BSA buffer (carrier protein), following the manufacturer’s protocol to generate stock aliquots.
• •Dilute the stock aliquots in microvessel or co-culture media to a desired microvessel treatment concentration (e.g., 10 ng/mL).

Store stock aliquots at −20°C or −80°C for up to 6 months or at 4°C for up to a week. Avoid repeated freeze-thaw cycles.

70 kDa dextran fluorescein permeability solution

• •Dilute 70 kilodalton (kDa) dextran fluorescein conjugate to a 10 mg/mL stock solution in sterile 1X DPBS following the manufacturer’s protocol.
• •Vortex or sonicate for 1 min to dissolve dextran conjugates in aqueous buffers.
• •For long-term storage, aliquot and store stock dextran solution at −20°C and protect from light. Use within one year.
• •For short-term storage, store the aqueous dextran stock solution at 4°C and protect from light. Use within one month.

Dilute the 70 kDa dextran fluorescein stock solution (10 mg/mL) at 1:100 (v/v) in microvessel or co-culture media. Use at least a 20-fold dilution for permeability assay.

Always prepare fresh dextran-containing media.

3.7% PFA and 0.008% glutaraldehyde fixative for microvessel-IE co-culture

• •Dilute the stock 20% PFA 1:5 (v/v) in 1X DPBS.
• •Add and dilute the stock glutaraldehyde to the 3.7% PFA solution to a final concentration of 0.008% glutaraldehyde.

Store the stock PFA and glutaraldehyde at 4°C once opened for up to 3 months and always prepare fresh 3.7% PFA and 0.008% glutaraldehyde fixative.

Blocking and permeabilization buffer

• •Prepare a 2% bovine serum albumin (BSA) working solution by adding 1 g of BSA to 50 mL of 1X DPBS.
• •Prepare a 1% Triton X-100 working solution by adding 0.5 mL of the stock Triton X-100 to 49.5 mL of 1X DPBS.

Note: Use a 1-mL syringe to transfer the viscous stock solution.

• •Filter-sterilize both solutions.
• •Prepare the final blocking and permeabilization buffer by mixing 1% Triton X-100 working solution and 2% BSA in a 1:9 (v/v) ratio.

Store the working solutions at 4°C and use within three months.

Staining solutions

• •Option 1: For microvessel and IE nuclei staining only, dilute Hoechst 33342 at 1:250 (v/v) in 1X DPBS.
• •Option 2: For visualization of actin cytoskeleton, cell adhesion molecules, and junctional proteins, dilute Alexa Fluor 568 Phalloidin, FITC-conjugated anti-human ICAM-1, and APC-conjugated anti-human VE-cadherin at 1:100 (v/v) in blocking and permeabilization buffer.
• •Option 3: For microvessel with adhered PBMC staining, dilute Hoechst 33342 at 1:250 (v/v), Alexa Fluor 488-conjugated anti-human CD14, and PE-conjugated anti-human CD45 at 1:50-1:100 (v/v) in blocking and permeabilization buffer.

Always prepare fresh solutions to ensure optimal staining quality.

Step-by-step method details

Fabrication of 3D brain microvessel

Timing: ∼1 day for material preparation, ∼1 h per device fabrication, and 3–5 days of microvessel culture post-fabrication

We provide an abbreviated overview of brain microvessel fabrication (Figure S1). The 3D perfusable brain microvessel is generated through soft lithography, collagen injection molding, and direct perfusion-based seeding of HBMECs. Detailed protocols for microvessel fabrication and gravity-driven culture are described elsewhere.^11,24

• 1.
  Prepare autoclaved and sterilized materials in advance:

  • a.
    Autoclave tweezers, a stainless-steel spatula, stainless-steel screws, stainless-steel dowel pins, glass coverslips, and water.

    Note: Match the dowel pin diameter, screw size and thread to the media reservoir ports and culture jig designs.

  • b.
    Clean the PDMS stamp surfaces with adhesive tape before autoclaving.
  • c.
    Sterilize the acrylic culture jigs:

    • i.
      Immerse the jigs in 10% bleach for at least 1 h.
    • ii.
      Thoroughly rinse the jigs with autoclaved water in a sterile beaker.
    • iii.
      Air dry the jigs completely.
  • d.
    Coat the molding wells (on top and bottom jigs) with poly(ethyleneimine)-glutaraldehyde (PEI-GA):

    • i.
      Prepare a well-mixed 1.0% (wt/vol) PEI solution and 0.1% (wt/vol) GA solution in molecular biology-grade water.
    • ii.
      Filter-sterilize both solutions using a 0.22-μm sterile syringe filter.
    • iii.
      Coat the surfaces of molding wells with PEI for 10 min.
    • iv.
      Aspirate the residual PEI solution.
    • v.
      Coat the molding wells with GA for 15 min.
    • vi.
      Rinse the jigs thoroughly with autoclaved water.

      CRITICAL: Ensure complete rinse and dry of the coated jigs as GA residue is cytotoxic and residual liquid may affect collagen gelation. Always prepare freshly coated jigs, as PEI-GA coating becomes ineffective after 2–3 days.

• 2.
  Prepare the neutralized collagen hydrogel for injection molding:

  • a.
    Determine the total volume (V_final) and the final concentration (C_final) of neutralized collagen hydrogel needed for microvessel fabrication.

    Note: V_final depends on the dimensions of the molding wells and the number of microvessels to be fabricated. We prepare 7.5 mg/mL collagen gel for collagen injection molding.

  • b.
    Transfer the calculated volume of stock collagen solution (V_stock) to a clean 30-mL freestanding centrifuge tube using a 1 mL syringe (see calculations in Table 1).

    CRITICAL: Always keep all reagents on ice or at 4°C. Use the final collagen solution for injection molding within 3 h after preparation (step 2d).

  • c.
    Prepare the neutralizing solution (see Table 1).

    Note: We use complete EGM-2MV growth media as the 1X cell culture media in this protocol. Replace 1X cell culture media with the appropriate media for the specific endothelial cell type seeded in the microvessel.

  • d.
    Mix and stir the neutralizing solution with stock collagen solution using a spatula without introducing air bubbles:

    • i.
      Mix thoroughly.
    • ii.
      Adjust the pH of the final collagen solution to 7.4 by adding additional 1 N NaOH in 5-μL increments.
    • iii.
      Centrifuge at 1,000 x g for 10 min at 4°C to remove air bubbles if needed.
• 3.
  Fabricate the micro-patterned top collagen piece (Figure S1B):

  • a.
    Place the top PDMS stamp (positive patterned surface facing up) in a Petri dish.
  • b.
    Plasma-treat the stamp surface for 30 s.

    CRITICAL: Plasma treatment modifies the PDMS surface to increase hydrophilicity. Surface wetting is essential to minimize bubble formation on the micro-patterned network PDMS surface during collagen injection.

  • c.
    Align the microvessel channels (patterned stamp) with the inlet/outlet ports (top jig).
  • d.
    Insert stainless-steel pins into the reservoirs.

    Note: Adjust the dowel pin size to match the media reservoir design. The purpose of inserting the dowel pin is to prevent collagen from blocking the inlet and outlet ports of the perfusion channel.

  • e.
    Slowly inject the prepared collagen solution through the injection port using a 1-mL syringe.

    CRITICAL: Both plasma treatment and slow injection are essential to minimize bubble entrapment and for preserving the imprinted microvessel network pattern. Adjust the collagen injection volume based on the molding well size.

  • f.
    Place the top piece assembly in a 37°C incubator for at least 30 min to allow collagen gelation.

    Note: Avoid exceeding 2 h of collagen gelation when fabricating multiple devices, as collagen can dry out and bubbles may form.

• 4.
  Fabricate the flat bottom collagen piece (Figure S1B):

  • a.
    Place a glass coverslip at the center of the molding well (bottom culture jig).
  • b.
    Slowly dispense the collagen gel evenly onto the glass coverslip and within the PEI-GA coated area.
  • c.
    Add one large collagen drop at a corner to create a “leading edge.”

    Note: Adjust the collagen injection volume based on the molding well size.

  • d.
    Gently lower the flat PDMS stamp from the “leading edge” over the collagen gel using two tweezers.
  • e.
    Gently scrape across the flat PDMS stamp surface.

    CRITICAL: Apply light pressure along the edges to ensure even and flat spreading of the collagen gel. This step ensures that the molded collagen bottom piece is level with the bottom jig surface. Uneven spreading can cause gel extrusion after gelation, leading to non-perfusable channels during device assembly.

  • f.
    Place the bottom piece assembly in a 37°C incubator for at least 30 min to allow collagen gelation.
• 5.
  Assemble the top and bottom jigs with molded collagen pieces (Figure S1B):

  • a.
    Prepare the bottom collagen piece for assembly:

    • i.
      Aspirate excess collagen around the PDMS stamp.
    • ii.
      Apply 1X DPBS to the bottom piece assembly.
    • iii.
      Carefully remove the flat PDMS stamp from the bottom collagen piece and jig.

      CRITICAL: Carefully slide the flat PDMS stamp off the collagen and bottom culture jig surface using pipette tips or tweezers. Direct lifting of the stamp can detach the collagen gel from the coated surface or shift the coverslip, leading to device leakage after assembly.

    • iv.
      Add 1X DPBS onto the bottom collagen piece to prevent dehydration and to create a liquid interface for assembly in step 5d.
  • b.
    Prepare the top collagen piece for assembly:

    • i.
      Grab the top culture jig using straight tweezers.
    • ii.
      Flip over the top piece assembly.
    • iii.
      Apply 1X DPBS to the top piece assembly around the stamp-jig interface using a 1-mL syringe.
    • iv.
      Carefully remove the micro-patterned PDMS stamp.

      CRITICAL: Lift the Petri dish straight up, perpendicular to the collagen surface in a single motion to avoid damaging the molded microvessel network. The top PDMS stamp should come off with the dish. Do not slide the top stamp.

    • v.
      Remove the stainless-steel pins and clean any excess collagen from the inlet and outlet openings.
    • vi.
      Add 1X DPBS to the top collagen piece to prevent dehydration.
  • c.
    Place four screws to the top culture jig (with the reservoir side facing up and the collagen piece facing down).
  • d.
    Align and slowly lower the top jig onto the bottom jig.

    CRITICAL: Add additional 1X DPBS to both the top and bottom collagen pieces to create a liquid interface during assembly. Avoid sliding the jigs against each other, as this may cause mechanical damage to the micro-pattered network features on the top collagen piece.

  • e.
    Gently tighten the screws at all four corners.

    CRITICAL: Do not overtighten the jigs as this may collapse the 3D microchannels.

  • f.
    Equilibrate the acellular device with EGM-2MV growth media (hereafter referred to as microvessel media) for 1 h:

    Note: This step is required to set optimal culture conditions and physiological pH for cell seeding.

    CRITICAL: Culture media, cells, and perfusates are removed from the microvessel at the inlet and outlet reservoirs using a micropipette only. Direct aspiration will damage collagen matrices and compromise microvessel viability or barrier integrity.

    • i.
      Remove residual 1X DPBS from the inlet/outlet reservoirs.
    • ii.
      Add 50 μL of pre-warmed microvessel media to the outlet reservoir.
    • iii.
      Add 150 μL of microvessel media to the inlet reservoir to establish gravity-driven flow.

      Note: Adjust media volume based on the media reservoir design.

    • iv.
      Place the assembled device in a 37°C incubator for at least 1 h to allow proper collagen sealing.

      Note: Perfusable devices should have balanced media levels by the end of the 1 h media equilibration. Non-perfusable devices cannot be seeded.

• 6.
  Seed microvessels with HBMECs under gravity-driven flow:

  • a.
    Trypsinize the HBMEC monolayer from the T-75 flask.
  • b.
    Count cells and resuspend the cell pellet to a final seeding density of 1X10⁷ cells/mL in microvessel media.

    Note: Gently agitate the cell suspension immediately before seeding the microvessel device to ensure a homogenous single-cell suspension, as HBMEC will quickly form cell pellet at high concentrations.

  • c.
    Perfuse 10 μL of cell suspension into the acellular microvessel channel via the inlet reservoirs under gravity-driven flow.

    Note: To ensure uniform cell coverage in the microvessel network, inspect the device under a bright-field microscope to confirm even seeding density and desired flow speed throughout the channel network (Methods video S1).
  • d.
    Adjust seeding density and flow speed by adding cell suspension to the inlet reservoir and/or changing the inlet-outlet volume difference in 5 μL increments.

    Note: Optimize flow speed based on microvessel channel geometry. Reduce seeding speed as the surface area-to-volume ratio decreases to maximize cell adhesion to the luminal collagen surface.

  • e.
    Place the seeded device in a 37°C incubator for 30 min to allow cell attachment.
  • f.
    Perfuse another 10 μL of cell suspension through the outlet reservoir under gravity-driven flow for 30 min (repeat steps 6 c-d in the reverse flow direction) to achieve uniform cell distribution throughout the network.
• 7.Remove unattached cells from the inlet and outlet reservoirs at least 6 h post-seeding.

Note: HBMECs should begin adhering to the collagen matrix immediately after seeding. To remove excess cells left around the reservoirs during seeding, dislodge cell aggregates with a single gentle pipetting using a P200 pipette.

• 8.Add fresh microvessel media to establish gravity-driven flow.

Table 1.

Neutralized collagen hydrogel for injection molding

Reagent	Formula for reagent volume
Stock collagen solution (Vstock)	Vstock = Vfinal x (Cfinal / Cstock)
Neutralizing solution
M199 10x media supplement	VM199 = 0.1 x Vfinal
1 N NaOH	VNaOH = 0.022 x Vstock
1x cell culture media	V1x = Vfinal - Vstock - VM199 - VNaOH

C_stock is the reconstituted stock collagen concentration from step 17 under the “Preparation of microvessel fabrication” section.

3D human brain microvessel culture and treatment before perfusion studies

Timing: 3–5 days

It is critical to culture microvessel for 3–5 days post-seeding to establish a resting endothelial state and a tight barrier integrity before perturbation with inflammatory stimuli or initiation of perfusion studies. The following section details the preparation of microvessels (resting or TNF-α-stimulated) prior to co-culture with P. falciparum-IEs or uninfected RBCs.

• 9.
  Culture the microvessels for 3 to 5 days at 37°C in a 5% CO₂ incubator:

  • a.
    Replenish media reservoirs every 12–15 h (day and night) with pre-warmed microvessel media to reset the gravity-driven flow:

    • i.
      Remove residual media from the inlet and outlet reservoirs.

      CRITICAL: Do not aspirate. Always use a pipette when handling microvessels.

    • ii.
      Add 50 μL of fresh media to the outlet reservoir.
    • iii.
      Add 150 μL of media to the inlet reservoir.

      CRITICAL: Newly fabricated microvessels are often leaky until the cell-cell junction and cell-matrix interactions establish and stabilize. Replace microvessel media at least twice per day during the first two days to prevent cell delamination from the collagen matrix and “artificial” barrier disruption due to microvessel drying out. Regular media changes are also essential to re-establish gravity-driven, unidirectional flow across the vessel network.

      Note: Flow rate gradually decays and equilibrates within approximately 1 h in our setup. The exact equilibration time may vary depending on the network geometry, initial media height difference, and reservoir volume.

  • b.
    Gently remove air bubbles from the inlet and outlet reservoirs using a P1000 pipette by slowly withdrawing the bubble to restore media flow through the network.

    CRITICAL: Always check for air bubbles at the reservoir ports to prevent flow blockage. Bubbles can form when residual media levels are low during media replenishment, likely due to air entrapment at the narrow inlet/outlet interface. Without removing the bubble, blocked flow can cause microvessel delamination within 12 h.

• 10.
  Monitor microvessels condition under bright-field microscopy during each media exchange session (Figure 2A):

  • a.
    Check for microvessel confluency, indicated by a continuous dark endothelial lining.
  • b.
    Remove any remaining cell debris from the inlet and outlet reservoirs.

Note: Several factors can cause delamination or partial retraction of the lumen surface during microvessel culture, including frequent media changes and temperature or pH fluctuations. Partially delaminated microvessels may recover with continued culture. We recommend monitoring the vessel for 1 day before discarding. If there is no sign of recovery or if delamination leaves a permanent retracted lumen wall, discard the microvessel.

• 11.
  Switch from standard microvessel media (5% FBS) to co-culture media (final concentration of 5% FBS + 5% human serum) at least one day after EC seeding and three days before IE perfusion studies (Figure 1B):

  • a.
    Prepare co-culture media by adding 5% pooled malaria-naïve human serum (v/v) to standard microvessel media.
  • b.
    Add 50 μL of co-culture media to the outlet reservoir.
  • c.
    Add 150 μL of co-culture media to the inlet reservoir.

CRITICAL: The IT4var19 P. falciparum line is grown in parasite culture media containing 10% human serum, but it grows poorly in standard microvessel culture media. While microvessel media works for short-term IE binding assay in microvessels, the presence of additional human serum is essential to support P. falciparum-IE asexual maturation in 3D brain microvessels. For details on microvessel-IE co-culture media optimization, see troubleshooting 1.

Note: Introducing co-culture media at least three days prior to IE perfusion studies provides a pre-adaptation period for the microvessel, as the same media is used for IE perfusion and microvessel-IE co-culture. In our experience, additional 5% human serum has no deleterious effect on microvessel culture (discussed in the Expected Outcomes section).

• 12.Continue to culture microvessel under gravity-driven flow, replenishing co-culture media every 12 h until the experiment is terminated.
• 13.Optional: Stimulate/prime microvessels with pro-inflammatory cytokines (e.g., 10 ng/mL TNF-α, see materials and equipment) in co-culture media for 18 h prior to IE perfusion studies (Figure 1B).

Note: This step is required to simulate the individual effects of circulating inflammatory mediators or the sequential effects of host and parasite stimuli on endothelial inflammation (e.g., elevated systemic TNF-α levels and parasite sequestration observed in CM patients).⁸

Figure 2.

Representative microscopic images of microvessel and trophozoite-stage IE binding

(A) Bright-field images showing normal (resting state) and defective microvessel conditions before IE perfusion. White arrows point at examples of specific microvessel defects. Exclude affected microvessels with persistent/irreversible defects labeled with red cross symbols. Scale bars indicate 100 μm.

(B) Fluorescence images illustrating fluorescent RBC membrane-labeled IT4var19-IE binding to the microvessel lumen following 1 h of perfusion, media washes, fixation, and Hoechst nuclei staining. Co-perfused trophozoite stage IT4var19-IEs are membrane labeled with either green PKH67 or red PKH26 dyes. White arrows indicate non-specific IE trapping and accumulation, typically observed in regions with microvessel delamination, retracted lumen surfaces, or imperfections in the grid pattern. Scale bars indicate 100 μm.

Parasite perfusion and co-culture in 3D brain microvessel

Timing: ∼11.5–27.5 h (2 h of IE enrichment + 1 h of IE/uninfected RBC perfusion + 0.5 h media wash + 8 h or 24 h of co-culture in 3D brain microvessels)

The following steps describe procedures for IE enrichment, perfusion, and co-culture in 3D brain microvessels through the intraerythrocytic cycle. These steps establish in vivo-like conditions for parasite binding, in situ maturation, and rupture on the microvessel wall. We used IT4var19, a brain-selected parasite line, as a representative parasite stimulus to study IE-induced brain endothelial activation and dysfunction in the context of CM pathogenesis.

CRITICAL: Always keep the microvessel at 37°C in a 5% CO₂ incubator, except for 5–10 min intervals during perfusate introduction and media exchanges. Even a 1-hour removal from the incubator can lead to possible microvessels delamination and stalled parasite maturation. Pre-warm all reagents to 37°C before adding them to the reservoirs. Before starting any experimental treatment or perfusion studies, always check microvessel conditions under bright-field microscope to ensure a consistent baseline. Exclude microvessels with major network imperfections, delamination, or unstable endothelial barrier from downstream studies (Figure 2).

• 14.
  Enrich trophozoite-stage IEs:

  • a.
    Use a MACS cell separator to purify mature IEs at the mid trophozoite-stage (∼25–33 h post-infection, hpi):

    Note: To ensure sufficient IEs for purification and subsequent perfusion and co-culture experiments, expand parasite cultures to a high parasitemia (at least 5% trophozoite-stage IEs) before performing magnetic purification. Culture-adapted parasite lines can usually tolerate a short-term higher parasitemia (5% or more). For more information on MACS purification, refer to Uhlemann et al. and Piatti et al.^10,33

    • i.
      Insert LD Columns on a QuadroMACS Separator.
    • ii.
      Equilibrate the column with 3 mL of parasite culture media.
    • iii.
      Discard the media flow-through.
    • iv.
      Load 5 mL of 1% hematocrit culture with 5% parasitemia to the column and allow it to flow through.

      Note: For high IE enrichment purity from uninfected RBCs, use parasite culture with at least 5% parasitemia.

    • v.
      Wash the column with 5 mL parasite culture media to remove unbound RBCs.
    • vi.
      Repeat the wash with 2 mL parasite culture media.
    • vii.
      Remove the column from the magnet.
    • viii.
      Add 2 mL of parasite culture media and collect the enriched IEs in a new tube using the plunger.
  • b.
    After MACS purification, count the total number of IEs using a hemocytometer.
  • c.
    Prepare a small, fixed blood smear on microscope slides and stain with 10% Giemsa solution for 10 min to confirm enrichment efficiency (i.e., parasitemia). See troubleshooting 2 if enrichment efficiency is low.

    Note: MACS purification typically yields synchronized mature stage IEs with greater than 90% purity.

  • d.
    Resuspend the IE pellet in 5 mL parasite culture media before preparing the IE perfusate.

    Note: Purified parasite samples can be stored at 18°C–23°C for up to 3 h, but they must be used in perfusion studies within that time.

• 15.
  Perfuse IEs in 3D brain microvessels:

  • a.
    Wash the purified parasite samples:

    Note: The same procedure is applicable for preparing uninfected RBC perfusate. Skip step 15b if fluorescent membrane labeling of the IEs is not required. For membrane labeling, use fluorescent cell linker kits (e.g., PKH67 green or PKH26 red). The fluorescent cell linker kit supplies dye solution (e.g., PKH67) and the labeling vehicle Diluent C. Refer to the manufacturer’s manual for detailed instructions.

    • i.
      Count IEs using a hemocytometer before centrifugation.
    • ii.
      Centrifuge the parasite sample at 400 x g for 2 min to obtain the cell pellet.
    • iii.
      Remove the supernatant and wash the cell pellet once with 2 mL of co-culture media if skipping membrane labeling.
    • iv.
      Wash instead with serum-free EGM-2MV growth media if proceeding with membrane labeling (step 15b).
    • v.
      Centrifuge at 400 x g for 2 min and discard the supernatant from the washing step.
  • b.
    Perform live cell membrane-labeling:

    Note: This step applies to both uninfected RBC and IE sample preparations, enabling real-time visualization of perfusion dynamics and straightforward binding quantification, discussed later. PKH dye is lipophilic, non-cytotoxic, and does not interfere with cellular functions such as adhesion, making it suitable for general cell labeling. The labeling volume provided below works well for a sample of 25 million erythrocytes. Optimize dye concentration for other cell types or cell concentrations.

    • i.
      Leave no more than 25 μL of supernatant from the washing step (step 15a(v)).
    • ii.
      Prepare 1 mL of staining solution in Diluent C at a working concentration of 4x10^-6 M (e.g., add 4 μL of ethanolic dye solution to 1 mL of Diluent C).
    • iii.
      Resuspend the cell pellet in 1 mL of Diluent C with gentle pipetting to ensure a single cell suspension shortly before staining.

      Note: A single cell suspension is required for uniform membrane staining.

    • iv.
      Add 1 mL of staining solution (step 15b(ii)) to 1 mL of cell suspension (step 15b(iii)) to initiate membrane labeling.
    • v.
      Incubate for 1-2 min at 18°C–23°C with periodic mixing.
    • vi.
      Stop the staining reaction immediately by adding an equal volume of human serum (i.e., 2 mL) and incubate for 1 min.
    • vii.
      Centrifuge at 400 x g for 2 min to pellet the cells and remove the supernatant.
    • viii.
      Wash the pellet by resuspending it in 2 mL of co-culture media.
    • ix.
      Centrifuge at 400 x g for 2 min and discard the supernatant after each wash.
    • x.
      Repeat the washing once.
  • c.
    Resuspend the IE pellet after the washing step (step 15a) or after membrane labeling (step 15b) to a final concentration of 5x10⁶ IEs/mL in co-culture media.

    Note: We do not recommend perfusing membrane-labeled IEs in microvessels assigned for subsequent RNA-sequencing or functional assays. Although we have observed no adverse effects on parasite binding or microvessel stability after labeling, we have noted that some membrane label can transfer to endothelial cells.

  • d.
    Remove the co-culture media from microvessel reservoirs before adding the IE suspension.

    CRITICAL: Before IE perfusion, check the microvessel condition under a bright-field microscope (Figure 2A). Microvessel defects can cause non-specific IE trapping in the surrounding collagen matrix (Figure 2B).

  • e.
    For gravity-driven perfusion, add 50 μL of co-culture media to the outlet reservoir.

    Optional: Gravity-driven perfusion is initiated by establishing a media height difference between the inlet/outlet reservoirs, which gradually equilibrates over 1 h as the hydrostatic pressure difference decreases. Alternatively, use a syringe pump and tubing to provide continuous flow through microvessels with a programmed/controlled flow rate.^9,10 Note that collagen hydrogel has limited tolerance to pressure higher than 40 mmHg tested across the microvessel network.

  • f.
    Add 150 μL of IE perfusate to the inlet reservoir.

    Note: Depending on the reservoir size and channel geometry, the perfusate volume added at the two reservoirs may vary. We recommend performing numerical simulation (e.g., using COMSOL Multiphysics software) to refine perfusion conditions (i.e., flow rate and wall shear stress). Adjust the pressure drop and corresponding flow through the microvessels as needed by adding or removing 5–10 μL of co-culture media at the inlet or outlet reservoirs.

  • g.
    Perform the same perfusion for the uninfected RBC control in a separate microvessel and for other experimental conditions (e.g., resting or TNF-α-stimulated microvessels).
  • h.
    Check the IE and uninfected RBC perfusion under bright-field microscopy before returning each device to the incubator.

    CRITICAL: This step ensures proper flow-based perfusate-microvessel interaction. If flow is obstructed by air bubbles, remove the perfusate from the inlet reservoir and reperfuse the microvessel using co-culture media first, as described in step 9b. Once perfusion is restored, reload fresh perfusate as outlined in step 15d-f.

  • i.
    Return the microvessel device to the incubator during the 1 h-perfusion to allow sufficient time for IEs to interact with and adhere to the microvessel lumen surface.

    CRITICAL: Always keep the microvessel-IE co-culture in the incubator to prevent microvessel delamination. For live imaging, use a microscope equipped with a cage incubator for temperature control and a micro-environmental chamber for humidity and gas control.

• 16.Optional: Perform live imaging during the 1-h IE perfusion to visualize microvessel-IE interactions in real time.

Note: If IEs are not membrane-labeled, monitor them using the bright-field channel. Perfusion in the 13-by-13 grid network can achieve a range of flow conditions that closely mimic flow velocities and wall shear stress observed in human post-capillary venules (Methods video S2).^9,34

• 17.
  Wash non-adherent P. falciparum-IEs from microvessels:

  • a.
    After the 1-h IE perfusion, prepare a Giemsa-stained smear with 10 μL of effluent collected at the outlet reservoir.

    Note: The smear collected at 1 h post-IE perfusion represents a baseline reference for IE morphology prior to intraerythrocytic growth.

  • b.
    Remove unbound IEs by washing the microvessels twice with pre-warmed co-culture media:

    • i.
      Gently remove the IE perfusate from both the inlet and outlet reservoirs.

      Note: Disinfect all consumables and media waste with a 10% bleach solution.

    • ii.
      Establish a gravity-driven media wash by adding 50 μL of co-culture media to the outlet reservoir.
    • iii.
      Add 150 μL of co-culture media to the inlet reservoir, maintaining the same flow rate used during IE perfusion.
    • iv.
      Allow each wash to proceed for 15 minutes.

      CRITICAL: Return the microvessel to the 37^oC incubator during media washes.

    • v.
      Remove media from both reservoirs and repeat the media wash once more.

      Note: Complete a total of two 15-min washes following the 1-h IE perfusion. If a visible uninfected RBC or IE pellet remains at the inlet or outlet reservoir, proceed with another media wash.

• 18.
  Optional: Terminate the experiment at the 1 h post-IE perfusion endpoint:

  • a.
    For direct assessment of IE binding, fix microvessels (step 22).
  • b.
    Stain microvessels and bound IEs for imaging (step 34):

    • i.
      Evaluate parasite stage based on nuclei and bright-field morphology.
    • ii.
      Perform binding quantification as discussed later (Binding quantification workflow).
  • c.
    Alternatively, proceed with sample preparation for ultrastructural analysis via scanning electron microscopy (step 35).
• 19.
  Continue the microvessel-IE co-culture with media feeding twice daily:

  • a.
    Add fresh co-culture media every 12 h without removing existing media from the reservoirs:

    • i.
      Add 50 μL of co-culture media to the outlet reservoir.
    • ii.
      Add 150 μL of co-culture media to the inlet reservoir.

CRITICAL: This step aims to maintain a CM-like inflammatory milieu after IE binding and during in situ maturation of IEs in 3D brain microvessels. Media volume will gradually decrease over the 12-h period as the 3D microvessel is not a fully sealed system (due to the porous nature of the collagen matrix and open reservoirs). Adding media without removing the existing volume minimizes dilution of parasite- and host-derived inflammatory factors while maintaining flow and nutrients for cell viability. Adjust the co-culture media volume as needed to account for device leakage, while keeping the hydrostatic pressure difference consistent with that used for IE perfusion.

• 20.Use Giemsa-stained effluent smears to monitor in situ parasite maturation through schizogony and IE rupture.

CRITICAL: Confirming the presence of matured parasites at later endpoints (e.g., 8 h and 24 h post-IE binding) compared to the 1-h post-perfusion baseline morphology is critical for validating the co-culture system (see expected outcomes).

• 21.
  Proceed with downstream assays at different experimental endpoints as described in later sections (Figure 1B):

  • a.
    Keep the co-culture in a cell culture incubator at 37°C and 5% CO₂.
  • b.
    Add fresh media to the co-culture twice daily.
  • c.
    Study microvessel inflammation in live microvessel-IE co-culture at the specified endpoints:

    • i.
      Perform functional assays and live imaging (steps 23–33).
    • ii.
      Perform bulk RNA sequencing (step 36).
  • d.
    Study microvessel inflammation in fixed microvessel-IE co-culture:

    • i.
      Perform microvessel staining and imaging (step 34).
    • ii.
      Perform ultrastructural imaging via scanning electron microscopy (step 35).

Note: Functional assays such as barrier permeability and leukocyte recruitment studies or RNA sequencing require live microvessel input. Evaluation of brain EC responses via fluorescence confocal or scanning electron microscopy requires fixed microvessels. Microvessels used for permeability assays or live imaging of parasite binding can be fixed for immunofluorescence imaging. Use the appropriate sample preparation methods and fixative reagents.

• 22.
  Fix the microvessels by perfusing fixative containing 3.7% PFA and 0.008% glutaraldehyde under gravity-driven flow (see troubleshooting 3 for issues with IE preservation and materials and equipment for preparation):

  Note: Add glutaraldehyde to preserve erythrocyte morphology.³⁵ Use a separate fixative for ultrastructural imaging (refer to step 35).

  • a.
    Slowly remove the co-culture media from both reservoirs.
  • b.
    Add 150 μL of pre-warmed fixative to the inlet and 50 μL to the outlet.
  • c.
    Keep the microvessels perfused with the fixative for 15 min at 18°C–23°C.
  • d.
    Remove the fixative from both reservoirs.

    Note: Dispose of PFA and glutaraldehyde as hazardous waste in appropriate waste containers.

  • e.
    Wash the microvessels by perfusing 1X DPBS three consecutive times, with each wash lasting 10 min.

    • i.
      For each wash, add 50 μL of 1X DPBS to the outlet.
    • ii.
      Add 150 μL of 1X DPBS to the inlet.
  • f.
    Replenish the inlet and outlet with 1X DPBS after the last wash.

    Pause point: To prevent dehydration of the collagen microvessel construct after fixation, fill the inlet and outlet reservoirs with 1X DPBS or submerge the device in 1X DPBS. Store fixed microvessel devices in cell culture dishes at 4°C for up to 1 month prior to downstream processing. Seal the dish with parafilm to prevent evaporation.

Methods video S2. Example of microvessel-IE interactions under flow, related to step 16

Zoomed-in video of enriched trophozoite-stage IT4var19 IEs perfused in the 3D microvessel under bright-field microscopy. Aggregated sequestered IEs remain bound to the luminal EC surface under a physiologically relevant high flow rate.

Characterization of barrier permeability in 3D microvessel

Timing: 30 min total for reagent preparation + 15 min per session of microvessel imaging + 10 min medium wash per device post-dextran perfusion

This section describes a functional assay to measure changes in endothelial barrier permeability using a live, 3D human brain microvessel co-cultured with adherent IEs at the 8-h endpoint. The procedures can be adapted to other endpoints and fluorophore-conjugated probes as needed. The goal is to characterize inflammatory stimuli-induced changes in barrier function by measuring the transport of perfused dextran tracer across the microvessel lumen into the collagen matrix.

Note: Use live-cell fluorescence microscopy equipped with a cage incubator for temperature control and a micro-environmental chamber for humidity and gas control to maintain microvessel barrier stability outside the cell culture incubator during procedures lasting longer than 5 min.

• 23.
  Perform baseline permeability assay (Figure 3):

  Note: Perform baseline permeability assays to assess endothelial barrier function when the microvessel has reached a confluent and stable state, generally 3 to 5 days after HBMEC seeding (Figure 3A). In addition, perform a baseline permeability assay 6–12 h before IE perfusion or cytokine stimulation (Figure 1B). Conducting this assay in advance minimizes potential overlap between transient endothelial responses to media changes or dextran exposure and subsequent perfusion studies. To further reduce variability, schedule the permeability assay at least 2 h after the most recent media change or device manipulation.

  If dextran leakage (either focal or diffuse) is observed, re-assess the microvessel to ensure a stable baseline barrier function after 1–2 days. Postpone the subsequent perfusion study or microvessel stimulation until the barrier integrity improves to minimize artifacts and reduce non-specific IE trapping at junctional gaps (Figures 2B and 3B).

  • a.
    Prepare 70 kDa dextran perfusate:

    Note: Use non-fixable dextran in this step. This ensures that residual tracers will be removed during media washes and will not become cross-linked to cells after vessel fixation with PFA. The same color channel can be reused for subsequent immunofluorescence imaging of microvessels.

    • i.
      Dilute the 10 mg/mL stock dextran solution 1:100 (v/v) in pre-warmed co-culture media to prepare a working solution of 100 μg/mL (see materials and equipment).

      Note: Prepare at least 200 μL of dextran perfusate per microvessel. This volume may vary depending on the reservoir design.

    • ii.
      Place the dextran solution in the microscope’s cage incubator or a similar microscope enclosure to equilibrate the temperature to 37°C.
  • b.
    Set up live cell time-lapse imaging to measure permeability (Figure 3A):

    Note: Optimize the acquisition setup according to the available instrument before performing the permeability assay.

    • i.
      Sterilize the stage holder.
    • ii.
      Select the appropriate objective and set the exposure time for the imaging channel.

      Note: We use 10x objective and 20 milliseconds (ms) exposure time for the 488 nm fluorescence channel.

    • iii.
      Configure time-lapse acquisition in imaging software (e.g., NIS-Elements used in this protocol) or establish an automated acquisition pipeline (e.g., using JOBS in NIS-Elements).

      Note: Define acquisition parameters, including the xyz position for each field of view (FOV), number of acquisition time points, time sequence definition (e.g., run every 1 s for a 1-min interval time-lapse sequence), and multi-channel definition (e.g., 488 nm+568 nm). Time-lapse data can then be used for both dynamic permeability coefficient calculations and intensity-based permeation quantification discussed later (Microvessel barrier permeation quantification workflow).

  • c.
    Transfer the microvessel device in a sealed cell culture dish to the microscope.

    Note: Check the media levels at the inlet/outlet reservoirs prior to dextran perfusion. If the media level is less than ∼20 μL in either the inlet or outlet reservoir, replenish the reservoir with pre-warmed co-culture media to prevent bubble formation during the media switch and dextran perfusion. Delay permeability measurement of that device for 2 h.

  • d.
    Define the desired FOV for the permeability assay:

    Note: Each FOV should capture a section of the bulk dextran flow in the vessel lumen, vessel wall-matrix interface, and the collagen matrix. We recommend using a 10x confocal objective for optimal resolution and FOV size (∼1300 x 1300 μm) in the current microvessel system. Ensure the microvessel device is securely mounted on the imaging stage to prevent movement during image acquisition. We recommend acquiring one FOV per microvessel device with a single inlet/outlet reservoir. The throughput/multiplexity of the permeability assay can be increased by using microvessel designs with multiple independent inlet/outlet channels or by co-perfusing dextran conjugates at different wavelengths (e.g., 10 kDa Alexa Fluor 568-conjugated and 70 kDa FITC-conjugated dextran) in equimolar amounts. We also recommend acquiring a maximum of three FOVs per microvessel device with multiple independent inlet/outlet channels in an imaging session lasting up to 8 min as dextran permeability measurement is time-sensitive and all acquisitions should be sampled before tracer equilibration and signal saturation to accurately capture the initial permeation kinetics across the endothelial barrier.

    • i.
      Use bright-field imaging to locate each FOV.
    • ii.
      Define and record the x and y coordinates of each FOV.
    • iii.
      Define and record the z position at the center of the microvessel lumen.

      Note: This is an estimated starting center position. The xyz positions may shift slightly after dextran introduction due to pipetting.

    • iv.
      Select the appropriate color channels for the perfused fluorescent tracers.
    • v.
      Acquire a background image of the collagen matrix prior to dextran perfusion (Figure 3B).

      Note: This image can be used to determine background fluorescence intensity from each FOV for background subtraction or for computing the permeability coefficient. It can also be used to segment the microvessel lumen in the quantification workflow discussed later.

  • e.
    Load fluorescent dextran:

    • i.
      Gently remove any residual co-culture media from the inlet and outlet reservoirs.
    • ii.
      Using a P200 pipette, load 50 μL of the dextran solution to the outlet reservoir.
    • iii.
      Quickly load another 150 μL of the dextran solution to the inlet reservoir (see troubleshooting 4 for issues regarding dextran perfusion).

      Optional: Temporarily enable live camera mode on the dextran fluorescence channel at the selected FOV during this step. This allows immediate visualization of dextran perfusion to detect potential flow blockages and to adjust the xyz position as soon as flow is established.

  • f.
    Initiate the first acquisition after reaching steady and uniform intraluminal tracer fluorescence:

    • i.
      During the defined waiting period, re-center the selected FOV using the fluorescent channel.
    • ii.
      During the defined waiting period, update the xy positions.

      Note: Adjust the wait time (required for flow to fully establish and for dextran to reach uniform, constant intravascular concentration post-loading). The precise timing will depend on different microvessel designs and perfusion setups. To optimize, measure the time it takes for the tracer signal to reach a stable maximum, intravascularly. The purpose is to minimize quantification variability caused by low signal-to-noise ratio (due to low tracer accumulation in the matrix) and potential photobleaching.

    • iii.
      Adjust and update the z focus at each FOV until the dextran-perfused lumen boundary becomes sharp.

      Note: This ensures consistent measurement from a comparable z position across devices.

    • iv.
      Initiate the first acquisition.

      CRITICAL: Ensure all images are acquired at consistent time points post-dextran perfusion for accurate comparison, since the permeability assay is time sensitive.

  • g.
    Initiate the second acquisition at the next time point (e.g., 5 min post-dextran perfusion) if the automated procedure is not used.

    Note: This permeability assay is designed to capture a permeation rate by fitting tracer leakage into the matrix during the short linear influx phase. We employed a simplified approach to estimate the permeation property using a linear fit between two discrete time points (i.e., 2 min and 5 min). Given that the square collagen grids of the microvessel network are only ∼300 μm wide (Figure 3B), a total assay duration up to 8 min post-dextran perfusion provides sufficient sensitivity to detect changes in endothelial barrier function before possible signal saturation and tracer equilibration. This is a faster assay compared to the standard 2D Transwell setup, which often requires longer tracer incubation (≥1 h) to detect signal changes.³⁶ Optimize acquisition time point selection and total assay duration by plotting tracer intensity from the collagen matrix over time in both an acellular device (lower bound) and a control endothelialized brain microvessel (with a tight barrier as the upper bound) using the same setup and tracer size. The optimal assay duration is expected to be proportional to matrix/gel size and concentration and inversely related to tracer size and the inherent hydrostatic pressure.

  • h.
    Perform media washes post-permeability imaging:

    • i.
      Bring the microvessel back to the cell culture hood.
    • ii.
      Remove the dextran perfusate from both the inlet and outlet reservoirs.
    • iii.
      Wash the microvessel once with fresh, pre-warmed co-culture media by adding 150 μL to the inlet and 50 μL to the outlet reservoir for 10 min in the incubator.
    • iv.
      Remove the media wash from both reservoirs.
    • v.
      Replenish both reservoirs with fresh co-culture media.

      Note: It is possible to perform multiple permeability assays on the same microvessel device, spaced at least 2 h apart. Media replenishment following dextran removal and media washes should effectively remove residual tracers from the microvessel lumen. This results in negligible background noise in the collagen matrix after media washes and no re-calibration is needed for previously defined acquisition time points (Figure S2).

• 24.Replenish the inlet and outlet media to re-establish gravity-driven flow.

CRITICAL: Check the media levels at the inlet and outlet reservoirs frequently during the first 2 h following the permeability assay as the microvessel barrier may become temporarily leaky due to dextran perfusion and device manipulation. Exclude any microvessels that show delamination after the baseline measurement from subsequent perfusion studies (Figure 3B).

• 25.Proceed with microvessel stimulation by perfusing cytokine (step 13) and IE (step 15–19) individually, sequentially, or in combination.
• 26.
  Assess endothelial barrier function post-stimulation:

  • a.
    Repeat the permeability assay outlined in step 23 at the treatment endpoint, e.g., the 8-h post-IE binding endpoint, to assess parasite-induced changes in endothelial barrier function.
  • b.
    Repeat the permeability assay to evaluate cytokine-induced barrier disruption (e.g., 18-h post-TNF stimulation).
• 27.Either fix the microvessels with 3.7% PFA and 0.008% glutaraldehyde for subsequent analysis or continue re-establishing gravity-driven flow with co-culture media until the next endpoint.

Note: This approach can also be used to assess the kinetics of barrier recovery. However, limit the number of dextran perfusions per microvessel to a maximum of three, including the baseline measurement, to prevent endothelial delamination and compromised barrier integrity from frequent changes in culture conditions.

Pause point: Store fixed microvessel devices in cell culture dishes at 4°C for up to 1 month before downstream analysis.

Figure 3.

Characterization of endothelial barrier permeability in a 3D human brain microvessel

(A) Schematic timeline of the permeability assay showing key steps and durations involved in a single-channel (FITC-conjugated 70 kDa dextran), one FOV measurement at a specified endpoint. The numbered boxes refer to protocol steps.

(B) Representative bright-field and dextran-perfused images of the same FOV, highlighting the structural and functional differences between a mature, stable brain microvessel with intact endothelial barrier and a delaminated microvessel. Delamination leads to significant leakage of 70 kDa dextran at 2 min post-perfusion, indicating compromised barrier integrity (white arrows). Scale bars indicate 200 μm.

Characterization of leukocyte recruitment in 3D microvessel

Timing: Total of 2 h for PBMC isolation and preparation + 1 h perfusion + 0.5 h media wash

Total of approximately 1 day per batch of device fixing and immunofluorescent labeling and another 1 h per microvessel imaging.

This section describes steps to assess leukocyte recruitment in 3D human brain microvessels co-cultured with bound IEs at the 8-h endpoint. These steps can be adapted to study other samples (e.g., leukocyte recruitment post-TNF-α stimulation).

CRITICAL: Perform fresh PBMC isolation on the day of the perfusion experiment (no later than 3 h before perfusion). Perform PBMC isolation and perfusion in sterile conditions. Ensure that all reagents and the centrifuge are at 18°C–23°C when processing blood samples to prevent platelet activation.

• 28.
  Obtain whole blood from healthy donors:

  • a.
    Collect whole blood in a vacutainer tube coated with sodium citrate.
  • b.
    Gently invert the tube 8–10 times immediately after blood collection.
  • c.
    Store the blood at 18°C–23°C until PBMC isolation.

CRITICAL: Ensure that blood is collected from consenting healthy donors under an Institutional Review Board-approved protocol. Avoid using K2 EDTA spray coated tubes as any EDTA residue may cause microvessel disruption during PBMC perfusion. Ensure sufficient mixing of the anticoagulant with the blood to prevent clot formation, but avoid excessive mechanical shaking to minimize platelet activation.

• 29.
  Isolate PBMCs using SepMate tubes according to the manufacturer’s instructions or following the isolation protocol described by Efthymiou et al:³⁷

  Note: Process the whole blood at 18°C–23°C during the entire procedure. Other PBMC purification methods are also applicable, such as Ficoll-Hypaque density gradient centrifugation.

  • a.
    Inject 15 mL of Histopaque gradient medium (Lymphoprep) into a 50-mL SepMate PBMC isolation tube.
  • b.
    Dilute whole blood (1:1, v/v) in clear RPMI medium before adding it into the SepMate tube.
  • c.
    Perform the initial centrifugation of the layered diluted blood in the Histopaque-containing SepMate tube at 1200 x g for 20 minutes.

    Note: Centrifuging for 20 minutes instead of 10 minutes improves RBC pelleting.

  • d.
    Pour off the enriched PBMCs and plasma layer into a new 50-mL conical tube.

    Optional: To reduce platelet contamination in the enriched PBMCs, pipette off the plasma layer above the PBMC layer before pouring.

  • e.
    Wash the PBMC pellet twice with 10 mL of RPMI 1640 medium:

    • i.
      Centrifuge at 300 x g for 10 min at 18°C–23°C with acceleration and deceleration set to 4.
    • ii.
      Remove the supernatant.
    • iii.
      Repeat the media wash and perform cell counting before the second centrifugation.

      Note: A pink cell pellet should appear at the bottom of the conical tube.

• 30.Resuspend the PBMC pellet in co-culture media to a final concentration of 1 × 10⁶ PBMCs/mL for perfusion.
• 31.
  Perfuse PBMCs in microvessels for 1 h:

  • a.
    Perfuse the microvessels with the PBMC suspension from step 30 by following the same IE perfusion procedure as described in steps 15.

Optional: Monitor the PBMC perfusion and leukocyte recruitment process in real time using bright-field live imaging.

• 32.
  Remove unbound PBMCs after 1 h of perfusion:

  • a.
    Remove PBMCs from the inlet and outlet reservoirs.
  • b.
    Media wash the microvessels twice using co-culture media as described in step 17b.
• 33.Proceed with microvessel fixation for subsequent imaging and analysis.

Note: Follow the same fixation procedure using 3.7% PFA as described in step 22, or follow the ultrastructural imaging protocol as discussed in step 35. Remove glutaraldehyde from the fixative to preserve immunoreactivity for PBMC immunostaining.

Pause point: Store fixed microvessel devices in cell culture dishes at 4°C for up to 1 month before downstream analysis.

Characterization of temporal changes in brain microvessel inflammation

Timing: Study within a 24 h window post-IE perfusion

Timing: ∼1 h per batch of nuclei staining and ∼1 h per microvessel imaging for IE binding. ∼1 day per batch of immunostaining and ∼1 h per confocal microvessel imaging (for step 34)

Timing: ∼2 days of sample preparation + ∼3 h of imaging (for step 35)

Timing: ∼15 min per lysate collection, ∼2.5 h per round of RNA purification, and ∼7 days of sequencing and analysis (for step 36)

This section describes methods to evaluate phenotypic, structural, and transcriptional responses of 3D human brain microvessels, in live or fixed states. Imaging and sequencing approaches can be adapted for other microvessel conditions, perfusate, and endpoints based on the hypothesis tested. To study extended parasite stimulation on brain microvessels mimicking in vivo conditions, a second round of IE perfusion as an alternative to parasite replication and invasion could be considered (see limitations).

Note: Microvessels following IE perfusion —whether or not they undergo permeability assays —can be used for immunofluorescence imaging. However, we recommend designating separate microvessels (without dextran perfusion) for ultrastructural imaging and bulk RNA sequencing to minimize minor artifacts or stress-induced changes and ensure the highest fidelity in ultrastructural and molecular readouts.

• 34.
  Perform microvessel staining and imaging:

  Note: For binding quantification of IEs (with or without membrane labeling), bright-field microscopy with Hoechst staining is sufficient to discriminate brain endothelial cells and adhered IEs based on their morphology and nuclear size. For labeling of other cell types (non-fluorescent and in mixed populations) as well as studying microvessel responses post-TNF-α and/or IE stimulation, immunostaining of relevant markers is needed.

  • a.
    Perform microvessel staining for parasite binding quantification:

    • i.
      Remove 1X DPBS from both reservoirs of the fixed microvessel.
    • ii.
      Perfuse Hoechst staining solution (see materials and equipment) to the inlet and outlet reservoirs under gravity-driven flow (150 μL and 50 μL, respectively).
    • iii.
      Incubate the staining solution for 30 min at 18°C–23°C.
    • iv.
      Wash with 1X DPBS three times consecutively, 10 min per wash.

      Note: Alternatively, include Hoechst stain in the antibody staining solution during immunostaining.

  • b.
    Perform microvessel immunofluorescence staining:

    Note: Perform microvessel immunostaining to evaluate, for example, changes in endothelial junctions or surface expression of adhesion molecules. Perform PBMC immunostaining to determine recruited immune cell types and their spatial distribution.

    • i.
      Perfuse blocking and permeabilization buffer (see materials and equipment) via the inlet and outlet reservoirs (150 μL and 50 μL, respectively) for 1 h at 18°C–23°C.
    • ii.
      During the blocking step, prepare the conjugated antibody staining solution (see examples listed in materials and equipment).
    • iii.
      Remove the blocking buffer after 1 h.
    • iv.
      Add the conjugated antibody solution (containing Hoechst stain) to the inlet and outlet reservoirs (150 μL and 50 μL, respectively) for 30 min at 18°C–23°C.
    • v.
      Reset the antibody staining solution levels to 150 μL at the inlet and 50 μL at the outlet reservoirs.
    • vi.
      Incubate the microvessel devices for 12–16 h at 4°C in cell culture dishes wrapped with aluminum foil and sealed with parafilm to prevent evaporation.
    • vii.
      Wash with 1X DPBS three times consecutively, 10 min per wash (refer to step 22e).
  • c.
    Alternatively, stain microvessels with primary and secondary antibodies:

    • i.
      After blocking, incubate the microvessel with primary antibodies diluted in blocking and permeabilization buffer at 4°C for 12–16 h.
    • ii.
      Wash three times with 1X DPBS.
    • iii.
      Perfuse secondary antibodies (with Hoechst stain) diluted in blocking and permeabilization buffer for 1 h at 18°C–23°C.

      Note: Use different host species or perform sequential staining when combining conjugated and unconjugated antibodies.

  • d.
    Image microvessels using wide-field or spinning disk confocal microscopy:

    Note: Apply the same acquisition settings across all samples to allow direct image comparison.

    • i.
      Define color channels.
    • ii.
      Define the z-range that covers the full lumen cross-section or luminal surface plane.

      Note: Acquire the full z-range using optical slices with a ≤5 μm z-step size.

    • iii.
      Acquire multi-channel z-stack images of a single microvessel lumen FOV.
    • iv.
      Alternatively, perform large-field tile scanning to capture the entire microvessel network.

      Note: Define the scan area to cover the entire microvessel network and define the z-range for the whole network. For example, for automated large image acquisition and stitching in NIS software, enable the ‘Large-image Acquisition’ and ‘Z series’ functions under the ND Acquisition panel. Capturing the entire 13-by-13 grid network (∼7 mm by 7 mm image dimension) with a 10x objective requires a total scan area of 6-by-6 fields. Define the z-range and refine it at all four corners of the scan area to ensure full coverage of the vessel lumen surfaces. At each corner, define the z-range first using the Hoechst staining (e.g., both the large endothelial nuclei and tiny IE nuclei) and confirm the setting using the membrane labels or other relevant markers.

    • v.
      Select ‘Large Images (Z series (Lambda))’ as the “Order of Experiment” to ensure that all color channels are acquired at each optical slice before proceeding to the next slice, and that the full z-stack is completed before moving to the next scan field.

      Note: This acquisition sequence helps to minimize artifacts caused by objective or stage movement.

    • vi.
      Initiate image acquisition.

      Note: IEs can bind to the top, side, and bottom surfaces of the microvessel lumen, but only limited binding information from the lumen sides can be captured in projected images. To reduce image file size during whole-microvessel tile scanning, we recommend acquiring separate z-stacks of the top and bottom luminal surfaces for binding quantification while skipping the middle z-range that captures the lumen sides.

• 35.
  Perform scanning electron microscopy (SEM) of microvessels:

  • a.
    Perfuse and fix the microvessels in half-strength Karnovsky’s fixative at 4°C for 12–16 h:

    • i.
      Add 50 μL of the fixative to the outlet.
    • ii.
      Add 150 μL of the fixative to the inlet.

      Note: Always work in a fume hood when handling Karnovsky's fixative because it can cause serious skin and eye irritation.

  • b.
    Disassemble the acrylic culture jigs and retrieve the 3D microvessel construct by the following steps (Figure S3):

    Note: Always double-glove with nitrile gloves when handling samples in Karnovsky's fixative.

    • i.
      Remove the screws from the microvessel device.
    • ii.
      Invert the microvessel jig with one hand holding the top acrylic jig and the other holding the bottom acrylic jigs.
    • iii.
      Gently tap on the glass coverslip to detach and remove the bottom acrylic jig from one side.

      Note: Ensure the collagen construct and coverslip remain attached to the top acrylic jig.

    • iv.
      Gently slide the coverslip away from the collagen construct with a fingertip.

      Note: Add or submerge the top acrylic jig in 1X DPBS if the coverslip does not come off easily due to surface tension.

    • v.
      Use a spatula to make multiple cuts along the periphery of the collagen construct from the top acrylic jigs.

      CRITICAL: Avoid making a single continuous cut with the spatula, as this may tear the thin collagen pieces.

    • vi.
      Use the spatula to gently lift and scoop the demolded collagen construct from the top acrylic jig.
  • c.
    Process the collagen construct on a clean petri dish based on imaging needs (Figure 4A):

    Note: Both processing methods can be applied to the same sample, for example, by first dividing the collagen construct in half.

    • i.
      For transverse sectioning, cut the collagen construct into halves or multiple pieces using a razor blade.

      Note: This approach is suitable for SEM imaging of microvessel lumen cross-sections.

    • ii.
      For top-down imaging of the microvessel or to assess IE binding on luminal surfaces, peel the collagen construct apart into top and bottom pieces.

      Note: This creates open channels for SEM imaging of the microvessel lumen surface. If peeling is difficult, trim excess collagen at the periphery and inspect for detachment areas. If the microvessel network is hardly visible by eye, operate under a microscope. Handle the collagen construct gently, as the thin bottom collagen piece is very fragile and prone to tearing.

  • d.
    Store processed collagen pieces at 4°C in 2 mL of half-strength Karnovsky’s fixative in a 6-well plate until imaging.
  • e.
    Submit samples to the imaging core facility for further processing:

    • i.
      Rinse the fixed construct three times with 0.1 M sodium cacodylate buffer.
    • ii.
      Dehydrate through a graded series of ethanol (50%, 70%, 95%, 100%).
    • iii.
      Perform critical point drying using the Autosamdri-815.
    • iv.
      Mount the sample onto the cylinder SEM mount.
    • v.
      Coat the sample with a thin (∼10 nm) gold/palladium layer using a sputter coater (Denton Desk IV).
  • f.
    Image the samples on a scanning electron microscope (JEOL JSM 6610LV):

    • i.
      Adjust magnification for FOV selection.
    • ii.
      Adjust focus, brightness, and accelerating voltage for optimal image acquisition.

      Note: Imaging at 5 kV provides clear resolution of EC surface morphology, bound IEs, and cross-sectional views of the vessel lumen in our setup. For transverse imaging, start with low magnification (e.g., x20 or x100) to locate the microvessel collagen construct and lumen. Use high magnification (x1000 or more) to capture detailed cross-sections of the vessel lumen and bound IEs along infoldings and outfoldings (Figure 4B). For top-down imaging, start with low magnification (x30-x100) to obtain an overview of the 13-by-13 grid network. Use ∼x250 to focus on a single vessel lumen. Increase magnification to ∼x2000 for detailed EC surface morphology with bound IEs, or to ∼×7500 for high-resolution imaging of a bound IE (Figure 4B).

• 36.
  Perform bulk RNA sequencing analysis for 3D microvessels:

  • a.
    Collect cell lysates from live microvessels:

    Note: The RNeasy Micro Kit is designed for RNA purification from low-input samples (< 5 x 10⁵ cells), such as these 3D microvessels. Refer to the manufacturer’s handbook for detailed instructions.

    • i.
      Remove culture media from the inlet and outlet reservoirs.
    • ii.
      Wash the microvessel for 2 min by gravity-driven perfusion with 150 μL of 1X DPBS to the inlet and 50 μL to the outlet.
    • iii.
      Meanwhile, prepare lysis buffer by adding 10 μL β-mercaptoethanol per 1 mL of Buffer RLT (Qiagen; RNeasy Micro Kit).
    • iv.
      Remove DPBS wash from the inlet and outlet reservoirs.
    • v.
      Perfuse 200 μL of RLT lysis buffer through the inlet for 1 min.
    • vi.
      Collect the effluents from the outlet and re-perfuse at least three times to ensure complete endothelial cell lysis.

      CRITICAL: Use a P1000 to flush and pipette the lysate through the vessel. Collect the effluent from the outlet reservoir and re-flush it through the inlet reservoir. Repeat this process to maximize collection efficiency.

    • vii.
      Collect the lysate in a 1.7-mL microcentrifuge tube on ice. See troubleshooting 5 if the total RNA yield is low.
    • viii.
      Adjust the final volume of each sample to 350 μL with lysis buffer.
    • ix.
      Homogenize the lysates by vortexing for 1 min.
    • x.
      Homogenize the lysates using a QIAshredder column for potentially higher RNA yield.

      Pause point: Proceed immediately with RNA isolation or store the lysate at −80°C for up to 6 months.

  • b.
    Purify and measure total RNA from lysate:

    • i.
      Purify RNA following the RNeasy Micro Kit protocol. Refer to the manufacturer’s handbook for detailed instructions.
    • ii.
      During the final elution step, add at least 14 μL of RNase-free water to the spin column.

      Note: A higher elution volume (e.g., 25 μL) may increase RNA yield.

    • iii.
      Centrifuge for 1 min at 20,000 x g to elute RNA.
    • iv.
      Measure RNA concentration and purity using a Nanodrop spectrophotometer.

      Note: An A260/280 ratio of around 2.0 is generally indicative of pure RNA. A significantly low A260/280 ratio at high RNA concentration may indicate protein contamination. However, at RNA concentrations below 10 ng/μL, the ratios may deviate from expected values and become less reliable. Always re-assess RNA quality and quantity using a TapeStation and Qubit Fluorometer prior to library preparation and sequencing.

  • c.
    Assess RNA concentration and quality:

    • i.
      Measure RNA concentration using the Qubit Fluorometer and assay kit.
    • ii.
      Perform RNA quality control using the Agilent High Sensitivity RNA ScreenTape System.

      CRITICAL: Proceed with library preparation only if all samples achieve a RIN score of ≥8.0 and a total RNA yield ≥50 ng with a concentration of at least 5 ng/μL.

  • d.
    Perform library preparation and sequencing:

    • i.
      Prepare sequencing libraries from quality-assessed RNA samples using the XT DNA Library Prep Kit (Nextera XT) and the SMART-Seq v4 Ultra Low Input RNA Kit (SMARTv4).
    • ii.
      Use a flow cell that yields approximately 30 million reads per sample.

      Note: NovaSeq 6000 SP with paired-end reads was chosen for the sequencing analysis in Howard et al.¹

  • e.
    Perform bulk RNA seq analysis:

    Note: Since parasites have much lower RNA concentrations than human cells, most transcripts are of human origin.

    • i.
      Align RNA-seq data to the hg38 human genome assembly using the TopHat or STAR algorithm, considering reads with >4 reads per million.

      Note: For additional details on TopHat alignment, refer to Trapnell et al., 2012.³⁸

    • ii.
      Normalize and analyze the aligned count data using iDEP.95.
    • iii.
      Perform differential expression analysis using DESeq2, considering genes with fold change >1.5 and false discovery rate (FDR) <0.05.
    • iv.
      Conduct gene ontology enrichment analysis using ShinyGo v0.75.

      Note: For an overview of RNA-seq data analysis workflows, including differential expression and pathway analysis, refer to Ge et al., 2018, Love et al., 2014, Ge et al., 2020.^29,30,31

Figure 4.

SEM imaging of the 3D microvessels

(A) Schematic illustration of two sample preparation methods for SEM, showing options for transverse and top-down imaging of the collagen construct.

(B) Representative SEM images of the two imaging methods using samples from 8 h and 24 h post-IE binding endpoints. Top row: Transverse imaging provides a cross-sectional view of the 3D rectangular to elliptical microvessel lumen with bound IEs after 8 h of trophozoite-stage adherence, while top-down imaging reveals endothelial surface morphology and IE binding details from both the top and bottom collagen pieces (i.e., the top and bottom surfaces of the microvessel lumen). Magnification power of 100 provides an overview of the microvessel construct and grid network geometry. Scale bars indicate 200 μm. Bottom row: Higher-magnification (x1000-2000) SEM images of the same sample preparation method show more resolved morphological changes in IE attachment and endothelial surface/junction than the x100 overviews. Asterisks denote the collagen matrix. The white arrow indicates remnants of a ruptured IE at 24 h post-binding. Scale bars indicate 20 μm for the x1000 image and 10 μm for the x2000 images.

Binding quantification workflow

Timing: 2–5 h of manual counting per vessel (varies depending on the sample complexity) + 2 h analysis per batch

The binding quantification of P. falciparum-IEs or PBMCs in 3D microvessels can be performed using z-stack fluorescence images acquired from step 34. Below are two complementary approaches to 1) visualize the spatial distribution of bound cells using a scatter plot, and 2) quantify binding levels across the microvessel network with a digitized heatmap that displays the binned cell counts in each microvessel segment and branching node (Figure 5).

• 37.
  Prepare image:

  • a.
    Load the hyperstack fluorescence image of the 3D microvessel in ImageJ.

    Note: This quantification workflow is compatible with hyperstack fluorescence images of 3D microvessels from a single FOV or a tile scan.

  • b.
    Use the “Rotate” function under “Transform” to adjust the microvessel image stack prior to counting.

    Note: This step is needed to best align microvessel features with rectangular-shaped regions of interest (ROIs) applied in later steps. For example, we oriented the horizontal and vertical vessel segments to be parallel to the display window. Rotation preview is not supported for multi-channel large stack images, so it is easier to duplicate a representative slice with a single color channel from the original stack and use a “grid line” overlay to guide image alignment.

• 38.
  Count cells (Figure 5A):

  Note: While we use the built-in plugin in ImageJ for manual cell counting, the overall process can be automated using other image analysis tools such as CellProfiler or Imaris. However, accuracy can vary depending on whether the algorithm is optimized for 2D or 3D datasets and on variations in marker intensity across the FOV.

  • a.
    Click on the “Multi-Point Tool” in ImageJ to start counting.
  • b.
    Double-click the “Multi-Point Tool” icon to configure its settings.

    • i.
      Use the drop-down list to assign a different counter number when counting multiple cell types.

      Note: When counting a new image, start with counter type 1 and skip counter type 0. This simplifies interpretation by aligning with standard indexing conventions and prevents default sequential numbering (i.e. 0,1,2,3,4…) in ImageJ. The assignment of numeric counter types is freely user-defined.

    • ii.
      Enable the “Show on all slices” option to overlay labeled points on all composited color channels and z-slices of the image stack.
  • c.
    Classify (co-) stained cells using relevant markers and count bound cells (Figure 5A).

    Note: IEs can be distinguished using a combination of membrane labeling and nuclei staining (e.g., IEs are PKH67+ nuclei+, Figure 5A). Nevertheless, nuclei and bright-field channels are sufficient for IE counting. For instance, cells are stained with large elliptical nuclei whereas parasites are often stained with fainter and multiple dotted nuclei (∼20-fold smaller in size), depending on the parasite stage. Always check one or two slices before and after the “in-focus” plane (luminal surface) for all IE markers. Occasionally, tiny parasite nuclei and membrane label are off by a few optical slices depending on the z-step size. Exclude those non-adhered IEs from counting. Adjust brightness/contrast when encountering parasites with weak nuclei staining or strong signal, usually from overlapping with EC nuclei.

    CRITICAL: Use the bright-field channel to verify cell morphology (e.g., to distinguish between IE or EC debris). Instead of small dotted and fainter parasite nuclei, EC debris is often composed of round-shaped cells with large and strong nuclear staining.

  • d.
    Save the image stack with the marked points of each cell type in TIFF format.

    Note: Save frequently. Most image processing functions in ImageJ are not compatible with “Point Selection.” Save point annotations before clearing them or perform the counting on a duplicated image stack.

• 39.
  Analyze the cell count:

  • a.
    Option 1: Display the spatial distribution of bound cells from the counted image (Figure 5B).

    • i.
      Load the image with counted cells (i.e., labeled points).
    • ii.
      Use Analyze>Measure in ImageJ to obtain a result window containing the xy coordinates of all counted cells.

      Note: The xy coordinates are relative to the origin at the top-left corner by default in ImageJ.

    • iii.
      Export the results as an Excel file.

      Note: Use the FILTER function to select specific cell types from the ‘Counter’ column if multiple cell types are analyzed.

    • iv.
      Create an XY scatter plot in GraphPad to show the spatial distribution of bound cells.

      Note: The use of the xy position data preserves the precise spatial information, which can be adapted to quantify proximity patterns (distance-based quantification) between cell types.

    • v.
      Use the “Multi-Point Selection” tool in ImageJ to obtain the total cell count per biological replicate.

      Note: Double-click the “Multi-Point Tool” symbol to select a counter number from the drop-down list. The total count for each type is displayed below the counter number and can be summarized using a bar chart in GraphPad Prism.

  • b.
    Option 2: Generate a binding heatmap per microvessel region (Figure 5B).

    Note: This ROI-based sub-sampling approach allows quantification of binned cell counts to capture local binding heterogeneity. Use nuclear staining or bright-field to guide ROI selection or define ROIs based on relevant markers and purposes.

    • i.
      Generate ROIs for binned binding quantification using the ROI Manager.

      Note: The microvessel has a 13-by-13 grid layout, which can be divided into a total of 481 straight vessel segments and branching nodes. The branching nature leads to varying flow velocities and shear stresses. To account for distinct flow dynamics, 481 rectangular-shaped ROIs are selected to evaluate binding distribution in high and low flow regions of the grid network. However, to accurately evaluate for flow-dependent binding across a range of wall shear stress and flow rates, create ROIs at the straight vessel segments located along the outermost edge of the network to quantify cell binding as described previously.⁹ The flow characteristics of a 13-by-13 grid network are symmetrical along the diagonal.

    • ii.
      Reload the image with counted cells.
    • iii.
      Extract the cell counts of each counter type within each ROI.

      Note: A publicly available BeanShell script can be used to extract cell counts from each ROI. The script takes the “Number of Point Types” as input and outputs a summary table of cell counts. To access the code, visit: https://forum.image.sc/t/how-to-combine-the-multipoint-tool-and-roi-manager/3075/2.

    • iv.
      Run the BeanShell script.

      Note: To use this code, the order of steps matters. Must complete step 39 b(i) before step 39b(ii). Ensure all labeled cells are registered and displayed on the image after reloading and that all selected ROIs are stored in the “ROI Manager” before running the code. “Number of Point Types” is the largest counter type plus one to account for zero-based indexing (e.g., for a combination of counters 1 and 6, set it to 7 point types).

    • v.
      Save the output table as an Excel file.

      Note: The output summary table contains a column vector of cell counts per counter type (cell type). Each row corresponds to the cell counts from a specific ROI. The rows are ordered according to the ROI index in the ROI Manager.

    • vi.
      Convert the cell count column of the selected cell type into a binned heatmap matrix.

      Note: This step is needed to register the cell count in the order of the ROI Manager to a standard matrix index. Refer to the initial ROI generation sequence for proper cell count mapping. For example, we created a heatmap index vector that maps cell counts from each ROI row to the correct positions in a 25 x 25 heatmap matrix for the 13-by-13 microvessel network, following standard linear matrix indexing in MATLAB.

    • vii.
      Use the ‘flip’ function in MATLAB to correct the orientation of all heatmap matrices.

      Note: All heatmap matrices must be adjusted to maintain the consistent inlet/outlet reservoir direction before performing matrix calculations.

    • viii.
      Perform data processing.

      Note: Unlike distance-based analysis using xy coordinates, binned heatmaps allow quantitative comparison of binding patterns. Compute element-wise averaging of heatmap matrices to generate a spatial map of average binding density across all biological replicates. Alternatively, assess co-localization of two cell types through correlation analysis.

    • ix.
      Visualize the results in GraphPad Prism in a Grouped table.

      Note: If IE-binding variability is high across individual microvessels, refer to troubleshooting 6 and 7. Ensure matrix elements representing the collagen grid contain a count of 0 unless transmigration is expected.

Figure 5.

Overview of binding quantification workflow

(A) Workflow for manual cell counting using the Multi-Point function in ImageJ. Left: Representative fluorescence image of a microvessel FOV consisting of four vessel segments, four branching nodes, and one collagen grid center. PKH67 membrane-labeled IT4var19-IEs (green) are assigned with counter 1 (magenta) as an example. Scale bar indicates 100 μm. Additional distinct counter types can be assigned to each cell population that is classified based on a combination of morphology and marker expression. Right: A zoom-in view showing nuclei of both endothelial cells (large nuclei, ∼15 μm) and parasites (small nuclei, 0.5 μm to 1 μm). Scale bar indicates 20 μm. The zoom-in view at the far-right shows binding of fluorescent PKH67-membrane labeled IEs. P. falciparum-IEs are PKH67+ nuclei+. Zoomed-in box inset shows a FOV of 32.5μm-by-32.5μm.

(B) Two binding quantification approaches. Option 1: Direct visualization of the spatial distribution of bound cells using scatter plot of measured xy positions on the microvessel luminal surface. Option 2: Generation of a binned heatmap to assess binding pattern across the microvessel network. First, eight ROIs are created from the same FOV in panel A, covering four vessel segments (labels #5-8) and four branching nodes (labels #1-4). Next, cell counts are extracted from the selected ROIs. Then, cell counts are converted into a digitized heatmap matrix (0 for matrix area without cell transmigration).

Microvessel barrier permeation quantification workflow

Timing: ∼3 h

Here we present a simplified approach to measure changes in microvessel barrier permeation properties between baseline and a selected experimental endpoint (i.e., 8 h post-IE binding) using imaging data from 70 kDa fluorescent dextran perfusion.

Note: The quantification below takes dextran accumulation in the collagen matrix after permeating across the endothelial barrier over a fixed time interval as a measurement of barrier function and assumes a linear relationship between fluorescence intensity and tracer concentration. An example FOV acquired in step 23 contains four individual collagen grids as technical replicates for measuring local endothelial barrier function (Figure 6). This workflow is applicable to the analysis of other endpoints and tracer sizes. Alternative permeability assay protocols and quantification methods (e.g., for calculating the permeability coefficient) can be adapted to this 3D microvessel system and the time-lapse dataset acquired from step 23.^39,40,41 However, a single, straight microvessel channel geometry or region is ideal for permeability analysis.

• 40.
  Open image pairs at the 2 min and 5 min time point in imageJ (Figure 6):

  • a.
    Load 2 min and 5 min time-lapse image stacks from the same endpoint and the same FOV.
  • b.
    Duplicate the image frames (single slices) of the 2 min and 5 min time points post-dextran perfusion.
• 41.
  Obtain raw intensity profiles (I_raw):

  • a.
    Generate binary masks of the collagen grid (Figure 6):

    • i.
      Duplicate the 2 min image frame.
    • ii.
      Use the “Threshold” function in ImageJ to generate a binary mask.

      Note: Generate separate masks for segmenting the collagen grid from different endpoints (e.g., baseline and 8 h post-IE binding) and different FOVs. The same segmentation can be reused for different time points (e.g., 2 min and 5 min frames from the same FOV), provided that the device position remains fixed between the two acquisitions. Always operate on duplicated raw images to prevent data loss.

  • b.
    Segment the collagen grid (technical replicate):

    • i.
      Use the “Analyze Particles” function on binary masks to add the segmented collagen grids to the “ROI Manager.”

      Note: If leakage at 2 min interferes with threshold-based automated segmentation, refer to the bright-field channel from the background image taken prior to dextran perfusion and manually use the “Rectangle” or “Polygon Selection” tool to draw the estimated collagen grid ROIs. Add ROIs to the “ROI Manager.”

    • ii.
      Exclude ROIs on the edges by setting an appropriate size threshold.
  • c.
    Define measurement windows (sampling ROIs) for each technical replicate (Figure 6):

    • i.
      Extract dimensions (e.g., coordinates and height) of the segmented collagen grid ROI.
    • ii.
      Measure the diameter of the microvessel branch (D_v).
    • iii.
      Generate one sampling ROI with scalable dimensions centered on each technical replicate.

      CRITICAL: A measurement window’s length should span the permeation direction across the tracer-perfused lumen (donor side), the vessel-matrix interface (endothelial barrier of interest), and the collagen matrix (receiver side). A measurement window’s height is a scalable value relative to the microvessel diameter and collagen grid width (e.g., 40% of D_v and 20% of W_c as used in our quantification example, Figure 6). Due to the grid geometry, we define a restricted height to limit interfering signal of dextran permeation from the adjacent microvessel lumens in quantification.

  • d.
    Select the raw dextran fluorescence image pair from step 40.
  • e.
    Obtain intensity measurements (Figure 6):

    • i.
      Remove the segmented collagen grid ROIs from the “ROI Manager.”
    • ii.
      Select all or selected sampling ROIs.
    • iii.
      Extract raw intensity profiles (I_raw) from the 2 min and 5 min time points using the “Multi Plot” function.
    • iv.
      Export or copy the raw intensity data points into an Excel file.
  • f.
    Repeat step 41a-e to obtain I_raw from both baseline and the selected experimental endpoint.
• 42.
  Quantify microvessel barrier permeation:

  • a.
    Calculate the mean peak dextran intensity value (I_lumenpeak) per sampling ROI.

    • i.
      Extract the maximum intensity value from the tracer-perfused microvessel lumen.
    • ii.
      Average the maximum intensity values from the 2 min and 5 min time points.

      Note: Once the flow is established, dextran intensity within the lumen should be stable during the assay session due to continued perfusion and the presence of a large dextran reservoir at the inlet. Signal loss from photobleaching or depletion into the collagen matrix is considered minimal.

  • b.
    Obtain the normalized intensity profile (I_norm) to account for potential perfusion variability across microvessel devices, defined as:

    $$I_{norm}\left(x,t,T\right)=\frac{I_{raw}\left(x,t,T\right)}{I_{lumenpeak}\left(T\right)}$$

    where x refers to the pixel position along the x-axis of a sampling ROI, t refers to the imaging time points (e.g., 2 min or 5 min frames post-dextran perfusion), and T refers to the experimental endpoints (e.g., baseline or 8 h post-IE binding). Both I_raw and I_norm are a N-by-1 intensity profile vectors.
  • c.
    Extract the mean dextran intensity value (I_{norm center}) from a defined collagen region (Δx, Figure S4).

    Note: We calculated the mean dextran intensity value over a region approximately Δx = 65 μm in length (i.e., 40% of D_v) centered one microvessel diameter away from the endothelial barrier of interest (Figure 6; Figure S4). The optimal size of Δx is expected to be proportionally adjusted and scaled to matrix/gel size and concentration, as well as tracer size, and inversely related to the set hydrostatic pressure during dextran perfusion.

  • d.
    Calculate the net dextran accumulation (ΔI_{norm center}) over the fixed time interval for each endpoint (Figure S4), as follows:

    $$\Delta I_{normcenter}\left(T\right)=I_{normcenter}\left(5\mathit{min},T\right)-I_{normcenter}\left(2\mathit{min},T\right)$$

    Note: This simplified approach approximates the net dextran accumulation between two discrete time points and accounts for background fluorescence signal from the same imaging session.

  • e.
    Determine the relative change in dextran accumulation from baseline.

    • i.
      Calculate the difference in net dextran accumulation between two endpoints: ΔI_{norm center}(post-IE binding -baseline).
    • ii.
      Divide the difference by the baseline level and multiply by 100 to assess the relative percentage change in dextran accumulation across two endpoints.

      Note: Assess the relative changes in dextran accumulation using paired measurements from the same FOV to ensure consistent comparison and the same level of flow effects. To calculate the dextran permeation rate based on step 42d, divide the net dextran accumulation value by the defined time interval (i.e., Δt = 5 min – 2 min = 3) for each endpoint.

    • iii.
      Visually verify the quantified relative change in barrier permeation against paired images for consistency.

      Optional: Perform one-to-one comparisons of the relative change in dextran accumulation from individual sampling ROIs with microvessel-IE images to evaluate localized changes in microvessel barrier function associated with IE binding density.

• 43.
  Aggregate data:

  • a.
    Average all technical replicates from one microvessel into one biological replicate.

    Note: Separate data averaging of technical replicates from high-flow and low-flow regions to account for flow-mediated barrier changes.

Figure 6.

Overview of microvessel barrier permeation quantification workflow

Top: Duplicated image frames of the FITC-70 kDa channel acquired at 2 min and 5 min post-dextran perfusion from the same FOV at the 8 h post-IE binding endpoint (Step 40). Bottom left: An auto-thresholded binary mask generated from the 2 min time point image frame was used to segment the collagen grid (Step 41a-b). The obtained collagen grid segmentation can be applied to other acquisition time points from the same FOV and treatment endpoint (e.g., the 5 min frame post-dextran perfusion shown on the right). Bottom middle: A sampling ROI with adjusted dimensions was generated and centered on the collagen grid. The sampling ROI was used to extract the fluorescence intensity profile across the perfused lumen, vessel-matrix interface, and collagen matrix (Step 41c). Bottom right: The “Multi-Plot” function in the ROI Manager was used to obtain raw fluorescence intensity profiles from the selected sampling ROIs at each time point (Step 41d-e). Representative 5 min raw fluorescence intensity profiles are shown from ROI-3 of microvessels at 8 h post-IE or uninfected RBC binding. Scale bars indicate 200 μm.

Expected outcomes

This protocol demonstrates the use of an engineered 3D perfusable human brain microvessel model as a valuable tool for in situ IE binding and maturation, enabling the study of inflammatory processes at the brain endothelial barrier during CM pathogenesis (Figure 1).

Establishing perfusable 3D human brain microvessels in vitro

Perfusable 3D microvessel networks (∼100-150 μm lumen diameter) are fabricated within Type I collagen hydrogels and seeded with primary HBMECs. By 24 h post-seeding, a successfully fabricated microvessel will display a stable 3D endothelial monolayer and be ready for switching to co-culture media. The addition of 5% human serum to co-culture media optimized for parasite growth will also enhance microvessel barrier integrity, as evidenced by improved media retention. In 7.5–10 mg/mL collagen hydrogels, a resting state brain microvessel will exhibit a smooth and continuous luminal surface under bright-field without angiogenic sprouting (Figures 2A and 3B). Significant remodeling or sprouting is likely due to inconsistent collagen density and/or fabrication defects (Refer to troubleshooting 6). After an additional 2–3 days of culture under gravity-driven flow, we expect formation of a mature 3D brain microvessel, characterized by a confluent EC monolayer lining the pre-patterned collagen grid network (Figure 7A). The endothelialized microvessels will remodel from a rectangular to an elliptical lumen as they mature (Figure 7A). Under resting condition, the microvessel should exhibit robust expression of endothelial junctional proteins such as VE-Cadherin (Figure 7A), and negligible expression of the inflammatory activation marker ICAM-1 by immunofluorescence (Figure 7B).

Figure 7.

Immunofluorescence imaging of the 3D microvessel system

(A) Left: Z-projection of confocal imaging of a 3D brain microvessel patterned with 13-by-13 grid network, showing confluent endothelialized lumens with robust vascular endothelial cadherin (VE-cadherin) staining. This is a 6-by-6 stitched-together image. Scale bar indicates 500 μm. Right: HBMECs remodel the collagen to form elliptical vessel lumens as depicted in the cross-sectional view from the zoom-in grid FOV. Scale bar indicates 100 μm.

(B) Immunofluorescence images showing upregulated ICAM-1 surface expression post-18 h TNF-α treatment as compared to the resting state. Actin filaments and cytoskeletal reorganization are visualized by phalloidin staining. Scale bars indicate 100 μm. Figure adapted from Howard et al., Cell Reports 2023.¹

Activating microvessels with inflammatory stimuli implicated in CM pathology

Microvessels can be stimulated by perfusing pathogen-derived factors (e.g., P. falciparum-IEs, secreted parasite products or products released during IE rupture) or host inflammatory factors such as TNF-α, alone or in combination. Gravity-driven perfusion can mimic flow conditions observed in the brain microcirculation (Methods video S2). After 1 h of IE perfusion (using ∼750,000 IEs from inlet reservoirs) and media wash, approximately 8,000-18,000 IEs should remain adhered per microvessel network. The 3D brain microvessel system tolerates high loads of bound IEs, and delamination of microvessel occurs only when excess unbound IEs and uninfected RBCs are not removed during post-perfusion media washes. The grid-patterned microvessel geometry leads to varying wall shear stress across the network and allows comparison of brain endothelial responses to high and low parasite binding density within a single microvessel device.^1,9 Overall, IT4var19-IE should display a flow-dependent binding pattern with greater binding in low-flow, low-shear regions, resulting in an estimated 0.7–2 IEs per HBMEC (Figure 8). Detailed characterization of IE-vessel wall interactions can be found in a previous study.⁹ The number of adherent IEs will also depend on the parasite strain and how recently it was panned on HBMEC monolayers to maintain the parasite binding tropism for brain endothelial cells.¹⁶ In contrast, uninfected RBCs should display negligible binding to 3D microvessels.

Figure 8.

Converting IE cytoadhesion in the 3D microvessel to a binding heatmap

Representative flow-dependent IE binding pattern in the microvessel after 1 h of perfusion followed by a media wash. The branching microvessel network leads to varying flow rates across the grid and differing cell binding levels. Left: Representative image of PKH67 membrane-labeled IT4var19-IEs (cyan) adhered to the microvessel wall, showing higher binding density in regions with lower flow and shear, located off-diagonally from the inlet and outlet. Scale bar indicates 500 μm. Right: Representative quantification of IE binding. Heatmap showing the spatial distribution of average number of bound IEs per vessel segment and branching node (n = 3 biological replicates). Figure adapted from Howard et al., Cell Reports 2023.¹

In situ maturation of parasites

The optimized brain microvessel system supplemented with 5% human serum in the co-culture media supports continued in situ maturation of some adherent IEs life cycle (Figure 9), allowing visualization of the microvessel response through intraerythrocytic maturation and IE rupture. To confirm parasite maturation during microvessel co-culture, Giemsa-stained smears of the media effluent and microscopic imaging of microvessels can be applied to detect brown hemozoin pigment from parasite catabolism of host hemoglobin, segmented nuclei in mature schizonts, and dispersed merozoites following IE rupture (Figure 9). IE maturation can also be verified by ultrastructural imaging, showing firm adhesion by 8 h post-binding and the presence of ghost IE membrane fragments on the EC surface at 24 h post-binding (Figure 4B).

Figure 9.

Optimized co-culture condition for intraerythrocytic maturation of adherent IEs on the microvessel wall

To examine the ability of adherent IEs to undergo maturation in 3D brain microvessels, infected erythrocytes were examined at 1, 8 and 24 h post-binding by making Giemsa-stained effluent smears from the outlet reservoir (left panels) and by bright-field microscopy (right panels, scale bars indicate 10 μm). By 8 h post-binding, brown hemozoin pigment (arrowhead) formed as a byproduct of parasite digestion of host hemoglobin. By 24 h post-binding, free merozoites (black arrowhead) and ruptured IEs (white arrowhead) were detected. Figure adapted from Howard et al., Cell Reports 2023.¹

Microvessel inflammation

Outcomes of microvessel inflammation can be assessed through microscopic imaging, functional assays, and transcriptional profiling as detailed in this protocol. These approaches are designed to capture both phenotypic alterations (e.g., junctional protein expression, surface adhesion molecule expression, and endothelial ultrastructure) and functional changes (e.g., barrier integrity and leukocyte recruitment) of brain microvessels in response to host and/or parasite stimuli. Overall, systemic TNF-α stimulation leads to widespread activation of brain microvessels, whereas adhered IEs induce more localized and limited endothelial responses. Transcriptional changes mediated by host factor and parasite stimuli in the brain microvessels can be assessed using bulk RNA sequencing and downstream bioinformatic analysis. Principal-component analysis of the RNA sequencing data—after removing outliers—should reveal distinct clusters corresponding to each treatment condition.

By integrating temporal kinetics of phenotypic, functional, and transcriptional responses across multiple experimental conditions (endpoint, microvessel stimulation, and perfusate), this 3D microvessel-IE co-culture system provides a robust and adaptable platform for studying CM-associated inflammatory mechanisms. This versatile 3D microvessel platform can also be readily adapted to study vascular inflammation in other disease- or organ-specific contexts.

Limitations

Simplistic model design

This protocol describes a simplified 3D perfusable co-culture system of P. falciparum IE and human brain microvascular endothelium in the vessel lumen patterned in a type I collagen hydrogel. A key limitation of this model is the lack of multicellular interactions and the absence of native brain ECM components that are characteristic of the BBB.

Our model does not include neurovascular cell types such as pericytes and astrocytes, which play critical roles in maintaining BBB homeostasis and modulating inflammatory responses during CM.⁴² Increasing cellular complexity in a 3D brain microvessel model could provide important insights into endothelial-perivascular crosstalk during vascular inflammation. However, incorporating additional cell types presents technical difficulties, particularly in optimizing culture conditions in a shared media channel.

This microvessel-IE co-culture model has been optimized to support full asexual maturation of adherent IT4var19-IEs on the microvessel wall, but other P. falciparum parasite strains may require further optimization for co-culture with primary HBMECs. Additionally, this protocol is designed to model only a single cycle of intraerythrocytic parasite growth. To mimic CM pathophysiology in patients, multiple rounds of trophozoite-stage IE binding and rupture may be required (e.g., successive rounds of IE perfusion). However, this approach has yet to be tested using the 3D microvessel system.

Type I collagen is a widely used natural matrix in vascular engineering. It supports endothelial adhesion and 3D vessel remodeling, but it does not fully recapitulate the biologically relevant endothelial-ECM interactions found at the brain basement membrane. This limitation may affect baseline brain endothelial phenotypes in engineered brain microvessels. Nevertheless, our previous work showed that incorporating human brain vascular pericytes in engineered microvessels promotes basement membrane matrix deposition.¹¹ Others have also demonstrated the feasibility of incorporating decellularized human brain ECM in a brain organoid culture system.⁴³ Development of a brain-specific ECM-based, multicellular, perfusable 3D co-culture model would enhance the biological relevance of the current system. Despite these limitations of a simplistic model design, this experimental system can still address key biological questions related to the parasite-human brain endothelial axis.

Cell types

Our protocol has been optimized for commercially sourced HBMECs and healthy donor PBMCs in short-term co-culture. Consequently, it is not possible to match the donor HBMEC and the donor PBMCs. If the user wishes to use other endothelial cell types, patient-specific cells, or study specific blood cell types, they will need to optimize the co-culture conditions of this protocol. Additionally, it is known that endothelial cells start to acquire phenotypic changes in vitro, thus this system relies on HBMECs in their early passages.

Fabrication variability

Despite the anatomical fidelity of the 3D perfusable microvessel for studying blood-endothelium interaction, minor vessel-to-vessel variability is expected, stemming from fabrication and vessel remodeling. The fabrication of a 3D microvessel model is inherently more complex than a 2D culture, and microvessels can acquire imperfections such as loss of the grid pattern during fabrication. After seeding, issues such as cell detachment or lumen surface retraction from the collagen matrix can also occur. Microvessel attrition from unsuccessful attempts is not uncommon, even for experienced users. Nevertheless, both morphological and transcriptional differences were readily distinguishable between resting and parasite-treated microvessels in our study.¹ For quantification of binding events and functional readouts, rigorous baseline normalization is needed to account for vessel-to-vessel variability.

Troubleshooting

Problem 1

Optimization of media for co-culture of P. falciparum in 3D brain microvessels. Related to Step 11.

Potential solution

Standard complete parasite culture media is based on RPMI 1640 (L-glutamine and 25 mM HEPES) supplemented with human serum (10%), 0.1 M hypoxanthine/1 M NaOH solution, glucose (45%), sodium bicarbonate (7.5%), gentamicin (50 mg/mL), double-distilled water, and 1 M HCl for pH adjustment (final pH of 7.16 to 7.19). While RPMI is commonly used for blood and suspension cultures, it does not fully support endothelial cell function. Thus, we recommend using complete EGM-2MV growth media as the base for microvessel-IE co-culture media. However, P. falciparum parasites can also stall their growth in non-optimal media conditions and not undergo a full round of intraerythrocytic development. To optimize the media for parasite-HBMEC co-culture, we tested various media formulations, evaluating parasite growth (parasitemia and intraerythrocytic development) and microvessel condition (Table 2). We found that 10% human serum is optimal for promoting the full intraerythrocytic parasite growth of the IT4var19 parasite line with RPMI supplemented media. To minimize modification of the microvessel culture condition, we supplemented standard EGM-2MV growth media (5% FBS) with an additional 5% malaria-naïve human serum (pool of 10 donors). This co-culture media supported parasite growth without causing delamination or noticeable morphological changes in the 3D microvessel system. We note that some P. falciparum parasite strains are routinely grown in albumax instead of human serum. Therefore, it may be necessary to tailor co-culture media for different P. falciparum strains. This troubleshooting strategy can be adapted to refine co-culture conditions for other models.

Table 2.

Media formulations tested for optimizing microvessel-IE co-culture

Base media	Exclusion of standard supplement from complete media for parasite culture	Parasite condition	Microvessel condition
EGM-2MV growth media (5% FBS)	(−) 0.1 M hypoxanthine/1 M NaOH solution	Comparable parasite growth to standard culture	N/A
	(−) 10% human serum	Stalled parasite growth	N/A
	(−) 7.5% sodium bicarbonate	Comparable parasite growth to standard culture	N/A
	(−) gentamicin	Comparable parasite growth to standard culture	N/A
	(−) 0.1 M hypoxanthine/1 M NaOH solution, 10% human serum, 7.5% sodium bicarbonate, gentamicin	Stalled parasite growth	Stable
	(−) 0.1 M hypoxanthine/1 M NaOH solution, 7.5% sodium bicarbonate, gentamicin (+) 5% human serum replacing 10% human serum	Comparable parasite growth to standard culture	Stable, matured
RPMI 1640 (L-glutamine and 25 mM HEPES)	(−) 0.1 M hypoxanthine/1 M NaOH solution, 10% human serum, 7.5% sodium bicarbonate, gentamicin	No parasite growth	N/A

Problem 2

Low trophozoite enrichment efficiency. Related to Step 14.

Potential solution

If enrichment yield is low, expand the parasite culture prior to magnetic enrichment. Eluates from multiple columns containing the same parasite strain can be combined into one collection tube. Alternatively, perform sorbitol synchronization or gelatin synchronization of mixed stage culture prior to magnetic enrichment.

Problem 3

Loss of sequestered IE after microvessel fixation. Related to Step 22.

Potential solution

This issue likely arises during the fixation step due to lysis of both uninfected RBCs and IEs by paraformaldehyde. To preserve antigenicity and erythrocyte morphology, fix the microvessels with a solution of 3.7% PFA supplemented with 0.008% glutaraldehyde in 1X DPBS (v/v).³⁵ Note that antigen retrieval may be needed if the fixative interferes with the subsequent immunofluorescence staining.

Problem 4

Perfusion issue with dextran during permeability assays. Related to Step 23.

Potential solution

This issue may be due to bubble formation at the inlet or outlet. First, record x, y, and z positions of the FOV in the acquisition pipeline, then switch to the fluorescence channel at a wavelength of 488 nm for FITC-dextran under live mode. Use a P200 pipette to remove a small amount of dextran solution from the reservoir and gently direct the pipette tip towards the inlet/outlet port to dislodge and remove any bubbles. Once dextran perfusion is visible on the 488 channel, start the 1 min countdown and adjust the z position within that time frame. If the dextran does not perfuse within 1 minute, wash the microvessel with fresh media once to re-establish the flow. The permeability assay may be repeated on the same microvessel, but to ensure optimal endothelial barrier conditions, it is recommended to return the microvessel to the incubator and resume the assay after 2 h at least.

Problem 5

Inability to collect cell lysate from outlet effluent microvessels for bulk RNA sequencing. Related to Step 36.

Potential solution

RNA collection from microvessels within the jig system can be challenging, depending on the integrity and experimental conditions of the microvessels. If total RNA yield via the current protocol proves low, unfixed microvessels can be removed from the acrylic jig (following the same disassembly procedure in step 35b, Figure S3) after a rinse in 1X DPBS. A biopsy punch can be used to extract the collagen construct containing the microvessel network. The microvessel cassette can then be chopped further and digested in RLT lysis buffer to maximize lysate/RNA yield. Given the relatively low endothelial cell fraction in microvessels compared to the surrounding dense collagen, this method is recommended only if direct perfusion-based RNA collection is unsuccessful.

Problem 6

Non-specific IE/uninfected RBC trapping due to microvessel defects. Related to Steps 15 and 17.

Potential solution

Pattern imperfection, lumen deformation, and delamination can alter fluid flow patterns and cause non-specific IE trapping during perfusion studies (Figure 2). For example, trapped bubbles in the lumen can expand over time, leading to pattern imperfection. Channel deformation can occur during device assembly, while lumen retraction may happen if culture conditions change or media flow is obstructed. To avoid these issues, ensure the microvessel pattern on the top collagen piece is properly cast during fabrication and maintained during culture. Additionally, prevent lumen deformation by avoiding collagen debris or swelling during fabrication and by not overtightening the screws. Consistent perfusion should result in fully equilibrated media level between the two reservoirs by 0.5–1 h post-feeding. Microvessel delamination can be minimized by regularly replenishing the media reservoir with fresh, pre-warmed media, and maintaining consistent perfusion. Also, avoid frequent handling of the device to reduce excessive mechanical stress on the microvessel. Refer to Morgan et al., for additional troubleshooting solutions on 3D microvessel fabrication.²⁴ If defects are irreversible (e.g., clogged channels or EC retraction), exclude the affected device from perfusion studies.

Problem 7

Loss of IE binding within 3D brain microvessels. Related to Step 15 and binding quantification.

Potential solution

This issue can arise due to changes in var gene expression during parasite culture or deletion of genes involved in parasite cytoadhesion (e.g., knob-assisted histidine rich protein). In this protocol, we use the IT4var19 parasite line, which was originally selected on human brain endothelial cells and expresses the var19 P. falciparum erythrocyte membrane protein 1 (PfEMP1) variant. To reinforce the brain endothelial cell tropism of this parasite line, it may be necessary to periodically reselect on a brain endothelial cell monolayer. After reselection on brain endothelial cells, it is important to confirm the var gene/PfEMP1 variant expression before perfusion via var transcriptional profiling using qRT-PCR.²⁸ Consistent binding in the brain microvessel relies on the predominant expression of the IT4var19 gene, as the expression of alternative var genes may alter the binding phenotype.¹⁶

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Ying Zheng (yingzy@uw.edu).

Technical contact

Technical questions on executing this protocol should be directed to and will be answered by the technical contact, Ruoqian Hu (ruoqih@uw.edu) and Ying Zheng (yingzy@uw.edu).

Materials availability

This study did not generate new unique reagents. All required reagents are commercially available. The parasite strain utilized in this research, specifically IT4var19, is available from the co-corresponding author, Dr. Joe Smith’s laboratory with a completed materials transfer agreement.

Data and code availability

The above protocol does not report any original code or use of any new datasets. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We would like to thank Dr. Samuel G. Rayner for the blood work and technical insights. We acknowledge the Lynn and Mike Garvey Imaging Laboratory at the Institute of Stem Cell and Regenerative Medicine and the Washington Nanofabrication Facility at the University of Washington. We also acknowledge the Electron Microscope facility and high-throughput sequencing lab at the Fred Hutchinson Cancer Research Institute. This work was supported by National Institutes of Health (NIH) grants AI148802 and AI141602 (to J.D.S. and Y.Z.) and 5T32EB032787-04 (R.H.).

Author contributions

Y.Z. and J.D.S. conceptualized and supervised the study; C.H., F.J., J.S., and Y.Z. designed the original study; C.H., R.H., and F.J. carried out experiments; and C.H. carried out bioinformatic analysis. R.H. and F.J. performed validation. R.H. and Y.J.S. drafted the protocol. R.H. compiled visualization used to explain steps in each figure. All authors contributed to methodology design. All authors contributed to writing and editing the protocol.

Declaration of interests

The authors declare no competing interests.