Movable typing of full‐lumen personalized Vein‐Chips to model cerebral venous sinus thrombosis

Cerebral venous sinus thrombosis (CVST) is a type of stroke associated with COVID‐19 vaccine‐induced immune thrombotic thrombocytopenia. The precise etiology of CVST often remains elusive due to the highly heterogeneous nature of its governing mechanisms, specifically, Virchow's triad that involves altered blood flow, endothelial dysfunction, and hypercoagulability, which varies substantially amongst individuals. Existing diagnostic and monitoring approaches lack the capability to reflect the combination of these patient‐specific thrombotic determinants. In response to this challenge, we introduce a Vein‐Chip platform that recapitulates the CVST vascular anatomy from magnetic resonance venography and the associated hemodynamic flow profile using the “Chinese Movable Type‐like” soft stereolithography technique. The resultant full‐lumen personalized Vein‐Chips, functionalized with endothelial cells, enable in‐vitro thrombosis assays that can elucidate distinct thrombogenic scenarios between normal vascular conditions and those of endothelial dysfunction. The former displayed minimal platelet aggregation and negligible fibrin deposition, while the latter presented significant fibrin extrusion from platelet aggregations. The low‐cost movable typing technique further enhances the potential for commercialization and broader utilization of personalized Vein‐Chips in surgical labs and at‐home monitoring. Future research and development in this direction will pave the way for improved management and prevention of CVST, ultimately benefiting both patients and healthcare systems.


INTRODUCTION
Cerebral venous sinus thrombosis (CVST) is a less common type of stroke that disproportionately affects the younger individuals and can be fatal. [1,2]Whilst recent attention has been drawn to CVST-related strokes in relation to COVID-19 vaccines, [3,4] our understanding of this condition remains limited and largely unexplored, primarily due to its heterogeneous nature among patients. [5]The pathophysiology of thrombosis-related stroke involves the complex interplay of Virchow's triad, encompassing altered blood flow, endothelial dysfunction, and hypercoagulability. [6]These factors exhibit considerable variability between patients, posing significant challenges for accurate diagnosis and surveillance. [1,5,7]Particularly concerning are the blood flow conditions, which existing diagnostic and surveillance strategies struggle to capture effectively. [1]Moreover, the nonspecific clinical manifestations observed in CVST patients further complicate early detection and monitoring. [8]Consequently, the development and availability of suitable animal models or in-vitro platforms for studying CVST remains a challenge.
Clinically, unenhanced magnetic resonance imaging (MRI) and contrast enhanced computerized tomography (CT) angiographic scans are the conventional approaches for diagnosing thrombosis-related stroke. [9]However, these methods often struggle to accurately reconstruct the intricate anatomical details of the cerebral venous sinuses (CVS) due to low contrast. [10]While more detailed images of the sinuses can be obtained with catheter-based angiography, this technique is invasive and not without risks.Magnetic resonance venography (MRV), [11] employing intravenous contrast dye, offers high-resolution visualizations of the veins, becoming more popular for diagnosing CVST and other nonacute thrombosis-related strokes. [10]14] To address these challenges, we drew inspiration from the traditional Chinese movable type methodology, which utilizes interchangeable stamping blocks to print diverse characters and form a newspaper in ancient time. [15]We have leveraged the low-cost movable typing-based soft lithography platform [16] to develop a full-lumen personalized Vein-Chip model, advancing our capacity to study CVST.This innovative Vein-Chip model facilitates the recreation of patient-specific CVS geometries, integrating two halves to accurately represent the full-lumen vessel geometry (Figure 1A).Additionally, our novel injection molding technique successfully achieved a full-lumen Vein-Chip with a thin base of 170 μm, providing excellent compatibility for real-time fluorescent imaging (Figure 1B).When further combined with computational fluidic dynamic (CFD) analysis, endothelialization, and coagulability modulation, we have systematically unveiled the relative contribution of Virchow's triad to CVST Vein-Chips.This model elucidates the mechanisms of fibrin formation and platelet aggregation, pioneering the way toward personalized, patient-specific, in-vitro diagnostic tools for thrombosis-related stroke.

CVS geometry reconstruction and full-lumen Vein-Chip fabrication
Conventional microfluidic fabrication techniques, such as photolithography, have been successful in producing channel models with straight and rectangular cross-sections.These models have been instrumental in the understanding of fluid dynamics and cellular behavior within these systems [17][18][19][20] (Figure 1A).However, generating fully personalized, 3D, closed-channel models using these traditional methods has been proven challenging.While 3D printing technology can form circular cross-sections, [16] most resinbased laser curing printers face limitations that prevent their use in creating full-lumen, personalized vessel-on-chip models (Figure 1A). [21,22]One significant hurdle is the viscosity of the resins, which hinders efficient flushing during the printing process, leading to resin accumulation and channel blockage.To overcome these obstacles, we developed an enhanced movable typing technology that enables cost-effective fabrication of full-lumen, personalized Vein-Chip models.
The sigmoid sinus was implicated in over 56% of CVST cases (Figure 1B). [23]Thus, we used our established algorithm to extract the anatomy of the left sigmoid sinus from an MRV image of a 70-year-old female patient. [16,24]The sigmoid sinus had a longitudinal length of 80 mm and a luminal diameter of 4 mm.We scaled down this luminal diameter to a minimum of 400 μm, a size suitable for stereolithography 3D printing.
To transfer the cerebrovascular geometry into a microfluidic device, the reconstructed sigmoid sinus was divided in two along the middle axial plane.Subsequently, a 30 × 20 × 6 mm block was virtually created followed by a Boolean subtraction to obtain the inner lumen of the vascular geometries for both halves, resulting in negative molds (Figure 1C, ①).Importing these negative mold designs into a stereolithography 3D printer and postprocessing for optical smoothness resulted in the desired molds (Figure 1C, ②).We then used our movable typing method to fabricate the positive molds efficiently (Figure 1C, ③).
To ensure the Vein-Chip was thin enough for imaging by standard fluorescent microscopy systems, we designed four spacers that were 170 μm higher than the vessel lumen of the positive mold.Placing a glass slide on the positive mold created a space into which polydimethylsiloxane (PDMS) was injected to form a thin film microfluidic device that incorporated the vessel geometry (Figure 1C, ④).After injection molding both halves of the thin film microfluidic, the two halves were permanently bonded following oxygen plasma treatment, yielding a full-lumen personalized microfluidic model (Figure 1C, ⑤).We selectively applied saline treatment on the molds for demolding (Supporting Information Figure S1).

F I G U R E 1
The evolution and fabrication of the full-lumen vein-chip using movable typing.(A) An overview of the transition from traditional straight, rectangular-cross-section microfluidic models to our advanced full-lumen Vein-Chip.Note that the idealized model produces a straight channel with a rectangular cross-section due to the restrictions of conventional photolithography.(B) The procedures of deriving a full-lumen Vein-Chip microfluidic device from patient-specific MRV images using an established algorithm for 3D reconstruction. [24](C) The schematics of the advanced movable typing process used for creating a personalized full-lumen Vein-Chip.This includes the step-by-step procedure from printing two halves of the vessel as negative molds, fabricating a positive mold using high-throughput movable typing, to the final step of injecting PDMS between the mold and a glass slide.The distance between the vessel geometry and the glass slide was controlled to be 170 μm by the spacer.Followed by plasma treatment combining the two halves of the lumen microfluidics.color-coded by WSR.For both the MRV scale and microfluidic scale, the input flow rate of 1.08 and 0.303 mL min −1 , respectively, were selected to achieve a bulk shear rate of γ 0 = 150 s −1 .(D) Comparative analysis of mean shear rates across five cross-section planes for both MRV scale and microfluidic scale models.

Vein-Chip fluid flow characterization at microfluidic scale
To characterize fluid flow profiles before and after the scaling down, CFD analyses were conducted following established methods [18,20,25] (Figure 2A).These analyses aimed to map shear rate distributions and velocity streamlines at both the original MRV scale (Figure 2B) and the microfluidic scale (Figure 2C).Besides, we set up our Vein-Chip with a venous flow condition, using a bulk shear rate of γ 0 = 150 s −1 .This physiological shear rate is widely used in modulating endothelial cell (ECs) functionality and subsequent thrombogenic responses. [1,18]t bulk shear rate γ 0 = 150 s −1 , the wall shear rate (WSR) distributions around the stenosis exhibited remarkable similarity between the MRV vessel geometry and the miniaturized Vein-Chip model (Figure 2D).For the MRV scale, the shear rates in the prestenotic and poststenotic cross-sections (Figure 2B, #1 and #5) were measured to be 28.1 and 133.8 s −1 , respectively.Similarly, for the microfluidic scale, these values were 33.0 and 114.0 s −1 (Figure 2C).As the narrowing of the vessel approached approximately 44%, the WSRs rapidly exceeded 280 s −1 for both scales.While there were variations in velocity streamlines observed between the two scales at cross-section #4, the mean shear rates were found to be similar, measuring 97.0 and 88.9 s −1 for the MRV and microfluidic scales, respectively.
The comparable mean shear rates between the microfluidic and MRV scales highlight the validity of our model.Moreover, the similar pattern of shear rate ranges observed at both scales captures dynamic changes across cross-sections 1 through 5.This consistency suggests that the microfluidic Vein-Chip model accurately recapitulates the fluid dynamics of the original MRV vessel geometry, validating its utility for studying thrombotic processes.

Vein-Chip endothelialization and proinflammatory injury modelling
To mimic the human vessel wall and its inflammatory state, we functionalized the Vein-Chip by seeding human umbilical ECs (HUVECs) within the microfluidic channels, as detailed in Section 4.4.One hour postseeding, we inverted the Vein-Chip to encourage ECs growth across the entire channel (Figure 3A).Overnight incubation led to the formation of a confluent EC monolayer (Figure 3A).
Confocal microscopy analysis verified the successful attachment and alignment of HUVECs as endothelium throughout the entire microfluidic channel (Figure 3B).It is important to note that due to the refractive index difference between the channel media and the surrounding PDMS material, the confocal microscope could not scan the complete 3D structure of the channel in a single pass.To address this limitation, we initially scanned the Vein-Chip's bottom using the confocal microscope and then flipped the chip to scan the ceiling (Figure 3A).We then 3D-stitched these to form a complete channel.This approach allowed comprehensive imaging of the endothelial monolayer within the Vein-Chip.The uniform CD31 expression at the integral endothelial junction further confirmed successful Vein-Chip functionalization in a healthy state (Figure 4B).
Inflammation within the vascular wall is recognized as a factor in endothelial activation, barrier function disruption, and increased susceptibility to venous thrombosis in vivo. [26,27]To replicate this inflammatory milieu, we subjected endothelialized Vein-Chips to short-term exposure of phorbol-12-myristate-13-acetate (PMA) for 1.5 h, thereby creating an inflamed vasculature within the microfluidic system (Figure 3A).To understand the impact of endothelial injury, we examined the expression levels of two inflammatory markers, intercellular adhesion molecule-1 (ICAM-1) and E-selectin, under both nontreated and PMA-treated conditions (Figure 3C).Comparative analysis revealed a significant upregulation in ICAM-1 expression on the ECs, exhibiting a remarkable 6.0-fold increase compared to nontreated conditions (Figure 3D).Similarly, E-selectin expression showed a substantial 3.8-fold elevation in response to the inflammatory stimuli (Figure 3D).

Whole blood perfusion in the CVS Vein-Chip for thrombotic analysis
To model venous thrombosis, we employed citrated human blood that was recalcified with a 6 mM CaCl 2 dose and subsequently perfused through the endothelialized Vein-Chips at a physiological venous bulk shear rate of γ 0 = 150 s −1 for a duration of 10 min (Figure 4A).To evaluate the thrombotic response within the microfluidic system, we employed confocal microscopy to visualize the immunostained platelet aggregates (Figure 4B, white) and fibrin formation (Figure 4B, magenta) under both nontreated and PMA-treated conditions in real-time (Supporting Information Video S1).Platelets and fibrin are crucial components in the complex coagulation cascade.Upon vascular injury, platelets adhere to the exposed subendothelial matrix and become activated and aggregate to form the primary hemostatic plug. [28]Simultaneously, the coagulation cascade is initiated, leading to the conversion of fibrinogen into fibrin by the action of thrombin, forming a meshwork around the platelet plug. [29]Given their central roles in the coagulation process, both platelets and fibrin are often used as indicators for the evaluation of coagulation. [30]ntriguingly, under the nontreated condition, minimal fibrin deposition was observed on the endothelium throughout the experimental duration (Figure 4B, row 1).In stark contrast, the introduction of PMA resulted in rapid and dynamic fibrin formation, exhibiting excessive growth between 5 and 10 min (Figure 5B, rows 2 and 4).Notably, fibrin deposition pre-dominantly occurred in areas where platelet aggregation was detected (Figure 4B).This observation suggests that platelet adhesion onto the endothelium serves as a trigger for fibrin formation under venous flow conditions, aligning with findings from other venous thrombosis models. [31]o quantitatively compare the patterns of fibrin formation and platelet aggregation under both nontreated and PMA-treated conditions, we analyzed the fluorescence intensity changes over time.Under the nontreated condition, both fibrin and platelet signals exhibited minimal fluctuations, remaining consistantly low throughout the experiment (Figure 4C and D).Conversely, under the PMA-treated condition, fibrin demonstrated a linear increase in fluorescence intensity from 0 to 10 min, while platelet signal displayed an initial rise within the first 3 min, followed by a stable signal from 3 to 10 min (Figure 4C and D).][34] Specifically, our results suggest that CVST is characterized by a prominent presence of fibrin and appears to be independent of biomechanical platelet aggregation. [17]

Investigate the thrombotic response of the endothelial injury at the Vein-Chip vascular bed
To further investigate the thrombotic response to vessel wall injury, we conducted 3D analyses using a confocal microscope (Figure 5A).A 3D scan through a cylindrical channel presents a unique set of challenges, primarily due to the refractive index differences between materials which prevent light from fully penetrating the channel. [35,36]Consequently, we chose to use the bottom side of the channel as a representative of the whole, providing detailed 3D data that corroborated the real-time imaging findings.To validate our approach, we injected PDMS fluid into one of the PMAtreated channels to equalize the refractive index across the entire microfluidic chip, thereby enabling a comprehensive 3D scan (Figure 5A).This validates that platelets and fibrin adhere uniformly to the top, bottom, and vertical aspects of the PDMS channel.Under the nontreated condition, minimal platelet aggregation and negligible fibrin deposition were observed in a 3D representation (Figure 5B, row 1).In contrast, the PMA-treated condition exhibited distinct features, with platelet aggregates forming within the lumen and fibrin exhibiting a fibrous structure that extruded from the platelet aggregations (Figure 5B, row 2).
To quantitatively assess the disparities in platelet aggregation and fibrin formation between the nontreated and PMA-treated conditions, we performed 3D reconstruction of the platelets and fibrin structures (Figure 5C).The volumes of platelets and fibrin were then calculated based on the reconstructed 3D data.The results demonstrated a significant increase in fibrin volume, with a fold change of 78.7, rising from 0.10 to approximately 80 million μm 3 .However, the volume of platelets only experienced a modest increase, growing from 0.45 to 2.42 million μm 3 (Figure 5D).These findings indicate that vessel wall injury predominantly influences fibrin formation, playing a pivotal role in triggering venous thrombosis.

DISCUSSION
In this study, we introduced a pioneering microfluidic Vein-Chip model, which faithfully emulates the fluid dynamics, cellular interactions, and thrombotic processes of native veins.The innovation of our work resides in the successful creation of full-lumen personalized microfluidic chip from patient-specific clinical data, along with enhanced imaging capabilities.Specifically, this advancement is achieved via the use of cutting-edge injection molding, precise alignment, and robust binding techniques.Compared to the previous semicircular chip, we significantly improved the reliability of thrombotic testing (Figure 5), which allows for determining potential "hot spots" of thrombosis in a 3D fashion.Despite the availability of sophisticated imaging techniques, diagnosing conditions like CVST and its associated stroke still poses significant challenges.The confirmation of CVST is often delayed, leading to inadequate treatment for patients.This situation underscores the need for alternative diagnostic approaches that are accurate, efficient, and readily accessible.The Vein-Chip model, fabricated using an upgraded movable typing method, addresses this need.By combining 3D stereolithography and PDMS injection techniques, we created microfluidic channels that accurately recapitulate patient-specific cerebral vessel sinus geometries.This advancement represents a significant leap forward in the organ-on-chip field, facilitating the creation of highly accurate and personalized models for CVST diagnosis and study.The utilization of movable typing enhances the cost-effectiveness and efficiency of Vein-Chip production, presenting a viable alternative to current diagnostic methods.Our innovative Vein-Chip model allows for a more nuanced investigation into the impact of hemodynamics on thrombosis and its progression, by faithfully representing patient-specific vasculatures.Importantly, our approach holds distinct advantages over traditional in vitro models, as these typically exhibit rectangular cross-sectional areas, having significantly different shear profiles when compared to circular vessels in vivo. [31,37,38]While embedded [39] and coaxial printing [40,41] techniques can generate micronsized blood vessels, replicating patient-specific geometries remains a challenge.Additionally, the substantial costs and single-use nature of these devices present further hurdles.On the other hand, our model makes use of readily available clinical imaging data like CT and MRI scans to generate personalized microfluidic devices.These devices are made from the durable material PDMS, allowing for repeated usage and, therefore, promoting both sustainability and cost-effectiveness.This transformative approach opens new pathways not only for research settings but also holds promise for future commercialization, given its potential for scalability and adaptability.
The Vein-Chip model also presents an unprecedented level of control over the cellular environment within the microfluidic channels.Using HUVEC, we simulated the vessel wall's endothelial layer, enhancing the model's physiological relevance.Furthermore, the capability to induce an inflammatory state within the endothelialized Vein-Chip model has provided a mechanism to replicate the increased susceptibility to venous thrombosis observed under inflammatory conditions in vivo.In terms of thrombosis modeling, the Vein-Chip demonstrated robust accuracy.We observed dynamic fibrin formation and platelet aggregation under venous flow conditions, mirroring in vivo thrombotic processes.Notably, the patterns of these processes closely mirrored those observed in patient-derived vascular geometries, underscoring the physiological relevance of the model.These observations suggest that CVST could be classified as a hypercoagulable state, where fibrin appears to be a major player.While platelets are crucial in the initial stages of thrombus formation, their biomechanical aggregation might not be as pivotal in the progression of CVST.Further investigations are needed to fully reveal the complex interplay of coagulation factors and platelet dynamics in different vascular thrombotic conditions.It is worth noting that HUVECs might not fully recapitulate the intricate properties of ECs in the CVS.
Looking forward, the low-cost nature of our fabrication method, facilitated by the movable typing technology, offers possibilities for commercialization and widespread adoption of personalized Vein-Chips.Our novel movable typing biofabrication can produce a single PDMS chip to approximately $10, factoring in machine use ($1), material costs ($4), and cells ($5).These could serve as valuable tools in surgical labs for preoperative planning and simulation, or even for at-home usage for individuals at risk of CVST to monitor their condition and take preventive measures.The commercialization of such personalized devices holds the potential to significantly improve patient care and outcomes while reducing healthcare costs associated with CVST management.
While our study focused on venous thrombosis, the principles and methodologies employed in the development of the Vein-Chip can be extended to model other vascular diseases such as atherosclerosis, aneurysms, and vasculitis.While fibrin is a critical protein involved in CVST coagulation and thrombus development, other proteins, such as von Willebrand Factor (VWF) and collagen, may substantially influence the pathophysiology of thrombogenesis.Our future research directions could examine VWF and collagen under high shear stress conditions for arterial thrombosis.This broad applicability, combined with the model's personalized nature, underscores the potential of the Vein-Chip as a valuable tool in the realms of biomedical research and personalized medicine.In conclusion, our innovative Vein-Chip model represents a significant advancement in microfluidic vascular modeling.It offers potential in understanding the complex interplay of fluid dynamics, cellular behavior, and pathological processes in the vascular system.It also sets the stage for personalized drug testing and the development of patient-specific therapeutic strategies.

Acquisition and processing of clinical MRV images
Clinical MRV data for a patient were obtained using a Philips Medical Systems (Achieva) as described previously. [16]The data acquisition followed the standard consent procedure.A dual-echo MRV pulse sequence was employed, with a readout bandwidth of ±15.6 kHz and a flip angle of 20 • .The matrix size for MRV data was 209 × 239 × 239 (k x × k y × k z ), with a field of view of 240 mm.The slice thickness was set at 1.4 mm, and a slice increment of 0.7 mm was applied.To optimize the echo times (TE), a 58.6% partial echo was utilized to decrease TE1 in the first echo and maximize TE2 in the second echo.
Postacquisition, images were reconstructed and further processed using Mimics Research 21.0 to isolate the relevant vessel geometry.0on a personal computer equipped with a 3.40 GHz CPU and 8 GB memory.To mitigate artifacts caused by signal wrap-around along the slice direction, the reconstructed area was manually segmented to isolate the vessel geometry relevant to this study.Subsequently, a venogram was generated by implementing a minimum intensity projection along the slice direction.

Fabrication of the Vein-Chip via 3D printing and movable typing
The blood vessel's standard triangle language (STL) file was obtained by extracting the relevant information from the raw MRV images of the sigmoid sinus (Figure 1B).The DICOM viewer was utilized to directly convert the image points into STL format for digital reconstruction.The converted STL file was further processed using SpaceClaim software (ANSYS Inc. 2020) to create curvature-based continuous surfaces, allowing for manual adjustments to refine the vessel's details (Figure 1B).The mold incorporating the vessel's features was then 3D printed using a stereolithography 3D printer (Form 3B; Formlabs) with "Clear V4" resin at a resolution of 25 μm.Postprint treatment involved subjecting the printed mold to UV light in a Formlabs UV light oven at room temperature for 240 min.
The postprocessing of the negative molds followed the aforementioned method. [16]Initially, the surface of the mold was filed using 1200-, 2500-, and 8000-grit sandpapers, controlling the filing depth by a specially designed rectangular trench incorporated in the mold.Finally, the surface was meticulously polished to attain a mirror-smooth finish using an acrylic polish paste and a polishing cotton wheel.During the polishing process, we engineered a safeguard to ensure the membrane structure remains uncompromised. [16]Specifically, the depth of the vessel lumen was set to be 100 μm more than its intended final depth.The polishing spacer, which mirrors this depth, acts as an effective guide during the polishing phase.When the polishing spacer was completely removed, it indicates that the vessel lumen has reached its predestined depth.Once all the polishing procedures were completed, the mold underwent cleaning in an ultrasonic cleaner to eliminate any residual polishing paste.Subsequently, the mold was dried using compressed air.
For high-throughput movable typing, the polished chips obtained from the previous stage were assembled into a large plate and treated with silane vapor (tridecafluoro 1,1,2,2tetrahydrooctyle-1-trichlorosilane; Sigma-Aldrich).To create the positive mold, a mixture of PDMS (Sylgard® 184 by Dow Corning) and the curing agent was prepared at a 10:1 ratio (w/w).The PDMS mixture was poured onto the assembled plate and heated in an oven at 60 • C for 4 h to facilitate solidification.Once solidified, the cured PDMS was peeled off from the plate and treated with silane vapor for further processing.

PDMS injection and Vein-Chip casting
To cast the Vein-Chips containing the desired microfluidic channels, we devised four spacers that were precisely 170 μm taller than the vessel lumen of the positive mold.Subsequently, a glass slide was positioned on top of the positive mold, effectively generating a narrow space between them.The microfluidic device was fabricated by injecting a PDMS mixture into this confined space, enabling formation of a thin film microfluidic device that seamlessly incorporated the vessel geometry.The PDMS-filled structure was then subjected to a heat treatment in an oven maintained at a temperature of 60 • C for a duration of 4 h.Once the PDMS solidified, it was carefully detached from the positive mold, while remaining adhered to the glass slide.Subsequently, the other half of the thin film microfluidic device was aligned with the first half, and the two halves were permanently bonded together following a plasma treatment of 3 min.Last, to facilitate static and flow experiments, the PDMS chips were punctured with biopsy punchers of diameters Ø6 mm and Ø1 mm for the creation of inlets and outlets, respectively.These modifications allowed for the successful implementation of both static and flow-based investigations within the microfluidic device.

Computational fluid dynamics simulation
CFD was conducted utilizing the ANSYS® Fluent 2021 R1 software, a widely used commercial platform for fluid dynamics analysis (version 21.1).Considering the unique and irregular shape of the Cerebral Venous Sinus, a tetrahedral dominant meshing technique was employed, with a maximum element size set at 50 μm.To ensure a smooth transition, an inflation layer was implemented throughout the entire body, with the vessel walls serving as the boundaries.The fluid flow was assumed to be steady and laminar, while the fluid itself was characterized as Newtonian with constant properties.To achieve a bulk shear rate γ 0 = 150 s −1 , a flow rate of 190 μL min −1 was selected.Finally, edge sizing conditions were applied to ensure 50 elemental divisions for the left and right edges stemming from the inlets and outlets, allowing for a more accurate representation of the flow behavior within the system.

Endothelialization of CVS Vein-Chips
The human umbilical ECs used in this study were obtained from Thermo Fisher Scientific and cultured in EGM-2 medium (EGM™−2 BulletKit™, Lonza), following the published methods. [16]Once the cells reached 80-90% confluency, they were washed with phosphate-buffered saline (PBS, ThermoFisher) and trypsinized using trypsin and EDTA solution (ThermoFisher).After centrifugation, the HUVECs were resuspended in EGM-2 medium at a seeding density of approximately 5 × 10 6 cells mL −1 .Cell density lower than 5 × 10 6 cells mL −1 led to nonconfluent monolayer, while higher density led to cell aggregates formed in the channel.The microfluidic channels were sterilized and coated with 100 μg mL −1 human plasma fibronectin (Thermo Fisher) and incubated at 37 • C for 1 h.The channels were then rinsed twice with PBS, and an 8 μL suspension of prepared HUVECs was injected from the outlet into the microchannel.The microfluidic chip was immediately flipped upside down to allow the HUVECs to attach to the dome of the microchannel for 1 h.Subsequently, the chip was flipped again to enable the attachment of HUVECs to the bottom surface for 1 h.A two-step seeding combined with flipping method overcame the incomplete cell coverage due to gravity and allowed cell attachment to both the top and bottom surfaces of the circular channel.Following this, EGM-2 medium was added to the inlet reservoir to culture the HUVECs statically overnight, completing the process of endothelialization of the microfluidic chip.
To investigate the effect of EC inflammation on platelet and fibrin deposition, the Vein-Chips were activated with PMA at a concentration of 50 ng mL −1 in serum-free medium EBM-2 for 1.5 h under static conditions.After activation, the Vein-Chips were washed twice with complete medium EGM-2 to remove any excess PMA.

Immunofluorescence staining within microfluidic Vein-Chips
Following the treatment with PMA, the microfluidic devices were fixed using a 4% paraformaldehyde solution.Subsequently, a blocking step was performed for 1 h in 5% bovine serum albumin (BSA) in PBS.The microfluidic chips were then incubated overnight at 4 • C with a primary antibody solution containing 2% BSA.Following the primary antibody incubation, secondary antibodies were applied at room temperature for 1-2 h in 2% BSA and PBS.Nuclei staining was performed for 10 min using PBS alone.The E-selectin (CD62E) and ICAM-1 (CD54) immunostaining was then performed as previously described. [16]

Blood collection
All procedures involving blood collection from healthy donors were approved by the University of Sydney Human

F I G U R E 2
Profiling wall shear rate (WSR) in MRV vessel geometry and Vein-Chip model.(A) Illustration of the ANSYS finite volume meshing scheme, used for computational fluid dynamics (CFD) analysis of the sigmoid sinus geometry, containing ∼800,000 tetrahedral elements.(B and C) CFD fluid streamlines and five cross-section planes in the MRV scale model (b; scale bar = 1000 μm) versus microfluidic scale Vein-Chip model (c; scale bar = 400 μm),

F I G U R E 3
Endothelialization and inducing proinflammatory injury in the Vein-Chips.(A) Procedure of seeding HUVECs in a bare Vein-Chip and treating with PMA to emulate the inflammatory state.(B) Confocal microscopy and 3D image stitching of functional endothelium lined inside a microfluidic channel.Anti-CD31 (green) showed endothelialization of the entire microfluidic channel.Scale bar = 200 μm.(C) Representative immunostaining images of ICAM-1 (yellow) and E-selectin (red) expression in the absence (−) or presence (+) of PMA stimulation.Scale bar = 200 μm.(D) The fluorescent intensity of ICAM-1 and E-selectin in the absence (−) or presence (+) of PMA stimulation.****p < 0.0001, assessed by unpaired, two-tailed Student's t-test with ROI randomizer n = 50 iterations.

F
I G U R E 4 Recalcified whole blood perfusion in the Vein-Chip for thrombotic analysis.(A) The Vein-Chips were subjected to perfusion with recalcified citrated whole blood at a bulk shear rate γ 0 = 150 s −1 for a duration of 10 min.Real-time confocal imaging was performed using a 10× objective at the bottom of the Vein-Chip.(B) Representative confocal images illustrating the non-treated condition (Row 1) and the PMA-treated condition (Rows 2, 3, and 4) at the vessel bed (green).Platelet aggregation (Row 2, white) and fibrin formation (Row 3, magenta) were examined specifically in the PMA-treated condition.The flow direction is from left to right.The displayed images represent n ≥ 3 independent experiments, with platelet and fibrin signal images captured at 3-, 5-, 8-, and 10-min time points under a 6 mM CaCl 2 dose.Scale bar = 200 μm.(C and D) Time-lapse quantification of fibrin formation (c) and platelet aggregation (d) in the Vein-Chip, covering the 0-to 10-min duration.The presented data represents the mean ± SEM of n ≥ 3 independent experiments in duplicate.

F I G U R E 5
Thrombotic response at the vascular bed of a Vein-Chip.(A) 3D confocal microscopy illustrates homogenous deposition of platelets and fibrins across the Vein-Chip endothelium.Scale bar = 150 μm.(B) Representative 3D confocal images depicting the nontreated condition (Row 1) and the PMA-treated condition (Row 2).Platelet aggregation (Column 1, white) and fibrin formation (Column 2, magenta) were examined.Scale bar = 200 μm.(C) 3D reconstruction of fibrin (magenta) and platelets (white) in the Vein-Chip, providing a visual representation of their spatial distribution.Scale bar = 150 μm.(D) Quantification of 3D volumes for reconstructed fibrin and platelets, enabling a quantitative assessment of their respective volume changes in response to the thrombotic process.All data presents mean ± SEM of n ≥ 3 independent experiments in duplicate.***p < 0.001, assessed by unpaired, two-tailed Student's t-test.