Endothelialized flow models for blood transfusion research

Introduction

Recent advances in cell culture and microfabrication technologies have enabled the development of perfusable endothelialized channels in vitro. To date, these techniques have primarily been applied to tissue engineering research. However, this set-up provides the unique opportunity to simulate blood product transfusion in a cost-effective, robust and reproducible manner – incorporating blood, endothelial and flow components in one unit. This Perspective describes the value of vascular models in transfusion research and discusses key decision points in the design process.

What happens to blood after transfusion?

On transfusion, blood products interact with blood cells and plasma components to alter platelet activation, leukocyte function and red blood cell (RBC) oxygen-carrying capacity. The pro-/anti-inflammatory balance hinges on whether neutrophil or macrophage responses dominate after blood product transfusion.31 These responses are dependent on the activation state of recipient neutrophils and macrophages which, in turn, is influenced by cytokines in the local microenvironment. RBCs and platelets modify immune system function by activating complement, releasing cytokines and participating in receptor-ligand interactions.4

Blood products and recipient blood are encased by endothelium in blood vessels, one of the largest organs in the body with a surface area of 350-1000 m5 The endothelium conveys blood to tissues, provides a surface that prevents improper clotting and cellular activation, acts as a selective barrier to macromolecule extravasation and regulates microvascular blood flow.6 Activated endothelium participates in inflammation by releasing chemotactic molecules (e.g., interleukin-8 and monocyte chemoattractant protein-1), generating reactive oxygen species and expressing adhesion molecules (CD62, CD106, CD54, CD31) to attract leukocytes and facilitate leukocyte transmigration.87 Furthermore, activated endothelium also enhances thrombosis by elaborating procoagulant surface molecules (von Willebrand factor, tissue factor) and microparticles.8 Endothelial dysfunction has been implicated in transfusion-related acute lung injury, sepsis and multiple organ dysfunction.98

The mixture of blood product and recipient blood is constantly mixed and propelled by the cardiac cycle – maximizing cellular interactions and minimizing inappropriate endothelial adhesion.10 After leaving the heart, blood flows through arteries to reach capillaries in the tissues and then veins before returning to the heart. The three types of blood vessels differ in structure, diameter, flow patterns and shear stress.10 In arteries and veins, RBCs and leukocytes flow in the center of the flow stream and platelets are distributed to the periphery of the stream.11 RBCs exhibit a parabolic velocity profile with shear-dependent rotation which continuously mixes the blood components. The configuration of cells in the flow stream can be modified by RBC plasticity, shear rate and fluid viscosity.1312 In the microvasculature, cells travel in single file with uniform distribution of platelets, RBCs and leukocytes in the flow stream.

Additionally, blood flow exerts shear stress on endothelium thereby altering endothelial gene expression, apoptosis, migration, permeability and alignment.146 Endothelial cells cultured under flow conditions demonstrate enhanced barrier function in conjunction with minimal adhesion molecule activation.1615 Physiological shear stress is protective against inappropriate endothelial cytokine release compared to low shear stress.9 Additionally, flow patterns (e.g. continuous versus pulsatile) influence endothelial adhesion protein expression, structure and alignment.6 Lastly, shear stress modifies endothelial interactions with blood cells. For example, monocytes perfused over endothelium activated by tumor necrosis factor-α expressed more tissue factor and CD11b compared to monocytes co-incubated with activated endothelium under static conditions.9

It follows that intravascular transfusion events are affected by blood product-recipient blood interactions and blood mixture-endothelium interactions under flow conditions. Simulation of the post-transfusion intravascular mileau in vitro requires recipient whole blood, blood product, endothelium and perfusate flow. To date, the majority of models used to test transfusion effects have involved static models in which blood products are co-incubated with specific cells such as neutrophils,21171 macrophages,3 platelets22 or endothelial cells23 to examine interactions between two cell types. While blood co-incubation experiments enable the investigation of multiple cellular interactions,242 they do not recapitulate the frequency and type of cellular interactions that occur under endogenous flow conditions. In vivo models are the “gold standard” for capturing the sum of intravascular interactions that occur after transfusion of stored, packed RBCs. However, such models are limited by increased variability (requiring larger sample sizes), reduced capacity to isolate parameters and longer set-up times. Endothelialized in vitro flow models circumvent these limitations by using cell culture methods, packed RBCs and fresh whole blood, which are easier to acquire with fewer associated ethical considerations, shorter model development time and lower cost.

What can we do with current manufacturing technology?

The potential methods to investigate blood mixture-endothelial interactions and current research outputs from these models are summarized in Table 1. Flow models can be divided into planar, microvascular and macrovascular depending on the arrangement of endothelial and blood cells. For planar models, endothelial cells were cultured on coverslips and subjected to various shear stresses generated by laminar flow.2725 These models were used to study blood-endothelial adhesion under macrovascular flow conditions with shear stresses of 0.3-10 dyn/cm2725 Notably, these models do not recapitulate macrovascular flow stream configurations (rectangular flow stream versus circular flow stream) and do not test physiological hematocrits (<10% whole blood/packed RBC suspensions versus undiluted blood).2725 The ex vivo artery model recapitulated physiological flow stream configurations by connecting human umbilical arteries in a perfusable circuit.28 However, the use of harvested arteries meant that the model was susceptible to inter-donor variation and ethical concerns. Synthetic arterial models, produced by seeding cells on a biodegradable tubular scaffold, can overcome these issues.29 To date, this model has not been perfused with blood to investigate blood-endothelial interactions.

Table 1.Available models for testing blood-endothelial interactions in vitro.

Microvascular models involve the formation of <200 μm channels in polydimethylsiloxane, subendothelium or hydrogels.31309 These channels are available in different branching geometries and can be lined by endothelial cells.3231 Acellular models can accommodate smaller vessel diameters (5-70 μm) and higher hematocrits (up to 0.4) without blockage.30 In contrast, endothelialized microvascular models are limited by vessel diameters of 50-200 μm with 10-20% (v/v) packed RBC perfusates.319 Vessels less than 50 μm in diameter are difficult to produce in hydrogels, while vessels greater than 200 μm present issues during endothelial seeding and culture which preclude the formation of a confluent monolayer. Manufacturing limitations associated with vessel diameter can be abrogated by using in vivo microvascular models such as muscle or dorsal skin flaps. However, these models lack the channel diameter and geometrical consistency resulting from in vitro-manufacturing techniques. The presence of endothelial cells in microvascular models enhances thrombosis, thereby hindering the hematocrits that can be tested. Physiological hematocrits are important for viscosity, shear stress and cell-to-cell interactions. For example, an increased hematocrit leads to more RBC-endothelial interactions which enhance shear stress at low flow rates. Furthermore, an increased hematocrit has been associated with altered RBC arrangement in the flow stream and margination of blood components.33

Designing vascular models

Perfusable in vitro macrovascular and microvascular models are composed of an endothelialized channel in a hydrogel scaffold. The unit is subsequently connected to a circuit and perfused with the desired “test” fluid. There are five key decision points in vascular model design (Figure 1): (i) channel geometry, (ii) scaffold moulding, (iii) endothelial cell seeding, (iv) circuit construction, and (v) perfusate selection.

Figure 1.Five key decision points in vascular model design. The five key factors are channel geometry, hydrogel scaffold, endothelial cells, circuit design and perfusate mixture. Branched channels less than 200 μm in diameter are produced using photolithography and soft lithography while straight channels greater than 120 μm in diameter are produced using rod/needle casting. The hydrogel scaffold can be produced in one or two pieces and may be seeded with perivascular cells. Endothelial cells are subsequently seeded onto the channels and cultured under flow conditions to form a confluent monolayer. The vascular model is then connected to a circuit with a pump with or without a gas trap. The last decision involves choosing the type of perfusate to be circulated through the circuit. CP: cryoprecipitate; EDTA: ethylenediamine-tetraacetic acid; FFP: fresh-frozen plasma; HAECs: human aortic endothelial cells; HDMECs: human dermal microvascular endothelial cells; HLMECs: human lung microvascular endothelial cells; HUVECs: human umbilical vein endothelial cells; PDMS: polydimethylsiloxane; Plt: platelet concentrates; PRBC: packed red blood cells; PSC: pluripotent stem cells.

The choice of channel size and shape determines the scaffold manufacturing method. Straight channels with diameters down to 120 μm can be injection-molded using rods, needles or wires stabilized on a mount. This method is cost-effective and technically easy to construct but has limited fidelity for endogenous vascular geometry. Branched channel geometries with diameters of 50-200 μm can be formed using a mixture of photolithography and soft lithography methods. Photolithography is used to create a hard negative-profile wafer. Complex channel geometries can be etched onto a photomask which is used to develop the desired pattern on a photoresist-coated crystalline silicon wafer.34 Notably, photo-lithography requires microfabrication facilities (e.g. clean room, photo pattern generator, spin coater) and specialized consumables (e.g. crystalline silicon wafer, photoresist, developer). The negative-profile wafer becomes the mold for the soft positive-profile stamp used to shape the hydrogel scaffold by soft lithography.31 The soft positive-profile stamp can be formed from polydimethylsiloxane in any standard laboratory using a vacuum degasser and oven. Polydimethylsiloxane can be printed directly in three dimensions: however, this method has inferior resolution compared to the photo/soft-lithography combination (~760 μm versus 50 μm, resolution limited by soft lithography).35

The hydrogel scaffold can be constructed from collagen, alginates, agarose, poly(ethylene glycol) dimethacrylate or methacrylated gelatin.5 These materials are chosen for their transparency, absence of toxicity, fidelity for micropatterning and transport properties. Perivascular and/or tissue cells can be incorporated into the hydrogel to enhance biological approximation. The hydrogel is subsequently injection-molded onto a stamp and allowed to solidify at specified temperatures. Straight channel scaffolds can be formed as one piece. Scaffolds involving the use of a stamp are formed as two pieces (stamped pieced, flat slab), which are subsequently combined to form closed channels. The key obstacles at this stage are bubble formation during injection molding and channel damage during scaffold assembly.

The next decision involves the type of endothelial cells to use to seed the channels. Commercially available multi-donor human umbilical vein endothelial cells are the most commonly used cell type due to their availability, robustness and proliferation consistency. However, multi-donor cells can express multiple antigenic profiles, complicating donor-recipient cross-matching in transfusion simulations. Furthermore, human umbilical vein endothelial cells may behave differently from other primary endothelial cells. Vessel-specific primary endothelial cells such as human lung microvascular endothelial cells and human aortic endothelial cells may improve physiological relevance but these cells are often more difficult to culture and susceptible to variability. Endothelial derivatives from human pluripotent stem cells can also be used.36 The desired endothelial cells can be seeded on the channel surface under static or low flow conditions and grown to confluence under laminar flow. Notably, high seeding flow rates will prevent endothelial cells from attaching to the hydrogel scaffold. We recommend seeding endothelial cells under static conditions and allowing up to 3 hours for the cells to attach, prior to incubating the cells overnight under laminar flow conditions. Endothelial cell growth under laminar flow is important to cultivate the barrier function and monolayer present in endogenous blood vessels.

Variables in circuit construction include circulation volume, tubing material and diameter, gas trap and pump. The circulation volume can be adjusted to accommodate the desired outcome measures (e.g. flow cytometry, enzyme-linked immunosorbent assay, mass spectrometry, biochemical analysis). Tubing length and diameter can be varied to reach the desired circulation volume. Notably, the more tubing is used to maximize circulation volume, the greater the ratio between non-endothelialized and endothelized surface area in the circuit. Medical grade tubing made out of Tygon, fluoropolymers or polyvinylchloride are used as they are inert, biocompatible and gas permeable. In our set-up, the gas trap is connected just before the inlet port of the vascular model to minimize bubbles passing through the endothelialized portion of the circuit. The pump chosen needs to support the desired flow rate (macrovascular: ≤70 mL/min, microvascular: ≤50 μL/min). Syringe pumps deliver consistent continuous flow rates. Peristaltic pumps can deliver continuous and pulsatile flow patterns. Due to the rotor mechanism of peristaltic pumps, the continuous flow setting may deliver a “dampened” pulsatile flow. Pumps which deliver inconsistent flow rates may lead to bubble formation in the circuit. The flow rate can be adjusted to simulate the desired environment (e.g. arteries, capillaries, veins) and shear stress. When calculating the shear rate, it is important to note that blood is a non-Newtonian fluid meaning that its viscosity and the shear stress that it exerts on the endothelium are dependent on the amount of pressure exerted on it. Lastly, the entire circuit can be placed in an incubator or the pump computer can be left outside the incubator (if there are ports available for the passage of cables).

Whole blood, blood product, whole blood-blood product mixtures and diluted blood can be perfused through the system to simulate blood flow. Undiluted blood best recapitulates the intravascular milieu by maintaining blood concentration and viscosity. However, undiluted anticoagulated whole blood still carries increased risks of thrombosis (particularly in microvascular channels) and is opaque – making it optically difficult to assess blood cell adhesion to endothelium. In these situations, blood can be diluted to maintain circulation and enable in situ microscopic analysis. Mixing whole blood with packed RBCs at a ratio of 9:1 simulates a one unit transfusion of packed RBCs. The whole blood should be diluted such that the hematocrit is below the transfusion threshold of 7 g/dL to prevent post-transfusion hyperviscosity. While this dilution simulates the hemodilution observed after crystalloid resuscitation, it is unlikely to approximate the blood viscosity observed in patients with chronic anemia. Notably, all blood and blood products circulated through in vitro models are anticoagulated as this treatment is necessary for ex vivo blood storage prior to experimentation. Sodium citrate, ethylenediaminetetraacetic acid (EDTA) and heparin anticoagulation are non-toxic and suitable for use. Sodium citrate and EDTA work by chelating calcium ions and their anticoagulant effect can be weakened by adding calcium. Sodium citrate is required for coagulation analyses. Furthermore, high levels of sodium citrate can alter the pH of blood, EDTA can activate endothelium and both chelators can interfere with microparticle formation. Heparin provides enduring anticoagulation, but it interferes with coagulation analyses and immune cell function. It follows that the requirement for anticoagulated blood represents a shortcoming of in vitro vascular models. This pitfall can be dealt with to some extent by using in vivo models to corroborate in vitro results.

Various outcome measures can be recorded from these in vitro vascular models. Real-time video microscopy can be used to visualize blood cell adhesion to the endothelium. The perfusant can be analyzed for soluble factors (e.g. cytokines, microparticles, biological response mediators) and cell surface markers. After blood circulation, the endothelial cells can be imaged in situ or removed from the scaffold with trypsin for further analysis. In our laboratory, a vascular model with a 3 mm, full-length channel seeded with human umbilical vein endothelial cells is primarily used to simulate transfusion reactions by circulating mixtures of recipient whole blood and donor blood products (Online Supplementary Material). Whole blood circulation leads to increased formation of annexin V-positive microparticles and erythrocyte microparticles compared to statically held whole blood (2.13 × 10 versus 6.08 × 10 microparticles/mL, P<0.0001 (Online Supplementary Material).

Conclusion

In vitro vascular models combine blood, endothelial and flow components into a single system. In this way, endothelialized flow models simulate the blood product-recipient blood and blood-endothelial interactions that occur under flow conditions after blood product transfusion. Five key factors of the experimental set-up (channel geometry, scaffold molding, endothelial cell seeding, circuit construction, and perfusate) can be manipulated depending on the desired organ system, transfusion scenario and outcome measures.

Footnotes

• Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/103/4/428

• Received October 31, 2018.
• Accepted January 15, 2019.