Acoustofluidic Engineering of Functional Vessel-on-a-Chip

Abstract

Construction of in vitro vascular models is of great significance to various biomedical research, such as pharmacokinetics and hemodynamics, and thus is an important direction in tissue engineering field. In this work, a standing surface acoustic wave field was constructed to spatially arrange suspended endothelial cells into a designated acoustofluidic patterning. The cell patterning was maintained after the acoustic field was withdrawn within solidified hydrogel. Then, interstitial flow was provided to activate vessel tube formation. In this way, a functional vessel network with specific vessel geometry was engineered on-chip. Vascular function, including perfusability and vascular barrier function, was characterized by microbeads loading and dextran diffusion, respectively. A computational atomistic simulation model was proposed to illustrate how solutes cross vascular lipid bilayer. The reported acoustofluidic methodology is capable of facile and reproducible fabrication of the functional vessel network with specific geometry and high resolution. It is promising to facilitate the development of both fundamental research and regenerative therapy.

INTRODUCTION

Vascular system is one of the most important circulatory systems in the human body¹. Blood vessels are not only a necessary pathway for physiological metabolism, material exchange, and nutrient delivery, but also an important route for most pharmacokinetic drug delivery^2,3. In order to match specific different structures and functions of each organ, the blood vessels have significant organ specificity^4,5. For example, the vascular structure in the hepatic lobule has a hexagonal distribution⁶, while the vessels in muscle tissue run in parallel lines⁷. Therefore, the in vitro reconstruction of functional vascular models has always been one of the key research topics in the field of tissue engineering⁸.

At present, self-assembly nature of endothelial cells (ECs) is mostly used for in vitro capillary vessel network construction^9–13, with inevitable randomness and low reproducibility¹⁴. In an attempt to control the vessel shape, researchers have been building blood vessels by coating the inner surface of hollow gel channels with ECs^15–17. However, limited by the template resolution and difficulty of removing mold template without destroying channel, it is difficult to achieve micron-scale blood vessels with slightly complex shapes^18,19. Multiphoton ablation technology reported by Prof. Zheng et al. can generate hollow gel channels with intricate patterns of a few microns, thus accurately control the shape and size of blood vessels at the same time^20–22. However, the complexity and cost of the laser machine system may limit its widespread adoption. Due to various strengths, such as remote control, biocompatibility, non-contact and easy operation, acoustofluidics has gradually developed into a popular biomedical technology^23,24, and widely used in the fields of disease diagnosis^25–30, drug delivery^31–34, tissue engineering^35–40 and biophysical characterization^41,42. Here, we report a standing surface acoustic wave (SSAW)-based acoustofluidic methodology as an alternate of engineering vessel-on-a-chip. First, the suspended endothelial cells were acoustically patterned into parallel lines topography in the hydrogel matrix. Then, the patterned endothelial cells were activated by the interstitial flow and developed into functional vessel tubes along the previous acoustic patterning geometry. The permeability difference between the mono-vessel and fibroblast-supported vessel was evaluated and a computational model was proposed for interpretation. Last, by characterizing the compound response of neurons in the pure matrix and in the vessel-associated matrix, the influence of blood vessels was demonstrated. The reported acoustofluidic vessel engineering demonstrates that without any physical channel as a guide, the HUVECs can still assembled into the vessel network with the designated specific geometry. Besides, the cost of the acoustofluidic system is more lab-friendly than photon laser facility. Thus, it is promising to be an alternative method for facile and high-resolution construction of in vitro vessel model.

MATERIALS AND METHODS

Acoustofluidic device fabrication

The SSAW device was fabricated through standard lithography and lift-off process⁴³. A 15-μm-thick photoresist layer (AZ512, Kayaku Advanced Materials, Inc.) was spin-coated on a 500-μm-thick, double-side polished, piezoelectric LiNbO₃ wafer (Precision Micro-Optics, Inc.). Then, the designed interdigital transducers (IDT) of double metal layers (Cr/100 Å, Au/600 Å) with a 125 μm finger width were transferred from the customized plastic mask to the substrate by metal deposition⁴⁴. The finger pair number of the IDT is 30. The resonant frequency of the obtained SSAW device was measured as around 8.21 MHz. The microfluidic chamber mold was fabricated with soft lithography⁴⁵. A 150-μm-thick SU8–2150 (Kayaku Advanced Materials, Inc.) photoresist mold was created on a 3-inch silicon wafer. After being peeled off from the mold, the polydimethylsiloxane (PDMS) chambers were plasma-bonded on a 1 mm thick glass coverslip⁴⁶. All the fabrication processes were finished in the Center for Photonics and Nanoelectronics (CPN) at Lehigh University.

Cell culture, experimental setup and immune-fluorescence staining

Human Umbilical Vein Endothelial Cells (HUVEC) and Normal Human Lung Fibroblasts (NHLF) were purchased from LONZA⁴⁷. HUVECs were cultured in an endothelial cell growth medium supplemented with EGM-2 SingleQuot kit supply and growth factors (EGM-2, LONZA). NHLF were cultured in fibroblast growth basal medium supplemented with FGM^™-2 BulletKitTM (FGM-2, LONZA)⁴⁸. The generation of excitatory human forebrain neurons was accomplished using over expression of Neurogenin-2 (NGN2) transcription factor, following the protocol^49,50. The neurons were cultured with complete mTeSR1 medium, differentiation to neuron happened with introducing dox to the culture (N2 supplement + doxycycline and growth factors in DMEM/F12 media) and maintained in culture with 50x B27 supplement and 2 mM L-Glutamine in Neurobasal medium (Thermo Fisher Scientific). All cells were cultured at 37 °C in a humidified 5% CO2 environment.

For the patterning experiment, bovine fibrinogen (CAS-9001-32-5, Sigma) was dissolved in Dulbecco’s phosphate buffered saline (DPBS) to 2 mg/ml solution. The cell-laden fibrinogen pre gel solution was mixed with 2U/ml bovine thrombin (CAS-9002-044, Sigma) and injected into the chamber through chamber loading ports⁵¹. For the pure HUVEC condition, the cell concentration was 5 million/ml. For the coculture situation, the HUVEC concentration retained with 5 million/ml, while the concentration of another cell was 0.2 million/ml. During the cell patterning experiment, a radio frequency signal from a function generator (AFG3102, Tektronix, USA) was amplified by an amplifier (ACS-230-25W, Com-Power, USA), and transferred to the IDT pair. The input voltages on the devices were from 50 Vpp. A coupling water layer was placed between the SSAW device and the PDMS-glass-based microfluidic chamber for acoustic wave conduction^52,53. The HUVECs were aligned into parallel lines in PDMS channels during 3 minutes of acoustic patterning process. Then the PDMS chambers were relocated into the incubator for the gel solidification and further in vitro cell culture⁵⁴.

For immunofluorescence staining of the vessel model, detailed method can be found in our previous publication¹³. For the neural cells-vessel coculture system, 1.0 mM monosodium glutamate (MSG) (Sigma-Aldrich) was dissolved in EGM medium and loaded into the vessel tube. c-Fos (sc-166940-AF488, Santa Cruz)⁵⁵ immunostaining was executed 10 min after loading the monosodium glutamate solution. All the fluorescent images and confocal scanning were acquired with a Nikon C2+ laser scanning confocal microscope in the Health Research Hub Center, Lehigh University. The image processing was conducted with ImageJ software from National Institutes of Health (USA).

On-chip vessel network construction and vascular permeability characterization

After the gel curing, a pipette tip containing 50 μL culture medium were inserted into one of the chamber loading ports to provide interstitial flow and hydrostatic, while an empty pipette tip was inserted into the other chamber loading port for medium collection^13,47. The culture medium in pipette tips were refreshed every day. The permeability of the vessels was measured with solute diffusion across the vessel wall. 70kDa FITC conjugated dextran (CAS-60842-46-8, Sigma) was dissolved in the EGM-2 medium to prepare the solution^56,57.

Computational simulation

COMSOL^® Multiphysics 5.6 was exploited to simulate the defined central area resulted from the interference of two opposing SAW. A “Pressure Acoustic, Frequency Domain” is used to analyze the pressure distribution of one-dimensional standing acoustic field^58–62:

$$\nabla \cdot \left(-\frac{1}{{\rho }_0}\nabla p\right)=\frac{{\omega }^2}{{\rho }_0{c}^2}$$

where $\mathrm{\omega }$ is the angular frequency, $c$ is the sound speed, ${\rho }_0$ is the density, and $p$ is the pressure. In the permeability computational model, the 70 kDa FITC-dextran solute is represented as a series of discrete beads uniformly distributed on one side of the simulation domain (Fig. 4A green beads). Cooke-Deserno three-bead membrane representation model is implemented to simulate vascular membrane, which offers an elegant balance between computational simplicity and biological realism⁶³. The interaction between these entities is defined by employing the Lennard-Jones potential, allowing for the manifestation of Brownian motion among the beads. The use of the Lennard-Jones potential not only provides a mechanism for interaction between the beads but also models the diffusion of these bead-represented solutes through the membrane^64,65. Within this framework, each lipid is depicted as three discrete beads, representing the hydrophilic head group(red in fig. 4) and the two hydrophobic tail groups(blue in fig. 4), effectively capturing the amphiphilic nature of lipids. These head and tail coarse-grained beads are interconnected via a finite extensible nonlinear elastic bond, thus capturing the essential amphiphilic character of lipids. Additionally, a harmonic angular potential is employed to maintain the straightened configuration of the lipid molecule. The lipid bilayer membrane representative of HUVECs (Human Umbilical Vein Endothelial Cells) and fibroblasts is modeled using two distinct membrane layers, with each layer consisting of three beads. To closely emulate the experimentally observed diffusion rates of membrane permeability, the area per lipid in the membrane was optimized by adjusting the Lennard-Jones parameters. Several foundational assumptions were made to enhance computational efficiency and ensure clarity of the simulation. we conceptualized the lipid bilayer as a seamless, uniform surface, intentionally omitting intricate molecular details reduce computational complexity. Furthermore, the complexities of lipid head-groups and tails are abstracted into coarse-grained beads, encapsulating essential interaction properties. Lastly, the behavior and dynamics of the membrane are primarily driven by a balance between curvature elasticity and thermal fluctuations. The membrane representation and the associated parameters was described in our previous publication⁶⁶.

Figure 4.

Computational vascular barrier function modeling with/without supporting fibroblasts. A) Modeling of the dextran diffusion process before (i), during (ii) and after (iii) the vessel membrane crossing. B) Modelling the final diffusion result of pure vessel condition within the time unit (i). Modelling the final diffusion result of fibroblast-supported condition within the time unit (ii). (iii) The detailed composition of simulated vascular membrane. (iv) Simulation result of the dextran permeability speed.

RESULTS AND DISCUSSION

Acoustofluidic patterning of the suspended HUVECs into acoustic pressure nodes

Figure. 1 illustrates the methodology workflow of the acoustofluidic vessel-on-a-chip engineering. Once signal of the resonant frequency from the signal generator was amplified and introduced to the acoustofluidic device, two identical but opposite-propagating traveling surface acoustic waves (SAW) were released from the IDT pair, and the SSAW field was constructed on the piezoelectric LiNbO₃ wafer surface^67,68. Thus, a periodic distribution of pressure nodes and antinodes with minimum and maximum pressure amplitudes was presenting respectively (Fig. 2B)^69–71. With the conduction of the coupling water layer, the acoustic field was introduced into the chamber⁷². As a result, the originally randomly distributed cells was relocated to the nearest pressure nodes, thus forming parallel linear arrays⁷³. Here, the half wavelength of the SSAW was designed as 250 μm, and the final patterned cell line spacing was approximately 150 μm (Fig. 2C). Fibrinogen is currently the most suitable hydrogel for culturing vascular tissue in vitro, and it was selected as the biological matrix in this experiment^74,75. The thrombin concentration was reduced to prolong the time required for the gel solidification, until the cells could be fully patterned. In the experiment, it was observed that the fibrinogen gel began to coagulate after being mixed with thrombin for 3 minutes at room temperature, at which time the cells rarely moved. Thus, after 3 minutes, the sample was moved into the incubator for fully solidification. After 3-minute-patterning, the microfluidic chambers were moved into the incubator to fully solidify the gel. In this way, the original cell patterning topography was maintained by the solidified gel could after the acoustic field was withdrawn⁷⁶.

Figure 1.

Schematics of acoustofluidic engineering functional vessel-on-a-chip. A) As shown schematically, standing surface acoustic wave is generated on-chip to pattern the cells and promote the desired vessel formation. B) Flow chart of the acoustofluidic engineering of vessel formation. The suspended cells are patterned into the acoustic pressure node and aligned into parrel line array. After the acoustic field disappears, the solidified gel can still maintain the original patterning shape of the cells. Due to the nature of adherent growth, some excess vascular cells will grow along the wall of the PDMS chamber, forming transverse vessels perpendicular to the patterned longitudinal vessel tubes, thereby connecting these parallel arrays of vascular tubes. In the end, functional vessel-on-chip can be formed with the interstitial flow stimulation.

Figure 2.

Acoustofluidic engineering of suspended HUVECs into functional vessel-on-a chip. A) The photo of the SSAW device and microfluidic chamber. B) The simulation of the SSAW distribution. C) Before patterning (i), the cells are randomly distributed. After patterning (ii), the cells are aligned into parallel straight-line distribution. D) The separated cells (i) self-assemble at hour 24 (ii) and formed parallel perfusable vessel tubes at hour 48 (iii). 10-um-fluorescent beads can be loaded into the vessel tube (iv). E) The geometry of vessel-on-a-chip(i). The HUVECs are labeled with 488 fluorescence-conjugated-Factin. (ii) The anastomosis structure between the parallel vessel tubes and the side tube perpendicular to them. Cell nuclei is labeled with DAPI. (iii) The vessel tube under static culture condition. (iv) The vessel tube under interstitial flow-stimulated culture condition. F) The vessel tubes are characterized with the immunostaining. Cell nuclei is labeled with DAPI (i). Vessel biomarker is labeled with red CD31 (ii). Cytoskeleton is labeled with green Factin (iii). The overlap image is presented as (iv).

Patterned vessel-on-a-chip formation under interstitial flow-assisted in vitro cell culture

Prior to this project, several papers have been published on the construction of vascular models by acoustic cell patterning^77–79. For example, Kang et al. acoustically patterned endothelial cells in hyaluronic acid to construct vascular tissue for ischemia therapy⁸⁰. The authors transplanted the in vitro vascular tissue into a mouse model. After the vascular tissue was anastomosed with the rat’s own blood vessels, the vessel persusability was verified in vivo. Besides, Zhang et al. reported that for in vitro vascular models, the key to making vessel tubes perfusable is to provide a stimulus of interstitial flow⁸¹. The interstitial flow can promote the expression of matrix metalloproteinase-2 protein to enhance the vasculogenic capacity of endothelial cells. Based on these inspiring works, we hypothesized that the perfusable patterned vascular models could be remodeled in vitro by delivering the patterned ECs with interstitial flow stimulation.

As control, a group of patterned HUVECs and random-seeded HUVECs were cultured in the static culture condition, respectively (Fig. S2). Although HUVECs could also self-assemble into vessel tubes in both groups, these blood vessels are essentially thin and non-perfused. For the experiment group, the patterned HUVECs were applied with hydraulic pressure by loading 50 μl of medium into a 200 μl pipette tip to induce interstitial flow. The patterned ECs began to interconnect in 24 hours (Fig. 2D). The ECs in the lumen not involved in vascularization were washed away in 48 hours (Fig. 2Diii). Parts of ECs migrated to the chamber wall, along which the ECs grew into the horizontal vessel tube to connect all the vertical parallel patterned vessel tubes (Fig. 2Eii). After 48 hours in vitro culture, a vascular network was formed, maintaining the shape of the previous acoustic patterning (Fig. 2Ei). The vessel perfusability was demonstrated by loaded the 10 μm red fluorescent beads into the vessel tube (Fig. 2Div). From the cytoskeleton characterization (Fig. 2E), it can be clearly seen that the blood vessels formed under hydraulic pressure stimuli were much wider than the static cultured blood vessels⁸² (Fig. 2Eiii&iv, Fig. S4). To verify the HUVEC identity, the patterned vessel tubes were stained with red fluorescence conjugated CD31 antibody (Fig. 2F)⁸³.

Vascular barrier function test of the vessel tube with/without the supporting fibroblasts

In addition to substance transportation, another important function of blood vessels is lateral permeability, which enables the exchange of substances with surrounding tissue, as known as vascular barrier function^84,85. To evaluate the vascular barrier function in different scenarios, the monocultured vessel was set as control and the fibroblast-supported vessel was set as the experiment group⁸⁶. First, the suspended fibroblasts were mixed into the HUVEC suspension, and went through the acoustofluidic engineering process (Fig. 3A). After 2 days culture, HUVECs could still interconnect and assemble into the straight vessel tubes while the fibroblasts were located on the outer wall of the vessel tube (Fig. S3A). Figure. 3B shows the confocal scanning of the fibroblast-supported vessel structure. The fibroblasts were labeled with red fluorescence conjugated α-smooth muscle actin antibody⁸⁷.

Figure 3.

Vascular barrier function characterization with/without supporting fibroblasts. A) Schematics drawing of the vessel formation from the suspended cells. (i) Pure HUVECs situation, and (ii) mixture of UVECs and fibroblasts. B) Coculture of the vessel (i) and the fibroblast (ii). The overlap image (iii) showed the fibroblast is on the outer wall of the vessel. C) Vessel permeability test in mono vessel tube (i) and fibroblast supported vessel tube (ii). Scale bar: 20um. D) (i)70 kDa dextran diffusion profiles at t = 3 min for vessels in various culture conditions. (ii)70 Kda dextran permeability values calculated for vessels in various culture conditions. Error bars show the standard deviation. t-test, *p< 0.05, **p< 0.01, ***p< 0.001, n= 3.

Then, the vascular barrier function was quantified by visualizing the diffusion of 70 KDa FITC-dextran solute⁸⁸. The dextran solution was perfused into the parallel patterned vessel tubes through the anastomosis zone (Fig. 2Eii) and permeated into the surrounding matrix. The monocultured vessel was leakier compared with the fibroblast-supported vessel (Fig. 3C). The lateral diffusion rate of dextran in the monocultured vessel was almost twice that of the fibroblast-supported vessel (Fig. 3D). For potential interpretation of the interaction between the solute and vascular cell membrane, an in silico atomistic model was proposed to visualize how the solute crossing the endothelial cell membrane (Fig. 4).

Vascular structure influence compound transport in the vessel-associated matrix

In addition to demonstrating that surrounding stromal cells can affect blood vessel, it was further shown that the vessel structure can also affect cells in the surrounding environment. Here, a functional assay was designed to show how vascular structure affects compound transport and influences neural activity. Induced pluripotent stem cells (iPSC)-derived neural cells were utilized⁸⁹, which express glutamate receptors⁵⁰ (Fig. 5). As a control, the iPSC-derived neural cells were resuspended in the pure fibrinogen gel in the microfluidic chamber (Fig. 5Ai). For the experiment group, the neural cells were cocultured with the patterned vessel network (Fig. 5Aii). Glutamate is one of the major excitatory neurotransmitters in the mammalian central nervous system⁹⁰. For the pure hydrogel matrix, solution containing glutamate was loaded into the microfluidic chamber through one inject port. The compound travels through the matrix to reach the suspended cells by passive diffusion⁹¹. For the vessel-associated matrix, the compound was delivered through the patterned vessel network channel. After 10 minutes of treatment, the samples were washed to remove compound solution. After 20 minutes, the immune-staining for c-Fos was performed. The c-fos gene expression can be activated by a wide range of stimuli and is a reliable marker of neural activity⁹². Its transcription was in rapid manner⁹³. In the group in which the compound was delivered by blood vessels, the level of C-fos expression in nerve cells was almost double that of the passive diffusion group (Fig. 5C). In a pure matrix without blood vessels, the substances are transported by passive diffusion, implying slow transport rates, limited transport distances and spatial substance gradients. In a matrix with blood vessel networks, substances are transported directly in vessel tubes, which serve as highway, and reaches the bulk of the matrix more quickly⁹⁴.

Figure 5.

C-Fos expression show reduced how glutamate influence the neural cell without/without vessel transport. A) The neural cells seeded in the hydrogel respond to glutamate stimuli without (i) and with (ii) the vessel transport. Scale bar: 50 microns. B) The neural cells grow in adherence state. C) The normalized c-Fos expression rate with/without vessel transport the compound.

Conclusion

The reported acoustofluidic engineering of vessel-on-a-chip not only inherited the advantages of previous acoustophoretic publications^80,95, but also incorporated technical points from the literatures about hydraulic pressure activating vessel formation^96,97. To demonstrate the function of the acoustofluidic engineered vascular structure, the perfusability and permeability of the vessel-on-a-chip was verified. Further, a molecular dynamic computational model was proposed to illustrate how the compound crossing the vascular lipid layer. Lastly, a vessel-neural cells coculture system was established to demonstrate that the vascular structure played the important role in compound delivery process. In general, we reported an acoustofluidic methodology to engineering in vitro vessel-on-a-chip model. It is believed that the proposed method can contribute to tissue engineering and regenerative medicine application. In the future, this method has the opportunity to further combine different patterning modes of the sound field (Fig. S5A) to create functional vascular tissues of the specific shapes^98,99.