Sound-induced morphogenesis of multicellular systems for rapid orchestration of vascular networks

Two crosslinkable hydrogels, namely gelatin methacryloyl (GelMA) and fibrin gel, were employed as matrices for the acoustic patterning of TCP particles or cells, respectively. The precursor gel was placed into an insert contained in a petri dish, and this system was positioned in the vibration chamber (figure S1 (available online at https://stacks.iop.org/BF/13/015004/mmedia)). After patterning with sound, the GelMA was crosslinked with UV light exposure, whereas fibrinogen underwent enzymatic crosslinking.

2.1.1. GelMA synthesis.

2.1.2. Fibrin gel.

A panel of different geometries was obtained by patterning tricalcium phosphate particles (TCP, Kuros Biosciences BV, Bilthoven, The Netherlands) with a diameter range of 125–250 μm in a photocrosslinkable methacrylated-gelatin hydrogel (GelMA). In particular, 70 mg TCP particles were mixed with 5% w v^–1 GelMA and 0.3% w v^–1 Irgacure 2959 within two different chamber geometries, the squared chamber (side of the inner square = 15 mm and h = 1 mm) and the circular chamber (diameter of the inner circle = 21 mm and h = 1 mm). These inserts were made of polycarbonate and were manufactured via conventional subtracting manufacturing. For both the squared and circular geometries, a range of frequencies and amplitudes was applied for 10 s in order to study the pattern reproducibility. Patterns were fixed via UV exposure using a Bio-Link BLX-365UV radiation chamber (Vilber Lourmat, Collegien, France) for 5 min (0.1 J cm⁻²). All samples were obtained in triplicates. Clips of TCP pattern formation were acquired with a camera integrated to the SIM prototype (mimiX Biotherapeutics Ltd., Neuchâtel, Switzerland).

The potential energy U of the system can be described by two different components, ${U_{vibr}}$ and ${U_{volume}}$ [29]:

$$\begin{equation}\begin{array}{*{20}{c}} {U = \mathop \sum \limits_{\text{n}} {{\text{U}}_{{\text{vibr}}}} + \mathop \sum \limits_{{\text{allpairs}}} {{\text{U}}_{{\text{volume}}}}} \end{array}\end{equation} \tag{ 1 }$$

The first component, ${U_{vibr}}$ applies to the n simulated particles and represents the effect of the vibrations that produce the Faraday waves on the liquid-air interface. Chen and colleagues have already described an equation for ${U_{vibr}}$ of a particle in the system, which is a position-dependent potential [30]:

$$\begin{align} {{\text{U}}_{{\text{vibr}}}} = & \left[ {\frac{4}{3}{\mathrm{\pi }}{{\text{R}}^3}{{{\rho }}_{{\text{par}}}} - {\mathrm{\pi }}{{{\rho }}_{{\text{liq}}}}{\text{R}}{{\mathrm{\delta }}^2}{\mathrm{\,}} + {\mathrm{\,\pi }}{{{\rho }}_{{\text{liq}}}}\frac{{{{\mathrm{\delta }}^3}}}{3}} \right]\frac{{{{\mathrm{\omega }}^2}}}{4}{{\left| {\mathrm{\zeta }} \right|}^2}\nonumber\\ &\quad + \frac{{3{{\mathrm{\pi }}^2}{\mathrm{\mu \delta \alpha }}}}{{\mathrm{\omega }}}\left| {{{\mathrm{\zeta }}_{90^\circ }}} \right| \end{align} \tag{ 2 }$$

where R and ${{{\rho }}_{{\text{par}}}}$ are the average radius and density of the particle respectively, ${{{\rho }}_{{\text{liq}}}}$ is the density of the liquid, μ is the dynamic viscosity of the liquid and δ is the submerged length of the particle; these values require parameterization according to the particles and the liquid used. Other parameters are function of the amplitude A and frequency f of the vertical vibrations of the system:

α is the acceleration of the system, ω is the angular frequency of the standing waves, ζ is the liquid surface deformation and ${{\mathrm{\zeta }}_{90^\circ }}$ is its 90° phase shift in x and y directions.

For the liquid surface deformation equation ζ, two different forms were used to simulate the patterning observed in two different chamber shapes, ${\zeta _{sq}}$ for a square chamber or ${\zeta _{cric}}$ for a circular chamber of radius ${R_c}$:

$$\begin{equation}\begin{array}{*{20}{c}} {{{\mathrm{\zeta }}_{{\text{sq}}}} = \cos \left( {\frac{{2{\mathrm{\pi }}}}{{\mathrm{\lambda }}}{\text{x}}} \right)\cos \left( {\frac{{2{\mathrm{\pi }}}}{{\mathrm{\lambda }}}{\text{y}}} \right){{\text{e}}^{\frac{{{\mathrm{i\omega t}}}}{2}}}A} \end{array}\end{equation} \tag{ 3 }$$

$$\begin{equation}\begin{array}{*{20}{c}} {{{\mathrm{\zeta }}_{{\text{circ}}}} = \cos \left( {\frac{{2{\mathrm{\pi }}}}{{\mathrm{\lambda }}}\sqrt {{{\left( {{\text{x\,}} - {\mathrm{\,}}{{\text{R}}_{\text{c}}}} \right)}^2} + {\mathrm{\,}}{{\left( {{\text{y\,}} - {\mathrm{\,}}{{\text{R}}_{\text{c}}}} \right)}^2}} } \right){{\text{e}}^{\frac{{{\mathrm{i\omega t}}}}{2}}}A} \end{array}\end{equation} \tag{ 4 }$$

After applying a vibration to the system, by using the linear dispersion relationship:

$$\begin{equation}\begin{array}{*{20}{c}} {{{\text{c}}^2} = \left[ {\frac{{{\mathrm{g\lambda }}}}{{2{\mathrm{\pi }}}} + {\mathrm{\,}}\frac{{{{\mathrm{\sigma }}_{{\text{liq}}}}}}{{{{{\rho }}_{{\text{liq}}}}}}\left( {\frac{{2{\mathrm{\pi }}}}{{\mathrm{\lambda }}}} \right)} \right]\tanh \left( {\frac{{2{\mathrm{\pi H}}}}{{\mathrm{\lambda }}}} \right)} \end{array}\end{equation} \tag{ 5 }$$

it is possible to calculate the wavelength λ of the Faraday waves by applying the limits of capillary waves and thin liquid films [11], where wavelength is equal or greater than the liquid thickness, to get the following equation:

$$\begin{equation}\begin{array}{*{20}{c}} {\lambda = \sqrt[4]{{\frac{{{\sigma _{liq}}}}{{{\rho _{liq}}}}{{\left( {\frac{{2\pi }}{{{f_\omega }}}} \right)}^2}H}}} \end{array}\end{equation} \tag{ 6 }$$

where ${{\mathrm{\sigma }}_{liq}}$ and H are the surface tension and the depth of the liquid, ${f_\omega }$ is the frequency of the Faraday waves and corresponds to half the value of the frequency f for the vibrations applied to the system.

The second component, ${{\text{U}}_{{\text{volume}}}}$, regulates the interaction between the particles and applies to all their possible pairs, preventing an intersection of their volumes. It is described by truncated cut-off (and purely repulsive) Lennard-Jones interactions between the pairs of particles:

$$\begin{equation} {{U_{volume}} = 4\epsilon \left[ {{{\left( {\frac{\sigma }{{{r_{ij}}}}} \right)}^{12}} - \,{{\left( {\frac{\sigma }{{{r_{ij}}}}} \right)}^6}} \right]} \end{equation} \tag{ 7 }$$

where ε is the well depth of the potential, σ is the distance at which the potential between the two particles is 0 (chosen as the sum of the radiuses of the two particles interacting) and ${r_{ij}}$ is the distance between particles i and j. To calculate the trajectories of the particles, a Verlet algorithm was used. This required the calculation of the components x, y and z of the forces F acting on the particles, which were obtained through the relationship:

$$\begin{equation}\begin{array}{*{20}{c}} {F = - \nabla U} \end{array}\end{equation} \tag{ 8 }$$

Non-conservative forces were applied to the system to simulate the friction of the moving particles with the bottom of the patterning chamber and with the fluid. These forces allowed the particles to stop. The simulation box was set up as a square or circle with solid borders and the simulated particles were placed randomly on its surface. This boundary conditions corresponded to the experimental settings. A C++ script was developed to calculate the simulations; the results and potential energy surfaces were visualised with Matlab^TM (MathWorks, Natick, MA, USA). The in silico simulations were performed by using parameters equivalent to those of GelMA 5% (density = 1.063 g cm⁻³), TCP particles approximated as spheres with a diameter of 125 μm and density = 3.14 g cm⁻³, and using 1000 particles for the square geometry and 2000 particles for the circular geometry. A timestep of 0.1 ms was used.

Green fluorescent protein-expressing Human Umbilical Vein Endothelial Cells (GFP-HUVEC) were purchased from Angio-Proteomie (Boston, MA, USA) and cultured in complete EGM-2 Endothelial Cell Growth Medium (Lonza, Basel, Switzerland) on tissue culture plastic pre-coated with Quick Coating Solution (Angio-Proteomie). Medium was refreshed every second day and cells were used up to passage 7.

Human mesenchymal stem cells (hMSC) were isolated using established protocol [31]. Bone marrow aspirates were harvested during routine procedures from the vertebral body of patients with full ethical approval (Bern Req-2016-00141). hMSC were subcultured maintaining a seeding density of 3 × 10³ cells·cm⁻² in α-MEM (Minimum Essential Medium Eagle - alpha modification, Gibco, Thermo Fisher, Zürich, Switzerland) with 10% MSC-qualified FBS (Fetal bovine serum, Pan-Biotech, Aidenbach, Germany), 100 U·ml⁻¹ penicillin, 100 μg·ml⁻¹ streptomycin (Gibco) and 5 ng·ml⁻¹ basic Fibroblast Growth Factor (bFGF, Fitzgerald Industries International, Acton, MA, USA). Medium was changed every second day. hMSC up to passage 4 were used for the experiments. Both GFP-HUVEC and hMSC were maintained at 37 °C in a 5% CO₂ humidified atmosphere.

For spheroid formation, cells were detached with 0.05% Trypsin-EDTA (Gibco, Thermo Fisher) and counted in a Neubauer chamber with Trypan Blue exclusion. Mixtures of 1:1 GFP-HUVEC:MSC (2 × 10⁶cells each) were seeded in 100 mm ultra-low attachment culture dishes (Corning, New York, NY, USA) to avoid cell adhesion to the substrate and favour cell-cell interactions and spheroid formation. The medium was composed of 1:1 mixture of EGM-2 and MSC growth medium as described above. The cells were incubated for 48 h at 37 °C—5% CO₂ to allow for spheroid formation, then the spheroids were used for further experiments.

hMSC were suspended in medium at a concentration of 5 × 10⁶cells·ml⁻¹ and placed in a 60 mm-Petri dish. A 3 G acceleration amplitude and a frequency of 157 Hz were used for 10 s in order to obtain the desired pattern. After patterning, the Petri dish containing the cell pattern was kept at rest for 1.5 h to ensure a stable adhesion of cells to the Petri dish. Samples were run in triplicates and non-patterned hMSC in medium were used as control. After 1 d of culture, both metabolic activity and cell viability were assessed respectively by CellTiter-Blue^® Assay (Promega, Dübendorf, Switzerland) and Live/Dead Assay (10 μM calcein AM and 1 μM ethidium homodimer, Sigma-Aldrich). Morphological evaluation of cell organization was performed using phalloidin staining. After an initial washing in PBS, patterned cells were permeabilized with 0.25% Triton X-100 for 15 min RT, then stained with 2 μg·ml⁻¹ of TRITC-conjugated phalloidin for 45 min at RT. HUVEC and hMSC were patterned after embedding within a fibrin gel to assess the formation of vascular networks. In particular, HUVEC alone or HUVEC and hMSC (ratio 10:1) were mixed in fibrinogen (final concentration 2.5 mgml). After the addition of thrombin at a 0.5 U·ml⁻¹ final concentration, cells were patterned in the circular insert (10.5 mm radius) with 1.2 g acceleration amplitude and 80 Hz frequency for 30 s. Three different HUVEC concentrations (0.4 × 10⁶, 0.8 × 10⁶ and 3 × 10⁶ cells·ml⁻¹) were tested in order to study the influence of cell density on the microvascular network formation. The resulted patterns were cultured up to 5 d in EGM-2 medium in a 37 °C and 5% CO₂ humidified atmosphere. The medium was changed every two days. At different time points (3, 24, 72 and 120 h) the vascular network was analysed for the following parameters: average number of branches, average junctions, total network length and vessel diameter. An ethidium homodimer staining (1 μM) was carried out to check the influence of the SIM patterning on the viability of cells embedded in a hydrogel matrix. All samples were prepared in triplicates.

A poly(dimethylsiloxane) (PDMS; Silgard 184, Dow Chemical, Midland, MI, USA) microfluidic device was produced as previously reported [32, 33] and bound to a glass coverslip with oxygen plasma in order to create channels with a 150 μm height. In particular, the device consisted of 4 channels for media supply (1300 μm wide × 10 000 μm long and 150 μm high) and 3 channels that hosted the fibrin gel (1000 μm wide × 8000 μm long and 150 μm high). Each hydrogel channel is designed with trapezoidal micro-posts (200 μm length and 100 μm distance) that allow surface-tension during the filling of the channel with the cell-laden fibrin hydrogel. The channels were accessible through inlets created with 4 mm—diameter biopsy puncher (for medium reservoir) and 1.2 mm biopsy puncher for hydrogel filling. GFP-HUVEC alone or in combination with hMSC were resuspended in complete EGM-2 medium mixed with thrombin and seeded in two of the 3 channels. Three different HUVEC concentrations (0.4 × 10⁶, 0.8 × 10⁶ and 3 × 10⁶ cells·ml⁻¹) were tested in order to study the influence of cell density on the microvascular network. Thrombin and fibrinogen were mixed at 1:1 v v^–1 ratio to obtain a final concentration of 0.5 U·ml⁻¹ and 2.5 mg·ml⁻¹, respectively. The microfluidic chips were kept in a humidity chamber for 15 min to allow the complete gelation and afterwards the medium channels of all microdevices were filled from one side with complete EGM-2 and placed into an incubator (humidified, 37 °C, 5% CO₂). The EGM-2 medium was changed at day 1 and day 3. At different time points (3, 24, 72, and 120 h) the vascular network was analysed for different parameters, such as average number of branches, average junctions, total network length and vessel diameter.

HUVEC-hMSC spheroids formed after 48 h of incubation were collected and mixed in thrombin solution, then mixed with fibrinogen solution immediately before patterning (final concentrations: 0.5 U·ml⁻¹ thrombin + 2.5 mg·ml⁻¹ fibrinogen). The spheroid-fibrinogen-thrombin solution was quickly transferred to a Petri dish containing the SIM circular insert (10.5 mm radius) and pattern formation was induced using a frequency of 80 Hz and 0.5 g acceleration amplitude for 10–15 s in order to obtain a pattern with concentric rings. The samples were incubated for 5 min at 37 °C to ensure complete fibrin crosslinking, then 1:1 EGM-2:MSC growth medium was added. Samples were maintained in culture in a 37 °C and 5% CO₂ humidified atmosphere up to 5 d and medium was refreshed every second day.

Polycarbonate inserts with a circular and square geometries were modified with two 1000 μm-diameter holes in the middle (figure S1(c)). The inserts were attached to a glass slides by using a double-sided tape and were sterilized with UV light for 1 h. The insert was designed to be placed directly into 60 mm-petri dishes. A sterile polyetheretherketone (PEEK) tube (external diameter 0.9 mm and internal diameter 0.8 mm) was inserted in the middle of the insert through two 1000 μm-diameter holes in order to recreate a macrovessel. Afterwards, spheroids of GFP-HUVEC (passage 6) and hMSC (passage 2) with a 1:1 ratio were embedded in a fibrin gel (2.5 mg·ml⁻¹ of fibrinogen and 0.5 U·ml⁻¹ of thrombin) that was polymerized in the insert keeping the PEEK tube in the middle to ensure interconnection between the macrovessel and the microvessels generated by the spheroids pattern. After crosslinking, the PEEK tube was removed, and the macrochannel with a resulting diameter of 800 μm was perfused with 10⁷ cells·ml⁻¹ GFP-HUVEC or RFP-HUVEC. Cell adhesion to the fibrin was ensured by keeping the cell suspension inside the macrochannel for 30 min per side. Finally, PBS was perfused through the macrochannel and the patterned sample was cultured with EGM-2 for 5 d. At day 5, the macrochannel was perfused with Texas-Red dextran (70 kDa, Thermo Fisher). Mosaic and z-stack images were acquired on days 0, 1 and every 2 d thereafter using a Zeiss LSM800 confocal microscope (Zeiss, Oberkochen, Germany). Perfusion tests were performed, and images were acquired in time sequence. Afterwards, the samples were fixed in 4% neutral buffered formalin.

After an initial washing in PBS, the gels with cell patterns were fixed in paraformaldehyde 4% for 15 min at RT. The non-specific binding sites were blocked with 10% FBS and permeabilization was performed with PBS/Tween20 for 30 min. Anti VE-cadherin primary antibody (CD144, Waltham, US) at a concentration of 5 μg·ml⁻¹ was added for 1 h at RT. Gels were washed with PBS, then the secondary antibody was added (Alexa Fluor 647 IgG 1:800, Thermo Fisher) for 1 h at RT. After washing with PBS, cell nuclei were counterstained with DAPI (4',6-diamidino-2-phenylindole) (1:800) and evaluated with a Zeiss LSM800 confocal microscope.

Phase-contrast and fluorescence images of cells viability and morphology were acquired using a Zeiss LSM800 confocal microscope equipped with a CCD camera (Axiocam 506 color) at magnification 5X (0.16), 10X (0.45), 25X (water immersion objective 0.8) and 40X (water immersion objective 1.1). Time-lapse images of cell growth and spatial organization were obtained at 0, 24, 72, 120 h after SIM patterning or random organization. Image analyses were performed using ImageJ software (NIH, Bethesda, MD, USA).

The grown kinetic of the pattern was evaluated by measuring the total green area in the time-lapse pictures between 0 and 24 h after the pattern formation. The image of the green channel was converted into 8-bit and a gaussian filter (1 pixel) was applied prior thresholding (minimum and maximum threshold values were set at 15 and 255, respectively). The total area was then measured and plotted against time.

The intensity profile was evaluated by applying the ImageJ built-in function 'Profile plot'. The orientation of the images was maintained constant and a median filter with neighbourhood of three pixels was applied in order to reduce the noise and smooth the image's features. The background was then removed with ImageJ built-in function 'Subtract background' with a rolling ball radius of 50 pixels. Next, a line with a thickness of seven pixels was drown through the image's center, horizontally and for all length of the sample image.

The vessel network was analysed by applying an ImageJ macro. Specifically, three different areas (1000 μm × 1000 μm) were selected for each sample and converted into 8-bit greyscale and adjusted through thresholding (minimum and maximum threshold values were set at 15 and 255, respectively). Images were smoothed with Gaussian blur to reduce image noise and redundant details. The images were then examined with the ImageJ function 'Analyze skeleton'. This plugin tags all pixels in a skeleton image and counts all its junctions, triple points (junctions with exactly three branches) and branches. The vessel diameter was calculated as average of ten measures per sample.

Polycarbonate contour chambers with a square (side = 15 mm, h = 1 mm) or circular (diameter = 21 mm, h = 1 mm) geometry were placed into 60 mm-diameter Petri dishes, which were accommodated within the circular vibration chamber on the SIM device (figure S1). These chambers were filled with selected matrices (cell culture medium, GelMA hydrogel precursor or fibrinogen) containing particles (TCP particles, cells or spheroids). Frequency and amplitude were applied through the electromagnetic vibration generator controlled by a sine wave generator integrated into an SIM prototype (mimiX Biotherapeutics Ltd., Neuchâtel, Switzerland) and the resulting acceleration amplitude, which is a function of both frequency and amplitude, was then measured by using an accelerometer.

Firstly, pattern formation was tested in the presence of 70 mg TCP particles (diameter = 125–250 μm) mixed within 2 ± 0.5 ml of GelMA precursor (5% w v^–1, G' = 240 Pa). Amplitude and frequency were kept constant at 0.5 g and 40 Hz, respectively. Patterns were generated after 10–15 s (square pattern showed in supplementary video 1). Afterwards, the exposure to UV light ensured the crosslinking of GelMA and shape stabilization.

A 3D visualization of the acoustic potential field (figure 2(a), i) and a particle dynamic simulation (figure 2(a), ii) were generated for both circular and square geometries in order to understand the influence of the chamber shape on pattern generation.

Zoom In Zoom Out Reset image size

The simulated pattern obtained in the square geometry was characterized by N × N spots, with N linearly increasing with the vibration frequency. The different zones in the petri dish correspond to anti-nodes and nodes of the combined standing waves on the liquid surface (supplementary video 2. The timing of the simulation does not reflect the time for real pattern formation). The simulated pattern was experimentally confirmed with TCP particles (figure 2(a), iii), showing comparable results in terms of spots distance and a high correlation between the experimental and simulated spots distribution (figure 2(a), iv). The simulation was optimized using the parameters of GelMA viscosity as matrix and TCP particles size and densities; the modelization with other hydrogels and particles/cells is feasible by setting up a different parameterization.

The circular geometry showed a pattern with N concentric rings, with N being related to the frequency applied (supplementary video 3). At 0.5 g amplitude and 40 Hz frequency, a central spot and 4 concentric rings were visible, both in the simulation and in experimental patterns (figure 2(a), i–iii). The comparison between SIM patterns obtained via simulation or experimentally showed similar values in terms of radii of the patterned concentric rings, in line with the results of the square geometry (figure 2(c), iv). We observed that the viscoelastic properties of the hydrogel matrix and its crosslinking chemistry play a key role in determining the final pattern geometry

These initial experiments allowed to understand the interplay between parameters such as frequency and amplitude, and their respective importance on obtaining the desired pattern and assembly.

With increasing the frequencies and fixing the amplitude at low values, a linear correlation was observed between the frequency and the number of spots (for the square geometry, figure 2(d), i) or of concentric rings (figure 2(e), i), showing a R² of 0.986 and 0.995, respectively. In contrast, both the thickness of the TCP rings or spots and their distance, calculated between the centers of two adjacent rings or spots, decreased with an increase of the frequency (figure 2(d), ii and (e), ii). The fraction 0.5 was used to identify the formation of a central spot in the circular pattern. The spot is visible only at specific frequency values.

Additionally, a panel of patterns were generated for circular geometry by varying both the frequency and the amplitude, as shown in figure 2(f). By increasing both frequency and amplitude, the complexity of the patterns increased, going from simple concentric rings to complex regular honeycomb-like structures.

After conducting the preliminary experiments by employing TCP particles, hMSCs were patterned as single cells in medium in a 60 mm Petri dish, by applying a frequency of 157 Hz and 3 g acceleration amplitude that resulted in a pattern of concentric circles. We assessed if the SIM technology may negatively affect cells. hMSCs showed a cell viability and a metabolic activity comparable to the non-patterned controls (figure S2). In particular, a cell viability of 93% and 92% at days 1 and 4 was observed for the patterned cells compared to the control, 98% and 92%, at day 1 and 4, respectively.

Next, GFP-HUVEC cells were mixed with hMSC at a ratio 10:1 to ensure a proper formation of the vascular network as shown in previous studies [34, 35]. The cells were patterned using concentrations of 0.4 × 10⁶, 0.8 × 10⁶ and 3 × 10⁶ cells·ml⁻¹, that are reported throughout the manuscript as 0.4 M, 0.8 M and 3 M, respectively. HUVEC alone were also patterned. A 2.5 mg·ml⁻¹ fibrin gel was selected as hydrogel matrix to support the formation of vessels thanks to its well-known natural angiogenic properties and its intrinsic ability to stimulate endothelial assembly [36, 37]. Thrombin concentration of 0.5 U·ml⁻¹ allowed to obtain a fast gelation (clotting time < 5 min [38] of the matrix right after the stabilization of the pattern). Without the use of sound patterning (Random group), GFP-HUVEC/hMSC cells were randomly distributed within the fibrin gel, whilst the SIM patterning (SIM group) allowed to organize cells into circular patterns by applying a frequency of 80 Hz and a 2.5 g amplitude within the circular insert. The amplitude and frequency were tuned in order to achieve the desired pattern shape considering the cell size and the viscoelastic properties of fibrin (G' = 264 Pa). In particular, the resulting pattern was characterized by concentric rings that subsequently connected to each other thanks to the sprouting of the HUVEC (figure 3(a)). This specific pattern geometry was selected to show the functionality of the generated vessel-like structures in terms of angiogenic properties, hence the branching of pre-existing structure represented by the patterned concentric circles. The vascular network formation was followed until day 5: the pattern evolution at early time points (2, 24 and 72 h) is shown in figure S3. The obtained pattern was confirmed by the quantification of the fluorescence intensity on confocal images. For the three cells concentrations the GFP intensity values of the patterned samples varied in correspondence with the cell density (figure 3(b)). The random samples showed a constant flat baseline intensity value throughout the surface areas.

Zoom In Zoom Out Reset image size

Furthermore, for both random and patterned group, the quantification of the green area over time for 0.4 M and 0.8 M HUVEC in co-culture with hMSC showed an increase in the green area at day 1 and these values were stable over time until day 5, whereas the group with 3 M showed a decrease in the green area from day 3 (figure 3(c)). As expected, the addition of hMSC in the fibrin gel guaranteed both a stability of the vascular network over time and a higher sprouting from the HUVEC. In particular, in the control groups with HUVEC alone, cells showed mostly a rounded morphology, with a low number of elongated cells both in the SIM and in the randomly distributed cells, as confirmed by confocal imaging (figure S4). For all concentrations, the addition of hMSC led to an increased stability of the cellular network and to a higher area covered by HUVEC: with 0.8 M HUVEC/hMSC, the green area value (47 mm²) doubled compared to the HUVEC alone, underlying an increase in cell sprouting. A percentage of dead cells below 1% was found for both random and sound-pattern groups as assessed by the ethidium homodimer staining (figure S5). These low values were confirmed over time within 5 d.

Additionally, SIM patterning increased proximity between cells, promoting the formation of areas with high cell density (HD) as shown in figures 4(a) and (b). HUVEC sprouted from a HD area into the low-density area (LD), connecting to the nearest HD area. A comparison between three selected areas (1000 μm × 1000 μm) of constructs generated via sound-patterning (HD and LD) and microfluidic devices was carried out to evaluate how the patterning capability of SIM influenced the formation of vascular networks (figure 4(c)) compared to state of the art technology (figure 4(d)). When co-culturing HUVEC and hMSC at HD (3 M), SIM LD and microfluidics showed a similar behaviour in vascular network formation, with an average number of junctions of 145 ± 22 and 140 ± 14, respectively (figure 4(f)). When HUVEC density was reduced to 0.8 M, the vessel network became less dense with average number of junctions of 104 ± 18 for SIM LD and 54 ± 7 for microfluidics. At 0.4 M HUVEC, no network formation was observed in either SIM LD areas, showing a low number of elongated cells and average number of junctions of 7.5 ± 0.7 for SIM LD and 6.5 ± 0.7 for microfluidics. Confirming this trend for both number of average branches (figure 4(e)) and total network length (figure 4(g)). However, at 0.4 M, the HD areas of the sound patterned samples showed the formation of a vascular network with 109 ± 30 average number of junctions, demonstrating the effect of the sound patterning and creation of local cell density gradients followed by self-assembly (SIM), compared to a self-assembly process alone (microfluidic) at low cell density. A similar behaviour was observed for the average branch numbers and the total network length (figure S4 shows quantification also for random controls). When co-culturing HUVEC and hMSC, the capillary-like structures showed diameters of about 20 μm with both 3 M and 0.8 M cells in the LD area, whereas in the microfluidic chip the average vessel diameter decreased from 17.5 μm to 10 μm when decreasing cell concentration (figure 4(h)).

Zoom In Zoom Out Reset image size

Our work proved that by employing SIM patterning, vascular networks could still be formed while starting from low cell concentrations (0.4 M). The advantage of SIM over self-assembly alone in generating vascular networks starting from low cell numbers could be of high translational relevance for cell therapy and tissue engineering. Indeed, in cell therapy the shortage of autologous cells, due to issues in harvesting, expansion, manipulation, represents one of the main bottlenecks [29]. Therefore, the use of SIM technology paves the way to novel approaches for clinical applications.

In addition to single cells, SIM can efficiently pattern bigger elements such as cell, as already showed for TCP particles. In the context of vascular network formation, spheroids have been shown to be versatile vascularization units both in vasculogenic and angiogenic approaches thanks to the production of growth factors for the paracrine stimulation of blood vessel development [39]. The sound-induced patterning is therefore a promising approach for 'organoid technology', where organoids are usually originated from epithelial cells and generally lack complex structures such as blood vessels. In this context, it was recently found that hMSC cell-driven contraction is essential for the dynamic condensation of heterotypic cell collectives with improved vascularization potential [40]. With this premise, the SIM technology was employed to pattern spheroids. GFP-HUVEC/hMSC spheroids (cell ratio 1:1) were produced in ultra-low attachment culture dishes (figure S6) and distributed in a regular pattern made of concentric rings by applying a frequency of 80 Hz and low amplitude resulting in a 0.5 g acceleration amplitude (figure 5(a), supplementary video 4). Though here we focused principally on the pattern evolution of endothelial cells, the relative amount of MSC on HUVEC was increased to a 1:1 ratio with the perspective of using such system for regenerative medicine/tissue engineering applications, where MSC can potentially differentiate towards the desired cell type, while HUVEC or other endothelial cells might provide the coupling to angiogenesis which is necessary especially if large constructs are required (as for the case of, but not limited to, bone tissue engineering) [41–43]. HUVEC sprouted from the patterned spheroids in capillary-like structures that, at day 5, fused together with the capillaries sprouted from the spheroids of the adjacent rings (figure 5(b)). As in the case of single cell patterning, the analysis of the area covered by HUVEC increased over time (figure S7(a) and (b)) suggesting a sustained cell growth.

Zoom In Zoom Out Reset image size

In order to verify the potency of the vessels and capillaries generated with the SIM technology, an endothelialized macrovessel accessible with a syringe needle was embedded into the construct (see schematic in figure 5(c) and experimental pattern in figure 5(d)) for perfusion with a Texas Red-dextran solution. A 3D reconstruction of the macrovessel showed the successful generation of a channel within the fibrin gel, uniformly seeded with HUVEC cells inside (figure S8b and supplementary video 5). As shown in figure 5(e), after perfusion dextran entered the macro-vessel and consequently perfused all the small vessels directly connected to it (figure 5(f) and supplementary video 6). The dextran (red in figures 5(e)–(f)) was clearly visible inside the lumen of the green vessels formed by the HUVEC. The perfusion of the network was confirmed with an ortho-analysis of the microvessels by Z-stack scanning that showed the presence of open lumina typical of a mature and organized structure.

Additionally, the integration between the macro- and the microvessels network, needed to ensure perfusion, was verified by using GFP-HUVEC for spheroid formation and RFP-HUVEC for the macrovessel generation. Figure 5(g) shows how the sprouting from the GFP-HUVEC spheroids connected with the RFP-HUVEC that formed the central macrovessel resulting in a multiscale vascular network. Mosaic confocal images of spheroids patterns that generated multiscale vascular networks are shown in figure S8. These patterns were generated by applying specific amplitude and frequency values.

Finally, in order to confirm the effective maturation of the vessels, an immunohistochemical analysis was performed by labelling the construct for VE-cadherin, a marker that allows to define mature endothelial junctions (figures 5(h) and S9). The expression of VE-cadherin confirmed the tight connection between HUVEC cells in the vascular networks.