Spatially controlled diffusion range of tumor-associated angiogenic factors to develop a tumor model using a microfluidic resistive circuit

Abstract

Developing a tumor model with vessels has been a challenge in microfluidics. This difficulty is because cancer cells can overgrow in a co-culture system. The up-regulation of anti-angiogenic factors during the initial tumor development can hinder neovascularization. The standard method is to develop a quiescent vessel network before loading a tumor construct in an adjacent chamber, which simulates the interaction between a tumor and its surrounding vessels. Here, we present a new method that allows a vessel network and a tumor to develop simultaneously in two linked chambers. The physiological environment of these two chambers is controlled by a microfluidic resistive circuit using two symmetric long microchannels. Applying the resistive circuit, a diffusion-dominated environment with a small 2-D pressure gradient is created across the two chambers with velocity <10.9 nm s⁻¹ and Péclet number <6.3 × 10⁻⁵. This 2-D pressure gradient creates a V-shaped velocity clamp to confine the tumor-associated angiogenic factors at pores between the two chambers, and it has two functions. At the early stage, vasculogenesis is stimulated to grow a vessel network in the vessel chamber with minimal influence from the tumor that is still developed in the adjacent chamber. At the post-tumor-development stage, the induced steep concentration gradient at pores mimics vessel–tumor interactions to stimulate angiogenesis to grow vessels toward the tumor. Applying this method, we demonstrate that vasculogenic vessels can grow first, followed by stimulating angiogenesis. Angiogenic vessels can grow into stroma tissue up to 1.3 mm long, and vessels can also grow into or wrap around a 625 μm tumor spheroid or a tumor tissue developed from a cell suspension. In summary, our study suggests that the interactions between a developing vasculature and a growing tumor must be controlled differently throughout the tissue development process, including at the early stage when vessels are still forming and at the later stage when the tumor needs to interact with the vessels.

---

1. Introduction

In recent years, microphysiological systems (MPSs) have been recognized as the next alternative method to replace current model systems for drug discovery.¹ Microfluidic technology is widely applied to create organ-specific structures and physiological environments in MPSs.^2,3 In particular, developing a perfused vascular network to create the primary route of drug delivery is crucial for mimicking in vivo tissues.^4,5 To grow functional vessels in MPSs, naturally grown human vessels have been realized using the monolayer and self-organization models. Monolayer models seed endothelial cells on the surface of a microchannel to create a connected single-layer cell sheet. These endothelial cells simulate an endothelium.^6,7 Angiogenic sprouts can be stimulated by introducing a pro-angiogenic factor gradient, such as culturing stromal cells in an adjacent chamber or flowing a pro-angiogenic factor in another microchannel. The self-organization model is to co-culture endothelial and stromal cells in a three-dimensional (3-D) matrix. The 3-D matrix keeps the endothelial and stromal cells suspended in the matrix instead of settling down to the bottom of the culture chamber. Hypoxia or a high interstitial flow is introduced in this 3-D construct to create a physiological environment that can stimulate vasculogenesis.⁸ The endothelial cells can self-organize into a 3-D vascular network through the vasculogenic process.^9,10

These monolayer and self-organization models have also been applied to develop tumor models with vessels. For example, cancer cells can co-culture with endothelial and stromal cells in a self-organization model. Cancer cells can develop into multiple small clusters, while endothelial cells grow into a perfused vascular network.^11,12 To prevent the outgrowth of cancer cells, the loading concentration is less than 2% of the total cell concentrations. On the other hand, co-culturing a tumor spheroid in a self-organization model has also been reported.^13,14 This approach can provide a better tumor structure with a vessel network growing to the proximity of the tumor spheroids. Most tumor spheroids made of different tumor cell lines did not realize an embedded vessel in the tumor spheroid and this only was very rarely for certain types of tumor cell lines.¹⁵ Hasse et al. demonstrated that significant vessel reduction and vessel barrier functions in the proximal region around the tumor spheroids can occur after a 7 day culture, suggesting that tumor-associated endothelial dysfunction can affect the development of a vascularized tumor model.¹⁴ Wan et al. reported that by co-culturing a tumor spheroid with its outer surface seeded with fibroblasts, it can directly co-culture with endothelial and stromal cells. Enhanced vessel formation around the tumor spheroid can be achieved.¹⁷ A tumor spheroid model with perfusable vessels was made possible using a co-cultured spheroid made of tumor and stromal cells or a tri-cultured spheroid made of tumor, stromal, and endothelial cells. It was found that tri-cultured spheroids had a higher success rate in developing perfusable vessels than a co-culture spheroid.^15,16 These results suggest that stromal cells can promote tumor vessel formation in a spheroid model system, and endothelial cells can create connections with surrounding vessels.

To mimic the interaction between a tumor and its surrounding vessels, the concept of developing a vessel network in advance of growing a tumor tissue in a separate chamber is reported. For example, a self-organization model is used to develop a vascular network for the first 7 days. Then, cancer cells are seeded into two adjacent chambers to stimulate angiogenesis.^18,19 A similar approach is also reported using an open-top microfluidic device. The monolayer^20–22 or the self-organization model^22,23 is applied to develop a vascular network in the bottom chamber. Then, a large tumor spheroid is loaded into the open chamber. Angiogenic sprouting can be stimulated by developing a vasculature before loading the tumor spheroid,^21–23 but not in the case that cells and tumor spheroid were seeded simultaneously.²⁰ The angiogenic vessels can wrap around the tumor spheroid, and migrated tumor cells can also create a partial fusion with vessels. A better tumor spheroid model with vessels can be achieved using a tri-cultured tumor spheroid made of tumor cells, fibroblasts, and HUVECs. For example, vasculogenic vessels can sprout and grow toward the tri-culture tumor spheroid and create perfusable vessels with HUVECs inside the spheroid.²¹ Nevertheless, angiogenic vessels have only successfully grown into a tumor spheroid if the spheroid is a co-culture or tri-cultured one.

To better mimic the tumor environment, Shirure et al. studied the interactions between the in vivo tumor and its surrounding vessels, which involve the diffusion of tumor-secreted morphogens and the interstitial fluid flow between vessels and the tumor chamber.¹⁸ This work suggests that the Péclet number (Pe) of an in vivo tumor is between 0.1 and 13. The lower limit of VEGF is Pe = 0.1 with an interstitial flow at 0.1 μm s⁻¹. Recreating these lower bond values, it is demonstrated that tumor angiogenesis can be stimulated.

In summary, developing a quiescent vessel network before seeding a cancer cell suspension or a tumor spheroid in a separated chamber is a more mature method for stimulating tumor angiogenesis on a chip. This two-step method is still a challenge in microfluidics due to the complex procedures and the necessity to create proper physical contact and ECM substrates at pores that link the developed vessels and the post-loaded tumor constructs.

In this paper, we report a one-step method that can load the vessel and tumor constructs on day-0. These two constructs can grow simultaneously and interact after their structures are developed. The complex loading and operational procedures of the two-step method can be eliminated. This one-step method was achieved by using the concept of the microfluidic resistive circuit to create a diffusion-dominated environment with Pe < 6.3 × 10⁻⁵. This design used diffusion to create a hypoxic environment in both microchambers. Vasculogenesis and angiogenesis can be stimulated to develop vessels. In a standard diffusive-dominated environment under Pe < 0.1, morphogen, like anti- and pro-angiogenic factors, released from a growing tumor to the adjacent vessel chamber follows Fick's law of diffusion. This makes it hard to control the interaction range between the growing vessels and the tumor in the vessel chamber. It is found that a matured vessel structure cannot grow, and tumor angiogenic sprouts cannot be stimulated.

To manipulate the diffusion pattern in such a low Pe range, we used the microfluidic resistive circuit to create a meticulously controlled 3-D pressure gradient inside the vessel chamber. It can spatially suppress the tumor-secreted morphogens to be just around the pores that connect the two chambers. Then, the vessel network and tumor can develop nearly independently at the early stage. Vasculogenesis can be stimulated to grow a vessel network in the vessel chamber. At the latter stage, the steep concentration gradient of tumor-secreted angiogenic factors can form near pores. The developed vessels and the tumor can interact to stimulate angiogenic sprouts in the tumor chamber. Experimental studies showed that this method can successfully develop a tumor model from a cancer cell suspension or a tumor spheroid. In addition, sprouted vessels can wrap around and grow into a tumor spheroid. Detailed design and experimental studies of the developed dual-chamber microphysiological system (DC-MPS) based on the one-step method are discussed. The capability of using the developed tumor model for drug screening is also presented.

2. Control the diffusion pattern and the physiological environment of the DC-MPS

The dual-chamber microphysiological system (DC-MPS) was made of poly(dimethylsiloxane) (PDMS), and its design is shown in Fig. 1A and B. Its equivalent circuit is shown in Fig. 1C, and the fabricated and assembled device is shown in Fig. 1D. The DC-MPS is separated into three devices, where the dual-chamber (DC) device was sandwiched between two long-microchannel (LM) devices. They were connected by silicone tubes (jumpers). The top vasculo-chamber (C_V, vessel chamber, LWH = 4.5 mm × 1.5 mm × 0.25 mm) was designed to develop a vessel network through the vasculogenic process. It was seeded with human umbilical vein endothelial cells (HUVECs) and normal human lung fibroblasts (NHLFs) in a fibrin gel. This co-culture system has been widely used to develop a vascular network through vasculogenesis and angiogenesis.^7–23 The bottom angio-chamber (C_A, tumor chamber, LWH = 3 mm × 1.13 mm × 0.25 mm) was designed to stimulate the angiogenic process and to grow a tumor model. It was seeded with a cell suspension composed of SW480 human colorectal cancer cells and NHLFs in a fibrin gel. Or a SW480 spheroid with a NHLF cell suspension in a fibrin gel. It was connected to the vasculo-chamber through five 50 μm wide pores. This design allowed the two constructs to have proper contact and fusion.

Fig. 1 An illustration of the long microchannel design for the dual-chamber microphysiological system (A and B), its equivalent resistive circuit (C), and an image of the fabricated DC-MPS device (D).

The physiological environment of these two microchambers was controlled by a microfluidic resistive circuit.²⁴ It is constructed by two symmetric long microchannels (SLM) (Fig. 1A). For ease of comparison with the equivalent circuit, we make each section of the long microchannel and its equivalent resistor have identical colors in Fig. 1A–C. The concept of long microchannels was previously reported to create different physiological environments for studying vasculogenesis.⁸ This method is based on the linear pressure drop of a long microchannel. By connecting a microchamber at different locations of a long microchannel, specific pressure can be applied to the connected pores to create a diffusion- or a convection-dominated culture chamber.^8,25 This paper applied this method to induce a small 2-D pressure gradient across the two chambers, and two diffusion mechanisms were introduced to stimulate neovascularization and control vessel–tumor interactions. The following sections describe the design concept.

2.1 Create a hypoxic environment for neovasculogenesis

To create hypoxia for stimulating vasculogenesis in the C_V chamber and angiogenesis in the C_A chamber, the pressure drop across the two chambers was made to be less than 0.1 Pa (10.2 μm H₂O) to create a diffusion-dominated environment. Then, hypoxia can be induced by placing the device in a 5% O₂ incubator, as verified in a previously reported paper.⁸

This low-level pressure drop was created by sandwiching the two chambers between two symmetric long microchannels in the DC device shown in Fig. 1A, including red channel and black jumpers (Z₁ & Z₄), gray (Z₂) and green (Z₃) microchannels. The total length was 64.5 mm with a 0.25 mm by 0.2 mm cross-section. The two ends of symmetric microchannels were merged and connected to two LM devices through jumpers. The LM device had a 263.5 mm long microchannel with a 100 μm by 100 μm cross-section (blue), creating a large equivalent resistance of Z_L. This microfluidic resistive circuit was driven by hydrostatic pressure. It was controlled by two glass reservoirs connected at the entrance and exit of the long microchannel (Fig. 1D). They are labeled as P_H and P_L in Fig. 1A. These LM devices provided two large resistances, significantly reducing the pressure drop across the DC device. The two LM devices also provided high hydraulic resistance to maintain a low volume flow rate for long-term culture. Using a P_H–P_L= 98 Pa (10 mm H₂O) driving pressure, the pressure drop along the section of Z₁ + Z₂ + Z₃ + Z₄ was only 0.164 Pa, which was only 0.17% of the total pressure drop. In addition, the symmetric connections created a much smaller pressure drop across the two chambers along the x-axis (Fig. 1B). The high resistance of this long microfluidic channel also maintained a relatively steady driving pressure over 24 hours. The 10 mm H₂O driving pressure only decreased to 9 mm H₂O in one day, and only 150 μl of media flowed through the device. The media height was adjusted daily to maintain the driving pressure and the resulting physiological condition.

This mechanism was verified using the finite element method (FEM). The simulated pressure distribution is shown in Fig. 2A. The pressure range was only 0.1 Pa, and the gap between contour lines was 0.001 Pa. It verified that the pressure applied to the two chambers can be nearly identical. The calculated Péclet number was lower than 6.3 × 10⁻⁵, suggesting a diffusion-dominated environment was successfully created. This result was verified by simulating FITC dextran delivered from the two microchannels. Fig. 2B shows the concentration profile at the 6-th h after flowing dextran in the symmetric microchannels. The symmetric diffusion profile is evident.

Fig. 2 Simulation results of the DC-MPS device (A–E) and the DC-MPS device without bottom pores (F–J): pressure distribution (A and F), diffusion patterns of dextran diffused from microchannels at 6 h (B and G), velocity profiles of V_y (C and H), diffusion patterns of dextran diffused from the C_A chamber at 12 h (D and I) and 24 h (E and J). The step values for the contour lines are 1 mPa (A and F), 0.05 mol m⁻³ (B and G), 2.5 nm s⁻¹ (C and H), and 0.05 mol m⁻³ (D and E, I and J). TP: top pores; BP: bottom pores; AP: angio-pores.

2.2 Create a 2-D pressure gradient to control vessel–tumor interactions

The culturing scenario of the one-step method is different from the two-step method. The two-step method requires having a quiescent vessel network before introducing a tumor construct for mimicking vessel–tumor interactions,^18–23 in which Pe number ranged between 0.1 and 13 with convection and diffusion contributed within one order of magnitude.¹⁸ For the one-step method, the vessel construct needs a few days to develop a vessel network while the tumor grows in the other chamber. At this early stage, the influence of tumor-associated anti-angiogenic factors needs to be suppressed in the C_V chamber to facilitate vessel formation. Once the vessels and tumor are developed, they need to interact to grow a tumor model with vessels through angiogenesis.

Due to the need to maintain hypoxia for neovascularisations, the diffusion-dominated environment had Pe < 6.3 × 10⁻⁵, in which the diffusion contributed four orders of magnitudes higher than convection. Under this condition, the diffusion pattern follows Fick's law of diffusion and usually cannot be manipulated.^8,25 To overcome this limitation, we report a new method to shape the diffusion pattern and range of the tumor-secreted morphogens from the C_A chamber into the C_V chamber. The concept was to confine the tumor-secreted morphogens to just around the angio-pores (APs in Fig. 2), which linked the two chambers. The approach was to design the spatial locations of pores connected to the C_V chamber. First, the two symmetric microchannels were connected to the top of the C_V chamber at the middle of each side with top pores (TPs). Then, the microchannels connected to the bottom side of the C_V chamber with bottom pores (BPs), which were placed 20% farther away from the TPs along the edges. Next, the microchannels were connected to the C_A chamber with two cancer pores (CPs). The width of these pores was 50 μm. Each microchamber can be equivalent to 3 parallelly connected resistors (Z_V and Z_A). The central resistor of the two microchambers was connected through a resistor Z_P, representing the APs connecting the two chambers. The high resistance of microchambers was due to small pore size and 3-D cell constructs.^8,9

To manipulate the diffusion pattern within 0.1 Pa range, the gray channel (Z₂) between TP and BP was designed to be much longer than the green channel (Z₃) between BP and CP. Their corresponding pressure drops were 94.2 mPa and 1.9 mPa. Combining the designed spatial offset between the TPs and BPs, a steep pressure gradient within 94.2 mPa was created in the C_V chamber, particularly next to the TPs and BPs. This effect can be observed from the simulated contour plot of pressure shown in Fig. 2A. This meticulously controlled 2-D pressure gradient created a V-shape velocity profile in the negative y-direction (V_y). It acted like a V-clamp (Fig. 2C) to suppress morphogen diffusion from the developing tumor in the C_A chamber at a velocity below 10 nm s⁻¹. Fig. 2D and E show the simulated diffusion pattern at 12 h and 24 h after loading dextran in the C_A chamber. The diffusion pattern was effectively suppressed (indicated by arrows), suggesting that the majority of tumor-secreted morphogens can be confined to just near the APs. This result also demonstrated that the diffusion range can be suppressed for more than 24 hours. The induced concentration gradient did not wipe out in a couple of hours like in the case of both diffusion and interstitial flow are involved.²⁶ Note that the diffusion range can further be narrowed down by moving both TPs and BPs toward the center of the C_V chamber, and the pressure gradient can be enhanced by increasing the length of the gray channel (Z₂).

Using this design, we can largely reduce the range of tumor co-opt vessels and tumor-secreted anti-angiogenic factors in the C_V chamber at the early stage. Thus, the anti-angiogenic factors secreted by co-opted vessels and tumors can be confined in this area, and vessels can properly form in the majority of the C_V chamber. At the latter stage, the steep concentration gradient near the APs created a directional gradient of tumor-associated pro-angiogenic factors for angiogenesis once the vessels are developed to interact with the tumor.

The effectiveness of this design can be verified by comparing it with the case of having both BPs removed and driving it with identical pressure. Fig. 2F–J show the simulation results. It is clear from the pressure contour plot that the 0.1 Pa pressure drop was gradually decreased from the C_V to the C_A chamber (Fig. 2F), and the V-clamp was not observed (Fig. 2H). The diffusion rate from the long channel was slower (Fig. 2G), and the diffusion range from the C_A chamber into the C_V chamber was much broader and deeper with a higher diffusion velocity, as shown in Fig. 2H–J. This result demonstrated that without a spatially designed pressure gradient to suppress the diffusion range, the signals released from a developing tumor can diffuse throughout the C_V chamber. It suggests that the tumor-associated anti-angiogenic factor can largely hinder vessel formation in the early stage. We also compared this diffusion pattern and range with the case of the DC-MPS device operated under reverse pressure. The pressure gradient was in the same direction as the diffusion of tumor-secreted factors. The FEM analysis results are shown in Fig. S1.† It was found that the diffusion range of the case without BPs was still larger than the one with reverse driving pressure.

2.3 Advantages of the one-step method

The advantage of the one-step method over the two-step method is simplicity and robustness. First, tumor cells or a tumor spheroid can be loaded into the C_A chamber right after loading the vessel construct in the C_V chamber. There is no need to wait 5 to 7 days to develop a quiescent vasculature. It can also bypass the challenge of needing to prevent media leakage or drying out through open-to-air pores during the initial vessel development stage. Second, the angiogenic sprouts will not be hindered due to trapped air bubbles at AP pores during the post-loading process. Media can eventually replace these air bubbles, but vessels in the C_V chamber cannot sprout into the C_A chamber due to the lack of extracellular matrix (ECM) in this region. This design can also solve the problem of fibrin gel degradation, which can almost be gone after 7 to 9 days. The fibrin substrate at angio-pores may be insufficient to support the angiogenic process after loading the tumor construct on day-7. In the present one-step method, the 3-D cell constructs in two chambers can grow simultaneously and form physical connections, followed by executing angiogenic sprouting to create fluidic connections. The initial cell concentrations of the two loaded 3-D cell constructs are designed to allow the developed tissue to fill both chambers and angio-pores in 7 days. Lastly, developed vasculature tissue of the two-step method has low mechanical strength and is vulnerable to elevated pressure during tumor construct loading.

2.4 Validation of the DC-MPS device

To connect the vasculogenic and angiogenic vessel network to the outside world, the side of the two symmetric long microchannels connected with the two chambers was lined with HUVECs (5 × 10⁶ cells per mL) to facilitate anastomosis between the developed vessel network and two long microchannels through pores. After completion of the connections, the two LM devices were disconnected from the DC device. Then, four pipette tips were tugged into the two entrances (P_A and P_B) and two exits (P_C and P_D) of the two symmetric long microchannels (Fig. 1A). Since the two symmetric long microchannels were placed on the opposite sides of the two microchambers and were not connected, a nearly steady pressure can be applied to the vessel network by making the media height of the two tips in one channel to be higher than the other two tips in the other channel.

To validate the performance of the tumor model developed in the DC device, different concentrations of the chemotherapy drug, FOLFOX, were delivered through the developed vasculogenic vessel network and then to the angiogenic vessels of the tumor by maintaining a 10 mm H₂O pressure difference between the two symmetric long microchannels. Since the DC-MPS device guided angiogenic vessels sprouting into the tumor chamber, these vessels did not connect to the outside world. Thus, the developed vessel network can simulate tumor vessels that usually do not connect to a venue and do not have an exit route. It can provide a more mimetic model for drug delivery.

A set of experimental conditions was conducted to verify the effectiveness of the DC-MPS system and the one-step method, as listed in Tables 1 and 2. The vessel construct was composed of HUVECs and NHLFs in a 1-to-1 ratio. Five tumor constructs were studied for the C_A chamber, including 3 different NHLF: SW480 ratios, a SW480 spheroid (340 μm) in a NHLF suspension, and a NHLF-only construct (F₁). The DC-MPS device was first cultured in a 5%O₂/37 °C incubator for 7 or 10 days with EGM™-2 medium without VEGF and bFGF. Then, HUVECs were lined in the two symmetric channels, and the media was replaced with EGM™-2 medium. The device was cultured until day-14 or day-17 in a 5%CO₂/37 °C or a 5%O₂/5%CO₂/37 °C incubator. This process allowed us to investigate the angiogenic process developed under the media saturated with 20% (Hyperoxia) or 5% oxygen (hypoxia). They are labeled as H and h in Table 2, respectively. Table 2 also summarizes the notations for all experimental conditions. For example, h₇H₁₄ represents the DC-MPS device cultured in a 5%O₂ incubator (h) for the first 7 days, and HUVECs were lined in the long microchannels on day-7. Then, this device was cultured in a 20%O₂ incubator (H) until day-14, followed by fixation and inspections.

Table 1 3-D construct seeding conditions

Experimental cond.	C V	C A	Notation
	H : F	S : F
Units: 10^7 cells per mL, H: HUVEC, F: NHLF, S: SW480, CV: vasculogenic chamber, CA: angiogenic chamber.
Vasculogenesis & angiogenesis	1 : 1	0 : 1.5	F1
Vasculogenesis & tumor angiogenesis	1 : 1	1.5 : 0	S1
	1 : 1	0.375 : 1.125	S1F3
	1 : 1	0.25 : 1.25	S1F5

---

Table 2 Experimental conditions

Seeding		HUVECs in SLM		Inspection	Notation
Day-0		Day-7		Day-14	h7h14, h7h17
		Day-10		Day-17	h10h14, h10h17
Day-0		Day-7		Day-14	h7H14, h7H17
				Day-17

---

3. Results

3.1 Effectiveness of the V-clamp in the DC-MPS

To demonstrate the effectiveness of the V-clamp induced by the 2-D pressure gradient, 70 kDa FITC dextran was added to the fibrin gel and loaded into the C_A chamber. The diffusion pattern in the C_V chamber was investigated. Fig. 3A and B show the intensity profiles of FITC dextran at the 12th and 24th hour in the C_V chamber. They are close to the simulated profiles shown in Fig. 2D and E. The FITC dextran was successfully suppressed and can be confined near the APs for at least 24 hours, as indicated by white arrows. This effect was further investigated by comparing the intensity profiles along the red dashed line across the two chambers (Fig. 3A and B). The experimental and simulated results are shown in Fig. 3C and D. These results clearly show that a steep concentration gradient was formed near the pore region, as indicated by black arrows. It suggests that the V-clamp can confine vessel–tumor interaction range just near the pore region.

Fig. 3 Experimental results of the concentration profiles of FITC dextran diffused from the C_A chamber toward the C_V chamber (A and B) and diffused from the symmetric microchannels (E and F). (Scale bar = 500 μm) measured (C and G) and simulated (D and H) intensity profiles across the chambers. Experimental studies on Ang-2 distributions next to the AP regions (bottom) in the C_V chamber at 72 h (day-3) after seeding, including (I) S₁, (J) S₁F₅, and (K) S₁F₅ with a reversed pressure gradient. Nuclei, vessels, and anti-angiogenic factor are stained with H33342 (blue), CD31 (green) and Ang-2 (yellow). (Scale bar = 250 μm) (L) the measured average Ang-2 intensity from the APs (0 μm) to the center of the C_V chamber (750 μm).

Fig. 3E and F show the experimental results of flowing FITC dextran in the symmetric long microchannels, and the diffusion patterns in both chambers were monitored at the 6th and 12th hour after loading. The diffusion pattern at the 6th hour was similar to the simulation result shown in Fig. 2B. The intensity profile along the red dashed line in Fig. 3E and F were measured and compared to FEM results, as summarized in Fig. 3G and H. The measured intensity profiles observed a nearly symmetric pattern (Fig. 3G). The slight asymmetric pattern at the 12th hour (blue line) could be attributed to fabricated long microchannels and pores were not entirely symmetric. It is close to the simulated symmetric diffusion patterns shown in Fig. 3H. These results verified that using a symmetric design of the long channel and the pores connected to the two chambers, the pressure drop across the two chambers in the x-axis can be largely suppressed.

3.2 Suppression of the anti-angiogenic factors

To verify that the V-clamp can effectively suppress anti-angiogenic factors in the C_V chamber, we investigated the distribution and expression level of angiopoietin-2 (Ang-2) on day-3 after seeding. Angiopoietin-2 is a cytokine that disrupts vessel formation.^28,29 We inspect 50% of the area in the C_V chamber next to the angio-pores (APs). Fig. 3I and J compare the experimental conditions of 100% SW480 (S₁) and 16.7% SW480 (S₁F₅) in the C_V chamber (Table 1). The Ang-2 distributions are shown in yellow fluorescence, and the developing vessels under vasculogenesis are stained with CD31 in green. The Ang-2 level was found to be much lower in S₁F₅ than in S₁, which also had a relatively thicker vessel structure. The averaged Ang-2 intensity of S₁F₅ and S₁ conditions measured from just next to the APs (0 μm) to the center of the C_V chamber are shown in Fig. 3L as green and blue lines. It shows that the level of Ang-2 for both conditions was successfully suppressed in the C_V chamber. The Ang-2 level of S₁F₅ and S₁ conditions was highest at 278.9 μm and 385.3 μm away from the AP regions, which were 17.7% and 24.4% deep into the C_V chamber. This experimental study shows that the V-clamp was more effective on the S₁F₅ than the S₁ case, suggesting that a stronger V-clamp was needed to suppress the S₁ condition since a much higher cell concentration was seeded.

To further verify the V-clamp function, we reversed the direction of the driving pressure of the S₁F₅ condition. That is, the P_H and P_L were switched in Fig. 1A. Simulated pressure distribution is shown in the Fig. S1A.† A small pressure gradient from the C_A to the C_V chamber was created, enhancing the diffusion range of tumor-secreted morphogens into the C_V chamber, as shown in Fig. S1C and D.† It also made the concentration gradient near the AP region much smoother than in Fig. 2D and E. The V-clamp was abolished, and the morphogens and cytokines expressed from the developing tumor were gradually pushed into the C_V chamber. The distribution of the Ang-2 and vessel structures are shown in Fig. 3K. The averaged Ang-2 level was much higher than the S₁ condition, and its level did not drop back to baseline in the middle of the C_V chamber, as indicated by the red lines in Fig. 3L. This result clearly shows that the V-clamp can effectively suppress the influence of a developing tumor on the initial vessel formation.

Finally, the vessel structure in the S₁F₅ condition also had a relatively larger structure than the other two conditions. These vessels can be stimulated to conduct the angiogenic process. Experimental studies will be discussed in the following sections.

3.3 Sequential stimulation of vasculogenesis and angiogenesis

To verify that the DC-MPS device can sequentially stimulate vasculogenesis in the C_V chamber and angiogenesis in the C_A chamber, we studied the vessel formation in the two chambers by seeding NHLFs (F₁) in the C_A chamber. Three sets of cultural conditions were studied, including h₇H₁₄ and h₇H₁₇, h₇h₁₄ and h₇h₁₇, and h₁₀h₁₄ and h₁₀h₁₇. Experimental results of these 3 sets of cultural conditions are shown in Fig. 4A–F. The vessel networks developed by vasculogenesis in the C_V chamber were first investigated and compared. It was found that vessel networks were properly formed in all conditions. The level of vessel formation was quantified by the percentage of vessel area and junction density. They are summarized in Fig. 4G and H. A_Top, A_Mid, and A_Bot represent the sampling areas in the C_V chamber taken from the farthest to the nearest location to APs. A_Tot is the average of all three regions. The statistical analysis shows that the percentage of vessel areas and junction densities did not significantly differ across the C_V chamber for each condition except the h₇H₁₇ condition. The value of A_Tot from larger to smaller was h₇h₁₄ > h₁₀h₁₄ > h₇H₁₄ and h₇h₁₇ > h₁₀h₁₇ > h₇H₁₇. On the other hand, h₇h₁₄ has a higher junction density than h₁₀h₁₄ and h₇H₁₄, and the h₁₀h₁₇ condition was also higher than the h₇H₁₇ condition. These results suggest that maintaining hypoxia throughout the 14 day or 17 day culture is an important physiological cue for the vasculogenic process.

Fig. 4 Fluorescent micrographs of developed vasculogenic vessel network and angiogenic vessels using F₁ (A) h₇H₁₄, (B) h₇H₁₇, (C) h₇h₁₄, (D) h₇h₁₇, (E) h₁₀h₁₄, and (F) h₁₀h₁₇ conditions, where vessels and nuclei were labelled with CD31 (green) and H33342 (blue). (Scale bar = 500 μm) statistical analysis of the area percentage (G) and junction density (H) of developed vasculogenic vessels in the C_V chamber, and the analysis of the length (I), diameter (J), and branch (K) of angiogenic sprouted vessels in the C_A chamber. Sequential fluorescent micrographs of angiogenic sprouted vessels in the C_A chamber under F₁(h₇h₁₄) condition (L). (Scale bar = 250 μm).

To investigate the effectiveness of the angiogenic process in the C_A chamber, the length, diameter, and branch of angiogenic sprouted vessels were investigated, as summarized in Fig. 4I–K. Fig. 4I shows that the length of sprouted vessels was longer for the 17 day culture than the 14 day culture in all conditions, suggesting that the angiogenic vessels continued to grow in the C_A chamber. In particular, the h₁₀h₁₇ condition (Fig. 4F) has the longest sprouting length. The average length was 1.3 mm, and sprouted vessels were nearly grown into the entire C_A chamber. This condition also had the largest average vessel diameters (Fig. 4J), 55.7 μm, while all other conditions ranged between 28.3 μm and 42.1 μm. Furthermore, it also had the largest average number of branches per sprouted vessel (Fig. 4K). It had 7.33 branches per vessel, while other conditions were less than 3 branches per vessel. It suggests that the h₁₀h₁₇ was the optimal condition to stimulate both vasculogenesis and angiogenesis in the DC-MPS device.

Finally, GFP-HUVECs and NHLFs were seeded in the C_V chamber to visualize the sequential stimulation of vasculogenesis and angiogenesis. Fig. 4L shows the experimental result using the h₇h₁₄ condition. It was found that vessels started to grow in the C_V chamber and the pore region (day-2). Angiogenic sprouting started on day-4 and grew into the C_A chamber until day-12. This result demonstrated the sequential stimulation of vasculogenesis and angiogenesis in the DC-MPS device.

3.4 Development of a tumor model using a SW480 and NHLFs suspension

Our experimental findings suggest that it is important to maintain hypoxia throughout the culturing process. Thus, we compare the effectiveness of using h₇h₁₄, h₇h₁₇, h₁₀h₁₄, and h₁₀h₁₇ conditions to develop a tumor model in the DC-MPS device. Three different SW480 : NHLF ratios were seeded in the C_A chamber with identical total cell concentrations, including 1 : 0 (S₁), 1 : 3 (S₁F₃), and 1 : 5 (S₁F₅). The corresponding SW480 percentage was 100%, 25%, and 16.7%, respectively. Fig. 5A–F show the vessel structures developed under conditions of S₁(h₇h₁₄), S₁F₃(h₇h₁₄), and S₁F₅(h₇h₁₄). Fig. 5B, D, and F are the corresponding zoomed-in images of SW480 tumors developed in C_A chambers. Experimental results showed no angiogenic sprouting in the 100% SW480 (S₁) condition (Fig. 5A and B), and vessels in the C_V chamber had a much smaller vessel area and were loosely connected. In contrast, both S₁F₃ (Fig. 5C and D) and S₁F₅ (Fig. 5E and F) successfully developed vasculogenic vessel networks and angiogenic sprouts. The sprouting length was longer in the S₁F₅ condition. The influence of the tumor on the vasculogenic vessels in the C_V chamber was inferred from the percentage of vessel area and junction densities, as shown in Fig. 5G and H. It clearly shows that vessel areas and junctions of the S₁ condition were much less than the other two conditions with lower SW480 concentrations. The vessel area of S₁F₃ and S₁F₅ were 4.2 and 3 times larger than the S₁(h₇h₁₄) condition, and they were 2.1 and 3 times larger than the S₁(h₁₀h₁₄) condition.

Fig. 5 Fluorescent micrographs of the vasculogenic vessel network and a tumor model with angiogenic vessels developed using (A and B) S₁, (C and D) S₁F₃, and (E and F) S₁F₅ conditions, where vessels, nuclei, and SW480s were labeled with CD31 (green), H33342 (blue), and EpCAM (red). (Scale bar = 500 μm) statistical analysis of area percentage (G) and junction density (H) of developed vasculogenic vessels in the C_V chamber, and the analysis of the length (I), diameter (J), branch (K) of angiogenic vessels, and the percentage of tumor area (L) in the C_A chamber. Sequential fluorescent micrographs of angiogenic sprouted vessels in the C_A chamber under S₁F₅(h₇h₁₄) condition (M). (Scale bar = 250 μm).

The length, diameter, and branch of angiogenic sprouted vessels were also used to quantify the level of the angiogenic process in the C_A chamber, as shown in Fig. 5I–K. Fig. 5I shows that the angiogenic vessel length of the S₁F₅(h₇h₁₄ & h₇h₁₇) conditions was about 3 times longer than the S₁F₃(h₇h₁₄ & h₇h₁₇) conditions. The vessels in S₁F₅(h₁₀h₁₄ & h₁₀h₁₇) conditions were 3.6 times longer than in the S₁F₃(h₁₀h₁₄ & h₁₀h₁₇) conditions. The average vessel length for S₁F₅ and S₁F₃ was 0.34 mm and 0.39 mm. As for the vessel diameters (Fig. 5J), there was no significant difference among the tested conditions except for the S₁F₃(h₁₀h₁₇) case. The S₁F₃(h₁₀h₁₄ & h₁₀h₁₇) result suggests that vessel regression could occur when SW480 cells grew into a larger tumor size. In addition, there was nearly no branch formation for angiogenic vessels grown in S₁F₅(h₁₀h₁₄), S₁F₅(h₁₀h₁₇), and all S₁F₃ conditions (Fig. 5K). In contrast, an average of 2.2 and 2.6 branches were identified for S₁F₅(h₇h₁₄) and S₁F₅(h₇h₁₇) cases. This result suggests that lining HUVECs in the long microchannels on day-7 can have an early completion of the anastomosis to allow media to flow into the vessel network. It could reduce the influence of the growing tumor on developed vessels.

The size of the developed SW480 tumor models was compared by the percentage of the projection area, as shown in Fig. 5L. The tumor projection area of S₁(h₇h₁₄), S₁F₃(h₇h₁₄), and S₁F₅(h₇h₁₄) conditions after 14-day culture were 91.8%, 80.6%, and 57.3%. The growth of the S₁(h₇h₁₄) and S₁F₅(h₇h₁₄) conditions also are summarized in the Fig. S2A.† The projection area of the developed tumor in S₁(h₇h₁₄) increased from 60.4% to 91.8%, and the area of S₁F₅(h₇h₁₄) increased from 42.1% to 57.3%. Since the 3-D tumor was grown three-dimensionally, the growth rates of the developing tumor were much larger. The SW480-only condition S₁(h₇h₁₄) had much higher tissue volume since its initial cell number was 6 times more than the S₁F₅(h₇h₁₄) condition. The high area percentage of the S₁(h₇h₁₄) tumor demonstrated that a high initial concentration of tumor cells could overgrow over the developing course of a tumor model.

Finally, we also use GFP-HUVECs to visualize the angiogenic process in the C_A chamber using the S₁F₅(h₇h₁₄) condition, as shown in Fig. 5M. It was found that the vasculogenic process occurred in the pore region on day-2, but the angiogenic process did not start until day-8. It was 4 days later than the 100% NHLFs F₁(h₇h₁₄) condition (Fig. 3L).

3.5 Development of a tumor model using a SW480 spheroid and NHLFs suspension

Fig. 6 shows the experimental results of using a SW480 spheroid and NHLF cell suspension as the cell construct in the C_A chamber. The optimal h₇h₁₄ condition found in section 3.4 was applied for this study. This experiment found that elevated shear stress can act on the surface of the SW480 spheroid during loading, and small pieces of SW480 clusters can peel off. This effect results in a cell construct composed of a large SW480 spheroid with multiple small SW480 clusters. We separated loaded constructs into two groups based on areas occupied by smaller SW480 clusters. One had a total area larger than 5% of the C_A chamber, and the other was less than 5%. They are labeled as SR₁ and SR₂, respectively. The growth rate of the large tumor spheroid is summarized in the Fig. S2.† The projection area increased by 54.5%. It suggests that the microfluidic resistive circuit created a physiological environment that can allow the tumor spheroid to grow in the C_A chamber. Fig. 6A shows the vessel structure of the SR₁ case. A vessel network can be developed adequately in the C_V chamber without a noticeable influence from the tumor spheroid and clusters in the C_A chamber. Angiogenic sprouts were successfully stimulated and grew into the C_A chamber. However, the sprouted vessels did not grow toward the tumor spheroid. In contrast, the vessel length was significantly longer for the SR₂ case, as shown in Fig. 6B. Vessels grew deeper into the C_A chamber and toward the 625 μm SW480 spheroid. These angiogenic vessels can wrap around the tumor spheroid and grow into the SW480 spheroid. This result can be observed from the 3-D confocal scanned images shown in Fig. 6C and D, and its recording in the Video-S1.mp4.†Fig. 6E and F summarize the analysis of the angiogenic vessel length and depth in the C_A chamber. It was found that the length and depth of SR₂ angiogenic vessels were 1.8 times longer and 2 times deeper than in the SR₁ case. The average vessel length of the SR₂ case was 0.86 mm long, and the average depth inside the C_A chamber was 0.71 mm.

Fig. 6 Fluorescent micrographs of the developed vessel network and a tumor model using (A) SR₁ and (B–D) SR₂ conditions, where vessels, nuclei, and tumor were labeled with CD31 (green), H33342 (blue), and EpCAM (red). Statistical analysis of the length (E) and depth (F) of angiogenic sprouted vessels in the C_A chamber [scale bar = 500 μm (A and B) & 250 μm (D)].

3.6 Validation of drug testing

To validate the capability to deliver drugs to the developed tumor through angiogenic sprouted vessels, we conducted a series of experiments to study the dosage effect of FOLFOX on the SW480 tumors. The optimal condition of S₁F₅(h₇h₁₄) was applied in this study, and the experimental protocol is listed in Table 3. To visualize the angiogenic process in the C_A chamber and the anastomosis at the pores connected to the long microchannels, GFP-HUVECs were used in this experiment. GFP-HUVECs lined the long microchannels on day-7, and FOLFOX was introduced on the day-16 and administrated for 48 hours. The dosage response was verified by measuring the level of EpCAM after fixation on day-18.

Table 3 Drug test

Seeding	HUVECs in SLM	FOLFOX	Notation
Day-0	Day-7	Day-16day-18	h7h16h18

---

To verify that the vasculogenic and angiogenic vessels were connected and perfused, we flew 70 kDa FITC dextran solution into the long microchannels for visualization. The fluorescent intensity of sprouted vessels in the C_A chamber was monitored. Fig. 7A and B show an example of the experimental result. It demonstrated that the FITC dextran solution can flow into angiogenic vessels. It was evident by the cross-sectional intensity of sprouted vessels shown in Fig. 7B, which is the intensity profile of the yellow line in Fig. 7A. The GFP-HUVECs contributed to the two peaks on two edges of the vessel, and the FITC solution filled the vessel cavity.

Fig. 7 A and B fluorescent micrograph and an intensity profile of FITC dextran-filled angiogenic sprouted vessels in the C_A chamber. Developed tumor model treated with 4% (C and D) and 200% (E and F) FOLFOX, where vessels, nuclei, and tumor were labeled with CD31 (green), H33342 (blue), and EpCAM (red). (Scale bar = 250 μm) comparison of normalized EpCAM intensity under different FOLFOX concentrations (G).

Fig. 7C–D and E–F show the EpCAM fluorescent intensity of SW480 tumors treated with 4% and 200% FOLFOX, respectively. The level of EpCAM expression was much reduced with a higher FOLFOX dosage. Fig. 7G compares the average intensity of positive EpCAM area under different dosage conditions. The intensity level was normalized with respect to the EpCAM level of the untreated control group for each set of experiments. It was found that the level of EpCAM was gradually decreased along with a higher dosage of FOLFOX, which met the standard effect of FOLFOX treatment.²⁷

4. Discussion

Studies have found that angiogenesis does not always involve the early stage of tumor development. Tumors may grow avascular or co-opt surrounding vasculature to support their growth. At this stage, anti-angiogenic factors, such as angiopoietin-2 (Ang-2), are upregulated. It causes vessel destabilization and regression, and they may dissociate from tumors.^28,29 At a later tumor development stage, pro-angiogenic factors, such as vascular endothelial growth factor (VEGF), are upregulated in the tumor margin to stimulate angiogenesis from surrounding vessels that have not regressed. Ang-2 can also upregulate in this region and become a pro-angiogenic factor to facilitate the angiogenic process.³⁰

Thus, a tumor interacts with its surrounding vessels differently during early and later tumor development stages. The vessel co-options of a growing tumor at its early stage could be the primary reason that vessels could not form adequately in a co-culture system. Thus, studies have found that the two-step method is a more reliable approach since the vessel structure is already formed, and a two-chamber system can better mimic the vessel–tumor interactions. However, as we discussed, the two-step method still has some fundamental challenges.

To bypass the inherent issues of the two-step method, we report the one-step method that allows the vessel and tumor constructs to be loaded and grow simultaneously in a DC-MPS device. Its operation procedure is more robust and simpler, and the pores are leak-proof and bubble-free. To spatially control the dynamic interactions between a developing vessel network and a growing tumor, the microfluidic resistive circuit is applied to create hypoxia for neovascularization and a V-clamp to confine tumor-secreted morphogens just near the APs between the two chambers. This unique design allows the vessel and tumor constructs to interact differently throughout development. At the early stage, vasculogenesis can be stimulated to grow a vessel network in the C_V chamber with minimal influence from the tumor that is still developing in the adjacent chamber. At the post-tumor-development stage, the induced steep concentration gradient at pores mimics vessel–tumor interactions to stimulate angiogenesis to grow vessels toward the tumor. The FEM and experimental studies (Fig. 2 and 3) verified this effect. These results also suggest that only the vessels near the APs could encounter tumor-secreted morphogens and become co-opt vessels. Thus, upregulated anti-angiogenic factors can be confined to this region.

To the best of our knowledge, this is the first microfluidic method that can dynamically adjust the vessel–tumor interactions along the course of the vessel and tumor development process. This dynamic mediation can be inferred from our experimental findings.

The experimental findings in Fig. 4G show that vessel areas in the C_V chamber did not significantly differ between 14 day and 17 day cultures using 100% NHLFs (F₁) in the C_A chamber. It suggests that the developed vasculature can be nearly stable. In contrast, similar vasculogenic vessels were only formed in S₁F₃ and S₁F₅ but not in the S₁ condition (Fig. 5G). The vessel area and junction density were smaller in the A_Bot region of the S₁F₃(h₁₀h₁₄) condition (Fig. 5G and H). It suggests that the anti-angiogenic factors could be higher near the pore region and can affect the level of vessel formations. In contrast, this effect was not observed in S₁F₅ conditions, suggesting that the V-clamp worked adequately. Note also that the S₁F₃(h₇h₁₄) vessel area was larger than the S₁F₃(h₁₀h₁₄) condition. It suggests that the influence of anti-angiogenic factors can be reduced by flowing media into developed vessels through earlier anastomosed vessels, which was achieved by lining HUVECs on day-7 instead of day-10. These results suggest that the induced V-clamp can confine anti-angiogenic factors at the APs region under a lower concentration of SW480s. A steeper and narrower V-camp is needed to have similar performance for the S₁ and S₁F₃ conditions. This requirement can be achieved by moving TPs and BPs toward the center of the C_V chamber and increasing the length of the gray channel (Z₂).

On the other hand, it was found that the angiogenic process was stimulated in the NHLF-only F₁(h₇h₁₄) condition on day-4 (Fig. 4L). In contrast, the angiogenic sprouting was not started until day-8 for the S₁F₅(h₇h₁₄) condition (Fig. 5M). These results suggest that anti-angiogenic factors were upregulated during the initial growing period of the SW480 tumor, and pro-angiogenic factors were upregulated at a later stage to stimulate the angiogenic process. This effect of the anti-angiogenic factors could be further inferred from the study on different SW480:NHLF ratios. It was found that there was no angiogenic sprouting in the S₁ condition. Active angiogenic sprouting was found in S₁F₃ and S₁F₅ conditions, and the activity was higher in S₁F₅ conditions. The S₁F₅ condition did not have a noticeable reduction of vessel formation in the C_V chamber, and much longer angiogenic vessels and more branches were observed. In contrast, the S₁F₃ condition had a lower vessel formation near the pore region on the side of the C_V chamber and less angiogenic activity (Fig. 5G). This correlation suggests that the V-clamp was sufficient to suppress the diffusion range of tumor-secreted signals in the S₁F₅ condition but not for S₁F₃ cases. This difference suggests that it is important to suppress the tumor-secreted morphogens at the early stage of the vasculogenic process. It can indirectly affect the following angiogenic process.

The importance of controlling the diffusion range of initial tumor-secreted morphogens was also evident in the SW480 spheroid cases. The different levels of shedding SW480 clusters show distinct patterns of sprouted angiogenic vessels. The SR₁ condition with many scattered SW480 clusters had a much shorter and shallower vessel formation (Fig. 6E and F). In contrast, the SR₂ condition with very few SW480 clusters had much longer angiogenic vessels, and these vessels grew into the outer region of the large SW480 spheroid (Fig. 6B and C). It is close to the in vivo tumor that the angiogenic process occurred at the tumor margin during the post-development stage. This result suggests that the SW480 cluster may still be in the early growing stage, and the large SW480 spheroid was at the post-development stage. The up-regulation of pro-angiogenic factors in the SR₂ case was high enough to compete with anti-angiogenic factors associated with the small SW480 clusters, and angiogenesis was successfully activated. The effectiveness of using the V-clamp to control the diffusion range from a SW480 spheroid at various locations in the C_A chamber was also simulated and is shown in Fig. S3.† It verified that a steep concentration gradient can also be created for at least 24 hours.

In summary, our study provides experimental evidence that reducing the influence of a growing tumor on developing vasculature in a co-culture system is important. The interactions between the vessel and tumor constructs must be treated differently at the early vessel development stage and the later tumor angiogenesis stage.

Lastly, the effectiveness of the DC-MPS system and the one-step method was experimentally verified. The one-step method can reduce the complexity of the loading procedure of the two 3-D constructs into the two microchambers. However, to facilitate anastomosis to the two symmetric long channels, suspended HUVECs need to flow into the channel and be seeded at pore regions after a 7- or 10-day culture. This step is relatively less labor intensive since suspended HUVECs can directly flow into the two long side channels and allow them to attach to the symmetric long channels. This step can potentially be removed by introducing interstitial flow across the two microchambers after the vessel network is completed since studies have shown that vessels developed from HUVECs can sprout toward the high-pressure side and the pores with a higher shear flow.³¹

5. Conclusions

This paper reports a new microfluidic method to develop a tumor model with angiogenic vessels. The core concept is to confine the vessel–tumor interaction range in a small area at the pore region (APs) between the vessel (C_V) and tumor (C_A) chamber. Thus, the majority of the vessel chamber can properly develop a vessel network through vasculogenesis at the early stage. This small area can also create a steep concentration gradient of pro-angiogenic factors to direct tumor angiogenesis at post tumor development stage. This design enables the one-step method to develop a tumor model with vessels.

Our approach differs from the previous method, which uses a low-level convective flow to counteract the diffusion of tumor-secreted morphogens. We applied the microfluidic resistive circuit and the location designs of pores connected to the dual-chamber, and a V-shape velocity clamp was embedded in the diffusion-dominated environment. A partition created by the steep concentration gradient was formed at pores between chambers. This partition allowed endothelial cells to form vessels and enabled tumor angiogenesis once the vessels and tumor were grown. This one-step method successfully created a co-culture system to grow a tumor model with angiogenic vessels. Its capability to conduct anti-cancer drug tests was also verified. In summary, the present method could provide new insight for developing a tumor model with better dynamic control of vessel–tumor interactions. A further study will verify this method's effectiveness in developing a tumor model using a tumor organoid or a patient-derived tumor xenograft.

6. Experimental section/methods

Cell culture

HUVECs and NHLFs were purchased from Lonza. They were cultured with EGM™-2 endothelial cell growth media bulletKit™ and FGM™ fibroblast growth medium bulletKit™, respectively. GFP expressing HUVECs (GFP-HUVECs) and its culture media (cAP-02) were purchased from Angio-proteomie. Their culture procedures followed the vendors' protocols. The SW480 human colorectal cancer cells were cultured in DMDM/F12 supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin (Gibco™). These cells were cultured in a 5% CO₂/37 °C incubator.

Experimental procedure

Two cell–matrix suspensions were prepared before loading the DC-MPS device. A fibrinogen solution was prepared at a 10 mg mL⁻¹ concentration in DPBS. Then, HUVECs and NHLFs were suspended in a 20 μL fibrinogen solution at a 1-to-1 ratio. This cell suspension was quickly mixed with a 4 μl thrombin solution and loaded into the C_V chamber. This process created a 3-D HUVEC-NHLF construct in a fibrin gel with a total cell concentration of 2 × 10⁷ cells per mL and a thrombin concentration of 4 U mL⁻¹. The second cell–matrix solution followed the same procedure for loading the C_A chamber using the 3 different SW480 : NHLF ratios listed in Table 1 or a SW480 spheroid in a suspension of NHLFs (1 × 10⁷ cells per ml). The SW480 spheroid was developed by loading 1200 SW480 cells in a 96-well low-attachment round-bottom plate (Thermo Scientific™). They were cultured for 6 days to reach an average diameter of 340 μm spheroid before loading. After each cell construct loading, the DC device was placed in an incubator for 30 minutes to allow the completion of gelation.

After loading both chambers, we flowed the EGM™-2 medium into the long microchannels of the DC device and connected two LM devices with jumpers. A 5 mm H₂O driving pressure (P_H–P_L) was applied, and the assembled DC-MPS device was placed in a 5%CO₂/37 °C incubator for 6 hours. Then, the culture media were replaced with EGM™-2 medium without VEGF and bFGF, and the driving pressure was increased to P_H–P_L = 10 mm H₂O. Different experimental conditions were conducted, as listed in Table 2. After experiments, standard fixation and staining protocols were conducted. The level and distribution of the Ang-2 cytokine were investigated using PE-labeled angiopoietin-2 antibody (CliniSciences, orb124448). This experiment was repeated 3 to 4 times and was studied at 72 h (day-3) after seeding. The vessels were immunostained with FITC-conjugated anti-human CD31 antibody (eBioscience™, WM59), and nuclei were stained with Hoechst® 33 342 (Invitrogen™, H21492).

Drug test

To evaluate the feasibility of using the DC-MPS to study drug dosage response. The chemotherapy drug, FOLFOX, was used.^11,24 It was composed of 5 μM oxaliplatin, 10 μM leucovorin, and 100 μM 5-fluorouracil. It was labeled as 100% FOLFOX in this study. Five different concentrations were studied, including 1%, 4%, 10%, 100%, and 200%. After developing the tumor model for 16 days, FOLFOX was administrated through the long microchannels for 48 hours, followed by the fixation process. The efficacy of the FOLFOX treatment was verified by investigating the fluorescent intensity of epithelial cell adhesion molecule (EpCAM) using PE-conjugated CD326 antibody (eBioscience™, 12-9326-42).

Quantification and analysis

To study the resultant mass transport in the DC-MPS device, a FITC labeled dextran solution (Sigma Aldrich, 70 kDa MW) was added into the long microchannels or the C_A chamber to investigate the diffusion pattern for 24 hours. The developed vessel network in the C_V chamber was analyzed by selecting three 250 μm by 250 μm square areas from the farthest to the nearest region of the APs. The percentage of vessel area and junction density for each selected region were quantified. The angiogenic sprouted vessels in the C_A chamber were quantified by measuring the length, diameter, and number of branches per sprouted vessel. Lastly, the evaluation of FOLFOX treatment was quantified by measuring the average EpCAM intensity per positive area. These studies were conducted using an inverted fluorescent microscope (Olympus X71) and a confocal microscope (Zeiss LSM780). Images were analyzed using the ImageJ software. Each condition was conducted at least 3 times, and Student's t-test was applied for statistical analysis.

Fabrication

4′′ silicon wafers were first cleaned with a Piranha solution, and the residual oxide layer was removed using a buffered-oxide etchant. After dehydration at 120 °C, standard photolithography was conducted to make SU-8 master using SU-8 2050. Two layers of SU-8 2050 were spun onto a silicon wafer to reach 250 μm thickness to make a master for the DC device. A single layer of 100 μm thick SU-8 2050 was spun on a silicon wafer to make the master for the LM device. After completing the SU-8 photolithography process, the SU-8 masters were hard-baked at 175 °C for 30 min. Then, PDMS prepolymer and curing agent in a 10-to-1 ratio were mixed and degassed before casting on the SU-8 masters. After curing in a 55 °C oven for 8 hours, the molded 2 mm thick PDMS was peeled off and then bound to another 1 mm thick PDMS sheet using O₂ plasma. Then, one DC device and two LM devices were placed on a 1 mm thick PMMA sheet to complete the device preparation process. PDMS devices and the PMMA sheet were sterilized using an autoclave and UVC light, respectively.

Finite element analysis

A 3-D model of the DC and two LM devices were built in commercial software, COMSOL Multiphysics®. Hydrostatic pressure P_H = 98 Pa and P_L = 0 Pa were applied at the entrance and exit of the long microchannel, respectively. Non-slip boundary conditions were set to all other surfaces. The steady-state analysis of the combined effect of convection and diffusion was conducted to simulate the pressure and velocity distributions in the C_V and C_A chambers. Then, the steady-state result was applied to conduct a time-dependent analysis of the nutrition delivery from the long channels, and the tumor-secreted morphogens diffused from the C_A to the C_V chamber. The density and dynamic viscosity of the media were 0.99 kg m⁻³ and 0.7 mPa. The temperature was 310.15 K. The porosity and permeability of the fibrin gel-filled dual-chamber were 0.99 and 1.5 × 10⁻¹³ m². The concentration of nutrition and anti- and pro-angiogenic factors was 1 mol m⁻³, and their diffusion coefficient was 7 × 10⁻¹¹ m² s⁻¹.

Author contributions

Y.-H. Hsu – conceptualization, supervision, funding acquisition, project administration, methodology, investigation, writing – original draft, review & editing; W.-C. Yang, Y.-T. Chen, C.-Y. Lin, C.-F. Yang – formal analysis, visualization, investigation, validation; W.-W. Liu, S. Shivani– formal analysis, visualization; P.-C. Li – conceptualization, supervision, funding acquisition, project administration, methodology.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Health Research Instituts, Taiwan (R.O.C) (NHRI-EX107-10624EI), and the National Science and Technology Council, Taiwan (R.O.C) (MOST 109-2221-E-002-045-).

References

1. V. Brancato, J. M. Oliveira, V. M. Correlo, R. L. Reis and S. C. Kundu, Biomaterials, 2020, 232, 119744,  DOI:10.1016/j.biomaterials.2019.119744.
2. L. A. Low, C. Mummery, B. R. Berridge, C. P. Austin and D. A. Tagle, Nat. Rev. Drug Discovery, 2021, 20, 345–361,  DOI:10.1038/s41573-020-0079-3.
3. A. E. Danku, E. Dulf, C. Braicu, A. Jurj and I. Berindan-Neagoe, Front. Bioeng. Biotechnol., 2022, 10, 840674,  DOI:10.3389/fbioe.2022.840674.
4. M. I. Bogorad, J. DeStefano, J. Karlsson, A. D. Wong, S. Gerecht and P. C. Searson, Lab Chip, 2015, 21, 4242–4255,  10.1039/c5lc00832h.
5. J. J. Tronolone and A. Jain, Adv. Funct. Mater., 2021, 1(31), 2007199,  DOI:10.1002/adfm.202007199.
6. V. van Duinen, D. Zhu, C. Ramakers, A. J. van Zonneveld, P. Vulto and T. Hankemeier, Angiogenesis, 2019, 22, 157–165,  DOI:10.1007/s10456-018-9647-0.
7. S. Kim, H. Lee, M. Chung and N. L. Jeon, Lab Chip, 2013, 13, 1489–1500,  10.1039/c3lc41320a.
8. Y. H. Hsu, M. L. Moya, P. Abiri, C. C. W. Hughes, S. C. George and A. P. Lee, Lab Chip, 2013, 13, 81–89,  10.1039/c2lc40787f.
9. Y. H. Hsu, M. L. Moya, C. C. W. Hughes, S. C. George and A. P. Lee, Lab Chip, 2013, 13, 2990–2998,  10.1039/ c3lc50424g.
10. M. L. Moya, Y. H. Hsu, A. P. Lee, C. C. W. Hughes and S. C. George, Tissue Eng., Part C, 2013, 19, 730–737,  DOI:10.1089/ten.TEC.2012.0430.
11. A. Sobrino, D. T. T. Phan, R. Datta, X. Wang, S. J. Hachey, M. Romero-López, E. Gratton, A. P. Lee, S. C. George and C. C. W. Hughes, Sci. Rep., 2016, 6, 31589,  DOI:10.1038/srep31589.
12. S. J. Hachey, S. Movsesyan, Q. H. Nguyen, G. Burton-Sojo, A. Tankazyan, J. Wu, T. Hoang, D. Zhao, S. Wang, M. M. Hatch, E. Celaya, S. Gomez, G. T. Chen, R. T. Davis, K. Nee, N. Pervolarakis, D. A. Lawson, K. Kessenbrock, A. P. Lee, J. Lowengrub, M. L. Waterman and C. C. W. Hughes, Lab Chip, 2021, 21, 1333–1351,  10.1039/d0lc01216e.
13. C. Li, S. Li, K. Du, P. Li, B. Qiu and W. Dings, ACS Appl. Mater. Interfaces, 2021, 13, 19768–19777,  DOI:10.1021/acsami.1c03740.
14. C. K. Haase, G. S. Offeddu, M. R. Gillrie and R. D. Kamm, Adv. Funct. Mater., 2020, 30, 2002444,  DOI:10.1002/adfm.202002444.
15. Y. Nashimoto, R. Okada, S. Hanada, Y. Arima, K. Nishiyama, T. Miura and R. Yokokawa, Biomaterials, 2020, 229, 11954,  DOI:10.1016/j.biomaterials.2019.119547.
16. Y. Nashimoto, R. Okada, S. Hanada, Y. Arima, K. Nishiyama, T. Miura and R. Yokokawa, Lab Chip, 2022, 22, 4335,  10.1039/d2lc00597b.
17. Z. Wan, M. A. Floryan, M. F. Coughlin, S. Zhang, A. X. Zhong, S. E. Shelton, X. Wang, C. Xu, D. A. Barbie and R. D. Kamm, Adv. Healthcare Mater., 2023, 12, e2201784,  DOI:10.1002/adhm.202201784.
18. V. S. Shirure, Y. Bi, M. B. Curtis, A. Lezia, M. M. Goedegebuure, S. P. Goedegebuure, R. Aft, R. C. Fields and S. C. George, Lab Chip, 2018, 18, 3687–3702,  10.1039/c8lc00596f.
19. Y. Bi, V. S. Shirur, R. Liu, C. Cunningham, L. Ding, J. M. Meacham, S. P. Goedegebuure, S. C. George and R. C. Fields, Integr. Biol., 2020, 12, 221–232,  DOI:10.1093/intbio/zyaa017.
20. Y. Nashimoto, T. Hayashi, I. Kunita, A. Nakamasu, Y. S. Torisawa, M. Nakayama, H. Takigawa-Imamura, H. Kotera, K. Nishiyama, T. Miura and R. Yokokawa, Integr. Biol., 2017, 9, 506–518,  10.1039/c7ib00024c.
21. Y. Kameda, S. Chuaychob, M. Tanaka, Y. Liu, R. Okada, K. Fujimoto, T. Nakamura and R. Yokokawa, Lab Chip, 2022, 22, 641–651,  10.1039/d1lc00751c.
22. S. Oh, H. Ryu, D. Tahk, J. Ko, Y. Chung, H. K. Lee, T. R. Lee and N. L. Jeon, Lab Chip, 2017, 17, 3405–3414,  10.1039/c7lc00646b.
23. J. Ko, J. Ahn, S. Kim, Y. Lee, J. Lee, D. Park and N. L. Jeon, Lab Chip, 2019, 19, 2822–2833,  10.1039/c9lc00140a.
24. K. W. Oh, K. Lee, B. Ahn and E. P. Furlani, Lab Chip, 2012, 12, 515–545,  10.1039/c2lc20799k.
25. Y. H. Hsu, W. W. Liu, T. H. Wu, C. J. Lee, Y. H. Chen and P. C. Li, Biomed. Microdevices, 2019, 21, 7,  DOI:10.1007/s10544-018-0356-7.
26. V. S. Shirure, A. Lezia, A. Tao, L. F. Alonzo and S. C. George, Angiogenesis, 2017, 20, 493–504,  DOI:10.1007/s10456-017-9559-4.
27. S. Soltanian, H. Riahirad, A. Pabarja, E. Jafari and B. K. Khandani, Daru, J. Pharm. Sci., 2018, 26, 19–29,  DOI:10.1007/s40199-018-0210-8.
28. J. Holash, P. C. Maisonpierre, D. Compton, P. Boland, C. R. Alexander, D. Zagzag, G. D. Yancopoulos and S. J. Wiegand, Science, 1999, 284, 1994–1998,  DOI:10.1126/science.284. 5422.1994.
29. P. Carmeliet and R. K. Jain, Nature, 2000, 14(407), 249–257,  DOI:10.1038/35025220.
30. Z. Huang and S. D. Bao, World J. Gastroenterol., 2004, 10, 463–470,  DOI:10.3748/wjg.v10.i4.463.
31. Y. Abe, M. Watanabe, S. Chung, R. D. Kamm, K. Tanishita and R. Sudo, APL Bioeng., 2019, 201(3), 036102,  DOI:10.1063/1.5094735.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3lc00891f

---