The Applications and Challenges of the Development of In Vitro Tumor Microenvironment Chips

Abstract

The tumor microenvironment (TME) plays a critical, yet mechanistically elusive role in tumor development and progression, as well as drug resistance. To better understand the pathophysiology of the complex TME, a reductionist approach has been employed to create in vitro microfluidic models called “tumor chips”. Herein, we review the fabrication processes, applications, and limitations of the tumor chips currently under development for use in cancer research. Tumor chips afford capabilities for real-time observation, precise control of microenvironment factors (e.g. stromal and cellular components), and application of physiologically relevant fluid shear stresses and perturbations. Applications for tumor chips include drug screening and toxicity testing, assessment of drug delivery modalities, and studies of transport and interactions of immune cells and circulating tumor cells with primary tumor sites. The utility of tumor chips is currently limited by the ability to recapitulate the nuances of tumor physiology, including extracellular matrix composition and stiffness, heterogeneity of cellular components, hypoxic gradients, and inclusion of blood cells and the coagulome in the blood microenvironment. Overcoming these challenges and improving the physiological relevance of in vitro tumor models could provide powerful testing platforms in cancer research and decrease the need for animal and clinical studies.

Introduction

Cancer is a disease involving abnormal cell growth with the potential to invade or spread to other parts of the body. Cancer can affect nearly any organ in the human body, of which the most lethal forms include glioblastomas,^33,45,69,123 lung carcinomas,^41,102,129 and colorectal cancers.^26,35,111 Some distinct characteristics of cancer include self-sufficiency in growth signals, insensitivity to anti-growth signals, apoptosis evasion, the ability to sustain angiogenesis, unlimited replicative potential, and tissue invasion and metastasis.³⁴ Another hallmark of cancer is the tumor microenvironment (TME). The TME is composed of both cellular (e.g. endothelial cells, fibroblasts, immune cells) and non-cellular (e.g. extracellular matrix (ECM), chemokines, cytokines, growth factors) components. Together, these components interact with various cancer cells to impact tumor proliferation, progression, migration, and drug resistance (Fig. 1).⁴⁵ It has also been shown that the TME has various nutrient and oxygen gradients, which can influence drug resistance and the proliferation of cancer stem cells (CSCs).^45,82 Therefore, the TME has emerged as a key area of cancer research that should be explored as a potential target for cancer therapy.

Figure 1.

Tumor microenvironment (TME).¹⁰⁷ The TME consists of multiple cell types, extracellular matrix (ECM), soluble factors, and hypoxic regions. There is an imbalance between oxygen/nutrient supply and demand. The TME contributes to the generation of cancer stem cells (CSCs) and tumor metastasis.

One major challenge that cancer research faces includes the need to develop an accurate and predictive experimental model of human tumor development that accounts for the complexities of the TME. Microfabrication and microfluidic approaches have propelled the development of artificial human tissues and have enabled the recreation of biomimetic systems to replicate in the in vivo tumor. Tumor chips using microfluidic devices complement animal models and are superior compared to simple 2D tissue culture systems. Indeed, this is because they can serve as reductionist platforms for well-controlled microscopic studies of the interaction among tumor cells, immune cells, and other cells in the TME.^70,109 The physical and chemical factors within tumor chips can be controlled and stromal cells can be co-cultured with cancer cells to produce responses that mirror those observed in tumors in vivo (Fig. 1).^72,87,98 For example, Mattei et al. and Ozcelikkale et al. demonstrated that their microfluidic devices produced similar cancer cell responses to those of immunocompromised mice with cell-type-dependent resistance and matching phenotypic changes.^75,87 Tumor migration, multi-organ response, and metastasis in tumor chips have also been evaluated in comparison to in vivo mouse models and have revealed a more consistent outcome than traditional 2D culturing methods.^72,115 These results suggest that in vitro tumor chips have the potential to complement, and even replace, aspects of mouse models for mechanistic studies or requisite pre-clinical data for translation studies. Microfluidic tumor devices enable the observation of unique molecular and sub-cellular processes in real-time by way of microscopy techniques integrated with biosensors.⁹⁷

Therefore, microfluidic tumor models hold potential for the study of cancer progression and metastasis,^6,16,27,110 including the capture and analysis of tumor cell dynamical interactions with stromal cells, immune cells, and other cells in the blood.^2,69,77,85 These models represent an increasingly valuable tool to advance cancer research and develop more effective and specific cancer treatments. The maturation of microfluidic systems has afforded the development of artificial human tumor models that replicate aspects of the TME found in vivo.

Current In Vitro Tumor Models

Fabrication of the In Vitro Tumor Chip

An important challenge in modeling TMEs is to replicate a three-dimensional (3D), multicultural, and microfluidic environment with complex ECM composition.⁹³ Three common methods for fabricating tumor microfluidic chips include lithography, 3D-bioprinting, and scaffolding (Table 1). In the process of photo- and soft-lithography, a silicon wafer is spin-coated with a layer of light-sensitive photoresists (e.g. SU-8 negative photoresist).^{27,43,110,121} By exposing the photoresists to UV light through a micropatterned mask, a structure is left on the silicon wafer. This structure serves as a mold for poured and cured polydimethylsiloxane (PDMS) or other biocompatible hydrogels.^10,33 The advantages of this method include the control and uniformity of thickness and the high resolution of the micropattern. Less common techniques for fabricating PDMS molds include micro-milling and 3D printing.^37,79,88,111

Table 1.

Characteristics and applications of current microfluidic tumor chips.

No	Main cell/organ type	Culture type	Duration	Device dimension	Flow property	Fabrication	Application	References
1	MDA-MB-231, MCF-7 (breast)	Co	5 days	Micropillar array (D x H) 50 × 100 μm2	0.05 μL/min	Soft-lithography	Drug screening/delivery	113
2	MCF-7, MDA-MB-231 (breast)	Mono	N/A	Top layer (W) 300 μm	~ 1 μm/s	Soft-lithography	Drug resistance testing	109
3	HepG2 (liver)	Mono	7 days	Culture chamber (W x L x H) 200 × 200 × 250 μm3, 300 × 300 × 250 μm3	> 100 μL/min	Soft-lithography	Drug testing	94
4	F98 (rat glioblastoma), EC	Co	4–5 days	Culture chamber (W x L x H) 2000 × 8000 × 130 μm3	N/A	Photolithography	Controlled drug release/chemotherapy	128
5	MCF-7 (breast), FB	Co	12 h	Chamber (W x L x H) 780 × 370 × 300 μm3	3.0 μL/min	Soft-lithography	Tumor progression	37
6	HMEpiC, MDA-MB-231 (breast)	Co	3 days	Microchannel array (W x L x H)  60 × 300 × 10 μm3	N/A	Soft-lithography	Cancer migration/drug screening	78
7	HUVEC, FB (lung)	Co	5–6 days	W × L × H 800 × 1300 × 110 μm3	N/A	Photo- and soft-lithography	Microvasculature assay	27
8	HCT-116 (colon), INT-407 (intestine), HepG2 (liver)	Co	24 days	Circular chamber (D x H) 10 × 3 mm2	5 μL/min	3D printing/soft-lithography	Tumor progression/drug screening	111
9	SKOV3 (ovarian)	Mono	9 days	W × L × H 7000 × 10,000 × 600 μm3	1 μL/min	3D printing/soft-lithography	Cancer therapy/drug efficacy	98
10	H1975, A549 and H460 (lung)	N/A	2 days	Channel (W x L x H) 1000 × 49,000 × 50 μm3	20, 30, 40, and 50 μL/min	Photo- and soft-lithography	Isolation of CTCs/drug potency evaluation	43
11	MDA-MB-231 (breast), THLE-3 (liver), C3Asub28	Co	3 days	Vessels (D) 711 and 435 μm	1 dyn/cm2	Micromilling and soft-lithography	Drug delivery/toxicity	88
12	MCF-7 (breast), leukocytes, erythrocytes	Co	1 h	D 35 μm	200 and 400 μL/min, Vol 320 μL	Photo- and soft-lithography	Isolation of CTCs	112
13	MDA‐MB‐231 (breast)	Mono	3 days	Channels D 100 and 200 μm W × H 200 × 200 μm2	15 μL/min, 0.01–0.1 dyn/cm2	CNC machining, soft-lithography	Microvascular tumor modeling	79
14	A549 (lung), MDA-MB-231, MDA-MB-231/BRMSI (breast)	Mono	1–2 days	W × L × H 1000 × 800 × 200 μm3	500 μL/min	Soft-lithography	Drug testing	129
15	BGC823 (gastric), HCT116 (colon), PC3 (prostate)	N/A	N/A	W × L × H 60,000 × 20,000 × 60 μm3	50, 100, 200, 300, 400, 600, 800 μL/min	Wet etching technique, thermal bonding	Isolation of CTCs	124
16	HT-29 (colon)	Mono	7–8 days	Cylindrical microwells (D x H) 880 × 400, 400 × 200, and 200 × 100 μm2	200 μL/min	Soft-lithography	Drug testing	35
17	PC9 (lung), 16HBE (bronchial), HFL1, HA-1800 (astrocytes)	Co	4–50 days	Vascular chambers (D x H) 300 × 100, 1200 × 100 μm2	0.1 μL/min	Soft-lithography	Tumor metastasis modeling	72
18	HepG2 (liver), MCF-7, MCF-7/ADR (breast), FB	Co	5 days	W × L × H 800 × 1500 × 270 μm3	N/A	Soft-lithography	Drug resistance testing	117
19	Airway and alveolar cells, H1975 NSCLC (lung)	Co	28 days	Top and bottom channels (W x L x H) 1000 × 16,700 × 1000, 1000 × 16,700 × 200 μm3	60 μL/h	Stereo- and soft-lithography	Drug testing	41
20	HepG2 (liver), U251 (glioma)	Mono	2–4 days	Microwell array (D x H) 100–600 × 3000 μm2	1 μL/min	Photo- and soft-lithography	Drug testing	120
21	MCF-7 (breast), MDA-MB-231 (breast)	N/A	N/A	Circular channel (D) 250 μm Trap channel (W x L x H) 100 × 400 × 130 μm3	150 μL/min	N/A	Isolation/CTC identification	60
22	HUVEC, U87 (brain)	Co	2 days	W × L × H 1200 × 12,000 × 700 μm3	0, 5, 10, 20 μL/min	Photo- and soft-lithography	Drug testing	71
23	GB3-RFP (brain), HUVEC	Co	3 days	D × H 1000, 2000, 3000 × 200 μm2	N/A	Photo- and soft-lithography	Vascular tumor modeling	115
24	MRC-5 (lung), MCF-10A, MCF-7 (breast), A549 (lung)	Co	Few days	Microchannels (W x H) 100 × 100 μm2 Chambers and microwells (D x H) 2700 × 100, 500 × 100 μm2	4.5 μL/min, Vol 100 μL	Soft-lithography	Drug testing	50
25	MCF-7 (breast), HMF	Mono and Co	4 days	Microchambers (D x H) 2700 × 100 μm2 Medium microchannels (W x L x H) 100 × 100 × 100 μm3	4.5 μL/min	Soft-lithography	Drug testing	3
26	MCF-7, MDA-MB-231 (breast)	Mono	3–5 days	Top and bottom channels (W x H) 300 × 100 μm2	1 mm/s	Photo- and soft-lithography	Drug testing/ delivery	87
27	MCF-7, MDA-MB-231 (breast)	In blood	Few hours	Trap and sample prep channels W × H 250 × 80, 800 × 80 μm2 L 500–4700 μm	Velocity 41.6 mm/s	Soft-lithography	Isolation/identification of CTCs	59
28	U937, MDA-MB-231 (breast)	Co	3 days	Microchannels (W x H) 1000, 1200, and 1500 × 120 μm2	0.1–10 μm/s	N/A	Tumor migration/invasion	24
29	HT-29 (colorectal), CCD-18Co	Co	3 days	Channel (W x H) 1000 × 190 μm2	N/A	Photo- and soft-lithography	Tumor modeling	47
30	LNCaP-C4-2 (Prostate)	N/A	Few hours	Channels and trap chambers (W x L x H) 8 × 100 × 8, 30 × 40 × 30 μm3	∼ 2.4 mL/h	Soft-lithography	Isolation of CTCs	99
31	HMSCs derived from the Wharton’s jelly of the UC	Mono	3 days	Channel (H) 135, 150, 165, or 200 μm	Vol 70 μL	Standard dry-film soft lithography	Tumor modeling	103
32	MDA-MB-231 (breast)	Mono	2 days	Channels (W x L x H) 15 × 1300 × 500 μm3	< 1.8 × 10−3 mL/min	Soft-lithography	Tumor hypoxia modeling	56
33	U-251 MG (glioblastoma)	Mono	2 days	Chamber (W x L x H) 2000 × 6000 × 150 μm3	N/A	Soft-lithography	Tumor hypoxia modeling	89
34	HeLa (cervical), HFL1 (lung), FB	N/A	5–7 days	Main and lateral channels (W x H) 1300 × 150, 1000 × 150 μm2	N/A	Soft-lithography	Tumor modeling	66

EC endothelial cells, FB fibroblasts, Mono monoculture, Co co-culture or multiple cultures, Vol volume, D diameter, H height, W width, L length

Another fabrication method is 3D-bioprinting. This method yields more realistic microenvironments and matrices. It also provides the ability to arrange cells in specific 3D structures.⁴⁰ In this technique, bioprinters print structures using bio-inks of various compositions to replicate native organ tissues and vessels.⁹³ Cao et al. designed a tumor-on-a-chip that incorporated bioprinted lymphatic vessels.²¹ These vessels were printed using a multilayer, concentric, and coaxial nozzle to achieve simultaneous printing of a bioink. The bioink utilized in this process was composed of alginate, GelMA, a photo-initiator combined with polyethylene glycol diacrylate (PEGDA), and a cross-linking agent (e.g. CaCl₂).²¹ By using 3D-bioprinting rather than conventional needles and microfabrication techniques, Cao et al. was able to model bloodlines and lymphatic vessels that more closely mimic the behavior and properties of their native counterparts.²¹

Finally, scaffolds can be used for fabrication processes. These scaffolds are typically made with polymeric biomaterials to provide structural support for cell attachment and tissue development.²² For instance, Lee et al. used a 3D scaffold integrated into a microfluidic device to create a transferable substrate that could be isolated afterwards for in vivo use as a transplanted tissue bed. The study showed that the hydrogel scaffold matrix filled the fluid chamber, increased the surface area to volume ratio, and was viewable underneath a microscope.⁶⁴ Furthermore, decellularized tissues were used as natural 3D scaffolds, which created a more realistic microenvironment to couple imaging with mechanistic studies.^76,93

2D Culture Model

Several tumor microfluidic devices still rely on a 2D culture configuration, wherein cells are grown in a monolayer.^1,20,49 Using such an approach, Chang et al. designed a microfluidic device with a serpentine channel to study the migration of tumor cells in response to chemokine and oxygen gradients.²³ Dami et al. also used a 2D monolayer system to develop an automated, high-throughput drug screening system based on a microfluidic cell culture array that was used to evaluate drug cytotoxicity of prostate cancer cells.⁷ Specifically, cells were treated with a sequential, combinatorial concentration of two different drugs generated by two microfluidic diffusive mixers in a device. While microfluidic models employing 2D cultures have been successfully utilized for screening and agonist or challenge studies,⁸ they are limited in use for mechanistic studies because they lack physiologically relevant intracellular interactions, nutrient gradients, and cellular activities that are associated with the native in vivo environment.^98,131

3D Tumor Spheroid Chip

Tumor spheroids overcome many of the limitations found in 2D culture systems because they mimic the physiological, structural, and behavioral characteristics of early-stage tumorigenesis in vivo (Fig. 2).^{3,8,50,55,94,101} This is mainly attributed to the natural hypoxic regions generated along the radii of tumor spheroids, with gradients of nutrients, metabolites, and oxygen that lead to pro-survival gene expression and drug resistance in cancer cells (Figs. 1, 2b, and 2c).^3,55,94,101 Furthermore, co-culturing different cell types (e.g. cancer cells and non-malignant mammary cells such as fibroblasts) within a tumor spheroid chip allows for intercellular interactions that more accurately reflect the in vivo environment (Fig. 2c).^3,50 These interactions include cancer cell–cancer cell, cancer cell–stromal cell, and cancer cell–ECM.^50,74 Due to these advantages, tumor spheroid chips have recently been employed to study the efficacy and efficiency of various cancer therapies such as photodynamic therapy (PDT). Tumor spheroid chips have also been used to screen potential anticancer drugs and compounds (Fig. 2).^55,94,101

Figure 2.

3D tumor spheroid chips. (a) An integrated microfluidic system with a gradient generator and a tumor manipulator to generate tumor spheroids and co-culture. Image was reproduced with permission, from Ref. 74. (b) Spheroids of human hepatocellular carcinoma cells (HepG2) that were generated, cultured, and treated with drugs in culture chambers (width × length, 200 µm × 200 µm) of a microfluidic device. Image was reproduced with permission, from Ref. 94. (c) A 3D multi-cellular spheroid chip to generate tumor spheroids in microwells of a microfluidic device for photothermal therapy in breast cancer research. Image was reproduced with permission, from Refs. 3,50.

Tumor spheroids have been generated through traditional methods including hanging droplets, cubical or semi-spherical chambers (e.g. 96-roundwell plates), and static liquid overlays. More recently, methods relying on microfluidics have been increasingly used to generate tumor spheroids.^{3,8,50,94,101} This is because factors such as volume and flow rate can be easily controlled.⁵⁰ For example, He et al. generated tumor spheroids using micro-cages that were then integrated into a microfluidic culture platform.¹²⁶

3D Organ Tumor Model Using a Multi-Culture System

Multi-cell coculture is the process of culturing two or more different types of interacting cells together to create biomimetic environments of natural tumor tissue (Table 1).³⁸ For instance, Liu et al. employed a 3D heterotypic co-culture platform to perform a parallel, large-scale, tissue-mimicking antitumor investigation of the effects of chemical gradients on cell physiology and outcomes. This system recapitulated several features observed in vivo, including comparable tumor–stromal composition and functional phenotypic gradients of metabolic activity and viability.⁶⁹

Tumor microfluidic devices have also been designed to recreate the 3D characteristics of tumors within infected organs. Lung,^41,72 breast,^17,28,88 and bladder⁷⁰ cancers have all been studied through the application of 3D multi-culture systems to observe cancer growth, dormancy, and response to various therapies such as tyrosine kinase inhibitor^28,41,88 and nanoparticle drug delivery (Fig. 3).^28,58 Within these organ TME models, cancer cells are cultured in ECM with stromal cells, fibroblasts, endothelial cells, and even monocytes and macrophages.⁷⁰ Organ tumor chips have also been used to assess multiple organ responses to treatment. For instance, Liu et al. developed a 3D multi-culture cancerous lung chip connected to a 3D brain chip with a blood–brain barrier to observe lung cancer metastasis to brain tissue in real-time (Fig. 4a).⁷² Similarly, a cancerous breast model was connected to either a cancerous- or noncancerous liver model to observe the toxicity of nanoparticles on the liver in breast cancer treatments.⁸⁸

Figure 3.

Tumor-vascular models for drug toxicity. (a) A biomimetic in vitro tumor chip with a bioprinted blood and lymphatic vessel pair. The blood vessels in a cell culture platform were printed with the bio-inks of PEGDA, alginate, GelMA, photoinitiator, and PEGOA. Image was reproduced with permission, from Ref. 21. (b) A tumor-vascular chip to evaluate the role of tumor environmental stress on NK cell exhaustion. Image was reproduced with permission, from Refs. 11,13. Breast cancer cells were embedded in a microchamber with a lumen lined with endothelial cells to mimic the vasculature. (c) A hydrogel-based 3D microfluidic device to simulate the tumor-vascular microenvironment. Image was reproduced with permission, from Ref. 117. The top layer of tumor cells was separated from the bottom layer of endothelial cells by a membrane. Drugs are introduced through the bottom channel of a blood vessel and diffuse into the cancer cells. (d) A tumor microfluidic device that contains three microchannels, representing a capillary vessel, the ECM, and tumor spheroids, respectively. Image was reproduced with permission, from Ref. 28. The U-shaped microchambers were used to capture cells and deliver nanoparticle-based drugs. (e) A 3D organotypic microfluidic device with hydrogel-based biomaterials to mimic the glioma stem cell (GSC) vascular niche. Image was reproduced with permission, from Ref. 115. The device was used to study GSC–vascular interaction and the effect of endothelial cells on GSC behavior.

Figure 4.

Metastasis models or organ tumors. (a) A multi-organ microfluidic device to study lung cancer-derived brain metastasis. The chip consists of a lung cell chamber simulating the respiratory rhythm to reconstitute the tumorigenesis and intravasation of lung cancer, and a brain parenchyma chamber circled by two vascular channels. Image was reproduced with permission, from Ref. 72. (b) A tumor-metastasis chip with multiple bioengineered 3D organoids to study metastasis of colorectal cancer to other tissues. Image was reproduced with permission, from Ref. 6. Colorectal cancer organoids in a microfluidic chamber are connected to multiple downstream chambers in the liver, lung, and endothelial constructs through the circulating channels.

Tumor Hypoxic Microenvironment Model

Oxygen concentration in the TME plays a significant role in cell metabolic activities, including drug resistance and tumor metastasis. Disruption of the homeostatic balance of oxygen in tissues often leads to hypoxic and, in some cases, anoxic microenvironments. A hypoxic microenvironment forces cells to differentiate and often increases their resistance to cancer therapies, thus promoting tumor progression and metastasis.⁹⁵ This occurrence is due partially to the increased production of hypoxia-inducible factor (HIF), as well as vascular endothelial growth factors (VEGF).⁶⁵ Many studies have created tumor devices to simulate hypoxic environments to model the rates of tumorigenesis and cancer therapy resistance.^44,56,89 For example, Lamberti et al. used a tumor chip to demonstrate that hypoxia affects cancer therapy resistance, specifically in PDT.^61,62 PDT uses light activation of a tumor-specific photosynthesizing drug in an aerobic process. This in turn initiates oxidative stress and reactive oxygen species (ROS) eventually promoting cancer cell death. Koens et al. created a double-layer microfluidic device, which allowed for the control of oxygen tension using two parallel gas channels that were located above media and gel channels to enhance gas exchange.⁵⁶ They also embedded a gas-impermeable polycarbonate film inside the tumor microfluidic device to prevent the diffusion of atmospheric oxygen.

Tumor Vascular Model

One of the most valuable and unique characteristics of microfluidic devices is the versatility of creating a TME that mimics complex vasculature. These models are often employed to observe tumor migration, angiogenesis, and drug delivery pathways (Fig. 3).^{21,28,53,63,80,115–117,121,123} With microchannel networks composed of hydrogel-based structures and/or chambers separated by porous membranes, vascular models can reveal the cellular interactions among endothelial cells, fibroblasts, and cancer cells and their effect on angiogenesis.^{53,80,115–117} It has been found that both angiogenic growth factors and VEGFs increase endothelial sprouting and contribute to the overall TME. These in vitro vascular models are well-suited devices for studying highly vascularized tumors, such as gliomas (Fig. 3e).^115,123 It has even been shown that antioxidants perfuse from the lumen through endothelial cells to glioma cells.⁷¹ As an example, Ayuso et al. developed a tumor vascular device to simulate a blood vessel with a lumen structure lined with endothelial cells. The chip was used to observe the effects of the suppressive environment generated by tumors on NK cell exhaustion. The suppressive tumor environment decreased the NK cell’s cytotoxic capacity and inhibited NK cell surveillance and tumor tolerance (Fig. 3b).¹³

Organ Tumor-Metastasis Model

While immense progress has been made over the past decade to understand tumor invasion and metastasis, many challenging questions remain unanswered. Microfluidic systems serve as novel experimental platforms that can offer fresh perspectives on the lymphatic or multistep metastatic processes (Figs. 3, 4).

Dissemination of tumor cells from the primary site often occurs via the lymphatic system, particularly in common cancers, such as breast and skin. As such, in vitro tumor models with lymphatic vessels have been created to study the pathophysiology behind this process. Some examples of tumor lymphatic models include a five-chamber device with individually controlled flow connecting tubes,⁹⁷ a 4-channel perfusion tri-culture device,¹⁸ and a bioprinted blood and lymphatic vessel pair (Fig. 3a).²¹ Ayuso et al. created hydrogel lumens in the microfluidic device by seeding primary human lymphatic endothelial cells (HLECs) to generate a 3D lymphatic vessel (Fig. 3b).¹¹ Such a model can be applied to study the complex transportation of drugs within a TME,^18,21 as well as the effect of trans-endothelial and luminal flows on cancer invasion and metastasis.⁹⁷

Several tumor microfluidic systems have been developed to study specific dynamic events within the metastasis cascade, including metastases, migration through gels, transmigration through subconfluent endothelial linings, and real-time imaging of invasion and extravasation steps with insight into tumor cell–endothelial cell interactions (Fig. 4; Table 1).^{6,15,16,27,43,68,72,88,110,111,115} Some devices can even connect multiple organ-on-a-chips for real-time monitoring of metastasis from a primary tumor to other organ sites. Examples include the device designed by Skardal et al. which evaluated the metastasis of colon cancer in a gut environment to a healthy liver environment,¹¹¹ and the tumor chip presented by Aleman et al. which observed the metastasis of colon cancer to the liver, lung, and various endothelial constructs (Fig. 4b).⁶

In Vitro CSCs Model

CSC research has emerged as an important area of discovery and one ideally suited for study using microfluidics. Recent findings regarding the impact of hypoxia and nutrient gradients on CSC drug resistance and quiescence have resulted in the natural application of microfluidic devices for gradient creation (Table 1).^{9,23,45,51,52} The incorporation of CSCs into tumor microfluidic chips has also provided a feasible platform for drug screening,^33,125 as well as the design and validation of chemotherapy strategies.⁵⁷ These chips have been used to study the metastasis patterns of CSCs, particularly for glioblastoma (Fig. 3e).¹¹⁵ Since glioma stem cells (GSCs) center around a vascular niche and thus fundamentally underpin the pathology of glioblastoma, Truong et al. developed a tumor chip co-culturing GSCs with endothelial cells in hydrogel-based biomaterials.¹¹⁵ They showed that endothelial cells increase the migration and invasive morphology of GSCs, while maintaining both GSC proliferation rates and phenotype due to the microvascular network. Kuo et al. also studied epithelial–mesenchymal transition (EMT) through the application of putative CSCs. They illustrated that the tumors in a microfluidic chip with a 3D spheroidal culture exhibited higher chemotherapeutic resistance and enhanced metastatic propensity compared to 2D culture.⁵⁷

Applications of Tumor Chips

Drug Toxicity and Screening

One of the most common applications of tumor chips is drug toxicity testing (Table 1). When compared to 2D culture systems, 3D multi-culture microfluidic devices presented drug responses that were more consistent with in vivo responses.⁹⁸ Pharmaceutical compounds such as doxorubicin,^{26,108,109,117,129} tirapazamine (TZP),^23,94 and paclitaxel^26,29 have been studied with various cancers. These cancers range anywhere from brain⁶⁹ to bladder cancers,¹⁴ with breast^{26,29,109,129} being commonly tested (Fig. 3). Additionally, some liposome-delivered drugs have been infused into a microfluidic device with ovarian cancer spheroids to monitor the uptake of different formations and evaluate effectiveness.⁹⁸ Most drug cytotoxicity testing uses live/dead cell viability assay kits to evaluate the effectiveness of a drug since removing the cells from most devices is not possible due to the irreversible bond of PDMS. However, Pitingolo et al. developed a reversible, magnetically sealed microfluidic device that allowed cancer spheroid surface deformation to be evaluated after drug treatment with a scanning electron microscope in addition to an in situ live/dead cell assay for cell viability.³⁵ These findings revealed a loss of structural spheroid integrity for colorectal spheroids treated with the chemotherapy camptothecin in a dose-dependent manner.

As multidrug resistance substantially contributes to the lethality of cancer, platforms such as tumor chips are in high demand. This is because these platforms are capable of efficient high-throughput drug screening.⁵⁵ For example, 3D co-cultures can be implemented into devices and automated with precisely controlled flow rates with gradients.^7,9,23,125 Brain, lung, prostate, and liver cancers have all undergone drug screening within tumor microfluidic chips with drugs such as pitavastatin, irinotecan, curcumin, and tumor necrosis factor-alpha-related apoptosis-inducing ligand (TRAIL).^7,8,33,125 Combination chemotherapy with TRAIL and curcumin was studied by An et al. with a 64-chamber microdevice that employed two microfluidic diffusion mixers with individualized pumps. This allowed the system to be automated and drug concentrations were optimized without using high volumes of reagent, thus increasing efficiency and decreasing cost.⁷ Li et al. created a gradient microfluidic device for multidrug resistance screening. This device was able to achieve high throughput, flexibility, stability, and low sample concentration when used to evaluate time-dependent drug efflux kinetics, chemo-sensitizing effect, and cytotoxicity of chemo-sensitizing agents in HepG2 cells.¹²⁵

Delivery of Drugs and Nanoparticles

Tumor microfluidic devices can be used to increase efficacy in drug delivery as they allow for the evaluation of nanoparticle transport in vitro and the optimization of nanoparticle size, shape, and surface designs (Table 1). Since nanoparticles can transport certain drugs such as doxorubicin, microfluidic devices represent a potentially optimal platform to evaluate both the transport success and efficacy of the transported drug or therapy. This is because complex microvascular geometries can be created to mimic the in vivo environment (Fig. 2c).^50,58,79,88 Breast cancer has been studied intensively with this application and it has been found that nanoparticles such as carbon dots diffuse relatively quickly (3 h) through the endothelium without harming the endothelial cells and selectively deliver doxorubicin to the tumor.^{3,28,50,58,79,87,88,109} Zuchowska et al. tested the effects of graphene oxide (GO) on cancerous tissues (e.g. liver, breast, and colon) using a simultaneous 2D and 3D microfluidic culturing device.¹³¹ The device consisted of a bottom PDMS plate with a concave microchannel and microchamber network, and a top plate that sealed the microstructural network and connected the device to a ventilation and perfusion system into which the GO suspensions were introduced. They found that cells in the 2D monolayer model presented more drug carrier resistance for GO nanoparticles than cells in the 3D spheroid model.¹³¹

Tumor chips have also shown that coupling nanoparticles with other therapies such as PDT (Fig. 2c),³ photothermal therapy,⁵⁰ and acoustofluidic therapy¹²⁸ increased transport control and allowed drugs to be exclusively released at the tumor site. Additionally, tumor microfluidic platforms provide the opportunity to evaluate the effect of nanoparticle delivery on other healthy organs, such as the liver, through the application of organ chip technology.⁸⁸ Palacio-Castañeda et al. used a tumor chip in combination with numeric modeling to observe the delivery of engineered proteins into the tumor. They found that the tumor chip more accurately determined the minimum concentration of the proteins delivered for killing extensive tumors than a 2D culture model.⁹⁰ Roh et al. developed both resin-based drop and pillar array chips with alignment stoppers to improve the alignment for uniform placement of spheroids. This eliminated the need for a stereomicroscope as it allowed for the facile and stable transfer of the spheroid array.¹⁰⁰

Biological Function Analysis and Biosensors

Another tumor-chip application is the analysis of intracellular biological functions, such as metabolism. A high metabolism in cancer cells combined with cell growth and poor neovascularization leads to the development of hypoxic regions.⁵⁵ Hypoxic regions can cause cancer cells to adopt a certain metabolism to increase survival. Tumor microfluidic chips are well-suited and optimizable for studying cell energy metabolism because both the oxygen supply and nutrient depletion can be in tumor cells.¹² Sengupta et al. used a tumor chip for the single-cell analysis of metabolites since higher levels of lactate in cancer cells have been found to correlate with increased metastasis.¹⁰⁴

Microfluidic devices can also be used for high-throughput readouts and on-chip monitoring of cancer cells, such as breast and brain cancers.^105,119 Integration of electro-, chemo-, and bio-sensors into the chip allows cellular functions to be monitored in real-time. Using such an approach, Cao et al. integrated an electrochemical cyto-sensor into their device to monitor the apoptosis of HeLa cells under various conditions.²⁰ By designing a device with chemo- and bio-sensors, Weltin et al. measured cell metabolism by dynamically monitoring the pH, oxygen, lactate, and glucose levels in the cancer microenvironment.¹¹⁹ Furthermore, realistic environments for cell–cell interactions in tumor–stromal assays, as well as real-time monitoring of drug delivery and cell response, can be achieved by micropatterning the devices.^73,113,119 Li et al. used a tumor chip to observe tumor progression at an extremely early stage. It assisted not only in identifying growth patterns and cellular behaviors of tumor spheroids of various sizes, but also helped elucidate the effect of tumor progression on peritumoral angiogenesis.⁶⁶

Tumor–Immune Response

Multi-culture tumor chips can serve as research platforms to study the tumor–immune response. Cell–cell communication and diapedesis can be facilitated by designing devices with multiple channels separated by membranes, endothelium, or other biological materials.^2,85,87 For example, Ozcelikkale et al. showed that MCF-7 and MDA-MB-231 cells (two types of non-tumorigenic and metastatic breast epithelial cell line) cultured in a tumor-microenvironment-on-chip (T-MOC) had higher resistivity to hyaluronic acid nanoparticles and doxorubicin than those cultured in traditional 2D cultures.⁹⁶ Furthermore, cell-type dependent resistivity and phenotypic changes displayed in the T-MOC matched those of tumors implanted in immunocompromised mice, which affirms improved accuracy using a T-MOC.⁸⁷ Using models similar to a T-MOC, immunocompromised splenocytes were found to have a less efficient and less effective immune response to murine metastatic melanoma.² Additionally, cancer-associated fibroblasts (CAFs) in breast cancer were discovered to antagonize the long cancer immune response promoted by trastuzumab.⁸⁵ Taken together, these results suggest that the state of the immune system could affect drug cytotoxicity. Jiang et al. developed an array of miniature bioreactors to model tumor–immune interactions. The chip included a high-throughput observation chamber to test the effect of programmed cell death protein 1 (PD-1) antibodies on cancer spheroids and T-cell interactions. The tumor immune interactions were quantified by measuring the concentration of IL-2 using a micro-pillar array.⁴⁸

Cancer Therapy

As chemoresistance arises seemingly inevitability in anti-cancer therapies,^31,32,61 researchers have recently begun investigating the role of the TME in this phenomenon.¹⁰⁶ This shift lends itself well to applications of microfluidic devices as 3D and dynamic tumor models. Chemotaxis has also gained attention within the microfluidic community as nutrient gradients can be created within one chamber of a device. Ovarian,^30,57 lung,^101,102 mouth,⁹ and bone⁴⁹ cancers have all been studied for either chemoresistance or chemotaxis using microfluidic devices. Lung cancer cultured in spheroids under constant perfusion exhibited higher chemotherapeutic resistance to cisplatin than that in 2D cultures. It is theorized that either the influx of new nutrients or the protection of pericytes that coat the spheroid surface caused the observed increase in chemoresistance.^101,102 This theory was further supported using gradient microfluidic devices to show natural migration/invasion of tumor cells in the direction of nutrients or fetal bovine serum.^9,49 Moreover, chemoresistance was exhibited at higher levels in cancer spheroids in a device coated with copolymer-based chains (e.g. cilia) which decrease adhesion to the substrate.⁵⁷

Isolation and Identification of Circulating Tumor Cells (CTCs)

Another application of tumor chips involves sorting and identifying CTCs that detached from primary tumors into blood vessels.⁸⁶ Tumor microdevices can either strictly isolate CTCs (Fig. 5a)^{5,42,43,84,99,118,124} or both isolate and identify CTCs (Table 1).^60,86,122 Many devices use magnetic nanoparticles to isolate CTCs. These devices can also identify CTCs^{42,60,118,112} using magnetic Fe₃O₄ nanospheres linked to antibodies specific for known cancerous antigens—the most common of which is epithelial-cell adhesion molecule (Ep-CAM). Blood samples are then run through a microfluidic device with a magnet incorporated in the design to isolate the CTCs (Fig. 5c). Other forms of isolation devices include direct antibody binding (Fig. 5b),^43,84,86 whereby the device surface is coated with antibodies or biophysical markers are applied. These biomarkers can include size¹²⁴ and comparative deformation⁹⁹ where the devices have decreasing channel size at varying points through the fluid flow to capture CTCs. Some devices even apply the microfluidic phenomena of centrifugal acceleration and varying channel expansion to increase turbulent fluid flow and mix the blood. This in turn makes nanoparticle binding more effective or increases exposure to the antibody-coated device surface (Fig. 5c).^43,112 Breast, lung, gastric, prostate, bladder, rectal, and colon cancers have all been analyzed through the application of CTC sorting devices, which has decreased the processing time to half an hour with an efficiency of about 95%.^99,122 Zhuang et al. also used microfluidic devices to identify CTCs and demonstrated that high metastatic CTCs have higher aggregation capacities than low metastatic CTCs.¹³⁰

Figure 5.

Tumor microfluidic devices for isolating circulating tumor cells (CTCs). (a) A microfluidic device for continuous separation of cancer cells from a heterogeneous mixture (e.g. blood samples) using dielectrophoresis (DEP). Image was reproduced with permission, from Ref. 5. (b) A size-based microfluidic device for isolating CTCs from the blood of patients with metastatic bladder and colorectal cancers. Image was reproduced with permission, from Ref. 84. (c) A microfluidic chip that comprises a micromixer to generate multivortexing flow and a magnetic sorter to remove magnet-coated cells. Image was reproduced with permission, from Ref. 112. The device was used to isolate CTCs from blood samples.

Challenges in Developing Tumor Tissue Chips and Future Perspectives

Mimicking Tumor Microenvironment with a 3D Multi-Culture System

It is crucial to recapitulate the in vivo environment as closely as possible to achieve physiologically relevant results. Novel microfluidic devices not only provide a platform for 3D culture methods, but they also provide a greater chance of viability for mono-, co-, and tri-cultured cell aggregates.^54,83 However, limitations in replicating the in vivo TME remain due to technical difficulties and the complexity and heterogeneity of the tumor environment. Currently, some microfluidic chips are commercially available to researchers, which allows them to create tumor models without the need to fabricate their own complex devices (Table 2).

Table 2.

A list of commercially available microfluidic chips for tumor modeling.

No	Manufacturer	Chips that are used in tumor model	References
1	AIM Biotech, Singapore	A 3D microfluidic chip for a drug test for human glioblastoma	Ref. 91, https://aimbiotech.com/products/
2	Ibidi, US	An Ibidi μ-slide chemotaxis microfluidic device	https://ibidi.com/19-channel-slides
3	PreciGenome, US	Microfluidic mixer, droplet generator, and organ on a chip	https://www.precigenome.com/products
4	ConductScience, Inc., US	Custom microfluidic chips including organ chips, tumor chips, and biosensors	https://conductscience.com/lab/custom-microfluidic-chips/?attribute_product-qty=1
5	Stellar Scientific, Inc., US	Organ on a chip with a perfused 12- or 48-well insert	https://www.stellarscientific.com/lena-biosciences-perfusionpal-12-well-organ-on-a-chip-1-pk/
6	Elve Flow, Inc., France	Microfluidic devices including organ-on-a-chips	https://www.elveflow.com/microfluidics-application-packs/biology-packs/organ-on-a-chip-pack/
7	SynVivo, Inc., US	A cell-based microfluidic platform for drug development and tumor chip	https://www.synvivobio.com/microfluidic-chips/#1593107172212-c852468e-7686
8	uFluidix, Canada	Custom fabrication of microfluidic devices	https://www.ufluidix.com/ufluidix-chips/
9	Dolomite, UK	Microdroplet generation, drug delivery, micromixer, membrane chip, and channels with multiple inlets and outlets	https://www.dolomite-microfluidics.com/product-category/microfluidic-components/microfluidic-chips/
10	Microfluidic ChipShop, Germany	Various types of devices including polymer chips, integrated chips, glass chips, silicone chips, assays, organ chips, and lab on a chip	https://www.microfluidic-chipshop.com/catalogue/
11	Fluigent, France	Droplet and particle generation and organ chips	https://www.fluigent.com/research/instruments/microfluidic-chips/

Multi-culture systems are challenging because adequate nutrients, essential growth factors, and sufficient oxygen must be supplied to maintain cell metabolism and proliferation.⁴ Moreover, each cell type requires specific conditions with respect to nutrients, oxygen, organotypic structure, cultural modalities, and ECM composition and stiffness. It is especially necessary to control the physical parameters of the culture (e.g. nutrient composition and balance, flow rate, shear stress, oxygen level, pH of the medium, temperature, waste accumulation). During the microfabrication processes, there are technical challenges associated with microfluidic device design. Some of these challenges include the ratio of cell-to-media volume, the cell population ratio, access for seeding cells and feeding media, sampling, and device geometry. For real-time analyses of biological function and metabolic parameters, it is also important to consider the sample collection method, access of a sample port, and transparent culture chamber when designing the device. It also is essential to establish methods to verify that distinct cell types are seeded into individual culture regions and that the desired multiple cell-layer configurations are formed.

Considering the hypoxic tumor environment, another challenge is to create a controlled and stable oxygen tension under spatiotemporal oxygen heterogeneity. Oxygen is a small, non-polar molecule; therefore, it readily diffuses through many materials used for the fabrication of microfluidic devices (e.g. PDMS). One way that researchers have worked around this challenge is to place the whole system in a hypoxic chamber^65,68,91,95 or using a low oxygen culture kit.⁸³ Others have covered the devices with an oxygen-impermeable substance layer, which selectively allows oxygen to diffuse.^56,89

In some cases, the generation of the spheroids without air bubbles, irregularities, or ruptures is difficult in itself. After the spheroids are generated, they are often transported to another tissue chip for culturing; however, one difficulty that may arise is the collection and transportation of tumor spheroids without deformation and damage. In other cases, structures such as microwells are used within the devices to form spheroids.¹³¹ For example, He et al. discovered a possible solution to the generation and transportation of tumor spheroids by creating a honeycomb frame cage paralleled by a monolayer of agarose-coated nanofibers and a mesh with relatively large openings. The device was used in the generation of tumor spheroids and then easily integrated into a microfluidic culture system.¹²⁶ Another challenge is associated with maintaining the consistency of both a uniform scaffold of ECM and spheroidal structure. This is imperative for insuring that there are reproducible communications, including cell–cell, cell–ECM, and cell–substrate interactions.

Other Challenges in Modeling Tumor Physiology

Other challenges in modeling tumor physiology are associated with controlling the ECM composition and its mechanical properties. The effects of ECM stiffness on cell behaviors have been extensively studied in 2D models.²⁵ However, because perturbation of matrix stiffness often affects cellular confinement, there have been challenges applying the same concepts to 3D cultures. Pathak et al. developed a matrix platform based on the microfabrication of channels with a defined wall stiffness and geometry to study the effect of ECM geometry on 3D culture. These channels allowed independent variation of ECM stiffness and channel width. They described that with a given ECM stiffness, cells confined in arrow channels migrate faster than wider channels or unconstrained 2D surfaces. They also found that in confinement, cell migration increases as ECM stiffness rises.⁹²

Another challenge is the type of cells being used. Most microfluidic devices use mature tumor cells for their cell cultures; therefore, not much is known about tumor behavior at an early stage. A solution to this problem would be to use stem cell-driven tumor tissue development. However, many issues arise when using stem cells. This is because stem cell differentiation is a complex process that requires the stem cell to be subjected to different forms of stimuli to determine its function. The controlled stimuli that the stem cell experiences and the timing of the stimuli are crucial to reach the desired function and avoid teratoma formation caused by undifferentiated cells. It is challenging to mimic the same signals the stem cells would experience in the extracellular environment due to the complexity.¹²⁷ In addition to considering the importance of the role of blood coagulation proteins in tumor progression, it is essential to include blood cells and the coagulome indicative of the blood microenvironment.^19,81 It remains a challenge to develop an authentic tumor progression model to predict the hemostatic complications occurring in cancer patients and tumor immune response.^36,67 The ultimate goal of tissue models in the near future is to provide a patient-specific tumor chip for personalized medicine and tumor therapy. To achieve this, it is necessary to overcome the challenges associated with device customization and culture parameters.

Future Perspectives

We have outlined several challenges associated with modeling the TME and physiology. Developing tumor chips and reproducing quantitative results remains a complex process. Despite this, however, new technologies are helping to advance and overcome some of the technical limitations. Moreover, tumor chip designs are evolving with improved physiological relevance to better recapitulate in vivo conditions. Indeed, a variety of trends have emerged as the technology for TME chips advances.

One of these trends includes the modification of PDMS, a commonly used material in microfluidic tumor chips. While the utility of PDMS is limited by its high absorption of hydrophobic drugs, recent studies have shown promise for modifying the hydrophobicity of PDMS using a copolymer or 3D bioprinting with biomaterials.³⁹

A further challenge in creating accurate tumor models includes replicating multiple organ interactions. Current TME chip models tend to focus on a single organ; however, there are an increasing number of multi-organ chips being developed. Although the development of these human-on-a-chips remains difficult due to the complex intercellular interactions, they would greatly improve upon the physiological relevance of tumor chips.

Tumor chips have the potential to be used as a platform for personalized tumor treatment.¹¹⁴ Advanced 3D bioprinting technology and the emergence of new biomaterials have largely contributed to creating personalized scaffolds. For example, gene editing technology (e.g. CRISPR-Cas9) has enabled us to create tumor organoids at different tumor development stages that can be personalized by editing the genes of specific patients. Furthermore, combining the personalized tumor organoids model with microfluidic technologies may assist with overcoming the clinical obstacle in drug treatment, building a deeper understanding of tumor progression in patients, and providing personalized therapy of tumors.

Another remarkable trend is to apply computer-aided software associated with artificial intelligence and machine learning to biology and medicine.⁴⁶ This approach has enabled us not only to analyze big data (e.g. cellular images, cell tracking, and histological images), but it has also allowed us to categorize cellular behaviors in response to treatment with various drugs or infections. When combined with microfluidics and tumor chip technology, this software may be useful for tumor progression predictions, cancer modeling, and drug stimulation.

Tumor chips have evolved with the advancement of new technologies, providing a powerful testing platform in cancer research, cancer therapy, and drug development. It is expected that a physiologically relevant in vitro tumor model would replace in vivo testing and minimize animal and clinical studies.

Conclusion

Tumor chips using microfluidic devices have been developed in response to challenges that researchers face using 2D culture systems and animal models in cancer research. This review covered the recent development of TME chips, along with their applications and challenges. The use of the in vitro tumor chip has brought many benefits. Some of these benefits include real-time observation capabilities, precise control of the TME, and increased accuracy as these tumor chips more closely recapitulate the in vivo tumor environment. Tumor chips have been fabricated using photolithography, soft-lithography, 3D bioprinting, and scaffolding. These have been used in various applications, including drug screening, drug toxicity testing, nanoparticle transport, biological function analyses, observation of the tumor immune response, and CTC isolation and identification. Despite the benefits of tumor chips, the ability of current models to model tumor physiology and pathophysiology is limited due to technical challenges incurred during the development processes and the complexity of the TME. These are associated with the replication of the TME including a 3D heterotypic co-culture, hypoxic environment, the transportation of adequate nutrients, ECM composition and stiffness, inclusion of blood cells, and the collection of 3D spheroid cells. However, new technologies are helping to advance and overcome some of the technical limitations. Tumor chips are evolving with improved physiological relevance to better recapitulate in vivo conditions.