Organoid-on-a-chip: Current challenges, trends, and future scope toward medicine

Abstract

In vitro organoid models, typically defined as 3D multicellular aggregates, have been extensively used as a promising tool in drug screening, disease progression research, and precision medicine. Combined with advanced microfluidics technique, organoid-on-a-chip can flexibly replicate in vivo organs within the biomimetic physiological microenvironment by accurately regulating different parameters, such as fluid conditions and concentration gradients of biochemical factors. Since engineered organ reconstruction has opened a new paradigm in biomedicine, innovative approaches are increasingly required in micro-nano fabrication, tissue construction, and development of pharmaceutical products. In this Perspective review, the advantages and characteristics of organoid-on-a-chip are first introduced. Challenges in current organoid culture, extracellular matrix building, and device manufacturing techniques are subsequently demonstrated, followed by potential alternative approaches, respectively. The future directions and emerging application scenarios of organoid-on-a-chip are finally prospected to further satisfy the clinical demands.

I. INTRODUCTION

Traditional tools for studying human biology include 2D monolayer cell culture and animal models. Although 2D cell growth models can be easily established, cells grown in culture dishes cannot achieve most of the complex cellular interactions, such as lack of overall tissue/organ functions and cellular responses to drug stimulation. Animal models, by contrast, serve as a tool to study overall 3D organ functions. However, in addition to ethical issues, limitations of animal studies include the level of organ metabolism, immune function, and tissue morphology that are not consistent with human biology. For example, placing patient-derived cancer cell xenografts into immunocompromised rodents to mimic tumor development would inevitably neglect an in vivo adaptive immune response to cancer.

To surmount these barriers, the self-organizing ability of mammalian cells has evolved from primary tissues to construct different 3D cell models, aiming to simulate the in vivo microphysiological system (MPS).^1–3 Organoid technology, as an innovative 3D model to bridge the gap between 2D culture and in vivo physiology, has gradually matured over recent years. Different from cellular spheroids which are monotypic 3D cellular aggregates, organoids are universally defined as 3D multicellular aggregates capable of recapitulating one or more fundamental organ functions. They offer substantial advantages compared to conventional two-dimensional cultures and animal models. Using the modeling of the nervous system as an example, 2D models capture the temporal sequence of events in corticogenesis within the cerebral cortex but fall short in replicating its critical architectural organization.⁴ This architecture is indispensable for comprehending the neural networks essential for normal function and the disruptions seen in disease states. In contrast, brain organoids mirror the multicellular, three-dimensional nature of central nervous system structures and can replicate early brain development stages.⁵ When it comes to animal models, there are challenges in replicating the function of the human nervous system due to the fact that neurons in the human cerebral cortex originate from a cell type known as the outer radial glial cell.⁶ This specific cell type either does not exist or is found only in minimal quantities in rodents. The sources of self-organized cell clusters include but not limited to various multipotent stem cells, such as induced pluripotent stem cells (iPSCs), embryonic stem cells (ESCs), and adult stem cells (aSCs). Organoid systems can be alternatively built using primary cells and cancer cells to reproduce organs from where they were derived. Tissue structures and functions, in this way, could be optimized to the greatest extent. Currently, organoids have been widely used to study organogenesis, drug discovery, and various inherited diseases, such as tumors, degenerative diseases, and infectious diseases.⁷ For example, organoid models have assisted researchers in discovering the pathogenesis of organs susceptible to SARS-COV-2 infection, which played a crucial role in the development of corresponding treatments.^8–11

Three commonly used techniques have been adopted to generate organoids in vitro. The first and simplest method is extracellular matrix (ECM) scaffolds consisting of highly biocompatible materials, such as Matrigel or collagen gel.^12–15 The second strategy is ESC or iPSC rotary action with spinning bioreactors, primarily aiming to create in vitro human neuronal growth conditions.^16,17 The air–liquid interface (ALI) as the third approach is mostly utilized to construct gastrointestinal organoids.^18,19 Nonetheless, the formation of organoids largely depends on arbitrary self-organization in the culture medium or ECM, without the precisely controllable stimulus, such as nutrient gradient, that is essential for in vivo microenvironment.

With recent advances in microfabrication technology and tissue engineering, a more advanced platform “organoids-on-a-chip” has emerged to take all the aforementioned considerations into account. This platform is generally defined as the 3D miniaturized biomimetic organ models, using microfluidic design lined with cells under the in vitro MPS to accelerate disease modeling and personalized medicine (Fig. 1). Typically, the establishment of organoid-on-a-chips follows the design principles of simplified analysis of target organs. The anatomy of the target organ requires to be understood first, followed by identifying key biological features, such as distinct cell types and organ-specific MPS. Microfluidic channels and tissue compartments are integrated into a microdevice to deliver nutrients to organoids and remove cellular metabolic waste. Most devices are fabricated using polydimethylsiloxane (PDMS) with preferable optical clarity and superior biocompatibility, thus facilitating visualization of dynamic biological activity with high spatiotemporal resolution. Organoid-on-a-chip systems can simulate various crucial parameters present in the physiological microenvironment of organs, including shear forces,²⁰ concentration gradients,²¹ cell morphology,²² tissue boundaries,²³ and tissue–organ interactions,²⁴ and provide the ability to flexibly regulate these factors, a capability unattainable within conventional well plates and Petri dishes. Therefore, the possibility of simulating more complete human biology is greatly increased. A variety of human organs have been simulated using microfluidics, such as lung,^25–27 kidney,^28–30 heart,^31–33 liver,^34–36 vascular network,^37–39 skin,^40–42 cornea,^43–45 and neuronal network.^46,47 Furthermore, multi-organoid models have been built in series or in parallel to investigate underlying interactions between different organs (e.g., liver and gut).^48–52

FIG. 1.

A schematic of the “organoids-on-a-chip” system and the summary description about its fabrication, function, and application.

Although organoids-on-a-chip can reproduce in vivo biological properties, the simulation regarding to functionality of the in vivo counterparts is far from being achieved. In this Perspective paper, we present the current technical issues regarding mainstream ECM preparation, representative device fabrication, and organoids culture, followed by proposing potential solutions, respectively. The objective of this paper is to narrow the maximum extent of the research gap between existing organoid-on-a-chips and the complexity of in vivo human organs. The opportunities of organoid development including integrated vascularization, immunotherapy, drug evaluation, and personalized medicine in the future are also discussed.

II. PROBLEMS IN ORGANOID CULTURE AND THE SOLUTIONS BROUGHT BY MICROFLUIDICS

Advances in stem cell areas provide a powerful source for in vitro modeling of organs and diseases.^53,54 Since Sato et al. developed intestinal organoids with small intestinal villi and crypt structures in 2009,⁵⁵ a series of organoids have been developed (e.g., brain,⁵⁶ kidney,⁵⁷ liver,⁵⁸ and pancreas⁵⁹). Somatic cells taken directly from patients can be induced into human induced pluripotent stem cells (hiPSCs) to customize drug screening platforms for patients. Also, researchers have introduced hiPSCs into organoid-on-a-chips because hiPSCs are available without ethical issues that typically exist in human embryonic stem cells (hESCs). However, organoids tend to suffer from low maturity, and stem cell-derived organoids more closely resemble fetal tissue than adult tissue.⁶⁰ For example, ESCs differentiate into insulin-secreting cells with rather low differentiation efficiency,⁶¹ and their insulin secretion levels in vitro are much lower than those of adult islet cells.⁶² Beyond that, the ESC-derived or iPSC-derived cardiovascular cells are more similar to fetal rather than adult cardiovascular tissue at the transcriptome level.⁶³ Both iPSC-derived cardiomyocytes and fetal cardiomyocytes share similarities in electrophysiological behavior, metabolism, morphology, and size.⁶⁴ Notably, commonly employed organoid preparation methods involve the ECM scaffold,⁶⁵ spinning bioreactor,⁶⁶ and low-adhesion well,⁶⁷ which are time-consuming and require cumbersome manual steps. Normally, iPSC purification from skin or blood samples takes at least several weeks, and iPSC differentiation into the desired cells would take longer.⁶⁸ This hinders the progress of organoids toward practical applications and mass production.

Considerable efforts have been devoted to enhancing the maturation of organoids. The application of biomechanical stimuli,^69,70 such as fluid stimulation,⁷¹ can facilitate organoid maturation. Promoting the resemblance of ECM to the in vivo tissue microenvironment also contributes to organoid construction, which will be discussed in detail in Sec. III.⁷² Furthermore, organoid maturation can be potentially enhanced by the integration of multiple tissues in the organoid-on-a-chip system.⁷³ The provision of diverse cues to the organoid, reflecting various aspects of the in vivo environment, can increase their developmental stage and optimize their function. However, as the fabrication process may confront more challenges with more complex chip designs, the potential trade-off between increasingly integrated organoid-on-a-chips and commercial mass production requires to be critically considered.

Microfluidics offers the possibility to increase the efficiency of organoid formation and reduce manual steps. Zhu et al. achieved in situ differentiation of iPSCs in a microfluidic device, followed by high-throughput production of brain organoids [Fig. 2(a)].⁷⁴ A micro-pillar array was designed for the natural aggregation of iPSCs and the formation of embryoid bodies. During the differentiation process, the culture medium in the microfluidic device was sequentially replaced with neural induction medium and neural differentiation medium, allowing the in situ formation of brain organoids. This approach significantly reduces the risk of cell contamination. Similarly, liver organoids were differentiated in situ from iPSCs on a chip using microfluidic elements.⁷⁸ Fluid stimulation eventually resulted in higher expression of mature hepatic genes in the liver organoids. Additionally, a novel microfluidic chip with a microwell structure was proposed to generate heterogeneous human islet organoids in bulk [Fig. 2(b)].⁷⁵ These organoids exhibited higher glucose-stimulated insulin secretion and Ca²⁺ flux compared to static cultures, indicating that endocrine cell differentiation could also be promoted under fluid stimulation.

FIG. 2.

Improvement in the efficiency of organoid culture using microfluidics. (a) Micropillar-based method for brain organoids, including embryoid body formation, neuroepithelial induction, and neuroepithelial expansion.⁷⁴ Reproduced with permission from Zhu et al., Lab Chip 17(17), 2941–2950 (2017). Copyright 2017 Clearance Center, Inc. (“CCC”). (b) In situ differentiation of hiPSCs into islet organoids-on-a-chip in perfusion conditions.⁷⁵ Reproduced with permission from Tao et al., Lab Chip 19(6), 948–958 (2019). Copyright 2019 Clearance Center, Inc. (“CCC”). (c) Automated culture of tumor organoids using droplet microfluidics.⁷⁶ Reproduced with permission from Jiang et al., Cell Rep. Med. 1(9), 100161 (2020). Copyright 2020 Creative Commons CC BY. (d) Droplet microfluidics-based mass production of hiPSC-derived islet organoids.⁷⁷ Reproduced with permission from Liu et al., Adv. Sci. 7(11), 1903739 (2020). Copyright 2020 Creative Commons CC BY.

Droplet microfluidics as another technique for the establishment of organoids has a great number of advantages, such as high efficiency,⁷⁹ desirable size consistency,⁸⁰ and low requirements of cells and solvents.^80,81 Currently, publications regarding droplet microfluidics have been presented for high-throughput generation of spheroids^82,83 and organoids.^84,85 Jiang et al. developed an organoid platform using droplet microfluidics and an automatic sampling device [Fig. 2(c)].⁷⁶ In this study, organoids from mouse tissues or patient tumors were embedded in Matrigel, followed by being automatically transferred to a 96-well plate for high-throughput drug screening. Liu et al. developed a droplet microfluidic device based on the aqueous two phase systems (ATPSs), which can bulk encapsulate pancreatic endocrine cells into sodium alginate-chitosan microcapsules [Fig. 2(d)].⁷⁷ After 7 days of culture in microcapsules, islet organoids with a diameter of 60–70 μm were formed. Notably, normal physiological functions, such as insulin secretion and glucose-induced responses in these organoids, were discovered.

Despite the aforementioned works aimed to optimize the iPSC differentiation protocol and enhance the efficiency of organoid formation, the issue of intermixed cell subtypes following the differentiation of iPSCs remains to be further investigated.

III. CHALLENGES IN EXISTING ORGANOID CULTURE MATRIX AND POTENTIAL ALTERNATIVE APPROACHES

Animal-derived ECM (e.g., Matrigel) has been widely used as the scaffold for organoid culture.^86,87 However, the composition of Matrigel is complex and unclear.^88–90 Meanwhile, differences in the biochemical and physical properties from different batches are not negligible,^91–93 making it difficult to confirm the ideal reproducibility of organoids in Matrigel. In addition, Matrigel contains growth factors,^93–95 transcription factors,⁹³ and cytokines⁹¹ that would interfere with mechanistic studies of cellular behavior in organoids. Thus, the role of various biochemical factors in organoid morphogenesis is difficult to be clarified. Matrigel is normally derived from mouse sarcoma cells and has immunogenicity issues in applications, such as human clinical transplantation.⁹⁶ However, pathogens, such as lactate dehydrogenase elevating virus (LDHV),⁹⁷ might present in the animal-derived ECM.^98,99 These defects limit the widespread use of organoids in Matrigel, as reproducibility and safety of organoid systems are critical in drug screening and regenerative medicine.

Tissue-specific decellularized extracellular matrix (dECM) is expected to be a strong alternative to animal-derived ECM.¹⁰⁰ As a biocompatible and non-immunogenic scaffold, tissue-specific dECM consists of enzymatic,^101,102 physical^103,104 or chemical removal^105,106 of immunogenic cellular components from human or animal organs/tissues. The dECM is capable of providing native ECM structures and biomechanical cues of specific target tissues for cell adhesion, differentiation, and function.⁸⁸ In addition, the biochemical and biomechanical properties of dECM can be modified by chemical modification^107,108 or hybridization with synthetic materials¹⁰⁹ to meet the requirements of organoid culture. Compared to collagen and Matrigel, tissue-specific dECM, especially for human-derived tissue-specific dECM, is more physiologically relevant to native tissues and circumvents immunogenic problems.

Giobbe et al. presented dECM derived from porcine small intestine (SI) for intestine organoids [Fig. 3(a)].¹¹⁰ Subsequent proteomic analysis demonstrated that several intestinal markers were expressed at a higher level in the intestine organoids generated in SI dECM compared to those in Matrigel. Cho et al. proposed a decellularized human brain tissue-derived brain extracellular matrix (BEM) [Fig. 3(b)] to culture human pluripotent stem cell (hPSC)-derived brain organoids.⁷² The results indicated that the BEM cues significantly enhanced the development of brain organoids, including spontaneous brain morphogenesis, mature neuronal properties, distinct brain cell populations, and electrophysiological activity. Bi et al. developed a detergent-free method for manufacturing decellularized rat pancreatic extracellular matrix (dpECM) hydrogels.¹¹¹ A higher quantity of ECM proteins was found in comparison to their Matrigel-based counterparts [Fig. 3(c)]. Additionally, dpECM hydrogels had more cell-ECM binding proteins and proteases involved in biochemical catalysis, which are required for pancreatic digestion.

FIG. 3.

Promotion in organoid maturation with dECM. (a) The preparation process of the small intestine-specific dECM.¹¹⁰ Reproduced with permission from Giobbe et al., Nat. Commun. 10(1), 5658 (2019). Copyright 2019 Creative Commons CC BY. (b) Bright-field images of (i) raw human brain tissue and decellularized human brain tissue (scale bar = 1 mm) and (ii) brain organoids in BEM and Matrigel at 75 days, respectively (scale bars = 500 μm).⁷² Reproduced with permission from Cho et al., Nat. Commun. 12(1), 4730 (2021). Copyright 2021 Creative Commons CC BY. (c) Enhancement of islet organoid functions using decellularized rat pancreatic ECM: (i) preparation steps of pancreatic-specific dECM of and islet organoids; (ii) immunofluorescence images of endocrine tissues on Matrigel (M)-coated, or Matrigel and Collagen V (M + C) mixed substrates-coated plates (CP is C-peptide; GCG is glucagon); (iii) immunofluorescence images of somatostatin (SST) and pancreatic polypeptide (PPY) showing physiologically relevant insulin-secreting β-cell distributions.¹¹¹ Reproduced with permission from Bi et al., Biomaterials 233, 119673 (2020). Copyright 2020 Creative Commons CC BY.

IV. CHALLENGES IN CURRENT PREPARATION METHODS OF ORGANOID-ON-A-CHIP AND POTENTIAL ALTERNATIVE APPROACHES

The current preparation of organoid-on-a-chip primarily relies on soft lithography using polydimethylsiloxane (PDMS), which is only suitable for laboratory use and not conducive to large-scale production.¹¹² This issue emerges because soft lithography requires equipment that can only be used in the clean room, and the fabrication process is typically time-consumable and high-cost. In addition, the manufacture of increasingly complex organoid-on-a-chip would inevitably involve multi-step lithography,^113,114 further exacerbating the cost disadvantage of soft lithography. Moreover, the adsorption of small molecules (e.g., estrogens¹¹⁵) by PDMS might also affect the quantitative analysis of organoid-on-a-chip.^116,117

Although soft lithography has been used for microfluidic chip preparation for decades, the process is still tedious and involves a large number of manual steps.¹¹⁸ To address this issue, three-dimensional (3D) printing has been presented to produce microfluidic chips in one step, contributing to the reduction of preparation time and minimization of manual operational errors. 3D printing can also be performed in a cost-effective and user-friendly way without an ultra-clean room.^119,120 Currently, three mainstream methods were adopted for 3D-printed microfluidic chips. The first technique involves the fabrication of a master die using a 3D printer as a substitute for the master die produced by conventional soft lithography.^121,122 The subsequent steps remain the same as those employed in soft lithography. The second method entails the 3D printing of a sacrificial layer in the form of flow channels, followed by the removal of this layer after casting a polymer layer, such as a PDMS membrane.¹²³ The third approach is the direct printing of microfluidic devices.^124,125

The preparation of complex structures can become more simplified through 3D printing. Alessandri et al. fabricated a funnel-shaped co-extrusion device [Fig. 4(a)], which was then used to batch-produce core-shell structured hydrogel microbeads with an average radius of 136 ± 28 μm.¹²⁶ The core of the microbeads consisted of Matrigel, while the shell was composed of sodium alginate. Neuronal stem cells cultured in the core of hydrogel microbeads spontaneously differentiated into a 3D neuronal network after 13 days. Kamei et al. presented 3D-printed molds as an alternative enabling rapid prototyping to construct microfluidic devices for cell stimulation with concentration gradient [Fig. 4(b)].¹²⁷ Männel et al. demonstrated the feasibility of 3D printing for non-planar droplet microfluidic chips with a minimum size of 75 μm.¹²⁴ Both single-emulsion and double-emulsion non-planar microfluidic chips were prepared, with droplet generation rates of up to 2800 per second.

FIG. 4.

High flexibility of 3D printing technology in the design and fabrication of microfluidic chips. (a) 3D-printed co-extrusion device (scale bar = 1 mm).¹²⁶ Reproduced with permission from Alessandri et al., Lab Chip 16(9), 1593–1604 (2016). Copyright 2016 Clearance Center, Inc. (“CCC”). (b) (i) Fabrication procedure and (ii) design of a 3D-printed soft lithography mold, and photos of (iii) microchannel-embedded mold and (iv) device.¹²⁷ Reproduced with permission from Kamei et al., Biomed. Microdev. 17(2), 36 (2015). Copyright 2015 Clearance Center, Inc. (“CCC”). (c) Preparation of spiral flow channel with 3D printing.¹²⁸ Reproduced with permission from Hwang et al., Sens. Actuators A 226, 137–142 (2015). Copyright 2015 Clearance Center, Inc. (“CCC”). (d) Production of complex flow channels by sacrificing 3D-printed ABS scaffold.¹²³ Reproduced with permission from Saggiomo and Velders, Adv. Sci. 2(9), 1500125 (2015). Copyright 2015 Creative Commons CC BY. (e) Schematic procedures of three primary molding methods in microfluidic fabrication. Reproduced with permission from Wu et al., J. Biomed. Opt. 16(8), 080901–080912 (2011). Copyright 2011 Creative Commons CC BY.

In 3D printing materials, acrylonitrile-butadiene-styrene copolymer (ABS), photocurable resin, and polyamide (PA) are commonly used. Yet, these materials lack desirable optic transparency and biocompatibility in organoid-on-a-chip. To solve this problem, Bressan et al. 3D printed biocompatible polylactic acid (PLA) on a transparent poly(methyl methacrylate) (PMMA) substrate.¹²⁹ The combination of PLA and PMMA can be achieved without additional adhesive coatings or surface treatments due to their similar molecular structures. 3D printing technology also enables the production of multi-layer chips and spiral flow channels [Fig. 4(c)],¹²⁸ which are always hindered in traditional soft lithography. Thus, design flexibility is greatly enhanced for the development of versatile organ-on-a-chips. Saggiomo et al. employed an innovative method to process three-dimensional twisted and complex flow channels.¹²³ The 3D-printed ABS scaffold was immersed into liquid PDMS. After the PDMS became solidified, the ABS scaffold was removed with acetone [Fig. 4(d)]. Nevertheless, it must be acknowledged that the high accuracy of 3D printing is not still achievable compared to soft lithography (around 1 μm).¹³⁰ The current smallest flow channel is limited to 18 μm,¹²⁵ which is accompanied by an associated degree of inaccuracy. Consequently, 3D printing is rendered inapplicable when components of the microfluidic chip are of dimensions less than 10 μm.

Another popular way to fabricate organoids-on-a-chips is injection molding, becoming the gold standard for mass production and commercialization of microfluidic chips that conventional PDMS production could not achieve. Briefly, the pre-heated and melted thermoplastic materials are injected into the processed molds, followed by being solidified after cooling down [Fig. 4(e)].^131,132 These materials are typically thermoplastics, including polymers and elastomers with outstanding optic transparency. Traditional molds fabricated by high-precision milling are durable, thus allowing the subsequent chip production with high reproducibility. However, the expenditure for the mold preparation would be extremely high, exceeding the budget for researchers.¹³³ To overcome this limitation, rapid injection molding was presented based on nickel-plated wafer or milled molds in a time-saving and cost-effective manner, and mold performance was comparable to hot embossing.^134–136 Nonetheless, the flexibility of rapid injection molding still requires to be improved for complicated structures of organoids-on-a-chips as well as material functions such as elasticity.

Beyond injection molding, hot embossing is also a growing technology in microfluidics, which is achieved by transferring micro-scale patterns to thermoplastic materials, such as PMMA and polycarbonate (PC).¹³⁷ Conventional hot embossing and roller to roller hot embossing are commonly adopted.¹³⁸ The general procedures for both include heating, pressing (molding), and cooling (unmolding), which could be fulfilled without expensive facilities for developing micro-structure [Fig. 4(e)]. The master molds in hot embossing are normally solid metal alloys, such as brass,^139–141 aluminum,¹⁴² nickel,^143–146 and stainless steel,^147–149 which are fabricated primarily through micro-milling, electroplating, lithography, etc. Thus, a single master mold with high-accuracy micropatterns can be reused for mass production. However, challenges still exist in the regulation of parameters to eliminate the possibility of incomplete pattern transfer to achieve maximum precision. Additionally, smooth demolding without any distortion or structural damage of thermoplastic materials after cooling is also required to be critically performed. Compared to PDMS, although the thermoplastic materials could be easily 3D printed or hot embossed, their own limitations, such as poor gas permeability and low flexibility, remain to be resolved. Therefore, the experimental purpose needs to be carefully considered prior to the selection of the optimum fabrication strategy.

V. FUTURE DIRECTIONS OF ORGANOID-ON-A-CHIP

Currently, the cellular composition of organoids tends to be relatively homogeneous. This situation leads to the inability of organoid models to accurately simulate the complexity of organs in the human body. In addition, single-organoid on a chip is unable to reproduce the crosstalk between human organs. Therefore, vascularization, immunization, and systematization of organoid-on-a-chip become increasingly important. Accordingly, vascular networks, immune cell populations, and comprehensive multi-organoid systems require to be further integrated together.

A. Vascularization of organoid-on-a-chip

Vascularization of organoid-on-a-chip is a requisite for successful in vitro modeling, due to the requirements of nutrient supply and further advancement of tissue functions.¹⁵⁰ In addition to supporting the metabolism of organoids, vascular structures also hold significant relevance in disease modeling within the field of biomedical engineering. The presence of vasculature enables the transport of drugs, immune cells, antibodies, immune factors, and other agents to specific cells within organoids, recapitulating the authentic distribution patterns of drugs in the human body. Several studies have demonstrated that vascularized tumor organoids exhibit greater drug resistance compared to non-vascularized ones.^82,151,152 Without the involvement of vasculature, the functions of organs, such as the blood–brain barrier¹⁵³ and the renal glomerular filtration barrier,¹⁵⁴ are impossible to replicate in vitro.

Vascularization has been implemented in brain,¹⁵⁵ kidney,⁵⁷ pancreas,¹⁵⁶ liver,¹⁵⁷ and other organoid models. Two strategies are generally adopted for organoid vascularization. The first is to co-culture a mixture of endothelial cells (ECs) and progenitor cells from the target organ. For example, Takebe et al. successfully co-cultured iPSC-derived intrahepatic cells, human umbilical vein endothelial cells (HUVECs), and umbilical cord-derived mesenchymal cells to construct vascularized liver buds.¹⁵⁷ Similarly, Sobrino et al. introduced colorectal cancer cells, lung fibroblasts, and HUVECs together into a microfluidic chip, followed by co-culture under fluid stimulation to establish a vascularized microtumor (VMT) platform.¹⁵¹ The endothelial cells self-assembled after 5 days to form a vascular network supporting tumor growth. A variety of common anti-cancer drugs were tested in the VMT, which showed higher drug resistance than single tumors.

In the co-culture strategy, HUVECs are widely employed as the preferred choice for the reconstruction of blood vessels. This preference stems from the relative ease of HUVEC isolation and high biocompatibility with other organoids. Notably, organ-specific endothelial cells derived from PSCs become advantageous to develop vascularized organoids. This is because organ-specific endothelial cells could effectively enhance the authenticity and resemblance of the resultant organoids to their native counterparts. Salmon et al. induced hPSCs differentiation toward the vascular and neural domains, which were subsequently co-cultured in a specifically designed 3D-printed microfluidic chip.¹⁵⁸ The co-cultured organoids eventually exhibited certain brain-specific properties [Fig. 5(a)].

FIG. 5.

Various strategies for the generation of vascularized organoids. (a) (i) Schematic of co-culture and (ii) angiogenesis in the 3D-printed microfluidic platform for vascularized organoid cultures on chip (scale bar = 2 mm).¹⁵⁸ Reproduced with permission from Salmon et al., Lab Chip 22(8), 1615–1629 (2022). Copyright 2022 Creative Commons CC BY. (b) (i) Vascular formation was induced within the brain organoid by over-expressing ETV2 and (ii) abundant vasculature after immunostaining at day 30.¹⁵⁹ Reproduced with permission from Cakir et al., Nat. Methods 16(11), 1169–1175 (2019). Copyright 2019 Clearance Center, Inc. (“CCC”). (c) (i) Schematic of preparation and (ii) whole-mount staining results in the fused vasculature and brain organoids (scale bar = 200 μm).¹⁵⁵ Reproduced with permission from Sun et al., eLife 11, e76707 (2022). Copyright 2022 Creative Commons CC BY.

The vascular network successfully invaded into the center organoid without influencing brain organoid growth and characteristics of vascular sprouting.

Another strategy is to obtain vascularized organoids directly from PSCs by regulating the differentiation process of PSCs. Takasato et al. reconstructed renal units by adjusting the exposure time to different growth factors and increasing the proportion of posterior renal mesenchymal formation.⁵⁷ Tubular lumens formed by endothelial cells were observed in the kidney organoid at day 18. This approach is relatively simple without involving exogenous cells or gene editing. However, the resultant vasculature was confined to the extent of the lumen with unknown vascular functionality. Additionally, in view of vasculature originating from the mesoderm, using the same set of PSCs to induce vascular tissues from other germ layers becomes impracticable. Consequently, this strategy is only applicable for constructing mesoderm-derived vascularized organoids. To raise the application potentiality of this strategy, Cakir et al. employed genetic engineering approaches to induce the overexpression of the EVT2 transcription factor in hESCs [Fig. 5(b)].¹⁵⁹ ETV2 serves as a pivotal regulator in the process of embryonic development, particularly in the generation of endothelial cells (ECs) and vascularization. Vascularized human cortical organoids with characteristics of the blood–brain barrier were successfully established. Another study separately developed brain organoids and vascular organoids using iPSCs.¹⁵⁵ Neurotrophic factors were introduced to the vascular organoids to enrich cerebrovascular features. Two types of organoids were combined upon the formation of vascular precursor cells and neural precursor cells [Fig. 5(c)], effectively overcoming the challenge of disparate germ layer origins between vasculature and the brain.

This section elucidates recent breakthroughs and pivotal implementation strategies in the domain of vascularized organoids. Notably, only a limited number of vascularized organoids successfully authenticated the comprehensive functionality of the vasculature (e.g., perfusion). With the exception of vascularized tumor models,¹⁵² vascularized organoids necessitate transplantation into animals to demonstrate their functionality after drug screening in vitro models. The new strategies are still required to be developed in the future for various vascularized organoids with better vascular performance.

B. Immunization of organoid-on-a-chip

The immune system is closely associated with blood vessels responsible for transporting immune cells, immune factors, and antibodies to targeted tissues. Due to the involvement in almost all disease processes, immune tissues become indispensable in organoid-on-a-chips for developing alternative drug testing. In vitro modeling of the immune system provides insight into immune mechanisms in the pathogenesis under controlled experimental conditions. Thus, the course of immune cell-tumor interactions, autoimmune diseases, allergies, and inflammation could be quantified. Isolation and application of immune cells or factors are likely to deepen our understanding of the underlying pathways for the development of new immunotherapies.

A number of organoid-on-a-chips incorporating the immune system have been developed for the recapitulation of different functions in the immune system. Exogenous immune components were mostly introduced such as lymphocytes from peripheral blood.^160,161 Sharma et al. constructed a human bladder organoid-on-a-chip for simulation of the infection process of uropathogenic E. coli.¹⁶² In this study, neutrophils rapidly transmigrated across the endothelial layer and infiltrated the bladder tissue, ultimately forming aggregates near the infection site. Moore et al. developed an EVIDENT microfluidic system capable of studying the interaction between tumor fragments.¹⁶³ This system can accommodate 12 tumor fragments sourced from different patient biopsy samples. The therapeutic effectiveness of immune inhibitor-treated tumor-infiltrating lymphocytes against tumors was verified. In contrast, Jenkins et al. produced tumor spheroids by enzymatically digesting patient-derived tumor tissues, followed by injecting them into a microfluidic chamber for culture.¹⁶⁴ Notably, this report integrated autologous bone marrow cells, lymphocytes, and cancer cells into the spheroids, thus achieving the in vivo high complexity of cancer tissues. Therefore, the direct integration of patient tumor biopsy samples into microfluidic chips is worth considering for establishing more physiologically relevant tumor-immune models.

Chimeric antigen receptor T-cell (CAR-T) therapy, emerging as an immunotherapeutic approach for cancer treatment, becomes widely appealing in the medical community and pharmaceutical companies. CAR-T therapy genetically modifies T cells from patients to express CARs that recognize specific antigens on cancer cells. Wan et al. effectively stimulated tumor vascularization with desirable perfusion ability by culturing fibroblasts alongside pre-formed tumor spheroids.¹⁶⁵ This group provided opportunities for evaluating the CAR-T cell functions on tumor cells in a controlled laboratory setting.

Despite immune system elements being crucial to be incorporated into organoid-on-a-chip platforms, progress in this area is still severely limited. This is partly due to the complexity of the immune system itself. The immune system encompasses an astonishingly vast number of lymphocytes around 5 × 10¹¹ comparable to the number of neurons in the human brain. Additionally, immune responses are intricately mediated by physicochemical factors secreted by lymphocytes. Furthermore, the advancement of this field is also possibly impeded by the competition from humanized mice, which are widely regarded as the more reliable model by immunologists. To overcome these obstacles and build comprehensive immune systems in organoid-on-a-chips, closer collaboration between experts in microfluidics and immunology is necessarily required.

C. Systematization of organoid-on-a-chip

Most physiological activities involve tissue-to-tissue and organ-to-organ interactions. However, inter-organ crosstalk mechanisms, such as the brain-gut axis, are still not fully understood. Single-organoid-on-a-chip can no longer meet the demand for in-depth study of these crosstalk mechanisms. Also, multiple organs in the human body are inevitably involved during the process of drug delivery and release. To simulate more comprehensive physiological conditions, the concept of multi-organoids-on-a-chip (MOOC), namely, body-on-a-chip, was proposed. Briefly, cells from different organs are co-cultured in MOOC with multiple chambers, and each chamber normally represents a specific organ. These chambers are interconnected by microchannels or external tubing to perform a function similar to that of an in vivo vascular network. The interconnected pathways enable the exchange of nutrients, metabolites, and biological factors between different organ modules. MOOC can be used to examine the interactions between different organ tissues or as a one-stop platform for evaluating drug efficacy.

MOOCs typically consist of a pump (e.g., pressure pump, injection pump, and peristaltic pump), and a medium reservoir. The culture medium is sequentially pumped through different organ modules, eventually flowing into a centrifuge tube for collection. Alternatively, a medium can circulate continuously within the MOOC to study the true interactions between organ modules. However, the external pump and reservoir in the MOOC might increase the consumption of medium and dilute biochemical factors (such as cytokines) of interest in the medium. Materne et al. designed a neural-liver MOOC device equipped with an on-chip micropump, significantly reducing the size of the system.¹⁶⁶ Compared to mono-culture, the co-culture mode in the MOOC exhibited higher sensitivity to the neurotoxic substance 2,5-hexanedione, probably due to signaling molecules released by necrotic cells in certain tissue impacted adjacent cells. Zhang et al. designed an automated modular MOOC platform for the long-term and short-term drug evaluation using a liver-heart chip and a liver cancer-heart chip, respectively.¹⁶⁷ This platform incorporates physical sensors to monitor extracellular microenvironment parameters, electrochemical biosensors to detect soluble protein biomarkers, a bubble trap, and a microfluidic breadboard for timed routing of fluids [Fig. 6(a)].

FIG. 6.

Representative MOOC platforms. (a) A MOOC model for real-time monitoring of organoids' behavior using embedded sensors.¹⁶⁷ Reproduced with permission from Zhang et al., Proc. Natl. Acad. Sci. U.S.A. 114(12), E2293–E2302 (2017). Copyright 2017 Creative Commons CC BY. (b) (i) Schematic of the biomimetic vascular system and (ii) Interconnection between biomimetic vascular network and organoids chambers in a human-on-leaf-chip.¹⁶⁸ Reproduced with permission from Mao et al., Small 16(22), 2000546 (2020). Copyright 2020 Clearance Center, Inc. (“CCC”). (c) Human 3X Gut-Liver-Brain (3XGLB) physiomimetic system of the gut-liver-brain axis: (i) pneumatic plates machined in acrylic; (ii) mesofluidic plate machined from monolithic polysulfone; (iii) three-way interaction that the center liver-specific MPS can be fluidically linked to two additional transwell-based MPSs.¹⁶⁹ Reproduced with permission from Trapecar et al., Sci. Adv. 7(5), eabd1707 (2021). Copyright 2021 Creative Commons CC BY.

Cancer metastasis is the process by which cancer cells spread from the primary tumor to other parts of the body. It involves various multi-organ mechanisms that facilitate the dissemination and establishment of secondary tumor sites. Kong et al. constructed a MOOC for the simulation of the lung, liver, bone, and muscle, which enabled the accurate simulation of organ-specific metastasis by circulating tumor cells.¹⁷⁰ Notably, the tendency of tumor cell stagnation and attachment to lung cells was found, primarily due to the elevated expression of CXCL12 chemokine within lung cells. Mao et al. engineered a bionic vascularized liver-bone MOOC [Fig. 6(b)],¹⁶⁸ which can replicate the metastatic characteristics of pancreatic cancer cells.

The gut-brain axis, referring to the intricate communication and interaction between the gut and the brain, can influence neurocognitive functions and the potential development of neurological disorders. Trapecar et al. integrated MPS with the gut, liver, and brain in a MOOC for the investigation of neurodegenerative diseases [Fig. 6(c)].¹⁶⁹ The brain MPS was built using iPSCs derived from Parkinson's patients with A53T gene mutation, a well-known genetic culprit of Parkinson's disease. Intriguingly, the authors incorporated short-chain fatty acids (SCFA) from intestinal bacteria into the chip, effectively mimicking the role of gut microbes in the gut-liver-brain axis. Another study reported a brain-gut-axis-on-a-chip to reproduce the sequential transport of exosomes across the intestinal epithelium and blood–brain barrier.¹⁷¹ These studies demonstrated the potential of MOOCs in the simulation of complex physiological interactions, providing valuable insights into disease mechanisms and drug effects.

VI. THE APPLICATION OF ORGANOID-ON-A-CHIP

A. Pharmacodynamic–pharmacokinetic modeling

Pharmacodynamic–pharmacokinetic (PD–PK) modeling is commonly used for evaluating the efficacy and safety of drugs in the human body. PD focuses on the relationship between drug concentrations and their effects on the target site. PD also analyzes how drugs interact with receptors, enzymes, and other biomolecules. On the other hand, PK primarily involves drug delivery, absorption, and metabolism. PK focuses on the process of drug delivery in vivo and the influence of parameters such as dosage, route of administration, and patient characteristics on drug concentrations.

Due to the drug action in multiple organs, PD–PK modeling typically requires MOOC toward two or more of the following organs: intestine (absorption), liver (metabolism), kidney (excretion), and a target organ (toxicity). Oleaga et al. developed a cardiac–hepatic MOOC to investigate the metabolic behavior of cyclophosphamide and terfenadine and impact of their metabolites on the cardiac and hepatic systems.¹⁷² The presence of hepatocytes amplified the effect of cyclophosphamide on cardiac function, while terfenadine significantly affected heart function in the absence of hepatocytes. In addition, PK parameters can be predicted using MOOC. Herland et al. proposed fluidically coupled vascularized organ chips integrating the gut, kidney, and liver to simulate drug absorption, metabolism, and excretion [Fig. 7(a)].¹⁷³ The authors predicted the PK parameters for nicotine and cisplatin, followed by comparing them to clinical data. An accurate and quantitative translation of in vitro results to in vivo PK parameters can be achieved, becoming crucial in the design of effective drug administration regimens for clinical trials. Tsamandouras et al. presented a gut-liver chip to examine the crosstalk between the gut and liver regarding PK-related processes [Fig. 7(b)].¹⁷⁴ The interaction between the gut and liver led to an augmented intrinsic metabolic clearance of the liver MPS.

FIG. 7.

Representative applications of organoid-on-a-chips. (a) Snapshots and schematics of the gut, liver, and kidney organoid-on-a-chips (scale bar = 5 mm). These chips consist of apical parenchymal and basal vascular compartments, which are separated by a porous matrix-coated membrane.¹⁷³ Reproduced with permission from Herland et al., Nat. Biomed. Eng. 4(4), 421–436 (2020). Copyright 2020 Clearance Center, Inc. (“CCC”). (b) (i) Exploded view of the microfluidic platform with gut and liver MPS and (ii) schematic of medium flow pattern.¹⁷⁴ Reproduced with permission from Tsamandouras et al., AAPS J. 19(5), 1499–1512 (2017). Copyright 2017 Creative Commons CC BY. (c) (i) Schematic of an organoids-on-a-chip for immunotherapies toward head and neck cancer patients and (ii) staining results of cancer and immune cells.¹⁷⁷ Reproduced with permission from Al-Samadi et al., Exp. Cell Res. 383(2), 111508 (2019). Copyright 2019 Clearance Center, Inc. (“CCC”).

B. Personalized medicine

Personalized medicine aims to tailor medical treatment to an individual patient. Significant inconsistency in disease manifestation and responses to treatment among patients always exists. The organoid-on-a-chip platform provides the possibility for generating organoids derived from patients' cells (biopsy samples or iPSCs obtained from normal tissues), followed by studying cellular response after treatments in a reproducible manner. In this way, organoid models can recapitulate critical characteristics of rare genetic diseases or cancer in vitro. To determine whether the genomic profile of the primary tumor could be preserved in organoids, Weeber et al. analyzed 1977 cancer-related genes in the organoids from 14 patients with metastatic colorectal cancer.¹⁷⁵ The findings revealed that 90% of the somatic mutations in the organoids were consistent with biopsies in the same patients. The results indicated that organoids effectively capture the genetic characteristics of the primary tumors, providing crucial evidence for their potential use in personalized medicine. Moreover, the patient-derived tumor organoid (PDTO) can be used to assess drug efficacy in glioblastoma.¹⁷⁶ The human tumor microenvironment can also be established using patient-derived tumor tissues, immune cells, and serum from individuals with head and neck squamous cell carcinoma [Fig. 7(c)].¹⁷⁷ In this study, immune cell migration toward cancer cells and cytotoxic activity were evaluated, providing clinicians with preliminary information regarding pharmaceutical effects for personalized treatments. Numerous tumor organoid biobanks have already been established,^178,179 some of which encompass samples from hundreds of patients.¹⁸⁰ In fact, there have been many studies utilizing drug screening results obtained from tumor organoids to aid in clinical diagnosis and treatment. For instance, Wang et al. conducted a blinded study and discovered that the PDTO model accurately predicted chemotherapy responses in patients with stage IV colorectal cancer,¹⁸¹ achieving an impressive accuracy rate of 79.69%. Jiang et al. developed a high-throughput tumor organoid generation platform based on droplet microfluidics technology and utilized it to mass-produce PDTOs. Subsequently, they conducted drug screening studies involving 29 frontline chemotherapy drugs and two targeted drugs. The results indicated that among 21 patients with various cancers, the PDTO screening outcomes were consistent with clinical treatment results for 17 patients, achieving an accuracy rate of 81.0%.⁷⁶ In another study, PDTOs achieved impressive results with an 88% positive prediction accuracy and a 100% negative prediction accuracy for metastatic gastric cancer.¹⁸² These studies demonstrate the ability of organoids in predicting drug responses in future personalized medicine. It is worth noting that in these studies, the entire experimental cycle, from constructing PDTOs to obtaining drug screening results, took only a few weeks. In fact, the majority of research endeavors involving the generation of PDTOs from biopsy specimens have successfully compressed the timeline from patient sample extraction to drug screening results to just a few weeks.^180,183 However, when utilizing normal patient tissues to construct PDTOs, an additional two to three months are typically required to induce normal tissues into iPSCs before organoid construction can commence.^68,179

Despite the advancement of organoids-on-a-chip in personalized medicine, high-throughput models for clinical diseases of interest still require to be improved in terms of automation, functionality, and accessibility. First, to shorten the time for new drug development becomes indispensable in preclinical trials, necessitating automation platforms for rapid tests. For instance, 3D printing has been commonly adopted to automatically construct sophisticated ECM scaffolds or microdevice modules in a programmable manner for organ-on-a-chips.^184–186 Additionally, low production yield in the traditional lithography technique could be enhanced by rapid injection molding or hot embossing, eliminating the user dependency in tests of newly developed drugs toward various biological targets. Nevertheless, due to the complexity of the physiopathological environment, how to reconstruct organoids-on-a-chip with better functionality becomes a major issue. The key point is to replicate the features of interest instead of the full biological conditions to study those responses not been discovered yet. For example, an intestine-on-a- chip was recently developed using patient-derived intestinal epithelial cells and immune cells to investigate the intestinal dynamics under coronavirus infection and relevant therapeutics.¹⁸⁷ Immune responses were characterized by peristaltic activity in this functional chip, and virus insult could be prevented by a protease inhibitor drug Nafamostat. Blood–brain–barrier on a chip was also recapitulated using astrocytes, neurons, and endothelial cells, which are derived from stem cells in a human patient.¹⁸⁸ Although the presented chip structures are relatively simple, training more laboratory technicians becomes necessary for cell culture operations once these chips get scaling up. Therefore, the feasibility of fully applying organoids-on-a-chip in bedside disease diagnosis or treatment remains questionable. The standardization of organoids-on-a-chip might be a practical solution to reduce labor and time costs. The design and construction methods of models targeting identical disease or organ need to be further unified, facilitating manufacture efficiency and accessibility in the future. Lab technicians, in this way, can perform experiments according to more comprehensive operating standards. While organoid-on-a-chip technology is still in its primary stage in personalized medicine, in-depth research will further functionalize organ models, replacing the conventional “one-size-fits-all” approach to treatment.¹⁸⁹

VII. CONCLUSIONS

In the past decade, a variety of organoids-on-a-chips have been presented with their prominent advantages, including miniaturization, flexible assembly, and cost-effective production. Despite our best efforts, minor improvements toward organoid formation may still play a crucial role with respect to cell sources, ECM selection, and microfabrication techniques, as discussed in Secs. I–VI. As the combination of multi-organs and vessel networks is more convenient for studying drug metabolism, the future mainstream developing trend is likely to be vascularization, immunization, and systematization of organoid-on-a-chip. With these micro-scale multi-functional platforms, the results could be obtained from organoid-on-a-chip more reliably, followed by being integrated with data from animal models. Drug testing can be eventually conducted in a safer manner for patient-derived tissues in pharmaceutical research, such as PD–PK modeling and personalized medicine. Both opportunities and obstacles in this development pattern are driving researchers to brainstorm and introduce more innovative ideas for organoids building from a wide range of disciplines. Cross fertilization of microfabrication and tissue engineering is breeding easily accessible and integrated organoids-on-a-chips for the ultimate goal of benefiting mankind.

AUTHOR DECLARATIONS

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

Z. Li and Q. Li contributed equally to this work.

Zhangjie Li: Writing – original draft (lead); Writing – review & editing (equal). Qinyu Li: Writing – original draft (supporting); Writing – review & editing (equal). Chenyang Zhou: Writing – original draft (supporting). Kangyi Lu: Writing – original draft (supporting). Yijun Liu: Writing – original draft (supporting). Lian Xuan: Resources (supporting). Xiaolin Wang: Conceptualization (lead); Funding acquisition (lead); Supervision (lead); Writing – review & editing (equal).

DATA AVAILABILITY

Data sharing is not applicable to this article as no new data were created or analyzed in this study.