Advances in multicellular spheroids formation

Abstract

Three-dimensional multicellular spheroids (MCSs) have a complex architectural structure, dynamic cell–cell/cell–matrix interactions and bio-mimicking in vivo microenvironment. As a fundamental building block for tissue reconstruction, MCSs have emerged as a powerful tool to narrow down the gap between the in vitro and in vivo model. In this review paper, we discussed the structure and biology of MCSs and detailed fabricating methods. Among these methods, the approach in microfluidics with hydrogel support for MCS formation is promising because it allows essential cell–cell/cell–matrix interactions in a confined space.

1. Introduction

Traditionally, cell culture on the two-dimensional tissue plastics has been used for cell biology studies and disease models for drug screening with the assumption that monolayer cells reflect the physiology of real tissues. However, in such a two-dimensional platform, cells are typically exposed to a rigid solid surface on the basal side and to a liquid at the apical surface. Inhabiting such a two-dimensional rigid substrate requires cytoskeletal adjustments by the surviving cells, because they lack exposure to the extracellular matrix (ECM) that is unique to each cell type, which may produce counterfeit polarity and cause abnormal cell metabolism and protein expression [1]. Furthermore, two-dimensional systems cannot provide a complex and dynamic microenvironment for cells, and thus lead to spurious findings to some extent by forcing cells to adjust to an artificial and rigid surface. Studies have shown that two-dimensional cell culture cannot replicate real microenvironment and cell behaviours in vivo because of lack of cell–cell and cell–matrix interactions and loss of tissue-specific architecture, mechanical and chemical cues, which are essential for unique functions of real tissues in the human body. For instance, a decrease in β1-integrin hindered multicellular spheroid (MCS) formation of PC3 prostate adenocarcinoma cells, while the decrease was not found in the two-dimensional culture [2]. Therefore, a new platform for in vitro cell culture has been pursued and three-dimensional cell culture has been demonstrated to recreate the in vivo microenvironment and physiological relevance [3].

A number of three-dimensional culture models have been explored. Three-dimensional MCS is an excellent in vitro three-dimensional model, mimicking the in vivo processes, such as embryogenesis, morphogenesis and organogenesis [4]. MCSs are cell aggregates that have complex cell-to-cell adhesion and cell-to-matrix interaction, which result in gradient generation for nutrients, gases, growth factors and signal factors. This structure recapitulates the microenvironment of cells that has been observed in the real tissues.

MCSs were first fabricated by Moscona & Moscona through self-assembling of cell suspensions, and they found these tissue-like aggregates were able to restore characteristics in the original tissue [5]. Since then, methods have been developed to form MCSs. Furthermore, a range of cells have been explored to form MCSs, including cancer cells [6], mesenchymal stem cells and human pluripotent stem cells [7]. MCSs exhibit similar features in the in vivo physiological conditions [8], for example, cardiomyocyte spheroids beat with heart-like rhythms [9], hepatocyte spheroids perform liver-like functions [9], and human endothelial cells vascularize fibroblast microtissues [10]. MSCs are considered to be the building blocks to bio-fabricate three-dimensional functional living macrotissues and organ constructs via organ printing, which could be a paradigm shift for tissue engineering [11]. Therefore, MCSs are considered as a bridge to bring in vitro and in vivo together, and they are becoming an emerging biomedical tool in many areas. This paper delivers a comprehensive review on the methods of MCSs fabrication. Special foci are devoted to recent progress in MCSs formation in the last 5 years.

2. Formation mechanism and structure of multicellular spheroids

MCS formation may be due to cell adhesion and/or cell differentiation. It has been reported that MCS formation involves three critical steps [12]. (i) Dispersed cells initially are drawn closer to form loose aggregates due to their long-chain ECM fibres with multiple RGD motifs that can bind tightly to the integrin on the cell membrane surface. Direct cell–cell contact due to initial aggregation results in upregulated cadherin expression. (ii) Cadherin is accumulated at the membrane surface. (iii) Cells are compacted into solid aggregates to form MCSs due to the homophilic cadherin–cadherin binding [13] (figure 1). The ECM fibres and cadherin type and concentration may vary for different type of cells [14–16]. For example, E-cadherin is responsible for tight packing of MCF7, BT-474, T-47D and MDA-MB-361 cells [15]. It was reported that β-catenin sequestrated by E-cadherin can initiate transcriptional activation of proteins such as cyclin D1 and c-Myc, which control the G1 to S-phase transition in the cell cycle to help cell proliferation which also result in cell differentiation within MCSs [17]. Meanwhile N-cadherin mediates the spontaneous formation of spheroids in MDA-MB-435S [15]. MDA-MB-231 and SK-BR-3 cells form spheroid structure due to collagen I/integrin β interaction without cadherin involvement [18].

Figure 1.

MCS formation process. Cells establish loose bonds through integrin–ECM. A delay stage is followed for cadherin expression and accumulation on the extracellular surface. A compact MCS forms due to homophilic cadherin–cadherin interactions. (Online version in colour.)

Intercellular adhesion proteins such as connexin and pannexins that influence cell–cell contacts contribute to the formation of MCSs. As a class of gap junction proteins [19], pannexins were introduced into the C6 glioma cells in order to accelerate the MCS formation process and obtain a more mature F-actin cytoskeleton. The result showed pannexin 1 as the conduit for ATP release initiated a signalling that dramatically accelerated the assembly of large MCSs [20].

The cytoskeleton also plays an important role in MCS formation (figure 2, the actin filaments in the cytoskeleton of the spheroids [21]). The actin filaments undergo significant changes during MCS formation. The expanded microfilaments as stress fibres become localized along the cell periphery. The cytoskeleton as a force generation structure performs as a continuous pre-stressed lattice that keeps cellular structural stability. Morphogenetic phenomena promote the emergence of ordered structures, resulting in the MCSs formation [22].

Figure 2.

The confocal images of the BT-474 spheroid [21]: (a) nuclei, (b) focal contacts (antivinculin antibody), (c) F-actin and (d) merged. The scale bar is 100 µm. (Online version in colour.)

Tumour MCSs in vivo have two major components: one is proliferating neoplastic parenchymal cells, and the other one is supportive stroma, constituting half the mass of most malignant tumours and including fibroblasts, dendritic cells, lymphocytes, blood vessels, macrophages as well as other myeloid cells [23]. The parenchymal cells determine growth and differentiation of tumour spheroids, while stroma contributes to tumour progression. Normal stroma delays or prevents MCS formation but the abnormal one promotes tumourgenesis [24]. Hence, stroma plays a crucial role in the MCS morphology and the evolution of tumour spheroids. Additional stroma elements promote tumour spread and invasion [25] and stroma composition or structure may be altered when the function of vascular cells or fibroblasts changes, or the influx of inflammatory cells increases [26]. The interaction between endothelial and tumour cells in fabricated MCSs offers information on formation of blood capillaries in tumours.

In vitro MCSs have been fabricated with different structures. The most common one is a three-layered structure with cells in asynchronously proliferative, quiescent and necrotic states (figure 3a). The proliferative rim contains cells with intact nuclei and abundant microvilli and these cells are active in proliferation and metabolism. Cells with shrunk nuclei and sunken-in membranes are distributed in the quiescent stage and these cells have minimum metabolic activities but become active after exposure to nutrients. Cells in the necrotic core have disintegrated nuclei and membranes and they lose their activities because of starvation of nutrients and accumulation of toxic wastes. The layer thickness is dependent on the size of the spheroids. The diffusion limitation of most small molecules like oxygen in MCSs is around 150–200 µm, which is similar to that in the avascular tissue and tumour mass in vivo. Beyond the diffusive limit, metabolic wastes are accumulated in the inner layer of MCSs [30–32], leading to a necrotic core. Proton magnetic resonance with pH-sensitive indicators [32] as well as microelectrodes [31,33] have been used to study the transportation of nutrients, wastes and oxygen transportation and confirm the concentration gradients inside the MCS.

Figure 3.

Microenvironment and structure of an MCS. (a) MCS has a spherical geometry with different regions for proliferating, quiescent and dead cells. Different mass transport rates for nutrient, O₂, ATP, waste, CO₂ and lactase are presented in the multilayer structure. (b) The live/dead image indicates the morphology of an MCS, green represents live/proliferating cells (proliferating zone), while red represents dead cells (quiescent and necrotic zone) [27]. (c) The scanning electron image of a human epithelial carcinoma A431 MCS grown for 3 days [28]. The smooth profile shows the final stage of spheroids formation. Cadherin–cadherin interactions lead to the deposition of ECM. As a result, a compact, smooth profile of the spheroid can be observed without distinguishable cell morphology [29]. (Online version in colour.)

A vascular lumen-like hollow structure has also been observed for endothelial MCSs due to the apoptosis caused by central cell polarization [34]. A unique phenotype of MCSs formed when bovine aortic endothelial cells grew within a matrix that was full of basal membrane constituents. A polarized epithelium was established in the outermost layer with an apical surface pointing inside and a lumen was formed due to gradual loss of cells in the sphere interior. The spheroid structure obtained from human mammary epithelia cells [35,36] have been used for studying apoptosis and tissue morphogenesis. This model is also employed for investigating the oncogenic pathway to affect cellular cycles and biochemical pathways to transmit signals from cell–matrix contact in order to increase the cell survival rate.

Heterotypic cell-to-cell interactions play an important role in maintaining the hierarchy architecture and physiological functions of tissues and heterotypic MCSs are generated for mimicking the architecture and functions of real tissues. The heterotypic MCSs have been generated from co-culturing fibroblasts (parenchymal hepatocytes with human dermal fibroblasts, rat hepatocytes with NIH/3T3 fibroblasts) [37,38], pancreatic islet cells with hepatocytes [39] or bone marrow stoma cells with hepatocytes [40]. Different architectures for normal and malignant breast epithelial cells have been fabricated to simulate the tumour micro-environment [41]. Co-culture of tumour cells with endothelial cells allows building small capillaries within the MCS structure to avoid a necrotic or quiescent core in the centre. Transportation of nutrients and oxygen from the outer layer to the centre by the endothelial cells is realized, which mimics the transportation in the in vivo tumours. MCSs were co-cultured with macrophages or lymphokine-activated killer cells in order to measure the cytotoxic and cytostatic activity and examine the immigration of the immune cells [42]. It was found that lactic acid generated from MCSs disturbed the migration of monocytes into MCSs, working as a potent immune suppressor for monocytes/dendritic in the tumour milieu. Tumour cell invasion and metastasis have also been investigated through interactions between tumour cells and fibroblasts [43,44]. Furthermore, epithelia cells in a heterotypic spheroid have been used to investigate cell–cell interaction for different types of cells [45]. A spheroid formed by co-culturing mesenchymal stem cells and primary liver cells behaved like hepatocyte cells, and was explored for the heterotypic cell co-culture for cell proliferation and differentiation with the potential for regenerative medicine since the co-culture system was more efficient in inducing differentiation of mesenchymal stem cells into hepatocyte-like cells [46].

The microenvironment of MCSs is crucial for cell behaviours and functions. The MSCs model has been used to investigate the impact of the architectural, physical and physiological microenvironment on the tumorigenic phenotypes. Intracellular gene expression is influenced by altering the chemical composition of the extracellular microenvironment [24]. Abnormal cell behaviours such as tumour progression are caused by inappropriate cell–microenvironment interactions [47]. Martin et al. [48] reported that cells performed normal embryogenesis if they were located in the uterus, while they were differentiated into malignant tissue cells when they were co-cultured with ectopic embryonic cells. Oral squamous carcinoma cells had similar behaviours in the synthetic poly (lactide-co-glycolic acid) (PLGA) scaffold in vitro and in vivo compared to two-dimensional culture as evidenced by increase in secretion of vascular endothelial growth factor (VEGF), interleukin 8 (IL-8), as well as basic fibroblast growth factor (FGF). An overall increase in the VEGF and FGF expression was able to mimic that in the in vivo tumours. The results also highlight that some factors were found in the three-dimensional culture but not in the two-dimensional culture, such as IL-8. Through analysis of the cumulative percentage of these factors, IL-8, instead of VEGF, was proven to play the most important role in angiogenesis, which has changed our understanding of the role of IL-8 from the two-dimensional study [49].

3. Conventional multicellular spheroid formation from dispersed cells

In vitro production of MCSs has several challenging issues. Economic production of MCSs in a short time frame needs to be considered as the first criteria for choosing an approach for MCS formation. The ability to produce uniform-size MCSs is also crucial. Damage of MCSs and the impact on the physiology should also be addressed. Many approaches have been developed to fabricate MCSs, such as dispersed cells (referred to as ‘matrix-free’ herein), cells embedded in hydrogel matrix (referred to as ‘matrix-based’ herein) and microfluidic platforms as shown in figure 4. The matrix-free approach has been commonly used to produce MCSs, such as suspension culture in non-adhesive surfaces [18,50–52], hanging drops [53,54], spinner flasks or rotational bioreactors [55–57] and external-force-driven MCS aggregation [21,58–62]. These methods have been detailed and reviewed in previous papers [24,63,64]. A summary of the matrix-free approach can be seen in table 1. The approach of using dispersed cells in microfabricated chambers for forming MCSs is discussed in more detail in §5.

Figure 4.

Methods for MCS formation. (a) Forced floating, (b) hanging drop, (c) spinning flask, (d) rotating vessel, (e) electrical-force assisted, (f) magnetic-force assisted and (g) matrix-based.

Table 1.

Matrix-free approach for MSCs fabrication.

methods		advantages	disadvantages	cell type	reference
forced-floating on non-adhesive surfaces	poly(HEMA) coating	—inexpensive cost —simple procedure —easy accessibility —reproducibility —large-scale production	—variations in MCS sizeand shape —intensive labourrequirement —lack of cell–matrixinteractions	MCF7-ADR,T47D, MCF-7	[18]
	agarose coating			HeLa	[50,51]
	poly (N-p-vinlbenzyl-d-lactonamide) coating			3T3colne A31, Type II lung alveolar cells	[52]
hanging drop	multi-wells	—uniform-size MCSs —easy operation —high reproducibility —controlled environment	—intensive labourrequirement —small MCSs —incompetency for massproduction	HepG2, MCF-7	[53]
	super hydrophobic chip			L929, SaOs-2	[54]
bioreactor	spinner flask	—simple culture method —mass production —long-term culture —dynamic controlled environment —excellent mixing fornutrient and oxygendelivery, and wasteremoval	—variations in MCS sizeand shape —shear damage of MCSs —lack of cell–matrixinteractions	Chinese hamster V79	[55,56]
	rotational culture (NASA)			PC3, LNCaP, DU 145, BT20	[57]
force driven	magnetic force	—controllable MCS sizesand shapes —applicability to a varietyof cell types for formingMCSs —co-culture of two ormore different types ofcells for MCS formation —rapid formation	—loss of aggregateintegrity after removal ofexternal forces —damage of cellmorphologies due toextra forces —lack of cell–matrixinteractions	NIH/3T3, human umbilical vein endothelial cells, bone marrow-derived human mesenchymal stem cells	[58,59]
	electric force			Jurkat cells	[60,61]
	acoustic force			HepG2, BT474	[21,62]

Overall, conventional approaches are easy, quick and cheap to operate. However, some approaches, such as hang drops, are labour consuming. In addition, it is quite challenging to control the size of the spheroid and to avoid shear damage to MCSs. Lack of matrix support results in no cell–matrix interactions that is essential in MCSs cell biology and functions.

4. Matrix supported multicellular spheroid formation

Hydrogel networks as scaffolds provide bio-mimicking matricies to MCSs to induce cell–matrix interactions for fabrication of MCSs. The scaffold structure, morphology and its components can be adjusted for tuning the microenvironment for MCSs so that the cellular responses to the microenvironment can be monitored during MCS formation [65] (table 2).

Table 2.

Typical polymers used as matrix supports for MCSs fabrication.

polymer type		features for MSC formation	cell types
natural polymers	collagen	—protein-based hydrogels for better cellattachment —weak mechanical strength	hESCs, hepatocytes [66], luminal cells [67], MDA-MA-231, HCTT116 [68], HepG2 [60]
	HA-	—excellent viscoelasticity —promoting cell proliferation	LNCap [69], U373-MG, U87-MG [70], MDA-MB231, MDA-MB-468 [71]
	CS	—carbohydrate structure similar to the ECMcomponent GAGs for better cell attachment —insoluble in aqueous solution atphysiological conditions	MCF-7 [72], adipose-derived stem cells [73]
	Matrigel	—rich in growth factors, signal factors for cellproliferation —expensive and animal derived	PC-3M, PrCa, NCI-H600 [74], HepG2 [75], A235, B16F10 [76]
	agar	—damage on cells by forces from gelformation inside the agar —excellent mechanical properties	human foreskin dermal fibroblasts [77], cortical cell [78]
	alginate	—rapid formation of hydrogel scaffolds —strong chemicals required to break hydrogelnetworks to harvest MCSs	human fat-derived stromal vascular fraction cells [79], human colon cancer cells (HCT116) [80]
	fibrin	—promoting cell attachment —high cell viability	human mesenchymal stem cell [81], B16-F1 cell [82]
synthesized polymers	PEG	—good biocompatibility —irreversible gelation leading to challenges inreleasing MCSs from gel	Huh7.5 cells [12], submadibular glands [83]
	PLG/PLGA	—biodegradable and biocompatible	hepatocytes [84], MCF-7, U87 [49]
	NIPAM-based hydrogel	—thermally reversible to release MSCs duringharvest —low viscoelasticity —porous structure with great nutrient,oxygen and waste transport	HepG2 [85], HeLa [51], human pluripotent stem cells [7]
	poly(ɛ-carpolacton)	—high porosity —a large surface area-to-volume ratio forcellular attachment —high reproducibility	TC-71 Ewing's sarcoma cells [86]

4.1. Natural polymers

Natural scaffolds from hydrogels, such as collagen, Matrigel and chitosan (CS), have been used to culture MCSs. Other natural polymers, such as agar for ovarian cancer cellular MCSs [1], fibrin [82] and silk fibroin protein [87] for B16-F1 cell and breast cancer cell line MDA-MB-231 cells, respectively, have also been explored. Natural hydrogels are favourable for cell proliferation due to its enriched ECM proteins or mimicking ECM components. They are also highly biocompatible [88].

Collagen is a main structural protein in human tissues. Collagen is often used as an excellent matrix for cell attachment due to its integrin binding sites [89]. The three-dimensional porous structure of collagen hydrogels allows sufficient exchange of oxygen, wastes and nutrients inside the scaffold [90] for short-term or long-term MCS culture [91]. The biodegradability due to enzymatic reactions offers a simple method of harvesting MCSs from the scaffold. Human embryonic stem cells (hESCs) were introduced into the collagen scaffold, and after 5 days of culture hESCs were differentiated into hepatocytes. The hepatocytes formed MCSs which possessed all characteristics and functions of human hepatocytes in vivo [66]. Luminal cells were co-cultured with myoepithelial cells and fibroblasts in the collagen type I matrix, and after 7 days of culture heterotypic MCSs were formed inside the matrix [67]. Fang et al. [68] cross-linked collagen with Matrigel to create a hydrogel scaffold to culture human breast carcinoma cells (MDA-MA-231) and colorectal carcinoma cells (HCTT116) for forming MCSs. The new hydrogel had better biocompatibility and reproduced the in vivo solid tumour microenvironment. One unique method of using collagen microparticles with a size of 10–20 µm to fabricate functional spheroids was reported [92]. The collagen micro-particles were fabricated via droplets in a non-equilibrium state in a micro-channel or membrane emulsification. Agarose-coated microchambers were used to prepare heterogeneous composite spheroids composed of hepatocytes and collagen particles. This approach recreates a similar in vivo microenvironment due to the incorporating ECM component collagen in the MCSs.

Hyaluronic acid (HA)-based hydrogels are also applied in MCS culture. Natural HA is found in the ECM of malignant tumours, and it promotes cancer cell proliferation and spheroid formation [69] because the excellent viscoelasticity of the HA hydrogel is able to mimic the stiffness of the natural environment. Prostate cancer cells (LNCaP PCa) were seeded into the HA hydrogel system, and after 7 days of culture MCSs were harvested with a size of 100 µm, and showed a significant upregulation of VEGF₁₆₅ and IL-8 expression. LNCap had a lower proliferation rate in the HA scaffold than in two-dimensional culture, which was similar to the LNCaP growth in vivo. Human glioblastoma U373-MG and U87-MG cells were found to form heterotypic MCSs after culturing them inside the HA scaffold for 24 h [70]. MDA-MB-231 and MDA-MB-468 cells mixed with the HA hydrogel were injected into mice to mimic the in vivo environment for tumour spheroid formation, and HA was found to stimulate cell proliferation and reduce the variability in the tumour size. This hydrogel was also found to facilitate tumour–tissue integration. Necrosis inside the formed MCSs was reduced, and vascularization in the structure was improved [71].

CS is the second most abundant material that can be produced from crustacean, molluscs, squid and insects. The biodegradability by lysozyme on the β-1, 4 glyosidic linkages enables mature MCS to release from the CS-based scaffold [93]. A porous CS scaffold was attempted to culture MCF-7 cells for producing MCSs and the scaffold with a high degree of deacetylation was found to improve cancer cell attachment and proliferation [72]. However, very low solubility of CS in water due to its strong hydrogen bonding hampers its wide application in MCS formation. CS co-polymers have been synthesized to improve the solubility. CS-alginate was synthesized as a hydrogel scaffold for culturing prostate cancer cell MCSs [94]. This co-polymer was more biocompatible, biodegradable and less immunogenic [95]. In addition, its chemical structure is similar to glycosaminoglycans (GAGs), a vital part of MCS ECMs. Instead of a hydrogel scaffold, CS-alginate was also fabricated as a fibrous scaffold coated with collagen to culture MCF-7 cells. Its porous structure optimally replicated the in vivo environment [96]. Small aggregates were observed after 2 days of culture. The size of spheroids was up to more than 100 µm after 6 days. Cell growth within this scaffold presented a spatial growth pattern with an improved growth rate and drug resistance. Apart from co-polymerized with natural polymers, CS is also combined with synthesized polymers for MCS formation. For instance, a poly(l-glutamic acid) and CS scaffold was developed to culture MCSs from adipose-derived stem cells (ADSCs) for cartilage regeneration shown in figure 5. This scaffold was repulsive to protein adsorption and facilitated ADSC MCSs formation by minimizing cell–matrix attachment to promote cell–cell contacts. MCSs of 100 µm were observed after 2 days of culture [73]. ASCs spheroids grown in this scaffold improved chondrogenic differentiation and dramatically decreased the deposition of collagen type I.

Figure 5.

Adipose stem cells (ASCs) cell–matrix interactions in poly(l-glutamic acid)/CS scaffold [73]. (a) Cells were stained with green fluorescent dye. The aggregate formation at the first 48 h within the scaffold was observed. After 48 h, MCS size achieved was around 80–110 µm. (b) SEM images indicated the multicellular aggregation progression and morphology change. Images from 12 to 48 h showed three steps of MCSs formation process. At 12 h cell-to-cell contact initiated, followed by cadherin expression and accumulation, then ECM deposition occurred, and the compact spheroids structure was obtained without distinguished sing cell morphology. Scale bar is 100 µm for (a) and 50 µm for (b). (Online version in colour.)

Matrigel is derived from the basement membrane of Engelbreth-Holm-Swarm(EHS) mouse tumour cells. Some key components for cell growth and attachment such as collagen and laminin are found in Matrigel [97]. As one of the popular ECM protein-enriched hydrogels, Matrigel has been widely used for MCS formation from PC-3M, PrCa and NCI-H600 cells [74]. The comparison of HepG2 MCS culture in collagen, gelatin and Matrigel hydrogels revealed cells proliferated rapidly in Matrigel, and larger MCSs were harvested from Matrigel [75]. Moreover, a Matrigel with lung cancer cell mixture injected into an animal model resulted in rapid formation of tumour MCSs in animals compared with cell injections without Matrigel. The similar results were also obtained from cancer cells of A253 and B16F10 [76]. However, three-dimensional tumour models grown on the Matrigel exhibited less similarity to the in vivo tumours compared with three-dimensional poly (lactide-co-glycolide) (PLG) engineered tumours because of a decreased level of IL-8 expression in the Matrigel [49]. Therefore, it is important to choose appropriate three-dimensional physical supports for fabricating MCSs.

The natural polymers share the structural or composition similarity with normal tissue ECM, have rich components, meet the amplified extracellular signalling needs and support cell behaviours. The impediments of using natural polymers for MCSs are the accessibility of the materials for the tissues of interest, dreary methodology and undesirable remnant proteins and confounding signalling proceedings [98]. In the case of multiple-component scaffolds, additional composition may be required for optimal support of cell growth.

4.2. Synthesized polymers

Synthesized polymers have better structural complexity and design flexibility that are tuned to mimic the in vivo environment for MCS culture and formation. In contrast to natural polymers, synthesized polymers are able to be modified to have desired ECM characteristics for the cell type of interest to promote cell aggregation and maintain tissue functionality [99].

Poly(ethylene glycol) (PEG) is one of the most popular polymers for three-dimensional cell culture due to its nontoxicity and non-immunogenicity. PEG is cross-linked via various means such as photo-polymerization and emulsion polymerization. PEG was cross-linked with other polymers such as poly(ethylene oxide) (PEO) to enhance polymer network performance for MCS fabrication. Through adjusting the PEG amount in the co-polymer, PEG-based hydrogels with different mechanical properties were used to culture hepatocellular carcinoma cell line (Huh7.5) for formation of MCSs to explore the impact of microenvironmental stiffness on cell aggregations [12]. Larger spheroids formed within the hydrogel network with better compliance or lower stiffness. Furthermore, cell proliferation, albumin secretion and CY450 expression in spheroids were found to perform better within compliant hydrogels, which may be due to better diffusion of oxygen and nutrients within a more compliant matrix. PEG-based hydrogels are formed via two different polymerization approaches: chain addition methacrylate-based and step-growth thiolene polymerization, were used for submandibular glands (SMG) MCS growth and formation. Step-growth thiolene polymerization had better performance for cell viability, proliferation and cell aggregation due to the reduced membrane peroxidation and intracellular reactive oxygen species formation [83].

Because of its excellent biodegradability, PLG has been widely investigated for three-dimensional cell culture. The key component of this synthesized polymer is extracted from natural metabolites and it has great biocompatibility. PLG has been used in different forms such as foam, fibres and sponges [100]. A porous PLG microsphere scaffold was developed to culture hepatocyte spheroids [84]. This approach accelerated MCS formation and the porous scaffold structure maximized cell attachment and transportation of nutrients, oxygen and wastes. MCF-7 and U87 cell lines were cultured in the PLG scaffold to form MCSs, and the PLG scaffold exhibited good performance in recreating the microenvironment for tumour engineering [43]. The tumour model obtained from the PLG scaffold possesses similar microenvironmental characteristics to in vivo tumours. The expression of IL-8, an angiogenetic factor, was found to be upregulated in cells cultured in the PLG scaffold [99]. The angiogenic feature was mapped with that in the in vivo tumours, and cells in this model were less sensitive to chemotherapy, which demonstrated the yielded tumours had improved malignant potential. PLG was mixed with hydroxyapatite (HA) to replicate a bone-like environment to culture breast cancer spheroids [101]. This scaffold promoted breast cancer cell proliferation and aggregation because HA encouraged the neoplastic and metastatic growth of breast carcinoma cells and stimulated IL-8 secretions.

Poly(N-isopropylacrylamide) (PNIPAM) hydrogel is another popular polymer network due to its thermal-reversibility which allows harvesting of MCSs without toxic or potent chemicals. Through co-polymerization with other monomers, the resultant copolymers can accommodate different types of cells, promote cell proliferation and aggregation, and maintain tissue functionality. This PNIPAM network was used for culture of HepG2 MCSs and a microgel of around 300 nm was found to have better cell proliferation and MCS formation [102]. The PNIPAM polymer was modified by acrylic acid (AA) for culturing HepG2 MCSs. PNIPAM-AA exhibited less shrinkage for long-term culture and maximally maintained the scaffold structure, and HepG2 cells proliferated best in the hydrogel with 1% AA in the copolymer [85]. PNIPAM-AA microgels were further galactosylated to culture HepG2 MCSs. The galactose ligands helped HepG2 MCSs in performing liver-specific functions [6]. PEG was also introduced to PNIPAM hydrogel for better cell attachment. Human pluripotent stem cells (hPSCs) were cultured in the hydrogels to form spheroids. hPSC-derived spheroids showed high proliferation, maintained the pluripotency in suboptimal culture conditions and had a high survival rate [7]. pNIPAM-PEG hydrogel was also used to culture HepG2 spheroids; using this method, HepG2 cells were pre-aggregated before they were cultured in the hydrogel [103]. HepG2 cells were treated by NaIO4 to produce surface aldehyde functionalities, and they were aggregated in the presence of an inter-cell linker, acrylic acid-modified CS. This method is very rapid, taking 1 day to form compact spheroids.

Others polymers have also been reported for MCS formation. Poy(ɛ-carpolacton) (PCL) was used to culture TC-71 Ewing's sarcoma cells to form MCSs and the MCSs exhibited greater chemo-resistance and different gene expressions from two-dimensional culture [86]. MCSs from breast, prostate and Lewis lung cancer cells were successfully demonstrated in the poly(lactic-co-glycolic acid)(PLGA) scaffold [99].

Overall, synthesized polymers show great design flexibility through co-polymerizing with other functional monomers. The physical and chemical properties are tuned to mimic the microenvironment for MCS formation. In addition, different from natural polymers with variations from batch to batch, synthesized polymers have high reproducibility and improved handling characters. However, their toxicity and degradability are two major concerns in the application of MCS formation.

One of the challenging issues is to release MCSs from natural or synthesized polymers. Most natural polymers except alginate may be degraded by introduction of specific enzymes. However, it may be necessary to remove any enzyme residues for certain biomedical applications. The alginate matrix becomes weak in an acidic environment due to gradual loss of ironic bonding and eventually the matrix is degraded to release MCSs. Synthesized polymers normally request toxic chemicals to break chemical bonds within their polymer structures. Thermal-sensitive polymer matrices can easily release spheroids by manipulating the cell culture temperature, which minimizes potential damage to MCSs. As a result, thermally sensitive polymers may be considered one of best candidates as a matrix to recover spheroids.

5. Multicellular spheroids formation on the microfluidic platform

The microenvironment that comprises complex chemical and mechanical cues is one of the most important factors for forming MCSs. Specific physico-chemical properties, such as oxygen tension, temperature, pH, osmolality and local concentration of soluble factors, can significantly influence cell-to-cell and cell–matrix interactions [104]. Conventional approaches for forming MCSs are less flexible in tuning the microenvironment and performing temporal and spatial stimuli on cells. Microfluidics, as an emerging tool to control the flow within micrometre-scale channels, is capable of manipulating the parameters dynamically and spatially, thus creating unique environments to meet the requirements for MCS formation. In addition, microfluidics can reduce the shear stress in order to minimize cell damage, and micrometre-scale chambers inside the microfluidics can be manipulated to control the size of MCSs [105]. Furthermore, integration of analytical tools with microfluidics (figure 6), such as Raman, fluorescence spectrometry and UV-Vis spectrometry, enables a rapid, in situ and dynamical analysis during MCS formation. To screen optimized culture conditions for forming MCSs from rare cells or primary cells the microfluidics, consumption of chemical or biochemical reagents is reduced since microfluidics often requires a very small amount of liquid [108]. Furthermore, oxygen supply to MCSs can be significantly improved through microfluidic chips. In conventional approaches, oxygen depletion in the centre core of a large spheroid leads to hypoxic conditions that cause cell necrosis at the heart of spheroids. Therefore, the maximum size of spheroids consisting of viable cells was reported to be 100–150 µm because of lack of oxygen [11]. A gas-permeable spheroid culture chip was developed to continuously supply oxygen to spheroids during culture [106]. As a result, the maximum size of spheroids was able to reach 600 µm with enhanced cell proliferation, viability, metabolic capacity as well as albumin secretion rates.

Figure 6.

Microfluidic methods for MCS fabrication. (a) MCS formation in micro moulds. (b) PC-3 prostate cancer MCSs formation in microwells [107]. (c) MCS formation in droplets generated from microfluidic channels: (1) T-junction cell encapsulation, (2) flow-focusing cell encapsulation, (3) double emulsion cell encapsulation and (4) human mesenchymal stem cells MCSs formation inside droplets [133]. (Online version in colour.)

5.1. Micro-moulding

The micro-moulding method has been explored to generate MSCs by fabricating complex structures using the lithography technique [109,110]. It has been documented that this micro moulding method can significantly reduce consumption of reagents, create desired concentration gradients of growth/signal factors and nutrients, allow high-density culture at a high cell-to-fluid volume ratio and decrease shear stress under the laminar condition in micro-scale structures [111].

Du et al. [112] proposed a method to fabricate different micro-scale structures with PEG-based cell laden hydrogels to form MCSs from NIH 3T3 mouse fibroblasts. The hydrophobic effects drove the ‘lock and key’ assembly of microgels to form cross- or rod-shaped structures. Three-dimensional tissue constructs containing MCSs were generated from these structures. Micro-patterned methacrylated HA hydrogels were applied to fabricate similar MCSs from NIH 3T3 mouse fibroblasts [29]. A patterned polydimethylsiloxane (PDMS) stamp was used to hold the HA precursor solution mixed with cells and the photo-initiator (figure 6a). After exposure to UV light, the HA solution was solidified and the PDMS mould was removed. The size and shape of MCSs were manipulated by the geometry of the PDMS mould. In addition, HA was biodegradable by enzyme and MCSs were released from the block during the harvest process. The PDMS membrane templates to encapsulate cells for forming MCSs have been developed [113,114]. The collagen hydrogel mixed with NIH 3T3 cells and the collagen/Matrigel mixture containing HepG2 were loaded into a mould from the PDMS membrane. After gelation, gel blocks were transferred into the culture medium for further forming MCSs. The MCS size was controlled by the size of the gel blocks.

A ‘bottom-up’ fabrication approach was developed for macroscale three-dimensional structures [115]. Collagen beads containing HepG2 cells were generated from an axisymmetric flow focusing device in a PDMS mould. The cell-laden beads were stacked to form a complex millimetre-thick tissue. After a couple of hours, the beads made contact with each other and were compacted into a designed shape. This approach can re-create the in vivo tissue environment to mimic the in vivo cell–matrix interactions.

Hardelauf et al. [116] designed an array of PDMS microwells to produce uniform tumour spheroids. Human colon carcinoma cells (HT29), BT474 breast carcinoma cells and NCI-H1792 lung carcinoma cells were pumped into the PDMS microwells and MCSs were observed after 3–4 days. The microwells were further modified to be concave in the bottom to culture hepatoshpere and hepG2 spheroids for drug screening [117]. Gong et al. [118] used the same concave microwell approach but replaced PDMS with agarose microplates to culture MCF-7 spheroids to test cancer drugs. Instead of a concave well at the bottom, a side-chamber design was used to culture MCSs [107]. In this device, two-layer PDMS microchannels were separated by a semi-permeable polycarbonate membrane. Twenty-eight dead-end side-chambers in the upper channel were designed to capture and stabilize PC-3 prostate cancer cells, meanwhile the lower channel allowed continuous medium flow for nutrient and waste diffusion. This design kept the spheroids stationary during media exchange and MCSs were monitored in situ during long-term cultures. Beside concave microwells, other similar designs for MCS culture have been proposed. Jin et al. [119] proposed a design to remove cell trapping barriers to facilitate MCS harvest. Cell suspension was supplied via the cell inlet port and distributed into four culture wells. Pressure from the membrane pressure port deformed the membrane at the bottom of the culture wells to form a horseshoe shape for trapping cells inside the wells. After spheroid formation, the membrane was deflated by reducing the pressure. Cell trapping barriers were removed and spheroids were harvested at the spheroid collection port. Kim et al. [120] developed a three-dimensional tumour spheroid chip with balanced droplet dispensing. The hydraulic-head difference between the nutrient stream inlet and the waste outlet triggered the removal of waste droplets at the outlet. A fresh medium droplet was supplied by the dispensing layer due to the decreased pressure caused by volume expansion. This design allowed automatic supply of fresh medium and removal of waste droplets. Lee et al. [121] proposed an approach for in situ MCSs formation and encapsulation. They used a PDMS mould to form uniform-sized HepG2 MCSs within the alginate hydrogel scaffold in the concave wells, and the nano-porous membrane to control the diffusion of cross-linker calcium ions for alginate gelation.

Instead of using a PDMS module or template, Zhao et al. [122] managed to employ patterned non-adhesive poly(2-hydroxyethyl methacrylate) hydrogel films in the cell culture plates to fabricate uniform-sized MCSs. The swelling-induced wrinkling of the film created a pattern composed of uniform concaves and cells were accumulated in the centre of the concaves to self-assemble into MCSs. Different from other microfabrication approaches, the hydrogel film is simple to be fabricated, and spheroids are much easier to be produced in the film.

A portable bioreactor was developed to maintain a sterile microenvironment and sustain cell growth, maturation and organ formation [111]. MCF-AT1 cells were encapsulated into the Matrigel that is placed at the bottom part of the PDMS bioreactor. Nutrients were perfused through the top section of the bioreactor. MCF-AT1 MCSs were harvested after couple of days. This bioreactor maintained cell viability for long-term culture, and also allowed visualization of the tumour spheroid formation progress through the transparent PDMS.

5.2. Micro-droplets-based multicellular spheroid formation

Uniform-size droplets generated from microfluidic devices provide a confined environment for cells and hydrogels inside the droplets offer a three-dimensional physical support for formation of functional MCSs. There are two configurations of microfluidic devices: flow focusing and T-junction are often employed to generate uniform-size droplets. In the flow-focusing device shown in figure 5, the aqueous phase with cells through the middle inlet as a dispersed fluid, and the immiscible organic solvent flows through both side inlets as a continuous fluid. The shear force from the continuous fluid squeezes the dispersed fluid to generate spherical individual droplets. On the other hand, in the T-junction device, the dispersed fluid penetrates the continuous fluid to form droplets and the newly formed droplets are swept by the continuous fluid. The hydrogels inside the droplets are often gelled due to external stimuli, such as temperature, light or ions. Thermal responsive hydrogels, such as agarose, NIPAM-based hydrogels or gelatin, solidify due to a temperature shift of the hydrogel system. Photoresistive hydrogels, such as PEG, form the physical gel by cross-linking triggered by exposure to UV light. Alginate is solidified by the introduction of divalent ion into the system (e.g. Ca²⁺). Normally, UV exposure or dramatic temperature change has a negative impact on cell viability as well as its physiology. Hence, the ion-based cross-linking gelation method is considered a safer alternative for cell encapsulation.

This droplet-based method is rapid and high throughput. Velasco et al. [123] demonstrated that cell-loaded hydrogel droplets significantly reduced the intensive labour needed to fabricate MCSs. In the porous micro droplets, oxygen and nutrients were diffused in and metabolic waste diffused out to maximize the cell viability [124]. In addition, when the formed MCSs encapsulated inside the hydrogel droplets are injected into the patient's body, they are protected from host immune responses, which minimizes administration of immunosuppressive drugs and increases the successful transplantation rate [125]. This approach also has a high degree of control over the morphological and physical properties [109].

Sakai et al. [126] used flow focusing microdevices to generate droplets for encapsulating HepG2 cells (figure 5c). Gelatin hydrogel with cells flew through from the top inlet, and paraffin oil flew through the left inlet. The hydrogel-containing droplets were heated to the melt point and the gelation was realized by cooling down to the gelation point when collecting the droplets. The size of the droplets from 300 to 100 µm was controlled by varying the paraffin flow rate. The HepG2 cells were cultured inside the droplets and they were aggregated to form MCSs in 24 h. This approach generated the desired size of MCSs since the size of droplet was controllable. Yoon et al. [127] proposed a similar approach, but they used alginate instead of gelatin to minimize the damage in the gelation process on the cell viability. Magnetic iron oxide nanoparticles were uptaken with the spheroids to achieve easy collection and separation of the spheroids during the harvest process. Instead of using alginate or gelatin, a mixture of alginate and Matrigel was introduced for MCS culture by Wang et al. [105]. This mixture gels performed better in the formation of HeLa spheroids compared with pure alginate beads. Tsuda et al. [128] used self-assembling peptides (SAP, Purmatrix RADA 16) to encapsulate bovine carotid artery endothelial cells [129,130]. This synthetic peptide was functionalized with other monomers to enhance cell attachment, increase cell proliferation and promote cell differentiation. The peptide hydrogels inside the droplets were solidified by exposure to the cross-linking agent in the continuous phase. After culturing the cells inside the droplets for 3 days, endothelial cells inside droplets migrated to form loose aggregates due to initial cell-to-cell contacts, spheroids were formed in long-term culture. The PuraMatrix hydrogels was solidified in the presence of certain ion concentrations. This hydrogel was attempted to encapsulate HepG2 cells in a double T-junction and the gelation was achieved by cross-linking with the ions in the cell medium [131]. Thermo-responsive microgel poly (N-Isopropylacrylamide-co-acrylic acid) (P(NIAM-AA)) droplets were also used to encapsulate and culture HeLa cell spheroids [132]. The porous structure of the hydrogel allowed sufficient mass transportation including nutrients, oxygen and bio-wastes, which was beneficial for long-term culture. In addition, the thermal reversibility of this hydrogel facilitated simple recovery of MCSs easily after long-term culture without introduction of toxic chemicals.

Chan et al. [133] developed a double emulsion system to fabricate MCSs from bone marrow-derived human mesenchymal stem cells (hMSCs). The double emulsion was performed in two flow-focusing devices. The water/oil emulsion was generated in the first device, which had an inner core containing a mixture of hMSCs and the culture medium and an out layer of oil. The water/oil emulsion went through the second device to generate the water/oil/water emulsion. The middle oil layer served as a barrier to prevent the inner core from mixing with the culture medium at the outer layer. The selective permeability of the oil layer allowed nutrient diffusion from the outer aqueous phase into the inner core and waste removal in the opposite direction.

6. Discussion and the future

With the advances in MCSs culture, the in vitro model is able to mimic the in vivo characteristics in many aspects. The MCS complex structure allows a better understanding of cell–cell and cell–matrix interactions. In addition, recapitulation of the in vivo microenvironment by MCSs enables fundamental research on cancer biology and tissue development and provides an opportunity for culturing functional tissues in vitro. However, there are a few challenges for fabricating MCSs. Firstly, a high initial cell density is required to form spheroids and cells in the MCSs structure are not required for further expansion while in vivo tumour spheroids form from a relatively low cell density and the size of the tumour spheroids increases due to cell expansion inside the structure. Secondly, the reproducibility and quality assurance of conventional means are low. Thirdly, the microenvironment and marcoenvironment for MCS formation from current methods are still different from in vivo, which results in different cellular behaviours. For instance, cell migration between MCSs is not realized from current methods. There is a gap in producing functional tissues through the in vitro culture due to lack of heterogeneous cell–cell interactions, cell–ECM organization and cellular signalling pathways within the tissue. Hence, the correlation between real tissues and MCSs should be thoroughly documented through gene expression profiling analysis. Moreover, cell behaviours may alter when they are exposed to different microenvironments/marcoenvironments and cells are able to adapt themselves for different types of MCSs in a real tissue, which are still missing in the in vitro MCS model. Furthermore, the heterogeneous metabolism and gene expression in MCSs can be influenced by the MCS size. Thus, a uniform-size of MCSs should be pursued.

The methods for fabrication of MCSs in vitro have been significantly developed in the past decade. The integration of microfluidics with scaffolds can be considered a promising new approach as it can not only produce controlled uniform-size MCSs, but also restore the complex cell–matrix/cell–cell interactions which are vital for MCSs morphology and functionality. The physical matrix support provided by the scaffold can facilitate MCSs in developing their own extracellular matrices that are crucial for cell functions. The chambers or droplets from the microfluidics allow fabrication of MCSs in a confined space to develop controllable uniform-sized MCSs. This approach is also able to tune the microenvironment for MCS formation and growth to mimic the in vivo conditions. In addition, the potential of commercialization makes the approach even more attractive. More elegant designs for microfluidics and scaffolds for MCSs formation are needed to create physiological relevant microenvironments and marcoenvironments to address the above challenges for MCSs formation.

Authors' contributions

X.C. organized, prepared and wrote this paper; Y.H. helped to organize figures and tables; H.Z. revised the manuscript.

Competing interests

We declare we have no competing interests.

Funding

H.Z. would like to acknowledge the financial support from ARC Discovery Project (DP160104632) and the Medical Advancement Without Animal (MAWA) Trust.