Design of Spherically Structured 3D In vitro Tumor Models -Advances and Prospects

Three-dimensional multicellular tumor models are receiving an ever-growing focus as preclinical drug-screening platforms due to their potential to recapitulate major physiological features of human tumors in vitro. In line with this momentum, the technologies for assembly of 3D microtumors are rapidly evolving towards a comprehensive inclusion of tumor microenvironment elements. Customized spherically structured platforms, including microparticles and microcapsules, provide a robust and scalable technology to imprint unique biomolecular tumor microenvironment hallmarks into 3D in vitro models. Herein, a comprehensive overview of novel advances on the integration of tumor-ECM components and biomechanical cues into 3D in vitro models assembled in spherical shaped platforms is provided. Future improvements regarding spatiotemporal/mechanical adaptability, and degradability, during microtumors in vitro 3D culture are also critically discussed considering the realistic potential of these platforms to mimic the dynamic tumor microenvironment. From a global perspective, the production of 3D multicellular spheroids with tumor ECM components included in spherical models will unlock their potential to be used in high-throughput screening of therapeutic compounds. It is envisioned, in a near future, that a combination of spherically structured 3D microtumor models with other advanced microfluidic technologies will properly recapitulate the flow dynamics of human tumors in vitro.


STATEMENT OF SIGNIFICANCE
The ability to correctly mimic the complexity of the tumor microenvironment in vitro is a key aspect for the development of evermore realistic in vitro models for drug-screening and fundamental cancer biology studies. In this regard, conventional spheroid-based 3D tumor models, combined with spherically structured biomaterials, opens the opportunity to precisely recapitulate complex cell-extracellular matrix interactions and tumor compartmentalization. This review provides an in-depth focus on current developments regarding spherically structured scaffolds engineered into in vitro 3D tumor models, and discusses future advances toward all-encompassing platforms that may provide an improved in vitro/in vivo correlation in a foreseeable future.


2D flat cell cultures of cancer cells remain the most commonly used in vitro
platforms for screening anti-cancer therapeutics, despite their recognized inability for mimicking three-dimensional (3D) cellular organization and tumor proliferation kinetics [1][2][3]. Moreover, these in vitro models lack the ability to correctly mimic tumor stromal heterogeneity and tumor-ECM components. Adding to these limitations, nutrients, oxygen, and pH gradients are not recapitulated, resulting in a recognized inability to realistically mimic in vivo tumors [3]. Overcoming the shortcomings of conventional 2D cultures through the engineering of more robust in vitro models capable of simulating in vivo solid tumors, could improve the efficacy of anti-cancer drug discovery and biological performance screening [1]. The development of such models could contribute to reduce the number of false-positive results obtained during preclinical validation of novel compounds and improve the in vitro/in vivo correlation.
In this context, three-dimensional (3D) in vitro models have been gaining increasing momentum in the field of drug-screening and cancer research, due to their improved ability for recapitulating the complexity of the tumor microenvironment (TME) [4,5]. 3D culture models are capable of recapitulating tumors cellular heterogeneity, cellcell interactions, and spatial architecture. In fact, the reproduction of such characteristics in 3D promotes the establishment of nutrient, oxygen, and signaling factor gradients, as well as the establishment of unique gene expression patterns similar to those observed in in vivo solid tumors [6].
From the currently available in vitro tumor models, 3D multicellular tumor spheroid models (3D-MCTS) remain one of the most commonly explored [7,8]. Their relative ease of assembly, reproducibility, and the ability to capture cellular heterogeneity (e.g., cocultures of cancer-stromal cells), renders them suitable tumor surrogates for preclinical validation of novel therapeutic compounds [9]. Up to date, various 3D-MCTS in vitro models have been used to modulate the cellular components present in the TME of different tumors including those of breast [10], colon [11], pancreas [12] and lung [13].
However, most of these models still lack a complete representation of tumor-specific and disease stage-specific ECM. This is a critical component which is recognized to extensively influence cancer evolution through key biochemical and biomechanical cues [14,15].
To overcome these limitations, various studies have attempted to include ECM mimetics in the form of spherically structured scaffolds, namely microparticles or microcapsules. These technologies have been extensively used in the field of tissue engineering and stem cell research [16][17][18], and offer further opportunities to mimic the complexity of the TME in vitro. In fact, this approach opens the possibility to study biochemical and tumor-ECM dependent mechanical cues through the inclusion of modular matrix-mimetic scaffolds [19]. The inclusion of tumor-ECM components in a 3D spherical geometry allows researchers to control various key parameters, such as: (i) pH, oxygen, and nutrient perfusion gradients, (ii) cell-cell interactions, (iii) morphology, and (iv) tumor models overall size. Importantly, previous studies have associated variations in 3D tumor microtissues size and morphology with variability in phenotype, gene expression profile, as well as with the degree of response to anti-cancer or anti-stromal therapeutics [20,21].
The integration of spherically structured scaffolds morphology into tumor modeling is herein demonstrated to not only allow an increased control over produced spheroids biophysical properties, but also to increase models' reproducibility in terms of shape and size, both of which are major aspects that must be considered in drug screening assays.
Furthermore, implementation of spherical designs facilitates incorporation of tumor-ECM components in well-established methodologies for spheroids analysis [22]. This unlocks the opportunity to model cell-ECM interactions and to evaluate the influence of ECM components inclusion in the response to anti-cancer therapeutics in a high-throughput compatible mode.
On this focus, this review showcases and critically discusses the recent advances in the field of complex 3D in vitro tumor models' assembly via spherical scaffolds. We begin by summarizing current scaffold-free and scaffold-based 3D microtumor production technologies and present up-to-date examples on the use of microparticles, microspheres, and microcapsules to assembly advanced 3D-MCTS. A critical perspective regarding future developments on new models that fully recapitulate in vitro the cellular and acellular components of the TME is also provided.

In vitro 3D Models Production
Ideally, 3D tumor models should be able to recapitulate the cell-cell and cell-tumor ECM crosstalk established during tumor progression. This communication is well recognized to contribute to the establishment of either pro-tumoral or anti-tumoral microenvironments [23,24], depending on the type of cells and ECM properties. Such dichotomy is well portrayed by the communications that can be established between immune cells and mesenchymal stem cells (MSCs) present in the TME. In fact, MSCsimmune cells communications can lead both to immunosuppression and promotion of angiogenesis, or to the increased recognition of cancer cells by infiltrating immune cells.
Such crosstalk is precisely mediated by the release of anti-and pro-inflammatory cytokines and growth factors, which can exert different effects depending on their amount and combinations [25,26]. Therefore, to recapitulate human tumors, 3D models must mimic this dichotomy, as well as ECM mechanical properties, cells spatial arrangement, and spatiotemporal biochemical composition of the TME. In vitro 3D microtumors should also be produced to dynamically recapitulate specific disease stages (e.g., from proliferation to metastasis) under highly reproducible conditions, so as to assure a direct correlation between in vitro and in vivo data [27].
3D multicellular tumor models production methodologies can be divided into three major categories: (i) scaffold-based models which take advantage of diverse natural, or synthetic materials that aim to mimic in vivo tumor-ECM [28][29][30]; (ii) scaffold-free models, which take advantage of cells suspension or hanging-drop techniques for assembly of 3D-MCTS [31][32][33][34]; and (iii) combinatorial hybrid technologies such as microencapsulation or microparticle-based approaches, which seek to combine the cell-cell aggregation obtained in scaffold-free based methods with the ECM representing capacities of 3D scaffold-based platforms. The following sections will discuss the differences and common advantages/disadvantages of currently employed technologies for the establishment of 3D-MCTS. In section 3, a focus is provided on spherically structured assembly of 3D in vitro models as these technologies have potential to emulate the different components of in vivo tumors TME.

Scaffold-free 3D Models Production
Scaffold-free methods are based on the implementation of cell cultures under nonadherent conditions (Figure 1). The main aim of this strategy is to promote the production of 3D spherical (spheroids) or more loosely aggregated microtissues (cellular aggregates) [35][36][37]. Despite scaffold-free models nomenclature remains non-regulated to date, in general terms it can be classified into three distinct categories: (i) multicellular tumor spheroid models (3D-MCTS), initially implemented in the 1970s by Sutherland and co-workers [38], (ii) organotypic multicellular spheroids (e.g., fragments of tumor tissue cultured in non-adherent conditions), and (iii) tissue-derived tumor spheres obtained by partial disruption of tumor tissues through mechanical and/or enzymatical dissociation [39]. These microtissues are easy to assemble, to culture in vitro, and have been extensively used in the field of 3D in vitro disease modeling for drug screening and more fundamental biology studies.
Static-based, scaffold-free methods such as forced-floating or hanging-drop techniques allow the assembly of highly reproducible 3D-MCTS in terms of size and morphology [40,41]. These techniques employ super-hydrophobic surfaces, or for example, poly(2-hydroxyethyl methacrylate) [42], or agarose coated multiwell plates [31,43,44] (Figure 1), to prevent adhesion to culture plate surfaces and promote instead cell-cell interactions that lead to intercellular adhesion and aggregation [32]. These 3D-MCTS can be formed either by monotypic or heterotypic co-cultures (e.g., comprising cancer and stromal cells) [45,46]. Over time, the newly formed 3D-MCTS start to secrete their own ECM, which increases their density and diminishes their size, further approaching these models to compact solid tumors similar to those obtained in vivo [47].
Dynamic, stirring-based technologies can be grouped into two classes: (i) stirring tank bioreactors in which the culture media is internally impelled (e.g., by spinning blades), and (ii) rotational (microgravity) bioreactors [48] (Figure 1). Such methods take advantage of mechanical forces that maintain cells in continuous suspension during culture [35]. useful to obtain 3D microtissue with highly reproducible sizes but are unable to mimic the mechanical forces that cells/cellular 3D aggregates experience when cultured under dynamic flow conditions. Image layout adapted from Picollet-D'hahan and co-workers [49], with permission from Elsevier.
By comparison to stirred tank bioreactors, rotational (microgravity) bioreactors have the advantage that the internal flow is generated by rotation of the container and not through blade mechanisms, thus imparting a lower shear-stress to the cultivated cells [50].
In comparison to static methods, stirring-based methodologies are able to effectively produce high amounts of 3D-MCTS, having the advantage of allowing easy culture medium exchange and modifications to cell culture conditions in situ [51].
However, regarding 3D aggregates reproducibility, stirring-based techniques lack a suitable level of control over 3D spheroids morphology and size, an important parameter that is generally present in static-based methodologies [52]. In fact, 3D spheroids and cellular aggregates obtained in bioreactors frequently exhibit variable shapes and density, leading to inconsistent responses to anti-cancer therapeutics [27,41]. Since cellular concentration is dictated for the entire batch, different internal dynamic flows can also result in aggregates with different sizes [53].
The combination of stirring and forced adhesion methods could overcome such limitations. For example, one could combine hanging-drop technique to obtain highly uniform 3D spheroids, and then translocate them into bioreactors or microfluidic platforms for evaluating fluid dynamics influence [54]. Overall, the main advantages of scaffold-free combination with dynamic culture methods are: (i) the ease of manufacturing multiple and reproducible 3D microtissues per batch; (ii) the ability to maintain prolonged culture times; (iii) the ability to modify the culture media and growth conditions (e.g., media perfusion flow rate, nutrient starvation); and (iv) the ability examine 3D models evolution in time either by single direct analysis in situ [55], or by placing them in hanging-drop superhydrophobic, on-chip arrays [33,56]. In a study performed by Oliveira and co-workers [57], such super-hydrophobic surfaces were employed to produce osteosarcoma spheroid arrays for anti-cancer drug-screening. Superhydrophobic surfaces patterned with wettable spots were successfully applied for osteosarcoma cells (Saos-2) 3D spheroids formation, and allowed the establishment of dynamic platforms capable of simulating chemotherapeutics clearance through means of dynamic media renewal [57]. In addition, these platforms were also recently explored by Oliveira and co-workers [58], for stem cellbased tissue engineering. In this study, human adipose-derived stem cells (hADSCs) 3D spheroids were established under either direct or indirect co-culture with 2D cell layers.
The cells were cultured in the flat microarray surface where a droplet of cells that formed the 3D spheroids was also formed. Co-culture of 3D hADSCs with 2D layers of HUVECs or Saos-2, resulted respectively in a significant decrease and slight increase in alkaline phosphatase (ALP) expression, an enzyme associated with hADSCs osteogenic differentiation, Therefore, such high-content imaging compatible, spheroid forming arrays, could be used in the future to study diverse stromal cells and 3D tumor spheroids direct or indirect interactions [58].

Scaffold-Based 3D Models Production
Tumor evolution in vivo is intimately correlated with the interactions between cells of the TME and their supporting ECM which provides both structural and signaling functions [59]. In a general perspective, the ECM is comprised mainly by fibronectin, collagen (types I-V), elastin, entactin, fibrillin, fibulin, vitronectin, laminin, as well as other glycoproteins and proteoglycans such as hyaluronic acid [60]. The exact composition of the ECM surrounding the primary tumor site can vary according to the type of tissue.
Hence, the design of 3D in vitro models must take into account tissue, and patient, ECM specificity, since this variability can lead to different response rates to candidate anticancer therapeutics [27]. Furthermore, during tumorigenesis, alterations in matrix composition and structure occur over time due to cancer and stromal cell-mediated ECM deposition or degradation (mainly through matrix metalloproteinases (MMP) enzymatic digestion (e.g., MMP-9, MMP-2). This matrix rearrangement often involves collagen deposition and matrix stiffening, a phenomenon that is associated to an increased metastatic potential [61]. Moreover, various reports indicate that increased hyaluronan deposition and degradation occurs in several cancers. Such increases angiogenesis, metastasis and possibly drug-resistance [62][63][64].
It is this dynamic nature of the interactions established between all the cells in the tumor microenvironment and the surrounding ECM that scaffold-based tumor models aim to fully recapitulate. Scaffold-based 3D in vitro tumor models take advantage of natural, synthetic or hybrid biomaterials to culture cells in a TME-like milieu [19,35]. Each of these classes presents its own advantages and disadvantages in mimicking in vivo tumors on an in vitro setting (Table 2). While naturally derived ECM mimetics have relatively low batch-to-batch reproducibility, scaffolds formulated with synthetic materials although more reproducible may require functionalization with bioactive molecules or inclusion of naturally-derived bioactive cues (e.g., peptides, adhesion proteins, etc), thus originating hybrid scaffolds. Despite the increased chemical modifications required in synthetic or hybrid scaffolds, the introduced biofunctionalizations can provide the opportunity to manipulate ECM mesh alignment, elasticity and swelling behavior, either by using static, or reversible precision chemistry reactions [68]. Hence, key events such as matrix stiffening can be simulated and precisely controlled [69,70].  [87,88] Poly-εcaprolactone (PCL) [89,90] Poly (lactic-coglycolic acid) PLGA [87,88] acidic by-products

Hybrid
Alginate-RGD [91,92] ▪ Combine the ease of chemical modification and the presence of ECM-like domains ▪ High-costs ▪ Representation of few ECM components PEG-RGD [93,94] PEG-fibrinogen [95,96] These ECM mimetic scaffolds can be manufactured into diverse structures such as: (i) fibrillar porous meshes, (ii) porous and non-porous microstructures (including microparticles or microcapsules), and (iii) micro-patterned surfaces, via 3D bioprinting technologies [28,35,[97][98][99][100][101]. The production methodologies of scaffold-based in vitro models have been extensively reviewed elsewhere [19,[102][103][104]. The different materials that are used for the manufacture of these scaffolds are selected by their specific characteristics, such as the rate of biodegradation, biocompatibility, elasticity, ease of manipulation and similarity to tumor-specific ECM. In this context, the following sections will review the most commonly used materials for these models, starting with naturalderived scaffolds and moving to synthetic and innovative combinatorial hybrid approaches used for correct ECM recapitulation.

Natural materials-based Scaffolds
From natural material-based scaffolds, the most commonly used hydrogel type for in vitro production of 3D-MCTS [28] is Matrigel TM , a hydrogel matrix comprised by basement-membrane extracts obtained from Engelbreth-Holm-Swarm mouse tumors [73].
The assembly of this hydrogel often requires laborious preparation, involving cooling of all the materials and thawing in ice to prevent premature polymerization, since the material is liquid at 4º C and jellifies above 10ºC [105]. When crosslinked, Matrigel TM forms a randomly weaved mesh of fibers which withholds a large amount of fluid. Several types of Matrigel TM formulations with various concentrations can be obtained commercially, such as those produced by Corning Life Sciences ® , BD Biosciences ® , and Trevigen ® [106]. This scaffold has been used to establish different types of 3D organoids ( Figure 2A) and 3D spheroid ( Figure 2B) models of various cancers, including those recognized by a highly aggressive progression such as pancreatic cancer [107]. Due to its in vivo origin, Matrigel TM -based scaffolds introduce ECM-specific signaling molecules and binding domains, such as laminin, collagen, elastin, entactin, fibronectin, fibrinogen and different growth factors (e.g., Epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), Platelet-derived growth factor (PDGF)), among others [109]. Such bioactive components provide for example integrin and MMP binding sites, both of them imperative in tissue organization and cancer metastasis [60]. As a result of their origin, natural-based scaffolds exhibit similar structural interactions to those found in humans providing a suitable in vivo-like matrix where different cell lines can proliferate and acquire a stem-like phenotype [109,110]. However, due to the animal origin of Matrigel TM , this material also exhibits significant batch-to-batch variability [109], which results in low data reproducibility [27]. Furthermore, the complex and variable protein composition reduces the possibility to mimic tissue-specific ECM environments [27]. For example, Matrigel TM does not contain proper ratios of collagen type I or hyaluronan as those found in the matrix of in vivo tumors [105]. Moreover, it is important to emphasize that dynamic control over matrix stiffness in a dynamic mode is not possible [111].
However, it is possible to obtain Matrigel TM formulations with different protein concentrations and thus manufacture gels with tunable elastic moduli. Such is an important aspect in 3D in vitro tumor models' establishment, especially since matrix stiffness has been directly correlated with cells migration processes [112].
Collagen-based 3D models can be assembled either through physical or chemical\enzymatic crosslinking [126], leading to the establishment of a fibrillary scaffold.
Moreover, through the variation of protein content, or addition of synthetic cross-linking agents scaffolds stiffness, elasticity and fiber alignment of collagen gels and scaffolds can be varied to simulate matrix stiffening that occurs during disease progression [86]. As such, collagen hydrogels have been extensively used by various researchers to establish in vitro models for evaluation of EMT and cellular migration. For example, collagen has been used to study breast cancer cell invasion through variations of fiber alignment and scaffold stiffness [71]. A study by Fraley and co-workers [117], analyzed the effect of distinct collagen gels crosslinking level, pore size and varying fibrillar alignment combinations on cancer cells mobility and ability to migrate in complex ECM mimetic matrices [117]. The main results demonstrated that according to fiber alignment the protrusion rate of cancer cells and their respective orientation was greatly influenced. Furthermore, the levels of matrix metalloproteinase activity and their respective inhibition were highly dependent on matrix structure and collagen density, mimicking the complexity of in vivo tumorassociated matrix. This capacity to modulate cancer progression and invasion highlights the necessity of standardizing ECM mimetic microstructure characterization and composition. However, given the origin and composition of collagen, these in vitro scaffolds are prone to batch-to-batch variations, similar to those obtained with Matrigel TM .
Other natural origin material that has been extensively investigated for assembly of 3D in vitro tumor models is hyaluronic acid (HyA). HyA is a major glycosaminoglycan found ubiquitously in normal and malignant tissues ECM, being comprised of Dglucuronic acid and D-N-acetylglucosamine residues [129]. HyA has been closely associated with cancer progression [130,131], and the presence of elevated quantities of HyA in the tumor stroma has been associated with poor patient outcome [132]. Given HyA importance in the TME a diverse array of studies has employed HyA-based scaffolds to establish 3D in vitro tumor models. Some of these studies take advantage of HyA unique chemical versatility that allows tailoring of its mechanical properties via precision chemical modification. One recent study that explores this possibility is that reported by Shen and co-workers [116], which developed a 3D scaffold based on acrylated hyaluronic acid hydrogels containing MMP-1 and MMP-2 sensitive peptides and bioadhesive RGD- Whereas, the establishment of hypoxic conditions lead to sprouting even in highly dense HyA hydrogel networks ( Figure 3F). These materials present low immunogenicity and suitable biocompatibility for in vitro cell culture [88]. Alginate is derived from brown algae and formed by repeating units of α-Lguluronic acid and β-D-mannuronic acid. Variations in the relative concentration of the two monomeric units produce changes in alginate physical and chemical properties, and the formation of gels with higher or lower water holding capacity and tunable porosity [133]. Alginate has been extensively used for cell encapsulation due to its capacity to rapidly crosslink under physiological conditions, easily forming a matrix that allows for medium and metabolite exchange [134,135]. However, it is important to mention that alginate possesses no cell adhesion properties, being often chemically modified with peptide-moieties or combined with other types of bioactive polymers [136].
Chitosan is another biopolymer that has been extensively used to form porous scaffolds for cell culture and tissue engineering due to its biodegradable, nonimmunogenic and cationic nature which also promotes a facile functionalization with anionic molecules such as anionic glycosaminoglycans (e.g., HyA, heparan sulfate, heparin, chondroitin sulfate, keratan) [137]. The overall cationic charge of chitosan is attributed to its polymeric backbone which is comprised of glucosamine and Nacetylglucosamine units [137]. Chitosan has been conjugated with other types of biopolymers (e.g., alginate, HyA) [138,139] or macromolecules (e.g., bioadhesion domains, signaling cytokines such as VEGF or bone morphogenetic protein 2 (BMP-2)) to form biofunctional 3D scaffolds for tumor in vitro modeling [140,141].
Combinations of both alginate and chitosan to produce 3D in vitro disease models are one of the most common. For example, in 2013, Kievit and co-workers [118], formulated a highly porous polyelectrolyte chitosan-alginate (CA) composite scaffold that was used to establish a 3D glioblastoma model to study cancer cells (U-87) migration. The manufactured CA scaffold presented well-defined and aligned-fibers with small diameters ranging from 200 nm to 1.1 µm. This promoted glioblastoma cells 3D migration and evidenced that this may be a cost-effective testing platform to screen for anti-metastatic therapies. Furthermore, later in 2016, Kievit and co-workers employed a similar scaffold to study glioblastoma cells interactions in diverse ECM mimetic or non-mimetic scenarios [119]. In this study, different CA scaffolds, namely those coated with poly-ε-caprolactone (PCL) (lacking bio-adhesion domains), or coated with HyA (allowing interaction with cell receptors -CD44/HyA interaction) were investigated ( Figure 4G). The produced scaffolds where then used for glioblastoma 3D cell culture in mono or co-culture conditions, using astrocytes and endothelial cells to better mimic glioblastoma TME and its complex cell populations ( Figure 4D, E, F) [119]. The obtained results showed that cells cultured in scaffolds coated with ECM mimetic HyA had an increased expression of stemness markers (CD44 and CD133) ( Figure 4G, H). Silk fibroin, has also been extensively used for the development of 3D in vitro models, particularly for the study of cancer-to-bone metastasis [120,121], due to its resistance to protease degradation, as well as mechanical and cell adhesion properties (e.g., resultant from the naturally occurring RGD sequences of the fibroin protein). Adding to this silk fibroin can also be conjugated with different biomolecular cues (e.g., BMP-2, EGFR, SDF-1) [142,143]. To date, these scaffolds have been used for establishing in vitro 3D models of primary tumors including those of prostate [85], and osteosarcoma [122].
Decellularized matrices (dECMs) such as those obtained from malignant/healthy tissues decellularization, or by through in vitro ECM production by 2D cultured cells [124] have been gaining increasing attention given their remarkable ability to provide native ECM components. However, the difficulty in producing such matrices in high-throughput compatible platforms, combined with the lack of structural and architectural control still hinders decellularized matrices widespread use as scaffolds for 3D in vitro tumor models assembly. Hence, recent studies have been mainly focused on the production and characterization of highly reproducibly cell-derived dECMs [124], or in the production of disease-specific decellularized matrices [125]. Using this concept, Rijal and co-workers described the development of a 3D tissue matrix scaffold, obtained from decellularized breast tissues, that was processed either into hydrogel or porous scaffolds which recapitulate native tissue architecture and resilience [125].

Synthetic Materials based Scaffolds
Alternatives to naturally derived scaffolds include synthetic, polymer-based, scaffolds that can be precisely manufactured to include ECM-mimicking cues and tweaked biophysical properties [103]. Several synthetic polymers exhibiting bioactive, biocompatible and biodegradable properties have been synthesized and reported in the literature in the recent decades, namely poly(ethylene glycol) (PEG), polylactic acid (PLA), poly(glycolic acid) (PGA), poly(lactic-co-glycolic acid) (PLGA) and poly-εcaprolactone (PCL) [5,133]. Overall, batch-to-batch variability and the lack of a precise control over scaffolds mechanical properties associated with natural scaffolds is eliminated when using well-defined, synthetic polymer-based scaffolds [144]. Moreover, when functionalized with bioactive molecules they serve both as bioinstructive and structural anchors until de novo matrix deposition by cancer and stromal cells occurs [28].
However, cells cultured in purely synthetic platforms can proliferate devoid of tumor-like gene expression patterns, presenting inconsistent tumorigenicity, metastatic potential or drug-resistant phenotypes, when compared to those of in vivo tumors [28,145].
Expectedly, such issues affect the production of robust tumor-mimicking 3D in vitro models. To overcome such drawbacks synthetic materials are often combined with other polymers such as polyvinyl alcohol (PVA), or with naturally derived biopolymers such as chitosan, hyaluronic acid or polydopamine, in order to attain more in vivo-like conditions [87,146,147].

Hybrid Scaffold Based Models
The development of hybrid-based scaffolds for the assembly of 3D models is based on the incorporation of natural bioactive molecules (e.g., growth factors), and bioadhesive moieties (e.g., peptides), into the highly tunable/controllable matrix of synthetic scaffolds [88,133]. Synthetic materials can act as a 'white-canvas' and be conjugated with natural polymers such as fibrin [148], HyA [74], or specific bioactive molecules (e.g., BMP-2, RGD peptides). PEG and its derivatives (PEG-diacrylate) have been the most widely used synthetic polymers for conjugation with bioactive molecules and to ultimately form 3D biofunctional hybrid hydrogel scaffolds for cancer in vitro modeling. As demonstrated by Weiss and co-workers, PEG polymeric backbone was successfully functionalized with a peptide containing RGD domains [149], which resulted in increased cellular adhesive properties. In addition, MMP or plasmin-sensitive sequences have also been chemically coupled into PEG [114,144]. The inclusion of these bioactive moieties increased cellular interactions (e.g., cell-ECM and cell-cell), and mimicked tumor-ECM specific degradability [150]. One example of this strategy is the study performed by Roudsari and co-workers [151], in which PEG-based hydrogels, containing both MMP sensitive (GGGPQGIWGQGK), and cell adhesion (RGD) peptides, were used to co-culture lung adenocarcinoma cells (3445Q) with endothelial and pericyte vascular cells. These proteolytically-degradable models promoted tubule-like network formation guided by interactions with cancer cells and provided a suitable platform to study tumor neovascularization.
Compared to conventional natural or synthetic scaffold-based approaches, advanced hybrid scaffold-based 3D in vitro models represent more robust platforms in which several aspects of tumor progression can be recapitulated. In fact, hybrid scaffolds could allow to evaluate the role of specific ECM components in events such as metastasis [152] or cancer cells proliferation [153]. However, the manufacture of hybrid scaffold-based models generally requires laborious production procedures, an important aspect that limits their use in high-throughput screening platforms [27]. Combination of hybrid scaffold-based methodologies with scaffold-free based techniques has provided novel and interesting platforms that can be used to modulate, with relative ease, several aspects of tumor progression, such as ECM deposition, metastasis, genetic drift and angiogenesis.
Recapitulating the diverse facets of tumor progression in multifactorial hybridbased approaches, allows more predictive models to be obtained. A recent study by Hirt responses [154].
Other promising models have combined microfluidic platforms with collagen matrix hydrogels [155]. Microfluidic systems entail the use of micrometer-sized channels that open the possibility to produce 3D microtumor models under flow perfusion conditions [156]. Ultimately, scaffold-microfluidic combinations allow for a precise control of cancer cells growth by dynamically controlling cell culture media composition and manipulating drugs mass transfer via modification of liquid flow rate [157]. These characteristics make microfluidic systems ideal to perform angiogenesis, migration, or flow perfusion studies in the context of tumor perfusion and tumor invasion, EMT, cells dissemination and metastasis [28,158,159].
From the abovementioned materials to assemble 3D tumor models for drug discovery, several have been translated into commercially available platforms in recent years. As summarized in table 2, these 3D in vitro cell culture tools take advantage of both scaffold-based and scaffold-free strategies to assemble 3D microtumors in TME mimicking environments and in a monotypic or heterotypic co-culture mode. Moreover, some of the most advanced platforms also allow to model nutrient flow dynamics and are amenable to screen anti-cancer candidates in high-throughput and high-content imaging settings ( Table 2).  Although a significant number of commercial platforms has been developed, to date the majority of scaffold-based models, particularly those based on hydrogels, fail to achieve a precise control over 3D microtumors spherical morphology in a sense similar to that obtained in 3D tumor spheroids assembled by scaffold-free approaches. However, despite the fact that 3D spheroids present reproducible morphological features and tunable size, the absence of pre-existing tumor ECM components also remains a significant drawback [19,190]. The following section will describe advanced technologies based on spherically structured 3D platforms that aim to bridge the gap between the lack of gelgrown microtumors reproducible morphological control and the absence or pre-existing ECM components in 3D scaffold-free assembled spheroids.

Spherically structured 3D In vitro Tumor Models
Considering the specificities of the drug-screening process, namely the necessity of high-throughput, ease of analysis, reliability and predictable potential of the preclinical validation models [39,191], an ideal approach could require the combination of scaffoldbased models to represent ECM biochemical and mechanical complexity, along with the ease of analysis obtained from simple scaffold-free spherical models. Following the example of stem cells research in tissue engineering [17], such a combination could be achieved for example through the inclusion of microparticles containing specific ECM mimetic components. These combinations would allow the inclusion of TME specific matrix and cellular components into a spherical scaffold ( Figure 5), thus leading to the formation of composite multicellular spheroids compatible with current analysis methodologies. An alternative methodology to the inclusion of microparticles would be the encapsulation of cancer cells inside hydrogel microcapsules [16]. Microencapsulation techniques have shown the capacity to confine cells and promote a reproducible spheroid growth, while providing ECM-like components that would otherwise be lacking [16,192,193]. Several spherical in vitro tumor mimicking cancer models have been developed so far, providing innovative platforms for the study of tumor biology and drugscreening assays ( Figure 5). Herein, microencapsulation and microparticle scaffold will be critically reviewed in light of recent reports.

Microparticles for 3D Tumor Models Assembly
As previously stated, 3D-MCTS are seen has golden standard in vitro models for performing drug-screening tests [194], due to their ability to correctly recapitulate several features of the tumor microenvironment, such as: (i) cell-cell interactions; (ii) matrix deposition; (iii) cell-ECM interactions; (iv) internal structure organization (responsible for hypoxia and consequently necrotic core formation) and (v) drug-resistance, stemming both from the acquisition of a resistant phenotype by continuous low drug dose stimulation, and drugs diffusional limitations in dense tumor masses [22]. These models provide a platform that can be easily assembled and facilitates high-throughput studies [195], in comparison with more complex scaffold or microfluidic-based models. However, as previously discussed in section 2.1, conventional spheroid-based models have inherent limitations that could be overcome by including biofunctional microparticles.
Microparticles have been extensively applied in the field of tissue engineering mainly in four areas of application: (i) delivery of incorporated or surface-attached molecular cues or tethered protein into tissues or cell aggregates; (ii) reporting changes in culture conditions; (iii) serving as scaffolds for cell attachment and providing necessary cues for cell differentiation or phenotype stimulation, and (iv) introduction or preservation of local targeted heterogeneity or homogeneity [17,196]. To date, most studies have explored of microparticles mainly for structural support of for providing molecular cues, with limited works exploring microparticles potential in the context of tumor modeling.
Studies involving microparticles-based scaffolds for the production of in vitro tumor models [87,95,134,[197][198][199][200], employing either non-modified synthetic polymers, or hybrid scaffolds such as the previously discussed PEG-Fibrinogen model, later developed by Pradhan and co-workers into the format of microspheres that allowed the assembly of spherical cancer models [95].
As mentioned, in comparison to scaffold based models, 3D-MCTS main limitation is the lack of a pre-existing ECM-like supporting structure. As a result, contrarily to what happens in vivo¸ the ECM will not be able to guide or influence cancer cells and stromal cells from the onset, ultimately failing to provide the necessary initial cues for characteristic cancer phenotypes to arise [201,202]. Consequently, the acquisition of phenotypes that do not resemble those found in vivo can be observed in certain culture settings. For instance, Brancato and co-workers [203], reported that for spheroids assembled either for cancer-associated fibroblasts (CAF) or normal fibroblasts, different cell metabolism, cell growth, matrix deposition rates and mechanical properties were observed when culturing cells with or without the support of porous gelatin microparticles [203]. Through the use of such microparticles, the authors were able to better replicate the functional and metabolic differences found in vivo between healthy and neoplastic tissues containing CAFs [203]. This study exemplifies that the introduction of microparticles into spheroid-based tumor models allows to surpass such limitations.
Emerging reports have described the use of microparticle-based scaffolds as a means of introducing previously lacking ECM components into 3D-MCTS, promoting stem-like or multidrug resistance profiles [197,199,203]. The production methodologies for these microparticles mainly involve the application of modified double emulsion methods and sieving, with the combination of both techniques allowing a high yield of microparticles in the desired size ranges [95,197,198,200,203]. A significant body of knowledge on the methodologies to prepare polymer and hydrogel microparticles has been accumulated in the field of drug delivery systems in the past decades [204], and could be transposed to support spherical cancer models production. One can envisage that simple compact particles to the surface of which cells can adhere, or porous micro-or macroparticle formulations (with enhanced surface area for cell attachment and adding additional volume for cell colonization) (Figure 5), can provide a structure capable of recapitulating aspects of the TME. Sahoo and co-workers [87], produced porous microparticles based scaffolds which allowed cells to interact with a semi-rigid or rigid ECM-like structure. The obtained PLGA/PLA microparticles exhibited diameters ranging from ~100-260 µm and consequently significant surface area for cell attachment. After a period of 5 days, microparticles were completely covered in cell layers, that ultimately formed a spherical aggregate in which compact cell-cell adhesion characteristics found in 3D scaffold-free techniques were reproduced. This approach allows the establishment of 3D-MCTS when used in combination forced-floating, hanging drop or stirring-based methodologies [87].
In the context of particle porosity, the work of Kang and co-workers [200], established a cryopreservable tumor model of MCF-7 using PLGA microspheres with an average particle diameter of 393±5 μm, an exterior pore size ranging from 10-70 μm with intertwined porosity (Figure 6). The particles were used for cancer cells culture in stirred suspension bioreactors, achieved an elevated growth rate (2.8-fold cell expansion over seven days), increased resistance to doxorubicin when compared to 2D counterparts, as well as maintained viability and metabolic profiles after the process of cryopreservation. Moreover, the model exhibited increased effectiveness in establishing tumors on athymic female mice, with MCF-7 cells cultured on microspheres presenting a 4-fold increase in tumor formation [200]. So far, most studies involving tumor modeling produced polymeric microparticles as supporting scaffolds for assembling breast cancer cell spheroids, but mostly restricted to the MCF-7 lineage [95,197,198,200,203,205]. Several studies made use of Microparticles scaffolds as a mean of culturing breast cancer cells for measuring the cytotoxic effect of diverse pharmacological compounds such as doxorubicin, cisplatin, gemcitabine, paclitaxel, and tamoxifen [197,198,200]. In this context, Horning and co-workers [198], used a combination of PLA and chitosan to create microparticles with diameters of 160-182 µm for culturing MCF-7 cells and evaluating its cytotoxicity profile against doxorubicin, paclitaxel, and tamoxifen [198]. The authors performed a comparison analysis between 2D and 3D models and observed that drug internalization was significantly delayed in the 3D model. (Figure 7). In fact, while in 3D models containing microparticles doxorubicin only reached the spheroid core region after 8h of incubation, in 2D models such observations were visible within the first hour. The ability to recapitulate in vivo arrangements and expression patterns may pave the future for screening novel therapeutics targeting specific TME hallmarks. Another study by Brancato and co-workers improved on the previous porous gelatin microparticlebased model of stroma through the addition of MCF-7 cancer cells. This breast cancer co-culture model containing porous gelatin Microparticles, was used to test a targeted nanoparticle drug delivery system [197]. A comparative analysis confirmed elevated expression of MMP-2 and other metalloproteinases in the 3D model versus 2D cultures, hence better mimicking in vivo overexpression by breast cancer cells in the TME.
Interestingly, this enzymatic overexpression was effectively exploited via an enzymeresponsive targeted delivery system, comprised by PLGA-PEG nanoparticles and a tumor targeting pro-drug activated by MMP-2 degradation. The obtained results evidenced increased specificity of targeting system in microparticles-based models, with the efficacy of the nanoparticles being confirmed through increased cytotoxicity in the 3D model ( Figure 8) [197].  [197], with permission from Elsevier.
Despite extensive implementation of microparticle-based scaffolds in tissue engineering applications, there is still a tremendous untapped potential for exploring these assemblies in the field of in vitro tumor modeling. In fact, most models containing microparticles merely focus their utilization as cell culture vehicles or for culture in bioreactors. Production of finely tuned microparticle structures using advanced 3D printing and micropatterning technologies [206], or through the use of flow-focusing microfluidic devices [207], could allow to study the role of specific signaling cues in in vitro expanded malignant cells. These novel approaches may shed light upon biomolecules specific roles and enhance our capacity to modulate the TME in vitro.

Microencapsulated 3D Models
Encapsulation of cancer cells or spheroids is a promising strategy for tumor modeling that has received considerable attention in recent years [5]. Microencapsulation can serve as a mean of representing spatially defined ECM mimicking scaffolds. This strategy allows cancer and stromal cells to grow (in mono or co-cultures), and establish both cell-cell and cell-ECM interactions in a confined, yet, not fully isolated environment.
Furthermore, the encapsulation of cancer cells, particularly in spherical-shaped, sizecontrolled microcapsules with semi-permeable membranes, allows bidirectional diffusion of nutrients, oxygen, therapeutic compounds and low/medium molecular weight signaling molecules (e.g., growth factors and cytokines). In addition, microencapsulation can be employed to prevent the penetration of high molecular weight objects such as antibodies and immune cells [208], having been originally used as a tool for cell transplantation and immune isolation. In the field of tumor modeling, microencapsulation has been employed in a diverse set of ways that will be discussed in the following examples [16].
The ability to restrain direct cellular contact makes microencapsulated 3D-MCTS as an ideal model to study the diverse paracrine interactions occurring in the TME between the different key cellular populations such as immune cells and mesenchymal stem cells [209]. This capacity of microencapsulated models was exploited by Yeung and co-workers [210], to study non-direct communication between neuroblastoma and bone-marrow derived mesenchymal stem cells. By using a collagen microsphere system, the authors demonstrated mesenchymal stem cells ability to promote neuroblastoma growth [210].
Such combinations of diverse cell populations in indirect contact can also be achieved in a hierarchically structured way, as demonstrated by the multilayered models produced by Fang and co-workers [192]. These authors manufactured hierarchically-assembled microencapsulated tumor models of prostate cancer cells (PKD1), and prostate cancer TME associated stroma cells (WPMY-1) in an alginate hydrogel. By separating the diverse cellular populations into different particle sub-layers, the authors achieved an ideal model in which to study the paracrine interactions established between both cell populations ( Figure 9). Adapted from Fang and co-workers [192], reproduced under Creative Commons License.
In a recent study by Lu and co-workers [211], compartmentalized hydrogel microparticles containing ECM-mimetic scaffolds were produced through the combination of multifluidic electrospray of hydrogel particles with ionic/thermal gelation mechanisms ( Figure   10). The ability to obtain compartmentalized platforms is important for several fields of tissue engineering and 3D in vitro disease models, a point which the authors demonstrated by utilizing the novel system to perform 3D cultures of small intestinal organoids, as well as tumor and hepatic microtissues. By using this platform, the authors were able to produce size-controlled microcapsules at a high rate (10 000 particles min -1 ), containing direct or  Layer-based microencapsulation can also be for the proliferation of key tumor subpopulations as demonstrated by Rao and co-workers [212], for prostate cancer stem-like cells (PCa-CSCs). By manufacturing liquid core microcapsules with an alginate hydrogel shell, the authors were able to accelerate PCa-CSCs production from 10 days to 2 days.
When compared to growth in ultra-low attachment plates, microencapsulated prostate spheroid models presented a higher degree of stem cell surface receptor markers and higher pluripotency, which combined with the rate of production, renders this platform suitable for the production of PCA-CSCs in vitro [212].
On a different perspective, microencapsulation can also provide a platform for enhanced recovery of both cells and cell-secreted factors [213,214]. A study by Cui and co-workers [215], demonstrated the feasibility of easy cell recovery through thermal-dissociation of microcapsules incorporating HeLa cells which were readily recovered as aggregates under specific temperature conditions. Another study, by Huang and coworkers [193], reported a microcapsule model formed by gelation of a newly discovered peptide for encapsulation of MCF-7 cells. The methodology employed for microcapsule formation allowed cells to be encapsulated at physiological pH and temperature, in minimum essential medium, decreasing cytotoxic effects sometimes associated with microencapsulation processes [16,216]. Through shear stress, caused by pipetting, the gel was easily converted back to its liquid form allowing recovery of breast cancer cells [193].
Furthermore, cytotoxicity assays with cisplatin revealed that the models were suitable for drug-screening assays, by allowing free penetration of the drug. Such ability to isolate specific cells further increases the capacity of studying genetic and phenotypic alterations in specific sub-sets of the cultured cells. In a recent study by Yang and co-workers [217], spherical alginate-based microcapsules were used to cultivate low passage human mucoepidermoid cells, and also to isolate angiogenesis-related molecules released from these cancer cells. The analysis of 3D cultures phenotypes and genotypes revealed a higher expression of pro-angiogenic genes and hypoxia associated factors in comparison to those obtained in standard 2D cultures [217].
Microencapsulation can also be exploited for the assembly of hierarchic 3D tumor models. In a recent report, Agarwal and co-workers [218], microencapsulated breast cancer cells (MCF-7) in collagen I and alginate core-shell semi-permeable microcapsules (~400 µm) that served units for the bottom-up assembly of a 3D microtumor. This hierarchical model was able to promote de novo vasculature establishment and organization when encased under perfusion in a collagen I hydrogel containing human umbilical vein endothelial cells (HUVECs) and hADSCs ( Figure 11B). This merged structure was then placed inside a microfluidic chip, being subjected to physiologic perfusion conditions. By introducing such physical and biochemical cues, through paracrine communication with encapsulated cancer cells, the authors effectively mimicked HUVECs vasculogenic morphogenesis. Furthermore, in vitro analysis revealed highly increased resistance to doxorubicin when compared to avascular and 2D models containing the same cellular ratios (4.7 and 139.5 times respectively) ( Figure 11D). Similarly, to microparticles, microcapsules can also serve as a technology for the inclusion of tumor ECM components. Several studies demonstrated that for neuroblastoma [210], lung [219] and breast [29,208,210,220] cancer microencapsulated spheroids establish cell-cell signaling interactions similar to those observed in vivo. Moreover, internal ECM matrix components deposition occurs inside the microcapsules, leading to increased resistance when compared to conventional 2D models. Such makes these 3D microencapsulated models possibly suitable for drug-screening assays and research in tumor drug resistance. Interestingly, microencapsulated 3D-MCTS models can mimic for example both solid tumor density [29], cell-matrix interactions, and the mechanical and physical pressures resultant from uncontrolled expansion of tumor masses, which can promote cancer metastasis and lead to profile alterations in cancer cells [216]. In fact, as demonstrated by Guzman and co-workers [221], depending on the elasticity of the chosen microcapsule, these can allow the study of the invasive processes carried out for example by invasive breast cancer [221] and other epithelial tumors [222], recreating the breaching of the involving basement membrane layer that surrounds the primary tumor site.
Furthermore, as elegantly demonstrated by Alessandri and co-workers [216], microcapsules can be used to study the buildup of intra-tumoral pressure, decurrent from the increasing of tumor mass generating increasing pressure on adjacent tissues and conversely compressing the tumor [223].
Microcapsule-encompassed spheroids are assembled through several methodologies, the most common of which being generation of liquid-core structures by employing microfluidic devices and hydrogel reticulation methods [29,208,216,[224][225][226].
Frequently assembled microcapsules present diameters in the order of a few 100 µm to 500 µm, an exception being the study produced by Pradhan and co-workers [205].  Microfluidic-based approaches use flow-focusing, T-junction chips or more complex channel designs [16,194] to generate droplets of cell suspensions mixed with preselected biocompatible polymers or hydrogels. In these strategies, encapsulation into spherical structures is achieved by exposure to a crosslinking agent, such as calcium bath solutions or UV light, which triggers gelation and produces microcapsules containing the desired cells. Encapsulated cells assemble over time to form matrix-encapsulated spheroids capable of establishing both cell-cell and cell-matrix interactions [227]. An excellent example of such application is the formerly mentioned study by Alessandri and co-workers [216]. In this study, a model of colon carcinoma based on murine CT26 colon cancer cell line was assembled through a simple and highly reproducible method, based on a microfluidic co-extrusion chip assembled by co-centering three glass capillaries extruding sequentially cell solution, calcium free solution and alginate solution into a calcium bath ( Figure 13) [216]. The authors were able to assemble highly elastic spherical microcapsules that acted as quantitative mechanical sensors to measure the internal pressure resultant from the expanding tumor cells. Moreover, the researchers found that peripheral cells inserted in the encapsulated model readily escaped the spheroid environment, while the spheroid invasive profile was not present in non-confinement 3D models ( Figure 13I) [216].   [16].
The work developed by Lee and co-workers [225], is an excellent example of the latter technique. In this study, the authors used diffusion-mediated encapsulation, performed in PDMS-micromolds where hepatocarcinoma spheroids were previously assembled. Such spheroids were subjected to posterior deposition of an encapsulating alginate hydrogel through nano-porous membranes, which allow a control over crosslinking agents deposition rates [225]. Lastly, 3D bio-printing has also been used by Xu and co-workers [228], to produce high-throughput automated encapsulation of ovarian cancer cells and fibroblast co-culture droplets in Matrigel™. This approach allows the study of co-culture interactions in diverse settings due to high control over initial cell density and spatial arrangement of the patterned structure of the model, paving the way for the development of more complex and precise spherical scaffold-based tumor models to be generated.

Tumor Models
Establishing reproducible and easy to analyze spherically structured in vitro 3D tumor models containing bioinstructive tumor-ECM moieties requires the manipulation of key parameters including: (i) the inclusion of multiple cellular components of the tumor stroma (e.g., fibroblasts, mesenchymal stem cells, adipocytes, etc) and their cell-cell ratios; (ii) type of culture medium; (iii) scaffolds biodegradability; (iv) selection and degree of functionalization of bioactive ECM-mimetics; (v) 3D spheroids size, as this parameter influences the formation of the characteristic necrotic core of solid tumors; and (vi) microparticles/microcapsules porosity due to its role in nutrients and cells diffusion through the scaffolds. Moreover, other parameters such as tumor ECM mechanical and viscoelastic properties must be controlled to closely mimic those of the native diseased tissues. This one of the most important parameters since as demonstrated by Alessandri and co-workers [216], the elasticity of the matrix in which cells adhere can influence the establishment of a pro-metastatic phenotype, simulating the pressure exerted by surrounding tissues over the tumor mass [222]. Careful consideration must also be given to the manufacturing processes required for the fabrication of spherical 3D models based on microparticles or microcapsules, since the crosslinking processes might require cells exposure into deleterious, non-physiological conditions such as acidic/basic pH [216], organic solvents, or exposure to high-intensity UV light during photo-crosslinking reactions, all of these resulting in loss of cellular viability [16].

Conclusions and Future Perspective
The need for expediting drug research both at the preclinical validation level and discovery of novel targets is crucial for the management of currently incurable diseases such as cancer. Research regarding the development of novel 3D in vitro models is increasingly contributing to this goal by providing innovative platforms capable of efficiently, predictively, and robustly mimicking in vitro the complex in vivo reality of the TME in what regards its cellular and ECM components. Among the vast array of 3D cell culture methodologies that have been developed to date for in vitro tumor modeling, 3D spheroid-based models are the most promising regarding the production of highthroughput usable and affordable tumor mimetics. Microencapsulation and microparticle-based production technologies are capable of recreating complex cell-cell, mechanical and physiological characteristics that recapitulate in vivo solid tumors, at both the cellular and ECM level. These are highly valuable characteristics since standard 3D spheroid models lack correct ECM representation and confinement of soluble mediators (e.g., growth factors, cytokines) in controlled environments such as those found in human tumors niche.
Overall, there is a tremendous potential for improving 3D spheroid-based drug screening platforms by combining the knowledge acquired in scaffold-based methodologies with microencapsulation or microparticle inclusion techniques to form spherical microtumor constructs compatible with already implemented analysis methodologies (e.g., high-content imaging) [16,17]. These healthy tissue constructs could be used to study the metastasis process from 3D tumor models in multi-compartment organ-on-a-chip platforms. This is envisioned to contribute to a faster discovery of more effective anti-cancer and anti-metastatic compounds or compound combinations.

Statement of Significance:
The ability to correctly mimic the complexity of the tumor microenvironment in vitro is a key aspect for the development of evermore realistic in vitro models for drug-screening and fundamental cancer biology studies. In this regard, conventional spheroid-based 3D tumor models, combined with spherically structured biomaterials, opens the opportunity to precisely recapitulate complex cell-extracellular matrix interactions and tumor compartmentalization. This review provides an in-depth focus on current developments regarding spherically structured scaffolds engineered into in vitro 3D tumor models, and discusses future advances toward all-encompassing platforms that may provide an improved in vitro/in vivo correlation in a foreseeable future.  Cellusponge Disc shaped Collagen type I or Galactose-based scaffold that allow cells to be cultured in easy to use 3D - [170] responsive gels, or gels tailored by the addition of specific cell adhesion domains

3D Perfusion Bioreactor
Combination of bioreactor technology with the 3D Biotek® PCL disc inserts for the formation of perfused microtissues - [160] 3DKUBE™ 3D cell scaffold-culture chambers which allows the establishment of independent scaffold-based cell cultures under perfusion - [161] Hybrid -

Microfluidicbased Platforms
Ibidi™ µ-Slide III 3D Perfusion, A set of microfluidic devices capable of working in static or fluid perfusion conditions, in which cells included in 3D scaffold-based models (e.g., Matrigel or other gel-based system) can be cultured. These platforms are suitable for simulating perfusion conditions, allowing for example drug administration under flow, chemotaxis and migration studies to be performed. The tumor models in chips/slides can be analyzed in real-time by microscopy-based analysis - [183,18 4] TissUse™ Organ-on-a-Chip Organ-on-a-chip microfluidic devices that can accurately mimic physiological flow in microchannels. These can work in either free-circulation or closed-loop setup, allowing communication between reservoirs that can contain 3D scaffold-based models of tumor and healthy tissues developed by the user Provides a specific service of chip design and tailored healthy tissue organoid integration, oriented for drug screening [185] MIMETAS™ Organplate Models High-throughput compatible, organ-ona-chip platforms that allows insertion of scaffold-based 3D models (e.g., gelbased), into close-loop microfluidic platforms. These platforms allow direct contact between scaffold containing sections and fluid containing channels by employing a patented phase guide system