Engineering strategies to capture the biological and biophysical tumor microenvironment in vitro

Abstract

Despite decades of research and advancements in diagnostic and treatment modalities, cancer remains a major global healthcare challenge. This is due in part to a lack of model systems that allow investigating the mechanisms underlying tumor development, progression, and therapy resistance under relevant conditions in vitro. Tumor cell interactions with their surroundings influence all stages of tumorigenesis and are shaped by both biological and biophysical cues including cell-cell and cell-extracellular matrix (ECM) interactions, tissue architecture and mechanics, and mass transport. Engineered tumor models provide promising platforms to elucidate the individual and combined contributions of these cues to tumor malignancy under controlled and physiologically relevant conditions. This review will summarize current knowledge of the biological and biophysical microenvironmental cues that influence tumor development and progression, present examples of in vitro model systems that are presently used to study these interactions and highlight advancements in tumor engineering approaches to further improve these technologies.

1. Introduction

Cancer presents a major global healthcare challenge as it accounts for 1 in 6 deaths according to the World Health Organization. In the United States (US) alone, the National Cancer Institute (NCI) estimates that over 1,800,000 new cases and 600,000 deaths will result from the disease in 2021 [1]. Although great strides have been made in treating early-stage cancer, the survival rate of patients dramatically decreases in late-stage disease, which is characterized by metastasis and therapeutic resistance [1]. Given these connections, there is a clear clinical need for treatments that can prevent metastasis and poor clinical outcomes, but the development of such treatments requires improved understanding of the mechanisms underlying tumor development, progression, and therapy resistance.

Cancer progresses in a series of distinct phases: following initial growth, tumor cells migrate into the adjacent stroma, intravasate into blood vessels, circulate to distant sites where they extravasate and finally form secondary tumors [2]. This process can take decades and it is difficult to predict which cells disseminate from the main tumor and go on to form metastatic tumors. To further complicate matters, disseminated tumor cells may become dormant, enabling them to evade therapy and contribute to relapse [3]. However, the mechanisms behind tumor progression are poorly understood due to an almost exclusive focus on the tumor cells although it is now clear that the cancer’s evolving interactions with host tissues and the local microenvironment are equally essential [4]. Indeed, Stephen Paget already proposed in 1889 that cancer cells, which he referred to as “bad seeds”, need the right microenvironment, or “soil”, to metastasize effectively [5,6]. Since then, the tumor microenvironment has been shown abundantly to influence not only metastasis, but all stages of tumorigenesis [7,8]. In a pivotal study underscoring the importance of the microenvironment in regulating tumorigenesis, cancerous cells reversed their malignant phenotype when prevented from interacting with their surroundings [9]. In order to truly understand the mechanisms that drive cancer progression, research must not only focus on tumor cell-intrinsic features, but also consider the effects of their surrounding environment.

The tumor microenvironment is shaped by both biological and biophysical cues, which synergistically affect tumorigenesis (Figure 1) [10]. For example, cancer-associated fibroblasts, endothelial cells, and immune cells regulate the growth and invasion of tumors through paracrine signaling [10], while extracellular matrix (ECM) properties and changes in fluid flow impact tumorigenesis by varying the mechanical and transport properties of the microenvironment [11,12]. As cancer cells move through different tissue and organ systems the biochemical and biophysical cues to which they encounter vary considerably, and, simultaneously, the tumor cells themselves modify these cues to support their malignant progression [2]. Finally, genotypic and phenotypic variability between patients influences how tumor cells interact with their microenvironment, which contributes to interpatient tumor heterogeneity and complicates the development of effective therapies [13]. A better understanding of how tumor-microenvironment interactions and patient-specific changes thereof impact tumor initiation and progression will be necessary to develop more effective treatments and ultimately improve disease management.

Figure 1: Biological and biophysical hallmarks of the tumor microenvironment.

Schematic of the cellular (cancer-associated fibroblasts [CAFs], endothelial cells, and immune cells) and acellular features (fibrotic remodeling of the extracellular matrix [ECM], metabolic/cytokine gradients, and interstitial fluid flow and pressure) of the tumor microenvironment. Increased recruitment of CAFs leads to compositional, structural, and mechanical changes of the ECM, while vascular dysfunction impacts mass transport of oxygen, nutrients, and paracrine signaling molecules. The latter results in varied gradient formation of solutes, elevated interstitial pressure and fluid flow as well as central necrosis. Collectively, these changes promote tumor progression and invasion where cancer cells evade immune surveillance, breach the basement membrane, migrate into the surrounding stroma, and eventually intravasate into an adjacent blood vessel to metastasize to distant organs. Image created with BioRender.com.

Although progress has been made in our mechanistic understanding of cancer, it is difficult to experimentally isolate the effect of specific microenvironmental conditions on human tumorigenesis with traditional models. Conventional monolayer culture does not recapitulate the 3D complexity of the microenvironment, while species-specific differences in tumorigenesis complicate translation of findings from mice to humans. Furthermore, mouse models do not allow for the use of human or patient-specific cells unless the animal is immune compromised, removing a major aspect of the tumor microenvironment. Engineered tumor models provide promising and constantly evolving alternatives that can be tailored to each experiment by adjusting their design parameters, including the utilized cell source, biomaterials, and fabrication technology. This review will summarize current knowledge of the biological and biophysical microenvironmental interactions that influence tumor development and progression, present examples of in vitro model systems that are presently used to study these interactions, and highlight advancements that could further improve these model systems. Ultimately, engineering-based cancer technologies provide innovative approaches to better understand how tumor-microenvironment interactions regulate cancer progression and enable screening of anti-cancer compounds under pathologically relevant conditions and in high-throughput settings.

2. Microenvironmental features regulating tumor progression

2.1. Cellular composition of the tumor microenvironment

2.1.1. Fibroblasts (and CAFs)

Cancer is often described as a wound that never heals and is associated with a sustained presence of reactive stromal fibroblasts termed cancer-associated fibroblasts (CAFs) [14–16]. CAFs comprise a heterogeneous population of activated fibroblasts of different origins with elongated morphology and expression of markers such as α-smooth muscle actin (α-SMA) [17,18]. CAFs can promote tumorigenesis by causing both biochemical and biophysical changes to the microenvironment (Figure 1). Biochemical changes are mediated by a broad spectrum of CAF-secreted cytokines and chemokines. For example, CAF secretion of stromal cell-derived factor 1 (SDF1), hepatocyte growth factor (HGF), epidermal growth factor (EGF), and interleukin 6 (IL-6) induces tumor cell phenotypical changes including proliferation and stem-like properties through altered paracrine signaling [19–22]. Biophysical changes in the microenvironment are introduced because CAFs are key drivers of fibrotic extracellular matrix (ECM) deposition and remodeling that represent hallmarks of most solid tumors (ECM) [18,23,24]. In particular, fibrotic ECM creates physical boundary conditions that directly and indirectly promote tumor progression by changing tumor cell behavior and by impeding drug delivery to desired sites, respectively, as described in more detail in section 2.2 below.

In addition to tumor cells, CAFs also influence other cell types. For example, CAFs can create immunosuppressive niches via secretion of IL-6 and transforming growth factor-β (TGFβ) [25] and promote the formation of dysfunctional vasculature by secreting vascular endothelial cell growth factor (VEGF) [26], which limits efficacious drug transport. Indeed, selective depletion of CAFs increases intratumoral drug uptake in breast and pancreatic cancers which, in turn, inhibits tumor growth and metastasis [27–29]. Nevertheless, CAFs impact tumor progression in manifold and sometimes opposing ways depending on CAF populations and tumor stages. For example, depletion of alpha-smooth muscle actin (α-SMA) positive CAFs in early or late stage pancreatic cancer can worsen rather than improve tumor progression and survival [30]. These discrepancies underscore the need for relevant model systems that can help elucidate the divergent roles of CAFs in tumorigenesis.

2.1.2. Endothelial cells

In addition to transporting nutrients, metabolic byproducts, and drugs, blood vessels enable intra- and extravasation of tumor cells and provide perivascular niches that control tumor cell stemness and dormancy (Figure 1) [31,32]. Tumor interactions with the vasculature have been widely studied in the context of tumor angiogenesis, during which new blood vessels form from endothelial cells [33]. When exponential tumor growth leads to hypoxia, tumor and stromal cells upregulate angiogenic factors including VEGF, fibroblast growth factor-2 (FGF2), platelet derived growth factor (PDGF), angiopoietins, and matrix metalloproteinases (MMPs) to activate angiogenesis [33–35]. In addition, dysregulated signaling pathways in tumor cells (e.g. overactive phosphoinositide 3-kinase [PI3K] signaling) alter endothelial cell behavior by increasing secretion of inflammatory cytokines such as IL-6 [36,37]. This combination of an overactivated angiogenic program and inflammatory response in endothelial cells results in blood vessels that are disorganized, leaky, and collapsed, leading to vascular dysfunction in the tumor microenvironment [38,39].

Aside from tumor-mediated effects on endothelial cells, endothelial cells reciprocally alter tumor cell behavior [40–42]. For example, endothelial cell-secreted VEGF, angiopoietin 2 (ANGPT2), interleukin-8 (IL-8), IL-6, and TGFβ regulate the growth and cancer stem-like cell (CSC) characteristics of tumor cells [41,43,44]. This endothelial cell-mediated activation of stemness, in turn, enhances the ability of tumor cells to invade, initiate the formation of distant metastases, and evade drug treatment [45–49]. Additionally, endothelial cells promote tumor cell dormancy and chemoresistance through direct cell contact [42,50,51]. Given these connections, engineered tumor models should be designed in a manner that allows the investigation of pheno- and genotypic changes in both tumor and endothelial cells simultaneously.

2.1.3. Immune cells

Immune cells are another key component of the microenvironment (Figure 1), and tumor associated macrophages (TAMs) are one of the most extensively studied of these cell types [52]. Depending on the particular subpopulation (e.g., M1-biased or M2-biased macrophages as two extreme examples, although there is a much broader range of phenotypes), TAMs can have either pro- or anti-inflammatory effects. While anti-inflammatory macrophages are most widely studied for their tumor-promoting effects, pro-inflammatory macrophages can similarly stimulate malignancy [53]. For example, anti-inflammatory TAMs express cytokines and chemokines (e.g. C-C motif chemokine ligand 2 [CCL2], C-X-C motif chemokine ligand 12 [CXCL12], IL-6, interleukin-1 beta [IL-1β], colony stimulating factor-1 [CSF-1], and VEGF) that can suppress antitumor immunity while promoting tumor progression [54]. This has been manifested in TAMs increasing tumor cell migration, invasion, and metastasis [55–57]. TAMs can further drive tumor growth and metastasis by altering other stromal cell types that, for example, induce tumor angiogenesis [58,59]. In addition to TAMs, other immune cells such as T cells and natural killer (NK) cells are significant contributors to the immune characteristics of the tumor microenvironment and are discussed in more detail elsewhere [60]. More generally, the tumor microenvironment often exhibits characteristics of an immunosuppressive milieu, as cancer cells and associated stromal cells express negative immune regulators such as checkpoint programmed death-1 (PD-1)/programmed cell death ligands (PD-Ls) [61]. In addition to these biochemical mediators, escape from immune surveillance can occur due to biophysical mechanisms. For example, dense layers of fibrotic ECM and dysfunctional vasculature often inhibit infiltration of immune effector cells into more central tumor regions [62,63], which may be further aggravated by the fact that immune cells can activate fibroblasts and thus, ECM remodeling [64]. Better understanding the diverse roles of immune cells as a function of biochemical and biophysical heterogeneity within the tumor microenvironment has the potential to broaden the effectiveness of immunotherapies.

2.2. Acellular and biophysical aspects of the tumor microenvironment

2.2.1. ECM – Composition, structure, mechanics

In healthy tissues, epithelial cells interact with the basement membrane (BM), a sheet like, nanoporous layer of ECM primarily composed of collagen type IV and laminins [65]. The BM contributes to the barrier function of epithelial tissues and is critical for their development, organization, and polarization [65,66]. However, during malignant transformation, cells break down the BM either proteolytically or mechanically to enable tumor invasion [65,66]. Tumor-associated fibrotic remodeling by CAFs (also termed desmoplasia) is characterized by increased deposition of collagen type I, III, and VI, fibronectin, and hyaluronan (HA) [18,64,67]. These changes can be detected clinically through mammographic screening and regulate the behavior of both tumor and stromal cells by varying adhesion-dependent cell signaling. For example, increased collagen I levels drive tumor cell invasion and metastasis by activating integrins and collagen I receptor discoidin domain receptor 2 (DDR2) [68,69], while increased fibronectin promotes tumor cell invasion and stromal cell secretion of proangiogenic factors in an integrin-dependent manner [70,71]. Changes in HA content regulate tumor cell stemness and interactions with the vasculature through CD44, but can also impact integrin signaling with consequences on invasion and metastasis [72–74]. Finally, changes in ECM composition alter biochemical signaling within the tumor microenvironment by influencing the sequestration, binding, and release of many bioactive molecules [23,75] including VEGF and FGF-2 to promote angiogenesis and tumor cell proliferation [76,77].

Compositional changes simultaneously alter the microarchitecture and mechanical properties of the ECM with functional consequences on tumor progression (Figure 1). In particular, stromal ECM consists of a porous network of fibers whose structure changes in the presence of a tumor [78]. While collagen fibers in healthy breast tissue are thin, relaxed, and curvy, fibers in tumors are thick, stretched, and straight [79,80]. Furthermore, fibers align parallel to the tumor and help prevent invasion during benign stages of cancer, while more advanced stages are characterized by fibers aligning perpendicular to the tumor providing guidance cues for invasion [79,81]. Importantly, collagen fiber alignment increases ECM stiffness, which promotes tumor malignancy through positive mechanical feedback as greater ECM stiffness increases cell traction forces mediating more fiber alignment and tumor progression [82]. While some of these changes are reversible, cross-linking through matrix processing enzymes such as lysyl oxidase (LOX) increases ECM stiffness irreversibly and influences cell behaviors via changes in mechanosignaling [69,83].

Cells translate altered ECM stiffness through mechanoreceptors including integrins, which alter gene expression by stimulating Rho/ROCK-mediated cell contractility, and activating the transcription factors YAP/TAZ [84–86]. These changes result in loss of cell polarity, reduced cell-cell adhesion, and increased proliferation ultimately contributing to metastatic invasion [85,87,88]. Importantly, stiffness-induced changes in cell contractility can influence the bioactivity of several ECM components as well as soluble signaling molecules. For example, cellular traction partially unfolds fibronectin with consequences on its bioactivity and fiber stiffness as well as subsequent deposition of collagen [89,90] and mechanoactivation of TGF-β signaling [91]. In summary, changes in tumor-associated ECM composition, structure, and mechanics collectively regulate cancer progression and thus, should guide the design of engineered tumor models.

2.2.2. Mass transport – Metabolism, hypoxia, and perfusion

In the tumor microenvironment, dysfunctional vasculature, lymphatics, and aberrant ECM remodeling, result in the disruption of mass transport homeostasis [92]. In contrast to normal vasculature, blood vessels in tumors are disorganized, leaky, and collapsed, leading to a unique spatiotemporal distribution of nutrients, oxygen, and metabolites (Figure 1) [38,93]. In combination with the forces exerted by the expansion of solid tumors and lack of fluid clearance due to impaired lymphatic function, mass transport due to convection and diffusion is further perturbed by abnormal interstitial pressure and fluid flow in the tumor microenvironment [94]. The metabolic and physical effects caused by this unique transport situation have been demonstrated to increase the malignancy of tumor cells [92].

One of the most extensively studied consequences of perturbed mass transport is hypoxia, which regulates all stages of cancer by stabilizing the transcriptional regulator hypoxia-inducible factor subunit 1α (HIF-1α) [95–97]. For example, hypoxia/HIF-1α upregulates dormancy markers in primary tumors, suggesting that transport limitations can pre-determine the fate of tumor cells in metastatic target organs even prior to invasion and dissemination [98]. By promoting ECM remodeling and the expression of integrins, hypoxia also contributes to the compositional and mechanical changes of tumor-associated ECM described above, which then work in concert to increase the metastatic potential of tumor cells [99–101]. Because of the importance of hypoxia in modulating cancer progression, therapeutic strategies to target HIF-1α or normalize blood vasculature have been proposed, and discoveries related to HIF-1α have been awarded a Nobel prize [37,102,103]. While HIF-1α is the most widely studied mediator of hypoxia responses, it should be noted that other signaling mechanisms including HIF-2α and the nuclear factor-κB (NF-κB) pathway can also mediate responses to hypoxia in tumor cells [104,105].

Mass transport-dependent spatial variations in hypoxia are also essential regulators of tumor metabolic reprogramming and thus, heterogeneity [93,106]. For example, the core of tumors is characterized by a decrease in glucose and oxygen and an increase in waste products such as lactate [107]. While a lactate-dependent reduction of pH can globally promote tumor progression [108], spatial variations thereof alter the metabolic cross-talk between cells in different regions of the tumor. More specifically, hypoxic cancer cells in the core generate excess lactate from upregulated glycolysis. This lactate can then utilized as a metabolic substrate to sustain the anabolic needs of oxygenated cancer cells at the periphery of the tumor, ensuring tumor growth and survival [93,109]. Similar scenarios also occur between cancer and stromal cells, as CAFs can support the growth and progression of tumor cells in a reverse Warburg effect during which they produce lactate by aerobic glycolysis that is then used by tumor cells [110]. It is worth noting that in addition to fulfilling energy needs, the presence of lactate, as well as other metabolites such as glutamine and asparagine, independently regulate tumor malignancy and aggressiveness [111,112].

Another feature of altered mass transport in the tumor microenvironment manifests in elevated interstitial fluid pressure, often resulting in fluid flow towards lymphatic vessels [94,113]. This imparts shear forces and mechanical stress on tumor cells which modulate their invasive and malignant potential [114]. For example, elevated interstitial pressure can promote collective invasion through increased tumor cell expression of markers involved in the epithelial-mesenchymal transition (EMT) such as Snail, vimentin, and E-cadherin [115]. Similarly, interstitial flow enhances tumor cell invasion and intravasation through factors such as flow direction (transmural vs. luminal with regard to lymphatic vessels), flow strength, as well as through the redistribution of chemokine gradients [116–118]. The effects of fluid flow are not limited to the tumor cells, but also impact the behavior of stromal cells including fibroblasts and macrophages, which can then independently enhance tumor cell invasion [119,120]. These results underscore the significance of including the physical traits of mass transport when studying tumor-microenvironment interactions.

3. Engineered tumor microenvironments

3.1. Model systems recapitulating the cellular complexity of tumors

3.1.1. Conventional in vitro approaches to study tumorigenesis

Preclinical models have been used for decades to model tumor progression, and animal studies remain the gold standard for screening potential drug candidates. However, many challenges remain as most animal models are expensive and the spatiotemporal dynamics of tumor progression are not easily identified or visualized, complicating mechanistic understanding. In vitro systems enable improved control over experimental variables while maintaining aspects of physiological relevance. Historically, the most commonly used in vitro models to study cancer are immortalized cell lines growing as adherent 2D cultures. These cultures are readily accessible, easy to maintain, and can be scaled up for rapid drug screenings. However, many drug candidates fail in clinical trials, despite promising in vitro and in vivo data [121]. This lack of predictive power and other gaps in performance during translational research are linked to many factors [122]. One important contributor is the lack of cellular heterogeneity that is a hallmark of tumors in vivo, in typical in vitro models [123,124]. For example, cell lines grown in conventional 2D culture are intrinsically homogenous, which is further reinforced by the lack of cellular and acellular microenvironmental cues described in the previous sections. In contrast, 3D model systems can integrate these components and thus provide a suitable intermediate between 2D cell culture and in vivo models that can mimic some of the intratumor heterogeneity impacting clinical prognosis. When developing physiologically relevant 3D models of the tumor microenvironment, several design parameters need to be considered: selection of cellular components, choice of relevant ECM materials, generation of fluid flow, and biochemical gradients, etc. These design principles and examples for corresponding models will be discussed in the following sections.

3.1.2. Multicellular spheroid cultures

Spheroid cultures refer to multicellular aggregates mimicking 3D cell-cell adhesions found in vivo [125,126]. Spheroids are most easily formed from cell lines using the “hanging drop” method, which involves adding a droplet of cell suspension on the lid of a tissue culture plate, which is then inverted to encourage cell-cell adhesion and spheroid formation due to surface tension [127]. However, this approach often results in aspherical aggregates, which need to be transferred to a secondary culture vessel. Both limitations can be circumvented by direct culture on non-adherent surfaces such as agar, agarose or ultra-low attachment culture plates with predefined well geometry (spherical or tapered) [128] and further improved by the addition of rotation culture and viscous media (cellulose) [129]. More recently, advanced micro-patterning techniques and microfluidic devices have been demonstrated to yield the large quantities of consistently sized spheroids necessary for large-scale drug screening [130–132]. The drawback to these methods is increased cost and labor-intensive maintenance.

Spheroid cultures provide 3D cell-cell communication and physiologically relevant structures that resemble in vivo tumor conditions [133]. As cells grow and spheroids expand in size, diffusion of nutrients and oxygen throughout the spheroid becomes limited, leading to nutrient starvation and central hypoxia. Indeed, spheroids > 400–500 μm in diameter exhibit characteristic zones of peripheral proliferating cells and necrotic centers mimicking the pathophysiological gradients found in vivo (Figure 2A)[128,134]. In addition to mono-culture spheroids, heterotypic cell–cell interactions can be recapitulated by co-culturing cancer and stromal cells [135]. Co-encapsulating cancer cells, CAFs, and monocytes in a single spheroid, for example, recapitulates tumor-immune compartments by driving monocytes to adopt a TAM-like phenotype [136]. Furthermore, this heterotypic spheroid model exhibited increased resistance to paclitaxel compared to mono-cultures, suggesting that reciprocal interactions between these different cellular components alter drug sensitivity and thus, should be incorporated into models used for in vitro drug testing (Figure 2B)[136].

Figure 2: Multicellular spheroid and organoid models.

A. Multicellular spheroids formed from a single cancer cell line recapitulate diffusion limited transport resulting in spatially distinct regions of proliferation and necrosis that mimic the pathophysiological heterogeneity of tumors in vivo. Reproduced from Impact Journals, LLC with an open-access copyright policy. [134]. B. Co-culture spheroids containing tumor cells (red), CAFs, and monocytes recapitulate stromal heterogeneity. Cisplatin treatment caused rapid apoptosis (green) in these cultures, while paclitaxel was less effective. Reproduced from Elsevier Ltd with an open-access copyright policy. [136]. C. Mouse pancreatic stellate cells (PSCs, red) and pancreatic cancer organoids (green) can be embedded in Matrigel to increase stromal populations. PSCs cultured in Matrigel alone remain quiescent, but PSCs in contact with tumor organoids assume CAF-like phenotypes. Reproduced with permission from Rockefeller University Press [139]. D. Different types of patient-derived tumor organoids (bottom row) can be formed via the air–liquid interface (ALI) method, a technique that recapitulates parental tumor histology (top row) and maintains immune cell populations. Reproduced with permission from Elsevier Ltd [143].

3.1.3. Organoid cultures

Patient-derived organoid cultures provide experimental platforms to recapitulate the cellular heterogeneity of tumors in patients. Tumor organoids are derived from dissociated tissue samples or circulating tumor cells that are encapsulated in bio-mimetic matrix to encourage self-assembly and differentiation into more complex structures [137]. The most commonly used method to generate organoids is to embed samples into Matrigel®, a BM extract derived from mouse Engelbreth-Holm-Swarm (EHS) sarcoma which preserves the heterogeneity of cancer cell populations [138]. Other stromal components such as CAFs [139], immune cells [140], and vascular networks [140] can be cultured together with tumor organoids to increase cellular complexity and further assess cell-cell interactions within the tumor microenvironment (Figure 2C) [139]. Air–liquid interface (ALI) culture is another method of forming organoids in which tissues fragments are embedded into collagen with one side facing the media and the other facing air via a transwell membrane [141,142]. The enhanced oxygenation afforded by this method improves the conservation of stromal populations from the original tissues (Figure 2D) [143]. To aid in the use of these models, tumor organoid bio-banks have been generated from patient-derived and mouse-derived tumor specimens as well as matching normal tissues, enabling more widespread applications of patient samples in research [138,144].

Organoid cultures enable biomarkers and other morphological changes to be monitored and measured across different time scales ex vivo, providing unprecedented insight into the dynamics of biological processes [137,145,146]. This is particularly useful for tumorigenesis studies and patient-specific drug screening, where matching organoids can be generated to compare cellular responses in tumor and healthy tissue [147]. In addition, pluripotent stem cell-derived normal tissue organoids can be modified via CRISPR-Cas9 genome editing to become tumor organoids, which provide a useful tool to track and decipher the genetic determinants of tissue morphogenesis and cancer progression [148]. However, the use of organoid cultures may be complicated due to variability resulting from tissue harvesting and formation efficiency. Moreover, expansion of organoids may lead to selective loss or overgrowth of stromal populations, limiting studies focused on tumor-specific heterogeneity [143,149,150]. Another limitation is the use of complex media that requires specific supplements and frequently uses fetal bovine serum, which could potentially influence cell signaling and drug efficacy testing, but could be overcome by using matched patient peripheral blood serum [151,152]. Not all organoid cultures can be expanded long term and tumor organoid biobanks are primarily focused on epithelial cancers, although tumor organoids can be generated from other cancers such as glioblastoma [153]. Collectively, organoid cultures are an invaluable technology to create patient-specific 3D in vitro models of cancer and are likely to replace cell line studies in the future.

3.2. Biomaterials-based model systems recapitulating tumor-ECM interactions

3.2.1. Natural biomaterials

Embedding cells into biomaterials mimicking the compositional and biophysical properties of tumor-associated ECM allows studying the resulting functional consequences on tumorigenesis. Given their intrinsic bioadhesivity, natural biomaterials isolated from mammalian tissues are particularly attractive. BM-mimicking materials such as Matrigel® and Cultrex® are widely used to study organotypic tissue morphogenesis and initial invasion of transformed epithelial cells through the BM [154–156]. For example, normal epithelial cells cultured in Matrigel® adopt spherical and acinar structures similar to mammary tissues, but activation of oncoproteins such as papillomavirus E7 (HPV E7) and erythroblastic oncogene B (ErbB2) can result in excessive proliferation and aberrant morphogenesis (Figure 3A)[156]. While Matrigel® mimics the biochemical composition of the BM, it is subject to batch-to-batch variations and lacks the structural and mechanical properties of the BM in vivo [157]. This shortcoming can be partially circumvented by using BM in its native and intact form (e.g., isolated from rat peritoneum), but limited quantities usually prevent large scale experiments [157,158]. In addition, commercial BM preparations cannot recapitulate the range of tissue stiffnesses present in normal versus diseased tissues (e.g. 10 times stiffer in breast cancer vs. normal breast tissue [85]) or different metastasis-prone organs (hundreds of pascals in brain versus gigapascals in bones [159,160]). Furthermore, commercial BM preparations contain growth factors and cytokines that affect cell behavior independently. One approach to avoid these limitations is to use single ECM components including collagen [82,161], fibronectin [162], and HA [163,164].

Figure 3: Biomaterials models to study cell-ECM interactions.

A. Culture in Matrigel® can be used to test the effect of malignant transformation on morphogenesis. Normal epithelial cells (MCF10A) form small acinar structures similar to human mammary tissues (top right) when cultured in Matrigel®, while transformed cells (human papillomavirus E7 [HPV E7] and erythroblastic oncogene B [ErbB2]) are larger and compose multiacinar structures. Reproduced with permission from Elsevier Ltd. [156]. B. (i) When embedded into collagen, co-culture spheroids composed of tumor cells and obesity-associated stromal cells invade collectively due to collagen displacement, a process that can be partly inhibited by MMP inhibition with Batimastat. Reproduced with permission from John Wiley & Sons, Inc. [135]. (ii) Culturing breast cancer cells in aligned collagen fiber matrices induced morphological changes (top images, cells in green, collagen in grey) that led to increased directional migration (bottom image, cells in cyan). Reproduced with permission from Elsevier [178]. C. Decellularized ECM can be derived from fibroblasts cultured in vitro. ECM derived from alpha-SMA (red) positive, desmoplastic human fibroblasts is more aligned and enriched for fibronectin (green) compared to ECM derived from normal fibroblasts. Reproduced with permission from Elsevier [182]. D. A hybrid alginate hydrogel modified with PEG spacers enables introduction of varied viscoelastic properties. Faster stress relaxation (left to right) promoted osteogenic differentiation of mesenchymal stem cells as indicated by increased matrix mineralization (Von Kossa staining, top panel) and type-1 collagen deposition (bottom panel). Reproduced with permission from Springer Nature [203]. E. Electrospun networks of RGD-modified dextran methacrylate (DexMA) fibers permit independent tuning of fiber and bulk mechanical properties and revealed that soft fiber stiffness (top panel) promotes cell spreading (cells outlined in magenta) typically associated with stiff matrices by enabling cellular recruitment of individual fibers. Reproduced with permission from Springer Nature [215].

Type I collagen has been extensively used to form defined ECM hydrogels. Collagen type I-based hydrogels have increased our understanding of how ECM in the tumor microenvironment regulates stromal cells [82,165], tumor cell proliferation and migration [166,167], and the collective behavior of both of these cell types (Figure 3B.i) [135]. These insights would not have been possible with BM-mimicking materials as cells embedded into Matrigel® undergo morphogenetic growth, while they exhibit protrusive migration in collagen gels [168]. The ability to adjust the mechanical and structural properties of collagen hydrogels via changes in concentration [169], adjustment of gelation parameters (e.g. temperature and pH) [82,169], as well as glycation [170,171] presents another advantage. Importantly, varying the gelation parameters simultaneously alters the fibrillar structure of collagen [172], which impacts the viscoelastic behavior of the resulting gels and enables cell-mediated local strain-stiffening through positive mechanical feedback [82]. The fibrillar structure of collagen can also be controlled via electrospinning [173], by applying a magnetic field [174], using guided cellular compaction [175], or through mechanical stretching [176] and shear flow [177]. For example, to study the effect of fiber alignment on tumor cells at the invasion front, one study used a strain device to create linearized ECM and showed that fiber alignment, rather than stiffness, promotes directional migration of breast cancer cells (Figure 3B.ii) [178].

While single ECM components such as collagen provide invaluable tools, natural ECM usually comprises many different ECM proteins that are biochemically and biophysically distinct. The compositional, mechanical, and architectural complexity introduced by these multiple ECM components synergistically regulate cellular behavior and are challenging to recapitulate in more defined ECM cultures. To evaluate how the ECM of specific stromal cell types affects tumorigenesis under conditions mimicking the compositional and mechanical complexity of native ECM, decellularized cell-derived ECMs are increasingly used. For example, CAF-derived ECMs retain desmoplastic features in vitro and drive cancer proliferation and drug resistance relative to ECMs derived from normal fibroblasts (Figure 3C) [179–182]. Similar approaches have also revealed that ECM changes drive tumorigenesis by altering the behavior of tumor and immune cells in the context of obesity, a condition known to be epidemiologically associated with increased risk and worse prognosis of cancer patients [183,184]. In the context of bone metastasis, decellularized ECMs produced by osteoblasts can mimic aspects of mineralized collagen, which is the basic building block of bone ECM and regulates the behavior of several bone-tropic cancer cell types including prostate and breast cancer cells [185,186]. To retain the compositional and architectural properties of native ECM even more fully, whole tissues or organs can be decellularized [67,187]. Indeed, this approach has been used for regenerative medicine applications [188], proteomic mapping of ECM composition [67,189], encapsulation of organoid cultures [190,191], and also offers promise for tumor engineering applications. For example, colorectal cancer cells seeded in decellularized ECM from liver and lungs can be used to model organ specific metastasis [190]. However, it is important to note that decellularization protocols influence which ECM components are retained, complicating comparisons across different studies [188]. Furthermore, the complexity of these models makes it difficult to isolate how specific compositional or physical properties of the ECM influence cancer behaviors.

3.2.2. Synthetic and composite biomaterials

Synthetic materials offer promising alternatives to natural biomaterials given their highly defined nature, reproducibility, and readily adjustable structural and mechanical properties. In particular, polyacrylamide (PA) hydrogels have been widely used to study tumorigenesis as a function of ECM stiffness as they can be designed to exhibit a wide range of stiffnesses while maintaining constant surface chemistry [192,193]. However, PA gels cannot be used for 3D culture studies due to cytotoxicity of their monomers and the polymerization procedure itself [188]. Polyethylene glycol (PEG) gels help circumvent this limitation, but these materials need to be covalently modified with adhesion ligands such as RGD found in fibronectin [194] or GFOGER in collagen [195] and/or MMP-degradable moieties [196] to enable cellular interactions reminiscent of native ECM. Alternatively, PEG hydrogels can be cross-linked with natural proteins such as HA to readily incorporate adhesion and degradation sites [197,198]. These modifications enable studies investigating the roles of ECM stiffness, cell adhesion, and migration [197–200].

Despite the highly versatile functions of synthetic ECM models, one major limitation is that synthetic hydrogels composed of only PA and PEG are typically linearly elastic and thus, do not capture the stress-relaxation properties of native viscoelastic interstitial ECM [201]. Nevertheless, it should be mentioned that it is possible to create viscoelastic PA gels by entrapping a linear PA solution in an elastic network of PA hydrogels [17]. Another complementary approach is to use PEG hydrogels cross-linked by reversible hydrazone bonds [202]. The dynamic stress-relaxing crosslinks in these systems permit complex cellular functions to occur while retaining the benefits of traditional covalently crosslinked hydrogels. Alternatively, it is possible to combine synthetic approaches with natural biomaterials in composite systems. For example, one system utilized ionically-crosslinked alginate hydrogels coupled with PEG spacers to enable introduction of varied viscoelastic properties independent of the initial elastic modulus and degradation [203]. It was then demonstrated that faster stress relaxation promoted fibroblast cell spreading and proliferation, and osteogenic differentiation of mesenchymal stem cells in this 3D hydrogel (Figure 3D) [203]. Being able to dynamically tune the viscoelastic properties of hydrogels provides additional avenues to study how ECM mechanics affect tumorigenesis.

In addition to hydrogels, synthetic materials such as poly(lactide-co-glycolide) (PLG), polylactic acid (PLA), poly(ester urethane) (PUR), and poly(ε-caprolactone) (PCL) can be fabricated using different strategies including phase separation, gas-foaming/particulate leaching and electrospinning to create porous scaffolds that are biocompatible and amenable to large-scale use [204–208]. Indeed, cancer cells seeded into porous PLG scaffolds formed tissues that mimic tumor histological features [209]. Furthermore, limited oxygen and nutrient transport in these engineered tumors led to increased malignancy compared to the same cells cultured in conventional 2D culture or 3D Matrigel [209]. The biophysical properties of PLG, PCL, and PUR scaffolds can be advantageous as they can achieve stiffnesses in the MPa range while providing substrates for mineralization that mimic the inorganic composition of skeletal ECM at bone metastatic sites [210–212]. In addition, implantation of microporous PCL scaffolds into mice promotes infiltration of circulating tumor and immune cells, and therefore creates a metastatic niche that can be sampled via biopsy to study disease progression and treatment response [213]. On the other hand, electrospun PCL fibrillar scaffolds can recapitulate aligned ECM substrates and have demonstrated that ECM alignment promotes epithelial-mesenchymal transition of breast cancer cells that is relevant to tumor invasion [214]. Interestingly, another study using electrospun dextran methacrylate (DexMA) fibers to independently tune individual fiber and bulk matrix mechanical properties indicated that softer fibers induce cell spreading and proliferation due to increased fiber recruitment and focal adhesion formation (Figure 3E) [215]. This finding was in contrast to previous observations using non-fibrous hydrogels and highlighted the importance of biomaterial structure in regulating cell behaviors. Despite their many advantages, synthetic scaffolds cannot be remodeled by cells and are subject to polymer degradation, which in the case of PGA and PLA result in the release of acidic byproducts that can independently influence cell behavior [216]. Collectively, biomaterials scaffolds can be tailored to recapitulate the compositional, structural, and mechanical properties of tumor-relevant ECM to study its impact on tumorigenesis, but careful consideration should be given to their choice of materials and fabrication method.

3.3. Microfabricated models to study transport considerations

3.3.1. Models to generate gradients and interstitial pressure

To understand how soluble factor gradients influence tumor progression, model systems are necessary to precisely control soluble factor delivery, interstitial flow, and pressure (Figure 4A). Microfluidic platforms made of polydimethylsiloxane (PDMS) are most commonly utilized to model physiological transport mechanics and generate controlled chemical gradients in a source and sink configuration by allowing modulation of geometries, materials, and inputs [217,218]. Such gradient generating microfluidics have enabled studies of cancer cell invasion and chemotaxis in response to exogenous or cell-generated chemokine gradients in both 2D and 3D contexts [219–223]. More recently, microfluidic models have been used to study the effects of metabolic limitations in tumors, including hypoxia and nutrient/waste gradients where nutrients are introduced in a single channel and gradients are generated intrinsically by cells seeded adjacent to the channel [224,225]. Such models demonstrated that tumor cells alter their metabolism and gene expression signatures in response to nutrient and oxygen gradients [224,225]. Recent work has utilized alternative methods to microfluidics, such as an O₂-controllable hydrogel that is able to generate hypoxic gradients without the need for complex microfabrication and flow setups [226,227]. Other innovative approaches to study the effect of chemokine or metabolic gradients on cancer include work from the McGuigan lab, which has pioneered a platform known as TRACER. In this work, cells are seeded onto a bio-composite strip that is then rolled onto an oxygen impermeable core, generating a naturally formed hypoxic gradient (Figure 4B) [228]. After culture, the strip can be unrolled to enable selective characterization of cell phenotype in specific regions of the culture as a function of hypoxia [228]. In addition to metabolic effects, this platform can be employed to characterize the effects of paracrine and non-soluble signaling between tumor and stromal cells [229,230].

Figure 4: Microfabricated models to study transport considerations.

A. Schematic demonstrating transport considerations in the tumor microenvironment, including gradients of secreted factors, nutrients and oxygen, and elevated interstitial flow and pressure. Microfabricated in vitro models can isolate these aspects to determine their contribution to tumorigenesis. Image was created with BioRender.com. B. A TRACER model consisting of a sheet of cells rolled onto an oxygen impermeable membrane creates a naturally derived oxygen and nutrient gradient enabling spatial mapping of cell metabolism and phenotype. Reproduced with permission from Springer Nature [228]. C. An in vitro 3D culture model consisting of a Boyden chamber containing a cell-laden hydrogel was used to mimic the effects of interstitial flow in conjunction with chemokine signaling on tumor cell migration through a 3D matrix. Reproduced with permission from Elsevier [237]. D. A PDMS-based microfluidic vascularized model composed of an endothelial cell-coated channel (green), a central hydrogel channel, and a tumor cell channel (red) was used to investigate how endothelial barrier function influences tumor cell intravasation. Reproduced with permission from the National Academy of Sciences [244]. E. Endothelial cell-coated microfluidic channels with and without pericyte stabilization can also be formed directly within a collagen bulk preventing cell-contact with PDMS surfaces that may cause artefacts. Reproduced with permission from Springer Nature [252].

Similar to models that enable studies in the presence of soluble factor gradients, microfluidic platforms have been designed to precisely control fluid flow regimes and geometry in 3D cultures and investigate their effects on tumor cell behavior and malignancy. To study interstitial fluid flow, microfluidic devices often contain a 3D biomaterial region that is seeded with cells, where input channels or ports can then introduce flow across the biomaterial region through pressure differentials or gravity-driven flow [118,231–233]. In one example, a PDMS microfluidic device consists of discrete cell-hydrogel compartments that are connected to microfluidic fluid channels, allowing for distinct spatial organization of different cell types and allowing for the investigation of interstitial flow on cellular communication [234]. Other examples include a modified Boyden chamber, in which cells embedded in a 3D hydrogel experience gravity driven-flow allowing for a relatively simple and reproducible alternative to investigate invasion due to interstitial flow (Figure 4C) [235–237]. Studies using these devices revealed that interstitial flow not only directly modulates the behavior of stromal cells, but also the distribution of growth-factors through MMP-mediated release of ECM-bound molecules to synergistically enhance tumor cell invasion [119]. Devices to study the effect of interstitial pressure can be created utilizing micromolded hydrogels that lack an output port and lead to accumulation of pressure at an ECM/liquid interface [115,238]. Such models suggested that increased pressure enhances tumor cell invasion.

3.3.2. Engineered models of the vasculature

Although the above models have been crucial to understanding the individual contributions of transport considerations in the tumor microenvironment, they lack functional vasculature. Tissue-engineered microvascular models provide platforms to investigate tumor-endothelial cell interactions in the context of transport phenomena [239]. Several technologies utilize PDMS-based microfluidic devices consisting of three parallel channels where the center channel can be filled with a hydrogel component such as collagen or fibrin to allow cell-matrix interactions. Endothelial cells can then be introduced into the side channel(s) to form an endothelial barrier at the matrix interface, or within the hydrogel to self-assemble into a vascular network [240,241]. These devices have been used to study the influence of growth factors and fluid forces in angiogenesis, finding that VEGF gradients and flow direction can influence angiogenic sprouting [240,242,243]. Tumor and other stromal cells could also be introduced into the hydrogel or perfused through the vascular network. This approach allows for the study of tumor cell intra- and extravasation [244–246] and has identified that macrophage-induced endothelial barrier impairment enhances tumor cell intravasation (Figure 4D) [244]. More recently, this design has been modified to incorporate a stromal region in between the tumor and endothelial regions, allowing for simultaneous investigation of cancer cell invasion into adjacent stroma and reciprocal tumor-vascular signaling under multiple ECM conditions [247,248]. One limitation of devices fabricated by soft-lithography is that cells may contact the PDMS surface, resulting in undesirable interactions at the device boundaries. Using a needle to form channels in a hydrogel for subsequent generation of a fully endothelialized lumen can help remedy this challenge [249,250]. However, as with the previous method, the architectures that can be obtained are limited to single tubes. Additionally, the size of endothelial lumens in these platforms is typically above a diameter of 100 μm due to device fabrication and stability constraints, whereas the average capillary diameter is an order of magnitude smaller [251]. By using a series of prefabricated molds or sacrificial scaffolds, more complex vascular architectures that include branched features reminiscent of physiological capillaries can be achieved (Figure 4E) [252–254].

The vascular platforms described in this section are particularly well suited to investigating how vessel permeability, angiogenic sprouting, and cellular composition can affect the delivery of nutrients and growth factors [255,256]. In addition, reciprocal tumor-endothelial interactions could be studied in the context of paracrine signaling, drug testing, and tumor metabolism [36,257,258]. The nature of these in vitro vascular platforms facilitates dynamic imaging approaches and has allowed visualization of both angiogenic processes and intra- and extravasation kinetics [242,244,259]. However, endothelial cells cultured in vitro require specialized media with growth factor supplementation to survive and self-assemble into networks. To this end, recent advancements have been made in genetically modifying endothelial cells to transiently reactivate ETS variant transcription factor 2 (ETV2), allowing them to self-organize into adaptable, 3D lumenized vascular networks without the use of exogenous growth factors that could confound cell-cell signaling studies [260]. By integrating these systems with biochemical gradients and/or interstitial fluid flow/pressure, these engineered vasculature models will play a critical role in delineating the interplay between cell signaling and mass transport in the tumor microenvironment.

3.4. Next generation engineered tumor models

3.4.1. Organ/Body-on-a-chip models

Organ- and body-on-a-chip models recapitulate organ level functions or even systemic interactions between multiple organs, enabling more detailed analysis of cancer as a multistage and multi-organ process [261]. The majority of organ-on-a-chip models are manufactured in a similar manner as the microfluidic systems discussed in the previous section, but their design includes more specialized components, biomaterials, and cell populations. For example, organ-on-a-chip models have been developed to recapitulate the blood brain barrier, liver, and the intestine [262–264], and have enabled studies of primary tumor growth in breast tissue, immune cell-mediated tumor cell migration, and tumor cell dormancy in liver [265–267]. Importantly, organ-on-a-chip systems enable investigations into how the cellular, physical, and structural heterogeneity of specific target organs regulate processes relevant to metastasis [268]. For example, incorporating endothelial cells, bone marrow stem cells, and mineralized bone matrix into a microfluidic flow chamber to investigate bone-specific metastatic colonization has demonstrated that interstitial flow promotes the formation of stable vasculature and mediates metastatic potential within the skeleton (Figure 5A) [269]. Organ-on-a-chip systems can also reproduce the unique mechanical function and structural organization of lung tissue to study lung cancer and metastasis [270]. One such model is composed of two microchannels separated by a thin, porous, and flexible PDMS membrane that is seeded with alveolar epithelial cells and pulmonary microvascular endothelial cells on opposite sides and stretched using actuated vacuum chambers to reproduce the mechanical effects of breathing [270]. Indeed, this setup recapitulates central functions of lung tissue and revealed that cyclic mechanical strain involved in breathing increased therapeutic resistance and dormancy characteristics of tumor cells [271]. In both of these examples, organ-on-a-chip systems revealed how the specialized biophysical function and cellular heterogeneity of metastasis-prone tissues regulate organ-specific metastasis, insights that would not be possible with traditional in vitro or in vivo models.

Figure 5: Organ/body-on-a-chip systems and 3d printed models.

A. Organ-on-a-chip models have been used to study bone-specific metastatic colonization by microfluidic integration of mineralized bone matrix, bone marrow stem cells, and endothelial cells. This device also enabled interstitial flow generation, revealing that flow increased cancer cell proliferation. Reproduced with permission from The National Academy of Sciences [269]. B. By integrating multiple organ-on-a-chip systems, a body-on-a-chip platform was used to measure anti-tumor efficacy and cardiotoxicity, demonstrating that the treatment responses of Ewing Sarcoma tumors and heart muscle to Linsitinib was closer to clinical trial results in this integrated system compared to treatment in isolated tissues. Reproduced with permission from The Royal Society of Chemistry [273]. C. Additive 3D bioprinting enables the fabrication of tissue scale constructs with complicated structures in conjunction with a temperature sensitive hydrogel support bath. The feasibility of this approach was demonstrated by printing, (i) a heart valve and (ii) a subregion of vascularized cardiac tissue. Reproduced with permission from The American Association for the Advancement of Science [282]. D. Negative-space bioprinting involves casting a material such as agarose around a 3D printed sacrificial template, which can generate complex architectures, for example, in vascular beds. Reproduced with permission from Springer Nature [287].

Body-on-a-chip platforms are even more complex and integrate cell types from multiple tissues or multiple organ-on-a-chip models to study how tissue-tissue interactions influence tumor-mediated systemic effects or anti-cancer drug toxicity [272]. Incorporating tumor cells, endothelial cells, and cardiomyocytes in microfluidic models, for example, permits simultaneous assessment of anti-tumor drug efficacy as well as cardiac safety [273,274]. Furthermore, integrating a bone Ewing sarcoma model and human cardiac tissue using microfluidics revealed that the drug Linsitinib had limited tumor response and reduced cardiotoxicity in the integrated system compared to either tissue type alone [273]. Importantly, these results were reminiscent of clinical trial results, suggesting that models integrating multiple tissue types could better predict clinical outcomes (Figure 5B).

In summary, organ- and body-on-a-chip systems represent the next generation of in vitro models that not only effectively capture the multiscale complexity of the tumor microenvironment, but also enable systems-level understanding of how cancer influences other organs and how these changes (or preconditions such as obesity or smoking), in turn, affect disease progression [272]. Combining these technologies with induced pluripotent stem cells (iPSCs) provides opportunities for experiments with human, organ-specific cells from a sustainable source [262,273–275] that could help address challenges associated with primary cells. Despite these advantages, organ-on-a-chip systems can be more difficult to implement and typically have low throughput compared to traditional cell culture. The development of automated technologies to fluidically integrate and operate large numbers of organ-on-a-chip platforms simultaneously can help circumvent this limitation [276,277]. Furthermore, many body-on-a-chip technologies incorporate relatively simplistic organ compartments. Given their modular nature, however, they can be readily advanced by designing each compartment based on the parameters outlined above. For example, organ-on-a-chip systems incorporating patient-derived breast cancer organoids could provide an exciting platform for advancing drug discovery in personalized medicine settings [278]. Advancements such as these promise to reduce the use of animal models in preclinical testing of anti-cancer drugs.

3.4.2. 3D bioprinted models

While microfabrication approaches have enabled sophisticated in vitro platforms for studies of cancer, their fabrication requires access to microfabrication equipment. Additionally, current platforms are limited to a planar geometry and are based on plastics, glass, or PDMS, which do not mimic the mechanical and chemical interfaces cells encounter in vivo. Moreover, PDMS can leach free oligomers into culture media that have been shown to affect cell behavior [279]. Tumor models developed using 3D bioprinting technologies provide opportunities to address these limitations. These technologies are derived from traditional 3D printing, a form of additive manufacturing where a material is deposited in a specified pattern layer by layer to form a 3D object [280]. Whereas traditional 3D printing typically uses a plastic material as its “ink”, 3D bioprinting combines biomaterials and cells as “bioinks” in order to manufacture 3D biological constructs [280]. Compared to other models discussed in this review, the increased resolution and thickness in the z-direction confers an advantage over traditional microfluidics and organ-on-a-chip models, which are relatively thin and planar. Moreover, the ability to 3D print different biomaterials and cells in discrete locations drastically improves the level of control over architecture and spatial organization of cells and matrix to better represent the tumor microenvironment.

A variety of 3D bioprinting techniques exists to generate in vitro models. Additive printing is a common method, where a bioink consisting of matrix material and/or cells is forced through a nozzle and deposited in discrete layers to build up a 3D structure layer by layer. The bioink is either extruded as a continuous stream (extrusion-based) or deposited as droplets in close contact (inkjet-based) and solidified through thermally-induced polymerization or post-print by chemical cross-linking [281,282]. Initially, printing soft biomaterials in complex shapes required the use of stiff support structures, which can now be circumvented by using of a support bath. This support bath is composed of a temperature sensitive hydrogel that provides mechanical support for 3D printed in situ polymerizing matrix materials such as collagen, fibrin, or alginate at low temperatures [282,283], but releases the fully printed structure when heated to 37°C (Figure 5C). A similar method utilizing a hydrogel support bath allows printing of multicellular spheroids that self-assemble into high-cell density microtissues with spatially controlled cell ratios [284]. However, extrusion and inkjet methods are currently limited in their ability to print microscale structures due to nozzle size limitations. Light-based methods such as digital light processing (DLP) and multiphoton scanning can photo-crosslink biomaterials and generate scaffolds with spatial resolutions of less than 10 μm [285]. In these methods, ultraviolet (UV) light in conjunction with a photomask or point-scanning lasers are used to photopolymerize a structure within a bulk material. A drawback of light-based methods is that they are limited to photopolymerizable materials and are unable to directly print cells due to phototoxicity and DNA damage from exposure to UV light or lasers [280]. Additionally, achieving high resolution patterning often requires expensive and complex microscopy equipment with multiphoton capabilities. To this end, a method that incorporates natural and synthetic food dyes as photoabsorbers in a photopolymerizable hydrogel has enabled projection stereolithography to create intricate vascular architectures without the use of multiphoton scanning [286].

Another alternative to additive 3D bioprinting is printing sacrificial templates in what is known as negative-space printing. In contrast to additive printing, this technique involves casting a material around a 3D printed object and then removing the object physically or dissolving it, leaving an empty “printed” shape within a bulk material. The sacrificial template is typically printed from a material that can be dissolved at physiological temperatures (37°C) or in aqueous media, such as carbohydrates or Pluronic-F127 [280]. This method has been extensively used for creating engineered vasculature with complex architectures, where the hollow lumens that are left after template removal are seeded with endothelial cells (Figure 5D) [253,287–289]. In comparison to additive bioprinting techniques, sacrificial scaffolds are limited in their material choice and cells can only be introduced after the printed template is removed, rather than being included in the printing material.

3D bioprinting holds great promise in advancing the field of tumor engineering. For example, 3D bioprinted models can recreate the irregular and tortuous architecture of tumor-associated vasculature enabling studies of the resulting consequences on mass transport in tumors [286,290]. Furthermore, the bioinks used in 3D printing could capture tumor specific heterogeneity, for example, by including multiple cell types, biochemical cues, and tumor-specific decellularized ECM in a spatially resolved manner to recapitulate the tumor-stroma interactions at the tumor/host interface [291–293]. Despite their advantages 3D bioprinted tumor models are not widely adopted in the field yet due to a number of limitations inherent to new technologies, such as the lack of standardized printers and the need for optimization of bioinks and biomaterials used for printing [280]. Additionally, 3D bioprinting is currently restricted to materials that can be polymerized in situ such as collagen, alginate, gelatin, and photopolymerizable polymers. As the technology continues to develop, 3D bioprinting will enable widespread fabrication of reproducible and complex in vitro models.

4. Future directions and conclusions

Tumor engineering strategies have increased understanding of the dynamic steps underlying cancer development, progression, and therapy response, but many opportunities exist to further advance the field and translate findings into improved therapies (Figure 6). For example, integrating the described models with advanced imaging methods will enable monitoring cell behavior in 3D without introducing artefacts. Multiphoton or single-photon autofluorescence imaging techniques in particular allow dynamic imaging without the use of fixation, labels, or dyes, and has been applied for rapid, non-destructive imaging of patient-derived cancer organoids [294,295]. In addition, optical coherence tomography (OCT) enables label-free imaging of cancer spheroids embedded into fibrillar matrices based on temporal sampling of back-scattered coherent light to correlate tumor-induced matrix deformation and tumor invasion [135,296]. To obtain mechanistic insights into how microenvironmental factors influence intratumoral heterogeneity multi-omic and sequencing technologies could be applied. One particularly promising technique is deterministic barcoding in tissue for spatial omics sequencing (DBiT-seq), which allows spatial barcoding of molecules in tissues for simultaneous proteomics, imaging, and genomics [297]. In contrast to traditional sequencing methods, DBiT-seq retains spatial information to correlate molecular profiles with tissue context [298]. Finally, computational methods can complement in vitro tumor models, especially by extrapolating the observed phenomena to timescales relevant to tumor development in patients. For example, computational approaches can predict tumor growth, invasion, and treatment outcomes in response to metabolic heterogeneity, with timescales ranging from days to years [299]. Such in silico models can be further enhanced by integrating microenvironmental information including spatial organization of cells, cell deformation, and changes in chemical gradients [300]. However, iterative coupling between experiments and computational modeling is required to increase the relevance of both approaches.

Figure 6: Opportunities to improve engineered tumor models.

The individual tumor models presented in this review vary in their complexity and span a range of engineering approaches to fabricate them. Integrating multiple of these different models and approaches within one integrated system represents an opportunity to develop the next generation of tumor models. Additional opportunities can be derived from integrating technologies from other disciplines such as imaging, advanced omics, and computational approaches as well as clinical/patient-oriented aspects such as using these models for cancer drug screening in precision medicine settings and improving scale-up manufacturing. Image was created with BioRender.com.

Given their ability to mimic aspects of cancer in a patient-specific manner, engineered tumor microenvironments have the potential to advance precision medicine applications. Indeed, incorporation of patient-derived or patient-specific materials such as organoids and patient-derived xenografts (PDXs) would enhance the translational value of engineered tumor models as they better capture the cellular heterogeneity and evolution of cancer seen in patients. For example, sequencing of organoid lines generated from advanced prostate cancer specimens revealed a landscape of mutations that were acquired during treatments and metastasis [144]. As these mutations could affect both cell-microenvironment interactions and therapy response, drug screening of such cells should be performed under conditions in which both parameters can be controlled independently. Engineered tumor microenvironments enable such studies and make it possible to more accurately predict therapy in the context of precision medicine [278]. Moving forward, integration of microfluidics, micropatterning and bioprinting technologies will create high throughput screening platforms that further advance personalized drug screening and ultimately improve patient prognosis.

In summary, cancer constitutes a multifaceted disease in which tumor cells’ interactions with the microenvironment are just as important as their molecular abnormalities. In vitro tumor models that provide controlled and physiologically relevant conditions to isolate the contribution of cellular and acellular microenvironmental factors to cancer progression continue to advance the field by defining the underlying mechanisms. By integrating them with advanced imaging, multi-omics, and computational approaches, engineered tumor models have the potential to yield unprecedented insights that will inform novel approaches to cancer therapy and precision medicine ultimately improving patient outcomes.