A Tissue Engineering Approach to Metastatic Colon Cancer

Summary

Colon cancer remains the third most common cause of cancer in the US, and the third most common cause of cancer death. Worldwide, colon cancer is the second most common cause of cancer and cancer deaths. At least 25% of patients still present with metastatic disease, and at least 25-30% will develop metastatic colon cancer in the course of their disease. While chemotherapy and surgery remain the mainstay of treatment, understanding the fundamental cellular niche and mechanical properties that result in metastases would facilitate both prevention and cure. Advances in biomaterials, novel 3D primary human cells, modelling using microfluidics and the ability to alter the physical environment, now offers a unique opportunity to develop and test impactful treatment.

Graphical Abstract

Bioengineering; Cancer; Disease

Metastatic Colorectal Cancer: Clinical Significance and Pathogenesis

Colorectal cancer (CRC) continues to grow in prevalence throughout the world and now accounts for nearly one in every ten cancers worldwide, constituting the second most common malignancy worldwide and the third most common in the United States (Bray et al., 2018) (Siegel et al., 2020). For colorectal cancer, stage at diagnosis is the most important predictor of survival with 5-year relative survival rates ranging from 90% in patients with local disease (stage I, II) to 14% for those diagnosed with distant (stage IV) disease (Bray et al., 2018; Siegel et al., 2020). Despite the plethora of research and the growing number of therapeutics, more than 700,000 deaths occur from CRC annually (Lozano et al., 2012).

In 1990, Fearon and Vogelstein outlined the earliest model of colorectal tumorigenesis, the adenoma-carcinoma sequence of sporadic CRC, describing an accumulation of mutations in chromosome 5q or loss of FAP followed by DNA hypomethylation, KRAS mutation, loss of DCC, and loss of TP53 (Fearon and Vogelstein, 1990) (Figure 1A). Through this sequence, normal intestinal epithelium evolves to adenoma and eventually to an invasive cancer. In the years since these initial findings, their model has been further refined with the discovery that most adenomas harbor constitutive Wnt signaling activation and accumulation of activating mutations in the EGF receptor signaling pathway and subsequent inactivating mutations in the TGF-β and P53 pathways (Fearon, 2011). Recently, early P53 loss has been shown to promote progression of CRC (Fumagalli et al., 2017).

Figure 1.

Colorectal Cancer Progression

(A) Normal colonic epithelium acquires an initial somatic mutation, often in the APC gene, leading to accelerated epithelial proliferation. Hyperproliferation of the epithelium leads to the formation of adenomas that eventually acquire additional mutations including those in KRAS and SMAD4. If an adenoma is not removed, it may undergo malignant transformation by acquiring further mutations, including P53 mutations, and invade through the submucosa to become an invasive cancer (fourth panel from left). Once invasion has occurred, the tumor cells have direct interaction with various elements of the microenvironment, including immune cells, blood vessels, and lymphatics. Tumor cells may then metastasize to distant sites, either hematogenously to the liver, lungs and other distant sites or via the lymphatics to local or regional lymph nodes (Right most panel). (B) In the metastatic process, in order to access the lymphatics, the cells must first invade through the basement membrane. The tumorigenic cells then intravasate into the blood vessel or the lymphatics, and enter the circulation. These cells then exit the circulation via extravasation, adhere, and will grow as microfoci as they colonize their new microenvironment.

Despite our advancement in understanding the early stages of CRC, the medical and scientific community has been unable to decipher a genetic underpinning for the transition from localized to metastatic disease. Using an in vivo orthotopic organoid model utilizing CRISPR-Cas9 genome editing to generate mutations in the Wnt, EGFR, TP53, and TGF-ß, Fumagalli and colleagues determined that stem cell niche-independent growth through the accumulation of mutations in all of these pathways was necessary for tumor cells to metastasize (Fumagalli et al., 2017).

Most historical tumor models rely on the view that tumor physiology can be understood through the analysis of a single clone of cells cultured on plates or implanted in vivo (Ahmed et al., 2013). These reductionist models served us well in helping to understand the basic underpinnings of colorectal cancer but do not provide an accurate representation of the complex dynamics of malignancy. As a whole, they fail to recapitulate the complex tumor microenvironment composed of epithelial cells, stromal cells, immune cells, and other components of the extracellular matrix (ECM) (Egeblad et al., 2010; Holzel et al., 2013). If we are to progress in our understanding of the physiology of the complex tumor microenvironment, the field requires reductionist models that can be combined to study the interplay between various components.

Metastasis: A Complicated Multistep Process

Metastasis is a series of progressive steps resulting in the formation of secondary tumors in other organs (Martin et al., 2013) (Figure 1B).Primary tumors of 1 cm³ (approximately 10⁹ cells) can disperse 1 million cancer cells into the circulatory system daily with less than 0.1% of these cells developing into metastatic disease In the late nineteenth century, Stephan Paget proposed that metastatic cells of a certain type tend to distribute to themselves to specific organs (Paget, 1989). Since his initial hypothesis, the scientific community has validated these early necropsy findings and coined the term “organotropism” to define the organ-specific predisposition (Nguyen et al., 2009).. Metastatic organotropism is believed to be governed by a variety of factors including the tumor microenvironment, local cytokine and chemokine profile, tissue biomechanics, physiology, and availability of lymphatics or blood vessels (Chambers et al., 2002; Gupta and Massague, 2006; Pretzsch et al., 2019).

“Invasion” (Figure 1B), the first step in the metastatic cascade occurs via infiltration of the tumor invasive front into the local tissues. It is during this stage that the widely discussed epithelial to mesenchymal transition (EMT), which is a biological program consisting of loss of epithelial markers, increases in mesenchymal gene expression, loss of adherence junctions, and motility (Cao et al., 2015; Cheung and Ewald, 2016). Once the tumor has invaded into local tissue, they are able to gain close proximity to tissue blood vessels and lymphatic tissue (Bockhorn et al., 2007; Fares et al., 2020). For “intravasation” (Figure 1B), the transendothelial migration of cancer cells into vessels, to occur, the invasive front is drawn toward blood or lymphatic vessels via chemotactic signals at which point there is a reliance on tumor macrophages and TGF-β signaling via EMT to facilitate intravasation into vessels (Denais et al., 2016; van Zijl et al., 2011).

Once cancer cells have entered the “hematogenous or lymphatic circulation” (Figure 1B), they are generally transported in clusters supported by neutrophils. These neutrophils function to decrease the activation of other elements of the immune system and platelet coating allowing for immune evasion and protection against physical stress (Fares et al., 2020; Gay and Felding-Habermann, 2011).

After traveling through circulation, the vessels must escape to reach a distant site through the process of “extravasation” (Figure 1B). Many studies have shown that vascular flow plays a major role in this process with the decreased flow in capillaries leading to circulating tumor cells becoming lodged and shear forces disrupting these cells leading to immune mediated extravasation (Follain et al., 2018; Goetz, 2018; Headley et al., 2016). Prior to the arrival of circulating tumor cells to distant sites, the primary tumor has already been preparing pre-metastatic niches through the secretion of exosomes and other factors, facilitating “colonization” (Figure 1B) (Hoshino et al., 2015). On arrival, tumor cells function in this niche generating favorable conditions for their outgrowth (Benedicto et al., 2017).

Current Approaches and Challenges

Biomedical research depends on the combination of in vitro cell culture models and animal models to understand and predict the human response. Animal models are expensive, have ethical issues, and are poor at predicting human responses owing to cross-species discrepancies (Eastwood et al., 2010). In vitro culture methods are important alternatives for discerning biomolecular and disease mechanisms. The culture model choice depends upon the type of research, questions being asked, desired physiology, fidelity representation of the human condition, and expense of culture system.

In Vitro Models of the Metastatic Process

Two-Dimensional and Three-Dimensional Cell Culture

Two-dimensional cultures are often established by growing the cells as a monolayer onto a surface (e.g., plastic, glass, membranes, or Transwells) (Figure 2A). Such 2D culture techniques are economical, convenient, and simple and provide preliminary screening, cell function, and cellular products. These 2D methods result in cultures that do not provide an adequate model of what occurs in vivo because they do not create complex interactions among organs and other tissues and fail to incorporate microenvironmental factors (Sensi et al., 2018). Two-dimensional cell cultures can be improved by aligning or organizing multiple cell types though differential plating, creating self-assembling monolayers, and using chemically modified surfaces resulting in more authentic phenotypes that retain functional responses of the human tissue or organs. The 2D culturing techniques can be used in a variety of assays, such as cell proliferation (kinetics of proliferation—well plates), cell clonogenicity (viabilities—well plates), wound healing (migratory capacity of cells cultured on petri dishes), permeability, migration, transendothelial migration, and invasion (Transwells) (Benedicto et al., 2017).

Figure 2.

Propagation of 2D and 3D Cultures

(A) Multiple approaches to 2D culturing. Propagation on cells, plates, wells, and flasks are shown.

(B) Multiple approaches to 3D cultures. Alternative vessels and encapsulation into hydrogels and onto scaffolds facilitate maintenance of 3D relationships in vitro.

Some cells easily grow in suspension, whereas others can be forced to grow in suspension by using non-adherent plates producing a dispersed 3D culture (Table 1). Other organized 3D cultures can be created by plating the cells in a hydrogel (e.g., collagen gels, agar, Matrigel, PG matrix) using a hanging droplet technique to establish the cells for development into spheroids or organoids, plating cells onto scaffolds or decellularized matrices, or using tissue slices (Figure 2B) (Crapo et al., 2011; Lu et al., 2018). These tissues may self-organize (especially organoids) or the tissue construct may be “engineered” by using a scaffold to spatially organize cells in the tissue mimic. The cellular diversity of organ microenvironments has been addressed through tissue engineering; knowledge of an organ's morphology, its microbiome or both (Jalili-Firoozinezhad et al., 2019; Shin et al., 2019); and layering various cell types together by using chemical modification, differential plating, or allowing the cells to self-organize to mimic the organ's function (Jalili-Firoozinezhad et al., 2019).

Table 1.

2D versus 3D Cultures

2D versus 3D	Advantages	Disadvantages
2D	• Well established • Economical, commercially available tests and media • Easy, convenient • High throughput capacity, mass screening • Repeatability and accessibility • Homogeneous growth and proliferation • Good for epithelial systems • Air-liquid interfaces • Difficult to model tumor cells • Unlimited access to oxygen, nutrients, metabolites, and signaling molecules • Changes gene expression, mRNA splicing, topology, and cell biochemistry	• Stiff substrate—plastic, glass • Lack cell-cell and cell-matrix interactions • Unable to assess tissue penetration • Unable to mimic in vivo microenvironment • Lack extracellular matrices • Lack natural tissue structures • Relies on adherence to flat surface • Do not control cell shape affecting bioactivity • Decrease in transcriptional genes • Morphological changes • Changes polarity • Changes in method of division
Customized 2D (e.g., microwells, micropillars, cell-adhesive islands, sandwich culture)	• Induce apical-basal polarity • Improve cell shape control • Improve morphology and function • Study uptake and efflux transport • Benefits to pharmacokinetics	• Induced polarity may alter function • Affects spreading, migrating, and sensing soluble factors and cues
3D	• Mimic the in vivo microenvironment • Cell-cell and cell-matrix interactions • Substrates that mimic natural extracellular matrix • Reduce overall cost and for pre-clinical in vivo testing • More reliable preclinical evaluation • Improve/effects protein expression, proliferation, differentiation, and metabolism • Improve architectural tissue structures • Better for studying cell migration • Suspension, in a gel • Preserves morphology • Diverse phenotype and polarity • Variable access to oxygen, nutrients, metabolites, and signaling molecules • Expression of genes, splicing, topology, and cell biochemistry	• More expensive (coated plates) • More complicated to establish culture models • Low throughput capacity • Overcome 2D limitations • Cell life expectancy can be short • Fewer commercially available tests

The generation of epithelial organoids as a 3D culture system to quickly grow and expand the colonic epithelium obtained from individuals with colon cancer was a breakthrough in in vitro modeling of the colon and facilitated functional studies of human intestinal epithelial cells (VanDussen et al., 2015). However, in both health and disease, the epithelia do not act in isolation and interactively communicate with microenvironmental factors, including the stromal compartment and ECM to maintain epithelial function (D'Angelo et al., 2020a; Neal et al., 2018). In the past, the ECM had been perceived to be passive. Now, tumor microenvironment elements, including the structural proteins and mechanical properties, have been shown to be interactive and crucially impact cancer progression and treatment responses (Romero-Lopez et al., 2017). For example, Baker et al. showed matrix stiffness increases in colorectal cancer tumors owing to overexpression of active LOX (Baker et al., 2013). Therefore, understanding ECM compositional changes in cancer and recapitulation of the mechanical properties of the native 3D cellular microenvironment are vital to reproduce the complexity of native tissues.

Ex Vivo Models: Biomaterials

For decades, cancer research has relied upon various in vitro models to investigate the mechanisms underlying development and progression of colorectal cancer. Among in vitro models, traditional monolayer cell cultures have been extensively used for research. Despite the advantages of these models—the cells are easily grown and maintained and biochemical assays are easily performed—traditional 2D cultures are often structurally disorganized and do not recapitulate the structural and pathophysiological complexity of the tumors. This significant inconsistency between monolayer cancer cell cultures and native tumors led to an essential need for more relevant in vitro tumor models to improve data for subsequent in vivo studies and clinical applications.

The need for more comprehensive in vitro models has led to a rapid growth in disease modeling with more structured, organized, and realistic tissue and organ models. Such models use both natural and synthetic biomaterials to engineer the biological and mechanical features of the native tumors. Natural biomaterials such as type I collagen (a natural matrix), a major component of the stromal matrix in colorectal cancer (Hilska et al., 1998), have been used individually or in various combinations to provide a supportive matrix for cell-cell and cell-ECM interactions, migration, and invasion in cancer (Figure 3). As an example, co-culture of LS174T colon cancer spheroids with normal colon fibroblast cells in a 3D collagen matrix reduced the mitotic potential of LS174T cells and resulted in contraction of collagen fibers around the fibroblasts, as is observed in the human colon adenocarcinoma. On the other hand, cancer-associated fibroblasts derived from colon adenocarcinoma promoted cancer cell invasion into the collagen matrix (Dolznig et al., 2011).

Figure 3.

Schematic Representation of 3D Matrices used in Cancer Biology Research

(A) Cellular and extracellular matrices are demonstrated with ultimate applications.

(B) Cartoon of epithelial cells with surrounding cellular and extracellular matrices of the microenvironment. Components of synthetic and commercially available extracellular matrices are listed.

However, the use of natural biomaterials in disease modeling includes some shortcomings such as inconsistent cross-linking, batch-to-batch variations, dependence on the tissue of origin, and lack of flexibility in adjustment of mechanical properties. On the other hand, synthetic biomaterials provide the biocompatibility of the natural materials along with an inert and defined structure with a wide range of mechanical properties to study the effect of each individual factor in disease progression. Chen et al. developed a thermosensitive poly (N-isopropylacrylamide)-based hydrogel as a drug delivery strategy for controlled release of doxorubicin toward CT-26 mouse colon carcinoma cells (Chen et al., 2018). In another study, a temperature-sensitive PEG-PCL-PEG triblock copolymer was synthesized to load 5-FU to improve the therapeutic effects of intraperitoneal chemotherapy (Wang et al., 2010).

Several studies have revealed that, in patient tumors, the ECM has a unique composition and in accordance with the “seed and soil hypothesis” emphasized the tumor microenvironment, including ECM composition, as a main contributor to tumor progression and drug response (Naba et al., 2014) (Table 2). Romero-López et al. highlighted the importance of recapitulating ECM composition as a vital aspect of tissue physiology by comparing ECM from normal human colon and colon tumor that had metastasized to liver; they found significant differences in protein composition, stiffness, and vascular network formation between normal and metastatic colon (Romero-Lopez et al., 2017). Recent introduction of the tissue de-cellularization technique (Ott et al., 2010) (Crapo et al., 2011), by which cells are removed while retaining the ECM structure and composition, is a substantial step forward in the field of tissue engineering and regenerative medicine. The advantage of this model compared with the other natural and synthetic 3D matrices is that the native structural proteins and soluble factors are largely preserved in the decellularized scaffold, resulting in a more comprehensive tissue recapitulation (Pinto et al., 2017). Piccoli et al. developed a technique to decellularize colorectal cancer tissue biopsies and generate a 3D “bioactive” model while maintaining the ultrastructural environment and biological activity of the native tissue (Piccoli et al., 2018). In another study, patient-derived tissue-specific decellularized ECM was introduced as a 3D culture system that recapitulated colorectal cancer liver metastasis (CRLM) to study CRLM progression and response to chemotherapy (D'Angelo et al., 2020b). Overall, the best model to recapitulate human CRC metastases remains patient-derived scaffolds. Our team has previously used decellularized human colonic tissue to model the process of intravasation and has discovered that, during intravasation, gene signatures emerge that are unique to the intravasation process (Chen et al., 2016).

Table 2.

Strengths and Weaknesses of Extracellular Matrix versus Decellularized Scaffolds

ECM versus Scaffolds	Types	Advantages	Disadvantages
ECM	• Hydrogels (e.g., Matrigel, PGmatrix) • Collagen or fibronectin • Polyacrylamide • Alginate • Soft agar	• Cell encapsulation • Tissue-like water content • Tunable biochemical and mechanical properties • Adding polyethylene glycol with photodegradable compounds to alter properties • Cells are easy to recover • Cells form tissue-like structures • Can study the aggressiveness of cells and metastatic potential	• Micron/nanometer-sized mesh too small for post-fabrication cellularization • Lack micro-topography for controlling cell shape • Gradual degradation • Time consuming • Low repeatability • Difficult to extract cells • Difficult for immunofluorescent staining • May influence structure formation
Scaffold	• Microsphere • Electrospun • Solvent-casting and particle leaching • Sphere-templated technique • Freeze drying • Gas foaming particle leaching • Solid freeform fabrication • Fiber meshes/fiber bonding • Melt molding • Solid freeform fabrication using computer-aided design (CAD), magnetic resonance imaging, computed tomography, stereolithography, fused deposition modeling, and laser sintering • Biopolymer scaffolds • Silk, collagen, laminin, alginate	• Sphere size between 1 and 1,000 μm • Applied easily with syringe • Biocompatibility, bioavailability • Porous or non-porous • Biodegradable or not • Can select polymer type • Can have ultrafine fibers • Provide cellular support • Control pore size • Porogens can help produce a uniform suspension • Can have high porosity and interconnectivity • Fine structures and complex geometries • Various sizes and/or shapes • Compatible with functional tests • Easy to prepare for immunohistochemical analysis	• Porosity control varies depending on method chosen • Leaching of particles due to method chosen • Porogens may be difficult to remove • Toxicity if organic solvents were used in production • Cells flatten and spread like 2D • Scaffold may cause various cellular behaviors • May affect adhesion and growth • Cell observation and extraction are restricted

ECM, extracellular matrix.

Advances in In Vitro Device Techniques

The 2D and 3D culturing techniques can be used in both single organ (Abaci et al., 2015; Esch et al., 2015; Wang et al., 2017) and multi-organ-interconnected microphysiological systems that are known as organ or body on a chip (OOC or BOC) (Abaci et al., 2015; Esch et al., 2016; Mahler et al., 2009; Maschmeyer et al., 2015; Miller and Shuler, 2016; Oleaga et al., 2016; Sin et al., 2004; Sung et al., 2010; Wagner et al., 2013; Wang et al., 2017) (Figure 4). These systems can be used to study metabolism, cell-cell and organ-organ interactions, physiological responses to chemical or biological stimuli, and drug development and toxicity, as well as for tumor metastasis. All in vitro 2D and 3D systems have advantages and limitations. The biological questions, desired physiology, fidelity of the representation of the human condition, and expense of the culture system are important considerations.

Figure 4.

Microfluidic devices used to Investigate Colon Cancer Metastases (Not to Scale)

(A) Configurations of various multi-chamber devices are diagrammed.

(B) Pumped systems can be external or integrated to provide recirculating systems or create a single pass flow.

(C) Examples of microfluidic devices that use rocker platforms to provide bidirectional or unidirectional flow.

Advances in tissue engineering and technology have progressed so that barrier and non-barrier tissues can be incorporated into the devices (Sakolish et al., 2016; Zheng et al., 2016) (Figure 5). Many choices are available for BOC system circulation, (for example, pump devices, a pumpless device that is gravity-induced on a rocker platform (bidirectional) (Abaci et al., 2015) or without rocking (single pass), or a pumpless device on a rocker platform that allows unidirectional flow while mimicking blood recirculation (Esch et al., 2016; Wang and Shuler, 2018). Fluid shear stress and mechanical stress from the laminar flow in OOC and BOC systems provide nutrients, remove waste, promote organ polarity, physical pressure on endothelial cells, and differentiation (Kshitiz et al., 2012; Wu et al., 2020). Cell patterning on the chip is partially controlled by microfluidics, but a researcher can make surface modifications, chip templates, and 3D-printer scaffolds (e.g., hydrogels, biocompatible poly (ethylene glycol) diacrylate) to create complex channels such as vascular networks and specific microenvironments like villi (Sung et al., 2011). Methods have been developed that even combine polyvinyl acetate coating and carbon dioxide laser ablation with continuous cell seeding techniques that enable controlled epithelial-mesenchymal interactions (Li et al., 2015).

Figure 5.

Advances in Microfluidic Devices Combine Current Technologies

By combining tissue engineering, hardware engineering, pharmacology, and computational modeling, one can create more advanced models that are more physiological.

Improvements in monitoring cells in a 3D microfluidic setting have been made using time-lapse imaging microscopy and microsensors (Kane et al., 2019). In particular, it is possible to measure not only biomarkers, especially those in circulation but also functional responses such as electrical activity using microelectrode arrays and cellular force generation using cantilevers or by monitoring movements of labeled micro- or nanoparticles.

Use of 3D Devices for the Study of Metastasis

Development of in vitro tumor models to study the progression of cancer and metastasis depends upon knowledge of tumor microenvironment, tumor cell biology, with realistic models at the cell, tissue, and organ levels, and potential use of BOC-type devices.

Tumor cells respond to biochemical/biophysical signals from surrounding stromal cells and ECM. Stromal components (e.g., fibroblasts, endothelial cells, pericytes, various specialized cells, matrix, and extracellular proteins) help determine biochemical and biophysical properties that affect cellular growth, differentiation, migration, and apoptosis (Discher et al., 2005; Engler et al., 2006). Organ dysfunction can be attributed to stress conditions that signal the fibroblasts to increase plasticity and transform into myofibroblasts, remodel the ECM framework, increasing the amounts of ECM; increase inflammatory cytokines, recruit immune cells, and promote neoangiogenesis, proliferation, invasion, migration, and metastasis (Klingberg et al., 2013; Roy et al., 2020). A hypoxic tumor environment (<8–10 mm Hg) leads to a decrease in the tissue oxygen consumption rate or ATP production rate (Hockel and Vaupel, 2001). Hypoxia-inducible factors mediate adaptive responses to reduce oxygen availability and induce the expression of target genes involved in angiogenesis, cell survival, energy metabolism, differentiation, and invasion. These cellular, physiological and environmental tumor factors need to be considered when developing tumor invasion models.

In sum, numerous models mimicking tissue and organ physiology have been developed, including microfluidic and Transwell systems, organoids, spheroids, hydrogels, perfused cultures, organ- or body-on-a-chip, bioreactors (devices that produce cellular constructs at large scale), and 3D bioprinting (Krishnan et al., 2011). These complex systems provide highly interactive environments, can better mimic native tumor behaviors, can carry the tumor hypoxic or necrotic core, mimic progression, and mimic invasion (Rodrigues et al., 2018). Some of these systems can incorporate barrier and non-barrier tissues, air/liquid interfaces, vascular and/or lymphatic components, and microvascular networks. Establishment of pre-vascular tissues and/or seeding lymphatic endothelial cells within tissue-engineered constructs (Gibot et al., 2017) can enhance colonization and are essential for studies of normal and tumoral mechanisms in 3D systems. Generally, devices that mimic tissue/organ structure can provide a better understanding of the tumor microenvironment and cancer behavior, development, and metastasis than unstructured 2D culture systems. New improvements in technologies offer many options to enhance the capacity of 3D culture systems to better mimic metastasis.

In Vivo Murine Models

Despite advances in vitro modeling capabilities, studying the complex whole-body determinants of cancer metastasis still requires in vivo models that recapitulate the anatomical route, tumor microenvironment, or both in human patients (Figure 6). Our understanding of CRC pathogenesis has been aided by the development of genetically engineered mouse models (GEMMs), which reproduce the molecular and morphological changes that occur during the dysplastic transformation of normal colonic epithelium into carcinoma (Romano et al., 2018; Roper and Hung, 2012). Inducing metastasis in these murine models has been challenging, with a few models able to reliably drive distant metastasis patterns (Burtin et al., 2020). Among these, spatial and temporal control over the expression of Apc, Kras, and Tp53 has been used to drive primary colon tumor formation and distant metastasis, whether using 4-OH-tamoxifen-ethanol enemas in iKAP mice (Apc^Lox/Lox; p53^Lox/Lox; Tet-O-LSL-Kras^G12D; VillinCre^ERT2) or surgical adenoviral Cre administration in Apc^CKO/CKOLSL-G12D; Kras^tm4tyj/+ mice (Chanrion et al., 2014; Haigis et al., 2008; Hung et al., 2010). However, despite their usefulness in studying carcinogenesis, GEMMs tend to be costly, slow, and not necessarily representative of the diverse genetic drivers of clinical metastasis (Francia et al., 2011).

Figure 6.

Murine Models of Colorectal Cancer Metastasis Rely Largely on Introduction of Cancer Cells into Circulation via Injection

Both orthotopic (red) sites and heterotopic sites (green) have been described, with different resulting patterns of metastatic spread (dotted lines).

Implanting tumor cells—whether in the form of cancer cell suspensions or tumor from a patient—into mice has long been an experimental model for studying cancer biology (Bresalier et al., 1987; Fu et al., 1991). To prevent graft rejection, this approach generally requires immunodeficient mice unless murine tumor cells are injected into syngeneic hosts (e.g., CT26 cells into the BALB/c mouse). Heterotopic transplants are particularly tractable, requiring little surgical expertise as they are implanted as a tumor cell suspension injected into the mouse flank. When patient tumor cells obtained at tumor excision or biopsy are implanted into immunodeficient mice, the resulting subcutaneous tumor is called a patient-derived xenograft and has been shown to maintain the histology and genetic profile of the donor over several passages (Ben-David et al., 2017; Brown et al., 2016; Cho et al., 2014; Guenot et al., 2006). In fact, the response of the donor to chemotherapy treatment is recapitulated in the “xenopatient” mice, indicating a promising application for precision therapeutics (Gao et al., 2015; Lazzari et al., 2019; Pauli et al., 2017). However, over passages in vivo, the heterogeneity of the original transplants is gradually lost as the donor tissue stroma is often replaced by murine architecture and the fastest-growing clones tend to take over. These models often fail to develop metastases (Oliveira et al., 2020). The underlying biology that prevents metastasis in heterotopic transplants is likely manifold—limited vascular routes for migrating cells, a lack of colonic microenvironmental stimuli, and minimal exposure to other niche elements (Fumagalli et al., 2017). A recent study using CRC PDX samples so provided evidence for a hybrid epithelial-mesenchymal state in the CRC clusters that seeded most effectively (Mizukoshi et al., 2020). Implantation into a more highly vascularized environment, like beneath the renal capsule, has also not shown significant advantages over subcutaneous injection, suggesting vasculature is not the only factor in driving metastatic spread. In fact, CRC cells grown on decellularized liver and lung scaffolds were shown to retain their tissue-specific tropism when injected into the spleen of mice (Tian et al., 2018), whereas intrasplenic injections are otherwise known to only cause metastases to the liver (Brand et al., 1996). Contrarily, intravenous injection into the tail vein leads to the formation of lung metastases, in accordance with hematogenous flow, although this approach captures only circulation, extravasation, and colonization while excluding the earlier steps occurring in primary tumors, including invasion and intravasation (Mittal et al., 2015). The predictability of these patterns indicates that these methods do not adequately capture the cellular mechanisms like matrix degradation that characterize physiological EMT.

Heterotopic tumors may be used as an intermediate in vivo growth platform prior to orthotopic implantation into the colon of another mouse, with the subcutaneous cell graft harvested from the flank moved to a more native tumor microenvironment. Numerous methods for orthotopic transplants have been described for inoculating tumor cells into the colon, including serosal tunneling, injection, and directly attaching the tumor pieces to the cecal wall (Lin et al., 2015; Wang et al., 2009). More recently, methods for subserosal injections into the cecal wall and intrarectal injections have been developed, both providing efficient tumor formation and metastatic behavior (Lee et al., 2014; Takahashi et al., 2004). These models have been shown to recapitulate genetic differences seen in advanced CRC in humans as well (Rajput et al., 2013; Sasaki et al., 2008). Thus, choice of orthotopic transplant method must be carefully made, particularly considering that slight variations such as injecting tumor in the mesenteric versus antimesenteric cecal wall have been shown to yield different metastatic patterns (Boni et al., 2005).

Although intracecal transplants remain a preferred model of CRC metastasis, recent evidence has suggested that intrarectal injections may provide similar metastatic capabilities, with liver and lung metastases occurring in 35%–100% of mice depending on the CRC cell line (Kishimoto et al., 2013; Moret et al., 2019; Takahashi et al., 2004). Although CRC metastasis was long thought to progress through lymph node invasion prior to liver spread, an interesting insight derived from using an intrarectal injection model showed that hepatic metastases can occur independently and in parallel with lymph node metastases (Enquist et al., 2014). Schölch et al. collected circulating tumor cells with stem cell-like features and performed a second intracecal injection, creating a model that reproducibly forms metastases in the liver and lung in a pattern recapitulating that found in humans (Kochall et al., 2017; Scholch et al., 2016). To increase the metastasis rate, many studies use CRC cell lines that are already metastatic, such as HCT116, HT29, and SW620 (Bu et al., 2018). Conversely, spontaneous models can be generated by combining GEMM with orthotopic transplantation. For example, organoids derived from the VillinCre-ER^T2; APC^fl/fl; KRAS^LSL−G12D; TP53^KO/KO model, which forms spontaneous CRC tumors and reaches the endpoint of localized tumorigenic growth without metastasis, can be orthotopically transplanted again to form a single primary CRC tumor that spontaneously metastasizes to the liver and lung (Fumagalli et al., 2020).

Tracking Metastasis in In Vivo Models

Observing cellular changes in the context of a systemic and dynamic phenomenon like metastasis poses a unique challenge. In essence, most advances in this area have been derived from fluorescent labeling of tumor cells prior to either in vitro assays of metastatic potential like invasion assays and EMT characterization or injection into murine models. Newer 3D culture methods, like organoid cultures and complex tumor-stroma co-culture systems, have begun to more faithfully recapitulate in vivo dynamics, and consequently implantation of these models into living systems has yielded metastatic features that more closely resemble those seen in patients (Devarasetty et al., 2020; Okazawa et al., 2018).

However, studying dynamic changes that occur during metastasis in vivo has remained challenging. Until recently, most studies tracking metastasis relied on imaging at discrete intervals, which required either sacrifice or significant disruption of the murine system at predetermined time points. These methods have generated interesting insights, like the presence of an extravascular migration pathway utilized by melanoma cells metastasizing to the brain (Bentolila et al., 2016; Kienast et al., 2010). In colon cancer models, Tanaka et al. performed intrasplenic injections of fluorescently tagged CRC cells, which were then imaged in the liver at intervals using high-resolution two photon laser scanning microscopy (Tanaka et al., 2012). Using this methodology, they were able to visualize changes in liver histology and stroma during metastatic colonization; however, this study was limited by the need to surgically exteriorize the hepatic lobes for repeated imaging.

Another popular method uses luciferase-expressing tumor cells, which emit light that can be captured by charge-coupled device cameras when luciferin is administered via intraperitoneal injection and subsequently oxidized (Thalheimer et al., 2013). However, bioluminescent imaging results are difficult to interpret because only a 10-fold change in measured intensity is possible for every added centimeter of tissue depth (Magistri et al., 2019). As such, using bioluminescent imaging in studying liver metastases, which are often deep within the peritoneal cavity, is often limited in its ability to generate new insights. Thus, beyond in vitro modeling systems like Transwell migration assays and tracing of characterized EMT markers, the dynamic cellular changes associated with metastasis remain largely unexplored in living experimental models.

However, intravital imaging in the recent years has provided a glimpse into the morphological and cellular changes involved in the metastatic process as it occurs in vivo. The development of the abdominal intravital window was a key advance in allowing visualization of peritoneal organs including the colon, liver, pancreas, and spleen for up to a month (Alieva et al., 2014; Ritsma et al., 2013). In a seminal application of this technology, Ritsma et al. imaged the metastasis of xenografted CT26 CRC cells using an abdominal intravital window in place for about 2 weeks (Ritsma et al., 2012). By visualizing the fluorescently tagged tumor cells as they metastasized to the liver, the authors showed that the seeding cells underwent clonal selection and distinct morphological changes from a “pre-micrometastatic” to a “micrometastatic” phenotype over the course of days.

Subsequent studies using similar intravital imaging methods have provided insights into the mechanistic basis underlying tumor cell invasion, generating novel hypotheses to be interrogated in future studies of CRC metastasis. In other cancer subtypes, intravital imaging has revealed novel interplay between components of the immune system and metastatic processes (Harney et al., 2015; Headley et al., 2016; Park et al., 2016). In CRC, intravital imaging of Lgr5 plasticity, a marker of CRC cancer stem cells, revealed that Lgr5-negative cells serve as the seeds of metastatic colonization before re-establishing the cellular hierarchy in the metastases (Fumagalli et al., 2020)

Although invasive intravital imaging is not feasible clinically, newer emerging technologies like microPET and SPECT have advanced our ability to image metastasis at the cellular resolution in humans as well. We expect that, as these technologies advance, we will be able to achieve resolution similar to that of the cellular imaging methods discussed here and validate that the findings in CRC murine models translate to patients with CRC.

Translational Applications

As our models of metastatic CRC expand, groups have begun to utilize these techniques to determine their feasibility to predict and model patient-specific treatment responses. In the recently published translational study, Vlachogiannis and colleagues generated a large biobank of patient-derived epithelial organoids from metastatic gastrointestinal cancers and observed that they recapitulate the chemotherapeutic response seen in the clinical setting (Vlachogiannis et al., 2018). Using in vitro drug screening of 16 unique colorectal primary and corresponding metastatic lesions, their group showed 100% sensitivity and 93% specificity between in vitro organoid response and patient clinical response. In an effort to overcome the lack of surrounding tissue structure in the organoid model, Sensi and colleagues recellularized tumor and normal colon from resected surgical specimens and seeded these with cell lines to assess the variation in response to chemotherapeutic agents (Sensi et al., 2020). To model chemotherapeutic responses for metastatic disease, one group combined several cell types that are normally found in the lung including NL20 epithelial cells, fibroblasts, lymphatic and blood vessel endothelial cells to generate a “primitive lung in a dish.” Primary colon cancer cells were then introduced into the dish leading to areas of metastatic colon carcinoma that modeled the chemotherapeutic response to that of the primary patient in vitro (Ramamoorthy et al., 2019). Similar to this model, another group developed a model of hepatic metastasis utilizing human hepatocytes, mesenchymal cells, and HCT116 cells in simulated microgravity rotating wall vessel bioreactors that allowed for assessment of dose-dependent cytotoxicity to 5-FU (Devarasetty et al., 2017; Skardal et al., 2015).