Tumor organoid models in precision medicine and investigating cancer-stromal interactions

Abstract

Tumor development and progression require chemical and mechanical cues derived from cellular and non-cellular components in the tumor microenvironment, including the extracellular matrix (ECM), cancer-associated fibroblasts (CAFs), endothelial cells, and immune cells. Therefore, it is crucial to develop tissue culture models that can mimic in vivo cancer cell-ECM and cancer-stromal cell interactions. Three-dimensional (3D) tumor culture models have been widely utilized to study cancer development and progression. A recent advance in 3D culture is the development of patient-derived tumor organoid (PDO) models from primary human cancer tissue. PDOs maintain the heterogeneity of the primary tumor, which makes them more relevant for identifying therapeutic targets and verifying drug response. Other significant advances include development of 3D co-culture assays to investigate cell-cell interactions and tissue/organ morphogenesis, and microfluidic technology that can be integrated into 3D culture to mimic vasculature and blood flow. These advances offer spatial and temporal insights into cancer cell-stromal interactions and represent novel techniques to study tumor progression and drug response. Here, we summarize the recent progress in 3D culture and tumor organoid models, and discuss future directions and the potential of utilizing these models to study cancer-stromal interactions and direct personalized treatment.

1. Introduction.

No cell in our body is an island. Local microenvironmental cues, such as adjacent cells, ECM, growth factors, and metabolites have profound effects on cell proliferation, differentiation, migration and invasion (Bissell & Hines, 2011). Therefore, restoration of these microenvironmental cues is critical for developing physiologically relevant culture models to study normal tissue function and disease progression. This has been shown to be particularly true in the development of relevant ex vivo cancer models. In the last several decades, scientists have generated numerous cancer cell lines to investigate cancer cell proliferation, migration, and drug response. Traditional 2D cell culture models based on these lines have been widely utilized in cancer cell biology and are still the mainstay for cancer drug discovery. However, very few of these potential new drugs targeting cancer make it through clinical trials. The gap between drug responses in 2D models and their actual in vivo efficacy indicates that a better tissue culture model needs to be developed for cancer drug discovery and evaluation (Breslin & O’Driscoll, 2013).

Cells in 2D culture grow as a monolayer on tissue culture plastic, where cell-cell and cell-extracellular interactions are distinct from those that occur in vivo. These interactions are critical for the regulation of signaling pathways and gene expression, which is essential for proper cellular function (Xu, Boudreau, & Bissell, 2009). For example, mammary epithelial cells in 2D culture cannot functionally differentiate and lose milk protein expression because basement membrane signaling and cell polarity are disrupted (Xu, Nelson, et al., 2009). Dr. Mina J Bissell and her colleagues developed a 3D organoid culture model three decades ago to investigate mammary gland differentiation and breast cancer progression (Barcellos-Hoff, Aggeler, Ram, & Bissell, 1989). In 3D culture, mammary epithelial cells are embedded in ECM such as collagen gel or Matrigel. In contrast to the 2D model, normal mammary epithelial cells in the 3D environment were shown to establish polarized acini-like spheroids with milk protein expression, and malignant cells formed disorganized grape or stellate-like structures (Barcellos-Hoff, et al., 1989; Petersen, Ronnov-Jessen, Howlett, & Bissell, 1992). In the last decade, many tumor organoid models, including PDOs, have been developed to mimic cancer progression (Fig. 1). Importantly, PDO models are emerging as a powerful preclinical tool for drug testing and guiding personalized treatment (Fig. 1). Here, we summarize advantages and limitations of these culture models and discuss future directions in developing 3D organoid assays.

Figure 1.

Graphs show the number of publications related to ‘three-dimensional culture’ or ‘patient derived organoid’ from 1950s to June 2020. The numbers were acquired from PubMed by searching with specific key words.

2. PDO models in cancer research and precision medicine.

2.1. Discrepancy between cancer cell lines and tumors in vivo.

Hundreds of cancer cell lines have been developed since the first cancer line HeLa was derived from cervical cancer tissue in the 1950s (Scherer, Syverton, & Gey, 1953). These cancer cell lines are common tools for cancer biology research and anticancer agent evaluation; however, accumulated evidence suggests that the clinical relevance of these models remain questionable (Gillet, et al., 2011; Gillet, Varma, & Gottesman, 2013). Most of the cancer cell lines have been cultured in monolayers for many passages in growth-promoting medium with 21% oxygen supply; and this traditional 2D culture condition selects for cells with rapid proliferation. Gene expression profile analyses showed that genes involved in cell cycle regulation and primary metabolic processes are upregulated in cancer cell lines compared with primary tumors; the upregulation of these gene is crucial for cell growth and survival in 2D culture (Gillet, et al., 2011; Lukk, et al., 2010). The long-term selection in 2D culture plays a major role in induction and maintenance of rapid growth and gene expression patterns, causing cells to become gradually less like their in vivo counterparts over time.

Intra-tumor heterogeneity is the driver of cancer progression and drug resistance (Marusyk, Almendro, & Polyak, 2012). Cancer cells are usually genetically unstable; cancer cell lines exhibit pronounced genomic heterogeneity, even after selection in culture conditions. Comparative genomic hybridization data showed that the genome abnormalities in cell lines are, on average, higher than that in primary tumors (Neve, et al., 2006). A recent study analyzed the genome, transcriptome, tumor microenvironment, and histopathological characteristics in lung squamous cell carcinoma by multi-region bulk and single-cell sequencing. Interestingly, genomic heterogeneity failed to indicate the heterogeneity at immune and transcriptomic levels. Therefore, phenotypic heterogeneity, including those related to proliferation, inflammation, cancer-stromal interaction, is the major indicator of intra-tumor heterogeneity and subtype classification (Sharma, Kilari, Cai, Simeon, & Misra, 2020). Unfortunately, cancer cell lines lose the majority of phenotypic heterogeneity after long-term selection in the growth promoting 2D culture conditions in the absence of cancer-stromal interactions. The key for developing a physiologically-relevant tumor culture model is to reestablish the intra-tumor heterogeneity and in vivo microenvironmental cues.

2.2. PDO models in drug testing and personalized treatment.

3D culture of normal and malignant cells isolated from mouse and human mammary glands has been utilized to study normal mammary tissue morphogenesis and breast cancer progression since the 1980s (Barcellos-Hoff, et al., 1989). Using 3D culture assays in Matrigel (laminin-rich reconstituted basement membrane), Peterson et al showed that normal primary human mammary epithelial cells were able to establish polarized and growth-arrested acini-like spheroids, while carcinoma cells continually grew and formed disorganized structures (Petersen, et al., 1992). This is the first study showing that primary normal mammary cells and cancer cells from biopsy cultured in 3D ECM gels recapitulate the structure and phenotype of these cells in vivo. Another significant improvement in the development of PDO models is the long-term expansion of murine and human intestinal organoids in 3D Matrigel by including key factors such as R-spondin-1, EGF, Noggin, nicotinamide, prostaglandin E2 (an Alk inhibitor), and a p38 inhibitor. In this culture condition, stem cells in the organoids retained the self-renewal ability and were genetically stable (Sato, et al., 2011). In the last decade, PDO models have been successfully developed for breast, colon, prostate, and pancreatic cancer (Table 1). Results from genetic profiling and histopathological analysis demonstrate that PDOs more accurately maintain the genetic diversity and phenotypic heterogeneity of in vivo tumors compared with cancer cell line models, thereby providing robust pre-clinical assays to predict in vivo drug sensitivity and to study cancer progression (Boj, et al., 2015; Fujii, et al., 2016; Jacob, et al., 2020; Sachs, et al., 2018; Vlachogiannis, et al., 2018). This evidence clearly shows that PDO models provide a solution to overcome many limitations in cancer cell lines.

Table 1.

Development and application of PDO models in cancer research

	Application	Culture material	Functional analysis	Citation
Breast Cancer	Drug testing	Basement membrane extract (BME; Cultrex); gelation	Western blot; Clonogenic assay; cell viability; Genomic analysis; Gene editing; Drug screen; imaging.	Whittle J.R. et al (2020), PMID: 32245900; Sachs N.et al (2018), PMID: 29224780; Walsh A.J. et al (2014), PMID: 25100563; Xiong G. et al (2018), PMID: 30367042.
	Model development	Basement membrane extract (BME; Cultrex); gelatin sponge platform; coculture assay in Transwell.	Immunohistochemistry (IHC); Genomic DNA analysis; Quantitative real-time PCR; ChlP-seq.	Petersen O.W. et al (1992), PMID: 1384042; Li X. et al (2020), PMID: 32206037; Mazzucchelli S. et al (2019), PMID: 31223292; Sachs N. et al (2018), PMID: 29224780; Centenera M.M. et al (2018), PMID: 30117261; Chung B. et al (2017); PMID: 28649646.
	Function research	Matrigel; Tumor-on-a- chip platform; floating 3D collagen- based hydrogel scaffolds	Cytokine ELISA array; Colony-formation assay; Quantitative PCR; western blot analysis; RNA sequencing.	Chatterjee S. et al (2019), PMID: 31419632; Shirure V.S.et al (2018), PMID: 30393802; Rocco S.A. et al (2018), PMID: 29741670.
	Personalized treatment	3D porous PCL scaffolds	Picogreen DNA quantification assay; Cell viability assay; drug response analysis.	Sachs N.et al (2018), PMID: 29224780; Nayak B. et al (2019), PMID: 30393802.
Colon and Gastrointe stinal Cancer	Personalized Medicine	Matrigel	Immunofluorescence (IF) imaging analysis; Phenotypic and genotypic profiling; Molecular profiling; cytotoxic assays; cell Viability assay; IHC.	Costales-Carrera A.et al (2019), PMID: 31752287; Pan Y. et al (2018), PMID: 30591933; Yao Y et al (2020), PMID: 31761724; Vlachogiannis G.(2018), PMID: 29472484.
	Drug testing and screening	Matrigel; collagen gel;	Imaging analysis; cell viability assay; tumor antigen-specific cytotoxicity; Western blotting analysis.	Schnalzger T.E. et al (2019), PMID: 31036555; Usui, T et al (2018), PMID: 29512123; Usui, T et al (2018), PMID: 29642386; Usui, T et al (2016), PMID: 28119740; Liu X. et al (2018), PMID: 30479121.
	Functional research	Matrigel	Imaging analysis; LC- MS/MS analysis.	Mousavi N. et al (2019), PMID: 30884019; Kondo J. et al (2018), PMID: 30343529; Abe Y et al (2018), PMID: 30061712.
	Model development	Matrigel; spheroids; agarose	Image analysis; microarray analysis; RNA and DNA Sequencing; real-time PCR; Growth kinetics; Cytotoxic assays; Histopathological analysis.	SATO T.et al (2011), PMID: 21889923; Wetering M.V.D. et al (2015), PMID: 25957691; Weeber F. et al (2015), PMID: 26460009; Weeber F.et al (2017), PMID: 28903445; Dngles-Marie et al (2007), PMID: 17210723; Fujii M. et al (2016), PMID: 27212702.
Glioblasto ma (GBM)	Drug screening	The collagen- hyaluronic acid (HA) bioink;	Cell viability assay; cell proliferation analysis	Maloney E. et al (2020), PMID: 32085455.
	Model development	Matrigel; gelatin and poly (ethylene glycol) (PEG) based hydrogel platforms; C(chitosan)- HA (hyaluronic acid) scaffold	Hematoxylin and eosin (H&E) and IHC staining; cell proliferation; quantitative PCR analysis; IF imaging; Invasion assay; Flow cytometry analysis.	Sailer V. et al (2017), PMID: 28373416; Pedron, S. et al.(2013), PMID: 23827186; (2015), PMID: 25521283; Wang, C. et al (2017), PMID: 27770562; Florczyk, S.J.et al (2013), PMID: 24075410; Kievit,F.M.et al (2014), PMID: 25109438.
	Function research	HA-Collagen hydrogel; Collagen I gel	Invasion assay; IHC; quantitative PCR; cell proliferation and viability assay; time-lapse confocal imaging; scanning electron microscopy (SEM) imaging; Gelatin zymography.	Cha, J. et al (2016), PMID: 27108713; Rao, S.S. et al (2013), PMID: 24010546; Pedron, S. et al (2013), PMID: 23559545; Vega, F.M. et al (2011), PMID: 21576392; Sarkar,S. et al (2006), PMID: 17178873.
Liver cancer	Model development and Drug testing	cross-linked MA-HPC hydrogel	Cell viability and growth assessment; mmunofluorescence staining; Whole exome sequencing and RNA- sequencing.	Eliza Li Shan Fong, et al (2018), PMID: 29353739.
Pancreatic cancer	Model development	Matrigl, embedded,	H&E staining, DNA and RNA analysis.	Tiriac H. et al (2020), PMID: 32009658; Sailer V. et al (2017), PMID: 28373416.
	Personalized treatment	Matrigel	Cell viability assay; redox metabolism analysis; Immunoblotting; extracellular metabolite analysis.	Broekgaarden M. et al (2019), PMID: 31494503.
	Function research	Matrigel	flow cytometric, IF and iIHC; quantitative PCR; Gene expression profiling; iTRAQ proteomic analysis.	Tsai S. et al (2018), PMID: 29587663; Johnson J. et al (2020), PMID: 31661714; Baker L.A. et al (2019), PMID: 30378047; Boj S.F. et al (2015), PMID: 25557080.
	Drug testing and screening;	Matrigel; suspension	Cell viability Assay; HTS Campaign and Data Processing; Transmission electron microscopy; IF and IHC; Gene expression analysis.	Burkhart R.A. et al (2018), PMID: 29736724; Huang L. et al (2015), PMID: 26501191; Hou S. et al (2018), PMID: 29673279.
Prostate cancer	Drug testing and screening	Floating basement membrane (Matrigel) spheroids; Matrigel	IHC and imaging analysis; western blot; RNA-seq; Methylation profiling; High-throughput drug assay.	Lee S. et al (2020), PMID: 32065165; Gao D, et al (2014), PMID: 25201530; Drost J. et al (2016), PMID: 26797458; Puca L. et al (2018), PMID: 29921838; W. R. et al (2014), PMID: 25201529.
	Modeldevelo pment	Spheroids; Matrigel	IHC; Whole exome sequencing and genetic analysis; drug sensitivity, Western blot.	Linxweiler J. et al (2019), PMID: 30474758.
	Function research	Matrigel	H&E staining.	Sailer V.et al (2017), PMID: 28373416.
gynecolog ic tumors	Drug screening and precision medicine	Matrigel Bilayer organoid culture; Spheroid cultures	Pathologicalanalysis; Cell proliferation and drug sensitivity; IHC; Drug sensitivity assay.	Maru Y. et al (2019), PMID: 31101504; Girda E. et al (2017), PMID: 28683005; Kiyohara Y. et al (2016), PMID: 26825848.
Renal cell carcinoma	Function research	Matrigel	H&E and IF staining; Cell proliferation assay (CCK- 8 assay); Quantitative PCR; Western blot.	Na J.C. et al (2020), PMID: 32158973.
Bladder cancer	Model development and drug testing	Matrigel	RNA sequencing; IF and imaging analysis; Organoid drug response assay; Western blotting.	Lee S.H. et al (2018), PMID: 29625057.

PDO models were first developed to mimic in vivo tumor phenotypes and investigate cancer progression ex vivo. After the development of long-term culture assays for PDOs, more and more studies began to utilize these modes for drug testing and personalized medicine applications (Fig. 2) (Table 1). PDO biobanks have also been established from samples derived from advanced rectal cancer and metastatic colorectal/gastroesophageal cancer. Phenotypic and genotypic profiling data showed that these PDOs are very similar to the original patient tumor (Vlachogiannis, et al., 2018). By comparing PDO responses to anticancer agents to the response of the patient in clinical trials, they found that PDOs subjected to anticancer treatment were able to recapitulate patient responses in the clinic trial (Vlachogiannis, et al., 2018). The co-clinical trial data from a study in advanced rectal cancer also confirm that chemoradiation responses in patients are highly matched to PDO responses (Yao, et al., 2020). PDO biobanks have also been established in breast cancer and glioblastoma (Jacob, et al., 2020; Sachs, et al., 2018). Breast organoids derived from primary and metastatic breast cancer recapitulated the diversity of the disease and matched the histopathology and hormone receptor/HER2 status of the original tumor. In addition, the response of breast cancer organoids to anti-hormone therapy indicates the potential use of breast cancer organoids as predictive surrogates for tumors in patients (Sachs, et al., 2018).

Figure. 2.

A scheme showing the utilization of PDO models in cancer biology, drug testing, and personalized medicine.

PDO models are a powerful tool for functional studies and cancer target verification (Fig. 2) (Table 1). Sylvia F. Boj et al described a method generating neoplastic pancreatic organoids using the CRISPR-Cas9 genome-editing system (Matano, et al., 2015). They introduced mutations in the tumor suppressor genes APC, SMAD4 and TP53, and in the oncogenes KRAS and/or PIK3CA in organoids derived from normal human intestinal epithelium. These organoids formed tumors after implantation under the kidney subcapsule in mice (Matano, et al., 2015). Xiong et al silenced RORα expression in primary mouse mammary organoids; these organoids were transplanted into mammary glands to study the function of RORα in branching morphogenesis. In a separate study using triple-negative breast cancer (TNBC) PDO models, Xu’s laboratory showed that targeting P4HA1 was a potential therapeutic strategy to sensitize TNBC to chemotherapeutic agents. The results of P4HA inhibitor treatment experiments in PDOs are consistent with the results from the TNBC patient-derived xenograft (PDX) model (Xiong, et al., 2018). These studies demonstrate the potential of utilizing PDOs to identify therapeutic targets and molecular pathways that modulate cancer progression.

PDO models can also be used for high throughput drug screening (Fig. 2) (Table 1). Dr. Clever’s laboratory tested a panel of 76 potential therapeutic agents in 30 PDOs from pancreatic tumors and identified the PRMT5 inhibitor EZP015556 as a potential drug to inhibit MTAP (a gene commonly lost in pancreatic cancer)-negative tumors and a subset of MTAP-positive tumors (Driehuis, et al., 2019). A high throughput screening-compatible organoids culture system has been established in standard flat-bottom 384- and 1536-well plates; patient-derived pancreatic cancer primary cells were tested in a pilot screening with well-characterized anticancer agents and FDA-approved drugs in this system (Hou, et al., 2018). One potential limitation for this type of screening is the amount of patient sample available. However, since organoids derived from PDX can largely retain the morphological and genomic status of the original tumor tissue, these organoids can be used to increase the relevant sample size. A recent study performed a high-throughput screening of 2427 drugs with PDX-derived organoids (Kondo, et al., 2019). These studies showed that PDO can be applied to support large-scale drug screening, which is an important step toward personalized medicine.

2.3. Limitations in current PDO models.

Organoids in POD models are cultured either in the natural or synthetic matrix hydrogel or in a suspension condition (Table 1). Laminin-rich ECM, including basement membrane extract and Matrigel, is widely used to culture breast, colon, prostate, and pancreatic cancer PDOs (Boj, et al., 2015; Fujii, et al., 2016; Sachs, et al., 2018). It provides an ideal culture condition for adenocarcinoma cells since epithelial cells usually adhere to the basement membrane in vivo (Yurchenco, 2011). Interestingly, hyaluronic acid-based assays have been successfully developed for the glioblastoma PDOs (Malhotra, et al., 2020; Rao, et al., 2013). Hyaluronic acid is one of the major components of brain ECM and is also enriched in glioblastoma tissue (Y. Kim & Kumar, 2014). Culture conditions for these PDOs suggest that mimicking ECM cues of primary tumors is crucial for the development of PDO models. Nevertheless, ECM composition in the tumor microenvironment is much more complicated than laminin- or HA-rich ECM gels, and the ECM-cancer cell interaction is oversimplified in current PDO culture assays.

Stromal cells, including fibroblast and immune cells, are present in primary PDOs; however, these cells are gradually lost during the long-term culture (Clinton & McWilliams-Koeppen, 2019). Most of the PDO models contain only malignant cells, which have unlimited proliferation capability. Although addition of growth factors and supplements may partially recapitulate stromal cell function, the lack of multi-cellular components in the tumor microenvironment is still considered a significant limitation of current PDO models. The 3D co-culture systems and decellularization techniques discussed in the next section may offer a promising solution to mimic in vivo cancer-stromal cell crosstalk and cancer-ECM interaction.

3. 3D culture models to establish cancer-stromal interaction.

3.1. Acellular tissue to mimic in vivo ECM microenvironment.

ECM is the major non-cellular stromal component in the tumor microenvironment; its complex and dynamic nature has a significant impact on cellular behavior. Hundreds of ECM and ECM-related proteins have been characterized in normal and malignant tissues. ECM deposition has been detected at the early stage of tissue morphogenesis (Loganathan, et al., 2016); these ECM molecules have profound effects on cell proliferation, migration, and differentiation (Lu, Weaver, & Werb, 2012). ECM remodeling in malignant tissue is crucial for cancer development and progression (Lu, et al., 2012). Overexpression and increased deposition of collagen have often been detected in the early stages of cancer (Peng, et al., 2017; Zhang, et al., 2018; Zhu, et al., 2015). It has been shown that increased collagen deposition and crosslink promotes cancer development by enhancing tissue stiffness to change biophysical cues (Levental, et al., 2009; Provenzano, et al., 2008), and inducing biochemical signaling through receptor DDRs (Coelho, et al., 2017; H. Huang, et al., 2016) and integrins (Zeltz & Gullberg, 2016).

Mass spectrometry-based proteomic analysis has been used to characterize ECM composition in normal and malignant tissues (Aebersold & Mann, 2016). These analyses identified the organ-specific ECM components in the lung, liver and colon (Naba, et al., 2016) (Table 2). Specific alterations in the ECM composition and organization were also identified in cancer compared with normal tissue, including changes in collagen expression patterns (Biondani, et al., 2018), alternative ECM-protein splice forms, and modified fibril organization (Bonnans, Chou, & Werb, 2014; Pickup, Mouw, & Weaver, 2014). These studies indicate the ECM composition is organ-specific and associated with cancer progression. The organ-specific ECM deposition suggest that cancer development and progression require specific matrix context. Although multiple 3D culture models including spheroids in suspension, collagen gels, Matrigel, chitosan, and artificial polymers have been developed and widely used for cancer biology research, most of these models lack native ECM composition from specific organs, and therefore fail to mimic biophysical and biochemical ECM cues in the tumor microenvironment. How to fill this gap between 3D culture models and native ECM cues in tumor tissue remains an important issue to be addressed.

Table 2:

Tissue-specific ECM deposition.

	Major components in ECM
Human Organ	collagens	Glycoproteins	Proteoglycans	ECM-associated protein	ref
Lung	COL1, COL4, COL5, COL12, COL14, COL17	Fbn1, EFEMP1, LAM (C2,B1,A5), NID1,FBLN1, MGP	LUM, HSPG2, BGN	ECM1	Calle E.A. et al (2016), PMID: 27693690.
Colon	COL1, COL3, COL4, COL5 COL6,	TGFBI, ELN, POSTN, FGB, EMILIN1, DPT, TINAGL1, FGG, FBN1, LAMC1 EFEMP1	Decorin HSPG2 LUM OGN PRG2 PRELP	Galectin-1 LGALS4 ANXA2 CTSG	Naba A. et al (2014), PMID: 25037231.
aorta	COL1 COL6	Tetranectin, Fibulin-3	Lum, Galectin-1	SOD3SAP ApoA-I TRYB1	Didangelos A. et al (2010), PMID: 20551380.
liver	COL1, COL3 COL5, COL6	ELN, FGG, FGB TGFBI, POSTN, EMILIN1	PRG2 HSPG2 PRG3 PRELP LUM	CTSG, CTSD, TGM2, CTSZ LGALS4, CTSB INHBE, TIMP3	Naba A. et al (2014), PMID: 25037231.
Myocardium	COL4, COL1 COL6	Lam (A2, B2, C1), Fbn1, SAMP	Heparin sulfate	ALBU, CILP1, MYH7	Guyette J.P. et al (2016), PMID: 26503464.
glomerular	COL4, COL18, COL6	LAM(A5, B2, C1), TINAGL1, NID(1,2), VTN,	Perlecan Biglycan, Asporin		Lennon R. et al (2014), PMID: 24436468.

Acellular tissue (decellularized matrix) provides a tool to bridge the gap between 3D culture assays and in vivo tumor models. Decellularization is a process that allows cell removal from organs or tissues but has little impact on ECM structure and composition (Fig. 3). This technique was first developed for tissue engineering to generate tissue such as skin (Livesey, Herndon, Hollyoak, Atkinson, & Nag, 1995), bladder (Sutherland, Baskin, Hayward, & Cunha, 1996), and heart valve (Drakos, et al., 2011). Decellularized tissue has recently been used to study cancer invasion and colonization (Piccoli, et al., 2018; Xiong, Flynn, Chen, Trinkle, & Xu, 2015). By analyzing decellularized normal and colon tumor tissues, scientists confirmed that the ECM composition and alignment are largely maintained after decellularization (Romero-Lopez, et al., 2017). To study the lung colonization of breast cancer cells, Xu’s laboratory developed an ex vivo lung colonization model using decellularized lung matrix (Xiong, et al., 2015). In this model, metastatic breast cancer cells invaded and grew, while silencing Zeb1, an epithelial-mesenchymal transition (EMT) inducer, significantly decreased invasion and lung colonization, which confirmed the roles of EMT during cancer cell invasion and metastasis. Similar models have also been established for pancreatic and colon cancer (Campbell, Cukierman, & Artym, 2014; H. J. Chen & Shuler, 2019). This technique may be combined with PDO models to better mimic cancer progression and predict drug response (Fig. 3).

Figure 3.

Development of 3D tumor organoid models with decellularized matrix.

3.2. 3D co-culture models to mimic cancer-stromal cell interaction

Stromal cells in the tumor microenvironment, including cancer-associated fibroblasts (CAFs), adipocytes, endothelial cells, and immune cells, are crucial for cancer development and progression (Goto & Nishioka, 2017; Lambrechts, et al., 2018). These cancer-associated stromal cells modulate cancer cell proliferation, migration, invasion, and apoptosis through direct cell-cell interaction and secretion of soluble factors, ECM, and small metabolites (Alkasalias, et al., 2014; Guo & Deng, 2018). Therefore, it is important to include these components in the ex vivo model in order to better mimic cancer progression and drug response in vivo.

A variety of co-culture models incorporating cancer cells, ECM, and stromal cells have been developed in the last two decades. As one of the most prominent stromal cells in tumor microenvironment, CAFs has been identified in almost all solid tumors. 2D and 3D co-culture assays have been widely used to investigate cancer-CAF interaction in different types of cancer. CAFs are mainly derived from mesenchymal stromal cells (MSCs) and local fibroblasts (Spaeth, et al., 2009). MSCs contribute to tissue homeostasis and injury repair (Kawada, et al., 2004; Spees, Lee, & Gregory, 2016), and are identified in most tissues, including tumors. 3D co-culture models with MSCs and cancer cells have been developed using hanging drop assays, or scaffolds involving chitosan, hyaluronan or other hydrogels to mimic in vivo cancer cell-fibroblast interaction (Avnet, et al., 2019; Bartosh, Ullah, Zeitouni, Beaver, & Prockop, 2016; Devarasetty, Wang, Soker, & Skardal, 2017; Shamai, Alperovich, Yakhini, Skorecki, & Tzukerman, 2019). These models have been utilized to study the formation and structure of cancer organoids/spheroids in the presence of MSCs. Interestingly, MSCs were shown to inhibit cancer progression through cell cannibalism in some studies (Bartosh, et al., 2016); while in other reports, the MSCs promoted the cancer growth, invasion, and chemoresistance by inducing stemness and E-cadherin degradation in cancer cells (Avnet, et al., 2019; Devarasetty, et al., 2017; Dittmer, Hohlfeld, Lutzkendorf, Muller, & Dittmer, 2009). The discrepancy among those studies may be attributed to the different status of MSCs in the 3D models.

Patient-derived cancer organoids and CAFs have recently been utilized in 3D co-culture assays to study cancer cell-CAF interaction (Fiorini, Veghini, & Corbo, 2020). To investigate CAF heterogeneity in the pancreatic cancer, Dr. Tuveson’s laboratory developed a 3D culture assay by co-culturing CAFs and PDOs in Matrigel or Transwells. They identified two subtypes of CAFs with different protein expression profiles in pancreatic ductal adenocarcinoma (Ohlund, et al., 2017). One distinct subpopulation of CAFs with reduced αSMA was identified from the co-cultures, which located more distantly from neoplastic cells and secreted IL6 and additional inflammatory mediators (Ohlund, et al., 2017). A 3D coculture system that includes the ECM and CAFs that recapitulates the progression of lung squamous carcinoma and has been established to investigate dynamic interactions between cancer cells and stromal component in tumor tissue. Using this model, scientists showed that CAFs suppressed the function of SOX2 and restored hyperplasia in patient-derived non-small cell lung cancer cells (S. Chen, et al., 2018). The ability to capture heterogeneity and complexity of primary tumors makes 3D co-culture models that integrate PDOs and CAFs to be a promising strategy to study the tumor microenvironment.

Obesity is a risk factor for many types of cancer, and roles of adipocytes in cancer development and progression are well characterized in mouse models (Incio, et al., 2016; O’Sullivan, Lysaght, Donohoe, & Reynolds, 2018). Co-culture assays have been developed to identify potential pathways and factors that mediates the crosstalk between adipocytes and cancer cells. Transwell plates are often utilized in these assays to physically segregate cancer cells and adipocytes, while allowing transmission of soluble factors mediating cancer cell-adipocyte crosstalk. Adipocyte IL-6 was identified as the important factor mediating adipocyte dependent cancer invasion with the Transwell co-culture assay. The same assay also reveal function of cancer cell-derived Wnt3a in inducing cancer-associated adipocyte phenotypes (Dirat, et al., 2011; C. K. Huang, et al., 2017). The direct interaction between cancer cells and adipocytes can be analyzed with 3D co-culture systems. A recent study showed that physical interaction between adipocytes and cancer cells induced MET in MDA-MB 231 and Hs578 cell in the 3D co-culture assay (Pallegar, Garland, Mahendralingam, Viloria-Petit, & Christian, 2019). The majority of results from co-culture models determined that cancer-associated adipocytes promoted cancer cell growth, epithelial-mesenchymal or mesenchymal-epithelial transition during cancer progression, and cancer cell colonization at secondary sites (Goodwin & Chlebowski, 2016; Quail & Dannenberg, 2019).

Macrophages are the most abundant immune-related stromal cells in the tumor microenvironment, and tumor-associated macrophage infiltration is associated with cancer progression and poor prognosis (Allavena, Sica, Garlanda, & Mantovani, 2008). Macrophage infiltration and differentiation suppresses immune response, promote angiogenesis, and stimulate cancer malignancy and metastasis (Pollard, 2004). The 2D co-culture assays are widely used to verify these in vivo findings and determine how these macrophages and cancer cells interact with each other. Using this model, scientists showed that tumor-associated macrophages suppressed Taxol-induced mitotic arrest and promotes cancer cell survival through modulating the MAPK/ERK pathway in cancer cells (Olson, Kim, Quail, Foley, & Joyce, 2017). By co-culturing THP1 cells or CD14-positive monocytes with HMT-3522 series cells in 3D Matrigel, Li et al found that both THP1 cells and monocytes invaded into Matrigel and adhered to non-polarized malignant epithelial cells but not polarized normal cells (Li, et al., 2017). Transwell has also been used in some co-culture assays to separate the macrophages and cancer cells (Che, et al., 2017; Lee, et al., 2018). It has been shown in this assay that macrophages induce EMT and promote invasion of lung cancer cells through the COX-2/PGE2/beta-catenin signalling pathway. These co-culture models are appropriate to study the indirect interaction between solid tumor cells and infiltrated immune cells.

3.3. Microfluidic devices in 3D culture

In vivo, significant interactions between tumors and the circulatory system occur at virtually every stage of development: from simple gas exchange and delivery of soluble factors via the circulation to neovascularization of existing tumors to the complex intravasation-circulation-extravasation cascade that occurs during metastasis. Additionally, immune cells in circulation experience fluid sheer stress and hydrostatic pressure, which can substantially influence immune cell function and cancer-immune cell interaction (Heldin, Rubin, Pietras, & Ostman, 2004; Polacheck, Charest, & Kamm, 2011). Therefore, it is crucial to include a way of mimicking these circulation parameters in ex vivo cancer models. The use of microfluidic devices provides a platform for developing these types of assays. These devices typically contain a series of microscale channels with characteristic dimensions of less than 1 mm that can be used to recreate dynamic fluid flow. Devices such as these have been used extensively with cell culture (Mehling & Tay, 2014; van Duinen, Trietsch, Joore, Vulto, & Hankemeier, 2015) and to provide high resolution; this technology has been widely used in biological analysis (Du, Fang, & den Toonder, 2016; Whitesides, 2006).

Over the last decade, a number of microfluidic devices have been integrated with 3D culture models to study cancer progression; the field of microfluidic-enabled 3D cancer models has evolved substantially since its infancy, and several excellent review papers have been written on the subject (Boussommier-Calleja, Li, Chen, Wong, & Kamm, 2016; Carvalho, Lima, Reis, Correlo, & Oliveira, 2015; Sung & Beebe, 2014). Some of the more recent device innovations focus on increased accessibility of microfluidic devices to researchers, such as making it easier to integrate hydrogel-based 3D cultures within microfluidic devices (Torabi, et al., 2019) or using materials that are easily integrated into typical cell culture protocols (Berthier, Young, & Beebe, 2012). Additionally, the ability to include dynamic fluid flow in cell culture makes microfluidic devices well suited for studying facets of both angiogenesis and metastasis, and devices of this type have been used extensively to elucidate details of both processes (Jeon, et al., 2015; Peela, et al., 2017).

The inclusion of microfluidic channels in 3D co-culture models provides a unique platform to investigate the crosstalk between multiple cell types, such as breast cancer cells and mammary fibroblasts in collagen gel (Montanez-Sauri, Sung, Berthier, & Beebe, 2013). Similarly, the dynamic interaction between cancer cells and macrophages was studied by incorporating monocytes and monocyte-derived macrophages into a microfluidic device (H. Kim, et al., 2019; Yu, et al., 2019). It has also been shown that circulating tumor cell (CTC) clusters establish more rapidly in a microfluidic platform compared with traditional biopsy enrichment; subsequent treatment with chemotherapy agents inhibited cluster formation in the device (Khoo, et al., 2016). These results suggest that microfluidic platforms have great potential for CTC culture and drug screening.

4. Current limitations in tumor organoid culture models and future directions

Cancer progression is a multistep process; it is important to develop ex vivo models to mimic different stages of cancer progression, especially cancer metastasis. Cancer metastasis involves cancer cell detachment from the primary tumor, survival in circulation, and colonization in secondary organs. CTCs are cancer cells that detach from primary tumors and enter the circulation system. Only a small number of CTCs that detached from the primary tumors survive from environmental stressors in blood stream. These stressors include detachment from the original ECM environment, shear force from blood flow, oxidative stress, and attack from immune cells (Micalizzi, Maheswaran, & Haber, 2017). Development of a CTC culture model that can mimic these stressors will greatly facilitate research in CTC biology and drug screening. A significant portion of disseminated cancer cells remains in dormancy at secondary organs. Although dormant cancer cell lines and culture assays have been developed to study cancer dormancy, the clinical relevance of these culture models is still questionable. Since dormant cancer cells at metastatic sites are mainly derived from CTCs, it is plausible to develop cancer dormancy models by integrating patient-derived CTCs with the tissue microenvironment at secondary organs.

PDOs largely maintain the genetic diversity of original primary tumors even after long-term culture, and accumulated evidence indicates that the PDO model is a promising tool for cancer biology research, drug screening, and personalized medicine. However, the phenotypic heterogeneity in cancer tissue is associated with and regulated by microenvironmental cues. Decellularization techniques, 3D co-culture assays, and microfluidic devices have been developed to mimic in vivo cancer cell-ECM and cancer-stromal cell interaction. Combining the PDO model with these new techniques and co-culture assays may provide more powerful platforms for cancer research and anti-cancer drug screening. A significant challenge in developing patient-derived co-culture models is the stromal compartment. Most of the co-culture models uses stromal cell lines or primary stromal cells isolated from mouse tissues; however, the clinical relevance of these cell lines remains questionable. The future direction for PDO co-culture models is to utilize matched patient-derived stromal cells and/or immune cells to mimic stromal-cancer cell interaction in patients.

Development of high-throughput platforms is crucial for drug screening and personalized treatment testing. To mimic the complicated nature of tumor microenvironment, multiple stromal cell types and microfluidic devices have been included in the co-culture models. The high-throughput potential of these 3D co-culture models is significantly reduced by increasing assay complexity. Therefore, it is important to balance the complexity and high-throughput potential in tumor organoid co-culture models. Microfluidic devices provide a powerful tool to study the cancer-immune cell interaction, but most of these devices are designed at the limited scale and not suitable for drug screening. Therefore, high-throughput platforms still need to be developed for microfluidic devices and co-culture models in the future.

Abbreviations

CAFs

cancer-associated fibroblasts

CTC

circulating tumor cell

ECM

extracellular matrix

MSCs

mesenchymal stromal cells

PDO

patient-derived organoid

3D

three-dimensional

TNBC

triple-negative breast cancer

PDX

patient-derived xenograft