There are four commonly used in vitro tumor models (Fig. 1). Each tumor model and its applications are further elaborated below.
2D cell lines
Cell lines are derived from cells that have acquired oncogenic mutations that permit them to grow indefinitely [14]. It is estimated that 1 out of 105 to 107 cells that are cultured from tissue form these immortal cell lines [15].
In 2012, Barretina et al. [16] published a paper characterizing 947 human cancer cell lines representing 36 tumor types and tested 24 anticancer drugs on 479 of the cell lines. This study was later enhanced to also include RNA splicing, DNA methylation, and histone modification for 1072 cell lines [17]. The study is significant because of its robust and comprehensive nature, broadly covering various aspects of characterization and a large number of cell lines studied. The results reveal that the studied cell lines have the potential to show genetic correlations with drug responses. Similarly, a study by Garnett et al. [18] used 639 cancer cell lines with 130 drugs to determine the genes associated with specific cellular responses. Though promising, this form of 2D cell culture, where cell lines are plated as a two-dimensional monolayer of cells, has shown several limitations in recapitulating the complexity of cells growing in the human body. For example, these models do not capture the same microenvironment that the cells thrive in, as they lack cellular heterogeneity and the three-dimensional environment that involves complex interactions with neighboring cells and the extracellular matrix [1, 19]. Furthermore, while many cancer cell lines carry important genetic mutations found in corresponding cancers, many cell lines do not necessarily contain these important genetic aberrations [16, 20]. For example, the seven established prostate cancer cell lines do not carry the TMPRSS2-ERG interstitial deletion, SPOP mutation, and FOXA1 mutation found in patients with this disease [20]. Moreover, the genetic makeup of these cell lines changes over passages and in different laboratory conditions, as studies have reported that cell lines grown in two different labs show extensive clonal diversity [21]. When 27 strains of the MCF7 cell line were tested against anti-cancer compounds, very different drug responses were found [21]. Such high variability in both genetic makeup and drug responses can result in false positives during clinical trials leading to a waste of time and resources [1, 22]. Therefore, organoids provide promising model systems with less genetic variability and greater reproducibility as further detailed in the PDO section.
3D cell lines
Several alternatives to 2D methods of cell culture have been established. One of these is the 3D culture of cell lines to form spheroids. Studies have shown that 3D cell cultures show improved cell morphology, mechanical properties, differentiation, and viability compared with those of 2D culture [1, 2]. Drug metabolism and secretion in 3D culture also makes the cells more suitable for drug screening [1, 2]. For example, in a study of MCF-7 cell lines, the researchers found that cells grown in 3D culture were more resistant to chemotherapy compared with those grown in 2D culture [23]. Cells grown in 3D cultures have different cell surface receptor expressions, cell densities, and metabolic functions that may affect drug effectiveness [23]. Considering that many drugs may pass initial screening using 2D cell lines, 3D cell lines can filter out drugs that may not be effective in vivo. In a study by Lee et al. [24] on 3D cultures of malignant and nonmalignant mammary cells, important signals were found to be lost when cells were grown as 2D cultures. The researchers reported phenotypic differences between the two cultures, with the nonmalignant 3D cultures showing polarized, growth-arrested colonies, and the malignant 2D culture showing disorganized, proliferative, non-polar colonies [24]. Therefore, 3D cultures may provide a better cancer model for use in testing the effectiveness of drugs on particular cancer types. To further compare 2D and 3D cell culture, Zoetemelk et al. [25] used seven human colorectal carcinoma cell lines to form 3D spheroids in vitro and found that the sensitivity to drugs differed between the 2D and 3D cultures of the same cell line. For example, 5-fluorouracil (5-FU) efficacy was found to be reduced in SW620 and HCT116 3D cultures while sensitivity to erlotinib treatment increased in DLD1 3D cultures compared with their 2D counterparts [25]. Such differences indicate that the way the cells are plated (2D monolayer vs 3D spheroid) has an effect on their reliability as cancer models. However, although the 3D spheroid models are promising because of more similar morphology to the tissue of interest, they still present with problems such as clonal diversity and cellular evolution resulting in varying drug responses.
Primary cell cultures
The development of primary cell cultures provides a more personalized way of culturing cells, as it allows scientists to use cells that correspond to individual patients, rather than using one patient sample to represent all patients with that particular disease. Culturing primary cells involves growing cells from fresh patient tissues and using those that grow successfully. Unfortunately, culturing primary cells presents many problems. Many of the cells stop growing shortly after culturing because they undergo a process called ‘senescence’ (mortality checkpoint 1), which is characterized by growth arrest, preserved chromosome integrity, and gradual death, or ‘crisis’ (mortality checkpoint 2), which is characterized by chromosome instability and high rates of cell death [15]. Fibroblasts are the only cell types that are persistent because they secrete the proteins needed for growth [14]. This decreases the heterogeneity that is important in accurately maintaining the cells in their native environment in the tissue. Furthermore, fibroblast cultures are only able to divide around 50 times before they stop growing, greatly limiting their use for downstream experiments and analyses [14]. Finally, studies show that after many passages, gene expression profiles differ between low passage and high passage cells [20, 21, 26].
Patient-derived organoids (PDOs)
The establishment of PDOs has offered great potential in cancer research, since the year 2013. A PubMed search shows that the term “patient-derived organoids” in published studies begin to appear (n > 4) in 2013, during which five papers had been published on the subject. The number of papers is steadily increasing over the years, and by 2019, about 260 papers had been published. PDOs can provide the advantages found in both 3D spheroids (the improved morphology), primary cell lines (each culture can represent the patient it came from), and many more.
Briefly, the culture of PDOs begins with mincing up the patient tissue and plating the cells in drops of a solid jelly-like substance that is usually Matrigel or Basement Membrane Extract. These substances are two kinds of solubilized basement membrane extracts derived from Engelbreth-Holm Swarm mouse sarcoma consisting of laminin, collagen IV, entactin, and heparan sulfate proteoglycans [6, 11]. Then, nutrient-filled media is added around the drops to allow for the PDOs’ continuous growth, and for the cells to grow into spheroid shapes that are termed ‘organoids’. If grown well, the shapes retain the genetic landscape and histological properties of the original tumor. The currently established patient-derived cancer organoids include liver cancer [6], prostate cancer [27], breast cancer [28], colon cancer [29], and pancreatic cancer [30].
To begin with, gene expression profiles tend to remain more stable in PDOs. To illustrate this, seven prostate organoid lines showed identical mutational landscapes to those of their corresponding tumor tissues after three months of in vitro culturing, harboring similar disease-specific genomic alterations as the tumor tissue [27]. In another study of PDOs of metastatic gastrointestinal cancer, drug sensitivities were tested and compared with patient response [5]. The study results showed 88% accuracy in predicting treatment response and 100% accuracy in predicting non-response to treatment [5].
The reproducibility of PDOs is also promising. Sachs et al. [28] generated reproducible dose–response curves using 21 concentrations per drug. In most of the cases, they found a homogeneous response to particular drugs identifying a single half maximal inhibitory concentration (IC50), even though several IC50 values were obtained [28]. The results indicate that some subpopulations may be more susceptible to the drugs than others [28]. Reproducibility was also seen in liver cancer organoids in a study by Broutier et al. [6]. Specifically, the study revealed a correlation between drug response and the two biological replicates (1 and 2), which represent different passages of the same organoid line [6]. This suggests that drug response remains consistent between different passages.