Owing to the reduced co-relationship between conventional flat Petri dish culture (two-dimensional) and the tumour microenvironment, there has been a shift towards three-dimensional culture systems that show an improved analogy to the same. In this work, an extracellular matrix (ECM)-mimicking three-dimensional scaffold based on chitosan and gelatin was fabricated and explored for its potential as a tumour model for lung cancer. It was demonstrated that the chitosan–gelatin (CG) scaffolds supported the formation of tumoroids that were similar to tumours grown in vivo for factors involved in tumour-cell–ECM interaction, invasion and metastasis, and response to anti-cancer drugs. On the other hand, the two-dimensional Petri dish surfaces did not demonstrate gene-expression profiles similar to tumours grown in vivo. Further, the three-dimensional CG scaffolds supported the formation of tumoroids, using other types of cancer cells such as breast, cervix and bone, indicating a possible wider potential for in vitro tumoroid generation. Overall, the results demonstrated that CG scaffolds can be an improved in vitro tool to study cancer progression and drug screening for solid tumours.
Malignant neoplasms are the second most common cause of death throughout the world . Their treatment is usually performed either through surgery, radiotherapy, chemotherapy or via a combination of these treatment modalities . Of these, chemotherapy involves the use of potent molecules that can cause cell death. Before a molecule can be used clinically as a chemotherapeutic agent, it needs to broadly undergo two stages of development: screening and testing. The most commonly used models for screening and early-stage testing include the two-dimensional culture of human-derived tumour cell lines and xenografts in small animals. The gold standard two-dimensional culture-based screening method offers the advantages of convenience and relatively simple protocols. However, two-dimensional screening techniques lack the right microenvironmental cues that lead to altered cell–cell and cell–matrix interactions, cell surface receptor expression, proliferation and aggressiveness (malignancy), thereby providing a poor correlation between two-dimensional and in vivo tumour sensitivity [3–6]. Moreover, it fails to provide hypoxic conditions, typical of tumours grown in vivo [7,8]. On the other hand, in vivo models provide a better match in terms of micro-environment for tumours occurring in the human body. For example, the mouse model has been the most attractive in vivo model for drug screening/testing because the xenografts in mice respond to drugs in a manner similar to that observed in human tumours . Apart from xenografts, other improved tumour models such as orthotopic models, metastatic models and genetically engineered mouse models may emerge as better in vivo systems for drug screening/testing . However, in vivo models can be associated with limitations such as increased duration of experimentation, ethical constraints and costs. Hence, the development of in vitro models that involve simple culturing practices while providing more in vivo-like conditions would be of immense value for screening/testing and for understanding tumour biology. If achieved, this, in many ways, would bridge the currently existing gap between conventional two-dimensional culture and in vivo tumours.
Three-dimensional approach to cell culture is one such strategy that has led to the development of physiologically more relevant tumour models to study tumour biology (including angiogenesis, drug resistance and metastasis) in vitro and perform drug screening/testing assays [11,12]. For example, the three-dimensional culture of cancer cells usually results in nutritional deficiency, leading to the development of hollow core and finally necrosis that mimic the necrotic regions of in vivo tumours . In addition to the physiological closeness of three-dimensional systems to in vivo tumours, they permit easy experimental manipulation (better real time imaging), reduced duration of experiments and reduced costs when compared with in vivo models  while maintaining ease of experimentation like the two-dimensional cultures.
Tissue engineering, a more recent approach, enables the use of three-dimensional scaffold-based cultures for generation of tumour models . Although a few studies have explored the possibility of three-dimensional scaffold-based tumour models, most of them have focused on comparing the three-dimensional scaffold-based system with two-dimensional culture systems or with other three-dimensional culture systems [14–20]. However, in order to establish the ability of a three-dimensional scaffold-based culture model to bridge the gap between conventional two-dimensional culture and in vivo tumours, its comparison with both two-dimensional culture systems and tumours grown in vivo would be more appropriate. Bridging this gap would be further facilitated by providing a near-native extracellular matrix (ECM)-mimicking micro-environment. Therefore, in this study, we hypothesized that a three-dimensional scaffold fabricated using ECM-mimicking materials—chitosan (glycosaminoglycan mimic) and gelatin (collagen mimic)—would facilitate formation of tumours analogously to solid tumours grown in vivo. Because lung cancer is the leading cause of cancer-related deaths throughout the world, this study intended to establish the validity of the chitosan–gelatin (CG) scaffold-based model for a non-small cell lung cancer (NSCLC) cell line, NCI-H460. For this purpose, a comparison between cells grown on two-dimensional substrates, on three-dimensional scaffolds and in vivo was performed for factors involved in tumour-cell–ECM interaction, invasion and metastasis, and drug resistance. Further, this model was analysed for its potential as a universal scaffold for the generation of tumoroid-like structures, using other types of cancer cell lines.
2. Material and methods
Chitosan of low viscosity (less than or equal to 200 mPa.s) and gelatin-type A (300 Bloom viscosity) were procured from Sigma-Aldrich, St Louis, MO, USA and used as received. Genipin was procured from Challenge Bioproducts, Taiwan and used as received. NCI-H460 (NSCLC), HeLa (cervical cancer), MCF-7 (breast cancer) and MG-63 (osteosarcoma) lines were procured from National Center for Cell Science, Pune, India. RPMI-1640 and minimum essential medium (MEM) were procured from Sigma-Aldrich. Foetal bovine serum (FBS) was procured from Invitrogen, Carlsbad, CA, USA. Penicillin-streptomycin, amphotericin, trypsin-EDTA, HEPES buffer and sodium pyruvate were procured from Hi-Media, India. NCI-H460 cells were maintained in RPMI-1640 supplemented with 10 per cent FBS, 10 mM HEPES buffer, 100 mM sodium pyruvate, 1 per cent penicillin-streptomycin, 0.1 per cent amphotericin at 37°C and 5 per cent CO2 in a fully humidified incubator. Other cell lines were maintained in MEM supplemented with 10 per cent FBS, 100 mM sodium pyruvate, 1 per cent penicillin-streptomycin at 37°C and 5 per cent CO2 in a fully humidified incubator.
2.2. Synthesis and characterization of chitosan–gelatin scaffolds
2.2.1. Synthesis of chitosan–gelatin scaffolds
CG scaffolds were synthesized using the freeze-drying technique . Briefly, chitosan and gelatin were dissolved in the ratio of 1 : 2 (w/w) (total solution concentration 3 per cent, total solution volume 50 ml) in 2 per cent glacial acetic acid (w/v), cross-linked using genipin (2.25% w/w of total polymer weight) and allowed to stir at room temperature for 2 h. Stirred CG solution was transferred into 2 ml disposable plastic syringe moulds and incubated at room temperature for 36 h for effective cross-linking. Post incubation, CG solution-containing moulds were transferred to a deep freezer (New Brunswick Scientific) at −80°C for a period of 12 h. The moulds were then lyophilized (Christ-Alpha 1-2 LD) for 36 h to obtain dried CG hydrogel scaffolds. The CG scaffolds were washed with double-distilled water to remove unused genipin or acetate ions, transferred to a deep freezer at −80°C for 2 h and then lyophilized for 24 h. Post lyophilization, the CG scaffolds were stored at room temperature and used for all future experiments. For CG films, following addition of genipin to the CG solution, the solution was poured into 90 mm Petri dishes, incubated at room temperature for 36 h and subjected to drying at 37°C, resulting in the formation of dry CG films.
2.2.2. Characterization of chitosan–gelatin scaffolds
Morphology. Gross morphology of the CG scaffolds was evaluated using scanning electron microscopy (SEM) (FEI Quanta 200). Transverse and longitudinal sections of approximately 1 mm thickness were cut, placed on a copper stub and sputter-coated with gold. The coated samples were imaged at 40× and 200× magnifications at an accelerating voltage of 20 kV and a working distance of 10 mm.
Pore size and porosity. Pore size distribution of the CG scaffolds was evaluated from scanning electron micrographs. Pores were measured in their longest dimension in three different micrographs (transverse sections: 30 pores per image × 3 = 90 pores) using ImageJ software. Average pore size was represented as mean value ± s.d. Mean pore size and porosity of CG scaffolds were evaluated using micro-computerized tomography (Micro-CT) (Skyscan 1076 scanner) at 18.43 µm resolution (n = 3). NRecon program (v. 18.104.22.168) was used for reconstruction using NRecon Server as reconstruction engine (v. 1.6.1). Post reconstruction, CT-analyser program from Skyscan was used to quantify porosity.
Swelling study. Swelling studies were performed in order to assess the water absorption capacity of CG scaffolds. Dry weight of the scaffolds (1 mm thick cylindrical discs) was recorded prior to immersion in 0.1 M phosphate-buffered saline (PBS). The CG scaffolds were then removed at pre-determined time points, gently blotted on a dry filter paper to remove excess surface water and their wet weight was recorded. Swelling ratio (SR) was calculated using the formula  where Hw denotes weight of wet scaffolds, and Hd stands for weight of dry scaffolds.
Degradation study. Degradation study of the CG scaffolds (1 mm thick CG cylindrical discs) was conducted at 37°C in 0.1 M PBS. Dry weight of the hydrogel samples was recorded, and the samples were then immersed in PBS and incubated at 37°C. Samples (n = 3) were then removed at pre-determined time points, rinsed with double-distilled water and lyophilized for 24 h before being subjected to morphological (as explained in ‘Morphology’, §2.2.2) and dry-weight analysis. Weight remaining percentage was then calculated using the formula  where Hd indicates dry weight of the sample before degradation, and Hs represents dry weight of the sample after degradation.
Mechanical testing of CG scaffolds. Mechanical testing studies were conducted using a Bose Electroforce 3220 mechanical testing system with a 225 N load cell and Wintest software. CG scaffolds were pre-wetted in 0.1 M PBS for 30 min followed by testing in 0.1 M PBS at 37°C in order to provide for a near physiological environment. Uniaxial compression was performed, and compressive modulus was calculated as the ratio of stress to strain (from initial linear data).
2.3. Cell seeding on chitosan–gelatin scaffolds
The following common cell-seeding procedure was used for all cell lines. The CG scaffolds were sterilized by serial dilutions of ethanol followed by three washes of sterile PBS (pH 7.4). The scaffolds were preconditioned (for equilibrium swelling kinetics and protein adsorption) by immersion in cell-specific culture media (§2.1) for 12–14 h. After removal of cell-culture media, 1 × 105 cells suspended in 10 µl volume were seeded on each scaffold in an agarose-coated well plate. Cells were allowed to adhere onto the scaffold (2–3 h) prior to the addition of 500 µl of media in each well, with media being replaced every 24 h.
2.4 Characterization of tumoroids generated on chitosan–gelatin scaffolds
2.4.1. Cell proliferation
Cell proliferation was performed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. For proliferation studies, scaffolds were transferred to uncoated 48-well plates and incubated with MTT solution (0.5 mg ml−1) at 37°C for 4 h. Following incubation, MTT solution was gently aspirated, and 0.9 ml of dimethyl sulphoxide (DMSO) was added to each well containing scaffold. DMSO was resuspended several times in order to dissolve the formazan crystals formed within the scaffold, and absorbance of the samples was recorded using a microplate reader (Tecan Sunrise) at 570 nm with a reference wavelength of 650 nm.
The proliferation experiments were performed on 1, 2 and 3 mm thick discs (8 mm diameter). In the first set of experiments, 1 × 105 NCI-H460 cells were seeded on all scaffolds of varying thickness, and MTT assay was performed at the end of day 9. In the second set, 1 × 105 NCI-H460 cells mm−1 were seeded on scaffolds of 1, 2 and 3 mm thickness and MTT assay was performed at the end of day 9 (i.e. normalized for thickness due to difference in surface area and volume). Further, 1 mm scaffolds were seeded with 1 × 105 NCI-H460 cells and proliferation assay was performed with respect to time at the end of day 3, 6 and 9. Similar studies for proliferation were performed for MCF-7, HeLa and MG-63 cells. All absorbance values were converted into cell number, using previously measured standard curves for each cell line.
2.4.2. Morphological analysis
Morphology of tumoroids generated on CG scaffolds was analysed using SEM (NCI-H460, HeLa, MCF-7 and MG-63 cells) and confocal microscopy (NCI-H460). For SEM analysis, the samples were fixed using 2.5 per cent glutaraldehyde (Merck, India) followed by dehydration in ethanol gradient. The samples were then dried and sputter-coated with gold before imaging. For confocal microscopy, the samples (NCI-H460 tumoroids generated on CG scaffolds) were fixed with 4 per cent formaldehyde (Merck) and FITC-phalloidin (Sigma-Aldrich) (a high-affinity probe for F-actin) staining was performed on 200 µm sections. The cells on the scaffolds were counter-stained with propidium iodide (Sigma-Aldrich), mounted using 1,4-diazabicyclo[2.2.2]octane (DABCO, Sigma-Aldrich) and imaged using a confocal microscope (Leica SP5).
2.4.3. Quantification of reactive oxygen species
Conversion of non-fluorescent 2′,7′-dichlorofluorescein diacetate (DCFH-DA) (Sigma-Aldrich) to a fluorescent molecule due to oxidation by reactive oxygen species (ROS) was used to quantify intracellular ROS in NCI-H460 cells grown on two-dimensional tissue culture polystyrene (TCPS) and three-dimensional CG scaffolds. Briefly, 1 × 105 cells were seeded on 1 mm CG scaffolds (discs) and two-dimensional TCPS surfaces. Cells grown on three-dimensional CG scaffolds (end of day 9) and two-dimensional TCPS surfaces (end of day 1 with 60–70% confluency) were washed with PBS (pH 7.4) and incubated with 400 µl serum-free cell culture media supplemented with 50 µM DCFH-DA for 4 h at 37°C. The media were aspirated, and the fluorescent product was harvested in lysis buffer (Sigma Aldrich). Fluorescence was recorded using a Perkin Elmer fluorescence spectrometer at λex 516 and λem 535 nm. The final value of intracellular ROS concentration was normalized with respect to cell number determined by MTT assay.
2.5. Comparison of gene expression of NCI-H460 cells grown on two-dimensional tissue culture polystyrene surfaces, on three-dimensional chitosan–gelatin scaffolds and in vivo
The gene expression levels of fibronectin, integrin αv, integrin β1, epithelial-cadherin (E-cad), neural-cadherin (N-cad) and vimentin were measured using semi-quantitative reverse-transcriptase polymer chain reaction (RT-PCR). Total RNA was isolated using TRI Reagent (Sigma-Aldrich). Total RNA from monolayer cultures was isolated when the tissue culture plates were 60–70% confluent. RNA was isolated from cells grown on three-dimensional scaffolds at the end of day 9. For in vivo samples, 2.5 × 106 cells were injected onto the flanks of six- to eight-week-old nude mice and tumours were resected after three to four weeks. All animal experiments were performed as per the ethical guidelines of Indian Institute of Science, Bangalore, India. The tumour samples were homogenized in TRI Reagent and processed for RNA isolation. Complementary DNA was synthesized using 1.5 µg of total RNA using Gene-Amp RNA PCR complementary DNA synthesis kit (Applied Biosystems, Carlsbad, CA, USA). Primers were designed using Primer 3 online tool. The expression values for N-cad, integrin αv, integrin β1 and vimentin were normalized with respect to β-2 microglobulin (β2M), whereas E-cad and fibronectin were normalized with respect to hypoxanthine phosphoribosyltransferase (HPRT) (house-keeping genes). The sequence of gene-specific primers used is listed in the electronic supplementary material, table S1.
2.6. Immunohistochemistry of NCI-H460 cells grown on two-dimensional tissue culture polystyrene surfaces, on three-dimensional chitosan–gelatin scaffolds and in vivo
Expression levels of p21, β-catenin, vimentin and cytokeratin-18 (CK-18) in NCI-H460 cells grown on two-dimensional TCPS surfaces, on three-dimensional CG scaffolds and in vivo were qualitatively analysed using immunohistochemistry (IHC) experiments. For two-dimensional, cells were grown in 35 mm dishes and fixed with 4 per cent paraformaldehyde (PFA) after they attained 60–70% confluency. Three-dimensional scaffolds containing tumoroids (at the end of day 9) and excised tumours (three to four weeks of cell injection) were fixed with 4 per cent PFA and embedded in optimal cutting temperature (OCT) compound. The scaffold and tumour samples were cut into 10 µm sections using a cryotome (Leica CM 1510S) and placed on silane-coated slides. For staining, cells grown on two-dimensional TCPS surfaces, three-dimensional CG scaffold sections and in vivo sections were washed with PBS (pH 7.4) to remove excess OCT, followed by antigen retrieval with trypsin-EDTA. The samples were immediately washed with PBS, permeabilized with 0.1 per cent Triton-X 100 and blocked with gelatin from cold water fish skin (Sigma-Aldrich), and stained overnight with primary antibody for p21 (Calbiochem) (1 : 200), β-catenin (Abchem) (1 : 100), vimentin (Sigma-Aldrich) (1 : 500) and CK-18 (Santa Cruze) (1 : 100) in separate experiments. Secondary anti-mouse, anti-rabbit antibodies (1 : 1000) (Jacksons ImmunoResearch Laboratories) and DAKO colour development kit were used for colour development. All samples were counterstained using haematoxylin (nuclear stain) and imaged using an Olympus microscope (IX 71).
2.7. In vitro drug response
NCI-H460 cells grown on two-dimensional TCPS surfaces and three-dimensional CG scaffolds were compared for their response to anti-cancer drugs. 1 × 105 cells grown under both two-dimensional (end of day 1 with 60–70% confluency) and three-dimensional (end of day 3, 6 and 9) conditions were treated individually with either 400 nM topotecan, 100 nM paclitaxel or 8 µM oxaliplatin (drug dose determination is based on cell line susceptibility). Viability assay (MTT) was performed at the end of 24 h on all samples, and viability of drug treated samples was normalized with respect to untreated control cells.
2.8. Statistical analysis
All quantitative results were expressed as mean ± s.d. Statistical analysis was performed using one-way analysis of variance (ANOVA) and student's t-test. All values of p < 0.05 were considered to be statistically significant.
3. Results and discussion
3.1. Scaffold synthesis and characterization
CG scaffolds were successfully prepared using freeze-drying technique. Figure 1a shows top and side views of CG scaffolds cross-linked using genipin that imparts a greenish colour to the scaffolds. Transverse and longitudinal sections of CG scaffold (figure 1b,c respectively) show that the scaffold was highly porous with a mean pore size greater than 200 µm. Scaffolds did not exhibit a homogeneous pore distribution (figure 1b) that was attributed to the freeze-drying technique, which enabled smaller pores towards the periphery and larger pores towards the centre. Figure 1c depicts a longitudinal section taken from the central zone, where the pore sizes were relatively larger. Micro-CT of CG scaffold (figure 1d) showed inter-connected porous architecture depicting scaffold walls (blue) and void spaces (green) between the scaffold walls. Porosity of the CG scaffolds, as determined by Micro-CT, was 86.7 ± 0.9%, with an average pore size of 250 ± 23 µm, which has been reported to be sufficient for cellular infiltration and aggregation . Further, the CG scaffolds showed water absorption of nearly 25 times their dry weight at the end of 1 h and obtained about 90 percent of this value in the first 2 min of swelling studies (figure 1e), demonstrating that the scaffolds had good water/media absorption capacity and were well suited for cell-culture studies. In vitro degradation studies demonstrated that the scaffolds maintained 82.8 ± 3.1% weight at the end of day 3, 70 ± 2.9% at the end of day 28 and 60.6 ± 1% at the end of day 56 (figure 2a). This observation was corroborated by their morphological analysis (figure 2b) that demonstrated a minimal change in morphology and architecture over a period of four weeks, thereby indicating their ability to support long-term culture of cells that may facilitate formation of clinically meaningful tumoroids. Further, it has been recently reported that rigid surfaces with a stiffness in the range of 5 kPa support epithelial-to-mesenchymal transition (EMT), a known hallmark of metastasis . The compressive modulus of CG scaffolds was determined to be 8.3 ± 0.7 kPa (see the electronic supplementary material, figure S1), thereby suggesting their possible advantage in promoting EMT of lung cancer cell line NCI-H460.
3.2. Characterization of NCI-H460 tumoroids generated on chitosan–gelatin scaffolds
When scaffolds of different thickness were seeded with the same cell number (1 × 105 cells/scaffold), it was observed that with an increase in scaffold thickness (1, 2, 3 mm), there was a decrease in cell viability at the end of day 9 (figure 3a(i)). Similar results were observed when the cell number was normalized for scaffold thickness, i.e. 1 × 105 cells mm−1 of the scaffold (figure 3a(ii)). In spite of higher initial seeding densities used on 2 mm and 3 mm scaffolds, the cell number at the end of day 9 still remained higher for 1 mm scaffolds and decreased with an increase in scaffold thickness. These results suggest that increasing thickness probably compromised the transport of nutrients and exchange of gases and wastes, and as a consequence, cell viability . From these first set of experiments, it was inferred that the CG scaffolds supported cell attachment and growth, thereby indicating their cytocompatibility. Because 1 mm scaffolds supported maximum cell growth, they were used for subsequent studies.
Further, 1 mm scaffolds were used to study the proliferation and morphology of cells as a function of time. It was observed that there was a significant increase in cell number with increase in time from day 3 to day 9 (figure 3b). NCI-H460 cells, when cultured on CG scaffolds, showed a rounded morphology in the form of tumoroids, whose size increased with an increase in duration of culture (figure 3c). Results from confocal microscopy (figure 3d) analysis demonstrated that NCI-H460 cells used the scaffold walls (pink due to scaffold autofluorescence) as a support structure while the three-dimensional structure and macro-porosity of the scaffold enabled cellular infiltration and aggregation. Two-dimensional films of CG or two-dimensional TCPS surfaces did not facilitate aggregate formation (figure 3c(ii)(iii)). Hence, it can be inferred that three-dimensional CG scaffolds provide a microenvironment that promotes tumoroid formation that was reminiscent of solid tumours in vivo .
ROS levels in cancer cells are elevated when compared with normal cells and are associated with stimulation of cell proliferation, mutations and resistance to chemotherapeutic agents . Hence, as a preliminary indicator, ROS levels were investigated in the tumoroids generated on CG scaffolds and compared with that on two-dimensional surfaces. Tumoroids led to ROS levels that were 22 ± 2.2 (p < 0.005) times higher when compared with those of cells grown on two-dimensional surfaces (figure 3e), and is indicative of a possible advantage of the three-dimensional CG scaffold over two-dimensional TCPS.
3.3. Phenotypic changes in NCI-H460 cells grown on two-dimensional tissue culture polystyrene, on three-dimensional chitosan–gelatin scaffolds and in vivo
3.3.1. Expression of tumour suppressor p21
Because cell-cycle regulators have direct implications on proliferation , cyclin-dependent kinase inhibitor p21 (a negative regulator of cell cycle) was studied. Previous biochemical and genetic studies have demonstrated that p21 is a master regulator of various tumour suppressor pathways for promoting anti-proliferative activities . Nuclear expression of p21 (brown colour in the inset of figure 4a) was observed in few NCI-H460 cells when grown on two-dimensional TCPS surfaces, whereas there was complete downregulation of p21 in cells grown on three-dimensional CG scaffolds and in vivo (figure 4b,c). The absence of p21 expression in cells grown on three-dimensional CG scaffolds and in vivo indicated its reduced anti-proliferative effect when compared with cells grown on two-dimensional TCPS surfaces.
3.3.2. Cell–extracellular matrix interaction
ECM protein fibronectin over-expression is associated with enhanced malignancy of cancer cells and resistance to anti-cancer drugs . Further, fibronectin has been shown to suppress p21 activity (figure 4) through the Erk and Rho signalling mechanism, and, as a consequence, it has been shown to cause higher proliferation in lung cancer cells . Hence, fibronectin expression levels were determined in NCI-H460 cells grown on two-dimensional TCPS surfaces, on three-dimensional CG scaffolds and in vivo. A significant increase in the fibronectin expression was observed in NCI-H460 cells grown on three-dimensional CG scaffolds and in vivo when compared with cells grown on two-dimensional surfaces (figure 5a). Further, there was no significant difference between fibronectin expression in tumoroids grown on scaffolds when compared with tumours formed in vivo, suggesting that the scaffolds maintained the malignant nature of NCI-H460 similar to those grown in vivo. This result, in combination with p21 data, is in agreement with previous literature suggesting that upregulation of fibronectin is accompanied with downregulation of p21 .
Further, it has been shown that ECM proteins (fibronectin, laminin and collagen) enhance tumour invasion and metastasis following interaction with hetero-dimeric integrins expressed on tumour cells. Integrin αvβ1 expression is essential for tumour progression and formation of a more malignant phenotype in lung cancer . Hence, expression levels of integrin αv and β1 were individually studied in NCI-H460 cells grown on two-dimensional TCPS surfaces, on three-dimensional CG scaffolds and in vivo.
The results of integrin αv expression indicated that there was a significant difference in expression levels in NCI-H460 cells when grown on two-dimensional TCPS surface when compared with aggregates formed on three-dimensional CG scaffolds and tumours grown in vivo (figure 5b). Further, it is interesting to note that even though there was a downregulation of integrin αv expression, there was no significant difference in the expression levels in cells grown on three-dimensional CG scaffolds and in vivo. This can probably be attributed to the expression of other types of integrin α subunits apart from integrin αv in NCI-H460 that may be responsible for the enhanced tumorigenicity of these cells .
It has also been reported that lung cancer cells adhere to endothelial cells through the β1 sub-family of integrins , indicating their ability to infiltrate endothelial cells, thereby facilitating metastasis. The β1 sub-family is also known to maintain the motile nature of NSCLC cells by activation of motility-promoting signals involving Fak and Src kinases . The results indicated that there was a significant upregulation in β1 expression in cells grown in vivo when compared with cells grown on two-dimensional TCPS (figure 5c). However, there was no significant difference in levels of β1 expression between cells grown on three-dimensional CG scaffolds and two-dimensional surfaces or tumours grown in vivo. This could probably be because integrin β1 is involved in functions such as early development, differentiation, supra-molecular assembly of ECM proteins and repair of wounded epithelia apart from cancer cell migration [35,36]. Hence, it is speculated that the diversity in function of β1 subunit combined with the change in microenvironment from in vivo to in vitro probably led to downregulation of β1 on two-dimensional surfaces and three-dimensional CG scaffolds.
Thus, results from cell–ECM interaction study indicated that the expression levels of fibronectin and integrin αv on three-dimensional CG scaffolds were not significantly different from those of cells grown in vivo. This further reinforced the similarity in behaviour of NCI-H460 cells grown on three-dimensional CG scaffolds and in vivo in terms of their interaction with ECM.
3.3.3. Invasiveness and metastatic potential
Tumour progression is a multi-step process that involves tumour growth along with cell migration, invasion and metastasis. EMT is a developmental phase associated with initiation of cancer invasion and metastasis . During this transition, epithelial cells lose their characteristics including cell–cell adhesion, planar and apical-basal polarity and motility, and take on a more mesenchymal phenotype. Further, these transformed epithelial cells gain the ability to attach loosely to ECM, thereby increasing the possibility to migrate and metastasize. Downregulation of E-cad, a cell adhesion molecule and an epithelial marker, has been implicated in the process of EMT. NCI-H460, being a malignant cell type, showed no expression of E-cad when grown on two-dimensional TCPS, on three-dimensional CG scaffolds or in vivo (see the electronic supplementary material, figure S2).
The mesenchymal phenotype, on the other hand, is associated with upregulation of N-cad, whose expression promotes a dynamic adhesion state and is associated with EMT , motility , metastasis  and angiogenesis . Further, a few studies have also reported a cadherin switch from E-cad to N-cad during cancer progression . Hence, expression levels of N-cad were studied in NCI-H460 cells grown on two-dimensional TCPS surfaces, on three-dimensional CG scaffolds and in vivo. Results indicated that there was a significant upregulation of N-cad in cells grown in vivo and on three-dimensional CG scaffolds when compared with two-dimensional TCPS surfaces (figure 6a). However, there was no significant difference in expression levels of N-cad in cells grown on three-dimensional CG scaffolds and in vivo, thereby indicating that the CG scaffolds maintained the invasive and metastatic potential of NCI-H460 cells as in vivo.
Vimentin, a class-III intermediate filament, is an important EMT marker and a regulator of mesenchymal cell migration that is associated with increased cell motility, invasive behaviour and poor prognosis in tumours of epithelial origin . Further, overexpression of vimentin is associated with the absence or reduced expression of E-cad, leading to enhanced metastasis . Hence, vimentin expression was studied in NCI-H460 cells grown on two-dimensional TCPS surfaces, on three-dimensional CG scaffolds and in vivo. It was observed that there was a significant upregulation of vimentin in cells grown in vivo when compared with two-dimensional TCPS surfaces (figure 6b). On the other hand, cells grown on three-dimensional CG scaffolds did not show any significant difference in vimentin expression when compared with tumours grown in vivo. Although there was no significant difference between two-dimensional TCPS and three-dimensional CG scaffolds, IHC studies confirmed higher vimentin expression in NCI-H460 cells grown on three-dimensional CG scaffolds (figure 6c(ii); similar to tumours grown in vivo—figure 6c(iii)) when compared with those grown on two-dimensional TCPS surfaces (figure 6c(i)), as indicated by the darker brown colour. Further, it is important to note that the brown colour in figure 6c(ii) is as a result of vimentin expression in NCI-H460 cells grown on three-dimensional CG scaffolds and not an artefact, as there was absence of brown colour in negative control (see the electronic supplementary material, figure S3). Thus, IHC data suggested that metastatic potential of NCI-H460 cells was maintained when grown on three-dimensional CG scaffolds.
Another class of intermediate filaments, cytokeratins (CK), are associated with migration and metastatic ability of cancer cells through interactions with ECM . Further, co-expression of vimentin along with CK-8/18 has been implicated in increased tumour invasion and metastasis in various cancers . Hence, as a follow-up to vimentin, CK-18 expression was studied in NCI-H460 cells when grown on two-dimensional TCPS surfaces, on three-dimensional CG scaffolds and in vivo. It was demonstrated that CK-18 expression was elevated in both NCI-H460 cells grown on three-dimensional CG scaffolds (figure 6d(ii)) and in vivo (figure 6d(iii)) when compared with cells grown on two-dimensional TCPS surfaces (figure 6d(i)). The elevated levels of both vimentin and CK-18 in cells grown on three-dimensional CG scaffolds and in vivo were in concurrence with previous literature [43–46] and indicated that cells grown on three-dimensional CG scaffolds maintained their invasive and metastatic ability.
In addition to the aforementioned EMT marker, Wnt/β-catenin signalling pathway is a well-defined oncogenic pathway, whose activation in lung cancer lines has been implicated in increased metastasis . Further, it has been suggested that increased β-catenin in the nucleus leads to reduction in adherens junctions causing decreased stability and degradation of E-cad which in turn promotes increased metastasis in lung tumours . Hence, expression of β-catenin in NCI-H460 cells grown on two-dimensional TCPS, on three-dimensional CG scaffolds and in vivo was also studied. Cells grown on three-dimensional CG scaffolds (figure 6e(ii)) showed higher expression of β-catenin when compared with those grown on two-dimensional TCPS surfaces (figure 6e(i)). Further, the expression levels of β-catenin in tumoroids were similar to those grown in vivo (figure 6e(iii)). Increased nuclear expression of β-catenin in NCI-H460 cells grown on three-dimensional CG scaffolds with a concomitant downregulation of E-cadherin (see the electronic supplementary material, figure S1) suggests the possibility of maintenance of metastatic potential.
Thus, it can be concluded that the three-dimensional CG scaffolds provide an environment that maintains the metastatic and invasive potential of NCI-H460 tumoroids with characteristics similar to tumours grown in vivo.
3.4. Drug responsiveness of tumoroids generated on the chitosan–gelatin scaffolds
Cancer cells grown using three-dimensional culture systems have been reported to show an altered response to chemotherapeutic agents when compared with those grown on two-dimensional TCPS surfaces [18,49]. Hence, NCI-H460 cells grown on two-dimensional TCPS surfaces and three-dimensional CG scaffolds were treated with chemotherapeutic agents that have been previously explored for lung cancer treatment, namely, topotecan, paclitaxel and oxaliplatin. It was observed that NCI-H460 cells cultured on two-dimensional surfaces were sensitive to these agents and showed a significant drop in cell viability when treated with either of the chemotherapeutic agents. In contrast, tumoroids generated on three-dimensional CG scaffolds (at the end of day 3, 6 and 9) when treated with these chemotherapeutic agents showed no significant difference in viability when compared with untreated counterparts (see figure 7 and the electronic supplementary material, figure S4). This observation is in concurrence with previously reported literature wherein it was demonstrated that cells grown in three-dimensional culture systems showed reduced susceptibility to chemotherapeutic agents when compared with those grown on two-dimensional surfaces [15,18,50]. A possible explanation for the reduced sensitivity of tumoroids to the chemotherapeutic agents is upregulation of multiple genes involved in drug resistance and the diffusional limitations associated with the tumoroids unlike the uniform distribution of drugs in two-dimensional culture systems. Cytokeratins (including CK18) are responsible for cellular protection against any kind of mechanical and chemical (chemotherapeutic drugs) stresses that induce cellular death [51,52]. NCI-H460 cells showed higher expression of CK-18 when seeded on CG scaffolds when compared with those cultured on two-dimensional surfaces, thereby conferring these cells with greater resistance to chemotherapeutic agents (figure 6d). Further, higher expression levels of ECM proteins such as fibronectin in cells grown on CG scaffolds (figure 5a) induce tyrosine phosphorylation, which blocks caspase activation, thereby preventing chemotherapeutic agent induced apoptosis . In addition, elevated levels of ROS have also been implicated in drug resistance . Increased ROS levels in cells grown on CG scaffolds (figure 3e) when compared with those grown on two-dimensional TCPS surfaces further corroborate the results obtained from CK-18 and fibronectin upregulation. Taken together, cells grown on three-dimensional CG scaffolds form tumoroids that demonstrated increased drug resistance and hence three-dimensional CG scaffolds show potential for use in drug-screening studies.
3.5. Generation of tumoroids of different cancer cells on chitosan–gelatin scaffolds
In this study, the CG scaffolds were explored for their ability to support tumoroid formation, using cancer cells of different origin. MCF-7, HeLa and MG-63 were chosen as MCF-7 and HeLa are cancers of epithelial origin and MG-63 is a cancer of connective tissue. Further, breast cancer and cervical cancer are among the top five in terms of estimated new cases in many countries, including developed countries such as the USA  and developing countries such as India . Although osteosarcomas do not have high occurrence rates like the cancers of lung, breast and cervix, the rationale for choosing MG-63 was its difference in origin. The results indicated that CG scaffolds supported the growth of all three cell lines. As observed for lung cancer cell line NCI-H460, MCF-7, HeLa and MG-63, when seeded on CG scaffolds, were able to generate tumoroids, the size of which increased with increase in incubation time (figure 8a(i)–(iii),b(i)–(iii),c(i)–(iii)). Similar to NCI-H460, tumoroid formation was not supported on two-dimensional TCPS surfaces for all three cell lines (figure 8a(iv),b(iv),c(iv)). Further, proliferation of the cells on CG scaffolds was investigated as a function of time, and it was observed that there was an increase in cell number with increase in time from day 3 to day 9 (figure 8d–f). From this study, it could be inferred that CG scaffolds support cellular aggregate formation of cancer cells belonging to different origin. While tumoroids of MCF-7, HeLa and MG-63 cells on CG scaffolds remain to be established for various factors involved in tumour growth, metastasis and drug responsiveness, the ability to form tumoroids is indicative of their possible promise and use as universal scaffolding system for in vitro tumoroid formation.
Three-dimensional culture systems, owing to their physiological relevance, have been strongly pursued in the area of tissue engineering and have resulted in the development of successful tissue-engineered grafts. Further, principles of tissue engineering can be applied towards the development of three-dimensional cancer models that are relevant for studying tumour biology and drug screening. In this work, three-dimensional CG scaffolds having an inter-connected pore structure were fabricated and evaluated for their potential as ECM-mimicking structures that support the formation of cancer cell line-based tumoroids in vitro. The work largely focused on studying a lung cancer cell line, NCI-H460, to understand various aspects of tumour growth, metastasis and drug resistance. Unlike when grown on two-dimensional surfaces (TCPS and CG films), NCI-H460 cells, when seeded on three-dimensional CG scaffolds, generated tumoroids that were reminiscent of solid tumours grown in vivo. These tumoroids showed gene-expression profiles similar to tumours grown in vivo for genes involved in tumour growth (p21); tumour-cell–ECM interaction (fibronectin and integrin αv); tumour invasion and metastasis (N-cad, vimentin, CK-18 and β-catenin); and drug resistance (fibronectin and CK-18). Further, these tumoroids showed a significant increase in intra-cellular ROS levels, which has been implicated in enhanced tumour proliferation and drug resistance. Results from the drug response study demonstrated reduced susceptibility of tumoroids when compared with those grown on two-dimensional surfaces. This observation is in concurrence with the upregulation of factors responsible for drug resistance (ROS, fibronectin and CK-18) in this study, as well as previously reported studies [15,18,49].
The results using NCI-H460 and ECM-mimicking chemistry prompted the possibility of exploring the CG scaffold system as a universal three-dimensional culture model. It was demonstrated that the CG scaffold system supported the growth of various cancer cells (carcinomas and sarcomas) and facilitated the formation of tumoroids similar to those observed for NCI-H460. Overall, these results demonstrated that the CG scaffolds show potential to be developed as a three-dimensional cancer model for studying various facets of tumorigenesis and drug testing/screening. It is expected that such three-dimensional scaffold-based tumour models will bridge the gap between the currently prevalent two-dimensional culture and in vivo small animal models, and pave the way for a more efficient and clinically relevant in vitro tumour model system.
N.A. and M.S. acknowledge Council of Scientific and Industrial Research (CSIR), India for financial support, and V.S. is grateful to Indian Institute of Technology, Kanpur (IIT-K) for financial support. D.S.K. is a ‘Class-of-1970 Research Fellow’ and acknowledges Department of Biotechnology (DBT), India for research funding, Dr Naibedya Chattopadhyay, Central Drug Research Institute, Lucknow, India for Micro-CT facility and Cipla, India for gift sample of anti-cancer drugs. A.R. is a Wellcome-DBT India Alliance Senior Research Fellow and acknowledges DBT and Indian Institute of Science, Bangalore (IISc) for research funding, and the IRIS imaging facility and animal facility at IISc, Bangalore.
- Received July 17, 2012.
- Accepted August 20, 2012.
- This journal is © 2012 The Royal Society