Recent advances in 3D models of tumor invasion

Abstract

This review presents recent advances in the design of in vitro cancer models to study tumor cell migration, metastasis, and invasion in three-dimensions (3D). These cancer models are divided into two categories based on the biophysiological processes and structures simulated, namely (i) spheroid invasion models or (ii) vascularization models. Some recent advances to spheroid invasion models include new methods to make them amenable to high-throughput settings. In vascularization models, cancer cell extravasation, intravasation, and angiogenesis have been emulated. Finally, 3D bioprinting and microfluidic technologies are allowing researchers to recapitulate some of the complex architectural and microenvironmental changes that can drive cancer cells migration from the extracellular matrix and basement membrane to blood vessels.

1. Introduction

Despite the tremendous amount of time and money spent on submitting new oncology drugs to clinical trials, most drug compounds fail during Phase III trials, and it is estimated that only 3.4% of investigational compounds obtain FDA approval [1]. One reason for this failure may lie in the discrepancy between the in vivo physiological environment and conventionally utilized, and oversimplified, two-dimensional (2D) platforms that are used for in vitro drug testing [2,3]. Currently, conventional 2D monolayer cultures lack some key features. First, they do not allow tumor cells to interact with cells in other tissues or organs. Next, they lack the matrix surrounding the 3D tumor microenvironment (TME). Finally, they do not possess the ability to introduce flow of blood or other fluids [2,3]. By developing preclinical testing models that more accurately replicate the characteristics of human tissue and cancer microenvironments, the total cost and time spent on the drug discovery process could be reduced and more drugs may successfully pass clinical trials.

During preclinical testing, the tumorigenicity of cells is often assessed based on tumor properties, which include increased levels of proliferation, invasiveness or other migration properties, and resistance to therapeutic treatments. These properties can be quantitatively measured using various biological assays, including immunofluorescence, ELISA, flow cytometry, western blotting, and qPCR techniques. However, most of these techniques are analyzed at single time points and can require intensive labor, sample preparation time, and expensive biochemical reagents. As a result, interest has grown in developing live-cell invasion and migration assays that allow cells to be monitored in real time in the presence or absence of treatments. Monitoring invasion and migration of cells can be done cost-effectively, since data can be collected from bright-field images using common laboratory optical microscopes [4]. Migration and invasion are involved in every step of cancer progression and many assays have been developed and are routinely used to evaluate 2D migration or unidirectional invasion, including trans-well migration assays, wound healing assays, cell exclusion zone assays, and time-lapse cell tracking [4]. However, more recent trends have focused on 3D models of invasion, but these are limited in number due to the time required for sample preparation, the challenges of 3D imaging, and the requirement for automation in high-throughput screening.

To address some of these limitations, advances in materials fabrication techniques (e.g., microfluidic devices, 3D bioprinting, and high-throughput robotic systems) are supporting the development of sophisticated 3D cancer invasion models. These models recapitulate biologically relevant tumor architectures and enable control of the 3D microenvironment to emulate changes that occur during cancer progression (Figure 1). For various high-throughput applications, strategies have been adopted, including uniform spheroid formation [5–8], automation in imaging [9–12], and improved microfluidic chip design [13,14]. In this review, we decided to focus on 3D tumor models that allow the study of the invasiveness of cancer cells, especially spheroid and vascularization models, and we highlight articles published within the last two years.

Figure 1.

Schematic illustration of cancer cell migration and invasion in the human body created with BioRender.com. In vivo tumor invasion during cancer cell growth, intravasation, extravasation, and metastatic development to distant organs can be reconstituted with in vitro tumor invasion models. High-throughput systems, 3D bioprinting, and microfluidics technology facilitate the development of a complex 3D (a) spheroid invasion model or (b) vascularization model. The fluorescence photographs of a spheroid invasion model and a vascularization model utilized in this illustration were reproduced with permission from Refs [15] and [16], respectively.

2. Spheroid invasion models.

Spheroid-based in vitro tumor models are frequently used to observe and monitor the real-time invasiveness of cancer cells. Spheroids are conventionally prepared using cell aggregation techniques. These in vitro methods include the hanging drop method, cell pelleting, spinner flasks and rotating wall vessels, and the liquid overlay method [17,18]. Cells in spheroids experience multiple physiologically relevant tumors characteristics, such as nutrient and oxygen gradients [19,20]. By embedding spheroids in various extracellular matrix (ECM)-derived materials (e.g., collagen or Matrigel) or in ECM mimetic matrices, the influence of the tumor microenvironment (TME) on cancer cell phenotype and therapeutic efficacy can be studied [21]. However, there are technical limitations to using conventional spheroid fabrication methods for high-throughput screening (HTS) applications [19]. First, the process of culturing and collecting spheroids is often labor intensive and time consuming, requiring several days. Second, cancer cell spheroids prepared with the above-mentioned methods are highly heterogeneous in size and shape, making data normalization and interpretation complicated [5]. Further, spheroids embedded in 3D matrices are located randomly throughout the 3D space, which makes imaging less efficient and decreases reproducibility [9].

To overcome some of these limitations and to meet the demands of biopharmaceutical industries to screen large amounts of putative drugs [22], recent reports are addressing some of the key design issues to create advanced spheroid invasion models that can be used in high-throughput settings. A recent article from the Burdick group offers one such example [23]. Daly et al. used self-healing hydrogels to pattern spheroids in 3D spaces at high resolution. In their cardiac disease model, spheroids composed of induced pluripotent stem cell-derived cardiomyocytes (iPSC-CM) and primary human cardiac fibroblasts (CF) were bioprinted at varied ratios (4:1 iPSC-CMs to CFs for “healthy”; 1:4 iPSC-CMs to CFs for “scarred”) in shear-thinning hydrogels. After injection of the spheroids, the hydrogels recovered through their self-healing properties and held the spheroids in place (Figure 2a, 2b). When multiple spheroids were printed in close proximity, the viscoelastic nature of the hydrogels allowed the fusion of adjacent spheroids. For example, when the spheroids were patterned into rings at a separation distance of 50 μm, spheroid fusion and the formation of more complex microstructures was observed over four days of culture (Figure 2c).

Figure 2.

Advanced spheroid invasion models using functional materials, 3D bioprinting technology, and high-throughput platform. (a,b,c) Spheroids prepared in self-healing hydrogels Figures were reproduced with permission from Ref. [23]. (a) Schematics showing reversible interactions between guest (adamantane, blue) and host (β-cyclodextrin, orange) modified hyaluronic acid. (b) MSC spheroid aspiration, transfer into a self-healing support hydrogel, and deposition within the support hydrogel. (c) Brightfield and fluorescence images of spheroid fusion into microtissue rings over the four days of the culture. All scalebars were 200 μm. (d) Schematic illustration showing 3D bio-dot printing process with in-situ formation of cell spheroids. Figures are reproduced with permission from Ref. [16]. (e) Process diagram of the tumor-tissue invasion model and fabrication system setup. Tumor cells suspended with Oligomer solution are pipetted onto the posts of the fabrication platform, and then posts are covered with a 96-well plate, inverted, and incubated. After tumor compartment polymerization, a 96-well plate is inverted, and tissue compartments are polymerized to form an array of 3D embedded spheroids. Figures were reproduced with permission from Ref. [10]. (f) Schematic diagram of the procedure for patterning multicellular spheroids in the 3D collagen matrix. Spheroids were formed on a PDMS mold and covered with collagen gels. After the solidification of the collagen, the array is inverted and covered with a second collagen. Figures were reproduced with permission from Ref. [27].

In a second example, Jeon et al. developed a 3D bio-dot printing technique to precisely locate spheroids in alginate hydrogels [16]. In this model, two types of bioinks were used: a matrix ink and a sacrificial bio-ink. The matrix ink was composed of alginate that could be selectively crosslinked with CaCl₂ solution. During culture at 37°C, the sacrificial bio-ink, which was composed of gelatin and hyaluronic acid, was dissolved and removed (Figure 2d). Once the sacrificial bio-ink dissolved, cells were further clustered in the empty space inside the matrix bio-ink and formed cohesive spheroids. Most importantly, 3D invasion of these spheroids by alginate lyase treatment was observed. For example, when MDAMB231 and NIH3T3 cells were printed, the invasion of both cells could be visualized, and the accelerated invasion of MDAMB231 cells was observed when co-cultured with NIH3T3 cells.

Other recent articles have developed time-saving methods for spheroid preparation, maturation, collection, and embedment into 3D ECM microenvironments [6–8,10–12,24]. For example, Puls et al. developed a tumor invasion model for HTS using a custom-designed fabrication platform [10]. This platform allowed efficient preparation and used an adapted hanging drop method to culture cells in 3D, directly allocating each spheroid into a 96-well plate after spheroid formation (Figure 2e). Magnetic nanoparticles have also been used to create a large number of uniform spheroids efficiently [7,8]. For example, Perez et al. reported a method to reduce the maturation time required for cohesive spheroid formation [7]. In this work, iron oxide nanoparticles, average diameter 8 nm, were produced by an iron salt co-precipitation method. CT26 murine colon carcinoma cells spontaneously up-took these nanoparticles after overnight culture. The iron oxide-loaded cells were then seeded into wells of defined diameters, which were created using steel beads to promote magnetic aggregation of the cells. After 24 hours, the cells formed cohesive, mature spheroids that could be removed and embedded in agarose gels using magnetic force. As a proof-of-concept, doxorubicin penetration, the relationship between spheroid size and invasion, and the relationship between spheroid maturation and invasion were all measured to demonstrate the application of these spheroids in drug screening and toxicity assessments. Overall, reducing the time required to prepare spheroids allowed for high-throughput sample preparation.

Utilizing a soft material approach, Kuo et al, use hydrophobic surfaces to promote high-throughput spheroid assembly [6,25]. Specifically, photo-polymerized microspheres were prepared with hydrophobic surfaces and cell development into spheroids was monitored on these materials [25]. Antunes et al. and Kim et al. used hydrophobic PDMS surfaces to generate cell droplets and formed spheroids by using the hanging drop method [6,26]. The initial study transferred the spheroids to a 2D substrate by inverting the PDMS array and subsequently monitored invasion. However, Ma et al. applied the method to study 3D invasion [27] by embedding the aggregates formed on the hydrophobic PDMS array into a 3D collagen gel (Figure 2f).

Many previous studies (including those mentioned above) have used versatile methods to generate spheroids by cell fusion. In these examples, cells do not directly invade the surrounding matrix, but rather, cells from one spheroid invade and relocate into adjacent spheroids [8,23,28,29]. Using these methods, complex biophysical architectures have been reconstructed in vitro. For example, Zhao et al. [28] reported that spheroids of tumor and fibroblast cells could be combined in a well-defined manner using two microwell array microfluidic devices. In this study, high-content imaging of the paired spheroids showed a core-shell-like structure and revealed that tumor cells tend to envelop fibroblast spheroids. Similarly, Cui et al. reported high-throughput methods for fusing spheroids using a droplet merging methods, modified from the traditional hanging drop method [29]. For application in high-throughput platforms, the researchers then developed a droplet microarray (DMA) consisting of 14 × 14 hydrophilic spots divided by hydrophobic barriers. Spheroid shape and size created by the DMA was controlled with the hydrophilic pattern of the spots and spheroids in neighboring spots were merged by adding more medium with the dispenser. Using this method, 2–4 spheroids were merged to form a programmed architecture.

3. Vascularization models

Primary tumor cells migrate, intravasate or extravasate through blood or lymphatic vessels, and develop metastases in distant tissues, such as brain or bone tissue. Endothelial cells from blood vessels can also migrate and sprout to form new blood vessels via angiogenesis to supply nutrients and oxygen to the tumor site. During these processes, proangiogenic factors, such as vascular endothelial growth factors (VEGFs), are secreted by cancer cells to recruit many other cell types to the tumor site [30,31]. As cancer cells migrate through basement membranes and blood vessels, they encounter different interfaces and are exposed to microenvironments that are biochemically and mechanically diverse and can change with time [32]. Tumor-on-chip models using microfluidic systems (e.g., microphysiological systems) have been developed to recapitulate these complex structures and interfaces. These devices integrate multiple cell types, heterogeneous TME, and micro-scale fluid flow [32–36]. Current trends in this field include the development of multicellular tumor-on-chip systems that are amendable to high-throughput analyses. Advances in fabrication of these devices exploit 3D bioprinting and stereolithography to replace or use in conjunction with conventional microfabrication methods that often rely on soft lithography and/or photolithography [13,33,36].

In one recent example, microfluidic tumor-on-chip models were designed to accommodate more than two types of cells in various extracellular matrices by the use of three channels, three layers, and/or an open-top chamber design [14,37,40]. For example Naguraju et al. and Chi et al. developed three-layer and three-channel microfluidic models composed of cancer cell regions, porous empty regions, and stromal cell regions [14,37] In the Nagaraju et al. tumor model, a microfluidic device was designed with three channels to introduce a tumor, vasculature, and a stromal region [37]. The tumor channel was filled with breast cancer cells (MDA-MB0231 or MCF7) in collagen, the stromal region was filled with collagen, and the vascular region was composed of endothelial cells (HUVECs) in fibrin (Figure 3a–i). The intravasation of cancer cells into the vascular network was recapitulated using this model and enhanced invasion of MDA-MB-231 cancer cells into the endothelial vascular network was observed. Meanwhile, Chi et al. reported on a similar three-layer model in which the three compartments were layered vertically (Figure 3a–ii) [14]. The authors developed an ‘L‐ TumorChip’ system that incorporated microvasculature and tumor-stromal interfaces, which also allowed for HTS. In the study, triple‐ negative breast cancer cells (TNBCs) were cultured with normal fibroblasts, mesenchymal stem cells, and cancer‐ associated fibroblasts (CAF) to investigate how different stromal cells might affect therapeutic treatment and cancer malignancy. In the presence of CAFs, drug pharmacokinetics were delayed and TNBCs had a higher drug resistance. In contrast, the apoptotic response measured by Caspase-3 activity was higher when TNBCs were co-cultured with normal fibroblasts.

Figure 3.

Microfluidics-based vascularization models and 3D bioprinting-based vascularization models. (a) Multicellular microfluidic vascularization models. (i) Schematic design and a photograph of the three‐channel microfluidic models reproduced with permission of Ref [37]. The cancer cells are labeled with red fluorescence to distinguish them from the non-fluorescent HUVECs. The photograph images show cancer cells invasion and vascular network formation after 6 days. (ii) Design of the three‐layer microfluidic device used in Ref [14]. The figure is reproduced with permission of Ref [14]. (b) Vascularized solid tumor-on-a-chip for testing the efficacy and vascular toxicity of chemotherapy. Reproduced with permission of Ref [15]. (i) Multicellular cancer spheroids, lung fibroblast and endothelial cells are cocultured to mimic a lung tumor architecture. (ii) Fluorescence micrographs of the cancer spheroids composed of A549 cells (green) and RFP-HUVECs (red) on day 1, 3 and 7. Scale bar, 50 μm. (iii) A micrograph of a vascularized tumor spheroid perfused with fluorescent microbeads (blue). Scale bar, 50 μm. (iv) Paclitaxel treatment for 2 days leads to significant reduction in the viability of cancer cells. Scale bar, 20 μm. (c) 3D bioprinted metastatic tumor model, reproduced with permission from Ref [38]. (i) Photo of a 3D printed culture chamber. (ii) Fluorescence images of a metastatic model on days 3, 6, 9 and 12, showing that A549s approach and enter the vasculature. (d) 3D printed breast cancer and bone model, reproduced with permission of Ref [39]. (i) Schematic of printed breast cancer model. (ii) Development of breast cancer cells metastasis toward bone over the culture period. B: bone tissue, V: Vessel, T: tumor tissue. The yellow arrows indicate the migration of invasive breast cancer cells.

The cancer angiogenesis-on-a-chip model developed by Paek et al. and Lee et al. combined both spheroid and vascularization models [15,40]. Specifically, Paek et al. developed a perfusable vascularization model based on microfluidic devices with an open-top chamber design (Figure 3b) [40]. This design allowed the culture of other cell types (e.g., endothelial cells), increased media supply, and increased transportation of nutrients, oxygen, and soluble factors. Using this design, a triculture system was developed as a human lung cancer model, which used 3D tumor spheroids of lung adenocarcinoma cells (A549), endothelial cells (HUVECs), and lung fibroblasts. In the Lee et al. study, nanomedicines based on siRNA (siVEGF or siVEGFR) were delivered via mesoporous silica nanoparticles (MSNs) to target VEGF signaling and to demonstrate anti-angiogenic potential. When fibroblasts, tumor cells, and endothelial cells were separated into different channels of the microfluidic chip, EC sprouted toward the cancer cells, but only in the present of the fibroblasts. Unlike 2D invasion models, this model showed directional EC sprouting and angiogenesis.

Of further note, there have been multiple papers that have fully or partially fabricated a microfluidic-based tumor-on-chip model using 3D printing technologies [38,41]. Whereas common methods for preparing microfluidic devices use microfabrication-based techniques, Meng et al. took advantage of 3D bioprinting technology to create a chamber to induce the vascularization process (Figure 3c) [38]. The 3D printing allows custom-designed and controlled spatial arrangement of each biological element. For example, epidermal growth factor (EGF) capsules that continuously EGF gradients have been generated using 3D printing and to study how EGF affects invasion. Similarly, Ouyang et al. used 3D printing technology to generate microfluidic channels for a porous endothelialized vascular structure [41]. In their approach, sacrificial templating bioinks based on gelatin and matrix bioinks based on photocrosslinkable gelatin methacroyloyl (GelMA) were printed side by side. After the GelMA was polymerized using UV light, the sacrificial templating gelatin bioink was removed by simple incubation at 37°C. The sacrificial templating ink improved structural fidelity of the final porous interconnected structures, which allows in-situ cell encapsulation without significant deformation or collapse of the 3D structure. While the invasion and vascularization of endothelial cells were not studied in this research, the model has potential for future application as a tumor vascularization model.

There are also many vascularized tumor models prepared with 3D printing technology that are not based on microfluidics [39,42]. For example, Cui et al. developed a breast cancer cell bone metastasis model with porous scaffolds using a 3D bioprinting technology without any adaptation of current microfluidics systems [39]. To form a vascularized bone and tumor matrix, nano-porous hydrogels were printed with a bio-composite ink made of a photo-crosslinkable GelMA ink, a poly(ethylene glycol) diacrylate (PEGDA) ink, and nanohydroxyapatites. The vessel-like structure was printed with GelMA ink between the bone and tumor matrix (Figure 3d–i), and human fetal osteoblasts and breast cancer cells were then seeded onto the bone and tumor matrix. Endothelial cells were then injected into the vessel channel; however, the model did not include flow. The authors observed faster migration of MDA-MB-231 cells in the presence of endothelial cells, as well as faster migration of MDA-MB-231 cells compared to MCF-7 cells toward the vascularized regions (Figure 3d–ii).

Microfluidic-based vascularization models can be applied to replicate aspects of human physiology. For example, the effects of immune cells on extravasation have also been studied [43–45]. Calleja et al. cultured endothelial cells, fibroblasts, and monocytes in a microfluidic vascularization model and observed that monocytes slowly transmigrated, similar to an in vivo extravasation process [43]. Using fluorescence-activated cell sorting (FACS), they also found that 84±7% of the inflammatory monocytes extravasated, while only 12±10% of the patrolling monocytes transmigrated. When analyzing cell displacement and speed, patrolling monocytes had a much slower migration compared to the migration of inflammatory monocytes. Further, after extravasation, most monocytes migrated slowly, changed from a spherical shape to an elongated shape with more protrusions, and exhibited changes toward a macrophage-like phenotype (e.g., higher CD206 expression).

4. Conclusion

Recent advances in in vitro models to study cancer invasion in 3D were discussed with a particular focus on two categories, namely spheroid invasion models and vascularization models. In spheroid invasion models, microfabrication and 3D printing techniques have been developed to save time, cost, and labor for preparing, maturing, imaging, and analyzing spheroids. In vascularization models, complex interfaces between multiple cell types embedded in various microenvironments can be prepared and cell regions can be separated by using multiple channels. When fabricating microfluidic or microstructured devices, many manufacturing techniques are used, including conventional lithographic techniques, novel 3D printing techniques, advanced hanging drop assay techniques, and automated spheroid generation techniques. While most tumor models developed have relied on naturally derived materials, such as collagen, Matrigel, and alginate, more recent studies have begun to utilize functional, synthetic biomaterials (e.g., photo-polymerizable, shear-thinning, thermo-responsive polymers) to achieve desired properties. Further, significant improvements in automation and high-throughput development have been applied to generate spheroid invasion models, but fewer studies have been published in automated microfluidic-vascularization models. Of note, some new approaches are employing 3D bioprinting technology to create vascularization models, but future directions may need to further address some of the complication of various biological, biochemical, and imaging assays when applied to 3D spheroid and vascularization models.