Models to Study Cancer Invasion and Intravasation

To date, there have been numerous in vivo animal models and in vitro assays to study tumor cell invasion and intravasation. Here, we present a brief overview of different experimental models, which have been utilized to investigate invasion and tumor endothelial interactions, while providing a comparison of published in vivo and in vitro experimental findings.

1.4.1. In vivo models:

Tumor initiation is an unscheduled hyper proliferation of cells, which occurs due to activation of either cell growth machinery or signaling pathways (Hanahan and Weinberg 2000). Throughout the process of cancer progression, there is a bi-directional relationship between tumor and its host TME (Hanahan and Weinberg 2011). In vivo studies have been very valuable to provide a comprehensive overview on the molecular and cellular basis of disease progression (Mak, Evaniew, and Ghert 2014). However, it is complicated to use in vivo models to dissect specific cell-cell interactions, and determine the cause and effect relationships as it challenging capture dynamic interactions and adaptive responses of cancer cells (Mak, Evaniew, and Ghert 2014). Additionally, visualizing the process in real-time may not be easily feasible, and it might perturb the pathophysiology of tumors as it requires specialized invasive steps. The other major challenge is using a genetically engineered mouse to incorporate human cancer cell lines, as there will be notable differences in histology (Brown et al. 2010). When using immune compromised mouse models, the major component in the TME like the immune or fibroblast cells and their role cannot be studied. In general, tumor cell intravasation in in vivo models is analyzed using various methods including:

1. Real-time imaging of tumor cells
2. Measuring the number of circulating tumor cells in the blood stream
3. Examining the tumor injected at the primary site and the tumor at metastatic site.

In a model developed by Xiao et al. (Xiao et al. 2015), a ChorioAllantoic Membrane (CAM) model was used to study the growth, invasion, neoangiogenesis, and metastasis by transplanting NasoPharyngeal Carcinoma (NPC) in to chick embryo. This model closely simulated the growth of carcinoma and elucidated the mechanism of invasion of NPC cells. This also helped analyzing tumor angiogenesis, intravasation and metastasis. Tumor invasion was assessed by detecting the extent penetration of basement membrane by cancer cells. Angiogenesis was studied by analyzing the area of formed neovascular networks, while metastasis was quantified by counting the number of tumor cells in distant organs. The results showed tumor formation after inoculating the CAM with NPC cells. Demonstrated the feasibility of applying the CAM model to visualize and evaluate tumor growth, invasion, and tumor angiogenesis. The limitation of this model is that the CAM model is naturally immunodeficient, hence it cannot be used to study the role of immune cells during metastasis (Figure 1.2 a).

Alternatively, Zebra fish models have been used for real time imaging of injected human cancer cells. The vascular system of the zebra fish is fully functional, which allows better understanding of cancer cell invasion and metastatic profile (Stoletov et al. 2010). This study showed dynamic extravasation process of cancer cells into the surrounding vasculature of zebrafish using intravital imaging.

Figure 1.2: Examples of in vivo models to study intravasation (a) (i) Chorioallantoic membranes (CAMs) inoculated with NPC cells (ii) 3D images showing invasion of cancer cells through the basement membrane. Adapted from Xiao et al. Plos one copyright (2015) (Xiao et al. 2015). (b) (i) Fluorescence images of a zebrafish embryo at tumor cell injection site (ii) confocal images of tumor cells either inside or outside (extravasated) the vessel lumen. Scale bars: 200 μm. Adapted from Stoletov et al. with permission from Company of Biologists [Journal of Cell Science], copyright (2010) (Stoletov et al. 2010).

Using real-time intravital imaging, the authors demonstrated that extravasation of cancer cells is a dynamic process that involving modulation of endothelial layer by tumor cells. Additionally, the findings showed that tumor cells induce vessel remodeling rather than damaging or inducing vascular leakage at the site of extravasation. Locomotion of tumor cells did not depend on the direction of blood flow but was dependent on β1-integrin-mediated adhesion to the blood vessel walls. Although this model is useful for understanding the biological process of extravasation, some of its major drawbacks include (1) The inability to visualize extravasation of tumor cells in real time, (2) The findings should be confirmed using mammalian models before using them in human clinical therapies (Figure 1.2 b).

1.4.2. In vitro models:

Along with in vivo models, in vitro models have been also proven to be more valuable for investigating the cellular interactions and visualizing real time the dynamic interactions between the tumor and its surrounding microenvironment. The major advantages using in vitro models over in vivo models could be summarized to: high throughput, the ability to perform mechanistic studies at cellular and molecular level, low experimental costs and faster results.

1.4.2.1 Two dimensional (2D) models:

Microfabricated 2D in vitro models have been used to investigate the dynamics of cancer cell behavior as well as for drug screening. The conventional studies on cancer migration have been conducted using two dimensional (2D) models owing to convenience and ease to set up experiments.

In this study by Kramer et al., the HT1O8O fibrosarcoma cells were seeded in a reconstituted basement membrane matrix containing type Iv collagen, laminin, entactin, nidogen, and heparan sulfate on a petri dish (Kramer, Bensch, and Wong 1986). The invasiveness of the cancer cells was examined and was compared against the normal skin fibroblasts. After 7-days incubation of cells seeded on the surface of the gel, the invasion into the matrix was examined by electron microscopy. The results demonstrated that the cancer cells proliferated and gradually migrated dissolving the matrix forming disorganized cell monolayer, electron microscopy images revealed the filipodia and lamellopodia projections from the cancer cells extending into the matrix while the normal skin fibrabalsts cells attached to matrix exhibited minimal invasion during the same period.

Boyden chamber assay is one of the well-known 2D models for intravasation. For instance, Li et al. (Li and Zhu 1999) have developed a Boyden chamber assay to study the migration and invasion of cancer cells. Bovine Aortic Endothelial Cell (BAEC) were used to form an endothelial monolayer and seven different cell lines both malignant and non-malignant were introduced to the chamber. Cells were radioactively labeled for visualization. The trans-migratory activity of cancer cells was correlated with the level of tumor cell induced endothelial monolayer disruption. Quantification was done by image analysis and direct visualization.

Although, the use of 2D assays has enabled addressing important biological questions and significant knowledge has been gained on different characteristics of cancer cells like cell motility and migration from 2D models. However, these models do not accurately depict the in vivo tissue structure and organization which is necessary to study cell-cell and cell-ECM interactions.

1.4.2.2 Three dimensional (3D) models:

Conventional 3D in vitro assays have been mainly based on spheroid or macroscale cell-laden hydrogels that have been widely utilized to perform fundamental biological studies on cancer cell invasion (Szot et al. 2013, Xu et al. 2012). However, these models do not replicate tissue-specific human pathophysiology. Recently, there have been significant initiatives in the use of microscale technologies (i.e. microfluidics) to develop 3D tumor models, with precise control over cell-cell and cell-soluble factor interactions for well-controlled studies on cancer cells behavior within each specific step of metastasis (van Duinen et al. 2015).

Significant progress has been made in the recent years in micro- and nanofabrication techniques (Mehrali et al. 2017, Kharaziha et al. 2016). These techniques can be applied to create 3D microenvironments to control cell-cell, cell-substrate and cell-soluble factors interactions (Park and Shuler 2003). Notably, microfabrication techniques have proven to be instrumental to control both tumor and surrounding microenvironmental factors to conduct studies at single cell level (Nikkhah et al. 2012, Peela et al. 2016, Truong et al. 2016, Peela et al. 2017).

For instance, in a set of studies by Nikkhah et al., silicon surfaces were etched to form 3D microstructures to identify the biomechanical signatures of cancer cells and normal cells on response to 3D topographical architecture. Human fibroblasts, normal breast epithelial cells as well as malignant breast cells were cultured in these micro structures (Nikkhah et al. 2010, Nikkhah, Strobl, Schmelz, and Agah 2011, Nikkhah et al. 2009, Nikkhah, Strobl, Schmelz, Roberts, et al. 2011). Their findings demonstrated that cancer cells adopted to the curved microengineered surfaces, while fibroblast cells stretched and normal mammary epithelial cells formed cellular sheets with tight intracellular junctions. Addition of Histone deacetylase (HDAC) inhibitor drugs interestingly imparted marked alteration is cytoskeleton of the cancer cells when interacting with 3D architecture and compared to their normal epithelial counterparts (Strobl, Nikkhah, and Agah 2010).

Other studies have utilized 3D micropatterned hydrogels or microfluidic technologies to assess cancer cell behavior within each specific stage of metastatic cascade (Peela et al. 2016). In fact, hydrogels have been proven to be excellent biomaterials for tissue engineering and disease modeling applications (Navaei et al. 2017, Navaei, Saini, et al. 2016, Saini et al. 2015, Zorlutuna et al. 2012, Navaei, Truong, et al. 2016, Cha et al. 2014).

In the following, we will provide a brief overview of the current existing technologies. The fischbach group developed a model to analyze the response of tumors to culture dimensionality and hypoxia (DelNero et al. 2015). They designed alginate-based model and cells were introduced as a layer to form 2D structure and 3D spheroids structures. Oxygen in culture was regulated to introduce hypoxic conditions. The extent of invasion was analyzed by counting the number of cells crossing the membrane. Gene transcript analysis of OSCC3 cells cultured in 2D and 3D, under hypoxic and normoxic conditions, was performed to determine the interdependence of hypoxia and dimensionality on gene expression. Results showed that dimensionality of culture affected a large number of genes in hypoxia than normoxia, suggesting that hypoxia response depends on whether culture conditions are performed on 2D or 3D models. Specifically, hypoxia might have triggered the cells to respond to dimensionality, or conversely, changes in dimensionality might have triggered response of cells to hypoxia. Altogether, the results suggest that there is a strong interdependence between hypoxia status and dimensionality of culture at gene expression level. As IL-8 is known to effect inflammation and angiogenesis further investigation on the gene was performed. And results demonstrated significantly higher levels of gene expression in 3D vs. 2D culture independent of oxygen concentration. These results show that there is an interdependence on dimensionality and hypoxia.

George group developed a 3D in vitro model called Pre-Vascularized Tumor (PVT) to study early events of tumor progression (Ehsan et al. 2014). In this model, spheroids of tumor cells and endothelial cells were embedded into fibrin matrix consisting of fibroblasts. This model was proven to be efficient in studying two mechanisms, vessel formation (i.e. angiogenesis) and intravasation. The study was carried out under hypoxic conditions. In the presence of tumor cells, the vessel sprouts were more irregular and shorter as compared to spheroids containing only endothelial cells and fibroblasts. Moreover, breast cancer cells intravasated into the lumens of the vessel.

Figure: 1.3 in vitro models to study intravasation (a) (i) Schematic of model showing tumor and endothelial cell embedded in fibrin matrix (ii) immunofluorescence images showing intravasation of tumor cells (green) into the blood vessel (red). Scale bars: 100 μm. Adapted from Ehsan et al. with permission from Royal Society of Chemistry [Integrative Biology], copyright (2014) (Ehsan et al. 2014). (b) (i) Schematic illustrating the fabrication of microscale alginate scaffolds. (ii) Representative images of a co-culture invasion assay. Adapted from DelNero et al. with permission from Elsevier [Biomaterials], copyright (2015) (DelNero et al. 2015).

Notably, this effect was enhanced in hypoxia conditions. Further analysis confirmed that Slug, a marker for EMT, was highly upregulated within the hypoxic environment and played a critical role in intravasation (Figure 1.3 a).

During Intravasation, cancer cells migrate in response to gradients of chemokines (Roussos, Condeelis, and Patsialou 2011). Limitations of the above models is that cytokine gradients are not established. Thus, advances in microfluidic systems have enabled us to develop novel models which can capture different mechanistic features like cytokine gradients, fluid flow different components of the tumor as well as interactions with different cell types. For instance, Han et al. developed a 3D microfluidic model to study the effect of oriented collagen fibers on tumor cell migration across the basement membrane surrounding the vessels during intravasation (Han et al. 2016). The device was loaded with Matrigel^® followed by collagen I leading to sandwich of two hydrogels. During polymerization Matrigel^® volume was swollen while collagen volume shrunk developing strain into the system which led to collagen fibers to orient vertically at the interface of collagen-Matrigel^®. MDA-MB-231 cells embedded within the sandwiched gel began intravasating to the gel interface (collagen- Matrigel^® interface) as early as 48 h. By 144 h most of the cells invaded and crossed the Matrigel^® region. On the other hand, when collagen fibers were not oriented homogenously, MDA-MB-231 cells could not invade into the Matrigel^®. These results demonstrated that the local fiber alignment enhanced cell–ECM interactions, where metastatic MDA-MB-231 breast cancer cells followed the local fiber alignment direction during the intravasation (Figure 1.4 a).

Figure: 1.4 Microfluidics based models to study intravasation (a) (i) Schematic representation design and fabrication of microfluidic model (ii) Representative images depicting tumor cell intravasation into ECM due to collagen fiber orientation. Adapted from Han et al. with permission from National Academy of Sciences [PNAS], copyright (2016) (Han et al. 2016). (b) (i) Schematic of the microfluidic device (ii) Representative images showing tumor cell (red) intravasation into endothelial lining (green). Scale bars: 30 μm. Adapted from Zervantonakis et al. with permission from National Academy of Sciences [PNAS], copyright (2012) (Zervantonakis et al. 2012). (c) (i) Schematic of the microfluidic device (ii) Representative images showing tumor cell (green) intravasation into endothelial lining (red). Adapted from Lee et al. with permission from AIP Biomicrofluidics copyright (2014) (lee et al. 2014).

Similarly, the Kamm group developed a 3D microfluidic model to study tumor cell intravasation and elucidate the role of TNF-α and TAMs in regulating endothelial barrier permeability and tumor cell intravasation (Zervantonakis et al. 2012). The device had one central stromal region filled with collagen surrounded by endothelial region to left and tumor region on right. TAMs were embedded within collagen hydrogel to study their effect on vessel permeability and tumor intravasation. This model was useful in studying the process of intravasation as it allowed real time visualization of cancer cell behavior (i.e. migration, intravasation). In the presence of TAMs, microvessels had higher permeability. The results also showed that the presence of TNF- resulted in increased endothelial barrier impairment, which facilitated cancer cell intravasation. To further confirm their findings, they knocked down TNF-α in the presence of TAMs and observed that intravasation levels did not decrease, suggesting that TAMs may secrete other factors, which might indeed enhanced permeability of the vessels and resulted in increased intravasation (Figure 1.4 b).

In another study, Noo Li Jeon group developed a microfluidic based metastasis chip to study cancer angiogenesis and intravasation (Lee et al. 2014a). Endothelial cells HUVECs (Human Umbilical Vein Endothelial Cells) and stromal cells normal human lung fibroblasts were introduced into the channels at close proximity. The vascular network formed had well formed boundaries with proper cell junctions. After 7 days in culture, cancer cells U87MG were introduced into the upper channel of the device (Figure 1.4 c i). The microvessel were regarded as pre-existing at the cancer site and the cancer cells in the perivascular region which would secrete angiogenic factors to produce angiogenic sprouts. The number and coverage area of angiogenic sprouts were quantified and the results demonstrated that microvessels with cancer cells showed more angiogenic sprouts compared to microvessels without cancer cells. Further, to model intravasation, MDA-MB-231 cancer cells were introduced into the upper chamber of the device and trans-endothelial migration of cancer cells into preexisting vessels was observed. After three days in culture, several cancer intravsating into the lumens could be observed (Figure 1.4 c ii).

2. OBJECTIVE OF THE THESIS

This thesis is aimed to investigate the role of tumor microenvironment and different biochemical factors involved during cancer cell invasion and intravasation. In particular, we developed and utilized a novel microfluidic assay to test the hypothesis that there are bidirectional interactions between cancer cells and endothelial cells during the process of invasion and intravasation. Addressing this question is not only important for understanding the molecular mechanisms, but also will validate the device as a potential tool for identifying potential targets for cancer therapy.

The first step was to create a first generation platform to optimize the process of vasculogenesis in our microfluidic model. This was achieved by incorporating HUVECs in fibrinogen hydrogel and analyzing the vascular network formation, growth and maturation. Upon validation of the platform, VEGF level in the microenvironment was increased and endothelial network growth was characterized. Further analysis was performed to investigate the effect of increased levels of VEGF on endothelial cell permeability.

We further developed a second generation three-layer microfluidic based tumor-vasculature model to characterize tumor-endothelial interactions. To explore the relationship between endothelial barrier and cancer cell intravasation, we simultaneously investigated tumor cell invasion and endothelial vascular formation and permeability as well as intravasation. Interestingly, our assay was designed in a way to enable direct observation of tumor cells during invasion through the stroma and intravasation through the endothelial barrier while accurately controlling the organization and transport of biomolecular components. We analyzed the number and morphology of cancer cells invading through stroma. We further assessed the influence of cancer cells on endothelial cell capillary formation and permeability and compared to endothelial mono-culture condition.