Unraveling Cancer Metastatic Cascade Using Microfluidics-based Technologies

Abstract

Cancer has long been a leading cause of death. The primary tumor, however, is not the main cause of death in more than 90% of cases. It is the complex process of metastasis that makes cancer deadly. The invasion metastasis cascade is the multi-step biological process of cancer cell dissemination to distant organ sites and adaptation to the new microenvironment site. Unraveling the metastasis process can provide great insight into cancer death prevention or even treatment. Microfluidics is a promising platform, that provides a wide range of applications in metastasis-related investigations. Cell culture microfluidic technologies for in vitro modeling of cancer tissues with fluid flow and the presence of mechanical factors have led to the organ-on-a-chip platforms. Moreover, microfluidic systems have also been exploited for capturing and characterization of circulating tumor cells (CTCs) that provide crucial information on the metastatic behavior of a tumor. We present a comprehensive review of the recent developments in the application of microfluidics-based systems for analysis and understanding of the metastasis cascade from a wider perspective.

Introduction

Cancer is a pervasive disease that is the second greatest cause of death worldwide. In 2019, there were more than 1.7 million new patients diagnosed with cancer and more than 0.6 million cancer deaths in the USA alone (Bray et al. 2018; Siegel et al. 2019). In comparison, there is no better situation in our country Iran. In 2020, according to the World Health Organization (WHO), 131 thousand new patients were diagnosed with cancer and 79 thousand died (Roshandel et al. 2021). Numerous investigations have been conducted to overcome this major health problem yet an efficient and reliable universal anticancer treatment has yet to be found (Lin et al. 2019). Metastasis is the primary culprit for the majority of cancer deaths. As such an in-depth understanding of the metastasis cascade is crucial for controlling and treating cancer. Metastasis is a multi-step process involving the dissemination of cancer cells to a distant site and the production of secondary cancer growths, known as metastases (Valastyan and Weinberg 2011). The metastatic cascade is a convoluted process that begins with the rapid proliferation of cancer cells at the primary tumor stage. New blood vessels are formed to provide tumors with sufficient nutrients and oxygen in a process called angiogenesis (Tetala and Vijayalakshmi 2016). During cancer progression, some tumor cells acquire specific properties that allow them to pass through a series of steps in order to detach from the primary tumor. They can either undergo a complete or partial epithelial to mesenchymal transition (EMT) which causes them to vary extents, to lose the epithelial polarity as well as cell–cell adhesions, thereby becoming motile (Bakir et al. 2020). Cancer cells with mesenchymal phenotype can invade adjacent tissues and enter the circulatory system through a process called intravasation (Esmaeili et al. 2019; Janni et al. 2011). An aggressive subpopulation of these cancer cells, the so-called circulating tumor cells (CTCs), can escape the immune system and survive in the hostile environment of blood. Upon arrival to a suitable niche, mesenchymal CTCs extravasate and recover their epithelial phenotype through mesenchymal to epithelial transition (MET) in order to establish metastatic lesions (Kang and Pantel 2013). Even epithelial cancer cells that have not gone through the EMT process can be collectively invasive with the help of other cells like cancer-associated fibroblasts (Ilina et al. 2018).

The details of each step in the metastasis process are not yet fully understood. Appropriate models are therefore needed to shed light on the complex multi-step process of metastasis. So far, animal models—mostly mouse models—have been exploited to study the mechanism of metastasis (Steeg 2016). These models, however, suffer from some major limitations. For instance, there is a difference between human and mice antitumor immune response, specifically, the technical difficulty of increasing the spectrum of immune cells engrafted while reducing the mouse innate immune response (Shultz et al. 2012). In vitro models, on the other hand, could not recapitulate the systemic nature of metastasis (Myung et al. 2016; Zieglschmid et al. 2005).

Microfluidic technology with its unique characteristics has become a promising tool in cancer research. Microfluidic devices have proven to be excellent experimental systems due to their low sample volume requirement, controllable environment, portability, and rapid analysis (Tetala and Vijayalakshmi 2016). Microfluidic platforms have been used to study tumor growth by mimicking tumor microenvironment and angiogenesis (Whitesides 2006). They have been utilized in the detection and identification of CTCs with aggressive phenotypes (Shang et al. 2017). For instance, our research group has recently fabricated a microfluidic device for CTCs deformability assessment, and we were able to find a direct relation between deformability of breast cancer CTCs and their metastatic aggression (Hakim et al. 2021). Moreover, intravasation of CTCs, extravasation, and secondary tumor formation have also been explored by microfluidic devices with sophisticated designs (Tsai et al. 2017).

The individual steps of the metastasis cascade (Fig. 1) have been reviewed previously by many different investigators (Shang et al. 2019). For a better understanding, however, the entire process has to be viewed in one context. Here, we present a comprehensive review of the application of microfluidic technology to understanding the tumor microenvironment (TME), CTC characterization, and cancer dissemination.

Fig. 1.

A A typical schematic structure of a microfluidic device. B Fabricated samples of a structurally simple microfluidic device and C more complicated chip (Hakim et al. 2021)

Microfluidic Technology

Microfluidic refers to the science and technology associated with devices capable of working with fluids at micrometer scales. Microfluidics is the miniature of conventional fluid methods. This technology utilizes channels with dimensions of tens to hundreds of micrometers (Kim et al. 2010). Microfluidic platforms were first used in laboratories for analytical purposes as a result of their unique properties. The unique features include low sample volume requirement, controllable environment, portability, and rapid analysis (Tetala and Vijayalakshmi 2016). Figure 2 presents the basics of microfluidic chips showing they can be designed with simple or sophisticated architecture in a small chamber for different purposes. Microfluidics has many applications in various sciences, from chemical synthesis and bioanalysis to optics and information technology. The emergence and development of microfluidic systems have recently been very rapid in the four areas of molecular analysis, molecular biology, bio-defense, and microelectronics (Whitesides 2006).

Fig. 2.

A schematic overview of metastasis cascade (1. primary tumor formation, 2. intravasation and circulation as CTCs, 3. extravasation, and 4. homing and secondary tumor formation)

Back in the 1950s, the first microfluidic devices such as gas-phase chromatography (GPC), high-performance liquid chromatography (HPLC), and capillary tube electrophoresis (CE) were designed (Frazier et al. 1995). The newly introduced microfluidic platforms revolutionized the field of analytical chemistry since the high resolution and precise molecular analysis were made possible by the application of just small amounts of samples and reagents (Azizipour et al. 2020; Reyes et al. 2002). Moreover, the boom in genomics in the 1980s, with the advent of other fields of bioassay in the micro-dimension such as high-yield DNA sequencing, required more accurate, sensitive, and efficient measurement methods. Microfluidics could offer reliable solutions to overcome such requirements, which also paved the way for the development of different microfluidic platforms (Paegel et al. 2003). The other impetus for the application of microfluidic systems came after the end of the Cold War when chemical and biological weapons inflicted irreparable damages. To address the consequences, the US Department of Defense Advanced Research Projects Agency (DARPA) in the 1990s took action to support projects to develop microfluidic devices used to identify chemical-biological threats (Duffy et al. 1998). These programs were one of the most important reasons for the rapid development of microfluidic technologies (Whitesides 2018, 2006). The last but not the least main factor for microfluidics development was microelectronics. Traditionally, the bulk material was used for electronic devices had inherent limitations (Choi and Mody 2009). With the introduction of microfluidics and microstructure to the field of electronics, smaller, faster, cheaper, and more accurate electronic devices were possible to produce by new manufacturing methods such as photolithography and the application of silicon substrates (Venkatesan et al. 2020).

Early microfluidic chips were made of silicon and glass using methods such as photolithography. These procedures were costly, time-consuming, and complicated. The common materials also had limitations and were expensive, gas-impermeable, and non-transparent (Whitesides 2006). The drawbacks hindered the progress and widespread application of microfluidics among different fields up to the late 1980s by the time the soft lithography method was developed. The comparatively cheaper method of soft lithography enabled patterning 3-dimension (3D) microscale structures on a wide range of substrates such as polymers, elastomers, and organic as well as non-organic substances (Xia and Whitesides 1998). Later, in the 1990s, the new microfabrication method of soft lithography and simultaneous application of transparent, gas permeable, and biocompatible cheap elastomer of polydimethyl siloxane (PDMS) were used for the fabrication of microfluidics. Consequently, such platforms were made accessible for many scientific areas, specifically cell biology (Duffy et al. 1998). The fabrication process consists of several steps of which the first main one is creating a master mold with silicon wafers and photoresist by application of the photolithography methods. The next step is pouring liquid-phase PDMS polymer on the mold and curing the polymer after being degassed. Finally, the PDMS cast should be removed from the master mold and bonded to a clean glass wafer. Nevertheless, this fabrication procedure has some drawbacks, such as being time-consuming and requisite of being conducted in a cleanroom (Faustino et al. 2016). Accordingly, the search for different methods has never stopped. One of these alternatives is called “Xurography” through which the fabrication will be carried out using a cutting plotter and adhesive vinyl films for creating master molds. The master mold is generated fast enough to eliminate the need for a cleanroom. The procedure can be used for fabricating microfluidic tools with elastomers like PDMS. Devices fabricated by such technique can have widths and thicknesses as low as 6 and 25 µm, respectively. However, the method fails to provide a good resolution for microdevices with such small dimensions as well as microdevices with complicated geometries (Bartholomeusz et al. 2005). Hot embossing is an alternate method in which a thermoplastic film is placed in a specific location between two molds being followed by compression and heating to become viscous. By doing so, a cast of the desired mold is formed and can be removed after the cooling process. PMMA, a biocompatible and transparent material, is commonly used for microfluidic fabrication by such a method. Nevertheless, hot embossing has its own constraints too, such as being time-consuming and expensive regarding mold fabrication (Giboz et al. 2007). One of the recently developed and most promising alternative methods for microfluidics fabrication is 3D printing. Through the fabrication process, layers of viscous polymers or the initial solution of polymers are deposited on top of the previous layer by a moving computer-controlled nozzle (Bhattacharjee et al. 2016). The rapid and low-cost procedure can also be mild enough to be applied for bio-inks such as hydrogels and even cell-loaded hydrogels. As a result, creating human organ models for a wide range of biomedical applications in the most physiologically relevant manner is possible by the application of 3D printing (Papaioannou et al. 2019). Nevertheless, this method also has some yet-to-be-resolved shortcomings, such as limited printable hydrogels and the sensitivity of cells to harsh procedures, making 3D printing fabrication method a subject of an explosively increasing number of studies (Prabhakar et al. 2021).

Tumor microenvironment

Cancer is a complex invasive disease with a heterogeneous nature that can present different responses to the same treatment in patients with similar tumors (Coventry and Ashdown 2012). Therefore, a better comprehension of tumor biology is critical in the development of effective cancer therapeutics (Hanahan and Weinberg 2011; Siegel et al. 2016). Many cancer therapeutics face failures in clinical trials due to the use of inappropriate models through preclinical studies (Day et al. 2015). The generation of traditional preclinical models has proven time-consuming and cost-intensive with a poor prediction of drug responses (Sajjad et al. 2021). Moreover, clinical trials are mostly conducted on patients during the final stages of the disease. Consequently, the early stages of cancer, which are considered vital periods for diagnosis and treatments, are missed. In order to address all of the challenges, the development of in vitro cancer models being capable of mimicking the main aspects of a tumor is crucial for an effective prediction of therapeutic responses. Among conventional models, 2D tumor models—in which cancer cells are cultivated as monolayers on a flat plastic substrate—and animal models have been widely used (Hoarau-Véchot et al. 2018). However, both model types suffer from some major shortcomings. For instance, 2D monolayers lack physiological relevance in regard to factors such as cytoskeletal conformations, cell–cell interactions, and cell-extracellular matrix (ECM) interactions (Chitcholtan et al. 2013). Animal models, fabricated by injection of human tumor cells into immunodeficient animals, also face safety and reproducibility limitations (Thibaudeau et al. 2014). As a consequence, there has been a growing interest in the development of 3D in vitro models which have the benefits of easy manipulation, low cost, and minimum invasiveness (Esch et al. 2015).

An in vivo TME consists of cancer cells, different stromal cells, and vasculatures surrounded by a critical substrate—called ECM—with a specific (Ingber 2002). The ECM is a protein-rich structure secreted by stromal cells that provide cells with mechanical support as well as different growth factors and chemokines such as vascular endothelial growth factor (VEGF) (Rijal and Li 2016). Accordingly, ECM plays a critical role in different cell activities, specifically through tumor progression phases (Walker et al. 2018).

During tumor progression, rapidly growing cancer cells reside far away from blood vessels. Therefore, the lack of sufficient oxygen and nutrients force the cells to enter the necrotic phase in which cell death occurs, especially at the core of the tumor (Muz et al. 2015). This so-called hypoxic condition leads to the expression of angiogenesis inductive factors like VEGF (Krock et al. 2011). Angiogenesis begins with the sprouting of new vessels from the existing vasculature to help provide essential nutrients for the tumor. Following the angiogenesis phase, metastasis occurs as cancer cells degrade the ECM in order to enter the circulatory system. Upon reaching a secondary site, formation of a secondary tumor initiates (Hanahan and Weinberg 2011).

In order to mimic the intricate biology of cancer and to generate a physiologically relevant in vitro model (including the biochemical and biophysical factors), three main aspects should be considered: (i) geometry and spatial control over the location of different contributing cells, (ii) chemical and mechanical properties of the ECM, and (iii) cellular interactions between different cell types. The tumor-on-a-chip approach has emerged as a promising method for fulfilling these requirements by coupling microfluidic technology with the principles of tissue engineering (Millet et al. 2019). Tumor-on-a-chip devices provide researchers with various advantages over conventional 2D and animal cancer models, including transparency and real-time visualization, consideration of cell–cell and cell-ECM complex interactions, controlled flow of required substrates, and precise manipulation of microstructure geometry (Park et al. 2019).

Geometry of a tumor

The geometry of a native tumor is complex and geometric factors can be highly influential on different aspects of the TME including effects upon the concentration gradient of different soluble factors as well as biochemical and biomechanical cues (Polacheck et al. 2013). For instance, oxygen concentration distribution and the shape of the necrotic region of the tumor are determined by the oxygen diffusion coefficient within the tumor (Gatenby et al. 1988). Furthermore, the configuration of the tumor and the blood vessels cause abnormal flow patterns within the tumor which affect cellular processes by exposing them to differential shear stress and mechanical forces (Ballermann et al. 1998). Accordingly, a 3D in vitro cancer model should be capable of mimicking such geometry in order to provide a comprehensive physiological relevance required for cancer biology investigations and drug validations. Recent advances in microdevice fabrication techniques have provided researchers with the capability of recapitulating complex microscale geometry and spatial cell distributions of the native TME (Hassell et al. 2017; Park et al. 2019).

One of the most common microfluidic TME models is based on PDMS microstructures coupled with hydrogel-based ECMs for mimicking in vivo tissue-tissue interfaces (Carrion et al. 2010; Song and Munn 2011). This design includes adjacent straight channels, resembling different layers, separated by PDMS micro-posts which navigate the liquid hydrogels through the chip. This method provides the capability for polymerization of the hydrogel at desired sections allocated to the ECM without invading the adjacent channels. The model is a robust and easy-to-operate design that allows for the widespread application of tumor-on-a-chip models in preclinical tests (Ayuso et al. 2016). A microfluidic device has also been proposed to investigate the effect of oxygen gradients on cancer cell behavior (Chen et al. 2011). That design consisted of three adjacent channels, of which the central channel was allocated to media, and the other channels were used as cell culture channels. The central and cell culture channels were separated by PDMS micro-posts, being located between two sections with different oxygen concentrations. A higher cell death rate was observed in the vicinity of the regions with low oxygen concentrations. The effect of the anticancer drug, tirapazamine (TPZ), was also explored on cells using this approach and demonstrated a similar hypoxia-induced cytotoxic effect (Chen et al. 2011). Despite the fact that these microfluidic devices have been useful in exploring certain specific effects upon cancer biology, the extremely complex conformation of in vivo tumors has tended to be vastly simplified (Hachey and Hughes 2018).

The geometry of a tumor has a great impact on the flow patterns within the tumor as well as around the periphery and can thereby play a pivotal role in controlling each cells' fate within the tumor (Buchanan et al. 2014). The architecture of the vascular network is a crucial factor which if ignored in the construction of a cancer model can result in oversimplifications and produce results that lack physiological relevance. A breast tumor model incorporating a vasculature based on a mouse tumor architecture has been proposed that includes a network of microchannels to mimic both high and low perfused regions within the tumor in order to assess the efficacy of an anticancer drug (Fig. 3A). It was revealed that variations in the flow pattern and perfusion profiles within the tumor model caused morphological differences in cancer cells. Furthermore, a significant dependency of drug responses on the tumor flow pattern was confirmed (Pradhan et al. 2018). Another group of popular microfluidics-based models for primary tumors are 3D multicellular tumor spheroids, in which cancer cells are aggregated to form tumor spheroids (Wu et al. 2008). Several microfluidic platforms have been used to form different types of cell aggregates such as emulsion cell aggregation and hydrodynamic trapping of cells in microstructures (Hsiao et al. 2009; McMillan et al. 2016). Taking advantage of droplet-based microfluidics, a microfluidic device was fabricated that consisted of a two-layer polydimethylsiloxane chip with arrays of adjoining wells and an automated droplet-manipulation device (Du et al. 2018). Droplets containing tumor spheroids and human umbilical vein endothelial cells (HUVECs) were co-cultured in interconnected wells and the angiogenic potential of different cancer cell types was investigated. Higher VEGF expression levels, that cause a stronger angiogenic response, were observed in highly metastatic cancer cells (Fig. 3B) (Du et al. 2018).

Fig. 3.

Microfluidic platforms developed for tumor microenvironment (TME) modeling. A Breast tumor mimetic microfluidic device and the schematic of two designs used for fabrication of high perfusion and low perfusion microfluidic devices. Reproduced with permission from (Pradhan et al. 2018). B A schematic representation of microfluidic tumor encapsulation and perfusion devices. Moreover, the schematic of micro-tumors being encapsulation by core–shell technology and perfusion of micro-tumors with the capability of creating vascularized tumors are depicted. Reproduced with permission from (Du et al. 2018). C. A schematic of a natural TME. Reproduced with permission from (Rodrigues et al. 2021)

In order to show the effect of ECM confinement on cancer cells, an investigation was carried out on breast cancer cell motility when cultured in cell chambers of different sizes. It was found that in contrast to small cell chambers, where MDA-MB-231 cells were not able to spread and migrate, large cell chambers promoted cell motility and division (Nuhn et al. 2018). Moreover, it was shown that cells sense the interface of the ECM-device wall and migrate toward the interface if the distance is less than 150 µm. Tumor architecture is one of the main physical factors contributing to the tumor progression affecting cell fate in different manners including exposure to different flow patterns and the resultant shear stresses. Such architectural features can be recreated using a microfluidic device with an in-vivo-like design. Moreover, in order to model other elements of a tumor, it is necessary to consider biophysical and biochemical aspects of the ECM while the cancer cells surround the substrate (Nuhn et al. 2018).

Extracellular matrix

ECM is the non-cellular structure providing cells with chemical and mechanical support (Yue 2014). During cancer progression, cancer cells constantly interact with the ECM in order to first grow into a tumor and then spread to other tissues later. ECM remodeling can assist cancer cells to invade adjacent tissues and enter circulatory systems. However, the interaction between cancer cells and ECM is still unclear. A deeper insight into the process is achieved through the use of in vitro models of tumor ECM (Holle et al. 2016). To reproduce the native ECM in microfluidic platforms, appropriate materials with controlled chemical and mechanical parameters should be selected. Two types of biocompatible substances are typically utilized for ECM reproduction; biomaterials obtained from natural resources and biomaterials manufactured synthetically (Frantz et al. 2010). Biomaterials can be used in different formats such as scaffolds and hydrogels (Aamodt and Grainger 2016). Due to their high-water content, hydrogels provide an ideal environment for cell growth. Natural hydrogels, composed of common tumors ECM proteins, such as fibronectin and collagen, provide specific sites that cells can adhere to, proliferate, and then degrade the structure while migrating (Park et al. 2017). The composition of hydrogels can also be modified to obtain desirable biochemical and mechanical characteristics (Li et al. 2018). A wide range of natural hydrogels has been employed as 3D matrices to support cell growth and migration, including fibronectin, basement membrane extract (BME), collagen, and matrigel (Tibbitt and Anseth 2009). Among the natural hydrogels, collagen type I is the most abundant protein in the ECM of most tissues; additionally, the protein has been widely used in microfluidic platforms (Rijal and Li 2016).

Given the significant role of ECM on cellular activities in the tissue, the effect of different ECM characteristics such as protein composition, chemical, and mechanical properties have been investigated in several studies (Groessner-Schreiber and Tuan 1992; Padhi and Nain 2020). The impact of ECM mechanical stiffness on the behavior of breast cancer cells using different concentrations of alginate has been reported (Cavo et al. 2016). Alginate is a widely used natural biomaterial that has strong mechanical properties but poor cell adhesion characteristics (Sarker et al. 2015). In a recent study, alginate was used in combination with collagen to manipulate ECM stiffness without changing the chemical properties. Upregulation of EMT in microscale 3D culture was observed by higher ECM stiffness (Agarwal et al. 2017). In another study, it was revealed that the invasive ability of both malignant or benign breast cancer cells was affected by the substrate stiffness and that the number of invading cancer cells and invasion distance decreased when cultured in softer ECM highlighting the importance of ECM physical factors on cellular activities (Azadi et al. 2020).

Even though natural hydrogels have physiologically relevant biological characteristics, they generally possess poor mechanical features (Zhao et al. 2013). In contrast, synthetic hydrogels have strong mechanical characteristics but lack cell adhesive sites (Rahmany and Van Dyke 2013). Coupling of synthetic hydrogels with cell adhesive ligands, like arginylglycylaspartic acid peptide motif (RGD), would therefore allow for the modification and control of mechanical and chemical properties of the hydrogel-based ECM. For instance, stiffness and cell adhesion sites can be controlled and modified independently of each other (Park et al. 2017; Rijal and Li 2016). In a different study, PDMS microstructures were used to develop a dual-layer multicellular tumor model (Roudsari et al. 2016). Cell encapsulation hydrogel was synthesized, using a mixture of PEG-PQ-PEG, PEG-RGDS, HBS-TEOA, 1-vinyl-2-pyrrolidinone, and eosin Y photoinitiator as polymer precursors to analyze the effect of vascular cells on lung tumor cells in 3D culture (Roudsari et al. 2016). It was demonstrated that vascular cells at close distances from tumor cells can induce tumor cell migration and tumor cluster morphology alteration. Inhibition of TGF-β1 was also observed to block ECs-driven tumor cell migration (Roudsari et al. 2016). Studies have also shown that the ECM and cancer cells have a mutual effect on each other as cancer cells can alter the ECM structure as well as the chemical and mechanical properties through tumor progression (Friedl and Alexander 2011). In another study, a microfluidic device was fabricated to assess the impact of tumor progression on the surrounding ECM whereby overexpression of fibronectin and hyaluronic acid and its release into the ECM was observed during tumor epithelial invasion (Gioiella et al. 2016).

Cellular interactions

The natural TME consists of cancer cells and different types of stromal cells surrounded by ECM in a complex 3D architecture (Fig. 3C) (Rodrigues et al. 2021). Many cellular interactions occur between cancer and stromal cells through the exchange of different biochemical and biomechanical cues contributing to hypoxia, angiogenesis, metastasis, and cancer cell drug resistance. The lack of cell–cell and cell–stroma interactions in in vitro tumor models can lead to weak predictions (Baghban et al. 2020). Therefore, there has been a growing interest in the fabrication of in vitro tumor models that recapitulate the in vivo TME by employing different cell types actually involved in the cellular microenvironment (Lee et al. 2018). Normal human lung fibroblasts (NHLFs), endothelial colony-forming cell-derived endothelial cells (ECFC- ECs), colorectal cancer cell line (Caco-2), normal breast fibroblasts (NBFs), CAFs, MDA-MB-231, MCF-7, and colorectal cell line 268 (CRC-268) were cultivated in a microfluidic device (Shirure et al. 2018). The device included three cell culture chambers separated by microporous walls to recapitulate the tumor microenvironment near the arterial end of a capillary. Once the model was generated, the potential of the device in supporting the growth of different cancer cell lines was analyzed. The invasive cell lines, CRC-268 and MDA-MB-231, displayed a significant growth rate while invading the surrounding ECM. The nonmalignant cell lines, Caco-2 and MCF-7, formed tumor spheroids with a slower growth rate. The ability of different stromal cells in angiogenesis induction was assessed. The results showed that angiogenic response to cancer-associated fibroblasts (CAF) was more than normal breast fibroblasts (NBF) and angiogenesis was directed toward the CAF (Shirure et al. 2018).

Mimicking main phases of tumor progression

In a physiologically relevant in vitro model of a tumor, all the prementioned aspects should be adjusted based on the target tumor progression phase (Katt et al. 2016). The functionality of some anticancer drugs, which are based on specific attributes of each tumor progression phase, highlights the importance of mimicking the cellular niches (Ayob and Ramasamy 2018). Among different cellular states studied using tumor-on-a-chip technology, necrosis, angiogenesis, and metastasis are the three main cellular processes related to tumor progression that have been investigated in TME modeling (Benien and Swami 2014).

Necrotic conditions arise when abnormal cell growth generates low oxygen and nutrient levels, especially at the center of the primary tumor. Consequently, cellular death or necrosis happens at the core of the tumor thereby altering the metabolic condition of the cancer cells to anaerobic metabolism (Lee et al. 2018). These conditions induce cancer cell heterogeneity within the tumor and improve cell survival as well as drug resistance. In order to study cell behavior under nutrient deficiency, a microfluidic tumor slice model with a central chamber for a 3D cultivation of tumor cells and a lumen for media perfusion was developed (Ayuso et al. 2019). The model revealed that nutrient starvation could induce the production of a necrotic cellular core contributing to alterations in gene expression of cancer cells, such as the upregulation of VEGF expression. Additionally, starving cancer cells switch the metabolism to anaerobic glycolysis known as the “Warburg effect” (Ayuso et al. 2019; Hanahan and Weinberg 2011). Consequently, lactate is produced at high concentrations decreasing the pH of the TME. Accordingly, the distance of tumor cells from the nutrient source is considered as an influential factor affecting nutrient concentration and pH gradient, which are significant sources of tumor heterogeneity. This microfluidic model can be used to develop effective therapeutic strategies targeting intratumor metabolic heterogeneity (Ayuso et al. 2019).

In the absence of required levels of nutrients and oxygen, new vascular sprouts emerge from the closest vessels to the tumor in order to fulfill unmet needs (Carmeliet and Jain 2011). This phenomenon, known as angiogenesis, assists invasive tumor cells to enter the circulatory system and initiate the metastatic cascade (Martin et al. 2013). In fact, angiogenesis acts as a bridge that interconnects the necrotic and metastatic niches (Zuazo-Gaztelu and Casanovas 2018). Given the unique role of angiogenesis through metastasis, many attempts have been made to study this process in vitro (Ruoslahti 2002). A tumor spheroid was developed with a perfusable vascular network for long-term perfusion culture of tumor cells in a microfluidic device (Nashimoto et al. 2020). In contrast to static cell culture, where the media remains still within the channels, a continuous flow of media is maintained within the media channels. Co-culturing tumor cells with fibroblasts induced angiogenesis in the tumor spheroid. Dose-dependent results for screening drug responses under continuous perfusion were found to be more relevant to the real condition of a tumor, in contrast to the static state (Nashimoto et al. 2020).

Such advances in modeling the native tumor microenvironment suggest that microfluidic platforms are promising tools for developing efficient disease models that provide more clinically relevant results. Researchers, however, still face some major challenges in this regard. The main challenge is to confirm the physiological relevance with a real tumor which determines the effectiveness of the tumor model. Moreover, there are some other technical challenges such as incorporation of different cell types as a co-culture under a precisely set chemical and mechanical condition, which means that TME modeling remains both quite difficult and elaborate. To address these challenges, a robust microdevice design that includes elements for flow control can be helpful in improving reproducibility and precision. Such types of design may prove useful when it comes to exploring other aspects of cancer biology, such as studying the CTCs (Miermont et al. 2019). An overview of the critical aspects of the above TME models is represented in Table 1.

Table 1.

An overview of key aspects considered in the reported TME models

Cancer Cells	Geometry	Matrix	Objective of the Research	Ref
Carcinomic human alveolar basal epithelial cells (A549)	Three adjacent channels separated by PDMS micro-posts, in between two microchannels designed for generation of different oxygen gradients	Fibronectin coating	Investigation of cancer cell activities under different oxygen concentrations	(Chen et al. 2011)
MCF-7 human mammary cancer cells, MDA-MB-231 Human mammary adenocarcinoma cells, BJ-5ta fibroblasts, HBTECs	Multi-chamber microdevice designed using mouse tumor vasculature architecture	Fibronectin and gelatin and PEG	Assessment of vascular structure driven flow pattern effects on tumor cell Anti-cancer Drug responses	(Pradhan et al. 2018)
Human embryonic kidney cell line (HEK293), Human umbilical vein endothelial cells (HUVECs)	Two-layer PDMS microdevice including 6 × 9 arrangements of paired wells and an automated droplet-manipulation device	Matrigel	Tumor angiogenesis evaluation	(Du et al. 2018)
MCF-7, HUVECs, Primary human adipose-derived stem cells (hADSCs)	Three adjacent channels separated by PDMS micro-posts and a microdevice for creation of micro-tumor	Type Ι collagen, Alginate	Drug resistance analysis in 3D vascularized tumor model	(Agarwal et al. 2017)
HUVECs, lung adenocarcinoma cancer cells (344SQ), Human vascular pericytes (HVP)	Tumor model in a dual-layer structure fabricated using 368 µm PDMS cylinders	A mixture of 4% w/v PEG-PQ-PEG, 3.5 mM PEG-RGDS, HBS-TEOA, 3.5 μL/mL 1-vinyl-2-pyrillidinone and 10 μ M eosin Y photoinitiator polymerized using UV light	Investigation of Tumor and stromal cell interactions in poly(ethylene) glycol-based hydrogels	(Roudsari et al. 2016)
Cancer activated fibroblasts (CAF), Normal fibroblasts (NF), MCF-7	Two adjacent cell culture chambers separated by PDMS micro-posts	Collagen and fibronectin	ECM-Activation in vitro during Tumor Progression	(Gioiella et al. 2016)
Normal human lung fibroblasts (NHLFs), Endothelial colony forming cell-derived endothelial cells (ECFC- ECs), Colorectal cancer cell line (Caco-2), Normal breast fibroblasts (NBFs), CAFs, MDA-MB-231, MCF-7, Colorectal cell line 268 (CRC-268)	Three parallel tissue chambers separated by microporous walls	Fibronectin	Recapitulating the tumor microenvironment near the arterial end of a capillary	(Shirure et al. 2018)
Colon cancer (HCT-116)	A tumor cell culture chamber including a lumen enabling media perfusion	Collagen hydrogel	Necrotic low nutrition niche	(Ayuso et al. 2019)
HUVECs, NhLFs, MCF-7, Human colon cancer cell line (SW620), MDA-MB-231, Hepatocellular carcinoma cells (HepG2)	Three adjacent channels separated by PDMS micro-posts with the central channel including a spheroid well	Fibrin–collagen gel	Evaluation of perfusion culture effects on tumor activities and drug responses	(Nashimoto et al. 2020)

Circulating tumor cell

CTCs were first discovered in 1886 by Thomas Ashworth in the blood of a patient with metastatic cancer far from the primary tumor (Ashworth 1869). Since then, CTCs have attracted a lot of attention as they share many characteristics with primary tumor cells and therefore can shed light on the molecular mechanism of metastasis (Plaks et al. 2013). They are also regarded as a predictive biomarker for cancer detection and prognosis (Chinen et al. 2017). At first, their number in blood—regarded as CTC enumeration—was proposed as a reliable prognostic biomarker (Toss et al. 2014). Later, surface proteins and physical characteristics of CTCs were also found crucial for characterization (Millner et al. 2013).

However, two main obstacles complicate CTC analysis: rarity and heterogeneity of CTCs. To overcome these obstacles, researchers have focused on developing new CTC separation methods and techniques. Achieving high efficiency (ratio of captured CTCs to CTCs in a sample), high purity (ratio of captured CTCs to background similar cells like (WBCs)), high throughput (amount of sample being processed through a specific time), and high viability of captured CTCs are the main challenges for CTC separation methods (Dong et al. 2013). After the development of CTC detection methods, it was observed not all the CTCs are aggressive and only a small subset are metastatic. Characterization methods have also emerged to address the heterogeneity of CTCs (Maestro et al. 2009).

Isolation and detection of circulating tumor cells

With the discovery of CTCs, scientists began to delve into genetic and diagnostic information that could be collected from these cells (Williams 2013). However, the rarity of CTCs posed an obstacle to study them. Only 1–10 CTCs are available in one milliliter of blood surrounded by billions of blood cells. Therefore, isolation and detection methods emerged to address this shortcoming (Tsutsuyama et al. 2019). The purpose of these techniques is to first enrich CTCs by eliminating background cells (Song et al. 2017). Later, the isolated CTCs are identified using complementary detection technologies (Shen et al. 2017). Different bases of isolation and detection methods have provided different classifications for CTC-related approaches (Esmaeilsabzali et al. 2013). For instance, some technologies distinguish CTCs based on their biological differences from red and white blood cells, while other technologies rely on their physical differences (Chen et al. 2017). Another classification is based on either enriching CTCs directly by labeling and isolating them (known as positive selection methods) or removing unwanted healthy blood cells (known as negative selection methods). Taking the bulk or microfluidic approach provides another category of CTC isolation and detection techniques (Bankó et al. 2019). Different approaches for detection such as immunostaining (Kalluri and Weinberg 2009) and fluorescence in situ hybridization (FISH) (Swennenhuis et al. 2009) as well as gene sequencing (Rossi and Zamarchi 2019) are examples for CTC detections. All these techniques can be implemented at both macro- or microscales.

The FDA-approved CellSearch™ is the most well-known macroscale technology for the isolation and enumeration of CTCs (Kagan et al. 2002). In CellSearch™, CTCs are isolated by means of magnetic particles conjugated to EpCAM antibodies. The dependence on EpCAM expression in CTCs has limited CellSearch™ application to epithelial cancers such as breast, prostate, and colorectal cancer. CellSearch™ also suffers from not being able to detect CTCs that are undergoing EMT (Allard et al. 2004). A comprehensive review of the clinical applications of CellSearch™ and CTC analysis has been reported (Riethdorf et al. 2018). Microfluidic platforms have been proposed as a promising tool for efficient CTC separation since they provide great control over the forces acting on the cells. In general, microfluidic systems employed for CTC enrichment are divided into two main groups: active and passive platforms (Wyatt Shields Iv et al. 2015). In the active techniques, external fields such as magnetic, acoustic, or electric fields are applied for the isolation of CTCs (Sivaramakrishnan et al. 2020). Alternatively, in passive methods, hydrodynamic forces imposed by the geometry of the microsystem enable cell separation based on the physical properties of the cells (Mahmoudifard et al. 2018; Rostami et al. 2019). Microfluidic CTC separation and detection methods have been reviewed by others (Gascoyne and Shim 2014).

Characterization of circulating tumor cells

Metastasis is responsible for the majority of cancer deaths (Pantel et al. 2009). During metastasis, CTCs invade blood vessels through intravasation, to initiate a secondary tumor in a new site (Hanahan and Weinberg 2011). Different characteristics of CTCs may be involved in each step of the metastasis (Balkwill 2004). It is known that only a small fraction of CTCs with more aggressive phenotypes possess the potential to initiate metastasis. It is of utmost importance, therefore, to identify CTCs with greater metastatic potential (Pantel and Speicher 2016).

Many studies have been conducted to investigate the metastatic properties of CTCs such as their morphology (Fiorelli et al. 2015; Lazar et al. 2012) and expression of different surface proteins on CTCs—including HER2 (Wang et al. 2017), androgen receptor variant 7 (ARv7) (Scher et al. 2016), and thyroid-stimulating hormone receptor (TSHR) (Lin et al. 2016). Microfluidic platforms have provided powerful tools for the characterization of CTCs. Originally, microfluidic devices were designed for mere enumeration of CTCs. The number of CTCs present in blood samples has proven to be a prognostic biomarker for metastasis in different types of carcinomas including breast (Lee et al. 2016), prostate (Doyen et al. 2012), colorectal (Huang et al. 2015), and lung (Wang et al. 2013). As most microfluidic and non-microfluidic methods have been designed for the enumeration of CTCs, there is a pressing need to develop platforms with the ability to indicate aggressive CTCs (Saucedo-Zeni et al. 2012; Sha et al. 2008).

In an experiment conducted on mice to investigate different metastasis potentials of CTCs in single and cluster form, the metastatic potential of CTC clusters was found to be 23- to 50-fold more than single cells (Aceto et al. 2014). Based on this observation, the assumption that CTC clusters were not able to pass through narrow vessels in order to reach a distant organ was challenged and a microfluidic device with 16 parallel channels of different cross sections was designed to mimic human capillary constrictions (Fig. 4A) (Au et al. 2016). CTC clusters containing up to 20 cells were obtained from cancer patients. Once the samples were injected into the microfluidic platform, 90% of the clusters traversed through 5–10 μm-sized constriction. Through this process, clusters reversibly changed into single-cell chain-like geometries and passed through the microchannels. This phenomenon can potentially enable CTC clusters to pass through human organ capillary microenvironments. In order to investigate the effects of the microenvironment on single CTCs, a circulatory microfluidics-based system was fabricated in which different shear stresses could be produced (Regmi et al. 2017). Breast, ovarian, lung, and leukemic cancer cells were tested under varying flow conditions. It was reported that relatively high shear stresses of 60 dynes/cm² resulted in greater CTC death than lower shear stresses. Such high shear stress caused necrosis in 90% of CTCs and apoptosis was observed in surviving cells. In a different study, a microfluidic single CTC isolation platform was designed followed by injection to polymerase chain reaction (PCR) tubes (Yeo et al. 2016). High throughput and selective isolation of single classes of CTCs up to 100% purity were achieved. Moreover, two different mutations of L858R and T790M, which are representative of the dominant EGFR gene and associated with TKI response and resistance, were found in clinical CTC samples through sequencing (Yeo et al. 2016). With a slightly similar approach, an EpCAM chip was exploited to capture CTCs which was followed by a quantitative reverse transcription PCR (qRT-PCR) chip for gene expression profiling of epithelial and EMT-like CTCs (Zeinali et al. 2018).

Fig. 4.

Microfluidic devices designed for CTC characterization. A Microfluidic device designed to evaluate the ability of CTC clusters passing through capillaries. The platform consisted of 16 parallel microchannels of 5 × 5, 7 × 7, or 10 × 10-μm square cross sections. Schematic and real photo of CTC clusters while entering the capillary microchannel is also depicted. Reproduced with permission from (Au et al. 2016). B A multi-zone velocity valley device, fabricated by Mohamadi et al. with four different linear velocities for capturing and characterizing CTCs with different EpCAM expression (Mohamadi et al. 2015). C A platform for profiling functional and biochemical phenotypes of CTCs by a two-dimensional characterization CTCs according to the levels of a surface marker followed by a chemotactic phenotype sorting. Reproduced with permission from (Poudineh et al. 2017b)

One of the major sources of heterogeneity among CTCs is EMT through which the levels of EpCAM expression are altered (Pantel et al. 2009). In general, CTCs with a mesenchymal phenotype and decreased EpCAM expression have been associated with possibly greater metastatic potential (Hyun et al. 2016). Comprehensive studies on microfluidic approaches for functional characterization of CTCs have been reported (Green et al. 2019). A microfluidics-based chip was designed for spatial sorting and profiling of phenotypically-distinct CTCs into discrete bins (Mohamadi et al. 2015). CTCs labeled with anti-EpCAM magnetic nanoparticles passed through a channel with four compartments having different linear velocities and were captured by applying an external magnetic field. An array of X-shaped structures was fabricated throughout the channel to create regions of low flow velocity (called velocity valleys, VVs) to facilitate cell capture. As a result of this design, cells with varying levels of EpCAM expression and hence different numbers of bound nanoparticles were isolated in dedicated zones (Mohamadi et al. 2015). After CTC isolation, on-chip immunostaining was carried out to detect cells that were trapped in each zone. As a result, cancer cells with known EpCAM expressions generated unique capture profiles. For example, VCap cells with high EpCAM expressions were mainly captured in high-velocity zones as a consequence of having high numbers of bound magnetic nanoparticles. While low EpCAM expressing MDA-MB-231 cells were captured in low-velocity zones. The device was successfully tested with blood samples from 21 prostate cancer patients and showed distinctive profiles of EpCAM expression. Employing this technology in further studies, phenotypically differentiated CTCs were extracted from the four capture zones and subjected to functional assays to identify CTCs with more aggressive phenotypes (Fig. 4B) (Mohamadi et al. 2015). A two-dimensional sorting platform was also designed in which EpCAM-sorted CTC subpopulations were characterized based on the expression of a secondary marker (Labib et al. 2016). A chemotaxis profiling step was added to the Velocity Valley chip and a 2D phenotypic sorting device was fabricated where EpCAM-sorted cells were directed via application of a chemotaxis chip in which one cell-loading channel was connected to two chemoattractant reservoirs by migration channels (Fig. 4C) (Poudineh et al. 2017b). Each migration channel was divided into three different regions of minimal, intermediate, and high migration to assess the migratory potential of CTCs. It was reported that the loss of epithelial characteristics correlated with enhanced chemotactic migration of CTCs and therefore increased invasiveness. In another 2D functional characterization of CTCs, EpCAM-sorted CTC subpopulations were characterized based on their ability to digest collagen matrix as a representative component of ECM. CTCs with low EpCAM expression demonstrated increased collagen uptake and greater potential for invasiveness (Green et al. 2017).

To enhance the resolution of phenotypic sorting of CTCs, the velocity valley principles were utilized and a magnetic ranking cytometry (MagRC) technique was proposed for sorting CTCs in one hundred zones (Poudineh et al. 2017a). Circular nickel micromagnets with 100 increasing diameters were patterned throughout the chip to manipulate the magnetic force acting on EpCAM labeled cells. As a result, CTCs with high loads of EpCAM nanoparticles were settled near smaller micromagnets (early zones) and cells with lower magnetic load would reside in the zones with larger micromagnets. To improve the practicality of the MagRC technique, a modified MagRC technology was designed with a 10-zone platform to investigate the phenotypic profiles of CTCs in vivo (Kermanshah et al. 2018). The principle of the X-shaped magnetic phenotypic profiling microfluidic systems was implemented as a PCR-free mRNA cytometry technique where iron oxide nanoparticles functionalized with DNA capture probes complementary to the desired mRNA sequence were employed to capture cancer cells (Labib et al. 2018). Six velocity zones were designed to analyze the expression of different mRNAs and the technique was used to study the expression of three prostate-cancer-specific mRNAs of full-length androgen receptor (AR-FL), ARv7, and TMPRSS2/ERG. Recently, a CTC phenotyping microfluidic device named TU-chip™ was developed where the chip consisted of thousands of elliptical micropillars with a triangular arrangement to capture CTCs (Sun et al. 2019). Micropillars were arranged with three different gap sizes to trap a wide range of CTCs. Immunostaining was further exploited to identify CTC phenotypes of E, M, and H. Phenotypes E and M were referred to the expression of prototypical markers including (E-cad) and vimentin, respectively. However, phenotype H was defined as CTCs with the expression of EMT and mesenchymal-epithelial transition (MET) markers (Sun et al. 2019). In yet another recent platform, a microfluidic method has been proposed for simultaneous capture and membrane marker analysis (Armbrecht et al. 2020). Application of the platform allows for quantification of the secretion level of granulocyte growth-stimulating factor (G‐CSF) for breast cancer cell lines. Furthermore, the design provided a multiplexed platform for more comprehensive functional CTC profiling and characterization.

CTCs undergoing EMT can lose the EpCAM marker, resulting in a mesenchymal-like phenotype. A double marker method targeting both EpCAM and fibroblast activation protein alpha (FAPα) markers was integrated into a microfluidic device (Witek et al. 2017). FAPα—a marker associated with mesenchymal characteristics—was used and proposed as an additional marker for selecting phenotypically distinct CTC subpopulations. A system consisting of two connected microfluidic devices was also used to isolate both EpCAM + and CD133 + CTCs making the clinical evaluation more achievable and reliable (Zeinali et al. 2018). Recently, our research group has designed a 2-stage microfluidic chip based on the principle of deformability difference and VV for separation of leucocytes and CTCs and metastatic aggression characterization of separated CTCs (Hakim et al. 2021). The deformability-based chip (D-Chip) consisted of two microfluidic sections of separation and characterization. Slanted weirs with a gap of 7 μm (based on a previous study by (Yoon et al. 2019)) were considered in each section to separate cells with different deformability characteristics. CTCs were initially separated from leucocytes by passing over the weirs in a hall of slanted weirs, as a result of their inherent higher deformability (Sundling and Lowe, 2019). Then, the separated CTCs were put through a VV with 4 different levels of velocity at which a slanted weir was considered to separate CTCs with a certain range of deformability. MCF7 and MDA-MB-231 were used for the D-Chip performance evaluation. Separation efficiency of higher than 93% and significantly high purity were reported. 15 clinical blood samples from breast cancer patients were analyzed by the D-Chip revealing the relation between aggressive behavior of breast cancer CTCs and higher deformability (Hakim et al. 2021).

A comprehensive review of microfluidic and non-microfluidic CTC characterization methods has been provided by (Brown et al. 2019).

Cancer dissemination

One of the essential requirements for the cancer cells to spread over the body is to successfully invade circulatory systems and initiate the intravasation process. The term intravasation refers to the disruption of the endothelial layer and the subsequent entry of the invasive tumor cells either into the vasculature (hematogenous intravasation) or into the adjacent lymphatic circulatory system (lymphatic intravasation) (Chiang et al. 2016). Conventional in vivo studies do not replicate the invasion and intravasation processes of cancer cells. However, state-of-the-art microdevices have become invaluable assets to understand the underlying causes of these critical processes, potentially enabling one to unravel metastasis and eventually develop an effective treatment for this systemic disease. In the following sections, microfluidic technologies developed to study different factors influencing both intravasation and extravasation stages as well as simplified miniaturized systems to study organotropism of some cancers are reviewed.

Intravasation

The inherent invasive characteristics of cancer cells mean that they can find their way to the circulatory systems via passage through the vessel wall barriers during angiogenesis, lymphangiogenesis, and vascular remodeling (San Juan et al. 2019). Intravasation is facilitated by many biochemical and physical factors affecting the TME. The expression of VEGF, VEGFB, and matrix metalloproteinases (MMPs) or the presence of other cells such as macrophages, neutrophils, and cancer-associated fibroblasts can affect TME (Chiang et al. 2016). To obtain a clear explanation for the underlying mechanism of cancer cell intravasation, several studies have focused on developing miniaturized in vitro platforms to simplify the in vivo complexities and clarify specific factors’ exact roles and effects.

In one of the earliest studies on in vitro intravasation, a multi-purpose microfluidic device was used to investigate cell migration (Fig. 5A) (Shin et al. 2011). The device consisted of two main chambers meant to replicate intravasation and extravasation processes that were interconnected by two main valves. With the unique design of the platform, the separate study of intravasation and extravasation was made feasible. In order to model invasion assay, cancer cells were cultured in a matrigel matrix in the intravasation chamber. The criterion of intravasation occurrence was the observation of cells flowing out of the intravasation chamber. To investigate the criterion, two factors of degradation by MMPs and physical washout were evaluated. MMPs are a group of proteases whose biological activity degrade the ECM. Cancer cells can make their way through the remodeled and cleaved areas to invade the adjacent tissues. To investigate the effect of MMPs, the matrix was treated with MMP inhibitors such as MMP-2, MMP-9, and GM6001 inhibitors which demonstrated a decreased number of intravasated cells by 67%, 76%, and 78%, respectively. Furthermore, to evaluate the physical washout of cells from the matrix, different buffer flow rates were introduced to the channel. The final result showed as the higher shear stress was induced to intravasated cells by higher flow rates, less than 1% difference was observed in physical washout. Consequently, matrix degradation by invasive cancer cells was suggested to be an important factor in producing high intravasation rates (Shin et al. 2011). Since numerous factors are modulating the surrounding ECM in the TME, exploring their functions in intravasation process would be worthwhile. To further investigate other factors affecting the ECM degradation and subsequent intravasation, the role of macrophages was explored in the modulation of endothelial barrier where the microdevice that was used consisted of two main microchannels: one for seeding cancer cells and the other for endothelial cells (Zervantonakis et al. 2012). The two microchannels were connected by means of a channel loaded with hydrogel. Stimulation of the device with macrophages resulted in an increase of the intravasated cells (a ninefold increase) and also in a decay in the endothelial layer (Fig. 5B). Furthermore, the addition of TNF-α, a proinflammatory cytokine that alters endothelial cells permeability, provided the same result, which was an increase in both permeability and the number of migrated cells.

Fig. 5.

Illustrations of the microfluidic devices used in intravasation studies. A Schematic of the device consisting of two integrated parts for both intravasation and extravasation analysis. At the intravasation chamber, cells are 3D cultured, and by degradation of Matrigel and quantified. At the reaction chamber, cancer cells are treated with inhibitors to analyze cancer cell adhesion within the extravasation chamber. At the extravasation chamber containing poly-L-lysine, fibronectin, and HUVEC layers, extravasated cancer cells are quantified. Reproduced with permission from (Shin et al. 2011). B Three microchannels are fabricated for the intravasation study. Tumor, ECM, and endothelial microchannel are represented with red, green, and gray colors. Macrophage and TNF-α stimulated endothelium layer was used to assess their effect on cell intravasation. Reproduced with permission from (Zervantonakis et al. 2012). C. Schematic of the microchannel used for angiogenesis and intravasation study. The microchip consists of a microvessel containing LF and a vessel containing lung fibroblasts and HUVEC. Cancer cells are introduced into the bridge channel and attach to the fibrin wall. Then the channels are filled with media and intravasation is assessed. Reproduced with permission from (Lee et al. 2014)

A major problem with previous studies was that endothelial cells were formed adjacent to hydrogels as a monolayer, making the cell intravasation-mimicking process preliminary. One study focussed on the formation of a more physiologically relevant endothelial barrier using engineered microvessels that provided a number of distinct advantages. First, their continuous vasculature network imitated a more in vivo condition (Fig. 5C). Second, with the generation of the smooth and constant vessel, intravasation of the cells could be observed in an unobstructed fashion, making it a suitable option for other studies. The results indicated that TNF-α stimulated microvascular provided about 1.4-fold higher permeability in comparison with the normal condition (Lee et al. 2014). To tackle inefficiencies associated with assessing the role of biophysical cues in tumor cell intravasation, a novel microfluidic platform was devised to evaluate the transmural and luminal flow effects on tumor cell migration through the lymphatic endothelial layer (Pisano et al. 2015; Zuela-Sopilniak and Lammerding 2019). The results verified that both flows can enhance the migration capability of the cells. Moreover, the results indicated that the flows showed synergetic effects and suggested that transmural flows upregulate CCL21 expression by human dermal lymphatic endothelial cells which increased intravasation of cells. Although there has been numerous efforts to develop an in vivo-like microfluidic device to analyze the early stages of cancer, there are still some ambiguities concerning cellular interactions. A recent study has addressed the current challenges by developing a physiologically relevant in vitro microdevice to model cell–cell and cell–ECM interactions (Nagaraju et al. 2018). The device was made up of three main tumor microenvironment layers including the tumor, stroma, and vascular regions. The interconnection between the sections provides a spatially organized matrix for studying intravasation distinguishing this device from prior ones. The results showed that the presence of the endothelial vascular network affected the invasion and intravasation of cancer cells in many ways including an escalation in the number of the intravasated cells, elongation, protrusions in the shape of cells, and emergence of more phenotypically invasive cancer cells. The main features, key factors, and cell lines of the above intravasation models are presented in Table 2.

Table 2.

Highlights of the reviewed intravasation models at a glance

Vascular Cells	Cancer Cells	Key Factor	Matrix	Ref
HUVEC	colon cancer cells	Shear stress, MMP2, MMP9, and GM6001 inhibitors	Matrigel 4 mg/mL	(Shin et al. 2011)
HUVEC	breast cancer cells	TNF—α	Collagen 2.5 mg/mL	(Zervantonakis et al. 2012)
HUVEC and lung fibroblasts	breast cancer cells	TNF—α	Fibrinogen 2.5 mg/mL	(Lee et al. 2014)
human dermal lymphatic endothelial cell	breast cancer cells	Transmural and luminal flows	Collagen type I 1.5 mg/mL	(Pisano et al. 2015)
HUVEC	breast cancer cells	-	Collagen type I 1.5 mg/mL	(Nagaraju et al. 2018)

Extravasation

One of the prominent steps in the cancer metastatic cascade is the extravasation of cancer cells into the surrounding tissues and microenvironments to initiate the secondary tumor. The whole process of extravasation is still somewhat obscure; however, the underlying principles of this phenomenon have been investigated in numerous studies. In general, three major steps should be considered for extravasation: (i) the adjacency of tumor cells to the endothelium layer, (ii) the adhesion of tumor cells to the endothelium, and (iii) the transmigration of the adhered tumor cells through the endothelial monolayer (Giavazzi et al. 1993).

The way that cells, which have overcome the physical forces applied by blood flow, migrate toward secondary sites can be explained by the classic "seed and soil" theory (Paget 1889). This theory states that a specific relation and interaction exists between tumor cells, described as "seed," and the congenial microenvironment, defined as "soil." In the secondary microenvironment, some chemoattractant (biochemical factors) are present to navigate CTCs and facilitate the extravasation and secondary tumor growth (Roussos et al. 2011). Once CTCs arrive at the endothelium, they adhere to endothelial cells by expression of some adhesion molecules like integrins on the cancer cells and their selectin ligands on the endothelial cells (Sökeland and Schumacher 2019). The advent of microdevices has had a great impact in this area of research and many microfluidic systems have been designed to shed light on this complex phenomenon. Some of these platforms share the same principal features with intravasation chips regarding the formation of an endothelial layer. A multi-step microfluidic device was developed to analyze the deformability and migration rate of cancer cells (Fig. 6A) (Chaw et al. 2007). The device consisted of two separate chambers, one for analyzing deformability and the other for transmigration. In the first microchannel, various cancer cell lines were introduced to assess and compare cell viability and proliferation rate after deformation caused by passing through the 10 µm gaps. Subsequently, the deformed cells were directed toward the second microchannel where they were compared with the control cell lines (which were not deformed) in terms of transmigration ability. It was concluded that the deformation of cancer cells did not cause a drastic change in the migration potential. However, the addition of matrigel and endothelial monolayer to the assay in the second chamber (transmigration) diminished the total cell migration (Chaw et al. 2007).

Fig. 6.

Recently developed microfluidic systems to study extravasation. A A multi-step microfluidic device that contains two separate chambers for investigation of deformation and extravasation of cancer. Reproduced with permission from (Chaw et al. 2007). B Microfluidic system consisting of three independently addressable media channels, separated by chambers into which an ECM-mimicking gel can be injected. Cancer cells in the central channel extravasated toward the gel matrix and the number of invaded cells and permeability of gel was quantified. Unrestricted reproduction and use is permitted by (Jeon et al. 2013). C Conceptual design of a microfluidic device for transendothelial migration. After optimization of the pore size, the extravasated cells were collected in the microchamber below the membrane for subsequent characterizations. Reproduced with permission from (Cui et al. 2017). D Schematic representation of the microfluidic chip designed to inspect chemotactic migration of cells. The main channel (red) and side channels (blue) are shown in the figure. ECM contained in the side channels was treated with CXCL12 to study the chemotactic factor in extravasation of cells. Reproduced with permission from (Zhang et al. 2012)

In extravasation microfluidic platforms, the formation of an intact layer of endothelial cells that mimics the blood vessel wall (endothelium layer) is important. In one of the earliest studies, a microfluidic device was fabricated that consisted of three parallel channels connected by collagen-loaded chambers that mimicked the ECM (Fig. 6B) (Jeon et al. 2013). Human microvascular endothelial cells (hMVECs) were seeded into the middle channel to form a 3D endothelial monolayer lining the inner surface of the channel. Cancer cells, derived from a breast cancer line, were subsequently seeded into the middle channel. The whole process of cancer cell extravasation through endothelial monolayer was then observed and the permeability was determined with the help of a fluorescent dextran solution (Jeon et al. 2013). In conventional platforms, a monolayer of human umbilical vein endothelial cells (HUVEC) is formed to assess the extravasation potential of the cancer cell. In reality, however, CTCs have to pass through other compartments of the basement membrane of blood vessels that are primarily composed of fibroblast cells (Kalluri and Zeisberg 2006). To recapitulate this physiological condition, a microfluidic chip was fabricated with parallel hydrogel-loaded microchannels for co-culturing of HUVEC and normal lung fibroblast (NHLF) cells to establish microvascular networks (Chen et al. 2013). To further elucidate the crosstalk between cancer cells and the host microenvironment, an inflammatory cytokine (TNF-α) was introduced to the endothelial channel. The extravasation of different cancer cell lines was assessed and the results showed that the treatment of microvascular networks with TNF-α increased the tumor cell invasion (EXTRAVASATION) by 2.3 times. A new microfluidic design was introduced to tackle the challenges in the formation of the endothelial monolayer (Fig. 6C) (Kühlbach et al. 2018). The device benefited from a proper monolayer forming technique, which means the platform was being seeded in a monolayer with primary endothelial cells from the target organ of the metastatic tumor cells. The new device was made of a porous polymer membrane sandwiched between two PDMS channels. The upper channel was used to seed endothelial cells (EC) on the membrane while the bottom reservoir was used to collect the migrated cell. Taking advantage of this platform, further research can be conducted by introducing various chemokines and homing factors. Another contribution in this field involved the design of a multi-layer microfluidic device to analyze the properties of transmigrated cancerous cells (Fig. 6D) (Cui et al. 2017). The device was similar to the microfluidic apparatus discussed earlier (Kühlbach et al. 2018) in terms of using a porous polymer membrane. The optimization of the porous membrane was done to prevent cell alteration due to physical confinements of the pores. Results demonstrated that cells with higher migration potency had different cytoskeletal characteristics (Cui et al. 2017). For instance, migrated breast cancer cells showed a more inconsistent distribution of the filamentous actin. Additionally, an increase in the planar migration speed of the transmigrated cells was observed as opposed to lower speeds of non-migrated cells (Cui et al. 2017). In a different study, a microfluidic in vitro model was designed to study the chemotactic migration of circulating tumor cells toward secondary sites (Zhang et al. 2012). Salivary gland adenoid cystic carcinoma cells (ACC-M), which express a leukocyte chemoattractant receptor CXCR4, were used to investigate the interplay between CXCR4 and the CXCL12 ligands in extravasation (Zhang et al. 2012). The microfluidic platform was composed of one main channel, where CTCs flow in, and side channels, which included perivascular matrix containing CXCL12. In order to mimic the endothelium monolayer, HUVEC cells were seeded at the interface of the main channel and side channels. The transendothelial migration of ACC-M was assessed in response to varying concentrations of CXCL12. In the absence of CXCL12, ACC-M cells adhered to the matrix were unable to penetrate. By increasing the amount of CXCL12 from 100 ng/mL to 200 ng/mL, cells tended to show a higher rate of migration. Finally, the addition of an inhibitor completely halted the extravasation of cells; CXCL12 was inhibited with the assistance of AMD3100. Consequently, CXCR4 antagonist capability of neutralizing the CXCR4/CXCL12 interaction was verified and no extravasation was observed (Zhang et al. 2012).

Cancer cells express proteins like integrins that enable them to adhere to the endothelium layer. Another crucial set of molecules playing a significant role in the transmigration of cells is the members of the integrin family (Mitroulis et al. 2015). Throughout cancer progression, particularly in highly metastatic breast and ovarian cancer cells, β1 integrin is up-regulated to facilitate the formation of tumor aggregates (Lahlou and Muller 2011). This molecule also has a vital application in cell adhesion to the ECM by extension of protrusions into the subendothelial matrix. A microfluidic device was utilized to demonstrate the role of integrins in cell invasion (Chen et al. 2013). It was discovered that knockdown of β1 integrin via shRNA and antibody in cancer cells would result in deficient cells not being able to extend protrusions into the subendothelial matrix. Consequently, failure in extravasation was observed (Chen et al. 2016). In addition to biochemical factors, (Wirtz et al. 2011). One physical factor that plays a role in cancer cell dissemination is the hydrodynamic force exerted by blood in the circulatory system. A multi-layer microfluidic device was designed to study the effect of these hydrodynamic forces on cancer cell extravasation. The platform consisted of three major parts, a microfluidic channel, an elastic membrane, and a pneumatic layer (Huang et al. 2015). With the assistance of media flow induced by the partial vacuum, a different combination of fluid shear stress (FSS) and cyclic stretch (CS) was applied to study the cell-epithelial monolayer adhesion. Results showed that under a capillary microenvironment condition, cells show greater adhesion capacity as well as a higher possibility of extravasation (Huang et al. 2015).

Given the important role of immune cells like T cells, such as T lymphocytes, through cancer progression, the role of these cells has been the subject of numerous investigations (Liu et al. 2018; Marshall et al. 2016; Sun et al. 2017). Surprisingly other immune cell types such as immune cells including monocytes and macrophages have not been placed under as much scrutiny. It is believed that different subpopulations of monocytes differentially affect cancer cell invasive behavior (Boussommier-Calleja et al. 2019). The microfluidic device had five channels, two of which were used for supplying the device with media while the other channels contained a mixture of hydrogel and cells (Boussommier-Calleja et al. 2019) was used to investigate the effect of monocytes on the extravasation of cancer cells, with a particular focus on two different subsets of monocytes, inflammatory, and patrolling monocytes. Simultaneous perfusion of cancer cells and monocytes caused a reduction in the transmigration of cancer cells (Boussommier-Calleja et al. 2019). However, drastic changes in the number of extravasated cancer cells were not observed upon the introduction of monocytes to the microvascular milieu in the first place (Boussommier-Calleja et al. 2019). Moreover, it was suggested that the decrease in transmigration was mainly a result of monocyte-associated paracrine signaling.

The specific role of neutrophils in the TME and the extravasation process is controversial; some believe that they contribute to cancer cell invasion through activating MMPs, while there is also evidence that they have tumor-suppressing effects (Balkwill et al. 2012; Powell and Huttenlocher 2016; Uribe-Querol and Rosales 2015). One study sought to clarify this issue by investigating the role of neutrophils in extravasation (Spiegel et al. 2016). Minor modifications were made to the geometry of the microfluidic device used in a previous investigation (Chen et al. 2013) by introducing splenocytes into the new system. It was demonstrated that neutrophil-mediated release of MMP8 and MMP9 renders the cancer cells more invasive. Table 3 summarizes the main features of microfluidic studies on cancer cell extravasation in terms of the implemented cancer cells, vascular cells, matrix components, and other key factors.

Table 3.

A summary of different features of microfluidic studies on cancer cell extravasation.

Vascular Cells	Cancer Cells	Key Factor	Matrix	Ref
Microvascular endothelial cells	Hepatocellular carcinoma; cervical carcinoma; breast carcinoma	—	Matrigel (micro-gap coating): no matrix	(Chaw et al. 2007)
Umbilical vein endothelial cells	Salivary gland adenoic cystic adenocarcinoma cells	CXCL12 and AMD3100	Basement membrane extract	(Zhang et al. 2012)
Microvascular endothelial cells	Breast cancer cells	—	Collagen type I (2 mg/mL)	(Jeon et al. 2013)
Fibrosarcoma cell line endothelial cells	Breast cancer cells	TNF-α	Fibrin gel (2.5 mg/mL)	(Chen et al. 2013)
HUVEC	HeLa cells	FSS and CS forces	—	(Huang et al. 2015)
HUVEC + NHLF	Breast cancer cells	β1 integrin	2.5 mg/mL fibrin and 0.24 mg/mL collagen I mixture	(Huang et al. 2015)
HUVEC + NHLF	Breast cancer cells	Ly6G + neutrophils from splenocytes	2.5 mg/mL fibrin and 0.24 mg/mL collagen I mixture	(Spiegel et al. 2016)
Primary human vascular endothelial cells	Breast cancer cells	CXCL12	Poly-D-lysine (4 lg/cm2) and 50 lg/mL fibronectin	(Cui et al. 2017)
Human primary pulmonary arterial endothelial cells (HPAEC)	H838; SK-Mel 28	—	Membrane (PET), thickness of 19 µm, pore size of 5 µm	(Kühlbach et al. 2018)
HUVEC	Breast cancer cells	Monocytes isolated from human blood	6 mg/mL fibrin gel	(Boussommier-Calleja et al. 2019)

Circulating tumor cells homing and secondary tumor

Clinical data suggest that there are specific patterns of metastasis, and different organs welcome particular types of cancer cells (Obenauf and Massagué 2015). For instance, prostate cancer generally metastasizes to the bone, while breast cancer shows a greater propensity to metastasize to the kidney, brain, and bone (Witzel et al. 2016). The intrinsic trait of organ tropism of cancer cells is rather a blurry phenomenon and significant effort has been put into clarifying this topic (Obenauf and Massagué 2015). In this regard, microfluidic models have been developed to represent the organotypic tissues and capillaries that are involved in the metastasis process in order to probe the organ-specific cell trafficking and to make specific homing of cancer cells more simple and clear. In a recent study, the specificity of breast cancer metastasis to bone was investigated using a microfluidic device containing a 3D osteo-cell-conditioned microenvironment with human osteo-differentiated bone marrow-derived mesenchymal stem cells (hBM-MSCs) which, in comparison with non-treated matrix, showed an approximately 40% higher cell extravasation and a farther distance of cell migration (Bersini et al. 2014). Further experiments revealed that the pathway between the secreted CXCR2 cancer cell receptor and bone-secreted chemokine CXCL5 was the main reason for specific homing as well as the driving force for breast cancer cell migration toward bone tissue (Bersini et al. 2014). Built upon this study, a 3D perfusable microvascularized bone-mimicking microenvironment was developed that overcame the deficiencies of an in vivo experiment in terms of controlling the variable parameters (Fig. 7A) (Jeon et al. 2015). The microenvironment embodied a co-culture of differentiated hBM-MSCs and human umbilical vein endothelial cells. The ability of breast cancer cells was tested with the help of the in vitro model that used a more physiologically relevant endothelial monolayer. The quantitative results demonstrated that the katushka-expressing bone-seeking clone (BOKL) of the MDA-MB-231 metastatic breast cancer cells exhibited a greater inclination toward extravasation to the bone-mimicking microenvironment. On the other hand, matrices conditioned with myoblast cells and matrices without stromal cells had a relative lack of cancer cell migration which was indirect proof for the antimetastatic roles of molecules like adenosine in the TME (Jeon et al. 2015).

Fig. 7.

Examples of microfluidic approaches for homing studies. A A representative of the microchannels to investigate the extravasation of the breast cancer cells to bone-mimicking microvascular network. Endothelial cells form a microvasculature in the central gel channel. Reproduced with permission from (Jeon et al. 2015). B The design of a microchip to simulate BBB. The device contains 16 independent parts with four BBB regions at each. At BBB regions physiologically relevant blood–brain barrier is formed. To simulate the specific extravasation process different cancer cells are introduced to the vascular channel and their extravasation to the brain compartment (shown in orange) is observed. Reproduced with permission from (Xu et al. 2016). C The illustration of the upstream and downstream organ-on-chip models designed to analyze the brain metastasis of lung cancer cells. At the lung microchip, bronchial epithelial cells and pulmonary vascular endothelial cells are adhered to each side of a thin porous membrane. Tumor cells extravasate through the membrane and enter the brain microvessel coated by human brain microvascular endothelial cells. Reproduced with permission from (Liu et al. 2019)

Patients suffering from lung cancer usually face a severe brain metastasis with an approximately 40 to 50 percent occurrence probability (D’Antonio et al. 2014). This is also observed to a lesser degree in melanoma and breast cancer patients (Lin et al. 2013; Vosoughi et al. 2018). To explore these probabilities, a 3D dynamic microfluidic device capable of recapitulating the physiological condition of the blood–brain barrier (BBB) was designed to evaluate extravasation of different cancer cells to the proposed in vivo-like brain microenvironment (Fig. 7B) (Xu et al. 2016). MDA-MB-231 (breast cancer), M624 (melanoma), A549 (lung cancer), and BEL-7402 (liver cancer) cell lines were evaluated by the platform (Xu et al. 2016). The final results were in agreement with the clinical observations in that even though breast, lung, and melanoma cancer cells can extravasate through the BBB, liver cancer cells were not able to extravasate (Xu et al. 2016). Moreover, reverse experiments were conducted by including an aggressive brain carcinoma, U87 glioma cells, to evaluate metastasis potential to the brain. In agreement with clinical data, the U87 cells were not able to pass through the BBB and remained mostly in the brain compartment of the device (Xu et al. 2016). For actual brain cancer, glioma cells stay in the cerebral spinal fluidic space. To further clarify the ambiguity in the case of lung cancer metastasis to the brain, a multi-organ-on-a-chip device to model the key elements in brain metastasis was designed that consisted of two major sections; the lung resembling chip in the upstream and the brain mimicking chip featuring the blood–brain barrier in the downstream (Fig. 7C) (Liu et al. 2019). The results showed that during the metastasis and extravasation of cancer cells, expression of proteins like MMP-2 and MMP-9 was upregulated, confirming what was suggested by many previous studies (Feng et al. 2011). The inclusive parametric study also revealed that the destruction of tight junctions in the blood–brain barrier structure leads to brain metastasis and that this process was related to the overexpression of the Aldo–keto reductase family 1 B10 (AKR1B10) protein-based release of the matrix of metalloproteinases. Table 4 summarizes the main features of studies involving microsystems for analysis of CTC homing and secondary tumors.

Table 4.

A summary of the main features of studies involving microsystems for analysis of CTC homing

Vascular Cells	Cancer Cells	Key Factor	Matrix	Ref
HUVEC, hBM-MSC	breast cancer cells	CXCR2/CXCL5	Collagen type I (6 mg/mL)	(Bersini et al. 2014)
HUVEC, primary and differentiated hBM-MSCs	breast cancer cells	A3AR, adenosine	Fibrin gel (5 mg/mL)	(Jeon et al. 2015)
Human microvascular endothelial cells, astrocytes	brain, lung, breast, and liver cancer cells	—	Collagen type I	(Xu et al. 2016)
Rat brain microvascular endothelial cells, astrocytes	PC9-BrM3 lung cancer cells	AKR1B10	Collagen type I,	(Liu et al. 2019)
			fibronectin (each 100 μg/mL)

Conclusions

Breakthroughs in microfabrication have led to the development of powerful tools in the field of precision oncology. In the current review paper, we focused on the application of microfluidic devices with the purpose of untangling the dynamic mechanisms behind the metastatic cascade. Microfluidic designs used to mimic the tumor microenvironment were reviewed. Tumor-on-a-chip devices enable researchers to study different progression phases of a tumor from the early stages of cell necrosis and angiogenesis to the latter metastatic state. An insight into the application of CTC-based liquid biopsy devices for the characterization of the molecular signatures and phenotypes was presented. The use of microfluidic devices in the study of intravasation and extravasation of cancer cells (as two of the most important stages during metastasis) was also reviewed. Clinical data has long indicated that cancers with different origins have intrinsic and unique organotropisms. Recent microfluidic devices that try to represent the organ-specific metastasis of specific cancers were also reviewed. Evaluating the particular features of each patient’s tumor and predicting a pattern of metastasis can be an innovative as well as practical approach in the treatment of metastasized cancer cases—since prevention is better than cure.

Funding

No fund was used for this review study.

Declarations

Ethical Approval

Not applicable.

Consent to Participate

Not applicable.

Consent to Publish

Not applicable.

Conflict of Interest

The authors declare no conflict of interest.