An Overview of Preclinical Solid Tumor Models to Study Novel Therapeutics in Brain Metastases

Abstract

Metastases are the most common malignancy of the adult central nervous system and are becoming an increasingly troubling problem in oncology largely due to the lack of successful therapeutic options. The limited selection of treatments is a result of the currently poor understanding behind the biological mechanisms of metastatic development, which in turn is difficult to achieve because of limited preclinical models that can accurately represent the clinical progression of metastasis. Described in this unit are in vitro and in vivo model systems that are used to enhance the understanding of metastasis, and for identifying new therapies for the treatment of brain metastasis.

Introduction

Metastases are secondary malignancies that form when cancer cells break away from a primary tumor and spread to other locations within the body (1). Of all organs for these escaped tumor cells to colonize, the brain accounts for an estimated 9–17% of cases in patients with systemic cancer (2), though the exact incidence varies with several factors, such as primary site, staging, and even subtype. Brain metastases are the most frequently occurring tumor type of the adult central nervous system (CNS), being 10-fold greater than primary malignant brain tumors (3, 4). The frequency of brain metastases development varies with the primary cancer, occur most frequently with lung (20%), melanoma (7%), and breast (5%) primary cancers (5). As treatments for primary cancers improve, metastases are quickly becoming the major cause of morbidity and mortality in patients with solid cancers (1, 6). This is due in part to improved control of the primary tumors, resulting in longer survival times and subsequently more time for brain metastases to develop, as well as enhanced detection methods (7). While limited treatment options are available for survival of brain metastasis patients beyond a few months (5, 7), groundbreaking therapeutic modalities for various patient subsets are improving control and overall prognosis of brain metastasis (8–10). Several target-specific therapies have been developed for brain metastasis possessing various molecular alterations, such as EGFR (11), HER-2 (12), and BRAF (13, 14). Although meaningful extensions of survival are being achieved, only a subset of patients respond to therapy, and many of these therapies are associated with limiting side effects.

The mechanisms regulating brain metastasis initiation remain elusive, partly due to the inability to identify early stages of metastatic progression with current detection and visualizing techniques, thereby limiting, the development of preventative therapies (15). It is therefore essential to have in vitro and in vivo models that can recapitulate clinical disease progression to gain insights into the metastatic process and for identifying new therapeutic targets.

The Metastatic Process

Metastasis occurs when cancer cells successfully migrate from a primary tumor to secondary sites in the body. The metastatic process follows a sequence of well-defined steps that allow cancer cells to grow beyond their primary tumor site (16, 17). To detach from the primary tumor mass, the cancer cells first downregulate intercellular adhesion molecules such as cadherins and integrins (18). Normal cells that detach from their surrounding tissue typically undergo cell death via anoikis (19). However, migrating cancer cells ensure their survival by auto-activating survival signaling pathways such as FAK and the PI3k/Akt (20). Without requiring anchorage to surrounding cells or the extracellular matrix, the migrating cancer cells can continue traveling towards vasculature.

Upon reaching blood or lymph vessels, migrating cancer cells intravasate past endothelial cells and enter the systemic circulation (21). They must avoid immune detection or risk elimination. Metastasizing cancer cells employ a variety of strategies, from recruiting immunosuppressive cell populations to hijacking the function of normal immune cells (22). Upon successfully reaching distant sites, the migrating cancer cells extravasate from the circulatory system and colonize into satellite tumors (23). Some cancers subtypes are particularly suited for particular secondary metastasis sites, a phenomenon known as the seed-soil hypothesis (24).

Blood Brain Barrier

For brain metastasis to occur, the migrating cancer cells must penetrate the blood-brain barrier (BBB), a semipermeable border in the brain’s blood vessels that tightly regulates the movement of cells and solutes from the bloodstream into the CNS (25). The BBB is primarily composed of endothelial cells, mural cells, basement membrane, and astrocytes that form a network to prevent pathogens and toxins from infiltrating neural tissue (25). However, this barrier also prevents most circulating immune cells from entering the brain (26). Metastasizing cancer cells increase leakage of the BBB by disrupting endothelial cell transporters such as Mfsd2a (27) and once inside the brain, cancer cells utilize the BBB to shield themselves from immune surveillance (26). Although the exact mechanisms regulating cancer cell infiltration of the BBB is unknown, the compromised BBB is restructured into what is referred to as the blood-tumor barrier (BTB) (28). The BTB contains heterogenous vasculature that actively nurtures metastasized cells before they develop their own blood vessel networks in the brain (29). The dual role of the BBB in protecting metastasized cancer cells from the immune system and preventing drugs from entering the brain makes treatment of brain metastases particularly challenging (30).

The exact mechanisms regulating the metastatic process remain elusive, partly due to the inability to identify the early stages of metastatic progression with available detection and visualizing techniques (31). The development of assays and models that recapitulate the features of clinical disease progression is a way to address this issue.

In vitro models of metastasis

An ideal in vitro assay should mimic at least one of the following features of the metastatic cell in vivo: proliferation, migration/invasion and morphological characteristics, cellular and phenotypic heterogeneity, microenvironment and drug response (Table 1) (41). However, despite the fact that complexity of the metastatic cascade makes it difficult to replicate each step precisely, a number of assays have been developed that have proven highly valuable. Cell proliferation and viability can be assessed by utilizing methods that quantify the metabolic conversion of a natural or synthetic substrate (42). Other tests have been developed to interrogate the migratory and invasive capacities of cells, such as scratch-wound and zone exclusion assays (43); an empty “zone” is created into which motile metastatic cells migrate. The addition of an extracellular matrix (ECM) layer, such as Matrigel or collagen, mimics a basement membrane layer for cells to “invade” through, as occurs with transwell and Boyden chamber assays (43). These assays can be exploited further by incorporating a chemoattractant factor or chemotaxis. The concentration gradient of small molecules or chemokines can be lightly regulated as with modified transwell or capillary assays, or tightly controlled temporally and spatially with microfluidics (44, 45). Several studies have shown metastatic cells will follow a gradient or chemokine trail to “home” to their secondary destination (46, 47). For instance, the release of CXCL12 and CXCL16 by metastatic brain stroma has been shown to attract breast cancer cells (48). Likewise, CXCR4, CXCR7 and CXCL12 expression has been implicated in the homing of cancer cells to the brain (49, 50). Other cytokines can assist other stages of metastasis, such as aiding the transmigration of metastatic cells across the BBB (51), promoting vascular remodeling (52), and remodeling the metastatic cell cytoskeleton to enhance mobility (53). Time lapse microscopy allows for the observation of the morphological minutiae in cell motility, from determination of cell front velocity and zone closure speed, to the morphological changes a metastatic cell attains to identifying the type of motility (46, 54).

Table 1:

Major Assays for Metastasis

Select in vitro models
Model	Description	Advantages	Limitations	Reference
Invasion/Migration Assays	Uses chambers with a wide range of chemical and properties to assess motility of cancers cell lines across physical distances	Low costs for high cell throughputs; allows for comparisons between cancer cells for their metastatic potential under a variety of conditions	Cells grow in 2D; cell migration in a petri dish may not reflect how metastasis occurs in living systems	Katt et al.(32)
Tumorspheres	Growing cancer cell lines in suspension or low binding plates to allow outgrowth in three dimensions	Since cells grow in 3D, this model more accurately represents how both mechanical and chemical factors can affect tumor formation; effective for screening anti-metastasis drugs	Many cell lines may not form tumorsphere structures, limiting this model’s use to only small subsets of cancer populations	Lee et al. (33)
Organoids	Three-dimensional cell culture technique: organoids are grown from embryonic or stem cells and resemble many of the phenotypes found in tissue of origin	intercellular communication networks and biological signaling are similar to that of original organ, allowing for more accurate drug screen assays	Random and uncontrolled growth of organoid cancer cells is common; does not accurately reflect cell communication from cells outside the original organ	Drost et al.(34); Hynds et al. (35); Kim et al. (36); Klein et al.(37)
Select in vitro models
Model	Description	Advantages	Limitations	Reference
Syngeneic Model	Engraft cancer cells that derive from the same species as the host animal	Animal retains fully competent immune system; effective for testing new immunotherapies	Does not always reflect human immune system dynamics, since animal immune system is fundamentally different	Saito et al. (38)
Humanized Mouse Model	Engraft human immune system issue into immune-deficient mice	Fast experimental kinetics; reflects human-like immune environment in animals	Humanized mice have high mortality rates than wild type mice. Grafting efficiency can also be variable from mouse to mouse.	Yin et al.(39)
Genetically Engineered Mouse Model	Knockdown/upregulation of select genes in a model mouse	Allows for isolation and subsequent study of specific genes/oncogenic pathways	Expensive; Genetically engineered mouse lines also take a long time to establish.	Walrath et al.(40)
Xenografts	Engraft human tumor tissue into humanized or immunodeficient mice	Reflects the tumor microenvironment of the original cancer sample accurately.	Many tumor types will not successfully engraft or metastasize after implantation	Saito et al. (38)

These in vitro assays are largely two-dimensional (2D), and so lack significant factors such as the microenvironmental influence and cell-cell interactions. These deficiencies limit the translational potential of these assays. Significant efforts have been expended on developing three-dimensional (3D) assays to retain these essential interactions, such as incorporation of stromal cells (immune cells, endothelial cells, fibroblasts, etc.)(55) and enhanced matrices and scaffolds that allow adhesion (56).

Tumorspheres are spherical structures generated by cancer stem cells (CSCs). Studies have shown that this particular subpopulation within the bulk tumor possesses self-renewal properties leading to tumor recurrence, differentiation, metastasis and drug resistance (33, 57). Methods involve culturing cell lines and allowing them to aggregate on low-binding plates or hanging in drops of serum-free media and a cocktail of growth factors. The screening of potential anti-metastatic agents is more translatable to in vivo anti-tumorigenic activity when conducted with tumorspheres as compared to 2D monolayer cultures (33). Many small molecule inhibitors of components of signaling pathways vital to CSCs have entered clinical trials, such as GDC-0084 and Regorafenib, and others have gained FDA approval, such as RO4929097 (58), sonidegib (59), and glasdegib (60, 61).

Organoids, which are derived from tumor biopsies or resections, are characterized by their ability to self-organize into hierarchies that reflect their derivative tissues (34). Organoids have greater similarities to patient samples as compared to cells grown under 2D culturing conditions (62). A recently described procedure utilizing human cancer cells and human embryonic stem cell-derived cerebral organoids can successfully reproduce metastatic cancer processes and serve as a useful tool for drug screening, with lung-specific X protein (LUNX) found to play an important role in cell proliferation and migration or invasion (63).

Organotypic assays evaluate tumor cell invasion into brain tissue and offer an alternative to in vivo models. A few variations of these models have been developed. Typically, a mouse brain is sliced to a specified thickness, cultured at an air/liquid interface, and tumor cells or spheroids either plated right beside, or implanted directly into, the slice. The objective of this assay is to track cell infiltration into the neural tissue and thereby provide a better evaluation of metastatic cell invasion as it encounters the extracellular matrix (64). Organotypic assays have proven superior to 2D models as they take into account the unique ECM composition in the brain, maintaining their normal cytoarchitecture, complex cell relationships and biochemical and electrophysiological properties, while avoiding the costly, labor intensive and lengthy study periods associated with in vivo models (41, 65, 66).

The type of cell line selected for in vitro assays driven by the experimental endpoint. Commercial cell lines are readily available, with culture conditions known and that are characterized by relatively simple maintenance and rapid growth rates. Commercial cell lines are also often easy use for genetic manipulation, such as for loss-of or gain-of function studies and to allow clonal selection and expansion to select for transduced populations (67). However, as these cell lines have been developed over years or decades in culture conditions lacking the original microenvironmental cues and factors, they represent only a small, homogenous subpopulation of the original patient samples, thereby lacking the original tumor heterogeneity and displaying different genetic and epigenetic characteristics (68). Additionally, while the use of these cell lines should theoretically enable the generation of results that are comparable across laboratories, this has not been found to be the case, with many reports of inconsistencies reported for drug sensitivity (69, 70). Primary patient-derived cell lines are low passage lines developed from patient tumor samples. These lines more accurately reflect the molecular characteristics of the original patient tumor than do commercial cell lines. Because these model systems display better accuracy at predicting the success rates of potential therapeutics, they may be useful in making clinical decisions and in improving the field of personalized medicine (67, 71). However, use of patient-derived cell lines is hindered by the difficulties associated with obtaining enough samples or cells for experiments and the long latency periods for tumor development in vivo.

These in vitro models are valuable tools for elucidating functional and biological metastatic properties that may not be detectible with in vivo models. These assays are rapid, easy to perform, relatively inexpensive and lend themselves to high throughput drug screening. Additionally, by focusing on single metastatic stages these assays allow for an assessment of the influence individual molecules and genes by making it possible to genetically manipulate specific steps of tumor progression and metastasis (67). However, given the limitations of in vitro assays, the resultant data may not accurately reflect what occurs in vivo.

In vivo models of metastasis

The value of an in vivo model in identifying novel therapeutic candidates is the ability of the model to replicate the clinical progression of the disease in question (Table 1). Rodents (typically mice) have proven to be especially valuable for developing models that advance our understanding of the pathophysiology and etiology of cancer (72) and for the evaluation of novel candidate therapies (73). A number of patient-derived xenograft (PDX) models have been generated for studying brain metastasis. With these in vivo models, PDXs can retain better tumor heterogeneity, biology and microenvironmental conditions as compared to in vitro models (74). Use of these in vivo models has led to the proposition of several key genes and pathways that are essential to metastatic progression(75) as well as the identification of novel drug candidates for the treatment of this condition (67). Several varieties of spontaneous, induced, and experimental brain metastasis models have been established with commercial and patient-derived cell lines. These include tissues of both human and mouse origins, and involve cancers induced by genetic engineering, chemical or UV exposure. The route of cell inoculation can also determine the location of metastatic formation (76).

1. Syngeneic

Several established preclinical models have led to the discovery of key genes and pathways that are essential to metastatic progression. The syngeneic mouse model is one of the oldest models of brain metastasis and involves injecting allografts from immortalized murine cancer cells directly into mice (76). Both the cell donor mouse and host mouse typically come from the same inbred strains. Since the species of the cell line and the host animal match, syngeneic models allow the mouse to retain a fully competent immune system (77). This feature ensures that syngeneic models are particularly well suited for studying the efficacy of immunotherapies as well as interactions between cancer and immune cells (77). Commonly used murine cell lines for assessing brain derived cancers include GL261, CT-2A, and SMA-560 for glioma syngeneic models, all of which have been used for studying response to immunotherapy(78–80).

With spontaneous syngeneic mouse models, the engrafted cell lines are derived directly from cancers that occur naturally in mice (81). The primary advantage of this model is quick progression of tumor metastasis in the host immunocompetent mouse. With a standard genetic mouse model of cancer progression, a brain metastasis assay could take months to complete (81). Spontaneous syngeneic mouse models for studying brain metastasis typically employ three neurotrophic cancer subtypes: melanoma, lung, and breast (82). Examples of commonly used spontaneous cancer cell lines used to study brain metastasis include 4T1 mammary gland tissue (83), WM239 melanoma (84), and 113/6–4L melanoma (84). The 4TI spontaneous model was used, for example, to establish ID2 as a promoter of breast cancer brain metastasis, a promising therapeutic target (83). Similarly, the WM239 melanoma model allowed for the identification of endothelin receptor B (ENDRB) as a potential target for treating melanoma brain metastasis (85). In addition to identifying genetic markers, spontaneous syngeneic models are used to discover novel drugs for treating brain metastasis patients. The tyrosine-kinase inhibitor Neratinib, used for treating patients with Her2 amplified breast cancer, inhibited brain metastasis when tested in spontaneous syngeneic experiments (86).

Induced syngeneic mouse models differ from spontaneous models in that the engrafted cancer cell lines come from healthy animals exposed to carcinogens (76). Chemical and UV exposure are two common techniques used to induce cancer in the donor animals (76). Chemical induction of lung cancer can be performed by injection of urethane and nicotine-derived nitrosamine ketone (NNK), while breast and skin cancers are induced by 7,12-Dimethylbenz(a)anthracene (DMBA) oral gavage and topical application, respectively (87). Ultraviolet exposure remains the most effective method for inducing murine melanoma, with the K-1735 cell line serving as a model system for studying brain metastasis (88). Syngeneic models with K-1735 cell lines were utilized to highlight the role of cytokines such as TGF-beta2 in facilitating melanoma metastasis (89), which ultimately led for TGF-beta2 inhibitors like Fresolimumab and Cemiplimab (90). Chemically induced syngeneic models have a number of advantages for preclinical testing, from their reproducibility to strong similarities with the human condition (91). Typically, syngeneic mouse model experiments for studying brain metastases typically use spontaneous or UV-induced cancer cell lines.

2. Humanized

Humanized mouse models are generated by engrafting human-derived immune system tissue and immune cells into immunodeficient mice. The model follows the fast kinetics of syngeneic mouse experiments while also reflecting a human-like immune environment in the mouse (92). The mouse and human immune system are fundamentally different, with mice having lower activation of the innate system and more naïve lymphocytes (93). As many immune system drugs are species-specific, humanized mouse models present clear advantages for pre-clinical testing by allowing for the laboratory testing of human immune system drugs (92). Because immunotherapy is a powerful treatment option for managing some brain metastases, humanized mouse models provide an excellent platform for testing new CNS metastasis drugs (94).

The Hu-PBL-NSG mouse model is a popular humanized model for evaluating immuno-oncology drugs. It is created by engrafting peripheral blood mononuclear cells (PBMCs) into immunodeficient NSG strain mice (95). Humanized mouse models were used to test the efficacy of anti-PD-1 immune checkpoint inhibitor drugs such as pembrolizumab in primary tumors of melanoma and lung cancer (95, 96). These immune checkpoint inhibitors are now employed for treating CNS metastases. In a study of patients with melanoma and lung CNS metastases, pembrolizumab showed promising activity in the brain (97, 98). Other immunotherapies such as ipilimumab and nivolumab were also initially developed using pre-clinical mouse models and are now used to treat melanoma brain CNS metastases (99, 100). These new immunotherapy drugs highlight the importance of humanized mouse models in shaping clinical practice.

3. GEMM

Genetically engineered mouse models (GEMMs) are divided into two categories: transgenic GEMM, where exogenous oncogenes are expressed through inoculation of a zygote or embryonic stem cells possessing the construct of the gene of interest, and targeted GEMM, incorporating homologous recombination in mouse embryonic stem cells (101, 102)). With GEMMs tumors develop de novo in a natural immune-competent environment, where the histopathological and molecular features of the resulting tumor closely resemble those found in humans and are even able to spontaneously metastasize (102). The first set of such “onco-mice” models provided unequivocal proof that oncogenic expression in normal cells could lead to tumor development (103). The subsequent advancement in gene-targeting technology permitted the study of tumor suppressor gene (TSG) knockout mice (104, 105). TSGs are divided into two groups: “gatekeepers” that regulate tumor cell growth through control of proliferation and apoptosis, and “caretakers” that regulate cellular processes responsible for genetic alterations (106). These models, one of the most frequently utilized models in cancer research, have been developed for several important TSGs, including Rb (107), p53 (108, 109), Ink4a/Arf (110, 111), Brca1/2 (112, 113). The benefit to GEMMs is the ability for spontaneous metastatic development in an immunocompetent host, thereby mimicking the clinical condition, but also allowing for the study of micro-environmental and immunology components of the metastatic process (114). As syngeneic cell lines can also be applied to GEMMs, these models are of particular interest to those involved in brain metastasis research in relation to the immune system involvement. They also form brain metastasis more quickly than in PDX models (82, 115–118).

4. Xenograft

The PDX models of brain metastasis are derived by transplanting fresh human tumor tissue or early passage patient-derived cells into immunodeficient or humanized mice; the lack of an immune system or immune system of the same species promotes higher tumor engraftment frequency. The benefit to PDX models is better retention of the genomic, histopathological and phenotypic heterogeneity of the original sample, thereby possibly improving the screening of potential therapies as well as their value in personalized medicine (119, 120). Currently, PDX models are utilized for large scale and high throughput drug screening studies to predict responses to clinical therapeutic candidates (121).

Routes of injection

The location of cell inoculation is key to the type of tumor that will develop and how closely the established model recapitulates the clinical condition. Tumor implantation in model locations other than the brain or primary organ are termed heterotopic or ectopic. Typically, the route of inoculation is intravenous (IV), where tumor cell dissemination will vary due to circulation. Delivery of cells through the tail vein will often result in metastases developing predominantly in the lung and subsequently to the CNS. Administration by this route has a very high survival rate and is easy to execute (122, 123).

To avoid pulmonary localization of cells, IV inoculation through the left heart ventricle, termed intracardiac, permits systemic distribution of cells throughout the body. Intracardiac models are useful in determining the homing propensity of tumor cells, where the seed vs. soil theory suggests that cells have a predisposition for certain secondary environments despite circulation (124). While not as robust as tail vein models, intracardiac models still have a high survival rate and the procedure is relatively simple to perform. To minimize the spread of tumor cells to tissues other than the brain, intracarotid artery inoculation is performed. This model can be time-intensive, requiring skills in microsurgery for the required ligation of the artery, and has a high intraprocedural mortality rate (76, 125).

For some IV studies, cells that have managed to metastasize to the brain are gathered and re-inoculated, creating cell lines that are “brain-trained”. This method ensures that cells home specifically to the brain despite the IV route used and can increase the frequency of metastatic formation in the brain with each cycle of inoculation (126, 127). Although the molecular mechanism for “brain-training” is unknown, these cells are likely more suited to undergo the same physical and chemical changes to cross through the blood-brain barrier again(128).

The model is termed orthotopic when cells are implanted into the originating tissue environment. In the case of brain metastasis, direct implantation of cells into the brain would more accurately be termed local growth models, whereas models that involve implanting cells into the primary organ of origin (lung, breast, skin, etc) more accurately reflect the metastatic cascade as metastases arise spontaneously (76). Experimental xenograft models have been developed for brain metastases derived from engrafting human lung (129, 130), breast (131, 132), and melanoma cells into mice (133, 134). However, local growth models of direct implantation into the brain are the most common, as generating tumor growth in the primary organ and subsequent brain metastasis is difficult to achieve. Direct implantation is typically accomplished utilizing stereotactic equipment to ensure accuracy of cell delivery into the neural tissue rather than the ventricles (122). Other techniques involve the use of a subarachnoid catheter placed into the cisterna magna along the spinal cord (135). Other techniques utilize a subarachnoid catheter placed into the cisterna magna along the spinal cord (117) or freehand delivery into landmarked areas of the right frontal lobe (136).

Subcutaneous inoculation, where cells are injected into the flank of a mouse, is a traditional route that is highly used in metastatic models because of the ease of the administration procedure. However, while this model is typically used to assess tumor engraftment, size, and therapy screening, it is only ideal in development of metastatic models from primary melanomas which originate in the skin. For any other model system, they do not reflect the complexity of the metastatic cascade or the multistage process of oncogenesis, and very frequently the cells injected do not originate from melanomas (137).

In vivo models for CNS drug discovery

With the currently limited understanding of the mechanisms and molecules involved in brain metastasis development, in vivo models are a vital tool in the discovery of new therapies for treating brain metastaseis. Identification of drug targets relevant to brain metastasis is a first step in developing effective therapies. While in vitro methods are more useful for the identification of targets and large-scale drug screening, in vivo methods are better suited for fine tuning target validation and compound efficacy (138). A

The BBB is a significant hindrance to developing clinically effective therapies for CNS disorders as the role of the BBB is to prevent access of toxins and other molecules, such as chemotherapeutics, into the CNS. This limits the number of therapies that can penetrate the BBB. Moreover, many of those that can have minimal effects on extending survival (139). In vivo models are critical in identifying effective and unique delivery systems to assist passage of chemotherapeutics across the BBB. Coating drugs with various nanoparticles (metallic, organic and biological) improves survival in mice with brain metastases by facilitating transfer into brain (140). Included in this group are nanoparticles with doxorubicin and lapatinib (141) or camptothecin/Herceptin (142), liposomes with simvastatin/gefitinib (143) or irinotecan (144), and dendrosomes with solanine (145) or curcumin (146). Gadolinium nanoparticles and radiotherapy are in Phase I clinical trials for patients with multiple brain metastases (Clinicaltrials.gov;NCT02820454). Another method involves disruption of the BBB itself with MRI-guided ultrasound disruption. This approach has been shown to improve delivery of antibodies such as trastuzumab (147), pertuzumab (148) and natural killer (NK-92) cells (149). Disruption of the BBB through the application of hypotonic solutions such as mannitol (150) are currently in clinical trial as a combined administration with methotrexate and carboplatin with/without trastuzumab (Clinicaltrials.gov; NCT00397501). Other therapies that have reached clinical trials based on preclinical brain metastasis data are the dual EGFR and HER2 targeting inhibitor lapatinib (151), ALTTO trial (152), (Clinicaltrials.gov;NCT00470847), the histone deacytelase HDAC inhibitor Vorinostat (153), (Clinicaltrials.gov; NCT00820222), a VEGFR-2 TKI Cediranib (154), (Clinicaltrials.gov;NCT00937482), a low-molecular-weight epothilone Sagolipone (155), (Clinicaltrials.gov; NCT00496379), and a Paclitaxel derivative ANG1005 (156), (Clinicaltrials.gov; NCT01679743)(157) (138).

Drug screening pitfalls

The success of in vivo models depends on their predictive value for the clinical setting. The most common reason for a compound to fail to demonstrate clinical efficacy is the lack of an association between the modulating activity of the specific target(s) and the disease state. Such a failure is essentially due to a lack of proper or accurate validation in the preclinical models or unclear relationship of the preclinical model to the condition for which the drug is tested (158). Approximately 95% of agents that pass preclinical validation fail in phase 1 of clinical development (159). As such, despite the wealth of knowledge and clinical advances achieved with in vivo modelling, there remain significant limitations to this approach in drug discovery. Most importantly, no preclinical model is known to mimic precisely the complete clinical presentation of brain metastases. In the screening of potential treatments, therapies are often administered very soon after tumor inoculation, which does not replicate when patients are treated with advanced metastatic disease of a size detectable by imaging (6, 160).

In contrast to in vitro assays, in vivo model experiments are much more difficult to perform on a large scale. The initial establishment of preclinical models is can also be time consuming and costly. Because the target transgenes are expressed in all cells of the specific tissue, TSG knockout mice often fail to mimic sporadically established cancers where the accumulation of genetic events in a single cell gives rise to a tumor in an otherwise normal organ. In syngeneic and GEMM models, the genetics of the model (promoter and oncogenes) are not truly representative of human disease and their specific regualtion and many chemically induced cell lines such as JB/MS melanoma do not spread to the brain readily and lose tumorigenicity rapidly. These models also require extensive and timely breeding programs (114).

The PDX models often involve long latency periods from tumor cell inoculation to metastatic initiation, or display inconsistent or no metastatic dissemination that results in a lack of robustness or reproducibility between cell lines. There are also concerns about genetic drift in PDX models generated with late passages of cell lines. However, several studies have shown no major genetic variation up to at least passage 10 (161, 162). Immunocompromised mice are necessary for tumor cell engraftment in PDX models, losing completely any influence the immune system would have on tumor growth, whereas syngeneic models replace the original human stroma and immune cells with mouse, thereby losing key similarities to human disease progression.

The route of inoculation plays a significant role in how accurately the model can represent and recapitulate the clinical condition. In the case of brain metastasis, local growth orthotopic models can replicate the later stages of metastasis of tumor engraftment in the brain and macro-metastatic growth, although the initial stages in the disease process are bypassed. Orthotopic models that involve implantation of the tumor cells into the primary tissue of origin (i.e., other than the brain) can facilitate spontaneous metastases formation. However, metastases are not always to the brain and there may be an extended lag in time from primary inoculation to metastases formation. These challenges are also common with intracardiac injections, where metastases can develop throughout the body and not solely in brain. Heterotopic models only incorporate the mid to end stages of metastasis, skipping the initial stages of escape from the primary tumor and intravasation. It remains to be seen whether models where metastases are generated chemically, genetically, or through UV exposure, accurately reflect clinical metastatic progression (76). The validity and accuracy of results comes into question when inappropriate models are utilized for drug screens. To date, few if any models properly replicate the progression of metastaic disease in humans (76).

Conclusion

In vitro and in vivo assays have led to significant advancements in deciphering the intricacies of metastasis. Cell culture methods are essential for interrogating key morphological and functional aspects of metastatic biology, with preclinical evaluation of novel anti-metastatic strategies in animal models being an essential step towards the development of novel therapeutics. However, the translation of experimental findings to the clinical arena remains a major bottleneck in the development of therapies for brain metastases. Despite these drawbacks, both in vitro and in vivo models provide significant contributions to the understanding of pathology of brain metastasis and remain a vital tool for the screening of novel therapeutics.