Two-Step Carcinogenesis: DMBA/TPA

Perhaps the best known model of chemically-induced metastasis is the two-step DMBA/ TPA strategy that drives squamous cell carcinoma (SCC) and metastasis in mice. Tumor initiation is achieved by topical administration of the polycyclic aromatic hydrocarbon DMBA (7,12-dimethylbenz[α]anthracene) to shaved dorsal epidermis at a sub-carcinogenic dose.³ One week post-DMBA treatment, bi-weekly applications of the phorbol ester TPA (12-Otetradecanoylphorbol-13-acetate) commence, to promote tumor growth (Fig. 1).

Figure 1.

DMBA/TPA two-stage skin carcinogenesis and metastasis in vivo. Topical administration of DMBA to FVB/n mice induces oncogenic mutations within epidermal stem cells (e.g., HRAS). Bi-weekly applications of TPA promote the development of benign epidermal (more...)

Mechanistically, metabolic activation/hydrolysis of DMBA forms DMBA-DNA adducts that damage DNA, increasing the risk of genetic mutations within the proliferating epidermal basal cells and the bulge region of hair follicles (invariably oncogenic mutation of HRas codon 61).^1,4 Multiple TPA applications subsequently promote tumor growth by deregulating several signaling networks, including the PKC-Ras/MAPK cascade that mediates proliferation, differentiation and inflammation.^1,5,6

DMBA/TPA induction typically predisposes to benign squamous cell papillomas (BSCPs) by 6-8 weeks post-DMBA treatment (Fig. 1). During the promotion phase (typically 6-12 mo) BSCPs increase in number/volume and may undergo malignant transformation to invasive squamous cell carcinoma (SCC).^3,7,8 Once the basement membrane has been breached, SCC cells invade into the dermis and subcutaneous tissues and metastasize to regional lymph nodes and distant sites such as the lung.^3,7,8 A single application of DMBA (referred to as the complete model of carcinogenesis) has been shown also to give rise to metastatic lesions. However, SCCs arising using this approach do not consistently develop from BSCPs.⁹ Although the majority of mouse strains develop BSCPs, their susceptibility to developing metastatic SCC varies significantly depending on the genetic background. SENCAR and FVB/n genetic backgrounds are more commonly used as they are most sensitive to DMBA/TPA-induced metastatic SCCs, whereas C57BL/6 mice often require higher DMBA/TPA doses.³

The major advantage of DMBA/TPA-induced skin carcinogenesis is that tumor onset, growth rate and metastasis incidence can be quantified easily by monitoring tumor burden and multiplicity on a weekly basis, as well as by histological assessment at end point.³ Importantly, this model closely mimics human SCC which frequently metastasizes to the lymph nodes and lungs.¹⁰ Furthermore, oncogenic HRAS mutations have been reported in 1246% of human SCCs.¹¹ Nonetheless, key differences include the fact that human SCCs do not commonly develop from papillomas, and the immunogenicity of the chemicals employed may not resemble the human disease.³

While the majority of studies utilizing the DMBA/TPA two-step skin carcinogenesis model have investigated the role of tumor suppressors and oncogenes, several important modulators of metastasis have been identified, including p53¹² and the type-1 membrane glycoprotein CD200 (Cluster of Differentiation 200), which regulates myeloid cell activity and T-cell proliferation.¹³ Stumpfova and colleagues found that expression of the immunosuppressive molecule CD200 in poorly differentiated invasive DMBA/TPA-induced SCCs correlates with metastatic potential.¹³ FACS analysis of DMBA/TPA-induced SCCs identified CD200 receptor 1 (CD200R1) positive immune cells (CD11b⁺Gr-1⁺ myeloid-derived suppressor cells) in the stroma, presenting a potential mechanism whereby metastatic cells may survive at secondary sites via the immunosuppressive and pro-metastatic functions instigated through CD200-CD200R (tumor-stroma) interactions.¹³ Importantly, this study highlights the ability of the DMBA/TPA two-step skin carcinogenesis model to address the role of the tumor microenvironment in metastasis initiation and progression. Interestingly, CD200 expression is also induced in human metastatic SCC,¹³ consistent with the notion that DMBA/TPA-induced metastatic SCCs display features analogous to the human disease.

The major disadvantages of employing the two-step DMBA/TPA skin carcinogenesis model to study metastasis include the long latency and low incidence of metastatic disease.^7,8 In general, only 5-10% of invasive SCCs disseminate, and metastases can take 6-12 mo nts to develop.³ Nevertheless, the two-step DMBA/TPA model, in combination with a conditional transgenic approach, has provided valuable molecular insights into the metastatic cascade. For example, TNF-related apoptosis-inducing ligand receptor (TRAIL-R) deficient mice show a significant 3.5-fold increase in metastatic incidence compared with WT controls without affecting primary tumor growth (1 y post-DMBA induction), indicating that TRAIL-R acts as a metastasis suppressor in this setting.⁷ These findings suggest that agonists of the apoptosis-inducing TRAIL receptors may reduce the incidence of SCC metastasis in the clinic.

In addition, the temporal assessment of gene function at distinct stages of DMBA/TPA-induced skin carcinogenesis can be studied using inducible transgenic mice. For instance, induced expression of TGF-β1 in BSCPs rapidly promotes metastasis whereas triggering TGF-β1 expression at an earlier stage of progression suppresses tumorigenesis and metastasis, demonstrating the dual tumor suppressive and oncogenic roles of TGF-β1.¹⁴ Importantly, TGF-β1-induced metastases displayed elevated matrix metalloproteinases (MMPs), increased angiogenesis and decreased E-cadherin expression, mimicking SCC metastasis in patients.¹⁴

Two-Step Carcinogenesis: DEN/NMOR

Human hepatocellular carcinomas (HCC) commonly metastasize to lung, lymph nodes, adrenal glands and bone, correlating with poor prognosis.¹⁵ To model this disease, a single dose of diethylnitrosamine (DEN) (100 mg/kg i.p. injection) initiates liver carcinogenesis in F344 rats, and upon oral administration of N-nitrosomorpholine (NMOR) (14-24 weeks at 120 ppm in the drinking water), these animals reproducibly display pulmonary lesions within 24 weeks (100% incidence).^16,17 This model closely resembles features of human HCC pulmonary metastasis and provides a useful tool to investigate the efficacy of anti-metastatic therapeutics in vivo. including nuclear factor kappa B (NFκB)¹⁷ and MMP¹⁸ inhibition. In contrast to previous experimental HCC metastasis models where HCC cells are inoculated directly into the vascular system of immunocompromised mice, the DEN/NMOR model has the advantage of an intact immune system (despite chemical immunogenicity), tumor-stroma interactions and all stages of the metastatic cascade from a spontaneous tumor can be assessed.

In summary, chemically-induced in vivo models of metastasis are valuable preclinical tools, owing to their reproducibility, intact tumor-stroma microenvironment and clinicopathological similarity to the human disease modeled. Consequently, these models have greatly contributed to our molecular understanding of metastasis and identified new avenues for therapeutic intervention. However, it is important to consider the long metastatic latency, genetic background and chemical immunogenicity when utilizing carcinogenic models to study the metastatic cascade or to test novel therapeutic compounds.

Syngeneic and Xenogeneic Transplant Models

Syngeneic transplantation involves the inoculation of murine tissue or cells into a recipient of the same species and genetic background. In contrast, xenograft transplantation generally refers to the inoculation of human tissue or cells into a murine host.²⁰ The selection of a syngeneic or xenogeneic model depends primarily upon the question(s) being addressed, the availability of cells/tissue and compromise between the strengths and weaknesses of each model, as syngeneic and xenogeneic grafting have distinct advantages and limitations (Table 1).

Table 1.

Comparison of syngeneic and xenogeneic metastasis models.

To prevent rejection of human cells in xenogeneic models, the recipients innate immune system is compromised. The recruitment and activation of innate immune cells by the humoral immune responses can impinge on several processes required for metastasis, including angiogenesis and tumor cell survival.^20,21 This represents a major limitation for xenograft models, as the effects of an intact immune system on the onset of metastasis and progression are not faithfully replicated. The most common immunocompromised mice include the nude, NOD (non-obese diabetic), SCID (severe combined immunodeficiency) and RAG (recombination activating gene) strains.²² Immunocompromised strains that harbor either single or multiple genetic mutations to impair immune function have varying degrees of innate and adaptive immune deficiencies.²² For example, NOD/SCID mice are diabetes free and deficient in B-cells and T-cells, natural killer (NK) cells, macrophages and complement activity, whereas nude mice are considered to have a limited level of B-cell, dendritic cell and granulocyte function, albeit associated with a compensatory increase in NK cell activity and macrophages.^21,22

Mice with almost no immunity are increasingly exploited to model metastatic disease as they are extremely efficient recipients of heterologous cells, such as NOD/SCID/IL-2Rγ mice that lack B-, T- and NK cells and harbor dysfunctional macrophages and dendritic cells.^22,23 Consequently, NOD/SCID/IL-2Rγ mice are commonly employed to assess the metastatic potential of cells/ tissues that do not engraft or rarely metastasize in less severely immunocompromised mice.²⁴

Furthermore, a large number of cells is commonly inoculated into NOD/SCID mice (> 10⁶ cells) to reproducibly obtain metastases, while only a small number of cells ( < 10²) is necessary in more immunocompromised NOD/SCID/IL-2Rγ^null mice, better reflecting human metastasis.²⁴ NOD/SCID/IL-2Rγ^null mice also have the added benefit of not developing thymic lymphomas that less severely immunocompromised mice are predisposed to, making them more suitable for long-term experiments.²² However, specific markers of metastasis and histological analysis should be performed to ensure spontaneous lymphatic lesions are not mistaken for metastases.²²

The severity of the hosts immunodeficiency can affect tumor growth rate and the frequency of tumor-initiating cells detected, which in turn may impinge on the clinical translation of therapeutic responses observed in xenograft models. For example, human melanoma cells transplanted from NOD/SCID xenografts into a NOD/SCID/IL-2Rγ^null mouse display accelerated growth and increased number of tumor-initiating cells.²⁵ In agreement with a heritable change in the frequency of tumor-initiating cells, melanoma cells recovered from NOD/SCID/IL-2Rγ^null xenografts show reduced frequency of melanoma-initiating cells when injected into NOD/SCID mice.²⁵ However, given that cancer patients generally exhibit intact immunity, it can be argued that the use of a mouse that almost completely lacks an immune system may not be a valid platform to study human metastatic disease. Conversely, the nature of the immune response to human cells in immunocompetent mice may differ from that seen in patients.²⁶ Clearly, these limitations of xenograft models in replicating tumors-host interactions can significantly impact on the phenotypic and clinical relevance of the data generated. Therefore, the immune deficiency of each model must be taken into account when interpreting experimental observations in xenografts. Of note, to better model human metastasis, partial human immunity may be conferred by reconstituting immunocompromised mice with human bone marrow or peripheral blood, thus improving the value of these models when used to predict drug responses in the clinic.²¹ In addition, the limited tumor-stroma compatibility between human tumors and murine hosts can be partially overcome by co-implanting tumor cells with human stromal cells into xenografts. This approach has been shown to provide a more relevant model of human cancer to assess the role of metastasis regulators and anti-metastatic therapeutics.²⁷

Despite their drawbacks, xenotransplantation models have expanded our understanding of the metastatic cascade and helped identify novel prognostic and therapeutic candidates. For example, Chen and colleagues employed a xenograft model of spontaneous metastasis to demonstrate the anti-metastatic potential of CNTO-95 (a human monoclonal antibody that recognizes the αv family of integrins) against breast cancer.^28,34 Specifically, the authors showed that CNTO-95 administration can significantly impede primary tumor growth and subsequent metastasis of human MDA-MB-231 breast carcinoma cells to the lung in SCID mice by blocking αvβ3 and αvβ5 integrin-dependent tumor growth and angiogenesis.²⁸ Indeed, altered expression of specific integrins has been associated with metastatic progression in most cancers, including melanoma. Consistent with this, several integrin inhibitors such as CNTO-95,²⁹ etaracizumab³⁰ and cilengitide³¹ have shown favorable safety profiles and some efficacy against tumor progression and metastatic disease in phase I and II clinical trials and are being further evaluated in additional clinical trials.³²

In contrast to xenotransplantation, syngeneic models involve inoculating mouse cells into recipients that harbor the same genetic background, advantageously permitting the assessment of compatible tumor-stroma microenvironmental interactions, endocrine signaling and immune responses. Consequently, unlike xenografts, syngeneic mice may be utilized to test active immunotherapy and vaccine efficacy.²² For instance, immunization with a human gp100-expressing viral vector, a melanoma-associated antigen, significantly reduced the number of spontaneous B16 melanoma lung metastases in syngeneic C57BL/6 recipients.³³ The clinical relevance of this experimental approach was demonstrated in patients recently in a randomized phase III clinical trial.³⁴

Syngeneic models are derived from homozygous inbred mouse strains and consequently, genetic and epigenetic tumor heterogeneity common to human cancers may be low. However, clear functional and phenotypic heterogeneity, including metastatic potential, has been observed within a single syngeneic tumor.^35-37 To better recapitulate metastatic distribution and genetic heterogeneity, sub-populations of murine tumor cells with distinct metastatic potentials can be isolated by serial selection, cloning or genetic manipulations before seeding into syngeneic hosts.³⁸ For example, selection of variants from 4T1 mouse mammary carcinoma cells using a single cell cloning approach resulted in the isolation of several metastatic cell lines of distinct metastatic abilities, some of which that closely mimic the complete metastatic distribution of human breast cancer, including to bone.^35,36 Syngeneic transplantation of a 4T1.2 clonal variant into the mammary fat pad leads to a significantly higher incidence of bone, lung, and lymph node metastases compared with parental 4T1 cells.^35,36 This model is particularly useful as primary tumors are established within 710 d post-implantation and metastasis is evident within 2-3 weeks. The 4T1.2 model has been extensively utilized to test novel therapeutics, including laminin-a5-derived peptides to interfere with laminin-511 activity,³⁹ cathepsin B inhibitors⁴⁰ and monoclonal antibody based therapy against CD73.⁴¹

Spontaneous vs. Experimental Models of Metastasis: Which One Do I Use?

The appropriate selection of a spontaneous or experimental model of metastasis depends primarily upon the specific steps of metastasis being studied, the metastatic potential of the cancer cells and the pro- or anti-metastatic gene being investigated. The advantages and disadvantages of these models are summarized in Table 2. Generally, studies addressing metastasis inducers/ repressors or therapies during the early phases of metastasis (e.g., dissemination and intravasation) rely on the spontaneous model, whereas extravasation and metastatic colonization steps can be addressed by either an experimental or a spontaneous approach.

Table 2.

Comparison of experimental and spontaneous transplant models of metastasis.

Spontaneous models of metastasis involve the inoculation of tumor cells into an orthotopic site (tissue of origin) or ectopically (foreign tissue, e.g., breast cancer cells injected subcutaneously). These models permit the analysis of both early and late stages of metastasis and importantly, make it possible to investigate the function of any given gene (by genetic manipulation or use of inhibitors) on tumor growth and metastasis simultaneously. Drawbacks include a longer latency for metastatic disease compared with experimental models, lower predictability of the pattern of dissemination and lower overall incidence of metastasis. Spontaneous models of metastasis are often coupled with a tumor resection approach to better replicate therapeutic interventions in the clinic and/ or to determine if the impact of a specific genetic alteration on metastasis is due to a direct effect on metastasis or a result of changes in primary tumor growth, or both. Removal of the primary tumor at a defined size also allows time for metastases to establish without the primary tumor growing beyond ethical size limits. While the impact of this procedure on metastatic progression has been questioned, a recent study demonstrated that the effect of tumor removal on the number and phenotype of circulating tumor cells (CTCs) is negligible.⁴²

Generally, mice with invasive lesions are more likely to develop metastatic disease earlier than those with benign lesions following primary tumor resection.³⁸ However, resection of tumors may be difficult in some spontaneous models that display variable degrees of local invasion or models that simultaneously develop multiple primary lesions. In those cases, poor reproducibility may be partly overcome by increasing the cohort size to enhance statistical power.

Compared with spontaneous metastasis models, experimental metastasis models are rapid and more reproducible, as cells are introduced directly into the blood stream of the host. They are therefore ideally suited for testing therapies specifically targeting late stage metastasis. They are also particularly attractive models to investigate human tumor lines that have limited metastatic potential in mice from orthotopic sites. For instance, human MDA-MB-435 breast tumor cells do not normally metastasize spontaneously to bone from the mammary gland but readily do so via the intracardiac route.⁴³ However, this approach presents two potential caveats; (i) experimental models dismiss the early phases of the metastatic cascade and (ii) injecting a large bolus of tumor cells into the circulatory system does not accurately mimic disease progression in the clinic.⁴⁴ In patients, the precise number of CTCs has not been well characterized (estimated 1:1 billion normal blood cells) and only a handful of these cells are considered viable metastatic precursors.⁴⁵ Although the number of CTCs introduced into the vasculature in experimental metastasis models may not mimic the human disease, these platforms have nevertheless generated many interesting insights into the molecular mechanisms underlying metastasis. In a pioneering study analyzing pulmonary metastases that develop within 12-14 d post intravenous injection of murine B16 metastatic melanoma cells into syngeneic C57BL/6 hosts, Fidler and colleagues provided the first evidence for metastatic heterogeneity in neoplasms.⁴⁶ Different cell clones, each derived from the same primary B16 melanoma tumor, displayed differential pulmonary metastatic potentials.⁴⁶ This study fuelled the notion that only a small subpopulation of CTCs survive to establish heterogeneous metastases.

Experimental models are also frequently employed to model metastatic dissemination to sites that take longer to develop, such as central nervous system (CNS) metastases. Most tumor lines that can metastasize to brain do so with low frequency, partly due to the development of visceral disease that diminishes survival before CNS metastases had time to establish.^47,48 To circumvent this issue, Cruz-Munoz and colleagues employed an innovative approach to increase the incidence of brain metastasis in a melanoma xenograft model. Through prolonged control of visceral melanoma metastasis using the chemotherapeutic vinblastine, a metastatic human melanoma cell line was found to metastasize to the brain with ~20% incidence. Clonal lines derived from these brain lesions gave rise to spontaneous CNS metastasis in over 60% of animals,⁴⁸ thus providing a rare preclinical platform to test anti-metastatic compounds and investigate and molecular events underpinning melanoma CNS metastases.

Combining findings from spontaneous and experimental models is particularly useful for improving our interpretation of molecular mechanisms of metastasis and therapeutic drug responses. For example, the frequency of pulmonary metastases that developed from either orthotopic or intravenous inoculation of MDA-MB-231-derived metastatic 4173 cells was significantly reduced in cohorts treated with antibodies against mouse or human CCL2 (monocyte chemoattractant protein-1, MCP-1) compared with control antibodies.⁴⁹ Mechanistically, inhibition of CCL2 secretion from both breast cancer cells and stromal cells impaired the recruitment of CCR2 expressing monocytes and their differentiation into macrophages that facilitate extravasation and colonization.⁴⁹ The combination of both experimental and spontaneous models in this study clearly defined the pro-metastatic function of CCL2 and demonstrated conclusively the contribution of CCL2 secreted by disseminated tumor cells as well as stromal cells to the process.

The Site of Tumor Implantation Influences Spontaneous and Experimental Metastatic Dissemination

The site of implantation in experimental models can significantly influence metastatic potential/distribution and consequently clinical relevance, depending on the initial capillary bed encountered.⁴⁴ While tumor cell injection into the lateral tail vein predominantly results in pulmonary metastasis,⁵⁰ intracardiac injection increases the frequency of hepatic, ovary, adrenal gland, bone and brain metastases^43,44,50 and intracarotid injection gives rise to brain metastases.⁵¹ Accordingly, the selection of an experimental site of transplantation to introduce tumor cells directly into the bloodstream should take into consideration the common metastatic distribution of the human cancer being investigated. For example, human prostate and breast cancers commonly metastasize to bone, and therefore, an intracardiac injection would be appropriate to experimentally model these cancers.^43,52

Subcutaneous injections are often utilized to model spontaneous metastasis (including non-epidermal cancers) as cell inoculation and monitoring of primary tumor growth using calipers are convenient and not technically challenging.⁵³ Potential drawbacks however include alterations in tumor growth rate, metastatic ability/distribution and poor therapeutic responses owing to microenvironment differences (e.g., vascularization and hormone dependency) compared with the organ of origin.^53,54 In general, orthotopic transplantation models display better tumor take rates and clinically relevant therapies targeting processes involved in local invasion (e.g., angiogenesis) can be undertaken.⁵³ However, orthotopic transplantation may be technically difficult for some tumor types (e.g., colon carcinomas) and monitoring primary tumor growth often requires more complex in vivo imaging techniques.

Monitoring and Quantitation of Metastatic Burden

A major advantage of using cell lines in transplantation models of metastasis is that they are amenable to genetic manipulations. This has enabled the use of various molecular markers to facilitate the visualization tumor cells in vivo without sacrificing animals and/or quantitation of metastatic burden at precise time points. In particular, the use of fluorescent biomarkers, led by the green fluorescent protein (GFP), has revolutionized the field and is now common practice in metastasis research.^55-57 The various applications for fluorescent biomarkers in cancer research have been extensively reviewed in the literature.^58,59

A potential pitfall associated with the introduction of exogenously expressed proteins into mice is their immunogenicity, which must be taken into consideration when utilizing syngeneic animal models of metastasis. For instance, high expression of Enhanced GFP (EGFP) in the Balb/c-BM185 pre-B leukemia and C57BL/6-EL-4 T cell lymphoma models trigger a T-cell mediated response that significantly reduces disease progression.⁶⁰ Lymphocyte cytotoxicity directed against EGFP was also reported in the CMS4 sarcoma transplant model, owing to a naturally occurring immunogenic epitope of EGFP (HYLSTQSAL, corresponding to EGFP200-208).⁶¹ Similar findings have been reported in a neuroblastoma model expressing GFP in immunocompetent mice⁶² and intracellular expression of GFP in lung carcinoma and mammary tumor cells was shown to generate a high-titer antigen-specific IgG antibody response in syngeneic Balb/c mice.⁶³

Several studies have employed TaqMan technology to quantitate tumor burden in tissues by real time quantitative polymerase chain reaction (RT-QPCR) detection of ALU sequences⁶⁴ or exogenously expressed reporter genes.³⁶ While this method provides a highly sensitive and accurate means to quantitate metastatic burden in any tissue, our experience in syngeneic models suggests that the specific marker genes must be carefully selected. For instance, while introduction of neomycin or puromycin-resistance genes in 4T1 mammary carcinoma cell lines does not impact tumor growth or metastatic ability, insertion of a hygromycin-resistance marker leads to frequent regression of primary mammary tumors (12-15 d post-implantation) and dramatically reduces spontaneous metastasis (NP unpublished observations). Similarly, we have observed that GFP and tandem dimer Tomato (tdTomato) fluorescent markers trigger an immune response and reduce overall metastasis in 4T1-derived spontaneous syngeneic Balb/c models supporting previous work, whereas the monomeric red fluorescent marker mCherry has no effect (NP unpublished observations).

Bioluminescence is rapidly becoming the method of choice for live imaging of metastatic progression in animal models.⁶⁵ However, evidence suggests that this approach is not without drawbacks. For example, Brutkiewicz and colleagues showed that high but not low expression of luciferase in the HEYC2 ovarian cancer cell line severely impaired tumor growth in vivo, even in immunocompromised nude mice.⁶⁶ Growth inhibition was associated with repeated luciferin injections compared with PBS controls, suggesting that the growth rate may be impaired by the luciferase-luciferin reaction (possibly owing to oxidative stress or hypoxia) rather than high expression of luciferase per se. Consequently, selection of low luciferase expressing cells may avoid this caveat.

Interestingly, another study reported that intracardiac injection of NT2.5 mammary carcinoma cells expressing luciferase into nude mice led to extensive bone metastases whereas the same cells metastasized poorly in immunocompetent neu-N mice.⁶⁷ Failure to metastasize in neu-N mice was attributed to immune rejection of luciferase expressing cells. The bi-phasic growth of 4T1 tumors expressing luciferase and regression occurring around week 2 post-implantation is consistent with an immune reaction against luciferase⁶⁸ and supports our unpublished observations in this model. Together these studies highlight the importance of careful selection of biomarkers for visualization and quantitation of syngeneic models of metastasis.

Primary Tumor Transplant Models of Metastasis

Implanting fresh tumor cells/tissue into immunocompromised recipients directly from a patient has been shown to better retain the clinicopathological features of the human disease beyond established cancer cell lines, dramatically improving clinical relevance and therapeutic predictions.⁶⁹ However, transduction of primary cells/tissue to manipulate gene expression is technically challenging, limiting functional studies and high-throughput screening approaches. The advantages and disadvantages of fresh cell/tissue transplants compared with established cancer cell lines are summarized in Table 3.

Table 3.

The advantages and disadvantages of utilizing established cell lines compared with primary cells/tissue to model metastasis.

While fresh human cells/tissue xenografts are labor intensive, costly and not well suited for high-throughput drug screens,⁶⁹ the high predictive value of these models justifies their use (particularly in orthotopic transplants).⁷⁰ They provide a heterogeneous platform to; (i) optimize therapies in a clinically relevant setting with stromal involvement, (ii) identify biomarkers and (iii) develop personalized routes of therapeutic intervention as multiple assays can be performed on the same tumor.^69,71 Since established human cancer cell lines are typically homogeneous owing to the selection pressure associated with prolonged in vitro culture,⁶⁹ their molecular signature is often distinct from that of the original patients cancer, as a study comparing fresh vs. cultured small cell lung cancer (SCLC) specimens recently demonstrated.⁷² Interestingly, serially propagating fresh specimens in immunocompromised mice did not significantly alter the molecular profile,⁷² suggesting that limiting in vitro culturing of fresh tissue is preferable for developing a clinically relevant transplant model.

One major drawback of transplanting primary specimens is the low rate of engraftment and tumorigenicity. However, improvements have been reported when fresh tissues and cell lines are co-injected with Matrigel²⁵ or when using chimeric grafts combining fresh tissue and stromal cells.^71,73 Moreover, Patel and colleagues recently reported that a humanized stromal bed improves the success of xenografting human primary squamous cell carcinoma cells into the dorsal epidermis of nude mice.⁷⁴

In summary, it is paramount that the advantages and limitations of transplant models are carefully considered when selecting this platform to address a specific experimental question. To date, the most clinically relevant transplant models are arguably those involving orthotopic implantation of patient-derived primary tumor tissue, which also provide a worthy platform to optimize personalized therapy. Nonetheless, the distinct advantages and reliability of well-characterized syngeneic and xenograft models promotes their continued use, often in combination, to attain a better molecular understanding of the factors regulating the metastatic cascade and/or for high-throughput screening of novel therapeutics.

Transgenic Mouse Models of Metastasis

A traditional transgenic model refers to an animal whose genetic material has been modified using genetic engineering to systemically insert (knock-in) or delete (knockout) genes. The major advantages of transgenic models include an intact immune system, species-specific tumor-stroma microenvironment interactions and all stages of metastatic progression can be assessed.²¹ Although knockout/in models have been proven to recapitulate disease progression and to validate chemotherapeutic strategies, these models have several limitations; (a) biallelic mutation of essential genes frequently causes embryonic/neonatal lethality, (b) the occurrence of concomitant tumors can cause lethality before metastatic disease prevails and (c) additional genetic/epigenetic oncogenic events may be necessary to initiate tumor growth and/or progression.²¹ To exemplify these issues, homozygous depletion of Pten (phosphatase and tensin homolog) that negatively regulates the PI3K/AKT pathway causes embryonic lethality. However, viable Pten heterozygous mice develop multiple spontaneous tumors (> 14 mo) in the lymph, intestine, endometrium, mammary gland, liver and prostate, reducing survival before metastasis can occur.^76,77

The acquisition of many genetic/epigenetic mutations over time can promote disease progression. Combining clinically relevant genetic mutations has led to the development of numerous compound models of metastatic disease and has functionally demonstrated the synergistic effect of multiple genetic mutations. For example, compound transgenic mice heterozygous for PTEN and the homeobox protein NKX3.1 that are commonly depleted in human prostate cancer develop high-grade prostate intra-epithelial neoplasia (PIN) by 6 mo of age, which significantly progresses to invasive carcinoma (84%) by 12-15 mo of age compared with Pten^+/- or Nkx3.1^+/- mice (54% and 0% incidence respectively).⁷⁸ Metastasis to lymph nodes (25% incidence) was also observed in compound mutants, which was not evident in single mutants.⁷⁸ Importantly, this model displays clinicopathological features of human prostate cancer and is androgen-independent. However, the low incidence of metastasis limits the use of this model for trialing anti-metastatic therapeutics.

In general, transgenic knockout/in models have a relatively low metastatic burden, predominantly due to the rapid growth of primary tumors. Consequently, these models are primarily employed to study the functional role of oncogenes and tumor suppressors during tumor formation and progression rather than mediators of metastasis.

Conditional Transgenic Mouse Models of Metastasis

Conditional transgenic mouse models are powerful tools for assessing clinically relevant genetic abnormalities at controlled stages of tumor formation and progression within a targeted tissue.⁷⁹ While non-conditional models harbor systemic engineered genetic alterations, conditional models target a specific tissue/cell compartment. This methodology has the benefit of extending the longevity of mice that are predisposed to multiple tumors and more accurately mimicking human cancer, where distinct subpopulations of cells are mutated.²¹

Several bacterial (Cre-LoxP, Flp/frt and Dre/rox), and viral vectors or protein transduction recombinase approaches have been employed to establish conditional transgenic models.^80,81 Recombination is driven by tissue/cell specific promoters to achieve spatial regulation of gene expression.^80,81 The Cre-LoxP strategy has been most widely exploited to generate models of metastatic disease. Upon Cre-mediated recombination, floxed target sequences flanked by two 34 bp unidirectional bacterial LoxP sites are excised to mediate gene ablation or expression.^81,82 Gene ablation is typically accomplished by flanking the activating domain of the gene of interest with LoxP sites, whereas gene expression involves excision of the deactivating domain of the target gene or insertion of a LoxP-STOP-LoxP (LSL) cassette upstream of the target gene. Cre-mediated excision of the STOP cassette permits transcription of the target gene, such as oncogenic KRas upon recombination of the LSL-KRAS-G12D construct.⁸³

As a classical example, the LSL-KRAS-G12D transgenic line has been widely used to assess the role of KRas activation in several tissues, including the pancreas,⁸⁴ lung⁸³ and prostate⁸⁵ in mice. Oncogenic KRAS and p53 mutations are frequently associated with pancreatic ductal adenocarcinoma (PDAC). To model this disease the Pdx promoter was employed to drive Cre-mediated expression of oncogenic KRAS in combination with mutant p53 (termed KPC mice). The KPC model is predisposed to pancreatic intraepithelial neoplasms (96% incidence) and invasive lesions that metastasize to the liver (63% incidence), lung (44% incidence) and diaphragm (37% incidence), representing a valuable tool to assess novel therapeutic strategies to target metastatic pancreatic cancer in a clinically relevant setting.⁸⁴

Ultimately, the strength of the conditional transgenic approach is limited by the availability of a strong transcriptional promoter and its specificity.^80,81 Although conditional transgenic technologies provide a means to circumvent some of the limitations of conventional gene knockouts, such as embryonic lethality, the "leakiness" of some promoters during early stages of gestation may still result in this complication.⁸¹ For example, BRAF^V600E mutations are common in human melanoma patients, yet expression of oncogenic BRaf in the murine melanocytes using the Tyrosinase promoter to drive Cre mediated recombination (TyrCre) causes multiple developmental defects leading to embryonic lethality.⁸⁶ To overcome this, inducible promoters that can be chemically stimulated have been generated to achieve a level of temporal control.⁸¹ Moreover, 4-hydroxytamoxifen induction of the TyrCre estrogen receptor fusion protein (TyrCre-ERT2) in adult mice to mediate BRaf^V600E expression avoids embryonic lethality.⁸⁷ Using this model, Dankort and colleagues showed that oncogenic BRaf^V600E expression in mature melanocytes predisposes to benign melanocytic hyperplasia and that Pten deficiency in this setting rapidly promotes robust metastasis to the lymph nodes and lung.⁸⁸ This compound model exemplifies the synergistic relationship between oncogenic BRaf and Pten signaling in facilitating malignant melanoma progression and provides a translational platform to test combinatorial therapeutic strategies and further investigate the molecular basis of metastatic melanoma onset and progression.⁸⁹ To functionally assess the role of a gene at precise stages of metastatic progression, reversible transgenic mouse models have also been generated using the bacterial-derived Tetracycline-inducible system.⁹⁰ This elegant approach permits the spatiotemporal control of gene expression in the presence of the antibiotic tetracycline (or its derivative doxycycline, dox).⁹⁰

Although huge advances in the generation of mouse models of metastasis have been made possible through the development of inducible promoters, they too have been found to leak. For example, Pearson and colleagues have shown that un-induced AhCre mice harboring floxed Lkb1(STK11) alleles are prone to prostate neoplasia owing to endogenous activation of the AhCre-driven depletion of Lkb1 in prostate epithelium.⁹¹ To this end, histopathological assessment and validation using tissue-specific markers (e.g., by immunohistochemistry) are essential for determining the origin of metastases observed. Furthermore, reporter transgenic lines are often crossed with conditional transgenic models to profile endogenous/induced recombination events, such as Rosa26-LacZ reporter system.⁹²

Transgenic T-Antigen (Tag) Conditional Models of Metastasis

Transgenic t-antigen models of metastasis typically involve the expression of the simian virus 40 (SV40) viral large T-antigen that acts as an oncogene through interactions with Rb and p53 and/or the small t-antigen, which interacts with protein phosphatase 2A (PP2A) to constitutively activate the MAPK cascade.⁹³ For example, the Transgenic Adenocarcinoma of the Mouse Prostate (TRAMP) autochthonous mouse is a widely exploited t-antigen driven prostate metastasis model.^93,94 The minimal PB promoter is used to drive the expression of the SV40 viral large T-antigen within prostate epithelium, causing androgen-dependent neoplasia by 12 weeks and metastasis to the lung, lymph node, kidney and adrenal gland by 12-30 weeks.^79,80 Although on certain genetic backgrounds metastatic burden is not highly reproducible, crossing TRAMP mice with other transgenic lines harboring clinically relevant genetic mutations has provided mechanistic insights into prostate cancer progression. For instance, TRAMP:Fg f2^-/- mutant mice show decreased tumor progression/metastasis and improved survival, implicating Fg f2-mediated angiogenesis and intranuclear activities in prostate cancer progression.⁹⁵

To model breast cancer progression, the mouse mammary tumor virus (MMTV) promoter is commonly used to drive the expression of polyoma middle T-antigen (PyMT), which modulates several signaling networks to promote growth and proliferation (e.g., PI3K/AKT signaling).⁹⁶ PyMT expression in the mammary gland results in multifocal mammary adenocarcinomas and reproducible pulmonary and lymph node metastasis.⁹⁶ The MMTV-PyMT model shares many aspects of human breast cancer progression and has been widely employed to identify metastatic regulators.⁹⁷ Indeed, MMTV-PyMT metastases have been shown to express a clinically relevant metastasis-specific gene expression signature⁹⁷ and this model has been used extensively to address the function of candidate genes thought to play a roles in metastasis⁹⁸ and to trial novel therapeutics.⁹⁹ However, the extent of metastasis in the MMTV-PyMT model is highly dependent on host genetics¹⁰⁰ and although t-antigen models can recapitulate human cancer metastatic progression, it is important to note that the viral antigens and the phenotype observed are not always associated with human cancer.

One major disadvantage of this model is that the MMTV promoter is endocrine sensitive (similar to the PB promoter in the TRAMP model), preventing analysis of hormone effects on tumor growth and progression. Several additional conditional transgenic approaches have been employed to overcome this issue. For example, expression of Cre-recombinase specifically within the mammary gland by the endogenous ErbB2 promoter has been employed to manipulate target genes, such as the clinically relevant oncogenic activation of Neu/ErbB2.¹⁰¹

In summary, while transplant models are currently the favored in vivo models for high-throughput drug screening, GEMMs are continuing to enhance our understanding of the molecular basis of the metastatic cascade and contribute to the identification of novel strategies for therapeutic intervention and prognosis.