**2. Current knowledge**

#### **2.1 Metastasis**

As early as 1889, Stephen Paget observed that some cancers metastasized preferentially to specific organs, and developed his theory of "seed and soil"(Paget 1889). The essential tenet of this theory was that cancer cells (seeds) disseminate throughout the body from their point of origin but can only develop metastatic satellites in appropriate stromal environments (soils). The many advances in our understanding of the molecular and cellular bases of breast cancer metastasis has led to a somewhat more complex picture, and the processes involved are still not completely understood. Breast cancer can spread to any secondary site in the body but metastases appear preferentially in bone, lung and liver (Rabbani and Mazar 2007). Presumably these sites provide a microenvironment favourable for the growth and development of breast cancer cells (Nguyen, Bos et al. 2009).

There are two prevailing models of breast cancer metastasis; one suggesting a linear progression and the other a parallel progression. The linear progression model advances the idea that cells in the primary tumour accumulate progressive mutations in a stepwise manner in genes regulating some aspect of cell growth and division such as oncogenes and tumour suppressor genes. Some cells eventually become able to proliferate autonomously; they expand clonally and leave the primary site to travel through lymphatic or vascular systems to a distant organ where they develop into a secondary metastatic growth. This

Breast Cancer Metastasis: Advances Through the Use of In Vitro Co-Culture Model Systems 513

discordant with respect to EGFR mutation status (Gow, Chang et al. 2009). This is of concern in a time of more personalized treatment, where often it is the genetic signature of the

Evidence suggests that less than 1% of breast cancer cells that enter the circulatory system are capable of generating metastatic foci (Fidler 1970; Fidler and Nicolson 1977). Often disseminated breast tumour cells that have settled in the microenvironment at the site of metastasis will lie dormant for years in patients with no evidence of disease before developing into a clinically significant metastatic focus, indicating they are capable of escaping early systemic therapies that target rapidly proliferating cells at the site of the primary tumour (Pantel, Schlimok et al. 1993; Klein, Seidl et al. 2002; Riethdorf, Wikman et al. 2008; Morgan, Lange et al. 2009). As they remain quiescent for some period of time this also suggests that they, or the cells in their microenvironment, or both undergo genetic changes which allow them to progress to a metastatic phenotype (Riethdorf, Wikman et al. 2008) (Riethdorf, Wikman et al. 2008; Klein 2009; Nguyen, Bos et al. 2009; Smith and Theodorescu 2009; Rose and Siegel 2010; Stoecklein and Klein 2010). A class of genes that has been implicated in the regulation of this process is metastasis suppressor genes (Smith and Theodorescu 2009). These are genes that inhibit metastasis but do not affect the ability of cells to produce a primary tumour, and they play key roles in invasion, dissemination, arrest, survival and colony formation. Their function must be lost or inhibited for a metastasis to develop and they represent fertile new ground for the development of anti-

A number of metastasis suppressor proteins have been reported to inhibit breast cancer metastasis. Reduced levels of nm23 family proteins in the primary tumour have been reported to correlate with more aggressive phenotype in breast cancer patients (Galani, Sgouros et al. 2002; Terasaki-Fukuzawa, Kijima et al. 2002; Steeg, Ouatas et al. 2003; Peihong and Perry 2007), although conflicting results have also been presented (Charpin, Garcia et al. 1998; Belev, Aleric et al. 2002; Sgouros, Galani et al. 2007). The results seen in mouse models are more straightforward, where breast cancer cells with low expression of nm23 are more metastatic than those with high levels (Leone, Flatow et al. 1993; Bhujwalla, Aboagye et al. 1999; Tseng, Vicent et al. 2001). *In vitro* models have revealed that nm23 acts by reducing breast cancer cell motility and invasiveness (MacDonald, Freije et al. 1996; Russell, Pedersen

For Breast Cancer Metastasis Suppressor-1 (BRMS1), the clinical data reporting it to be a metastasis suppressor protein in breast cancer tumour samples is also conflicting (Kelly, Buggy et al. 2005; Hicks, Yoder et al. 2006; Lombardi, Di Cristofano et al. 2007). Again, its role in mouse models is clearer, where higher expression in breast cancer xenografts clearly resulted in reduced metastasis (Hedley, Vaidya et al. 2008; Hurst, Xie et al. 2008; Phadke, Vaidya et al. 2008). The stage at which BRMS1 suppresses metastasis is less clear, as it appears to affect a number of steps in the process of metastasis (Stafford, Vaidya et al. 2008). At least two of its functions appear to be increasing anoikis of cells free in the vascular system and inhibition of colonization of disseminated cells (Phadke, Vaidya et al. 2008). KAI1 (CD82, Tetraspannin), has also been clearly verified as a breast cancer metastasis suppressor in clinical samples, where decreased expression correlates with poor outcome (Yang, Welch et al. 1997; Christgen, Christgen et al. 2009; Malik, Sanders et al. 2009). Similar to BRMS1, KAI1 appears to act in multiple ways to inhibit metastasis and reduce breast

et al. 1998; Steeg, Ouatas et al. 2003; Horak, Lee et al. 2007).

primary tumour alone on which outcome predictions or treatment options are based.

**2.2 Metastasis suppressor genes** 

metastatic therapeutics.

model implies that cells at the primary site must undergo a number of rounds of division before they become autonomous and so development of metastasis is linked to primary tumour size with metastases more likely to develop from larger primary tumours. In support of this model it has long been known that there is a close association between tumour size and the possibility of development of metastasis, and tumour size is used as part of histological classification (1983; Rakha, Reis-Filho et al. 2010). The model also suggests that cells being shed by the primary tumour are fully metastatic and that cells that have metastasized to a secondary site should also be able to leave that site to set up at a tertiary site (Klein 1998; Klein 2009). Mutations in genes such as BRCA1, BRCA2, p53 and RB and amplification of the HER-2 receptor at the site of the primary tumour have been identified as being predictive of poorer outcome for breast cancer patients, consistent with this model (Slamon, Clark et al. 1987; Ross and Fletcher 1999; Bordeleau, Lipa et al. 2007; Bosco and Knudsen 2007; Kumar, Walia et al. 2007; Baker, Quinlan et al. 2010).

The parallel progression model suggests that tumour cells may disseminate from the site of the primary tumour very early in its development and may be subsequently genetically modified in the metastatic niche where they later settle (Klein 2009). This model predicts that disseminated tumour cells in the blood or lymph should be detectable very early in development of the primary tumour and that cells at the site of metastasis could be genetically divergent from those at the site of the primary tumour. In support of this model it has been shown in a HER-2 mouse model and in women with ductal carcinoma *in situ*, that disseminated tumour cells in bone and micro metastases could be detected from the time of earliest epithelial alterations at the site of the primary tumour. The numbers of disseminated tumour cells in this study were found to be the same for small and large tumours (Husemann, Geigl et al. 2008), suggesting that shedding of cells from the tumour mass was independent of primary tumour size. In a qualitative and quantitative study of 12,423 women with breast cancer, J. Engel *et. al.* (Engel, Eckel et al. 2003) determined that systemic disease was already present at the time of diagnosis in women who went on to develop metastases, again suggesting cells left the primary tumour early during its development.

The advent of single-cell genomics has allowed comparison of the characteristics of disseminated tumour cells in the blood and lymph and cells at the site of the primary tumour and these have been found to be genetically divergent in some cases (Klein, Seidl et al. 2002; Klein 2003; Fuhrmann, Schmidt-Kittler et al. 2008; Klein 2009; Klein and Stoecklein 2009), indicating that early clonal divergence and parallel progression may occur in some breast cancers. Disseminated tumour cells may also differ genetically from cells that eventually develop into a metastasis in the same patient (Stoecklein and Klein 2010). This could reflect the requirement for the disseminated tumour cells to undergo whatever genetic changes are necessary for them to adapt and be able to successfully grow in the new microenvironment. If that is the case it follows that the genetic aberrations found in the primary tumour may not reflect those seen in the metastasis and this has been found to occur (Tortola, Steinert et al. 2001; Albanese, Scibetta et al. 2004; Gow, Chang et al. 2009; Stoecklein and Klein 2010). In colorectal cancer, mutations in B-raf, K-ras and p53 seen in the primary tumour may be absent or altered in the metastasis. In some cases mutations in the metastasis may be absent in the primary tumour (Tortola, Steinert et al. 2001; Albanese, Scibetta et al. 2004; Stoecklein and Klein 2010). In a study of non-small-cell lung cancer where EGFR mutation status is used as a determinant for treatment with tyrosine kinase inhibitors, 27% of paired primary/metastasis samples (n=67 patients) were found to be

discordant with respect to EGFR mutation status (Gow, Chang et al. 2009). This is of concern in a time of more personalized treatment, where often it is the genetic signature of the primary tumour alone on which outcome predictions or treatment options are based.

### **2.2 Metastasis suppressor genes**

512 Breast Cancer – Focusing Tumor Microenvironment, Stem Cells and Metastasis

model implies that cells at the primary site must undergo a number of rounds of division before they become autonomous and so development of metastasis is linked to primary tumour size with metastases more likely to develop from larger primary tumours. In support of this model it has long been known that there is a close association between tumour size and the possibility of development of metastasis, and tumour size is used as part of histological classification (1983; Rakha, Reis-Filho et al. 2010). The model also suggests that cells being shed by the primary tumour are fully metastatic and that cells that have metastasized to a secondary site should also be able to leave that site to set up at a tertiary site (Klein 1998; Klein 2009). Mutations in genes such as BRCA1, BRCA2, p53 and RB and amplification of the HER-2 receptor at the site of the primary tumour have been identified as being predictive of poorer outcome for breast cancer patients, consistent with this model (Slamon, Clark et al. 1987; Ross and Fletcher 1999; Bordeleau, Lipa et al. 2007;

The parallel progression model suggests that tumour cells may disseminate from the site of the primary tumour very early in its development and may be subsequently genetically modified in the metastatic niche where they later settle (Klein 2009). This model predicts that disseminated tumour cells in the blood or lymph should be detectable very early in development of the primary tumour and that cells at the site of metastasis could be genetically divergent from those at the site of the primary tumour. In support of this model it has been shown in a HER-2 mouse model and in women with ductal carcinoma *in situ*, that disseminated tumour cells in bone and micro metastases could be detected from the time of earliest epithelial alterations at the site of the primary tumour. The numbers of disseminated tumour cells in this study were found to be the same for small and large tumours (Husemann, Geigl et al. 2008), suggesting that shedding of cells from the tumour mass was independent of primary tumour size. In a qualitative and quantitative study of 12,423 women with breast cancer, J. Engel *et. al.* (Engel, Eckel et al. 2003) determined that systemic disease was already present at the time of diagnosis in women who went on to develop metastases, again suggesting cells left the primary tumour early during its

The advent of single-cell genomics has allowed comparison of the characteristics of disseminated tumour cells in the blood and lymph and cells at the site of the primary tumour and these have been found to be genetically divergent in some cases (Klein, Seidl et al. 2002; Klein 2003; Fuhrmann, Schmidt-Kittler et al. 2008; Klein 2009; Klein and Stoecklein 2009), indicating that early clonal divergence and parallel progression may occur in some breast cancers. Disseminated tumour cells may also differ genetically from cells that eventually develop into a metastasis in the same patient (Stoecklein and Klein 2010). This could reflect the requirement for the disseminated tumour cells to undergo whatever genetic changes are necessary for them to adapt and be able to successfully grow in the new microenvironment. If that is the case it follows that the genetic aberrations found in the primary tumour may not reflect those seen in the metastasis and this has been found to occur (Tortola, Steinert et al. 2001; Albanese, Scibetta et al. 2004; Gow, Chang et al. 2009; Stoecklein and Klein 2010). In colorectal cancer, mutations in B-raf, K-ras and p53 seen in the primary tumour may be absent or altered in the metastasis. In some cases mutations in the metastasis may be absent in the primary tumour (Tortola, Steinert et al. 2001; Albanese, Scibetta et al. 2004; Stoecklein and Klein 2010). In a study of non-small-cell lung cancer where EGFR mutation status is used as a determinant for treatment with tyrosine kinase inhibitors, 27% of paired primary/metastasis samples (n=67 patients) were found to be

Bosco and Knudsen 2007; Kumar, Walia et al. 2007; Baker, Quinlan et al. 2010).

development.

Evidence suggests that less than 1% of breast cancer cells that enter the circulatory system are capable of generating metastatic foci (Fidler 1970; Fidler and Nicolson 1977). Often disseminated breast tumour cells that have settled in the microenvironment at the site of metastasis will lie dormant for years in patients with no evidence of disease before developing into a clinically significant metastatic focus, indicating they are capable of escaping early systemic therapies that target rapidly proliferating cells at the site of the primary tumour (Pantel, Schlimok et al. 1993; Klein, Seidl et al. 2002; Riethdorf, Wikman et al. 2008; Morgan, Lange et al. 2009). As they remain quiescent for some period of time this also suggests that they, or the cells in their microenvironment, or both undergo genetic changes which allow them to progress to a metastatic phenotype (Riethdorf, Wikman et al. 2008) (Riethdorf, Wikman et al. 2008; Klein 2009; Nguyen, Bos et al. 2009; Smith and Theodorescu 2009; Rose and Siegel 2010; Stoecklein and Klein 2010). A class of genes that has been implicated in the regulation of this process is metastasis suppressor genes (Smith and Theodorescu 2009). These are genes that inhibit metastasis but do not affect the ability of cells to produce a primary tumour, and they play key roles in invasion, dissemination, arrest, survival and colony formation. Their function must be lost or inhibited for a metastasis to develop and they represent fertile new ground for the development of antimetastatic therapeutics.

A number of metastasis suppressor proteins have been reported to inhibit breast cancer metastasis. Reduced levels of nm23 family proteins in the primary tumour have been reported to correlate with more aggressive phenotype in breast cancer patients (Galani, Sgouros et al. 2002; Terasaki-Fukuzawa, Kijima et al. 2002; Steeg, Ouatas et al. 2003; Peihong and Perry 2007), although conflicting results have also been presented (Charpin, Garcia et al. 1998; Belev, Aleric et al. 2002; Sgouros, Galani et al. 2007). The results seen in mouse models are more straightforward, where breast cancer cells with low expression of nm23 are more metastatic than those with high levels (Leone, Flatow et al. 1993; Bhujwalla, Aboagye et al. 1999; Tseng, Vicent et al. 2001). *In vitro* models have revealed that nm23 acts by reducing breast cancer cell motility and invasiveness (MacDonald, Freije et al. 1996; Russell, Pedersen et al. 1998; Steeg, Ouatas et al. 2003; Horak, Lee et al. 2007).

For Breast Cancer Metastasis Suppressor-1 (BRMS1), the clinical data reporting it to be a metastasis suppressor protein in breast cancer tumour samples is also conflicting (Kelly, Buggy et al. 2005; Hicks, Yoder et al. 2006; Lombardi, Di Cristofano et al. 2007). Again, its role in mouse models is clearer, where higher expression in breast cancer xenografts clearly resulted in reduced metastasis (Hedley, Vaidya et al. 2008; Hurst, Xie et al. 2008; Phadke, Vaidya et al. 2008). The stage at which BRMS1 suppresses metastasis is less clear, as it appears to affect a number of steps in the process of metastasis (Stafford, Vaidya et al. 2008). At least two of its functions appear to be increasing anoikis of cells free in the vascular system and inhibition of colonization of disseminated cells (Phadke, Vaidya et al. 2008). KAI1 (CD82, Tetraspannin), has also been clearly verified as a breast cancer metastasis suppressor in clinical samples, where decreased expression correlates with poor outcome (Yang, Welch et al. 1997; Christgen, Christgen et al. 2009; Malik, Sanders et al. 2009). Similar to BRMS1, KAI1 appears to act in multiple ways to inhibit metastasis and reduce breast

Breast Cancer Metastasis: Advances Through the Use of In Vitro Co-Culture Model Systems 515

points. One of the major disadvantages in using mouse models to study human cancer metastasis is that mice are not human, and there is no guarantee that the metastasis will develop in them in a way that recapitulates what happens in a human body. Mice do develop breast cancer as a heterogeneous disease, similar to humans (Andrechek and Nevins 2010), but there are significant differences between mice and humans in the capacity of the primary cells for transformation, the size of tumours, expression of hormone receptors and preferential sites of homing for breast cancer. For researching metastasis of human cells, immunodeficient mouse strains need to be used, taking the model a further step away from what happens in a human host. In addition human breast cancer cell lines may not accurately reflect the biological characteristics of in-vivo breast cancer such as natural evolution and tumor diversity. Given that caveat, mouse models are very important for testing pre-clinical data before moving on to clinical trials or human tumour tissue samples. There are many technical issues to take into account when considering the use of a mouse model to study breast cancer metastasis. Those are beyond the scope of this chapter but are very fully reviewed by Danny Welch (Welch 1997). Mouse models have been particularly useful in identifying molecules important for a number of steps in metastasis, such as epithelial to mesenchymal transition (EMT), invasion, extravasation and intravasation (Vernon, Bakewell et al. 2007). One approach to modeling metastasis is to use xenografts, where human tumour cells are injected subcutaneously or into the mammary fat pad of a mouse and allowed to develop a primary tumour that spontaneously metastasizes. A second approach is to inject tumor cells directly into the venous system, using tail vein injection or cardiac puncture. Tail vein injection results primarily in metastasis to the lung, but cardiac puncture results preferentially in bone metastasis. This approach obviates the need for development of the primary tumour but is not useful for studying some of the early steps of metastasis. The artificial injection of tumor cells directly into the venous system may produce pseudometastasis through a process of embolization rather than true physiological metastasis. A third approach is to utilize genetically engineered mice that have had a tumour suppressor gene deleted or an oncogene activated in an organ specific manner. Xenograft models and venous injections most generally use breast cancer cell lines, many of which are maintained and sold by the American Type Culture Collection (ATCC). These cell lines have a variety of gene expression profiles that identify them as similar to luminal, basal A or basal B [subtypes initially defined in tumour samples in 2006 (Fridlyand, Snijders et al. 2006)] and they show a variety of receptor and p53 profiles (Neve, Chin et al. 2006). Gene expression in these cell lines can be modified by over expression or deletion and the effect of the altered gene on metastasis can be monitored following injection. Although the resulting tumours are considered to metastasize "spontaneously", the injected cell lines are an artificial starting material as they have been cultured for long periods of time *in vitro* and do not resemble a spontaneously arising tumour. One of the advantages of xenograft models is that the cells of the primary tumour must interact with the stromal cells surrounding the tumour and must also interact with the stromal cells at the site of the metastasis for a productive metastasis to develop. A great deal of information about the interactions of human tumour cells and stromal cells has been accumulated by injecting human tumour cells and human mesenchymal stem stromal cells together in xenograft models [reviewed in (El-Haibi and Karnoub 2010)]. By using the same cell lines and different routes of injection it is possible to determine whether a gene is necessary for early steps of metastasis or whether

it is involved in later steps (Chabottaux, Ricaud et al. 2009).

cancer cell adhesion, migration and invasion *in vitro* (Malik, Sanders et al. 2009) and metastasis in mouse models *in vivo* (Yang, Wei et al. 2001). Other metastasis suppressor genes implicated in inhibiting breast cancer metastasis include KISS1 (Harms, Welch et al. 2003), MTSS1 (Parr and Jiang 2009) and alpha2beta1 integrin (Ramirez, Zhang et al. 2011), although their roles, at least in breast cancer have been less well studied.

As can be seen, the determination of the role of metastasis suppressor genes in metastasis using clinical samples is often confusing. This seemingly conflicting data may be a result of the many different experimental approaches to examining clinical samples; whether the samples are frozen or paraffin embedded and formalin fixed, whether mRNA or protein levels are the final determinant of expression (and these do not always correlate well), the type of extraction procedures used, and the source of the antibodies and staining methods for immunohistochemistry. The other difficulty with clinical samples is that they are almost exclusively derived from the primary tumour site, as biopsies of metastases are rarely carried out. Metastasis suppressor genes by definition do not inhibit events at the site of the primary tumour but must be inhibited for metastasis to take place. This inhibition may allow invasion of the circulatory system from the site of the primary tumour, survival through the process of transportation to the site of metastasis and evasion of the immune system, arrest within the metastatic niche, extravasion from the circulatory system or growth in the new environment (Kaplan, Psaila et al. 2006; Rabbani and Mazar 2007). Inhibition of expression at any step following detachment from the primary tumour would not likely be detected in the primary tumour.
