Abstract
Metastasis, the spread and growth of tumors at secondary sites, is an extremely important clinical event, since the majority of cancer mortality is associated with the metastatic tumors, rather than the primary tumor. In spite of the importance of metastasis in the clinical setting, the actual process is extremely inefficient. Millions of tumor cells can be shed into the vasculature daily yet few secondary tumors are formed. To successfully colonize a distant site tumor cells must overcome a series of barriers. Failure to complete any single step in the metastatic cascade abrogates the ability to form a secondary lesion. A variety of theories have been proposed to explain the inefficiency of the metastatic process. The most commonly accepted, the progression theory, posits a series of random mutational occurs within a primary tumor to generate a small subpopulation that acquires full metastatic capability. While significant evidence supports this model, recent discoveries demonstrating the ability to predict metastatic propensity from gene expression profiles in bulk tumor tissue are not consistent with only a small subpopulation of cells in the primary tumor acquiring metastatic ability. A second theory of metastatic inefficiency, the transient compartment theory, is more consistent with the microarray data, but does not completely explain observations like metastasis associated loss-of-heterozygosity events. To reconcile the observed results additional variables need to be added to the model of metastatic inefficiency. One possible variable that might explain the discrepancies is genetic background effects. Studies have demonstrated that the genetic background a tumor arises on can have significant affects on the ability of the tumor to metastasize and on gene expression profiles. Thus the observations could be reconciled by combining the theories, with genetic background influencing both metastatic efficiency and predictive gene expression profiles, upon which subsequently occur metastasis-promoting mutational and epigenetic events. If the genetic background is an important determinant of metastatic efficiency it would have significant implications for the clinical prediction and treatment of metastatic disease, as well as for the design of potential prevention strategies.

Keywords: Metastasis; Cancer; Progression; Genetics; Microarray; Gene expression; Inefficiency; Mouse models; Modifier; Quantitative traits; Genetic background

1. Metastasis
The process of metastasis is an extraordinarily complex process. To successfully colonize a secondary site a cancer cell must complete a sequential series of steps before it becomes a clinically detectable lesion. These steps include separation from the primary tumor, invasion through surrounding tissues and basement membranes, entry and survival in the circulation, lymphatics or peritoneal space, arresting in a distant target organ, usually followed by extravasation into the surrounding tissue, survival in the foreign microenvironment, proliferation, and induction of angiogenesis, all the while evading apoptotic death or immunological response (Figure 1; reviewed in [1]).

Figure 1: Simplified schematic representation of the steps of hemotagenous metastasis. Tumors grow and invade surrounding tissues then intravasate into the vasculature. Subsequently tumor cells disseminate through the body until they arrest, extravasate, then proliferate at the secondary site.

This process is of great importance to the clinical management of cancer since the majority of cancer mortality is associated with metastatic disease rather than the primary tumor [1]. In most cases cancer patients with localized tumors have significantly better prognoses than those with disseminated tumors. Since it has been estimated that 60–70% of patients have initiated the metastatic process by the time of diagnosis [2] better understanding of the factors leading to tumor dissemination is of vital importance. However, even patients that have no evidence of tumor dissemination at primary diagnosis are at risk for metastatic disease. Approximately one-third of women who are sentinel lymph node negative at the time of surgical resection of the primary breast tumor will subsequently develop clinically detectable secondary tumors [3]. Even patients with small primary tumors and node negative status (T1N0) at surgery have a significant (15-25%) chance of developing distant metastases [4].

In spite of the prevalence of secondary tumors in cancer patients, the metastatic process is an extremely inefficient process. To successfully colonize a distant site, a cancer cell must complete all of the steps of the cascade. Failure to complete any step results in the failure to colonize and proliferate. As a result, tumors can shed millions of cells into the bloodstream daily [5], yet very few clinically relevant metastases are formed [6, 7]. Although many steps in the metastatic process are thought to contribute to metastatic inefficiency, our incomplete understanding of this process suggests that we are aware of some but not all of these key regulatory points. For instance, killing of intravasated cells by hemodynamic forces and sheering has been thought be a major source of metastatic inefficiency [8]. However, recent evidence suggests that the destruction of tumor cells by hemodynamic force in the vasculature may not always be a major source of metastatic inefficiency. Cells in the bloodstream have been shown to arrest in capillary beds and extravasate with high efficiency and reside dormant in the secondary sites for long periods of time [9-11], sometime for years [12]. Micrometastases may form, but the bulk of these pre-clinical lesion appear to regress [10], probably due to apoptosis [13].

2. Theories of metastatic inefficiency
What then actually modulates metastatic inefficiency? To date, metastasis research as well as cancer research, has tended to focus on single genes, or at most, small collections of genes in an experimental model. This is often driven both by traditional scientific reductionist training as well as practical limitations. For example, research using genetically engineered mice usually involves the overexpression or mutation of a single gene. Although vast amounts of information have been gleaned from these studies and increasingly sophisticated models are available [14], they rarely accurately represent the complex milieu of genetic interactions that occur in naturally occurring cancer. For example, until the advent of inducible promoters, oncogenic constructs were expressed in all the cells of the target tissue in transgenic animals, rather than single clonal populations. Similarly, loss of gene function in knockouts was present in all of the tissues of an animal. Just as there is a growing understanding of the important role of the interactions of normal surrounding stroma and infiltrating tissues such as vascular epithelia in tumor formation, complex interaction between the genes identified by these techniques and other genomic elements, i.e. the genetic background that the tumor arises on, are likely to be important in metastatic potential.

2.1 Genetic modulation of metastasis inefficiency
The first suggestion of the role of genetic background as a critical determinant of metastatic potential was derived from transfection experiments. Introduction of proto-oncogenes can induce tumorigenicity and metastatic potential when transfected into NIH-3T3 cells. However, when the same oncogenes were transfected into cell lines derived from different strains of mice, metastatic potential, but not tumorigenicity, was lost [15, 16]. These results suggested either secondary mutations in metastasis promoting or suppressing genes were differentially present among the cell lines, or that allelic differences derived from the inbred strain progenitor were capable of modulating metastatic potential.

More compelling evidence for the existence of allelic variation influencing metastatic efficiency comes from experiments from our laboratory. These studies are based on the use of highly metastatic mouse mammary model, the FVB/N-TgN(MMTV-PyVT)^634Mul mouse [17]. This animal carries the mouse polyoma virus middle T antigen expressed from the mouse mammary tumor virus enhancer and promoter. Expression of the transgene induces synchronous multi-focal mammary tumors in all of the mammary glands of virgin female animals, and greater than 85% of these animals develop pulmonary metastases by 100 days of age [17].

Figure 2: Graphical representation of relative metastatic ability of the PyVT tumor in different genetic backgrounds. The Y-axis represents the density of pulmonary metastases of each genotype relative to the homozygous FVB/NJ background. The maternal genotype of each outcross is indicated across the X-axis. The FVB/NJ homozygous genotype is represented by the hatched bar. Only those genotypes in the strain survey [20] that did not significantly reduce the tumor burden are depicted here.

To determine whether there was genetic modulation of metastatic progression the genetic background that the tumor arose on was varied by a simple breeding strategy. The PyVT mouse was bred to a variety of different inbred strains selected from different branches of the mouse phylogenic tree [18, 19] to survey a broad range of the allelic diversity captured in the inbred strains. The F₁ progeny were aged to permit tumor induction and potential metastatic dissemination. Subsequently the lungs were examined to determine whether introduction of allelic variation had an affect on the density of pulmonary metastases. As can be observed in figure 2, a wide variation in metastatic efficiency was observed [20]. Since all of the tumors were induced by the same genetic event, expression of PyVT, the most likely explanation for this variation is that subtle genetic differences between the strains are affecting the metastasis process [21, 22].

Further evidence of the effect of background on metastatic efficiency was obtained by genetic mapping experiments. Using quantitative trait mapping strategies, three backcross mapping experiments and a recombinant inbred backcross were analyzed to identify chromosomal regions associated with metastatic efficiency. Two statistically significant associations were observed, one on chromosomes 6 and the other on 19 [23]. In addition, suggestive associations were observed for several other chromosomal regions. The ability to map metastasis efficiency loci within an inbred strain genome argues against random somatic mutations being the major determinant of metastatic efficiency since each individual animal would retain different sets of alterations, precluding meiotic mapping. Furthermore, the coincident mapping of metastasis efficiency modifying loci in independent experiments is consistent with inheritance by descent of a common allele during the genesis of inbred strains. This interpretation is further strengthened by the recent description of the limited haplotype diversity that is present in the inbred mouse genome [24-26].

Why do we need to add the additional component of complexity of genomic diversity to the theories explaining metastatic inefficiency? Other hypotheses have been developed over the decades to explain the phenomenon. Among the theories proposed are the transient compartment model [27] and the conventional progression model [28, 29]. Both of these theories are based on the supposition that it is a series of random events with in the tumor or tumor cell that are the primary determinants of the low efficiency of secondary tumor formation. Evidence exists for both models, however neither completely explains all of the observations.

2.2. The Progression Model
The progression model suggests that series of mutational events occurs either in subpopulations of the primary tumor or disseminated cells resulting in a small fraction of cells that the acquisition of full metastatic potential (see figure 3) [28, 29]. Alterations of the genome resulting in the silencing of metastasis suppressors and the up regulation of metastasis promoting elements, either by deletion or gene silencing events, has provided evidence for the progression theory of metastasis. The rapidly growing collection of metastasis suppressors, those genes whose reintroduction into tumor cells specifically interfere with metastatic colonization without affecting primary tumor initiation or growth kinetics, impact virtually every known step in the metastatic process [30, 31]. The statistical likelihood of stochastic events predicted by the model resulting in the appropriate combination of metastasis-associated genomic alterations is small, consistent with the poor efficiency of the process. However, in spite of the accumulation of evidence supporting this model, paradoxes remain. One paradox is the existence of patients with unknown-primary cancer (UPC) metastatic disease. These patients, estimated at approximately 5% of cases, present with disseminated disease but have no clinically detectable primary tumor or only a small well differentiated lesion found at autopsy [12]. The lack of large primary tumor mass could suggest that there was insufficient numbers of cells to achieve the necessary sequence of events predicted by the stochastically driven progression model.

Figure 3: Comparison of the progression and the transient compartment models of metastasis efficiency. In the progression model a series of heritable random mutational events occurs within a tumor resulting in a small subpopulation that acquires all of the necessary alterations for metastatic competency. In the transient model, all of the tumor cells have the basic ability, but due to reversible epigenetic events and position within the tumor, not all cells maintain the ability at all times.

A second paradox observed in the data contained in two recently published gene expression papers. Using either primary breast cancers [32] or comparisons between primary and metastatic human solid tumors [22] these investigators have identified gene signatures that are predictive of metastatic progression from bulk tumor tissue. The progression model predicts that only a small subpopulation of the tumor would attain metastatic capacity and therefore would not be expected to dominate the gene expression profile of bulk tumors (figure 4). As a result, it has been suggested that metastatic potential must be encoded early in tumorigenesis by the specific collection of initiating mutations rather than secondary metastasis promoting events [21, 22].

Figure 4: The paradox of microarray metastasis signature gene expression profiles. Recent studies have demonstrated that predictive gene profiles can be obtained from bulk tumor tissue (top, yellow tumor). Since the progression and transient compartment models predict only small subpopulations of metastatic competent cells at any given time, the expression profile of the metastatic competent cells (depicted in yellow) would be expected to be dominated by the signal of the non-metastatic bulk tumor tissue.

While this explanation accounts for small tumors with extensive metastatic behavior and metastatic expression signatures in bulk tumor tissue, it creates its own paradox. If every cell in a large tumor, or at least the majority of cells, is primed for metastatic capacity, why is the efficiency of the process still so low? Certainly different collections of oncogenic mutations might have different metastatic efficiencies, and interactions with stroma at secondary sites may play a critical role [33]. However, the potential for large numbers of circulating tumor cells in the vasculature [5] and the ability of the majority of those cells to arrest and successfully extravasate [10] in target tissues would predict more efficient colonization at distant sites than is, fortunately, observed.

2.3. The Transient Compartment Model
The transient metastatic compartment model was proposed by Weiss [27] to explain the lack of consistent increases in metastatic capacity of the secondary tumors compared to primary tumors. If metastatic capacity was due to a series of heritable mutation events, as predicted in the progression theory, it might be expected that cells that had successfully completed the metastatic cascade would be more efficient establishing new metastatic tumors than the primary tumor. However, this was not consistently observed in a number of experimental systems (reviewed in [27]). The transient metastatic compartment model proposed by Weiss suggests instead that all viable cells in a tumor acquire metastatic capacity, but due to positional and/or random epigenetic events only a small fraction are capable of completing the process at a given moment in time (figure 3) [27]. Thus, although a tumor may been a derived from a cell that successfully completed the metastatic process, but cells within the tumor no longer retain the capacity to colonize secondary sites due to random, or microenviromentally induced, epigenetic events or inadequate access to vasculature.

Support for this model comes from studies demonstrating that methylation inhibitors can modulate metastatic capacity of cell lines [34-38]. However, while global demethylation may mimic some of the proposed epigenetic events, these agents can cause chromosomal aberrations [39], opening up the possibility that the modulation of metastatic capacity was due to mutational rather than epigenetic events. In addition, genomic instability is a hallmark of solid tumors and increases in the numbers of chromosome aberrations often reflect poor prognosis [40]. The inability of cells isolated from metastases to be consistently more metastatic than the primary tumor could easily be explained by additional genomic events within cells that disrupt the delicate balance of molecular events required to successfully complete the metastatic cascade [41]. Furthermore, the transient compartment model does not explain the clonal nature of metastases [42-44]. Since primary tumors are known to be heterogeneous [45], if every cell had metastatic ability that was modulated only by transient epigenetic events then it is less likely that significant proportions of secondary tumors would appear to be of clonal origin [29, 46, 47].

3. Genetic background affects potentially resolves paradoxes
The influence of genetic background provides at least one method to reconcile the different observations associated with metastatic inefficiency. Some aspects of the genetic efficiency model are similar to the transient compartment model proposed by Weiss [27] and the early oncogenesis model posited recently by Bernards and Weinberg [21], with subtle but important differences. Like both of these models, the genetic efficiency model predicts metastatic capacity of the tumor would be encoded early (figure 5). However, unlike these models, the allelic composition of the host genome is a major determinant in metastatic efficiency, not activations or inactivation of genes by genomic rearrangements or epigenetic events. The combination of subtle changes in gene function, mediated by polymorphisms in coding sequence, splice sites, promoters and enhancers, would determine what specific secondary events (deletions, amplifications, epigenetic modulation), as supported by the progression model, would be necessary in each individual host genotype to successfully complete the metastatic cascade.

Figure 5: Influence of genetic background on metastatic efficiency. The allelic composition of the host genome is a major determinant of metastatic efficiency and predictive gene expression patterns. Low metastatic genotypes do not express the predictive pattern and are poorly metastatic. High metastatic backgrounds produce the predictive gene expression patterns in the bulk tumor tissue. Subsequently progressive events occur on this background producing metastatic capable subpopulations.

Importantly, the genetic efficiency model not only exerts its affects within the tumor cell itself, but also in the primary tumor stroma as well as the microenvironment at distant sites. Target organ microenvironment is known to play an important role in metastasis formation [48, 49]. Tumor cells are known to require normal stroma for important signaling events [50]. As a result, polymorphisms that alter the function of normal tissue functions, for example promoter polymorphisms altering cytokine levels, missense polymorphisms affecting adhesion molecule function, alterations in signaling cascades etc., may be as important a barrier to successful metastatic colonization as alterations occurring within the tumor cell itself. The growing evidence suggesting that the majority of tumor cells are capable of extravasating [10, 11] suggest that proliferation in the secondary sites may in fact be one of the most important determinants to whether cells proliferate into a secondary tumor or undergo apoptosis. Furthermore, it is conceivable that allelic variation may affect escape from immune surveillance. Subtle variations in the ability of the host to mount an effective cytolytic defense, coupled with the ability of highly malignant cells to down-regulate tumor specific antigens [51], might also play an important role in metastatic efficiency.

4. Implications
The genetic efficiency model provides a method to potentially resolve the paradoxes. For example, the paradox surrounding the predictive expression profiles observed can be explained if the microarray expression patterns are primarily influenced by genetic background and not the initiating oncogenic events. A number of studies support this possibility and have demonstrated that gene expression can vary due to genetic background [52-54]. To assess this possibility, we compared the expression of the 17-gene metastasis signature set described by Ramaswamy et al. [22] between the high metastatic FVB/NJ background with the low metastatic [NZB/B1NJ x FVB/NJ]F₁ background. 16 of the 17 of the mouse orthologs were expressed in the PyVT tumors. 15 out of 16 showed the same direction of expression as observed in the human primary versus metastasis [22, 54, 55]. Similar results are observed comparing FVB/NJ tumors with another low metastatic genotype ([DBA/2J x FVB/NJ]F₁; K. Hunter, unpublished results).

These observations, particularly the microarray data, have important implications for metastasis detection and management. If genetic background is a major influence on metastatic potential, as measured by predictive gene expression patterns in normal and tumor tissue, it suggests that like cancer susceptibility, there may be individuals or families present in the human population that are more susceptible to disseminated disease. It may therefore be possible to identify these individuals before they develop neoplastic disease so that they might be more aggressively treated with neo-adjuvant therapies immediately upon diagnosis of the primary tumor. Alternatively, since tumor dissemination often appears to be an early event, it is theoretically possible that a chemo-prevention regime might be developed that would prevent tumor metastasis before the primary tumor was clinically apparent, enabling the bulk of human cancer to be cured by surgical resection.

5. Summary
In conclusion, the identity of the genomic elements in the host background are modifying metastatic efficiency are currently unknown. They clearly warrant further investigations since the majority of the genetically defined regions are not associated with known metastasis suppressor genes. The metastasis suppressors that are associated with our genetically defined regions to not have any apparent molecular defects nor expression level differences between the high and low metastatic genotypes [54, 56]. Identification and characterization of these metastasis efficiency modifier genes may therefore yield novel targets to develop chemo-prevention agents or anti-metastatic therapies. The new strategies could be developed to either kill occult metastases or possibly by increasing the inefficiency of the myriad tasks necessary to generate a clinically relevant metastasis to the point where the odds of solitary, dispersed cancer cells successfully completing the metastatic cascade approaches zero.

Acknowledgements
I am grateful to Drs. Pat Steeg, Lalage Wakefield, Danny Welch and Warren Kruger for their critical comments on this manuscript.

References
[1] L.A. Liotta and W.G. Stetler-Stevenson, Principles of molecular cell biology of cancer: Cancer metastasis. Fourth Edition ed. Cancer: Principles & Practice of Oncology, (eds. DeVita, S.H.V., and Rosenberg, S.A.), Philadelphia, Lippincott Co. (1993) 134-149.

[2] S.A. Eccles, G. Box, W. Court, J. Sandle, and C.J. Dean, Preclinical models for the evaluation of targeted therapies of metastatic disease, Cell Biophys. 25 (1994) 279-291.

[3] R. Heimann, F. Lan, R. McBride, and S. Hellman, Separating favorable from unfavorable prognostic markers in breast cancer: the role of E-cadherin, Cancer Res. 60 (2000) 298-304.

[4] R. Heimann and S. Hellman, Clinical progression of breast cancer malignant behavior: what to expect and when to expect it, J. Clin. Oncol. 18 (2000) 591-599.

[5] T.P. Butler and P.M. Gullino, Quantitation of cell shedding into efferent blood of mammary adenocarcinoma, Cancer Res. 35 (1975) 512-516.

[6] D. Tarin, J.E. Price, M.G. Kettlewell, R.G. Souter, A.C. Vass, and B. Crossley, Mechanisms of human tumor metastasis studied in patients with peritoneovenous shunts, Cancer Res. 44 (1984) 3584-3592.

[7] D. Glaves, R.P. Huben, and L. Weiss, Haematogenous dissemination of cells from human renal adenocarcinomas, Br. J. Cancer 57 (1988) 32-35.

[8] L. Weiss, U. Nannmark, B.R. Johansson, and U. Bagge, Lethal deformation of cancer cells in the microcirculation: a potential rate regulator of hematogenous metastasis, Int. J. Cancer 50 (1992) 103-7.

[9] L. Holmgren, Antiangiogenis restricted tumor dormancy, Cancer Metastasis Rev. 15 (1996) 241-245.

[10] K.J. Luzzi, I.C. MacDonald, E.E. Schmidt, N. Kerkvliet, V.L. Morris, A.F. Chambers, and A.C. Groom, Multistep nature of metastatic inefficiency: dormancy of solitary cells after successful extravasation and limited survival of early micrometastases, Am. J. Pathol. 153 (1998) 865-873.

[11] G.N. Naumov, I.C. MacDonald, P.M. Weinmeister, N. Kerkvliet, K.V. Nadkarni, S.M. Wilson, V.L. Morris, A.C. Groom, and A.F. Chambers, Persistence of solitary mammary carcinoma cells in a secondary site: a possible contributor to dormancy, Cancer Res. 62 (2002) 2162-2168.

[12] G. Riethmuller and C.A. Klein, Early cancer cell dissemination and late metastatic relapse: clinical reflections and biological approaches to the dormancy problem in patients, Semin. Cancer Biol. 11 (2001) 307-311.

[13] C.W. Wong, A. Lee, L. Shientag, J. Yu, Y. Dong, G. Kao, A.B. Al-Mehdi, E.J. Bernhard, and R.J. Muschel, Apoptosis: an early event in metastatic inefficiency, Cancer Res. 61 (2001) 333-338.

[14] K.W. Hunter and D. Cozma, The Mouse in Cancer Research; Past, Present and Future(?), Current Genomics 2 (2001) 1-26.

[15] R.J. Muschel, J.E. Williams, D.R. Lowy, and L.A. Liotta, Harvey ras induction of metastatic potential depends upon oncogene activation and the type of recipient cell, Am. J. Pathol. 121 (1985) 1-8.

[16] A.B. Tuck, Wilson, S.M., and Chambers, A.F., Ras transfection and expression does not induce progression from tumorigenicity to metastatic abiiity in mouse LTA cells, Clin. Exp. Metastasis 8 (1990) 417-431.

[17] C.T. Guy, Cardiff, R.D., and Muller, W.J., Induction of mammary tumors by expression of polyomavirus middle T oncogene: A transgenic mouse model for metastatic disease, MCB. 12 (1992) 954-961.

[18] W.R. Atchley and W.M. Fitch, Gene trees and the origins of inbred strains of mice, Science 254 (1991) 554-558.

[19] J.A. Beck, S. Lloyd, M. Hafezparast, M. Lennon-Pierce, J.T. Eppig, M.F. Festing, and E.M. Fisher, Genealogies of mouse inbred strains, Nat Genet 24 (2000) 23-5.

[20] T. Lifsted, T. Le Voyer, M. Williams, W. Muller, A. Klein-Szanto, K.H. Buetow, and K.W. Hunter, Identification of inbred mouse strains harboring genetic modifiers of mammary tumor age of onset and metastatic progression, Int. J. Cancer 77 (1998) 640-644.

[21] R. Bernards and R.A. Weinberg, A progression puzzle, Nature 418 (2002) 823.

[22] S. Ramaswamy, K.N. Ross, E.S. Lander, and T.R. Golub, A molecular signature of metastasis in primary solid tumors, Nat. Genet. 33 (2003) 49-54.

[23] K.W. Hunter, K.W. Broman, T.L. Voyer, L. Lukes, D. Cozma, M.T. Debies, J. Rouse, and D.R. Welch, Predisposition to efficient mammary tumor metastatic progression is linked to the breast cancer metastasis suppressor gene Brms1, Cancer Res. 61 (2001) 8866-8872.

[24] Y.G. Park, R. Clifford, K.H. Buetow, and K.W. Hunter, Multiple cross and inbred strain haplotype mapping of complex-trait candidate genes, Genome Res. 13 (2003) 118-121.

[25] T. Wiltshire, M.T. Pletcher, S. Batalov, S.W. Barnes, L.M. Tarantino, M.P. Cooke, H. Wu, K. Smylie, A. Santrosyan, N.G. Copeland, et al., Genome-wide single-nucleotide polymorphism analysis defines haplotype patterns in mouse, Proc. Natl. Acad. Sci. U S A 100 (2003) 3380-3385.

[26] C.M. Wade, E.J. Kulbokas, 3rd, A.W. Kirby, M.C. Zody, J.C. Mullikin, E.S. Lander, K. Lindblad-Toh, and M.J. Daly, The mosaic structure of variation in the laboratory mouse genome, Nature 420 (2002) 574-578.

[27] L. Weiss, Metastatic Inefficiency, Advances in Cancer Research 54 (1990) 159-211.

[28] P.C. Nowell, The clonal evolution of tumor cell populations, Science 194 (1976) 23-28.

[29] I.J. Fidler and M.L. Kripke, Metastasis results from preexisting variant cells within a malignant tumor, Science 197 (1977) 893-895.

[30] E.C. Kauffman, V.L. Robinson, W.M. Stadler, M.H. Sokoloff, and C.W. Rinker-Schaeffer, Metastasis suppression: the evolving role of metastasis suppressor genes for regulating cancer cell growth at the secondary site, J Urol 169 (2003) 1122-1133.

[31] P.S. Steeg, Metastasis suppressors alter the signal transduction of cancer cells, Nat. Rev. Cancer 3 (2003) 55-63.

[32] L.J. van 't Veer, H. Dai, M.J. van de Vijver, Y.D. He, A.A. Hart, M. Mao, H.L. Peterse, K. van der Kooy, M.J. Marton, A.T. Witteveen, et al., Gene expression profiling predicts clinical outcome of breast cancer, Nature 415 (2002) 530-6.

[33] D.J. Ruiter, J.H. van Krieken, G.N. van Muijen, and R.M. de Waal, Tumour metastasis: is tissue an issue?, Lancet Oncol 2 (2001) 109-112.

[34] D.L. Trainer, T. Kline, G. Hensler, R. Greig, and G. Poste, Clonal analysis of the malignant properties of B16 melanoma cells treated with the DNA hypomethylating agent 5-azacytidine, Clin. Exp. Metastasis 6 (1988) 185-200.

[35] M. Ishikawa, F. Okada, J. Hamada, M. Hosokawa, and H. Kobayashi, Changes in the tumorigenic and metastatic properties of tumor cells treated with quercetin or 5-azacytidine, Int. J. Cancer 39 (1987) 338-342.

[36] R.S. Kerbel, P. Frost, R. Liteplo, D.A. Carlow, and B.E. Elliott, Possible epigenetic mechanisms of tumor progression: induction of high-frequency heritable but phenotypically unstable changes in the tumorigenic and metastatic properties of tumor cell populations by 5-azacytidine treatment, J. Cell Physiol. Suppl 3 (1984) 87-97.

[37] L. Olsson and J. Forchhammer, Induction of the metastatic phenotype in a mouse tumor model by 5-azacytidine, and characterization of an antigen associated with metastatic activity, Proc. Natl. Acad. Sci. U S A 81 (1984) 3389-3393.

[38] H. Stopper, R. Pechan, and D. Schiffmann, 5-azacytidine induces micronuclei in and morphological transformation of Syrian hamster embryo fibroblasts in the absence of unscheduled DNA synthesis, Mutat. Res. 283 (1992) 21-28.

[39] P. Frost, R.S. Kerbel, B. Hunt, S. Man, and S. Pathak, Selection of metastatic variants with identifiable karyotypic changes from a nonmetastatic murine tumor after treatment with 2'-deoxy-5-azacytidine or hydroxyurea: implications for the mechanisms of tumor progression, Cancer Res. 47 (1987) 2690-2695.

[40] T. Ried, K. Heselmeyer-Haddad, H. Blegen, E. Schrock, and G. Auer, Genomic changes defining the genesis, progression, and malignancy potential in solid human tumors: a phenotype/genotype correlation, Genes Chromosomes Cancer 25 (1999) 195-204.

[41] I.J. Fidler, Critical determinants of cancer metastasis: rationale for therapy, Cancer Chemother. Pharmacol. 43 Suppl (1999) S3-10.

[42] T. Nakayama, B. Taback, R. Turner, D.L. Morton, and D.S. Hoon, Molecular clonality of in-transit melanoma metastasis, Am J Pathol 158 (2001) 1371-1378.

[43] A.F. Chambers and S. Wilson, Use of NeoR B16F1 murine melanoma cells to assess clonality of experimental metastases in the immune-deficient chick embryo, Clin. Exp. Metastasis 6 (1988) 171-182.

[44] S.T. Cheung, X. Chen, X.Y. Guan, S.Y. Wong, L.S. Tai, I.O. Ng, S. So, and S.T. Fan, Identify metastasis-associated genes in hepatocellular carcinoma through clonality delineation for multinodular tumor, Cancer Res 62 (2002) 4711-4721.

[45] I.J. Fidler, S. Yano, R.D. Zhang, T. Fujimaki, and C.D. Bucana, The seed and soil hypothesis: vascularisation and brain metastases, Lancet Oncol. 3 (2002) 53-57.

[46] J.E. Talmadge, S.R. Wolman, and I.J. Fidler, Evidence for the clonal origin of spontaneous metastases, Science 217 (1982) 361-363.

[47] I.J. Fidler and J.E. Talmadge, Evidence that intravenously derived murine pulmonary melanoma metastases can originate from the expansion of a single tumor cell, Cancer Res. 46 (1986) 5167-5171.

[48] I.J. Fidler, Critical determinants of metastasis, Semin. Cancer Biol. 12 (2002) 89-96.

[49] V.K. Puduvalli, Brain metastases: biology and the role of the brain microenvironment, Curr. Oncol. Rep. 3 (2001) 467-475.

[50] R. Alessandro and E.C. Kohn, Signal transduction targets in invasion, Clin. Exp. Metastasis 19 (2002) 265-273.

[51] V. Schirrmacher, M. Fogel, E. Russmann, K. Bosslet, P. Altevogt, and L. Beck, Antigenic variation in cancer metastasis: immune escape versus immune control, Cancer Metastasis Rev. 1 (1982) 241-274.

[52] I.A. Eaves, L.S. Wicker, G. Ghandour, P.A. Lyons, L.B. Peterson, J.A. Todd, and R.J. Glynne, Combining mouse congenic strains and microarray gene expression analyses to study a complex trait: the NOD model of type 1 diabetes, Genome Res. 12 (2002) 232-243.

[53] E.E. Schadt, S.A. Monks, T.A. Drake, A.J. Lusis, N. Che, V. Colinayo, T.G. Ruff, S.B. Milligan, J.R. Lamb, G. Cavet, et al., Genetics of gene expression surveyed in maize, mouse and man, Nature 422 (2003) 297-302.

[54] T.H. Qiu, G.V.R. Chandramouli, N.W. Alkharouf, K.W. Hunter, P. Long, and E.T. Liu, Analysis of Gene Expression Profiles in MMTV-PyMT Transgenic Mice in Different Genomic Background, Manuscript in preparation (2003).

[55] K.W. Hunter, D.R. Welch, and E.T. Liu, Genetic background is an important determinant of metastatic potential, Nat. Genet. 34 (2003) 23-24.

[56] Y.G. Park, L. Lukes, H. Yang, M.T. Debies, R.S. Samant, D.R. Welch, M. Lee, and K.W. Hunter, Comparative sequence analysis in eight inbred strains of the metastasis modifier QTL candidate gene Brms1, Mamm. Genome 13 (2002) 289-292.

Back . . .