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 circulation, lymphatics, or peritoneal space, arresting in a distant target organ, usually, but not always (Al-Mehdi et al. 2000), followed by extravasation into the surrounding tissue, survival in the foreign microenvironment, proliferation, and induction of angiogenesis, all the while evading apoptotic death or immunologic response (reviewed in Liotta and Stetler-Stevenson 1993).

In spite of the prevalence of secondary tumors in cancer patients, the metastatic process is an extremely inefficient one. 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 (Butler and Gullino 1975), yet very few clinically relevant metastases are formed (Tarin et al. 1984). Although many steps in the metastatic process are thought to contribute to metastatic inefficiency, our incomplete understanding of tumor progression 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 to be a major source of metastatic inefficiency (Weiss et al. 1992). 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 (Luzzi et al. 1998), sometime for years (Riethmuller and Klein 2001). Micrometastases may form, but the bulk of these preclinical lesions appear to regress (Luzzi et al. 1998), probably because of apoptosis (Wong. et al. 2001).

Previously we demonstrated that a potentially major factor that might be modulating metastatic efficiency and other phenotypes could be attributed to polymorphisms present in the genetic background of the individual. Using the highly metastatic mammary tumor transgenic tumor model FVB/N–TgN(MMTV–PyMT)634mul (PyMT) and a simple breeding scheme, we were able to significantly alter the tumor latency, growth rate or metastatic capacity of the transgene-induced mammary cancer by altering the genetic background of the host animal (Lifsted et al. 1998; Le Voyer et al. 2001). Since normal tissues and autochthonously arising tumors have the same constitutional genetic information, genetic polymorphism that modifies tumor phenotypes can operate in the tumor cell, the surrounding stroma, the stroma of the secondary site, intermediate tissues during dissemination, or in any combination of the above simultaneously. Thus, genetic constitutional polymorphism affects tumorigenesis at the level of the whole organism, possibly modulating the function or efficiency of any or all of the critical determinants in the efficiency of neoplastic transformation and progression.

Identification and characterization of the genes underlying the tumor modifier, therefore, would likely provide interesting and important insights into tumor heterogeneity in the human population. Using a number of experimental genetic crosses and quantitative trait analysis, a number of loci that modulated each of these phenotypes were mapped in the genome (Le Voyer et al. 2000, 2001; Hunter et al. 2001). In addition, a number of suggestive but statistically non-significant loci were also observed. Quantitative trait analysis and positional cloning strategies have been performed (Cozma, et al. 2002) or are currently ongoing to identify the statistically significant modifier loci. However, before undertaking laborious attempts to identify reproducible but statistically insignificant modifier loci, validation of these loci was required.

This study therefore describes the efforts to validate the putative presence of five different suggestive modifier loci: 1 accelerating tumor latency, on Chromosome 7, and four that modulate tumor progression or metastasis, on Chromosomes 9, 13, and 17, as well as confirm the presence of Mtes1 on Chromosome 19. By constructing and analyzing chromosomal substitution strains (Pyle 1978; Carracedo et al. 1995; Nadeau et al. 2000), we have confirmed the presence of all of the suspected loci. Identifying and characterizing the genes that underlie these quantitative traits will hopefully gain important insights into the biological mechanisms that modulate the ability of tumors to progress to the mortality-associated metastatic state.

Material and methods

Animals

FVB/N-TgN(MMTVPyVT)634Mul mice were obtained from W. Muller, McGill University, Montreal, Quebec, Canada (Guy et al. 1992). FVB/NJ, NZB/BINJ, and I/LnJ mice were purchased from The Jackson Laboratory (Bar Harbour, ME). Inheritance of the polyoma transgene was determined by PCR amplification of weanling tail biopsy DNA with the following primers: 5′-AAC GGC GGA GCG AGG AAC TG-3′; 5′-ATC GGG CTC AGC AAC ACA AG-3′.

Construction of the chromosomal substitution strains

The I.F. (1/LnJ × FVB/NJ) Chromosomes 7 and 9 substitution strains were constructed using The Jackson Laboratory Custom Breeding Service. The N.F. (NZB/B1NJ × FVB/NJ) chromosomes 9, 17, and 19 chromosomal substitution strains were generated in our laboratory by marker-assisted breeding, essentially as described previously (Markel et al. 1997). Briefly, I/LnJ or NZB/B1NJ animals were bred to FVB/NJ, and the progeny backcrossed to FVB/NJ. Backcross progeny were genotyped with loci at approximately 15–20–cM spacing (loci available on request) to identify those animals containing the chromosome, of interest and the minimal amount of unlinked donor DNA. These animals were selected for the next round of backcrossing to FVB/NJ and the process repeated more than 5 generations until no unlinked donor DNA was detected in the chromosomal substitution strains. Sibling animals were then interbred to achieve homozygosity for the donor chromosome of interest.

Determination of tumor phenotypes

Transgene-positive females were maintained at 3–5 animals per cage and screened by palpation three times a week for the presence of the primary mammary tumor. Diagnosis was performed by a single operator to minimize interpersonnel variance. The location of the tumor was recorded and the animals were examined for an additional week to confirm diagnosis, and then aged for 40 days diagnosis to permit development of metastases. After 40 days the animals were anesthetized with Avertin and then sacrificed by cervical dislocation. Total tumor mass was determined by complete dissection of all tumors in all animals. Lungs were harvested, fixed in 10% paraformaldehyde, embedded in paraffin, sectioned, and stained with hemotoxylin–eosin. Three coronal nonadjacent sections of both lungs, each separated by 100 μm, were prepared from each animal. The slides were examined with a Leica M420 Macroviewer with an Apozoom lens under 12 × magnification. Three fields were scored for each slide, for a total of 9 fields per animal. Pulmonary metastatic density was determined utilizing a Leica Q500MC Image Analysis System. The metastasis index was measured as the number of multicellular metastatic lesions observed per square micron of lung tissue. The Q500MC system was utilized to eliminate alveolar space from the measurement of lung tissue area to thereby control for various degrees of lung inflation at sacrifice. Average metastasis size was calculated based on the total area of metastatic tissue on the slide divided by the number of metastases observed. All slides were read blind and analyzed by a single operator to improve technical consistency.

Genotyping

Tail biopsy DNA was used as a template for PCR reactions. Microsatellite primers were purchased from Research Genetics (Huntsville, AL). PCR reactions were performed basically as described previously (Le Voyer et al. 2000). Reactions were performed in a PTC200 Thermocycler (MJ Research, Watertown, MA) and analyzed on 4% agarose TAE gels.

Statistical analysis

Statistical analyses were performed by the Mann–Whitney U-test using the statistical program StatisticaTM (StatSoft, Tulsa, OK).

Complex trait analysis

Genetic analysis of backcrosses was performed using Map Manager (http://www.mapmanager.org/mmQTX.html) QTXb20. Composite mapping for the Chromosome 17 tumor growth QTL was performed by marker regression after controlling for the three known tumor growth modifiers (Mmtg1–3, linked to markers D4Mit17, D4Mit308, and D7Mit76). Significance of the results was found by permutation testing, permutating the data 10,000 times in 1-cM steps using an additive regression model.

Results

Construction of chromosomal substitution str-ains

Chromosomal substitution strains (CSS) were generated using either I/LnJ or NZB/B1NJ as donors to test for the presence of putative mammary rumor modifier genes on Chromosomes 7, 9, 13, 17, and 19. A total of four CSS lines were generated: one from an I/LnJ donor, FVB/NJ–Chr7I/LnJ, and three from the NZB/B1NJ donor, FVB/NJ–Chr9NZB/B1NJ, FVB/NJ–Chr17NZB/B1NJ, and FVB/NJ-Chr19NZB/B1NJ_206. The NZB/BINJ Chr 13 substitution strain was lost during breeding, possibly due to incompatibility with the FVB/NJ background.

Confirmation of putative tumor latency modifier

The CSS lines were then bred to the PyMT transgenic animal to confirm or refute the presence of the previously identified suggestive tumor modifier loci. As seen in Table 1, the Chrs 9, 17, and 19 CSS stains did not exhibit any significant changes in tumor latency. The FVB/NJ–Chr7I/LnJ CSS subline, however, demonstrated a very significant acceleration of tumor latency, as predicted by our earlier mapping studies (Le Voyer et al. 2000). Therefore, the suggestive latency modifier linked to the locus D7Mit246 was validated in this study.

Table 1 Tumor latency
Full size table

Effect of chromosomal substitutions on tumor growth potential

The effect of the chromosomal substitution on the accumulation of the total tumor tissue mass in each animal during the course of the experiment was also examined. As anticipated from previous mapping studies (Le Voyer et al. 2001), a significant reduction in tumor mass was observed in the FVB/NJ-Chr7I/LnJCSS (see Table 2), confirming the presence of the tumor growth modifier locus Mtmg3 (Le Voyer et al. 2001). Unexpectedly, significant reductions in tumor burden were also observed in the rest of the CSS. The Chr 9 CSS also demonstrated significant tumor reduction of approximately 20%, while significant reductions of 44% or 26% were observed for the Chrs 17 and 19 CSSs, respectively. Reexamination of quantitative trait mapping data in our crosses did not reveal any evidence for tumor growth modifiers on Chr 9 or 19 (data not shown). However, composite interval mapping of the I/LnJ cross revealed a statistically significant tumor growth modifier linked to D17Mit81 (see Fig. 1). The NZB Chr 17 CSS tumor modifier might therefore represent this locus.

Table 2 Total tumor burden
Full size table
Fig. 1
figure 1

Composite interval mapping result for tumor growth phenotype in the I/LnJ backcross. The chromosome is oriented with the centromere at the top. The black line indicates the LRS statistical score for the presence of a modifier at each point along the chromosome. The red line denotes the additive effect of potential modifier genes, with positive additive effect to the right of the chromosome, negative to the left. The green vertical lines represent statistically suggestive, significant, and highly sign-ificant results (p values of 0.63, 0.05, and 0.001, respectively) as determined by genomewide permutation testing. Yellow bars indicate the most likely position for the modifier, based on bootstrap analysis.

Full size image

Effect of chromosomal substitutions on pulmonary metastases

Lungs were harvested from each of the chromosomal substitution strains, sectioned, and the average density and size of pulmonary metastases determined. As expected from previous studies, the Chr 19 strain showed significant reduction in both the number and the size of metastases, consistent with the presence of the metastasis efficiency suppressor locus Mtes1 (see Tables 3 and 4). Similarly, Chrs and 17 demonstrated significant reductions in both metastatic phenotypes (see Fig. 2), consistent with the presence of metastasis modifiers on these chromosomes. The NZB/B1NJ-derived Chr 9 CSS also displayed significant reduction in both the number and the size of the metastatic lesions, consistent with the preliminary genetic mapping results (Hunter et al. 2001).

Table 3 Relative density of pulmonary metastases
Full size table
Table 4 Relative Metastasis size
Full size table
Fig. 2
figure 2

Examples of histological sections for scoring pulmonary metastases. (A) FVB/NJ homozygous PyMT control; (B) FVB/NJ–Chr7I/LnJ. Both the average density and the size of pulmonary metastases are reduced by the presence of the I/LnJ-derived Chromosome 7 compared with the wild-type control. Arrows indicate examples of metastatic lesions in the FVB/NJ–Chr7I/LnJ sample.

Full size image

Discussion

Previous mapping studies suggested the presence of modifiers of PyMT tumorigenicity on Chrs 7, 9,13, and 17 in addition to the previously described Apmt, Mtmg, and Mtes loci (Le Voyer et al. 2000, 2001; Hunter et al. 2001). To validate these putative loci before initiating modifier gene cloning strategies, CSSs were constructed to assay the phenotypic effect of each chromosome in isolation. As anticipated from the genetic mapping studies, all of the CSSs replicated the predicted phenotypic modifications. A locus on Chr 7 affecting the latency of frank, palpable tumors was observed, as predicted. Since this locus mapped to a different, albeit, location adjacent to the mammary growth modifier locus Mtmg3, also located on this chromosome, and no correlation was observed between latency and tumor growth rate (Le Voyer et al. 2001), we believe that these are likely to be independent modifier loci.

Earlier work also suggested the presence of a very weak tumor growth modifier linked to D9Mit207 in an NZB × PyMT backcross (Hunter et al. 2001 and unpublished data). The marginally significant though modest reduction of tumor growth in the NZB/B1NJ-derived CSS is consistent with this observation. Significant tumor reduction was observed in the I/LnJ-derived Chr 9 CSS. No evidence of a tumor growth modifier on Chr 9 had previously been demonstrated in the I/LnJ cross used in our studies of the Mtmg loci. However since this cross is segregating three previously known major tumor growth modifiers [Mtmg1-3 (Le Voyer et al. 2001)] it is possible that this locus was overlooked due to a lack of statistical power to detect minor modifiers.

In contrast, no evidence was observed for tumor growth modifiers on Chrs 17 and 19 in any of the previously analyzed crosses (Hunter et al. 2001; Le Voyer et al. 2001). In the case of the Chr 17 locus, this is due primarily to the fact that the composite interval mapping had been overlooked in the previously described work. On Chr 19 a locus thought to specifically modify metastatic progression [Mtes1 (Hunter et al. 2001)] had already been identified through genetic mapping, but no evidence for tumor growth modification had been linked to this locus. However, it is possible that the Mtes1 locus might also have some effect on tumor growth since tumor burden and metastatic capacity are positively correlated (Lifsted et al. 1998). Alternatively, an independent tumor growth modifier may exist on this chromosome that did not segregate with the metastasis modifier locus. Resolution of this question will require analysis of subcongenic intervals.

All of the CSSs also had significant effects on metastatic density and the size of the metastatic lesions. Although the intent of these studies was to identify novel metastasis-related modifier genes, the fact that all of the CSSs also significantly affect tumor growth and tumor growth and metastasis have been positively correlated in our previous studies precludes separation of the metastatic and tumor growth modifiers. Therefore, we have chosen to designate the newly validated tumor modifiers as modifiers of tumor progression rather than specifically of the metastatic process. At this time it is not possible to distinguish whether these phenotypes are being modulated by single loci or multiple loci on each chromosome. Higher-resolution genetic analysis of subchromosomal regions will be required to resolve this question and to begin to narrow down candidate intervals for positional cloning strategies.

Identification and confirmation of the presence of tumor latency and progression loci on these chromosomes will aid in discovering the factors that modulate tumor phenotypes in the human population. At present, although we know a great deal about the oncogenic events that induce tumorigenesis, little is known about the polygenic factors that influence penetrance or expressivity of the tumors. These modifiers are likely to be important for the understanding of human cancer. For example, only about 50% of the Ashkenazi women who carry mutations in the BRCA1 gene develop breast cancer during their lifetime (Struewing et al. 1997). While environmental factors and exposures undoubtedly play an important role in the induction of these tumors, these results also suggest that modifiers like the Apmt loci and other cancer modifiers (Threadgill et al. 2004) may play an important role in the etiology of human neoplastic disease.

Likewise, the metastasis and tumor growth modifying loci described in this study may also be relevant for the understanding of human tumor progression and metastatic disease. Understanding the events and factors that influence tumor progression is clearly of great importance for the development of more effective prevention or clinical interventions. Several studies were published that demonstrated the ability to classify primary tumors as metastatic or nonmetastatic based on gene expression from bulk tumor tissue (van’t Veer et al. 2002; Ramaswamy et al. 2003). Since a substantial portion of the tumor must exhibit a particular expression pattern to be detectable in microarray experiments, the authors interpret their data to suggest that metastatic capacity is likely to be encoded early in tumorigenesis by the particular collections of oncogenic events that initiate the tumor. In contrast, we previously demonstrated that the genetic background on which a cancer arises has a significant affect on the ability of mammary tumors to successfully colonize the lung. In addition, we and others (Eaves et al. 2002; Qiu et al. 2004) have demonstrated that genetic background significantly influences gene expression, including the metastasis signature genes (Hunter et al. 2003; Qiu et al. 2004). These observations suggest that the propensity of a tumor to metastasize and the predictive gene expression profile are at least in part set by the combination of subtle changes in gene function, mediated by polymorphisms in coding sequence, splice sites, promoters, and enhancers before tumor initiation. Subsequently, progressive events such as translocations and deletions occur to produce rare cells that are capable of completing the metastatic process.

These results, if extrapolated make a very specific prediction, that is, that the metastasis predictive gene expression signature set should be diff-erentially regulated in normal tissues before the onset of oncogenesis. In support of this hypothesis, we have recently demonstrated that at least the majority of the described metastasis signature genes are differentially regulated in normal mammary tissue when compared between high and low metastatic genotypes (Hunter et al. unpublished). Therefore, it may be possible to identify these individuals before they develop neoplastic disease so that they might be treated more aggressively with neoadjuvant 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 chemoprevention 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. Identification of those genes underlying the propensity to metastasize may therefore yield valuable tools for identifying those patients at high risk for metastatic disease, as well as potentially providing new targets for chemoprophylaxis or novel antimetastasis therapeutics.