Breast cancer is a major public health concern among Western female populations. In 2004 breast cancer was the most common form of malignancy diagnosed in females in the USA and the second leading cause of cancer mortality in women 1. The great majority of these deaths are related to complications caused by metastatic disease, and in spite of therapeutic advances metastatic breast cancer is currently incurable 2. It has been estimated that 24–30% of women with node-negative disease and at least 50–60% of those with node-positive disease at diagnosis will relapse 3. Furthermore, approximately 6–10% of breast cancer patients present with clinical evidence of metastasis, and the median survival of those with metastatic disease is estimated to be 2–4 years 3.

The impact of metastatic disease on prognosis underscores the importance of improving upon currently available means of diagnosing and treating advanced breast cancer. One potentially powerful approach would be to identify early those patients who are at high risk for disseminated disease and then administer tailored treatment to reduce or eliminate the emergence of secondary lesions. As our knowledge of both the genetic and molecular mechanisms underpinning metastasis improves, it is becoming apparent that use of simple, reliable and robust assays that rely upon stable, constitutional fingerprints, including germline DNA polymorphisms, may provide an alternative means of identifying patients at risk for advanced disease. Recent studies have demonstrated that low penetrance germline polymorphisms influence overall susceptibility to breast cancer (for review, see the report by Houlston and Peto 4) as well as the likelihood of tumor progression and/or recurrence (for review, see the report by Foulkes and coworkers 5). Probably the best documented genes with respect to the latter are BRCA1 and BRCA2, with specific germline mutations being demonstrated to be associated with various indicators of poor outcome 6789.

Previously, our laboratory provided evidence that the host genetic background upon which a tumor arises can significantly alter metastatic efficiency. This was demonstrated by a breeding scheme in which the highly metastatic polyoma middle T transgenic mouse mammary tumor model was bred to a variety of inbred strains and the metastatic capacity of the tumors in each of the F₁ hybrid populations determined 10. Wide variations in metastatic efficiency for each of the different F₁ progeny were observed following mammary tumor development 10, and because all of the tumors were induced by the same transgenic event – namely the activation of the polyoma middle T antigen 11 – it was concluded that the observed difference in metastatic efficiency in F₁ progeny was a consequence of germline genetic variation between each of the different inbred strains.

Quantitative trait mapping experiments revealed the presence of a metastasis efficiency modifier locus linked to the proximal end of mouse chromosome 19 12. Subsequent evaluation of this locus identified a gene called Sipa1 ('signal-induced proliferation-associated gene 1'; also known as Spa1) as a likely candidate for this modifier locus. Sequence analysis of Sipa1 revealed an alanine to threonine mutation in the PDZ domain in mice with high metastatic potential (FVB strain, AKR strain) when compared with low metastatic phenotype mice (NZB strain, DBA strain) 13. Computer modeling of the wild-type and mutant forms of Sipa1 indicated that this mutation occurs on the open face of an α-helix, and in vitro analysis demonstrated that the polymorphism functionally impacts on protein-protein interactions. Furthermore, these data indicated that the metastatic capacity of mammary tumors in spontaneous metastasis assays was significantly altered by modulation of Sipa1 expression. Metastatic capacity was increased by Sipa1 upregulation as a consequence of ectopic gene expression 13, and a decrease in metastasis was seen when Sipa1 expression was reduced by RNA interference utilizing Sipa1-specific short-hairpin RNAs.

The aim of this study is to determine whether SIPA1 (MIM# 602180) plays a similar role in the modulation of metastatic potential in human breast cancer populations. We hypothesized that polymorphisms or haplotypes within key regulatory or coding elements of the human gene encoding SIPA1, located on chromosome 11q13.3, are positively correlated with the presence of metastases in breast cancer patients. To test this hypothesis, we designed a case-only association study in which a number of single nucleotide polymorphisms (SNPs) in SIPA1 were characterized in patients with metastatic or nonmetastatic breast cancer.

In this preliminary study we observed strong associations between the three SIPA1 SNPs and indicators of aggressive breast cancer in this population. Specifically, we report a number of particularly interesting findings regarding nodal status at the time of diagnosis and tumor sex hormone receptor status; these findings suggest that SIPA1 germline variation contributes to metastatic potential in breast cancer. With regard to lymph node involvement, the variant -313G>A and 2760G>A alleles were strongly associated with axillary node involvement at the time of diagnosis on univariate analysis. Indeed, the apparent effect of these SNPs on the frequency of metastasis that was evident upon univariate analysis was complemented by the outcome of multivariate and haplotype frequency estimation. Multivariate analysis of genotype frequencies showed that possession of either of the -313G>A or 2760G>A variant alleles was associated with a significantly increased likelihood of nodal involvement. Similarly, haplotype analyses demonstrated that the haplotype consisting of the wild-type allele of 545C>T and the variant alleles of the two other studied SNPs (haplotype 3; Table 5) is associated with nodal involvement at the time of diagnosis. This apparent association with nodal involvement at the time of diagnosis is particularly interesting given that there was no evidence of association of SIPA1 genotype status and tumor size. It was previously demonstrated that a linear relation exists between tumor diameter and the percentage of cases with positive lymph node involvement 18. One would therefore expect similar associations with tumor size, given the apparent relationship with nodal status at diagnosis. However, this was not the case in the present study, which may well be a consequence of inadequate statistical power to detect such an association. Alternatively, SIPA1 polymorphism might facilitate nodal metastasis independent of tumor size. Whichever is the case, further analysis of the significance of these polymorphisms in different populations is required to clarify the situation.

SIPA1 polymorphisms in this population were also associated with other features of tumors with aggressive behavior. Positive sex hormone receptor status correlates with favorable prognostic features, including a lower rate of cell proliferation and histological evidence of tumor differentiation 19. The 545C>T variant allele was associated with ER-negative and PR-negative tumors in this cohort. In addition, haplotype 4, which is defined by the variant allele 545C>T, is strongly associated with ER and PR tumor status (Table 5). This again highlights the deleterious effect that SIPA1 germline variation has on prognosis in primary breast carcinoma and implies that variant forms of this gene are associated with more aggressive disease forms.

The importance of these data is heightened when one considers the outcome of our linkage and functional studies of Sipa1 in a murine population 121320, coupled with consideration of the known physiologic functions of SIPA1. Traditional dogma has stated that cancerous cells gain the ability to metastasize as a consequence of somatic mutation (for example, through loss of heterozygosity or epigenetic events). However, it has become increasingly accepted that hereditary factors influence all stages of tumor pathogenesis in apparently sporadic cancers. These heritable defects will be of low penetrance, but hypothetically they could affect any part of the neoplastic cascade, including modulation of metastatic efficiency. Evaluation of quantitative trait loci in inbred mouse models has proven a particularly powerful means of studying the effects of this type of genetic variation. The present study is an additional example of how mouse quantitative trait locus data, when used in an appropriate manner, can help to unravel the complexities of hereditary influences in sporadic human cancer 21.

Although the specific mechanism by which SIPA1 modulates metastatic efficiency is currently unclear, it is worthwhile to speculate on how observations of SIPA1 functionality in mice and humans may relate to the process of tumor progression and characterization. Specifically, tumor development in a background of germline-encoded SIPA1 dysfunction may enhance the ability of tumor cells to escape the primary lesion through alteration in cell morphology/polarity and/or by weakening the strength of intercellular contacts. This argument gains further potency when one considers the negative regulatory effect of SIPA1 on Rap1, and that Rap1 has been implicated in the maintenance of epithelial polarization 22 as well as in the maintenance of intercellular adherens junctions 23. Further studies will be required to confirm whether this is the case.

SIPA1 has been shown to have GTPase activity that is specific for Rap1 and Rap2, both of which are members of the Ras family of GTPases, and are involved in regulating cell proliferation, differentiation, and adhesion 24. Indeed, SIPA1 appears to play a prominent role in regulating cell adhesion, and transient expression of this gene in HeLa cells induces cellular rounding up and detachment from the culture surface 25. It appears to act by modulating a number of cellular adhesion molecules, for example β₁-integrin, and thus regulates the interactions between these molecules and other cellular and extracellular matrix proteins involved in the process of adhesion 262728. It has been hypothesized that SIPA1 negatively regulates integrin-mediated cell adhesion via Rap1 GAP activity 27. Specifically, SIPA1 interacts with the cytoskeletal-anchoring protein AF6 through a common protein-protein interaction motif known as a 'PDZ domain' 27.

In addition to the effects on intercellular adhesion, disruption of SIPA1 has been shown to have potent effects on cellular proliferation. For example, SIPA1-deficient mice develop a spectrum of symptoms resembling those seen in human myelodysplastic and myeloproliferative disorders 2930. This is especially noteworthy because metastatic cells must be able to proliferate effectively at a distant site in what must be considered a foreign microenvironment, if they are to become a clinically relevant metastatic lesion. When all this evidence is considered, we argue that functional data from both mouse and human studies strongly support the findings of the present case-only study, in which we demonstrate that germline polymorphisms and haplotypes of SIPA1 are associated with enhanced metastatic efficiency in primary breast carcinoma.

However, further experiments will be required to confirm the role of SIPA1 germline variation in modulating tumor progression efficiency, the most obvious of which is to replicate the associations seen here in other populations. It is also vital to resequence various elements of the SIPA1 gene to attempt to identify novel polymorphisms, the most critical of which is that encoding the PDZ protein-protein interaction domain – a region that is critical to SIPA1 functionality 27. This endeavor is especially important given that the exon 12 2760G>A SNP – a noncoding, synonymous polymorphism – is strongly associated with aggressive disease in this study. This may well be a reflection of the physical proximity of this SNP to the other polymorphisms characterized in the study, which could have a more plausible effect on SIPA1 functionality. This hypothesis is somewhat supported when one considers that a haplotype containing the 2760G>A variant allele is associated with more aggressive disease at diagnosis, as measured by positive lymph node involvement. Conversely, it may well be that the 2760G>A SNP is in linkage disequilibrium with a nearby yet uncharacterized SNP with a more obvious functional effect. Indeed, one could postulate that the associations observed in the present study are in fact a consequence of another gene in linkage disequilibrium with SIPA1, because the functional significance (if any) of each of the polymorphisms has not been assessed. However, we argue that this is unlikely because we investigated genes in the chromosomal regions surrounding Sipa1 in the mouse and found no data supporting the hypothesis that surrounding genes are involved in modulating metastatic efficiency 13. This, in combination with observations relating to Sipa1 functionality in the mouse and the fact that human haplotype blocks are probably significantly smaller than in inbred mice, leads to the conclusion that the observed associations are a result of SIPA1 polymorphism.

The implications of this work are potentially far reaching in that it may well prove possible to assess metastatic potential through genotyping polymorphisms in germline DNA. The ability to evaluate the risk for metastatic disease from readily available tissues such as blood could be a powerful means to augment currently available means of assessing prognosis in breast cancer. However, we do acknowledge that the effect of germline SIPA1 variation in isolation is probably too low to allow for the development of an assay that would possess sufficient sensitivity and specificity to determine metastatic propensity accurately. However, it is plausible that a panel of polymorphisms derived from a number of metastasis 'susceptibility' genes would possess the power and accuracy to enable the development of a clinically relevant prognostic assay, which will necessitate the identification of further metastasis efficiency modifier genes.

Our results imply that SIPA1 germline polymorphisms are associated with aggressive disease behavior in the present cohort. If the results of this study hold true in other populations, then the potential for this to have a positive impact on current breast cancer staging and treatment protocols may be great. For example, staging of primary breast cancer in cases in which there is no palpable evidence of nodal metastasis typically requires axillary dissection to assess nodal status. However, controversy exists as to the actual benefit of nodal sampling in those patients with small tumors because of the morbidity frequently associated with this procedure. For example, the incidence of axillary node relapse in patients with stage T1a tumors (tumor >0.1 cm but not >0.5 cm in greatest dimension) treated without axillary dissection was 2% 31. It is plausible that analysis of SIPA1 polymorphisms could be used as an additional element of the staging process that might allow axillary node-negative patients with small primary tumors (for example, T1-2 N0) to be classed as being at increased risk for development of metastasis. In turn, this could then become a powerful influence on whether to initiate adjuvant therapy in patients who would not currently be considered for this type of treatment.

CI = confidence interval; D' = Lewontin's D prime; ER = estrogen receptor; NCBI = National Center for Biotechnology Information; OR = odds ratio; PCR = polymerase chain reaction; PR = progesterone receptor; SNP = single nucleotide polymorphism.

The authors declare that they have no competing interests.

NC was responsible for genotyping polymorphisms, contributed to study design, and drafted the manuscript. The University of California Irvine investigators played several roles in this report. HA-C contributed to the study design, the recruitment of subjects and the interviewing and biospecimen collection, and drafting and editing of the manuscript. AZ contributed to the study design, recruitment of subjects and data analysis, and drafting and editing of the manuscript. DP contributed to study design, biospecimen processing, and drafting and editing of the manuscript. JH participated in data processing and analysis, and editing of the manuscript. KH participated in study design, was responsible for supervising laboratory work, and was involved in manuscript preparation. All authors read and approved the final manuscript.