The diverse techniques employed in this study, including genetical genomics [21
], functional genomics, sequence analysis, and molecular epidemiology, have allowed us to identify Rrp1b
as a candidate for both a tumor progression and metastasis modifier in mice, and a marker of inherited breast cancer metastasis susceptibility in humans. Furthermore, functional data provides evidence that Rrp1b
regulates ECM gene expression, and that a nonsynonymous SNP in RRP1B
is associated with tumor progression and disease-specific survival in pilot epidemiology experiments.
was identified through two distinct experimental approaches designed to address two independent questions: (a) what is the molecular mechanism(s) by which Sipa1
modulates metastasis, and (b) what drives ECM dysregulation in metastasis-prone primary tumors? Yeast two-hybrid assays and functional genomic studies addressed the first question, which identified Rrp1b as binding to the polymorphic PDZ domain of Sipa1. The second question probed the origins of metastasis-predictive gene expression signatures. ECM genes are components of all metastasis-predictive gene expression signatures in both humans [9
] and mice [12
], a finding that may well be explained in part by constitutional variation [13
]. To test this question, we examined whether ECM eQTLs and metastasis modifiers might be the same entities by analyzing ECM gene expression in a RI mouse panel. This led to the identification of several eQTLs, with a locus on proximal chromosome 17 displaying the strongest linkage. The peak linkage region of this locus encompasses both Rrp1b
and a tumor growth and progression QTL [6
], and rather remarkably, when we examined expression of transcripts within the peak eQTL linkage region, Rrp1b
was highly correlated with metastasis-predictive ECM gene expression. Taken together, these data suggest that Rrp1b
is a potential dual ECM and tumor progression candidate.
Further experimentation demonstrated that ectopic expression of Rrp1b
in two highly metastatic mammary tumor cell lines and a mouse fibroblast cell line modulates ECM gene expression, a finding concurrent with our initial hypothesis that Rrp1b
is indeed the chromosome 17 ECM eQTL. It should be noted however, that the directionality of the ECM expression changes observed in response to Rrp1b
activation in these cell lines are not directly comparable to the changes in ECM expression observed in metastasis-predictive gene expression signatures [9
]. Primary tumors are composed of a variety of cell and tissue types, and this cellular and microenvironmental heterogeneity is not accurately reflected by in vitro growth conditions of single cell types. Nevertheless, we do argue that our in vitro experimentation demonstrates that Rrp1b
modulates the expression of metastasis-predictive ECM genes in a variety of individual cell lines. It is also evident that further work will be necessary to define the complex microenvironmental relationships that modulate ECM gene expression in bulk tumor tissue and their relationship to overall levels of Rrp1b
In addition, it should be noted that eQTL analysis is most commonly associated with expression changes in normal tissues, and not the neoplastic tissues analyzed in this study. However, eQTLs owe their existence to germline polymorphism, and such variation will be present in tumor tissue in addition to normal tissues. With this in mind, it is therefore not unreasonable to assume that the phenotypic effects of eQTLs will be observed in tumors as well as normal tissue. Yet neoplastic tissues possess an inherent genomic instability, and it is therefore possible that variations in tumor gene expression patterns could also arise from somatic mutation. However, we have demonstrated that eQTLs can be genetically mapped in tumors, which suggests either that similar somatic mutations consistently occur in the majority of the tumor tissue in a subset of the RI strains, or that the observed eQTLs result from inherited polymorphisms. Our previous demonstration that the same ECM genes used to define tumor eQTLs are differentially expressed in normal mammary tissues derived from high- and low-metastatic mouse genotypes [13
] suggests that such differential expression may be partially regulated by germline polymorphism. At this stage, however, we cannot formally dismiss a role for somatic mutation in ECM gene expression variation within RI mammary tumors.
We also described a number of important functional differences in Rrp1b between the progenitors of the AKXD RI panel, the high metastatic AKR/J, and low metastatic DBA/2J genotype mice. Specifically, the proximal promoter of Rrp1b in AKR/J mice contained polymorphisms that may reduce Rrp1b expression. This difference in promoter activity is one possible explanation for the observed differences in Rrp1b expression in the normal mammary tissue of these mice, with the high metastatic capacity AKR/J mouse having significantly lower levels of Rrp1b activity than the low metastatic DBA/2J mouse. Furthermore, ectopic expression led to reduced metastatic potential and primary tumor growth following tumor cell implantation into mice. In combination, these observations suggest, at least in this mouse model that increased Rrp1b expression correlate with a better outcome.
We have used a dual approach to try to address the importance of RRP1B
in human breast cancer progression. The first of these approaches was to use data derived from microarray expression analysis of the Mvt-1/Rrp1b
cells to address one of the central goals of our research: the translation of experimental data from mouse models of human breast cancer into potentially clinically relevant observations. Based on the work of Bild et al. [20
] we identified an RRP1B
gene expression signature and demonstrated it that predicts outcome in a publicly available and well-characterized breast cancer cohort [10
]. This gene signature not only held strong prognostic value in the Dutch study cohort [10
], but was also able to stratify those patients with ER positive tumors and LN negative disease at presentation into high and low risk categories. There has been significant interest in using gene expression profiles for improved patient stratification [23
] in the clinic since it raises the possibility of improvements in breast cancer subtype classification, which in turn could enable clinicians to tailor treatment to individual patients. Whether the RRP1B
signature proves of clinical value in this respect is at present unclear, and further testing of its prognostic value in different cohorts will be required to address this possibility. The significance of this study, however, is not the identification of yet another prognostic signature, but the fact that the underlying casual element is known. Identification of other genetic elements that drive the predictive gene expression patterns may provide a more robust means of complementing currently available tests used for the assessment of prognosis in breast cancer. Furthermore, this type of study also provides us with potentially novel and important insights into the mechanisms underlying the metastatic process.
Further supportive evidence for the role of RRP1B in human breast cancer progression was evident in the two pilot epidemiological studies, both of which found an inverse relationship between the variant A allele of the 1421G→A RRP1B SNP and poor outcome markers. These consistent findings indicate that 1421G→A is a marker for disease progression, and patients who carry the A allele are less likely to present with advanced disease than homozygous carriers of the more common G allele. These data are consistent with the results of functional analysis of Rrp1b, and associate RRP1B with disease outcome in human breast cancer. It should be noted that the variant 1421G→A allele was associated with improved outcome in those individuals with ER positive tumors, which may permit better stratification of patients who are currently thought to be in a low risk category. A particularly intriguing question is if and how a patient's 1421G→A genotype affects expression of the 33-gene RRP1B expression signature, and whether polymorphisms in the promoter of RRP1B in linkage disequilibrium with the 1421G→A SNP are more important in this respect. These studies are currently ongoing in this laboratory. Indeed, a link between constitutional polymorphism and bulk tumor gene expression would be particularly significant given the technical difficulties associated with tumor gene expression profiling and the relative ease of SNP genotyping. These results, while consistent between the studies and in support of our hypothesis, must be considered only as preliminary. Further investigations in larger epidemiology studies specifically designed to address tumor progression and outcome, rather than tumor incidence, will be necessary to gain further support for the role of germline polymorphism in RRP1B in breast cancer progression.
Several differences were evident between the epidemiology study cohorts, the most notable of which was that the 1421G→A SNP was associated with breast cancer-specific survival only in the Baltimore cohort, a discrepancy that is likely due to several factors. First, both cohorts are relatively small (n
< 300), thus some differences might arise from statistical power issues. More importantly, the study population compositions differed: the Orange County cohort was derived from a population-based case-cohort study, including all cancer patients, regardless of stage, with more than 10 years of follow up, whereas the Baltimore cohort is a surgical breast cancer population, and therefore is biased against patients with metastatic disease at presentation. Furthermore, unlike the Orange County cohort, the Baltimore cohort contained African-American and Caucasian women. Differences in race/ethnicity influence allele frequency and disease outcome, and some variability in results is therefore expected. While the participation of patients from different race/ethnicities also strengthens study design, stratification is required. Stratification on race/ethnicity in the Baltimore cohort indicated that race/ethnicity was not a significant confounding factor. It is interesting to note, however, that the variant allele distribution is different in African-Americans and Caucasians in the Greater Baltimore cohort (0.099 in African-Americans versus 0.411 in Caucasians). It is known that African-American women have a poorer prognosis compared to other breast cancer patients [25
], and given the protective effect exerted by the variant 1421G→A, it is interesting to speculate that polymorphisms in genes such as RRP1B
may be driving these ethnicity-specific differences in outcome. Thus, further characterization of this SNP in larger and ethnically diverse cohorts is required to determine the influence of race/ethnicity upon the association between this SNP and breast cancer survival.
The function of Rrp1b
and its exact role in metastasis remain unclear at this time. Sipa1
was originally cloned as a mitogen-inducible protein [26
] that was subsequently shown to be a negative regulator of Rap1 by serving as a GAP for Rap1 [27
has significant effects on cellular adhesion [28
], primarily related to its effects on Rap1, which has been implicated in maintaining the integrity of polarized epithelia [29
] and intercellular adherens junctions [30
], and potentially integrating signaling between cadherins and integrins [31
]. Rrp1b may therefore mediate tumor cell adhesion properties by altering intercellular and cell–ECM contacts in a Sipa1-dependent and Rap1-dependent manner. It should be noted, however, that the human polymorphism in RRP1B
falls outside of the domain that directly interacts with the PDZ domain of Sipa1. Whether this polymorphism impacts the enzymatic function of Sipa1 or mediates metastatic potential through some other mechanism is unclear and currently under investigation. Similarly, it is unclear whether the amino acid substitution in human RRP1B directly affects function. Based on the mouse model, where increased expression confers protection against malignant progression, the variant leucine in the human ortholog may phenocopy the mouse situation by activating some function of RRP1B.
Further in vitro analysis will be required to clarify the different situations in the two species and is currently under investigation in our laboratory.
In addition to the negative regulatory role on Sipa1
function, there is some evidence to suggest that Rrp1b
may be involved in RNA metabolism. Protein homology analysis has shown Rrp1b to contain a Nop52 domain, a motif found in proteins critical to 28S rRNA generation [32
], and a previous yeast two-hybrid analysis has shown that Rrp1b may also interact with Lsm1, which is a protein involved in regulation of mRNA degradation [33
]. Publicly available databases show that RRP1B
is ubiquitously expressed at a somewhat low level, although it is expressed at a slightly higher level in lymph nodes in humans (http://smd.stanford.edu/cgi-bin/source/sourceSearch
). Differential expression of RRP1B
has been reported in fibroblasts from patients with systemic sclerosis, an autoimmune disorder characterized by dysregulation of a variety of ECM genes, including procollagens I, III, and VI [34
], consistent with our results. Further research, however, is clearly needed to fully explore the role of Rrp1b
in human breast cancer and other tumor types. Unraveling the mechanisms of action and the molecular pathways that they regulate are likely to provide novel and valuable insights into tumor dissemination and metastasis.