Introduction
Breast cancer is the most common malignancy in women, with the mortality rate being especially high in patients in whom brain metastases develop. Approximately 15% to 20% of breast cancers metastasize to the brain, although incidence rates are increasing [
1]. The incidence of metastases is thought to be increasing because of the improved treatment of metastases at other distant sites and advances in imaging techniques, leading to improved detection of central nervous system (CNS) metastases [
2].
Metastasis formation is a highly selective, multistep process, involving complex interactions between tumor and host cells. To metastasize, tumor cells must disengage from the primary tumor, invade the stroma, and penetrate into vessels, where they disseminate, extravasate, and start to grow at distant organ sites. As a distant metastatic site, the brain forms a special challenge for tumor cells because of the blood-brain barrier [
3]. In addition, all other steps have to be successfully completed for the tumor cell to survive and expand. The molecular basis for all of these steps is still unclear, and several models have been suggested [
4,
5].
Comparative gene-expression analyses on primary breast tumors and lymph node metastases have indicated that, in general, metastases have very similar expression signatures compared with their parent tumors [
6,
7]. However, detailed analyses have also revealed that a number of genes are consistently differentially expressed between primary tumors and metastases [
8‐
10] and that metastases often show a greater variety of aberrations than the primary tumor [
11,
12]. At the chromosomal level, even greater differences have been described between primary breast tumors and their derived metastases. Most of the relevant studies compared the chromosomal aberrations in matched primary breast tumors and lymph node metastases [
13‐
16], and only a very few studies on distant metastases exist [
17‐
19]. In general, all of these studies showed that metastases harbor more and also new aberrations that could not be found in the corresponding primary tumors (reviewed in [
20]). These results imply that the clonal evolution of a tumor is more complex than would be predicted by linear models, highlighting the importance of investigating distant metastases as the end point of the metastatic cascade.
In this study, the patterns of chromosomal aberrations of primary tumors and brain metastases from breast cancer patients were compared with array-comparative genomic hybridization (CGH) and microsatellite analysis. The goal was to identify genetic alterations in the primary breast tumors associated with metastatic spread to the brain to be able to define subgroups of high-risk breast cancer patients. Our results indicate that loss of 10q and especially phosphatase and tensin homologue (
PTEN) could be predictive factors for the development of brain metastases. Interestingly, whereas loss of PTEN is generally very rarely seen in most epithelial tumors, it is one of the most frequent aberrations found in primary glioblastomas [
21] and other CNS malignancies [
22,
23], indicating that loss of PTEN might be an important factor for breast tumor cell survival in the CNS environment.
Discussion
Central nervous system metastases are a frequent complication of many solid tumors. Approximately 15% of all epithelial tumors metastasize to the brain, with incidence rates highly dependent on the primary tumor type. Whereas prostate cancer very rarely metastasizes to the brain (1% to 5%), small-cell lung cancer (40%) and breast cancer (15% to 20%) commonly metastasize to the brain [
38]. Apparently, the brain microenvironment is especially permissive for the growth of disseminated tumor cells from some carcinomas but not from others. The mechanisms by which metastatic tumor cells adapt to the selection pressure exerted by the brain microenvironment are still unknown.
In this study, our aim was to identify putative molecular markers associated with the development of brain metastases in breast cancer. First, we screened for chromosomal aberrations by array CGH. The most prominent finding of the loss of 10q in brain metastases was validated in a larger study population, and the tumor-suppressor gene PTEN was found to be the potential target gene in this region.
Overall, the array CGH results of the primary breast tumors were in agreement with those described previously [
39‐
41]. In general, the brain metastases showed aberration patterns similar to those of the primary tumors. However, in the brain metastases, a remarkably higher frequency of gains and losses was found at almost every chromosomal locus. Only a gain at 1p and a loss at 16q, described as being typical of luminal breast tumors and as markers of a favorable prognosis, were more common in the primary tumors [
39‐
41]. This finding is not surprising, as most of the primary tumors were hormone receptor (HR) positive, whereas 38% of the brain metastases were HR negative. Statistically significant differences were found between the primary breast tumors and brain metastases at nine different loci on six different chromosomes.
The most significant differences were found at chromosomes 7 and 10. Chromosome 7 contains two regions, 7p22-p15 and 7p11.2, that were gained or amplified in more than 70% of the metastases and gained in only 3% to 13% of the primary tumors. Whereas the 22-Mbp region 7p22.1-p15.3 contains many genes, the second gained region on chromosome 7 contains only one gene, the epidermal growth factor receptor (
EGFR);
EGFR is a well-known and important gene in breast cancer initiation and progression [
42]. Recently, Gaedcke
et al. [
43] reported a
de novo protein expression of EGFR in the brain metastases of matched primary and metastatic breast cancer cases, and mouse models have shown the EGFR ligand HBEGF to be one of the key mediators of cancer-cell passage through the blood-brain barrier [
5].
Most of the long arm of chromosome 10 was lost in 60% to 90% of the brain metastases, but only in none to 13% of the primary tumors. Numerous previously published large-array CGH datasets, also including locally advanced breast cancer, seldom showed a loss of 10q [
39‐
41]. The loss of 10q is, in general, rarely seen in most epithelial tumors, whereas it is the most common aberration found in glioblastomas, and it also is common in other CNS malignancies [
21,
22]. Similarly, the gain of 7p is among the most frequently found gains in astrocytomas [
44]. Both aberrations also are present in melanomas, which often metastasize to the brain (Additional file
8). The origin of these brain-specific aberrations in metastases could be explained by different hypotheses [
45]. The first one is that only few cells carried these genetic alterations in the primary tumors, and thus they could not be detected when the tumors were analyzed in bulk. Subsequently, cells with these alterations selectively metastasized to the brain and formed the bulk of the metastases. Alternatively, these additional alterations might have occurred at the distant site of metastasis, and therefore represent
de novo mutations that were not present in the primary tumors. The third scenario is that only a small fraction of the primary tumors contained these aberrations and that these tumors are specifically prone to relapse in the brain. Finally, another possibility is that several cells metastasize to the brain (for example, as a tumor-thrombus), but only those with alterations at 10q and/or 7p are able to survive in the brain environment, giving rise to metastases. Recently, by massively parallel DNA sequencing, Ding
et al. [
46] showed that metastases were indeed significantly enriched for shared mutations, which supports the last model.
Matched primary and metastatic tumor samples could be investigated in four cases by microsatellite analysis. Three cases showed identical aberration patterns, whereas in the fourth case, the AI imbalance was larger (that is, in the metastases, the distal marker was also affected by AI). Furthermore, primary tumor samples from patients in whom brain metastasis later developed showed a frequency of AI at 10q that was almost as high as in the brain metastases, but AI was rarely seen in primary tumors without brain relapse or other distant metastases. These results indicate that the loss of 10q does exist in a fraction of primary tumors with a high risk of developing brain metastases. The size of the aberration can expand in metastases and thus become more detectable by, for example, CGH. Because aberrations in 10q were not associated with any clinical factor other than brain relapse, this implies that loss of 10q is a specific marker of brain metastasis and is thus needed for the outgrowth of the breast tumor in the brain. However, this hypothesis must be validated in future studies, both functionally and on independent larger cohorts of patients.
The AIs were concentrated around two core regions, the first one around markers D10S1765 and D10S541 containing the
PTEN locus, and the second around markers D10S1236 and D10S190 at 10q26. The
PTEN gene located at 10q23.31 is a well-described tumor-suppressor gene, also in breast cancer; PTEN functions as an important tumor suppressor by negatively regulating the PI3K-mediated cell-signaling pathway [
47]. The present microarray analysis showed that PTEN is significantly downregulated in brain metastases compared with nonmetastatic primary tumors. Furthermore, mutation screening of the
PTEN gene in brain tumors showed that the frequency of the mutation was much higher (15%) in primary breast tumors (none to 5%) than previously described [
48,
49].
Several studies have shown that ERBB2/HER2 and the basal subtype of breast cancer are the predominant types of breast cancer that metastasize to the brain [
43,
50,
51]. To avoid misleading results by analyzing unmatched samples, the primary tumor sample cohort was matched for the main clinicopathologic characteristics in our AI analysis. When we classified the primary tumors as being ER/PR-positive, triple-negative, and HER2-positive tumors, no association was found between PTEN status (or loss of 10q) and breast cancer subtype, indicating that loss of PTEN is an independent predictor of brain metastases. In a recent publication, the protein expression of PTEN was analyzed in 54 brain metastasis samples from breast cancer patients [
52]; no correlation was found between PTEN loss and subtype in this study. Furthermore, a high concordance rate (83%) of PTEN expression was found among the 12 matched pairs. The authors also investigated the role of PTEN expression in primary breast cancer progression by using
in silico expression datasets with 855 patients [
52]. They showed that, in all patients, lower levels of PTEN expression were associated with a poor prognosis and shorter time to brain recurrence, irrespective of hormone-receptor and HER2 status after 5 years. Also, this analysis was independent of subtype. In addition, we found, by using the same data set, a significant downregulation of PTEN expression among primary tumor patients with brain relapse compared with patients with bone relapse, but not to the lung. This finding is in line with the findings from Bos
et al. [
4,
5], who found a significant overlap of brain with lung-relapse signature, but not with the bone signature, which argues for the important role of environment interaction in metastasis formation in different organs.
Interestingly, the two predominant genes derived from this breast cancer study on brain metastases also play a prominent role in the development and progression of primary brain tumors. The
PTEN gene is one of the key tumor-suppressor genes found in primary glioblastomas, and it is often (15% to 40%) silenced through mutations (reviewed in [
21]). Interestingly, also in glioblastomas, PTEN inactivation does not seem to be required for tumor initiation, but its loss is a hallmark for progression to highly malignant cancer [
53]. Together with
EGFR amplification, the loss of PTEN is the most frequent alteration observed in primary glioblastomas. Thus, these two genes, which are both involved in the PI3K kinase pathway, may play a key role in the growth of malignant cells in the brain environment and therefore might be suitable targets for therapeutic intervention.
Competing interests
The authors declare that they have no competing interests. The work was funded by DFG (PA 341/15-2, HW, KP) and EU (DISMAL; HW, KE, KP, MZ), ERC Advanced Investigator grant no. 269081 (DISSECT, KP), VFK Krebsforschung GmbH (TIME, HW, KE, KP).
Authors' contributions
HW, ND, LU, MW, IH, and KY carried out the molecular genetic studies, CB, DK, and MZ carried out the statistical analyses. DK and BY performed the microarray analyses. HW, KL, NW, GS, MG, VM, and KP participated in the design of the study and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.