Background
Breast cancer is the most common carcinoma detected in women, accounting for about one fifth of new cancer cases in females [
1]. Surgical removal of the cancer represents the standard of care followed by radiation and/or adjuvant therapy in patients considered to be at particular risk for persistent local or systemic disease. Histopathological parameters are of particular importance for assessing tumor aggressiveness. This especially applies for pathological stage (pT), histologic grade, and nodal stage (pN). Although powerful in statistical analyses, these parameters are not sufficient to predict the prognosis of individual patients reliably enough in all cases. Analysis of molecular features in cancer cells bears the potential of providing a better estimation of the prognosis than classical pathological parameters alone [
2‐
4].
The
PTEN gene at 10q23 encodes a lipid phosphatase that functions as a direct antagonist of phosphatidylinositol 3-kinase and is involved in the regulation of the AKT pathway. Inactivation of
PTEN leads to constitutively activated levels of AKT, thus promoting cell growth, proliferation, survival and migration through multiple downstream effectors [
5].
PTEN is one of the most frequently deleted genes in various human cancer types [
6], and alterations of
PTEN were reported to have prognostic relevance in gastric cancer [
7], colorectal cancer [
8], non-small cell lung cancer [
9], diffuse large B-cell lymphoma [
10], mesothelioma [
11] and prostate cancer [
12].
In breast cancer - despite various earlier studies - frequency and relevance of
PTEN alterations is unclear. The frequency of
PTEN deletions or reduced expression varies from 4 % to 63 % in the literature [
13,
14]. Some studies on 99–151 breast cancer patients have suggested associations between
PTEN inactivation and poor prognosis [
15‐
17], but this could not be confirmed in other studies involving 212–670 cancers [
18‐
20]. In addition,
PTEN deletion has been intensively discussed as a potential predictor for failure of anti-HER2 therapy [
21‐
23].
To better understand the clinical relevance of PTEN deletions in breast cancer we analyzed more then 2,100 breast cancers with clinical follow-up data. Fluorescence in-situ hybridization (FISH) was applied for PTEN analysis because FISH is the gold standard for analyzing DNA copy number changes. Moreover, to better understand the role of PTEN deletions in cancers with HER2 amplification we analyzed a historical cohort of cancers that was collected before anti-HER2 treatments were routinely applied to women with HER2 positive breast cancer. Our data show that PTEN deletion is tightly linked to poor disease outcome and they suggest this also applies to the subgroup of HER2 positive cancers not treated with trastuzumab.
Discussion
Analyzing more than 1,200 informative breast cancers using a FISH probe directed against the known tumor suppressor gene PTEN at 10q23, we found that PTEN deletion is strongly linked to poor patient prognosis.
PTEN deletion was found in 18.8 % of breast cancers. These results are in the range of two earlier studies reporting
PTEN deletions in 12 %–19 % of unselected breast cancers using FISH [
27] or array CGH [
16]. In our study,
PTEN deletion was defined as “fewer
PTEN signals than centromere 10 signals in at least 60 % of all tumor cells”. These stringent criteria resulted in a 100 % concordance of results found by FISH and comparative genomic hybridization in a previous
PTEN study of our group in prostate cancer [
12]. A higher PTEN deletion rate by FISH (41 %) was only reported from a study on 199 highly selected metastatic breast cancers [
28], as well as from heterozygosity (LOH) studies, reporting
PTEN deletion in 30 %–41 % of 22–105 analyzed breast cancers [
29‐
33]. LOH induced by unequal allele distribution in the case of triploid or aneuploidy cancers or tumors with overrepresentation of chromosome 10 may contribute to the relatively high rates of
PTEN losses in these studies. FISH represents the gold standard for gene copy number analysis, because FISH is independent of the purity of cancer tissue and chromosomal aberrations such as polysomy. Deletions can be analyzed on a cell-by-cell basis, and abnormalities can be detected in a few cells or even a single cell.
Numerous studies measured alterations of PTEN protein expression by IHC in breast cancer. These studies show highly discordant results, with PTEN loss ranging from 4 %–63 % in 33–670 analyzed breast cancers [
13,
14,
18,
31]. Reasons for this variability might include inherent issues of IHC, as no validated antibody, threshold, or protocol is available yet for PTEN expression analysis by immunohistochemistry on FFPE tissue [
34]. In a previous study, we tested a series of seven anti-PTEN antibodies, but found all of them unsuitable for reliable measurement of PTEN expression in formalin fixed prostate cancer tissues since no meaningful association was found between the staining level of these antibodies and presence of
PTEN deletions or tumor phenotype [
12]. Discordant results were also reported for RNA expression analysis by RT-PCR. Here studies reported loss of PTEN expression in 20 % of 135 and 77 % of 59 analyzed breast cancers [
16,
27]. These discrepancies are not surprising, given that the results of such studies are substantially dependent on the purity of cancer tissue and the PTEN expression levels in normal breast epithelium, as well as in inflammatory or stromal cells.
PTEN deletions were unevenly distributed between breast cancer subtypes. Lobular cancers had a particularly low and cancers with medullary features had a markedly high rate of
PTEN deletions as compared to NST cancers. This is a novel and unexpected observation, given that gene mutations of
PTEN had been reported at similar rates in lobular (2 %) [
35] and NST (2.3 %) carcinomas [
36]. The high rate of
PTEN deletions in cancers with medullary features fits well to the increased proliferation rate, which is one of the hallmarks of this kind of tumor [
37]. However, it has been suggested that the favorable clinical course of cancers with medullary features may be strongly driven by a lower tendency for lymphovascular invasion and an activated immune defense that might overrule adverse genetic features [
38,
39].
The data of our study show, that – at least in the largest subgroups of NST carcinomas -
PTEN deletions are strongly linked to features of unfavorable tumor phenotype and to shortened overall survival. While most of the earlier studies on 22–258 breast cancers had also found associations with at least some phenotypic features of aggressive cancer [
15‐
17,
30‐
33,
40], data were more conflicting with respect to patient outcome. The vast majority of studies analyzed the prognostic impact of
PTEN alterations in NST cancers by immunohistochemistry [
15‐
20,
41‐
43]. Half of these studies suggested a prognostic impact of lost or decreased PTEN expression in 99–182 tumors using overall survival, metastasis free-survival, disease-related death or recurrence-free survival as clinical endpoints [
15‐
17,
43]. Other studies could not confirm a significant association between
PTEN alterations and patient outcome using these clinical endpoints in 97–670 patients [
18‐
20,
41,
42]. Only one study analyzed
PTEN deletions by FISH in a set of 135 tumors [
16]. In line with our analysis, this study revealed a strong link between
PTEN deletion and reduced metastasis-free interval. Therefore our findings, in combination with the ease of measuring
PTEN deletions by next generation sequencing methods, strongly prompt for including
PTEN deletion measurement in future multi-parametrical tests.
PTEN deletions were strongly linked to
HER2 and
MYC gene amplification in our patient set. Both HER2 and MYC are closely connected to AKT/PTEN signaling. MYC is a downstream effector of AKT-depended growth signaling but also controls AKT pathway activity in a negative feedback loop by up-regulation of PTEN [
44]. Obviously, impairment of this control loop by deletion of
PTEN can be expected to provide a growth advantage to cancer cells beyond
MYC amplification alone. HER2 is one of the receptor tyrosine kinases upstream of PTEN. Since
HER2 amplification strongly activates AKT signaling [
45], the additional deletion of
PTEN might confer a similar growth advantage to
HER2 amplified cells as to
MYC-amplified cells. In line with this notion, we found at least a strong trend towards an inferior prognosis for
HER2 and/or
MYC-amplified cancers with co-deletion of
PTEN as compared to tumors harboring only one of these alterations.
Further in line with an important functional interaction between PTEN and HER2, it has been suggested that
PTEN inactivation confers resistance to anti-HER2 therapies in cancers harboring
HER2 amplification. Given that studies analyzing the impact of
PTEN loss on trastuzumab response reported massively conflicting results [
46],
PTEN analysis has not been established as a clinical routine test prior to administration of trastuzumab. Some studies have suggested, that PTEN loss or reduced PTEN expression is sufficient to cause a decreased response to trastuzumab [
47‐
49], whereas other studies found that an additional
PIK3 mutation is required to confer resistance to anti-HER2 treatment [
50,
51]. Additional studies did not find a relationship between
PTEN alterations and sensitivity to trastuzumab [
51,
52].
Whether or not PTEN alterations are of relevance for response to anti-HER2 therapies is of utmost relevance. Since PTEN deletion is found in a subset of 27 % of HER2 amplified cancers that are eligible for HER2 treatment, it can be estimated that PTEN deletion could potentially account for therapy failure in about one quarter of breast cancer patients receiving HER2 inhibitors. It is of note, however, that the generally poor prognosis of patients with the combination of both PTEN deletion and HER2 amplification as determined in a historical breast cancer patient set predating the era of anti-HER2 therapies demonstrates that the poor success of trastuzumab in PTEN deleted cancers may also be driven by increased cancer aggressiveness irrespective of therapy response.
The strong association of
PTEN deletions with
HER2 and
MYC gene amplifications also fits well with the suggested role of PTEN in maintenance of genomic stability. PTEN is required for double strand breakage (DSB) repair by homologous recombination [
53]. Consequently, PTEN has been termed “A new Guardian of the Genome” in reference to the crucial role of the p53 tumor suppressor for genome integrity [
54]. Since erroneous DSB repair is one important prerequisite of DNA amplification by fusion bridge breakage [
55], it seems likely that
PTEN deletion contributes to development of gene amplification. Evidence for specific factors (such as for example
PTEN inactivation) facilitating the development of genomic instability comes also from previous studies by others and us showing a nonrandom accumulation of amplifications of different genomic regions, including
HER2 and
MYC, in a subset of breast cancers that are considered to show an “amplifier” phenotype [
26,
56‐
59].
It is thus of note, that
CCND1 amplifications were inversely linked to
PTEN deletions in our study. This finding provides further evidence for the existence of molecularly distinct subsets of breast cancers, one of which may be linked to generalized genetic instability with accumulation of numerous genomic amplifications (including
HER2 and
MYC) as well as deletions (including
PTEN), and another one that might develop
CCND1 amplification by a more targeted mechanism. This noting is supported by the fact that
CCND1 amplification is strongly linked to hormone receptor (ER/PR) positive breast cancers [
60], while
HER2 and/or
MYC (like
PTEN deletions in our study) are more often found in ER-negative than in ER-positive cancers [
26,
61]. Such a putative instability-independent amplification mechanism may be driven by the synergistic actions of ER signaling and CCND1 in promoting cell cycle progression [
62‐
64], and may include specific selection of cells with increased
CCND1 copy numbers.
The results of our study demonstrate a marked impact of the deletion of one
PTEN allele on breast cancer aggressiveness and prognosis. This finding is in line with several in-vivo studies demonstrating that deletion of one
PTEN allele is sufficient to have substantial impact on cell biology. For example in mouse models, heterozygosity for
PTEN leads to massively increased susceptibility to multiple tumor types [
65‐
67], increased cell proliferation in the thyroid and prostate gland [
65], and has been shown to cooperate with other genetic events, such as
ERG fusion in murine prostate cancer [
68,
69]. In an earlier study of 2,266 prostate cancers we had been able to demonstrate that no clinical difference existed between 225 patients with biallelic and 152 patients with monoallelic deletion of
PTEN [
12]. It was traditionally thought that chromosomal deletion is a mechanism to inactivate tumor suppressor genes in a two hit model for example in combination with a mutation of the other allele. However, large scale next generation sequencing of breast cancer genomes is increasingly demonstrating that mutations of “classical” tumor suppressor genes do often not accompany large deletions, suggesting that large deletions may not only serve for inactivation of a single gene [
70]. Increasing evidence suggests that deletions impact affected cells by reduced dosage of multiple different genes residing on a certain chromosomal area. Studies in prostate cancer have demonstrated that various chromosomal deletions have marked prognostic relevance [
12,
71‐
74]. As deletions are particularly frequent in breast cancer, it will be interesting to evaluate their clinical potential. This is all the more true, as deletions can relatively easy be studied by NGS even from formalin fixed material [
75].
Competing interest
The authors have disclosed that they have no significant relationships with, or financial interest in, any commercial companies pertaining this article.
Authors’ contributions
PL, EB, RS, and GS designed the study, and drafted the manuscript. UH, VM, AA and FJ participated in study design. VK, MK, ÖK, KH, BT performed FISH analysis and scoring. LT and AL participated in pathology data analysis. CH-M and RS performed statistical analysis. IW, LW, SM, SG, pp and CW participated in data interpretation, and helped to draft the manuscript. AM, RK, participated in data interpretation. All authors read and approved the final manuscript.