Introduction
The retinoblastoma tumour suppressor gene (RB1) encodes a nuclear phosphoprotein that plays a central role in regulating the cell cycle [
1]. Inactivation of both alleles of this gene is involved in the development of retinoblastoma, which is a rare childhood malignancy. The loss of RB1 is also a well-characterised occurrence in many other human tumour types and it is probable that the p16
INK4a-CDK4/6-RB pathway is disrupted in most human tumours [
2]. RB1 regulates progression through the G1 to S-phase transition of the cell cycle. In cells entering the cell cycle, extracellular signals induce the expression of D-type cyclins, which bind to and activate cyclin-dependent kinases (CDK4 and CDK6); these complexes in turn lead to the phosphorylation of RB and its dissociation from E2F family members that then transcriptionally activate many genes required for the S phase [
1]. The INK4 family of CDK inhibitors (p16
INK4a, p15
INK4b, p18
INK4c and p19
INK4d) inhibits CDK4 and CDK6, retaining RB in its hypo-phosphorylated E2F-associated state, thereby preventing G1 to S-phase progression. It has recently been shown that CDK4 and CDK6 (and CDK2) are dispensable for driving the essential cell cycle; however, they are required in specialised tissues and possibly to achieve higher levels of proliferation [
3].
Inactivation of the RB1 gene in breast cancer was originally shown using a series of cell lines [
4]. Subsequently, loss of heterozygosity (LOH) has been observed in primary tumours, but does not necessarily correlate with low RB1 protein expression as assessed by immunohistochemistry [
5,
6]. LOH has, however, been shown to correlate with low RB1 mRNA expression [
5]. There are also genetic events upstream of RB1 that may be present in breast tumours, which can negatively impact RB1 function by promoting its phosphorylation, that include p16
INK4a loss [
7] and cyclin D1 amplification/overexpression [
8].
Breast cancer is a heterogeneous disease, which can be separated into clinically significant subtypes as defined by molecular profiling [
9,
10]. In addition to reproducible gene expression differences between these subtypes, specific molecular alterations continue to be identified that correlate with each subtype. Tumours of the basal-like subtype generally have a high mitotic index, tend to be p53 mutated [
11] and highly express the proliferation signature, which is a gene cluster shown to contain many E2F target genes [
12,
13]. Here we report that LOH at the RB1 locus occurs at a high frequency in human basal-like and luminal B tumours, while occurring infrequently in luminal A and human epidermal growth factor receptor 2 (HER2)-enriched tumours. p16
INK4a is also highly expressed both by microarray and by immunohistochemistry in most of the RB1 LOH basal-like tumours, presumably due to a feedback caused by RB1 loss. These results further illustrate the unique biology of each breast cancer subtype.
Discussion
Our understanding of breast cancer biology has been improved by the identification of genomically defined tumour subtypes. These subtypes are defined by distinct gene expression patterns, molecular changes and potentially distinct developmental cell types of origin adding up to observed differences in outcome and responses to therapy. In this report, we show that the frequency of RB1 LOH varied significantly according to 'intrinsic' subtype. RB1 LOH occurs at a frequency of 72.2% in basal-like breast tumours and 61.5% in luminal B tumours, both of which are observed in retinoblastomas in the frequency range of 60 to 75% [
37‐
40]. RB1 protein staining as assessed by immunohistochemistry, however, did not correlate with RB1 LOH in our study (as has been reported before); however, a SAM analysis supervised by RB1 LOH robustly identified many E2F genes, and E2F-regulated target genes, while the RB1 protein-guided SAM analysis did not. This suggests that RB1 LOH is a better biomarker of RB-pathway function than immunohistochemistry staining for total RB1 protein.
Additional support for the functional loss of RB1 in basal-like tumours comes from the correlation with high p16
INK4a message and protein expression. The inverse relationship between p16
INK4a and RB1 expression in breast cancers has been previously reported [
41,
42]; however, this relationship and its association with basal-like tumours is new in this report. Another intriguing link between p16
INK4a and basal-like cells comes from studies on human mammary epithelial cells, which have been shown to resemble the basal-like subtype by gene expression analysis [
43,
44]. It has been shown that in order for human mammary epithelial cells to proliferate
in vitro in culture for an extended period they must overcome an RB-mediated stress associated senescence barrier (stasis), which usually involves spontaneously losing p16
INK4a expression by promoter methylation [
45]. The gene expression changes associated with this
in vitro transition are similar to those we previously reported to occur with RB1 LOH [
46].
In basal-like tumours
in vivo, however, the exact opposite seems to occur in that the RB-pathway barrier appears to be RB1 functional loss with a concomitant feedback loop that induces p16
INK4a gene and protein expression. The link between high p16
INK4a expression being caused by RB1 loss is known; RB1 recruits Polycomb repression complexes to the p16
INK4a locus, which silence p16
INK4a transcription [
47]. It is also well-known that cell cycle inhibition by p16
INK4a is RB-dependent [
48] and, therefore, these RB1-deficient breast tumours would be expected to be refractory to the high levels of p16
INK4a. This explains their high proliferation rates in the presence of high levels of p16
INK4a. High p16
INK4a expression has reproducibly been shown to be associated with poor prognosis [
41,
49‐
51] and in a recent study by Grupka and colleagues, p16
INK4a staining of sentinel lymph nodes was predictive in determining the presence of non-sentinel node metastases [
52]. Gauthier and colleagues have recently shown that in ductal carcinoma
in situ (DCIS) lesions, high p16
INK4a together with low Ki-67 (proliferation) appears to not promote tumour progression, while high p16
INK4a and high Ki-67 lead to subsequent tumours [
24]. Lastly, it has been demonstrated that the deletion of RB1 in the murine mammary gland is capable of initiating tumourigenesis and that many of these resulting tumours have basal-like features (E. Zacksenhaus, personal communication). In total, these data strongly argue that the RB-pathway lesion that occurs in most basal-like tumours is RB1 loss, possibly with a compensatory activation of p16
INK4a.
Basal-like tumours also highly express a recently published RB-loss gene expression signature [
29,
32], which we have shown to have significant similarity to a previously defined proliferation signature and our newly described RB1 LOH signature. Lastly, the p16
INK4a expression seen in DCIS by Gauthier and colleagues, and the elevated Ki-67 index seen in basal-like DCIS lesions, suggests that RB1 loss may be an early event for this tumour type [
53].
Similar to basal-like tumours, luminal B tumours also showed a high frequency (61.5%) of RB1 LOH in our study, but this was not associated with induction of p16
INK4a. The differential effect of RB1 loss on p16
INK4a expression in luminal B versus basal-like tumour cells implicates other transcription factors in addition to pRb-E2F in the regulation of this CDK inhibitor in luminal tumours. In the recent study by Bosco and colleagues, luminal tumour-derived cell lines were shown to be more proliferative and resistant to hormone therapy after knockdown of RB1 [
32], both of which are signatures of luminal B tumours [
9‐
11]. The RB-loss signature was shown to be predictive of outcome in a data set containing only oestrogen receptor positive (ER+) breast tumours treated with tamoxifen monotherapy. Therefore, the loss of RB1 function may also play a substantial role in the increased proliferation, possible resistance to hormonal therapies and poor prognosis that is seen in luminal B tumours. In addition, the knockdown of RB1 in established breast cancer cell lines has recently been shown to increase sensitivity to a variety of DNA-damaging therapeutic agents [
32,
54]. While these experiments were performed with ER+ tumour cell lines, it does open the possibility that the RB1-defect in basal-like tumours plays a role in their increased chemosensitivity compared with most luminal tumours [
55,
56]. This is supported by our findings of increased neoadjuvant response in patients expressing high levels of the four different RB1-proliferation associated signatures.
The presence of LOH is typically thought to indicate that a mutated allele is present on the other chromosome and that the LOH makes the cell homozygous or hemizygous for the mutated allele. There is little published evidence to suggest that dramatic structural changes aside from LOH are occurring at the RB1 locus in breast tumours. There are a few reports of alterations in RB1 in breast cancer with two reports showing structural changes as assessed by Southern blotting in 7% and 19% of primary tumours [
57,
58], and no published reports to our knowledge of point mutations. Interestingly, a study by Kallioniemi and colleagues looking at RB1 loss in clinical breast cancer samples by fluorescent in situ hybridisation showed that most of the cells within these tumours contained two copies of the RB1 gene even when they showed LOH by restriction fragment length polymorphism at the RB1 locus [
59]. When these studies are considered with the data presented here, they suggest a complex scenario where one allele is lost by LOH and the remaining allele/residual protein is compromised by an as yet to be identified mechanism(s) that potentially varies between tumour subtype, and potentially varies even within basal-like tumours. For example, some basal-like tumours with LOH show complete loss of RB protein, while others show expression and both types show high proliferation. As opposed to breast cancer, there is a great deal known about the mechanisms of RB1 loss in retinoblastoma [
60‐
62]. As in retinoblastoma, it is clear that a combination of techniques will need to be applied in order to identify the precise mechanisms of RB1 inactivation in breast cancer.
Competing interests
CMP has stock ownership in University Genomics. Other authors declare no competing interests.
Authors' contributions
JIH participated in the data analysis, scoring of immunohistochemistry staining, development of the figures and the writing of the manuscript. XH prepared DNA from tumours and normal patient lymphocytes, and immunohistochemistry staining and scoring. CF performed statistical analysis. CMP was the principal investigator, conceived and designed the study, and drafted the paper. All authors read and approved the final manuscript.