Retinoblastoma tumor suppressor pathway in breast cancer: prognosis, precision medicine, and therapeutic interventions

The retinoblastoma tumor suppressor (RB) is a potent regulator of cellular proliferation whose status provides critical information related to breast cancer prognosis and therapeutic interventions.

Although initially identified in a pediatric eye tumor, the last 30 years of research have demonstrated that RB plays an important role in many cancers. Loss of heterozygosity at the Rb1 locus represents the seminal basis for the development of retinoblastoma, and was the basis through which the gene encoding RB was identified [1‐3]. RB has no known enzymatic activities, but functions in a host of processes by mediating protein interactions that are important for multiple phenotypes [4‐7]. It is well known that RB binds to the E2F family of transcription factors and can repress the activity of E2F, leading to the attenuation of many genes that are required for cell cycle progression. This is very much the canonical view of RB in cell cycle control, although one should note that RB exerts many important affects that are independent/complementary. For example, RB can control chromatin cohesion, chromatin structure, tumorigenic proliferation, differentiation, and cell death through mechanisms beyond the dogmatic influence on E2F activity [4, 7, 8].

In spite of this complexity, there is wide agreement that RB must be inactivated for cell cycle progression, and thereby cell division, to occur. In normal physiology this is achieved via the phosphorylation of RB, which is catalyzed by cyclin-dependent kinases (CDKs) [9‐11]. In particular, CDK4 and CDK6 are activated in response to the accumulation of D-type cyclins by mitogenic signaling and initiate the phosphorylation of RB [9, 12, 13]. CDK2 also plays a role in the phospho-inactivation of RB [14, 15]. The biochemical importance of these processes was demonstrated using mutants of RB that cannot be phosphorylated and are potent inhibitors of cell cycle progression in the vast majority of tumor cells [10, 11, 16]. Furthermore, the CDK4/6 inhibitor p16ink4a is dependent on RB for function in suppressing cell cycle progression and inducing senescence [17‐20]. Together, the pathway of CDK4/6, RB, and p16ink4a define the proximal RB pathway (Figure 1). Importantly, this pathway demonstrates mutually exclusive disruption in cancer [21‐24]. For example, tumors with loss of p16ink4a will retain wild-type RB, while tumors mutant for RB will express p16ink4a at very high levels (Figure 2).

Although the RB pathway is often simplified to the level that all perturbations are viewed as being equivalent, there are clearly unique features of CDK4/6, p16ink4a, and RB biology. For example, CDK4/6 and cyclin D₁ have been shown to have important targets beyond the phosphorylation of RB [25‐27]. p16ink4a and RB are members of gene families, and compensatory mechanisms can mitigate the effects of loss of these tumor suppressors in specific settings, most notably in mouse models [28‐30]. Additionally, phosphorylation of RB can be regulated by a plethora of mechanisms beyond CDK4/6 [9, 31] and, due to its downstream position in the pathway, RB loss has a particularly significant effect on cell cycle control.

The analysis of many tumor types has indicated that the RB pathway is perturbed in some form or another in most cancers. However, there is often a tumor-specific tropism for the mechanisms of pathway inactivation [21, 23, 24, 32]. For instance, pancreatic cancers frequently lose p16ink4a, while osteosarcomas often lose RB. There is thus both common loss of the pathway and intrinsic diversity based on the mechanism through which the pathway is inactivated.

In breast cancer, different subtypes of disease are dominated by differential mechanisms of RB pathway inactivation. Estrogen receptor (ER)-positive breast cancers generally exhibit deregulation of the kinase components CDK4/6 as a result of aberrant cyclin D₁ expression or amplification [33‐37]. Her2-positive breast cancers typically push on the pathway through D-type cyclins, and similarly there are few cases that exhibit RB loss. In contrast, loss of the Rb1 gene and the RB protein has been documented at high frequency in triple-negative breast cancer (TNBC) [38, 39] . In spite of these generalities, one should note that any breast cancer can exhibit loss of RB, loss of p16ink4a, or amplification of cyclin D₁; there is thus the opportunity to evaluate how these events impinge on the underlying biology of disease and the prognostic and therapeutic implications in the clinic.

The majority of invasive breast cancers are believed to develop from precursor lesions. In particular, ductal carcinoma in situ (DCIS) is considered the precursor to the majority of breast cancers [40, 41]. With standard use of mammography, the frequency of DCIS diagnosis has increased over 20-fold in the last 20 years [39]. The control rates for DCIS are very good and women with a DCIS diagnosis are generally treated with minimally invasive surgery (that is, lumpectomy) coupled with adjuvant radiation therapy [42, 43]. However, it is apparent that most DCIS cases do not require radiation, and in fact most women are overtreated [40]. In a review of large clinical trials on the treatment of DCIS, the recurrence rate is approximately 30% with surgery alone but approximately 15% with the inclusion of radiation. This means radiation induces a significant clinical benefit. However, ~70% of the women who were treated with radiation would have not had their cancer return; they were therefore overtreated. In contrast, there are ~15% of women for whom an even more effective treatment is needed. For these reasons there has been a lot of interest in understanding determinants of recurrence and progression to invasive disease in DCIS.

Early functional studies from Tlsty’s group and others suggested that the CDK4/6 inhibitor p16ink4a could be a particularly important factor in suppressing the progression of DCIS [44‐46]. Such a model is consistent with the finding that high levels of p16ink4a represent a significant barrier to oncogenic conversion. For example, high levels of p16ink4a in benign Nevi are believed to contribute to potent suppression of melanoma [18]. Paradoxically, high levels of p16ink4a, particularly in conjunction with a high proliferation index, were associated with disease recurrence and progression [47]. Such a combination of markers (high p16ink4a and high proliferation) is indicative of the loss of RB. This is supported by a multitude of studies showing that p16ink4a levels are very high in tumors that have lost RB by mutation or through the action of viral oncoproteins [48]. Furthermore, only through the loss of RB can the cytostatic effect of p16ink4a be bypassed [17]. Subsequent work validated the primary findings in independent cohorts [49, 50]. Importantly, subsequent direct analysis of RB loss in DCIS by optimized immunohistochemistry revealed that RB loss is one of the strongest markers of DCIS recurrence and progression that has been identified and does occur in tumors that express high levels of p16ink4a [51] (Figure 3). The prognostic significance of RB-pathway deregulation is significant in multivariate models, and is true both as a single marker and in combination with other determinants of DCIS biology, including Her2 levels, Cox2 levels, and PTEN levels [49‐52].

Defining the mechanisms underlying the progression of DCIS has been the subject of recent intense study. Functionally, the transition between DCIS and invasive breast cancer represents invasion through ductal myoepithelium and basement membrane into the surrounding tissue. Molecular analysis comparing DCIS with invasive breast cancer demonstrated that one of the key differences between these disease states is the presence of epithelial–mesenchymal transition (EMT) in invasive cancer [53, 54]. This finding emerged from independent groups using unbiased gene expression profiling on microdissected tissues. Interestingly, several groups have demonstrated that, in addition to its canonical effects on proliferation, RB loss can lead to EMT or a partial EMT [52, 55, 56]. Particularly in a variety of breast cancer models, knockdown of RB led to altered morphology and the expression of specific markers of EMT (for example, vimentin) [56]. These outcomes were ostensibly driven through the induction of the activity of EMT-mediators Zeb1 and Slug. Similarly, in mouse models it was observed that tumors which arise with Rb1 deletion are particularly characterized by aspects of EMT [57]. Lastly, in models of DCIS progression using three-dimensional cultures of MCF10A cells, RB deficiency drove an altered gene expression program indicative of loss of epithelial characteristics [52].

The results from these multiple lines of investigation support a critical role for RB in DCIS progression. Most probably, RB loss represents a particularly potent hit for DCIS as it both relieves control over the cell cycle allowing for expansion of DCIS cells and promotes EMT-like processes that promote invasion. These effects of RB loss clearly occur in the context of other oncogenic events (for example, loss of PTEN or Her2) that cooperate in the genesis of DCIS and its ultimate progression to invasive disease.

These provocative findings support the concept that RB pathway status could be utilized to direct the care of patients with DCIS. For example, cases of DCIS that are p16ink4a high and RB deficient would be expected to have the most benefit from adjuvant radiation therapy. In contrast, tumors that are RB proficient would have a low risk of recurrence/progression and therefore may not require radiation treatment. Unfortunately, in spite of the data published thus far, no analysis of clinical specimens has addressed whether the RB pathway status can be used specifically to direct the care of DCIS. Additional investigation is therefore required to bring the RB status to clinical application in DCIS.

Constituting ~70% of diagnosed breast cancer, ER-positive breast cancer is the most prevalent form of breast cancer and is treated based on the presence of the ER. As mentioned above, ER-positive breast cancer is dominated by deregulation of CDK4/6 that is driven by amplication or overexpression of cyclin D₁[36, 60] (Figure 4). This event has been hypothesized to contribute to more aggressive forms of ER-positive breast cancer, although this hypothesis remains controversial.

Her2-positive breast cancer constitutes ~20% of diagnosed breast cancers. In general, Her2-positive disease is more aggressive than ER-positive disease, although the treatment paradigm has many parallels with ER-positive breast cancer. For example, Her2-positive disease is treated based on an addiction to Her2 signaling, and a plethora of active Her2-targeted agents are now employed in the clinic [75]. Like ER-positive disease, the principle clinical issue remains recurrence or intrinsic resistance to the front-line therapy. However, other than Her2 itself, there are no markers to direct treatment decisions, and there is no clear indication that the status of the RB pathway impinges on the sensitivity to Her2-targeted therapy. In spite of the relatively limited analysis of the RB pathway in clinical specimens, there are compelling reasons to believe that the pathway will be relevant as a therapeutic target in Her2-positive breast cancer.

Genetically engineered mouse models provide an important means to evaluate genetic dependencies in tumor development. Sicinski’s laboratory investigated the relationship of cyclin D₁ genetic loss with the development of mammary tumors in mice whose tumors were driven by multiple different genetic stresses [76]. While Myc or Wnt oncogenes can effectively drive tumor development irrespective of cyclin D₁, genetic deletion of cyclin D₁ specifically prevented tumor development driven by Her2 [77]. More recently, a conditional knockout approach showed that cyclin D₁ does represent a therapeutic target in Her2-positive breast cancer [78]. Parallel investigation has shown that Her2-positive models are sensitive to CDK4/6 inhibitors in preclinical models; interestingly, in such models CDK4/6 inhibition not only prevents proliferation but also limits the invasive potential of such models [52]. Therefore, in addition to preventing tumor growth, such an arrest compromises invasive potential. Presumably, as with ER-positive breast cancer, the efficacy of CDK4/6 inhibition in this context is due to its position downstream of the canonical Her2 signaling. Clinical trials are now being initiated to specifically address the ability of CDK4/6 inhibitors to add to the control of Her2-positive cancers.

TNBC is widely accepted to represent the biggest clinical challenge in breast cancer [78, 79]. TNBC is highly aggressive and almost certainly represents several different diseases that have different sensitivities to therapies. Currently, all patients with a TNBC diagnosis are treated with cytotoxic chemotherapy. Such therapy can be incredibly effective for a subset of patients (~30% ), while many patients do not effectively respond and will undergo rapid recurrence [80]. Unfortunately, there are minimal therapeutic interventions for such patients; thus, while TNBC constitutes a relatively minor fraction of breast cancer cases (~20% ), almost 50% of cancer-associated deaths are associated with TNBC. Additional targeted approaches are therefore urgently needed.

Unlike Her2-positive or ER-positive breast cancers that have relatively limited loss of RB, TNBC exhibits frequent loss of RB as determined by histological analysis [38, 39]. Such a finding is also supported by the high levels of p16ink4a observed in many TNBC cases [81]. Lastly, TNBC has very high levels of RB/E2F signature genes relative to other tumor subtypes [62, 63]. Representative images of RB-positive and RB-deficient TNBC are shown in Figure 7. Gene expression analysis and immunohistochemical approaches have shown that tumors that lack RB have a good response to chemotherapy, as indicated by a pathological complete response in neoadjuvant studies or improved overall outcome [62, 63, 82]. This finding is counterintuitive because it suggests that the most aggressive rapidly growing tumors in fact have the best prognosis. The prevailing view of this paradox is that such rapidly proliferating tumors lack critical RB-mediated cell cycle checkpoints and are thus very sensitive to chemotherapy [61]. Such a concept is supported by preclinical data from multiple laboratories showing that RB loss is associated with sensitivity to chemotherapeutic agents. Recent drug screening efforts indicate a complex combination of responses that is conditioned by the therapeutic intervention employed [83]. If more extensively validated, these data would suggest that RB loss could be specifically used to define patients whose tumors would be most apt to respond to optimized chemotherapy regimens.

These findings from RB pathway-based analysis are largely consistent with recent findings from Pietenpol’s group that TNBC represents several intrinsic subtypes [84]. Those subtypes with the highest expression of RB/E2F-regulated genes are generally more sensitive to chemotherapy. In contrast, other subtypes can be treated with other agents. For example, the luminal AR subtype can be treated with androgen receptor antagonists [84]. What remains unclear is whether the analysis of RB/E2F does in fact define subtypes that have improved response to chemotherapy (for example, basal-like cancers) or whether it provides additional information to the subtypes that have been elucidated.

Because RB is lost in ~40% of TNBC, one would expect that it could be used to define unique sensitivities upon which to direct treatment. Anecdotal clinical data suggest that CDK4/6 inhibition may not be particularly effective in TNBC. There are multiple possible explanations for this, including the fact that TNBC is inherently heterogeneous, and most probably single targeted agents will never have much impact on advanced disease that has failed prior chemotherapy. Approaches to define drugs that interact positively with CDK4/6 inhibitors are therefore incredibly important. Unlike endocrine therapy or Her2-targeted therapies that interact positively with CDK4/6 inhibitors, multiple laboratories have shown antagonism with chemotherapy [85, 86]. As such, either metronomic approaches will be needed or it will be critical to define targeted agents that have positive interactions with CDK4/6 inhibitors in the treatment of TNBC.

While RB-deficient tumors do respond better to conventional chemotherapy, the unfortunate reality is that many RB-deficient tumors fail to respond or undergo recurrence. These findings have lead to the pursuit of drugs that exploit the vulnerabilities associated with loss of RB. Several groups have explicitly screened for genes/drugs that kill RB-deficient tumor cells [83, 87]. Such agents could presumably be effective for RB-deficient TNBC that fail front-line chemotherapy.

The preceding volumes of data provide a compelling basis to actively target the RB pathway in breast cancer. However, several key questions in particular emerge in realizing the capability of utilizing this information clinically.

In total, the hope is that the importance of the RB pathway will be utilized to inform treatment in concert with a full understanding of breast cancer biology to improve disease outcomes.