Rational targeting of Notch signaling in breast cancer

Recent evidence indicates that Notch signaling plays an important oncogenic role in breast cancer. Furthermore, a number of reports show that Notch regulates the fate of breast cancer stem cells as well as tumor angiogenesis. Consequently, there is increasing interest in targeting Notch signaling therapeutically in breast cancer patients. Notch inhibitors, particularly γ-secretase inhibitors (GSIs), are in early clinical development. However, rational targeting of Notch signaling in breast cancer will require a systematic exploration of several areas that remain incompletely understood. A clear understanding of cross-talk between Notch signaling and other pathways that play important roles in breast cancer is essential to determine which agents may be most effectively combined with Notch inhibitors. To determine which breast cancer subsets and, ultimately, which patients will benefit the most from Notch inhibitors, we will need to assess the role of Notch in different breast cancers. This, in turn, will require accurate molecular tests that measure Notch pathway activity in clinical specimens. Finally, therapeutic regimens will have to be optimized to reduce or eliminate mechanism-based toxicities. The payoff for such efforts is likely to be a novel, highly promising class of antineoplastic agents for the rational, individualized treatment of breast cancer.

What is the evidence for a role of Notch in breast cancer?

The basic features of Notch signaling and its multiple roles in oncogenesis and tumor suppression have been discussed in several recent reviews [1–8]. Briefly, Notch signaling is a short-range intercellular communication system in which transmembrane ligands bind and activate transmembrane receptors in contiguous cells. This triggers the proteolytic cleavage of Notch receptors to generate intracellular fragments (Notch^IC) that are potent transcriptional coactivators and regulate hundreds of target genes through the ubiquitous transcription factor CSL [6,7,9]. Humans and rodents have four homologous Notch receptors (Notch-1–4) and five ligands (Delta-1, Delta-3, Delta-4, Jagged-1 and Jagged-2). Notch receptors are integral type I membrane, noncovalent heterodimers consisting of an inhibitory extracellular subunit (Notch^EC) bound to a transmembrane subunit (Notch™). Ligand binding to a Notch receptor induces the separation of Notch^EC from Notch™. Notch^EC is transendocytosed with the ligand into ligand-expressing cells. Subunit separation exposes an extracellular cleavage site in Notch™ that is cleaved by disintegrin metalloproteases ADAM17 or ADAM10. The ADAM-cleaved receptor is endocytosed and cleaved by multisubunit aspartyl protease γ-secretase to generate Notch^IC. The latter migrates to the nucleus, binds to CSL and regulates transcription at CSL-responsive elements. In human tumors, Notch signaling can transmit bidirectional signals among cancer cells that express both ligands and receptors, or mediate bidirectional tumor–stroma interations and tumor–endothelium interactions [6,7,9].

Among solid tumors, breast cancer may be the disease in which a pathogenetic role of Notch is supported by the strongest evidence. Notch signaling is a key cell fate regulator during normal mammary gland development [10]. Constitutively active forms (Notch^IC) of Notch-1, [11–14], Notch-3 [14] and Notch-4 [15,16] cause mammary tumors in transgenic mice. Similarly, constitutively active Notch-1 [17] and Notch-4 [18] transform human mammary epithelial cells in vitro. Interestingly, Notch-2 appears to antagonize signals by the other three homologs in breast cancer cells [19]. In breast cancer clinical specimens, mRNA expression of Notch-1 and Notch ligand Jagged-1 have been shown to correlate strongly with poor prognosis [20,21]. Loss of Notch-negative regulator Numb, a protein that triggers endocytosis and degradation of Notch receptors, has been described in approximately 50% of human breast cancers [22]. Notch-4 expression, as detected by immunohistochemistry, correlates with Ki67, a well-known proliferation marker in infiltrating breast carcinomas of ductal or lobular histologies [23]. Conversely, and consistent with published in vitro data, expression of Notch-2 appears to have a positive prognostic significance [24].

What makes Notch signaling an attractive therapeutic target in breast cancer?

The fact that a pathway is associated with poor prognosis in a particular cancer type does not necessarily mean that this pathway is a potentially promising therapeutic target. However, several characteristics of Notch signaling make it a particularly interesting potential target in breast cancer. First, the Notch pathway regulates survival and proliferation not only in ‘bulk’ breast cancer cells [17,23] but also in breast cancer stem cells [25–28]. At the same time, Notch signals play a distinct pro-angiogenic role in tumor endothelial cells, which is largely dependent on ligand Delta-4 [29–32]. Thus, in principle, pharmacological inhibition of Notch signaling may have significant therapeutic effects in breast cancer primary lesions, prevent the self-renewal of breast cancer stem cells responsible for tumor recurrence and block tumor angiogenesis, thus preventing or ameliorating metastatic disease. At the molecular level, a feature that Notch signaling shares with few other evolutionarily ancient ‘elite’ pathways such as Hedgehog and Wnt, is the great variety of its effects on multiple proliferation, survival and differentiation pathways [6,8]. While virtually all pathways are engaged in cross-talk interactions, ancient developmental pathways such as Notch are especially pleiotropic in their effects. Thus, inhibiting Notch signaling is likely to simultaneously affect numerous secondary therapeutic targets, achieving what amounts to ‘multitargeted’ therapy with a single agent. This does not necessarily mean that monotherapy with Notch inhibitors should be viewed as an achievable goal in the clinic. However, it does mean that these agents have the potential to synergize with multiple classes of drugs, thereby potentially maximizing efficacy and decreasing the likelihood of resistance. Seeking the safest and most effective among combinations including Notch inhibitors is a very promising strategy.

What pharmacological tools do we have at our disposal?

At the time of this writing, GSIs are in early clinical development for breast cancer. These drugs inhibit the final proteolytic cleavage of Notch receptors that generates Notch^IC. Monoclonal antibodies (mAb) that prevent Notch activation by binding the extracellular ‘negative regulatory region’ are in preclinical development [33]. These mAbs prevent ligand-induced subunit separation, essentially ‘locking’ the receptors in their heterodimeric inactive conformation. Furthermore, mAbs that target ligand Delta-4 [30], preventing it from binding Notch receptors, are being developed as antiangiogenic agents [29,31,32]. Numerous other ways of modulating Notch signaling are possible in principle, and have been recently reviewed [8].

In preclinical studies, GSIs are active in estrogen receptor (ER)α-negative breast cancer xenografts and, in combination with endocrine therapy, in ERα-positive xenografts [23]. GSIs block the activation of all four Notch homologs. This may be an advantage in indications like breast cancer, where at least three Notch homologs have pro-oncogenic effects. Off-target effects are a potential concern with GSIs, since γ-secretase has numerous substrates other than Notch receptors and a rather promiscuous cleavage specificity [34]. However, off-target effects are not necessarily an obstacle to clinical development, unless they are shown to reduce the safety or efficacy of these agents. The recent observation that some NSAIDs and structurally related compounds can allosterically modulate the substrate specificity of γ-secretase [35] raises the prospect that more Notch-selective γ-secretase modifiers could be developed in the near future.

Under what conditions would Notch-inhibitory mAbs be preferable to GSIs? One possible situation may be the case in which a therapeutically relevant mAb target has a more restricted expression pattern compared with other Notch pathway components. Such a target could be affected selectively, thereby potentially reducing mechanism-based toxicities. In the case of breast cancer, Notch-4 may be a potentially attractive target for selective mAbs. Notch-4 is expressed in breast cancers, and its knockdown inhibits the proliferation and survival of some breast cancer cell lines as effectively as Notch-1 knockdown or GSI treatment [23]. Notch-4 expression in normal tissues is more restricted than Notch-1 expression. Published data suggest that it is limited to the vascular endothelium [36,37], epidermis [38], and ovarian blood vessels during folliculogenesis [39].

In addition, a more general potential advantage of mAbs is the possibility of conjugating them with radionuclides or toxins to selectively target cells that overexpress their targets. Conversely, a potential disadvantage of mAbs as Notch-targeting agents is their generally long biological half-life. If intermittent inhibition of Notch signaling is desirable to minimize adverse effects, using a mAb that will remain in circulation for days or weeks may be a disadvantage. F(ab)2s, F(ab)s or single chain Fvs may be useful to circumvent this problem.

For therapeutic purposes, some features of Notch signaling are of particular interest. First, the effects of Notch activation are notoriously dose-dependent [7]. In other words, different amounts of Notch have different effects, implying that complete blockade of Notch signaling in target cells may not be necessary to achieve a therapeutic effect. Second, the duration of Notch signaling events at the cellular level is short, consisting essentially of a series of brief ‘pulses’ of gene regulation that are extinguished by Notch degradation. This suggests that intermittent pharmacological inhibition may be effective in vivo. In fact, intermittent administration regimens for GSIs reduce toxicity in animals and humans without compromising activity. Third, the effects of Notch are remarkably context-dependent. Different Notch homologs have different effects in different cell types. Even the same Notch receptor can have different target genes and different biological effects in different cell types and under different conditions in the same cell type. This is largely due to bidirectional cross-talk with other pathways, which modulates the intensity, duration and effects of Notch signals. Given the biological and genetic heterogeneity of breast cancer, this implies that different disease subtypes need to be considered separately in order to design the most promising rational combinations including Notch inhibitors.

Rationally designed combination regimens

Our group and others are systematically exploring the cross-talk of Notch with other pathways relevant to breast cancer, in order to develop rationally designed therapeutic combinations for different disease subtypes. Rizzo et al. have recently demonstrated that in ERα-positive breast cancer cells estrogen causes accumulation of inactive Notch-1 and inhibits Notch signaling [23]. Conversely, estrogen deprivation, such as might be achieved in humans by administration of aromatase inhibitors, as well as selective estrogen-receptor modulators like tamoxifen, cause ERα-positive breast cancer cells to re-activate Notch signaling and become more dependent on Notch for proliferation and survival. A combination regimen including tamoxifen and a GSI was more effective than either agent alone in T47D xenografts. Based on these observations, a pilot clinical trial of aromatase inhibitors or tamoxifen in combination with GSI has been designed and will be opened soon.

Osipo et al. have recently reported that Her2/neu overexpression inhibits Notch signaling, possibly by altering ligand availability at the cell membrane [40]. Treatment of Her2/Neu overexpressing breast cancer cells with trastuzumab or with a dual EGF receptor-Her2/neu tyrosine kinase inhibitor caused re-activation of Notch signaling, and increased dependence on Notch for proliferation and survival. Trastuzumab-GSI and tyrosine kinase inhibitor-GSI combinations were at least additive in vitro and are currently being tested in vivo. Importantly, when Her2/neu-overexpressing BT474 cells were rendered trastuzumab-resistant after 6 months of continued culture in the presence of trastuzumab, these cells became exquisitely sensitive to GSI, and GSI reversed their trastuzumab resistance. Consistent with these data, recent observations by Yamaguchi et al. indicate that in the absence of Her2/neu-inhibitory treatment, Notch-3 plays an important role in Her2/neu-nonoverexpressing, but not in Her2/neu-overexpressing cells [41].

From these observations, it would appear that both estrogen deprivation in ERα-positive breast cancers and Her2/neu inhibition in Her2/neu-overexpressing cancers cause cancer cells to ‘fall back’ onto a more primitive pathway, Notch, and become functionally addicted to it. Hence, combination regimens including endocrine therapy plus Notch inhibitors or Her2/neu-targeted drugs plus Notch inhibitors may deserve further investigation. The possible role of Notch signaling in tamoxifen-resistant breast cancer models is currently being studied.

A corollary of these observations is that ‘triple-negative’ (ERα, progesterone receptor-negative, Her2/neu-nonoverexpressing) breast cancers may be particularly dependent on Notch signaling. While this hypothesis has not been rigorously tested yet, MDA-MB231 (triple-negative) cells require Notch-1 or Notch-4 for proliferation, and MDA-MB231 xenografts are highly sensitive to single-agent GSI treatment [23]. Given the fact that triple-negative tumors are particularly difficult to treat, this subset of tumors may be a promising indication for Notch inhibitors. Rational combinations have not been thoroughly explored in this setting, but it’s possible to make a few educated guesses. Inhibitors of the PI3-kinase-AKT-mTOR pathway may be useful in combination with Notch inhibitors. This pathway is frequently overactive in breast cancer due to PI3K mutations, loss of PTEN expression or activity or activation of growth factor receptors, and there is evidence that this combination is effective in GSI-resistant T-cell acute lymphoblastic leukemia cells that carry PTEN-inactivating mutations [42]. The complex cross-talk between Notch and NF-κB suggests that NF-κB inhibitors and Notch inhibitors could be successfully combined [43]. Given the importance of NF-κB in breast cancer, particularly endocrine therapy resistant cases [44,45], such combinations deserve further investigation. Notch^IC is degraded by the proteasome and accumulates in cells treated with proteasome inhibitors. Proteasome inhibitors are thought to be potentially useful in breast cancer [46], and GSI-proteasome inhibitor combinations may warrant experimental scrutiny. Finally, Delta-4 mAb appear to be effective as antiangiogenic agents independently of VEGF [29,30]. Thus, these agents may be useful in combination with bevacizumab or other VEGF inhibitors.

What adverse effects can we expect?

The most common adverse event in patients treated with GSIs is secretory diarrhea caused by goblet cell metaplasia of the small intestine. This effect is mechanism-based and is observed in mouse models as well [47]. Intermittent oral administration of GSIs significantly reduces intestinal toxicity, and in our hands parenteral administration of GSIs in mouse xenograft models at doses that caused significant antineoplastic effects did not cause diarrhea or weight loss [48,23]. Importantly, myelotoxicity is not observed in preclinical or clinical studies. In mice, other adverse effects of systemic GSI treatment include reversible suppression of lymphopoiesis [47], reversible hair depigmentation and immunosuppressive effects that may be undesirable under some circumstances but may have therapeutic applications of their own [49]. Whether prolonged Notch inhibition by agents that pass the blood–brain barrier can cause neurological toxicity is currently unknown, but should be considered based on mouse data [50]. In summary, current experience indicates that systemic inhibition of Notch signaling is reasonably well tolerated for a period of weeks, particularly if intermittent administration regimens are used. Not unlike other antineoplastic agents, it is possible that multiple cycles of administration separated by breaks from drugs will prove to be the most practical way of using Notch inhibitors, at least until more selective agents or better drug delivery systems are developed.

Five-year view

Our knowledge of Notch signaling in breast cancer is still incomplete, but evidence accumulated so far suggests that rationally designed combination regimens including Notch inhibitors may be a novel, highly promising strategy to treat breast cancer. If inhibition of breast cancer stem cell self-renewal is achieved in the clinic, the benefits may be fully appreciated only after long-term follow-up studies are conducted, with disease-free survival as an end point. Which regimens will be most effective will likely depend on the breast cancer subtypes. In order to identify the groups of patients and/or subtypes of breast cancer that will benefit the most from Notch inhibitors, it will be necessary to develop accurate molecular tests that measure the level of Notch pathway activity in clinical specimens such as core biopsies. These tests are likely to require parallel quantitative reverse transcription (Q-RT)-PCR measurements of expression levels of multiple Notch target genes, with strategies similar to the Oncotype^DX test. These target genes are likely to be only partly overlapping in different breast cancer subsets. The ultimate goal will be the individualized design of therapeutic regimens to maximize safety and efficacy, while minimizing the likelihood of resistance and disease recurrence. Current molecular tools put this goal within our reach.

Acknowledgements

We are grateful to Barbara Osborne, Todd Golde, Kathy Albain, Paola Rizzo, Clodia Osipo, Kimberly Foreman and Antonio Pannuti for helpful discussions.

Financial & competing interests disclosure

Our work was supported by National Institutes of Health grant P01 AG2553101 and DOD IDEA grant W81XWH-04–1-0478. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.