Biological Determinants of Endocrine Resistance in Breast Cancer

REVIEWS

TherapeuTic resisTance Biological determinants of endocrine resistance in breast cancer

Elizabeth A. Musgrove*‡ and Robert L. Sutherland*‡ Abstract | Endocrine therapies targeting oestrogen action (anti-oestrogens, such as tamoxifen, and aromatase inhibitors) decrease mortality from breast cancer, but their efficacy is limited by intrinsic and acquired therapeutic resistance. Candidate molecular biomarkers and gene expression signatures of tamoxifen response emphasize the importance of deregulation of proliferation and survival signalling in endocrine resistance. However, definition of the specific genetic lesions and molecular processes that determine clinical endocrine resistance is incomplete. The development of large-scale computational and genetic approaches offers the promise of identifying the mediators of endocrine resistance that may be exploited as potential therapeutic targets and biomarkers of response in the clinic.

Aromatase inhibitors The steroid hormone oestrogen is central to normal One-third of women treated with tamoxifen for Drugs that function by blocking female physiology, reproduction and behaviour, through 5 years will have recurrent disease within 15 years9, aromatase, the enzyme that its effects on cellular processes including cell proliferation and so endocrine-resistant disease may represent up to converts androgens to oestrogens in tissues including and cell survival. These effects are mediated by nuclear one-quarter of all breast cancers. Therefore, two major the breast and adipose tissue. oestrogen receptors (ERα and ERβ; BOX 1). ERα is respon- challenges for the successful treatment of breast can- Examples include anastrazole, sible for many of the effects of oestrogen on normal and cer are the development of more specific biomarkers letrozole and exemestane. cancerous breast tissue, through ligand-activated trans- that predict therapeutic response to endocrine therapy criptional regulation (genomic actions) and by acting as and the identification of new therapeutic targets for ER-positive breast cancers In current clinical practice, a component of membrane and cytoplasmic signalling endocrine-resistant disease. This Review summarizes 1 ER‑positive breast cancers are cascades (non-genomic actions) (FIG. 1). and evaluates the recent insights into the mechanisms those with immunohisto‑ Sustained exposure to endogenous or exogenous of endocrine resistance that have been made through chemically detectable ERα oestrogen is a well-established cause of breast cancer2,3, candidate gene approaches, as well as more global gene levels. underpinning the use of anti-oestrogens and aromatase expression profiling and functional genetic screens. inhibitors in breast cancer prevention4–6. At least 70% of We necessarily focus on tamoxifen resistance, as the breast cancers are classified as ER‑positive breast cancers7, experience with this drug is more extensive and the and interfering with oestrogen action has been a main- clinical data more mature than for other drugs. Many stay of breast cancer treatment for more than a cen- of the broad concepts discussed will probably also tury. Early therapies included surgical removal of the apply to resistance to aromatase inhibitors and other ovaries, but the synthesis of competitive inhibitors of anti-oestrogens, although the lack of clinical cross- oestrogen–ER binding during the 1970s led to the first, resistance10–12 indicates that some resistance mechanisms *Cancer Research Program, and to date most successful, targeted cancer therapy: are independent. Garvan Institute of Medical the selective oestrogen receptor modulator (SERM) Research, Sydney, New South tamoxifen8. Adjuvant therapy with tamoxifen almost Molecular mechanisms of resistance Wales 2010, Australia. intrinsic resistance ‡St Vincent’s Clinical School, halves the rate of disease recurrence and reduces the The primary mechanism of de novo or Faculty of Medicine, annual breast cancer death rate by one-third, mak- to tamoxifen is lack of expression of ERα. Recently, a University of New South ing a significant contribution to the 25–30% decrease second intrinsic mechanism has been documented in Wales, New South Wales in breast cancer mortality in the past two decades9. which patients carrying inactive alleles of cytochrome 2052, Australia. Subsequently, other new, effective endocrine therapies P450 2D6 (CYP2D6) (approximately 8% of Caucasian e‑mails: [email protected]; have been developed that target oestrogen synthesis women) fail to convert tamoxifen to its active meta- 10 [email protected] (such as aromatase inhibitors ) or ER signalling (such bolite, endoxifen, and are consequently less responsive doi:10.1038/nrc2713 as other SERMs and ‘pure’ anti‑oestrogens11). to tamoxifen13. By contrast, a plethora of mechanisms

a t a glance As detailed information on the biology of the many molecules implicated in tamoxifen resistance in vitro • Endocrine therapies that target oestrogen action (anti-oestrogens and aromatase and the means by which they cause resistance is sum- inhibitors) are widely used and successful breast cancer therapies, but many women marized in a series of excellent recent reviews11,12,14–21, treated with these therapies will relapse with endocrine-resistant disease. we focus here on recent developments that shed light • Mechanisms of endocrine resistance in oestrogen receptor (ER)-positive breast on potential mechanisms of prognostic or predictive cancers include loss of ER expression and expression of truncated isoforms of α importance. As shown by the examples in TABLE 1, most ERα and ERβ, post-translational modifications of ERα, increased AP1 activity and deregulation of ER co-activators, increased receptor tyrosine kinase signalling of the molecules that modulate tamoxifen sensitivity in leading to the activation of the Erk and PI3K pathways, and deregulation of the experimental models are correlated with disease out- cell cycle and apoptotic machinery. come not only in women treated with tamoxifen but • Gene expression signatures that are predictive of poor outcome in women treated also in a wider population of patients with breast can- with tamoxifen commonly contain ER target genes, as well as genes involved in cer. Furthermore, because adjuvant tamoxifen has been proliferation, apoptosis, and invasion and metastasis. Many of these signatures are the ‘therapy of choice’ in ER-positive breast cancer for also predictive of outcome in women who have not been treated with tamoxifen and more than 25 years, most available patient cohorts do so are markers of intrinsic biology rather than specific to tamoxifen responsiveness. not allow the comparison of therapeutic responsiveness • Gene expression signatures representing particular biological processes (for and outcome in well-matched control populations that example, cell cycle progression, cell death and invasion) or pathways (for example, differ only with respect to tamoxifen therapy. Therefore, RB deregulation, MYC overexpression and E2f activation) can also predict outcome it is difficult to distinguish specific differences in tumour in women treated with tamoxifen and point towards possible mechanisms for reponse to tamoxifen from the broader effects of the endocrine resistance. underlying biology of the disease. • Functional genetic screens have successfully identified several genes, the loss or overexpression of which can reduce anti-oestrogen sensitivity in cell lines and is ER and co-regulators. Response to tamoxifen is rare associated with clinical endocrine resistance. in ER-negative breast cancer, and so ERα expression is • Insights into the mechanisms of resistance have suggested possible therapeutic currently the principal biomarker of response to endo- approaches for endocrine-resistant ER-positive breast cancer, for example tyrosine crine therapy. Early studies implicated the loss of ERα kinase inhibitors. Further potential therapeutic targets may emerge from combining expression or ERα mutations as potential mechanisms large-scale genomic and transcriptomic data with large-scale functional analyses. of acquired resistance. However, loss of ERα expression occurs in only a minority (15–20%) of resistant breast cancers22 and <1% of ER-positive tumours have have been postulated to account for acquired resistance ERα mutations14,20,23. More recently, expression of a following prolonged exposure to tamoxifen, some of new truncated variant of ERα, ERα36, in the presence which may also account for intrinsic resistance in the of full-length ERα has been associated with reduced clinic. Much of the published information on these responsiveness24. The development of antibodies that potential molecular mechanisms has been derived can distinguish between ERα, ERβ and naturally occur- from ERα-positive breast cancer cell lines and from ring ERβ variants (BOX 1) has led to the identification of Adjuvant therapy A drug treatment (for example, variants of these cell lines selected for adaptation to responses in ERβ-positive but ERα-negative cancers and chemotherapy or endocrine sustained exposure to anti-oestrogens or withdrawal of a potential role for the carboxy-terminally truncated var- therapy) that is given after the oestrogen. Such models identify mechanisms that can iants of ERβ (ERβ2/cx and ERβ5) in tamoxifen respon- primary therapy (for example, induce tamoxifen resistance in vitro rather than those siveness25,26. in addition, the oestrogen-related receptor surgery and/or radiotherapy), that actually mediate resistance in patients with breast ERRγ is overexpressed and mediates tamoxifen resistance with the aim of increasing the 27 overall effectiveness of cancer, and they have several other potential limita- in lobular invasive breast cancer models . treatment. tions. These include the degree to which the few ERα- One mechanism by which ERα regulates gene expres- positive breast cancer cell lines studied reflect the range sion is through protein–protein interactions with other SERMs of ER-positive phenotypes in situ and the absence of transcription factors — for example, activator protein 1 Drugs such as tamoxifen that bind the oestrogen receptor epithelial–stromal and tumour–host interactions that (Ap1), specificity protein 1 (Sp1) and nuclear factor-κB and thereby block the effects probably modulate sensitivity in vivo. Furthermore, the (nF-κB) (FIG. 1). increased Ap1 and nF-κB transcriptional of oestrogen on tissues such as mechanisms that are responsible for the clinical obser- activity are also associated with endocrine resistance28–30. the breast but that function vation that tamoxifen-resistant cancers often respond to ERα function is regulated by post-translational modifica- similarly to oestrogen in other second-line endocrine therapies10–12 remain unclear. tions (phosphorylation, methylation and sumoylation) tissues such as bone. Unlike oestrogen, these drugs are not notwithstanding these potential limitations, stud- that influence interactions with other proteins, including 19 steroidal in structure. ies of candidate genes involved in oestrogen signalling transcriptional co-regulators and cytoplasmic signalling (FIG. 1) or anti-oestrogen regulation of cell proliferation molecules (FIG. 1). There is significant evidence to show ‘Pure’ anti-oestrogens and survival (FIG. 2) and more global unbiased approaches that effects on these end points contribute to endocrine Drugs that bind the oestrogen 18,20,31 receptor, thereby blocking the using cell line models have yielded important concepts resistance . Overexpression and increased phosphor- effect of oestrogen, but have and hypotheses that correlate with tamoxifen resist- ylation of ERα co-activators, particularly nuclear receptor no detectable oestrogen‑like ance in the clinic, and they have provided a basis for co-activator 3 (nCOA3; also known as AiB1 or SRC3), effects. Most have a steroidal new therapeutic approaches. Deregulation of various leads to constitutive ERα-mediated transcription, which structure. aspects of oestrogen signalling is a common mechanism confers resistance in vitro and in xenograft models12,16 and Intrinsic resistance for resistance, but unrelated mechanisms that provide is associated with reduced responsiveness to tamoxifen 32 The failure to respond to initial tumour cells with alternative proliferative and survival in patients . Transient methylation of ERα at R260 by drug therapy. stimuli also confer resistance. protein arginine N-methyltransferase 1 (pRMT1) results

Cytochrome P450 2D6 in the formation of cytoplasmic complexes that contain with the loss of PTEN expression or overexpression of (CYP2D6) ERα, pi3K, the tyrosine kinase SRC and focal adhesion ERBB2 (REFS 20,21) and activation of iGF1R and ERBB3 A member of the large and kinase (FAK; also known as pTK2) and that activate Akt following the loss of PTEN40. How these events mediate diverse superfamily of (FIG. 1). However, it is not known whether this methyla- tamoxifen resistance has not been fully elucidated, but cytochrome P450 enzymes. 33 CYP2D6 catalyses the tion event, which is frequent in breast cancer , is associ- several potential contributing factors have been sug- conversion of tamoxifen into its ated with the endocrine response. Another example is gested (FIGS 1,3): decreased ERα expression mediated by active metabolites, endoxifen the ER co-activator pElp1, which in many breast cancers ERK activation; loss of ER-mediated repression of EGFR and 4‑hydroxytamoxifen. It is localizes to the cytoplasm, where it can confer tamoxifen and ERBB2 and consequent activation of mitogenic sig- highly polymorphic, so its resistance31. pElp1 functions as a scaffold that modu- nalling cascades; ligand-independent activation of ER or activity is variable between individuals. lates ER interaction with SRC, leading to activation of its co-activators through phosphorylation; upregulation SRC and the Erk family kinases and also promotes of key cell cycle regulators, for example MYC and the Acquired resistance oestrogen activation of pi3K31 (FIG. 1). D-type and E-type cyclins, through constitutive activa- In contrast to intrinsic tion of mitogenic signalling pathways; and the inhibition resistance, an initial response to drug therapy followed by Receptor tyrosine kinase signalling. The bidirectional of apoptosis through constitutive activation of survival subsequent disease crosstalk between ER and receptor tyrosine kinase sig- signalling. progression. nalling is evidenced by the early observations of recip- Overexpression of ERBB2 is one of the best- rocal expression of ER and members of the epidermal characterized mechanisms of endocrine resistance21. growth factor receptor (Egfr) family such as EGFR and Recent evidence implicates the loss of transcriptional ERBB2 (also known as HER2)34. Growth factors of the repressors and amplification of ERBB2 as mechanisms Egf and insulin-like growth factor (igf) families can that are responsible for increased expression of this recep- modulate tamoxifen sensitivity in vitro35; although breast tor. The X-linked tumour suppressor forkhead box p3 cancer cells are quiescent and insensitive to growth fac- (FOXp3) and the zinc finger transcription factor GATA4 tor stimulation following treatment with the pure anti- can repress ERBB2 expression, even in a cell line with an oestrogen iCi 182780 (fulvestrant), tamoxifen treatment approximately tenfold amplification of ERBB2, and their does not lead to growth factor insensitivity36. This has expression is negatively correlated with ERBB2 expres- focused attention on receptor tyrosine kinase expression sion in breast cancer44,45. in addition, a recent pivotal and function as potential mediators of endocrine resist- study showed that ERα-mediated repression of ERBB2 is ance. increased expression of EGFR, ERBB2 and iGF1 dependent on competition between the paired-domain receptor (iGF1R) can elicit tamoxifen resistance37–40, as transcription factor pAX2 and the ERα co-activator can activation of the components of their downstream nCOA3 for binding and regulation of ERBB2 transcrip- signalling pathways, particularly the Erk and pi3K path- tion and, in turn, tamoxifen responsiveness46. A direct ways41–43. in some cases, deregulation of these signalling relationship between FOXp3 or GATA4 expression pathways occurs as a result of genetic or epigenetic and tamoxifen responsiveness has not been established. modifications, such as amplification of ERBB2, activat- However, increased pAX2 expression and consequent ing mutations in PIK3CA, which encodes a catalytic repression of ERBB2 was associated with increased sur- subunit of type i pi3Ks, and loss of heterozygosity or vival following tamoxifen treatment, and loss of pAX2 methylation of PTEN, a tumour suppressor that inhibits expression in the presence of increased nCOA3 expres- the pi3K pathway20,21. in other cases, however, deregula- sion predicted a poor outcome46, indicating that this tion of these pathways reflects aberrations in upstream mechanism is of direct clinical relevance. regulators, such as the activation of Akt in association Members of the Src family of tyrosine kinases, particularly SRC itself, and their downstream targets are also commonly overexpressed in breast cancer and have been implicated in resistance. The Src substrate Box 1 | Oestrogen receptors BCAR1 (also known as CAS) is a focal adhesion adap- The oestrogen receptors ERα and ERβ are encoded by separate genes located on tor protein that activates proliferative, survival and different chromosomes. They have a similar overall domain structure, consisting of a invasion pathways. it can induce tamoxifen resist- central DNA-binding domain flanked by two autonomous transcriptional activation ance when overexpressed in vitro47, and BCAR1- domains, one of which, AF-2, is positioned in the ligand-binding domain and is ligand overexpressing breast cancers are less responsive to dependent (reviewed in REF. 11). The AF-2 domain also mediates interactions with tamoxifen48. BCAR1 binds and activates SRC with con- co-activators that increase ER transcriptional activity19,127. Isoforms of both ER α sequent phosphorylation of the Src substrates EGFR and ERβ have been described, but the most relevant in the context of endocrine and signal transducer and activator of transcription 5B resistance are a truncated variant of ERα, ERα36 (REF. 24), and two isoforms of ERβ, ERβ2/cx and ERβ5, in which amino-terminal truncations remove some of the (STAT5B) and effects on downstream signalling path- ligand-binding domain26. ways20. However, recent data suggest that the ability of In the breast, ERα-positive epithelial cells located in the ductal lumen facilitate BCAR1 to confer anti-oestrogen resistance may not the development of a branched ductal tree, which functions as a scaffold for require interaction with SRC49. The putative guanine milk-producing alveoli. Not all luminal cells express ERα, but ERα-null mice develop nucleotide-exchange factor BCAR3, which synergizes only a rudimentary mammary ductal tree and are unable to lactate128, indicating that with BCAR1 to activate SRC50, also causes tamoxifen the ERα-positive cells make an essential contribution to mammary development. resistance in vitro51. in addition, BCAR3 activates 128 By contrast, the mammary glands of ERβ-null mice develop normally . When the two Rac and p21-activated kinase 1 (pAK1)52; the latter is receptors are co-expressed in breast cancer cell lines, ER functions as an antagonist β itself implicated in tamoxifen resistance through ERα of ERα, impairing the ability of oestrogen to stimulate proliferation129. phosphorylation53.

Neoadjuvant Cell cycle regulators. Data from experimental model anti-oestrogen targets confers resistance in vitro and is A drug treatment that is given systems, supported by clinical correlations, indicate associated with reduced tamoxifen responsiveness in weeks to months before that anti-oestrogens are both cytostatic and cytotoxic. patients. Overexpression of MYC, cyclin E1, cyclin D1 surgery, often to reduce the Neoadjuvant endocrine therapy leads to decreased pro- or the cyclin D1 splice variant cyclin D1b, or the inac- size of tumours before surgery. liferation54, and in cell culture anti-oestrogen treat- tivation of the RB tumour suppressor — an important Cyclin E1 ment leads to a G1 phase-specific cell cycle arrest and substrate for cyclin-dependent kinases (CDKs) that are Cyclin E1 and cyclin E2 are a consequent reduction in growth rate55. not unexpect- active in G1 phase — and the decreased expression of regulatory subunits of kinase edly, the molecules pivotal to the anti-oestrogen effects the CDK inhibitors p21 or p27, results in decreased anti- complexes that contain CDK2 on cell cycle progression (FIG. 2a) have central roles in oestrogen sensitivity in vitro56–64. MYC overexpression as their catalytic subunit and regulate the G1 to S phase cell the control of G1 phase progression downstream of and consequent tamoxifen resistance is accompanied by cycle transition. polypeptide growth factor mitogens, as well as oestro- transcriptional repression of CDKN1A (which encodes gen. Aberrant expression of several such oestrogen and p21)65, relieving the inhibitory effect of p21 on cyclin E1–CDK2 complexes. Cyclin D1 overexpression leads to an increased abundance of cyclin D1–CDK4 complexes Oestrogen (which indirectly activate cyclin E1–CDK2 (REF. 58) by RTKs: sequestering p21 and p27) and to the activation of cyclin EGFR, Growth ERBB2 E2–CDK2 by increased transcription of CCNE2 (which a factors and IGFR c encodes cyclin E2)66. in addition to its cell cycle regula- Plasma tory role, cyclin D1 interacts with several transcription membrane b ER factors, including ERα and STAT3 (REF. 67). Τamoxifen induces cyclin D1 binding to ERα at the expense of Ras PI3K cyclin D1–STAT3 binding, activating both STAT3 and M ERα — this is an additional mechanism by which cyclin ER deER ER D1 overexpression can affect the tumour’s response to Erk Akt PI3KFAK PELP1 Src Src tamoxifen68. Therefore, MYC and cyclin D1 overexpres- Src sion can potentially affect anti-oestrogen sensitivity at P P several levels. ER Akt Erk P P Supporting the clinical relevance of the aberrant expression of these molecules, there is accumulating evi- Cytoplasm dence that overexpression of MYC or cyclin D1 is asso- 69,70 ER ER TFs TFs ciated with tamoxifen resistance in patients . There is also more limited evidence for a relationship between the Nucleus overexpression of cyclin E1, RB inactivation and reduced CoAHAT CoA ER expression of p27 and clinical response62,69–71. in breast ER ER Ap1 Sp1 TF TF cancer, overexpression of MYC, cyclin D1 and cyclin E1 is at least two to three times more common than amplifi- ERE SRE RE 69 mRNA cation of the corresponding genes , and RB inactivation AAAAA is also more common than RB1 deletion or mutation62. MYC, cyclin D1, The reasons for this probably include the activation of cyclin E1 and cyclin E2 upstream mitogenic signalling pathways and the deregulation of transcriptional regulators, including the E2f fam- Growth, proliferation and survival ily. The gene encoding p27, CDKN1B, is rarely mutated or Figure 1 | Oestrogen action at the molecular level. Three distinct pathways of Nature Reviews | Cancer deleted in breast cancer, but p27 expression is frequently oestrogen regulation of gene expression (reviewed in REFS 1,11,31) are shown. First, in reduced by oncogenic activation of mitogenic signalling classic oestrogen signalling, ligand-bound oestrogen receptor (ER) activates gene (for example, by ERBB2 overexpression or Src activa- expression — either through direct binding of dimeric ER to specific DNA response 71 elements, EREs, in complexes including co-activators (CoAs) and histone acetyl tion) and increased p27 degradation . The microRnAs transferases (HATs), or through protein–protein interactions with other transcription (miRnAs) miR-221 and miR-222 reduce p27 expression factors, particularly members of the activation protein 1 (Ap1) and specificity protein 1 and confer resistance to tamoxifen, although the precise (Sp1) families — to facilitate binding to serum response elements (SREs) and activation of mechanism of resistance is unclear, as these miRnAs also transcription (a). Second, ER can also be activated as a consequence of signalling events reduce ERα expression and are overexpressed in ERBB2- downstream of receptor tyrosine kinases (RTKs) such as the epidermal growth factor overexpressing breast cancers72,73. There are conflicting receptor (EGFR), ERBB2 (also known as HER2) and the insulin-like growth factor receptor data on the relationship between p21 expression and out- (IGFR) (b). Phosphorylation (P) by the Erk or Akt serine/threonine kinases leads to come in breast cancer69, and the role of p21 in tamoxifen ligand-independent activation of the ER. Third, signalling can be mediated through response in breast cancer has not been studied extensively, non-genomic mechanisms by ER that is localized at the cell membrane or in the although the ERBB2 repressor FOXp3 is essential for p21 cytoplasm (c). Ligand binding induces the assembly of functional protein complexes that 74 involve other signalling molecules and that activate intracellular signalling cascades, expression and there is some evidence for p21 deregula- resulting in transcription factor (TF) activation. Two recently characterized mechanisms tion in cancers with ERBB2 overexpression or Akt acti- 70 that ultimately activate transcription independently of ER binding to DNA are illustrated: vation . Akt activation results in the mislocalization of ligand-induced methylation (M) of ER and formation of an ER–PI3K–Src–focal adhesion p21 to the cytoplasm; predominantly cytoplasmic locali- kinase (FAK) complex that activates Akt (d), and activation of Erk by ER–Src–PELP1 zation of p21 has been associated with poor response to complexes (e). tamoxifen75.

17,76 (FIG. 2b) a p21 P messenger ceramide . Crosstalk between the Cyclin p27 P P E2F4 apoptotic effects of anti-oestrogens and the tumour necro- E1 CDK2 sis factor (TnF) pathway, as well as anti-oestrogen effects G1 G0 p27 MYC p21 RB E2F4 on survival signalling through the pi3K–Akt, nF-κB and interferon pathways, is also likely to contribute to p27 p130 Cyclin anti-oestrogen-mediated apoptosis17 (FIG. 2b). Finally, ER AE E1 CDK2 intriguing recent observations indicate that autophagy is p21 Cyclin Cyclin p27 a mechanism of cell survival in breast cancer cells that are D1 D1 resistant to apoptotic concentrations of tamoxifen77. CDK4 P P P it has been difficult to establish the role of apop- P RB P S tosis in the clinical setting. neoadjuvant studies have b BCL-2 yielded conflicting data and have been limited by small patient numbers and the methodological challenges of BIK BAD BAX measuring apoptosis in vivo78. nevertheless, many signatures of response to endocrine therapy include genes BAK JNK Mitochondrial with roles in apoptosis, as discussed below. As tumour Gene Cellular TNF cytochrome c transcription stress growth reflects the balance between cell proliferation release and cell death, disruption of this balance by effects on p38 MAPK survival signalling and apoptosis are expected to affect Synergy Caspase clinical response. There is accumulating evidence for with TNF activation the increased expression of anti-apoptotic molecules, IRF1 NF-gB for example BCl-2 and BCl-Xl, and decreased expres- JAK or Ceramide sion of pro-apoptotic molecules, for example BAK, BiK Stat Akt 17 IFN and caspase 9, in attenuated responses to tamoxifen . PI3K Although many of these responses are probably conse- Autophagy Apoptosis RTK quences of the activation of survival signalling through Figure 2 | anti-oestrogen action on the cell cycle and apoptotic pathways. the pi3K–Akt pathway, as a consequence of overex- a | Anti-oestrogen (AE) treatment of cultured breast cancer cells leads to oestrogen Nature Reviews | Cancer pression of receptor tyrosine kinases and increased receptor (ER) binding and subsequent rapid decreases in the expression of MYC, followed ‘non-genomic’ signalling from cytoplasmic ER, by decreased expression of cyclin D1. Downregulation of MYC leads to de-repression of other pathways have been documented. For example, CDKN1A (which encodes p21) transcription. In addition, because cyclin D1–cyclin- increased DnA-binding and transcriptional activity dependent kinase 4 (CDK4) complexes function as a cellular ‘sink’ for the CDK inhibitors 30 p21 and p27, the reduction in cyclin D1–CDK4 abundance makes p21 and p27 available of nF-κB are features of tamoxifen-resistant cells , and for cyclin E1–CDK2 binding, and so indirectly contributes to the inhibition of cyclin tamoxifen sensitivity can be restored by parthenolide, 30,79 E1–CDK2 activity. The decrease in activity of both CDK2 and CDK4 prevents RB a specific nF-κB inhibitor . Tamoxifen insensitivity phosphorylation (P) and therefore impedes transition into S phase. Treatment with the in vitro is also associated with the downregulation of pure anti-oestrogen ICI 182780, but not tamoxifen, leads to an increase in the expression iRF1, an interferon-responsive putative tumour sup- of p27 and molecular markers that are characteristic of quiescence (G0), that is, the pressor that binds nF-κB and is essential for apoptosis. formation of p130–E2F4 complexes and the accumulation of hyperphosphorylated E2F4. Furthermore, overexpression of a splice variant of human These effects are reviewed in REF. 55. b | Proteins and processes that are upregulated X-box-binding protein 1 (XBp1), a transcription factor during anti-oestrogen-induced apoptosis are indicated in green (reviewed in REF. 76); red that controls the unfolded protein response, is also associ- crosses indicate proteins that are downregulated. Apoptotic concentrations of tamoxifen ated with tamoxifen resistance in vitro and poor survival elicit caspase activation downstream of responses such as activation of the stress kinases 80–82 Jun N-terminal kinase (JNK) and p38 MAPK, activation of the intracellular second in patients with breast cancer treated with tamoxifen . messenger ceramide, transcriptional downregulation of anti-apoptotic molecules nF-κB, XBp1 and iRF1 expression are correlated in 83 including BCL-2, and upregulation of pro-apoptotic molecules such as IRF1, BIK and patients with breast cancer , which may indicate that possibly BAK. In addition, anti-oestrogens have effects on the interferon (IFN) and nuclear these molecules function in a common pathway14. factor-κB (NF-κB) pathways, and on survival signalling through Akt downstream of receptor tyrosine kinases (RTKs), as well as synergistic effects on tumour necrosis factor signatures of tamoxifen responsiveness (TNF)-mediated apoptosis. These pro-apoptotic effects of tamoxifen are opposed by The advent of genome-wide gene expression analy- 77 autophagy . JAK, janus kinase; Stat, signal transducer and transcription activator. sis allowed clinical material from patients of known responsiveness to tamoxifen to be used as a means of gaining broad insights into the potential mechanisms Cell survival signalling and apoptosis. Treatment with of endocrine resistance. it also helped in the develop- high (micromolar) concentrations of anti-oestrogen, ment of clinically relevant markers of response and

Cyclin D1 oestrogen withdrawal (mimicking the effects of aroma- potential mechanisms of resistance. This is not without The regulatory subunit of a tase inhibitors) or aromatase inhibitor treatment of cells its own limitations, for example the difficulty of obtaining kinase complex that functions transfected with aromatase leads to the activation of the tumour tissue at the time when resistance has developed, as a growth factor sensor to cellular stress response and apoptosis in breast cancer rather than before therapy. regulate G1 phase cell cycle cells17,76. The molecular mechanisms are not well defined, progression. The catalytic subunits of cyclin but several molecular consequences that promote apop- Selection on the basis of disease outcome. The identifica- D1‑dependent kinases are tosis have been documented, including the regulation of tion of women who are unlikely to respond to endocrine CDK4 and CDK6. Bcl‑2 family members and increases in the apoptotic second therapy, but who may benefit from chemotherapy, is a

Table 1 | pathways associated with tamoxifen resistance in vitro Pathway Molecular aberration Clinical correlates Poor outcome* Tamoxifen resistance‡ ER signalling ERs and ERRs ERα loss (methylation) Yes 14 Yes 14 ERα36 (truncated variant) No24 Yes 24 ERα phosphorylation Yes 130 Yes 130 ERα methylation (by PRMT1) ND ND ERβ, ERβ2/cx and ERβ5 Yes 23,25 Yes 23,25 ERRγ ND ND ER-associated AP1 overexpression Yes 28–30 Yes 28–30

transcription factors 131 131 and co-activators NF-κB activation Yes Yes NCOA1 overexpression Yes 132 Yes 132 NCOA3 amplification Yes 32,133 Yes 32,132 PELP1 ND ND CBP and p300 overexpression Yes 134 Yes 134 Growth factor receptors and cytoplasmic signalling Receptors EGFR overexpression and ERBB2 amplification Yes 21,135 Yes 21,135 PAX2 loss leading to ERBB2 de-repression Yes 46 Yes 46 IGF1R overexpression Yes 136 ND FGFR overexpression ND Yes (only FGFR4)137 MAPK signalling Mek and Erk activation Yes 138 Yes 138 CDK10 methylation ND Yes 114 PI3K signalling Akt activation and overexpression Yes 139,140 Yes 139,140 PTEN loss Yes 141 Yes 141 SRC and SRC activation ND ND

SRC-interacting 111 48 proteins BCAR1 No Yes BCAR3 overexpression No111 No — associated with favourable response111

pressing clinical need that has driven much of the work importance of these molecules in anti-oestrogen effects on aimed at identifying clinically useful markers of tamoxifen proliferation. However, signatures predominantly com- response. Gene selection on the basis of correlations posed of MYC-responsive or RB- and E2f-responsive between expression and patient outcome, with the goal of genes are associated with poor outcome in women identifying a minimum gene set that retains robust pre- treated with tamoxifen62,86, suggesting that the absence dictive ability in women treated with tamoxifen, has led of these genes reflects the limitations of assessing their to the development of several gene signatures (TABLE 2), activity on the basis of their expression alone. Bcl-2 family some of which are the subject of testing in prospective Because of their shared biological features, many of the A protein family of up to 25 clinical trials (reviewed in REF. 84). Other signatures signatures selected on the basis of correlations with out- members that are classified derived using a broader group of patients with breast come are largely concordant, both with each other and with according to their structure cancer can also distinguish prognostic groups within more generally derived prognostic signatures that have not and function as anti‑apoptotic (TABLE 2) (BCL2‑like) or pro‑apoptotic ER-positive cancers treated with tamoxifen . been tested for association with response to tamoxifen, 87–89 (multidomain BAX‑like and Consistent with the idea that deregulation of ER sig- for example the ‘Mammaprint’ signature . in a recent ‘BH3‑only’) proteins. nalling and upregulation of alternative mitogenic path- meta-analysis that compared nine prognostic signatures, ways is a common mechanism of resistance, essentially the subset of genes related to proliferation within each Autophagy all these signatures contain a substantial proportion of signature was at least as good a predictor as the whole sig- A cellular response in which 90 the cell metabolizes its own genes that are ER targets, are involved in ER action or nature . However, the remaining genes were also often 90 contents and organelles to have roles in cell proliferation and survival (TABLE 2). significantly correlated with outcome , indicating that maintain energy production, Genes involved in apoptosis, invasion and cell motil- biological processes other than proliferation may also be often in response to stressful ity are also consistently included85 (TABLE 2). perhaps important. These probably include apoptosis and invasion stimuli. Although such a process can eventually result in surprisingly, MYC, CCND1 (which encodes cyclin D1) or metastasis, as signatures predominantly composed of cell death, it can also be used and RB1 are rarely included in the signatures, despite genes with roles in these cellular processes are predictive to maintain cell survival. the predominance of proliferation markers and the of poor outcome after tamoxifen treatment86,91,92.

Table 1 (cont.) | pathways associated with tamoxifen resistance in vitro Pathway Molecular aberration Clinical correlates Poor outcome* Tamoxifen resistance‡ Cell cycle Cyclins Cyclin D1 amplification or overexpression Yes 69 Yes 69 Cyclin D1b overexpression Yes 142 ND Cyclin E1 overexpression Yes 69 Yes 143 MYC MYC amplification and overexpression Yes 69,144 Yes 69 CDK inhibitors Cytoplasmic p21 expression ND Yes 75 Low p27 expression Yes 71,145 Yes 71,145 RB RB inactivation ND Yes 62 Apoptosis and cell survival signalling Bcl-2 family members Low BCL-2 expression Yes 146 Yes 147 BIK ND ND Low BAD expression Yes 148 Yes 148 Survival signalling (IRF1) ND ND XBP1 (spliced variant) ND Yes 82 *Yes indicates poor outcome in women with breast cancer, whether or not treated with tamoxifen.‡Yes indicates poor outcome in women treated with tamoxifen. AP1, activator protein 1; CBP, CREB-binding protein; CDK, cyclin-dependent kinase; EGFR, epidermal growth factor receptor; ER, oestrogen receptor; ERRγ, oestrogen-related receptor-γ; FGFR, fibroblast growth factor receptor; IGF1R, insulin-like growth factor receptor; IRF1, interferon regulatory factor 1; ND, not done; NF-κB, nuclear factor-κB; NCOA, nuclear receptor co-activator; PRMT1, protein arginine N-methyltransferase 1; XBP1, X-box-binding protein 1.

Biology as a selection criterion. Biomarkers of therapeutic of oestrogen-induced genes selected on the basis of response are expected to include genes that are mecha- co-expression with pR is correlated with longer survival nistically involved in conferring drug resistance, but in tamoxifen-treated women96, many oestrogen-induced signatures of tamoxifen response derived solely from genes are correlated with poor outcome86,94,95. it is clear clinical outcome data have provided only limited that oestrogen-regulated genes are not homogeneous mechanistic insights. There is little overlap between but are instead regulated by different mechanisms their component genes and, consequently, there are (for example, direct ERα–DnA binding at oestrogen few obvious starting points for functional studies. response elements compared with indirect ERα tether- potential contributing factors include dependence ing to DnA through interactions with other transcription of the signature composition on the patient set used factors (FIG. 1)) and can be divided into biologically distinct to derive it, and minimization of the signature size subsets86,91 that may have independent relationships with to increase clinical practicality, with the result that response to therapy. multiple non-overlapping signatures of equal predictive power could potentially be derived from the same Intrinsic biology or altered response? An important analysis93. Therefore, any single prognostic or predic- clinical question concerns the degree to which the cor- tive signature contains a somewhat arbitrary set of relation between individual signatures and outcome genes, and this selectivity can confound attempts to in women treated with tamoxifen is a consequence of understand the underlying biology. response to therapy rather than inherently aggressive A potentially more mechanistically revealing or indolent biology. As with individual candidate genes approach is the inclusion of facets of the underlying (TABLE 1), many outcome signatures, even those derived biology in the selection criteria for genes that consti- specifically from women treated with tamoxifen, are cor- Unfolded protein response tute a signature. Because of the importance of oestro- related with outcome irrespective of treatment (TABLE 2) A cellular response to stress that senses the misfolding of gen action in breast cancer biology and the evidence and may therefore reflect the underlying biology as well proteins in the endoplasmic for deregulation of oestrogen signalling as a major as, or instead of, response to treatment. Differential sen- reticulum. It activates a series mechanism of resistance, oestrogen-regulated genes sitivity to endocrine therapy is observed in the different of pathways that help the cells are common starting points for the derivation of signa- subtypes of ER-positive breast cancer that are defined by survive proteotoxicity that is 98,99 caused by unfolded proteins tures that are predictive of outcome in women treated common patterns of gene expression . Of these, the 86,94–96 or activate mechanisms of cell with tamoxifen (TABLE 2). The progesterone recep- best characterized are the luminal A and B subtypes. The death. tor (pR) is a well-established marker of response to luminal A subtype has higher ER expression and lower endocrine therapy97; pR expression occurs rarely in the proliferation indices than the remaining luminal cancers Concordant absence of functional ER, implying that pR-positive have, and a better outcome independent of tamoxifen Clinical biomarkers and 90,98–100 signatures are concordant if cancers, and cancers expressing high levels of other therapy . The subtypes are characterized by dif- they classify the same patients oestrogen-induced genes, are likely to be more depend- ferent patterns and degrees of copy number abnormali- as ‘high risk’. ent on ER signalling. However, although a signature ties101, indicating that at least some differences in clinical

Multivariate analysis response to tamoxifen may arise through differences in it might be informative to distinguish different mecha- A statistical analysis of the intrinsic tumour biology rather than through differences nisms of deregulation of proliferation or different types relationship between multiple in tamoxifen response per se. of proliferative defects. parameters (variables) to Differences in response also arise from differences identify those that have a insights from computational approaches dominant effect on outcome in the rate of cell proliferation. Although proliferation is (termed independent reduced in almost all ER-positive breast cancers follow- Dissection of biological processes. A limitation of many predictors of outcome) and ing neoadjuvant endocrine therapy, the level of prolifera- clustering algorithms used to derive gene signatures is those that are dependant or tion following short-term therapy is a better predictor of their dependence on global behaviour across the entire redundant. outcome than its pretreatment level or the magnitude test set of profiles, at the expense of identifying more 102,103 Biological concepts analysis of the decrease in proliferation following treatment . specific features that may be strongly related to only a A bioinformatic approach in This suggests that poor treatment response does not small group of patients or conditions. in addition, an which related information is result from intrinsically high proliferation but rather individual gene is often allocated to a particular cluster, grouped together into a from an ability to maintain high levels of proliferation although genes commonly contribute to different bio- ‘biological concept’, and associations between different in the presence of tamoxifen. increased proliferation that logical processes or are regulated by multiple stimuli. ‘concepts’ are sought. occurs as a consequence of deregulated mitogenic sig- However, some algorithms do allow the identification nalling pathways may be more readily inhibited by of partially overlapping gene sets that show coordinate tamoxifen than is increased proliferation that results expression in only particular contexts, and so these from genetic or epigenetic events targeting individual tend to be functionally coherent. One such algorithm genes directly involved in cell cycle progression (for yielded eight biologically consistent gene set ‘modules’ example, MYC or CCND1 amplification). in addition, in breast cancer expression profiles, including one the effects of the deregulation of cell cycle regulatory module that was largely composed of genes with roles genes are likely to be gene specific. Consequently, to bet- in apoptosis91. These genes were downregulated when ter understand the mechanisms of tamoxifen resistance MCF-7 breast cancer cells were treated with tamoxifen, suggesting that their expression may reflect active ER signalling91. Consistent with this idea, high expression Nuclear RTK signalling Stress of this module was associated with longer survival in receptors Growth factors responses several patient populations, including in women treated EGF, TGFα, IGF1, HRG and FGF ROS with tamoxifen. A complementary approach is to identify biologi- NCOA3 RTKs cally coherent signatures on the basis of a candidate ERBB2 EGFR, IGFR and FGFR Src gene approach or a functional annotation. For example, the high expression of a gene signature consisting ER PAX2 P PAK1 P of genes that are differentially expressed following RB P loss, or the expression of constitutively active (growth- BCAR3 inhibitory) RB mutants in rodent fibroblasts, was associ- P BCAR1 ated with shorter disease-free survival in women treated 62 CDK10 Erk PI3K or Akt JNK with tamoxifen . in another study, a combination of gene ontology and pathway analysis that used a curated EERα database of functional annotations led to the subdivi- P IRF1 sion of oestrogen-regulated genes into networks that represented specific biological processes, for example NCOA3 P mTOR NF-κB XBP1 two distinct aspects of cell proliferation (cell cycle progression and increased cell size (cell growth)), apoptosis and survival signalling, and transcriptional regulation86. The networks representing cell proliferation and apop- ↑ Proliferation ↓ Apoptosis tosis were predictive of poor outcome in women treated MYC, cyclin D1, cyclin E1, BCL-2, BIK, with tamoxifen86. previous analyses had pointed to both RB, p21 and p27 BCLXL and BAK proliferation and apoptosis genes as potential markers of Figure 3 | Molecular mechanisms of endocrine resistance. AcquiredNatur etamoxifen Reviews | Cancer outcome in women treated with tamoxifen. This study resistance in breast cancer cells is mediated by either modulation of the oestrogen provides evidence that these may identify distinct sub- receptor (ER) pathway or aberrant or compensatory cellular signalling pathways populations with different mechanisms of resistance, as controlled by growth factor receptors that negate the anti-proliferative and the cell proliferation-related signatures and the apopto- 11,12,14–21 pro-apoptotic effects of tamoxifen . In tamoxifen-resistant cells, these end points sis signature were independent predictors of outcome in are regulated by nuclear receptors (ER and EERα) and their co-activators, for example multivariate analysis86. nuclear receptor co-activator 3 (NCOA3; also known as AIB1 or SRC3); receptor tyrosine kinases (RTKs), including the epidermal growth factor receptor (EGFR), ERBB2, insulin-like growth factor receptor (IGFR) and fibroblast growth factor receptor (FGFR); RTK ligands, Molecular mechanisms. Although data from large- including transforming growth factor-α (TGFα) and heregulin (HRG), and downstream scale genomic studies of breast cancer are increasingly signalling pathways; and cellular stress responses, including those downstream of reactive available, few studies have adopted global approaches oxygen species (ROS). CDK10, cyclin-dependent kinase 10; JNK, Jun N-terminal kinase; to identify specific molecular aberrations that might be P, phosphorylation; PAK1, p21-activated kinase 1; PAX2, paired box 2 transcription factor; associated with response to endocrine therapy. However, NF-κB, nuclear factor-κB; XBP1, X-box-binding protein 1. a global biological concept analysis identified associations

Table 2 | signatures predicting response to endocrine therapies Signature Selection criteria Clinical correlates Biological processes Poor Tamoxifen outcome* resistance‡ Breast cancer Hierarchical clustering of breast cancers Yes Yes (luminal A Proliferation (luminal A compared with subtypes98,100 compared with luminal B) luminal B) HOXB13/IL17RB Differential expression in recurrent Yes Yes ER targets and invasion (HOXB13) expression ratio149 compared with non-recurrent breast cancers treated with adjuvant tamoxifen; association with recurrence 21-gene signature:16 250 manually selected genes tested Yes Yes ER targets, proliferation85, apoptosis85, cancer-related and 5 for association with outcome following invasion and motility85, signalling85 and controls (Oncotype adjuvant tamoxifen ERBB2 amplification Dx)150 81-gene discriminatory Differential expression in progressive No89 Yes ER action and ER targets, apoptosis, signature and 44-gene disease compared with objective extracellular matrix formation, immune predictive signature151 response in recurrent disease treated response and 17q21–22 amplification with tamoxifen 76-gene signature152 Correlation with outcome in breast Yes Possibly Proliferation and cell cycle85, and cancer apoptosis85 REF. 94 Oestrogen-induced genes used to cluster Yes Yes ER targets, proliferation and cell cycle, ER-positive breast cancer: groups have and apoptosis differing outcomes Genomic grade Distinguishes grade 1 and grade 3 breast Yes Yes 88 Proliferation and cell cycle85, apoptosis85, signature153 cancer motility85 and immune response85 ‘TuM1’ 33 genes91 Modular analysis linked to biological No Yes Apoptosis function 59-gene RB-deficiency Genes deregulated by RB1 loss or ND Yes Proliferation and E2f targets signature62 repressed by RB activation 36-gene signature154 Segregation of relapse and relapse-free ND Yes ER action and ER targets, proliferation, breast cancers treated with adjuvant apoptosis, cell survival, adhesion and tamoxifen motility, signalling, metabolism and immune response REFS 155,156 Segregation of responsive and ND Yes ER targets, cell cycle, cell movement, non-responsive breast cancers apoptosis, and TGFβ and ERBB2 treated with neoadjuvant aromatase signalling inhibitor with or without tamoxifen: oestrogen-responsive gene subset identified 36 genes96 Oestrogen-induced, underexpressed No Yes ER targets in ER-negative breast cancer and correlated with PR 181 genes in 13 Clusters of co-expressed genes in breast ND Yes — adjuvant Cell cycle and proliferation, apoptosis, biological clusters157 cancers treated with adjuvant tamoxifen; tamoxifen; no motility, and signalling of TGFβ, EGF, IGF 13 were combined to predict outcome — tamoxifen at and PDGF after adjuvant tamoxifen relapse except for one cluster alone 256 genes in 4 Functional annotation of oestrogen- ND Yes ER targets, Myc targets, cell cycle, signatures86 regulated genes proliferation, cell growth, apoptosis, survival signalling, and invasion and motility 47 genes158 Differential expression at the time ND Yes — some ER targets, immune response, of progression on tamoxifen therapy individual genes transcriptional regulation, proliferation, compared with non-recurrent cancers invasion and adhesion, and apoptosis *Yes indicates poor outcome in women with breast cancer, whether or not treated with tamoxifen. ‡Yes indicates poor outcome in women treated with tamoxifen. EGF, epidermal growth factor; ER, oestrogen receptor; HOXB13, homeobox B13; IGF, insulin-like growth factor; IL17RB, interleukin 17 receptor B; ND, not done; PDGF, platelet-derived growth factor; PR, progesterone receptor; TGFβ, transforming growth factor-β.

between a signature of early relapse in ER-positive breast The E2f and MYC biological concepts seemed to make cancer and the activation of MYC- and E2f-responsive independent contributions to the likelihood of disease pathways, together with chromosomal aberrations at progression104, despite the considerable overlap in the loci including 8q24 (the location of the MYC gene)104. known functions of these genes. This may be because

Synthetic lethal aberrations in these pathways occur independently and implications for therapy In genetics, a phenomenon in characterize biologically distinct phenotypes, as a sub- Endocrine-resistant ER-positive disease accounts for which the combination of two set of ER-positive breast cancers is characterized by a approximately one in four breast cancers but, similarly otherwise non‑lethal mutations high probability of MYC pathway deregulation but a low to the ‘triple-negative’ phenotype that does not express results in an inviable cell. Used 105,106 in the context of functional probability of E2f pathway deregulation . ER, pR or ERBB2, it is not well served by current screens to indicate a screen in targeted therapies. The possibility of identifying new which the end point is Functional genetic screens targets for therapy in resistant disease, or patients who apparent in only some in parallel with the emergence of increasingly sophis- may benefit from additional treatment with existing conditions, for example in the ticated bioinformatic and genomic tools, there has therapies, provides a strong impetus to identify markers presence of a specific genetic lesion. been an increase in the feasibility of global functional and mediators of therapeutic resistance. Some candi- screens. Gain-of-function screens in breast cancer date endocrine resistance genes can also affect response cells have identified >40 candidate tamoxifen resist- to other therapies; for example, BCAR1 confers resistance genes47,51,107–110. These include genes encoding cell ance to adriamycin116, and cyclin D1 confers resistance surface receptors (EGFR, ERBB2, platelet-derived growth to the EGFR-targeted therapy gefitinib117. This adds to factor receptor-β (PDGFRB) and colony-stimulating fac- the challenge of identifying alternative therapies in tor 1 receptor (CSF1R)) and their ligands (neuregulin, endocrine-resistant disease. Recent data suggest that which is a ligand for some Erbb receptors, and FGF17), integrative bioinformatics and/or targeted large-scale intracellular signalling molecules (BCAR1, BCAR3, screening may assist in meeting this challenge. One AKT1, AKT2, SRC and GRB7) and transcriptional regu- bioinformatic study identified highly significant links lators (BCAR2 (also known as TRERF1) and the ER co- between genes repressed by inhibition of pi3K and genes repressor NCOR2 (also known as SMRT)), as well as activated by MYC, suggesting that pi3K inhibition may genes of poorly understood function (for example, be an effective therapy for MYC-dependent cancers104. BCAR4 and TLE3). Many have roles in signalling Together with evidence from the combined small- pathways that are implicated in endocrine resistance, molecule and RnA interference screen discussed above such as the Erk and pi3K–Akt pathways (as summa- that highlighted inhibition of Akt as a means of poten- rized above and in TABLE 1), illustrating the potential tiating tamoxifen responses in vitro115, this observation for identifying functionally relevant pathways using argues that therapies targeting the pi3K–Akt pathway this approach. Genes derived from unbiased func- may be useful in endocrine-resistant breast cancer. Similar tional screens will not necessarily be deregulated in integrative analyses including data relating specifically to cancer, and even if they do show differential expres- endocrine resistance may yield new targets for further sion there is no guarantee that this will correlate with analysis and testing. Synthetic lethal genetic and/or small- a therapeutic response. However, some of the genes molecule screens using cell lines that are tamoxifen resist- that confer tamoxifen resistance in functional genetic ant, either because of adaption in culture or because of the screens are overexpressed in breast cancer and are inde- introduction of specific genetic lesions, may also identify pendent predictors of response in women treated with pathways specifically required in tamoxifen-resistant cells tamoxifen at relapse48,111–113. and the means of targeting them. An RnA interference screen to identify kinases Current models of endocrine resistance (FIG. 3) iden- for which decreased expression can reduce tamoxifen tify many individual molecules that can affect anti- sensitivity yielded 20 candidates, including multi- oestrogen sensitivity in vitro and are therefore potential ple components of the Erk signalling pathway, other targets for therapy in endocrine-resistant disease. The intracellular signalling molecules and several kinases convergence of many potential resistance genes on the related to CDC2, the prototypical CDK114. One of these, Erk and pi3K pathways suggests that inhibitors of these CDK10, was examined in more detail. The CDK10 gene pathways that are currently in development may be use- was silenced by methylation in 7 of the 34 breast cancers ful in this setting. Data from cell line models indicate examined, and low CDK10 expression was associated that inhibiting Src118, BCAR1 (REF. 119), Mek–Erk120, with poor outcome in women treated with adjuvant Akt38,115,121,122, mTOR39,123 or nF-κB30,79 can restore or tamoxifen114. Downregulation of CDK10 in cultured potentiate tamoxifen sensitivity. However, early clini- breast cancer cells led to an ETS2-dependent increase cal trials of EGFR- and ERBB2-targeted agents that in RAF1 transcription and activation of the Erk pathway. reduce Erk signalling (gefitinib, erlotinib, trastuzamab Therefore, downregulation of CDK10 seems to modulate and lapatinib) or mTOR inhibitors (everolimus and tamoxifen sensitivity through effects on intracellular sig- temsirolimus) in combination with endocrine therapies nalling, rather than through more direct effects on cell have yielded mixed results, which may be partly due to cycle progression. This screen also identified sensitizers difficulties in identifying tumours that are dependent on of tamoxifen response, many of which were in the pi3K these pathways and so are most likely to benefit from the pathway, including PIK3C2B, PDK1 (which is activated additional therapy124. in addition, crosstalk and negative by pi3K) and the pDK1 substrates PRKCZ and AKT1, feedback loops between different signal transduction as well as several other activators of Akt115. The sensi- pathways may lead to cellular resistance to individual tizers identified in a parallel small-molecule screen also inhibitors. One approach to this problem is the use included many inhibitors of Akt115, again emphasizing of inhibitors that target more than one kinase or the the importance of the pi3K pathway as a determinant use of combinations of therapies to simultaneously of the response to tamoxifen. target multiple pathways. Alternatively, end points shared

by multiple pathways — for example MYC and cyclin towards areas for further study. The next generation D1–CDK4 — could be targeted, which would have the of high-throughput technologies allows genome-scale benefit of more directly targeting proliferation, a cen- genetic and synthetic lethal screens, large-scale pro- tral component in essentially all gene signatures that teomic and phospho-proteomic profiling, and parallel are predictive of endocrine resistance. Recent evidence transcriptome, epigenome and genome sequencing of indicates that targeting MYC may be an effective cancer both experimental models and well-annotated clinical therapy125, and some small-molecule CDK inhibitors samples. The further use of integrative bioinformatic have entered clinical trials126. approaches combined with these new technologies offers the potential for a ‘systems biology’ approach, in concluding remarks which cellular responses and pathways are considered To date, much of the information on the mechanisms as a whole. This may identify previously unconsidered of endocrine resistance has come from studies that mechanisms for the modulation of therapeutic response, have considered relatively few genes. The application of as suggested by recent data implicating miRnAs in integrative approaches that examine gene expression in tamoxifen resistance72, and thereby greatly increase the context of global genomic, proteomic or functional the depth of our understanding of the mechanisms of data to understand mechanisms of endocrine resistance endocrine-resistant disease. This, in turn, should aid in is in its infancy and has relied on small-scale genomic the identification of the specific genetic events that cause and transcriptional data sets. nevertheless, as summa- resistance and the means by which these events can be rized above, this approach has provided some pointers targeted therapeutically.

1. McDonnell, D. P. & Norris, J. D. Connections and 19. Jordan, V. C. & O’Malley, B. W. Selective estrogen- 35. Koga, M., Musgrove, E. A. & Sutherland, R. L. regulation of the human estrogen receptor. Science receptor modulators and antihormonal resistance in Modulation of the growth-inhibitory effects of 296, 1642–1644 (2002). breast cancer. J. Clin. Oncol. 25, 5815–5824 (2007). progestins and the antiestrogen hydroxyclomiphene 2. Colditz, G. A. Relationship between estrogen levels, 20. Riggins, R. B., Schrecengost, R. S., Guerrero, M. S. & on human breast cancer cells by epidermal growth use of hormone replacement therapy, and breast Bouton, A. H. Pathways to tamoxifen resistance. factor and insulin. Cancer Res. 49, 112–116 (1989). cancer. J. Natl Cancer Inst. 90, 814–823 (1998). Cancer Lett. 256, 1–24 (2007). 36. Carroll, J. S. et al. p27Kip1 induces quiescence and 3. Hankinson, S. E., Colditz, G. A. & Willett, W. C. 21. Arpino, G., Wiechmann, L., Osborne, C. K. & Schiff, R. growth factor insensitivity in tamoxifen-treated Towards an integrated model for breast cancer Crosstalk between the estrogen receptor and the HER breast cancer cells. Cancer Res. 63, 4322–4326 etiology: the lifelong interplay of genes, lifestyle, and tyrosine kinase receptor family: molecular mechanism (2003). hormones. Breast Cancer Res. 6, 213–218 (2004). and clinical implications for endocrine therapy This paper shows that although cells treated with 4. Cuzick, J. Aromatase inhibitors for breast cancer resistance. Endocr. Rev. 29, 217–233 (2008). tamoxifen and other SERMs remain sensitive to prevention. J. Clin. Oncol. 23, 1636–1643 (2005). 22. Gutierrez, M. C. et al. Molecular changes in tamoxifen- growth factor stimulation, cells treated with ‘pure’ 5. Jordan, V. C. Chemoprevention of breast cancer with resistant breast cancer: relationship between estrogen anti-oestrogens do not, implicating the CDK selective oestrogen-receptor modulators. Nature Rev. receptor, HER-2, and p38 mitogen-activated protein inhibitor p27 in modulating growth factor Cancer 7, 46–53 (2007). kinase. J. Clin. Oncol. 23, 2469–2476 (2005). sensitivity in this context. 6. Howell, A. The endocrine prevention of breast cancer. 23. Herynk, M. H. & Fuqua, S. A. Estrogen receptor 37. Faridi, J., Wang, L., Endemann, G. & Roth, R. A. Best Pract. Res. Clin. Endocrinol. Metab. 22, mutations in human disease. Endocr. Rev. 25, Expression of constitutively active Akt-3 in MCF-7 615–623 (2008). 869–898 (2004). breast cancer cells reverses the estrogen and 7. Harvey, J. M., Clark, G. M., Osborne, C. K. & Allred, 24. Shi, L. et al. Expression of ER-α36, a novel variant of tamoxifen responsivity of these cells in vivo. Clin. D. C. Estrogen receptor status by estrogen receptor α, and resistance to tamoxifen Cancer Res. 9, 2933–2939 (2003). immunohistochemistry is superior to the ligand- treatment in breast cancer. J. Clin. Oncol. 27, 38. deGraffenried, L. A. et al. Eicosapentaenoic acid binding assay for predicting response to adjuvant 3423–3429 (2009). restores tamoxifen sensitivity in breast cancer cells endocrine therapy in breast cancer. J. Clin. Oncol. 17, 25. Honma, N. et al. Clinical importance of estrogen with high Akt activity. Ann. Oncol. 14, 1051–1056 1474–1481 (1999). receptor-β evaluation in breast cancer patients treated (2003). 8. Jensen, E. V. & Jordan, V. C. The estrogen receptor: a with adjuvant tamoxifen therapy. J. Clin. Oncol. 26, 39. deGraffenried, L. A. et al. Inhibition of mTOR activity model for molecular medicine. Clin. Cancer Res. 9, 3727–3734 (2008). restores tamoxifen response in breast cancer cells with 1980–1989 (2003). 26. Murphy, L. C. & Watson, P. H. Is oestrogen receptor-β aberrant Akt activity. Clin. Cancer Res. 10, 9. Early Breast Cancer Triallists’ Collaborative Group a predictor of endocrine therapy responsiveness in 8059–8067 (2004). (EBCTCG). Effects of chemotherapy and hormonal human breast cancer? Endocr. Relat. Cancer 13, 40. Miller, T. W. et al. Loss of phosphatase and tensin therapy for early breast cancer on recurrence and 327–334 (2006). homologue deleted on chromosome 10 engages 15-year survival: an overview of the randomised trials. 27. Riggins, R. B. et al. ERRγ mediates tamoxifen ErbB3 and insulin-like growth factor-I receptor Lancet 365, 1687–1717 (2005). resistance in novel models of invasive lobular breast signaling to promote antiestrogen resistance in breast 10. Johnston, S. R. & Dowsett, M. Aromatase inhibitors cancer. Cancer Res. 68, 8908–8917 (2008). cancer. Cancer Res. 69, 4192–4201 (2009). for breast cancer: lessons from the laboratory. Nature 28. Johnston, S. R. et al. Increased activator protein-1 DNA 41. McClelland, R. A. et al. Enhanced epidermal growth Rev. Cancer 3, 821–831 (2003). binding and c-Jun NH2-terminal kinase activity in factor receptor signaling in MCF7 breast cancer cells 11. MacGregor, J. I. & Jordan, V. C. Basic guide to the human breast tumors with acquired tamoxifen after long-term culture in the presence of the pure mechanisms of antiestrogen action. Pharmacol. Rev. resistance. Clin. Cancer Res. 5, 251–256 (1999). antiestrogen ICI 182,780 (Faslodex). Endocrinology 50, 151–196 (1998). 29. Schiff, R. et al. Oxidative stress and AP-1 activity in 142, 2776–2788 (2001). 12. Ali, S. & Coombes, R. C. Endocrine-responsive breast tamoxifen-resistant breast tumors in vivo. J. Natl 42. Knowlden, J. M. et al. Elevated levels of epidermal cancer and strategies for combating resistance. Cancer Inst. 92, 1926–1934 (2000). growth factor receptor/c-erbB2 heterodimers mediate Nature Rev. Cancer 2, 101–112 (2002). 30. Zhou, Y. et al. Enhanced NFκB and AP-1 an autocrine growth regulatory pathway in tamoxifen- 13. Hoskins, J. M., Carey, L. A. & McLeod, H. L. CYP2D6 transcriptional activity associated with antiestrogen resistant MCF-7 cells. Endocrinology 144, and tamoxifen: DNA matters in breast cancer. resistant breast cancer. BMC Cancer 7, 59 (2007). 1032–1044 (2003). Nature Rev. Cancer 9, 576–586 (2009). 31. Gururaj, A. E., Rayala, S. K., Vadlamudi, R. K. & 43. Hutcheson, I. R. et al. Oestrogen receptor-mediated 14. Clarke, R. et al. Antiestrogen resistance in breast Kumar, R. Novel mechanisms of resistance to modulation of the EGFR/MAPK pathway in tamoxifen- cancer and the role of estrogen receptor signaling. endocrine therapy: genomic and nongenomic resistant MCF-7 cells. Breast Cancer Res. Treat. 81, Oncogene 22, 7316–7339 (2003). considerations. Clin. Cancer Res. 12, 1001s–1007s 81–93 (2003). 15. Jordan, V. C. Selective estrogen receptor modulation: (2006). References 41–43 provide the first detailed concept and consequences in cancer. Cancer Cell 5, 32. Osborne, C. K. et al. Role of the estrogen receptor mechanistic evidence that activation of EGFR and 207–213 (2004). coactivator AIB1 (SRC-3) and HER-2/neu in tamoxifen ERBB2 signalling leads to anti-oestrogen 16. Ring, A. & Dowsett, M. Mechanisms of tamoxifen resistance in breast cancer. J. Natl Cancer Inst. 95, resistance. resistance. Endocr. Relat. Cancer 11, 643–658 (2004). 353–361 (2003). 44. Hua, G. et al. A negative feedback regulatory loop 17. Riggins, R. B., Bouton, A. H., Liu, M. C. & Clarke, R. 33. Le Romancer, M. et al. Regulation of estrogen rapid associates the tyrosine kinase receptor ERBB2 and Antiestrogens, aromatase inhibitors, and apoptosis in signaling through arginine methylation by PRMT1. the transcription factor GATA4 in breast cancer cells. breast cancer. Vitam. Horm. 71, 201–237 (2005). Mol. Cell 31, 212–221 (2008). Mol. Cancer Res. 7, 402–414 (2009). 18. Nicholson, R. I. et al. Growth factor signalling in 34. Sainsbury, J. R., Farndon, J. R., Sherbet, G. V. & 45. Zuo, T. et al. FOXP3 is an X-linked breast cancer endocrine and anti-growth factor resistant breast Harris, A. L. Epidermal-growth-factor receptors and suppressor gene and an important repressor of the cancer. Rev. Endocr. Metab. Disord. 8, 241–253 oestrogen receptors in human breast cancer. Lancet 1, HER‑2/ErbB2 oncogene. Cell 129, 1275–1286 (2007). 364–366 (1985). (2007).

46. Hurtado, A. et al. Regulation of ERBB2 by oestrogen 67. Coqueret, O. Linking cyclins to transcriptional control. 89. Kok, M. et al. Comparison of gene expression profiles receptor-PAX2 determines response to tamoxifen. Gene 299, 35–55 (2002). predicting progression in breast cancer patients Nature 456, 663–666 (2008). 68. Ishii, Y., Waxman, S. & Germain, D. Tamoxifen treated with tamoxifen. Breast Cancer Res. Treat. 113, This paper shows that the transcription factor stimulates the growth of cyclin D1-overexpressing 275–283 (2009). PAX2 competes with the ERα co-activator AIB1, breast cancer cells by promoting the activation of 90. Wirapati, P. et al. Meta-analysis of gene expression thereby modulating ER-mediated repression of signal transducer and activator of transcription 3. profiles in breast cancer: toward a unified ERBB2 and tamoxifen responsiveness, and that Cancer Res. 68, 852–860 (2008). understanding of breast cancer subtyping and this mechanism of ERBB2 repression is of clinical 69. Butt, A. J., McNeil, C. M., Musgrove, E. A. & prognosis signatures. Breast Cancer Res. 10, R65 significance. Sutherland, R. L. Downstream targets of growth factor (2008). 47. Dorssers, L. C., van Agthoven, T., Dekker, A., van and oestrogen signalling and endocrine resistance: the 91. Yu, K., Ganesan, K., Miller, L. D. & Tan, P. A modular Agthoven, T. L. & Kok, E. M. Induction of antiestrogen potential roles of c-Myc, cyclin D1 and cyclin E. analysis of breast cancer reveals a novel low-grade resistance in human breast cancer cells by random Endocr. Relat. Cancer 12, S47–S59 (2005). molecular signature in estrogen receptor-positive insertional mutagenesis using defective retroviruses: 70. Caldon, C. E., Daly, R. J., Sutherland, R. L. & tumors. Clin. Cancer Res. 12, 3288–3296 (2006). identification of bcar-1, a common integration site. Musgrove, E. A. Cell cycle control in breast cancer 92. Helleman, J. et al. Association of an extracellular Mol. Endocrinol. 7, 870–878 (1993). cells. J. Cell. Biochem. 97, 261–274 (2006). matrix gene cluster with breast cancer prognosis and 48. van der Flier, S. et al. Bcar1/p130Cas protein and 71. Chu, I. M., Hengst, L. & Slingerland, J. M. The Cdk endocrine therapy response. Clin. Cancer Res. 14, primary breast cancer: prognosis and response to inhibitor p27 in human cancer: prognostic potential 5555–5564 (2008). tamoxifen treatment. J. Natl Cancer Inst. 92, and relevance to anticancer therapy. Nature Rev. 93. Ein-Dor, L., Kela, I., Getz, G., Givol, D. & Domany, E. 120–127 (2000). Cancer 8, 253–267 (2008). Outcome signature genes in breast cancer: is there a 49. Brinkman, A. et al. The substrate domain of BCAR1 is 72. Miller, T. E. et al. MicroRNA-221/222 confers unique set? Bioinformatics 21, 171–178 (2005). essential for anti-estrogen-resistant proliferation of tamoxifen resistance in breast cancer by targeting 94. Oh, D. S. et al. Estrogen-regulated genes predict human breast cancer cells. Breast Cancer Res. Treat. 3 p27Kip1. J. Biol. Chem. 283, 29897–29903 (2008). survival in hormone receptor-positive breast cancers. May 2009 (doi:10.1007/s10549-009-0403–0404). 73. Zhao, J. J. et al. MicroRNA-221/222 negatively J. Clin. Oncol. 24, 1656–1664 (2006). 50. Riggins, R. B., Quilliam, L. A. & Bouton, A. H. regulates estrogen receptorα and is associated with 95. Yu, J. et al. A transcriptional fingerprint of estrogen in Synergistic promotion of c-Src activation and cell tamoxifen resistance in breast cancer. J. Biol. Chem. human breast cancer predicts patient survival. migration by Cas and AND-34/BCAR3. J. Biol. Chem. 283, 31079–31086 (2008). Neoplasia 10, 79–88 (2008). 278, 28264–28273 (2003). 74. Liu, R. et al. FOXP3 up-regulates p21 expression by 96. Lippman, M. E., Rae, J. M. & Chinnaiyan, A. M. An 51. van Agthoven, T. et al. Identification of BCAR3 by a site-specific inhibition of histone deacetylase 2/histone expression signature of estrogen-regulated genes random search for genes involved in antiestrogen deacetylase 4 association to the locus. Cancer Res. predicts disease-free survival in tamoxifen-treated resistance of human breast cancer cells. EMBO J. 17, 69, 2252–2259 (2009). patients better than progesterone receptor status. 2799–2808 (1998). 75. Perez-Tenorio, G. et al. Cytoplasmic p21WAF1/CIP1 Trans. Am. Clin. Climatol. Assoc. 119, 77–90 52. Cai, D. et al. AND-34/BCAR3, a GDP exchange factor correlates with Akt activation and poor response to (2008). whose overexpression confers antiestrogen resistance, tamoxifen in breast cancer. Int. J. Oncol. 28, References 86 and 94–96 examine the relationship activates Rac, PAK1, and the cyclin D1 promoter. 1031–1042 (2006). between networks of oestrogen-regulated genes Cancer Res. 63, 6802–6808 (2003). 76. Mandlekar, S. & Kong, A. N. Mechanisms of tamoxifen- and response to tamoxifen therapy. 53. Rayala, S. K., Molli, P. R. & Kumar, R. Nuclear p21- induced apoptosis. Apoptosis 6, 469–477 (2001). 97. Clark, G. M., McGuire, W. L., Hubay, C. A., Pearson, activated kinase 1 in breast cancer packs off 77. Schoenlein, P. V., Periyasamy-Thandavan, S., O. H. & Marshall, J. S. Progesterone receptors as a tamoxifen sensitivity. Cancer Res. 66, 5985–5988 Samaddar, J. S., Jackson, W. H. & Barrett, J. T. prognostic factor in Stage II breast cancer. N. Engl. (2006). Autophagy facilitates the progression of ERα-positive J. Med. 309, 1343–1347 (1983). 54. Dowsett, M. et al. Proliferation and apoptosis as breast cancer cells to antiestrogen resistance. 98. Sorlie, T. et al. Gene expression patterns of breast markers of benefit in neoadjuvant endocrine therapy Autophagy 5, 400–403 (2009). carcinomas distinguish tumor subclasses with clinical of breast cancer. Clin. Cancer Res. 12, 1024s–1030s 78. Butt, A. J., Sutherland, R. L. & Musgrove, E. A. Live or implications. Proc. Natl Acad. Sci. USA 98, (2006). let die: oestrogen regulation of survival signalling in 10869–10874 (2001). 55. Doisneau-Sixou, S. F. et al. Estrogen and antiestrogen endocrine response. Breast Cancer Res. 9, 306 99. Sotiriou, C. et al. Breast cancer classification and regulation of cell cycle progression in breast cancer (2007). prognosis based on gene expression profiles from a cells. Endocr. Relat. Cancer 10, 179–186 (2003). 79. Riggins, R. B., Zwart, A., Nehra, R. & Clarke, R. The population-based study. Proc. Natl Acad. Sci. USA 56. Prall, O. W. J., Rogan, E. M., Musgrove, E. A., Watts, nuclear factor κB inhibitor parthenolide restores ICI 100, 10393–10398 (2003). C. K. W. & Sutherland, R. L. c-Myc or cyclin D1 mimics 182,780 (Faslodex; fulvestrant)-induced apoptosis in 100. Sorlie, T. et al. Repeated observation of breast estrogen effects on cyclin E-Cdk2 activation and cell antiestrogen-resistant breast cancer cells. Mol. Cancer tumor subtypes in independent gene expression cycle reentry. Mol. Cell. Biol. 18, 4499–4508 (1998). Ther. 4, 33–41 (2005). data sets. Proc. Natl Acad. Sci. USA 100, 57. Venditti, M., Iwasiow, B., Orr, F. W. & Shiu, R. P. C-myc 80. Gu, Z. et al. Association of interferon regulatory 8418–8423 (2003). gene expression alone is sufficient to confer resistance factor-1, nucleophosmin, nuclear factor-κB, and cyclic 101. Chin, K. et al. Genomic and transcriptional aberrations to antiestrogen in human breast cancer cells. Int. AMP response element binding with acquired linked to breast cancer pathophysiologies. Cancer Cell J. Cancer 99, 35–42 (2002). resistance to Faslodex (ICI 182,780). Cancer Res. 62, 10, 529–541 (2006). 58. Hui, R. et al. Constitutive overexpression of cyclin D1 3428–3437 (2002). 102. Dowsett, M. et al. Prognostic value of Ki67 expression but not cyclin E confers acute resistance to 81. Gomez, B. P. et al. Human X-box binding protein-1 after short-term presurgical endocrine therapy for antiestrogens in T-47D breast cancer cells. Cancer confers both estrogen independence and antiestrogen primary breast cancer. J. Natl Cancer Inst. 99, Res. 62, 6916–6923 (2002). resistance in breast cancer cell lines. FASEB J. 21, 167–170 (2007). 59. Dhillon, N. K. & Mudryj, M. Ectopic expression of 4013–4027 (2007). A neoadjuvant study providing the first definitive cyclin E in estrogen responsive cells abrogates 82. Davies, M. P. et al. Expression and splicing of the evidence that a marker of proliferation, Ki67, antiestrogen mediated growth arrest. Oncogene 21, unfolded protein response gene XBP-1 are predicts responsiveness to endocrine therapy in 4626–4634 (2002). significantly associated with clinical outcome of patients. 60. Cariou, S. et al. Down-regulation of p21WAF1/CIP1 or endocrine-treated breast cancer. Int. J. Cancer 123, 103. Ellis, M. J. et al. Outcome prediction for estrogen p27Kip1 abrogates antiestrogen-mediated cell cycle 85–88 (2008). receptor-positive breast cancer based on arrest in human breast cancer cells. Proc. Natl Acad. 83. Zhu, Y. et al. Expression patterns among interferon postneoadjuvant endocrine therapy tumor Sci. USA 97, 9042–9046 (2000). regulatory factor-1, human X-box binding protein-1, characteristics. J. Natl Cancer Inst. 100, 1380–1388 61. Carroll, J. S., Prall, O. W. J., Musgrove, E. A. & nuclear factor kappa B, nucleophosmin, estrogen (2008). Sutherland, R. L. A pure estrogen antagonist inhibits receptor-alpha and progesterone receptor proteins in 104. Rhodes, D. R. et al. Molecular concepts analysis links cyclin E-Cdk2 activity in MCF-7 breast cancer cells and breast cancer tissue microarrays. Int. J. Oncol. 28, tumors, pathways, mechanisms, and drugs. Neoplasia induces accumulation of p130-E2F4 complexes 67–76 (2006). 9, 443–454 (2007). characteristic of quiescence. J. Biol. Chem. 275, 84. Sotiriou, C. & Pusztai, L. Gene-expression signatures 105. Acharya, C. R. et al. Gene expression signatures, 38221–38229 (2000). in breast cancer. N. Engl. J. Med. 360, 790–800 clinicopathological features, and individualized 62. Bosco, E. E. et al. The retinoblastoma tumor (2009). therapy in breast cancer. JAMA 299, 1574–1587 suppressor modifies the therapeutic response of 85. Yu, J. X. et al. Pathway analysis of gene signatures (2008). breast cancer. J. Clin. Invest. 117, 218–228 (2007). predicting metastasis of node-negative primary breast 106. Bild, A. H., Potti, A. & Nevins, J. R. Linking oncogenic 63. Varma, H., Skildum, A. J. & Conrad, S. E. Functional cancer. BMC Cancer 7, 182 (2007). pathways with therapeutic opportunities. Nature Rev. ablation of pRb activates Cdk2 and causes 86. Musgrove, E. A. et al. Identification of functional Cancer 6, 735–471 (2006). antiestrogen resistance in human breast cancer cells. networks of estrogen- and c-Myc-responsive genes and 107. van Agthoven, T. et al. Functional identification of PLoS ONE 2, e1256 (2007). their relationship to response to tamoxifen therapy in genes causing estrogen independence of human 64. Wang, Y. et al. Cyclin D1b is aberrantly regulated in breast cancer. PLoS ONE 3, e2987 (2008). breast cancer cells. Breast Cancer Res. Treat. 114, response to therapeutic challenge and promotes This paper and reference 62 derive gene networks 23–30 (2009). resistance to estrogen antagonists. Cancer Res. 68, on the basis of biological function that are 108. Dorssers, L. C. & Veldscholte, J. Identification of a 5628–5638 (2008). correlated with outcome in women treated with novel breast-cancer-anti-estrogen-resistance (BCAR2) 65. Mukherjee, S. & Conrad, S. E. c-Myc suppresses tamoxifen. locus by cell-fusion-mediated gene transfer in human p21WAF1/CIP1 expression during estrogen 87. Fan, C. et al. Concordance among breast-cancer cells. Int. J. Cancer 72, 700–705 signaling and antiestrogen resistance in human gene-expression-based predictors for breast cancer. (1997). breast cancer cells. J. Biol. Chem. 280, N. Engl. J. Med. 355, 560–569 (2006). 109. Brinkman, A., van der Flier, S., Kok, E. M. & Dorssers, 17617–17625 (2005). 88. Loi, S. et al. Definition of clinically distinct molecular L. C. BCAR1, a human homologue of the adapter 66. Caldon, C. E. et al. Estrogen regulation of cyclin E2 subtypes in estrogen receptor-positive breast protein p130Cas, and antiestrogen resistance in requires cyclin D1, but not c-Myc. Mol. Cell. Biol. 29 carcinomas through genomic grade. J. Clin. Oncol. 25, breast cancer cells. J. Natl Cancer Inst. 92, 112–120 4623–4639 (2009). 1239–1246 (2007). (2000).

110. Meijer, D., van Agthoven, T., Bosma, P. T., Nooter, K. & 127. Perissi, V. & Rosenfeld, M. G. Controlling nuclear 146. Callagy, G. M. et al. Bcl-2 is a prognostic marker in Dorssers, L. C. Functional screen for genes responsible receptors: the circular logic of cofactor cycles. Nature breast cancer independently of the Nottingham for tamoxifen resistance in human breast cancer cells. Rev. Mol. Cell Biol. 6, 542–554 (2005). Prognostic Index. Clin. Cancer Res. 12, 2468–2475 Mol. Cancer Res. 4, 379–386 (2006). 128. Couse, J. F. & Korach, K. S. Estrogen receptor null (2006). This paper and references 47, 51 and 108 mice: what have we learned and where will they lead 147. Gasparini, G. et al. Expression of bcl-2 protein document the first genetic screens aimed at us. Endocr. Rev. 20, 358–417 (1999). predicts efficacy of adjuvant treatments in operable identifying genes that can confer resistance to 129. Strom, A. et al. Estrogen receptor β inhibits node-positive breast cancer. Clin. Cancer Res. 1, anti-oestrogens in culture. 17β-estradiol-stimulated proliferation of the breast 189–198 (1995). 111. van Agthoven, T. et al. Relevance of breast cancer cancer cell line T47D. Proc. Natl Acad. Sci. USA 101, 148. Cannings, E. et al. Bad expression predicts outcome in antiestrogen resistance genes in human breast cancer 1566–1571 (2004). patients treated with tamoxifen. Breast Cancer Res. progression and tamoxifen resistance. J. Clin. Oncol. 130. Murphy, L. C., Niu, Y., Snell, L. & Watson, P. Treat. 102, 173–179 (2007). 27, 542–529 (2009). Phospho-serine-118 estrogen receptor-α expression is 149. Ma, X. J. et al. A two-gene expression ratio predicts 112. Meijer, D. et al. TSC22D1 and PSAP predict clinical associated with better disease outcome in women clinical outcome in breast cancer patients treated with outcome of tamoxifen treatment in patients with treated with tamoxifen. Clin. Cancer Res. 10, tamoxifen. Cancer Cell 5, 607–616 (2004). recurrent breast cancer. Breast Cancer Res. Treat. 5902–5906 (2004). 150. Paik, S. et al. A multigene assay to predict recurrence 113, 253–260 (2009). 131. Zhou, Y., Eppenberger-Castori, S., Eppenberger, U. & of tamoxifen-treated, node-negative breast cancer. 113. Dorssers, L. C. et al. The prognostic value of BCAR1 in Benz, C. C. The NFκB pathway and endocrine-resistant N. Engl. J. Med. 351, 2817–2826 (2004). patients with primary breast cancer. Clin. Cancer Res. breast cancer. Endocr. Relat. Cancer 12 (Suppl. 1), 151. Jansen, M. P. et al. Molecular classification of 10, 6194–6202 (2004). S37–S46 (2005). tamoxifen-resistant breast carcinomas by gene 114. Iorns, E. et al. Identification of CDK10 as an important 132. Redmond, A. M. et al. Coassociation of estrogen expression profiling. J. Clin. Oncol. 23, 732–740 determinant of resistance to endocrine therapy for receptor and p160 proteins predicts resistance to (2005). breast cancer. Cancer Cell 13, 91–104 (2008). endocrine treatment; SRC-1 is an independent References 149–151 were the first studies to use This publication describes the first RNA predictor of breast cancer recurrence. Clin. Cancer clinical material to derive gene expression interference screen for anti-oestrogen resistance. Res. 15, 2098–2106 (2009). signatures that were specifically aimed at It shows that decreased expression of CDK10 is 133. Harigopal, M. et al. Estrogen receptor co-activator predicting response to endocrine therapy. associated with anti-oestrogen resistance in culture (AIB1) protein expression by automated quantitative 152. Wang, Y. et al. Gene-expression profiles to predict and in women treated with tamoxifen. analysis (AQUA) in a breast cancer tissue microarray distant metastasis of lymph-node-negative primary 115. Iorns, E., Lord, C. J. & Ashworth, A. Parallel RNAi and and association with patient outcome. Breast Cancer breast cancer. Lancet 365, 671–679 (2005). compound screens identify the PDK1 pathway as a Res. Treat. 115, 77–85 (2009). 153. Sotiriou, C. et al. Gene expression profiling in breast target for tamoxifen sensitization. Biochem. J. 417, 134. Green, A. R. et al. The prognostic significance of cancer: understanding the molecular basis of 361–370 (2009). steroid receptor co-regulators in breast cancer: histologic grade to improve prognosis. J. Natl Cancer 116. Ta, H. Q., Thomas, K. S., Schrecengost, R. S. & co-repressor NCOR2/SMRT is an independent Inst. 98, 262–272 (2006). Bouton, A. H. A novel association between p130Cas indicator of poor outcome. Breast Cancer Res. Treat. 154. Chanrion, M. et al. A gene expression signature that and resistance to the chemotherapeutic drug 110, 427–437 (2008). can predict the recurrence of tamoxifen-treated adriamycin in human breast cancer cells. Cancer Res. 135. Arpino, G. et al. HER-2 amplification, HER-1 primary breast cancer. Clin. Cancer Res. 14, 68, 8796–8804 (2008). expression, and tamoxifen response in estrogen 1744–1752 (2008). 117. Kalish, L. H. et al. Deregulated cyclin D1 expression is receptor-positive metastatic breast cancer: a 155. Harvell, D. M. et al. Estrogen regulated gene associated with decreased efficacy of the selective southwest oncology group study. Clin. Cancer Res. 10, expression in response to neoadjuvant endocrine epidermal growth factor receptor tyrosine kinase 5670–5676 (2004). therapy of breast cancers: tamoxifen agonist effects inhibitor gefitinib in head and neck squamous cell 136. Creighton, C. J. et al. Insulin-like growth factor-I dominate in the presence of an aromatase inhibitor. carcinoma cell lines. Clin. Cancer Res. 10, activates gene transcription programs strongly Breast Cancer Res. Treat. 112, 489–501 (2008). 7764–7774 (2004). associated with poor breast cancer prognosis. J. Clin. 156. Harvell, D. M. et al. Molecular signatures of 118. Planas-Silva, M. D. & Hamilton, K. N. Targeting c-Src Oncol. 26, 4078–4085 (2008). neoadjuvant endocrine therapy for breast cancer: kinase enhances tamoxifen’s inhibitory effect on cell 137. Meijer, D. et al. Fibroblast growth factor receptor 4 characteristics of response or intrinsic resistance. growth by modulating expression of cell cycle and predicts failure on tamoxifen therapy in patients with Breast Cancer Res. Treat. 112, 475–488 (2008). survival proteins. Cancer Chemother. Pharmacol. 60, recurrent breast cancer. Endocr. Relat. Cancer 15, 157. Loi, S. et al. Predicting prognosis using molecular 535–543 (2007). 101–111 (2008). profiling in estrogen receptor-positive breast cancer 119. Soni, S., Lin, B. T., August, A., Nicholson, R. I. & 138. Bergqvist, J. et al. Activated ERK1/2 and treated with tamoxifen. BMC Genomics 9, 239 Kirsch, K. H. Expression of a phosphorylated phosphorylated oestrogen receptor α are associated (2008). p130(Cas) substrate domain attenuates the with improved breast cancer survival in women 158. Vendrell, J. A. et al. A candidate molecular signature phosphatidylinositol 3-kinase/Akt survival pathway in treated with tamoxifen. Eur. J. Cancer 42, 1104–1112 associated with tamoxifen failure in primary breast tamoxifen resistant breast cancer cells. J. Cell. (2006). cancer. Breast Cancer Res. 10, R88 (2008). Biochem. 107, 364–375 (2009). 139. Kirkegaard, T. et al. AKT activation predicts outcome 120. Kurokawa, H. et al. Inhibition of HER2/neu (erbB-2) in breast cancer patients treated with tamoxifen. Acknowledgements and mitogen-activated protein kinases enhances J. Pathol. 207, 139–146 (2005). We are grateful to R. J. Daly, C. E. Caldon and A. J. Butt for tamoxifen action against HER2-overexpressing, 140. Tokunaga, E. et al. Activation of PI3K/Akt signaling helpful discussions and comments, and C. E. Caldon for tamoxifen-resistant breast cancer cells. Cancer Res. and hormone resistance in breast cancer. Breast expert assistance with the figures. Research in the authors’ 60, 5887–5894 (2000). Cancer 13, 137–144 (2006). laboratories is supported by the National Health and Medical 121. Clark, A. S., West, K., Streicher, S. & Dennis, P. A. 141. Shoman, N. et al. Reduced PTEN expression Research Council of Australia, the Cancer Institute New South Constitutive and inducible Akt activity promotes predicts relapse in patients with breast carcinoma Wales, the Australian Cancer Research Foundation, the Petre resistance to chemotherapy, trastuzumab, or treated by tamoxifen. Mod. Pathol. 18, 250–259 Foundation and the RT Hall Trust. tamoxifen in breast cancer cells. Mol. Cancer Ther. 1, (2005). 707–717 (2002). 142. Millar, E. K. et al. Cyclin D1b protein expression in 122. deGraffenried, L. A. et al. NF-kappa B inhibition breast cancer is independent of cyclin D1a and DaTaBases markedly enhances sensitivity of resistant breast associated with poor disease outcome. Oncogene 28, National Cancer Institute Drug Dictionary: cancer tumor cells to tamoxifen. Ann. Oncol. 15, 1812–1820 (2009). http://www.cancer.gov/drugdictionary/ 885–890 (2004). 143. Span, P. N., Tjan-Heijnen, V. C., Manders, P., Beex, adriamycin | erlotinib | everolimus | gefitinib | lapatinib | 123. Boulay, A. et al. Dual inhibition of mTOR and estrogen L. V. & Sweep, C. G. Cyclin-E is a strong predictor of tamoxifen | temsirolimus receptor signaling in vitro induces cell death in models endocrine therapy failure in human breast cancer. UniProtKB: http://www.uniprot.org of breast cancer. Clin. Cancer Res. 11, 5319–5328 Oncogene 22, 4898–4904 (2003). AP1| BCAR1 | EGFR | ERα | ERβ | ERBB2 | FOXP3 | GATA4 | IRF1 | (2005). 144. Denning, S. L., Nass, S. J., Dickson, R. B. & Trock, B. J. NCOA3 | PAK1 | PAX2 | PELP1 | PR | PRMT1 | SP1 | STAT5B 124. Leary, A. F., Sirohi, B. & Johnston, S. R. Clinical trials C-myc amplification in breast cancer: a meta-analysis update: endocrine and biological therapy of its occurrence and prognostic relevance. Br. FurTher inFOrMaTiOn combinations in the treatment of breast cancer. Breast J. Cancer 83, 1688–1695 (2000). Elizabeth A. Musgrove’s homepage: http://www.garvan. Cancer Res. 9, 112 (2007). 145. Pohl, G. et al. High p27Kip1 expression predicts org.au/about-us/our-people/associate-professor-liz- 125. Soucek, L. et al. Modelling Myc inhibition as a cancer superior relapse-free and overall survival for musgrove therapy. Nature 455, 679–683 (2008). premenopausal women with early-stage breast Robert L. Sutherland’s homepage: http://www.garvan.org. 126. Shapiro, G. I. Cyclin-dependent kinase pathways as cancer receiving adjuvant treatment with tamoxifen au/about-us/our-people/professor-rob-sutherland targets for cancer treatment. J. Clin. Oncol. 24, plus goserelin. J. Clin. Oncol. 21, 3594–3600 all linkS are aCTive in The Online Pdf 1770–1783 (2006). (2003).