Luminal-B breast cancer and novel therapeutic targets

Gene expression profiling has led to a new molecular classification of breast cancer characterized by four intrinsic subtypes: basal-like, HER2-positive, luminal A, and luminal B. Despite expressing estrogen receptor, the luminal-B subtype confers increased risk of early relapse with endocrine therapy compared with the luminal-A subtype. Although luminal-B definitions vary, the hallmark appears to be increased expression of proliferation-related genes. Several biological pathways are identified as possible contributors to the poor outcomes, and novel agents targeting these pathways are being developed with aims to improve survival. We review the definition of luminal-B breast cancer, its pathological and clinical features, and potential targets for treatment.


Introduction
Th e National Cancer Institute defi nes personalized medicine as 'a form of medicine that uses information about a person's genes, proteins, and environment to prevent, diagnose, and treat disease' [1]. Personalized cancer medicine has existed in breast cancer since the late 1980s when benefi ts of tamoxifen were found to be limited to patients with tumors expressing estrogen receptors (ERs) [2]. Th is personalized treatment has advanced further in recent times through the discovery of erbB2/HER2 gene amplifi cation and its subsequent targeted treatments such as trastuzumab and lapatinib [3,4].
Until very recently, personalized cancer medicine in breast cancer relied on only two predictive markers, ER and erbB2/HER2. Th e advent of gene expression profi ling, however, has led to a paradigm shift in breast cancer medicine. Breast cancer is now recognized not as a single disease with variable morphology, but as at least four molecularly distinct neoplastic disorders: basal-like breast cancer, HER2-positive breast cancer, luminal-A breast cancer, and luminal-B breast cancer [5][6][7][8]. Although the immediate additional clinical value of this molecular classifi cation is limited by its close correlation to traditional methods of testing for ER and HER2, the identifi cation of genetic aberrations that underlie molecularly distinct subclasses of breast cancer has revealed new therapeutic targets and has reshaped breast cancer clinical trial design.
Th e subtypes most in need of therapeutic advances are basal-like breast cancer and luminal-B breast cancer, where therapeutic resistance is common and where advances in molecular profi ling have identifi ed promising new therapeutic targets. In the present review article, we discuss the defi nition of luminal-B breast cancer, the clinical behavior and pathological features of luminal-B breast cancer, and emerging molecular targets for improved therapy (see Table 1 for a summary).

Defi ning luminal-B breast cancer
Microarray technology has enabled better understanding of cancer biology at a molecular level through the interrogation of tens of thousands of expressed genes simultaneously. In breast cancer, hierarchical clustering of a series of breast cancers based upon a set of diff erently expressed intrinsic genes between individual patients led to the identifi cation of a novel molecular classifi cation of breast cancer [7]. Th e so-called intrinsic molecular classifi cation of human breast cancer includes basal-like, HER2-positive, luminal-A and luminal-B subtypes. Th ese subtypes have been associated with distinct pathological features and clinical outcome: basal-like breast cancer is predominantly triple-negative, with absent expression of ER, progesterone receptor (PR) and normal erbB2/HER2 gene copy number; HER2-positive breast cancer is erbB2/ HER2 gene amplifi ed and is associated with poorer outcomes when untreated; and both luminal-A and luminal-B breast cancers are ER-positive, although luminal-B cancers have poorer outcomes [9].
Th e seminal work of Perou and colleagues initially identifi ed molecular portraits of breast cancer based upon gene expression profi ling of 65 breast cancer samples from 42 individual patients using cDNA microarrays [7]. Th eir classifi cation was based upon the premise that individual diff erences in gene expression should be greater than diff erences in gene expression from paired tumor samples derived from the same patient. Th ey identifi ed a set of 496 genes that demonstrated signifi cantly greater variation between individual tumors than within paired tumor samples from the same individual. When this intrinsic gene set was used to perform hierarchical clustering of their tumor samples, four subgroups were identifi ed: basallike, based upon similari ties in gene expression to basal epithelial cells in the normal breast; Erb-B2 positive, based upon increased expression of genes in the erbB2/ HER2 gene amplicon on chromosome 17q12; luminal, based upon similarities in gene expression to luminal epithelial cells in the normal breast; and normal breastlike, based upon the inclusion of three normal, nonmalignant breast samples. In this initial study, no distinction between luminal-A and luminal-B breast cancers was identifi ed.
A subsequent study from the same group extended the sample size to 78 breast cancers (including 40 from the original publication) using hierarchical clustering with an intrinsic gene set of 456 cDNA clones. Extension of the sample size allowed for the identifi cation of subsets within the luminal cluster (47 tumors): luminal A (32 tumors), luminal B (fi ve tumors), and luminal C (10 tumors) [8]. Luminal B and luminal C demonstrated lower expression of ER-related genes compared with luminal-A tumors, while luminal C was further distinguished from luminal A and luminal B by high expression of a set of genes shared with basal-like and HER2-positive subtypes, but of unknown function [8]. Compared with luminal-A tumors, poorer outcomes were observed in luminal-B and luminal-C tumors. It is well recognized that hierarchical clustering based on a small sample number can result in unstable molecular classifi cations. Later studies failed to reproduce the luminal-C subtype and the luminal classifi cation was collapsed into two subtypes: luminal-A with high expression of ER-regulated genes and favorable longterm outcome, and luminal-B with lower expression of ER-regulated genes and poorer long-term outcome.
Multiple gene expression studies have reproduced luminal-A and luminal-B subtypes. Both subtypes have expression patterns reminiscent of the luminal epithelial component of the breast, including expression of luminal cytokeratins 8/18, ER and genes associated with ER activation such as CCND1 (cyclin D1). Th e major molecular distinction between the two luminal subtypes is that, in general, luminal B has lower expression of ERrelated genes and higher expression of proliferative genes [6,8,10]. Luminal-B tumors also demonstrate increased expression of growth receptor signaling genes, although only 10% of tumors were HER2-positive by immunohistochemistry [11]. A review of several gene expression studies noted that approximately 20% of luminal-B breast cancers were HER2-positive by immunohistochemistry [5]. Since HER2-positive breast cancers are treated very diff erently from HER2-negative breast cancers, a clinically meaningful classifi er of luminal-B breast cancer should not include HER2-positive breast cancers [9]. Approximately 30% of HER2-positive tumors defi ned by immunohistochemistry are assigned to the luminal-B subtype. Most of the tumors are also ER-positive by immunohistochemistry or ESR1 gene expression [6,9]. Th e clinical relevance of whether an ER-positive breast cancer with overexpression of HER2 is classifi ed as HER2-positive or as luminal B by the intrinsic molecular classifi cation remains to be determined.
In many subsequent studies, luminal-B breast cancer has been defi ned as ER-positive breast cancer with increased proliferation [5]. In gene expression studies, proliferation genes such as CCNB1, MKI67 and MYBL2 are more highly expressed in luminal-B compared with luminal-A subtypes [12], correlating with a higher proportion of histological grade III also observed in luminal-B cancers [9].
Since the seminal paper of Perou and colleagues fi rst identifi ed the intrinsic molecular subtypes of breast cancer, there have been various single subtype predictors (SSPs) that have been developed to identify the molecular subtype of an individual breast cancer [6,13,14]. Th ese SSPs diff er in the intrinsic gene list that is used to defi ne molecular classifi cation. Recently, the reproducibility of subtype assignment across these three SSPs was Relative insensitivity to endocrine therapy compared with luminal-A subtype evaluated by retrieving expression data from three publicly available datasets involving nearly 800 patients and performing two-way average-linkage hierarchical cluster analysis using fi ve distinct intrinsic gene lists. Whilst the basal-like and HER2 subtypes could be reproducibly identifi ed by independent observers, none of the classifi cation systems could produce substantial agreement in subdividing luminal cancers [15]. A similar study by Weigelt and colleagues produced similar conclusions [16]. Although this lack of agreement is troublesome, it is perhaps not surprising as the initial molecular classifi cation was based upon only 42 individuals with breast cancer [7].
Proliferation has been consistently identifi ed as the most important feature of several prognostic multigene signatures, including the intrinsic molecular classifi cation [5,17]. In ER-positive/HER2-negative tumors, prolifera tion is the strongest predictor of early relapse risk that diff eren tiates high-risk luminal-B tumors from low-risk luminal-A tumors [5,18]. Whilst ER is bimodally expressed (meaning the overwhelming majority of cases are either completely ER-positive or unambiguously ERnegative) in breast cancer, thus allowing a meaningful cut-off point to be applied [19], proliferation-related genes are expressed along a unimodal continuum. Th is makes it extremely diffi cult to apply any meaningful cutoff point that diff erentiates between high and low proliferative tumors in a reproducible manner. Th is is evident in the diff erences in subtype assignment between luminal-B and luminal-A tumors across SSPs, where tumors with a level of proliferation around the median value may be inconsis tently classifi ed by SSPs that use diff erent proliferation-driven intrinsic gene lists.

Clinical features of luminal-B breast cancer
Since the earliest studies of the intrinsic molecular subtypes in breast cancer, the defi ning feature of luminal-B breast cancer has been its poor outcome compared with the luminal-A subtype [7]. Overall survival in untreated luminal-B breast cancer is similar to the basallike and HER2-positive subgroups, which are widely recognized as high risk. One study used a 50-gene classifi er to assign intrinsic subtypes to 761 untreated breast cancer patients, and correlated subtype with outcome. In a multivariate analysis of untreated early breast cancer, using the luminal-A subtype as a reference, luminal-B breast cancers were demonstrated to have a hazard ratio of 2.43 (P <0.0001) for relapse-free survival (RFS), similar to hazard ratios for erbB2/HER2 amplifi ed (2.53, P = 0.00012) tumors [13].
Th e increased relapse risk associated with the luminal-B phenotype appears to be limited to the early period after surgery. Since increased proliferation is the hallmark of luminal-B cancer [20], it is not surprising that increased relapse rates observed in luminal-B tumors are limited to the fi rst 5 years after diagnosis, with no diff erence in distant relapse beyond 5 years [21]. In a series of 831 untreated node-negative breast cancers, curated from fi ve publicly available gene expression datasets, we found the hazard ratio for distant metastases of luminal-B subtype compared with luminal-A subtype to be 2.86 (P <0.01) for early metastases (<5 years) and 0.65 (P = nonsignifi cant) for late metastases (≥5 years) ( Table 2).
Th ere are diff erences in the anatomic sites of relapse according to molecular subtypes. Th e increased incidence of brain metastases in HER2-positive and basallike breast cancer is well recognized [22,23]. Luminal breast cancers appear to have a predilection for metastasis to bone and pleura. In a small study of 81 patients with metastatic breast cancer, no diff erences in sites of metastasis were observed between luminal-B and luminal-A breast cancers [24].
Several studies have suggested luminal-B breast cancer is relatively insensitive to endocrine therapy compared with luminal-A breast cancer, and to chemotherapy compared with HER2-enriched and basal-like breast cancers. Five studies examined the pathological complete response (pCR) rate following preoperative chemotherapy according to molecular subtype. Table 3 compares the pCR for each molecular subtype in each study and demonstrates that the pCR rate is consistently lower in luminal-B breast cancer when compared with HER2 and basal-like subtypes [25][26][27][28][29]. Th ere were important methodological diff erences in these studies, including the method of subtype defi nition of the luminal-B subgroup (particularly the inclusion of ER-positive and HER2positive breast cancer in some studies), and diff erences in chemotherapy received.
Although luminal-B tumors are characterized by high proliferation, the likelihood of achieving pCR with preoperative chemotherapy is exceedingly low. In other high-risk breast cancer subtypes, pCR is a robust surrogate endpoint for disease-free survival and overall survival [30,31]. It is not clear whether pCR is a meaningful surrogate endpoint in luminal tumors. Paradoxically, large, low proliferative ER-positive tumors categorized with a low recurrence score by the OncotypeDx™ assay that fail to achieve pCR with preoperative chemotherapy experience excellent long-term survival [32]. In this study, there was no diff erence in long-term outcome for low recurrence score tumors that achieved pCR with preoperative chemotherapy compared with tumors in which there is residual invasive disease, although there were few low recurrence score tumors in this study that achieved pCR.
Response to endocrine therapy in the preoperative setting has also been explored as a surrogate marker for long-term outcomes. In ER-positive tumors, the level of residual proliferation after 10 to 14 days of preoperative endocrine therapy is prognostic for long-term RFS [33]. A correlative substudy of the IMPACT trial analyzed 158 patients with paired biopsies (baseline and 2 weeks after endocrine therapy), and found that the absolute value of residual proliferation after short-term endocrine therapy, as assessed by the percentage of Ki-67 immunostaining, was strongly predictive of RFS; Ki-67 index <2.7% was asso ciated with favorable RFS. Interestingly, the Ki-67 index measured after 10 to 14 days of endocrine therapy was more predictive of long-term RFS than the pre treatment Ki-67 index [34]. pCR after preoperative endocrine therapy is rare [33,[35][36][37]. Whether clinical or radiographic response to preoperative endocrine therapy is predictive of long-term outcome in ER-positive disease is not fi rmly established.
Despite the problems with subtype classifi cation, the luminal-B subtype remains a clinically important classi fication of breast cancer with prognostic and potential predictive implications. Weigelt and colleagues suggest that standardized methods and defi nitions for identi fication of breast cancer molecular subtypes are necessary to incorporate molecular subtype classifi cation into routine clinical practice [16]. HER2 and basal-like subtypes can already be identifi ed using fl uorescence in situ hybridization and immunostaining for ER, PR and HER2. With regard to diff erentiating between luminal-A and luminal-B subtypes, various authors have tried to defi ne more pragmatic criteria that can broadly be applied to clinical practice. Some studies have used the level of ER expression to diff erentiate luminal-B from luminal-A subtypes [27], but this does not take into account the level of proliferation.   [29] clustering (included some HER2 + ) ER, estrogen receptor; IHC, immunohistochemistry. a In Rouzier and colleagues [29], luminal-A and luminal-B subtypes were grouped together as the luminal subtype.
One study explored the use of the Ki-67 index as a potential unidimensional proliferation marker that could successfully diff erentiate luminal-B tumors from luminal-A tumors in a clinically practical way [12]. Subtypes were assigned for a cohort of 357 breast cancers using microarray-based gene expression profi ling, and the Ki-67 status, hormone receptor status and HER2 status were concurrently determined by immunohistochemistry. Th e authors used receiver operating characteristic curves to determine the Ki-67 cut-off point that distinguished luminal-A from luminal-B tumors, then applied it to an independent microarray series of 4,046 breast cancers. Th ey were able to successfully demonstrate using immuno histochemistry that determining the Ki-67 index can distinguish between the two subtypes [12]. However, arbitrarily applying a clinically relevant cut-off point to a continuous variable such as Ki-67 that is unimodally distributed is problem atic. Ki-67 immunohistochemistry is also limited by low reproducibility between laboratories, ongoing debate over both the optimal antibody for testing and the method for cell counting (manual versus automated), in addition to potential problems resulting from tumor heterogeneity [38]. Multigene prognostic assays, such as OncotypeDx™, are currently used to assess proliferation in providing independent prognostic information in early breast cancer [39]. Given their level of reproducibility and less potential for infl uence by tumor heterogeneity, these assays may have potential advantages over a uni dimensional marker such as Ki-67 in assigning subtype classifi cation.

Insulin-like growth factor signaling
In cancer, ligand activation of the insulin-like growth factor 1 receptor (IGF-1R) and its downstream pathways (phosphoinositide 3-kinase (PI3K)/AKT/mammalian target of rapamycin (mTOR) and Ras/Raf/MEK/ERK) stimulates tumor proliferation, survival, transformation, meta stasis and angiogenesis [40]. Laboratory and epidemio logical studies have demonstrated a link between cancer and insulin-like growth factor 1 (IGF-1) signaling [41]. IGF-1R is expressed in the majority of breast cancer (90 to 95%) and is often co-expressed with ER [42]. Cross-talk between ER and IGF-1R plays a critical role in tamoxifen resistance. Increased circulating plasma levels of IGF-1, a ligand for IGF-1R, identify women at increased risk of relapse on adjuvant tamoxifen [41]. Activation of IGF-1R signaling is associated with loss of PR expression, which itself is associated with high proliferative luminal-B breast cancer, and with resistance to tamoxifeninduced apoptosis [43]. Creighton and colleagues derived a signature of more than 800 genes whose expression was signifi cantly altered after exogenous IGF-1 stimulation in ER-positive MCF7 cells [44]. Activation of this IGF-1 signature was seen in approximately 25% of ER-positive breast cancers and was associated with an increased risk of recurrence. Similar fi ndings were reported by Ignatiadis and colleagues with their 142-gene in silico signature of IGF-1 activation that was more commonly found in luminal-B tumors compared with luminal-A tumors [45].
Inhibition of IGF-1R signaling demonstrates synergistic activity in combination with endocrine therapy in preclinical models of ER-positive breast cancer [43,46]. Various approaches to interrupting the IGF-1 signaling axis have been developed. Although potential targets include growth hormone and growth-hormone-releasing hormone, the most advanced therapeutic approach has been the development of antibodies against IGF-1R that block IGF-1 ligand-mediated activation and smallmolecule inhibitors of the IGF-1R tyrosine kinase domain [47]. Table 4 outlines IGF-1R-targeted therapies that are being investigated in ER-positive breast cancer and other solid tumors. Th e fi rst preliminary report of a randomized phase II trial of exemestane or fulvestrant and AMG-479, a fully human monoclonal antibody against IGF-1R, or matching placebo was presented at the 2010 San Antonio Breast Cancer Symposium [48]. Th is study involved 156 patients with ER-positive metastatic breast cancer who had progressed following fi rst-line endocrine therapy for advanced disease or who had relapsed within 12 months of completing adjuvant endocrine therapy. Th is study failed to meet its primary endpoint, as the addition of AMG-479 did not improve progression-free survival. Th e median progression-free survival for AMG-479 + endocrine therapy was 3.9 months, versus 5.7 months for placebo + endocrine therapy (hazard ratio = 1.17, P = 0.44). Correlative studies of this trial and other ongoing studies will be essential to determine whether there is a signal of activity for IGF-1R inhibition in patients with luminal-B-like features, such as increased proliferation measured by Ki-67 immunostaining.

Fibroblast growth factor signaling
Th e fi broblast growth factor (FGF) signaling system includes 22 ligands and four receptors [49], and is a highly complex growth factor signaling pathway that is responsible for many functions, including cell proliferation, survival and migration, through diff ering downstream molecules or pathways [49]. Multiple studies indicate that FGF may also be involved in angiogenesis. One study demonstrated that the ligand FGF2 stimulates migration and proliferation of endothelial cells, whilst another study demonstrated that, under anti-vascular endothelial growth factor therapy, tumor angiogenesis can switch from vascular endothelial growth factor receptor dependence to fi broblast growth factor receptor (FGFR) dependence via upregulating FGF2, possibly explaining resistance to vascular endothelial growth factor-targeted agents [50,51]. Whether it is through cell proliferation, survival, migration or angiogenesis, the FGF pathway clearly has oncogenic roles in many cancers. Th ese roles occur through various genetic aberra tions that include amplifi cations, activating mutations, chromosomal translocations, SNPs and aberrant splicing at the post-transcriptional level.
In breast cancer, changes to FGF signaling are considered important for oncogenesis, mainly through amplifi cation of FGFR1 and FGFR2. Following erbB2/ HER2, FGFR1 is amongst the most commonly amplifi ed genes in breast cancer, present in up to 10% of all breast cancers [52]. Various reports have shown that FGFR1 amplifi cation is most commonly associated with ER expres sion, the absence of HER2 overexpression and lobular histology [53,54]. Additionally, the FGFR2 gene is amplifi ed in approximately 1 to 2% of breast cancers [52]. Genome-wide association studies have also shown that inherited SNPs in the FGFR2 gene are associated with an increased risk of developing ER-positive breast cancer, probably through an increase in FGFR2 transcription [52]. Although activating mutations in FGFR3 and FGFR4 occur in many types of human tumors, they seem to be rare in breast cancer [49].
Recent data suggest that the luminal-B subtype is enriched for FGFR1 gene amplifi cation [52]. One study examined tumors from two independent series of breast cancer for FGFR1 amplifi cation, demonstrating that FGFR1-amplifi ed cancers are frequently PR-negative, have a high proliferative rate assessed by Ki-67 immunostaining and are present in 16 to 27% of luminal-B breast cancer [53]. Furthermore, the same study demonstrated that FGFR1-amplifi ed breast cancer cell lines have both enhanced ligand-dependent and ligand-independent signaling, and are dependent upon FGFR signaling for anchorage-independent growth [53]. Th ese authors also demonstrated that FGFR1-amplifi ed cells were resistant to endocrine therapy, but this could be reversed by knockdown of FGFR1 [53]. Other studies have also observed that resistance to endo crine therapy can be reversed through both knockdown of FGFR1 expression and the use of a small molecule FGFR tyrosine kinase inhibitor [55]. Th ese fi ndings all suggest that the FGF pathway, and more specifi cally FGFR1 gene amplifi cation, may be a major contributor to the poor prognosis observed in luminal-B breast cancer, through increased proliferation and resistance to endo crine therapy. Preclinical models of breast cancer cells amplifi ed for FGFR1 or FGFR2 have demonstrated sensitivity to inhibition of FGFR [49]. Th is has led to several proof-ofconcept early-phase clinical trials using FGFR inhibitors. Several antibodies and small-molecule inhibitors of FGFR are currently in clinical testing. First-generation tyrosine kinase inhibitors also inhibit VEGFR2 due to structural similarity between the two tyrosine kinase domains [52]. Table 4 lists some of the current agents targeting the FGF pathway in breast cancer clinical trials. An important challenge for all of these studies is the identifi cation of patients whose tumors harbor genetic amplifi cation of FGFR1 or FGFR2. Like erbB2/HER2, chromosome in situ hybridization and fl uorescence in situ hybridization are the dominant methods used to identify gene amplifi ca tion in paraffi n-embedded tumor samples.

Phosphoinositide 3-kinase signaling
Deregulated PI3K signaling has been implicated in many aspects of carcinogenesis [56]. Genetic aberrations along the pathway can occur anywhere from the upstream growth factor receptors to downstream target molecules, regulatory molecules and PI3K itself [56]. Th ese genetic aberrations have the potential to change a number of cell functions that contribute to the transformed phenotype, including cell growth and proliferation, diff erentiation, cell survival, adhesion and cell motility [56]. Subsequently, the PI3K pathway -including its enzymes, targets and regulators -is considered an important potential therapeutic target in cancer.
In breast cancer, the PI3K pathway is frequently activated. Amplifi cation of upstream receptors such as erbB2/HER2, loss of negative regulators such as PTEN, amplifi cation of downstream targets such as Akt, and activating mutations or genetic amplifi cation of the αcatalytic subunit of PI3K (PIK3CA) have all been described in breast cancer. PIK3CA somatic mutations occur in approximately 25% of breast cancer [57,58]. In luminal-B breast cancer, the role of PI3K signaling is being defi ned. Th ere appear to be no diff erences in the frequency of PIK3CA mutation between luminal-A and luminal-B breast cancers [58]. A recent preclinical study has demon strated that increased expression of PI3K pathway genes is a feature of luminal-B breast cancer [59]. Growth inhibition induced by endocrine therapy in luminal-B breast cancer cell lines could be signifi cantly increased by adding a selective PI3K inhibitor, suggesting that PI3K inhibitors may have a role in luminal-B breast cancer [59].
In breast cancer, the initial studies targeting the PI3K pathway involved rapamycin analogs or mTOR inhibitors. A phase II study of ER-positive breast cancer in the neoadjuvant setting compared 4 months of letrozole treatment with 4 months of letrozole and everolimus treat ment, an oral mTOR inhibitor [36]. Although the rate of sonographic response was only marginally improved with the addition of everolimus to letrozole (68% vs. 59%, P = 0.062), there was a much greater improvement in antiproliferative response, defi ned as day 15 Ki-67 immuno staining <2.7% (57% vs. 30%, P <0.01). Th e authors also noted that the rate of anti-proliferative response in the everolimus and letrozole arm was higher in tumors with PIK3CA mutations [36].
Recently, a negative intracellular signaling feedback loop between the mTOR complex 1 and the IGF-1 signaling axis has been discovered [60]. Intracellular levels of IRS1, a key mediator of IGF-1R signaling, are increased when mTOR complex 1 is inhibited by everolimus and other similar mTOR inhibitors, leading to paradoxical activation of Akt [61]. Preclinical models suggest robust activity for dual IGF-1R and mTOR inhibition [62,63]. A recent phase I study demonstrated that this combined therapy may be eff ective in breast cancer, where fi ve out of 23 breast cancer patients had either partial response, prolonged stable disease or partial metabolic response [64]. Th e combination appeared particularly active in luminal-B like breast cancer -defi ned as ER-positive with Ki-67 immuno staining ≥15% -as three out of 10 patients in this trial, all of whom were heavily pretreated, achieved a partial response by the Response Evaluation Criteria in Solid Tumors.
Th ere is great enthusiasm for highly specifi c PI3K inhibitors that are currently in early development. Th ey range from pure pan-PI3K inhibitors to isoform-specifi c PI3K inhibitors, to dual inhibitors of both PI3K and mTOR, to Akt inhibitors. Currently, most early-phase studies using these agents attempt to select patients with genetic aberrations in this pathway. A recent preclinical study, however, suggested that PIK3CA mutation, but not PTEN loss, confers sensitivity to mTOR inhibitors [65]. In the clinical setting, one study retrospectively correlated the PIK3CA mutation status of patients with response rates from PI3K/Akt/mTOR inhibition. Th ese authors observed higher responses in the PIK3CA mutant population (35% partial responses) compared with the PIK3CA wild-type population (6% partial responses) [66]. Targeting the PI3K pathway appears very promising, although more extensive study is required -particularly in identifying patients who will benefi t. Novel agents targeting this pathway are listed in Table 4.

Other potential targets
Cyclin D1 is amplifi ed in approximately 10% of breast cancer and is known to have a role in driving proliferation through its interaction with cyclin-dependent kinases such as CDK4 [67]; subsequently, it has been suggested that cyclin-directed therapies may have a role in luminal-B subtypes, where proliferation is an important factor [67]. Cyclin inhibitors are currently in early-phase development.
Recent preclinical research has identifi ed a potential breast cancer oncogene, ZNF703, implicated in the luminal-B subtype [68]. In this study ZNF703 was signifi cantly amplifi ed in luminal-B tumors, and its overexpression was associated with poor clinical outcome [68]. In cell lines, overexpression of ZNF703 induced cell proliferation independent of estradiol stimulation [68]. Th e investigators also observed that ZNF703 is ER regulated and may have a role in cancer stem cell selfrenewal [68], suggesting a potential role for ZNF703 inhibition in luminal-B breast cancer.
Another recent study has indicated an association between luminal-B tumors and overexpression of the scaff old protein NHERF1 (sodium-hydrogen exchanger regulatory factor 1). NHERF1 expression is associated with poorer survival and resistance to endocrine therapy in ER-positive breast cancer [69]. Further study is required to determine whether NHERF1 is an appropriate candidate for targeted therapy.

Conclusion
Gene expression studies have led to the identifi cation of luminal-B breast cancer, a subtype of ER-positive breast cancer defi ned by increased proliferation, relative resistance to chemotherapy compared with other highly proliferative breast cancers, and poor outcome with endocrine therapy. Assigning the luminal-B subtype to individual breast cancers has been problematic, however, as the robustness of single subtype classifi ers is suboptimal. Rather than approaching luminal-B cancer as a fi xed biological entity, it is more clinically useful to consider the luminal-B phenotype as a conceptual framework, recognizing that proliferation in ER-positive/ HER2-negative tumors exists along a continuum. Identifi cation of highly proliferative ER-positive/HER2negative tumors -whether through histological grading, the Ki-67 labeling index, or a multigene signature -is useful to separate aggressive luminal-B-like tumors with a risk of early relapse from more indolent luminal-A-like tumors that are adequately treated with endocrine therapy alone. In an eff ort to improve survival in luminal-B breast cancer, there has been a recent focus on particular molecular pathways where development of effi cacious therapeutic agents may alter the natural history of the disease. For these novel treatments to have their desired eff ect, however, additional work is needed to characterize the drivers of aggressive biology, and future trials should acknowledge the molecular heterogeneity of ER-positive breast cancer and separate the more indolent luminal-A breast cancers from their more proliferative luminal-B-like counterparts.

Competing interests
BT declares that he has no competing interests. PLB declares research funding from Novartis, BristolMyersSquibb, and GlaxoSmithKline, and consultancy funding from Roche, Sanofi -Aventis and Johnson and Johnson.