PTEN Loss Mediates Clinical Cross-Resistance to CDK4/6 and PI3Kα Inhibitors in Breast Cancer

Three different CDK4/6 inhibitors in combination with endocrine therapy are approved by the FDA as the standard of care for women with estrogen receptor (ER)–positive, HER2-negative advanced breast cancer (1–3). CDK4/6 forms a complex with D-cyclins and phosphorylates RB, thereby inactivating it and releasing cells to progress through the G₁ checkpoint of the cell cycle. ER modulates cyclin D1 expression, and cyclin D1 in turn increases ER transcriptional activity, likely reinforcing the unique dependence of ER-positive cancer cells on cyclin D1 to initiate the G₁ to S-phase transition (4). Despite markedly improved clinical outcomes with this combination, acquired resistance emerges similarly to other targeted therapies (5, 6).

The combination of CDK4/6 inhibitors and antiestrogen therapy has been tested in a phase Ib trial using ribociclib and the aromatase inhibitor letrozole in postmenopausal women with ER-positive, HER2-negative advanced breast cancer (NCT01872260; ref. 7). Of the 7 patients treated on this protocol at our institution, we were able to collect paired pretreatment and progression biopsies from 5 patients (Table 1). To understand reasons for acquired resistance to ribociclib, all the collected samples were analyzed for RB expression, because RB loss can confer resistance to CDK4/6 inhibition in preclinical models (8). Loss of RB expression at the time of progression was observed in patients 1, 2, and 3, whereas RB expression was preserved in patients 4 and 5 (Table 1; Fig. 1A–C; Supplementary Fig. S1). Thus, in agreement with preclinical models (8) and circulating tumor DNA analyses (9, 10), RB loss does appear to drive resistance to CDK4/6 inhibitors in the clinic.

Notably, patient 1 was subsequently treated with the combination of selective ER degrader fulvestrant with a beta-sparing PI3K inhibitor taselisib on the clinical trial NCT01296555, but developed rapidly progressive disease after only one cycle of therapy (Fig. 1A). Because PTEN loss mediates clinical resistance to PI3Kα inhibitors (6), we tested whether PTEN expression was also lost during the treatment with ribociclib to explain the primary resistance to PI3Kα inhibition. Of note, to analyze PTEN and RB expression, we utilized IHC because multiple studies have shown that their loss can occur through epigenetic mechanisms. To our surprise, tumor biopsies from patient 1 (acquired at the time of resistance to ribociclib but prior to the treatment with taselisib) revealed acquired concomitant loss of PTEN and RB, raising the possibility that both alterations may contribute to acquired resistance to CDK4/6 inhibition (Fig. 1A). In support of this hypothesis, analysis of the progressing lesion in patient 4 revealed discrete PTEN loss (Fig. 1C), suggesting that PTEN loss alone confers resistance to CDK4/6 inhibition. Patients 2, 3, and 5 had preserved PTEN expression in their biopsies at the time of progression (Fig. 1B; Supplementary Fig. S1A and S1B). Only patient 5 showed loss of PTEN at baseline in the lymph node lesion (Supplementary Fig. S1B). Consistent with our hypothesis, this lesion, which responded to the prior line of therapy with taselisib (Supplementary Fig. S1B, CT images on the top left), did not reduce in size during the ribociclib-based therapy (Supplementary Fig. S1B, CT images on the top right), suggesting that PTEN loss can drive both primary and acquired resistance to CDK4/6 inhibitors. In contrast, this patient had a liver lesion that initially responded to a much higher extent (and was the primary reason for the patient to be deemed to have a partial response per RECIST of −45.3%), but ultimately progressed and was sampled at the end of treatment (Supplementary Fig. S1B, CT images on the bottom). At the time of progression of this lesion, both PTEN and RB were intact, indicating that other mechanisms can contribute to resistance to CDK4/6 inhibitors. Overall, we observed acquired loss of either RB or PTEN or both as potential drivers of resistance to ribociclib in 4 of 5 analyzed cases (Table 1). In addition, for patients 3 and 4, rapid autopsies were performed in an effort to understand intratumoral heterogeneity of their refractory disease. This revealed RB loss in 17 of 18 sampled lesions for patient 3 (Fig. 2A) and PTEN loss in 7 of 10 collected lesions for patient 4 (Fig. 2B). Thus, multiple lesions revealed loss of PTEN as a potential mechanism of resistance. High prevalence of lesions that lack PTEN or RB expression among all progressive lesions is consistent with the preexisting nature of RB- or PTEN-deficient cells with their rapid selection under treatment pressure.

Importantly, although RB and PTEN loss appeared in the context of combination therapy with ribociclib and letrozole, all studied patients had previously progressed on an aromatase inhibitor (Table 1), suggesting, although not proving, that selection for cells harboring the loss of RB or PTEN was primarily driven by ribociclib.

We aimed to determine whether PTEN loss is a bona fide mechanism of resistance to CDK4/6 inhibitors, because the effect of RB loss has already been evaluated in other preclinical studies (8). We generated CRISPR-based PTEN knockout T47D cells, an ER-positive breast cancer model. Multiple clones for each guide (named A and B) showed complete loss of PTEN expression (Supplementary Fig. S2A). As expected, PTEN-deficient cells had increased levels of phosphorylated AKT (Supplementary Fig. S2A) and were resistant to the p110α-selective inhibitor BYL719 (Supplementary Fig. S2B; ref. 6). Of note, because different clones displayed a slightly different proliferation rate (Supplementary Fig. S2C) and this could potentially affect their sensitivity to cell-cycle inhibitors in viability assays (11), the duration of cell-growth assays were performed so that each of the untreated clones (and untreated parental control) underwent the same number of cell doublings. All PTEN-null clones showed decreased sensitivity to two distinct CDK4/6 inhibitors (ribociclib and palbociclib) in comparison with CAS9 control cells (Fig. 3A; Supplementary Fig. S2C). Results were confirmed in a second ER-positive breast cancer model, MCF7 (Fig. 3B; Supplementary Fig. S2A and S2C). In addition, ectopic expression of the wild-type (WT) form of PTEN, but not of a catalytically inactive version, increased the sensitivity to CDK4/6 inhibitors of a PTEN-deficient breast cancer model (Fig. 3C; Supplementary Fig. S2A). Moreover, T47D PTEN-null cells were also more resistant to the combination of CDK4/6 inhibitors with fulvestrant (Fig. 3D) or when treated with CDK4/6 inhibitors in the absence of estrogen (Supplementary Fig. S2D). In accordance with in vitro results, PTEN-deficient T47D xenograft models were resistant to ribociclib in vivo (Fig. 3E). Overall, these data suggest that loss of PTEN expression alone reduces sensitivity to multiple CDK4/6 inhibitors.

We next investigated the mechanism by which PTEN loss promotes resistance to CDK 4/6 inhibitors. Cell-cycle analysis revealed that CAS9 control in both T47D and MCF7 cells underwent a dramatic arrest in G₁ induced after 24 hours of treatment with CDK4/6 inhibitors (Fig. 3F; Supplementary Fig. S2E–S2G). For T47D, a modest increase in the number of cycling cells was observed at 72 hours, as previously reported (8). The isogenic PTEN-deficient cells initially showed a similar strong arrest in G₁ after 24 hours of treatment; however, unlike the control cells, a large fraction of the PTEN-deficient cells reentered S-phase by 48 hours (Fig. 3F; Supplementary Fig. S2F and S2G). In CAS9 control cells, CDK4/6 inhibition resulted in decreased RB phosphorylation and expression of multiple cyclins (such as cyclin D3 and E2) that are normally regulated by RB. In contrast, CDK4/6 inhibition was correlated with increased expression of cyclin D1 and E1, likely reflecting a stabilization of these proteins in cells arrested in G₁ (8, 12; Fig. 3G; Supplementary Fig. S2E–S2G). In contrast, PTEN-null cells initially showed a decrease in pRB (24 hours) but rephosphorylation of RB was clearly apparent after 48 hours (Fig. 3G; Supplementary Fig. S2E–S2G). To differentiate whether the increased phosphorylation of RB was either a cause or a consequence of PTEN-deficient cells rapidly reentering the cell cycle, a dose–response experiment with ribociclib was performed (Fig. 4A). In PTEN-deficient T47D cells, a substantially higher dose of ribociclib was necessary to suppress RB phosphorylation. Interestingly, 6 μmol/L of ribociclib reduced pRB in PTEN-deficient cells to the same extent as 1 μmol/L in control cells (Fig. 4A). This higher dose also completely blocked these PTEN-deficient cells from reentering the cell cycle (Fig. 4B). The same concentration of drug (6 μmol/L) was unable to induce cell-cycle arrest in T47D cells in which RB was knocked down (Fig. 4C), indicating that at this concentration the effect of ribociclib is still mediated by on-target suppression of RB. This finding also underscores that PTEN loss leads to a relative decrease in potency of ribociclib, but RB loss causes complete resistance. However, at clinically relevant doses of ribociclib (1 μmol/L; ref. 13), PTEN-null (Fig. 3A) and RB knockdown (Fig. 4D) cells showed a comparable level of resistance to ribociclib. Altogether, these data suggest that the loss of PTEN expression is sufficient to reduce sensitivity to CDK4/6 inhibitors by maintaining RB phosphorylation and thereby alleviating cell-cycle arrest.

In PTEN-deficient cells, the addition of MK2206, an allosteric AKT inhibitor, restored the capacity of CDK4/6 inhibitors to reduce RB phosphorylation (Supplementary Fig. S3A), demonstrating that increased AKT activity is a critical mediator of the resistance induced by PTEN loss. Accordingly, the combination of CDK4/6 and two distinct AKT inhibitors resulted in cell-cycle block in G₁ (Fig. 4E; Supplementary Fig. S3B and S3C) and decreased cell proliferation (Fig. 4F; Supplementary Fig. S3D–S3F). The effect of the combination was comparable to that achieved with single-agent CDK4/6 inhibitor in the Cas9 control cells. In addition, the combination of MK2206 and ribociclib induced tumor regression in the PTEN-deficient T47D xenograft model (Fig. 4G). As observed previously, the combination of CDK4/6 and fulvestrant was less effective at inhibiting the proliferation of PTEN-deficient cells (Fig. 3D; Supplementary Fig. S3G). However, the triple combination of CDK4/6 and AKT inhibitors with fulvestrant was sufficient to suppress proliferation of PTEN-deficient cells. Importantly, both the CDK4/6 and AKT inhibitors' doses could be lowered and still yield efficacy in the triple combination, compared with the doses of single-agent CDK4/6 or AKT inhibitors required for efficacy in combination with fulvestrant (Supplementary Fig. S3G). Overall, inhibiting AKT effectively reverses the effect of PTEN loss on resistance to CDK4/6 inhibitors and restores their capacity to suppress RB phosphorylation and cell-cycle progression.

We next aimed to understand how increased AKT activation leads to resistance to CDK4/6 inhibitors. Previous studies in breast cancer have shown that AKT can induce cell-cycle progression by activating CDK2, which in turn phosphorylates RB during the G₁ to S-phase transition (4). AKT directly phosphorylates threonine 157 of p27, an inhibitor of CDK2. Phosphorylation of p27 impairs its nuclear import, thereby blocking access to CDK2 and resulting in increased CDK2 activation (14–16). We therefore initially tested whether p27 localization was affected by PTEN loss in the paired biopsies (Fig. 5A). Although loss of RB did not alter p27 localization (patient 2), a significant decrease in p27 nuclear staining was observed in the lesions that did not express PTEN (patients 4 and 5). These results are consistent with the hypothesis of PTEN loss causing decreased p27 expression in the nucleus.

To explore whether modulation of p27 localization could affect the resistance to CDK4/6 inhibitors, both T47D and MCF7 cells were transduced with either WT or a mutant form of p27 (T157A) that is unable to be phosphorylated by AKT (Supplementary Fig. S4A). Consistent with increased levels of AKT, PTEN-deficient cells showed decreased localization of WT p27 in the nucleus in comparison with control cells (Supplementary Fig. S4B). However, the T157A mutant p27 displayed a more prominent nuclear localization in PTEN-deficient cells, comparable to WT p27 in control cells (Supplementary Fig. S4C). The expression of the T157A-p27 in PTEN-deficient cells restored the capacity of ribociclib to suppress RB phosphorylation and cell growth (Fig. 5B and C; Supplementary Fig. S4D–S4F), indicating that relocalization of p27 in the nucleus is sufficient to partially resensitize PTEN-deficient cells to CDK4/6 inhibition. In support of the hypothesis of PTEN loss causing p27-mediated increased activity of CDK2, in T47D cells we observed a decreased association of p27 with CDK2 in the absence of PTEN (Fig. 5D), also in line with previously reported findings (17). Moreover, PTEN-deficient T47D cells demonstrated an enhanced association of CDK2 with cyclin D3 and cyclin E1 before and after 24-hour treatment with ribociclib (Fig. 5E; Supplementary Fig. S4G). This was likely due to an upregulation in the total levels of expression of cyclins D3 and E1, in line with previous reports showing that AKT activation can increase cyclin D3 protein stability and cyclin E1 expression (17, 18). Of note, we compared cells after 24 hours of treatment, when all cells were equally arrested in G₁ (Fig. 3F), and thus the expression levels of cyclins were not a mere consequence of the cells being in a different phase of the cell cycle. Consistently, although only approximately 50% suppression of CDK2 via RNAi was achieved, this was sufficient to partially resensitize the PTEN-null T47D and MCF7 cells to ribociclib with respect to RB phosphorylation and progression into the S-phase of the cell cycle (Supplementary Fig. S4H–S4M).

Although p27 was initially identified as a CDK2 inhibitor, studies have shown that it also promotes assembly of the cyclin D–CDK4 complex (19). When AKT phosphorylates p27, this leads to increased cytoplasmic localization and binding to cyclin D1–CDK4 (but not CDK2) and promotes the assembly and/or stabilization of this ternary complex (20). Consistently, an increased association of CDK4 with p27 in the absence of PTEN was observed (Supplementary Fig. S5A). To determine whether PTEN loss affects cyclin D1–CDK4 complex formation, immunoprecipitation experiments were performed. In PTEN-deficient cells, we observed an increase in association between CDK4 and cyclin D1 (after 24 hours of treatment with ribociclib) and D3 (at baseline; Fig. 5F; Supplementary Fig. S5B). Although increased complex formation with cyclin D3 may simply be a consequence of increased cyclin D3 expression levels, cyclin D1 levels were not increased in PTEN-deficient cells, further supporting the notion that PTEN loss promotes the formation and/or stabilization of the CDK4/cyclin D1 complex. Notably, whereas downregulation of cyclin D1 resensitized PTEN-deficient T47D and MCF7 cells to ribociclib (Fig. 5G; Supplementary Fig. S5C and S5D), downregulation of cyclin D3 showed a more modest and consistent effect only in MCF7 cells (Fig. 5H; Supplementary Fig. S5E and S5F), suggesting that the relative contribution of the cyclins D in mediating CDK4 activity might be different among breast cancer cell models. In addition, siRNA was used to decrease the expression of CDK4/6, and although only a modest decrease in their expression levels was achieved, this partially resensitized the PTEN-deficient cells to ribociclib (Supplementary Fig. S6A–S6C). These data suggest that in PTEN-null cells, CDK4/6 enzymes are still partially active despite the presence of clinically relevant doses of ribociclib, and contribute to RB phosphorylation and progression through the cell cycle. This is also consistent with the finding that higher doses of ribociclib were necessary to fully suppress RB phosphorylation in the PTEN-deficient cancer cells (Fig. 4A). Overall, these mechanistic data indicate that loss of PTEN, through increased AKT activity, induces delocalization of p27 outside the nucleus and therefore causes increased activity of both CDK4/6 and CDK2, which together contribute to overcome the blockade in G₁ induced by CDK4/6 inhibitors.

In this study, we observed that RB and PTEN loss are both mechanisms of resistance to CDK4/6 inhibitors in ER-positive metastatic breast cancer. Although CDK4/6 inhibitors are commonly used in patients with advanced ER-positive breast cancer, the PI3Kα-selective inhibitor alpelisib was only recently approved as a second line of treatment for PIK3CA-mutated ER-positive breast cancer (21). Significantly, previous studies established that PTEN loss also causes resistance to alpelisib (6). These partially overlapping resistance mechanisms (i.e., PTEN loss for both CDK4/6 and PI3Kα inhibitors; RB loss just for CDK4/6 inhibitors) have potential clinical implications, as these drugs are sequentially deployed in the management of patients with advanced breast cancer. Indeed, because PTEN loss is a mechanism of resistance to both PI3Kα and CDK4/6 inhibitors, a subset of patients progressing on CDK4/6 inhibitors harboring this alteration may have diminished response to PI3Kα inhibitors (see patient 1). In contrast, this study (Fig. 4D) and others (22) have shown that RB loss does not confer resistance to PI3Kα inhibitors in breast cancer models in vitro. Thus, cancers with this mechanism of resistance to CDK4/6 inhibitors may derive similar benefit from PI3Kα inhibitors as cancers that had not been exposed to CDK4/6 inhibitors. Notably, RB or PTEN expression levels are currently not being used for patient selection in second-line clinical trials with PI3Kα inhibitors, especially because both RB and PTEN loss are rare in treatment-naïve ER-positive tumors (23, 24). However, the findings in this study suggest that the prevalence of this alteration is likely to increase under selective pressure with CDK4/6 inhibitors and highlight the importance of investigating RB and PTEN expression in prospective clinical trials in the post-CDK4/6 setting. Of note, with a relatively small number of samples, we cannot determine the true prevalence of each of these alterations, and larger studies are needed to assess the generalizability of these findings. Nevertheless, for one of the five patients, the lesion at the time of progression retained both RB and PTEN expression, indicating that other mechanisms contribute to resistance to CDK4/6 inhibitors, as recent publications have demonstrated (10, 25–27).

To understand how loss of PTEN promotes resistance, we used gene-editing approaches to develop isogenic ER-positive breast cancer models. In PTEN-deficient cells, increased AKT activation mediates resistance to CDK4/6 inhibitors, and selective AKT inhibitors can restore sensitivity in vitro and in vivo. This mechanistic link to resistance is consistent with a recent study showing that increased AKT activation, in that case mediated by PDK1, is sufficient to drive resistance to ribociclib in breast cancer cells (28). Notably, four of five pretreatment biopsies (for one, patient data were not available) showed that these tumors harbor additional PI3K pathway alterations (either in PIK3CA or AKT1; Table 1), similar to the cell lines we utilized for in vitro studies. It is conceivable that clinical resistance to CDK4/6 inhibitors is achieved only through substantial activation of the PI3K–AKT axis, mediated by multiple alterations, such as PIK3CA or AKT1 and concomitant PTEN loss. However, future studies should clarify whether PTEN loss can cause resistance to CDK4/6 inhibitors also in tumors that do not carry additional mutations in the PI3K pathway. AKT activation itself may drive resistance by many potential mechanisms. Several groups have reported that AKT-dependent phosphorylation of p27 increases p27 cytoplasmic localization (14–16). This cytoplasmic retention of p27 may help drive resistance by at least two mechanisms. First, the AKT-dependent delocalization of p27 may reduce nuclear p27 availability to bind and inhibit cyclin E/CDK2, thus contributing to cyclin E/CDK2 activation (29). Accordingly, in our paired biopsies, we observed that loss of PTEN expression correlates with exclusion of p27 from the nucleus. Notably, PTEN-deficient cells also had a modest but consistent increased expression of cyclin E1, which may also partially contribute to increased CDK2 activity (25). Second, p27 can increase the assembly/stability of cyclin D1–CDK4 complexes in the cytoplasm (19). Indeed, we observed increased complex formation in the PTEN-deficient cells. It is possible, however, that the PTEN-mediated effect on CDK4–cyclin D complex might also be partially p27-independent. Of note, in the paired biopsies, despite observing loss of nuclear PTEN staining, we were not able to observe an accumulation of p27 in the cytoplasm when PTEN was lost. This might be due to multiple reasons, including a technical artifact related to the specific antibody we used. Indeed, this antibody has been generated by immunization with an epitope that contains the first 200 amino acids of p27, which includes the residue phosphorylated by AKT. It is thus conceivable that the phosphorylated form of p27 that accumulates in the cytoplasm in the absence of PTEN is not efficiently recognized by this antibody. Alternatively, we cannot exclude the possibility that the cytoplasmic fraction of p27 is less stable than the nuclear fraction in these needle biopsies.

The findings in this article suggest that PTEN loss causes diminished sensitivity to clinically relevant doses of CDK4/6 inhibitors by initiating signaling cascades that induce hyperactivation of cyclins/CDKs. This suggests that these resistant cancers still require RB inactivation for G₁-to-S progression. The fact that PTEN-null cells are still dependent on inactivating RB to progress through the cell cycle suggests that these cells could gain additional proliferative advantage by concomitant loss of RB, as observed in patient 1.

Previous preclinical studies have shown that PTEN loss also affects the activity of BRAF (30), EGFR (31), and immune checkpoint inhibitors (32) in other settings, suggesting that PTEN loss may emerge as a resistance mechanism in many therapeutic paradigms. Furthermore, this study provides a proof of concept of how a single genetic event can cause cross-resistance to multiple targeted therapies and affect subsequent lines of treatment in the same patient population.

For SpCas9 expression and generation of single guide RNA for human PTEN, two 20-nt target sequences were selected. A: CCAAATTTAATTGCAGAGGT; B: AGAGGCCCTAGATTTCTATG. Oligonucleotides were annealed and cloned into the BsmbI-BsmBI sites downstream from the human U6 promoter in LentiCRISPR v2 plasmid (Addgene, # 52961). The doxycycline-inducible system pINDUCER11 was used to express shRNA that targets RB1 at the sequence 5′-CAGAGATCGTGTATTGAGATT-3′. Stable T47D cell lines were generated by lentiviral transduction with the pINDUCER11-shRB1, followed by FACS for GFP expression. Plasmid expressing p27 cDNA with N-terminus–fused eGFP was purchased from GeneCopoeia (# EX-C0696-Lv182). Mutation of Threonine 157 to Alanine was introduced by site-directed mutagenesis using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs, #E0554S). The following oligonucleotide primers were used for mutagenesis: T157A-FW: 5′-GCGACCTGCAgCCGACGATTC-3′; T157A-RV: 5′-TTCCTTATTCCTGCGCATTGC-3′. PTEN-WT and PTEN-C124S constructs (Addgene, #10785 and #10931) were cloned in pTREX doxycycline-inducible vector kindly provided by Novartis.

T47D cells were cultured in RPMI-1640 with 10% FBS. EVSA-T and MCF7 cells were cultured in DMEM with 10% FBS. Cancer cell lines were obtained from the Center for Molecular Therapeutics at the Massachusetts General Hospital Cancer Center, which performs routine SNP and short tandem repeat authentication. In addition, cell lines underwent SNP validation at the completion of the project. Cell lines were routinely tested for Mycoplasma during experimental use. For in vitro experiments, the following drugs were used at the following concentrations (unless otherwise specified): ribociclib (1 μmol/L), palbociclib (250 nmol/L), MK2206 (1 μmol/L), GDC0068 (500 nmol/L), fulvestrant (10 nmol/L), BYL719 (1 μmol/L), all obtained from Selleckchem.

Antibodies against RB (#9309), pRB (Ser608; #8147), pRB (Ser780; #9307), pRB (Ser807/811; #8516), pAKT (T308; #4056), pAKT (S473; #4060), cyclin A2 (#4656), cyclin E1 (#4129), cyclin E2 (#4132), cyclin D1 (#2922), cyclin D3 (#2936), PTEN (#9552), CDK6 (#13331), CDK4 (#12790), CDK2 (#2546), p27 (#3686), Vinculin (#13901), and Actin (#4967) were purchased from Cell Signaling Technology. GAPDH antibody was purchased from Millipore. Cells were seeded into 6-well plates; 24 hours later, cells were transfected with ON-TARGETplus SMARTpool siRNA against CDK2, CDK4, CDK6, cyclin D1, and cyclin D3, or ON-TARGET plus Non-Targeting Pool as negative control (Dharmacon) using HiPerFect Transfection Reagent (Qiagen) according to the manufacturer's instructions. Transfected cells were cultured for 24 hours before drug treatment.

Cells were seeded and treated the following day with the indicated drugs for the specified period of time. After treatment, cells were washed twice with ice-cold PBS and homogenized in immunoprecipitation (IP) lysis buffer. Soluble cell lysates were collected after 13,500 rpm centrifuge at 4°C for 15 minutes. An equal amount of total protein was incubated with anti-CDK2 (#07-631, EMD Millipore; 1 μg antibody/200 μg lysate) or anti-p27 (#3686, Cell Signaling Technology; 0.4 μg antibody/200 μg lysate) rabbit antibody plus protein A-Sepharose beads (#17-0963-03, GE Healthcare), or anti-CDK4 mouse antibody (#sc-23896, Santa Cruz Biotechnology; 1 μg antibody/200 μg lysate) plus protein G-Sepharose beads (#17-0618-01, GE Healthcare) for more than 3 hours at 4°C. Beads were washed at least three times with IP lysis buffer after incubation and centrifuged at 4,000 rpm at 4°C for 3 minutes to remove supernatant. After washing, SDS loading buffer was added and boiled at 95°C for 10 minutes.

Viability assays were performed by plating cells in 96-well plates and treating them the following day with the indicated drug(s) for 5 to 8 days, according to the cell doubling. At the end of all experiments, cells were fixed and stained with Hoechst (1 μg/mL). Ninety-six–well plates were imaged with ImageXpress Micro (Molecular Devices) and nuclei counts were performed with MetaXpress program. For experiment with estrogen depletion, cells were cultured with Phenol red-free RPMI-1640 and 10% charcoal-stripped FBS for two days, after cells were seeded in the same medium with or without 0.2 nmol/L β-estradiol.

After drug treatment, cells were incubated with 20 μmol/L 5-ethynyl-2′-deoxyuridine (Life Technologies) for 2 hours, fixed with 3.7% formaldehyde for 15 minutes, blocked with 3% BSA for 5 minutes, permeabilized with 0.5% Triton X-100 in PBS for 30 minutes, and stained with ClickIT reaction kit (100 mmol/L Tris-HCl pH 7.5, 3 mmol/L CuSO₄, 50 mmol/L ascorbic acid, 2.5 μmol/L Alexa-647 azide, Life Technologies) for 30 minutes and 3 μmol/L DAPI (Life Technologies) in staining buffer (100 mmol/L Tris-HCl pH 7.5, 150 mmol/L NaCl, 1 mmol/L CaCl₂, 0.5 mmol/L MgCl₂, 0.1% NP-40) for 15 minutes. FACS analysis was performed with LSR II Flow Cytometer (BD Biosciences).

All mouse studies were conducted through Institutional Animal Care and Use Committee–approved animal protocols in accordance with institutional guidelines. For xenograft experiments, female nude mice were implanted with estrogen (0.72 mg) of 17β-estradiol pellets (Innovative Research of America). A suspension of 2 × 10⁷ cells was inoculated subcutaneously into the left flank of 6- to 8-week-old female athymic nude mice. Tumors were monitored until they reached an average size of 100 mm³, after roughly 6 to 8 weeks, at which point mice were randomized into the different groups (n = 4 mice/group) and treatments were begun. Ribociclib, kindly provided by Novartis, was dissolved in 0.5% methylcellulose and administered at 75 mg/kg once a day by oral gavage. Xenografts derived from PTEN-null T47D cells were treated with vehicle, ribociclib (60 mg/kg daily), MK-2206 (60 mg/kg twice a week, dissolved in 30% captisol), or both inhibitors in combination. Tumors were measured twice weekly using calipers.

Cells were seeded into 96-well plates with a density of 3,000 cells per well and treated the following day with the indicated drugs for the specified period of time. After treatment, cells were fixed with 2.25% paraformaldehyde in 0.1% PBS-Tween at room temperature for 30 minutes followed by permeabilization with 0.1% Triton X-100 in PBS at room temperature for 10 minutes. The cells were then washed with PBS twice and blocked with 2% BSA at room temperature for 1 hour. After blocking, cells were subjected to immunofluorescence staining with anti-GFP antibody (#A11122, Thermo Fisher Scientific) overnight at 4°C. Then, the cells were rinsed with PBS with 0.1% Tween-20 twice and incubated with Alexa 488–labeled anti-rabbit secondary antibody (#A21206, Life Technologies) at room temperature for 1 hour. After that cells were incubated with Alexa 568–labeled Phalloidin to stain the cytoplasm (#A12380, Thermo Fisher Scientific) at room temperature for 20 minutes followed by staining with Hoechst (1 μg/mL) for 30 minutes. Images were acquired with Opera Phenix (PerkinElmer) and analyzed with Harmony software.

Patients were enrolled in the phase I clinical study of ribociclib in combination with letrozole (NCT01872260), and the response was assessed using routine cross-sectional imaging as per RECIST v1.1 criteria. All clinical data and tumor specimens were collected and analyzed in accordance with Institutional Review Board–approved protocols, to which patients provided written informed consent, and all studies were conducted in accordance with the Declaration of Helsinki.

IHC was performed on a 5-μm section using an automated stainer (Bond Rx, Leica Microsystems). Following deparaffinization and rehydration of the slides, antigen retrieval was performed using citrate buffer (pH 6.0) for 20 minutes. After endogenous peroxidase activity was blocked with 3% to 4% hydrogen peroxide, samples were incubated at room temperature with RB mouse mAb (clone 1F8, Bio SB), PTEN mouse mAb (Biocare Medical), and p27 mouse antibody (#610242, BD Biosciences). Horseradish peroxidase polymer for detection was subsequently applied onto the slides. Sections were then incubated with DAB refine (10 minutes) for visualization and counterstained with hematoxylin (5 minutes). As defined previously (33), RB protein expression was considered lost if >95% of tumor nuclei showed lack of staining (0+). Staining of endothelial cells served as internal positive control; cases without this endothelial cell staining were excluded. PTEN protein expression was considered lost if the staining intensity of the tumor cell cytoplasm was markedly decreased or absent across all tumor cells, compared with adjacent benign glands/stroma (internal positive control). Cases were excluded if benign tissue lacked PTEN staining.

Unless otherwise specified, statistical comparison among groups was carried out with one-way ANOVA, with Sidak multiple group comparison test (GraphPad Prism; ****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05).

C. Costa is an employee of Novartis. Y. Wang is an employee of Novartis. A. Bardia is a consultant/advisory board member at Novartis, Radius Pharma, Pfizer, Daiichi, Sanofi, Genentech, Merck, Immunomedics, Spectrum, and Taiho. C.H. Benes reports receiving a commercial research grant from Novartis and other commercial research support from Amgen. J.A. Engelman is a global head, oncology, and has ownership interest (including patents) in Novartis. D. Juric is a scientific advisory board member for Novartis, Eisai, Genentech, Petra Pharma, EMD Serono, Ipsen, Syros, and Guardant and reports receiving commercial research support from Novartis, Eisai, Genentech, EMD Serono, Syros, Takeda, Placon Therapeutics, Celgene, and Amgen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Costa, A. Bardia, J.A. Engelman, D. Juric

Development of methodology: Y. Wang, C. Healy, J.R. Stone, H. Ebi, C.H. Benes, D. Juric

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Costa, Y. Wang, A. Ly, Y. Hosono, E. Murchie, T. Huynh, C. Healy, R. Peterson, S. Yanase, L.E. Henderson, L.J. Damon, D. Timonina, I. Sanidas, J.R. Stone, N.J. Dyson, A. Bardia, H. Ebi, C.H. Benes, D. Juric

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Costa, Y. Wang, A. Ly, Y. Hosono, C. Healy, S. Yanase, L.J. Damon, M. Mino-Kenudson, H. Ebi, J.A. Engelman, D. Juric

Writing, review, and/or revision of the manuscript: C. Costa, Y. Wang, A. Ly, M. Mino-Kenudson, J.R. Stone, L.W. Ellisen, A. Bardia, H. Ebi, J.A. Engelman, D. Juric

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Wang, C.S. Walmsley, R. Peterson, C.T. Jakubik, L.E. Henderson, C.J. Pinto, H. Ebi, D. Juric

Study supervision: C. Costa, C.H. Benes, J.A. Engelman, D. Juric

We would like to thank Silvia Goldoni, Tamara Gilbert, and Lesley Griner from Novartis for the help with image acquisition and analysis with the Opera Phenix. This work was funded by the HMS Laboratory for Systems Pharmacology Grant (P50GM107618), Susan Eid Tumor Heterogeneity Initiative, Dr. Jerry Younger Grant for Clinical and Translational Breast Cancer Research, the Jonathan Kraft Translational Award for Innovation in Cancer Research, and generous philanthropic support from Andrew Hertneky, and Stephen and Kathleen Chubb. Y. Hosono, S. Yanase, and H. Ebi's contribution was supported by the Fund for the Promotion of Joint International Research (15KK0303) from the Japan Society for the Promotion of Science (to H. Ebi).