Dual Targeting of CDK4/6 and BCL2 Pathways Augments Tumor Response in Estrogen Receptor–Positive Breast Cancer

Estrogen receptor–positive (ER⁺) breast cancers frequently exhibit deregulation of the cyclin-dependent kinase 4 and 6 (CDK4/6)/cyclin D1 (CCND1)/retinoblastoma (Rb) signaling pathway, resulting in uncontrolled cellular proliferation (1, 2). CDK4/6 inhibitors are active in breast cancer where they have demonstrated synergistic activity with endocrine therapy, leading to improvements in progression-free survival and overall survival (3–5). As a result, combination treatment with endocrine therapy and a CDK4/6 inhibitor is now considered standard of care for patients with early-line metastatic ER⁺ breast cancer. Nevertheless, de novo or acquired resistance to therapy is almost inevitable, underscoring the need for novel targeted therapies for this group of patients.

While the key mechanism of action of CDK4/6 inhibitors is to provoke a cell-cycle arrest, they likely contribute clinically relevant antitumoral effects through modulation of the immune response (6, 7), induction of senescence (7, 8), or through other noncanonical functions (9). Immune effects include immunogenic activation of tumor cells and promotion of a switch to a less immunosuppressive state through reduced regulatory T cell (Treg) numbers (7). In mammary tumor models, the profound cell-cycle arrest has been linked to induction of a senescence program (10). Despite their potent antiproliferative effects and demonstrated efficacy in the clinic, CDK4/6 inhibitors do not induce apoptotic cell death in breast cancer cells (7, 10, 11). Indeed, recent findings suggest that combination therapy may actually reduce apoptosis in treatment-naïve tumors (12). These findings are consistent with the senescent state, characterized by relatively irreversible replicative arrest and resistance to apoptosis (13, 14).

Evasion of cell death, a hallmark of cancer, can result from the overexpression of antiapoptotic BCL2 family members. BCL2 is overexpressed in the majority of primary and metastatic ER⁺ breast cancer (15, 16). The targeting of BCL2 and other antiapoptotic proteins has emerged as a viable therapeutic option due to the recent development of BH3-mimetic drugs that mimic endogenous antagonists of BCL2 and its related family members (17–19). Venetoclax (ABT-199/GDC-0199), a potent and highly selective inhibitor of BCL2 (20), has demonstrated single-agent activity in chronic lymphocytic leukemia (21) and in combination therapy for a number of hematologic malignancies (18). In preclinical models of ER⁺ BCL2⁺ breast cancer, ABT-199 improved tumor response to endocrine therapy with tamoxifen by enhancing apoptosis (22). These findings appear to be clinically relevant as the addition of venetoclax to tamoxifen elicited promising clinical activity in a phase I study in women with metastatic ER⁺ BCL2⁺ breast cancer (16).

On the basis of these observations, we have explored a role for dual targeting of CDK4/6 and BCL2 signaling pathways in preclinical models of ER⁺ BCL2⁺ breast cancer and found that the addition of ABT-199 to endocrine therapy with fulvestrant and the CDK4/6 inhibitor palbociclib substantially augmented tumor responsiveness in vivo. Combination therapy was accompanied by a deeper cell-cycle arrest and induction of apoptosis, including cells exhibiting a senescence-associated β-galactosidase phenotype. Moreover, the favorable palbociclib-mediated immunogenic activation of tumor cells together with reduction in Treg numbers and proliferation was not abrogated by the addition of ABT-199.

This study was designed to evaluate the response of breast cancer cells to the combination of endocrine therapy with ABT-199 and palbociclib. We evaluated the response to single-agent or combination therapy in breast cancer cell lines, patient-derived breast tumor organoids and xenograft models, as well as a syngeneic mouse mammary tumor model. Experiments were designed to investigate the mechanisms of tumor response. As outlined below, all mouse studies included randomization and blinding. The numbers of replicates performed for each experiment are included in the figure legends.

The studies were conducted in accordance with recognized ethical guidelines (Declaration of Helsinki). Human breast cancer tissues were obtained from consenting patients through the Royal Melbourne Hospital Tissue Bank and the Victorian Cancer Biobank with relevant institutional review board approval. Human Ethics approval was obtained from the Walter and Eliza Hall Institute (WEHI, 05/06) Human Research Ethics Committee.

ER⁺ PDX 50 and 315 models have been reported previously (22). PDX 50 originated from a patient following neoadjuvant chemotherapy (FEC-D) for her primary tumor; PDX 315 was derived from a primary tumor arising in a patient with a past history of a contralateral breast tumor. She had previously received adjuvant chemotherapy (FEC x 6) and endocrine therapy (tamoxifen followed by letrozole) and was taking letrozole at the time of her second primary tumor diagnosis. The resulting PDX 315 is strongly ER⁺ as reported previously (22). PDX 1105 and 1232 were derived from primary tumors in treatment-naïve patients.

NOD-SCID-IL2Rγ^−/− (NSG) mice were bred and maintained according to institutional guidelines. All animal experiments were approved by the WEHI Animal Ethics Committee (2017.002). Cohorts of 50–60 female mice were seeded with thawed single-cell suspensions of early-passage human breast tumors (passage 2 or 3). Briefly, 150,000–250,000 cells were resuspended in 10 μL of transplantation buffer (50% FCS, 10% of 0.04% Trypan blue solution and 40% PBS) and growth factor–reduced Matrigel (BD Pharmingen) at a ratio of 3:1 and injected into the cleared mammary fat pads of 3- to 4-week-old NSG female mice. Mice were monitored for tumor development three times weekly and tumor size measured using electronic Vernier calipers. Tumor volume was estimated by measuring the minimum and maximum tumor diameters using the formula: (minimum diameter)²(maximum diameter)/2. Once tumors arose, mice were randomized into treatment arms. Treatment was initiated when the tumor volume reached 80–120 mm³. Randomization and tumor measurements were managed using the Study Director software (v 3.0, studylog). Mice were sacrificed at the first measurement where tumor volume exceeded 600 mm³, or if their health deteriorated for reasons other than disease progression or drug toxicity (censored event).

Mice were treated for 21 days in a 28-day cycle. Fulvestrant (Clifford Hallam) 5 mg or its vehicle was injected intraperitoneally weekly for 3 of 4 weeks. ABT-199 (Active Biochem) 100 mg/kg or its vehicle was prepared as described previously (22) and given via oral gavage on days 1–5 for three out of four weeks. Palbociclib 100 mg/kg or its vehicle was given via oral gavage on days 1–5 for 3 out of 4 weeks. Palbociclib was dissolved in 50 mmol/L sodium lactate, pH 4. Palbociclib was administered at least 4 hours after ABT-199 to avoid negative interaction between the drugs and vehicles. Anti-PD1 (BioXcell) 200 μg/mouse or isotype control was given intraperitoneally on days 1, 3, and 5. Mice were monitored as per institutional guidelines.

Early-passage 67NR cells were maintained as described in Supplementary Methods and cells (5 × 10³) transplanted into the cleared mammary fat pads of 3- to 4-week-old BALB/c mice. Mice bearing 67NR tumors were euthanized at the experimental endpoint and tumors excised. To obtain a single-cell suspension for detection of cell surface and intracellular proteins, tumors and spleens were processed as described previously (23). BALB/c mice were bred and maintained according to institutional guidelines. Experiments were approved by the WEHI Animal Ethics Committee (2017.002).

Analysis of immune cells was performed as described previously (23). Antibodies to cell surface proteins were purchased from BioLegend unless otherwise stated: CD4 (clone GK1.5), CD8 (clone 53-6.7), TCRβ (clone H57-597), CD45 (clone 30-FII), Ki67 (clone B56), PD1 (clone 29F.1A12), and FoxP3 (eBiosciences, clone FJK-16). For intracellular detection of TNFα and IFNγ, cells were isolated and stimulated as described previously (23). Sample data were acquired on an LSRFortessa X-20 flow cytometer (BD Biosciences) and analyzed using FlowJo software (TreeStar).

Breast cancer organoids were generated from patient samples as described previously (24, 25). Briefly, breast cancer cell suspensions from patients undergoing surgery were prepared through mechanical disruption and enzymatic digestion. Isolated cells were then plated in adherent basement membrane extract (BME; Cultrex) drops in uncoated 24-well plates and cultured in optimized human organoid medium (HOM). For PDX-derived breast cancer organoids, we obtained single-cell suspensions by digestion of primary tumors and then sorted these by flow cytometry using a FACSARIA III (BD Biosciences), as described previously (22, 26). Sorted human EpCAM⁺ cells were then plated in BME drops and cultured in HOM.

For 3D colony assays, organoids were dissociated and then seeded in BME in 24-well plates (four drops of 8 μL containing 1,000 cells). After 24 hours, organoids were treated with fulvestrant (Sigma-Aldrich), ABT-199 (Active Biochemicals), palbociclib (Active Biochemicals), vehicle, or the indicated treatments. Treatments were replaced every 3–4 days. After 14 days in culture, cell viability was assessed using CellTiter Glo Luminescent Assay (Promega).

Statistical analyses were performed in the GraphPad Prism software version 8.0a. Kaplan–Meier (log rank test) was used to test for significant differences in the survival of mice (using the ethical end point for tumor size as a surrogate for death). Unpaired t tests were used to test the significance of differences in column means between treatments.

For details, see Supplementary Methods.

To gain insight into the short-term effects of combined endocrine and CDK4/6 inhibitor therapy on BCL2 prosurvival family members in treatment-naïve breast cancer, we evaluated gene expression data from the NeoPalAna window study (27), in which patients with newly diagnosed ER⁺ breast cancer were treated with the aromatase inhibitor anastrozole followed by the CDK4/6 inhibitor palbociclib prior to surgery (Fig. 1A; Supplementary Fig. S1A). As reported previously, expression of MKI67 (encodes the proliferation marker Ki67), was reduced by therapy but recovered when palbociclib was ceased. BCL2 and MCL1 levels similarly declined, albeit modestly, in response to combination therapy and rebounded once palbociclib was withdrawn. Reciprocal changes in BCL2L1 (encoding BCLXL) were observed, consistent with adaptive changes due to altered BCL2 and MCL1 expression. The effect of longer-term, continuous palbociclib treatment on BCL2 family members was next evaluated in MCF7 and T47D breast cancer cells, which were treated for 3 months at different concentrations (Supplementary Fig. S1B). A modest decrease in BCL2 and MCL1 proteins was observed in MCF7 cells, while expression was largely unchanged in T47D cells. BCLXL levels rose in both cell lines. Despite this, palbociclib-treated cells (continuous treatment at high concentration) exhibited greater sensitivity to BCL2 inhibition with ABT-199, consistent with either heightened cellular stress and/or lower BCL2 levels (Supplementary Fig. S1C). Together, these findings suggest that BCL2 could represent a potential target in palbociclib-treated breast cancer cells.

The link between estrogen receptor (ER) signaling and G₁–S phase cell-cycle progression has informed the combined use of CDK4/6 inhibitors with endocrine therapy (11). To address whether combination therapy was synergistic or additive, we compared the impact of adding either palbociclib or ABT-199 to the selective ER degrader fulvestrant in MCF7 cells (Fig. 1B). Doses were selected on the basis of the induction of on-target effects, including reduced ER (for fulvestrant), Rb phosphorylation (for palbociclib and fulvestrant) and generation of cleaved PARP (for ABT-199) at 24 hours and/or 7 days (Supplementary Fig. S2A and S2B). As reported previously (11), palbociclib showed synergy when combined with endocrine therapy, as determined by cell viability using CellTiter-Glo and BLISS score analysis (Fig 1B, top). Synergy was also observed between fulvestrant and ABT-199 at higher doses, consistent with prior findings for tamoxifen (22). Adding ABT-199 to fulvestrant and palbociclib was neither synergistic nor antagonistic (Fig 1B, bottom), suggesting that cotargeting of the BCL2 pathway might enhance response to fulvestrant/palbociclib.

To investigate the effect of drug treatment on key target proteins, MCF7 cells were treated for 72 hours with fulvestrant, palbociclib, ABT-199 or various combinations. As expected, combination therapy with fulvestrant and palbociclib reduced ER and phospho-Rb (pRb) levels, but did not induce apoptosis, as determined by cleaved PARP and cleaved caspase-3 expression (Fig. 1C). ABT-199 alone (or in combination) increased expression of pro-apoptotic BIM, a measure of primed cells (28) and induced cleaved PARP and cleaved caspase-3 (CC3; Fig. 1C). This was unexpectedly accompanied by reduced levels of pRb, the G₁–S cyclins D1 and E, and ER (Fig. 1C). Triple therapy with fulvestrant and palbociclib induced similar changes. Interestingly, the adaptive increase in cyclin D1 and E seen with palbociclib alone was abrogated by ABT-199 cotreatment (Fig. 1C). Together, these findings indicate that combining ABT-199 with endocrine therapy and palbociclib can trigger apoptosis in breast cancer cells that are undergoing a palbociclib-mediated cell-cycle arrest. The unexpected downregulation of pRb and cyclins D1 and E could reflect selective induction of apoptosis in cycling cells. Decreased ER expression is also consistent, as lower ERα levels are observed in G1-arrested cells (29).

The unanticipated finding that ABT-199 leads to reduced G₁–S phase cyclins and Rb dephosphorylation was next investigated by cell-cycle analysis using flow cytometry (Fig. 2A). MCF7 cells were treated for 72 hours with fulvestrant (100 nmol/L), ABT-199 (5 μmol/L), and palbociclib (250 nmol/L) alone or in combination. Fulvestrant and palbociclib both efficiently reduced the percentage of S-phase cells, consistent with their known mechanism of action. Notably, ABT-199 induced similar changes, albeit to a lesser extent than observed with palbociclib. This effect was more pronounced when ABT-199 was combined with palbociclib or fulvestrant (Fig. 2A).

We next investigated whether other BH3 mimetics could impact the cell cycle. ABT-737 (which targets BCLXL, BCL2, and BCLW) and the MCL1 inhibitor S63845 (30) exerted similar dose-dependent effects on pRb, cyclin D1 and ER, with all three BH3 mimetics inducing cPARP (Fig. 2B). As expected, MCL1 levels increased in response to ABT-199, ABT-737, and S63845, the latter attributable to stabilization of the protein by the inhibitor (30). Minimal changes were observed in BCL2 or BCL-XL levels. To address whether the changes in pRb, cyclin D1, or ER reflected an “on-target” effect of BH3 mimetics or a consequence of apoptosis, MCF7 cells were treated with BH3 mimetics (including the BCLXL inhibitor A1331852; ref. 31) in the presence or absence of the pan-caspase inhibitor Q-VD-Oph (Fig. 2C). While apoptosis was efficiently blocked by Q-VD-Oph, as demonstrated by lack of PARP cleavage, the effects on pRb, cyclin D1, and ER were maintained, implying they are caspase-independent. Furthermore, cell-cycle analysis revealed that each BH3 mimetic could inhibit proliferation independent of cell death (Fig. 2D). Together, these results suggest that ABT-199 (and other BH3 mimetics) may have unexpected effects as inhibitors of cell-cycle progression in breast cancer cells, beyond their canonical role as proapoptotic agents.

Given the potent effect of ABT-199 on apoptosis and the cell cycle, we next evaluated the effect of single or combination therapy in clonogenic cellular assays. For this, lower doses of fulvestrant and palbociclib were used to address synergy, because clonogenic activity was readily inhibited by single-agent treatment (Supplementary Fig. S3A and S3B). T47D and MCF7 cells were treated for two weeks with single, doublet, or triplet therapy comprising fulvestrant, palbociclib, and/or ABT-199. The addition of ABT-199 to fulvestrant and palbociclib significantly reduced the number and size of colonies, compared with doublet therapy (Fig. 3A and B; Supplementary Fig. S3C). Following removal of inhibitors at day 14, regrowth was apparent in cells that had been treated with doublet therapy. In contrast, minimal growth was observed in cells that had received triplet therapy (Fig. 3A and B; Supplementary Fig. S3C). At the lower doses used for clonogenic assays, the previously observed effects on key target proteins were maintained (Supplementary Fig. S4A and S4B). These findings, together with the adaptive changes in BCL2 family expression in breast tumor cells treated with palbociclib (Fig. 1A; Supplementary Fig. S1B), suggest that palbociclib might prime cells for apoptotic cell death with ABT-199.

To further investigate the combinatorial effect of ABT-199 and palbociclib, MCF7 cells were treated with either vehicle or palbociclib (500 nmol/L) for four days prior to switching drug treatment to vehicle, ABT-199, ABT-737, or both ABT-199 and palbociclib (Fig. 3C and D). Palbociclib pretreated cells were more responsive to either ABT-199 or ABT-737, with increased CC3 (Fig. 3C) and reduced viability as measured by propidium iodide exclusion (Fig. 3D). As expected, palbociclib-treated cells exhibited less PCNA expression and thus proliferation. Consistent with cell-cycle analysis (Fig. 2A and D), ABT-199–treated cells also exhibited lower PCNA levels compared with vehicle-treated controls (Fig. 3C).

CDK4/6 inhibitors elicit a senescence phenotype (7, 8) that can result in cells becoming more resistant to apoptosis (13, 14). In keeping with this finding, palbociclib induced a dose-dependent senescence-like phenotype characterized by increased cell size, cell flattening, and β-galactosidase expression (Supplementary Fig. S5A), while no discernible change in morphology or β-galactosidase expression was observed with ABT-199 alone. Cells treated with palbociclib for 4 days maintained this senescent-like phenotype following drug withdrawal (Fig. 3E and F). This phenotype, accompanied by the increased vulnerability to apoptosis observed following treatment with a BH3 mimetic (Fig. 3C), suggests that ABT-199 (and ABT-737) can augment CDK4/6 inhibition through the induction of senolysis of cells that have undergone growth arrest.

To address potential mechanisms of response or resistance to various combinations, we performed a genome-wide CRISPR/Cas9 screen in T47D cells (Fig. 3G; Supplementary Fig. S6A and S6B; Supplementary Tables S1 and S2). Cas9-expressing T47D cells were transduced with a pooled human genome-wide small guide RNA (sgRNA) lentiviral library containing 123,411 unique sgRNAs targeting 19,050 genes, and 1,864 miRNAs (6 sgRNAs per gene and 4 sgRNAs per miRNA) at low multiplicity of infection in five independent infections. Transduced cells were continuously treated with either fulvestrant, fulvestrant and palbociclib, fulvestrant and ABT-199, or the combination of fulvestrant, palbociclib, and ABT-199 for 2 months to identify sgRNA guides that helped to promote the growth of resistant cells. Doses chosen (fulvestrant 1 μmol/L, ABT-199 10 μmol/L, and palbociclib 1 μmol/L) were based on single-agent activity that exerted high selective pressure (Supplementary Fig. S3).

When the frequency of reads for guides was compared between different treatments, sgRNAs for RB1 were found to be enriched in fulvestrant–palbociclib–treated cells, with the greatest differential observed between fulvestrant-only and fulvestrant–palbociclib treated cells (Supplementary Tables S1 and S2). In contrast, RB1 sgRNAs were not over-represented in cells treated with fulvestrant–ABT-199 (Supplementary Tables S1 and S2). For fulvestrant and fulvestrant/palbociclib treatment–resistant cells, differentially represented guides (with an FDR of < 0.05) against genes and molecular pathways previously associated with CDK4/6 inhibitor resistance were identified such as PIK3C isoforms, CDK4 and FGFR members (32–34), mir-432-5p (35), members of the IL6–JAK–Stat pathway (IL6ST/GP130, JAK1, JAK3, STAT2, and STAT4; ref. 36), E2F family members (E2F3, E2F4, E2F6, E2F7, and E2F8; ref. 37), and Hippo pathway family members TEAD3 and TEAD4 (38). Interestingly, the histone H3 acetyltransferase epigenetic modifier genes from the MOZ/MORF complex (KAT6B, BPRF1, and ING5) were also implicated as genes involved in resistance. Interrogation of BCL2 family members revealed sgRNAs targeting MCL1 were enriched in fulvestrant-treated cells compared with those treated with fulvestrant–palbociclib or fulvestrant–ABT-199 (Supplementary Tables S1 and S2), suggesting that this could be a target for use in combination or upon progression.

To determine whether the addition of ABT-199 to fulvestrant and palbociclib modified mechanisms of resistance, we examined gene knockouts that were enriched or depleted between fulvestrant–palbociclib and fulvestrant–palbociclib–ABT-199 triplet therapy. Notably, RB1 was identified as the top candidate for cells treated with triple therapy versus fulvestrant–palbociclib while the genes identified above did not feature (Fig. 3G; Supplementary Table S2). These observations suggest that treatment with ABT-199 could mitigate loss of Rb and potentially other mechanisms of acquired resistance to palbociclib (34).

Given their distinct molecular effects and augmented response observed with CDK4/6 and BCL2 inhibition, we next studied patient-derived breast tumor organoid models, which more accurately reflect tumor heterogeneity and are emerging as promising preclinical models to predict drug response (24, 25). ER⁺ tumor organoids were generated from primary ER⁺ breast tumors or patient-derived xenograft (PDX) models (Fig. 4A; Supplementary Table S3). Primary breast tumor organoids were generated as recently described (24, 25) and were found to retain ER and BCL2 expression (Supplementary Fig. S7). They exhibited mutational profiles expected in luminal breast tumors (Supplementary Table S4). For PDX organoids, single-cell suspensions were prepared from PDX tumors, then sorted by flow cytometry and plated in organoid medium optimized for breast tumors (24, 25). These retained the histologic features of the parental tumor, including ER and BCL2 expression (Supplementary Fig. S8A and S8B). The four PDX tumors exhibited variable levels of BCL2 and cell-cycle proteins, with notable absence of pRb expression in PDX 50 (Fig. 4B). PDX 50 was found to harbor a heterozygous RB1 mutation (c.1215+1G>A) known to be pathogenic (ClinVar).

PDX-derived breast tumor organoids were treated with fulvestrant, ABT-199, palbociclib, or combination therapy to evaluate tumor response (Fig. 4; Supplementary Fig. S9) using a parallel approach to that described for MCF7 and T47D cells. Doses were selected on the basis of the dose–response observed for single-agent treatment (Supplementary Fig. S9A and S9B). As illustrated for PDX 1105 (Fig. 4C–F), treatment for 14 days reduced tumor organoid formation (Fig. 4C) and cell viability (Fig. 4D), with the greatest reduction observed upon triple therapy. Recovery in growth and viability was observed after a 1 week wash-out period for organoids treated with doublet combinations, whereas minimal tumor organoids were observed after triple therapy (Fig. 4C and D). Similar findings were observed in the other PDX organoid models (Supplementary Fig. S9C). As expected, fulvestrant and palbociclib reduced cellular proliferation, as measured by Ki67 (Fig. 4E; Supplementary Fig. S9D), while ABT-199 (as a single agent or in combination) augmented cell death, as determined by CC3 expression by Western blot analysis (Fig. 4F) or immunostaining (Supplementary Fig. S9E). Similar effects of fulvestrant, ABT-199, and palbociclib on cell viability were observed in three tumor organoid models that were directly derived from primary breast tumors (Fig. 4G).

The in vivo response to combination therapy was next evaluated in the four ER⁺ BCL2⁺ PDX models, whereby NSG mice bearing PDX tumors were treated with up to 3 cycles of combination therapy. A treatment cycle was defined as 28 days, with fulvestrant administered weekly for 3 weeks (on day 1, 8, 15). Palbociclib and ABT-199 were administered for 3 weeks (on days 1–5, 8–12, 15–19) followed by a 1-week rest. Fulvestrant alone marginally delayed tumor growth compared with vehicle in all models. Tumor response was improved with either fulvestrant–palbociclib or fulvestrant–ABT-199 combinations. Triple therapy that included ABT-199 further attenuated tumor growth during the treatment period (3 cycles of therapy for PDX 315 and 1232) and improved survival to the experimental ethical endpoint, including the 315 model, which was derived from a patient with endocrine refractory breast cancer (Fig. 5A; Supplementary Fig. S10A). Mechanistically, this response was accompanied by reduced cellular proliferation and increased apoptosis, as determined by Ki67 and CC3 levels, respectively (Fig. 5B–D). As expected in the RB1-deficient PDX 50 model, palbociclib did not improve tumor response to fulvestrant, although the fulvestrant–ABT-199 combination was efficacious in vivo (Supplementary Fig. S10B).

Because palbociclib is associated with neutropenia (3–5) and venetoclax produces “on-target” lymphopenia and low-grade neutropenia (16), we next determined whether myelotoxicity was problematic in vivo. As NSG mice are leukopenic, safety studies were conducted on healthy BALB/c mice (Supplementary Fig. S11). Mice were treated with the same regimen used for the PDX models, comprising 3 weeks of therapy followed by a 1-week “rest” period. Triple therapy was well tolerated, with mice maintaining normal body weight (Supplementary Fig. S11A, left), similar to the findings in NSG tumor-bearing mice (Supplementary Fig. S11A, right). As anticipated, a reduction in neutrophils and lymphocytes was observed at three weeks (cycles 1 and 2, day 21), but these recovered after a 1-week rest period (cycles 1 and 2, day 28; Supplementary Fig. S11B). No perturbation in urea, creatinine, or liver enzymes was observed at day 28 of cycles 1 and 2 (Supplementary Fig. S11C). These findings suggest that combination therapy might be safely administered in humans.

CDK4/6 inhibitors have been recently shown to promote antitumor immunity (6, 7), while venetoclax has been associated with lymphopenia in patients with breast cancer (16), raising the possibility that combination therapy could adversely impact the immune response. We therefore evaluated immune modulation in a syngeneic mouse mammary tumor model using the 67NR cell line, which is a nonmetastasizing subclone derived from a spontaneous BALB/cfC3H mammary tumor (39). This cell line expresses nuclear ER (40) and is BCL2 positive (Supplementary Fig. S12A). We first confirmed that the growth of 67NR cells was endocrine dependent through siRNA-mediated knockdown of ESR1, which significantly reduced cell viability (Fig. 6A; Supplementary Fig. S12B). These cells also displayed a dose-dependent response to fulvestrant (Supplementary Fig. S12C). Thus, 67NR cells represent a potentially useful syngeneic model to investigate the in vivo immune effects of palbociclib and ABT-199.

To determine whether ABT-199 modulated immunogenic activation of tumor cells by a CDK4/6 inhibitor, 67NR cells were treated with fulvestrant or combinations containing palbociclib and/or ABT-199. As previously described in human cells for MHCI, and in a mouse mammary tumor model (7), palbociclib in combination with fulvestrant increased expression of MHCI molecules (H-2Kd). Intriguingly, H-2Kd levels were also increased by ABT-199 and were further elevated when both drugs were combined (Fig. 6B), with similar findings for β2-microglobulin (Supplementary Fig. S12D and S12E). In keeping with enhanced IFN signaling, we also observed increased expression of phospho-STAT1 in 67NR and MCF7 cells in response to ABT-199 treatment (Supplementary Fig. S12F and S12G). These data suggest that ABT-199 further enhances the immunogenicity of 67NR cells when combined with palbociclib.

We next transplanted 67NR cells into the cleared mammary fat pads of BALB/c recipient mice. Mice bearing small tumors 10–14 days after engraftment were randomized into one of five treatment arms: (i) vehicle, (ii) fulvestrant, (iii) fulvestrant plus ABT-199, (iv) fulvestrant plus palbociclib, or (v) fulvestrant plus ABT-199 and palbociclib. While single-agent treatment with fulvestrant produced modest attenuation in growth, there was no benefit to the addition of palbociclib or ABT-199. However, similar to PDX models (Fig. 5), triple therapy elicited an improved response and survival benefit compared with doublet therapy with either fulvestrant–palbociclib or fulvestrant–ABT-199 (Fig. 6C; Supplementary Fig. S12H).

To explore the intratumoral immune response to combination therapy, analysis was performed after a short treatment of 9 days, when most vehicle-treated tumors reached ethical endpoint (Fig. 6D). Flow cytometric analysis of the tumor-infiltrating cells showed that combinations of fulvestrant–palbociclib resulted in a reduction in the proportion and absolute number of FOXP3⁺ Tregs (Fig. 6E; Supplementary Fig. S13A–S13C), elevated levels of which are associated with poor survival in ER⁺ breast cancer (41). Consistent with this observation, Tregs were also less proliferative (Fig. 6F), suggesting the impact of pRb and BCL2 inhibition in this setting. The reduction in Tregs resulted in a modest but significant increase in the CD8/Treg ratio (Supplementary Fig. S13D). Reduced expression of the immune checkpoint receptor PD1 in Tregs was also observed (Fig. 6G), consistent with previous findings in a breast cancer mouse model using the CDK4/6 inhibitor abemaciclib (7). The systemic impact of BCL2 and CDK4/6 inhibition was most pronounced in the tumor microenvironment, but modest effects of combination therapy on Treg cells were observed in the spleen (Supplementary Fig. S14A–S14D). Furthermore, increased production of tumor necrosis factor (TNFα) by CD4⁺ and CD8⁺ T cells (Fig. 6H) and a trend toward increased IFNγ were evident (Supplementary Fig. S14E and S14F), indicative of enhanced effector function. Together, these findings confirm that increased tumor immunogenicity and intratumoral immune cell activation are provoked by palbociclib and indicate that ABT-199 does not antagonize these effects, but may further enhance them.

To further investigate the immunomodulatory effects of CDK4/6 and BCL2 inhibition, we measured PDL1 expression in 67NR cells in vitro and noted that ABT-199 increased PDL1 levels in 67NR cells (Fig. 6I). A similar effect was observed in MCF7 cells (Supplementary Fig S14G). The favorable intratumoral immune microenvironment and upregulation of PDL1 expression in 67NR cells suggested that immune checkpoint blockade might further improve the efficacy of triple therapy. We therefore treated 67NR tumor–bearing mice with either (i) vehicle (which included an isotype control antibody); (ii) fulvestrant and ABT-199; (iii) fulvestrant, ABT-199, and anti-PD1; (iv) fulvestrant plus ABT-199 and palbociclib; or (v) fulvestrant, ABT-199, palbociclib, and anti-PD1. Consistent with the data shown in Fig. 6C, triple therapy improved tumor response and survival, compared with fulvestrant plus ABT-199. The addition of anti-PD1 therapy to this triple combination further attenuated tumor growth and improved survival (Fig. 6J; Supplementary Fig. S14H).

Despite the majority of patients with ER⁺ breast cancer responding to endocrine therapy, the duration of response is highly variable due to multiple mechanisms that result in innate or acquired resistance (42). The recent emergence of CDK4/6 inhibitors has changed the landscape for treatment of ER⁺ breast cancer, with remarkable improvements in survival (3–5) for a subset of patients. Resistance to therapy, however, is almost inevitable. In patients with breast cancer, the combination of endocrine therapy with CDK4/6 inhibition is predominantly cytostatic, likely accounting for reduced apoptosis (12). Combination therapy that overcomes resistance through the induction of tumor cell death could therefore be beneficial. Using breast cancer cell lines, organoids, and PDX models, we show that inhibition of BCL2 and CDK4/6 in combination with endocrine therapy can inhibit proliferation and induce apoptosis of cancer cells, including of phenotypically senescent cells, leading to enhanced responsiveness in vivo. Notably, we also show that ABT-199 augments the CDK4/6 inhibitor response through dephosphorylation of Rb and reduced G₁–S cyclins. In addition, pharmacologic inhibition of both CDK4/6 and BCL2 in a syngeneic model reduced the proportion and proliferation of Tregs in the tumor microenvironment. This was accompanied by T-cell activation, with improved antitumor activity observed in vivo upon PD1 blockade.

The unexpected observation that ABT-199 and other BH3 mimetics led to Rb dephosphorylation (or possibly degradation of pRb), and reduced G₁–S cyclins potentially accounts for ABT-199 enhancing the effect of fulvestrant and palbociclib. As this was more pronounced at higher doses, the clinical significance of this finding is unclear. Although this was a caspase-independent effect (indicating that the phenotype is independent of the final “demolition” stage of apoptosis), it remains possible that the effect is due to mitochondrial stress imposed by BH3 mimetics and preferential killing of cycling cells. Further studies will be required to clarify this, including on BAX/BAK-deficient cells. The downregulation of ER following ABT-199 treatment would be consistent with the resultant G₁ arrest. The precise mechanism underpinning the reduction in G₁ cyclins (cyclin D and E) is unclear but this could alleviate resistance to CDK4/6 inhibitors given that elevated cyclin E levels have been clinically reported (43). These findings indicate the potential of triple therapy as early line therapy for patients with endocrine sensitive tumors, aimed at delaying adaptive resistance.

As reported previously (7, 8), palbociclib induced a senescence-like phenotype, characterized by distinctive morphologic changes that included cell flattening and enlargement of growth-arrested cells that were ß-gal–positive and more resistant to apoptosis. Notably, short- and long-term palbociclib-treated cells appeared to be more vulnerable to ABT-199–mediated apoptotic cell death. Previous reports have shown that BCL-XL is a senolytic target (14). Our findings indicate that BCL2 is also a target and suggest that venetoclax could be a senolytic agent when combined with CDK4/6 inhibitor therapy in breast tumors.

Data from clinical trial samples are helping to elucidate the pleiotropic mechanisms of acquired resistance to CDK4/6 inhibitors. Due to the requirement for intact Rb tumor suppressor protein function, RB1 loss or mutation has been identified as a robust marker of resistance to CDK4/6 inhibition. Notably, patients who progressed on fulvestrant and palbociclib in the PALOMA-2 clinical trial developed RB1 mutations, although at relatively low prevalence (34). ABT-199 may provide an alternate means of targeting these tumors, as demonstrated in vitro and in vivo in PDX model 50, which harbored a RB1 mutation. Other resistance mechanisms that have been identified in patient tumors include perturbations in the PI3K/AKT and mTOR, FGFR, IL6–JAK–Stat, E2F, and Hippo signaling pathways (32–38). Our CRISPR/Cas9 screen of fulvestrant–palbociclib versus fulvestrant-treated cells produced differential enrichment of sgRNA guides in these pathways, supporting those findings. Interestingly, we observed differentially enriched sgRNA guides for epigenetic modifier genes in the MOZ/MORF complex, such as KAT6B, suggesting that these may represent novel targets. When we evaluated sgRNA guides that were enriched in fulvestrant–palbociclib–venetoclax versus fulvestrant–palbociclib–treated cells, RB1 mutations were under-represented with triple therapy, while targets identified above did not feature. This, together with the reduction in cyclin E levels noted above, suggests that cotreatment with venetoclax could mitigate some of the mechanisms of acquired resistance to CDK4/6 inhibitors.

While immune checkpoint inhibitors are showing promise for triple-negative breast cancer, ER⁺ breast cancers contain fewer TILs and are considered to be far less immunogenic (44). It is noteworthy, however, that anti-PDL1 therapy led to responses in a subset of patients with advanced ER⁺ PDL1⁺ breast cancer (45). Intriguingly, CDK4/6 inhibitors have been shown to modulate the tumor immune microenvironment by increasing antigen presentation, reducing inhibitory Treg cells and stimulating increased PDL1 expression (6, 7). These findings imply that the therapeutic effect of these agents may be mediated, in part, through overcoming immune evasion, although the magnitude of this effect remains uncertain. Utilizing a syngeneic model of ER⁺ breast cancer, we confirmed these observations and showed that ABT-199 augments CD4 and CD8 T-cell activation and does not abrogate palbociclib-mediated effects in reducing Tregs. Although ABT-199 induces a peripheral lymphopenia, this is largely due to a reduction in B cells (16). Our data suggest that ABT-199 may favorably contribute to the intratumoral immune microenvironment rather than adversely impact immune signaling systemically. The additional response we observed with PD1 checkpoint inhibition supports this observation. These data are also consistent with recent findings using a MYC-driven model of breast cancer where adjuvant ABT-199 plus anti-PD1 therapy produced a survival benefit (46).

Despite the promising responses observed in preclinical models of ER⁺ breast cancer, it remains to be determined if the triple combination will be effective and can be safely delivered in the clinic. Both CDK4/6 inhibitors and venetoclax are metabolized by CYP3A4 and can induce neutropenia and lymphopenia, although this has not proven to be problematic when each drug is administered alone with endocrine therapy (16, 47–49). The 28-day regimen used in our studies included a 1-week break from both drugs (currently in clinical use for palbociclib and ribociclib) and appeared to facilitate leukocyte recovery. This regimen could also potentially prime tumor cells for cell death, given that BCL2 levels rose following palbociclib withdrawal.

The specificity of ABT-199 for BCL2 is advantageous for reducing possible toxicity, given the limitation of current BCLXL inhibitors due to their induction of on-target thrombocytopenia (50). However, this could increase the risk of resistance due to upregulation of prosurvival proteins, including BCLXL and MCL1. We observed that prolonged exposure to high concentrations of palbociclib increased BCLXL expression in both MCF7 and T47D cells. However, they remained sensitive to ABT-199, presumably due to increased cellular stress. The downregulation of MCL1 following palbociclib treatment, together with its potential role in inducing resistance to standard therapy observed in our CRISPR/Cas9 screen, indicates that MCL1 is also a possible target. This is consistent with findings by others that MCL1 is a potential target in ER⁺ breast cancer (51, 52). Further investigation of combinatorial (if clinically tolerated) or sequential BH3 mimetic therapy may therefore be warranted.

It will be important to determine whether our preclinical findings translate in the clinic for patients with ER⁺ breast cancer. BCL2 is often expressed at high levels in ER⁺ tumors, presumably because BCL2 is an estrogen-responsive gene. This could account for the apparent paradox of BCL2 being considered a favorable prognostic marker, while also contributing to treatment resistance (19). Importantly, the majority of relapsing ER⁺ tumors are BCL2 positive (16), underscoring its potential utility as a therapeutic target.

The combination of ABT-199 with tamoxifen has shown promising clinical activity in patients with metastatic ER⁺ breast cancer (16). These findings are currently being further investigated in a randomized phase II study of fulvestrant with or without venetoclax in patients with ER⁺ breast cancer who progress on a CDK4/6 inhibitor (NCT03584009). Here we show that combining BCL2 and CDK4/6 inhibition with endocrine therapy elicits potent activity in endocrine-sensitive and refractory models of ER⁺ breast cancer. This combination overcomes the cytostatic effects of endocrine and CDK4/6 inhibitor therapy to trigger cell death in vivo, and induces a favorable tumor immune microenvironment, together supporting further investigation of this combination in the clinic, currently underway (NCT03900884).

The Walter and Eliza Hall Institute receives milestone and royalty payments related to venetoclax, and employees (J.R. Whittle, F. Vaillant, E. Surgenor, A.N. Policheni, G. Giner, B.D. Capaldo, H. Chen, H.K. Liu, M.J. Herold, G.K. Smyth, D.H.D. Gray, J.E. Visvader, and G.J. Lindeman) may be eligible for benefits related to these payments. J.R. Whittle is an employee of The Walter and Eliza Hall Institute of Medical Research and reports receiving benefits related to venetoclax. N. Sachs and H. Clevers are listed as co-inventors on two patents that relate to improved culture methods for expanding epithelial stem cells and obtaining organoids, which is owned by the Royal Netherlands Academy of Arts and Sciences (KNAW, Koninklijke Nederlandse Akademie van Wetenschappen) and licensed to the HUB Foundation for Organoid Technology. H. Clevers is a paid advisory board member for Roche Holding. D.H.D. Gray reports receiving commercial research grants from Servier Pharmaceuticals. J.E. Visvader is an employee of The Walter and Eliza Hall Institute of Medical Research and reports receiving benefits related to venetoclax. G.J. Lindeman is an employee of The Walter and Eliza Hall Institute of Medical research and reports receiving benefits related to venetoclax; is a paid advisory board member of AbbVie; reports receiving commercial research grants from AbbVie, Roche/Genentech, Pfizer, and Servier; reports receiving speakers bureau honoraria from Amgen and Genentech; and is an unpaid consultant/advisory board member for Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.R. Whittle, F. Vaillant, J.E. Visvader, G.J. Lindeman

Development of methodology: J.R. Whittle, N. Sachs, D.H.D. Gray, J.E. Visvader, G.J. Lindeman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.R. Whittle, F. Vaillant, E. Surgenor, A.N. Policheni, B.D. Capaldo, H.-R. Chen, H.K. Liu, J.F. Dekkers, A. Fellowes, T. Green, H. Xu, S.B. Fox, J.E. Visvader, G.J. Lindeman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.R. Whittle, F. Vaillant, A.N. Policheni, G. Giner, A. Fellowes, T. Green, H. Xu, M.J. Herold, G.K. Smyth, D.H.D. Gray, J.E. Visvader, G.J. Lindeman

Writing, review, and/or revision of the manuscript: J.R. Whittle, F. Vaillant, G. Giner, H.K. Liu, H. Clevers, A. Fellowes, S.B. Fox, G.K. Smyth, D.H.D. Gray, J.E. Visvader, G.J. Lindeman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.R. Whittle, S.B. Fox, J.E. Visvader, G.J. Lindeman

Study supervision: J.E. Visvader, G.J. Lindeman

We are grateful to G.B. Mann, L. Graham, and staff at the Royal Melbourne Hospital Tissue Bank for support and thank the Animal, Histology and FACS services at WEHI. We thank A. Strasser for helpful discussions. Coded breast tumor samples were provided by the Victorian Cancer Biobank (which is supported by the Victorian Government). This work was supported by the National Health and Medical Research Council, Australia (NHMRC 1054618, 1113133, 1153049), NHMRC IRIISS, Victorian State Government Operational Infrastructure Support, Australian Cancer Research Foundation, National Breast Cancer Foundation (NT-13-06, IIRS-19-004), Breast Cancer Research Foundation (BCRF-18-182), Cancer Australia (1165878), Qualtrough Cancer Research Fund, Joan Marshall Breast Cancer Research Fund, and Collie Foundation (2017F001). J.R. Whittle was supported by an NHMRC/NBCF Research Fellowship and the Royal Australasian College of Physicians, A.N. Policheni by a Cancer Council Victoria Postdoctoral Research Fellowship, J.F. Dekkers by a Marie-Curie Postdoctoral Fellowship, and NHMRC Research Fellowships to D.H.D. Gray (1090236 and 1158024), G.K. Smyth (1058892), J.E. Visvader (1037230), and G.J. Lindeman (1078730 and 1175960).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.