Combined CDK4/6 and PI3Kα Inhibition Is Synergistic and Immunogenic in Triple-Negative Breast Cancer Free

Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer that is clinically characterized by its lack of expression of estrogen- and progesterone receptors (ER/PR) and HER2 (1). TNBC has the worst outcome of all breast cancer subtypes (1). Cytotoxic chemotherapy is the mainstay of current treatment for TNBC patients. However, although these regimens can prove effective for a subgroup of TNBC patients, in general, once relapsed, remissions are brief and are frequently followed by rapid disease progression and death. In contrast, the existence of a tumor immune infiltrate has been reported to be a robust prognostic factor in TNBC (2). However, the causal mechanisms underlying this immune response are unclear. It has been proposed that DNA damage with resultant activation of stimulator of interferon genes (STING) and/or mutational load and neoantigen production may be responsible (3, 4). Regardless, this immune response seems not to be sufficient to induce primary tumor clearance in humans.

Cell-cycle control is frequently dysregulated in breast cancer (5). The transition from G₁ to S phase of the cell cycle is controlled by interactions between cyclin-dependent kinases 4 and 6 (CDK4/6), cyclin D1, and retinoblastoma protein (RB). CDK4/6 inhibition has been shown to be an effective therapeutic strategy for ER-positive breast cancers in clinical trials (6–10). However, TNBC is a molecularly heterogeneous disease characterized by genomic instability along with high expression of cell-cycle genes including cyclin E1 (11) and has shown resistance to single-agent CDK4/6 inhibition (6, 12).

The complex biology of TNBC suggests that combination treatments will be required to achieve effective and durable disease control. Combined treatment with PI3K and CDK4/6 inhibitors has been shown to be effective in mitigating early adaption responses to single-agent PI3K and CDK4/6 inhibitors for overcoming single-agent inhibitor resistance in ER-positive breast cancers (13, 14). We sought to determine whether combined PI3K and CDK4/6 inhibition would be a similarly useful strategy for TNBC. We also hypothesized that induction of cell death could promote an immune response, given that TNBC may be amendable to immune approaches (2, 15–17).

Here, we show that synergistic interactions of PI3Kα and CDK4/6 inhibitors resulted in more effective disease control of TNBC both in vitro and in vivo. Combination treatment resulted in increased cell-cycle arrest, apoptosis, calreticulin cell-surface expression, and tumor immunogenicity. Finally, we show that combination of PI3Kα and CDK4/6 inhibitors with immune-checkpoint blockade resulted in long-lasting tumor regression in a syngeneic immunocompetent mouse model of TNBC.

All culture media were supplemented with 10% heat-inactivated FBS. HCC70, HCC1806, and MDA-MB-468 were kindly provided by Professor Roger Daly from Monash University, Australia. All other human cell lines were obtained from the ATCC. The human cell lines have been authenticated by short tandem repeat analysis and were maintained in DMEM or RPMI1640 (Gibco). AT3OVA (18) mouse TNBC cell line was maintained in SAFC DMEM. We have previously characterized AT3OVA and found that it was wild-type for Rb1, Pik3ca, and Trp53 (19). All human and mouse cell lines have been verified to be negative for mycoplasma contamination and were maintained at 37°C in a 5% CO₂ incubator. Characteristics of the human cell lines are further detailed in Supplementary Table S1.

BYL719 (Alpelisib) and LEE011 (Ribociclib) were kindly supplied by Novartis. PD991 (Palbociclib; PD-0332991) was purchased from SelleckChem. All drugs for in vitro use were reconstituted in DMSO at 10 mmol/L. BYL719, LEE011, and PD991 were reconstituted in 0.5% methylcellulose (Sigma Aldrich) for in vivo experiments. Immune-checkpoint antibodies and the isotype control used in in vivo experiments were obtained from BioXCell: anti-mouse PD-1 mAb (RMP1-14), anti-mouse CTLA-4 mAb (9H10), and Rat IgG2a isotype control mAb (2A3). Doses used are detailed in figure legends.

Drug combination studies were performed according to the Chou–Talalay method of synergy quantitation (20). MDA-MB-468, HCC1143, HCC70, HCC1806, MDA-MB-231, HS578T, and MDA-MB-453 cells were treated in vitro for 72 hours with the combination of BYL719 and LEE011 over a range of concentrations held at a fixed ratio based on the GI50 (drug concentration required for 50% cell growth inhibition) of each drug specific for each cell line. CalcuSyn 2.0 (Biosoft) performs multiple drug dose-effect calculations using the Median Effects methods described by Chou and Talalay (21) and was used to determine the combination index (CI), which offers quantitative definition for additive effect (CI = 1), synergism (CI < 1), and antagonism (CI > 1) of drug combinations.

Cells were plated in 24-well plates and treated the following day with the indicated agents. Cells and tissues were lysed and processed as described in Supplementary Methods. FACS analysis was performed on an LSR II flow cytometer (BD Biosciences).

Cell lines were plated in 6-well plates and treated the following day as indicated. Cells and tissues were lysed and processed as described in Supplementary Methods. Signal intensities for each protein of interest were quantitated using Image J (https://imagej.net) and presented in Supplementary Fig. S1.

MDA-MB-231 cells were plated in 6-well plates and treated the following day as indicated. Two replicates were included. After 24 hours, total cell RNA was extracted using a PureLink RNA Mini kit (ThermoFisher Scientific) following the manufacturer's instructions. The quantity and integrity of the total RNA were determined using TapeStation 2200 system (Agilent Technologies) and Qubit RNA High Sensitivity assay kit (ThermoFisher Scientific). Note that 500 ng total RNA was used for library preparation according to the manufacturer's instructions (QuantSeq 3′ mRNA-Seq FWD; Lexogen). Indexed libraries were pooled and sequenced on a NextSeq500 (Illumina). Briefly, the library was generated with an oligo-dT containing the Illumina Read2 linker and a random forward primer containing the Illumina Read1 linker. The library was then amplified with PCR primers containing sample indices and the Illumina clustering sequences. Five to 15 million single-end 75 bp reads were generated per sample.

Gene set enrichment analysis (GSEA) was performed using the GSEA tool (22) on the entire normalized RNA expression count matrix without limiting the input to only differentially expressed genes. The Hallmark collection of 50 predefined gene sets and C5 Gene Ontology (GO) collection of 5,917 predefined gene sets from the Molecular Signatures Database (MSigDB, Broad Institute) were used for analysis. The gene sets (3,576 C5 and 49 Hallmark gene sets) included in the analysis were limited to those that contained between 15 and 500 genes. Permutation (by gene sets) was conducted 1,000 times according to default weighted enrichment statistics and using difference of class metrics to calculate and rank genes according to their differential expression levels between two treatment groups. Significant gene sets were those defined with a nominal P < 0.05. Calculation of the false discovery rate (FDR) was used to correct for multiple comparisons and gene set sizes. Significantly enriched gene sets with FDR < 0.25 were selected for hypothesis generation.

Six- to 8-week-old female Nod SCID γ (NSG) mice were used for MDA-MB-231, HCC1806, and patient-derived xenograft models. For MDA-MB-231 and HCC1806 xenografts, 2 × 10⁶ cells (suspended in PBS:Matrigel, 1:1) were injected into the fourth mammary fat pad of the NSG mice. For experiments with the patient-derived xenograft (PDX) model 12006, a single tumor from a previously established NSG mouse (passage 5) was excised and small fragments were implanted directly into the fourth mammary fat pads of 6- to 8-week-old female NSG mice. AT3OVA (5 × 10⁵) cells were suspended in PBS and injected into the fourth mammary fat pads of 6- to 8-week-old female immune-competent C57BL/6 mice.

For all models, tumor volume (length × width² × 0.5) was assessed by caliper measurements every 3 to 4 days. Once tumors reached an average of 100 to 150 mm³ (human xenograft models) or 50 to 100 mm³ (mouse syngeneic model), mice were randomized to commence drug treatment (day 1). BYL719, LEE011, and PD991 were administered via oral gavage. BYL719 was given daily throughout the duration of the experiment. LEE011 or PD991 were given daily on days 1 to 21 of each 28-day cycle (3 weeks on, 1 week off). Anti–PD-1 and/or anti–CTLA-4 were administered via intraperitoneal injections on days 1, 5, 9, and 13 only. For 7-day immune response analysis, BYL719 and LEE011 were administered daily with no breaks with dosing. For these experiments, anti–PD-1 and/or anti–CTLA-4 were only administered on days 1 and 5.

Mice were euthanized if the tumors reached ethical limit of 1,400 mm³ or if the animals displayed health indicators that met the institutional criteria for sacrifice. All animal experiments were approved by the Peter MacCallum Cancer Centre Animal Experimentation Ethics Committee (E539 and E556) and conducted in accordance with the National Health and Medical Research Council Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.

One-way ANOVA with Tukey multiple comparison test was used to compare between treatment groups. Tumor growth curves were analyzed using two-way repeated measures ANOVA with Tukey multiple comparison test. Survival differences between treatment groups were determined using the Kaplan–Meier log rank analysis. Statistical analyses (not including differential gene expression or GSEA analyses) were performed using Prism 7 (GraphPad). A two-tailed P < 0.05 was considered statistically significant. All data are expressed as mean ± SEM.

We quantitated the level of synergy between BYL719 (PI3Kα inhibitor) and LEE011 (CDK4/6 inhibitor) in seven human TNBC cell lines including six RB1-wild-type and one RB1-mutant cell line. Of the six RB1-wild-type cell lines, three were classified as basal-like, two mesenchymal-like, and one was of the luminal androgen receptor (LAR) subtype (23). We demonstrated that the interaction between BYL719 and LEE011 was synergistic in all six RB1-wild-type TNBC cell lines but was antagonistic in the RB1-mutant cell line, MDA-MB-468 (Fig. 1A).

As expected from the above results, combination treatment with BYL719 and LEE011 significantly inhibited cell growth compared with single-agent and vehicle treatments in the synergistic lines (Fig. 1B). Consistent with this, we show in HCC1806 and MDA-MB-231 cell lines that BYL719 and LEE011 single-agent treatments resulted in reduction of AKT and RB phosphorylation, respectively, compared with vehicle treatment, and both remained suppressed with combination treatment (Fig. 1C).

The efficacy of combined PI3Kα and CDK4/6 inhibition was next examined in vivo. HCC1806 and MDA-MB-231 xenograft tumor models were treated with BYL719, CDK4/6 inhibitor (LEE011 or PD991), the combination, or vehicle control. In both models, strong synergism was observed with significant reduction of tumor growth of mice treated with combination therapy compared with single-agent and vehicle treatments (Fig. 1D).

We next treated a patient-derived TNBC xenograft (PDX12006) with the combination of BYL719 and LEE011. This tumor sample had oncogenic TP53 p.S241C (COSM10709) and ARID1A p.R2232W (COSM3724474) mutations and was classified as mesenchymal subtype (P < 0.001). Similar to what we observed in HCC1806 and MDA-MB-231 xenograft models, combination treatment in the PDX model exhibited significant tumor growth inhibition (Fig. 1D).

We next examined combined BYL719 and LEE011 treatment on cell cycle and apoptosis in vitro. Combined treatment significantly increased the effect on G₁ cell-cycle arrest compared with single-agent BYL719 and vehicle treatments in both basal-like and mesenchymal-like TNBC cell lines (Fig. 2A). In HCC70 and MDA-MB-231 lines, G₁ cell-cycle arrest was significantly greater following combination treatment compared with BYL719 and LEE011 single-agent treatments as well as vehicle treatment.

Notably, combined treatment with BYL719 and LEE011 significantly increased apoptosis as indicated by an increase in percentage of Annexin V⁺/propidium iodide⁻ cells compared with single-agent and vehicle treatments (Fig. 2B).

We next analyzed protein components of the PI3K and CDK4/6 pathways by Western blot analysis in HCC1806 and MDA-MB-231 cells treated with vehicle, BYL719, LEE011, or the combination for 72 hours in vitro (Fig. 2C). Treatment with LEE011 was associated with increased expression of cyclin D1, consistent with previous reports (13, 14). Phosphorylation levels of eukaryotic translation initiation factor 4E-binding protein-1 (4EBP1 Thr37/40), known to be a direct substrate of mTOR and one of the regulators of cyclin D1 expression (24, 25), were maintained across treatment groups in MDA-MB-231 and only slightly decreased with BYL719 and LEE011 combined treatment in HCC1806 compared with vehicle treatment, suggesting incomplete inhibition of mTOR activity. Nonetheless, we observed that combination treatment removed AKT inhibition on GSK3β and p21^Cip1/WAF1 (p21) tumor suppressors. p-GSK3β Ser9 and p-p21 Thr145 expression levels were suppressed with combined BYL719 and LEE011 treatment compared with single-agent and vehicle treatments in MDA-MB-231 cells, whereas in HCC1806 cells, suppression was observed with single-agent BYL719 and combination therapy compared with vehicle treatment.

Next, we assessed the CDK2/cyclin E complex, which associates with CDK2 and initiates CDK2 activation shortly before entry into S phase (26). We demonstrate that p-CDK2 Thr160 levels were reduced with LEE011 treatment and further decreased with combination treatment in both cell lines. Furthermore, cyclin E2 levels were markedly decreased with LEE011 and combination treatments compared with vehicle control; however, this was also accompanied by an increase in cyclin E1 expression with LEE011 and combination treatments.

These results suggest that decreased cell survival and proliferation due to combined PI3Kα and CDK4/6 blockade is likely mediated through multiple cell-cycle–related pathways in TNBC.

To further understand the mechanism by which BYL719 and LEE011 were effective in combination in an unbiased manner, we performed next-generation RNA sequencing and GSEA using MDA-MB-231 cells treated with vehicle BYL719, LEE011, or the combination. BYL719 and LEE011 single-agent and the combination treatment groups were each compared with vehicle treatment.

Not surprisingly, we observed that the “E2F targets” hallmark gene set was the top enriched (P < 0.05) downregulated gene set across all comparisons (Fig. 3A). Compared with single-agent and vehicle treatment groups, the combined therapy was associated with significant downregulation of E2F target genes involved in cell-cycle progression, consistent with our results on cell-cycle inhibition above.

The top enriched (P < 0.05) downregulated GO processes for each comparison were all related to cell division processes, with particular enrichment for downregulation of genes in chromosome/chromatid segregation processes as shown in Fig. 3B.

A heatmap of the “core-enriched” genes in the top enriched (P < 0.05) hallmark and GO process gene sets shows striking downregulation of these genes (log fold change, –0.39 to –3.74; median log fold change = –1.66) with the combination treatment compared with vehicle treatment (Fig. 3C). Interestingly, we noted that genes involved in DNA damage and repair processes (highlighted in yellow in Fig. 3C) made up a sizeable proportion among those downregulated.

Given these findings, we next investigated if there was objective evidence of DNA damage. Validating our observations above, we found that both LEE011 single-agent and combination treatments were associated with increased levels of γH2AX phosphorylation on Ser139 (γH2AX, a marker for DNA damage; Fig. 3D).

In our GSEA analysis, we observed that immune-related pathways were also among the top enriched (P < 0.05) upregulated hallmarks. These included upregulation of IFNα, IFNγ, TNFα, and innate immune (complement) responses (Fig. 4A and B). We also observed significant upregulation of genes (log fold change, 0.24 to 1.51; median log fold change = 0.55) involved in antigen presentation that included HLAs HLA-A (human major histocompatibility complex class I; MHCI), HLA-DMA (human MHCII), CTSD, ICAM, RELB, PSME1, and TAPBP, in response to combination therapy compared with vehicle treatment (highlighted in yellow in Fig. 4B).

We next investigated whether combined BYL719 and LEE011 treatment could modulate tumor immunogenicity given the transcriptional findings above. We examined tumor cell-surface expression of HLA antigens as well as the immune-checkpoint ligands, programmed death ligand 1 (PD-L1), and CD86 in HCC1806 and MDA-MB-231 cell lines (Fig. 4C). We found that LEE011 single-agent as well as the combination treatments significantly increased expression of HLA antigens (expression was further increased in the presence of IFNγ), as well as CD86, the ligand for immune-checkpoint cytotoxic T-lymphocyte–associated protein 4 (CTLA-4), in both cell lines. BYL719 seemed to have an inhibitory effect on the expression of PD-L1, consistent with the literature that this ligand is regulated through the PI3K/AKT signaling pathway (27–29).

Given the above results, along with observations of increased DNA damage and tumor immunogenicity, we next assessed whether combined therapy could induce an increase in “immunogenic” cell death. We examined levels of tumor cell-surface calreticulin expression, given that this protein, usually present on the endoplasmic reticulum, is translocated to the plasma membrane surface of cells undergoing immunogenic cell death in the early, preapoptotic phase, facilitating engulfment of dying cells by dendritic cells (30). Consistent with our hypothesis, we observed pronounced increased expression of cell-surface calreticulin following combination therapy on both HCC1806 and MDA-MB-231 cells compared with single-agent and vehicle treatments (Fig. 4D).

Collectively, these findings suggest that the combined BYL719 and LEE011 treatment enhanced immunogenic cell death and could promote tumor immunogenicity. These findings provided the rationale to evaluate whether combination therapy could enhance immune responses in vivo.

AT3OVA is a well-characterized syngeneic mouse model of TNBC and was used to examine whether immune responses were important for the therapeutic effect following combined PI3Kα and CDK4/6 inhibition in vivo. AT3OVA tumor–bearing immune-competent mice were treated with BYL719, LEE011, the combination, or vehicle control. We observed that combined treatment exerted greater control of AT3OVA tumor growth compared with single-agent or vehicle treatments (Fig. 5A).

We next characterized the effect of treatment on immune cell populations in the tumor microenvironment by flow cytometry. Compared with single-agent and vehicle treatment cohorts, the combined treatment markedly enhanced infiltrating CD8⁺ and CD4⁺ T-cell activity as indicated by an increase in CD69, granzyme B, and IFNγ expression (Fig. 5B and C). Further evidence of enhanced T-cell activation following combined therapy was indicated by a significant increase in coexpression of programmed cell death protein 1 (PD-1) and CTLA-4 immune-checkpoint proteins in CD8⁺ and CD4⁺ T cells compared with single-agent and vehicle treatment groups (Fig. 5B and C). Single-agent or combination treatments were not observed to affect proliferation or frequency of tumor-infiltrating CD8⁺ and CD4⁺ T cells (Supplementary Fig. S2A and S2B).

In addition to activating adaptive immune cells, combined BYL719 and LEE011 treatment significantly increased the frequency of tumor-infiltrating natural killer T (NKT) cells, and there was a trend for increased mature NK cells (NK1.1⁺/CD11b^high/CD27^low) compared with single-agent LEE011 and vehicle treatments (Supplementary Fig. S2C). Granzyme B expression by mature NK cells was also observed to be significantly greater with combination treatment compared with BYL719 and LEE011 single-agent and vehicle treatment groups, indicating an increase in their cytotoxic potential (Supplementary Fig. S2C).

Interestingly, combined BYL719 and LEE011 treatment was effective in significantly reducing the frequency of immunosuppressive monocytic myeloid-derived suppressor cells (mMDSC; Fig. 5D). We also observed a significant decrease in proliferation of CD4⁺ FOXP3⁺ regulatory T cells (Treg) following combination therapy as indicated by the significant decrease in expression of the Ki67 proliferative marker (Supplementary Fig. S2B).

In summary, combined PI3Kα and CDK4/6 inhibition resulted in enhanced innate and adaptive antitumor immune responses. The increase in expression of both PD-1 and CTLA-4 on T cells suggested that the addition of immune-checkpoint blockade could further augment the antitumor therapeutic effect of PI3Kα and CDK4/6 inhibition.

We next examined the efficacy of combining BYL719 and LEE011 with immune-checkpoint blockade in the AT3OVA model. We tested anti–PD-1 and anti–CTLA-4 mAbs given that these immune checkpoints were found to be upregulated on T cells following PI3Kα and CDK4/6 inhibition and that they are currently being evaluated in clinical trials. Anti–PD-1 and anti–CTLA-4 treatments were administered on day 1 along with BYL719 and LEE011 treatments in tumor-bearing mice. All treatments ceased by day 50.

We observed that tumors treated with anti–PD-1 and anti–CTLA-4 as single agents performed worse in terms of tumor growth control and survival compared with BYL719 and LEE011 combination treatment (Fig. 6A). However, we found that the quadruple combination treatment group (comprising BYL719, LEE011, anti–PD-1, and anti–CTLA-4) performed the best among all treatment groups (Fig. 6A). All tumors (10/10) treated with the quadruple combination remained regressed during the 50 days of treatment; whereas tumor growth continued in all other treatment groups despite being under constant drug pressure (Supplementary Fig. S3). Of note, 50% (5/10) of tumors treated with the quadruple combination remained regressed for more than a year, even after treatment cessation, with resultant significant improvement in survival compared with BYL719 and LEE011 as well as anti–PD-1 and anti–CTLA-4 treatment groups (Fig. 6A). Notably, there was no toxicity observed in mice following the full treatment schedule.

In terms of mechanism, we observed a significant increase in MHCI (H2KB) expression on AT3OVA tumor cells following combined BYL719 and LEE011 treatment in vivo (Fig. 6B), validating our results that BYL719 and LEE011 treatment increases tumor immunogenicity in vitro (Fig. 4C). Although changes in MHCII expression were less impressive with combined BYL719 and LEE011 treatment, similar to results obtained in vitro, we show that addition of immune-checkpoint blockade in the quadruple combination treatment group resulted in a pronounced and significant increase in MHCII as well as PDL-1 expression on AT3OVA tumor cells compared with other treatment groups (Fig. 6B). Moreover, we found that granzyme B expression in CD8⁺ and CD4⁺ T cells was significantly increased following quadruple combination treatment compared with all other treatment groups (Fig. 6C). These results support that the addition of immune-checkpoint blockade further augments tumor immunogenicity and subsequent T-cell cytotoxicity responses following PI3Kα and CDK4/6 inhibition in TNBC.

TNBC remains the most challenging breast cancer subtype to treat with the poorest prognosis and median survival after relapse. There are currently no targeted therapies approved for the treatment of TNBC, although Olaparib in BRCA1 germline-deficient breast cancers may soon see this change (31). The majority of TNBC, however, are BRCA1/2 wild-type, and their prognosis remains poor particularly after relapse. New therapies or drug combinations that can achieve durable disease control are needed. Immunotherapy, in particular immune-checkpoint blockade, has shown remarkable success in other solid cancer types; however, response rates in metastatic TNBC patients remain low compared with those observed for immunogenic diseases such as melanoma (32, 33).

In this study, we have shown synergistic antitumor effects for the combination of PI3Kα (BYL719) with CDK4/6 (LEE011, PD991) inhibitors in a variety of TNBC preclinical models including basal-like, mesenchymal, mesenchymal-like, and LAR TNBC subtypes. Cooperative increase in cell-cycle arrest, DNA damage, replicative stress, and apoptosis all contributed to the antitumor effects of combined PI3Kα and CDK4/6 inhibition in TNBC.

A recent report suggested synergism for the CDK4/6 inhibitor (PD991) with two PI3K inhibitors (Pictilisib, GDC-0941 and Taselisib, GDC-0032) for only TNBC cell lines that are of the luminal-like subtype or those that harbor activating mutations in PIK3CA (encoding PI3 kinase catalytic α subunit, PI3Kα; ref. 34). However, our data suggest wide-ranging synergism in a variety of TNBC models in vivo and in vitro, regardless of PIK3CA mutation status, with PI3Kα inhibitor (BYL719) together with either CDK4/6 inhibitors (LEE011 or PD991). The PI3K inhibitors vary markedly in their potency across the PI3K catalytic isoforms (PI3Kα/β/γ/δ), and we hypothesize that this may explain the differences in findings. GDC-0941 is a pan-isoform PI3K inhibitor, whereas GDC-0032 is a β isoform–sparing PI3K inhibitor (IC₅₀ against PI3Kβ = 9.1 nmol/L) and has potent inhibitory effects on PI3Kα (IC₅₀ = 0.29 nmol/L), PI3Kγ (IC₅₀ = 0.97 nmol/L), and PI3Kδ (IC₅₀ = 0.12 nmol/L) isoforms (35, 36). Whereas BYL719 is a selective PI3Kα inhibitor, exhibiting the highest potency against PI3Kα activation (IC₅₀ against PI3Kα = 19 nmol/L) compared with other isoforms such as PI3Kβ (IC₅₀ = 1,156 nmol/L), PI3Kγ (IC₅₀ = 320 nmol/L), and PI3Kδ (IC₅₀ = 290 nmol/L; ref. 37). Furthermore, we differed in our approaches in compound synergy assessment and have additionally evaluated synergistic effects on apoptosis and cell-cycle arrest.

The LAR subtype of TNBC has been previously shown to be susceptible to combined PI3K and CDK4/6 pathway inhibition due to the presence of PIK3CA mutations (14, 34). Moreover, tumors of the LAR subtype have been shown to be heavily enriched in hormonally regulated pathways that include androgen/estrogen metabolism as well as display luminal gene expression patterns, similar to ER-positive luminal breast cancers (23). RB has been shown to have a highly specific and indispensable role in mediating antitumor responses to CDK4/6 inhibition (6, 38, 39). Consistent with other reports (34), the antagonistic interaction of the two compounds in the RB1-mutant cell line MDA-MB-468 further suggests that RB remains essential for synergistic interactions following PI3K and CDK4/6 pathway inhibitions in TNBC. In The Cancer Genome Atlas study, the percentage of TNBC tumors where RB1 was mutated or lost was reported to be only 20% (11), indicating that combined PI3Kα and CDK4/6 inhibitor treatment could be beneficial to the majority of TNBC patients.

Gene expression analyses revealed that combined PI3Kα and CDK4/6 inhibition in TNBC significantly downregulates genes involved in multiple key prosurvival cellular processes including DNA damage and repair responses. We have shown that this corresponds with an increase in DNA damage. The observed increased cyclin E1 expression in addition to reduced expression of DNA repair genes with combined PI3Kα and CDK4/6 inhibitor treatment could have contributed to further catastrophic genomic instability in our models (40, 41).

The clinical importance of tumor-infiltrating lymphocytes (TIL) has been an emerging area of research in breast cancer, particularly in TNBC where higher levels of TILs are associated with both increase in response to chemotherapy and improved survival (2, 15). However, many patients with advanced TNBC have few or no TILs present at diagnosis. Therefore, our novel observation that combined PI3Kα and CDK4/6 inhibition could effectively promote antitumor immunity in TNBC was of high interest. To our knowledge, we have shown for the first time that combined PI3Kα and CDK4/6 inhibition increases tumor antigen presentation as well as immunogenic cell death, which would potentially facilitate the eradication of tumor cells by the immune system (30).

Further evaluation of immune responses to PI3Kα and CDK4/6 inhibition in mouse immunocompetent environment revealed that combination treatment in vivo results in increased activation and cytotoxicity of both adaptive and innate immune cell populations as well as decreased frequency of immune-suppressive MDSCs within the tumor environment. The observation of increased antitumor immunity as a result of suppressing PI3K and CDK4/6 pathways was intriguing as both these pathways have been long thought to be important for immune cell function and proliferation (42, 43). Nonetheless, recent studies have shown that adaptive and innate effector cell function and survival were not affected with PI3Kα inhibition (44, 45). Moreover, our work, along with other reports, has highlighted that various PI3K isoforms as well as some CDKs could be essential for modulation of the immune-suppressive tumor microenvironment (46–49). These studies have demonstrated that blockade of PI3Kδ or γ isoforms reduces the function of Tregs and MDSCs to a greater extent compared with effector T cells, thereby maintaining better control of tumor growth (48, 49). Interestingly, in the clinical setting, neutropenia is reported as an on-target side effect of CDK4/6 inhibition (50, 51). This effect could conceivably contribute to our novel observation of decreased MDSC frequency with PI3Kα and CDK4/6 inhibition. The functional effect of combined therapy on regulatory cell types such MDSCs and Tregs warrants further investigation in future studies.

Taken together, our work suggests that dual inhibition of PI3Kα and CDK4/6 acts in concert to promote synergistic disease control, apoptosis, and antitumor immunity. Importantly, further combination with immune-checkpoint blockade significantly increased tumor immunogenicity and T-cell cytotoxicity, resulting in effective and durable (>1 year) tumor regression with significantly better overall survival compared with all other treatment groups.

In summary, our findings suggest that combined PI3Kα and CDK4/6 inhibition is synergistic and immunogenic against a diverse range of PIK3CA- and RB1-wild-type preclinical TNBC models. We also conclude that antitumor efficacy can be further enhanced with the addition of immune-checkpoint blockade. Our results have identified a novel approach that warrants evaluation in clinical trials for patients in critical need of new effective therapies.

S. Loi has received funding to her institution from Novartis and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Z.L. Teo, W.A. Phillips, P.K. Darcy, S. Loi

Development of methodology: Z.L. Teo, B. Virassamy, S. Loi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z.L. Teo, S. Versaci, S. Dushyanthen, C.P. Mintoff, B. Virassamy, S. Loi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z.L. Teo, S. Dushyanthen, F. Caramia, P. Savas, M. Zethoven, B. Virassamy, G.A. McArthur, P.K. Darcy, S. Loi

Writing, review, and/or revision of the manuscript: Z.L. Teo, P. Savas, M. Zethoven, S.J. Luen, G.A. McArthur, W.A. Phillips, P.K. Darcy, S. Loi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z.L. Teo, F. Caramia, B. Virassamy, S. Loi

Study supervision: P.K. Darcy, S. Loi

The authors wish to acknowledge staff from the Peter MacCallum Cancer Centre Animal Facility, FACS Facility, and Molecular Genomics Core Facility for their assistance.

This work was supported by a National Health and Medical Research Council (NHMRC) of Australia Project Grant (1123208; S. Loi, P.K. Darcy, W.A. Phillips). Z.L. Teo was supported by a NHMRC Early Career Fellowship (1106967). P.K. Darcy was supported by a NHMRC Senior Research Fellowship (1041828). S. Loi was supported by the Cancer Council Victoria Australia John Colebatch Fellowship as well as the Breast Cancer Research Foundation NY. S. Loi has received funding to her institution from Novartis and Pfizer.

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.