Depleting the Action of EZH2 through PI3K-mTOR Inhibition to Overcome Metastasis and Immunotherapy Resistance in Triple-Negative Breast Cancer

Abstract

Almost half of patients with triple-negative breast cancer develop distant metastases, heralding unfavorable outcomes. Here, we provide novel insights into the contribution of the PI3K-mTOR pathway to the triple-negative breast cancer phenotypes that promote growth, migration, metastasis, and therapy resistance. Specifically, we demonstrate that dual targeting of PI3K and mTOR but not PI3K alone inhibits cancer cell proliferation and migration in vitro. Dual PI3K-mTOR inhibition with paxalisib not only promotes a favorable mesenchymal-to-epithelial phenotype but also inhibits signatures associated with metastasis-initiating cells, including the highly aggressive cancer stem cell phenotype, persister cancer cell phenotype (p65, FOXQ1, NRF2, and NNMT), and a cancer drug resistance signature (ABCB5, SNAIL, and ALDH1). In vivo, paxalisib overcomes immunotherapy resistance to reduce primary tumor burden, circulating tumor cells, and direct and indirect indicators of metastasis with a favorable toxicity profile. Gene expression and spatial analyses show that paxalisib profoundly affects the immune microenvironment in tumors, reducing adaptive immune phenotypes associated with immunotherapy resistance (exhausted T cells and regulatory T cells) and protumor innate immune populations such as mast cells. PI3K-mTOR blockade acts upstream of EZH2, impacting both the classic repressive catalytic p85β-EZH2-H27ME3 and active EZH2-NF-κB pathways. Our data suggest that dual targeting of the PI3K-mTOR pathway disrupts both the catalytic and noncatalytic axes of EZH2 to inhibit metastasis and enhance cancer immune visibility, potentially increasing the utility of immunotherapy in resistant individuals.

Figure 1.

Dual PI3K–mTOR blockade with paxalisib inhibits proliferation and migration, induces mesenchymal-to-epithelial transition, and reduces MIC signatures. A, WST proliferation reagent was added to MDA-MB-231 cells pretreated with PI3K and PI3K–mTOR inhibitors for 72 hours in a dose-response design. Cell proliferation (%) was measured indirectly by the formation of formazan and the absorbance recorded at 450 nm. B, Representative images from MDA-MB-231 cells scratched prior to treatment with paxalisib for 6, 12, and 18 hours. Bar graph comparing the relative wound densities (%) observed for PI3K–mTOR inhibition at each time point, ****, P < 0.0001, vs. vehicle, two-way ANOVA with the Bonferroni multiple comparisons test, n = 8/group. Representative images and IF analysis of (C) vimentin and (D) E-cadherin in 24 hour paxalisib-treated MDA-MB-231 cells, ****, P < 0.0001; ***, P < 0.001 vs. vehicle, unpaired t test, n ≥ 100 cells/group. E, Flow cytometry analysis showing CD44:CD24 expression (%vehicle) in 24-hour paxalisib-treated MDA-MB-231 cells. F, IF analysis of ALDH1, ABCB5, and Snail protein expression in 24-hour paxalisib-treated MDA-MB-231 cells, ****, P < 0.0001, vs. vehicle, unpaired t test, n ≥ 100 cells/group. G, IF analysis of nuclear and cytoplasmic NF-κB p65, FOXQ1, NRF2, and NNMT protein expression in 24-hour paxalisib-treated MDA-MB-231 cells, *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, vs. vehicle, unpaired t test, n ≥ 100 cells/group. H, IF analysis of IL-6 protein expression (%total cells) in paxalisib-treated ABCB5⁺/EpCAM⁺ CTCs isolated from patients with TNBC, *, P < 0.05, vs. vehicle, paired t test, n = 2/group. I, qPCR analysis of viral mimicry-associated gene expression in 24-hour paxalisib-treated MCF-7 iEMT cells, *, P < 0.05, vs. vehicle, unpaired t test, n = 2/group. J, Comparison of GBP2 mRNA expression (%vehicle) following 24-hour PI3K and PI3K–mTOR treatment in MCF-7 iEMT cells, vs. vehicle, Dunnett multiple comparisons test, n = 2/group.

Figure 2.

Paxalisib reduces primary tumor burden and metastasis with a favorable toxicity profile. A, Treatment regimens using the BALB/c 4T1 TNBC breast cancer model (paxalisib with ± anti-PD1, with or without Abraxane). B, Primary tumor volumes (%vehicle) and (F) weights of individual mice from each experimental group prior to harvest, *, P < 0.05; **, P < 0.01; ***, P < 0.001, vs. vehicle, Dunnett, n = 4–5/group. C, Left, Changes in liver inflammation and hepatocyte damage (metabolic and/or degeneration) were scored 1 = mild, 2 = moderate, or 3 = severe for each parameter, *, P < 0.05; ****, P < 0.0001, vs. vehicle, Dunnett, n = 4–5/group. Middle, changes in lung weights and leukocytosis, *, P < 0.05; ***, P < 0.001; ****, P < 0.0001, vs. vehicle, Dunnett, n = 4–5/group. Right, changes in spleen weights and spleen extramedullary hematopoiesis (EMH), *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, vs. vehicle, Dunnett, n = 4–5/group. D, Paxalisib combined with αPD1 experiments using the BALB/c 4T1 intravenous metastasis breast cancer model. E, Mouse body weights were monitored throughout the experiments. F, IF images of CD45, vimentin (VIM), and SNAIL1 in MDA-MB-231 cells spiked into human healthy PBMCs. BF, Brightfield (BF). G, CTCs were isolated using ScreenCell Cyto kits from mouse blood collected from the 4T1 intravenous metastasis model, followed by Giemsa staining. H, Captured CTCs were counted on Cyto IS membranes. I, Images of Indian ink–stained lungs highlighting metastases, including large, protruding and small, flat nodules and graph shows the number of flat and protruding metastatic lung nodules, **, P < 0.01; ****, P < 0.0001, vs. vehicle, two-way ANOVA with Tukey multiple comparisons test, n = 2/group.

Figure 3.

CODEX multiplexed imaging reveals changes in immune cell landscapes when paxalisib is combined with immunotherapy. A, Normalized expression of 23 proteins in 14 identified cell types. B, Percentages of indicated cells types out of total cells in stromal, adaptive immune, innate immune, and tumor populations in αPD1 and paxalisib + αPD1–treated tumors. C, Differential location of stromal, tumor, adaptive, and innate cell populations in αPD1 and paxalisib + αPD1–treated tumor tissues. D, Representative images of neighborhoods mapped to tumor tissues from αPD1 and paxalisib + αPD1–treated tumors. E, Quantification of neighborhood fraction in αPD1 and paxalisib + αPD1–treated tumors. APC, APC, antigen-presenting cell.

Figure 4.

PI3K–mTOR inhibition enhances antitumor immune profiles in combination with immunotherapy. A, NanoString nCounter cell abundance analysis of DC, cytotoxic cell, NK cell, and T- and B-cell immune populations from paxalisib ± anti-PD1–treated tumors, n = 3/group. B, NanoString nCounter cell abundance analysis of tumor-infiltrating lymphocyte populations from paxalisib ± anti-PD1–treated tumors, n = 3/group. C, NanoString nCounter gene expression levels of cytotoxicity-related cytokines, including IFNG and granzyme B (GZMB), n = 3/group. D, NanoString nCounter cell abundance analysis of mast cell immune populations from paxalisib ± anti-PD1–treated tumors, n = 3/group. E, Toluidine blue staining for mast cell detection in anti-PD1 and paxalisib + anti-PD1–treated tumors.

Figure 5.

PI3K–mTOR inhibition targets p85β:H3K27Me3 switch. A, IF images of p85β and H3K27Me3 in MDA-MB-231 cells. B, IF analysis of p85β and H3K27Me3 nuclear intensity staining in paxalisib-treated MDA-MB-231 cells and MDA-MB-468 cells, *, P < 0.05; ****, P < 0.0001, vs. vehicle, unpaired t test.

Figure 6.

PI3K–mTOR inhibition targets the dual role of EZH2. A, qPCR analysis of EZH2 mRNA expression in paxalisib-treated MDA-MB-231 cells ****, P < 0.0001, unpaired t test. B, IF analysis of EZH2 nuclear intensity staining in paxalisib-treated MDA-MB-231 cells ****, P < 0.0001, unpaired t test. C, Duolink analysis of the p85β:EZH2 interactions in paxalisib-treated MDA-MB-231 cells, ****, P < 0.0001, unpaired t test. NS, non-stimulated. D, Duolink analysis of the NF-κB:EZH2 interactions in paxalisib-treated MDA-MB-231 cells, ****, P < 0.0001, unpaired t test. E, Overlap between the EZH2 bound upregulated and downregulated genes from the three comparisons (anti-PD1, paxalisib, and paxalisib + anti-PD1, relative to control). DE genes were defined as log₂ fold change > 0.5 and P value < 0.05. Blue * indicated genes were also bound with H3K27me3. F, Distribution of EZH2-bound genomic locations that are part of the nCounter Tumor 360 Signaling Panel. G, EZH2 and NF-κB2 ChIP-seq tracks at IL-6 promoters in MDA-MB-231 cells. UTR, untranslated region.

Figure 7.

Impact of dual PI3K–mTOR pathway inhibition in TNBC. Mechanism 1. PI3K–mTOR blockade inhibits the resistant cancer cell populations (dormant tumor cells, CSCs, and persister cells), thereby overcoming resistance and reducing inflammation, including key proinflammatory cytokines such as IL-6. PI3K–mTOR inhibition also increases immune reinvigoration and activates viral mimicry signatures, making the cancer cells more immune visible. Using this biphasic approach of targeting resistance signatures and enhancing cancer immune visibility, when combined with immunotherapy or PARP inhibitors, PI3K–mTOR inhibition reduces primary tumor burden and metastases. Mechanism 2. PI3K–mTOR blockade inhibits the dual role of EZH2. PI3K–mTOR blockade (e.g., with paxalisib) acts upstream of EZH2, thereby targeting its dual role in resistance. PI3K–mTOR inhibitors potentially target EZH2 by two mechanisms. First, by inhibiting p85β translocation into the nucleus, which diminishes the p85:EZH2 interaction, thereby inhibiting the catalytic repressive role of the epigenetic enzyme resulting in reduced H3K27 trimethylation. By targeting upstream of EZH2, at the same time PI3K–mTOR blockade inhibits the noncatalytic inducible role of EZH2, thereby reducing NF-κB signaling and downstream targets. Alternatively, PI3K–mTOR inhibition directly impacts EZH2 transcription, which in turn inhibits its dual catalytic and noncatalytic roles.