A bedside-to-bench translational analysis of NF1 alterations and CDK4/6 inhibitor resistance in hormone receptor-positive metastatic breast cancer

Abstract

Background: CDK4/6 inhibitors (CDK4/6i) are used for management of hormone receptor-positive (HR+) metastatic breast cancer (MBC), and activation of the RAS/MAPK and PI3K/AKT signalling pathways has been implicated in resistance to these agents. Pathogenic NF1 mutations (pNF1m) dysregulate RAS signalling, but NF1 has not been linked to CDK4/6i resistance. We analysed multi-institutional data, real-world evidence, and preclinical models to characterise the impact of pNF1m on CDK4/6i sensitivity.

Methods: A retrospective cohort of patients with pNF1m tumours were identified from 4 institutions between 2/2015-5/2023 and evaluated for progression-free survival and intrinsic/acquired resistance on CDK4/6i. Real-world clinical-genomic data from GuardantINFORM between 6/2014 and 3/2023 was analysed for associations between pNF1m and time-to-next-treatment or overall survival following CDK4/6i, adjusted using propensity score weighting. We used CRISPR/Cas9 to delete NF1 in MCF7 and T47D breast cancer cells in vitro. NF1-knockout (NF1-KO) and -wild-type (WT) cells were analysed with respect to CDK4/6i sensitivity, MAPK and PI3K pathway activation, and sensitivity to MAPK and PI3K pathway inhibitors. In parallel, we assessed treatment response in a patient-derived organoid (PDO) harbouring NF1 loss, established from an HR+/HER2- breast tumor following progression on a CDK4/6i.

Findings: Among 1962 multicentre patients, we identified 38 with HR+/HER2- MBC, pNF1m, and exposure to CDK4/6i. NF1-associated intrinsic or acquired resistance to CDK4/6i was observed in a majority of tumours, and in those with baseline pNF1m on first-line CDK4/6i, a median progression-free survival of 6.2 months was much less than expected in routine practice. Real-world weighted analysis of 1161 patients comparing 28 pNF1m to 1133 NF1 non-altered tumours demonstrated shorter time-to-next-treatment on CDK4/6i regimens (4.2 vs. 12.4 months, hazard ratio 3.14, 95% confidence interval 2.01-4.93) and overall survival (15.8 vs. 45.2 months, hazard ratio 2.04, 95% confidence interval 1.09-3.82). NF1-deleted cells exhibited reduced sensitivity to CDK4/6i with or without oestrogen suppression, which was accompanied by induction of both MAPK and PI3K pathways, the latter of which was exacerbated by CDK4/6i. Blockade of RAS or AKT, but not MEK or ERK, reversed CDK4/6i resistance mediated by NF1 loss in cell lines and the PDO.

Interpretation: NF1 mutations are associated with shorter therapy duration on CDK4/6i in MBC. A causal link between NF1 loss and CDK4/6i resistance was supported by experiments in HR + breast cancer cells. NF1 deletion was accompanied by activation of ERK and AKT, and blockade of RAS or AKT combined with CDK4/6i was effective in NF1-deleted cells and an NF1-mutant PDO.

Funding: Breast Cancer Research Foundation DRC-20-001, National Cancer Institute R01CA273246, National Institute of Health P30 CA142543, Susan G. Komen Breast Cancer Foundation SAB1800010, Department of Defence BC 210406, Mary Kay Ash Foundation International Postdoctoral Scholars in Cancer Research Fellowship.

Keywords: CDK4/6 inhibitor; HR+ breast cancer; Metastatic breast cancer; NF1 mutation; Precision medicine; Resistance mechanisms.

Fig. 1

Study flow diagrams for patient identification, inclusion, and exclusion in the multicentre clinical cohort (a) and real-world cohort (b). CDK4/6i, CDK4/6 inhibitor; G360, Guardant 360; HR+, hormone receptor-positive; MBC, metastatic breast cancer; NGS, next-generation sequencing; pNF1m, pathogenic NF1 mutation.

Fig. 2

Clinical outcomes in multicentre cohort patients with a pathogenic NF1 mutation detected within 3 months of starting a CDK4/6 inhibitor therapy regimen. a. Progression-free survival in 22 patients treated with a CDK4/6 inhibitor during any line of therapy for metastatic disease. b-c. Subgroup results of progression-free survival in 12 patients who received a CDK4/6 inhibitor first-line for metastatic breast cancer (b), and in 10 patients treated in the second-or-later-line setting (c). Progression-free survival (with 95% CI) was estimated using the Kaplan–Meier method. Vertical dashed lines indicate censored data. d. Detailed clinical-genomic vignettes for 6 patients with intrinsic resistance (<6 months) on a first-line CDK4/6i regimen. Vignettes include time of primary and metastatic diagnosis, treatment sequencing, and key genomic information from ctDNA and/or tumour tissue biopsy specimens. 95% CI, 95% confidence interval; Bx, biopsy; ctDNA, circulating-tumour DNA; CDK4/6i, CDK4/6 inhibitor; mo, months; NR, not reached; PFS, progression-free survival.

Fig. 3

Clinical-genomic characteristics of 38 patients with pNF1m metastatic breast tumours in the multicentre clinical cohort. a. Tile plot depicting the concurrent genomic alterations detected on next-generation sequencing of tissue and/or blood specimens. Genes were included in the plot if > 2 patients had a tumour that harboured an alteration in that gene. Patients are broadly organized by NF1-associated CDK4/6 inhibitor resistance phenotypes (intrinsic resistance, acquired resistance, sensitive), and recursive ordering for gene alteration frequency. b. Bar graph demonstrating the percentage of NF1 alterations by type found in the 27 IR/AR tumours. AR, acquired resistance; IR, intrinsic resistance; NGS, next-generation sequencing; Pct, percent; pNF1m, pathogenic NF1 mutation; S, sensitive.

Fig. 4

Propensity score weighted time to event analysis of 1161 real-world cohort patients. Survival curves are depicted comparing rwTTNT (a) and rwOS (b) between 28 patients with pNF1m breast tumours vs. a weighted, non-NF1m control group treated with CDK4/6i-containing therapy. Propensity score weighting used covariates of age, sex, line-of-treatment, and year of G360 test to achieve balance. Survival curves were compared using the weighted log-rank test, and weighted HRs with 95% CIs were estimated using a Cox proportional hazards model. Vertical dashed lines indicate censored data. 95% CI, 95% confidence interval; HR, hazard ratio; mo, months; non-NF1m, NF1 non-mutant; pNF1m, pathogenic NF1 mutation; PSW, propensity score weighted; rwTTNT, real-world time-to-next-treatment; rwOS, real-world overall survival.

Fig. 5

NF1 deletion is causal to CDK4/6i resistance in HR + breast cancer cell lines. a. Immunoblot analysis of NF1^WT and NF1^KO MCF7 and T47D cells cultured in oestrogen-free media (0% CSS). Lysates were probed with the indicates antibodies. b. Cell proliferation of NF1^WT and NF1^KO MCF7 and T47D cells grown in oestrogen-free media (10% CSS) for 6 days was measured by Incucyte live-cell analysis. c-f.NF1^WT and NF1^KO MCF7 and T47D cells were treated with 9 concentrations of palbociclib in full media (c, d) or in oestrogen-depleted media (10% CSS) (e, f) for 6 days. The number of cells per well at the end of treatment was measured by Incucyte live-cell analysis. Data represent mean ± SD (n = 3 technical replicates).

Fig. 6

NF1-deficient breast cancer cells are sensitive to RAS or AKT inhibition in combination with palbociclib. a-d.NF1^KO MCF7 (a and b) and T47D (c and d) cells were treated for 6 days with increasing concentrations (3-fold) of palbociclib, capivasertib, RMC-6236 or trametinib up to 10 μM. For combination assays, all cells were treated with 250 nM palbociclib and increasing concentrations of the second drug. The number of cells per well at the end of treatment was measured by Incucyte live-cell analysis. Data represent mean ± SD (n = 3 technical replicates). Data from palbociclib vs. palbociclib in combination with capivasertib or RMC-6236, capivasertib vs. capivasertib + palbociclib 250 nM, and RMC-6236 vs. RMC-6236+palbociclib 250 nM were analysed with ANOVA multiple comparisons. e-h. Growth of a patient-derived organoid (PDO) 33682 under different treatment conditions. PDOs from an NF1-mutant, CDK4/6 inhibitor-resistant patient were cultured for 22 days in the presence of vehicle (DMSO), palbociclib 250 nM or 1000 nM, and the combination of palbociclib 250 nM with increasing concentrations of capivasertib, RMC-6236, or trametinib. Organoid proliferation in 3D was assessed using the CellTiter-Glo 3D assay. Bar graph represents mean organoid growth relative to vehicle control (e) or to palbociclib 250 nM (f–h), with error bars indicating ±SD. Data correspond to four technical replicates. ∗, p < .05; ∗∗, p < .01; ∗∗∗, p < .001; ∗∗∗∗, p < .0001.