Inhibition of ACAA1 Restrains Proliferation and Potentiates the Response to CDK4/6 Inhibitors in Triple-Negative Breast Cancer

Triple-negative breast cancer (TNBC) is characterized by the absence of the estrogen receptor (ER) and progesterone receptor and the lack of HER2 overexpression (1). TNBC accounts for approximately 10%–20% of newly diagnosed breast cancers and is associated with aggressive behavior and poor prognosis (2). In the past decades, advances in omics technologies have led to a comprehensive view of TNBC tumors and microenvironment heterogeneity (3). Although multiple compounds have been evaluated in clinical trials, chemotherapy remains the mainstay systemic treatment for TNBC (3). Currently, only a limited number of targeted agents have been approved for patients with breast cancer, including TNBCs. For example, PARP inhibitors olaparib and talazoparib for patients with HER2⁻ breast cancer carrying deleterious germline in breast cancer mutations (4–6), immune checkpoint inhibitors pembrolizumab and atezolizumab for PD-L1–positive patients with TNBC (7–9), antibody–drug conjugate (ADC) sacituzumab govitecan-hziy (Trodelvy) for third-line treatment of unresectable locally advanced or metastatic TNBC (10), and fam-trastuzumab-deruxtecan-nxki (Enhertu) for HER2-low breast cancer (11). Given that the current therapeutic scenario does not benefit all TNBCs and treatment efficacy requires further improvement, exploring novel treatment strategies in TNBC remains a major challenge.

In our previous study, we constructed a cohort of 465 patients with primary TNBC and classified TNBC into four transcriptome-based subtypes (12). We further revealed the potential therapeutic targets of each subtype and conducted a phase Ib/II clinical trial (FUTURE, ClinicalTrials.gov, number: NCT03805399) to evaluate the efficacy and safety of the proposed treatment strategies (13). Although some patients experienced favorable outcomes, a low objective response rate was observed in patients with the luminal androgen receptor (LAR) subtype treated with androgen receptor (AR) inhibitor and cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitor. Currently, CDK4/6 inhibitors are not the standard of care for TNBC, and the application of CDK4/6 inhibitors is more challenging in patients with TNBC than in ER⁺ patients with breast cancer due to the high frequency of RB transcriptional corepressor 1 (RB1) loss/mutations and high expression of CDK2, CDK6, CCNE1, and CDKN2A in TNBCs (14). However, increasing evidence suggests that a proportion of patients with TNBC may benefit from CDK4/6 inhibitors, although the underlying mechanism remains unclear (15–17). Therefore, there is an urgent need for studies exploring potential biomarkers to select patients with TNBC most likely to benefit from CDK4/6 inhibitors and to identify novel therapeutic targets.

We conducted a proteome-centric study to explore potential protein targets for TNBC, especially for the LAR subtype, that showed a low response rate to AR inhibitor and CDK4/6 inhibitor treatment in the FUTURE trial (13, 18). We observed high expression of acetyl-CoA acyltransferase 1 (ACAA1) and other key molecules associated with fatty acid metabolism in the integrative proteomic (iP)-2 subtype, that is mainly comprised of the transcriptomic LAR subtype. In this study, we uncovered the biological functions and mechanisms by which ACAA1 promotes TNBC proliferation and diminishes the tumor response to the CDK4/6 inhibitor abemaciclib and explored the potential of ACAA1 as a therapeutic target in TNBC.

Proteomic profiles of 82 TNBC tissues and 66 paired adjacent normal tissues were obtained from patients who were treated at Fudan University Shanghai Cancer Center (FUSCC; Shanghai, China) from July 2010 to September 2014 (18). Ten pairs of primary breast tumor tissues and adjacent normal tissues were obtained from patients with TNBC who underwent surgery at FUSCC from April 2017 to August 2018. In addition, fresh tumor samples from three patients with TNBC who underwent surgery at FUSCC between October 13, 2021, and December 2, 2021, were collected for the generation of patient-derived organoids. Public datasets were obtained from Kaplan–Meier plotter (http://kmplot.com/analysis/), FUSCC TNBC, and The Cancer Genome Atlas (TCGA) for analysis (12, 14, 18–21).

The human mammary epithelial cell lines MCF10A and HMEC, human TNBC cell lines BT-549, HCC70, Hs-578T, MDA-MB-453, HCC38, HCC1143, MDA-MB-157, MDA-MB-436, MDA-MB-468, BT-20, HCC1937, CAL-148, and human embryonic kidney cells (HEK293T) were obtained from ATCC. MFM-223 cells were obtained from the European Collection of Authenticated Cell Cultures. The human breast cancer cell line MDA-MB-231 and its lung-tropic derivative LM2–4175 were gifted from Professor Guohong Hu (Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences, Shanghai, China). Only cells within 6 months of being thawed were used and maintained at 37°C with 5% carbon dioxide (CO₂). All cell lines were authenticated by short tandem repeat profiling and tested free of Mycoplasma contamination.

Generation of patient-derived organoids was performed as described in a previous study (22). Breast cancer tissues were cut into 1–3 mm³ pieces and digested in collagenase (Sigma).

Abemaciclib mesylate (S7158), trimetazidine dihydrochloride (S4543), MG-132 (S2619), bafilomycin A1 (Baf-A1, S1413), and 3-methyladenine (3-MA, S2767) were purchased from Selleck. Ammonium chloride (NH₄Cl, 213330) was purchased from Sigma. Protease and phosphatase inhibitors (Roche) were purchased from Merck. All primary and secondary antibodies used in this study are summarized in Supplementary Table S1.

Short-term viability assays were performed by using an IncuCyte ZOOM HD/2CLR system or Cell Counting Kit-8 (CCK-8) assays. For the IncuCyte system, cells were seeded in 96-well plates and imaged using the IncuCyte ZOOM system (Essen BioScience). Frames were captured at 12-hour intervals in three separate regions per well. Cultures were maintained in a 37°C incubator with 5% CO₂, and the growth rate was analyzed using IncuCyte software (2013A Rev2). For the CCK-8 assay, cells were seeded in 96-well plates and incubated with 10% CCK-8 solution (Dojindo, CK04) at 37°C for 2 hours, and then the absorbance was measured at 450 nm.

For colony formation assays, cells were seeded into 6-well or 24-well plates and cultured for 2 weeks. The cells were rinsed with PBS, fixed and stained with a 0.25% crystal violet solution in methanol for 30 minutes for colony counting.

The cells were stained with propidium iodide (BD Biosciences) for 30 minutes at 37°C and were analyzed using a CytoFLEX S Flow Cytometer (Beckman Coulter) with Modfit LT5.0 software.

Cells were seeded into 96-well plates (1–6 × 10³ cells/well) and incubated 24 hours before the administration of drugs. The growth medium was then replaced with a medium containing sterile water or increasing concentrations of abemaciclib (9 doses ranging from 0.078 μmol/L to 20 μmol/L, 2-fold dilution) or trimetazidine (9 doses ranging from 0.156 mmol/L to 40 mmol/L, 2-fold dilution), and plates were incubated for three days. Growth was quantified using the CCK-8 assay (Dojindo, CK04) according to the manufacturer’s instructions. The IC₅₀ values were determined by probit regression using the Logit model with SPSS version 22.0 (SPSS). The combination index (CI) value was calculated to evaluate the effects of the two drugs using the Chou–Talalay method with CompuSyn software (CI > 1, antagonism; CI = 1, addition; CI < 1, synergism; ref. 23).

The pcDNA3.1-Flag-ACAA1 (CN24874) and pcDNA3.1-HA-CDK4 (CN26246) vectors were purchased from Synbio Technologies. Various truncated mutants of CDK4 and ACAA1 were constructed from the full-length vector. The ACAA1 cDNA (NM_001607) was subcloned into the pCDH-CMV-MCS-EF1-puro vector to generate the Flag-ACAA1 expression vector. Human ACAA1 shRNAs in the U6-MCS-Ubiquitin-EGFP-IRES-puromycin vector (GV248) were purchased from GeneChem. The sense sequences of shRNAs are listed in Supplementary Table S2.

Total RNA was extracted from cells using TRIzol Reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol. First-strand cDNAs were synthesized from 1 μg of total RNA using HiScript III RT SuperMix for qPCR with genomic DNA wiper (Vazyme, R323–01). qRT-PCR was performed using ChamQ SYBR qPCR Master Mix (Vazyme, R311–02) on the QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific). The primer sequences are listed in Supplementary Table S3. The results were analyzed with QuantStudio Design & Analysis software (v1.5.1) and the 2^−ΔΔC_t method.

Proteins were separated by SDS-PAGE and transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore) that were incubated with the indicated primary and secondary antibodies. Corresponding antibody-specific signals were detected with enhanced chemiluminescence substrate (Thermo Fisher Scientific). The signals were captured by an Amersham Imager 600 with AJ600 control software and quantified with ImageJ software.

Cells were homogenized in lysis buffer (50 mmol/L Tris; pH 7.4, 150 mmol/L NaCl, 1% NP-40, 0.25% sodium pyrophosphate, 0.02% NaN₃, 5 mmol/L EDTA, 50 mmol/L NaF, 1 mmol/L Na₃VO₄) containing protease and phosphatase inhibitor cocktails (Roche). For immunoprecipitation (IP) assays, cell extracts were incubated with a rabbit anti-ACAA1 antibody or normal rabbit IgG (Cell Signaling Technology, catalog no. 2729) overnight at 4°C on a rotating platform, followed by incubation with Dynabeads Protein A (Thermo Fisher Scientific) for more than 4 hours at 4°C. For coimmunoprecipitation assays, cell extracts were immunoprecipitated with anti-Flag M2 magnetic beads (Sigma) or anti-HA magnetic beads (Thermo Fisher Scientific) at 4°C overnight. The immunoprecipitates were thoroughly washed with lysis buffer five times, eluted with SDS loading buffer, boiled at 99°C for 5 minutes, and then immunoblotted with the indicated antibodies.

Cells were rinsed in PBS three times, fixed with 4% paraformaldehyde (PFA) at room temperature for 30 minutes, and permeabilized using 0.1% Triton X-100 in 1% BSA in PBS on ice for 10 minutes. After being blocked with 5% BSA at room temperature for 1 hour, the cells were incubated overnight at 4°C with an anti-ACAA1 mouse antibody and an anti-CDK4 rabbit antibody. After incubation with the fluorescein-conjugated secondary antibody, the cells were mounted with ProLong Gold Antifade Mountant with 4',6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific).

Animals were used in accordance with the Institutional Animal Care and Use Committee of SLAC Guide for Care and Use of Laboratory Animals (protocol number: 20210810012 and 20220808016). For the in vivo mammary fat pad xenograft model, 8 × 10⁵ LM2–4175 cells stably expressing negative control (shNC) or shACAA1 vectors were injected into the mammary fat pads of six-week-old female NOD/SCID mice (Shanghai SLAC Laboratory Animal Co., Ltd.). Tumors were monitored until they reached an average size of 50 mm³, at which point, mice were randomly grouped (n = 8 per group) and treatments were initiated. Abemaciclib and trimetazidine were freshly dissolved in PBS at concentrations of 5 mg/mL and 50 mg/mL, respectively. Mice were administered abemaciclib (25 mg/kg; oral gavage) and trimetazidine (250 mg/kg, i.p. injection) once daily, alone or in combination. PBS was administered by oral gavage, i.p. injection, or both routes to obtain consistency among the groups. Tumor diameters were measured every other day with electric calipers, and tumor volumes were calculated as follows: tumor volume = 0.5 × (length) × (width)². Animal weights were recorded every other day after drug administration. When the mean tumor volumes in the control group reached approximately 1000 mm³, the experiment was ceased, and tumors were harvested and imaged.

Endogenous peroxidase activity was blocked with 3% hydrogen peroxide after deparaffinization and rehydration of the slides. Antigen retrieval was performed by incubating the slides with citrate buffer (pH 6.0) at 100°C for 15 minutes. The slides were blocked with 10% normal goat serum for 1 hour at room temperature and incubated overnight at 4°C with a primary antibody. The slides were incubated for 30 minutes with horseradish peroxidase (HRP)-conjugated secondary antibody (Gene Tech) at room temperature. Sections were then incubated with 3,3′-diaminobenzidine substrate (Gene Tech) for visualization and counterstained with hematoxylin for 2 minutes. The slides were then dehydrated in an ascending ethanol series before clearing with xylene and mounting under a coverslip. Adjacent tumor sections were selected and stained with hematoxylin and eosin. The positive staining density was measured using a computerized imaging system composed of an Olympus DP27 charge-coupled device camera connected to a phase-contrast Olympus BX43 light microscope (×40/0.65 magnification). Staining scores were determined using the H-score system that takes both intensity and percentage positivity into account using the formula: 3 × (% of 3+ cells) + 2 × (% of 2+ cells) + 1 × (% of 1+ cells), with a final score ranging from 0 to 300.

All procedures involving patients were conducted in accordance with the Declaration of Helsinki and were approved by the Institutional Ethics Review Board of FUSCC. Written informed consent was obtained from all of the patients.

Quantification and statistical analyses were performed with GraphPad Prism 8.0 (GraphPad Software, Inc.) and SPSS version 22.0 (SPSS). A two-tailed t test, ANOVA, or Mann–Whitney Wilcoxon test was employed for in vitro and in vivo studies (indicated in the figure legends). The correlation coefficients between ACAA1 protein and mRNA expression, as well as between ACAA1 protein levels and IC₅₀ values of trimetazidine, were calculated using Spearman rank correlation analysis. The survival curves were constructed using the Kaplan–Meier method and compared with the log-rank test using the autoselected best cutoff value to define high and low ACAA1 expression. A P < 0.05 was considered statistically significant.

The previously published RNA sequencing (RNA-seq) and proteomic data of the FUSCC TNBC cohort analyzed in this study were obtained from the data repository The National Omics Data Encyclopedia (NODE) at http://www.biosino.org/node/project/detail/OEP000155 and OEP002770. The ACAA1 mRNA expression data and survival information of the TCGA TNBC cohort were downloaded from cBioPortal for Cancer Genomics (http://www.cbioportal.org/). The data of the Kaplan–Meier plotter online tool analyzed in this study were obtained at http://kmplot.com/analysis/. The RNA-seq data generated in this study are publicly available in the NCBI sequence read archive (SRA) at https://www.ncbi.nlm.nih.gov/sra/PRJNA908232. The survival data of the FUSCC TNBC cohort are not publicly available due to patient privacy requirements but are available upon reasonable request from the corresponding author.

Methods used for Supplementary figures are provided in the Supplementary Methods.

A total of 7,531 proteins were obtained from the proteomic profiles of 82 TNBC tumors and 66 paired adjacent normal tissues after quantification assessment and quality control using a two-component Gaussian mixed model as described in our previous study (18). Data from the FUSCC TNBC cohort were screened to uncover potential therapeutic targets for TNBC, particularly for the LAR subtype (Fig. 1A). First, proteins present in less than 30% of samples were eliminated, leaving 1,777 proteins. Then, 26 proteins with higher expression levels in the LAR subtype than in the other subtypes (log₂-fold change ≥ 1, FDR ≤ 0.05) were chosen. Finally, after comparing the protein abundance of the 26 candidates between tumors and normal tissues, the top three candidates were nicotinamide phosphoribosyltransferase (NAMPT), sterol carrier protein-2 (SCP2), and ACAA1. Previous studies have reported that NAMPT and SCP2 could promote tumor progression (24, 25). Thus, we chose ACAA1 for further study, as it was an FDA-approved drug target with no TNBC-related basic research previously reported (26).

The protein abundance and mRNA expression levels of ACAA1 were higher in the LAR subtype (n = 19 for protein; n = 81 for mRNA) than in other subtypes and were higher in TNBC tumors than in adjacent normal tissues (Fig. 1B–E, Supplementary Fig. S1A and S1B). The ACAA1 protein abundance and mRNA expression levels were positively correlated according to Spearman rank correlation analysis (R = 0.386, P < 0.001; Fig. 1F). Kaplan–Meier survival analysis using public databases indicated that higher ACAA1 mRNA expression levels were associated with worse relapse-free survival, distant metastasis-free survival and overall survival (OS) rates in patients with TNBC (Fig. 1G–I). Similarly, higher ACAA1 protein abundance was associated with reduced OS in the publicly available breast cancer cohort (Supplementary Fig. S1C), and higher ACAA1 mRNA expression in the TCGA TNBC cohort was associated with shorter relapse-free survival and OS (Supplementary Fig. S1D and S1E). To confirm the ACAA1 protein expression pattern, we collected 10 pairs of primary TNBC tumors and adjacent normal tissues. Quantitative analysis showed that ACAA1 protein levels were much higher in TNBC tumors than in the corresponding normal tissues (Fig. 1J; Supplementary Fig. S1F). We examined ACAA1 protein levels in two human mammary epithelial cell lines and 12 TNBC cell lines using immunoblotting. We observed that ACAA1 was highly expressed in several TNBC cell lines compared with that in HMECs and MCF10A cells (Fig. 1K; Supplementary Fig. S1G). In our following experiments, different cell lines were chosen on the basis of the purpose of each experiment and the expression pattern of ACAA1 or other certain proteins. In summary, our findings demonstrate that ACAA1 is highly expressed in TNBC and is associated with poor prognosis.

We performed in vitro cell proliferation and colony formation assays in ACAA1-overexpressing and ACAA1-downregulated TNBC cells to explore the biological functions of ACAA1 in TNBC. ACAA1 was found to be expressed at relatively higher levels in LM2–4175, MFM-223 and MDA-MB-453 cell lines (Fig. 1K; Supplementary Figs. S1G and S2A), leading us to choose these cell lines for downregulation experiments, while ACAA1 low-expressing MDA-MB-231 and BT-549 cell lines were used for overexpression experiments. Overexpression of ACAA1 increased the growth rate and clonogenic ability of MDA-MB-231 and BT-549 cells (Fig. 2A–C; Supplementary Fig. S2B), whereas downregulation of ACAA1 with shRNAs significantly impaired the proliferative potential and clonogenic ability of TNBC cells (Fig. 2D–F; Supplementary Fig. S2C). Flow cytometry revealed a decreased proportion of S-phase cells and an increased proportion of G₁ phase cells when ACAA1 was downregulated (Fig. 2G). RB1 plays a key role in regulating G₁–S cell-cycle progression (27). Immunoblotting was used to analyze the expression pattern and phosphorylation status of RB1. We observed higher RB1 phosphorylation (Ser780) in cells stably expressing ACAA1 and decreased levels of total and phosphorylated RB1 when ACAA1 was downregulated (Fig. 2H and I; Supplementary Fig. S2D and S2E).

Given the similarity between ACAA1 and ACAA2, we next investigated the effects of trimetazidine, an approved ACAA2 inhibitor, on TNBC cell lines (28). We detected the IC₅₀ values of trimetazidine in 12 TNBC cell lines and observed that several TNBC cell lines were vulnerable to trimetazidine (Supplementary Fig. S3A). However, no significant correlation was observed between the ACAA1 protein level and sensitivity to trimetazidine (Supplementary Fig. S3B).

To explore the role of trimetazidine on ACAA1, we performed qRT-PCR and immunoblotting analyses in ACAA1 high-expressing MFM-223 and HCC1143 cells treated with trimetazidine. Interestingly, we observed a dose-dependent decrease in the protein level but not in the mRNA expression of ACAA1 in trimetazidine-treated cells (Fig. 2J and K; Supplementary Fig. S4A). ACAA1 high-expressing MFM-223 and HCC70 cells were treated with the proteasome inhibitor MG-132 to determine the degradation pathway of ACAA1. ACAA1 did not accumulate, but a significant increase in p21, a known substrate of the ubiquitin–proteasome system (29), was detected (Supplementary Fig. S4B). Meanwhile, the lysosomal inhibitors Baf-A1 and NH₄Cl resulted in the accumulation of ACAA1 in MFM-223 cells (Supplementary Fig. S4C and S4D), suggesting that ACAA1 was degraded in the lysosomes. Three types of autophagy are involved in lysosomal degradation: macroautophagy, microautophagy, and chaperone-mediated autophagy (CMA; ref. 30). CMA may be induced by serum starvation (31). Further studies showed that serum starvation and siRNA-mediated knockdown of two key molecules involved in CMA, HSP family A (HSP70) member 8 (HSPA8) and lysosome-associated membrane protein type 2a (LAMP2A), had no significant effects on ACAA1 protein levels, indicating that CMA was not responsible for ACAA1 protein degradation (Supplementary Fig. S4E–G). Furthermore, treatment with 1 mmol/L 3-MA, a selective inhibitor of macroautophagy (32), resulted in a significant increase in ACAA1 protein levels in MFM-223 and HCC70 cells, indicating that macroautophagy was responsible for ACAA1 protein degradation (Supplementary Fig. S4H).

We further explored whether trimetazidine could inhibit tumor growth through regulation of ACAA2. First, we explored the role of ACAA2 in TNBC and evaluated the effect of trimetazidine on ACAA2. The siRNA-mediated knockdown of ACAA2 had no significant effect on the growth of Hs-578T cells with high ACAA2 expression (Supplementary Fig. S5A–S5C). In addition, no decreases in ACAA2 mRNA expression or protein levels in trimetazidine-treated Hs-578T, MFM-223, or HCC1143 cells were observed (Supplementary Fig. S5D and E). In summary, ACAA2 had no effect on TNBC cell growth, and trimetazidine had no effect on the protein level or mRNA expression of ACAA2.

Overall, our findings suggest that ACAA1 promotes TNBC progression through the G₁–S cell-cycle transition and is a potential therapeutic target for TNBC, because treatment with trimetazidine causes ACAA1 protein degradation and inhibits TNBC cell proliferation.

We performed stable isotope labeling by amino acids in cell culture (SILAC)-based quantitative proteomics to identify the ACAA1-associated proteins and uncover the underlying molecular mechanisms of ACAA1 in TNBC (Fig. 3A). After excluding three HSPs that are well-known regulators of the stress response (33), four potential ACAA1-associated proteins (PEX7, FKBP10, CDK4, and ARHGAP21) were identified on the basis of the unique peptide numbers, log₂ (ratio H/L) and “significance B” values (Fig. 3B). We focused on CDK4 as the key ACAA1-associated protein because its role in cell-cycle progression is similar to that of ACAA1 in TNBC (34). We verified the interaction between ACAA1 and CDK4 through IP assays in HEK293T, BT-549, HCC70, and HCC38 cells but observed no interaction between ACAA1 and D-cyclins (cyclin D1, D2, and D3; Fig. 3C and D; Supplementary Fig. S6A). Immunofluorescence (IF) staining showed that CDK4 localized primarily in the nucleus, while ACAA1 localized both in the cytoplasm and in the nucleus (Fig. 3E).

We constructed five truncated mutations of CDK4 and four truncated mutations of ACAA1 to identify the region of each protein responsible for this interaction (Fig. 3F and G). Next, we performed coimmunoprecipitation assays in HEK293T cells and showed that deleting the domain (residues 161–245) following the Asp₁₅₈-Phe₁₅₉-Gly₁₆₀ (Deutsche Forschungsgemeinschaft; DFG) motif in CDK4 abolished the binding between CDK4 and ACAA1 (Fig. 3H). This domain contains both the T-loop region (residues 161–171) and CDK4 phosphorylation site (Thr172), which is critical to the function of CDK4 (35). In addition, we identified the N-terminal thiolase domain (residues 38–161) of ACAA1 to be critical in its interaction with CDK4 (Fig. 3I). However, further immunoblotting analyses showed that overexpression or downregulation of ACAA1 did not alter CDK4 protein expression, phosphorylation (Thr172) or nucleoplasmic distribution (Supplementary Fig. S6B–D). Likewise, overexpression of CDK4 did not affect ACAA1 protein expression (Supplementary Fig. S6E). Altogether, these data suggest that ACAA1 binds to CDK4 but has no effect on its protein expression, phosphorylation status, or nucleoplasmic distribution.

We detected the IC₅₀ values of abemaciclib and analyzed the protein expression pattern by immunoblotting using a panel of 13 TNBC cell lines with different ACAA1 protein levels and RB1 statuses [wild-type (WT), deleted, or mutant] to determine the sensitivity to abemaciclib (Fig. 4A and B). We observed heterogeneous sensitivities to abemaciclib, with IC₅₀ values ranging from 0.157 μmol/L (Hs-578T) to 11.658 μmol/L (HCC1937; Fig. 4A). Consistent with a previous study, we observed that cells with deleted or mutant RB1 (the group shown in black) were relatively resistant to abemaciclib (36). Among the WT cell lines, cells with lower ACAA1 protein levels (ACAA1^low, the blue group) had lower IC₅₀ values than those with higher ACAA1 protein levels (ACAA1^high, the red group). These data suggested a possible correlation between ACAA1 expression and sensitivity to the CDK4/6 inhibitor abemaciclib.

We performed growth inhibition assays using LM2–4175 and HCC1143 cells and found that downregulation of ACAA1 sensitized RB1-WT TNBC cells to abemaciclib (Fig. 4C; Supplementary Fig. S7A). Consistently, a substantially lower dose of abemaciclib suppressed RB1 phosphorylation and E2F-targeted protein (PLK1 and cyclin A2) levels in ACAA1 knockdown LM2–4175 and HCC1143 cells (Fig. 4D; Supplementary Fig. S7B and C). However, the potentiating effect was not observed in RB1-mutant HCC70 cell lines (Supplementary Fig. S7D and S7E). We used the mammary fat pad injection model in female NOD/SCID mice to investigate the effects of ACAA1 on TNBC proliferation and the response to abemaciclib in vivo. In accordance with the in vitro results, downregulation of ACAA1 inhibited xenograft tumor growth, and the tumors were more sensitive to abemaciclib (Fig. 4E–G). We observed a 24.08% decrease in tumor weight after treatment with abemaciclib and a 51.75% decrease in tumor weight after abemaciclib treatment combined with ACAA1 knockdown (Fig. 4F). Likewise, we observed a 55.83% decrease in the tumor volume and a 43.21% decrease in the tumor volume following abemaciclib treatment with or without downregulation of ACAA1 (Fig. 4G). We also observed that treatment with abemaciclib delayed tumor growth for 6 days in LM2–4175 xenograft models and extended to 15 days when abemaciclib was combined with ACAA1 knockdown (Supplementary Fig. S8A). Treatment toxicity evaluated by body weight, white blood cell (WBC) counts, hemoglobin levels, and platelet counts was limited (Supplementary Fig. S8B–S8E). We subsequently performed 3D cell viability assays using two TNBC patient-derived organoid models treated with or without 0.5 μmol/L abemaciclib. The transcriptomic subtype of those two patients was defined as the basal-like and immune-suppressed (BLIS) subtype for PDO811 and the LAR subtype for PDO845 on the basis of our previously defined IHC method (37). We observed that the organoid (PDO811) derived from the patient with weaker ACAA1 and RB1 phosphorylation (Ser780) IHC staining was more sensitive to abemaciclib (Fig. 4H and I). Overall, our data reveal that ACAA1 may be used to stratify RB1-proficient TNBC for CDK4/6 inhibitor treatment and that inhibition of ACAA1 potentiates the tumor response to the CDK4/6 inhibitor abemaciclib.

As downregulation of ACAA1 potentiates the response to the CDK4/6 inhibitor abemaciclib and ACAA1 protein levels are decreased by trimetazidine, we performed drug combination assays and calculated the CI values using the Chou–Talalay method (38). As expected, the combination of abemaciclib with trimetazidine exhibited significant synergy in three RB1 WT TNBC cell lines: MFM-223 (CI = 0.48 ± 0.21), LM2–4175 (CI = 0.52 ± 0.21), and HCC1143 (CI = 0.29 ± 0.21; Fig. 5A; Supplementary Fig. S9A and S9B). The growth rate of MFM-223, LM2–4175, and HCC1143 cells was slightly or moderately inhibited using a single agent (apart from 1 mmol/L trimetazidine in MFM-223 cells) but was further inhibited by the combination of abemaciclib and trimetazidine (Fig. 5B; Supplementary Fig. S9C–D). Notably, a synergistic effect was observed even on RB1-deficient HCC1937 cells (Supplementary Fig. S9E). Moreover, immunoblot analysis showed that RB1 and PLK1 protein levels and RB1 phosphorylation (Ser780) levels decreased significantly following treatment with trimetazidine and abemaciclib, indicating that combined treatment further enhanced cell-cycle arrest compared with single-agent treatment (Fig. 5C; Supplementary Fig. S9F). Next, we evaluated the efficacy and toxicity of trimetazidine and abemaciclib alone or in combination in vivo. The combination of abemaciclib and trimetazidine induced greater tumor regression in LM2–4175 xenograft models than abemaciclib or trimetazidine alone (Fig. 5D–F). We observed 31.46%, 35.93%, and 62.13% decreases in tumor weight following treatment with abemaciclib, trimetazidine, and their combination, respectively (Fig. 5E). Likewise, we observed 43.25%, 45.75%, and 74.26% decreases in tumor volumes after treatment with abemaciclib, trimetazidine and their combination, respectively (Fig. 5F). Importantly, no significant body weight loss was observed in the combination group compared with the single-agent groups (Fig. 5G). We also observed that treatment with abemaciclib or trimetazidine alone delayed tumor growth for 6 days, that extended to 15 days when these drugs were administered in combination (Supplementary Fig. S10A). Toxicity evaluated by body weight, WBC counts, hemoglobin levels, platelet counts, and hepatorenal indicators was limited (Supplementary Fig. S10B–S10H). A significant decrease in phosphorylated RB1 (Ser780) was observed in the IHC staining of xenograft tumors from mice that underwent combined treatment (Fig. 5H; Supplementary Fig. S11A). Likewise, significantly weaker IHC staining for ACAA1 was observed after in vivo trimetazidine treatment (Fig. 5I; Supplementary Fig. S11B). We performed RNA-seq and gene set enrichment analysis (GSEA) on xenograft tumors to investigate the influence of trimetazidine in vivo. Consistent with the in vitro results, no significant decrease in ACAA1 mRNA expression was observed (Supplementary Fig. S12A and S12B). GSEA revealed downregulation of cell-cycle–related gene sets (such as the mitotic G₁ phase, G₁–S transition, and E2F targets gene sets) after in vivo trimetazidine treatment (Supplementary Fig. S12C and S12D). Finally, we verified the synergy of 1 μmol/L abemaciclib and 1 mmol/L trimetazidine in an organoid model derived from one patient with TNBC (Fig. 5J). Altogether, these findings reveal that trimetazidine downregulates ACAA1 protein levels and the G₁–S cell-cycle transition in vivo, supporting trimetazidine combined with abemaciclib as a potential strategy for TNBC with high ACAA1 expression levels.

In this study, we screened ACAA1 on the basis of our proteomic and transcriptomic data and identified that ACAA1 inhibition restrained tumor proliferation and facilitated the response to the CDK4/6 inhibitor abemaciclib in TNBC. Further study indicated that trimetazidine downregulates ACAA1 protein levels and gene sets of G₁–S transition and downstream E2F targets. Mechanistic investigations demonstrated the role of ACAA1 in regulating the G₁–S cell-cycle transition and the interaction between ACAA1 and CDK4 in the nucleus (Fig. 6). Importantly, we revealed that trimetazidine potentiated the antitumor efficacy of abemaciclib in TNBC cell lines, patient-derived organoids, and xenograft models.

The LAR subtype remains a challenge in TNBC with a relatively poor prognosis, as evidenced by the low pathologic complete response rates after neoadjuvant chemotherapy and low objective response rates in metastatic TNBC after targeted therapy, including an AR inhibitor (13, 39). The LAR subtype and proteomic iP-2 subtype overlap significantly, with 60% of the iP-2 subtype clustered in the FUSCC subtype LAR (18). We demonstrated that ACAA1 was expressed relatively higher in both the LAR subtype and proteomic iP-2 subtype and confirmed its association with poor survival in TNBCs. ACAA1 encodes human peroxisomal 3-oxoacyl-CoA thiolase, an enzyme responsible for the thiolytic cleavage of straight-chain 3-oxoacyl-CoAs and the generation of acetyl-CoA and acyl-CoA in the peroxisomal beta-oxidation system (40). ACAA1 is a key member of the PPAR and fatty acid metabolism pathways, and its expression is upregulated or downregulated in various tumor tissues compared with the corresponding normal tissues (41–44). A previous study showed that five ACAA1 SNPs are associated with breast cancer susceptibility (45).

Three pharmacologic inhibitors of CDK4/6 (palbociclib, ribociclib, and abemaciclib) have been widely implemented in clinical practice as treatments for hormone receptor–positive and HER2-negative advanced or metastatic breast cancer (46). As a well-known tumor suppressor, RB1 plays an important role in restricting cell-cycle progression by regulating E2F transcription factors, and RB1 aberrations have been reported as mechanisms of preexisting and acquired resistance to CDK4/6 inhibitors (47–50). Comprehensive profiles of breast cancer revealed that a high frequency of RB1 loss and mutations were detected in a considerable proportion of TNBC tissues (12, 14). In this study, we observed total and phosphorylated blockages of RB1 following ACAA1 downregulation and discovered an interaction between ACAA1 and CDK4 in TNBC. CDK4 is a key regulator of the G₁–S cell-cycle transition and is crucial in the growth and development of many cancer types, including breast cancer (51). Many studies have documented the synergistic effects of combining CDK4/6 inhibitors with other targeted therapies (such as inhibitors of PARP, AR, BET, EGFR, and PI3Kα) in certain TNBCs (52–57). However, the translation of these preclinical studies into clinical benefits is an ongoing challenge.

Trimetazidine, an anti-ischemic agent, reduces long-chain fatty acid oxidation by inhibiting the 3-ketoacyl-coenzyme A thiolase enzyme (58). Previous studies reported that the combination of trimetazidine with other drugs, such as the PPARγ agonist troglitazone, the glutaminase C inhibitor C.968, and chemotherapy, induces apoptosis in cancer cells (59, 60). In this study, we utilized trimetazidine’s ability to reduce ACAA1 protein levels to target ACAA1. We observed a similar role of trimetazidine to CDK4/6 inhibitors and discovered its effect on potentiating the efficacy of abemaciclib in RB1-proficient TNBC models. We demonstrate that ACAA1 regulates TNBC proliferation via the CDK4/6-RB1-E2F signaling pathway and that inhibiting ACAA1 with trimetazidine enhances the response to the CDK4/6 inhibitor abemaciclib. However, we speculate that other mechanisms beyond the CDK4-RB1-E2F pathway may also be involved, and further studies are still required to investigate the role of trimetazidine in treating breast cancer.

Several limitations of this study need to be noted. First, ACAA1 was detected in the nucleus of TNBC cell lines. It lacks the canonical nuclear localization signal sequences, and the mechanisms underlying its nuclear localization and potential effects remain unclear, warranting further investigation. Second, the dosage of trimetazidine used in our study was higher than the clinical dose (20 mg, orally, three times a day) used to treat heart-related conditions. Finally, our study lacks paired biopsies and detection of ACAA1 in patients with TNBC before and after treatment with CDK4/6 inhibitors, as CDK4/6 inhibitors are not currently used to treat patients with TNBC. Hence, a well-designed clinical trial enrolling patients with TNBC with heterogeneous ACAA1 expression status is required to verify our conclusions in the clinic.

In conclusion, our data suggest that the combination of trimetazidine with abemaciclib may be a promising treatment strategy for ACAA1-high TNBCs. Furthermore, our study highlights the importance of investigating ACAA1 protein levels in further clinical trials involving CDK4/6 inhibitors in TNBC.

No disclosures were reported.

W.T. Peng: Conceptualization, resources, data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. X. Jin: Conceptualization, resources, data curation, software, formal analysis, funding acquisition, visualization, methodology, writing–original draft, writing–review and editing. X.E. Xu: Data curation, methodology, project administration, writing–review and editing. Y.S. Yang: Software, investigation, methodology, writing–review and editing. D. Ma: Data curation, formal analysis, methodology, writing–review and editing. Z.M. Shao: Conceptualization, resources, supervision, funding acquisition, validation, investigation, project administration, writing–review and editing. Y.Z. Jiang: Conceptualization, data curation, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing.

This work was supported by grants from the National Key Research and Development Project of China (2020YFA0112304 to Z.-M. Shao), the National Natural Science Foundation of China (91959207 and 92159301 to Z.-M. Shao), the Program of Shanghai Academic/Technology Research Leader (20XD1421100 to Y.-Z. Jiang), the Natural Science Foundation of Shanghai (22ZR1479200 to Y.-Z. Jiang), the Shanghai Key Laboratory of Breast Cancer (12DZ2260100 to Z.-M. Shao), Youth Talent Program of Shanghai Health Commission (2022YQ012 to X. Jin), the SHDC Municipal Project for Developing Emerging and Frontier Technology in Shanghai Hospitals (SHDC12021103 to Z.-M. Shao), and the Clinical Research Plan of SHDC (SHDC2020CR4002 to Y.-Z. Jiang; SHDC2020CR5005 to Z.-M. Shao).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.