Materials

Detailed information regarding the materials is provided in the supplementary material.

Cell culture

MDA-MB-231, MDA-MB-468, MCF-7, T-47D, SK-BR-3 (Chinese Academy of Sciences, Kunming, Yunnan, China), MCF-10A, MCF-12A, and WI-38 cell lines (American Type Culture Collection, Manassas, VA, USA), and immortalized primary CAFs from TNBC tissues (BeNa Culture Collection, Beijing, China) were cultured as described in the Supplementary Material. The patient provided consent for use of the tissue and had not received any prior therapy. All cell lines were tested for mycoplasma contamination and authenticated using short tandem repeat profiling. CAFs were identified using the CAF-related biomarkers alpha-smooth muscle actin (α-SMA) and fibronectin, confirming a purity of > 99% (Fig. S1) and utilized at 3–15 passages. CAFs were cultured to approximately 70% confluence and then subjected to serum-free medium treatment for 48 h. The conditioned medium (CM) was then collected and filtered for subsequent investigation. WI-38 cells were transformed into CAFs by culturing in medium from various subtypes of breast cancer cells for 48 h, as described previously [23, 24]. Subsequently, CAF-CM from activated WI-38 cells was then used to stimulate breast cancer cells for an additional 48 h.

Wound healing assay

TNBC cells were stained with CellTracker™ Green dye for 30 min at 37 ℃, digested, and co-cultured with CAFs at a ratio of 2:1 in 6-well plates. Once the cells reached approximately 90% confluence, the cell monolayers were scraped using a micropipette tip and treated with or without shikonin in serum-free medium. Alternatively, TNBC cells were seeded in 6-well plates, scraped, and treated with or without CAF-CM. The scratches in the cell monolayers were photographed under a microscope (DMI8; Leica, Wetzlar, Germany) at 0 and 48 h after treatment.

Cell invasion assay

Transwell chambers, pre-coated with Matrigel (8-μm pore size) and placed in 24-well plates, were used to assess cell invasion. TNBC cells, pre-treated with or without shikonin for 6 h, were seeded in the upper compartment with serum-free media, and CAFs in serum-free medium were included in the lower compartment. Alternatively, TNBC cells were treated with CAF-CM with or without shikonin for 48 h, then suspended with serum-free media in the upper compartment, while media containing 10% FBS was added into the lower compartment. The Transwell systems were placed in a humidified incubator for 24 h (5% CO2, 37 ℃). The infiltrated cells were fixed with methanol, stained with hematoxylin/eosin, and photographed using an inverted microscope (DMI1; Leica) followed by quantification with ImageJ software (v.1.8.0).

Detection of filamentous actin (F-actin)

TNBC cells and CAFs (total 9 × 103 cells per well, at a ratio of 2:1) were seeded into 96-well black plates. The cells were fixed (4%, paraformaldehyde, 25 ℃, 10 min), permeabilized (0.2% Triton X-100, 4 ℃, 10 min), blocked (5%, goat serum, 25 ℃, 30 min), and then incubated with Alexa Fluor 488™ phalloidin (1:50, 37 ℃, 30 min) for F-actin staining. The cells were then incubated with anti-EpCAM antibody (1:500, 4 ℃, overnight) followed by Alexa Fluor 647 secondary antibody (1:250, 25 ℃, 1.5 h), and stained with 4’,6-diamidino-2-phenylindole (DAPI, 1 μg/mL, 25 ℃, 30 min) for nuclear staining. Fluorescence signals were detected using a high-content imaging system.

3D-culture assay

The 24-well plates were pre-coated with Matrigel (growth factor reduced, GFR, 300 μL/well). TNBC cells were collected and resuspended in media containing 2% GFR Matrigel, and then located in the Matrigel-coated plates. The cells were then treated with CAF-CM with or without shikonin, and the treatment medium was replaced every day for 3 days. The cells were photographed under a DMI1 microscope (Leica).

Detection of anoikis

TNBC cells were inoculated in poly-2-hydroxyethyl methacrylate coated (anchorage-resistant) plates for 24 h, incubated with propidium iodide (1:100) and Annexin V-fluorescein isothiocyanate (1:100) reagents at 25℃ for 15 min, and then analyzed using flow cytometry (NovoCyte; Ex/Em = 488/630; 488/530 nm) with NovoExpress software (ACEA Biosciences). Alternatively, TNBC cells were incubated with calcein AM (1:5) and ethidium homodimer (EthD-1, 1:5) from a CytoSelect™ 96-Well Anoikis Assay kit for 30 min at 37 ℃. The fluorescence intensities of Calcein AM and EthD-1 were measured using a high-content analysis system at Ex/Em = 485/515 and 525/590 nm, respectively, with Harmony Software (v.4.9; PerkinElmer, Berlin, Germany).

Detection of mitochondrial numbers and distribution

Mitochondrial numbers were assessed by staining the cells with MitoTracker Green FM (100 nM, 37 ℃, 30 min), followed by flow cytometry (NovoCyte; Ex/Em = 488/530 nm). Mitochondrial distribution was determined by staining the cells with Alexa Fluor 488™ phalloidin (1:50, 37 ℃, 30 min), MitoTracker Red (50 nM, 37 ℃, 30 min), and Hoechst 33342 (1:100, 25 ℃, 30 min) after fixation with 4% paraformaldehyde (25 ℃, 10 min). Images were captured using the high-content analysis system.

Quantification of mitochondrial DNA (mtDNA) levels

This experiment was performed using a genomic DNA extraction kit, TB Green Premix, and Human Mitochondrial DNA Primer kit, as described previously [25].

Transmission electron microscopy (TEM)

Cells were fixed in 2.5% glutaraldehyde (4 ℃, overnight), incubated in 1% osmium tetroxide (25 ℃, 2 h), dehydrated in a graded ethanol series, and then embedded with Epon 812. Ultrathin sections (70 nm) were obtained using an EM UC7 ultramicrotome (Leica) and counterstained with 2% uranyl acetate and lead citrate for 15 min. The sections were viewed using a compact digital transmission electron microscope at 80 kV (HT7800, Hitachi, Tokyo, Japan). Mitochondrial morphology was classified into three types based on the ratio of length to width using ImageJ software: ≤1.5: round; 1.5–3.0: intermediate; >3.0: elongated.

Blue-native polyacrylamide gel electrophoresis (PAGE) analysis

Blue-native PAGE analysis was conducted as described previously [26]. Cells were collected and suspended in phosphate-buffered saline supplemented with digitalis glycoside (8 mg/mL), phenylmethanesulfonyl fluoride (PMSF, 1 mM), and protease inhibitor cocktail (PIC, 1%) at 4 ℃ for 15 min, followed by centrifugation (21,130 × g, 4 ℃, 20 min). The resulting pellets were resuspended in native PAGE sample buffer (containing 5% digitalis glycoside, 1 mM PMSF, and 1% PIC), vortexed (every 5 min for 30 min), and centrifuged (21,130 × g, 4 ℃, 30 min). The protein complexes were quantified using a BCA protein assay kit, separated using a native PAGE 4–16% gel, 10-well kit, and transferred to polyvinylidene fluoride membranes. After blocking with 5% nonfat-dried milk (4 ℃, 1 h), they were incubated with OXPHOS Rodent WB antibody cocktail (1:1000, 4 ℃, overnight) and secondary antibody. The results were imaged using an enhanced chemiluminescence kit in a ChemiDoc system (v.5.2 Image Lab Software; XRS+; Bio-Rad).

Determination of mitochondrial reactive oxygen species (ROS) level

Mitochondrial ROS levels were assessed by staining cells with MitoSOX (4 μM) in serum-free medium (37 ℃, 30 min). The fluorescence intensities of MitoSOX (Ex/Em = 488/572 nm) were measured using flow cytometry with NovoExpress software (ACEA Biosciences).

Detection of ATP levels

Cells were lysed in lysis buffer supplemented with 1 mM PMSF and 1% PIC. ATP levels were measured using an ATP determination kit and quantified with a microplate spectrophotometer (Varioskan, Thermo) at 560 nm. The ATP content was normalized to the cell protein concentration.

Western blotting and immunoprecipitation assay

Total, nuclear, and mitochondrial proteins were extracted using radioimmunoprecipitation assay (RIPA) lysis buffer, a nuclear/cytoplasmic protein kit, and cell mitochondria isolation kit, respectively. The lysis buffer was supplemented with 1 mM PMSF and 1% PIC. Immunoprecipitation was performed using an immunoprecipitation kit. Western blotting was performed as described previously [25]. Immunoblots were detected using an enhanced chemiluminescence kit and a ChemiDoc system.

Immunofluorescence staining

TNBC cells (6 × 103 cells/well) were seeded into 96-well black plates, fixed (paraformaldehyde, 4%, 25 ℃, 10 min), permeabilized (Triton X-100, 0.2%, 4 ℃, 10 min), and blocked with 5% goat serum (25 ℃, 30 min). The cells were then incubated with anti-PGC-1α antibody (1:300, 4 ℃, overnight), followed by Alexa Fluor 488 secondary antibody (1:500, 25 ℃, 1.5 h), anti-ERRα antibody (1:200, 4 ℃, overnight), Alexa Fluor 647 secondary antibody (1:250, 25 ℃, 1.5 h), and DAPI (1 μg/mL, 25 ℃, 30 min). The results were detected using a high-content system.

Dual-luciferase reporter assay

Cells were inoculated into 12-well plates and transfected with pRL Renilla luciferase control reporter vectors (pRL-TK) and ERRE-luciferase reporter plasmids (pGL3-3×ERRE) at a ratio of 1:10 using Lipofectamine 2000 reagent. After treatment, ERRE-luciferase activity was detected using a dual-luciferase reporter system and normalized to Renilla luciferase activity.

Chromatin immunoprecipitation (ChIP) assay

ChIP assay was conducted with 2 × 107 cells per group using a ChIP kit and anti-PGC-1α (5 μg) antibody according to the standard protocol. Quantification of PGC-1α ChIP enrichment was performed using a quantitative polymerase chain reaction (Q-PCR) detection system (Bio-Rad) and specific primers for the ERRSA promoter (Table S1). ChIP enrichment was calculated against the IgG group and subsequently normalized against the control group.

Q-PCR

Q-PCR was performed using an RNA extraction kit, cDNA synthesis kit, and TB Green Premix in a CFX96 RT-PCR detection system (Bio-Rad), as described previously [25]. The primer sequences are listed in Table S1.

Phosphorylation site and binding partners of PGC-1α

Cells were lysed using RIPA lysis buffer and immunoprecipitation-mass spectrometry (IP-MS) assays were performed. PGC-1α was immunoprecipitated from cellular extracts using an anti-PGC-1α antibody, and the immunoprecipitates were separated by sodium dodecyl sulfate-PAGE. The indicated protein was excised, digested with trypsin, fractionated, and then analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) using Easy-nLC 1000 and LTQ Orbitrap ETD (Thermo). Mass spectrometric data were identified using Proteome Discoverer software (Thermo; v1.4) and the NCBI database.

Plasmid transfection

Lentiviruses harboring pGPU6/GFP/Neo-shRNA-LDHA vector (LDHA shRNA: 5′- CTCTGGCAAAGACTATAATGT-3′), pGPU6/GFP/Neo-shRNA-NC vector (5′- TTCTCCGAACGTGTCACGT-3′), LV2 (U6/Puro)-shRNA-PGC-1α vector (PGC-1α shRNA: 5′- GGTGCAGTGACCAATCAGAAA-3′), LV2-shRNA-NC vector (5′- TTCTCCGAACGTGTCACGT-3′), and pcDNA3.1 plasmids containing PGC-1αWT or PGC-1αA295 were purchased form GenePharma (Shanghai, China). All transfection procedures were performed according to the manufacturer’s instructions. TNBC cells were transfected with shRNA-PGC-1α to knockdown PGC-1α and then transfected with pc DNA3.1 (+) (CMV/MCS/AmpR) plasmids to express PGC-1αWT- or PGC-1αA295. Cells with stable knockdown of PGC-1α and expression of PGC-1αWT or PGC-1αA295 were selected by incubation with 0.2 μg/mL puromycin and 100 μg/mL ampicillin for 21 days.

Animals and treatment

Female NOD/SCID mice (6–8 weeks old, 18–22 g, Charles River, Beijing, China) were raised in 12:12 h light/dark-cycle standard conditions and assigned randomly to different groups. MDA-MB-231 cells (MDA-MB-231/PGC-1αWT, MDA-MB-231/PGC-1αA295, 1 × 106 cells/mouse) and CAF cells (0.5 × 106 cells/mouse) were co-injected into the left fourth inguinal mammary fat pads (n = 10/group). After 4 weeks, tumors were removed by mammary gland lumpectomy. The following day, the mice were treated intragastrically with shikonin (10 mg/kg/2 days) or vehicle (olive oil) for 30 days. After the experiment, the mice were sacrificed under anesthesia using CO2 inhalation.

Tissues (5-μm sections) were stained with hematoxylin/eosin. TOM20 was detected in lung tissues by immunofluorescence using a fluorescence staining kit, anti-TOM20 antibody (1:200), and DAPI (1 μg/mL). Immunohistochemistry staining of lung tissues was conducted using a Biotin-Streptavidin HRP Detection kit with anti-PGC-1α (1:200), anti-ERRα (1:200), and anti-COXIV (1:200) antibodies. Tissue images were photographed using a DMI8 microscope and analyzed with ImageJ software. The investigators were blinded to the assignments and outcomes.

Multiplex immunofluorescence staining

We carried out multiplex immunofluorescence staining of breast cancer (n = 230) and TNBC (n = 38) [27] tissue microarrays from patients with primary cancer (Shanghai Outdo Biotech, Shanghai, China). The clinicopathological parameters of these specimens are listed in Supplementary Tables S2 and S3. PGC-1α and ERRα expression in tumor tissues were evaluated using a PANO 7-plex immunohistochemistry kit, anti-PGC-1α, anti-ERRα, and anti-pan-cytokeratin (Pan-CK) antibodies, and DAPI. Fluorescence signals were captured using a Mantra System (PerkinElmer). PGC-1α and ERRα expression levels in breast cancer tissues were calculated by mean cell intensity, and PGC-1α expression in TNBC tissues was calculated by H-score using inForm image software. The pathologists who collated and analyzed the clinical parameters were blinded to the study results. The results were analyzed and represented using R script (v.4.0.2) and ggplot2 package.

Kaplan–Meier plot analysis

The correlations between PPARGC1A, ESRRA, and PPARGC1A/ESSRA mRNA levels and RFS in patients with different breast cancer subtypes were analyzed using the Kaplan–Meier plotter database (http://kmplot.com, ER-positive/HER2-negative patients: n = 2301; HER-2-positive/ER-negative patients: n = 323; TNBC patients: n = 534) [28]. PPARGC1A and ESRRA mRNA levels were analyzed using the autoselect best cutoff mode, while PPARGC1A/ESSRA mRNA levels were analyzed using the mean expression levels of the selected genes. Statistical significance was calculated by log-rank test and indicated as p < 0.05. The correlations of PPARGC1A mRNA levels and mRNA levels of ERRα-targeted mitochondrial genes were evaluated using R- and p-values, calculated by Pearson’s correlation analysis. The results were represented using R script (v.4.0.2) and the ggplot2 package.

Statistical analysis

The results are presented as the mean ± standard deviation. Differences among multiple groups were analyzed by one way-ANOVA and Bonferroni’s post hoc test, while differences between two groups were analyzed by Student’s t-test. Differences in Kaplan–Meier survival curves were evaluated using the log-rank test. A p value < 0.05 was considered statistically significant in all tests conducted in this study.