TGF-β induces cholesterol accumulation to regulate the secretion of tumor-derived extracellular vesicles

Abstract

Background: Cancer cells are avid extracellular vesicle (EV) producers. EVs transport transforming growth factor-β (TGF-β), which is commonly activated under late stages of cancer progression. Nevertheless, whether TGF-β signaling coordinates EV biogenesis is a relevant topic that remains minimally explored.

Method: We sought after specific TGF-β pathway mediators that could regulate EV release. To this end, we used a large number of cancer cell models, coupled to EV cell biological assays, unbiased proteomic and transcriptomic screens, followed by signaling and cancer biology analyses, including drug resistance assays.

Results: We report that TGF-β, by activating its type I receptor and MEK-ERK1/2 signaling, increased the numbers of EVs released by human cancer cells. Upon examining cholesterol as a mediator of EV biogenesis, we delineated a pathway whereby ERK1/2 acted by phosphorylating sterol regulatory element-binding protein-2 that transcriptionally induced 7-dehydrocholesterol reductase expression, thus raising cholesterol abundance at both cellular and EV levels. Notably, inhibition of MEK or cholesterol synthesis, which impaired TGF-β-induced EV secretion, sensitized cancer cells to chemotherapeutic drugs. Furthermore, proteomic profiling of two distinct EV populations revealed that EVs secreted by TGF-β-stimulated cells were either depleted or enriched for different sets of cargo proteins. Among these, latent-TGF-β1 present in the EVs was not affected by TGF-β signaling, while TGF-β pathway-related molecules (e.g., matrix metalloproteinases, including MMP9) were either uniquely enriched on EVs or strongly enhanced after TGF-β stimulation. EV-associated latent-TGF-β1 activated SMAD signaling, even when EV uptake was blocked by heparin, indicating competent signaling capacity from target cell surface receptors. MMP inhibitor or proteinase treatment blocked EV-mediated SMAD signaling, suggesting that EVs require MMP activity to release the active TGF-β from its latent complex, a function also linked to the EV-mediated transfer of pro-migratory potential and ability of cancer cells to survive in the presence of cytotoxic drugs.

Conclusion: Hence, we delineated a novel signaling cascade that leads to high rates of EV generation by cancer cells in response to TGF-β, with cholesterol being a key intermediate step in this mechanism.

Keywords: Cancer; Cholesterol; Extracellular; Matrix metalloproteinase; Transforming growth factor β; Vesicles.

Fig. 1

TGF-β induces EV secretion via activation of the MEK/ERK pathway. (A, B) EVs released by the three indicated cell models were quantified by NTA in terms of particle size (A) and particle number after normalization to the total cell number (B). The cells were stimulated with vehicle (Ctrl), 5 ng/mL TGF-β1 for 48–120 h, 5 µM LY2157299 TGF-β type I receptor inhibitor (TβRi) or combination of TGF-β with TβRi for 120 h. (C) Pearson correlation of nanoparticle numbers per cell with corresponding cellular TGFB1 mRNA expression from the same cell. Each data point represents one of the seven cell lines analyzed. (D) Protein expression levels of the indicated EV-specific proteins in EV extracts (isolated as VSF or after CD81-specific enrichment) derived from MDA-MB-231 cells stimulated with 5 ng/mL TGF-β1 in the absence or presence of 5 µM TβRi for the indicated time periods, and densitometric values were normalized to the vehicle control. (E, F) Representative cryo-TEM (E) and SEM (F) pictures of EVs isolated as VSF (E) or CD81-enriched fraction (F) from MDA-MB-231 cells stimulated with 5 ng/mL TGF-β1 for 48–120 h or not (Ctrl). A negative control image of CD81-specific immunobeads alone is also shown (F). Scale bars are included and red arrows mark individual or clustered EVs. (G) Protein expression levels of SMAD2, SMAD3 and β-ACTIN (as loading control) in MDA-MB-231 protein extracts transiently transfected with the indicated siRNAs, and densitometric values were normalized to the control siRNA. (H) EVs released by MDA-MB-231 cells, which were first transiently transfected with SMAD-specific siRNAs, were quantified by NTA in terms of particle size (left) and particle number after normalization to the total cell number (right). The cells were stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 for 48 h. (I) Protein expression levels of the indicated signaling proteins and β-TUBULIN (as loading control) in MDA-MB-231 cells stimulated with 5 ng/mL TGF-β1 in the absence or presence of 5 µM MEK inhibitor (MEKi; U0126) for the indicated time periods, and densitometric values were normalized to the vehicle control. (J) EVs released by MDA-MB-231 cells treated with vehicle (DMSO) or 5 µM MEKi and quantified by NTA in terms of particle size (left) and particle number after normalization to the total cell number (right). The cells were stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 for 48 h. (K) Expression levels of the indicated proteins and β-ACTIN (as loading control) in A549 cells stimulated with 5 ng/mL TGF-β1 in the absence or presence of 5 µM gefitinib or 0.5 µM lapatinib for 48 h, and densitometric values normalized to the vehicle control. (L) EVs released by A549 cells treated with vehicle (DMSO) or 5 µM gefitinib or 0.5 µM lapatinib and quantified by NTA in terms of particle size (left) and particle number after normalization to the total cell number (right). The cells were stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 for 48 h. Data in (A, B, H, J and L) are presented as mean values of three biological replicates ± SEM, each in technical duplicates and p-values are shown based on two-way ANOVA, followed by multiple paired comparisons conducted by means of Bonferroni’s post-test method. p-values: **p ≤ 0.01; ***p ≤ 0.001. The data in (D, G, I and K) show representative immunoblots of three independent biological replicates along with molecular mass markers in kDa

Fig. 2

TGF-β induces cholesterol synthesis and ERK1/2-SREBP2 activation leading to DHCR7 expression. (A, B) Pearson correlation analysis of mRNA expression in BRCA and LUAD of gene signatures for TGF-β signaling (A) and cholesterol homeostasis (B) relative to the TGFB1 mRNA expression, measured as transcripts per million (TPM) transformed by log₂; data was obtained from TCGA. P- and R-values are listed. (C, D) Quantification of total cholesterol levels in MDA-MB-231 (C) and A549 (D) cells stimulated with 5 ng/mL TGF-β1 for 48 h in the absence or presence of 5 µM MEKi. (E) Quantification of total cholesterol levels in VSF and CD81-EVs enriched from MDA-MB-231 cells stimulated with 5 ng/mL TGF-β1 for 120 h. (F, G) RT-qPCR analysis of the indicated mRNA levels in MDA-MB-231 cells upon stimulation with 5 ng/mL TGF-β1 for 48–120 h (F) or in MDA-MB-231 cells stimulated with 5 ng/mL TGF-β1 in the absence or presence of 5 µM MEKi for 48 h (G). Values represent fold-change of mRNA expression normalized to GAPDH and expressed relative to the level at 0 h TGF-β1 (Ctrl). (H) Expression levels of the indicated cellular proteins and GAPDH serving as loading control from MDA-MB-231 cells treated as in panels C and G, and densitometric values were normalized to the vehicle control. (I) Expression levels of pSREBP2, SREBP2, PAI1 and GAPDH serving as loading control from MDA-MB-231 cells treated with TGF-β1 for 6, 24 and 48 h. Densitometric values were normalized to the vehicle control. (J) RT-qPCR analysis of DHCR7 mRNA levels in MDA-MB-231 cells upon stimulation with 5 ng/mL TGF-β1 for the indicated time points. Values represent mRNA expression normalized to GAPDH. (K) Expression levels of SREBP2 protein with GAPDH serving as loading control in MDA-MB-231 cells transiently transfected with control (siCtrl) or specific siRNA targeting SREBP2 (siSREBP2). Densitometric values were normalized to the siCtrl. (L) Quantification of total cholesterol levels in MDA-MB-231 cells transiently transfected with control siRNA (siCtrl) or siSREBP2 and stimulated with 5 ng/mL TGF-β1 for 48 h. (M) EVs released by MDA-MB-231 cells, which were first transiently transfected with siCtrl or siSREBP2, were quantified by NTA in terms of particle size (left) and particle number after normalization to the total cell number (right). The cells were stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 for 48 h. (N) RT-qPCR analysis of DHCR7 and SQLE mRNA in MDA-MB-231 cells after transient transfection with siCtrl or siSREBP2 after stimulation of the transfected cells with 5 ng/mL TGF-β1 for 48 h. Values represent fold-change of mRNA expression normalized to GAPDH and expressed relative to the level at siCtrl. The cholesterol level data (C-E, L) and the RT-qPCR data (F, G, K and N) are presented as mean values of three biological replicates ± SEM, in technical triplicates and p-values are shown based on two-way ANOVA, followed by multiple paired comparisons conducted by means of Bonferroni’s post-test method: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001. The protein data (H, I and K) show representative immunoblots of at least two independent biological replicates along with molecular mass markers in kDa

Fig. 3

TGF-β induces EV release via DHCR7 and the cholesterol pathway. (A, B) Quantification of total cholesterol levels in MDA-MB-231 (A) and ZR-75-1 (B) cells stimulated with 5 ng/mL TGF-β1 for 48 h after transient transfection with control or siRNA targeting DHCR7. (C, D) EVs released by MDA-MB-231 (C) and ZR-75-1 (D) cells transiently transfected with control or siRNA targeting DHCR7, quantified by NTA in terms of particle size (left) and particle number after normalization to the total cell number (right). The cells were stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 for 48 h. (E) Quantification of total cholesterol levels in MDA-MB-231 cells stimulated with 5 ng/mL TGF-β1 for 48 h in the absence or presence of 20 µM DHCR7i (AY9944) or 1.25 µM simvastatin. (F) EVs released by the MDA-MB-231 cells quantified by NTA in terms of particle size (left) and particle number after normalization to the total cell number (right). The cells were stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 in the absence or presence of 20 µM DHCR7i or 1.25 µM simvastatin for 48 h. (G) Cell viability assay with MDA-MB-231 cells incubated with vehicle DMSO (Ctrl), 5 µM MEKi, 20 µM DHCR7i or 1.25 µM simvastatin for 48 h, in the presence of co-incubation with DMSO, 0.5 µM Dox or 0.25 µM Taxol. (H) Representative immunoblot of at least two independent biological replicates of cleaved PARP1 and Caspase-3 (CASP3) in MDA-MB-231 cells treated with 5 µM MEKi, 20 µM DHCR7i or 1.25 µM simvastatin for 48 h, in the presence of DMSO or 0.5 µM Dox. GAPDH was used as a loading control, and molecular mass (kDa) markers are indicated along with densitometric values of normalized band intensity only in the lanes where the relevant protein markers scored positively. (I) Schematic model created with Biorender.com, summarizing the effect of TGF-β signaling on EV secretion. TGF-β from the extracellular matrix binds to its receptors (TβR) on the plasma membrane activating phosphorylation-dependent (p) SMAD and ERK signaling. In the nucleus, ERK induces SREBP2 phosphorylation, which enhances expression of cholesterol biosynthesis genes (DHCR7, SQLE). SMAD signaling induces invasion and adhesion genes (MMP2, MMP9, ITGA6). The cholesterol synthesis genes lead to cholesterol increase, causing EV secretion as shown in this paper. In addition to EV secretion (invaginated cell membrane), microvesicles are shown to emanate from the cell surface. The cholesterol level data (A, B and E), EV/NTA data (C, D and F) and cell viability (G) are presented as mean values of three biological replicates ± SEM, in technical triplicates and p-values are shown based on two-way ANOVA, followed by multiple paired comparisons conducted by means of Bonferroni’s post-test method: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001

Fig. 4

Analysis of the EV cargo proteome. (A, B) PCA of statistically significant protein expression differences between EVs enriched by the CD81-specific (A) or the CTB-specific (B) method in the supernatant of MDA-MB-231 cells stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1 for 120 h. Three independent biological repeats were analyzed per condition. (C, D) Venn diagram summarizing the total number of significant and differentially expressed proteins in at least two biological conditions shown in panel A and B. (E, F) Volcano plot of the significant (log₂(p-value); y-axis) and differentially expressed (log₂(fold-change/FC); x-axis) common proteins between the same two biological conditions shown in panels A-D respectively, expressed as proteins enriched (Up) in the EVs of control or TGF-β1-stimulated cells. Selected protein IDs are shown. Dotted lines indicate the filtering levels along each axis. (G, H) Tables of highly significant gene ontology (GO) and Reactome (REAC) terms represented in the two biological conditions analyzed using the CD81-specific EV isolation method, indicating the term name and associated false discovery rate (FDR) q-value. (I) RT-qPCR analysis of the indicated mRNAs, selected based on the corresponding proteins that scored significantly in the proteomic analysis of MDA-MB-231 cells after stimulation with 5 ng/mL TGF-β1 for 0 (Ctrl), 48 and 120 h. (J) Active and total TGF-β1 concentration was measured through ELISA-sandwich in EVs derived from control cells (VSF^Ctrl) or cells stimulated with 5 ng/mL TGF-β1 for 120 h (VSF^{+ TGF−β1}). TGF-β1 concentrations are presented based on two-way ANOVA, followed by multiple paired comparisons conducted by means of Bonferroni’s post-test method: **p ≤ 0.01; ****p ≤ 0.0001

Fig. 5

EV cargo MMPs and TGF-β activate TGF-β signaling in recipient cancer cells. (A) Representative immunofluorescence microscopy pictures of MDA-MB-231 cells incubated with EVs derived from control cells (VSF^Ctrl) or cells stimulated with 5 ng/mL TGF-β1 for 120 h (VSF^{+ TGF−β1}) for 3–18 h. The early endosome protein EEA1 (green), the EVs (red) and nuclei (blue) are labeled. (B) Relative luciferase activity generated in MDA-MB-231 cells by transfecting the TGF-β-inducible CAGA₁₂-luc reporter, normalized to β-galactosidase activity generated by a co-transfected reporter, after stimulation of the transfected cells with vehicle (Ctrl), 5 ng/mL TGF-β1, or incubation with 1 × 10⁹ nanoparticles per mL of VSF^Ctrl or VSF^{+ TGF−β1} for 8 h. The cells were also treated with vehicle (Ctrl), 50 µg/ml heparin, 5 µg/mL neutralizing anti-TGF-β antibody or 5 µM TβRi. Data are presented as mean values of three biological replicates ± SEM, each in technical duplicates. (C) Relative CAGA₁₂-luciferase activity generated in MDA-MB-231 cells after incubation of the transfected cells for 8 h with increasing nanoparticle numbers per mL (5 × 10⁷, 1 × 10⁸ and 1 × 10⁹) of VSF^Ctrl or VSF^{+ TGF−β1}. Data are presented as mean values of three biological replicates ± SEM, each in technical duplicates. (D, E) Protein expression levels of the indicated signaling proteins in cellular extracts of MDA-MB-231 cells stimulated or not with 5 ng/mL TGF-β1 or incubated with 1 × 10⁹ nanoparticles per mL of VSF^Ctrl or VSF^{+ TGF−β1} for the indicated time periods, and densitometric values normalized to the control 0 h. Representative immunoblots of three independent biological replicates along with molecular mass markers in kDa are shown. (F) RT-qPCR analysis of the indicated mRNA levels in MDA-MB-231 cells stimulated with vehicle (Ctrl) or 5 ng/mL TGF-β1, or incubated with 1 × 10⁹ nanoparticles per mL of VSF^Ctrl or VSF^{+ TGF−β1} for 48 h. The data are presented as mean values of three biological replicates ± SEM, in technical triplicates. (G) Quantification of active TGF-β1 released from a constant amount (2.5 ng) latent-(L)-TGF-β1 was measured through ELISA-sandwich upon acidification with 1 N HCl for 10 min at room temperature or after incubation for 30 min at room temperature with 1 × 10⁹ nanoparticles per mL of VSF^Ctrl or VSF^{+ TGF−β1}. Note that 2.5 ng L-TGF-β1 contained a portion of mature TGF-β1 (Ctrl). (H) MMP proteolytic activity was measured in 1 × 10⁹ nanoparticles per mL from VSF^Ctrl or VSF^{+ TGF−β1} using EnzChek Gelatinase/Collagenase Assay kit. The VSFs were pre-treated with vehicle DMSO (Ctrl), 25 µM MMP inhibitor (MMPi) or depleted from CD81-positive EVs. (I) Relative luciferase activity generated in MDA-MB-231 cells by transfecting the TGF-β-inducible CAGA₁₂-luc reporter, normalized to β-galactosidase activity generated by a co-transfected reporter, after stimulation of the transfected cells with vehicle (Ctrl), 5 ng/mL TGF-β1, or incubation with 1 × 10⁹ nanoparticles per mL of VSF^Ctrl or VSF^{+ TGF−β1} for 8 h. The VSFs were pre-treated with vehicle DMSO (Ctrl), 25 µM MMPi or 40 µg/ml proteinase K for 1 h. (J) Matrigel invasion assay in trans-wells with MDA-MB-231 cells stimulated with vehicle (Ctrl) or incubated with 1 × 10⁹ nanoparticles per mL of EVs derived from control cells (VSF^Ctrl) or cells stimulated with 5 ng/mL TGF-β1 (VSF^{+ TGF−β1}) for 48 h in the presence of vehicle (DMSO) or 25 µM MMP inhibitor (MMPi). (K, L) In vivo extravasation and collagenous tail-fin invasion assay in zebrafish embryos injected with fluorescently labelled (red) MDA-MB-231 cells stimulated with vehicle (Ctrl) or incubated with 1 × 10⁹ nanoparticles per mL of VSF^Ctrl or VSF^{+ TGF−β1} for 48 h prior to their microinjection in the duct of Cuvier of transgenic zebrafish with GFP-tagged (green) vasculature. Microinjected embryos were then incubated in water in the absence (K) or presence (L) of 5 µM Dox. Images were captured 24 h after microinjection. The data in K and L represent the extravasated cells that invaded the collagenous tail-fin cells as numbers of red fluorescent cell clusters and are plotted as mean values of at least 20 biological replicates (individual embryos) ± SEM and p-values are shown based on Wilcoxon matched-pairs test. Comparisons in panels B, C, F-J are presented as mean values of at least two biological replicates ± SEM. The p-values in B, F, I and J are based on two-way ANOVA, followed by multiple paired comparisons conducted by means of Bonferroni’s post-test method, while in G and H, each in technical triplicates, are shown based on one-way ANOVA, followed by multiple paired comparisons conducted by means of Bonferroni’s post-test method: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001; **** p ≤ 0.0001. (M) Schematic model summarizing the effect of TGF-β signaling on higher EV secretion, assisting tumorigenesis. TGF-β signaling activation enhances EV secretion as summarized in Fig. 3I. TGF-β signaling activated by EVs does not require heparan sulphate proteoglycan-mediated internalization, suggesting that the EV TGF-β signals directly at the cell surface. Secreted EVs act in an autocrine or paracrine manner on tumor cells or other cells in the tumor microenvironment, inducing multiple phenotypes: TGF-β signaling, EMT, cell motility and viability that contributes to resistance of cancer cells to cytotoxic drugs