Cancer Cell-Secreted miR-33a Reduces Stress Granule Formation by Targeting Polyamine Metabolism in Stroma to Promote Tumourigenesis

Abstract

Tumour progression depends on the bidirectional interactions between cancer and stroma in the heterogeneous tumour microenvironment (TME) partially through extracellular vesicles (EVs). However, the secretary mechanism and biological effect of cancer cell derived EVs on tumour survival under starvation is poorly defined. Here, we identify cancer cells selectively secrete miR-33a with the assistance of aconitase 1 (ACO1), an iron-responsive RNA binding protein, under glucose starvation and lower iron level, which affiliates the binding capability of miR-33a and ACO1. Exosomal miR-33a suppresses putrescine biosynthesis by targeting AGMAT in cancer-associated fibroblasts (CAFs) from tumour core region, where putrescine inhibits the expression of demethylase KDM5C. TIA1 gene, stress granule (SG) marker, is tightly regulated by miR-33a/KDM5C/H3K4me3 axis and exosomal miR-33a diminishes the formation of stromal SGs in CAFs. Collectively, our study reveals tumour selectively secretes miR-33a-5p through EVs to remodel the stromal SG formation and gain survival possibility for cancer cells in tumour core region, highlighting a novel regulatory mechanism of iron and nutrient level on EV secretion and the function of polyamine metabolism in reshaping epigenetic profiles.

Keywords: epigenetic reprogramming; extracellular vesicles; polyamine metabolism; stress granules; tumour microenvironment.

FIGURE 1

Polyamine metabolism spatially alters in tumour microenvironment. (A) Schematic diagram of the tumour spatial structure. Color deepening area, tumour core region. (B) GSEA analysis identified the cellular response to starvation signature in phenotype tumour core and the arginine metabolic process signature in phenotype tumour margin. (C) Schematic diagram depicting polyamine metabolic pathway. ARG1, arginase 1; ADC, arginine decarboxylase; ODC1, ornithine decarboxylase 1; AGMAT, agmatinase; SRM, spermidine synthase; SMS, spermine synthase; SAT1, diamine acetyltransferase 1; PAOX, polyamine oxidase. (D) Representative mass spectrometry imaging for spermidine and spermine abundance in 231 and 4T1 wild‐type tumours and 231/Rab27a KD tumour. (E) 4T1 xenograft tumours were harvested and separated into core and margin samples. Metabolites were extracted from each sample and polyamine concentration was measured by targeted LC‐MS/MS. Data are presented as mean ± s.d., n = 6 biological replicates, paired two‐tailed Student's t‐test. (F) Relative putrescine levels within tumour core and margin regional mouse cancer‐associated fibroblasts (mCAFs) derived from 4T1/WT tumours (left) and mCAFs or tumour cells from tumour core region (right). Putrescine concentrations were measured by ELISA kit. Data are presented as mean ± s.d., n = 5 biological replicates, paired two‐tailed Student's t‐test. (G) RNA levels of Agmat in mCAFs derived from tumour core and marginal regions from 231/WT and 231/Rab27a KD xenograft tumours were detected by RT‐qPCR. Data are presented as mean ± s.d., n = 5 biological replicates, paired two‐tailed Student's t‐test. (H) 231/WT and 231/Rab27a KD tumour tissues were harvested from NSG mice. Tissues from the core regions were separated to isolate mCAFs followed by western blots analysis. (I) Representative mIHC images showing AGMAT and putrescine staining of NSG 231/WT and 231/Rab27a KD xenograft tumour in core region mCAFs (α‐SMA). α‐SMA (green), AGMAT/Putrescine (red) and DAPI (blue). Quantification of AGMAT/Putrescine and α‐SMA colocalization presented as mean ± s.d., n = 3 mice, unpaired two‐tailed Student's t‐test.

FIGURE 2

Cancer cell‐secreted miR‐33a inhibits putrescine production by targeting AGMAT in CAFs. (A) Venn diagram of upregulated exosomal miRNA from 231 ‐G EV versus 231 EV by miRNA‐seq (fold change > 1.5) and predicted miRNAs targeting AGMAT (TargetScan). ‐G indicates glucose free. (B) Relative intracellular or exosomal miR‐33a levels from glucose starvation cultured MDA‐MB‐231 compared to those for complete medium cultured cells were determined by RT‐qPCR. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. (C) RT‐qPCR‐determined levels of exosomal miR‐33a from different breast cancer cell lines under glucose starvation treatment compared with complete medium cultured. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. (D) RNA levels of AGMAT in human cancer‐associated fibroblasts (hCAFs) transfected with miR‐33a or control mimic. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. (E) Protein levels of AGMAT in hCAFs transfected with miR‐33a or control mimic. (F) Predicted miR‐33a targeting human and mouse AGMAT 3'UTR binding sites. The corresponding sequences in WT and mutated reporters are shown. (G) Responsiveness of the WT and mutant AGMAT/Agmat 3'UTR reporters to miR‐33a in stable transfected cell lines. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. (H) AGMAT levels in 231 transfected with AGMAT‐Myc‐WT 3′UTR or AGMAT‐Myc‐Mut 3′UTR were detected followed by miR‐33a mimic or negative control mimic treatment. (I) RNA levels of AGMAT in 231 transfected with AGMAT‐Myc‐WT 3′UTR or AGMAT‐Myc‐Mut 3′UTR were detected followed by miR‐33a mimic or negative control mimic treatment. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. (J) Human CAFs were treated with EVs for 24 h to detect AGMAT level. Data are presented as mean ± s.d., n  =  3 biological replicates, one‐way analysis of variance (ANOVA), Tukey's multiple comparisons test. (K) AGMAT protein level of EVs treated hCAFs was detected by western blots. (L) The intracellular putrescine levels of EVs treated mCAFs measured by LC‐MS/MS. (succinate‐1,4‐¹³C₂ as internal standard). Data are presented as mean ± s.d., n  =  3 biological replicates, one‐way ANOVA, Tukey's multiple comparisons test. (M) Representative mIHC images showing AGMAT and putrescine staining of NSG 231/WT tumour and 231ΔmiR‐33a tumour in core region mCAFs (α‐SMA). α‐SMA (green), AGMAT/Putrescine (red) and DAPI (blue). Quantification of AGMAT/Putrescine and α‐SMA colocalization presented as mean ± s.d., n = 3 mice, unpaired two‐tailed Student's t‐test. (N) Tumour volume of 231/WT and 231ΔmiR‐33a xenograft in NSG mice. Data are presented as mean ± s.d., n = 7 biological replicates, two‐way ANOVA, Sidak's multiple comparisons test.

FIGURE 3

Putrescine enhances H3K4 tri‐methylation by inhibiting KDM5C expression. (A) GSEA analysis identified the regulation of histone H3K4 methylation signature and histone H3K4 trimethylation upregulated in phenotype tumour margin. (B) Mouse CAFs from core region of 4T1/WT and 4T1/Rab27a KO xenograft tumours were used for histone extraction. Histone methylation levels were assessed by western blots. (C) The schematic diagram showing the histone lysine methyltransferase and demethylases. (D) RNA levels of histone methyltransferase and demethylase of mCAFs treated with EVs. Data are presented as mean ± s.d., n  =  3 biological replicates, one‐way ANOVA, Tukey's multiple comparisons test. (E) KDM5C and H3K4me3 protein levels of mCAFs with EVs treatment were examined by western blots. (F) Western blots showing the expression levels of KDM5C and H3K4me3 in mCAFs treated with PBS or polyamine (25 µM). Arginine free medium was supplied. (G) Responsiveness of the human KDM5C 5'UTR reporters to polyamine (25 µM). Data are presented as mean ± s.d., n  =  3 biological replicates, one‐way ANOVA, Dunnett's multiple comparisons test. (H) RNA levels of histone methyltransferases and demethylases of mCAFs treated with gradient concentration of putrescine. Arginine free medium was supplied. Data are presented as mean ± s.d., n  =  3 biological replicates, one‐way ANOVA, Dunnett's multiple comparisons test. (I) Western blots showing the expression levels of KDM5C and H3K4me3 in mCAFs treated with gradient concentration of putrescine. Arginine free medium was supplied. (J) KDM5C levels in mCAFs from the core regions of 4T1/WT or 4T1/Rab27a KO mice were assessed by western blots. (K) Representative mIHC images showing KDM5C staining of 4T1/WT tumour and 4T1ΔmiR‐33 tumours in core region mCAFs (α‐SMA). α‐SMA (green), KDM5C (red) and DAPI (blue). Quantification of KDM5C and α‐SMA colocalization presented as mean ± s.d., n = 3 mice, unpaired two‐tailed Student's t‐test.

FIGURE 4

MiR‐33a reprograms epigenetic profile to induce stress granule disassembling in CAF. (A) Experimental workflow for studying the H3K4me3 landscapes of EV treated mCAFs. (B) Heatmap of ChIP‐seq experiments using antibody against H3K4me3 in EV treated mCAFs. (C) ChIP‐qPCR was used to detect the binding of H3K4me3 with putative elements near Tia1, Itln1 and Tnfsf18 in mCAFs after EV treatment. Relative enrichment score was presented as the ratio of corresponding group to 231 EV treatment. Data are presented as mean ± s.d., n  =  3 biological replicates, one‐way ANOVA, Tukey's multiple comparisons test. (D) Immunofluorescence of mCAFs treated with EVs under low glucose medium or exposed to sodium arsenite (AS, 400 µM; 1 h) and stained for the SG markers TIA1 and G3BP1. Scale bar, 50 µm. Quantification of the number and diameters of stress granules (SGs) were presented as mean ± s.d., n  =  6 different cells, one‐way ANOVA, Tukey's multiple comparisons test. (E) Immunofluorescence of EV treated mCAFs transfected with TIA1 or Vec under low glucose medium followed by staining for SG marker TIA1 and G3BP1. Scale bar, 50 µm. Quantification of the number and diameters of stress granules (SGs) were presented as mean ± s.d., n  =  6 different cells, unpaired two‐tailed Student's t‐test. (F) Representative scatter plots of PI versus Annexin V showed the apoptosis of EVs treated CAF/Vec or CAF/AGMAT by flow cytometry. Quantification of apoptosis and necrosis are presented as mean ± s.d., n  =  3 biological replicates, one‐way ANOVA, Tukey's multiple comparisons test.

FIGURE 5

Nutrient supply blocks the effect of miR‐33a in tumour core. (A) The strategy of BALB/c 4T1 xenograft tumours with PBS or glucose injection into core region was shown. Glucose (1 g/L) was administered into tumour core every other day at 25 µL per tumour 3 weeks. (B) The glucose concentrations of core and margin regions of BALB/c 4T1 glucose injected tumours were detected. Data are presented as dots, n = 5 mice, paired two‐tailed Student's t‐test. (C) Representative in situ hybridization images showing the miR‐33a level in the core and margin regions of glucose‐injected 4T1 tumour. Scale bar, 200 µm. Quantification of miR‐33a level presented as mean ±s.d., n = 5 mice, paired two‐tailed Student's t‐test. (D) RT‐qPCR‐determined miR‐33a levels of EVs from 4T1 glucose injection tumour in core and margin regions. Data are presented as mean ± s.d., n = 5 biological replicates, paired two‐tailed Student's t‐test. (E) Samples from core or margin mCAFs of the 4T1 glucose injected tumours as indicated were harvested and gene expression levels were detected by western blots. (F) The injection strategy of BALB/c 4T1 xenograft tumours with polyamine injection into core region was shown. Putrescine, spermidine and spermine (25 µM) was administered into tumour core every other day at 25 µL per tumour. (G) Core mCAFs of the 4T1 PBS or polyamine injected tumours as indicated were harvested and protein levels were detected by western blots. (H) Tumour volume of 4T1 PBS or polyamine injected tumours. Data are presented as mean ± s.d., n = 5 mice, two‐way ANOVA, Sidak's multiple comparisons test.

FIGURE 6

RNA binding protein ACO1 assists miR‐33a secretion under low iron level. (A) MiR‐33a levels were detected in EVs derived from MCF‐10A (left) or 231 (right) stably transfected with Lenti‐miR‐33a by RT‐qPCR. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. (B) Diagram shows potential RNA binding protein levels in 231 control or ‐G EV detected by EVs Mass Spectrum. (C) MiR‐33a levels in EVs derived from 231/miR‐33a or putative RNA binding protein knockdown cells were measured by RT‐qPCR (Top). Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. Representative images of immuno‐EM analysis of small EVs purified from 231/WT or 231/miR‐33a and visualized with 10‐nm gold‐gold particles (Bottom). Scale bar, 100 nm. (D) Volcano plot of expression differences between 231 EVs and 231 ACO1 knockdown EVs as determined via miRNA‐seq. Pink‐colored points and cerulean blue‐colored points represent gene expression fold change > 1.5 and < −1.5, respectively (p values < 0.05). (E) Schematic diagram showed the sequence of miR‐33a and the structure of ACO1 binding miR‐33a. Color band was 3D structure of ACO1 while the orange chain showed the structure of hsa‐miR‐33a‐5p, red band represented the domain 4 of ACO1, docking result between hsa‐miR‐33a and ACO1. The purple band was the binding sequence CAUUG. The visualization was done by PyMOL 2.5.2. The structure of human ACO1 protein was obtained from AlphaFold Protein Structure Database based on PDB 2B3X. (F) Iron levels within tumour core and margin regions derived from patients (left) and mice (right). Data are presented as mean ± s.d., n = 8 biological replicates, paired two‐tailed Student's t‐test. (G) RIP assays with anti‐ACO1 antibody (with IgG as negative control) were performed in the cell (left) and EV (right) lysates from miR‐33a overexpressed in MDA‐MB‐231 cell lines. miR‐33a levels in immunoprecipitated samples were detected by RT‐qPCR and normalized to the corresponding input samples. Data are presented as mean ± s.d., n = 3 biological replicates, unpaired two‐tailed Student's t‐test. FeCM, ferric carboxymaltose. (H) Western blots analysis of ACO1 expression in exosome lysates subjected to miRNA pulldown with biotinylated miR‐33a or biotinylated miR‐33a mutant probes. MiR‐33a‐5p: GUGCAUUGUAGUUGCAUUGCA, miR‐33a‐5p mut1: GUCGUAACUAGUUGCAUUGCA, miR‐33a‐5p mut2: GUGCAUUGUAGUUGGUAACGA. (I) Relative intracellular (left) or exosomal (right) miR‐33a‐5p levels from core and margin regions of 4T1 FeCM injected tumours were detected by RT–qPCR. Data are presented as mean ± s.d., n = 5 biological replicates, paired two‐tailed Student's t‐test. (J) Tumour volume of 4T1/WT tumours with PBS or ferric carboxymaltose (FeCM) injection (0.3 mg) into tumour core region. Data are presented as mean ± s.d., n = 5 mice, two‐way ANOVA, Sidak's multiple comparisons test. (K) Representative tumour images in each group were shown.

FIGURE 7

MiR‐33a/AGMAT axis widely exists in primary breast tumours and other cancer types. (A) RNA levels of indicated genes in core and margin hCAFs of patient primary tumours were detected. Data are presented as mean ± s.d., n = 5 patients, paired two‐tailed Student's t‐test. (B) Protein levels of indicated genes in core and margin hCAFs of patient primary tumours were detected. (C) Relative hsa‐miR‐33a‐5p levels in BRCA tumours and normal mammary tissues from TCGA databases. miR‐33a‐5p is upregulated in BRCAs. Statistical significance was assessed using two‐tailed Mann–Whitney test (left) and paired two‐tailed Student's t‐test (right). (D) Relative hsa‐miR‐33a‐5p levels of BRCA patients at different clinical stages by means of TNM stages from TCGA databases. Data are presented as mean ± s.d., one‐way ANOVA, Dunnett's multiple comparisons test. (E) Representative immunofluorescence images showing patient tissue SGs staining of core and margin primary tumour. Scale bar, 50 µm. Quantification of colocalization presented as mean ± s.d., n = 3 for core or margin of patient breast cancer samples, paired two‐tailed Student's t‐test. (F) Protein levels of indicated genes in Hep3B, SW480 and A549 tumour core and margin mCAFs were detected through western blots. (G) BC tissues analyzed for correlations among selected gene expression patterns by IHC/in situ hybridization‐determined scores. The Pearson's r, sample size (N; number of independent tissue samples) and p value (paired two‐tailed t‐test) are indicated.