Volume 19, Issue 10, Pages 2026-2032 (June 2017) MicroRNA-Mediated Dynamic Bidirectional Shift between the Subclasses of Glioblastoma Stem-like Cells  Arun K. Rooj, Franz Ricklefs, Marco Mineo, Ichiro Nakano, E. Antonio Chiocca, Agnieszka Bronisz, Jakub Godlewski  Cell Reports  Volume 19, Issue 10, Pages 2026-2032 (June 2017) DOI: 10.1016/j.celrep.2017.05.040 Copyright © 2017 The Author(s) Terms and Conditions

Cell Reports 2017 19, 2026-2032DOI: (10.1016/j.celrep.2017.05.040) Copyright © 2017 The Author(s) Terms and Conditions

Figure 1 MiR-128 Expression Drives GSC Subtype Shift (A) MiR-128 is diversely expressed in GSC subclasses. Shown is qPCR analysis of mature miR-128 in GSCs (n = 8). Data are shown as mean ± SD (∗∗p < 0.001). (B) Putative miR-128 targets are diversely expressed in GSCs subclasses. 5,666 putative miR-128 target genes were assessed based on the value of expression (p < 0.05, fold > 2 cutoff is indicated) (left). The expression of miR-128 putative target genes in GSCs stratifies P from M GSCs (n = 3). Targets significantly deregulated in M GSCs (n = 569) and P GSCs (n = 741) were queried with a gene signature retrieved from TCGA and identified by clustering with subtype prediction (classical [C], blue; mesenchymal [M], red; proneural [P], magenta; neural [N], green) (right). (C) Engineered GSCs with overexpression and knock-down of miR-128. Shown is qPCR analysis of mature miR-128 in GSCs (n = 3). Data are shown as ± SD (∗∗p < 0.001) (left). Deregulation of miR-128 in GSCs shifts the subtype-specific gene signature. Shown is qPCR analysis of subtype-specific ten-gene signatures in M GSCs (center) and P GSCs (right). Data are shown as a heatmap with cell and gene clustering. NC, negative control (empty pCDH); 128, overexpression of miR-128 (pCDH-miR-128); aNC, negative control (empty MIR-ZIP); a128, knockdown of miR-128 (MIR-ZIP-anti-miR-128). (D) Deregulation of miR-128 shifts GSC transcriptome clustering. Shown is gene microarray analysis of 3,017 genes significantly (p < 0.05, fold > 2) deregulated in GSCs (n = 3) in an miR-128-dependent fashion. Data are shown as a heatmap with cell clustering (left) and as a principal-component analysis (PCA) (center). The expression of genes deregulated by miR-128 in GSCs results in a transitional signature resembling the classical subtype. Targets significantly deregulated in M and P GSCs (n = 3,017) were queried with a gene signature retrieved from TCGA and identified by clustering with subtype prediction (n = 485, overall error rate > 0.05) (C, blue; M, red; P, magenta; N, green) (right). Cell Reports 2017 19, 2026-2032DOI: (10.1016/j.celrep.2017.05.040) Copyright © 2017 The Author(s) Terms and Conditions

Figure 2 PRC Proteins and Their Effectors Are miR-128-Dependent Controllers of GSC Subtype Shift (A) MiR-128 deregulation affects the expression of PRCs, their chromatin marks, and PRC-dependent genes in GSCs. Shown is Western blot analysis of selected factors in control and miR-128 de-regulated GSCs (n = 3). (B) Expression of PRC-dependent genes is shifted by miR-128. Shown is microarray platform analysis of a 30-PRC-dependent gene signature in control and miR-128 de-regulated GSCs. Data are shown as a heatmap with cell and gene clustering (n = 3). (C) Expression of PRC-dependent genes is diversely shifted by Bmi1/Suz12 knockdown and miR-128. Shown is microarray platform analysis of a 30-PRC-dependent gene signature in control, miR-128-overexpressing, and Bmi1/Suz12 siRNA-transfected M GSCs. Data are shown as a heatmap with cell and gene clustering (n = 3). (D) Bmi1/Suz12 knockdown deregulates the GSC transcriptome in a subtype-dependent manner. Shown is gene microarray analysis of genes significantly (p < 0.05, fold > 2) deregulated in GSCs in a Bmi1/Suz12-dependent fashion (n = 3). Data are shown as a heatmap with cell and gene clustering (left) and as a PCA (center). The expression of genes deregulated by Bmi1/Suz12 knockdown in M GSCs results in a transitional signature resembling the classical subtype. Genes significantly deregulated in M and P GSCs (n = 937) were queried with a gene signature retrieved from TCGA and identified by clustering with subtype prediction (n = 380, overall error rate > 0.05) (right). Cell Reports 2017 19, 2026-2032DOI: (10.1016/j.celrep.2017.05.040) Copyright © 2017 The Author(s) Terms and Conditions

Figure 3 MiR-128 Mitigates the Aggressive Phenotype of M GSCs Both In Vivo and In Vitro (A) MiR-128 overexpression reduces the M GSC tumor burden. Shown are MRI (left; scale bar, 2.5 mm) and tumor volume quantification (right) (n = 3). Data are mean ± SD. ∗∗p < 0.01. (B) MiR-128 overexpression in M GSC tumors is associated with prolonged survival. Median survival: M GSC control, 11 days; M GSC miR-128, 16 days. n = 8 mice/group. (C) MiR-128 targets PRCs’ components in vivo. Shown is western blot analysis of M GSC cells (day 0) and in M GSC tumors (day 10) (left). Also shown are western blot quantification (center, n = 3) and qPCR analysis of miR-128 expression in tumors (day 10, right, n = 3). Data are mean ± SD. ∗p < 0.05, ∗∗p < 0.01. (D) MiR-128 overexpression shifts the subtype-specific gene signature in vivo. Shown is qPCR analysis of a ten-gene subtype-specific signature in M GSC tumors (left). MiR-128 overexpression shifts the expression of PRC-dependent genes in vivo. Also shown is microarray platform analysis of a 30-PRC gene signature in M GSC tumors (n = 3). Data are shown as a heatmap with cell/gene clustering (right). Cell Reports 2017 19, 2026-2032DOI: (10.1016/j.celrep.2017.05.040) Copyright © 2017 The Author(s) Terms and Conditions

Figure 4 Targets of miR-128 and miR-128-Dependent PRC-Associated Genes Predict Patient Outcome in GBM Subtypes (A) MiR-128 knockdown in P GSC tumors is associated with shorter survival. Micrographs of tumors (left) and survival (right) are shown. n = 4 or 5 mice/group. Average survival: P GSC control, 168 days; P GSC amiR-128, 143 days. Scale bars, 1 mm (top) and 200 μm (bottom). (B) Expression of miR-128 affects the survival of heterogeneous GSC-originated tumors. Micrographs of tumors (left) and survival (right) are shown. n = 4 mice/group. Average survival: M GSC miR-128/P GSC control, 19 days; M GSC miR-128/P GSC amiR-128, 9 days. Scale bars, 500 μm (top) and 50 μm (bottom). (C) The miR-128-dependent transcriptome separates patients’ survival prediction. Genes deregulated in GSCs queried with a TCGA gene signature were identified by clustering with class (top bar) and subtype (bottom bar) (left), miR-128 deregulated target genes (center; value of expression, p < 0.05, fold > 2, n = 43), and class survival prediction (right). (D) MiR-128 deregulation in M GSCs and P GSCs shifts the transcriptome toward a transitional hybrid classical-like subtype. Cell Reports 2017 19, 2026-2032DOI: (10.1016/j.celrep.2017.05.040) Copyright © 2017 The Author(s) Terms and Conditions