Two-way hierarchical clustering grouped samples as either NHSCs, mixed benign dNFSC and pNFSC or malignant tumour (MPNST) cell lines. Two classes of NFSCs are identifiable, with Class 1 NFSC (green bar beneath the heat map) expression levels attenuated relative to Class 2 NFSC (blue bar beneath the heat map) levels. Genes in clusters C1 and C4, which show decreased or increased expression in benign tumours, respectively, and opposite expression in MPNST cell lines (indicated with asterisks), are abundant in genes associated with cell cycle (AURKA, CDC25B, CDKN2A, CNAP1, INHBA, MCM7, PDGFB) and cell differentiation (ADAM12, ANGPTL4, BMP1, CHL1, IL11, INHBA, PPL, SERPINE2). The bar to the right of the heat map shows five clusters (C1–C5), corresponding to k-means functional clusters listed in Tables S2 and S3 of Supporting Information.

A. Boxplot of SOX9 gene expression microarray data in each of the seven sample types normalized to Schwann cell gene expression (NHSC). Horizontal lines in each bar represent the median SOX9 expression within each sample type and the error bars indicate the range of non-outlier measurements.
B. SOX9 mRNA expression measured by qPCR of individual samples within each sample type and normalized to expression in Schwann cells (NHSC). Fold-change values were transformed to log 10 scale in order to display MPNST samples (ranging from 1,448- to 1,72,950-fold) and NF samples (ranging from 20- to 6,654-fold) on the same graph. Validation of RNA samples that were analysed by microarray (Technical) include: 1–3, NHSCs; 4, dNFSC+/−; 5, dNFSC−/−; 6, dNF; 7, pNFSC; 13, pNF; 14, 22, MPNSTs. Validation of independent RNA samples (Biological) include: 8–12, pNFSCs; 15–21, MPNST cell lines; 15, STS26T; 16, ST8814; 17, S462; 18, T265; 19, 90-8; 20, 88-3; 21, YST1.
C. SOX9 protein expression in Schwann cells (NHSC), MPNST cell lines (26T, 8814, S462, T265) and pNFSC by Western blot (top panel) analysis. β-actin was used as a loading control.
D, E. Immunohistochemical detection of SOX9 in a representative NF and MPNST tissue section with extensive SOX9 expression. Cells stained brown are positive for SOX9. (e) The percentage distribution of SOX9-positive cells in a panel of dNF, pNF and MPNST sections. The majority of samples screened were represented in the microarray analysis with the addition of ten independent samples, including three dNFs, five pNFs and two MPNSTs.
F. Immunofluorescent detection of SOX9 (TRITC = red) and nuclei (DAPI = blue) in cultured NHSCs, NFSC and ST8814 NF1-derived MPNST cells. Similar results were obtained in STS26T sporadic MPNST cells (data not shown). Fields shown are representative of each population. Merging fluorescent images detecting DAPI and TRITC (Merge) highlights nuclear localization of SOX9, mainly in MPNST.

SOX9 expression throughout the Schwann cell lineage is inferred from work in species other than human. Red represents high SOX9 expression, yellow low SOX9 expression. (1) dNF and pNF Schwann cells express intermediate levels of SOX9 and show gene signature characteristic of Schwann cell progenitors/immature Schwann cells. (2A) A neural crest gene signature is characteristic of MPNST cells, and neural crest cells are known to express higher levels of SOX9 than more mature cells. (2A, 2B) Neural crest cells may give rise directly to MPNST, or MPNST may form indirectly via a NF-like cell intermediate.

Differentially expressed transcripts in NF1 peripheral nerve cell culture samples were filtered to identify genes with similar patterns of expression in solid tumours. A total of 1,708 transcripts (60%) were identified and clustered across NHSCs, dNFSCs, pNFSCs, MPNST cell lines (annotated as in Fig 1), and dNFs, pNFs and MPNSTs. The bar to the right of the heat map shows five clusters (C6–C11), corresponding to k-means functional clusters listed in Table S5 of Supporting Information.

A. Confirmation of reduction in SOX9 RNA expression in NFSCs.
B. NFSCs infected with lentivirus expressing SOX9 shRNA or non-specific shGFP control were plated 7 days post-selection in puromycin to measure cell survival in 4 days using an MTS assay. A trend towards a decrease in cell survival was observed in shSOX9-expressing cells compared to shGFP-expressing cells but was not statistically significant. The corresponding phase contrast images to the right show NFSCs infected with shGFP (NFSC shGFP) or SOX9 shRNA (NFSC shSOX9).
C–E. Confirmation of reduction in SOX9 RNA expression in MPNST cells by qPCR (c) and SOX9 protein expression by western blot (d, e). Expression levels were measured 1 day (d) or 7 days (c, e) post-selection in puromycin and normalized to the non-specific shGFP control. Expression of β-actin was used as a control.
F. Reduction in MPNST cell number (18-fold) in the presence of SOX9 shRNA relative to shGFP control. Cells were plated in triplicate and represent three independent infections (*p = 0.05). The top right panel represents the confluent dish of MPNST cells infected with shGFP lentiviral particles three days post-selection with puromycin; bottom right panel represents dying MPNST cells infected with shSOX9 lentiviral particles three days post-selection with puromycin. Similar results were observed with three different SOX9 shRNAs.
G. MPNST cells were infected with shGFP or shSOX9 lentiviral particles for MTS analysis of cell accumulation during a time course of 1–4 days post-selection in puromycin. Cells were plated in triplicate in the presence or absence of puromycin to account for infection efficiency. All shRNAs infect with similar efficiency at greater than 90%. Values are corrected for infection efficiency and represent three independent infections.
H. MPNST cells treated as described in (c–e) and assayed for apoptosis by TUNEL staining three days post-selection in puromycin. Uninfected and shGFP-infected cells have similar numbers of TUNEL-positive cells with a significant increase in TUNEL-positive cells in the presence of shSOX9 (**p = 0.002).