letters
PRC2 is recurrently inactivated through EED or SUZ12 loss in malignant peripheral nerve sheath tumors
npg
© 2014 Nature America, Inc. All rights reserved.
William Lee1,2,17, Sewit Teckie2,3,17, Thomas Wiesner3,17, Leili Ran3,17, Carlos N Prieto Granada4, Mingyan Lin5, Sinan Zhu3, Zhen Cao3, Yupu Liang3, Andrea Sboner6–8, William D Tap9,10, Jonathan A Fletcher11, Kety H Huberman12, Li-Xuan Qin13, Agnes Viale12, Samuel Singer14, Deyou Zheng5,15,16, Michael F Berger3,4, Yu Chen3,9,10, Cristina R Antonescu4 & Ping Chi3,9,10 Malignant peripheral nerve sheath tumors (MPNSTs) represent a group of highly aggressive soft-tissue sarcomas that may occur sporadically, in association with neurofibromatosis type I (NF1 associated) or after radiotherapy1–3. Using comprehensive genomic approaches, we identified lossof-function somatic alterations of the Polycomb repressive complex 2 (PRC2) components (EED or SUZ12) in 92% of sporadic, 70% of NF1-associated and 90% of radiotherapyassociated MPNSTs. MPNSTs with PRC2 loss showed complete loss of trimethylation at lysine 27 of histone H3 (H3K27me3) and aberrant transcriptional activation of multiple PRC2-repressed homeobox master regulators and their regulated developmental pathways. Introduction of the lost PRC2 component in a PRC2-deficient MPNST cell line restored H3K27me3 levels and decreased cell growth. Additionally, we identified frequent somatic alterations of CDKN2A (81% of all MPNSTs) and NF1 (72% of non-NF1-associated MPNSTs), both of which significantly co-occur with PRC2 alterations. The highly recurrent and specific inactivation of PRC2 components, NF1 and CDKN2A highlights their critical and potentially cooperative roles in MPNST pathogenesis. MPNSTs arise from peripheral nerves and associated cellular components and represent a highly aggressive subtype of soft-tissue sarcoma1. MPNSTs metastasize early and are often resistant to radiotherapy and chemotherapy. Conventional MPNSTs present in three distinct clinical settings: sporadically, in association with neurofibromatosis type I (NF1 associated) or in association with previous radiotherapy (radiotherapy associated), accounting for approximately
45%, 45% and 10% of cases, respectively2,4,5. Histologically, MPNSTs are characterized by intersecting fascicles of monotonous spindle cells with hyperchromatic nuclei and high mitotic counts with focal areas of necrosis, but accurate diagnosis remains challenging because of the lack of specific immunohistochemical and molecular biomarkers5,6. Among individuals with NF1, loss of the non-mutant allele of NF1 is thought to be the key driver in benign NF1-associated neurofibroma7. Little is known of the genetic alterations that mediate progression from neurofibroma to MPNST in individuals with NF1 or of the molecular pathogenesis of sporadic and radiotherapyassociated MPNST. To investigate the molecular basis of MPNST, we performed wholeexome sequencing, DNA copy number and loss-of-heterozygosity (LOH) profiling, and whole-transcriptome sequencing (RNA-seq) on a discovery cohort consisting of paired normal-tumor tissues for 15 MPNSTs from 12 individuals (6 NF1 associated, 4 sporadic, 4 radiotherapy associated and 1 epithelioid) (Supplementary Tables 1 and 2). Epithelioid MPNST is a rare histological variant of MPNST that is composed of exclusively epithelioid malignant cells with diffuse immunoreactivity for the S100 protein and is not associated with NF1 (ref. 6). We identified four frameshift mutations and one splice-site mutation in EED (Fig. 1a,c and Supplementary Fig. 1). RNA-seq validated aberrant EED splicing in the sample with the mutated splice site (Supplementary Fig. 2a). All five samples showed LOH at the EED locus, three samples (11T, 12T and 14T) by heterozygous deletion of the normal allele (Supplementary Fig. 1b) and two samples (15T and 16T) by copy-neutral LOH (Supplementary Fig. 2b). These data suggest that samples with EED mutation have complete loss of EED function.
1Computational
Biology Program, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 2Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 3Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 4Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 5Department of Genetics, Albert Einstein College of Medicine, Bronx, New York, USA. 6Department of Pathology and Laboratory Medicine, Weill Cornell Medical College, New York, New York, USA. 7Institute for Computational Biomedicine, Weill Cornell Medical College, New York, New York, USA. 8Institute for Precision Medicine, Weill Cornell Medical College, New York, New York, USA. 9Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 10Department of Medicine, Weill Cornell Medical College, New York, New York, USA. 11Department of Pathology, Brigham and Women’s Hospital, Boston, Massachusetts, USA. 12Genomics Core Laboratory, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 13Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 14Department of Surgery, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 15Department of Neurology, Albert Einstein College of Medicine, Bronx, New York, USA. 16Department of Neuroscience, Albert Einstein College of Medicine, Bronx, New York, USA. 17These authors contributed equally to this work. Correspondence should be addressed to P.C. ([email protected]). Received 21 April; accepted 26 August; published online 21 September 2014; doi:10.1038/ng.3095
Nature Genetics VOLUME 46 | NUMBER 11 | NOVEMBER 2014
1227
letters
npg
© 2014 Nature America, Inc. All rights reserved.
b
NF1
CDKN2A
Germline Germline Germline Germline Germline Germline D1237_splice D173Y, hom deletion Hom deletion WT Hom deletion Hom deletion K99*, het loss K99*, het loss WT
Sample ID MPNST 10 MPNST 22 MPNST 28 MPNST 51 MPNST 04.2 MPNST 14 MPNST 46.1 MPNST 46.3 MPNST 43.1 MPNST 13.1 MPNST 19 MPNST 06.1 MPNST 06.2 MPNST 04.1 MPNST 05.1 MPNST 12.1 5T MPNST 21.1 MPNST 03 MPNST 25 MPNST 47.1 MPNST 23 MPNST 18 MPNST 01 MPNST 27 MPNST 29 MPNST 02.2 MPNST 02.1 MPNST 48 MPNST 02.3 MPNST 45 MPNST 24 MPNST 52 MPNST 44 MPNST 53 MPNST 26 MPNST 50 MPNST 12.2 MPNST 47.2 MPNST 21.2 MPNST 46.2 MPNST 13.2 MPNST 43.2 MPNST 05.2
NF1 CDKN2A F1740fs, L43P Het loss I1979fs, het loss WT L1590*, het loss Hom deletion Q1188*, C2278* Het loss R1534*, het loss Hom deletion R1534* Het loss R1769* Het loss R1769* Hom deletion R304*, Y1989* Het loss R440*, het loss WT T165fs, I1482T WT V1957fs Het loss V1957fs Hom deletion W837*, R1534* WT V1231fs, Y2285* Het loss F1593del Het loss T2206R, het loss Het loss W1559*, het loss Hom deletion T676fs, het loss Hom deletion Q2434*, hom deletion Het loss Germline WT Hom deletion Hom deletion Hom deletion Het loss L1622fs, het loss Het loss K286fs, het loss Hom deletion Het loss Hom deletion Het loss V51_splice WT V51_splice WT WT Het loss V51_splice K195fs, hom deletion Hom deletion Het loss Hom deletion Q2234* Hom deletion V567fs, het loss Hom deletion WT Hom deletion WT WT WT WT F1593del Het loss I2057fs Het loss D2662fs, het loss WT R1769* WT R440*, Q1336* Het loss R681*, Y1989* WT R1534*, Y2285* WT
NF1 associated
Sporadic
Hom deletion Hom deletion Hom deletion Hom deletion Het loss Hom deletion Het loss Hom deletion Hom deletion Hom deletion Hom deletion WT Hom deletion Hom deletion WT
TP53 WT WT WT Het loss WT Het loss WT R273C WT WT Het loss T118fs WT WT WT
EED F289fs H213fs H213fs WT WT R287_splice E249fs WT WT WT WT WT WT WT WT
SUZ12 WT WT WT Het loss WT WT WT Hom deletion Hom deletion WT SV + het loss SV + het loss SV + het loss SV + het loss WT
TP53 EED SUZ12 WT WT WT C135fs WT H471R, het loss Het loss V121D, het loss WT WT A114P WT WT H213fs Het loss Het loss WT Het loss Het loss WT G291fs, het loss Y220H, het loss WT WT P190L, het loss WT Het loss WT G167R, het loss Het loss Q144fs Het loss WT V173A, het loss WT WT V173A, het loss WT WT Het loss WT Het loss R158H, het loss WT WT WT WT WT WT WT Hom deletion WT WT Q504fs, het loss WT WT Het loss Hom deletion WT Hom deletion WT WT WT WT WT Hom deletion WT H92fs, het loss Hom deletion WT WT Hom deletion Het loss Het loss Het loss N131I WT N651fs, het loss V157fs, het loss Hom deletion WT WT D317fs WT E285K, het loss Hom deletion WT V157fs Hom deletion WT Het loss Hom deletion Y565fs, het loss WT Hom deletion WT WT WT WT WT WT L564fs, het loss WT Hom deletion WT WT WT WT WT WT WT WT WT WT WT WT WT WT WT Het loss WT WT WT WT WT WT Het loss WT WT WT WT WT
Radiotherapy associated
Epithelioid
c PRC2 core components H3K27me2 and H3K27me3
EZH1-EZH2 EED
SUZ12 RbAp46/48
EED 5
Number of mutations
Sample ID 15T 11T 12T 9T 19T 14T 16T 8T 10T 4T 2T 13T 7T 18T 20T
H92fs
0
A114P
V121D
H213fs G167R E249fs R287_splice F289fs
WD40 0
100
D317fs
WD40
200
300
400
441
SUZ12 5
Number of mutations
a
G291fs 0
H471R
Q504fs L564fs Y565fs
N651fs
VEFS-Box 0
200
400
d
600
739
NF1 All WT 3
1
CDKN2A 8 1
PRC2 5 1
31 P = 0.001 2
Neurofibroma
Figure 1 The most frequent genetic alterations in MPNST (NF1 associated, sporadic, radiotherapy associated and epithelioid) and neurofibroma. (a,b) Nonsynonymous single-nucleotide variants (SNVs) and copy number variants (CNVs) in 15 MPNSTs with matched normal samples identified by whole-exome sequencing, SNP6.0 array analysis and RNA-seq (a) and in 37 MPNSTs and 7 neurofibromas identified by targeted sequencing (IMPACT) (b). Brackets indicate two different tumor samples from the same individual. WT, wild type; hom, homozygous; het, heterozygous; SV, structural variation (light shading indicates heterozygous loss and dark shading indicates homozygous loss). (c) Schematics of the nonsynonymous SNVs observed in the PRC2 core components EED and SUZ12 in 15 MPNST samples analyzed by whole-exome sequencing and 37 MPNST samples analyzed by custom IMPACT. (d) Schematic of the overlap of mutations affecting NF1, PRC2 components (EED or SUZ12) and CDKN2A in all MPNSTs (NF1 associated, sporadic, radiotherapy associated and epithelioid). With Fleiss’ κ statistics, three-way comparison of NF1, CDKN2A and PRC2 component (EED or SUZ12) genetic alteration suggested that they significantly co-occur: κ = 0.21, P = 0.001.
We further identified two homozygous and five heterozygous deletions of SUZ12 (Fig. 1a,c and Supplementary Figs. 1 and 3a). We examined RNA-seq profiles for the SUZ12 transcript among the five samples with heterozygous loss of the gene. Two samples, 9T and 12T (with EED mutations encoding p.His213fs), expressed full-length SUZ12 transcript, indicating that the remaining SUZ12 copy is intact (Supplementary Fig. 1b and data not shown). Remarkably, the other three samples displayed structural alterations of the SUZ12 transcript, starting at exon 6, exon 10 and exon 4 in samples 2T, 7T and 13T, respectively (Supplementary Fig. 3b–d). These structural changes are likely due to local genomic rearrangements of the remaining copy, which were not identified by standard whole-exome sequencing analysis. Indeed, for 7T and 18T, derived from two tumors from the same individual, there was a DNA break in exon 10 identified through manual examination of whole-exome sequencing data (Supplementary Fig. 3c). We designated these cases as having structural variation and 1228
heterozygous loss at the SUZ12 locus, and, intriguingly, all constituted radiotherapy-associated MPNSTs (Fig. 1a). EED and SUZ12 are the core components of PRC2 and, together with EZH1 and EZH2, establish and maintain the di- and trimethylation of lysine 27 of histone H3 (H3K27me2 and H3K27me3)8. EED and SUZ12 genetic alterations were mutually exclusive and were collectively found in 80% (12/15) of all MPNSTs (Fig. 1a,c). We did not observe any genetic alterations in other PRC2 core components, including EZH1 and EZH2 (Supplementary Table 3). We found recurrent nonsense mutations and homozygous deletions in NF1 in 87.5% (7/8) of sporadic and radiotherapy-associated MPNSTs (Fig. 1a and Supplementary Fig. 1). These data, in combination with the germline mutations in NF1 in NF1-associated MPNSTs, suggest that NF1 is a uniquely important tumor suppressor in MPNST. Alterations of the CDKN2A locus and of TP53 have been reported in MPNST9–12. We observed homozygous deletion and VOLUME 46 | NUMBER 11 | NOVEMBER 2014 Nature Genetics
letters b
PRC2 WT SUZ12 loss EED loss
PRC2 WT SUZ12 loss EED loss
c 0
–4
4
19
S8 S2 S7 S5 SS13 S16 S11 S10 S12 S14 S4 S9 S19 S22 S21 S20
100 ●
50
PC2
14
11 ● 2 ●●● 5 7 ● 16
0
9
12
●●● 13 ● 10● 8 ●
4
–50
●
●
●
● 20 21 22
–100
npg
50
PRC2 WT (n = 6) SUZ12 loss (n = 6) EED loss (n = 4)
15 10 5 0 –2
100
e
f
Bernstein neural progenitor cell H3K27me3
Orkin PRC2 module 0.6
Enrichment score
20
0 PC1
EM FO X2 XN H 4 H MX O 1 X H A1 O 3 XD 11 IG F O 2 TX PA 2 X PA 2 X PI 7 TX TL 1 X1
Normalized expression (log2)
© 2014 Nature America, Inc. All rights reserved.
d
–50
Enrichment score
–100
NES = 3.24 FDR = 0 P=0
0.4 0.2 0
PRC2 loss
PRC2 WT
DNA-binding region:homeobox Homeobox IPR017970:homeobox, conserved site IPR001356:homeobox Developmental protein IPR012287:Homeodomain-related SM00389:HOX GO:0030182:neuron differentiation GO:0043565:sequence-specific DNA binding GO:0003700:transcription factor activity GO:0030528:transcription regulator activity GO:0048598:embryonic morphogenesis GO:0048736:appendage development GO:0060173:limb development GO:0035108:limb morphogenesis GO:0035107:appendage morphogenesis GO:0001501:skeletal system development GO:0048666:neuron development GO:0030900:forebrain development GO:0051216:cartilage development glycoprotein GO:0007389:pattern specification process GO:0042471:ear morphogenesis Cell membrane DNA binding GO:0048562:embryonic organ morphogenesis
0.6
NES = 3.26 FDR = 0 P=0
0.4 0.2 0
PRC2 loss
PRC2 WT
0 5 10 15 20 25 –log10 FDR (Benjamini)
g Enrichment score
a
Lander brain H3K27me3 0.6
NES = 3.34 FDR = 0 P=0
0.4 0.2 0
PRC2 loss
PRC2 WT
Figure 2 MPNSTs with PRC2 loss exhibit distinct gene expression pattern from MPNSTs with wild-type PRC2, signifying activation of developmentally suppressed pathways. (a) PCA of the MPNST whole-transcriptome data showed that the samples with wild-type PRC2 and those with PRC2 component (SUZ12 and EED) loss segregate by principal component 1 (PC1). Each sample is colored on the basis of its corresponding PRC2 mutational status derived from whole-exome sequencing data, except for samples 21 and 22, which are colored on the basis of manual examination of the RNA-seq data for mutations in PRC2 components. (b) Heat map of genes with significantly different expression in MPNSTs with PRC2 loss and those with wild-type PRC2 identified by RNA-seq. Clustering was based on the 479 most differentially expressed genes with a false discovery rate (FDR) of 8.0 (Supplementary Table 2). Samples are colored on the basis of PRC2 mutational status. The key shows mean log 2-transofrmed normalized fold change. (c) GO analysis of the differentially upregulated genes in MPNSTs with PRC2 loss in comparison to those with wild-type PRC2. (d) Gene expression by RNAseq of a representative group of developmental master regulators and imprinted genes in MPNSTs with PRC2 loss and those with wild-type PRC2. Error bars, s.e.m. (e–g) GSEA plots of the ranked list of differentially expressed genes in MPNSTs with PRC2 loss and those with wild-type PRC2 generated using three gene sets: PRC2 module (e) and H3K27me3 targets in brain (g) and neural precursor cells (f). NES, normalized enrichment score.
heterozygous loss of the CDKN2A locus in 73% (11/15) and 13% (2/15) of MPNSTs, respectively. We also observed nonsynonymous mutations and heterozygous loss of TP53 in 13% (2/15) and 20% (3/15) of MPNSTs. We did not identify other recurrent somatic alterations with relatively high frequency (Supplementary Table 3). Next, we used a targeted sequencing approach (IMPACT13; Supplementary Table 4) to characterize a validation cohort of formalin-fixed, paraffin-embedded samples consisting of 37 MPNSTs and 7 neurofibromas from 32 individuals (Fig. 1b and Supplementary Table 1). Combining the discovery and validation cohorts, we observed PRC2 mutations in 70% (19/27) of NF1-associated, 92% (12/13) of sporadic and 90% (8/9) of radiotherapy-associated MPNSTs (Fig. 1a–d). Genetic alterations in NF1 were identified in 82% (18/22) of sporadic and radiotherapy-associated MPNSTs. Genetic alterations in CDKN2A and TP53 were found in 81% (42/52) and 42% (22/52) of all MPNSTs, respectively (Fig. 1a,b). There was a significant cooccurrence of NF1, CDKN2A and PRC2 genetic alterations (Fleiss’ κ statistics, k = 0.21, P = 0.001)14, suggesting that these are three critical pathways in the pathogenesis of conventional MPNST. In the seven NF1-associated neurofibromas, we observed few PRC2 and CDKN2A alterations, suggesting that they might be associated with malignant progression to MPNST. None of the three epithelioid MPNSTs had genetic alterations in the three critical pathways, suggesting that they represent a distinct entity. Nature Genetics VOLUME 46 | NUMBER 11 | NOVEMBER 2014
To understand the effect of PRC2 loss, we performed gene expression analysis in 16 MPNSTs (Supplementary Table 1). In principalcomponent analysis (PCA), all samples with EED mutation or SUZ12 homozygous deletion clustered together and were separated from the other samples by the first principal component (PC1) (Figs. 1a and 2a). Of the five samples with heterozygous loss of SUZ12, three with structural variation of the remaining SUZ12 copy (2T, 7T and 13T) and one with an SUZ12 mutation encoding p.His213fs (12T) clustered with the group with PRC2 loss and one with an intact SUZ12 transcript (9T) clustered with the group with wild-type PRC2. These observations highlight the complexity in identifying structural variations that accurately determine PRC2 status in cases of heterozygous EED loss. To explore the transcriptional consequences of PRC2 loss, we generated a gene set composed of genes that were differentially expressed in MPNSTs with loss of PRC2 and those with wild-type PRC2 (Supplementary Table 5). Hierarchical clustering of these genes robustly separated the MPNSTs with loss of PRC2 and those with wild-type PRC2. The vast majority of differentially expressed genes (455/479; 95%) were upregulated in MPNSTs with PRC2 loss, consistent with the role of PRC2 in transcriptional repression (Fig. 2b). In Gene Ontology (GO) analysis, known PRC2-suppressed targets, including homeobox transcription factors and genes associated with development and morphogenesis, were highly enriched 1229
letters
npg
NF1 associated
Sporadic
Radiotherapy associated
b
PRC2 genetic status
lo ss H om lo s
H et
+
18
4
0
–
0
7
30
H3K27me3 IHC status
T W
H&E H3K27me3 IHC
d P = 0.035 P = 0.001
30
15 1
20 10
PRC2 positive PRC2 negative
P = 0.009
1
19
15
11
0
3 8 Neurofibroma
Plexiform neurofibroma
40 Number of cases
H&E
MPNST
Epithelioid
MPNST-neurofibroma interface
Radiotherapy associated
MPNST
Sporadic
Neurofibroma
NF1 associated
c
among the genes upregulated in MPNSTs with PRC2 loss (Fig. 2c and Supplementary Table 6). The expression of several prototypical PRC2-suppressed genes confirmed this difference (Fig. 2d). In Gene Set Enrichment Analysis (GSEA), the most significantly enriched gene sets upregulated in MPNSTs with PRC2 loss included the ‘PRC2 module’ defined by genes bound by PRC2 components in mouse embryonic stem (ES) cells15 (Fig. 2e and Supplementary Table 7) and H3K27me3 target genes in neural precursor cells 16 (Fig. 2f) and brain tissue17 (Fig. 2g). These data indicate that loss of PRC2 function results in distinct transcriptome changes, including activation of developmentally regulated master regulators and imprinted genes (for example, IGF2). We evaluated H3K27me3 levels by immunohistochemistry in formalin-fixed, paraffin-embedded samples of MPNST and neurofibroma. Whereas MPNSTs with wild-type PRC2 showed robust staining for H3K27me3, MPNSTs with PRC2 loss showed complete loss of H3K27me3 in tumor cells and preservation of H3K27me3 staining in stromal cells (Fig. 3a). Positive and negative immunostaining for H3K27me3 were highly concordant with a genetic status of wild-type PRC2 and homozygous PRC2 loss, respectively (Fig. 3b). However, heterozygous loss of PRC2 components was not predictive of H3K27me3 staining in immunohistochemistry (Fig. 3b). Among the samples with heterozygous loss for which associated RNA-seq data were available, clustering by transcriptional pattern matched H3K27me3 immunohistochemistry results. This finding suggests that DNA sequencing (exome or IMPACT) alone cannot predict PRC2 functional status in all MPNSTs and that H3K27me3 immunohistochemistry might be more accurate. All neurofibromas (7/7), which were wild type for PRC2 except for one sample with heterozygous loss of SUZ12, retained H3K27me3 immunostaining (Figs. 1b and 3c,d). In specimens that contained the interface of MPNSTs arising from preexisting benign plexiform neurofibromas, we observed a transition from robust H3K27me3 staining in the plexiform neurofibroma to a clear loss of this staining in the MPNST. These data suggest that PRC2 loss is involved in the malignant progression of benign plexiform neurofibroma into MPNST. Indeed, 56% (19/34) of the NF1-associated MPNSTs had lost H3K27me3 (Fig. 3d). Curiously, a significantly greater percentage 1230
NF1 associated
s
a
H3K27me3 IHC H3K27me3 IHC
© 2014 Nature America, Inc. All rights reserved.
Figure 3 H3K27me3 immunohistochemistry correlates with PRC2 genetic status, and H3K27me3 loss characterizes progression from neurofibroma to MPNST. (a) Representative hematoxylin and eosin (H&E) staining and H3K37me3 immunohistochemistry (IHC) images of NF1-associated, sporadic and radiotherapy-associated MPNSTs. Scale bars, 100 µm. (b) Correlation of PRC2 genetic status by whole-exome sequencing, RNA-seq and custom targeted sequencing and H3K27me3 immunohistochemistry status. (c) Representative hematoxylin and eosin staining and H3K27me3 immunohistochemistry images of neurofibroma, NF1-associated MPNST and the interface of plexiform neurofibroma transitioning into MPNST. Scale bars, 100 μm. (d) Distribution of PRC2 loss and PRC2 presence by H3K27me3 immunohistochemistry in NF1-associated, sporadic, radiotherapy-associated and epithelioid MPNSTs and in neurofibromas. Fisher’s exact test was used to calculate P values.
MPNST
(>90%) of sporadic and radiotherapy-associated MPNSTs had lost H3K27me3 staining (Fig. 3d), suggesting that the progression of disease and sequence of genetic inactivation of NF1, CDKN2A and PRC2 components might be different in MPNSTs that arise in distinct clinical settings. Unlike NF1-associated MPNSTs that universally arise from preexisting neurofibromas, sporadic and radiotherapyassociated MPNSTs rarely have identifiable preexisting benign nerve sheath tumors6. In one sporadic MPNST sample (16T), the presence of nonsynonymous mutations in both NF1 (encoding a p.Asp1237_ splice alteration) and EED (encoding p.Glu249fs) allowed us to use the prevalence of these mutations to infer the sequence of genetic events18. The largest subpopulation of cells (84%) contained the NF1 mutation, whereas a smaller subpopulation (57%) contained the EED mutation, suggesting that the NF1 mutation occurred first during progression of this sporadic MPNST (Supplementary Fig. 4). The sequence of NF1, PRC2 component and CDKN2A inactivation described here is largely correlative. The precise sequence of events will require experimental validation with sequential inactivation of each pathway in cell line and mouse models. To determine whether PRC2 loss is required for MPNST oncogenesis, we screened available human MPNST cell lines using immunoblotting for H3K27me3. We identified one MPNST cell line (ST88-14; derived from an NF1-associated MPNST) that had lost H3K27me3 marks. RNA-seq analysis showed that ST88-14 cells had lost expression of SUZ12, and immunoblotting confirmed loss of the SUZ12 protein (Fig. 4a,b and Supplementary Fig. 5). We next introduced Flag-HA–tagged wild-type SUZ12 (FH-SUZ12) or EED (FH-EED) into the ST88-14 cell line and into an MPNST cell line with wild-type PRC2 (MPNST724) that maintained H3K27me3 levels (Fig. 4a,b). FH-SUZ12 but not FH-EED restored H3K27me3 levels in ST88-14 cells and substantially decreased cell growth (Fig. 4a–c). In MPNST724 cells, there was a mild increase in H3K27me3 levels with the introduction of either FH-SUZ12 or FH-EED (Fig. 4a), but neither had any effect on cell growth (Fig. 4c). These data suggest that PRC2 loss contributes to oncogenesis at least in part by promoting cell proliferation and growth. VOLUME 46 | NUMBER 11 | NOVEMBER 2014 Nature Genetics
letters
0.5
0 15
0 30
20
20
10
20
10
10
5
10
0 30
0 30
0 15
0 30
20
20
10
20
10
10
5
10
0 80 60 40 20 0 15
0 80 60 40 20 0 40 30 20 10 0 80 60 40 20 0 10 8 6 4 2 0
0 100 80 60 40 20 0 20 15 10 5 0 10 8 6 4 2 0 10 8 6 4 2 0
0 50 40 30 20 10 0 30
4
4
72 ST
88
PN M
ST
72
-1
4
4 ST
ST
0 25 20 15 10 5 0 10 8 6 4 2 0
-1
-1 88
PN M
ST
10
88
0 20 15 10 5 0 10 8 6 4 2 0
20
PN
5
ST
SUZ12 ChIP IgG ChIP
DAPI
10
M
FH-SUZ12
H3K27me3
FH-EED
4
ST88-14 Parental
72
25
ST
20
-1
10 15 Days
88
5
PN
0
4
0.5
0 30
Merge
© 2014 Nature America, Inc. All rights reserved.
npg
0.5
0 30
ST
MPNST724 FH-SUZ12
0.5
M
1.0 0
FH-EED
mRNA
2.0
Total H3
Parental
ST88-14
3.0
H3K27ac
b
1.0
4
H3K27me3
25
TLX1
1.5
1.0
72
Actin
* **
20
PAX2
1.5
1.0
4
EED
* **
10 15 Days
1.5
FH-SUZ12
1.0
0 5
IGF2
EZH2 ChIP
SUZ12
1.5
1.0 0
Vector FOXN4
H3K27me3 ChIP
EZH1
2.0
H3K4me3 ChIP
EZH2
d
MPNST724 Parental FH-EED FH-SUZ12
3.0
H3K27ac ChIP
NF1
c Normalized fluorescence units
Parental
FH-SUZ12
ST88-14
FH-SUZ12
FH-EED
Parental
MPNST724
FH-EED
a
Figure 4 PRC2 loss promotes cell proliferation and growth in MPNST with PRC2 loss. (a) Immunoblots demonstrating SUZ12 loss and corresponding loss of H3K27me3 in ST88-14 cells, a cell line derived from NF1-associated human MPNST, in comparison to a cell line derived from sporadic human MPNST, MPNST724, with intact PRC2 and retained H3K27me3 levels. Introduction of exogenous Flag-HA–tagged SUZ12 (FH-SUZ12) but not FlagHA–tagged EED (FH-EED) in ST88-14 cells restores H3K27me3 protein levels. *, exogenous FH-SUZ12 or FH-EED; **, endogenous SUZ12 or EED. (b) Immunofluorescence of H3K27me3 demonstrating the restoration of H3K27me3 levels at the cellular level by introducing FH-SUZ12 in the ST88-14 MPNST cell line with SUZ12 loss. Scale bars, 100 µm. DAPI, 4′,6-diamidino-2-phenylindole. (c) Representative growth curves for the MPNST724 and ST88-14 cell lines demonstrating that the introduction of FH-SUZ12 but not FH-EED in SUZ12-deficient ST88-14 cells leads to growth retardation, whereas it had no effect in MPNST724 cells. Similar results have been obtained in at least three independent experiments. (d) ST88-14 and MPNST724 cells were infected with vector control virus or virus expressing FH-SUZ12. Plots of expression determined by quantitative RT-PCR (expressed as 2 −∆∆Ct) and promoter localization determined by chromatin immunoprecipitation (ChIP) and quantitative PCR (expressed as percent input) of SUZ12, EZH2, H3K27me3, H3K4me3, H3K27ac and IgG control for the FOXN4, IGF2, PAX2 and TLX1 genes are shown. Error bars, s.e.m. n = 3 technical replicates.
We next examined the transcriptional and chromatin changes in ST88-14 and MPNST724 cells after the introduction of FH-SUZ12, focusing on several known PRC2-regulated genes (FOXN4, IGF2, PAX2 and TLX1) that are significantly upregulated in MPNST samples with PRC2 loss in comparison to MPNST samples with wild-type PRC2 (Fig. 2c). At baseline, ST88-14 cells exhibited increased expression of FOXN4, IGF2, PAX2 and TLX1 accompanied by loss of PRC2 components (SUZ12 and EZH2) and the PRC2-repressive mark (H3K27me3) and reciprocal gain of the activation marks trimethylation at lysine 4 of histone H3 (H3K4me3) and acetylation of lysine 27 of histone H3 (H3K27ac) at their promoters (Fig. 4d and Supplementary Fig. 6). After the introduction of FH-SUZ12 in ST88-14 cells, FH-SUZ12 localized to the promoters of these genes. This localization was accompanied by increased levels of EZH2 and H3K27me3 and decreased levels of H3K4me3 and H3K27ac at the promoter regions, as well as by decreased transcript levels of these PRC2 target genes. These data suggest that PRC2 loss has a direct impact on transcriptional regulation, and introduction of the missing PRC2 component has the ability to at least partially restore PRC2 function. MPNSTs often exhibit divergent differentiation, including rhabdomyoblastic, glandular, squamous and neuroendocrine elements6. Our study identified a high frequency of loss-of-function genetic Nature Genetics VOLUME 46 | NUMBER 11 | NOVEMBER 2014
alterations in NF1, CDKN2A and PRC2 components (EED or SUZ12), demonstrating that MPNSTs share common molecular pathogenic pathways despite clinical and histological diversity. PRC2 loss activates multiple developmentally suppressed pathways, which might explain the frequent observation of divergent differentiation in MPNST. The high frequency of PRC2 loss suggests that PRC2 mutational status and, more specifically, H3K27me3 immunohistochemistry could be used as biomarkers for the more acute diagnosis of MPNST. PRC2 was initially thought to be oncogenic: PRC2 components have higher expression in dividing cells and are important in maintaining stemness. EZH2 is overexpressed in a variety of cancers19,20, and activating EZH2 mutations are found in a subset of lymphomas21. Paradoxically, recent work suggests that PRC2 can be tumor suppressive in distinct contexts, with loss-of-function genetic alterations found in up to 25% of myeloid disorders and T cell acute lymphoblastic leukemia (ALL)22–24 and 42% of early T cell ALL25. Notably, the majority of the loss-of-function alterations are found in EZH2 (refs. 22–25). Cellular studies and mouse models have shown that Ezh1 can maintain suppression of Polycomb-regulated genes in the setting of Ezh2 loss, and combined Ezh1 and Ezh2 loss or individual loss of Eed or Suz12 derepresses Polycomb-regulated genes and causes Cdkn2a-mediated growth arrest26–29. These findings suggest that MPNST is unique in 1231
letters that complete loss of PRC2 function is important for tumorigenesis, and loss of CDKN2A might be a critical cooperative event in addition to NF1 loss. URLs. GENE-E, https://www.broadinstitute.org/cancer/software/ GENE-E/; Picard, http://picard.sourceforge.net/; aroma.affymetrix, http://aroma-project.org/; GSEA, http://www.broadinstitute.org/gsea/; DAVID, http://david.abcc.ncifcrf.gov/; STAR, https://code.google. com/p/rna-star/; DESeq, http://bioconductor.org/packages/release/ bioc/html/DESeq.html; HTSeq, https://pypi.python.org/pypi/HTSeq; IGV, http://www.broadinstitute.org/igv/IGV. Methods Methods and any associated references are available in the online version of the paper.
npg
© 2014 Nature America, Inc. All rights reserved.
Accession codes. Exome sequencing, RNA-seq and DNA copy number data have been deposited in the database of Genotypes and Phenotypes (dbGaP) under accession phs000792.v1.p1. Note: Any Supplementary Information and Source Data files are available in the online version of the paper. Acknowledgments Next-generation sequencing and SNP6.0 analysis were performed at the Memorial Sloan Kettering Cancer Center (MSKCC) Genomics Core Facility, Center for Molecular Oncology (CMO). The authors thank members of the cBioPortal team (J. Gao, B. Gross, N. Schultz and C. Sander) for assistance with data analysis and visualization. The authors also thank T. Chan (MSKCC) and HOPP informatics for assistance with data analysis. This work was supported in part by a grant from the US National Institutes of Health (NIH) to W.L. (NCI U24-CA143840), the Harry J. Lloyd Trust–Translational Research grant to T.W., the Charles H. Revson Senior Fellowship to T.W., Jubilaeumsfonds of the Oesterreichische Nationalbank to T.W., grants from the US NIH to P.C. (P50CA140146, Career Development Award and Developmental Research Project.; DP2CA174499; K08CA151660), Y.C. (K08CA140946), L.-X.Q. (P50CA140146), C.R.A. (P50CA140146) and S.S. (P50CA140146), an award from the Sidney Kimmel Foundation to P.C. (Kimmel Scholar Award) and funding from the Cycle for Survival Fund to P.C. AUTHOR CONTRIBUTIONS Project planning and experimental design: P.C., C.R.A., Y.C., T.W., W.L., S.T. and L.R. Sample collection, clinical database and cell lines: S.T., S.S., C.R.A., C.N.P.G., J.A.F. and W.D.T. Pathology review: C.R.A. and C.N.P.G. Preparation of DNA, RNA and next-generation sequencing libraries: T.W., K.H.H., S.T. and A.V. Sequence data analysis: W.L., Y.C., S.T., P.C., M.F.B., M.L., D.Z., Y.L. and A.S. Immunohistochemistry: T.W. Protein blots, immunofluorescence, growth curves and all cellular assays: L.R. and S.Z. Generation of the expression vectors: Z.C. and L.R. Biostatistics: L.-X.Q. Manuscript writing: P.C., Y.C., W.L., T.W. and L.R. The final manuscript was reviewed by all authors. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html.
1232
1. Brennan, M.F., Antonescu, C.R. & Maki, R.G. in Management of Soft Tissue Sarcoma 149–160 (Springer, New York, 2013). 2. LaFemina, J. et al. Oncologic outcomes of sporadic, neurofibromatosis-associated, and radiation-induced malignant peripheral nerve sheath tumors. Ann. Surg. Oncol. 20, 66–72 (2013). 3. Stucky, C.C. et al. Malignant peripheral nerve sheath tumors (MPNST): the Mayo Clinic experience. Ann. Surg. Oncol. 19, 878–885 (2012). 4. Eilber, F.C. et al. Validation of the postoperative nomogram for 12-year sarcomaspecific mortality. Cancer 101, 2270–2275 (2004). 5. Rodriguez, F.J. Peripheral nerve sheath tumors: the elegant chapter in surgical neuropathology. Acta Neuropathol. 123, 293–294 (2012). 6. Antonescu, C.R., Scheithauer, B.W. & Woodruff, J.M. in Tumors of the Peripheral Nervous System 553 (The American Registry of Pathology, Silver Spring, Maryland, 2013). 7. Taylor, B.S. et al. Advances in sarcoma genomics and new therapeutic targets. Nat. Rev. Cancer 11, 541–557 (2011). 8. Margueron, R. & Reinberg, D. The Polycomb complex PRC2 and its mark in life. Nature 469, 343–349 (2011). 9. Kourea, H.P., Orlow, I., Scheithauer, B.W., Cordon-Cardo, C. & Woodruff, J.M. Deletions of the INK4A gene occur in malignant peripheral nerve sheath tumors but not in neurofibromas. Am. J. Pathol. 155, 1855–1860 (1999). 10. Menon, A.G. et al. Chromosome 17p deletions and p53 gene mutations associated with the formation of malignant neurofibrosarcomas in von Recklinghausen neurofibromatosis. Proc. Natl. Acad. Sci. USA 87, 5435–5439 (1990). 11. Nielsen, G.P. et al. Malignant transformation of neurofibromas in neurofibromatosis 1 is associated with CDKN2A/p16 inactivation. Am. J. Pathol. 155, 1879–1884 (1999). 12. Perrone, F. et al. p15INK4b, p14ARF, and p16INK4a inactivation in sporadic and neurofibromatosis type 1–related malignant peripheral nerve sheath tumors. Clin. Cancer Res. 9, 4132–4138 (2003). 13. Won, H.H., Scott, S.N., Brannon, A.R., Shah, R.H. & Berger, M.F. Detecting somatic genetic alterations in tumor specimens by exon capture and massively parallel sequencing. J. Vis. Exp. (80), e50710 (2013). 14. Fleiss, J.L. Measuring nominal scale agreement among many raters. Psychol. Bull. 76, 378–382 (1971). 15. Kim, J. et al. A Myc network accounts for similarities between embryonic stem and cancer cell transcription programs. Cell 143, 313–324 (2010). 16. Mikkelsen, T.S. et al. Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553–560 (2007). 17. Meissner, A. et al. Genome-scale DNA methylation maps of pluripotent and differentiated cells. Nature 454, 766–770 (2008). 18. Andor, N., Harness, J.V., Muller, S., Mewes, H.W. & Petritsch, C. EXPANDS: expanding ploidy and allele frequency on nested subpopulations. Bioinformatics 30, 50–60 (2014). 19. Varambally, S. et al. The polycomb group protein EZH2 is involved in progression of prostate cancer. Nature 419, 624–629 (2002). 20. Simon, J.A. & Lange, C.A. Roles of the EZH2 histone methyltransferase in cancer epigenetics. Mutat. Res. 647, 21–29 (2008). 21. Morin, R.D. et al. Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat. Genet. 42, 181–185 (2010). 22. Ernst, T. et al. Inactivating mutations of the histone methyltransferase gene EZH2 in myeloid disorders. Nat. Genet. 42, 722–726 (2010). 23. Nikoloski, G. et al. Somatic mutations of the histone methyltransferase gene EZH2 in myelodysplastic syndromes. Nat. Genet. 42, 665–667 (2010). 24. Ntziachristos, P. et al. Genetic inactivation of the polycomb repressive complex 2 in T cell acute lymphoblastic leukemia. Nat. Med. 18, 298–301 (2012). 25. Zhang, J. et al. The genetic basis of early T-cell precursor acute lymphoblastic leukaemia. Nature 481, 157–163 (2012). 26. Ezhkova, E. et al. EZH1 and EZH2 cogovern histone H3K27 trimethylation and are essential for hair follicle homeostasis and wound repair. Genes Dev. 25, 485–498 (2011). 27. Shen, X. et al. EZH1 mediates methylation on histone H3 lysine 27 and complements EZH2 in maintaining stem cell identity and executing pluripotency. Mol. Cell 32, 491–502 (2008). 28. Xie, H. et al. Polycomb repressive complex 2 regulates normal hematopoietic stem cell function in a developmental-stage-specific manner. Cell Stem Cell 14, 68–80 (2014). 29. Margueron, R. et al. Ezh1 and Ezh2 maintain repressive chromatin through different mechanisms. Mol. Cell 32, 503–518 (2008).
VOLUME 46 | NUMBER 11 | NOVEMBER 2014 Nature Genetics
npg
© 2014 Nature America, Inc. All rights reserved.
ONLINE METHODS
Human tumor tissue collection. Biospecimens were collected during surgical resection from patients with pathologic diagnosis of MPNSTs or neurofibromas. Material was collected under institutional review board (IRB)-approved protocols (IRB 06-107) at MSKCC. All patients provided informed consent. Pathologic diagnosis was confirmed by at least two pathologists using diagnostic formalin-fixed and paraffin-embedded sections to select cases with an estimated tumor content of >70%. The majority of the tumors were collected from different patients. In some cases, more than one tumor was selected from the same patient, and these tumors were resected from distinct anatomical locations in different surgeries. Sample annotation is shown in Supplementary Table 1. For the discovery cohort, the goal was to obtain normal DNA and tumor DNA for whole-exome sequencing and SNP6.0 array analysis and to obtain tumor RNA for RNA-seq analysis. A total of 15 fresh-frozen paired MPNST tumor–normal samples were identified. Tumor and adjacent normal tissue specimens were embedded in optimal cutting temperature (OCT) medium, and a histological section was obtained for review. Cryomolds of both tumor and normal tissues were macrodissected to minimize contamination before RNA and DNA preparation. Sample processing was designed to secure samples and minimize the availability of identifying information. Specimens with insufficient tissue amount or severely degraded nucleic acids were excluded. The IMPACT assay is a hybridization capture, next-generation sequencing platform amenable to DNA from both fresh-frozen and formalin-fixed, paraffin-embedded samples for targeted sequencing. The panel includes NF1, SUZ12, EED, CDKN2A and TP53 (ref. 13). We validated the somatic mutational findings from the discovery cohort by performing IMPACT assays on the same DNA isolated from tumor tissue. In addition, we performed IMPACT assays on DNA from formalin-fixed, paraffin-embedded samples derived from a second cohort of 37 MPNSTs and 7 neurofibromas from individuals with NF1 who were diagnosed with concurrent MPNST (Supplementary Table 1). Sample preparation and quality control. RNA was extracted from tumor and normal tissues using a modification of the protocol for the DNA/RNA AllPrep kit (Qiagen). DNA from fresh-frozen tissues was extracted from tumor and normal tissue specimens using the DNeasy Blood and Tissue kit (Qiagen). DNA from formalin-fixed, paraffin-embedded samples was isolated using the QIAamp DNA FFPE Tissue kit (Qiagen). DNA from each specimen was initially quantified using the NanoDrop UV spectrophotometer and was further quantified with the Bioanalyzer assay (Agilent Technologies). RNA sequencing and analysis. The isolated RNA was processed using the TruSeq RNA Sample Prep kit (15026495, Illumina) according to the manufacturer’s protocol. Briefly, RNA was polyA selected and reverse transcribed, and the cDNA obtained underwent end repair, A-tailing, ligation of the indexes and adaptors, and PCR enrichment. Libraries were sequenced on the Illumina HiSeq 2500 platform with 51-bp paired-end reads to obtain a minimum yield of 40 million reads per sample. Sequence data were processed and mapped to the human reference genome (hg19) using STAR (v2.3)30. Gene expression levels were quantified with htseq-count31 and normalized using DESeq32. Variance in expression levels was calculated for all genes across samples, and the 75th percentile was set as a cutoff. PCA was performed on the set of genes with variance greater than this cutoff. We used ANOVA to define differentially expressed genes between samples with PRC2 loss and wild-type PRC2. Genes that showed a >8-fold difference in expression and a corrected FDR of
PRC2 is recurrently inactivated through EED or SUZ12 loss in malignant peripheral nerve sheath tumors
npg
© 2014 Nature America, Inc. All rights reserved.
William Lee1,2,17, Sewit Teckie2,3,17, Thomas Wiesner3,17, Leili Ran3,17, Carlos N Prieto Granada4, Mingyan Lin5, Sinan Zhu3, Zhen Cao3, Yupu Liang3, Andrea Sboner6–8, William D Tap9,10, Jonathan A Fletcher11, Kety H Huberman12, Li-Xuan Qin13, Agnes Viale12, Samuel Singer14, Deyou Zheng5,15,16, Michael F Berger3,4, Yu Chen3,9,10, Cristina R Antonescu4 & Ping Chi3,9,10 Malignant peripheral nerve sheath tumors (MPNSTs) represent a group of highly aggressive soft-tissue sarcomas that may occur sporadically, in association with neurofibromatosis type I (NF1 associated) or after radiotherapy1–3. Using comprehensive genomic approaches, we identified lossof-function somatic alterations of the Polycomb repressive complex 2 (PRC2) components (EED or SUZ12) in 92% of sporadic, 70% of NF1-associated and 90% of radiotherapyassociated MPNSTs. MPNSTs with PRC2 loss showed complete loss of trimethylation at lysine 27 of histone H3 (H3K27me3) and aberrant transcriptional activation of multiple PRC2-repressed homeobox master regulators and their regulated developmental pathways. Introduction of the lost PRC2 component in a PRC2-deficient MPNST cell line restored H3K27me3 levels and decreased cell growth. Additionally, we identified frequent somatic alterations of CDKN2A (81% of all MPNSTs) and NF1 (72% of non-NF1-associated MPNSTs), both of which significantly co-occur with PRC2 alterations. The highly recurrent and specific inactivation of PRC2 components, NF1 and CDKN2A highlights their critical and potentially cooperative roles in MPNST pathogenesis. MPNSTs arise from peripheral nerves and associated cellular components and represent a highly aggressive subtype of soft-tissue sarcoma1. MPNSTs metastasize early and are often resistant to radiotherapy and chemotherapy. Conventional MPNSTs present in three distinct clinical settings: sporadically, in association with neurofibromatosis type I (NF1 associated) or in association with previous radiotherapy (radiotherapy associated), accounting for approximately
45%, 45% and 10% of cases, respectively2,4,5. Histologically, MPNSTs are characterized by intersecting fascicles of monotonous spindle cells with hyperchromatic nuclei and high mitotic counts with focal areas of necrosis, but accurate diagnosis remains challenging because of the lack of specific immunohistochemical and molecular biomarkers5,6. Among individuals with NF1, loss of the non-mutant allele of NF1 is thought to be the key driver in benign NF1-associated neurofibroma7. Little is known of the genetic alterations that mediate progression from neurofibroma to MPNST in individuals with NF1 or of the molecular pathogenesis of sporadic and radiotherapyassociated MPNST. To investigate the molecular basis of MPNST, we performed wholeexome sequencing, DNA copy number and loss-of-heterozygosity (LOH) profiling, and whole-transcriptome sequencing (RNA-seq) on a discovery cohort consisting of paired normal-tumor tissues for 15 MPNSTs from 12 individuals (6 NF1 associated, 4 sporadic, 4 radiotherapy associated and 1 epithelioid) (Supplementary Tables 1 and 2). Epithelioid MPNST is a rare histological variant of MPNST that is composed of exclusively epithelioid malignant cells with diffuse immunoreactivity for the S100 protein and is not associated with NF1 (ref. 6). We identified four frameshift mutations and one splice-site mutation in EED (Fig. 1a,c and Supplementary Fig. 1). RNA-seq validated aberrant EED splicing in the sample with the mutated splice site (Supplementary Fig. 2a). All five samples showed LOH at the EED locus, three samples (11T, 12T and 14T) by heterozygous deletion of the normal allele (Supplementary Fig. 1b) and two samples (15T and 16T) by copy-neutral LOH (Supplementary Fig. 2b). These data suggest that samples with EED mutation have complete loss of EED function.
1Computational
Biology Program, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 2Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 3Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 4Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 5Department of Genetics, Albert Einstein College of Medicine, Bronx, New York, USA. 6Department of Pathology and Laboratory Medicine, Weill Cornell Medical College, New York, New York, USA. 7Institute for Computational Biomedicine, Weill Cornell Medical College, New York, New York, USA. 8Institute for Precision Medicine, Weill Cornell Medical College, New York, New York, USA. 9Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 10Department of Medicine, Weill Cornell Medical College, New York, New York, USA. 11Department of Pathology, Brigham and Women’s Hospital, Boston, Massachusetts, USA. 12Genomics Core Laboratory, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 13Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 14Department of Surgery, Memorial Sloan Kettering Cancer Center, New York, New York, USA. 15Department of Neurology, Albert Einstein College of Medicine, Bronx, New York, USA. 16Department of Neuroscience, Albert Einstein College of Medicine, Bronx, New York, USA. 17These authors contributed equally to this work. Correspondence should be addressed to P.C. ([email protected]). Received 21 April; accepted 26 August; published online 21 September 2014; doi:10.1038/ng.3095
Nature Genetics VOLUME 46 | NUMBER 11 | NOVEMBER 2014
1227
letters
npg
© 2014 Nature America, Inc. All rights reserved.
b
NF1
CDKN2A
Germline Germline Germline Germline Germline Germline D1237_splice D173Y, hom deletion Hom deletion WT Hom deletion Hom deletion K99*, het loss K99*, het loss WT
Sample ID MPNST 10 MPNST 22 MPNST 28 MPNST 51 MPNST 04.2 MPNST 14 MPNST 46.1 MPNST 46.3 MPNST 43.1 MPNST 13.1 MPNST 19 MPNST 06.1 MPNST 06.2 MPNST 04.1 MPNST 05.1 MPNST 12.1 5T MPNST 21.1 MPNST 03 MPNST 25 MPNST 47.1 MPNST 23 MPNST 18 MPNST 01 MPNST 27 MPNST 29 MPNST 02.2 MPNST 02.1 MPNST 48 MPNST 02.3 MPNST 45 MPNST 24 MPNST 52 MPNST 44 MPNST 53 MPNST 26 MPNST 50 MPNST 12.2 MPNST 47.2 MPNST 21.2 MPNST 46.2 MPNST 13.2 MPNST 43.2 MPNST 05.2
NF1 CDKN2A F1740fs, L43P Het loss I1979fs, het loss WT L1590*, het loss Hom deletion Q1188*, C2278* Het loss R1534*, het loss Hom deletion R1534* Het loss R1769* Het loss R1769* Hom deletion R304*, Y1989* Het loss R440*, het loss WT T165fs, I1482T WT V1957fs Het loss V1957fs Hom deletion W837*, R1534* WT V1231fs, Y2285* Het loss F1593del Het loss T2206R, het loss Het loss W1559*, het loss Hom deletion T676fs, het loss Hom deletion Q2434*, hom deletion Het loss Germline WT Hom deletion Hom deletion Hom deletion Het loss L1622fs, het loss Het loss K286fs, het loss Hom deletion Het loss Hom deletion Het loss V51_splice WT V51_splice WT WT Het loss V51_splice K195fs, hom deletion Hom deletion Het loss Hom deletion Q2234* Hom deletion V567fs, het loss Hom deletion WT Hom deletion WT WT WT WT F1593del Het loss I2057fs Het loss D2662fs, het loss WT R1769* WT R440*, Q1336* Het loss R681*, Y1989* WT R1534*, Y2285* WT
NF1 associated
Sporadic
Hom deletion Hom deletion Hom deletion Hom deletion Het loss Hom deletion Het loss Hom deletion Hom deletion Hom deletion Hom deletion WT Hom deletion Hom deletion WT
TP53 WT WT WT Het loss WT Het loss WT R273C WT WT Het loss T118fs WT WT WT
EED F289fs H213fs H213fs WT WT R287_splice E249fs WT WT WT WT WT WT WT WT
SUZ12 WT WT WT Het loss WT WT WT Hom deletion Hom deletion WT SV + het loss SV + het loss SV + het loss SV + het loss WT
TP53 EED SUZ12 WT WT WT C135fs WT H471R, het loss Het loss V121D, het loss WT WT A114P WT WT H213fs Het loss Het loss WT Het loss Het loss WT G291fs, het loss Y220H, het loss WT WT P190L, het loss WT Het loss WT G167R, het loss Het loss Q144fs Het loss WT V173A, het loss WT WT V173A, het loss WT WT Het loss WT Het loss R158H, het loss WT WT WT WT WT WT WT Hom deletion WT WT Q504fs, het loss WT WT Het loss Hom deletion WT Hom deletion WT WT WT WT WT Hom deletion WT H92fs, het loss Hom deletion WT WT Hom deletion Het loss Het loss Het loss N131I WT N651fs, het loss V157fs, het loss Hom deletion WT WT D317fs WT E285K, het loss Hom deletion WT V157fs Hom deletion WT Het loss Hom deletion Y565fs, het loss WT Hom deletion WT WT WT WT WT WT L564fs, het loss WT Hom deletion WT WT WT WT WT WT WT WT WT WT WT WT WT WT WT Het loss WT WT WT WT WT WT Het loss WT WT WT WT WT
Radiotherapy associated
Epithelioid
c PRC2 core components H3K27me2 and H3K27me3
EZH1-EZH2 EED
SUZ12 RbAp46/48
EED 5
Number of mutations
Sample ID 15T 11T 12T 9T 19T 14T 16T 8T 10T 4T 2T 13T 7T 18T 20T
H92fs
0
A114P
V121D
H213fs G167R E249fs R287_splice F289fs
WD40 0
100
D317fs
WD40
200
300
400
441
SUZ12 5
Number of mutations
a
G291fs 0
H471R
Q504fs L564fs Y565fs
N651fs
VEFS-Box 0
200
400
d
600
739
NF1 All WT 3
1
CDKN2A 8 1
PRC2 5 1
31 P = 0.001 2
Neurofibroma
Figure 1 The most frequent genetic alterations in MPNST (NF1 associated, sporadic, radiotherapy associated and epithelioid) and neurofibroma. (a,b) Nonsynonymous single-nucleotide variants (SNVs) and copy number variants (CNVs) in 15 MPNSTs with matched normal samples identified by whole-exome sequencing, SNP6.0 array analysis and RNA-seq (a) and in 37 MPNSTs and 7 neurofibromas identified by targeted sequencing (IMPACT) (b). Brackets indicate two different tumor samples from the same individual. WT, wild type; hom, homozygous; het, heterozygous; SV, structural variation (light shading indicates heterozygous loss and dark shading indicates homozygous loss). (c) Schematics of the nonsynonymous SNVs observed in the PRC2 core components EED and SUZ12 in 15 MPNST samples analyzed by whole-exome sequencing and 37 MPNST samples analyzed by custom IMPACT. (d) Schematic of the overlap of mutations affecting NF1, PRC2 components (EED or SUZ12) and CDKN2A in all MPNSTs (NF1 associated, sporadic, radiotherapy associated and epithelioid). With Fleiss’ κ statistics, three-way comparison of NF1, CDKN2A and PRC2 component (EED or SUZ12) genetic alteration suggested that they significantly co-occur: κ = 0.21, P = 0.001.
We further identified two homozygous and five heterozygous deletions of SUZ12 (Fig. 1a,c and Supplementary Figs. 1 and 3a). We examined RNA-seq profiles for the SUZ12 transcript among the five samples with heterozygous loss of the gene. Two samples, 9T and 12T (with EED mutations encoding p.His213fs), expressed full-length SUZ12 transcript, indicating that the remaining SUZ12 copy is intact (Supplementary Fig. 1b and data not shown). Remarkably, the other three samples displayed structural alterations of the SUZ12 transcript, starting at exon 6, exon 10 and exon 4 in samples 2T, 7T and 13T, respectively (Supplementary Fig. 3b–d). These structural changes are likely due to local genomic rearrangements of the remaining copy, which were not identified by standard whole-exome sequencing analysis. Indeed, for 7T and 18T, derived from two tumors from the same individual, there was a DNA break in exon 10 identified through manual examination of whole-exome sequencing data (Supplementary Fig. 3c). We designated these cases as having structural variation and 1228
heterozygous loss at the SUZ12 locus, and, intriguingly, all constituted radiotherapy-associated MPNSTs (Fig. 1a). EED and SUZ12 are the core components of PRC2 and, together with EZH1 and EZH2, establish and maintain the di- and trimethylation of lysine 27 of histone H3 (H3K27me2 and H3K27me3)8. EED and SUZ12 genetic alterations were mutually exclusive and were collectively found in 80% (12/15) of all MPNSTs (Fig. 1a,c). We did not observe any genetic alterations in other PRC2 core components, including EZH1 and EZH2 (Supplementary Table 3). We found recurrent nonsense mutations and homozygous deletions in NF1 in 87.5% (7/8) of sporadic and radiotherapy-associated MPNSTs (Fig. 1a and Supplementary Fig. 1). These data, in combination with the germline mutations in NF1 in NF1-associated MPNSTs, suggest that NF1 is a uniquely important tumor suppressor in MPNST. Alterations of the CDKN2A locus and of TP53 have been reported in MPNST9–12. We observed homozygous deletion and VOLUME 46 | NUMBER 11 | NOVEMBER 2014 Nature Genetics
letters b
PRC2 WT SUZ12 loss EED loss
PRC2 WT SUZ12 loss EED loss
c 0
–4
4
19
S8 S2 S7 S5 SS13 S16 S11 S10 S12 S14 S4 S9 S19 S22 S21 S20
100 ●
50
PC2
14
11 ● 2 ●●● 5 7 ● 16
0
9
12
●●● 13 ● 10● 8 ●
4
–50
●
●
●
● 20 21 22
–100
npg
50
PRC2 WT (n = 6) SUZ12 loss (n = 6) EED loss (n = 4)
15 10 5 0 –2
100
e
f
Bernstein neural progenitor cell H3K27me3
Orkin PRC2 module 0.6
Enrichment score
20
0 PC1
EM FO X2 XN H 4 H MX O 1 X H A1 O 3 XD 11 IG F O 2 TX PA 2 X PA 2 X PI 7 TX TL 1 X1
Normalized expression (log2)
© 2014 Nature America, Inc. All rights reserved.
d
–50
Enrichment score
–100
NES = 3.24 FDR = 0 P=0
0.4 0.2 0
PRC2 loss
PRC2 WT
DNA-binding region:homeobox Homeobox IPR017970:homeobox, conserved site IPR001356:homeobox Developmental protein IPR012287:Homeodomain-related SM00389:HOX GO:0030182:neuron differentiation GO:0043565:sequence-specific DNA binding GO:0003700:transcription factor activity GO:0030528:transcription regulator activity GO:0048598:embryonic morphogenesis GO:0048736:appendage development GO:0060173:limb development GO:0035108:limb morphogenesis GO:0035107:appendage morphogenesis GO:0001501:skeletal system development GO:0048666:neuron development GO:0030900:forebrain development GO:0051216:cartilage development glycoprotein GO:0007389:pattern specification process GO:0042471:ear morphogenesis Cell membrane DNA binding GO:0048562:embryonic organ morphogenesis
0.6
NES = 3.26 FDR = 0 P=0
0.4 0.2 0
PRC2 loss
PRC2 WT
0 5 10 15 20 25 –log10 FDR (Benjamini)
g Enrichment score
a
Lander brain H3K27me3 0.6
NES = 3.34 FDR = 0 P=0
0.4 0.2 0
PRC2 loss
PRC2 WT
Figure 2 MPNSTs with PRC2 loss exhibit distinct gene expression pattern from MPNSTs with wild-type PRC2, signifying activation of developmentally suppressed pathways. (a) PCA of the MPNST whole-transcriptome data showed that the samples with wild-type PRC2 and those with PRC2 component (SUZ12 and EED) loss segregate by principal component 1 (PC1). Each sample is colored on the basis of its corresponding PRC2 mutational status derived from whole-exome sequencing data, except for samples 21 and 22, which are colored on the basis of manual examination of the RNA-seq data for mutations in PRC2 components. (b) Heat map of genes with significantly different expression in MPNSTs with PRC2 loss and those with wild-type PRC2 identified by RNA-seq. Clustering was based on the 479 most differentially expressed genes with a false discovery rate (FDR) of 8.0 (Supplementary Table 2). Samples are colored on the basis of PRC2 mutational status. The key shows mean log 2-transofrmed normalized fold change. (c) GO analysis of the differentially upregulated genes in MPNSTs with PRC2 loss in comparison to those with wild-type PRC2. (d) Gene expression by RNAseq of a representative group of developmental master regulators and imprinted genes in MPNSTs with PRC2 loss and those with wild-type PRC2. Error bars, s.e.m. (e–g) GSEA plots of the ranked list of differentially expressed genes in MPNSTs with PRC2 loss and those with wild-type PRC2 generated using three gene sets: PRC2 module (e) and H3K27me3 targets in brain (g) and neural precursor cells (f). NES, normalized enrichment score.
heterozygous loss of the CDKN2A locus in 73% (11/15) and 13% (2/15) of MPNSTs, respectively. We also observed nonsynonymous mutations and heterozygous loss of TP53 in 13% (2/15) and 20% (3/15) of MPNSTs. We did not identify other recurrent somatic alterations with relatively high frequency (Supplementary Table 3). Next, we used a targeted sequencing approach (IMPACT13; Supplementary Table 4) to characterize a validation cohort of formalin-fixed, paraffin-embedded samples consisting of 37 MPNSTs and 7 neurofibromas from 32 individuals (Fig. 1b and Supplementary Table 1). Combining the discovery and validation cohorts, we observed PRC2 mutations in 70% (19/27) of NF1-associated, 92% (12/13) of sporadic and 90% (8/9) of radiotherapy-associated MPNSTs (Fig. 1a–d). Genetic alterations in NF1 were identified in 82% (18/22) of sporadic and radiotherapy-associated MPNSTs. Genetic alterations in CDKN2A and TP53 were found in 81% (42/52) and 42% (22/52) of all MPNSTs, respectively (Fig. 1a,b). There was a significant cooccurrence of NF1, CDKN2A and PRC2 genetic alterations (Fleiss’ κ statistics, k = 0.21, P = 0.001)14, suggesting that these are three critical pathways in the pathogenesis of conventional MPNST. In the seven NF1-associated neurofibromas, we observed few PRC2 and CDKN2A alterations, suggesting that they might be associated with malignant progression to MPNST. None of the three epithelioid MPNSTs had genetic alterations in the three critical pathways, suggesting that they represent a distinct entity. Nature Genetics VOLUME 46 | NUMBER 11 | NOVEMBER 2014
To understand the effect of PRC2 loss, we performed gene expression analysis in 16 MPNSTs (Supplementary Table 1). In principalcomponent analysis (PCA), all samples with EED mutation or SUZ12 homozygous deletion clustered together and were separated from the other samples by the first principal component (PC1) (Figs. 1a and 2a). Of the five samples with heterozygous loss of SUZ12, three with structural variation of the remaining SUZ12 copy (2T, 7T and 13T) and one with an SUZ12 mutation encoding p.His213fs (12T) clustered with the group with PRC2 loss and one with an intact SUZ12 transcript (9T) clustered with the group with wild-type PRC2. These observations highlight the complexity in identifying structural variations that accurately determine PRC2 status in cases of heterozygous EED loss. To explore the transcriptional consequences of PRC2 loss, we generated a gene set composed of genes that were differentially expressed in MPNSTs with loss of PRC2 and those with wild-type PRC2 (Supplementary Table 5). Hierarchical clustering of these genes robustly separated the MPNSTs with loss of PRC2 and those with wild-type PRC2. The vast majority of differentially expressed genes (455/479; 95%) were upregulated in MPNSTs with PRC2 loss, consistent with the role of PRC2 in transcriptional repression (Fig. 2b). In Gene Ontology (GO) analysis, known PRC2-suppressed targets, including homeobox transcription factors and genes associated with development and morphogenesis, were highly enriched 1229
letters
npg
NF1 associated
Sporadic
Radiotherapy associated
b
PRC2 genetic status
lo ss H om lo s
H et
+
18
4
0
–
0
7
30
H3K27me3 IHC status
T W
H&E H3K27me3 IHC
d P = 0.035 P = 0.001
30
15 1
20 10
PRC2 positive PRC2 negative
P = 0.009
1
19
15
11
0
3 8 Neurofibroma
Plexiform neurofibroma
40 Number of cases
H&E
MPNST
Epithelioid
MPNST-neurofibroma interface
Radiotherapy associated
MPNST
Sporadic
Neurofibroma
NF1 associated
c
among the genes upregulated in MPNSTs with PRC2 loss (Fig. 2c and Supplementary Table 6). The expression of several prototypical PRC2-suppressed genes confirmed this difference (Fig. 2d). In Gene Set Enrichment Analysis (GSEA), the most significantly enriched gene sets upregulated in MPNSTs with PRC2 loss included the ‘PRC2 module’ defined by genes bound by PRC2 components in mouse embryonic stem (ES) cells15 (Fig. 2e and Supplementary Table 7) and H3K27me3 target genes in neural precursor cells 16 (Fig. 2f) and brain tissue17 (Fig. 2g). These data indicate that loss of PRC2 function results in distinct transcriptome changes, including activation of developmentally regulated master regulators and imprinted genes (for example, IGF2). We evaluated H3K27me3 levels by immunohistochemistry in formalin-fixed, paraffin-embedded samples of MPNST and neurofibroma. Whereas MPNSTs with wild-type PRC2 showed robust staining for H3K27me3, MPNSTs with PRC2 loss showed complete loss of H3K27me3 in tumor cells and preservation of H3K27me3 staining in stromal cells (Fig. 3a). Positive and negative immunostaining for H3K27me3 were highly concordant with a genetic status of wild-type PRC2 and homozygous PRC2 loss, respectively (Fig. 3b). However, heterozygous loss of PRC2 components was not predictive of H3K27me3 staining in immunohistochemistry (Fig. 3b). Among the samples with heterozygous loss for which associated RNA-seq data were available, clustering by transcriptional pattern matched H3K27me3 immunohistochemistry results. This finding suggests that DNA sequencing (exome or IMPACT) alone cannot predict PRC2 functional status in all MPNSTs and that H3K27me3 immunohistochemistry might be more accurate. All neurofibromas (7/7), which were wild type for PRC2 except for one sample with heterozygous loss of SUZ12, retained H3K27me3 immunostaining (Figs. 1b and 3c,d). In specimens that contained the interface of MPNSTs arising from preexisting benign plexiform neurofibromas, we observed a transition from robust H3K27me3 staining in the plexiform neurofibroma to a clear loss of this staining in the MPNST. These data suggest that PRC2 loss is involved in the malignant progression of benign plexiform neurofibroma into MPNST. Indeed, 56% (19/34) of the NF1-associated MPNSTs had lost H3K27me3 (Fig. 3d). Curiously, a significantly greater percentage 1230
NF1 associated
s
a
H3K27me3 IHC H3K27me3 IHC
© 2014 Nature America, Inc. All rights reserved.
Figure 3 H3K27me3 immunohistochemistry correlates with PRC2 genetic status, and H3K27me3 loss characterizes progression from neurofibroma to MPNST. (a) Representative hematoxylin and eosin (H&E) staining and H3K37me3 immunohistochemistry (IHC) images of NF1-associated, sporadic and radiotherapy-associated MPNSTs. Scale bars, 100 µm. (b) Correlation of PRC2 genetic status by whole-exome sequencing, RNA-seq and custom targeted sequencing and H3K27me3 immunohistochemistry status. (c) Representative hematoxylin and eosin staining and H3K27me3 immunohistochemistry images of neurofibroma, NF1-associated MPNST and the interface of plexiform neurofibroma transitioning into MPNST. Scale bars, 100 μm. (d) Distribution of PRC2 loss and PRC2 presence by H3K27me3 immunohistochemistry in NF1-associated, sporadic, radiotherapy-associated and epithelioid MPNSTs and in neurofibromas. Fisher’s exact test was used to calculate P values.
MPNST
(>90%) of sporadic and radiotherapy-associated MPNSTs had lost H3K27me3 staining (Fig. 3d), suggesting that the progression of disease and sequence of genetic inactivation of NF1, CDKN2A and PRC2 components might be different in MPNSTs that arise in distinct clinical settings. Unlike NF1-associated MPNSTs that universally arise from preexisting neurofibromas, sporadic and radiotherapyassociated MPNSTs rarely have identifiable preexisting benign nerve sheath tumors6. In one sporadic MPNST sample (16T), the presence of nonsynonymous mutations in both NF1 (encoding a p.Asp1237_ splice alteration) and EED (encoding p.Glu249fs) allowed us to use the prevalence of these mutations to infer the sequence of genetic events18. The largest subpopulation of cells (84%) contained the NF1 mutation, whereas a smaller subpopulation (57%) contained the EED mutation, suggesting that the NF1 mutation occurred first during progression of this sporadic MPNST (Supplementary Fig. 4). The sequence of NF1, PRC2 component and CDKN2A inactivation described here is largely correlative. The precise sequence of events will require experimental validation with sequential inactivation of each pathway in cell line and mouse models. To determine whether PRC2 loss is required for MPNST oncogenesis, we screened available human MPNST cell lines using immunoblotting for H3K27me3. We identified one MPNST cell line (ST88-14; derived from an NF1-associated MPNST) that had lost H3K27me3 marks. RNA-seq analysis showed that ST88-14 cells had lost expression of SUZ12, and immunoblotting confirmed loss of the SUZ12 protein (Fig. 4a,b and Supplementary Fig. 5). We next introduced Flag-HA–tagged wild-type SUZ12 (FH-SUZ12) or EED (FH-EED) into the ST88-14 cell line and into an MPNST cell line with wild-type PRC2 (MPNST724) that maintained H3K27me3 levels (Fig. 4a,b). FH-SUZ12 but not FH-EED restored H3K27me3 levels in ST88-14 cells and substantially decreased cell growth (Fig. 4a–c). In MPNST724 cells, there was a mild increase in H3K27me3 levels with the introduction of either FH-SUZ12 or FH-EED (Fig. 4a), but neither had any effect on cell growth (Fig. 4c). These data suggest that PRC2 loss contributes to oncogenesis at least in part by promoting cell proliferation and growth. VOLUME 46 | NUMBER 11 | NOVEMBER 2014 Nature Genetics
letters
0.5
0 15
0 30
20
20
10
20
10
10
5
10
0 30
0 30
0 15
0 30
20
20
10
20
10
10
5
10
0 80 60 40 20 0 15
0 80 60 40 20 0 40 30 20 10 0 80 60 40 20 0 10 8 6 4 2 0
0 100 80 60 40 20 0 20 15 10 5 0 10 8 6 4 2 0 10 8 6 4 2 0
0 50 40 30 20 10 0 30
4
4
72 ST
88
PN M
ST
72
-1
4
4 ST
ST
0 25 20 15 10 5 0 10 8 6 4 2 0
-1
-1 88
PN M
ST
10
88
0 20 15 10 5 0 10 8 6 4 2 0
20
PN
5
ST
SUZ12 ChIP IgG ChIP
DAPI
10
M
FH-SUZ12
H3K27me3
FH-EED
4
ST88-14 Parental
72
25
ST
20
-1
10 15 Days
88
5
PN
0
4
0.5
0 30
Merge
© 2014 Nature America, Inc. All rights reserved.
npg
0.5
0 30
ST
MPNST724 FH-SUZ12
0.5
M
1.0 0
FH-EED
mRNA
2.0
Total H3
Parental
ST88-14
3.0
H3K27ac
b
1.0
4
H3K27me3
25
TLX1
1.5
1.0
72
Actin
* **
20
PAX2
1.5
1.0
4
EED
* **
10 15 Days
1.5
FH-SUZ12
1.0
0 5
IGF2
EZH2 ChIP
SUZ12
1.5
1.0 0
Vector FOXN4
H3K27me3 ChIP
EZH1
2.0
H3K4me3 ChIP
EZH2
d
MPNST724 Parental FH-EED FH-SUZ12
3.0
H3K27ac ChIP
NF1
c Normalized fluorescence units
Parental
FH-SUZ12
ST88-14
FH-SUZ12
FH-EED
Parental
MPNST724
FH-EED
a
Figure 4 PRC2 loss promotes cell proliferation and growth in MPNST with PRC2 loss. (a) Immunoblots demonstrating SUZ12 loss and corresponding loss of H3K27me3 in ST88-14 cells, a cell line derived from NF1-associated human MPNST, in comparison to a cell line derived from sporadic human MPNST, MPNST724, with intact PRC2 and retained H3K27me3 levels. Introduction of exogenous Flag-HA–tagged SUZ12 (FH-SUZ12) but not FlagHA–tagged EED (FH-EED) in ST88-14 cells restores H3K27me3 protein levels. *, exogenous FH-SUZ12 or FH-EED; **, endogenous SUZ12 or EED. (b) Immunofluorescence of H3K27me3 demonstrating the restoration of H3K27me3 levels at the cellular level by introducing FH-SUZ12 in the ST88-14 MPNST cell line with SUZ12 loss. Scale bars, 100 µm. DAPI, 4′,6-diamidino-2-phenylindole. (c) Representative growth curves for the MPNST724 and ST88-14 cell lines demonstrating that the introduction of FH-SUZ12 but not FH-EED in SUZ12-deficient ST88-14 cells leads to growth retardation, whereas it had no effect in MPNST724 cells. Similar results have been obtained in at least three independent experiments. (d) ST88-14 and MPNST724 cells were infected with vector control virus or virus expressing FH-SUZ12. Plots of expression determined by quantitative RT-PCR (expressed as 2 −∆∆Ct) and promoter localization determined by chromatin immunoprecipitation (ChIP) and quantitative PCR (expressed as percent input) of SUZ12, EZH2, H3K27me3, H3K4me3, H3K27ac and IgG control for the FOXN4, IGF2, PAX2 and TLX1 genes are shown. Error bars, s.e.m. n = 3 technical replicates.
We next examined the transcriptional and chromatin changes in ST88-14 and MPNST724 cells after the introduction of FH-SUZ12, focusing on several known PRC2-regulated genes (FOXN4, IGF2, PAX2 and TLX1) that are significantly upregulated in MPNST samples with PRC2 loss in comparison to MPNST samples with wild-type PRC2 (Fig. 2c). At baseline, ST88-14 cells exhibited increased expression of FOXN4, IGF2, PAX2 and TLX1 accompanied by loss of PRC2 components (SUZ12 and EZH2) and the PRC2-repressive mark (H3K27me3) and reciprocal gain of the activation marks trimethylation at lysine 4 of histone H3 (H3K4me3) and acetylation of lysine 27 of histone H3 (H3K27ac) at their promoters (Fig. 4d and Supplementary Fig. 6). After the introduction of FH-SUZ12 in ST88-14 cells, FH-SUZ12 localized to the promoters of these genes. This localization was accompanied by increased levels of EZH2 and H3K27me3 and decreased levels of H3K4me3 and H3K27ac at the promoter regions, as well as by decreased transcript levels of these PRC2 target genes. These data suggest that PRC2 loss has a direct impact on transcriptional regulation, and introduction of the missing PRC2 component has the ability to at least partially restore PRC2 function. MPNSTs often exhibit divergent differentiation, including rhabdomyoblastic, glandular, squamous and neuroendocrine elements6. Our study identified a high frequency of loss-of-function genetic Nature Genetics VOLUME 46 | NUMBER 11 | NOVEMBER 2014
alterations in NF1, CDKN2A and PRC2 components (EED or SUZ12), demonstrating that MPNSTs share common molecular pathogenic pathways despite clinical and histological diversity. PRC2 loss activates multiple developmentally suppressed pathways, which might explain the frequent observation of divergent differentiation in MPNST. The high frequency of PRC2 loss suggests that PRC2 mutational status and, more specifically, H3K27me3 immunohistochemistry could be used as biomarkers for the more acute diagnosis of MPNST. PRC2 was initially thought to be oncogenic: PRC2 components have higher expression in dividing cells and are important in maintaining stemness. EZH2 is overexpressed in a variety of cancers19,20, and activating EZH2 mutations are found in a subset of lymphomas21. Paradoxically, recent work suggests that PRC2 can be tumor suppressive in distinct contexts, with loss-of-function genetic alterations found in up to 25% of myeloid disorders and T cell acute lymphoblastic leukemia (ALL)22–24 and 42% of early T cell ALL25. Notably, the majority of the loss-of-function alterations are found in EZH2 (refs. 22–25). Cellular studies and mouse models have shown that Ezh1 can maintain suppression of Polycomb-regulated genes in the setting of Ezh2 loss, and combined Ezh1 and Ezh2 loss or individual loss of Eed or Suz12 derepresses Polycomb-regulated genes and causes Cdkn2a-mediated growth arrest26–29. These findings suggest that MPNST is unique in 1231
letters that complete loss of PRC2 function is important for tumorigenesis, and loss of CDKN2A might be a critical cooperative event in addition to NF1 loss. URLs. GENE-E, https://www.broadinstitute.org/cancer/software/ GENE-E/; Picard, http://picard.sourceforge.net/; aroma.affymetrix, http://aroma-project.org/; GSEA, http://www.broadinstitute.org/gsea/; DAVID, http://david.abcc.ncifcrf.gov/; STAR, https://code.google. com/p/rna-star/; DESeq, http://bioconductor.org/packages/release/ bioc/html/DESeq.html; HTSeq, https://pypi.python.org/pypi/HTSeq; IGV, http://www.broadinstitute.org/igv/IGV. Methods Methods and any associated references are available in the online version of the paper.
npg
© 2014 Nature America, Inc. All rights reserved.
Accession codes. Exome sequencing, RNA-seq and DNA copy number data have been deposited in the database of Genotypes and Phenotypes (dbGaP) under accession phs000792.v1.p1. Note: Any Supplementary Information and Source Data files are available in the online version of the paper. Acknowledgments Next-generation sequencing and SNP6.0 analysis were performed at the Memorial Sloan Kettering Cancer Center (MSKCC) Genomics Core Facility, Center for Molecular Oncology (CMO). The authors thank members of the cBioPortal team (J. Gao, B. Gross, N. Schultz and C. Sander) for assistance with data analysis and visualization. The authors also thank T. Chan (MSKCC) and HOPP informatics for assistance with data analysis. This work was supported in part by a grant from the US National Institutes of Health (NIH) to W.L. (NCI U24-CA143840), the Harry J. Lloyd Trust–Translational Research grant to T.W., the Charles H. Revson Senior Fellowship to T.W., Jubilaeumsfonds of the Oesterreichische Nationalbank to T.W., grants from the US NIH to P.C. (P50CA140146, Career Development Award and Developmental Research Project.; DP2CA174499; K08CA151660), Y.C. (K08CA140946), L.-X.Q. (P50CA140146), C.R.A. (P50CA140146) and S.S. (P50CA140146), an award from the Sidney Kimmel Foundation to P.C. (Kimmel Scholar Award) and funding from the Cycle for Survival Fund to P.C. AUTHOR CONTRIBUTIONS Project planning and experimental design: P.C., C.R.A., Y.C., T.W., W.L., S.T. and L.R. Sample collection, clinical database and cell lines: S.T., S.S., C.R.A., C.N.P.G., J.A.F. and W.D.T. Pathology review: C.R.A. and C.N.P.G. Preparation of DNA, RNA and next-generation sequencing libraries: T.W., K.H.H., S.T. and A.V. Sequence data analysis: W.L., Y.C., S.T., P.C., M.F.B., M.L., D.Z., Y.L. and A.S. Immunohistochemistry: T.W. Protein blots, immunofluorescence, growth curves and all cellular assays: L.R. and S.Z. Generation of the expression vectors: Z.C. and L.R. Biostatistics: L.-X.Q. Manuscript writing: P.C., Y.C., W.L., T.W. and L.R. The final manuscript was reviewed by all authors. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html.
1232
1. Brennan, M.F., Antonescu, C.R. & Maki, R.G. in Management of Soft Tissue Sarcoma 149–160 (Springer, New York, 2013). 2. LaFemina, J. et al. Oncologic outcomes of sporadic, neurofibromatosis-associated, and radiation-induced malignant peripheral nerve sheath tumors. Ann. Surg. Oncol. 20, 66–72 (2013). 3. Stucky, C.C. et al. Malignant peripheral nerve sheath tumors (MPNST): the Mayo Clinic experience. Ann. Surg. Oncol. 19, 878–885 (2012). 4. Eilber, F.C. et al. Validation of the postoperative nomogram for 12-year sarcomaspecific mortality. Cancer 101, 2270–2275 (2004). 5. Rodriguez, F.J. Peripheral nerve sheath tumors: the elegant chapter in surgical neuropathology. Acta Neuropathol. 123, 293–294 (2012). 6. Antonescu, C.R., Scheithauer, B.W. & Woodruff, J.M. in Tumors of the Peripheral Nervous System 553 (The American Registry of Pathology, Silver Spring, Maryland, 2013). 7. Taylor, B.S. et al. Advances in sarcoma genomics and new therapeutic targets. Nat. Rev. Cancer 11, 541–557 (2011). 8. Margueron, R. & Reinberg, D. The Polycomb complex PRC2 and its mark in life. Nature 469, 343–349 (2011). 9. Kourea, H.P., Orlow, I., Scheithauer, B.W., Cordon-Cardo, C. & Woodruff, J.M. Deletions of the INK4A gene occur in malignant peripheral nerve sheath tumors but not in neurofibromas. Am. J. Pathol. 155, 1855–1860 (1999). 10. Menon, A.G. et al. Chromosome 17p deletions and p53 gene mutations associated with the formation of malignant neurofibrosarcomas in von Recklinghausen neurofibromatosis. Proc. Natl. Acad. Sci. USA 87, 5435–5439 (1990). 11. Nielsen, G.P. et al. Malignant transformation of neurofibromas in neurofibromatosis 1 is associated with CDKN2A/p16 inactivation. Am. J. Pathol. 155, 1879–1884 (1999). 12. Perrone, F. et al. p15INK4b, p14ARF, and p16INK4a inactivation in sporadic and neurofibromatosis type 1–related malignant peripheral nerve sheath tumors. Clin. Cancer Res. 9, 4132–4138 (2003). 13. Won, H.H., Scott, S.N., Brannon, A.R., Shah, R.H. & Berger, M.F. Detecting somatic genetic alterations in tumor specimens by exon capture and massively parallel sequencing. J. Vis. Exp. (80), e50710 (2013). 14. Fleiss, J.L. Measuring nominal scale agreement among many raters. Psychol. Bull. 76, 378–382 (1971). 15. Kim, J. et al. A Myc network accounts for similarities between embryonic stem and cancer cell transcription programs. Cell 143, 313–324 (2010). 16. Mikkelsen, T.S. et al. Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553–560 (2007). 17. Meissner, A. et al. Genome-scale DNA methylation maps of pluripotent and differentiated cells. Nature 454, 766–770 (2008). 18. Andor, N., Harness, J.V., Muller, S., Mewes, H.W. & Petritsch, C. EXPANDS: expanding ploidy and allele frequency on nested subpopulations. Bioinformatics 30, 50–60 (2014). 19. Varambally, S. et al. The polycomb group protein EZH2 is involved in progression of prostate cancer. Nature 419, 624–629 (2002). 20. Simon, J.A. & Lange, C.A. Roles of the EZH2 histone methyltransferase in cancer epigenetics. Mutat. Res. 647, 21–29 (2008). 21. Morin, R.D. et al. Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat. Genet. 42, 181–185 (2010). 22. Ernst, T. et al. Inactivating mutations of the histone methyltransferase gene EZH2 in myeloid disorders. Nat. Genet. 42, 722–726 (2010). 23. Nikoloski, G. et al. Somatic mutations of the histone methyltransferase gene EZH2 in myelodysplastic syndromes. Nat. Genet. 42, 665–667 (2010). 24. Ntziachristos, P. et al. Genetic inactivation of the polycomb repressive complex 2 in T cell acute lymphoblastic leukemia. Nat. Med. 18, 298–301 (2012). 25. Zhang, J. et al. The genetic basis of early T-cell precursor acute lymphoblastic leukaemia. Nature 481, 157–163 (2012). 26. Ezhkova, E. et al. EZH1 and EZH2 cogovern histone H3K27 trimethylation and are essential for hair follicle homeostasis and wound repair. Genes Dev. 25, 485–498 (2011). 27. Shen, X. et al. EZH1 mediates methylation on histone H3 lysine 27 and complements EZH2 in maintaining stem cell identity and executing pluripotency. Mol. Cell 32, 491–502 (2008). 28. Xie, H. et al. Polycomb repressive complex 2 regulates normal hematopoietic stem cell function in a developmental-stage-specific manner. Cell Stem Cell 14, 68–80 (2014). 29. Margueron, R. et al. Ezh1 and Ezh2 maintain repressive chromatin through different mechanisms. Mol. Cell 32, 503–518 (2008).
VOLUME 46 | NUMBER 11 | NOVEMBER 2014 Nature Genetics
npg
© 2014 Nature America, Inc. All rights reserved.
ONLINE METHODS
Human tumor tissue collection. Biospecimens were collected during surgical resection from patients with pathologic diagnosis of MPNSTs or neurofibromas. Material was collected under institutional review board (IRB)-approved protocols (IRB 06-107) at MSKCC. All patients provided informed consent. Pathologic diagnosis was confirmed by at least two pathologists using diagnostic formalin-fixed and paraffin-embedded sections to select cases with an estimated tumor content of >70%. The majority of the tumors were collected from different patients. In some cases, more than one tumor was selected from the same patient, and these tumors were resected from distinct anatomical locations in different surgeries. Sample annotation is shown in Supplementary Table 1. For the discovery cohort, the goal was to obtain normal DNA and tumor DNA for whole-exome sequencing and SNP6.0 array analysis and to obtain tumor RNA for RNA-seq analysis. A total of 15 fresh-frozen paired MPNST tumor–normal samples were identified. Tumor and adjacent normal tissue specimens were embedded in optimal cutting temperature (OCT) medium, and a histological section was obtained for review. Cryomolds of both tumor and normal tissues were macrodissected to minimize contamination before RNA and DNA preparation. Sample processing was designed to secure samples and minimize the availability of identifying information. Specimens with insufficient tissue amount or severely degraded nucleic acids were excluded. The IMPACT assay is a hybridization capture, next-generation sequencing platform amenable to DNA from both fresh-frozen and formalin-fixed, paraffin-embedded samples for targeted sequencing. The panel includes NF1, SUZ12, EED, CDKN2A and TP53 (ref. 13). We validated the somatic mutational findings from the discovery cohort by performing IMPACT assays on the same DNA isolated from tumor tissue. In addition, we performed IMPACT assays on DNA from formalin-fixed, paraffin-embedded samples derived from a second cohort of 37 MPNSTs and 7 neurofibromas from individuals with NF1 who were diagnosed with concurrent MPNST (Supplementary Table 1). Sample preparation and quality control. RNA was extracted from tumor and normal tissues using a modification of the protocol for the DNA/RNA AllPrep kit (Qiagen). DNA from fresh-frozen tissues was extracted from tumor and normal tissue specimens using the DNeasy Blood and Tissue kit (Qiagen). DNA from formalin-fixed, paraffin-embedded samples was isolated using the QIAamp DNA FFPE Tissue kit (Qiagen). DNA from each specimen was initially quantified using the NanoDrop UV spectrophotometer and was further quantified with the Bioanalyzer assay (Agilent Technologies). RNA sequencing and analysis. The isolated RNA was processed using the TruSeq RNA Sample Prep kit (15026495, Illumina) according to the manufacturer’s protocol. Briefly, RNA was polyA selected and reverse transcribed, and the cDNA obtained underwent end repair, A-tailing, ligation of the indexes and adaptors, and PCR enrichment. Libraries were sequenced on the Illumina HiSeq 2500 platform with 51-bp paired-end reads to obtain a minimum yield of 40 million reads per sample. Sequence data were processed and mapped to the human reference genome (hg19) using STAR (v2.3)30. Gene expression levels were quantified with htseq-count31 and normalized using DESeq32. Variance in expression levels was calculated for all genes across samples, and the 75th percentile was set as a cutoff. PCA was performed on the set of genes with variance greater than this cutoff. We used ANOVA to define differentially expressed genes between samples with PRC2 loss and wild-type PRC2. Genes that showed a >8-fold difference in expression and a corrected FDR of
