Purpose: Malignant peripheral nerve sheath tumors (MPNST) are highly aggressive sarcomas with variable patient survival and few known prognostically relevant genomic biomarkers. To identify survival-associated genomic biomarkers, we performed high-resolution array-based comparative genomic hybridization (aCGH) on a large set of MPNSTs.

Experimental Design: Candidate gene alterations identified by aCGH in 38 MPNSTs were validated at the DNA, RNA, and protein levels on these same tumors and an independent set of 87 MPNST specimens.

Results: aCGH revealed highly complex copy number alterations, including both previously reported and completely novel loci. Four regions of copy number gain were associated with poor patient survival. Candidate genes in these regions include SOX5 (12p12.1), NOL1 and MLF2 (12p13.31), FOXM1 and FKBP1 (12p13.33), and CDK4 and TSPAN31 (12q14.1). Alterations of these candidate genes and several others of interest (ERBB2, MYC and TP53) were confirmed by at least 1 complementary methodology, including DNA and mRNA quantitative real-time PCR, mRNA expression profiling, and tissue microarray-based fluorescence in situ hybridization and immunohistochemistry. Multivariate analysis showed that CDK4 gain/amplification and increased FOXM1 protein expression were the most significant independent predictors for poor survival in MPNST patients (P < 0.05).

Conclusions: Our study provides new and independently confirmed candidate genes that could serve as genomic biomarkers for overall survival in MPNST patients. Clin Cancer Res; 17(7); 1924–34. ©2011 AACR.

Translational Relevance

Malignant peripheral nerve sheath tumor (MPNST) is an aggressive malignancy with considerable biological variability, a variable and often lethal clinical course, and limited response to current therapeutic approaches. A major impediment to more effective disease management has been a lack of available prognostic or predictive biomarkers. Through initial identification of several survival-associated copy number alterations, followed by confirmatory assays and a comprehensive multivariate analysis, our present study demonstrates that CDK4 gain/amplification and increased FOXM1 protein expression are significant independent predictors for poor survival in MPNST patients. If confirmed in additional prospective studies, these biomarkers may provide clinically useful information for managing patients with this aggressive malignancy. Interestingly, a previous report has demonstrated that the multi-kinase inhibitor sorafenib can inhibit MPNST growth, suggesting the potential of CDK4 as a therapeutic target and also as a potential predictive biomarker of response to sorafenib therapy.

Malignant peripheral nerve sheath tumors (MPNST) are highly aggressive and frequently lethal Schwann cell-derived neoplasms associated with poor survival (1–5). Roughly half of MPNSTs develop sporadically and the other half arise in the setting of neurofibromatosis type 1 (NF1). The 5-year survival rate for NF1-associated MPNST patients is reportedly half of that for sporadic cases (21% vs. 42%; ref 5). However, other studies suggest that a diagnosis of NF1 does not affect survival (6). Clinicopathologic factors, such as tumor grade and anatomic site reportedly influence MPNST patient outcome, although the clinical course remains highly variable and difficult to predict in individual patients.

Traditional cytogenetic studies have reported that MPNSTs have remarkably complex karyotypes with a wide spectrum of chromosomal alterations including translocations, duplications, and numerical gains and losses, making it difficult to discern which of these alterations are biologically relevant (7–9). Unfortunately, no single alteration has consistently proven to be diagnostically or prognostically useful across multiple studies. For example, alterations on chromosome 17 (where the NF1 gene is located) are expected, given that MPNSTs are highly over-represented among NF1 patients. However, NF1 losses are not always encountered (8–10), suggesting that other tumorigenic pathways also exist. A major limitation of traditional cytogenetic studies, however, has been their limited resolution of ∼10 Mb. This has complicated the discovery of specific, recurring genetic alterations involving small regions.

High-resolution genome-wide DNA microarray analyses have identified many novel tumor-specific microdeletions and amplifications in leukemia, lung, and ovarian cancers (11–13). To date, a similar approach applied to MPNSTs has been limited to few studies with either less comprehensive or lower resolution array platforms (∼1 Mb; refs 14–16). In the present study, we have analyzed a larger cohort of MPNSTs utilizing both Affymetrix and Nimblegen array-based comparative genome hybridization (aCGH) platforms (0.25–2.5 kb of resolution with whole genome coverage) to identify recurrent DNA copy number alterations (CNA) that are associated with reduced patient survival, independent of traditional clinical and pathological parameters in patients with this extremely heterogeneous tumor. The results from this study provide a short list of candidate genes with CNAs which may prove clinically useful as prognostic biomarkers for patients with MPNST.

Patients and specimens

A cohort of 38 patients with histologically confirmed MPNST was included in this study for microarray analysis. Among the 38 patients, 23 (60%) individuals carried a previous diagnosis of NF1 and 15 (40%) were sporadic cases. At the time of last follow-up, 14 (37%) cases had documented metastases, whereas 16 (42%) had no evidence of metastatic disease. Recurrence data were unavailable for 8 of the cases. The majority (33/38) of cases subjected to genomic analysis was primary tumors, although 5 cases of recurrent tumors were also included. A full accounting of clinical and pathological data is provided in Supplemental Table T1. DNAs and RNAs were extracted from frozen tumors, using commercial reagents (QIAamp DNA Mini Kit, Qiagen; Trizol reagent, Invitrogen) and following manufacturers' standard protocols. These same specimens were previously utilized for gene expression profiling (17). For biomarker validation studies, another independent set of 121 formalin-fixed paraffin-embedded MPNST tumors represented in a tissue microarray (TMA) was utilized; 87 of these patients had clinical annotation and follow-up data (18). Of these 87 annotated cases, 25 (29%) were NF1-associated and 62 (71%) were sporadic MPNSTs. A total of 68 primary, 12 locally recurrent, and 7 metastatic tumors were represented. The majority of the tumors (78/87) were high grade and 9/87 were classified as low grade. Supplemental Table T2 provides additional data regarding the validation set. All tissue specimens and clinical data from these patients were collected and utilized in accordance with Institutional Review Board–approved protocols.

Array-based comparative genome hybridization

The Affymetrix Genome-Wide 500K SNP Mapping array set was employed to detect genomic alterations in the set of 38 MPNSTs. For aCGH analysis, relative copy number values of the MPNSTs were generated using a reference set of 37 nonmalignant DNA samples derived from 22 males and 15 females. To identify meaningful CNAs, 2 DNA copy number analysis tools, Partek Genomics Suite (PARTEK, www.partek.com) and Copy Number Analyzer for GeneChip (CNAG, www.genome.umin.jp; ref 19) were applied to the aCGH data. CNAs identified with each of the 2 tools were compared with identify recurrent regions with minimal overlap. Physical positions of CNAs were annotated according to human genome build 35 (www.ncbi.nlm.nih.gov). All CNAs were filtered against the human normal copy number variation (CNV) database (http://projects.tcag.ca/variation) prior to further analysis. CNAs identified on the 500K SNP platform were also confirmed in a subset of 8 patients where patient-matched tumor and germline DNAs were available, using a dual-color high-density NimbleGen 135K aCGH platform (see Supplemental Text for details).

Quantitative real-time PCR

SYBR Green chemistry was utilized in quantitative real-time PCR (qRT-PCR) assays; qRT-PCR primer sequences are listed in Supplemental Table T3. The qRT-PCR was performed in 10-μL reactions containing 10 ng of cDNA or genomic DNA and gene-specific primers each at 200 nmol/L of final concentration, as previously described (20), on the ABI 7900HT system with the following program: 50°C for 2 minutes, 95°C for 10 minutes, and 40 cycles of 95°C, 56°C, and 72°C for 15, 30, and 30 seconds, respectively. All sets of primers yielded a single peak in dissociation curve analysis and had an efficiency of greater than 0.9.

Immunohistochemistry

Fresh TMA sections were obtained from the Washington University/Siteman Cancer Center Tissue Procurement Core Facility, and immunohistochemistry (IHC) was performed using the streptavidin-biotin-peroxidase complex technique as described previously (21). The primary antibodies included FOXM1 (Santa Cruz; Cat. #: sc-502; 1:200 dilution), MYC (Zymed; Cat. #: 9E10; 1:200 dilution), SOX5 (Abcam; Cat. #: ab26041; 1:500 dilution), and p53 (Biogenex; Cat. #: AM240-5M; 1:200 dilution). Evaluation of each core on the TMA was made independently by 2 pathologists (L.S., A.P.), and the final score for each tumor was derived from the average score of replicate cores on the arrays (see Supplemental Text). CDK4 TMA IHC was also performed, but was not interpretable because of poor staining quality of this antibody reagent on the TMA spots.

FISH

FISH was performed on TMA slides as previously described (22, 23). Freshly cut slides were subjected to dual-color hybridization with the following probe pairings: centromere enumerating probe 12 (CEP12)/CDK4 (12q14.1), CEP17/HER2 (ERBB2; 17q21.1), and CEP8/MYC (8q24.21), respectively. Five of the probes were obtained commercially and were labeled with SpectrumGreen (CEP8, CEP12, and CEP17) or SpectrumOrange (MYC and ERBB2) fluorochromes (Vysis). The CDK4 locus was targeted with a locally developed probe, utilizing the RP11-571M6 human BAC clone as previously published (24; see Supplemental Text).

Statistical analysis

Each of the minimally overlapping genomic regions was tested for association with clinical variables, for which the Kaplan–Meier model was used for survival analysis. Differential expression of candidate genes was detected between patients with and without CNAs where positive correlations with survival were seen using the Mann–Whitney U test. Survival probability was first estimated using the Breslow test (univariate Cox regression) for each of the genomic and relevant clinical variables, and then the significant predictors in univariate models were entered into multivariate models. The stepwise selection method was utilized to build the most significant models. The cut-off P value was set as 0.05. All computations were performed in SAS (v9.1.3).

Characterization of MPNST genomic alterations

As shown by high-resolution aCGH (Fig. 1) and further quantified by calculation of the “breakpoint index” (a surrogate measure of genomic instability calculated based upon the total number of independently amplified and deleted genomic segments identified in a tumor based on aCGH analysis, see ref. 13), MPNSTs demonstrate a highly complex and disrupted genome. MPNST genomes had higher overall complexity (mean “breakpoint index” = 346) compared with similarly characterized genomes of other tumor types: leukemia (mean = 88; ref 25), ovarian cancer (median = 95; ref 13), and pilocytic astrocytoma (mean = 111; ref 20). The indices were, however, similar to lymphoma (mean = 374; ref 26). The percentage of the total genome affected by CNAs was also calculated to assess genome complexity of individual tumors. This ranged widely from 0.2% to 97% (median 34%) among the 38 MPNSTs (Fig. 1C). This observation no doubt reflects the pathological diversity of tumors characterized in this cohort, although a direct association between reduced survival time and overall high genome complexity (defined here as ≥ 34% of the genome affected by CNA) did not quite reach statistical significance (Fig. 2A, P = 0.071).

Figure 1.

Demonstration of high complexity genome alterations in 38 MPNSTs, defined by PARTEK Genomic Segmentation algorithm in aCGH analysis. A, heat map of SNP-level copy number data. Copy number loss is shown in blue and gain in red on each of 22 autosomal chromosomes. A 2-color bar at the top of the map indicates the data range in log2 ratio. “Good outcome” is defined as survival more than 24 months, “moderate outcome” as survival 13 to 24 months (or “indetermined” as lost to follow-up while alive, with recorded survival <24 months), and “poor outcome” as survival of 12 months or less. Note that in general, patients with “poor outcome” had more CNVs than patients with “good outcome.” B, frequency of individual CNAs, across all autosomal chromosomes in 38 MPNST patients. Blue represents copy number losses and red denotes copy number gains. C, distribution of percentage of the total genome affected by CNAs among 38 MPNSTs analyzed by aCGH. Calculation of “percentage” is based upon NCBI Build 35, May 2004 Assembly (hg17) for human genome size (2864255932 bp for autosome).

Figure 1.

Demonstration of high complexity genome alterations in 38 MPNSTs, defined by PARTEK Genomic Segmentation algorithm in aCGH analysis. A, heat map of SNP-level copy number data. Copy number loss is shown in blue and gain in red on each of 22 autosomal chromosomes. A 2-color bar at the top of the map indicates the data range in log2 ratio. “Good outcome” is defined as survival more than 24 months, “moderate outcome” as survival 13 to 24 months (or “indetermined” as lost to follow-up while alive, with recorded survival <24 months), and “poor outcome” as survival of 12 months or less. Note that in general, patients with “poor outcome” had more CNVs than patients with “good outcome.” B, frequency of individual CNAs, across all autosomal chromosomes in 38 MPNST patients. Blue represents copy number losses and red denotes copy number gains. C, distribution of percentage of the total genome affected by CNAs among 38 MPNSTs analyzed by aCGH. Calculation of “percentage” is based upon NCBI Build 35, May 2004 Assembly (hg17) for human genome size (2864255932 bp for autosome).

Close modal
Figure 2.

Kaplan–Meier survival curves in MPNST patients of 2 independent cohorts, stratified by specific genome CNAs or clinical variables, as indicated in: A–F, initial set of 38 MPNSTs; G–L, confirmation set of 87 MPNSTs. Less than the total number of tumors in each cohort was analyzed when individual biomarker or clinical data were not available.

Figure 2.

Kaplan–Meier survival curves in MPNST patients of 2 independent cohorts, stratified by specific genome CNAs or clinical variables, as indicated in: A–F, initial set of 38 MPNSTs; G–L, confirmation set of 87 MPNSTs. Less than the total number of tumors in each cohort was analyzed when individual biomarker or clinical data were not available.

Close modal

Many of the genomic loci associated with recurrent CNAs harbor genes previously implicated in MPNST biology and were observed with frequencies comparable to those reported in previous studies (14; Supplemental Table T4). For example, 19 of 38 (50%) cases had deletion at the NF1 locus. Interestingly, however, this event was independent of clinical diagnosis as 13 of 23 NF1 patients (56%) as well as 6 of 15 sporadic cases (40%) demonstrated deletions. Several other recurrent CNAs of genes not previously associated with MPNSTs were also identified. These included gains of PDCD1 at 2p25.3 (47% of cases), SOX5 at 12p12.1 (34% of cases), FOXM1 at 12p13.33 (29% of cases), and NOL1 at 12p13.31 (29% of cases) as well as losses of PSIP1/p75 at 9p22.3 (50% of cases; ref 27), PTPRD at 9p23 to 9p24.1 (47% of cases; ref 28), and PTEN at 10q23.31 (34% cases). Although it is likely that NF1-associated and sporadic MPNSTs arise in the context of different genetic backgrounds, we found that none of these nor any other loci identified in this analysis was present with statistical significance between NF1-associated and sporadic MPNSTs.

Identification of survival-associated genomic alterations

Minimally overlapping regions of genome alteration (1,564 copy number gains and 252 copy number losses) identified using 2 alternate segment calling algorithms were correlated with clinical variables. Each genome region evaluated contained at least 1 annotated gene and recurred in greater than 10% of 38 tumors. A detection frequency of 10% (4 tumors) was chosen as a minimum cut-off to provide sufficient statistical power given the sample size and recognizing the already extreme diversity of this tumor type. Using either Mann–Whitney U test or chi-square test, no significant association was observed between these regions and clinical variables such as age, gender, tumor status (i.e., primary or recurrent), anatomic location and grade, metastasis, recurrence, or NF1 association. Using the Kaplan–Meier model, however, 36 minimally overlapping regions with copy number gains demonstrated significant associations with decreased survival times (Supplemental Table T5). No regions with copy number losses were statistically associated with patient survival.

When ranked by both frequency and P value, the most significant region of copy number gain was a 173 kb minimally overlapping region at 12p12.1 that includes only a portion of the SOX5 gene (Fig. 2B). This region demonstrated copy number gains in 13 tumors, but with variability in the size of the locus affected (Fig. 3A). Three other genomic regions associated with survival were also identified (Fig. 2C and E), where differential mRNA expression was seen in at least 1 overlapping gene for each of the regions, based on previously published expression profiling data of the same tumors (Table 1; ref 17). The sizes of these 3 regions were 265 kb at 12p13.31, 551 kb at 12p13.33, and 420 kb at 12q14.1, and included the entire coding sequence of NOL1/MLF2 (9 cases; 25%), FOXM1/FKBP4 (11 cases; 31%), and CDK4/TSPAN31 (9 cases; 25%), respectively (Fig. 3B–D). In comparison, the 8q24.21 locus harboring MYC, a gene frequently amplified in neuroectodermal tumors, was also amplified in 13 (34%) MPNSTs (Fig. 3E) and was associated with a trend toward decreased survival (P > 0.05). As described below, this association reached statistical significance on subsequent confirmation experiments that evaluated MYC protein expression (Fig. 2G).

Figure 3.

Five minimally overlapping regions of copy number gain identified by cross-comparison of segmentation data using 2 independent algorithms in copy number analysis of a set of 38 MPNSTs. For each of 38 tumors, 2 segments are shown in adjacent columns: left—segmentation detected by Hidden Markov Model and right—segmentation detected by Genomic Segmentation algorithm. Categories of CNAs are designated by a gray scale from white (neutral), light-gray (deletion), and dark-gray (low copy number gain) to black (high-level DNA amplification). The dotted lines in each of the 5 regions indicate the minimally overlapping region in cross-comparison. The genomic location is shown at the top and bottom within each region, along with representative genes at specific cytobands. Patients outcome: “good” as survival more than 24 months, “moderate” as survival 13 to 24 months (or “indetermined” as lost to follow-up while alive, with recorded survival <24 months), and “poor” as survival of 12 months or less.

Figure 3.

Five minimally overlapping regions of copy number gain identified by cross-comparison of segmentation data using 2 independent algorithms in copy number analysis of a set of 38 MPNSTs. For each of 38 tumors, 2 segments are shown in adjacent columns: left—segmentation detected by Hidden Markov Model and right—segmentation detected by Genomic Segmentation algorithm. Categories of CNAs are designated by a gray scale from white (neutral), light-gray (deletion), and dark-gray (low copy number gain) to black (high-level DNA amplification). The dotted lines in each of the 5 regions indicate the minimally overlapping region in cross-comparison. The genomic location is shown at the top and bottom within each region, along with representative genes at specific cytobands. Patients outcome: “good” as survival more than 24 months, “moderate” as survival 13 to 24 months (or “indetermined” as lost to follow-up while alive, with recorded survival <24 months), and “poor” as survival of 12 months or less.

Close modal
Table 1.

Validation summary on potential survival marker genes

MarkersCopy number gains (losses) are associated to survival in 38 MPNSTs?Differential mRNA expression exists in arrays between tumorswith and without CNAs?CNAs identified by aCGH are confirmed by qRT-PCR in same MPNSTs?Differential mRNA expression is confirmed by qRT-PCR in same MPNSTs?Gene amplification in FISH is associated with survival of 87 MPNSTs?High protein expression in IHC is associated with survival of 87 MPNSTs?
SOX5 Significant No Significant Not significant n/a Not significant 
NOL1 Significant Yes Significant Significant n/a n/a 
MLF2 Significant Yes Significant Significant n/a n/a 
FOXM1 Significant Yes Significant Significant n/a Significant 
FKBP4 Significant Yes Significant Significant n/a n/a 
CDK4 Significant Yes Significant Significant Significant n/a 
TSPAN31 Significant Yes Significant Not significant n/a n/a 
ERBB1 Not significant No n/a n/a Not significant n/a 
MYC Not significant Yes n/a n/a Not significant Significant 
TP53 Not significant No n/a n/a n/a Not significant 
MarkersCopy number gains (losses) are associated to survival in 38 MPNSTs?Differential mRNA expression exists in arrays between tumorswith and without CNAs?CNAs identified by aCGH are confirmed by qRT-PCR in same MPNSTs?Differential mRNA expression is confirmed by qRT-PCR in same MPNSTs?Gene amplification in FISH is associated with survival of 87 MPNSTs?High protein expression in IHC is associated with survival of 87 MPNSTs?
SOX5 Significant No Significant Not significant n/a Not significant 
NOL1 Significant Yes Significant Significant n/a n/a 
MLF2 Significant Yes Significant Significant n/a n/a 
FOXM1 Significant Yes Significant Significant n/a Significant 
FKBP4 Significant Yes Significant Significant n/a n/a 
CDK4 Significant Yes Significant Significant Significant n/a 
TSPAN31 Significant Yes Significant Not significant n/a n/a 
ERBB1 Not significant No n/a n/a Not significant n/a 
MYC Not significant Yes n/a n/a Not significant Significant 
TP53 Not significant No n/a n/a n/a Not significant 

Confirmation of copy number alterations associated with survival

To confirm results obtained from aCGH data sets, qRT-PCR was performed to quantify DNA copy number for 7 specific candidate genes using the same 38 MPNST DNAs employed for aCGH analysis. DNA qRT-PCR results confirmed associations between decreased survival and DNA copy number gains for all 7 genes (Table 1). Using the same MPNST RNAs from our previously reported gene expression microarray study (17), qRT-PCR analysis confirmed differential expression for 5 of the 7 genes (NOL1, MLF2, FOXM1, FKBP4, and CDK4), whereas SOX5 and TSPAN31 showed no differential expression (Table 1). Given that only a portion of the SOX5 gene was involved in the amplified region, we questioned whether this amplification reflected a qualitative rearrangement of the locus, rather than SOX5 amplification. However, sequencing of RNAs from tumors that exhibited SOX5 amplification using both 3′ and 5′ RACE (rapid amplification of cDNA ends) failed to identify a novel fusion gene transcript in these tumors (data not shown).

We also employed an independent cohort of 121 patients tumors (87 with clinical follow-up) to construct a TMA and evaluate gene amplification and protein expression of several candidate genes (CDK4, FOXM1, and SOX5) that were identified and confirmed by the genomic approaches described above. For comparison, we also examined several other previously reported biomarkers of survival (ERBB2, MYC, and TP53) in this same cohort. Using Kaplan–Meier survival analysis, CDK4 amplification in 12 of 79 (15%) cases (scored as both true gene amplification and high-level polysomies/gains by FISH analysis) and increased protein expression of FOXM1 and MYC were statistically associated with significantly shorter overall survival in this cohort (Figs. 2G–I and 4). In contrast, MYC and ERBB2 gene amplification as well as TP53 (Fig. 2J) and SOX5 protein immunoreactivity were not associated with overall survival.

Figure 4.

A–D, representative TMA FISH analysis of CDK4 copy number gain (red, CDK4 probe; green, chromosome 12 probe). A, CDK4 gene amplification; B, high-level polysomy 12; C, polysomy 12; D, normal copy numbers of CDK4 and chromosome 12. For survival analysis, A and B were grouped into a single category of CDK4 amplification, with C and D classified as CDK4 nonamplification. E–H, representative TMA IHC analysis of FOXM1 (E, F) and MYC (G, H) protein expression. High expression (E, G; scores of +++ and ++) versus low or negative expression (F, H; scores of + and −) for each antibody is shown.

Figure 4.

A–D, representative TMA FISH analysis of CDK4 copy number gain (red, CDK4 probe; green, chromosome 12 probe). A, CDK4 gene amplification; B, high-level polysomy 12; C, polysomy 12; D, normal copy numbers of CDK4 and chromosome 12. For survival analysis, A and B were grouped into a single category of CDK4 amplification, with C and D classified as CDK4 nonamplification. E–H, representative TMA IHC analysis of FOXM1 (E, F) and MYC (G, H) protein expression. High expression (E, G; scores of +++ and ++) versus low or negative expression (F, H; scores of + and −) for each antibody is shown.

Close modal

Finally, a multivariate analysis was carried out to determine whether any of the genomic biomarkers identified could serve as independent predictors for survival (Table 2). Among all genomic and clinical variables examined, only 4 CNA regions (representatively named as CDK4, FOXM1, NOL1, and SOX5) and anatomic site (Fig. 2F) showed significant associations with survival time by univariate regression analysis in the 38 MPNSTs. In addition, CDK4 gene amplification by FISH analysis, FOXM1 and MYC protein expression, tumor type (i.e., primary, recurrent, or metastatic; Fig. 2K) and tumor grade (Fig. 2L) also demonstrated statistically significant association with overall survival. In multivariate analysis, however, the most significant survival predictor was CDK4 copy number gain by aCGH in the 38 patient cohort and corresponding CDK4 gene amplification by FISH in the 87-patient cohort. Tumor anatomic site in the 38 patient cohort, and high FOXM1 protein expression and tumor type in the 87-patient cohort were also independently significant in multivariate analyses.

Table 2.

Multivariate analysis of both clinical and genomic variables for independent survival predictors in MPNST patients

aCGH test cohort (N = 36)TMA validation cohort (N = 87)
Parameter   n Median survival Pa HRb Pb n Median survival Pa HRb Pb 
Location Central 17 12 (5–18) 0.052 2.82 (1.15–6.92) 0.024 32 27 (12–39) 0.879   
Peripheral 19 32 (15–112)    55 19 (11–52)    
Tumor type Primary 32 16 (12–52) 0.653   68 30 (16–52) 0.045 1.41 (1.02–1.94) 0.038 
Recurrent 12 (6–30)    12 18 (6–42)    
Metastatic      8 (4–10)    
Grade Low 18 (5–31) 0.429   16 (11–31) 0.026   
High 31 15 (12–32)    78 14 (6–124)    
Metastasis Yes 13 15 (12–32) 0.223        
No 15 10 (6–60)         
Recurrence Yes 20 24 (10–112) 0.248   20 28 (10–72) 0.555   
No 16 15 (12–32)    67 19 (12–37)    
NF1 associated Yes 22 18 (6–52) 0.919   25 33 (11–92) 0.402   
No 14 15 (12–112)    62 17 (12–39)    
Age ≤20 42 (10–52) 0.126   15 83 (15–132) 0.124   
21–30 5 (4–112)    17 26 (6–54)    
31–40 12 15 (6–48)    20 15 (5–142)    
>40 13 12 (4–24)    35 16 (4–140)    
Gender Male 16 12 (10–112) 0.281   39 19 (10–42) 0.468   
Female 20 18 (15–52)    48 28 (13–51)    
CDK4 gain Yes 12 (4–15) 0.021 4.22 (1.43–12.44) 0.009      
No 27 24 (12–112)         
CDK4 amplification Yes      12 10 (5–16) 0.023 2.04 (1.03–4.04) 0.041 
No      67 31 (16–51)    
FOXM1 gain Yes 11 12 (6–17) 0.016        
No 25 32 (12–112)         
FOXM1 expression High      34 15 (8–29) 0.005 1.93 (1.12–3.31) 0.017 
Low      48 39 (14–84)    
NOL1 gain Yes 12 (4–15) 0.034        
No 27 24 (12–112)         
SOX5 gain Yes 11 12 (6–17) 0.024        
No 25 32 (12–112)         
SOX5 expression High      23 18 (11–42) 0.553   
Low      37 14 (9–31)    
MYC gain Yes 16 15 (6–52) 0.459        
No 20 24 (12–112)         
MYC expression High      25 16 (10–30) 0.045   
Low      52 28 (12–54)    
TP53 expression High      27 33 (15–66) 0.126   
Low      54 15 (10–29)    
aCGH test cohort (N = 36)TMA validation cohort (N = 87)
Parameter   n Median survival Pa HRb Pb n Median survival Pa HRb Pb 
Location Central 17 12 (5–18) 0.052 2.82 (1.15–6.92) 0.024 32 27 (12–39) 0.879   
Peripheral 19 32 (15–112)    55 19 (11–52)    
Tumor type Primary 32 16 (12–52) 0.653   68 30 (16–52) 0.045 1.41 (1.02–1.94) 0.038 
Recurrent 12 (6–30)    12 18 (6–42)    
Metastatic      8 (4–10)    
Grade Low 18 (5–31) 0.429   16 (11–31) 0.026   
High 31 15 (12–32)    78 14 (6–124)    
Metastasis Yes 13 15 (12–32) 0.223        
No 15 10 (6–60)         
Recurrence Yes 20 24 (10–112) 0.248   20 28 (10–72) 0.555   
No 16 15 (12–32)    67 19 (12–37)    
NF1 associated Yes 22 18 (6–52) 0.919   25 33 (11–92) 0.402   
No 14 15 (12–112)    62 17 (12–39)    
Age ≤20 42 (10–52) 0.126   15 83 (15–132) 0.124   
21–30 5 (4–112)    17 26 (6–54)    
31–40 12 15 (6–48)    20 15 (5–142)    
>40 13 12 (4–24)    35 16 (4–140)    
Gender Male 16 12 (10–112) 0.281   39 19 (10–42) 0.468   
Female 20 18 (15–52)    48 28 (13–51)    
CDK4 gain Yes 12 (4–15) 0.021 4.22 (1.43–12.44) 0.009      
No 27 24 (12–112)         
CDK4 amplification Yes      12 10 (5–16) 0.023 2.04 (1.03–4.04) 0.041 
No      67 31 (16–51)    
FOXM1 gain Yes 11 12 (6–17) 0.016        
No 25 32 (12–112)         
FOXM1 expression High      34 15 (8–29) 0.005 1.93 (1.12–3.31) 0.017 
Low      48 39 (14–84)    
NOL1 gain Yes 12 (4–15) 0.034        
No 27 24 (12–112)         
SOX5 gain Yes 11 12 (6–17) 0.024        
No 25 32 (12–112)         
SOX5 expression High      23 18 (11–42) 0.553   
Low      37 14 (9–31)    
MYC gain Yes 16 15 (6–52) 0.459        
No 20 24 (12–112)         
MYC expression High      25 16 (10–30) 0.045   
Low      52 28 (12–54)    
TP53 expression High      27 33 (15–66) 0.126   
Low      54 15 (10–29)    

NOTE: Numbers in parenthesis are 95% CI.

aUnivariate Breslow method.

bMultivariate Cox regression.

MPNST is an aggressive malignancy with considerable biological variability and limited response to current therapeutic approaches. This phenomenon is well demonstrated by the clinical heterogeneity of patient outcome in the current study, with overall survival ranging from less than a month to over 13 years. A major impediment to more effective disease management has been a lack of available prognostic or predictive biomarkers. While other sarcomas exhibit characteristic translocations that lead to signature fusion transcripts (29), a definitive set of clinically relevant genomic alterations has been difficult to identify for MPNST, likely because of the large number of gross structural genomic aberrations identified in individual tumors.

Our current study provides one of the first global, high-resolution genomic views of MPNST with independent validation of several biologically relevant genes. Not surprisingly, the overall level of genomic complexity is higher than previously observed by conventional cytogenetic analyses. Identifying the presumably small number of biologically relevant genomic alterations among a large number of secondary changes resulting from genomic instability remains a daunting task. For this reason, it was reassuring that analysis of these highly complex data sets nevertheless identified a number of loci already known to be important in MPNST tumorigenesis and malignant progression. These included losses of NF1 on 17q11.2 and CDKN2A on 9p21, as well as common gains on 7p, 8q, 12, and 17q. However, a relatively larger sample size, the use of a high-resolution genomic platform, and correlation with clinical follow-up data in this study identified a number of novel genes whose copy number gain could serve as clinically useful prognostic markers. In contrast to single gene, high-level amplification that is seen at MYC and EGFR loci in some tumor types, aCGH analysis typically identifies larger genomic regions exhibiting more subtle copy number increases. This may be attributed to both the limited dynamic range of microarray-based technology as well as biological heterogeneity of the tumors themselves.

Interestingly, some of the genome CNAs that we identified by aCGH were discordant when measured at the level of gene transcription and/or protein expression. For example, SOX5 copy number gains were associated with diminished survival times, but increased mRNA and protein levels were not. There are a number of explanations for such discrepancies, including the possibility of translocations that result in biologically meaningful alterations unassociated with increased expression. Although sequencing of SOX5 transcripts in tumors with amplification at this locus was performed in this study, our inability to identify chimeric transcripts by this targeted sequencing approach does not entirely exclude this possibility. Conversely, while MYC protein and RNA expression levels correlated with survival, we did not observe a similar relationship at the genome level, as determined by either FISH or aCGH. This suggests that mechanisms of increased expression at the transcriptional level may be more important for this particular gene. These repeated observations underscore the need for a comprehensive evaluation of all candidate biomarkers at the genome, transcriptome, and protein levels in independent sample sets, particularly as they are being considered as prognostic tools.

For example, in the recent literature, p53 expression has been proposed as an important predictor of survival time in MPNST patients (30, 31). Our data, however, did not confirm this finding at the level of protein expression, gene expression, or genome deletion. Although 50% of MPNSTs had deletions at the TP53 locus in our present study, differential mRNA expression between TP53 deleted and nondeleted tumors was not seen (P > 0.05) and TP53 deletion itself did not stratify MPNST patients by survival. Furthermore, in the second cohort of 87 MPNSTs, p53 immunoreactivity was not statistically associated with survival time. One reason for this discrepancy between our current study and previous studies could reflect the presence of p53 substitution mutations that would obviously not be detected by aCGH but that could affect protein expression and stability. However, the analytical and interpretive techniques applied to IHC in the current study were similar to those of previous reports and as such, this discrepancy is more difficult to explain. In contrast, we did identify recurrent copy number gains involving loci at 12p13.31 and 12p13.33 that were associated with survival on univariate analyses. Although these 2 regions did not remain statistically significant in our primary cohort of 38 MPNST patients by multivariate analysis, we were able to subsequently demonstrate concurrent mRNA and protein overexpression of FOXM1 in this region. In a second larger cohort of 87 MPNSTs, FOXM1 over-expression by IHC was associated with poor survival, a significance which was retained on multivariate analysis. FOXM1 is a transcription factor that is only expressed in proliferating cells and has critical functions in tumor development and progression (32, 33). Expression has been associated with enhanced invasion and metastasis in pancreatic and prostate cancers (32, 34). Increased expression of FOXM1 also correlates with increased tumorigenicity of cultured glioma cells and in human glioblastomas, its expression levels are inversely correlated with patient survival (35). Our current data suggest that increased FOXM1 protein expression is a significant predictor of poor survival in MPNST patients.

Finally, of the numerous candidate loci and genes identified in this study, CDK4 appears to be the most clinically and biologically significant. This gene was initially identified as a region at 12q14.1 that demonstrated recurrent copy number gains associated with patient survival. Of the 12 genes within this locus, only CDK4 demonstrated concordant copy number gains and mRNA overexpression by qRT-PCR. Furthermore, by FISH analysis in an independent patient cohort, we found a significant association between CDK4 amplification (15% of tumors) and decreased survival in a multivariate analysis. Studies have demonstrated amplification of the CDK4 region in many human cancers (36–38), a finding variably associated with tumorigenesis and/or survival. This 12q amplicon has long been recognized for its role in sarcoma tumorigenesis. Interestingly, however, in contrast to what has been reported in dedifferentiated liposarcoma (39), MDM2 coamplification with CDK4 was uncommon in MPNSTs analyzed in this study. To our knowledge, our study is the first to provide specific evidence for a significant association between CDK4 amplification and poor patient survival in MPNSTs. Interestingly, a previous report has demonstrated that the multi-kinase inhibitor sorafenib can inhibit MPNST growth, in part through hypophosphorylation of pRB at CDK4-specific sites (40), suggesting the potential of CDK4 as a therapeutic target and also as a potential predictive biomarker of response to sorafenib therapy.

In conclusion, CDK4 gain/amplification and increased FOXM1 protein expression were the most significant independent predictors for poor survival in MPNST patients after accounting for well established clinicopathologic prognostic variables in MPNSTs. If confirmed in prospective studies, these prognostic and potentially predictive biomarkers may provide clinically useful information for managing patients with this aggressive malignancy.

No potential conflicts of interest were disclosed.

This study is supported in part by Department of Defense (DOD)/Neurofibromatosis Research Program Concept Award W81XWH-06-1-0201 (A. Perry).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Grobmyer
SR
,
Reith
JD
,
Shahlaee
A
,
Bush
CH
,
Hochwald
SN
. 
Malignant peripheral nerve sheath tumor: molecular pathogenesis and current management considerations
J Surg Oncol
2008
;
97
:
340
9
.
2.
Arun
D
,
Gutmann
DH
. 
Recent advances in neurofibromatosis type 1
Curr Opin Neurol
2004
;
17
:
101
5
.
3.
Kar
M
,
Deo
SV
,
Shukla
NK
,
Malik
A
,
DattaGupta
S
,
Mohanti
BK
, et al
Malignant peripheral nerve sheath tumors (MPNST)—clinicopathological study and treatment outcome of twenty-four cases
World J Surg Oncol
2006
;
4
:
55
62
.
4.
Kim
DH
,
Murovic
JA
,
Tiel
RL
,
Moes
G
,
Kline
DG
. 
A series of 397 peripheral neural sheath tumors: 30-year experience at Louisiana State University Health Sciences Center
J Neurosurg
2005
;
102
:
246
55
.
5.
Evans
DG
,
Baser
ME
,
McGaughran
J
,
Sharif
S
,
Howard
E
,
Moran
A
. 
Malignant peripheral nerve sheath tumours in neurofibromatosis 1
J Med Genet
2002
;
39
:
311
4
.
6.
Anghileri
M
,
Miceli
R
,
Fiore
M
,
Mariani
L
,
Ferrari
A
,
Mussi
C
, et al
Malignant peripheral nerve sheath tumors: prognostic factors and survival in a series of patients treated at a single institution
Cancer
2006
;
107
:
1065
74
.
7.
Upadhyaya
M
,
Han
S
,
Consoli
C
,
Majounie
E
,
Horan
M
,
Thomas
NS
, et al
Characterization of the somatic mutational spectrum of the neurofibromatosis type 1 (NF1) gene in neurofibromatosis patients with benign and malignant tumors
Hum Mutat
2004
;
23
:
134
46
.
8.
Bridge
RS
 Jr
,
Bridge
JA
,
Neff
JR
,
Naumann
S
,
Althof
P
,
Bruch
LA
. 
Recurrent chromosomal imbalances and structurally abnormal breakpoints within complex karyotypes of malignant peripheral nerve sheath tumour and malignant triton tumour: a cytogenetic and molecular cytogenetic study
J Clin Pathol
2004
;
57
:
1172
8
.
9.
Plaat
BE
,
Molenaar
WM
,
Mastik
MF
,
Hoekstra
HJ
,
te Meerman
GJ
,
Van Den Berg
E
. 
Computer-assisted cytogenetic analysis of 51 malignant peripheral-nerve-sheath tumors: sporadic vs. neurofibromatosis-type-1-associated malignant schwannomas
Int J Cancer
1999
;
83
:
171
8
.
10.
Frank
D
,
Gunawan
B
,
Holtrup
M
,
Fuzesi
L
. 
Cytogenetic characterization of three malignant peripheral nerve sheath tumors
Cancer Genet Cytogenet
2003
;
144
:
18
22
.
11.
Weir
BA
,
Woo
MS
,
Getz
G
,
Perner
S
,
Ding
L
,
Beroukhim
R
, et al
Characterizing the cancer genome in lung adenocarcinoma
Nature
2007
;
450
:
893
8
.
12.
Paulsson
K
,
Cazier
JB
,
Macdougall
F
,
Stevens
J
,
Stasevich
I
,
Vrcelj
N
, et al
Microdeletions are a general feature of adult and adolescent acute lymphoblastic leukemia: Unexpected similarities with pediatric disease
Proc Natl Acad Sci U S A
2008
;
105
:
6708
13
.
13.
Gorringe
KL
,
Jacobs
S
,
Thompson
ER
,
Sridhar
A
,
Qiu
W
,
Choong
DY
, et al
High-resolution single nucleotide polymorphism array analysis of epithelial ovarian cancer reveals numerous microdeletions and amplifications
Clin Cancer Res
2007
;
13
:
4731
9
.
14.
Mantripragada
KK
,
Spurlock
G
,
Kluwe
L
,
Chuzhanova
N
,
Ferner
RE
,
Frayling
IM
, et al
High-resolution DNA copy number profiling of malignant peripheral nerve sheath tumors using targeted microarray-based comparative genomic hybridization
Clin Cancer Res
2008
;
14
:
1015
24
.
15.
Mantripragada
KK
,
de Stahl
TD
,
Patridge
C
,
Menzel
U
,
Andersson
R
,
Chuzhanova
N
, et al
Genome-wide high-resolution analysis of DNA copy number alterations in NF1-associated malignant peripheral nerve sheath tumors using 32K BAC array
Genes Chromosomes Cancer
2009
;
48
:
897
907
.
16.
Brekke
HR
,
Ribeiro
FR
,
Kolberg
M
,
Agesen
TH
,
Lind
GE
,
Eknaes
M
, et al
Genomic changes in chromosomes 10, 16, and X in malignant peripheral nerve sheath tumors identify a high-risk patient group
J Clin Oncol
2010
;
28
:
1573
82
.
17.
Watson
MA
,
Perry
A
,
Tihan
T
,
Prayson
RA
,
Guha
A
,
Bridge
J
, et al
Gene expression profiling reveals unique molecular subtypes of neurofibromatosis type I-associated and sporadic malignant peripheral nerve sheath tumors
Brain Pathol
2004
;
14
:
297
303
.
18.
Mawrin
C
,
Schulz
S
,
Hellwig-Patyk
A
,
Kirches
E
,
Roessner
A
,
Lendeckel
U
, et al
Expression and function of somatostatin receptors in peripheral nerve sheath tumors
J Neuropathol Exp Neurol
2005
;
64
:
1080
8
.
19.
Nannya
Y
,
Sanada
M
,
Nakazaki
K
,
Hosoya
N
,
Wang
L
,
Hangaishi
A
, et al
A robust algorithm for copy number detection using high-density oligonucleotide single nucleotide polymorphism genotyping arrays
Cancer Res
2005
;
65
:
6071
9
.
20.
Deshmukh
H
,
Yeh
TH
,
Yu
J
,
Sharma
MK
,
Perry
A
,
Leonard
JR
, et al
High-resolution, dual-platform aCGH analysis reveals frequent HIPK2 amplification and increased expression in pilocytic astrocytomas
Oncogene
2008
;
27
:
4745
51
.
21.
Wang
H
,
Wang
H
,
Zhang
W
,
Fuller
GN
. 
Tissue microarrays: applications in neuropathology research, diagnosis, and education
Brain Pathol
2002
;
12
:
95
107
.
22.
Perry
A
,
Kunz
SN
,
Fuller
CE
,
Banerjee
R
,
Marley
EF
,
Liapis
H
, et al
Differential NF1, p16, and EGFR patterns by interphase cytogenetics (FISH) in malignant peripheral nerve sheath tumor (MPNST) and morphologically similar spindle cell neoplasms
J Neuropathol Exp Neurol
2002
;
61
:
702
9
.
23.
Lusis
EA
,
Chicoine
MR
,
Perry
A
. 
High throughput screening of meningioma biomarkers using a tissue microarray
J Neurooncol
2005
;
73
:
219
23
.
24.
Hoischen
A
,
Ehrler
M
,
Fassunke
J
,
Simon
M
,
Baudis
M
,
Landwehr
C
, et al
Comprehensive characterization of genomic aberrations in gangliogliomas by CGH, array-based CGH and interphase FISH
Brain Pathol
2008
;
18
:
326
37
.
25.
Walter
MJ
,
Payton
JE
,
Ries
RE
,
Shannon
WD
,
Deshmukh
H
,
Zhao
Y
, et al
Acquired copy number alterations in adult acute myeloid leukemia genomes
Proc Natl Acad Sci U S A
2009
;
106
:
12950
5
.
26.
Dijkman
R
,
Tensen
CP
,
Jordanova
ES
,
Knijnenburg
J
,
Hoefnagel
JJ
,
Mulder
AA
, et al
Array-based comparative genomic hybridization analysis reveals recurrent chromosomal alterations and prognostic parameters in primary cutaneous large B-cell lymphoma
J Clin Oncol
2006
;
24
:
296
305
.
27.
Dimaras
H
,
Gallie
BL
. 
The p75 NTR neurotrophin receptor is a tumor suppressor in human and murine retinoblastoma development
Int J Cancer
2008
;
122
:
2023
9
.
28.
Veeriah
S
,
Brennan
C
,
Meng
S
,
Singh
B
,
Fagin
JA
,
Solit
DB
, et al
The tyrosine phosphatase PTPRD is a tumor suppressor that is frequently inactivated and mutated in glioblastoma and other human cancers
Proc Natl Acad Sci U S A
2009
;
106
:
9435
40
.
29.
Mertens
F
,
Antonescu
CR
,
Hohenberger
P
,
Ladanyi
M
,
Modena
P
,
D'Incalci
M
, et al
Translocation-related sarcomas
Semin Oncol
2009
;
36
:
312
23
.
30.
Brekke
HR
,
Kolberg
M
,
Skotheim
RI
,
Hall
KS
,
Bjerkehagen
B
,
Risberg
B
, et al
Identification of p53 as a strong predictor of survival for patients with malignant peripheral nerve sheath tumors
Neuro Oncol
2009
;
11
:
514
28
.
31.
Zou
C
,
Smith
KD
,
Liu
J
,
Lahat
G
,
Myers
S
,
Wang
WL
, et al
Clinical, pathological, and molecular variables predictive of malignant peripheral nerve sheath tumor outcome
Ann Surg
2009
;
249
:
1014
22
.
32.
Wang
IC
,
Chen
YJ
,
Hughes
D
,
Petrovic
V
,
Major
ML
,
Park
HJ
, et al
Forkhead box M1 regulates the transcriptional network of genes essential for mitotic progression and genes encoding the SCF (Skp2-Cks1) ubiquitin ligase
Mol Cell Biol
2005
;
25
:
10875
94
.
33.
Laoukili
J
,
Kooistra
MR
,
Bras
A
,
Kauw
J
,
Kerkhoven
RM
,
Morrison
A
, et al
FoxM1 is required for execution of the mitotic programme and chromosome stability
Nat Cell Biol
2005
;
7
:
126
36
.
34.
Chandran
UR
,
Ma
C
,
Dhir
R
,
Bisceglia
M
,
Lyons-Weiler
M
,
Liang
W
, et al
Gene expression profiles of prostate cancer reveal involvement of multiple molecular pathways in the metastatic process
BMC Cancer
2007
;
7
:
64
84
.
35.
Liu
M
,
Dai
B
,
Kang
SH
,
Ban
K
,
Huang
FJ
,
Lang
FF
, et al
FoxM1B is overexpressed in human glioblastomas and critically regulates the tumorigenicity of glioma cells
Cancer Res
2006
;
66
:
3593
602
.
36.
Idbaih
A
,
Criniere
E
,
Marie
Y
,
Rousseau
A
,
Mokhtari
K
,
Kujas
M
, et al
Gene amplification is a poor prognostic factor in anaplastic oligodendrogliomas
Neuro Oncol
2008
;
10
:
540
7
.
37.
Wikman
H
,
Nymark
P
,
Vayrynen
A
,
Jarmalaite
S
,
Kallioniemi
A
,
Salmenkivi
K
, et al
CDK4 is a probable target gene in a novel amplicon at 12q13.3-q14.1 in lung cancer
Genes Chromosomes Cancer
2005
;
42
:
193
9
.
38.
Ragazzini
P
,
Gamberi
G
,
Pazzaglia
L
,
Serra
M
,
Magagnoli
G
,
Ponticelli
F
, et al
Amplification of CDK4, MDM2, SAS and GLI genes in leiomyosarcoma, alveolar and embryonal rhabdomyosarcoma
Histol Histopathol
2004
;
19
:
401
11
.
39.
Italiano
A
,
Bianchini
L
,
Gjernes
E
,
Keslair
F
,
Ranchere-Vince
D
,
Dumollard
JM
, et al
Clinical and biological significance of CDK4 amplification in well-differentiated and dedifferentiated liposarcomas
Clin Cancer Res
2009
;
15
:
5696
703
.
40.
Ambrosini
G
,
Cheema
HS
,
Seelman
S
,
Teed
A
,
Sambol
EB
,
Singer
S
, et al
Sorafenib inhibits growth and mitogen-activated protein kinase signaling in malignant peripheral nerve sheath cells
Mol Cancer Ther
2008
;
7
:
890
6
.