spinalmyxopapillaryependymomasdemonstrate a warburg … · pathway analysis revealed that mpe are...
TRANSCRIPT
Biology of Human Tumors
SpinalMyxopapillary EpendymomasDemonstratea Warburg PhenotypeStephen C. Mack1,2,3, Sameer Agnihotri1,3, Kelsey C. Bertrand1,2,3, Xin Wang1,2,3,David J. Shih1,2,3, Hendrik Witt4,5,6, Nadia Hill1,3, Kory Zayne1,3, Mark Barszczyk1,2,3,Vijay Ramaswamy1,2,3, Marc Remke1,2,3, Yuan Thompson1,2,3, Marina Ryzhova4,5,6,Luca Massimi7,Wieslawa Grajkowska8, Boleslaw Lach9, Nalin Gupta10,William A.Weiss10,Abhijit Guha1,3, Cynthia Hawkins1,2,3, Sidney Croul1,2,3, James T. Rutka1,2,3,Stefan M. Pfister4,5,6, Andrey Korshunov6,11, Melike Pekmezci10, Tarik Tihan10,Joanna J. Philips10, Nada Jabado12, Gelareh Zadeh1,3, and Michael D. Taylor1,2,3
Abstract
Purpose: Myxopapillary ependymoma (MPE) is a distincthistologic variant of ependymoma arising commonly in thespinal cord.Despite an overall favorable prognosis, distantmetas-tases, subarachnoid dissemination, and late recurrences havebeen reported. Currently, the only effective treatment for MPE isgross-total resection.We characterized the genomic and transcrip-tional landscape of spinal ependymomas in an effort to delineatethe genetic basis of this disease and identify new leads for therapy.
Experimental Design: Gene expression profiling was per-formed on 35 spinal ependymomas, and copy number profilingwas done on an overlapping cohort of 46 spinal ependymomas.Functional validation experiments were performed on tumorlysates consisting of assays measuring pyruvate kinase M activity(PKM), hexokinase activity (HK), and lactate production.
Results: At a gene expression level, we demonstrate thatspinal grade II and MPE are molecularly and biologically
distinct. These are supported by specific copy number altera-tions occurring in each histologic variant. Pathway analysisrevealed that MPE are characterized by increased cellularmetabolism, associated with upregulation of HIF1a. Thesefindings were validated by Western blot analysis demonstrat-ing increased protein expression of HIF1a, HK2, PDK1, andphosphorylation of PDHE1A. Functional assays were per-formed on MPE lysates, which demonstrated decreasedPKM activity, increased HK activity, and elevated lactateproduction.
Conclusions:Ourfindings suggest thatMPEmaybedrivenby aWarburg metabolic phenotype. The key enzymes promoting theWarburg phenotype: HK2, PKM2, and PDK are targetable bysmall-molecule inhibitors/activators, and should be consideredfor evaluation in future clinical trials for MPE. Clin Cancer Res; 1–9.�2015 AACR.
IntroductionEpendymoma is an incurable malignancy in 45% of patients.
It can arise within the entire central nervous system and canmanifest in both children and adults. While intracranial ependy-momas predominate in childhood, spinal ependymomas arecommonly observed from adolescence through to adulthood(1). Spinal ependymomas are generally slow-growing tumorsand are divided mainly into grade I tumors, with myxopapillaryhistology, or grade II tumors with more classic histologic features(2). Myxopapillary ependymomas (MPE) are a distinct histologicvariant, arising in specific regions of the spine including the conusmedullaris, cauda equina, or filum terminale (3). Despite anoverall favorable prognosis, and classification as a grade I tumor,MPE have been associated with distant metastases, subarachnoiddisseminations, and late recurrences particularly in the pediatricpopulation (4–10). Currently, the only effective treatment forMPE is complete resection, while the benefits of radiation andchemotherapy remain unclear (11). Furthermore, given the loca-tion of these tumors in the spinal cord, the morbidity associatedwith surgery is high, placing patients at significant risk of incon-tinence, impotence, and paralysis. Compared with other ependy-moma subtypes, the genetic basis of MPE is poorly characterized,thereby hindering the identification of pathways and "actionable"
1Developmental &StemCellBiologyProgram,ArthurandSoniaLabattBrainTumourResearchCentre,TheHospital for SickChildren,Toronto,Ontario, Canada. 2Laboratory Medicine and Pathobiology, Universityof Toronto, Toronto, Ontario, Canada. 3Division of Neurosurgery, Uni-versity of Toronto, Toronto, Ontario, Canada. 4Division of PediatricNeuro-Oncology, German Cancer Research Center (DKFZ), Germany.5Department of Pediatric Oncology, Hematology and Immunology,University of Heidelberg, Heidelberg,Germany. 6German Cancer Con-sortium (DKTK), Heidelberg,Germany. 7Pediatric Neurosurgery,Cath-olic University Medical School, Gemelli Hospital, Rome, Italy. 8Depart-ment of Pathology University of Warsaw, Children's Memorial HealthInstitute University of Warsaw,Warsaw, Poland. 9Division of Anatom-ical Pathology, Department of Pathology and Molecular Medicine,McMaster University, Hamilton General Hospital, Hamilton, Ontario,Canada. 10Departments of Neurology, Pediatrics, Neuro-Pathologyand Neurosurgery, University of California, San Francisco, The HelenDiller Family Cancer Research Building, San Francisco, California.11CCU Neuropathology, German Cancer Research Center (DKFZ),Heidelberg,Germany. 12Departments of Pediatrics and Human Genet-ics, McGill University and theMcGill UniversityHealth Center ResearchInstitute, Montreal, Quebec, Canada.
Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).
Corresponding Author: Michael D. Taylor, The Hospital for Sick Children, Suite1503, 555 University Ave, Toronto, Ontario, M5G 1X8, Canada. Phone: 416-813-7564; Fax: 416-813-6169; E-mail: [email protected]
doi: 10.1158/1078-0432.CCR-14-2650
�2015 American Association for Cancer Research.
ClinicalCancerResearch
www.aacrjournals.org OF1
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
drug targets (12–15). This challenge is further complicated by adifficulty in establishing in vitro and in vivomodels ofMPE. To thisend, we characterized the genomic (n¼ 46) and transcriptional (n¼ 35) landscape of a total of 52 primary spinal ependymomas, inan effort to elucidate themolecular underpinnings ofMPE, and toidentify new leads for rationale targeted therapy.
Materials and MethodsTumor sample isolation and preparation
Clinical samples and data were utilized in accordance withresearch ethics board approval from both The Hospital of SickChildren (Toronto, ON, Canada) and DKFZ. Informed consentwas obtained from all patients in this study. Adult and fetalspine protein samples were purchased from Biochain. Detailedpatient and sample information can be found in the Supple-mentary Table S1.
Copy number data processing and analysisGenomicDNAandRNA from fresh frozen tumorswere isolated
according to the same procedures described by Witt and collea-gues. Genomic DNAwas hybridized to Affymetrix SNP6.0micro-arrays according to manufacturer's instructions and preprocessedaccording to methods described in Witt and colleagues. Mediancentering of copy number probes was performed before summa-rization and visualization using Integrated Genome Viewer(Broad Institute, Cambridge, MA). Significant focal regions ofgain or loss were identified by GISTIC2 (16).
Gene expression data processing and analysisRNA was hybridized to Affymetrix Gene 1.0ST microarrays
according to the manufacturer's instructions. Array data waspreprocessed using the same methods described by Witt andcolleagues. Consensus hierarchical clustering (HCL, R package:ConsensusClusterPlus) was performed using 1,000 genes exhibit-ing the greatest median absolute deviation, and 5,000 genes forconsensus non-negative matrix factorization (R package: NMF).Silhouette analysis was used to evaluate sample membershipfollowing consensus HCL, and SigClust was used to determine
statistical significance of subgroups. A comparison was madebetween consensus HCL and NMF using a Rand Index, andassessed statistically by permutation of sample labels and repe-titionof theRand Index calculation to generate a null distribution.
Pathway analysis of gene expression dataGene set enrichment analysis was performed using gene sets
described in Witt and colleagues, 2011 and visualized usingCytoscape: EnrichmentMap (17, 18). Single sample GSEA wasalso performed (Broad: GenePattern) to evaluate pathways andbiologic samples over-represented in individual samples (19). AWilcoxon rank-sum test was used, with FDR correction (Benja-mini–Hochberg method), to compare the pathways/processesdifferentially activated betweenmyxopapillary and grade II spinalependymoma.
Western blot analysisTumor samples were lysed in PLC lysis buffer containing
protease and phosphatase inhibitors (Sigma-Aldrich). Proteinconcentration was determined using the bicinchoninic acid assay(Thermo Fisher Scientific). Thirty micrograms of protein lysatewere loaded into 10 or 12% SDS-PAGE gels. Proteins were thentransferred onto polyvinylidene difluoride membrane (NENResearch Products) using a semi-dry transfer apparatus (Bio-RadLaboratories). Membranes were blocked in 5% milk TBST or 5%BSA TBST as per manufacturer's instructions for an hour andprobed for varying proteins at 4�C overnight. See SupplementaryTable S2 for dilutions and suppliers. After incubation,membraneswere washed in TBST (3� 10minute washes) and incubated withhorseradish peroxidase–conjugated antibodies against the speciesthe primary antibody was raised against (Bio-Rad Laboratories).Protein detection and quantification was performed by usingChemiluminescence Reagent Plus (PerkinElmer) using the AlphaImager HP imaging system for nonsaturated densitometric anal-ysis and exposure to X-ray film.
Immunohistochemistry stainingA nonoverlapping cohort of 39 spinal ependymomas was
analyzed by immunohistochemistry (IHC) for PKM2 proteinexpression (Schebo Bio) as previously reported (20). Tumorswere assigned a score from 0 to 3 based upon the followingcriteria: 0, � 5% positivity; 1, >5% but < 25% positivity; 2, 25%to 75% positivity; 3, �75% positivity. Because our initialhypothesis was that PKM2 expression is elevated in spinalMPEs, we used a one-sided Wilcoxon rank-sum test to comparethe scoring results in our independent cohort of spinal tumorsanalyzed by IHC.
Hexokinase and pyruvate kinase assayTumor samples were lysed in 100 mL of the following buffer:
50 mmol/L potassium phosphate, 2 mmol/L dithiothreitol(DTT), 2 mmol/L EDTA, and 20 mmol/L sodium fluoride.Tumor homogenate was incubated on ice for 30 minutes,followed by centrifugation at 1,000 � g at 4�C for 10 minutes.Twenty micrograms of fresh lysate was used to measure hexo-kinase activity using the BioVision Hexokinase ColorimetricAssay Kit (Catalog # K789-100). Twenty micrograms of freshlysate was also used to measure pyruvate kinase activity (Cat-alog #K709-100).
Translational Relevance
Myxopapillary ependymoma (MPE) is a distinct tumorentity arising predominantly in the spinal cord. Despite anoverall favorable prognosis, distant metastases and late recur-rences have been reported, especially in the pediatric popu-lation. Currently, the only effective treatment for MPE is acomplete resection. Thus, in an effort to delineate the geneticbasis of MPE, and identify novel therapeutic leads, we charac-terized the genomic and transcriptional landscape of 52 pri-mary spinal ependymomas. In this study, we demonstrate thatMPE may be defined by a "Warburg" phenotype as evidencedby increased protein expression of HIF1a, HK2, PDK1, andPKM2, concordant with elevated activity of numerous meta-bolic processes. These findings were supported functionally bydecreased PKM activity, increased HK activity, and elevatedlactate production. Our findings suggest that MPE may becharacterized by a "Warburg" phenotype providing rationalefor evaluation of glycolysis inhibitors in future clinical trials.
Mack et al.
Clin Cancer Res; 2015 Clinical Cancer ResearchOF2
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
Lactate measurementsLactate measurements of frozen tumor samples were per-
formed according to manufacturer's protocol and normalized tomicrogram lysate (Eton Bioscience).
Statistics relating to Western blots and functional assaysWestern blot analysis, chemiluminescent quantification of
protein, lactate measurements, hexokinase, and pyruvate kinaseassays were performed in triplicate with mean and SEM reported.ANOVA was performed for multiple comparisons with post-Tukey analysis for pairwise comparisons. The Student t test wasused for direct pairwise comparisons. Significancewas establishedas P < 0.05.
Microarray data are deposited on Gene Expression Omnibus(GEO) under the identification number: GSE66787.
ResultsGrade II and myxopapillary spinal ependymoma aremolecularly distinct
Our cohort of 52 spinal ependymomas consisted of 24 tumors,which were histologically classified as MPE, 20 tumors as grade IIependymomas, 1 tumor as grade III ependymoma, and 7 tumorsfor which a histologic diagnosis was unavailable. To delineate thetranscriptional heterogeneity between patients with spinal epen-dymoma, we performed gene expression profiling on 35 primary
tumors (Supplementary Tables S1 and S3). Using two distinctunsupervised consensus clustering approaches, hierarchical clus-tering andnon-negativematrix factorization,wedemonstrate thatgrade II ependymoma and MPE of the spine are transcriptionallydistinct (Fig. 1A–D and Supplementary Fig. S1). We next charac-terized the somatic copynumber landscapeof grade II andMPEbyprofiling 46 primary spinal tumors using the Affymetrix SNP 6.0DNA microarrays. In both grade II and MPE, we determined thatmajority of chromosomal aberrations involved whole chromo-some gains and losses, suggesting aneuploidy, while chromo-somal arm and focal aberrationswere less common (Fig. 2A and Band Supplementary Fig. S2). Despite convergent chromosomalgains and losses, grade II and MPE spinal ependymoma werecharacterized by distinct somatic copy number alterations(SCNA), with grade II ependymomas harboring specific loss ofchr16 and gainof chr12, andwithMPEharboring specific losses ofchr2 and chr12 and gains of chr4, chr9, and chr18 (Fig. 2A and B).In addition, we found that the majority of spinal ependymomaswere characterized by loss of chromosome 22, with an increasetoward homozygous loss in spinal grade II versus MPE (Fig. 2Aand B). In a pattern similar to pediatric ependymoma, focal copynumber alterations were highly infrequent (Supplementary Fig.S2A; Supplementary Tables S4 and S5). The only focal andrecurrent amplification was observed specifically in MPE (2/22cases) encompassing the entire uncharacterized transcriptC15ORF54 (Supplementary Fig. S2A and S2C). In line with
n = 35
108642
0.5
0.4
0.3
0.2
0.1
0.0
Rel
ativ
e ch
ange
in a
rea
unde
r C
DF
cur
ve
Silhouette width 0.80.40.0
Sigclust P value <0.0001
BA
DC
Number of clusters
Grade II Myxopapillary
Rand index = 0.8890756P < 0.0001
HCL NMFM
yxop
apill
ary
Hierarchical clustering Non-negative matrix factorization
MyxopapillaryGrade II
Gra
de II
Figure 1.Grade II and myxopapillary spinalependymomas are transcriptionally distinct. A,unsupervised consensus hierarchical clusteringof 35 primary spinal ependymoma geneexpression profiles generated by AffymetrixGene 1.0ST microarrays. Shown is a rank 2classification. Significance of clustering wasmeasured using SigClust. B, unsupervised non-negative matrix factorization of 35 spinalependymoma gene expression profiles. Shownis a rank 2 classification. Sample overlapbetween consensus HCL and NMF wasmeasured using a Rand index. C, area under thecumulative distribution function curveindicating 2 principal subgroups of spinalependymomas. D, silhouette analysis identifies"core" representative samples within the 2subgroups of spinal ependymoma.
Spinal Myxopapillary Ependymomas Exhibit a Warburg Phenotype
www.aacrjournals.org Clin Cancer Res; 2015 OF3
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
previous studies, nonrecurrent statistically significant gains wereobserved in EGFR, and nonrecurrent chromosomal losses wereobserved in AKAP12, TGIF, and UBB (Supplementary Fig. S2B).We conclude that spinal grade II and MPE are ependymomaentities harboring molecularly distinct transcriptomic and SCNAprofiles.
Spinal myxopapillary ependymoma are characterized byincreased gene expression of metabolic networks
To identify the biologic processes and signaling pathwaysdistinguishing spinal grade II and MPE, we performed gene setenrichment analysis and visualized enriched pathways usingCytoscape EnrichmentMap (14, 17, 18). We found that path-ways involved in photoreceptor development, tight junctionformation, and ciliogenesis/microtubule assembly defined spi-nal grade II ependymomas (Fig. 3A; Supplementary Table S6).In spinal MPE, we discovered an unexpected link to pathwaysinvolving angiogenesis, HIF1a/hypoxia signaling, and cellularmetabolism, which represented 33% of all gene sets, enrichedin the subgroup (Fig. 3A and B and Supplementary Fig. S3A andS3B; Supplementary Table S7). Using single-sample GSEA, wenext attempted to identify patients who harbored elevatedmetabolic gene expression and whether there was a correlationwith demographic parameters. Comparing between spinal MPEand grade II ependymomas, we demonstrated significant over-expression of numerous pathways involving HIF1a/hypoxiasignaling, PI3K/AKT/MTOR signaling, MYC signaling, reactiveoxygen species production, glycolysis, citric acid cycle, mito-chondrial electron transport, and amino acid, vitamin, andlipid metabolism (Fig. 3C). Furthermore, these pathways wereenriched in the youngest patients, who represent the age groupassociated with increased incidence of relapse and metastaticdissemination (Fig. 3C; ref. 7). We conclude that spinal MPEsare characterized by increased gene expression of metabolicnetworks occurring preferentially in the pediatric populationand young adulthood.
Spinalmyxopapillary ependymoma are defined by a "Warburg"phenotype
To validate the metabolic signature observed transcriptionallyin spinal MPE, we performed Western blot analysis coupledwith linear protein quantification by chemiluminescence. Wefirst examined a central metabolic transcription factor, HIF1a,and demonstrated that spinal MPE exhibited increased HIF1aexpression compared with spinal grade II ependymomas andadult normal spinal tissue (Fig. 4A and B and SupplementaryFig. S3A and S3B). These results were supported by increasedprotein expression of pyruvate dehydrogenase lipoamide kinaseisozyme 1 (PDK1), increased hexokinase 2 (HK2) expression, anddecreased hexokinase 1 (HK1; Fig. 4A and B and SupplementaryFig. S4).
Given the lack of established and available cell lines, short-term cultures, and in vivo models of myxopapillary ependy-moma, we used the matched primary samples to analyze theenzymatic activity levels associated with overexpression of"Warburg" signature metabolic proteins. Concordant with anincrease in HK2 protein expression, we observed an increase intotal HK activity specifically in spinal MPE (Fig. 4C). Together,these findings predict a shift towards elevated glycolysis andpossible lactate accumulation, described as a "Warburg" phe-notype (21).
Spinal myxopapillary ependymoma demonstrate a "Warburg"phenotype through elevated PKM2 expression and lactateproduction
The pyruvate dehydrogenase complex catalyzes the overallconversion of pyruvate to acetyl-CoA and carbon dioxide, andthereby links the glycolytic pathway to the tricarboxylic cycle.These enzymes are frequently phosphorylated and inhibited incancers by PDK1 thus promoting the "Warburg" effect. In spinalMPE, we observed a specific increase in phosphorylation ofpyruvate dehydrogenase E1a subunit (PDHE1a), thus validatingthe metabolic networks over-represented in Fig. 3 (Supplemen-tary Fig. S5).
A characteristic feature of tumors exhibiting a "Warburg" phe-notype is ametabolic switch frompyruvate kinasemuscle isozymeisoform M1 to M2 expression. In MPE, we observed a specificincrease in pyruvate kinase muscle isozyme isoform M2 (PKM2)expression compared with total PKM protein levels, providingsupportive evidence of a "Warburg" phenotype (2) (Fig. 5A–C).Using immunohistochemical staining of PKM2 protein, in anonoverlapping cohort of spinal ependymomas (n ¼ 39),we observed a significant enrichment of PKM2 expression inspinal MPE tumors (Fig. 5C). Finally, we observed a decreasein total PKM activity in spinal MPE, which overexpress PKM2protein (Fig. 5D). This is consistent with PKM2 being a less activeisoform, thereby allowing cells to accumulate metabolites forcellular growth and division (2). Finally, we demonstrate thatspinalMPE exhibit increased lactate accumulation consistentwithincreased glycolysis and a "Warburg" phenotype (Fig. 6 andSupplementary Fig. S4).
DiscussionSpinal MPE although having a generally favorable prognosis,
are refractory to radiotherapy, and depend largely upon surgicalresection to reduce the odds of tumor relapse. Furthermore,spinal MPEs in the pediatric population have been shown toexhibit an increased incidence of relapse and metastatic dissem-ination (7). Given the difficulty in establishing in vitro and in vivomodels of spinal MPE, evaluating biologic mechanisms andtargets for therapy have been challenging.
In this report, we have leveraged transciptomic and genomictechnologies to examine a cohort of 52 spinal ependymomas inan effort to delineate the genetic basis of this disease, with anultimate goal of identifying novel treatment modalities. Wedemonstrate that spinal grade II and MPE are histologically,transcriptionally, and genetically distinct tumor entities. Whileboth tumors demonstrate similar patterns of whole-chromo-some loss, suggesting aneuploidy, they are characterized bydisparate genomic alterations with grade II tumors character-ized by loss of chr16 and gain of chr12, and spinal MPE by lossof chr2 and chr12, and gain of chr4, chr9, and chr18. In linewith previous genomic characterizations of ependymoma, focaland recurrent copy number alterations were rare, with theexception of C15ORF54 amplification found exclusively inspinal MPE (12, 14, 22, 23). The only other amplificationencompassing an entire gene, albeit occurring in a single tumor,involved EGFR, which has been shown in previous reports to beamplified and potentially associated with poor clinical out-come in spinal MPE (24). Focal and statistically significantdeletions were also observed in single tumors involvingAKAP12, TGIF, and UBB.
Mack et al.
Clin Cancer Res; 2015 Clinical Cancer ResearchOF4
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
AMyxopapillaryGrade II
chr1chr2chr3chr4chr5chr6chr7chr8chr9chr10chr11chr12chr13chr14chr15chr16chr17chr18chr19chr20chr21chr22
*
*
*
*
*
*
*
0.0005
0.0010
0.0056
0.0138
0.0100
0.0026
0.0052
0.2 0.4 0.60.8 1.00.20.40.60.81.0 0 00.2 0.4 0.60.81.00.20.40.60.81.0 0 0Proportion of samples Proportion of samples
MyxopapillaryGrade IIP value P value
ns
ns
ns
ns
nsns
ns
nsns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
nsns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
ns
Whole-chromosome losses Whole-chromosome gains
B
y rallipa poxyM
II edarG
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1819
2021
22Chr:
GainLoss
Figure 2.Grade II and myxopapillary spinal ependymomas harbor distinct copy number landscapes. A, bar graphs of whole chromosomal gains and losses in grade II versusspinalmyxopapillary ependymomas. Gains are show in red and losses in blue. B, genome-wide viewof copy number alterations in 46 spinal ependymomasgeneratedby Affymetrix SNP6.0 DNA microarrays sub-divided by grade II and myxopapillary spinal ependymoma. Gains are show in red and losses in blue.
Spinal Myxopapillary Ependymomas Exhibit a Warburg Phenotype
www.aacrjournals.org Clin Cancer Res; 2015 OF5
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
Grade II
Myxopapillary
TCA cycle/glycolysis / oxidative metabloism
Cell cycle
Angiogenesis
Golgi vessicle transport
Lyase activity
Embryonic morphogenesis
Response to hypoxia /HIF1-alpha pathway
Ciliogenesis/microtubule assembly
Photoreceptor development
Tight junction formation
FDR<0.25P value <0.01
Pathways in cancer
A
B
33%
67%
Hypoxia, HIF-1a signaling, and cellular metabolism
Other pathways and biologic processes
n = 69
n = 137
CMyxopapillary Grade II
HIF-1alpha signaling hypoxia response
PI3K/AKT/MTOR signaling
MYC targets
Amino acid metabolism
Lipid metabolism
Vitamin metabolsmROS production
Glycolysis
Citric acid cycle
Electron transport chain
Metabolic P
rocesses
FDR < 0.05
Increased pathwayactivity
Decreasedpathwayactivity20
40 60
Age
NA
Figure 3.Pathway analysis identifies over-representation of metabolic gene sets in myxopapillary ependymoma. A, cytoscape enrichment map of significant genesetsdistinguishing grade II versusmyxopapillary spinal ependymomas identifiedbyGSEAand visualized inCytoscape. A statistical significance cutoff ofP <0.01 and FDR< 0.25 was used for the pathway analysis. B, Donut plot demonstrating significant over-representation of pathways and biologic processes in involving hypoxiasignaling and cellular metabolism in myxopapillary spinal ependymoma. C, single sample GSEA demonstrating subgroup-specific over-representation of pathwaysand biologic processes involving cellular metabolism and hypoxia signaling. Bar graph of the ages of patients is shown in the bottom plot.
Mack et al.
Clin Cancer Res; 2015 Clinical Cancer ResearchOF6
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
Spinal grade II ependymomas harbor a variety of NF2 muta-tions, and in our study, were found to exhibit increased homo-zygous or clonal loss of chromosome 22 as compared withspinal MPE (25, 26). In addition to the genomic differences, wedemonstrate that spinal grade II ependymoma and MPE aretranscriptionally distinct. Specifically, spinal grade II ependy-moma are characterized largely by pathways involved in cilio-genesis and microtubule assembly consistent with our previousfindings (14). Conversely, we found that spinal MPE aredefined by upregulation of metabolic networks involvingHIF1a/hypoxia signaling, PI3K/AKT/MTOR signaling, MYCsignaling, reactive oxygen species production, glycolysis, citricacid cycle, mitochondrial electron transport, and amino acid,vitamin, and lipid metabolism. These MPE-specific pathwayswere enriched in younger patients, who may be at greater risk oftumor dissemination. Our transcriptional findings were con-firmed by increased protein expression of HIF1a, HK2, PDK1,and a decrease in HK1. Furthermore, an increase in HK2expression was associated with elevated hexokinase activity,an indication of elevated glycolysis in spinal MPE. Theseproteins may represent potential avenues for drug inhibition
in spinal MPE such as Lonidamine targeting hexokinase activityand Dichloroacetate targeting PDK1 (27).
HIF1a protein is a central mediator of the hypoxic response innormal cells and its expression is regulated predominantly byoxygen-dependent hydroxylation, a modification necessary forproteosomal degradation. In spinal MPE, we observed consistentupregulation of HIF1a transcript despite varied protein stability(Supplementary Fig. S3A and S3B). This suggests that themechan-isms regulating HIF1a protein stability are still active in at least asubset of spinal MPEs.
We also observed a specific increase in protein expression ofPKM2 comparedwith total PKM levels in spinalMPE, ametabolicswitch observed in tumors characterized by a "Warburg" pheno-type (2, 21). Thiswas supported by activity assays demonstrating adecrease in overall PKM activity, associated with potential accu-mulation of metabolites needed for macromolecule and nucle-otide synthesis. PKM2 activators have also been identified, such asTEPP46andDASA58,whichmay represent additional therapeuticleads for treatment of spinalMPE (ref. 28; Supplementary Fig. S6).Our findings demonstrate that targeting tumormetabolism repre-sents a novel therapeutic strategy for treatment of spinal MPEs. It
0.00
0.50
1.00
HIF1a protein
HIF
1a/B
-AC
TIN
rat
io
**
Grade II Myxopapillary0.00
0.50
1.00
1.50
HK1 protein
HK
2/B
-AC
TIN
rat
io*
0.00
0.50
1.00
1.50
HK2 protein
HK
1/B
-AC
TIN
rat
io
**
0.00
0.50
1.00
1.50
PDK1 protein
PD
K1/
B-A
CT
IN r
atio
**
A
B
PDK1
HIF1a
HK2
HK1
Adu
lt sp
ine
Fet
al s
pine
1 2 3 4 5 6 1 2 3 4 5 6
Grade II Myxopapillary
β-ACTIN
Grade II MyxopapillaryGrade II Myxopapillary
Grade II Myxopapillary
Spinal ependymomas
0
50
100
150
200
250
300
Hex
okin
ase
act
ivity
(%
)
**n = 12
Grade II Myxopapillary
C
Figure 4.Validation of pathway analysischaracterizing a metabolic signatureenriched in myxopapillary spinalependymomas. A, Western blotanalysis validation of upregulation ofHIF1a, HK2, PDK1, and decreasedexpression of HK1. B, linearquantification of protein expressionlevels detected relating to proteins in A.C, increased HK2 activity inmyxopapillary spinal ependymomasmeasured from primary tumorlysates as compared with grade IItumors (n ¼ 12).
Spinal Myxopapillary Ependymomas Exhibit a Warburg Phenotype
www.aacrjournals.org Clin Cancer Res; 2015 OF7
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
should also be noted that many of these agents such as loni-damine, dichloroacetate, DASA58, and TEPP46, despite prom-ise in various preclinical cancer models, have only recentlyentered clinical trials and efficacy in patients is still underevaluation.
Together, our findings suggest that spinal MPE may becharacterized by a Warburg phenotype as demonstrated by aspecific increase in tumor lactate production. In addition, thekey enzymes promoting the Warburg phenotype: HK2, PKM2,and PDK are targetable by specific small-molecule inhibitors/activators. This may represent a novel treatment strategy thatshould be evaluated in preclinical studies as potential therapyfor spinal MPEs.
Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.
Authors' ContributionsConception and design: S.C. Mack, A. Guha, A. Korshunov, M.D. TaylorDevelopment of methodology: S.C. Mack, M.D. TaylorAcquisition of data (provided animals, acquired and managed patients, pro-vided facilities, etc.): S.C. Mack, S. Agnihotri, X. Wang, H.Witt, N. Hill, K. Zayne,M. Barszczyk, V. Ramaswamy, Y. Thompson, M Rhyzova, L. Massimi,W. Grajkowska, B. Lach, N. Gupta, W.A. Weiss, C. Hawkins, S. Croul,A.Korshunov,M.Pekmezci, T. Tihan, J.J. Philips,N. Jabado,G.Zadeh,M.D.TaylorAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): S.C. Mack, S. Agnihotri, K.C. Bertrand, X. Wang,D.J. Shih, H. Witt, V. Ramaswamy, M. Remke, Y. Thompson, S.M. Pfister,A. Korshunov, J.J. Philips, M.D. TaylorWriting, review, and/or revision of the manuscript: S.C. Mack, S. Agnihotri,X. Wang, H. Witt, Y. Thompson, N. Gupta, J.T. Rutka, S.M. Pfister,A. Korshunov, M. Pekmezci, T. Tihan, J.J. Philips, N. Jabado, M.D. TaylorAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): S.C. Mack, Y. Thompson, B. Lach, A. Guha,A. Korshunov, M.D. TaylorStudy supervision: J.T. Rutka, M.D. Taylor
Grant SupportM.D. Taylor holds a Canadian Institutes of Health Research Clinician-
Scientist Phase II Award, was a Sontag Foundation Distinguished Scholar, andis supported by The Garron Family Chair in Childhood Cancer Research. M.D.Taylor is also supported by grants from the Cure Search Foundation, TheYounger Foundation, the NIH (R01CA148699, and R01CA159859), The Pedi-atric Brain Tumor Foundation, The Canadian Cancer Society, The Terry FoxResearch Institute, and Brainchild. S.C. Mack and X. Wang are supported byVanier Scholarships from the Canadian Institutes of Health Research.
The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received October 15, 2014; revised March 25, 2015; accepted April 16, 2015;published OnlineFirst May 8, 2015.
0
50
100
150
200
250
300
(mm
ol/L
) L
acat
e/(m
g)
lysa
te
**n = 12
Grade II Myxopapillary
Figure 6.Spinal MPE exhibit increased lactate production consistent with anaerobicglycolysis and a "Warburg" phenotype. Increased lactate production inmyxopapillary spinal ependymomasmeasured fromprimary tumor lysates ascompared with grade II tumors (n ¼ 12).
3
2
1
0
n = 39P = 0.0357
PK
M a
ctiv
ity (%
)
Adu
lt sp
ine
Feta
l spi
ne
PK
M2/
Tota
l PK
M ra
tio
A
C D
B
PK
M IH
C s
core
Figure 5.Spinal MPEs demonstrate increasedPKM2 expression and decreasedoverall PKM activity consistent with a"Warburg" phenotype. A, proteinvalidation of increased PKM2 proteinexpression relative to total PKM1 andPKM2 proteins as demonstrated byWestern blot analysis. B, linearquantification of PKM2 levels relativeto total PKM1/2 demonstrated in A. C,immunohistochemistry for PKM2expression in a representative spinalgrade II and MPE (left). Bar graphcomparing immunohistochemistryscores in a cohort of 39 non-overlapping spinal ependymomas(right). D, decreased PKM activity inmyxopapillary spinal ependymomasmeasured from primary tumor lysatesas compared with grade II tumors(n ¼ 12).
Clin Cancer Res; 2015 Clinical Cancer ResearchOF8
Mack et al.
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
References1. Vera-Bolanos E, Aldape K, Yuan Y, Wu J, Wani K, Necesito-Reyes MJ, et al.
Clinical course and progression-free survival of adult intracranial andspinal ependymoma patients. Neuro Oncol. 2014 Aug 13. [Epub aheadof Print].
2. ChristofkHR, VanderHeidenMG,HarrisMH, RamanathanA,Gerszten RE,Wei R, et al. The M2 splice isoform of pyruvate kinase is important forcancer metabolism and tumour growth. Nature 2008;452:230–3.
3. Louis DN, Ohgaki H, Wiestler OD, CaveneeWK, Burger PC, Jouvet A, et al.The 2007 WHO classification of tumours of the central nervous system.Acta Neuropathol 2007;114:97–109.
4. Mridha A, SharmaM, Sarkar C, Suri V, Rishi A, Garg A, et al. Myxopapillaryependymoma of lumbosacral region with metastasis to both cerebello-pontine angles: report of a rare case. Childs Nerv Syst 2007;23:1209–13.
5. Ilhan A, Furtner J, Birner P, R€ossler K, Marosi C, Preusser M.Myxopapillaryependymoma with pleuropulmonary metastases and high plasma glialfibrillary acidic protein levels. J Clin Oncol 2011;29:7.
6. Higgins G, Smith C, Summers D, Statham P, Erridge S. Myxopapillaryependymomawith intracranialmetastases. Br J Neurosurg 2005;19:356–8.
7. Fassett D, Pingree J, Kestle J. The high incidence of tumor dissemination inmyxopapillary ependymoma in pediatric patients. Report of five cases andreview of the literature. J Neurosurg 2005;102:59–64.
8. Al-Hussaini M, Herron B. Metastasizing myxopapillary ependymoma.Histopathology 2005;46:469–70.
9. Weber DC, Wang Y, Miller R, Villa S, Zaucha R, Pica A, et al. Long-termoutcome of patients with spinal myxopapillary ependymoma: treatmentresults from theMDAndersonCancerCenter and institutions from theRareCancer Network. Neuro Oncol 2014;17:588–95.
10. Fegerl G, Marosi C. Stabilization of metastatic myxopapillary ependy-moma with sorafenib. Rare Tumors 2012;4:e42.
11. Al-Habib A, Al-Radi O, Shannon P, Al-Ahmadi H, Petrenko Y, Fehlings M.Myxopapillary ependymoma: correlation of clinical and imaging featureswith surgical resectability in a series with long-term follow-up. Spinal Cord2011;49:1073–8.
12. Johnson RA, Wright KD, Poppleton H, Mohankumar KM, Finkelstein D,Pounds SB, et al. Cross-species genomicsmatches drivermutations and cellcompartments to model ependymoma. Nature 2010;466:632–6.
13. Wani K, Armstrong TS, Vera-Bolanos E, Raghunathan A, Ellison D, Gil-bertson R, et al. A prognostic gene expression signature in infratentorialependymoma. Acta Neuropathol 2012;123:727–38.
14. Witt H, Mack SC, Ryzhova M, Bender S, Sill M, Isserlin R, et al. Delineationof two clinically and molecularly distinct subgroups of posterior fossaependymoma. Cancer Cell 2011;20:143–57.
15. Mack SC, Witt H, Wang X, Milde T, Yao Y, Bertrand KC, et al. Emerginginsights into the ependymoma epigenome. Brain Pathol 2013;23:206–9.
16. Mermel CH, Schumacher SE, Hill B, Meyerson ML, Beroukhim R, Getz G.GISTIC2.0 facilitates sensitive and confident localization of the targets offocal somatic copy-number alteration in human cancers. Genome Biol2011;12:R41.
17. Merico D, Isserlin R, Stueker O, Emili A, Bader GD. Enrichment map: anetwork-based method for gene-set enrichment visualization and inter-pretation. PLoS ONE 2010;5:e13984.
18. SubramanianA, TamayoP,Mootha VK,Mukherjee S, Ebert BL,GilletteMA,et al. Gene set enrichment analysis: a knowledge-based approach forinterpreting genome-wide expression profiles. Proc Natl Acad Sci U S A2005;102:15545–50.
19. Reich M, Liefeld T, Gould J, Lerner J, Tamayo P, Mesirov JP. GenePattern2.0. Nat Genet 2006;38:500–1.
20. Mukherjee J, Phillips JJ, Zheng S,Wiencke J, Ronen SM, Pieper RO. Pyruvatekinase M2 expression, but not pyruvate kinase activity, is up-regulated in agrade-specific manner in human glioma. PLoS ONE 2013;8:e57610.
21. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the War-burg effect: the metabolic requirements of cell proliferation. Science2009;324:1029–33.
22. Mack SC, Taylor MD. The genetic and epigenetic basis of ependymoma.Childs Nerv Syst 2009;25:1195–201.
23. Taylor MD, Poppleton H, Fuller C, Su X, Liu Y, Jensen P, et al. Radial gliacells are candidate stem cells of ependymoma. Cancer Cell 2005;8:323–35.
24. Mendrzyk F, Korshunov A, Benner A, Toedt G, Pfister S, Radlwimmer B,et al. Identification of gains on 1q and epidermal growth factor receptoroverexpression as independent prognostic markers in intracranial ependy-moma. Clin Cancer Res 2006;12:2070–9.
25. Bettegowda C, Agrawal N, Jiao Y, Wang Y, Wood LD, Rodriguez FJ, et al.Exomic sequencing of four rare central nervous system tumor types.Oncotarget 2013;4:572–83.
26. Ebert C, von Haken M, Meyer-Puttlitz B, Wiestler OD, Reifenberger G,Pietsch T, et al. Molecular genetic analysis of ependymal tumors. NF2mutations and chromosome 22q loss occur preferentially in intramedul-lary spinal ependymomas. Am J Pathol 1999;155:627–32.
27. Tennant DA, Duran RV, Gottlieb E. Targetingmetabolic transformation forcancer therapy. Nat Rev Cancer 2010;10:267–77.
28. Anastasiou D, Yu Y, Israelsen WJ, Jiang JK, Boxer MB, Hong BS, et al.Pyruvate kinase M2 activators promote tetramer formation and suppresstumorigenesis. Nat Chem Biol 2012;8:839–47.
www.aacrjournals.org Clin Cancer Res; 2015 OF9
Spinal Myxopapillary Ependymomas Exhibit a Warburg Phenotype
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650
Published OnlineFirst May 8, 2015.Clin Cancer Res Stephen C. Mack, Sameer Agnihotri, Kelsey C. Bertrand, et al. PhenotypeSpinal Myxopapillary Ependymomas Demonstrate a Warburg
Updated version
10.1158/1078-0432.CCR-14-2650doi:
Access the most recent version of this article at:
Material
Supplementary
http://clincancerres.aacrjournals.org/content/suppl/2015/05/08/1078-0432.CCR-14-2650.DC1Access the most recent supplemental material at:
E-mail alerts related to this article or journal.Sign up to receive free email-alerts
Subscriptions
Reprints and
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's
.http://clincancerres.aacrjournals.org/content/early/2015/06/29/1078-0432.CCR-14-2650To request permission to re-use all or part of this article, use this link
Research. on March 6, 2020. © 2015 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Published OnlineFirst May 8, 2015; DOI: 10.1158/1078-0432.CCR-14-2650