Methylthioadenosine phosphorylase and activated insulin-like growth factor-1 receptor/insulin receptor: potential therapeutic targets in chordoma

Journal of PathologyJ Pathol 2010; 220: 608–617Published online 22 December 2009 in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/path.2679

Original Paper

Methylthioadenosine phosphorylase and activatedinsulin-like growth factor-1 receptor/insulin receptor:potential therapeutic targets in chordoma

Josh Sommer,1* Doha M Itani,2 Kelly C Homlar,3 Vicki L Keedy,4 Jennifer L Halpern,3 Ginger E Holt,3

Herbert S Schwartz,3 Cheryl M Coffin,2 Michael J Kelley5 and Justin MM Cates2

1Chordoma Foundation, Greensboro, NC, USA, and Division of Medical Oncology, Department of Medicine, Duke University Medical Center andDurham Veterans Affairs Medical Center, Durham, NC, USA2Department of Pathology, Vanderbilt University Medical Center, Nashville, TN, USA3Department of Orthopaedic Surgery and Rehabilitation, Vanderbilt Orthopaedic Institute, Nashville, TN, USA4Division of Hematology/Oncology, Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, USA5Division of Medical Oncology, Department of Medicine, Duke University Medical Center and Durham Veterans Affairs Medical Center, Durham, NC,USA

*Correspondence to:Josh Sommer, ChordomaFoundation, P.O. Box 4562,Greensboro, NC 27404, USA,and Division of MedicalOncology, Department ofMedicine, Duke UniversityMedical Center and DurhamVeterans Affairs Medical Center,Durham, NC 27710, USA.E-mail:[email protected]

No conflicts of interest weredeclared.

Received: 20 October 2009Revised: 7 December 2009Accepted: 16 December 2009

AbstractCurrently there is no effective chemotherapy for chordoma. Recent studies report co-expression of insulin-like growth factor-1 receptor (IGF1R) and its cognate ligand in chor-doma, but it is unknown whether this receptor tyrosine kinase is activated in these tumours.Additionally, genetic studies have confirmed frequent deletions of chromosome 9p in chordo-mas, which encompasses the cyclin-dependent kinase inhibitor 2A (CDKN2A) locus. Anothergene in this region, methylthioadenosine phosphorylase (MTAP), is an essential enzyme ofthe purine salvage pathway and has therapeutic relevance because MTAP-deficient cells areparticularly sensitive to inhibitors of de novo purine synthesis. We investigated whetherthese pathways might be potential therapeutic targets for chordoma. Paraffin-embedded tis-sue samples from 30 chordomas were analysed by immunohistochemistry for expression ofthe phosphorylated isoforms of IGF1R or the insulin receptor (pIGF1R/pIR) and selecteddownstream signalling molecules, including BCL2-associated agonist of cell death protein(BAD). Expression of CDKN2A and MTAP proteins was also assessed. Skeletal chondrosar-comas, benign notochordal cell tumours, and fetal notochord were studied for comparison.Phosphorylated IGF1R/IR was detected in 41% of chordomas, together with activated down-stream signalling molecules, and pIGF1R/pIR was absent in benign notochordal cell tumoursand fetal notochord. Thirty-nine per cent of chordomas were negative for MTAP immunore-activity. Patients with pIGF1R/pIR-positive tumours showed significantly decreased mediandisease-free survival in multivariate survival analysis (p = 0.036), whereas phosphorylationof BAD at serine-99 was found to be associated with a favourable prognosis (p = 0.002).Approximately 40% of chordomas demonstrate evidence of activation of the IGF1R/IR sig-nalling pathway or loss of a key enzyme in the purine salvage pathway. Aberrant signallingcascades and disrupted metabolic pathways such as these may represent opportunities fornovel targeted therapeutic approaches for the treatment of chordoma.Copyright 2009 Pathological Society of Great Britain and Ireland. Published by JohnWiley & Sons, Ltd.

Keywords: chordoma; notochord; skull base; sacrum; immunohistochemistry; insulin-like growth factor-1 receptor; insulin receptor; methylthioadenosine phosphorylase; BAD;CDKN2A

Introduction

Chordomas are rare, malignant bone tumours believedto arise from remnants of embryonic notochord [1].Local recurrences are frequent and often multiple,and most patients succumb to their disease within5–10 years of diagnosis, due to involvement of criticaladjacent anatomical structures [2,3]. Complete surgical

resection followed by high-dose radiation therapyoffers the best chance of long-term survival [4,5].Unfortunately, existing chemotherapies are largelyineffective [6,7].

Advances in targeted molecular therapy, coupledwith increased understanding of the molecular patho-genesis of chordoma, offer new possibilities for sys-temic treatment of this aggressive neoplasm. Recent

Copyright 2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.www.pathsoc.org.uk

MTAP and pIGF1R/pIR in chordoma 609

studies have reported co-expression of the insulin-likegrowth factor-1 receptor (IGF1R) and its cognate lig-and in chordoma [8,9]. Inhibition of IGF1R throughantibodies and small molecules is an emerging ther-apeutic strategy for sarcomas and other cancers[10–12]. The presence of phosphorylated and therebyactivated IGF1R can be detected in formalin-fixedand demineralized tumour tissues by immunohisto-chemistry using an antiserum raised against a syn-thetic phospho-peptide derived from the amino acidsequence surrounding the tyrosine 1161 (Y1161) phos-phorylation site of human IGF1R. However, this anti-serum cross-reacts with the homologous phosphory-lation site of the activated human insulin receptor(tyrosine 1185). Since the activation status and clin-ical significance of these receptor tyrosine kinasesin chordoma is uncertain, we examined a series ofchordomas for the presence of phosphorylated IGF1Ror insulin receptor (pIGF1R/pIR) and the phosphory-lated isoforms of key downstream signal transductionmolecules involved in these signalling pathways [13].

Other genetic pathways may also have biologicaland therapeutic significance for chordoma. Homozy-gous or heterozygous deletion of the cyclin-dependentkinase inhibitor 2A and 2B (CDKN2A and CDKN2B)locus on chromosome 9p21 has been demonstratedin up to 70% of chordomas by array comparativegenomic hybridization [14]. One of the other genesin this region is methylthioadenosine phosphorylase(MTAP), an essential enzyme in the purine salvagepathway that is frequently co-deleted with CDKN2Aand CDKN2B in a variety of cancers [15,16]. MTAPdeficiency renders cells entirely dependent on denovo purine synthesis and therefore highly sensitiveto purine synthesis inhibitors [17,18]. Co-deletion ofMTAP and CDKN2A was therefore also examined inchordoma.

Materials and methods

Case retrieval

All cases of chordoma or benign notochordal celltumour (BNCT) diagnosed at Vanderbilt UniversityMedical Center (Nashville, TN, USA) between 1986and 2008 were identified by a computerized search ofthe surgical pathology database. All available haema-toxylin and eosin (H&E)-stained slides were reviewedand chordomas were subtyped as conventional orchondroid type using World Health Organization cri-teria [1]; dedifferentiated chordomas were excludedfrom this study. Twenty-four conventional chordo-mas, six chondroid chordomas, and three BNCTs wereretrieved. All study cases demonstrated co-expressionof S-100 protein and cytokeratin (AE1/AE3), confirm-ing the diagnoses. Patient charts were reviewed torecord pathological data (tumour location, tumour size,and status of surgical resection margins), demographicdata (patient age and sex), and follow-up information

(time to local recurrence, distant metastasis, and deathdue to disease). For comparison, 17 skeletal chon-drosarcomas (including six skull base tumours) andsix human fetal notochord samples were also retrievedfrom the files. The study protocol was approved by theVanderbilt University Institutional Review Board; therequirement for informed consent was waived by thiscommittee.

Tissue microarray and immunohistochemicalanalysis

Twenty cases were represented on a tissue microar-ray (TMA) described previously [19]. Whole sec-tions from an additional ten cases were also avail-able for study. Tissues were fixed in 10% neutral-buffered formalin and demineralized overnight inRDO Rapid Decalcifier Solution (Apex Engineer-ing Products, Aurora, IL, USA). Immunohistochem-istry (IHC) was performed on paraffin-embedded tis-sues using standard techniques with antibodies andconditions detailed in Table 1. Appropriate formalin-fixed and demineralized positive and negative controlsections were included for each antibody and assayrun. IHC stains were scored as positive if any specificstaining was observed in the appropriate subcellularlocation(s).

Statistical analysis

IHC data were compared to clinicopathological find-ings using standard univariate methods. Measure-ments for tumour size and tumour volume were trans-formed by natural logarithm prior to statistical anal-ysis. Disease-free survival was defined as time fromdiagnosis to death or radiographic documentation oflocal recurrence or distant metastasis. Overall survivalwas defined as time from diagnosis to death from com-plications of disease and was censored for death dueto unrelated causes or loss to follow-up. All statisti-cal analyses were performed using R software (v2.9.0)[20]. Univariate analysis of Kaplan–Meier survivalplots was performed using the log-rank test functionfrom the R Survival package [21]. Biomarkers andclinicopathological factors with p values less than orequal to 0.10 were included in Cox proportional haz-ards regression models. Two-tailed p values less than0.05 were considered statistically significant.

Results

Clinicopathological data

The cohort of patients with chordoma consisted of19 males and 11 females (median age 53 years;range 3–77 years). Fourteen tumours involved theclivus/skull base region, 12 tumours arose in thesacrum, and four were located in the mobile spine.Negative surgical margins were achieved in nine ofthe 12 initial surgical resections of tumours involving

J Pathol 2010; 220: 608–617 DOI: 10.1002/pathCopyright 2009 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

610 J Sommer et al

Table 1. Antibodies and immunohistochemistry protocols

Antibody Clone Dilution∗ Antigen retrieval†† Detection system‡‡ Source§ Cytolocalization

MTAP Polyclonal 1 : 100 Citrate EnVision+ ProteinTech CytoplasmicCDKN2A F-12 1 : 100 Citrate EnVision+ Santa Cruz Nuclear or cytoplasmicpIGF1R (Y1161)/ Polyclonal 1 : 200 Proteinase K EnVision+ Abcam Membranous

pIR (Y1185)pAkt (S473) 736E11 1 : 40¶ Citrate LSAB2 Cell Signaling Nuclear or cytoplasmicpMAPK (T202/ Polyclonal 1 : 50 Trilogy LSAB2 Cell Signaling Nuclear or cytoplasmic

Y204)pBAD (S75) Polyclonal 1 : 50 Trilogy EnVision+ Cell Signaling Nuclear or cytoplasmicpBAD (S99) Polyclonal 1 : 100 Citrate EnVision+ Spring Bioscience Nuclear or cytoplasmic

∗ Incubated for 1 h at room temperature unless otherwise noted.† Citrate, 1× Citra buffer, pH 6.0 (Dako, Carpinteria, CA, USA); Proteinase K, pre-diluted (Dako); Trilogy, pH 8.0, pre-diluted (Cell Marque, HotSprings, AR, USA).‡ EnVision+ (Dako); LSAB2 (Dako).§ ProteinTech Group (Chicago, IL, USA); Santa Cruz Biotechnology (Santa Cruz, CA, USA); Abcam Inc (Cambridge, MA, USA); Cell SignalingTechnology (Danvers, MA, USA); Spring Bioscience (Pleasonton, CA, USA).¶ Overnight incubation at 4 ◦C.

the sacrum or mobile spine. The status of the final mar-gins was not documented in one case. One skull basetumour appeared to be resected in contiguity with a rimof normal tissue, including the documentation of mul-tiple, separately submitted, negative tissue ‘margins’taken from around the tumour bed. Two skull basetumours were resected with positive surgical resectionmargins confirmed on microscopic examination. Theadequacy of the resection margins for the four otherskull base resections were not evaluable. In six cases,tumours were biopsied and treated with radiotherapyonly. An additional five cases represented re-excisionsof locally recurrent disease (one sacral tumour withpositive margins, one sacral tumour with negative mar-gins, and three skull base tumours with unevaluableresection margins). Tumour size ranged from 0.5 cmto 35 cm in greatest dimension (median 4.0 cm); thesedata were not available for seven cases. Tumour vol-ume was calculable in 20 cases. Median tumour vol-ume was 12.9 cm3 (range 0.04–3920 cm3). Adjuvantradiation therapy was delivered post-operatively in 11of 25 cases for which this history was available (44%).Median duration of follow-up for this cohort was36 months (range 1–162 months), during which time12 patients (40%) suffered local recurrences, three(10%) developed metastatic disease, and 17 (57%)patients died from complications of disease. Over-all and disease-free survival at 2 years was 82% and58%, respectively. The corresponding rates 5 yearsafter diagnosis were 42% and 35%.

MTAP and CDKN2A expression

MTAP is constitutively expressed in normal cells, butwas not detected by IHC in 11 of 28 (39%) chor-domas, similar to the rate of loss in skeletal chon-drosarcoma (Figure 1 and Table 2). Two of the threeBNCTs studied also showed loss of MTAP staining.As expected, MTAP expression was preserved in fiveof six samples (83%) of human notochord. CDKN2Aexpression was absent in 20 of 27 (74%) chordomas,a significantly larger proportion of cases than MTAP

Table 2. Loss of MTAP expression in chordoma, chondrosar-coma, BNCT, and notochord

Diagnosis

Totalnumberof cases

NumberMTAP-

negative

PercentageMTAP-

negative(%)

Chordoma 22 9 41Chondroid chordoma 6 2 33Chondrosarcoma 16 5 31BNCT 3 2 67Notochord 6 1 17

Table 3. Association between MTAP and CDKN2A expressionin chordoma

CDKN2A status

Negative Positive

MTAP status Negative 10 (37%) 0 (0%)Positive 10 (37%) 7 (26%)

Fisher’s exact test, p = 0.026.

(95% confidence interval for difference in proportions0.09–0.61). All three BNCTs studied were also neg-ative for CDKN2A. In contrast, CDKN2A stainingwas observed in four of six chondrosarcoma samples.Whereas CDKN2A was detected in three notochordsamples obtained from fetuses 24–34 weeks of gesta-tional age, it was not seen in two other samples fromearlier stages of fetal development (12 and 17 weeks’gestational age) (Figure 1).

Ten chordomas lacked both CDKN2A and MTAPstaining, and another ten cases demonstrated loss ofCDKN2A staining with preservation of MTAP expres-sion (Table 3). Conversely, loss of MTAP expres-sion was never observed without concurrent loss ofCDKN2A staining. The association between MTAPand CDKN2A IHC staining was statistically signifi-cant (Fisher’s exact test; p = 0.026).



Figure 1. Representative cases of human fetal notochord, benign notochordal cell tumour (BNCT), conventional chordoma,chondroid chordoma, and skeletal chondrosarcoma (A,D,G,J,M, H&E-stained sections). Immunohistochemical stains formethylthioadenosine phosphorylase (MTAP) (B,E,H,K,N) and cyclin-dependent kinase inhibitor 2A (CDKN2A) (C,F,I,L,O).Note preservation of CDKN2A staining in notochord and chondrosarcoma and loss of MTAP staining in chordoma and benignnotochordal cell tumour

IGF1R/IR pathway activation

Phosphorylated IGF1R/IR was detected in 11 of 27cases (41%) of chordoma and a similar proportion(35%) of chondrosarcomas. In contrast, pIGF1R/pIRwas not detected in any of the BNCT or notochordalsamples tested (Table 4 and Figure 2). That the dif-ferences in the proportions of pIGF1R/pIR-positivecases did not reach statistical significance betweenany of the groups is most likely due to the smallsample sizes available for study. The activation sta-tus of the IGF1R/IR signalling pathway was furtherinvestigated by determining whether phosphorylatedisoforms of downstream signal transduction moleculessuch as v-akt murine thymoma viral oncogene homo-logue (pAkt), mitogen-activated protein kinases 1 and3 (pMAPK), and BCL2-associated agonist of celldeath protein (pBAD) correlated with pIGF1R/pIRimmunoreactivity.

Phosphorylated isoforms of Akt and MAPK weredetected in 22 (79%) and 25 (89%) of 28 chordomas,respectively (Figure 2 and data not shown). AlthoughpAkt was not detected in BNCT, it was present inall five human notochord samples tested (Table 4).

Phosphorylated MAPK was present in similar pro-portions among all diagnostic groups. Activation ofthe pro-apoptotic protein BAD was assessed by IHCdetection of phosphorylation at serine residues S75and S99 (Figure 2). Phosphorylated BAD (S99) wasdetected in 77% of chordomas and in two of fournotochord samples, but not in any of the BNCTs.The proportion of pBAD (S99)-positive chordomaswas significantly greater than in chondrosarcoma andBNCT. In contrast to S99, phosphorylation of BAD atS75 was observed only in a small subset of conven-tional chordomas and one of two BNCTs (Table 4).

Correlation of IGF1R/IR activation withdownstream markers

Fisher’s exact test was used to evaluate potential asso-ciations between immunoreactivity for pIGF1R/pIRand activated downstream signal transduction mole-cules in chordoma. Only pBAD (S75) showed a sta-tistically significant association with pIGF1R/pIR pos-itivity (Table 5). Post-hoc analysis showed a mod-erate correlation coefficient of association betweenpIGF1R/pIR and pBAD (S75) (� = +0.54). Although


612 J Sommer et al

Figure 2. Representative cases of human fetal notochord, benign notochordal cell tumour (BNCT), conventional chordoma,chondroid chordoma, and skeletal chondrosarcoma stained for phosphorylated insulin-like growth factor-1 receptor/insulinreceptor (pIGF1R/pIR), phosphorylated v-akt murine thymoma viral oncogene homologue 1/2/3 (pAkt) and phosphorylatedBCL2-associated agonist of cell death protein, Ser-75 [pBAD (S75)] and Ser-99 [pBAD (S99)]. Note positive staining forpIGF1R/pIR in chordoma. Phosphorylated Akt is not seen in BNCT, but is present in chordoma and fetal notochord. NuclearpBAD (S99) is observed in chordomas and fetal notochord, but not in BNCT. In contrast, pBAD (S75) is detected in only aminority of chordomas and one of two BNCTs

Table 4. Phosphorylated IGF1R/IR and downstream signal transduction molecules in chordoma, chondrosarcoma, and chordoidtissues

Diagnosis pIGF1R/pIR pAkt pMAPK pBAD (S75) pBAD (S99)

Chordoma 10/22 (46%)∗ 18/22 (82%) 19/22 (86%) 8/23 (35%) 16/21 (76%)Chondroid chordoma 1/5 (20%) 4/6 (67%) 6/6 (100%) 1/5 (20%) 4/5 (80%)Chondrosarcoma 6/17 (35%) 9/12 (75%) 10/12 (83%) 0/12 (0%) 5/14 (36%)†

BNCT 0/3 (0%) 0/3 (0%)† 2/3 (67%) 1/2 (50%) 0/3 (0%)†

Notochord 0/6 (0%) 5/5 (100%) 2/4 (50%) 0/2 (0%) 2/4 (50%)

∗ Number of positive cases/total number of cases stained (%).† Significant difference in proportion of positive cases compared with combined conventional and chondroid chordoma (Fisher’s exact test,p < 0.05).

pAkt and pMAPK were detected in all chordomas withconcurrent activation of IGF1R/IR, these associationsdid not reach statistical significance.

Since BAD is directly regulated by the Akt kinase,detection of pBAD in chordoma was also examined inrelation to the presence of pAkt (Table 5). AlthoughpBAD (S75 or S99) was not significantly associatedwith concurrent expression of pAkt, phosphorylationat S75 was detected only in tumours that were positivefor pAkt. Phosphorylated MAPK was co-expressedwith pAkt in 75% of chordomas, but the associ-ation between these two signalling molecules wasnot statistically significant; neither were associationsbetween pMAPK and pBAD (S75) or (S99) (data notshown).

Correlation with clinical outcomes

With the exception of pBAD (S99), none of the mark-ers studied was associated with any of the clinico-pathological parameters recorded (histological sub-type of chordoma, patient age, sex, anatomical site,tumour size, tumour volume, adjuvant radiation ther-apy, surgical procedure performed, or status of finalsurgical resection margins) (data not shown). How-ever, negative staining for pBAD (S99) was notedmore often in re-excised chordomas (2 of 5) andtumours that were only biopsied (3 of 5) comparedwith resection specimens (1 of 15) (Fisher’s exact test,p = 0.028).

Log-rank analysis of Kaplan–Meier survival curvesdemonstrated that patients with pIGF1R/pIR-positive



Table 5. Associations between pIGF1R/pIR and pAkt status and activated downstream signalling molecules

Downstream marker pIGF1R/pIR-negative pIGF1R/pIR-positive p∗ pAkt-negative pAkt-positive p∗

pAkt Negative 5 0 0.12 — — —Positive 11 10 — —

pMAPK Negative 3 0 0.26 2 1 0.11Positive 13 10 4 21

pBAD (S75) Negative 15 5 0.009 5 14 0.28Positive 1 6 0 7

pBAD (S99) Negative 4 2 0.66 2 4 0.57Positive 10 9 3 15

∗ Fisher’s exact test.

Table 6. Log-rank tests for overall and disease-free survival in30 chordoma patients

VariableOverall

survival (p)Disease-freesurvival (p)

Histological diagnosis∗ 0.29 0.48Age (by decade) 0.44 0.35Sex 0.57 0.28Anatomical site† 0.55 0.76Procedure‡ <0.001 0.01Surgical resection margin 0.48 0.14Adjuvant radiation therapy 0.54 0.81Tumour size§ 0.20 0.17Tumour volume¶ 0.37 0.47MTAP 0.22 0.22CDKN2A 0.08 0.03pIGF1R/pIR 0.09 0.01pAkt 0.60 0.98pMAPK 0.07 0.07pBAD (S75) 0.54 0.45pBAD (S99) 0.02 0.002

∗ Conventional versus chondroid chordoma.† Sacral versus clival/skull base versus mobile spine.‡ Resection, biopsy only, or re-excision of locally recurrent disease.§ Greatest tumour dimension <1.0 cm, 1.0–2.4 cm, 2.5–7.5 cm,>7.5 cm.¶ Tumour volume ≤ 1.0 cm3, 1.1–20.0 cm3, >20.0 cm3.

chordomas had significantly shorter disease-free sur-vival intervals than those without evidence ofIGF1R/IR activation (Table 6 and Figure 3). Loss of

CDKN2A expression or negative staining for pBAD(S99) also correlated with decreased disease-free sur-vival in univariate analysis. Only the type of surgicalprocedure performed (biopsy only, resection, or re-excision of locally recurrent disease) and pBAD (S99)status were associated with significant differences inoverall survival intervals. No significant differencesin overall or disease-free survival were detected forcohorts categorized by age, sex, histological subtype,anatomical site of tumour, greatest tumour dimension,tumour volume, adjuvant radiation therapy, or the sta-tus of surgical resection margins. Molecular biomark-ers associated with p values less than 0.1 by log-rankanalysis were subsequently evaluated in Cox propor-tional hazards regression models together with the typeof surgical procedure performed.

Multivariate survival analysis revealed that posi-tive staining for pIGF1R/pIR and negative stainingfor pBAD (S99) were associated with increased risk ofadverse events, independent of the type of surgical pro-cedure performed (Table 7). The difference in mediandisease-free survival for patients with CDKN2A-positive tumours (90 months) and negative tumours(24 months) was not statistically significant upon mul-tivariate analysis. None of the prognostic markersincluded in the regression model for overall survival,including the type of surgical procedure performed,correlated with increased risk of death. Althoughmedian overall survival of the pIGF1R/pIR-negative

Figure 3. Kaplan–Meier disease-free survival curves for chordomas categorized by phosphorylation status of insulin-like growthfactor-1 receptor/insulin receptor (pIGF1R/pIR) (A) and BCL2-associated agonist of cell death protein at Ser-99 [pBAD (S99)] (B).Phosphorylation of IGF1R/IR is associated with a significant decrease in median disease-free survival (17 months versus 75 months;p = 0.01). The absence of IHC staining for pBAD (S99) was also associated with significantly shortened disease-free survival(13 months versus 40 months; p = 0.002)


614 J Sommer et al

Table 7. Multivariate survival analysis of prognostic factors inchordoma

Parameter∗ HR 95% CI p

Overall survivalBiopsy only versus resection 4.63 0.49–43.6 0.18Re-excision versus resection 1.54 0.36–6.61 0.56CDKN2A 0.17 0.02–1.45 0.10pIGF1R/pIR 1.48 0.33–6.60 0.61pMAPK 4.42 0.49–40.1 0.19pBAD (S99) 0.53 0.10–2.98 0.47

Disease-free survivalBiopsy only versus resection 0.73 0.11–4.71 0.74Re-excision versus resection 1.11 0.28–4.35 0.88CDKN2A 0.15 0.02–1.32 0.087pIGF1R/pIR 4.42 1.10–17.8 0.036pMAPK 3.46 0.37–32.4 0.28pBAD (S99) 0.11 0.02–0.67 0.017

HR = hazard ratio; CI = confidence interval.∗ For IHC markers, the reported HR and CI are for positive casescompared with negative cases.

cohort was 2.5-fold longer than that of patients withpIGF1R/pIR-positive chordomas (131 months versus49 months), this difference failed to reach statisti-cal significance in univariate or multivariate survivalanalysis. Similar differences were observed for pBAD(S99) and CDKN2A.

Discussion

Limited knowledge about the molecular pathogenesisof chordoma and lack of effective chemotherapeu-tic agents pose a major challenge in treating patientswhose chordomas recur or progress after surgeryand/or radiation therapy. Since the human notochordinvolutes during late fetal and early postnatal devel-opment, there exists no definitive adult tissue coun-terpart for comparative studies [22]. Chordomas arenow believed to arise from persistent intraosseousnotochordal rests or from benign precursor lesions[23–25]. Understanding the molecular differencesbetween chordomas, BNCTs, and the fetal noto-chord might elucidate molecular events critical tochordomagenesis. Therefore, we examined the expres-sion patterns of several therapeutically relevant pro-teins including MTAP, IGF1R/IR, and selected down-stream signalling molecules in a series of chordomasin comparison to skeletal chondrosarcoma, BNCT, andhuman fetal notochord.

IGF1R/IR and downstream signalling molecules

Phosphorylated IGF1R/IR was detected in 41% ofchordomas, but not in any of the BNCTs or notochordsamples studied. Positive staining for pIGF1R/pIR inchordoma was associated with significantly decreaseddisease-free survival in multivariate survival analysis,which suggests that pIGF1R/pIR is an independentprognostic marker for poor outcome in chordoma and

that aberrant activation of the IGF1R/IR signallingpathway in chordoma is an important determinant ofaggressive biological potential. Tyrosine kinase sig-nalling through the IGF1R/IR results in activation ofthe PI3K/Akt and MAPK signalling pathways, whichstimulate cell proliferation and inhibit apoptosis [26].Phosphorylated isoforms of Akt and MAPK werepresent in most chordoma samples, confirming ourprevious observations [27]. Detection of pAkt andpMAPK in pIGF1R/pIR-negative cases suggests thateither (1) these markers are activated by signal trans-duction pathways other than the IGF1R/IR pathwaysin chordoma, (2) loss of phosphatase and tensin homo-logue (PTEN) expression may deregulate these path-ways in some chordomas [28], or/and (3) the kineticsof the receptor tyrosine kinase signalling response arenot accurately reflected at a single point in time byIHC.

The proportions of pAkt and pMAPK-positive caseswere not significantly different among chordoma,skeletal chondrosarcoma, or notochordal tissue. Pre-liminary evidence suggests that Akt activity is minimalin BNCTs compared with chordoma, but the numberof BNCT samples available for study was limited. Theconsistent detection of pAkt in notochord is interestingand may raise questions regarding the pathogenic roleof Akt in chordoma. However, there are at least threeAkt homologues which appear to have distinct, cell-specific functions [29,30]. Investigation of the activa-tion status of the individual Akt proteins in additionalsamples of human notochord and BNCT would benecessary to further address the significance of thisfinding.

Among other possible mechanisms, inhibition ofapoptosis in response to IGF1R/IR signalling is medi-ated by phosphorylation and inactivation of BAD,a pro-apoptotic member of the BCL2 family [31].The S75 residue of BAD is a known phosphoryla-tion target of ribosomal protein S6 kinases of theMAPK pathway, cAMP-dependent protein kinase A,and possibly Akt, whereas S99 is phosphorylatedprimarily by the Akt kinase [32,33]. Phospho-BAD(S99) was detected in most chordomas, consistentwith the corresponding high prevalence of pAkt inthese tumours. The proportion of pBAD (S99)-positivechordomas was significantly greater compared withBNCT, but not notochord, similar to the results forpAkt (see above). Although phosphorylation of BADat S75 was detected in only a small subset of con-ventional chordomas, staining for pBAD (S75) cor-related with activation of IGF1R/IR. These data,together with those of other investigators, suggestthat pIGF1R/pIR/Akt/mTOR and MAPK-dependentpathways are both active in a subset of chordomas[28,34–36].

These findings also provide evidence that alteredsignalling through BAD may be involved in thedysregulation of apoptosis in chordoma. Whereas itmight be hypothesized that phosphorylation of BADin chordoma is indicative of anti-apoptotic signalling



and a worse prognosis, the opposite relationship wasobserved. The presence of pBAD (S99) was associ-ated with a decreased risk of adverse events in ourseries of chordoma patients. Favourable associationsbetween pBAD and prognosis have been reported pre-viously in patients with breast carcinoma [37]. Perhapsloss of BAD regulatory pathways (possibly by muta-tion or deletion of BAD) render the anti-apoptoticBCL2 pathway independent of normal cellular con-trol mechanisms. In any case, further study of BADand other BCL2 family members may be warranted inchordoma.

CDKN2A and MTAP

Recently, Hallor et al documented loss of the CDKN2Agene locus in 70% of chordomas by array comparativegenomic hybridization [14]. Our IHC results confirmthe absence of CDKN2A protein expression in a sim-ilar proportion of chordomas. CDKN2A is a cyclin-dependent kinase inhibitor that regulates retinoblas-toma protein phosphorylation and blocks progressionof the cell cycle at the critical G1/S checkpoint [38].Consequently, inactivation of CDKN2A is frequentlyan early and critical event during tumourigenesis[38,39]. Loss of CDKN2A in normal cells is associatedwith centrosome dysfunction and genomic instabil-ity, ultimately leading to aneuploidy and malignanttransformation [39,40]. CDKN2A was not detectedin any of the three BNCTs tested, suggesting thatloss of this tumour suppressor gene may be an earlyevent in chordomagenesis, as it is for other cancers.MTAP is frequently co-deleted with CDKN2A in sev-eral different types of malignancies [17,41]. Therefore,we examined the prevalence of MTAP expression inchordoma by IHC. Cytoplasmic MTAP immunoreac-tivity was not detectable in 39% of chordomas, similarto proportions reported for other skeletal sarcomas[15,16].

Interestingly, whereas CDKN2A was detected innotochords greater than 24 weeks’ gestational age,notochordal cells were negative for CDKN2A ear-lier in fetal development. This suggests that CDKN2Aexpression in the human notochord is temporally reg-ulated during development. Expression of the T-boxgenes Tbx2 and Tbx3, which repress CDKN2A tran-scription, may account for the absence of CDKN2Ain the early embryonic notochord [42,43]. Additionalstudy of human notochord sampled at various gesta-tional ages is necessary to test this hypothesis further.

Implications for pharmacological therapy ofchordoma

The loss of MTAP activity in a subset of chordo-mas suggests that this group of patients might benefitfrom chemoselective inhibition of the de novo purinesynthesis pathway. Several anti-metabolites that blockpurine synthesis are currently clinically available andadditional agents are in development [41,44]. There-fore, clinical trials using purine synthesis inhibitors

should be considered for chordoma patients withMTAP-deficient tumours.

Although chordoma patients with CDKN2A-ne-gative tumours demonstrated poor prognosis by uni-variate survival analysis, this association was not sig-nificant in a multivariate regression model. Regard-less, loss of CDKN2A expression in chordoma seemedindicative of a more aggressive clinical course.Although CDKN2A deficiency may not have imme-diate therapeutic relevance, pharmaceutical inhibitorsof cyclin-dependent kinases may be able to neutral-ize the consequences of functional loss of CDKN2A[45,46].

Recently, other investigators have shown that thePI3K/Akt/mTOR pathway is hyperactivated in mostchordomas and that inhibition of PI3K or mTOR cansuppress proliferation of chordoma cells in vitro, sug-gesting that this pathway represents a valuable ther-apeutic target in chordoma [34–36]. That inhibitionof mTOR results in feedback activation of Akt viaIGF1R-dependent mechanisms provides further ratio-nale for simultaneous inhibition of both pathways [47].The increased efficacy of combination chemother-apy is supported by a recent study showing thateight of nine chordoma patients with tumour pro-gression on imatinib alone achieved stable diseasewith addition of an mTOR inhibitor [48]. The presentstudy demonstrates that pIGF1R/pIR in chordoma isa poor prognostic factor and may be a potential ther-apeutic target. In the future, clinical investigation ofIGF1R inhibitors, possibly in combination with mTORinhibitors, may be warranted in the subset of chordomapatients with activated IGF1R/IR.

Conclusion

Aberrant signalling cascades and disrupted metabolicpathways in chordoma may represent opportunitiesfor targeted molecular pharmacotherapy in selectedpatients. In this study, we have presented evidencesupporting the theoretical rationale for novel thera-peutic strategies involving inhibition of the IGF1R/IRsignalling pathway and de novo purine synthesis inchordoma. Pharmaceutical agents that inhibit thesemetabolic pathways are currently being investigatedfor use as anti-neoplastic agents. Inclusion of chor-doma patients in these types of clinical trials shouldbe considered. In the interim, the development ofmodel systems that accurately replicate the dysfunc-tional molecular physiology of chordoma cells in vivowould be useful to further define the potential thera-peutic value of these compounds.

Acknowledgements

We thank Sandy Olson for technical assistance and Les-ley Albert and Jean McClure for editorial assistance. Thisstudy was supported by the Division of Anatomic Pathol-ogy, Department of Pathology, Vanderbilt University MedicalCenter.


616 J Sommer et al

References

1. Mirra J, Nelson S, Della Rocca C, Mertens F. Chordoma. InWorld Health Organization Classification of Tumours. Pathologyand Genetics of Tumours of Soft Tissue and Bone, Fletcher CD,Unni KK, Mertens F (eds). IARC Press: Lyon, 2002; 316–317.

2. Baratti D, Gronchi A, Pennacchioli E, Lozza L, Colecchia M,Fiore M, et al. Chordoma: natural history and results in 28 patientstreated at a single institution. Ann Surg Oncol 2003;10:291–296.

3. McMaster ML, Goldstein AM, Bromley CM, Ishibe N, Parry DM.Chordoma: incidence and survival patterns in the United States,1973–1995. Cancer Causes Control 2001;12:1–11.

4. Munzenrider JE, Liebsch NJ. Proton therapy for tumors of theskull base. Strahlenther Onkol 1999;175(Suppl 2):57–63.

5. Park L, Delaney TF, Liebsch NJ, Hornicek FJ, Goldberg S,Mankin H, et al. Sacral chordomas: impact of high-doseproton/photon-beam radiation therapy combined with or withoutsurgery for primary versus recurrent tumor. Int J Radiat OncolBiol Phys 2006;65:1514–1521.

6. Chugh R, Tawbi H, Lucas DR, Biermann JS, Schuetze SM,Baker LH, et al. Chordoma: the nonsarcoma primary bone tumor.Oncologist 2007;12:1344–1350.

7. Casali PG, Stacchiotti S, Sangalli C, Olmi P, Gronchi A. Chor-doma. Curr Opin Oncol 2007;19:367–370.

8. Mitsuhashi T, Watanabe M, Sasano H, Ono M. The expressionof insulin-like growth factor-1 (IGF-1), IGF-1 receptor andtransforming growth factor-beta in chordoma. Mod Pathol2006;19:6A.

9. Mitsuhashi T, Asanuma H, Hasegawa T. Insulin-like growth fac-tor (IGF)-I and IGF-I receptor (IGF-IR) are consistently expressedin the most of chordomas. Mod Pathol 2008;21:15A–16A.

10. Scotlandi K, Picci P. Targeting insulin-like growth factor 1receptor in sarcomas. Curr Opin Oncol 2008;20:419–427.

11. Samani AA, Yakar S, LeRoith D, Brodt P. The role of the IGFsystem in cancer growth and metastasis: overview and recentinsights. Endocr Rev 2007;28:20–47.

12. Anderson P, Kopp L, Anderson N, Cornelius K, Herzog C,Hughes D, et al. Novel bone cancer drugs: investigationalagents and control paradigms for primary bone sarcomas(Ewing’s sarcoma and osteosarcoma). Expert Opin Invest Drugs2008;17:1703–1715.

13. Chitnis MM, Yuen JSP, Protheroe AS, Pollak M, Macaulay VM.The type 1 insulin-like growth factor receptor pathway. ClinCancer Res 2008;14:6364–6370.

14. Hallor KH, Staaf J, Jonsson G, Heidenblad M, Vult von Stey-ern F, Bauer HCF, et al. Frequent deletion of the CDKN2Alocus in chordoma: analysis of chromosomal imbalancesusing array comparative genomic hybridisation. Br J Cancer2008;98:434–442.

15. Garcıa-Castellano JM, Villanueva A, Healey JH, Sowers R,Cordon-Cardo C, Huvos A, et al. Methylthioadenosine phospho-rylase gene deletions are common in osteosarcoma. Clin CancerRes 2002;8:782–787.

16. Chow WA, Bedell V, Gaytan P, Borden E, Goldblum J, Hicks D,et al. Methylthioadenosine phosphorylase gene deletions are fre-quently detected by fluorescence in situ hybridization in con-ventional chondrosarcomas. Cancer Genet Cytogenet 2006;166:95–100.

17. Chen ZH, Zhang H, Savarese TM. Gene deletion chemoselectiv-ity: codeletion of the genes for p16(INK4), methylthioadenosinephosphorylase, and the alpha- and beta-interferons in human pan-creatic cell carcinoma lines and its implications for chemotherapy.Cancer Res 1996;56:1083–1090.

18. Hori H, Tran P, Carrera CJ, Hori Y, Rosenbach MD, Carson DA,et al. Methylthioadenosine phosphorylase cDNA transfection alterssensitivity to depletion of purine and methionine in A549 lungcancer cells. Cancer Res 1996;56:5653–5658.

19. Fasig JH, Dupont WD, Olson SJ, LaFleur BJ, Cates JM. Steroidhormone receptor and COX-2 expression in chordoma. Am J ClinPathol 2007;128:375–381.

20. R Development Core Team. R: a language and environment forstatistical computing . http://www.R-project.org [4 July 2009].

21. Therneau T, Lumley L. Survival: survival analysis, includingpenalised likelihood. R package version 2.34-1. http://cran.r-project.org/web/packages/survival/index.html [4 July 2009].

22. Horwitz T. The Human Notochord: A Study of its Developmentand Regression, Variations and Pathologic Derivative, Chordoma.Private printing: Horwitz, Indianapolis, 1977; 206.

23. Choi KS, Cohn MJ, Harfe BD. Identification of nucleus pulposusprecursor cells and notochordal remnants in the mouse:implications for disk degeneration and chordoma formation. DevDyn 2008;237:3953–3958.

24. Yamaguchi T, Suzuki S, Ishiiwa H, Ueda Y. Intraosseous benignnotochordal cell tumours: overlooked precursors of classicchordomas? Histopathology 2004;44:597–602.

25. Yamaguchi T, Watanabe-Ishiiwa H, Suzuki S, Igarashi Y, Ueda Y.Incipient chordoma: a report of two cases of early-stage chor-doma arising from benign notochordal cell tumors. Mod Pathol2005;18:1005–1010.

26. Wang Y, Sun Y. Insulin-like growth factor receptor-1 as an anti-cancer target: blocking transformation and inducing apoptosis.Curr Cancer Drug Targets 2002;2:191–207.

27. Fasig JH, Dupont WD, LaFleur BJ, Olson SJ, Cates JMM.Immunohistochemical analysis of receptor tyrosine kinase signaltransduction activity in chordoma. Neuropathol Appl Neurobiol2008;34:95–104.

28. Han S, Polizzano C, Nielsen GP, Hornicek FJ, Rosenberg AE,Ramesh V, et al. Aberrant hyperactivation of Akt and mammaliantarget of rapamycin complex 1 signaling in sporadic chordomas.Clin Cancer Res 2009;15:1940–1946.

29. Dummler B, Hemmings BA. Physiological roles of PKB/Aktisoforms in development and disease. Biochem Soc Trans2007;35:231–235.

30. Gonzalez E, McGraw TE. The Akt kinases: isoform specificity inmetabolism and cancer. Cell Cycle 2009;8:2502–2508.

31. Datta SR, Dudek H, Tao X, Masters S, Fu H, Gotoh Y, et al. Aktphosphorylation of BAD couples survival signals to the cell-intrinsic death machinery. Cell 1997;91:231–241.

32. Protein Modification Site. http://www.phosphosite.org/homeAction.do [2 September 2009].

33. Phospho.ELM Search. http://phospho.elm.eu.org/ [2 September2009].

34. Park MJ, Kassam A, Gardner P, Prevedello DM, Soeda A,Lieberman F, et al. Platelet-derived growth factor stimulateschordoma growth and invasion by activating the PI3K/AKT/mTORsignaling pathway. Neuro Oncol 2008;10:764A.

35. Presneau N, Shalaby A, Idowu B, Gikas P, Cannon SR, Gout I,et al. Potential therapeutic targets for chordoma: PI3K/AKT/TSC1/TSC2/mTOR pathway. Br J Cancer 2009;100:1406–1414.

36. Schwab J, Antonescu C, Boland P, Healey J, Rosenberg A,Nielsen P, et al. Combination of PI3K/mTOR inhibition demon-strates efficacy in human chordoma. Anticancer Res 2009;29:1867–1871.

37. Al-Bazz Y, Underwood J, Brown B, Dobson P. Prognosticsignificance of Akt, phospho-Akt and BAD expression in primarybreast cancer. Eur J Cancer 2009;45:694–704.

38. Rocco JW, Sidransky D. p16(MTS-1/CDKN2/INK4a) in cancerprogression. Exp Cell Res 2001;264:42–55.

39. Berman H, Zhang J, Crawford YG, Gauthier ML, Fordyce CA,McDermott KM, et al. Genetic and epigenetic changes inmammary epithelial cells identify a subpopulation of cells involvedin early carcinogenesis. Cold Spring Harb Symp Quant Biol2005;70:317–327.

40. McDermott KM, Zhang J, Holst CR, Kozakiewicz BK, Singla V,Tlsty TD, et al. p16(INK4a) prevents centrosome dysfunction andgenomic instability in primary cells. PLoS Biol 2006;4:e51.

41. Lubin M, Lubin A. Selective killing of tumors deficient inmethylthioadenosine phosphorylase: a novel strategy. PloS one2009;4:e5735.

42. Takabatake Y, Takabatake T, Takeshima K. Conserved anddivergent expression of T-box genes Tbx2–Tbx5 in Xenopus .Mech Dev 2000;91:433–437.



43. Vance KW, Carreira S, Brosch G, Goding CR. Tbx2 is overex-pressed and plays an important role in maintaining prolifera-tion and suppression of senescence in melanomas. Cancer Res2005;65:2260–2268.

44. Kindler HL, Burris HA, Sandler AB, Oliff IA. A phase IImulticenter study of L-alanosine, a potent inhibitor of adeninebiosynthesis, in patients with MTAP-deficient cancer. Invest NewDrugs 2009;27:75–81.

45. Hirai H, Kawanishi N, Iwasawa Y. Recent advances in thedevelopment of selective small molecule inhibitors for cyclin-dependent kinases. Curr Top Med Chem 2005;5:167–179.

46. DePinto W, Chu XJ, Yin X, Smith M, Packman K, Goelzer P,et al. In vitro and in vivo activity of R547: a potent and selectivecyclin-dependent kinase inhibitor currently in phase I clinical trials.Mol Cancer Ther 2006;5:2644–2658.

47. O’Reilly K, Rojo F, She Q, Solit D, Mills G, Smith D, et al.mTOR inhibition induces upstream receptor tyrosine kinasesignaling and activates Akt. Cancer Res 2006;66:1500–1508.

48. Stacchiotti S, Marrari A, Tamborini E, Palassini E, Virdis E,Messina A, et al. Response to imatinib plus sirolimus in advancedchordoma. Ann Oncol 2009;20:1886–1894.


Documents

Methylthioadenosine phosphorylase and activated insulin-like growth factor-1 receptor/insulin receptor: potential therapeutic targets in chordoma