dysregulated trk signalling is a therapeutic target in cyld defective tumours

ORIGINAL ARTICLE

Dysregulated TRK signalling is a therapeutic target in CYLD defective

tumours

N Rajan1, R Elliott2, O Clewes1, A Mackay2, JS Reis-Filho2, J Burn1, J Langtry3, M Sieber-Blum1,CJ Lord2 and A Ashworth2

1Institute of Human Genetics, University of Newcastle upon Tyne, Newcastle upon Tyne, UK; 2The Breakthrough Breast CancerResearch Centre, The Institute of Cancer Research, London, UK and 3Department of Dermatology, Royal Victoria Infirmary,Newcastle upon Tyne, Newcastle upon Tyne, UK

Individuals with germline mutations in the tumour-suppressor gene CYLD are at high risk of developingdisfiguring cutaneous appendageal tumours, the definingtumour being the highly organised cylindroma. Here, weanalysed CYLD mutant tumour genomes by arraycomparative genomic hybridisation and gene expressionmicroarray analysis. CYLD mutant tumours were char-acterised by an absence of copy-number aberrations apartfrom LOH chromosome 16q, the genomic location of theCYLD gene. Gene expression profiling of CYLD mutanttumours showed dysregulated tropomyosin kinase (TRK)signalling, with overexpression of TRKB and TRKC intumours when compared with perilesional skin. Immuno-histochemical analysis of a tumour microarray showedstrong membranous TRKB and TRKC staining incylindromas, as well as elevated levels of ERK phosphor-ylation and BCL2 expression. Membranous TRKC over-expression was also observed in 70% of sporadic BCCs.RNA interference-mediated silencing of TRKB andTRKC, as well as treatment with the small-moleculeTRK inhibitor lestaurtinib, reduced colony formation andproliferation in 3D primary cell cultures established fromCYLD mutant tumours. These results suggest that TRKinhibition could be used as a strategy to treat tumours withloss of functional CYLD.Oncogene advance online publication, 9 May 2011;doi:10.1038/onc.2011.133

Keywords: cylindroma; tropomyosin receptor kinase;TRK; microarray; comparative genomic hybridization;CYLD

Introduction

Germline truncating mutations in the tumour-suppres-sor gene CYLD have been associated with threedisfiguring hair follicle tumour syndromes: (i) familial

cylindromatosis, (ii) Brooke–Spiegler syndrome and (iii)multiple familial trichoepitheliomas (Bowen et al.,2005). CYLD mutant patients develop three distincttypes of cutaneous tumours, namely cylindromas,spiradenomas and trichoepitheliomas. The presence ofpainful spiradenomas and truncal and genital tumourscan have a significant impact on quality of life, as canthe surgery used to remove lesions as this may culminatein entire scalp removal. As such, development of non-surgical approaches to control tumour burden arerequired (Rajan et al., 2009). Novel therapeuticapproaches using inhibitors of the nuclear factor-kB(NFkB) pathway such as aspirin were proposed whenCYLD was found to encode a ubiquitin hydrolase thatnegatively regulated this pathway (Brummelkamp et al.,2003; Kovalenko et al., 2003; Trompouki et al., 2003).However, initial clinical studies with topical salicylicacid showed a significant response in only a smallproportion of treated tumours (Oosterkamp et al.,2006). Therefore, a more detailed understanding of themolecular dysregulation caused by loss of functionalCYLD is warranted, and this may inform the develop-ment of novel therapeutic approaches.

CYLD encodes a ubiquitin hydrolase that cleaveslysine-63 (K63)-linked ubiquitin chains (Brummelkampet al., 2003; Kovalenko et al., 2003; Trompouki et al.,2003), the specificity of which depends on the presenceof a B-box domain within the CYLD protein (Koman-der et al., 2008). In patients with germline CYLDmutations, the majority (94%) of mutations result inpremature termination codons and predict translatedtruncated proteins that have reduced catalytic activity(Saggar et al., 2008), which in turn is thought to perturbthe regulation of the NFkB pathway. CYLD substrates,including TRAF2, TRAF6 and NEMO, which regulatecanonical NFkB signalling, are regulated by K63ubiquitin tagging, and lack of functional CYLD resultsin constitutively active NFkB signalling (Brummelkampet al., 2003; Kovalenko et al., 2003; Trompouki et al.,2003). CYLD interacts with and negatively regulatesTAK1, reducing TAK1-mediated stimulation of IKKand hence activation of NFkB (Reiley et al., 2007).TAK1 also phosphorylates and activates MKK6 andMKK7, leading to the activation of the p38 and JNKkinase pathways, which may also contribute to disease

Received 3 September 2010; revised 17 February 2011; accepted 24February 2011

Correspondence: Dr N Rajan, Institute of Human Genetics, Uni-versity of Newcastle upon Tyne, Newcastle upon Tyne NE1 3BZ, UK.E-mail: [email protected]

Oncogene (2011) 1–18& 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11

www.nature.com/onc

http://dx.doi.org/10.1038/onc.2011.133

http://www.nature.com/onc

pathogenesis (Reiley et al., 2004; Liu et al., 2006).CYLD negatively regulates BCL3, preventing nuclearentry where it forms a dimer with p50/p52, resulting inthe transcription of genes involved in proliferation,including cyclin-D1 (Massoumi et al., 2006). CYLD wasalso shown to negatively regulate Wnt signalling,recognised to have a role in tumorigenesis, by de-ubiquitinating Dishevelled (Tauriello et al., 2010).Additional roles for CYLD have been suggested in theregulation of immunity, microtubule assembly, calciumion channel regulation and facilitating entry intomitosis. (Reiley et al., 2006; Stokes et al., 2006;Stegmeier et al., 2007; Gao et al., 2008). How perturba-tion of these pathways by loss of functional CYLDcontributes to tumour formation is not yet clear.

Apart from cylindromas, other cancers that lackfunctioning CYLD or have reduced expression ofCYLD include hepatocellular and colonic carcinoma(Hellerbrand et al., 2007), multiple myeloma (Annun-ziata et al., 2007; Jenner et al., 2007; Keats et al., 2007;Demchenko et al., 2010), lung cancer (Zhong et al.,2007), prostate cancer (Kikuno et al., 2008) andmalignant melanoma (Massoumi et al., 2009). In multi-ple myeloma, loss of heterozygosity (LOH) at 16q, thegenomic location of the CYLD gene, is an indicator ofpoor prognosis and is associated with reduced overallsurvival (Jenner et al., 2007). Both deletions involvingthe locus of CYLD resulting in reduced copy number aswell as biallelic events, comprising a 16q deletion withmutations within the coding exons of CYLD in theremaining allele, have been shown (Demchenko et al.,2010). Overexpression of NFkB target genes has beenshown in myeloma consistent with the proposed role ofCYLD (Jenner et al., 2007).

CYLD-deficient mouse models with complete loss ofCYLD expression do not develop spontaneous tumours,but show an increased susceptibility to cancer. Thisincludes an increased susceptibility to chemical carcino-gen-induced cutaneous tumours (Massoumi, 2006) aswell as colorectal tumours following dextran sulphate-induced colitis (Zhang et al., 2006). Mice expressingtruncated CYLD mutations that mimic human muta-tions, by contrast, are not viable beyond a few hoursafter birth (Trompouki et al., 2009). This difference inthe two phenotypes is intriguing and suggests thatfurther work is required to understand genotype–phenotype correlations.

CYLD has been shown to interact with tropomyosinkinase-A (TRKA) and has a role in receptor internalisa-tion and signal transduction (Geetha et al., 2005). TRKreceptors were first recognised to have a tumorigenicrole when an oncogenic TRK fusion protein was foundin a proportion of colonic carcinomas (Martin-Zancaet al., 1986) and subsequently thyroid carcinomas (Buttiet al., 1995). Dysregulation of TRK signalling has alsobeen shown in a number of epithelial tumours (Weer-aratna et al., 2000; Ricci et al., 2001; Lagadec et al.,2009).

Here, we have performed a detailed molecularanalysis of tumours from patients with CYLD muta-tions. This showed upregulation of TRK signalling in

CYLD-defective tumours. We went on to develop athree-dimensional (3D) cell culture model from primarytumour cells to show the potential of therapeuticallytargeting TRK for the treatment of these tumours.

Results

Cylindroma and spiradenomas tumours are similar andare genomically stableFresh frozen tumours from patients with germlineCYLD mutations were microdissected to isolate tumourcells and perilesional control cells. Genomic analysis ofcylindroma and spiradenoma tissue was performed on12 tumours and seven perilesional skin samples using a32K bacterial artificial chromosome (BAC) tiling patharray (Natrajan et al., 2009). This BAC array platformhas been shown to be as robust as, and to have acomparable resolution with, high-density oligonucleo-tide arrays (Coe et al., 2007; Tan et al., 2007;Gunnarsson et al., 2008). Comparative genomic hybri-disation was performed using DNA extracted fromperipheral blood leukocytes from the same patient as thereference. No obvious regional genomic amplification ordeletion was noted in the 12 tumour samples(Figure 1a). Moreover, no difference was seen betweencylindromas (n¼ 7) and spiradenomas (n¼ 5).

Given the apparent absence of genomic copy-numbervariation observed using BAC arrays, we went on toperform high-resolution single-nucleotide polymorph-ism (SNP) typing using a 550K SNP array platform(Illumina, San Diego, CA, USA) on four cylindromatumours and one control sample. This showed a copy-number silent LOH for the entire arm of chromosome16q in all four tumours but no other significant changes(Figure 1b).

The LOH status at the site of the mutation wasassessed in a further 27 tumours used in this study,which was possible as the genotype of each patient wasknown. LOH was assessed using genomic DNA frommicrodissected tumours by using restriction enzymesthat specifically were unable to cut only the mutatedallele in each tumour. PCR amplification was performedusing primers that flanked exon-18, and the PCRproduct was subject to restriction enzyme digestionusing enzymes chosen for pedigree-specific mutations:HpyCH4V (c.2460delc) and SmlI (c.2469þ 1G4A).This showed LOH at the site of the mutation in 75%of the 27 tumours, suggesting the mechanism resultingin LOH to be similar in the majority of tumours.

TRK signalling is dysregulated in cylindroma andspiradenoma tissueGene expression profiling was performed on total RNAextracted from 42 microdissected samples, consisting of19 cylindromas, 9 spiradenomas, 4 trichoepitheliomasand 10 perilesional controls. A bead array platform(Illumina) that is capable of assaying the expression of24 526 transcripts for each sample was used (April et al.,2009). The average signal values from replicate beads

Targeting TRK in CYLD defective tumoursN Rajan et al

2

Oncogene

were summarised, log-transformed and normalisedusing quantile normalisation across all samples (Work-man et al., 2002; Dunning et al., 2008). Unsupervisedclustering analysis of the individual transcriptomes(24 526 probes) by Euclidean distance showed closeclustering of the majority of cylindroma and spiradeno-ma tissue, and distinct clustering of the perilesional skinand trichoepithelioma tissue (Figure 2). This wascorroborated with clustering by an independent techni-que that used multiple bootstrapping to determine therobustness of the clusters. Clustering of transcriptomesusing individual probes (24 526) was used and dissim-ilarity was measured using the correlation method usinga statistical package (pvclust) (Suzuki and Shimodaira,2006). Clusters were maintained; however, because ofthe small numbers of tumours used, statistical signi-ficance was seen between tumours and controls(Po0.01), but not between tumour subsets (Supplemen-tary Figure 3). The unsupervised clustering analysis wasfurther corroborated by a supervised clustering analysis,

allowing class comparison between cylindroma andspiradenoma tumours. Nineteen cylindromas and ninespiradenomas were subject to significance of micro-arrays analysis, which showed no significant differencebetween the tumour types when standard criteria wereapplied (Supplementary Figure 4). (Differentially ex-pressed genes between the two groups had to pass asignificance threshold of Po0.01 and 100 permutationsof the data were analysed.)

Quantitation of differential gene expression wasdetermined using the custom algorithm of the beadarray manufacturer (Illumina custom method; IlluminaInc, 2009). The statistical significance of differential geneexpression between tumours and samples was calculatedas a differential score. This calculation incorporated thedifference in signal values between the two groupscompared, corrected for appropriate negative controlbead values, and was corrected for multiple hypothesistesting using a Bonferroni post test. We used stringentthresholds for filtering genes, with only transcripts that

Figure 1 Cylindroma and spiradenomas tumours are genomically similar and show stability. (a) Genomic DNA was extracted fromseven cylindroma and five spiradenoma microdissected tumours, and competitively hybridised against patient-matched peripherallymphocyte DNA on a 32K BAC tiling path array. No significant copy-number change was seen between tumours and perilesionalskin; an indicative result is shown. (b) To examine for genomic changes at higher resolution, SNP typing was performed using a 550KSNP array in a further 4 cylindromas. LOH was seen across the entire arm of chromosome 16q in all 4 samples, indicated by the redarrow. In panel a, circular binary segmentation (cbs)-smoothed Log2 ratios are plotted on the Y axis against each BAC clone accordingto genomic location on the x-axis. BACs categorised as showing genomic gains or amplification are plotted in green, and thosecategorised as genomic losses in red. BAC, bacterial artificial chromosome; SNP, single-nucleotide polymorphism.


3

Oncogene

were differentially expressed with a P-value of o0.01included for subsequent analysis (n¼ 4492). Given thegenomic and transcriptomic similarity seen in cylindro-mas and spiradenomas, these were pooled and analysedagainst pooled controls to provide enhanced statisticalrobustness.

Transcripts of proteins that were previously noted tobe overexpressed in cylindromas were also present andsupported the enrichment of cylindroma tissue followingmicrodissection. These included numerous laminins(Table 1) (Tunggal et al., 2002), collagen-4 and 7 (Timplet al., 1984; Bruckner-Tuderman et al., 1991), andcytokeratins (Tellechea et al., 1995; Meybehm andFischer, 1997), which were preferentially expressed inthe tumours compared with perilesional epidermis.

Upregulation of NFkB target genes (http://people.bu.edu/gilmore/nf-kb/target/index.html) was present asexpected and served as an internal positive control. Alltumours were compared to all controls, and the NFkBtarget genes that were differentially expressed with aP-value ofo0.01 have been tabulated in Table 1. To detectlow-level changes in gene expression between all tumoursand all control tissue, gene set enrichment analysis wasperformed on data filtered at a threshold of Po0.05(Subramanian et al., 2005). This highlighted multiple genesets that were involved in apoptosis, and genes that werecommon to these sets were found to be members of or

target genes of the NFkB and JNK signalling pathways(Figure 3 and Supplementary Figure 1).

Fold change (FC) analysis (Table 2) showed over-expression of several members of the TRK signallingpathway, namely, TRKC (fold change¼ 4.75� ), neu-rotrophin-3 (fold change¼ 3.00� ) and neurotrophin-4/5 (fold change¼ 3.15� ), in the tumour group whencompared with perilesional skin. This finding wascorroborated by connectivity mapping analysis (Lamb,2006) using Ingenuity Pathways Analysis (IngenuitySystems, Redwood City, CA, USA; www.ingenuity.com). This highlighted the concerted overexpression ofmultiple members of the neurotrophin/TRK signallingpathway in tumour tissue compared with control tissue(P¼ 4.02� 10�3). In addition to the transcripts relatedto the TRK signalling pathway detected by the foldchange analysis above, TRKB, brain-derived neuro-trophic factor (BDNF) and PIK3R1, which encodes aregulatory subunit of phosphatidylinositol-3-kinase, adownstream mediator of TRK signalling, were elevatedin expression (Figure 4). Moreover, TRKA, a previouslydescribed CYLD-interacting protein (Geetha et al.,2005), was found to be downregulated, whereas thecognate ligand nerve growth factor (NGF) was upregu-lated. No obvious changes in copy number were foundat the chromosomal location of the different TRKreceptors and their respective cognate ligands, suggest-

Figure 2 Cylindromas and spiradenomas share similar transcriptomes, and cluster distinctly from trichoepitheliomas and perilesionalcontrol skin. Total RNA was extracted from 32 microdissected tumours and 10 perilesional controls, and the gene expression levels of24 526 transcripts for each sample were assayed. Unsupervised clustering of the 42 transcriptomes was performed, and cylindromas andspiradenomas clustered together, with trichoepitheliomas and perilesional skin clustering separately. No clustering by gender, genotypeor tumour location was seen.


4

Oncogene

ing that genomic amplification or deletion was notdirectly responsible for these changes in transcript level.

TRKB and TRKC proteins are overexpressed in humancylindromas and spiradenomasWe investigated whether the overexpression of TRKBand TRKC observed at the mRNA level was reflected inelevated protein expression in CYLD tumours. Weconstructed a tumour tissue microarray (TMA) ofCYLD-deficient tumours cylindromas, spiradenomas,trichoepitheliomas and perilesional skin, and analysedit by immunohistochemistry. Expression of TRKB andTRKC was elevated in cylindroma, spiradenoma and toa lesser extent in trichoepithelioma tissue compared withperilesional skin (Figure 5a). Strong membranousstaining was apparent in cylindroma cells, whereasdiffuse cytoplasmic staining was noted in overlyingkeratinocytes. TRKC expression was mainly noted overthe basophilic peripheral cells in cylindroma islands,whereas TRKB overexpression was non-discriminatorywithin tumour islands (Figure 5a). Quantification ofmembranous TRK expression was performed on eachcore of the TMA. A scoring scale, which assessed theintensity of membranous TRK staining and the pro-

portion of tumour cells stained, was used (Figures 5band c), and the difference in expression between tumoursand perilesional keratinocytes was found to be statis-tically significant (Po0.05) (Tables 3 and 4). Further-more, cylindroma cells were strongly positive withantibodies against phosphorylated ERK (p44/42), amarker of active TRK signalling, mainly at the peri-phery of each tumour island. The antiapoptotic proteinBCL2, a downstream target of TRK signalling, was alsooverexpressed in cylindroma tissue, showing strongperinuclear staining (Figure 5a).

TRKB and TRKC silencing reduces cell viability andcolony formationTo assess the functional significance of TRKB andTRKC in cylindromas, primary cell cultures wereestablished from fresh tumour samples obtained frompatients who contributed samples used for the molecularprofiling and functional studies (Figure 6a). Cylindromatissue was microdissected from normal perilesionaltissue under a stereomicroscope, subject to enzymaticdigestion to obtain a single-cell suspension and culturedunder standard conditions (5% CO2, 21% O2). Theseprimary cell cultures were used within 4 weeks of surgeryto protect against culture-induced genomic alterations.To validate these cultures, CYLD expression wascharacterised and these cells did not express full-lengthCYLD, which was observed in perilesional fibroblastculture (Figure 6b). Cylindroma cells in culture ex-pressed cytokeratin-6, 14, 17 and smooth-muscle actin,markers that were shown to be present in cylindromatumours in vivo, confirming the purity of the cultures(Figures 6c and d). These primary cells were transducedwith lentiviruses expressing short-hairpin RNAs target-ing TRKB and TRKC. The short-hairpin RNA vectorscarried a green fluorescent protein (GFP) reporterallowing monitoring of delivery and a puromycin geneallowing selection—cells were grown in media contain-ing puromycin for 48 h after transfection, resulting incultures that were approximately 100% GFP-positivewhen analysed. TRKB and TRKC knockdown wasassayed using immunoblotting (Figures 6e and f) andafter 10–14 days in culture, cells were fixed, stained andcolonies were counted. Primary cells with TRKB andTRKC knockdown showed a modest reduction (20–40%; Po0.05) in colony formation (Figure 6g).

3D primary cell cultures of cylindroma cells are sensitiveto TRK inhibitorsThe primary cell cultures described above expressedonly low levels of TRK, an observation made inprevious studies (Lagadec et al., 2009). To establish amore biologically relevant cylindroma culture model, weseeded primary cells on 3D polystyrene tissue culturescaffolds to allow enhanced cell–cell contact (Figure 7a).This caused increased the expression of TRKB andTRKC when compared with matched cells grown instandard two-dimensional (2D) culture (Figures 7b andc). To determine whether TRK signalling was active, weexamined the level of phosphorylated ERK and BCL2after stimulating the cultured cells with the cognate

Table 1 NFkB target genes are overexpressed in cylindroma tumours

Gene Foldchange

Gene name

NFkB target genes that are upregulated in tumour tissue (Po0.01)SAA1 5.92 Serum amyloid-a2BCL2A1 5.74 BCL2-related protein-a1SAA2 4.58 Serum amyloid-a2BDNF 2.72 Brain-derived neurotrophic factorOLR1 4.53 Oxidized low-density lipoprotein (lectin-like)

receptor-1CD83 2.53 CD83 moleculeTIFA 2.54 Traf-interacting protein with forkhead-asso-

ciated domainTF 2.77 TransferrinCXCL10 2.67 Chemokine (c-x-c motif) ligand-10HLA-G 3.44 Major histocompatibility complex, class-I, gRELB 2.36 V-rel reticuloendotheliosis viral oncogene

homologue-bMYB 2.17 V-myb myeloblastosis viral oncogene homo-

logue (avian)

Genes that are upregulated in tumour tissue that have been previouslydescribed to be overexpressed in cylindroma tumours (Po0.01)KRT18 1.74 Keratin-18KRT13 4.6 Keratin-13KRT8 3.71 Keratin-8 pseudogene-9; similar to keratin-8KRT7 2.1 Keratin-7LAMA1 4.14 Laminin, a-1LAMA3 3.08 Laminin, a-3LAMB3 2.79 Laminin, b-3LAMC2 1.86 Laminin, g-2COL7A1 1.94 Collagen, type-VII, a-1COL4A1 1.5 Collagen, type-IV, a-1COL4A2 1.73 Collagen, type-IV, a-2

Abbreviation: NFkB, nuclear factor-kB.Transcripts that were differentially expressed between tumours andperilesional controls were filtered with the criteria of a 4� fold changeand P-value of o0.01. Transcripts that fulfilled these criteria werecompared against known NFkB target genes and tabulated. The purityof tumour microdissection was supported by overexpression oftranscripts in cylindroma tissue described previously in the literature.


5

Oncogene

ligands of the TRK receptors (NGF, BDNF and NT3).Stimulation indeed resulted in increased phosphorylatedERK above basal levels (Figure 7d). The antiapoptoticfactor BCL2 has been shown to be upregulated inresponse to NT3 in oligodendrocyte progenitor cells(Saini et al., 2004). This was recapitulated in our model,where BCL2 expression was increased following stimu-lation with TRK receptor ligands (Figure 7e). Todetermine whether these cultures were dependent onTRK signalling for survival, they were treated withsmall-molecule TRK kinase inhibitors. As control theywere also exposed to salicylic acid, which has beenshown previously to effect survival in CYLD-deficientcells (Brummelkamp et al., 2003). Cells were onlymodestly sensitive to salicylic acid (surviving fraction(SF50) 1.5mM), but were much more so to the pan-TRKinhibitors lestaurtinib (CEP-701) (SF50, 4.57 mM),AG879 (SF50, 4.21 mM) and K252a (SF50, 366 nM)(Figure 8). It seems possible that loss of CYLD catalyticactivity accounts for the TRK dependency. To addressthis, we introduced a cDNA construct expressing CYLDinto primary tumour cells, with the aim of assessing

whether CYLD expression could rescue the phenotypesobserved. However, as is often the case when genes areartificially expressed from cDNA constructs, reintroduc-tion of CYLD cDNA resulted in reduced cell viability,precluding further mechanistic studies (SupplementaryFigure 5).

Membrane-localised TRKC is overexpressed in sporadicBCCsCutaneous appendageal tumours seen in CYLD mutationcarriers share similarities in cytokeratin profiles withsporadic basal cell carcinoma (BCC), which in humansare thought to be also derived from the hair follicle(Donovan, 2009). Dysregulation of several mutualoncogenic pathways are seen in cylindromas and BCCs.Wnt signalling is thought to have an oncogenic role inboth, with expression of nuclear b-catenin shown in bothcylindromas (Tauriello et al., 2010) and human BCCs(Kriegl et al., 2009). Also, the Sonic hedgehog pathway,recognised to depend on such canonical Wnt signalling forcutaneous tumour formation (Hoseong Yang et al., 2008),is an important final common pathway in both tumours.

Figure 3 NFkB target genes are overexpressed in CYLD mutant tumours. The expression levels of 24 526 genes from 32microdissected tumours were pooled and compared against 10 pooled perilesional controls and subject to gene set expression analysis.Sets of genes recognised to modulate apoptosis were found to be overexpressed in the tumour tissue (see also Supplementary Figure 1).A heat map of the members of one set (HSA04210_APOPTOSIS) is shown here, with overexpressed genes shown in red and under-expressed genes in blue. The overexpressed genes included several NFkB target genes. NFkB, nuclear factor-kB.


6

Oncogene

Table 2 Altered gene expression in a human CYLDtrunc/trunc model

Gene Fold change Gene name

FGF3 8.38 Fibroblast growth factor-3 (murine mammary tumour virus integration site (v-int-2) oncogene homologue)(FGF3), mrna

APIN 6.74 Apin protein (APIN), mrnaABTB2 5.65 Ankyrin repeat and BTB (POZ) domain-containing-2 (ABTB2), mrnaIQCA 5.53 IQ motif containing with AAA domain (IQCA), mrnaSAA1 5.13 Serum amyloid-A1 (SAA1), transcript variant-2, mrnaRPRM 5.07 Reprimo, TP53-dependent G2 arrest mediator candidate (RPRM), mrnaRARB 5.06 Retinoic acid receptor, b (RARB), transcript variant-1, mrnaCOL9A2 4.97 Collagen, type-IX, a-2 (COL9A2), mrnaUSP6 4.84 Ubiquitin-specific peptidase-6 (Tre-2 oncogene) (USP6), mrnaIL17RB 4.83 Interleukin-17 receptor-B (IL17RB), transcript variant-2, mrnaRARB 4.82 Retinoic acid receptor, b (RARB), transcript variant-2, mrnaC1ORF61 4.82 Chromosome-1 open reading frame 61 (c1orf61), mrnaBCL2A1 4.81 BCL2-related protein-A1 (BCL2A1), mrnaREPS2 4.76 RALBP1-associated Eps domain-containing-2 (REPS2), mrnaCPA6 4.70 Carboxypeptidase-A6 (CPA6), mrnaNTRK3 4.66 Neurotrophic tyrosine kinase, receptor, type-3 (NTRK3), transcript variant-3, mrnaRAB6B 4.64 RAB6B, member RAS oncogene family (RAB6B), mrnaLRRN5 4.59 Leucine-rich repeat neuronal-2 (LRRN2), transcript variant-1, mrnaFLJ45803 4.58 FLJ45803 protein (FLJ45803), mrnaARNT2 4.51 Aryl-hydrocarbon receptor nuclear translocator-2 (ARNT2), mrnaEDAR 4.37 Ectodysplasin-A receptor (EDAR), mrnaKIAA1622 4.34 KIAA1622 (KIAA1622), transcript variant-1, mrnaMIA 4.34 Melanoma-inhibitory activity (MIA), mrnaCLUL1 4.30 Clusterin-like-1 (retinal) (CLUL1), transcript variant-2, mrnaDACT2 4.30 Dapper, antagonist of b-catenin, homologue-2 (Xenopus laevis) (DACT2), mrnaADM2 4.25 Adrenomedullin-2 (ADM2), mrnaSNX22 4.15 Sorting nexin-22 (SNX22), mrnaADAMTS18 4.12 ADAM metallopeptidase with thrombospondin type-1 motif, 18 (ADAMTS18), transcript variant-1, mrnaLOC130940 4.10 Hypothetical protein BC015395 (LOC130940), mrnaCOL22A1 4.06 Collagen, type-XXII, a-1 (COL22A1), mrnaMMP7 4.02 Matrix metallopeptidase-7 (matrilysin, uterine) (MMP7), mrnaENPP6 4.00 Ectonucleotide pyrophosphatase/phosphodiesterase-6 (ENPP6), mrnaC6ORF85 0.25 Chromosome-6 open reading frame 85 (c6orf85), mrnaGRHL3 0.25 Grainyhead-like-3 (Drosophila) (GRHL3), transcript variant-1, mrnaKRT10 0.24 Keratin-10 (epidermolytic hyperkeratosis; keratosis palmaris et plantaris) (KRT10), mrnaZD52F10 0.24 Dermokine (DMKN), transcript variant-2, mrnaTLCD1 0.24 TLC domain-containing-1 (TLCD1), mrnaMUC15 0.23 Mucin-15, cell-surface-associated (MUC15), mrnaSLC19A3 0.23 Solute carrier family-19, member-3 (SLC19A3), mrnaRP11-19J3.3 0.23 Centromere protein-P (CENPP), mrnaCPA4 0.23 Carboxypeptidase-A4 (CPA4), mrnaLGALS7 0.23 Lectin, galactoside-binding, soluble, 7 (galectin-7) (LGALS7), mrnaALDH3A1 0.22 Aldehyde dehydrogenase-3 family, member-a1 (ALDH3A1), mrnaXKRX 0.22 XK, Kell blood group complex subunit-related, X-linked (XKRX), mrnaBNIPL 0.22 BCL2/adenovirus E1B 19-kDa interacting protein like (BNIPL), transcript variant-1, mrnaALDH3B2 0.21 Aldehyde dehydrogenase-3 family, member-B2 (ALDH3B2), transcript variant 2, mrnaANXA8 0.21 Annexin-A8-like-2 (ANXA8L2), mrnaLOC144501 0.20 Hypothetical protein LOC144501 (LOC144501), mrnaCST6 0.20 Cystatin-E/M (CST6), mrnaMICALCL 0.20 MICAL C-terminal like (MICALCL), mrnaHOP 0.20 Homeodomain-only protein (HOP), transcript variant-3, mrnaMAP2 0.20 Microtubule-associated protein-2 (MAP2), transcript variant-2, mrnaGGT6 0.20 g-Glutamyltransferase-6 homologue (rat) (GGT6), mrnaPSORS1C2 0.20 Psoriasis susceptibility-1 candidate-2 (PSORS1C2), mrnaDCT 0.20 Dopachrome tautomerase (dopachrome d-isomerase, tyrosine-related protein-2) (DCT), mrnaSERPINB13 0.20 Serpin peptidase inhibitor, clade-B (ovalbumin), member-13 (SERPINB13), mrnaMAP2 0.20 Microtubule-associated protein-2 (MAP2), transcript variant-1, mrnaTMEM79 0.19 Transmembrane protein-79 (TMEM79), mrnaFGFR3 0.19 Fibroblast growth factor receptor-3 (achondroplasia, thanatophoric dwarfism) (FGFR3), transcript variant-1, mrnaC14ORF29 0.19 Abhydrolase domain containing-12B (ABHD12B), transcript variant-2, mrnaDSG1 0.19 Desmoglein-1 (DSG1), mrnaKRT1 0.19 Keratin-1 (epidermolytic hyperkeratosis) (KRT1), mrnaTRIM7 0.19 Tripartite motif-containing-7 (TRIM7), transcript variant-6, mrnaSLC39A2 0.18 Solute carrier family-39 (zinc transporter), member-2 (SLC39A2), mrnaFLJ10847 0.17 Solute carrier family-47, member-1 (SLC47A1), mrnaCST6 0.17 Cystatin-E/M (CST6), mrnaLOC342897 0.16 Similar to F-box only protein-2 (LOC342897), mrnaATP13A4 0.16 ATPase type-13A4 (ATP13A4), mrnaGPR115 0.15 G-protein-coupled receptor-115 (GPR115), mrna


7

Oncogene

In a murine model that overexpresses GLI1 (Nilssonet al., 2000), tumours that had histological appearancesthat resembled both human BCCs and cylindroma wereseen. Diffuse light expression of TRK receptors haspreviously been noted in a small series of BCCs using apan-TRK antibody (Chen-Tsai et al., 2004). Therefore,we went on to investigate if the membranous TRKB andTRKC overexpression we observed in CYLD mutanttumours was a feature that extended to sporadic skincancers (Table 5). We examined the expression of TRKreceptors in a human skin cancer TMA that includedsamples of squamous cell carcinoma (SCC; n¼ 19),malignant melanoma (MM; n¼ 5) and BCC (n¼ 23),using TRKA-, TRKB- and TRKC-specific antibodies.Interestingly, membranous TRKC was overexpressed in70% of the BCC (Figure 9) and 5% of SCC tumours butnot in the melanoma samples. In some samples, TRKCstaining was strongest at the invasive front where thetumour cells apposed stroma (Supplementary Figure 2).No membranous TRKA staining was seen; however,there was some perinuclear staining, a feature associatedwith activated TRKA signalling (Wu et al., 2007), whichwas seen in 39% of cases of BCC and 16% of cases ofSCC. TRKB staining was seen in the cytoplasm in 13% ofBCC cases only, where it was seen in a few cells, but not atthe cell membrane.

Discussion

CYLD mutation carriers face repeated, disfiguringsurgery to control tumour burden, and in a study of

two large pedigrees, up to one in four affected patientsunderwent total scalp removal (Rajan et al., 2009).Painful eccrine spiradenomas, development of conduc-tive deafness caused by occlusive tumours in theear canal and the risk of malignant transformation(Gerretsen et al., 1993) justify the development of non-surgical alternatives for this patient group. Here, weperformed an unbiased approach using genomic andtranscriptomic analysis of CYLD-defective tumours.We showed that cylindroma and spiradenoma tumourswere genomically similar and stable, with LOH at 16qfound in the majority of the tumours in this study. Thegenomic stability seen across 16 tumours could suggesthomogeneity in the mechanisms of pathogenesis andpotentially in the response to therapeutic agents. More-over, it seems possible that genomically stable tumoursmay be less liable to the acquisition of resistancemechanisms that arise through genomic diversity. Wedetected the upregulation of transcripts of TRKB andTRKC, and their cognate ligands NT3/NT4 and BDNFin cylindroma and spiradenoma samples. This was ofparticular interest as CYLD, a de-ubiquitinase withK63-linked ubiquitin specificity, has previously beenshown to de-ubiquitinate TRKA and inhibit its inter-nalisation from the cell membrane and subsequentdownstream signalling through the phosphatidylinosi-tol-3-kinase pathway (Geetha et al., 2005). Further-more, recent studies have shown that BDNF is an NFkBtarget gene that may afford a neuroprotective role inastrocytes, supporting the finding that this TRK ligandis overexpressed by the cylindroma cells resulting in anautocrine loop (Saha et al., 2006). Validation of protein

Table 2 Continued

Gene Fold change Gene name

UNC93A 0.15 Unc-93 homolog-A (Caenorhabditis elegans) (UNC93A), mrnaSBSN 0.15 Suprabasin (SBSN), mrnaLOC144501 0.14 Hypothetical protein LOC144501 (LOC144501), mrnaTYRP1 0.14 Tyrosinase-related protein-1 (TYRP1), mrnaGSDM1 0.14 Gasdermin-1 (GSDM1), mrnaLOC130576 0.13 Hypothetical protein LOC130576 (LOC130576), mrnaFLJ21511 0.13 Hypothetical protein FLJ21511 (FLJ21511), mrnaC3ORF52 0.13 Chromosome-3 open reading frame 52 (c3orf52), mrnaWFDC5 0.12 WAP four-disulfide core domain-5 (WFDC5), mrnaSLC5A1 0.12 Solute carrier family-5 (sodium/glucose co-transporter), member-1 (SLC5A1), mrnaMLANA 0.12 Melan-A (MLANA), mrnaTGM5 0.11 Transglutaminase-5 (TGM5), transcript variant-1, mrnaCYP3A5 0.10 Cytochrome-P450, family-3, subfamily-A, polypeptide-5 (CYP3A5), mrnaAADACL2 0.10 Arylacetamide deacetylase-like-2 (AADACL2), mrnaBPIL2 0.10 Bactericidal/permeability-increasing protein-like-2 (BPIL2), mrnaCDSN 0.10 Corneodesmosin (CDSN), mrnaK5B 0.10 Keratin-5b (K5B), mrnaRDH12 0.09 Retinol dehydrogenase-12 (all-trans/9-cis/11-cis) (RDH12), mrnaDSC1 0.08 Desmocollin-1 (DSC1), transcript variant-Dsc1b, mrnaTGM3 0.07 Transglutaminase-3 (E-polypeptide, protein-glutamine-g-glutamyltransferase) (TGM3), mrnaPPP2R2C 0.07 Protein phosphatase-2 (formerly-2A), regulatory subunit-B, g-isoform (PPP2R2C), transcript variant-2, mrnaIL1F10 0.07 Interleukin-1 family, member-10 (y) (IL1F10), transcript variant-2, mrnaCA12 0.07 Carbonic anhydrase-XII (CA12), transcript variant-1, mrnaCCL27 0.07 Chemokine (C-C motif) ligand-27 (CCL27), mrnaPLA2G4F 0.07 Phospholipase-A2, group-IVF (PLA2G4F), mrnaK6IRS3 0.06 Keratin-73 (KRT73), mrna

Transcripts that were differentially expressed between tumours and perilesional controls were filtered with the criteria of a 4� fold change andP-value of less than 0.01. Overexpressed transcripts are shown in red and underexpressed transcripts are shown in green.


8

Oncogene

expression using TRK-specific antibodies showed strongmembranous expression of TRKB and TRKC in cylin-dromas and spiradenomas, when compared with perile-sional epidermis and hair follicles. Finally, downstreamtargets of TRK signalling such as phosphorylated ERKand BCL2 were overexpressed in the tumours.

TRK receptors have received considerable attentionas mediators of neuronal growth, survival and main-tenance (Huang and Reichardt, 2001), and have beenshown to be expressed in hair follicles (Adly et al., 2005;Blasing et al., 2005). Their role in cancer was initiallyshown by uncommon oncogenic fusions involvingC-terminal sequences of the NTRK1/NGF receptor genewith 5-terminal sequences of various activating genessuch as TPM3, TPR and TFG in colon and papillarythyroid carcinoma (Martin-Zanca et al., 1986; Buttiet al., 1995). Exploitation of the constitutively activeprosurvival signalling that is usually restricted toneurons was thought to confer an advantage for thecancer cells. Recent advances in the understanding ofcomplex signalling pathways involved in cancer (Luoet al., 2009) has suggested that some neural-derived

cancer cells, such as neuroblastoma, may becomedependent on TRK signalling for survival, independentof mutations in TRK (Matsumoto et al., 1995; Scalaet al., 1996). Subsequent demonstration of TRKexpression associated with a survival advantage inepithelial tumours, including breast (Lagadec et al.,2009), prostate (Weeraratna et al., 2000) and lung (Ricciet al., 2001), extended the role of TRK signalling to non-neural-derived tumours. At a genomic level, we estab-lished that neither TRKB and TRKC amplification, northat of BDNF and NT3/4, was shown to account for theoverexpression seen in cylindroma tumours, sugges-ting that CYLD dysfunction perturbs normal TRKhomeostasis.

The differential expression of TRKA, TRKB andTRKC in these tumours is intriguing. Whereas TRKA isregulated by CYLD, the interaction of CYLD withTRKB and TRKC remains to be clarified. The lowperinuclear expression of TRKA, but conversely thehigh membranous expression of TRKB and TRKC,suggests that not all TRKs are regulated by CYLD inthe same manner. This has been shown to be true withother forms of ubiquitin modification, including mono-ubiquitination (Arevalo et al., 2006) or Lys48 poly-ubiquitination (Geetha et al., 2005). Arevalo et al.showed that an E3 ubiquitin ligase, Nedd4-2, associatedwith the TRKA receptor and was phosphorylated uponNGF binding, whereas Nedd4-2 did not bind to orubiquitinate related TRKB receptors, because of thelack of a consensus PPXY motif. These results indicatethat TRK receptors are differentially regulated byubiquitination to modulate the survival of neurons(Arevalo et al., 2006). Interestingly, Geetha et al. (2008)have shown that the scaffold protein p62, whichfacilitates the interaction between CYLD and TRKA,is able to bind to specific lysine residues on TRKB (Lys811) and TRKC (Lys 602 and 815), and mutation ofthese sites disrupted downstream signalling. This sug-gests the possibility that CYLD could interact withTRKB and TRKC, but the precise mechanism of TRKspecific homeostasis remains to be clarified.

We established that TRK signalling was effective incylindroma primary cell culture using recombinanthuman BDNF and NT3 for TRKB and TRKC,respectively, and showed increased expression of phos-phorylated ERK and BCL2. To delineate the advantageconferred by TRKB and TRKC overexpression, we useda lentivirus-mediated knockdown system. This showedreduced colony formation in TRKB- and TRKC-deficient cells, supporting the pathogenic role of thesereceptors.

We attempted to investigate if the catalytic activity ofCYLD was implicated in the perturbed TRK signallingseen in these cultures. However, reintroduction of wild-type-CYLD resulted in reduced cell viability, preventingfurther investigation in this model (SupplementaryFigure 5). The dependence of perturbed TRK signallingseen on NFkB and JNK signalling was investigated bygrowing primary cultures in NFkB and JNK inhibitors;however, we were unable to conclusively show a cleareffect on TRK levels (data not shown).

Figure 4 TRK signalling is dysregulated in cylindroma andspiradenoma tissue. The expression levels of 24 526 genes from 32microdissected tumours were pooled and compared against 10pooled perilesional controls, and genes that were differentiallyexpressed (P-value o0.01) were analysed by Ingenuity pathwayanalysis. Overexpressed transcripts are shown in red and under-expressed transcripts are shown in green. TRKB (NTRK2) andTRKC (NTRK3), members of the TRK signalling pathway, wereoverexpressed in tumour tissue compared with control samples.Expression of TRKA (NTRK1) however was reduced. Down-stream members such as phosphatidylinositol-3-kinase and AKTand BCL2 were also over expressed. TRK, tropomyosin kinase.


9

Oncogene

To investigate whether dysregulated TRK signalling inpatients with truncating CYLD mutations represented adruggable pathway, we used a 3D primary cell culturemodel. As seen in breast cancer (Lagadec et al., 2009),TRK expression in vitro was low when 2D cylindroma cellculture was compared with tumours. 3D culture wasassociated with an increase in the expression of TRKBand TRKC, allowing us to use this as an in vitromodel toassay cell viability in the presence of TRK inhibitors. Thereduction in viability at micromolar concentrations ofTRK inhibitors is encouraging, particularly as thesetumours are amenable to transcutaneous or intra-lesionaldrug delivery. We attempted to determine whether themechanism of action of TRK inhibition in our primary

cell culture model was mediated by ERK or BCL2. Thisanalysis proved not to be possible because of the highsensitivity of the primary cells to TRK inhibitors,resulting in only very few live cells after exposure toTRK inhibitors. The result seen with lestaurtinib isparticularly interesting as this agent was recently foundin a screen of 2800 compounds to be a potent inhibitor ofNFkB, suggesting that it may target two key pathways inCYLD mutant tumours (Miller et al., 2010). Lestaurtinibis currently in phase-3 clinical trials (Shabbir and Stuart,2010), and serum levels in patients on an 80-mg oral dosetwice a day was 17–25mm after taking the drug for 28days (Marshall et al., 2005), suggesting that the sensitivityof CYLD-defective tumours suggested by our model may

Figure 5 TRKB and TRKC proteins are overexpressed in human cylindromas and spiradenomas. (a) TMAs of cylindromas andspiradenomas were probed with antibodies as indicated and visualised using a horseradish peroxidase–diaminobenzidene system, withhaematoxylin used as a nuclear counterstain. Membranous TRKB and TRKC were overexpressed in the tumours, exemplified by thecylindroma tumours illustrated here. Patchy perinuclear TRKA staining was seen in some tumour cells. Expression of downstreamtargets phosphorylated ERK and BCL2 was also seen. The scale bars indicate 100 mm; the insets are shown to show specificmembranous staining of TRKB and TRKC (2� enlargement of related image). (b, c) Membranous immunostaining scores of TRKCand TRKB were made on all cores on the array, with P-values of the difference in score (t-test) between tumour and control indicatedin Table 3. The error bars indicate the standard error of the mean. TMA, tissue microarray; TRK, tropomyosin kinase.

Table 3 Membranous TRKB is highly expressed in CYLD mutanttumours

Membranous TRKB Mean score and s.e.m. P-value

Normal tissue (n¼ 3) 2.125±0.1250Cylindroma (n¼ 6) 5.900±0.05571 o0.0001Spiradenoma (n¼ 5) 4.800±0.1414 o0.0001Cylindrospiradenoma (n¼ 1) 5.800±0.2000 o0.0001Trichoepithelioma (n¼ 1) 3.000±0.3162 0.012

Abbreviation: TRK, tropomyosin kinase.Membranous TRKB expression scores in cylindroma tumourscompared with perilesional unaffected tissue. A custom tumourmicroarray was scored for membranous expression of TRKB. Meanscores were grouped by tumour type and compared with the scores ofunaffected perilesional tissue.

Table 4 Membranous TRKC is highly expressed in CYLD mutanttumours

Membranous TRKC Mean score and s.e.m. P-value

Normal tissue (n¼ 2) 0.0005±0.0005Cylindroma (n¼ 4) 5.350±0.1500 o0.0001Spiradenoma (n¼ 4) 3.850±0.1094 o0.0001Cylindrospiradenoma (n¼ 1) 4.600±0.2449 o0.0001Trichoepithelioma (n¼ 1) 3.400±0.4000 0.0038

Abbreviation: TRK, tropomyosin kinase.Membranous TRKC expression scores in cylindroma tumourscompared with perilesional unaffected tissue. A custom tumourmicroarray was scored for membranous expression of TRKC. Meanscores were grouped by tumour type and compared with the scores ofunaffected perilesional tissue.


10

Oncogene

be achievable in patients. As novel agents with increasedpan-TRK specificity such as AZ-23 become available(Thress et al., 2009), patients with truncating CYLDmutations may be another step closer to non-surgicalalternatives for this disfiguring disease.

We chose to investigate the expression of TRK inother human skin cancer models. The overexpression ofmembranous TRKC in a high proportion of BCCs isinteresting. In sporadic BCC, LOH at 9p, the locus forPTCH1, has been found in the majority (93% in thisseries) of BCC (Teh et al., 2005), whereas 16q, the locusfor CYLD, appears rarely affected. TRKC signalling innon-melanoma skin cancer has not been characterised,but there are data from other models to support thatTRK signalling could confer a proliferative advantageto cells where constitutively active SHH signallingfollowing loss of functional PTCH could occur. Synergybetween NT3, the cognate ligand for TRKC, and SHHhas already been shown in murine motorneurons whereboth are important for growth, and use of NT3antibodies in this system reduces the growth of theseneurons (Dutton et al., 1999) . TRKC also directlyinteracts with bone morphogenetic signalling receptors(BMPRI) in murine epithelial cells and inhibits thetumour-suppressor effects of BMP signalling (Jin et al.,2007) in a mouse colon cancer model. BMP and SHHsignalling are tightly balanced, and BMP2, the ligand ofBMPRI, has been shown to oppose SHH-mediatedproliferation in murine cerebellar granular precursorcells through regulation of N-myc (Alvarez-Rodrıguezet al., 2007). Furthermore, BMP signalling is protectivein cutaneous carcinoma, and a SMAD-4 conditionalknockout expressed in murine keratinocytes has indeedshown to result in spontaneous SCCs and BCCs (Qiaoet al., 2006). It is hence conceivable that TRKC-mediated inhibition of BMP signalling, on a mutatedPTCH background in BCC, may have a pathogenic role.The localisation of TRKC at the invasive margin ofBCC tumour islands suggests that this role may be ininvasion (Supplementary Figure 2). The mechanism thatresults in TRKC overexpression in BCC howeverremains to be explored. However, TRK receptors mayrepresent a novel therapeutic target both in tumourswith loss of CYLD function, but also for BCC, the mostcommon cancer in humans.

Materials and methods

Tumour and blood samplesTumour tissue was obtained from patients undergoing surgeryindicated to control tumour burden or for symptomatic relief,under regional ethical committee approval (REC REF:06/1001/59). Patients provided blood samples from which genomic DNAwas extracted. Patients had been genotyped from all threepedigrees, and were confirmed to have the same mutation whenentered into the study. These mutations were c.2460delc,c.2469þ 1G4A and c.2290del5 (accession no. NM_015247).

Sample microdissectionAll samples for microarray studies were snap frozen immedi-ately after surgery, with detailed location maps of the tumours

made at the time. Tumour blocks were mounted in an optimalcutting temperature compound and sectioned on a cryostat at12 mm onto RNAse-free Superfrost coated slides (VWR). Allsubsequent steps were performed in RNAse-free solutions.Tissue sections were fixed in 100% ethanol at �20 1C for 2minand then washed in 70% DEPC-treated ethanol, and treatedwith 50% DEPC-treated ethanol for a further 2min. Thesections were stained with a crystal violet solution (Ambion,Austin, TX, USA) for 40 s and then dehydrated in increasingconcentrations of ethanol. Slides were then stored on dry iceuntil needle microdissection was performed. Microdissectionwas performed at 4 1C using a Leica dissecting microscope.The dissected samples were placed directly into Trizol(Invitrogen, Paisley, UK). Needle homogenisation was per-formed and the sample was frozen at �80 1C until RNA andDNA extraction was performed.

RNA and DNA extraction from tissueTrizol was used to extract RNA and DNA from the samples.The standard supplied protocol was used, with glycogen addedto aid precipitation of RNA into the aqueous phase. RNA wasquantified using a Nanodrop spectrophotometer (Nanodrop,Wilmington, DE, USA). RNA integrity analysis was per-formed using a chip-based electrophoresis kit to confirmpreservation of the 18S and 28S ribosomal bands (RNA 6000Nano and Pico Chip kits; Agilent 2100 Bioanalyzer).

DNA extraction from peripheral bloodGenomic DNA was extracted from peripheral blood using theQiagen whole-blood kit (Qiagen, Crawley, UK) according tothe manufacturer’s protocols.

Array CGHThe 32K BAC re-array collection (CHORI) tiling path arraycomparative genomic hybridisation platform was constructedat the Breakthrough Breast Cancer Research Centre, asdescribed earlier by Marchio et al. (2008). This type of BACarray platform has been shown to be as robust as and to havecomparable resolution with high- density oligonucleotidearrays. DNA labelling, array hybridisations and imageacquisition were performed as described earlier by Turneret al. (2010). Array comparative genomic hybridisation datawere pre-processed and analysed using an in-house R script(BACE.R) in R version 2.9.0 as previously described byNatrajan et al. (2009). After filtering polymorphic BACs, afinal data set of 31 544 clones with unambiguous mappinginformation according to the August 2009 build (hg19) of thehuman genome (http://www.ensembl.org) was smoothed usingthe circular binary segmentation (cbs) algorithm (Natrajanet al., 2009). A categorical analysis was applied to the BACsafter classifying them as representing amplification (40.45),gain (40.12 and o0.44), loss (o�0.12) or no-changeaccording to their cbs-smoothed log2 ratio values. ArrayCGHdata were deposited at ROCK (http://brcabase.icr.ac.uk/index.jsp), at the following URL: http://rock.icr.ac.uk/collaborations/Rajan/Cylindroma.

SNP typingGenomic DNA was extracted from microdissected tissueand assayed on an Illumina 550K SNP array according tothe manufacturer’s protocols. SNP array data were depositedat ROCK (http://brcabase.icr.ac.uk/index.jsp), at the followingURL: http://rock.icr.ac.uk/collaborations/Rajan/Cylindroma.


11

Oncogene


12

Oncogene

Restriction enzyme allelotypingPCR was performed using primers (50–30) CYLD18F GAGAGCTTAAGCAGATGGAA and CYLD18R TAAACAGAAAAGGCAAAAGC that flanked exon-18, and the PCRproduct was subject to restriction enzyme digestion usingHpyCH4V (c.2460delc) and SmlI (c.2469þ 1G4A). Thedigested products were separated on agarose gel and visualisedusing ethidium bromide using a UV transilluminator. LOHwas determined by the absence of a band that correspondedwith loss of the wild-type allele in both cases.

Gene expression profilingThe Illumina WG-DASL platform was used for each samplethat allowed monitoring of the gene expression of 24 526transcripts. A 50-ng weight of total RNA from each samplewas converted to cDNA as per the supplied protocol and thenhybridised to Illumina Human 8 v3 chips. A confocal IlluminaInfinuim bead reader then scanned the chips and the resultswere extracted using the Illumina Beadstudio software. Themicroarray gene expression profiles were deposited at Array-express (www.ebi.ac.uk) under accession number E-MTAB-352.

Figure 6 TRKB and TRKC knockdown reduces primary cell culture colony formation. (a) Cylindroma tumour tissue wasenzymatically digested and converted to a single-cell suspension, and grown on collagen-coated tissue culture plastic. (b) Total celllysates from control HeLa cells, three cylindroma primary cell cultures and a matched patient fibroblast culture were subject to sodiumdodecyl sulphate–PAGE and probed with anti-CYLD antibodies. Lack of full-length CYLD is seen in primary cell cultures.(c) Cylindroma primary cell cultures and tissue were fixed and probed with antibodies against the indicated cytokeratins and labelledwith fluorescent secondary antibodies for visualization. DAPI was used as a nuclear counterstain. CYLD expression is not seen inprimary cell culture. Cylindroma primary cell culture cells expressed cytokeratins that are seen (d) in cylindroma tumours.(e, f) Cylindroma cells were subject to lentivirus-mediated short-hairpin RNA knockdown of TRKB and TRKC, and these were comparedagainst a control hairpin. Knockdown was assessed at the protein level by immunoblotting. (g, h) Transcript expression of TRKB andTRKC is reduced following lentiviral knockdown. (i) Following puromycin selection, GFP-positive cells were plated and grown over12 days. TRKB and TRKC knockdown resulted in reduced colony formation. (j) Similar results were seen in a short-term cell viability assay.All experiments were repeated at least three times. The error bars indicate the standard error of the mean and the scale bars represent adistance of 100mm. DAPI, 40,6-diamidino-2-phenylindole; GFP, green fluorescent protein; TRK, tropomyosin kinase.

Figure 7 3D primary cell culture of cylindroma cells express TRK proteins and respond to stimulation with cognate TRK ligands.(a) Cylindroma primary cell cultures were plated onto collagen-coated tissue culture scaffolds (3D cell culture) and grown for 28 daysin the same conditions as standard 2D cell culture. Tissue culture scaffolds were fixed and stained with haematoxylin and eosin(� 20 original magnification expression; scale bar 100mm) (b, c) Total cell lysates from fresh tumours, 3D and 2D cell cultures wereseparated using sodium dodecyl sulphate–PAGE, and immunoblots were probed using (b) TRKB and (c) TRKC antibodies. Lanes 1and 2 were loaded with 5mg of cell lysate from fresh tumours, labelled ‘Cylindroma 1 and 2’, rest 20 mg of lysate from primary cellcultures in 3D (PCC 2,4 3D) and cultures in 2D (PCC 2,4 2D). TRKB and TRKC were expressed at higher levels in 3D cell cultures(lanes 3 and 5) compared with matched 2D cultures (lanes 4 and 6). (d, e) Cylindroma cells were grown on 3D scaffolds and stimulatedwith NGF, BDNF and NT3 for the periods indicated, and then transferred to ice-cold phosphate-buffered saline, and cell lysates wereextracted for immunoblotting. Membranes were probed with antibodies against phosphorylated ERK and BCL2. Increased levels ofphosphorylated ERK and BCL2 were seen. All experiments were repeated at least three times. 2D, two-dimensional; 3D, three-dimensional;BDNF, brain-derived neurotrophic factor; NGF, nerve growth factor; PCC, primary cell culture; TRK, tropomyosin kinase.


13

Oncogene

Data normalisation and analysisRaw gene expression values were normalised using quantilenormalisation. Differential expression analysis was performedusing the Beadstudio software to detect significantly differentexpressed genes. Differences in signal values for each genewere scored with a differential score. Bonferroni testing wasperformed to correct for multiple hypothesis testing, and theresults were ranked in order of differential score, with adifferential score ranging from þ 374 to �374, with ±65corresponding to a P-value cut off of o0.01. Signal valuesfrom genes that fulfilled this level of significance were analysedfor fold change and results have been tabulated in Table 1.Pathway discovery was further aided by the Ingenuity pathwayanalysis software. Finally, a parallel analysis was performed bygene set expression analysis. For this analysis, probe values(24 526) were collapsed to gene symbols (18 630), with themaximum probe value used for replicate probes. Afterfiltering, 867 gene sets were used in the analysis. The gene

sets were permutated 1000 times, before being ranked in orderusing a ‘difference of classes’ metric, with a weightedenrichment statistic applied. Fifty-six gene sets were signifi-cantly enriched at a nominal P-value of o0.01.

Quantitative PCRTotal RNA was extracted from cylindroma primary cellcultures as described above, subject to DNAse digestion andthen reverse-transcribed using the Superscript III kit (Invitro-gen) according to the manufacturer’s instructions. cDNA wasthen used for quantitative PCR using the following primers(50–30): GAPDH F, ATGGGGAAGGTGAAGGTCG; GAPDHR, GGGGTCATTGATGGCAACAATA; NTRK2 F, TGTTCAGCACATCAAGCGACA; NTRK2 R, GCTCAGGACAGAGGTTATAGCAT; NTRK3 F, TCCGTCAGGGACACAACTG; NTRK3 R, GCACACTCCATAGAACTTGACA, usingan SYBR green-containing Taq polymerase master mix(Sigma, Dorset, UK), in an ABI 7900HT thermal cycler. Geneexpression was normalised to GAPDH and comparisons inexpression were made using the 2�DDCT formula as the PCRreactions had similar efficiencies of amplification.

ImmunoblottingHomogenised tumour samples and lysates were sonicated inlysis buffer (65mM Tris, 50mM NaCl, 5mM EDTA, 1mM

phenylmethyl-sulphonyl fluoride, 1mM sodium orthovanadate,0.1% sodium dodecyl sulphate, 2% Triton X and 1% NP-40)with added phosphatase and protease inhibitors (Roche,Hertfordshire, UK). A 5- to 20-mg weight of total protein wasloaded onto sodium dodecyl sulphate–PAGE gels and thentransferred to polyvinylidene difluoride membranes. Immuno-

Figure 8 Cylindroma 3D primary cell cultures show increased sensitivity to TRK inhibitors relative to aspirin. (a–d) Cylindroma 3Dcell cultures were grown for 2 weeks in the presence of salicylic acid, K252a, tyrphostin AG879 and lestaurtinib (CEP-701) in triplicatein a 96-well format. This was repeated using at least three different tumours from different body sites of three patients. Cell viabilitywas assessed using a luminescent cell viability assay at the end of this period and dose response curves were generated. Cylindroma cellswere sensitive to lestaurtinib at micromolar levels, whereas salicylic acid had a similar effect at millimolar levels. 3D, three-dimensional;TRK, tropomyosin kinase.

Table 5 Membranous TRKA, TRKB and TRKC expression insporadic skin cancers

BCC SCC MM

TRKA 0/23 0/19 0/5TRKB 0/23 0/19 0/5TRKC 16/23 1/19 0/5

Abbreviations: BCC, basal cell carcinoma; MM, malignant melanoma;SCC, squamous cell carcinoma; TRK, tropomyosin kinase.A skin cancer TMA was scored for the presence of membranous TRKstaining.


14

Oncogene

blots were probed with the following antibodies: TRKB,TRKC, BCL2, p44/42, b-actin (Cell Signaling Technology,Hertfordshire, UK), and visualised using appropriate horse-radish peroxidase-conjugated secondary antibodies using ECL-Plus (GE Healthcare, Bucks, UK). The immunoblots wereexposed to a Biomax XR film and processed routinely.

Tissue microarrayA custom-made TMA was made from single 2-mm cores fromrepresentative tumour areas and was used to validate proteinexpression in tissue sections. The TMA contained 10cylindroma samples, four spiradenoma samples and twotrichoepithelioma samples. The cylindromas and spiradeno-

Figure 9 Membranous TRKC is overexpressed in BCC. A skin cancer tumour microarray with 23 BCCs, 19 SCCs and 5 malignantmelanomas, with two cores per case, was immunostained with antibodies against TRKA, TRKB and TRKC. (a) Strongcircumferential membranous immunostaining of TRKC is seen in 16 out of 23 cases (70%) of BCC on a TMA, whereas there isabsence of membranous TRKA and TRKB staining. (b) One SCC showed a similar pattern of staining. Malignant melanomas did notshow membranous staining. (c, d) A total of 39% of BCCs and 16% of SCCs also showed perinuclear TRKA staining, and 13% ofBCCs only showed faint cytoplasmic TRKB staining. (e) Percentage of cases with membranous TRKC staining. (f) Mean membranousstaining scores for TRKC were performed. Membranous TRKC staining is seen in 70% of the BCCs in this series and 5% of SCCs, butnot in malignant melanoma. The error bars indicate the standard error of the mean. BCC, basal cell carcinoma; SCC, squamous cellcarcinoma; TMA, tissue microarray; TRK, tropomyosin kinase.


15

Oncogene

mas were chosen from both scalp and torso, and trichoepithe-liomas were from face. Human skin cancer TMA slides (A216)were purchased from Stretton Scientfic (Stretton, UK).

Immunohistochemistry on paraffin-embedded sections andTMAsTissue sections from the TMA were de-waxed with xylene andrehydrated in graded ethanols before undergoing microwaveantigen retrieval in sodium citrate buffer (pH 6.0). The tissuesections were then blocked with a peroxidase-blocking agent(Dako, Cambridgeshire, UK) and then incubated with thefollowing primary antibodies: TRKA, TRKB, TRKC, p44/42(phosphorylated ERK) and BCL2. The tissue sections werethen washed in phosphate-buffered saline and probed withsecondary horseradish peroxidase-conjugated antibodies, andstaining was visualised with 3,30-diaminobenzidine. Haema-toxylin was used as a nuclear counterstain. The perilesionalepidermis was used as negative control, and anagen hairfollicles in perilesional tissue served as positive control forTRK antibodies.

Immunohistochemistry scoring of TRK expressionTissue sections were assessed for staining in five randomlyselected high-power fields per core on the TMA. A modifiedversion of an established score to assess membranous stainingof Her-2 was developed (Kavanagh et al., 2009).To quantifyTRK staining, tumour cells were scored as having circumfer-ential strong membranous staining (3), strong non-circumfer-ential staining (2), weak staining (1) or no staining (0). Theproportion of such cells stained was counted per high-powerfield under a � 63 objective using a Leica DMR microscopeand scored as follows: 75–100% of cells (3), 50–75% (2) andless than 50% (1). A total score of up to 6 was possible per fieldper core and scores were compared with perilesional kerati-nocytes for statistical significance using t-test.

Cylindroma primary cell culture and drug inhibition experimentsCylindroma samples were micro-dissected using a Zeissmicroscope after surgery and tumour pieces were enzymati-cally digested in 2.5% trypsin. A cell suspension was thencounted for viable cells using Trypan blue and the cells wereplated in collagen-coated tissue culture plasticware and grownin keratinocyte serum-free medium. Lentiviral transfection wasperformed using short-hairpin RNA against TRKB andTRKC, and compared with a non-silencing control (Open-biosystems, Thermo Scientific, Surrey, UK). Cells were grownin puromycin selection media till all cells were GFP-positiveand then used for experiments. For colony-forming assays,cells were plated at a low density of 4000 cells per well of a six-well plate. These cells were maintained for 10–14 days, beforethey were fixed in 3.3% trichloroacetic acid, and stained with

0.57% sulphorhodamine in 1% acetic acid. The plates werethen scanned and colonies with more than 10 cells werecounted. For short-term viability assays, cells were grown in a96-well format for 96 h before cell viability was assayed using aluminescent assay that lysed cells and used liberated ATP asreadout for viability (CellTiter Glo; Promega, Southampton,UK). This was performed according to standard manufac-turer’s protocols and luminescence was read using a lumin-ometer (Luminoskan–Ascent, Thermo Scientific). For 3Dcultures, cells were seeded on a polystyrene collagen-coatedscaffold (Reinnervate, Co., Durham, UK), grown for 28 daysand then treated with drugs. Cells grown in the followinginhibitors were used, diluted in dimethylsulphoxide, with afinal concentration of dimethylsulphoxide at 0.1%: K252a andtyrphostin AG879, salicylic acid (Sigma), lestaurtinib (TorontoResearch Chemicals, Ontario, Canada). After 14 days CellTi-ter Glo (Promega) was added to the media according tostandard manufacturer’s protocols and luminescence was readusing a luminometer (Luminoskan–Ascent). All experimentswere performed in triplicate.

Fluorescence-activated cell sorting analysisCells were trypsinised, resuspended in phosphate-bufferedsaline and fixed by adding ice-cold ethanol while gentlyvortexing the tube, until a final concentration of 70% ethanolwas reached. The cells were fixed for 1 h at 4 1C. Cells werethen pelleted and resuspended in Cystain Solution (Partech,Kent, UK), which stained the nuclear DNA with 40,6-diamidino-2-phenylindole (DAPI). Cells were stained for 1 hin the dark at 4 1C before analysis on a FACSCanto II flowcytometer (Becton Dickinson, Oxford, UK) using a UV laser.The cells were analysed for total DNA content and a minimumof 10 000 events were used for cell-cycle data. All cell-cycledata were analysed using Modfit (Verity Software House,Topsham, ME, USA).

Conflict of interest

The authors declare no conflict of interest.

Acknowledgements

This work was funded in part by grants from the NewcastleHealthcare Charities Trust, the North East Skin ResearchFund, Breakthrough Breast Cancer, Cancer Research UK andthe Medical Research Council. N Rajan is a MRC ClinicalTraining Fellow.

References

Adly MA, Assaf HA, Nada EA, Soliman M, Hussein M. (2005).Human scalp skin and hair follicles express neurotrophin-3 and itshigh-affinity receptor tyrosine kinase C, and show hair cycle-dependent alterations in expression. Br J Dermatol 153: 514–520.

Alvarez-Rodrıguez R, Barzi M, Berenguer J, Pons S. (2007). Bonemorphogenetic protein 2 opposes Shh-mediated proliferation incerebellar granule cells through a TIEG-1-based regulation ofNmyc. J Biol Chem 282: 37170–37180.

Annunziata CM, Davis RE, Demchenko Y, Bellamy W, Gabrea A,Zhan F et al. (2007). Frequent engagement of the classical and

alternative NF-kappaB pathways by diverse genetic abnormalitiesin multiple myeloma. Cancer Cell 12: 115–130.

April C, Klotzle B, Royce T, Wickham-Garcia E, Boyaniwsky T, IzzoJ et al. (2009). Whole-genome gene expression profiling of formalin-fixed, paraffin-embedded tissue samples. PLoS One 4: e8162.

Arevalo JC, Waite J, Rajagopal R, Beyna M, Chen Z-Y, Lee FS et al.(2006). Cell survival through Trk neurotrophin receptors isdifferentially regulated by ubiquitination. Neuron 50: 549–559.

Blasing H, Hendrix S, Paus R. (2005). Pro-inflammatory cytokinesupregulate the skin immunoreactivity for NGF, NT-3, NT-4 and


16

Oncogene

their receptor, p75NTR in vivo: a preliminary report. Arch Dermatol

Res 296: 580–584.Bowen S, Gill M, Lee DA, Fisher G, Geronemus RG, Vazquez ME

et al. (2005). Mutations in the CYLD gene in Brooke–Spieglersyndrome, familial cylindromatosis, and multiple familial trichoe-pithelioma: lack of genotype–phenotype correlation. J Invest

Dermatol 124: 919–920.Bruckner-Tuderman L, Pfaltz M, Schnyder UW. (1991). Cylindroma

overexpresses collagen VII, the major anchoring fibril protein.J Invest Dermatol 96: 729–734.

Brummelkamp TR, Nijman SM, Dirac AM, Bernards R. (2003). Lossof the cylindromatosis tumour suppressor inhibits apoptosis byactivating NF-kappaB. Nature 424: 797–801.

Butti MG, Bongarzone I, Ferraresi G, Mondellini P, Borrello MG,Pierotti MA. (1995). A sequence analysis of the genomic regionsinvolved in the rearrangements between TPM3 and NTRK1 genesproducing TRK oncogenes in papillary thyroid carcinomas.Genomics 28: 15–24.

Chen-Tsai CP, Colome-Grimmer M, Wagner RF. (2004). Correlationsamong neural cell adhesion molecule, nerve growth factor, and itsreceptors, TrkA, TrkB, TrkC, and p75, in perineural invasion bybasal cell and cutaneous squamous cell carcinomas. Dermatologic

Surgery: Official Publication for American Society for Dermatologic

Surgery (et al) 30: 1009–1016.Coe BP, Ylstra B, Carvalho B, Meijer GA, Macaulay C, Lam WL.

(2007). Resolving the resolution of array CGH. Genomics 89:647–653.

Demchenko YN, Glebov OK, Zingone A, Keats JJ, Bergsagel PL,Kuehl WM. (2010). Classical and/or alternative NF B pathwayactivation in multiple myeloma. Blood 115: 3541–3552.

Donovan J. (2009). Review of the hair follicle origin hypothesis forbasal cell carcinoma. Dermatologic Surgery: Official Publication for

American Society for Dermatologic Surgery (et al) 35: 1311–1323.Dunning MJ, Barbosa-Morais NL, Lynch AG, Tavare S, Ritchie ME.

(2008). Statistical issues in the analysis of Illumina data. BMC

Bioinformatics 9: 85.Dutton R, Yamada T, Turnley A, Bartlett PF, Murphy M. (1999).

Regulation of spinal motoneuron differentiation by the combinedaction of Sonic hedgehog and neurotrophin 3. Clin Exp Pharmacol

Physiol 26: 746–748.Gao J, Huo L, Sun X, Liu M, Li D, Dong JT et al. (2008). The tumor

suppressor CYLD regulates microtubule dynamics and plays a rolein cell migration. J Biol Chem 283: 8802–8809.

Geetha T, Jiang J, Wooten MW. (2005). Lysine 63 polyubiquitinationof the nerve growth factor receptor TrkA directs internalization andsignaling. Mol Cell 20: 301–312.

Geetha T, Seibenhener ML, Chen L, Madura K, Wooten MW. (2008).p62 serves as a shuttling factor for TrkA interaction with theproteasome. Biochem Biophys Res Commun 374: 33–37.

Gerretsen AL, van der Putte SC, Deenstra W, van Vloten WA. (1993).Cutaneous cylindroma with malignant transformation. Cancer 72:1618–1623.

Gunnarsson R, Staaf J, Jansson M, Ottesen AM, Goransson H,Liljedahl U et al. (2008). Screening for copy-number alterations andloss of heterozygosity in chronic lymphocytic leukemia—a com-parative study of four differently designed, high resolutionmicroarray platforms. Genes Chromosomes Cancer 47: 697–711.

Hellerbrand C, Bumes E, Bataille F, Dietmaier W, Massoumi R,Bosserhoff AK. (2007). Reduced expression of CYLD in humancolon and hepatocellular carcinomas. Carcinogenesis 28: 21–27.

Hoseong Yang S, Andl T, Grachtchouk V, Wang A, Liu J, Syu L et al.(2008). Pathological responses to oncogenic Hedgehog signaling inskin are dependent on canonical Wnt/b-catenin signaling. Nat Genet

40: 1130–1135.Huang EJ, Reichardt LF. (2001). Neurotrophins: roles in neuronal

development and function. Annu Rev Neurosci 24: 677–736.Illumina Inc. (2009). Beadstudio Gene Expression Module v3.4 User

Guide (113172615 Rev A) Illumina Inc.: San Diego, CA.Jenner MW, Leone PE, Walker BA, Ross FM, Johnson DC,

Gonzalez D et al. (2007). Gene mapping and expression analysis

of 16q loss of heterozygosity identifies WWOX and CYLD as beingimportant in determining clinical outcome in multiple myeloma.Blood 110: 3291–3300.

Jin W, Yun C, Kim H-S, Kim S-J. (2007). TrkC binds to the bonemorphogenetic protein type II receptor to suppress bone morpho-genetic protein signaling. Cancer Res 67: 9869–9877.

Kavanagh DO, Chambers G, O’Grady L, Barry KM, Waldron RP,Bennani F et al. (2009). Is overexpression of HER-2 a predictor ofprognosis in colorectal cancer? BMC Cancer 9: 1.

Keats J, Fonseca R, Chesi M, Schop R, Baker A, Chng W et al. (2007).Promiscuous mutations activate the noncanonical NF-kB Pathwayin multiple myeloma. Cancer Cell 12: 131–144.

Kikuno N, Shiina H, Urakami S, Kawamoto K, Hirata H, Tanaka Yet al. (2008). Genistein mediated histone acetylation and demethyla-tion activates tumor suppressor genes in prostate cancer cells.Int J Cancer 123: 552–560.

Komander D, Lord CJ, Scheel H, Swift S, Hofmann K, Ashworth Aet al. (2008). The structure of the CYLD USP domain explains itsspecificity for Lys63-linked polyubiquitin and reveals a B boxmodule. Mol Cell 29: 451–464.

Kovalenko A, Chable-Bessia C, Cantarella G, Israel A, Wallach D,Courtois G. (2003). The tumour suppressor CYLD negativelyregulates NF-kappaB signalling by deubiquitination. Nature 424:801–805.

Kriegl L, Horst D, Kirchner T, Jung A. (2009). LEF-1 expression inbasal cell carcinomas. Br J Dermatol 160: 1353–1356.

Lagadec C, Meignan S, Adriaenssens E, Foveau B, Vanhecke E,Romon R et al. (2009). TrkA overexpression enhances growth andmetastasis of breast cancer cells. Oncogene 28: 1960–1970.

Lamb J. (2006). The connectivity map: using gene-expressionsignatures to connect small molecules, genes, and disease. Science

313: 1929–1935.Liu H-H, Xie M, Schneider MD, Chen ZJ. (2006). Essential role of

TAK1 in thymocyte development and activation. Proc Natl Acad

Sci USA 103: 11677–11682.Luo J, Solimini NL, Elledge SJ. (2009). Principles of cancer therapy:

oncogene and non-oncogene addiction. Cell 136: 823–837.Marchio C, Natrajan R, Shiu KK, Lambros MB, Rodriguez-Pinilla

SM, Tan DS et al. (2008). The genomic profile of HER2-amplified breast cancers: the influence of ER status. J Pathol 216:399–407.

Marshall JL, Kindler H, Deeken J, Bhargava P, Vogelzang NJ,Rizvi N et al. (2005). Phase I trial of orally administered CEP-701, anovel neurotrophin receptor-linked tyrosine kinase inhibitor. Invest

New Drugs 23: 31–37.Martin-Zanca D, Hughes SH, Barbacid M. (1986). A human oncogene

formed by the fusion of truncated tropomyosin and protein tyrosinekinase sequences. Nature 319: 743–748.

Massoumi R. (2006). Cyld inhibits tumor cell proliferation by blockingBcl-3-dependent NF-kB signaling. Cell 125: 665–677.

Massoumi R, Chmielarska K, Hennecke K, Pfeifer A, Fassler R.(2006). Cyld inhibits tumor cell proliferation by blocking Bcl-3-dependent NF-kappaB signaling. Cell 125: 665–677.

Massoumi R, Kuphal S, Hellerbrand C, Haas B, Wild P, Spruss Tet al. (2009). Downregulation of CYLD expression by Snailpromotes tumor progression in malignant melanoma. J Exp Med

206: 221–232.Matsumoto K, Wada RK, Yamashiro JM, Kaplan DR, Thiele CJ.

(1995). Expression of brain-derived neurotrophic factor andp145TrkB affects survival, differentiation, and invasiveness ofhuman neuroblastoma cells. Cancer Res 55: 1798–1806.

Meybehm M, Fischer HP. (1997). Spiradenoma and dermal cylin-droma: comparative immunohistochemical analysis and histoge-netic considerations. Am J Dermatopathol 19: 154–161.

Miller SC, Huang R, Sakamuru S, Shukla SJ, Attene-Ramos MS,Shinn P et al. (2010). Identification of known drugs that act asinhibitors of NF-kappaB signaling and their mechanism of action.Biochem Pharmacol 79: 1272–1280.

Natrajan R, Weigelt B, Mackay A, Geyer FC, Grigoriadis A, Tan DSPet al. (2009). An integrative genomic and transcriptomic analysis


17

Oncogene

reveals molecular pathways and networks regulated by copy numberaberrations in basal-like, HER2 and luminal cancers. Breast Cancer

Res Treat 121: 575–589.Nilsson M, Unden AB, Krause D, Malmqwist U, Raza K,

Zaphiropoulos PG et al. (2000). Induction of basal cell carcinomasand trichoepitheliomas in mice overexpressing GLI-1. Proc Natl

Acad Sci USA 97: 3438–3443.Oosterkamp HM, Neering H, Nijman SM, Dirac AM, Mooi WJ,

Bernards R et al. (2006). An evaluation of the efficacy of topicalapplication of salicylic acid for the treatment of familial cylindro-matosis. Br J Dermatol 155: 182–185.

Qiao W, Li AG, Owens P, Xu X, Wang X-J, Deng C-X. (2006). Hairfollicle defects and squamous cell carcinoma formation in Smad4conditional knockout mouse skin. Oncogene 25: 207–217.

Rajan N, Langtry JAA, Ashworth A, Roberts C, Chapman P, Burn Jet al. (2009). Tumor mapping in 2 large multigenerational familieswith CYLD mutations: implications for disease management andtumor induction. Arch Dermatol 145: 1277–1284.

Reiley WW, Jin W, Lee AJ, Wright A, Wu X, Tewalt EF et al. (2007).Deubiquitinating enzyme CYLD negatively regulates the ubiquitin-dependent kinase Tak1 and prevents abnormal T cell responses.J Exp Med 204: 1475–1485.

Reiley W, Zhang M, Sun S-C. (2004). Negative regulation ofJNK signaling by the tumor suppressor CYLD. J Biol Chem 279:55161–55167.

Reiley WW, Zhang M, Jin W, Losiewicz M, Donohue KB, NorburyCC et al. (2006). Regulation of T cell development by thedeubiquitinating enzyme CYLD. Nat Immunol 7: 411–417.

Ricci A, Greco S, Mariotta S, Felici L, Bronzetti E, Cavazzana A et al.(2001). Neurotrophins and neurotrophin receptors in human lungcancer. Am J Respir Cell Mol Biol 25: 439–446.

Saggar S, Chernoff KA, Lodha S, Horev L, Kohl S, Honjo RS et al.(2008). CYLD mutations in familial skin appendage tumours.J Med Genet 45: 298–302.

Saha RN, Liu X, Pahan K. (2006). Upregulation of BDNF inastrocytes by TNF-alpha: a case for the neuroprotective role ofcytokine. J Neuroimmune Pharmacol 1: 212–222.

Saini HS, Gorse KM, Boxer LM, Sato-Bigbee C. (2004). Neurotro-phin-3 and a CREB-mediated signaling pathway regulate Bcl-2expression in oligodendrocyte progenitor cells. J Neurochem 89:951–961.

Scala S, Wosikowski K, Giannakakou P, Valle P, Biedler JL, SpenglerBA et al. (1996). Brain-derived neurotrophic factor protectsneuroblastoma cells from vinblastine toxicity. Cancer Res 56:3737–3742.

Shabbir M, Stuart R. (2010). Lestaurtinib, a multitargeted tyrosinekinase inhibitor: from bench to bedside. Expert Opin Invest Drugs

19: 427–436.Stegmeier F, Sowa ME, Nalepa G, Gygi SP, Harper JW, Elledge SJ.

(2007). The tumor suppressor CYLD regulates entry into mitosis.Proc Natl Acad Sci USA 104: 8869–8874.

Stokes A, Wakano C, Koblan-Huberson M, Adra CN, Fleig A,Turner H. (2006). TRPA1 is a substrate for de-ubiquitination by thetumor suppressor CYLD. Cell Signal 18: 1584–1594.

Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL,Gillette MA et al. (2005). Gene set enrichment analysis: aknowledge-based approach for interpreting genome-wide expressionprofiles. Proc Natl Acad Sci USA 102: 15545–15550.

Suzuki R, Shimodaira H. (2006). Pvclust: an R package for assessing theuncertainty in hierarchical clustering. Bioinformatics 22: 1540–1542.

Tan DS, Lambros MB, Natrajan R, Reis-Filho JS. (2007). Getting itright: designing microarray (and not ‘microawry’) comparativegenomic hybridization studies for cancer research. Lab Invest 87:737–754.

Tauriello DVF, Haegebarth A, Kuper I, Edelmann MJ, Henraat M,Canninga-van Dijk MR et al. (2010). Loss of the tumor suppressorCYLD enhances Wnt/beta-catenin signaling through K63-linkedubiquitination of Dvl. Mol Cell 37: 607–619.

Teh M-T, Blaydon D, Chaplin T, Foot NJ, Skoulakis S, Raghavan Met al. (2005). Genomewide single nucleotide polymorphism micro-array mapping in basal cell carcinomas unveils uniparental disomyas a key somatic event. Cancer Res 65: 8597–8603.

Tellechea O, Reis JP, Ilheu O, Baptista AP. (1995). Dermalcylindroma. An immunohistochemical study of thirteen cases.Am J Dermatopathol 17: 260–265.

Thress K, Macintyre T, Wang H, Whitston D, Liu ZY, Hoffmann Eet al. (2009). Identification and preclinical characterization of AZ-23, a novel, selective, and orally bioavailable inhibitor of the Trkkinase pathway. Mol Cancer Ther 8: 1818–1827.

Timpl R, Fujiwara S, Dziadek M, Aumailley M, Weber S, Engel J.(1984). Laminin, proteoglycan, nidogen and collagen IV: structuralmodels and molecular interactions. Ciba Found Symp 108: 25–43.

Trompouki E, Hatzivassiliou E, Tsichritzis T, Farmer H, Ashworth A,Mosialos G. (2003). CYLD is a deubiquitinating enzyme thatnegatively regulates NF-kappaB activation by TNFR familymembers. Nature 424: 793–796.

Trompouki E, Tsagaratou A, Kosmidis SK, Dolle P, Qian J,Kontoyiannis DL et al. (2009). Truncation of the catalytic domainof the cylindromatosis tumor suppressor impairs lung maturation.Neoplasia 11: 469–476.

Tunggal L, Ravaux J, Pesch M, Smola H, Krieg T, Gaill F et al.(2002). Defective laminin 5 processing in cylindroma cells. Am J

Pathol 160: 459–468.Turner N, Lambros MB, Horlings HM, Pearson A, Sharpe R,

Natrajan R et al. (2010). Integrative molecular profiling of triplenegative breast cancers identifies amplicon drivers and potentialtherapeutic targets. Oncogene 29: 2013–2023.

Weeraratna AT, Arnold JT, George DJ, DeMarzo A, Isaacs JT.(2000). Rational basis for Trk inhibition therapy for prostatecancer. Prostate 45: 140–148.

Workman C, Jensen LJ, Jarmer H, Berka R, Gautier L, Nielser HB et al.(2002). A new non-linear normalization method for reducing variabilityin DNA microarray experiments. Genome Biol 3: research0048.

Wu C, Ramirez A, Cui B, Ding J, Delcroix J-DM, Valletta JSet al. (2007). A functional dynein-microtubule network is requiredfor NGF signaling through the Rap1/MAPK pathway. Traffic 8:1503–1520.

Zhang J, Stirling B, Temmerman ST, Ma CA, Fuss IJ, Derry JMJ et al.(2006). Impaired regulation of NF-kB and increased susceptibilityto colitis-associated tumorigenesis in CYLD-deficient mice. J Clin

Invest 116: 3042–3049.Zhong S, Fields CR, Su N, Pan Y-X, Robertson KD. (2007).

Pharmacologic inhibition of epigenetic modifications, coupled withgene expression profiling, reveals novel targets of aberrant DNAmethylation and histone deacetylation in lung cancer. Oncogene 26:2621–2634.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)


18

Oncogene

dysregulated trk signalling is a therapeutic target in cyld defective tumours

Documents