Clinical association analysis of ependymomas and pilocytic astrocytomas reveals elevated FGFR3 and FGFR1 expression in aggressive ependymomas

RESEARCH ARTICLE Open Access Clinical associa ion analysis o ependymomas and pilocy ic as ocy omas e eals ele a ed FGFR3 and FGFR1 exp ession in agg essi e ependymomas Bi gi a Leh inen 1† , Annina Rai a 2,3† , Juha Kesseli 1 , Ma i Annala 1 , K is iina No d o s 2,4 , Olli Yli-Ha ja 5 , Wei Zhang 5,6 , Tapio Visako pi 1,2 , Ma i Nyk e 1,7 , Hannu Haapasalo 2,3* and Ki si J. G anbe g 1,5,7* Abs ac Backg ound: Fib oblas g ow h ac o ecep o s (FGFRs) a e well-known p o o-oncogenes in se e al human malignancies and a e cu en ly he apeu ically a ge ed in clinical ials. Among glioma sub ypes, ac i a ing FGFR1 al e a ions ha e been obse ed in a subpopula ion o pilocy ic as ocy omas while FGFR3 usions occu in IDH wild- ype di use gliomas, esul ing in high FGFR3 p o ein exp ession. The pu pose o his s udy was o associa e FGFR1 and FGFR3 p o ein le els wi h clinical ea u es and gene ic al e a ions in ependymoma and pilocy ic as ocy oma. Me hods: FGFR1 and FGFR3 exp ession le els we e de ec ed in ependymoma and pilocy ic as ocy oma issues using immunohis ochemis y. Selec ed cases we e u he analyzed using a ge ed sequencing. Resul s: Exp ession o bo h FGFR1 and FGFR3 a ied wi hin all umo ypes. In ependymomas, inc eased FGFR3 o FGFR1 exp ession was associa ed wi h high umo g ade, ce eb al loca ion, young pa ien age, and poo p ognosis. Mode a e- o-s ong exp ession o FGFR1 and/o FGFR3 was obse ed in 76% o ce eb al ependymomas. Cases wi h mode a e- o-s ong exp ession o bo h p o eins had poo clinical p ognosis. In pilocy ic as ocy omas, mode a e- o-s ong FGFR3 exp ession was de ec ed p edominan ly in non-pedia ic pa ien s. Ta ge ed sequencing o 12 umo s ound no p o ein-al e ing mu a ions o usions in FGFR1 o FGFR3. Conclusions: Ele a ed FGFR3 and FGFR1 p o ein exp ession is common in agg essi e ependymomas bu likely no d i en by gene ic al e a ions. Fu he s udies a e wa an ed o e alua e whe he ependymoma pa ien s wi h high FGFR3 and/o FGFR1 exp ession could bene i om ea men wi h FGFR inhibi o based he apeu ic app oaches cu en ly unde e alua ion in clinical ials. Keywo ds: Tissue mic oa ay, Deep-sequencing, FGFR inhibi ion, Immunohis ochemis y s aining Backg ound Fib oblas g ow h ac o ecep o s (FGFRs) a e a amily o ecep o y osine kinases ha a e ac i a ed in a a ie y o cance s and ha e well-es ablished oncogenic p ope ies [1, 2]. Since he disco e y o ecu en FGFR gene usions in glioblas oma [3, 4], FGFR inhibi o based ea men egimens ha e been iewed as a p omising he apeu ic op- ion o b ain umo s wi h FGFR al e a ions. The mecha- nisms o FGFR ac i a ion in b ain umo s a y by umo ype, bu include oncogenic FGFR3 and FGFR1 usions, FGFR1 ea angemen s, and FGFR1 mu a ions [2–8]. Mo eo e , gene usions appea o be he sole ecu en oncogenic FGFR3 al e a ion in b ain umo s. Al hough FGFR3 is commonly used o a ans o ming acidic coiled- coil-con aining p o ein 3 (TACC3) gene, o he usion pa ne s exis . Fo example, ecu en FGFR3–BAIAP2L1 usions ha e been de ec ed in bladde cance [9]. Se e al FGFR inhibi o s a e cu en ly unde p e-clinical and * Co espondence: hannu.haa[email p o ec ed];[email p o ec ed] † Equal con ibu o s 2 Fimlab Labo a o ies L d., Tampe e Uni e si y Hospi al, Bioka u 4, 33520 Tampe e, Finland 1 BioMediTech Ins i u e and Facul y o Medicine and Li e Sciences, Bioka u 8, 33520 Tampe e, Finland Full lis o au ho in o ma ion is a ailable a he end o he a icle © The Au ho (s). 2017 Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. The C ea i e Commons Public Domain Dedica ion wai e (h p://c ea i ecommons.o g/publicdomain/ze o/1.0/) applies o he da a made a ailable in his a icle, unless o he wise s a ed. Leh inen e al. BMC Cance (2017) 17:310 DOI 10.1186/s12885-017-3274-9 clinical e alua ion, and ecen epo s ha e shown good ea men esponses in FGFR3 usion posi i e cells and umo s [8, 10, 11]. While mos o he FGFR inhibi o s udies, o da e, ha e been pe o med in cases in ol ing ca cinomas, esponses o FGFR inhibi o s ha e also been epo ed in cases wi h glioblas oma [8, 12]. Ependymomas and pilocy ic as ocy omas a e nondi - use gliomas, in which neoplas ic cells do no subs an- ially in il a e in o su ounding no mal issue. They ep esen di e en g ades, ypes o g ow h and clinical cou ses. Nondi use g ow h pa e n acili a es e icien su gical emo al o he umo , which pa ly explains he be e p ognosis in hese pa ien s ela i e o hose wi h di use gliomas. Howe e , umo ecu s in some o he pa ien s, and o e all su i al a es a e wo se wi h mo e agg essi e ependymomas [13]. Ependymomas a e he hi d mos common b ain umo in child en, ep esen ing 8–10% o pedia ic in a- c anial umo s and app oxima ely 4% o all adul b ain umo s [13]. Ependymomas a e ound in all loca ions o he cen al ne ous sys em, and may be in ac anial (in a en o ial o sup a en o ial) o spinal. In a en o ial pos e io ossa ependymomas can be u he subclassi- ied in o pos e io ossa g oup A (PFA) and g oup B (PFB) umo s [14]. Adul ependymomas a e ypically g ade I myxopapilla y ependymomas localized in he spinal co d, while pedia ic ependymomas a e ypically in ac anial g ade II–III umo s [13, 15]. Al hough epen- dymomas in young child en a e ypically associa ed wi h poo p ognosis [15, 16], adul sup a en o ial ependymo- mas a e also associa ed wi h lowe su i al a es [13]. Apa om copy numbe al e a ions [13], signi ican gene ic and epigene ic d i e s o ependymoma de elop- men ha e been ecen ly epo ed. C11o 95–RELA u- sions ha e been obse ed o occu in wo- hi ds o pedia ic cases o sup a en o ial ependymomas and a e belie ed o be oncogenic due o inc eased NF-kB signaling [17]. Fu he mo e, a sub ype o ce ebella ependymomas ha is associa ed wi h young pa ien age and poo p og- nosis is cha ac e ized by a CpG island me hyla o pheno- ype (CIMP) and Polycomb ep essi e complex 2 d i en ime hyla ion o H3K27. These umo s a e esponsi e o pha macological he apies a ge ing epigene ic egula o s [18]. The au ho s also highligh ed he low a e o ecu - en mu a ions and copy numbe al e a ions in ce ebella ependymomas. Fu he mo e, FGFR al e a ions ha e no been epo ed in high- h oughpu sequencing s udies wi h he excep ion o FGFR1 missense mu a ion N544 K [17] localized o he y osine kinase domain o FGFR1. Pilocy ic as ocy oma (PA), he mos common b ain neoplasm in he pedia ic popula ion, is classi ied as WHO g ade I [19, 20]. They a ise mos commonly in he ce ebellum, b ains em and he op ic ne e. Familial PAs a e cha ac e ized by inac i a ion o he neu o ib oma osis 1(NF1) umo supp esso gene, while ac i a ing BRAF usions and mu a ions a e ypical o spo adic PAs [19]. BRAF al e a ions subsequen ly lead o ac i a ion o he MEK-ERK pa hway [19], which is also an impo an downs eam signalling pa hway o FGFR-induced sig- naling [19, 21]. Addi ionally, FGFR1-TACC1 usion has been epo ed in a BRAF wild- ype pilocy ic as ocy- oma o he diencephalon and se e al s udies ha e epo ed oncogenic s uc u al FGFR1 a ian s wi h du- plica ion o he y osine kinase domain [6, 7]. Fu he - mo e, app oxima ely 5% o PAs ha bo an FGFR1 mu a ion a ge ing codons Asn546 o Lys656 in he kinase domain [7]. The Lys656 mu a ion has been associa ed wi h dec eased pa ien su i al [22]. Mos FGFR1-mu an umo s s udied ha e been ex a- ce ebella , loca ed mos ly in midline loca ions, and mu- ually exclusi e wi h BRAF, NF1, and o he ecu en MAPK pa hway al e a ions [7, 22]. Al hough hese s udies did no epo mu a ions o s uc u al a ian s in FGFR3, hey emphasized he u ili y o FGFR1 as a ma ke o PA sub yping. In di use gliomas, FGFR3 p o ein le el is an in o m- a i e ma ke o usion s a us [34]. Mos umo s in a co- ho o 791 cases did no ha e any de ec able FGFR3 p o ein exp ession, and all he usion-posi i e cases we e s ongly s ained (s aining sensi i i y 100% and speci ici y 88% in he a ge ed sequencing coho ). In non-di use gliomas, FGFR1 al e a ions a e commonly p esen in a subg oup o pilocy ic as ocy omas ha lack o he yp- ical MAPK pa hway al e a ions [6, 7], bu FGFR1 and FGFR3 exp ession le els ha e no been sys ema ically e alua ed. Fu he mo e, FGFR usions o inc eased FGFR p o ein exp ession le els ha e no , o da e, been e- po ed o occu in ependymomas. In he p esen s udy, we sough o in es iga e he clinical signi icance o FGFR3 and FGFR1 exp ession in wo di e en nondi - use gliomas: ependymomas and pilocy ic as ocy omas. We used immunohis ochemis y o de ec FGFR1 and FGFR3 p o ein le els in ependymomas and pilocy ic as ocy omas, and e alua ed he ela ionship be ween p o ein exp ession le els, clinical ea u es and selec ed gene ic al e a ions. Me hods Pa ien samples This s udy was app o ed by he E hical Commi ee o Tampe eUni e si yHospi aland heNa ionalAu- ho i y o Medico-legal A ai s in Finland. The s udy coho included 108 ependymal umo s om 88 pa ien s, 97 pilocy ic as ocy omas om 97 pa ien s (Table 1). Ependymoma pa ien s unde wen neu osu gical ope - a ion wi h he in en ion o g oss adical umo esec ion be ween 1984 and 2009 a Tampe e Uni e si y Hospi al, Leh inen e al. BMC Cance (2017) 17:310 Page 2 o 12 be ween 1979 and 1998 a Kuopio Uni e si y Hospi al, and be ween 1986 and 1999 a Tu ku Uni e si y Hospi al, Finland. The clinical da a de ail abou adicali y o umo esec ion is impe ec , bu adical esec ion has always been pe o med when possible o he pa ien . G ade I umo s included 17 myxopapilla y ependymomas and 1 subependymoma. G ade II umo s included 68 epen- dymomas, while G ade III umo s included 22 anaplas- ic ependymomas, as classi ied acco ding o WHO c i e ia [23]. Pilocy ic as ocy oma pa ien s unde wen umo su ge y a he Tampe e Uni e si y Hospi al be ween 1985 and 1999, a he Kuopio Uni e si y Hospi al be ween 1980 and 1992, a he Tu ku Uni e si y Hospi al be ween 1981 and 1992, and a he Helsinki Uni e si y Hospi al be ween 1986 and 1993. Tissue his opa hology and mic oa ays Tumo samples we e ixed in o maldehyde (bu e ed wi h 4% phospha e) and embedded in pa a in. The samples we e p ocessed in o pa a in blocks and sec ions we e s ained wi h hema oxylin and eosin (H&E). His o- pa hological yping and g ading, e alua ion, and iden i i- ca ion o his ologically ep esen a i e umo egions on each slide we e pe o med by an expe ienced neu o- pa hologis . Tissue mic oa ay (TMA) blocks we e cons uc ed using ep esen a i e sample egions and a cus om-buil ins umen (Beeche Ins umen s, Sil e Sp ing, MD, USA). The diame e o he issue co e on he mic oa ay block was 0.6 o 1 mm, depending on he TMA ype. Fi e-mic ome e - hick sec ions we e cu om ep esen a i e a ay pa a in blocks. Immunohis ochemis y Pa a in was emo ed wi h hexane. A e ehyd a ion in e hanol, he p e-p ocessing s age was pe o med using Ta ge Re ie al Solu ion ci a e bu e (Dako). The sam- ples we e s ained using abbi monoclonal FGFR1 an i- body (#9740, Cell Signaling Technology, 1:100 dilu ion) and mouse monoclonal FGFR3 an ibody (sc-13,121, San a C uz Bio echnology, 1:600 dilu ion). ‘En ision + Sys em- ho se adish pe oxidase and diaminobenzidine (DAB)’ki (Dako) was used o FGFR3. The nuclei we e s ained wi h hema oxylin. A mouse monoclonal an ibody MIB- 1 (Ki-67 an igen, dilu ion 1:40, Immuno ech, S.A. Ma seille, F ance) was used o analyze cell p oli e a ion. The issue sec ions we e coun e s ained wi h me hyl g een. The pe cen age o issue MIB-1-posi i e nuclei was quan i a i ely e alua ed using a compu e -assis ed image analysis sys em (CAS-200 TM So wa e, Bec on Dickinson & Co., USA) and ImmunoRa io analysis. Only neoplas ic cells we e included in he analysis (nec o ic and hemo hagic a eas we e omi ed). The in ensi y o FGFR3 and FGFR1 immunoposi- i i y was sco ed by wo obse e s (HH and KG) on a scale om 0 o 3 as ollows: 0 (no s aining), 1 (weak immunos aining), 2 (mode a e immunos aining), o 3 (s ong immunos aining). S a is ical analysis All da a we e analyzed using R packages o IBM SPSS s a is ics 21.0 so wa e (SPSS Inc., Chicago, IL, USA) o Windows. Tes s o pai wise associa ion be ween disc e e a iables we e pe o med using Fishe ’s exac es o coun da a. Fo ables la ge han 2 × 2, he p- alues o Fishe ’s exac es s we e calcula ed using Mon e Ca lo simula ion wi h 2.5*10^7 eplica es. p- alues we e no co ec ed o mul iple es ing. Log- Table 1 Pa ien demog aphics and clinical cha ac e is ics wi hin ependymoma and pilocy ic as ocy oma umo pa ien coho s Ependymomas Pilocy ic as ocy omas Pa ien s 88 80 Male 48 42 Female 40 38 Age (yea s) Median (Mean ± SD) 37 (35 ± 21) 9 (14 ± 14) Minimum 1 0 Maximum 73 58 Follow-up o p ima y umo pa ien s Su i o s in he end o he ollow-up 60 69 Follow-up ime o su i o s (m) (median (mean ± SD)) 125 (135 ± 82) 70 (111 ± 89) 5-yea esidi e- ee su i al (%) 71 82 5-yea su i al (%) 82 93 Tumo s 108 80 P ima y 74 73 Second 14 5 Thi d 14 1 Fou h-six h 6 1 His ological g ade I1880 II 68 0 III 22 0 Topog aphy Sup a en o ial 35 3 In a en o ial 28 69 Spinal 43 2 C anial ne e 0 6 Pa ien age and ollow-up in o ma ion we e calcula ed using p ima y cases. Follow-up imes a e shown in mon hs (m) SD s anda d de ia ion Leh inen e al. BMC Cance (2017) 17:310 Page 3 o 12 ank es was used o he analysis o p ognos ic ac o s. In cox eg ession analysis, cox model was buil using a s epwise o wa d likehood- a io es ing. Ta ge ed sequencing All he issue samples we e o malin ixed and pa a in embedded (FFPE). A u XTRAC FFPE DNA ki (Co a is) o AllP ep DNA/RNA Mini Ki (Qiagen) was used o DNA isola ion. We used 1 μg o ex ac ed DNA o a ge ed sequencing using he Su eselec XT Ta ge en ichmen sys em oge he wi h cus om- designed RNA p obes (Addi ional ile 1: Table S1). The sequencing lib a y was p epa ed acco ding o he ki ins uc ions (200 ng o DNA samples) wi h a sho e DNA-shea ing p o ocol (220 s) and sequenced wi h MiSeq (Illumina). Tumo s Epe002 and Epe003 we e de i ed om he i s and he hi d umo su ge y (a e second ecu ence) o one pa ien . In addi ion, he u- mo s Epe004 (1s umo su ge y) and Epe005 (2nd umo su ge y) we e de i ed om a sepa a e ependymoma pa ien . The esul ing da a we e aligned agains he GRCh37 human e e ence genome using Bow ie 2.2.4 [24]. Mu a- ions we e iden i ied in umo samples by sea ching o si es wi h an al e na e allele ac ion o a leas 10%, and a leas 5 eads wi h he mu a ion. Addi ionally, he al- lele ac ion was equi ed o be 20 imes highe han he backg ound e o a e (i.e., he a e age allele ac ion ac oss con ol blood samples om heal hy pa ien s). P o ein-le el consequences o a ian s we e p edic ed using ANNOVAR so wa e ool [25]. Mu a ions wi h a known o suspec ed pa hological unc ion we e iden i- ied manually. To disco e ch omosomal ea angemen s o usion de ec ion, unaligned eads om each sample we e spli in o wo 30 bp ancho s (one om bo h ends) ha we e aligned o he hg38 genome using Bow ie- 1.1.2. Disco dan ancho pai s we e g ouped by posi ion, and g oups wi h 8 o mo e suppo ing eads we e lagged as ea angemen candida es and manually cu- a ed using IGV and BLAT. Log a ios o amplicon ead coun s we e used o DNA copy numbe calling. Di e ences in a e age co e - age be ween samples we e co ec ed on he basis o con- ol amplicons in ch omosomes 5, 8, 11, and 18 (14–21 amplicons pe ch omosome), posi ioned in egions wi h he lowes a e o epo ed copy numbe al e a ions. Blood-de i ed DNA om heal hy indi iduals was used as a nega i e con ol o he copy numbe analysis. Resul s We used an an ibody ha a ge s amino acids 25–124 in he FGFR3 N- e minus o pe o m immunohis ochemi- cal (IHC) s aining on 188 cases including ependymomas o pilocy ic as ocy omas (Table 1). FGFR3 s aining was localized o he cy oplasm and plasma memb ane (Fig. 1). S aining was ypically he e ogeneous in all umo ypes s udied. Nega i ely s ained blood essels p o ided an in e nal con ol o an ibody speci ici y. No mal b ain issue was immunonega i e, wi h he excep ion o he ce ebella and ce eb al molecula laye s, whe e weak- o-mode a e s aining was obse ed (Addi ional ile 1: Figu e S1a). In ependymomas, FGFR3 s aining is associa ed wi h disease agg essi eness Immunohis ochemis y was used o in es iga e FGFR3 exp ession le els in 108 ependymal umo samples ap- plied o TMAs. The TMA coho (Table 1), ep esen ing di e en g ades o ependymomas and disease sub ypes, has been pa ly epo ed p e iously [26]. FGFR3 immu- no eac i i y was de ec ed in 27 (37%) o he cases; 11 (15%) showed weak immunos aining, 11 (15%) showed mode a e immunos aining and 5 (7%) we e s ongly immunoposi i e. Inc eased s aining was also obse ed in pseudo ose e s uc u es (Addi ional ile 1: Figu e S1b). Recu en umo s showed ypically simila s aining le els as he p ima y umo . Wi h espec o he associa ion analysis (Addi ional ile 1: Figu e S2), FGFR3 s aining was signi ican ly associa ed wi h a highe umo g ade (p< 0.01, Fishe ’s exac es , Fig. 1b, Table 2). None o he g ade I cases showed de ec able FGFR3 exp ession. Mode a e- o-s ong FGFR3 immunos aining was p e- dominan ly de ec ed in ce eb al umo s as compa ed o o he loca ions (p< 0.001, Fishe ’s exac es , Fig. 1c, Table 2). Ele a ed FGFR3 immunoposi i i y in high- g ade ce eb al umo s sugges s ha FGFR3 immuno- s aining may be ypical o pedia ic ependymomas. Indeed, pa ien s wi h age < 20 yea s a umo onse had a highe equency o FGFR3 immunoposi i e s aining (p< 0.05, Fishe ’s exac es , Fig. 1d). Cases wi h mode a e- o-s ong FGFR3 immunos aining end o show a high p oli e a ion a e (Fig. 1e), al hough his associa ion was no s a is ically signi ican (p= 0.07, Fishe ’s exac es ). Impo an ly, mode a e- o-s ong FGFR3 immunos aining was signi ican ly associa ed wi h sho e o e all pa ien su i al (p< 0.05, log- ank es , Fig. 1 ) and sho e ime o umo ecu ence (p< 0.01, log- ank es , Fig. 1g). The associa ion wi h disease- ee su i al emained signi ican a e adjus men o umo loca ion, g ade, and p oli e a ion (p= 0.003, RR = 1.82, 95% CI 1.23–2.68 o FGFR3, o he a iables no signi ican in he inal equa ion, N= 77, s epwise Cox eg ession), bu only umo loca ion (p= 0.022, RR = 2.47, 95% CI 1.42– 5.34, N= 77, s epwise Cox eg ession) was a signi ican p ognos ic p edic o o disease-speci ic su i al in mul i- ac o ial analysis. I is ele an o no e he pa ien numbe s (N= 77) a e a he low o mul i ac o ial analysis using ou di e en a iables. S ill, he ob ained esul s sugges ha Leh inen e al. BMC Cance (2017) 17:310 Page 4 o 12 a) b) c) d) e) ) g) Fig. 1 Mode a e- o-s ong FGFR3 immunos aining was p edic i e o poo pa ien su i al in ependymomas. aRep esen a i e s aining images. bDis ibu ion o FGFR3 immunos aining in g ade I–III ependymomas. FGFR3 immunos aining was posi i ely associa ed wi h umo g ade (p< 0.01, Fishe ’s exac es ). cMode a e- o-s ong FGFR3 immunos aining was associa ed wi h ce eb al umo loca ion (p< 0.0001, Fishe ’s exac es ). To al numbe o umo s o each loca ion is ma ked in o he igu e. dMode a e- o-s ong FGFR3 exp ession was mo e common in younge pa ien s (p < 0.05, Fishe ’sexac es ). Only newly-diagnosed cases we e included in he analysis and hese we e di ided in o hose wi h nega i e- o-weak s. mode a e- o-s ong FGFR3 immunos aining. eCases wi h mode a e- o-s ong FGFR3 exp ession ended o ha e highe p oli e a ion index (p= 0.07, Fishe ’sexac es ).Sampleswe e di ided based on FGFR3 s aining and p oli e a ion a e (1: low, 2: in e media e, and 3: high p oli e a ion index). -gMode a e- o-s ong FGFR3 immunos aining was associa ed wi h wo se g) disease-speci ic su i al (N= 73, p< 0.05, log- ank es ) and g) ecu ence- ee su i al (N= 70, p< 0.01, log- ank es ). Only newly-diagnosed cases we e included in o he analysis Leh inen e al. BMC Cance (2017) 17:310 Page 5 o 12 FGFR3 immunoposi i i y is associa ed wi h mo e agg essi e ependymomas. As pedia ic and adul ependymomas di e in many espec s and he age associa ion migh in luence he ob- se ed associa ions, we analyzed he pedia ic and adul sample coho s independen ly. Pa ien s ha we e a leas 16 yea s old we e conside ed as adul s acco ding o gen- e al p ac ice in Finnish pedia ic clinics. The e we e 35 pedia ic and 73 adul samples in ou coho . Mode a e- o-s ong FGFR3 s aining was sligh ly mo e common in pedia ic han adul samples (34.3% s 13.7%, p= 0.055, Fishe ’s exac es , Table 2). In pedia ic pa ien s, mode - a e FGFR3 immunos aining was obse ed in ce ebella (31%, n= 16) and ce eb al (29%, n= 14) umo s and s ong FGFR3 s aining only in ce eb al umo s (21%, n= 14), whe eas all he spinal cases (n= 5) we e nega- i e o FGFR3 (p= 0.065, Fishe ’s exac es ). FGFR3 s aining was no associa ed wi h umo g ade o p oli e - a ion index in pedia ic ependymomas. In adul s, FGFR3 associa ions we e la gely e y simila as in he whole sample coho : s onge FGFR3 s aining was associa ed wi h umo g ade (p< 0.01, n= 73, Fishe ’s exac es ), umo loca ion (p< 0.001, n= 71, Fishe ’s exac es ) and he e was a close- o-signi ican associa ion wi h p o- li e a ion index (p= 0.095, n= 66, Fishe ’s exac es ). P ognos ic associa ions we e mos ly nonsigni ican in sepa a e su i al analyses in pedia ic (n= 14) and adul (n= 30) sample coho s, bu his was likely due o low sample coun in he analysis, as he end emain he simi- la . O no e, when FGFR3 s aining was di ided in o ou g oups, i was associa ed wi h wo se disease-speci ic (p< 0.01, log- ank es ) and disease- ee (p<0.001,log- ank es ) su i al in pedia ic pa ien s. FGFR1 s aining is associa ed wi h highe umo g ade and ce eb al loca ion The in e p e a ion o he FGFR1 immunos aining da a was no as s aigh o wa d as FGFR3 s aining, pa ly be- cause mac ophages, neu ons, and nec o ic a eas showed immunoposi i e s aining. The e o e, FGFR1 immunohis- ochemical sco ing was based on he p esence o FGFR1-posi i e malignan cell clus e s o la ge umo a eas (i.e. di use s aining), and sco ing o indi idual cells was omi ed in he analysis. Spo adic mode a e- o- s ong FGFR1 immunoposi i i y was also de ec ed and cha ac e ized by high ou lie exp ession in indi idual malignan cells. These obse a ions suppo hose om p e ious epo s [27]. FGFR1 s aining was de ec ed in he cy oplasm and memb ane compa men s, while g anula s aining was also obse ed in a subpopula ion o posi i ely-s ained samples. In e es ingly, mode a e- o- s ong FGFR1 immunos aining was also obse ed in ependymal ose es (Addi ional ile 1: Figu e S3). Di use FGFR1 immuno eac i i y was de ec ed in 42 (58%) o ependymal umo s. Twen y- ou cases (33%) showed weak immunos aining, 15 (21%) cases showed mode a e immuno eac i i y, and 3 (4%) cases showed s ong immunoposi i i y (Fig. 2a). Consis en wi h FGFR3 exp ession, FGFR1 immunos aining was signi i- can ly associa ed wi h a highe umo g ade (p< 0.05, Fishe ’s exac es , Fig. 2b, Table 2) and ce eb al loca ion (p< 0.01, Fishe ’s exac es , Fig. 2c, Table 2). Di use FGFR1 s aining was no signi ican ly associa ed wi h o e all o ecu ence- ee su i al bu cases wi h high FGFR1 exp ession had a endency owa d dec eased su - i al a es in his coho (Addi ional ile 1: Figu e S4). When ependymomas we e di ided in o pedia ic (n=34) and adul (n= 72) pa ien s, no associa ions we e obse ed o FGFR1 in he pedia ic coho . Howe e , FGFR1 s aining was simila ly associa ed wi h umo loca ion (p< 0.001, n= 70, Fishe ’sexac es )andhighe umo g ade (p<0.01,n= 72, Fishe ’s exac es ) in he adul coho as in he whole sample coho . Fu - he mo e, a weak associa ion was obse ed be ween s onge FGFR1 s aining and highe umo p oli e a ion index (p=0.061,n=68,Fishe ’s exac es ) among adul pa ien s. FGFR1 and/o FGFR3 le els a e ele a ed in majo i y o he ce eb al ependymomas Among ependymomas, ma ked (mode a e- o-s ong) immunos aining o FGFR1, FGFR3, o bo h p o eins oc- cu ed mo e equen ly in ce eb al han in non-ce eb al umo s (76, 32, and 19% in ce eb al, ce ebella , and spinal umo s, espec i ely, p< 0.001, Fishe ’s exac es , Fig. 2d). Table 2 Samples numbe s in FGFR1 low, FGFR1 high, FGFR3 low, and FGFR3 high g oups in espec o umo loca ion, umo g ade and pa ien age FGFR1 low FGFR1 high FGFR3 low FGFR3 high Tumo loca ion Spinal 37 6 41 2 Ce ebella 21 4 23 5 Ce eb al 16 20 20 15 p- alue 0.0001 0.0002 Tumo g ade I 16 2 18 0 II 50 15 54 14 III 10 13 14 8 p- alue 0.002 0.013 Pa ien age <162212 2312 > =16 50 18 61 10 p- alue 0.15 0.055 p- alues ha e been calcula ed using Fishe ’s exac es . High: Mode a e- o-s ong immunos aining, Low: Nega i e- o-low immunos aining Leh inen e al. BMC Cance (2017) 17:310 Page 6 o 12 0 20406080100 G ade 1 G ade 2 G ade 3 % o cases n=23 n=65 n=12 Spinal Ce ebella Ce eb al 0 20 40 60 80 100 nega i e weak mode a e s ong %o cases n=43 n=25 n=36 n=42 n=25 n=34 1.0 0.8 0.6 0.4 0.2 0.0 0 100 200 300 400 p<0.05 1.0 0.8 0.6 0.4 0.2 0.0 0 100 200 300 400 O e all su i al Recu ence- ee su i al FGFR1+FGFR3 low (N = 43) FGFR1 high (N = 11) FGFR3 high (N = 8) FGFR1+FGFR3 high (N = 7) a) b) c ) d) e) ) p < 0.05 Spinal Ce ebella Ce eb al 0 20 40 60 80 100 %o cases nega i e weak mode a e s ong FGFR1+FGFR3 high FGFR3 high FGFR1 high FGFR1+FGFR3 low FGFR1 FGFR1+FGFR3 low (N = 41) FGFR1 high (N=11) FGFR3 high (N=7) FGFR1+FGFR3 high (N = 7) Fig. 2 Mode a e- o-s ong FGFR1 and/o FGFR3 exp ession is cha ac e is ic o agg essi e ependymomas. aRep esen a i e images o FGFR1 s aining in ependymomas. bThe dis ibu ion o FGFR1 immunos aining in g ade I-III ependymomas. FGFR1 s aining was associa ed wi h highe umo g ade (p< 0.05, Fishe ’s exac es ). cMode a e- o-s ong FGFR1 immunos aining was associa ed wi h ce eb al umo loca ion (p< 0.01, Fishe ’s exac es ). To al numbe o umo s o each loca ion is ma ked in o he igu e. dMode a e- o-s ong immunos aining o FGFR1 and/o FGFR3 was de ec ed in a majo i y o ce eb al ependymoma samples (p< 0.0001, Fishe ’s exac es ). e- ). Mode a e- o-s ong immunos aining o bo h FGFR3 and FGFR1 was associa ed wi h e) poo disease-speci ic su i al (N= 69, p< 0.05, log- ank es ) and wo se ecu ence- ee su i al (N= 66, p< 0.05, log- ank es ). Newly diagnosed cases we e di ided in o ou ca ego ies based on he exp ession o bo h FGFR1 and FGFR3. High: Mode a e- o-s ong immunos aining, Low: Nega i e- o-low immunos aining Leh inen e al. BMC Cance (2017) 17:310 Page 7 o 12 Inc eased FGFR1 and/o FGFR3 exp ession was he e o e a common cha ac e is ic o ce eb al umo s. S ikingly, umo issues exp essing ma ked (mode a e- o-s ong) le els o bo h FGFR1 and FGFR3 we e associa ed wi h sig- ni ican ly wo se pa ien su i al han issues ob ained om o he cases, in e ms o bo h o e all mo ali y (p< 0.05, log- ank es , Fig. 2e) and ecu ence- ee su - i al (p< 0.05, log- ank es , Fig. 2 ). Fu he mo e, he combined a iable o FGFR1 and FGFR3 (bo h a e nega i e- o weak, ei he s aining is mode a e- o-s ong o bo h a e mode a e- o-s ong) was he only signi ican p e- dic o o he disease-speci ic su i al (p= 0.014, RR = 1.91, 95% CI 1.14–3.20, N= 77, s epwise Cox eg ession) and disease- ee su i al (p= 0.007, RR = 1.75, 95% CI 1.17– 2.62, N= 77, s epwise Cox eg ession), when i was com- bined oge he wi h umo loca ion, g ade, and p oli e a- ion index as explana o y ac o s in he mul i ac o ial analysis. I is good o emembe ha he pa ien numbe s (N= 77) a e a he low o mul i ac o ial analysis using ou di e en a iables when in e p e ing hese esul s. S ill, he ob ained esul s suppo he agg essi e na u e o umo s wi h mode a e- o-s ong s aining o bo h FGFR1 and FGFR3. Ou esul s a e also conco dan wi h p e ious no ions (e.g. [28]) ha sup a en o ial and in a en o ial ependymomas a e la gely di e en and appea o ep e- sen dis inc umo en i ies. FGFR3 s aining is associa ed wi h inc eased pa ien age in pilocy ic as ocy oma In he pilocy ic as ocy oma coho , 60 (82%) samples we e nega i e o FGFR3 exp ession, while only 21 cases (22%) ailed o show any FGFR1 exp ession (Fig. 3c-d). Among samples wi h FGFR3 immuno eac i i y, 7 sam- ples (9%) showed weak immunos aining, 5 samples (6%) showed mode a e immunos aining, and 2 samples (3%) we e s ongly immunoposi i e. Immunoposi i e FGFR3 s aining was de ec ed in bo h mic ocys ic and pilocy ic a eas. Among samples wi h posi i e FGFR1 s aining, 59 samples (61%) showed weak immunoposi i i y, 16 sam- ples (16%) samples showed mode a e immunoposi i i y, and 1 sample (1%) was s ongly immunoposi i e. Mode a e- o-s ong FGFR1 immunos aining was de- ec ed p edominan ly in mic ocys ic a eas. Clinical asso- cia ion analysis (Addi ional ile 1: Figu e S5) did no e eal any signi ican associa ions be ween FGFR1 s aining and o he clinical ac o s. In e es ingly, mode a e- o-s ong FGFR3 p o ein le els we e associ- a ed wi h inc eased pa ien age (≥16 yea s, p< 0.01, Fishe ’s exac es , Fig. 3e). All bu one o he six p i- ma y cases showing mode a e- o-s ong FGFR3 im- munos aining we e om pa ien s who we e a leas 15 yea s old. FGFR3 immunos aining was no associ- a ed wi h umo loca ion o aneuploidy. Absence o FGFR1 o FGFR3 usions in a ge ed sequencing coho Ten umo s showing mode a e- o-s ong FGFR1 o FGFR3 immunos aining we e selec ed o a ge ed sequencing analysis. All analyzed ependymomas we e sup a en o ial. In addi ion o FGFR3 and FGFR1, he sequencing panel in- co po a ed genes wi h epo ed al e a ions in gliomas, in- cluding IDH1, IDH2,TP53,ATRX,CIC,CDKN2A, RB1, RELA, and BRAF (Addi ional ile 1: Table S1). We did no de ec FGFR coding mu a ions o usions in any o he samples (Fig. 4, Addi ional ile 2: Table S2, Addi ional ile 1: Figu e S6). FGFR3 usions we e de ec ed wi h high sensi i i y om la ge di use glioma coho using he same sequencing panel and me hodology [34], sug- ges ing ha he lack o de ec able FGFR usions was no due o me hodological limi a ions. The umo s se- lec ed o analysis con ained many known al e a ions, in- cluding a C11o 95-RELA usion and CDKN2A al e a ions in ependymoma umo s (Epe001, Epe002 and Epe003). RELA usions and loss o CDKN2A ha e been ou inely obse ed in agg essi e ependymomas [17, 29, 30]. A TERT p omo e mu a ion was obse ed in umo s Epe004 and Epe005 ob ained om he same ependymoma pa ien . In addi ion, one pilocy ic as ocy oma umo ha bo ed he KIAA1549-BRAF usion, which is he mos equen MAPK pa hway al e a ion in his umo ype [7]. I is in- e es ing ha majo i y o sequenced PA samples did no ca y any BRAF o FGFR1 al e a ions, bu limi ed sample size does no allow ull gene aliza ion o his esul . A o al o 4 cases in ou coho did no ca y any al e a ions in a ge ed genes. This may be due, in pa , o he ac ha all genomic egions we e no co e ed du ing a ge ed se- quencing. In addi ion, pilocy ic as ocy omas a e known o ha bo e y ew al e a ions [7]. Discussion Ou esul s demons a e ha mode a e- o-s ong FGFR3 and/o FGFR1 immunos aining was de ec able in mos o he sup a en o ial ependymomas. In ependymoma, mode a e- o-s ong FGFR3 s aining was associa ed wi h umo loca ion, highe p oli e a ion index, and highe g ade. Simila associa ions we e ob ained when only adul pa ien s we e included in o he analysis. Mode a e- o-s ong FGFR3 s aining was mo e equen ly obse ed among pedia ic pa ien s han among adul s, bu only he associa ion be ween FGFR3 and umo loca ion emained signi ican in he pedia ic coho . This migh be pa ly due o a small numbe o pedia ic cases (n= 35) and sho age o g ade I umo s (n= 1) among child en. In any case, he da a sugges ha clinical asso- cia ions o FGFR3 we e no solely due o age- ela ed di e ences. The si ua ion was simila o FGFR1: mode a e- o-s ong s aining was associa ed wi h umo loca ion and highe g ade in bo h he whole and he Leh inen e al. BMC Cance (2017) 17:310 Page 8 o 12 adul coho , despi e he lack o clinical associa ions in he pedia ic coho . Tumo s wi h high exp ession o bo h FGFR3 and FGFR1 we e associa ed wi h poo clinical p ognosis in ependymoma, sugges ing ha agg essi e sup a en o ial ependymomas may bene i om ea men egimens based on FGFR inhibi ion. Addi ional wo k is equi ed o elucida e he signi icance o high FGFR1 and/o FGFR3 exp ession as independen p ognos ic ac o s o ea men esponse. The absence o FGFR al e a ions in hese umo s does no ule ou he possibili y o ea - men esponse. In head and neck squamous cell cance s and a ious lung cance s, FGFR1 exp ession has, in ac , been shown o p edic ea men esponses be e han 0 20406080100 FGFR1 FGFR3 nega i e weak mode a e s ong % o cases 0 20 40 60 80 100 nega i e weak mode a e s ong % o cases 16 yea s <16 yea s FGFR3 FGFR1 FGFR3 Posi i e Nega i e Posi i e Nega i e a) b) c) N=22 N=54 n=80 n=97 Fig. 3 FGFR3 and FGFR1 s aining in pilocy ic as ocy oma. aRep esen a i e immunohis ochemical images in pilocy ic as ocy oma. bDis ibu ion o immunohis ochemis y sco es. The majo i y o samples we e nega i e o FGFR3. cNea ly all o he pilocy ic as ocy oma samples showing mode a e- o-s ong FGFR3 immunos aining we e ob ained om non-pedia ic pa ien s (p< 0.01, Fishe ’s exac es ). Only newly-diagnosed umo s we e included in o his analysis Fig. 4 Summa y o gene ic al e a ions in he cases ha we e analyzed using a ge ed sequencing. No coding mu a ions o gene usions we e de ec ed in FGFR3 o FGFR1. FGFR1 and FGFR3 immunohis ochemical s aining sco es a e shown abo e he igu e. I s ained whole-moun issue slides we e a ailable, hey we e used o sco ing. Pilocy ic: pilocy ic as ocy oma Leh inen e al. BMC Cance (2017) 17:310 Page 9 o 12