Loss of CADM1 expression is associated with poor prognosis and brain metastasis in breast cancer patients

Onco a ge 3076 www.impac jou nals.com/onco a ge www.impac jou nals.com/onco a ge / Onco a ge , Vol. 5, No. 10 Loss o CADM1 exp ession is associa ed wi h poo p ognosis and b ain me as asis in b eas cance pa ien s Ha ie Wikman1,*, Lau a Wes phal1,*, Felici as Schmid1,2, Si kku Polla i3, Jolan he K opidlowski1, Be ina Siela -F impong1, Ma kus Gla zel4, Jakob Ma schke4, Man ed Wes phal5, K is iina Iljin3, Heini Huh ala6, Luigi Te acciano7, Anne Kallioniemi8, Guido Sau e 9, Volkma Mülle 10, Isabell Wi zel10, Ka in Lamszus5, Di k Kemming11, Klaus Pan el1 1 Ins i u e o Tumo Biology, Uni e si y Medical Cen e Hambu g-Eppendo , Hambu g, Ge many 2 Max-Delb ück Cen e o Molecula Medicine, Be lin, Ge many 3 Medical Bio echnology, VTT Technical Resea ch Cen e o Finland and Tu ku Cen e o Bio echnology, Uni e si y o Tu ku, Tu ku, Finland 4 Ins i u e o Neu opa hology, Uni e si y Medical Cen e Hambu g-Eppendo , Hambu g-Eppendo , Hambu g, Ge many 5 Depa men o Neu ological Su ge y, Uni e si y Medical Cen e Hambu g-Eppendo , Hambu g-Eppendo , Hambu g, Ge many 6 School o Heal h Sciences, Uni e si y o Tampe e, Tampe e, Finland 7 Depa men o Pa hology, Basel Uni e si y Clinics, Basel, Swi ze land 8 Ins i u e o Biomedical Technology, Uni e si y o Tampe e and BioMediTech, Tampe e, Fimlab Labo a o ies, Tampe e, Finland 9 Ins i u e o Pa hology, Uni e si y Medical Cen e Hambu g-Eppendo , Hambu g, Ge many 10 Depa men o Gynecology, Uni e si y Medical Cen e Hambu g-Eppendo , Hambu g, Ge many 11 Eu opean Labo a o y Associa ion, Ibbenbue en, Ge many * These au ho s sha e i s -au ho ship Co espondence o: Ha ie Wikman, email: [email p o ec ed] Keywo ds: b eas cance , b ain me as ases, CADM1, me hyla ion Recei ed: Janua y 22, 2014 Accep ed: Ma ch 16, 2014 Published: Ma ch 16, 2014 This is an open-access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal au ho and sou ce a e c edi ed. ABSTRACT: B eas cance b ain me as ases (BCBM) a e de ec ed wi h inc easing incidence. In o de o de ec po en ial genes in ol ed in BCBM, we i s sc eened o genes down- egula ed by me hyla ion in cell lines wi h si e-speci ic me as a ic abili y. The exp ession o i e genes, CADM1, SPARC, RECK, TNFAIP3 and CXCL14, which we e also ound down- egula ed in gene exp ession p o iling analyses o BCBM issue samples, was e i ied by qRT-PCR in a la ge pa ien coho . CADM1 was chosen o u he down-s eam analyses. A highe incidence o CADM1 me hyla ion, co ela ing wi h lowe exp ession le els, was ound in BCBM as compa ed o p ima y BC. Loss o CADM1 p o ein exp ession was de ec ed mos commonly among BCBM samples as well as among p ima y umo s wi h subsequen b ain elapse. The p ognos ic ole o CADM1 exp ession was inally e i ied in ou la ge independen b eas cance coho s (n=2136). Loss o CADM1 p o ein exp ession was associa ed wi h disease s age, lymph node s a us, and umo size in p ima y BC. Fu he mo e, all analyses e ealed a signi ican associa ion be ween loss o CADM1 and sho e su i al. In mul i a ia e analyses, su i al was signi ican ly sho e among pa ien s wi h CADM1-nega i e umo s. Loss o CADM1 exp ession is an independen p ognos ic ac o especially associa ed wi h he de elopmen o b ain me as ases in b eas cance pa ien s. Onco a ge 3077 www.impac jou nals.com/onco a ge INTRODUCTION B eas cance (BC) is he mos common non-skin malignancy in women a ec ing abou 1.2 million women in he wo ld each yea . The sp ead o malignan cells om he p ima y umo o dis an o gans such as he b ain is he main cause o BC ela ed dea hs. B eas cance is he second mos common cause o he de elopmen o cen al ne ous sys em (CNS) me as ases, which a e, i espec i ely o he p ima y umo o igin, mo e commonly diagnosed han p ima y b ain umo s [1, 2]. CNS-me as ases a e diagnosed in 15-20% o pa ien s wi h me as a ic BC and a e usually occu ing wi h a median ime pe iod o 31 mon hs a e he diagnosis o b eas cance [2]. Due o imp o ed sys emic ea men op ions o b eas cance , pa ien s bene i om p olonged su i al a es. Howe e , consecu i ely o he longe su i al and an in ensi ied use o sensi i e de ec ion me hods such as con as -enhanced magne ic esonance imaging o he b ain (cMRI), he incidence a es o b ain me as ases a e ising. B ain me as ases a e o en associa ed wi h he agg essi e iple nega i e b eas cance (TNBC), being ho mone ecep o (HR) and HER2 nega i e, and also wi h HER2 posi i e p ima y b eas cance s [3, 4]. Pa ien s su e ing om b eas cance b ain me as ases (BCBM) ha e an ex emely poo p ognosis wi h a median su i al o only 7.8 mon hs [3]. This b ings o h o in es iga e new p ognos ic ma ke s, which may be help ul o he iden i ica ion o pa ien s wi h an inc eased isk o subsequen CNS-in il a ion and o de elop new he apeu ic app oaches. The loss o umo supp esso gene (TSG) exp ession is known o cons i u e a c ucial s ep in cance o ma ion. Fo he dissemina ion and ou g ow h o me as ases, ano he se o me as asis supp esso and ac i a o genes is needed. Me as asis supp esso genes (MSGs) usually do no in luence umo g ow h a he p ima y si e, bu con ol he umo cells’ capaci y o escape om he p ima y umo and o m o e me as ases a dis an si es [5]. MSGs con ibu e o do mancy con ol and/o ou g ow h a seconda y si es and hus egula e he inal s ep o me as asis, he me as a ic coloniza ion. In e es ingly, in con as o TSGs, MSGs seem o be mo e o en cance ype speci ic. Fu he mo e, MSGs a e a ely mu a ed, and epigene ic e en s, such as me hyla ion, a e hus likely he main cause o hei loss o unc ion [6]. Since epigene ic e en s a e e e sible, do mancy con ol by MSGs is po en ially a new o m o a ge ed gene he apy ( e iewed in [6, 7 ]). The aim o his s udy was o iden i y no el genes in ol ed in BCBM o ma ion. We i s sc eened o genes down- egula ed by me hyla ion in b eas cance cell lines (pa en al MDA-MB-231 cell line and he b ain- and bone- speci ic sub cell lines) known o si e-speci ic me as asis [8]. Gene exp ession p o iling and quan i a i e eal- ime RT-PCR (qRT-PCR) alida ion o he candida e genes we e he ea e conduc ed on umo issues om p ima y BC and BCBM in o de o iden i y he mos clinically ele an genes. The me hyla ion pa e n o CADM1 was u he mo e cha ac e ized, and CADM1 p o ein exp ession was alida ed in wo la ge independen p ima y umo coho s as well as in BCBM samples and co ela ed wi h clinico-pa hological pa ame e s. RESULTS Me hyla ion a ay sc eening o b ain me as ases ela ed genes A subclone o MDA-MB-231 wi h a high me as a ic po en ial o he b ain, MDA-MB-231 BR, was compa ed o he pa en al MDA-MB-231 and o a bone-seeking a ian MDA-MB-231 SA in o de o iden i y genes, which migh be speci ically in ol ed in b ain me as asis o ma ion. The cell lines we e ea ed wi h 5-Aza-2’-deoxycy idine, a deme hyla ing agen , in o de o ind genes po en ially down- egula ed by me hyla ion. Mic oa ay analysis was pe o med on pooled iplica e expe imen s and he non- umo igenic epi helial cell line MCF 10A was used o con ol o s ess esponse a e he ea men wi h 5-Aza- 2’-deoxycy idine. The gene exp ession p o iling a e 5-Aza-2’- deoxycy idine ea men e ealed 914 di e en ansc ip s, which we e signi ican ly up- egula ed in one o he MDA- MB-231 cell lines bu no al e ed in MCF 10A (Figu e 1A). The la ges numbe o up- egula ed genes (691 ansc ip s) and cell line-speci ic up- egula ed genes (20%) was ound in he MDA-MB-231 BR cell line. Mos o he genes we e, howe e , up- egula ed in all o hese cell lines (30%, 279/914). In gene al, he mo e agg essi e subclones BR and SA we e mo e simila o each o he han o he pa en al cell line, indica ing a di e en ia ion in o gene ally mo e agg essi e o ms. Gene exp ession sc eening o me hyla ion- ela ed genes in p ima y and me as a ic b eas umo s Gene exp ession p o iling o p ima y non- me as asized b eas umo s (n=32) and b ain me as ases (n=9) was pe o med o ind he mos ele an genes among he 690 po en ially b ain me as ases de e mining genes de ec ed in me hyla ion a ay sc eening. 110 ansc ip s (16%) o hose up- egula ed in he MDA- MB-231 BR subclone in esponse o 5-Aza2’- deoxycy idine ea men we e ound signi ican ly lowe exp essed among he BCBM samples as compa ed o he samples om non- elapsed p ima y BC pa ien s (Supplemen a y Table 2). Twen y- ou (22%) o hese genes we e ound Onco a ge 3078 www.impac jou nals.com/onco a ge exclusi ely in he BR subclone, implica ing a b ain speci ic down- egula ion o hese genes. 28 genes we e in common be ween he highly me as a ic BR and SA a ian s, indica ing ha hese genes may media e a mo e agg essi e beha io in gene al. Valida ion o po en ial me as asis-supp essing genes in p ima y BC and BCBM issue samples The exp ession o i e genes, which we ound down- egula ed in he b eas cance da a se and in he cell line analyses, was u he in es iga ed by qRT-PCR in 39 p ima y BC samples wi hou b ain me as ases and 20 BCBM issue samples. SPARC, RECK, TNFAIP3 and CXCL14 we e down- egula ed (p < 0.05) in BCBM as compa ed o p ima y BC samples i espec i ely o he cance sub ype, while o CADM1, his co ela ion was ound in HR posi i e and TNBC samples (Figu e 1B and Supplemen a y Figu e 1). Wi hin he p ima y BC samples, 84% o he HER2 nega i e pa ien s showed an ele a ed CADM1 mRNA exp ession ( op 75% pe cen ile), whe eas only 29% o he HER2 posi i e pa ien s had a high CADM1 exp ession (p = 0.003) (Supplemen a y Figu e 1). Among he p ima y umo s, high RECK exp ession was also associa ed wi h HR posi i e s a us, whe eas he highes SPARC exp ession was s a is ically signi ican ly linked o iple nega i e (TNBC) samples. CXCL14 exp ession was no associa ed wi h a sub ype, bu i s low exp ession was associa ed wi h bo h posi i e lymph node s a us (p = 0.011) and high g ade (p = 0.029) (Supplemen a y Table 3). F equency o CADM1 me hyla ion in p ima y b eas cance and BCBM issue samples The me hyla ion s a us o CADM1 was de e mined by MSP in 17 BCBM and 14 p ima y BC samples. CADM1 was homozygously me hyla ed in 17.5% (3/17), Figu e 1:A) Gene exp ession changes in esponse o 5-Aza-2’-deoxycy idine ea men in pa en al MDA-MB-231 b eas cance cell line and MDA-MB-231 BR and MDA-MB-231 SA a ian s. The numbe be o e b acke s de ines he genes up- egula ed in each cell line a e 5-Aza-2’-deoxycy idine ea men . The numbe in b acke s de ines he numbe o genes also down- egula ed in he b ain me as ases as compa ed o non- elapsed p ima y umo s. Figu e 1B) Exp ession o CADM1, RECK, CXCL14, SPARC and TNFAIP3 in BCBM samples ela i e o p ima y b eas umo s (PT). HR pos.: es ogen and p oges e one posi i e ecep o , TNBC: iple nega i e b eas cance . P- alues we e de e mined by he log ank es . Onco a ge 3079 www.impac jou nals.com/onco a ge he e ozygously me hyla ed in 17.5% (3/17) and no me hyla ed in 65% (11/17) o he BCBM samples. In con as , p ima y BC samples showed no homozygous CADM1 me hyla ion, only 7% (1/14) he e ozygous and 93% (13/14) WT CADM1 s a us (Figu e 2). Due o he low equency o me hyla ion in he p ima y BC samples, no associa ion be ween clinico-pa hological ac o s and he me hyla ion s a us o CADM1 could be ound. CADM1 p o ein exp ession in p ima y and me as a ic b eas cance CADM1 p o ein exp ession could be assessed in a la ge numbe o p ima y b eas umo s (TMA I, n= 1331) and 27 BCBM samples (Table 1, and Figu e 3). No mal b eas duc showed an in ensi e memb anous s aining o CADM1, whe eas he umo s oma was CADM1 nega i e. The CADM1 exp ession in he plasma memb ane was los in 42% o he p ima y BC and in 68% o he BCBM samples. S ong exp ession was eco ded in 22% o he p ima y BC and in 11% o he BCBM cases. The CADM1 p o ein exp ession di e ed be ween p ima y umo and BCBM wi h mo e nega i e s aining among he BCBM samples (p = 0.011). Su p isingly, a nuclea s aining o CADM1 was seen in a small subse o pa ien s. Nuclea s aining has been p e iously epo ed o occu in ce ical epi helium wi h he localiza ion o s aining being dependen on epi helial o igin [9]. 86 (6.5%) o he p ima y BC samples on he p ognos ic TMA (TMA I) had a clea CADM1 s aining in he nucleus. The nuclea s aining highly co ela ed wi h he memb anous s aining (p < 0.001). In 70 o he 86 pa ien s wi h nuclea s aining, also he memb ane was s ained o CADM1 (81% conco dance). In addi ion, 43 ma ched pai s o p ima y umo s and lymph node me as asis samples (TMA III) we e analyzed o CADM1 p o ein exp ession. 70% (30/43) o he ma ched pai s showed a conco dan CADM1 exp ession wi h 51% (22/43) o he p ima y umo and lymph node samples being CADM1 nega i e and 19% (8/43) CADM1 posi i e in bo h ypes o issues. Only 7% (3/43) o he samples had a highe CADM1 exp ession in he lymph nodes as compa ed o he ma ched p ima y umo s, while down- egula ion o CADM1 exp ession in he lymph node me as ases was obse ed in 23% (10/43) o he ma ched pai s, indica ing a loss o CADM1 p o ein exp ession du ing he me as a ic cascade. Compa ison o gene exp ession, me hyla ion pa e n and p o ein exp ession s a us o CADM1 in p ima y BC and BCBM In o de o ind ou how much o he silencing o exp ession is go e ned by p omo e me hyla ion, we compa ed he esul s om he qRT-PCR, IHC and me hyla ion analyses. In 22 samples esul s om he gene exp ession, me hyla ion and p o ein exp ession s a us o CADM1 we e a ailable (Table 2). Homozygous me hyla ion was associa ed wi h a nega i e CADM1 s aining and low o in e media e mRNA exp ession. He e ozygo e me hyla ion pa e n showed a nega i e IHC esul in 50% o he cases and o he es Figu e 2: P omo e CpG me hyla ion o CADM1. A) Gel elec opho esis pic u e o CADM1 in 5 ep esen a i e cases, MW= molecula weigh ma ke , UM= unme hyla ed PCR, M= me hyla ed PCR. B) Tables o MSP esul s o CADM1 o p ima y umo s and b eas cance b ain me as ases. WT= wild ype, HET= he e ozygous me hyla ion, MET= homozygous me hyla ion. Figu e 3: CADM1 immunos aining in p ima y b eas cance and BCBM samples. A) P ima y BC sample wi h homogenous posi i e CADM1 memb ane and nega i e nuclea s aining, B) P ima y BC sample wi h nega i e CADM1 memb ane and weak nuclea s aining, C) BCBM sample wi h nega i e CADM1 memb ane and nuclea s aining wi h posi i ely s ained e y h ocy es, D) BCBM sample wi h weak he e ogeneous CADM1 memb ane and nega i e nuclea s aining. Onco a ge 3080 www.impac jou nals.com/onco a ge Table 1: CADM1 memb ane p o ein exp ession in co ela ion o clinical pa ame e s in p ima y BC (TMA I) and BCBM. CADM1 nega i e CADM1 weak CADM1 s ong p- alue n % n % n % B ain me as ases All 19 67.9 6 21.4 3 10.7 P ima y umo s All 557 42.1 488 36.9 286 21.6 His ology n.s. Duc al 408 42.1 361 37.2 201 20.7 Lobula 71 46.4 51 33.3 31 20.3 o he s 78 37.7 76 35.2 54 27.1 Tumo s age <0.001 pT1 118 27.6 185 43.0 125 29.0 pT2 290 44.8 224 34.6 133 20.6 pT3 40 52.6 25 32.9 11 14.5 pT4 106 60.6 52 29.7 17 9.7 n.a. 3 2 0 Lymph node s a us 0.003 pN0 193 35.8 219 40.6 127 23.6 pN1 207 44.6 159 34.3 98 21.1 pN2 43 53.8 28 35.0 9 11.3 n.a. 114 82 52 G ade n.s. 1 130 42.6 113 37.0 62 20.3 2 201 40.4 188 37.8 108 21.7 3 203 45.8 151 34.1 89 20.1 n.a. 23 36 27 Tumo size <0.001 < 2.0 cm 120 27.9 185 43.0 125 29.1 > 2.0 cm 421 48.4 290 33.3 159 18.3 n.a. 16 2.9 13 2.7 2 0.7 Ho mone ecep o n.s. nega i e 126 44.3 87 32.8 65 22.9 posi i e 410 41.0 379 37.9 210 21.0 n.a. 21 22 11 HER2 n.s. nega i e 438 40.4 403 37.2 242 22.3 posi i e 93 45.6 73 35.8 38 18.6 n.a. 26 12 6 Sub ype 0.036 ER/PR pos. 350 40.4 332 38.3 185 21.3 TNBC 81 42.4 55 28.8 55 28.8 HER2 pos. 93 45.6 73 35.8 38 18.6 n.a. 33 28 8 Age 0.028 < 50 yea s 85 37.1 85 37.1 59 25.8 > 50 yea s 379 44.8 310 36.6 157 18.6 n.a. 93 93 70 Cou se o disease <0.001 dead 238 51.7 145 31.7 76 16.6 ali e 319 36.6 343 39.3 210 24.0 n.a. = no a ailable; n.s. = no signi ican Onco a ge 3081 www.impac jou nals.com/onco a ge a weak posi i e CADM1 p o ein s aining was obse ed. Nega i e CADM1 p o ein exp ession was always associa ed wi h ei he low/ nega i e o in e media e mRNA exp ession. Clinical signi icance o CADM1 p o ein exp ession The clinical signi icance o CADM1 exp ession was assessed in wo publicly a ailable mRNA exp ession da a se s (GSE3494 and GSE6532) and on wo p ognos ic TMAs o p ima y BC samples (Figu e 4). In bo h exp ession a ay da a se s a signi ican associa ion be ween low CADM1 exp ession and bad p ognosis was ound (p = 0.033 and p = 0.001). The wo p ognos ic TMAs (TMA I and TMA II) o p ima y BC ga e highly consis en esul s. Clinico- pa hologic examina ion o bo h independen pa ien coho s consis ing o 1718 p ima y b eas umo samples e ealed a signi ican associa ion be ween nega i e CADM1 s a us and ad anced umo s age, posi i e lymph node s a us and la ge umo size (all p < 0.05; Table 1 and Supplemen a y Table 4). In he TMA I a signi ican associa ion be ween CADM1 and age, sub ype and cou se o disease was also de ec ed. The equency dis ibu ion o CADM1 was compa able on he TMA II bu signi icance was no eached due o smalle sample numbe s. G ade Figu e 4: Kaplan–Meie analysis o CADM1 mRNA and p o ein exp ession in p ima y b eas cance . A) Su i al analyses in wo publicly a ailable exp ession da a se s. B) Su i al analyses o he p ognos ic TMA (TMA I) o he whole s udy popula ion and among HR-posi i e pa ien s. C) Su i al analyses in he second TMA (TMA II) o he whole s udy popula ion and among HR-posi i e pa ien s. Su i al di e ences we e analyzed by he log ank es . Onco a ge 3082 www.impac jou nals.com/onco a ge was associa ed wi h CADM1 loss on he second TMA, whe eas no di e ence be ween he di e en his ological sub ypes could be ound on ei he TMA. The si e o elapse was eco ded o 353 p ima y umo pa ien s (TMA II). Among pa ien s wi hou elapse a loss o CADM1 was seen in 58.6%, whe eas he highes equency o CADM1 loss was seen among pa ien s su e ing om b ain elapse wi h 81.1%. Also o he ypes o elapse had a highe equency o CADM1 loss compa ed o he non- elapsed (bone 73.8% (p = 0.023), li e 69.7% and lung 65.5%; Supplemen a y able 4). Consis en wi h he memb anous s aining, a s a is ically signi ican associa ion be ween loss o CADM1 in he nucleus and mo e agg essi e o ad anced umo s age and size, as well as o he ho mone ecep o s a us and he cou se o disease (p < 0.05) was ound (only assessed on TMA I, Supplemen a y Table 5). In addi ion, he majo i y o cases wi h nuclea CADM1 s aining was ound among lobula umo s (11% s. 6% in duc al, p = 0.042). A highly signi ican associa ion be ween wo se pa ien ou come (o e all su i al and disease ee su i al) and nega i e CADM1 p o ein exp ession could be ound in bo h da a se s (p < 0.001 and p = 0.006) (Figu e 4). Loss o CADM1 was signi ican ly associa ed wi h wo se pa ien ou come among duc al ca cinomas (p = 0.001 and p = 0.001), whe eas a signi ican associa ion was ound o lobula ca cinomas only in TMA II (p = 0.026; TMA I p = 0.068) (da a no shown). Fu he mo e, loss o CADM1 exp ession was signi ican ly associa ed wi h sho e su i al in HR posi i e (bo h da a se s p = 0.001 and p = 0.022, Figu e 4), bu did no ha e a p ognos ic ele ance in HER2 posi i e pa ien s, indica ing a umo supp essing e ec o CADM1 in HER2 nega i e pa ien s only (da a no shown). Mul i a ia e analysis showed ha loss o CADM1 exp ession was a signi ican independen nega i e p ognos ic ac o (TMA I p = 0.045 and TMA II p = 0.01). Fo he i s la ge s udy coho (TMA I) he i e-yea su i al was 64.8% (median 105 mon hs) o pa ien s wi h no CADM1 p o ein exp ession and 75.9% (154 mon hs) and 77.7% o pa ien s wi h weak and s ong CADM1 exp ession espec i ely. Fo TMA II he i e-yea su i al was 84.7% (median 179.7 mon hs) o pa ien s wi h no CADM1 p o ein exp ession and 92.3% (189 mon hs) and 95.2% (221 mon hs) o pa ien s wi h weak and s ong CADM1 exp ession, espec i ely. DISCUSSION The incidence o b ain me as ases in b eas cance pa ien s is inc easing, wi h no s anda d diagnos ic managemen o hese pa ien s applied in clinical ou ine. In iew o he inc easing success o a ge ed he apies in ea ing BC pa ien s wi h dis an me as ases, imp o ed insigh s in o he pheno ype o b ain me as ases could ha e impo an he apeu ic implica ions. In he p esen s udy, we i s sc eened o genes silenced by me hyla ion in a b eas cance cell line wi h a high po en ial o o m b ain me as ases. Fi e o he iden i ied genes, also down- egula ed in clinical samples, CADM1, RECK, SPARC, CXCL14 and TNFAIP3, we e chosen o u he exp ession analyses. These genes ha e p e iously no been associa ed wi h b ain me as asis o ma ion. Howe e , hey ha e been implica ed as cance - ele an genes in epi helial umo s [10-14]. Fu he mo e, he me hyla ion s a us o CADM1 in p ima y BC and BCBM was in es iga ed and CADM1 p o ein exp ession was examined in co ela ion wi h clinico-pa hological pa ame e s in BC and compa ed o BCBM issue samples. We iden i ied CADM1 as an impo an p ognos ic ac o , whose loss was no only associa ed wi h mo e agg essi e p ima y BC wi h wo se ou come, bu also wi h an inc eased isk o b ain me as asis o ma ion. Table 2: Combined esul s om he qRT-PCR, MSP, and IHC analyses o CADM1 CADM1 Pa . No. Sample qPCR me hyla ion IHC- esul B M-9 BCBM INT MET NEG B M-10 BCBM LOW MET NEG B M-8 BCBM INT HET NEG B M-7 BCBM LOW HET NEG B M-5 BCBM LOW WT NEG B M-25 BCBM LOW WT NEG B M-19 BCBM LOW WT NEG B M-1 BCBM INT WT NEG B M-16 BCBM INT WT NEG B M-18 BCBM INT WT NEG B M-22 BCBM INT WT NEG B M-2 BCBM INT WT NEG PT-95 PT INT WT NEG PT-103 PT INT WT NEG B M-13 BCBM n.d. WT NEG B M-21 BCBM INT HET WEAK PT-88 PT INT HET WEAK B M-15 BCBM INT n.d. WEAK B M-14 BCBM HIGH n.d. WEAK B M-24 BCBM LOW WT WEAK PT-101 PT INT WT STRONG B M-11 BCBM HIGH WT STRONG BCBM: b eas cance b ain me as ase; PT: p ima y umo ; HIGH: CT exp ession alue in uppe qua ile in all pa ien s analyzed; LOW: CT exp ession alue in bo om qua ile in all pa ien s analyzed n.d.: no de ined; WT: wild ype; HET: he e ozygous me hyla ion; MET: homozygous me hyla ion; NEG: nega i e p o ein s aining Onco a ge 3083 www.impac jou nals.com/onco a ge CADM1/TSLC1 (cell adhesion molecule 1) is a memb ane-spanning glycop o ein belonging o he supe amily o immunoglobulin cell adhesion molecules. I was i s ecognized as a umo supp esso in non-small cell lung cance (NSCLC) [10]. I has been sugges ed ha he dis up ion o cell adhesion h ough he loss o CADM1 is a mechanism leading o cance cell in asion and me as asis [15, 16]. CADM1 is known o be in ol ed in inhibi ion o cell p oli e a ion and induc ion o apop osis wi h a epo ed loss o exp ession in a a ie y o cance s o epi helial cell o igin such as b eas , p os a e, panc ea ic, hepa ocellula and colo ec al cance , bu also in neu oblas oma ( e iewed in [16]). Recen ly, CADM1 was iden i ied by Fa aji e al. (2012) o be a me as asis suscep ibili y gene, i.e. an inhe i ed ac o ha supp esses me as asis by sensi izing umo cells o immunosu eillance by CD8+ T-cells [17]. We ound a down- egula ion o CADM1 bo h on mRNA and p o ein le el in BCBM issue samples as compa ed o p ima y BC. I was p e iously epo ed ha CADM1 p o ein is de ec ed on he cell memb ane in no mal epi helial cells o he b eas and ha nega i e CADM1 s aining, de ec ed in mo e han hal o he p ima y BCs, is associa ed wi h ad anced disease s ages [18, 19]. He e, we ound ha in p ima y BC, loss o CADM1 was mos commonly seen among TNBC and HER2 posi i e pa ien s, wo sub ypes mo e closely associa ed wi h b ain me as ases. Fu he mo e, loss o CADM1 p o ein exp ession was associa ed wi h isk ac o s such as high umo s age, posi i e lymph node s a us and la ge umo size (all p < 0.05) in wo la ge independen sample coho s consis ing o mo e han 1 700 umo samples, indica ing a p ognos ic ole o CADM1 in p e en ing b eas cance p og ession. Loss o CADM1 exp ession was ound mo e commonly among p ima y umo pa ien s wi h lung, li e , bone o b ain elapse compa ed o pa ien s wi h no elapse. Impo an ly, loss o CADM1 exp ession was mos commonly seen among pa ien s wi h b ain elapse indica ing a special bu no exclusi e ole o CADM1 in b ain me as asis o ma ion. Simila ly, ma ched pai s o p ima y BC and lymph node (LN) me as asis samples we e addi ionally analyzed o CADM1 p o ein exp ession, showing mo e equen ly a down- egula ion o CADM1 exp ession in he LN me as ases. The obse ed down- egula ion o CADM1 no only in BCBM bu also in LN me as ases as compa ed o he p ima y umo suppo s he ole o CADM1 as a me as asis suscep ibili y gene in e ms o educing he me as a ic capabili y (25). This hypo hesis is u he suppo ed by ou esul s showing an associa ion be ween CADM1 posi i e s aining in p ima y umo s and longe o e all su i al. Mul i a ia e analysis showed ha CADM1 exp ession is an independen p ognos ic ma ke , s eng hening he hypo hesis ha CADM1 may play an impo an ole in p e en ing he me as a ic p og ession. In line wi h he esul s om ou exp ession s udies, he CADM1 p omo e egion was ound me hyla ed in 35% o he BCBM samples, whe eas in p ima y BC samples wi h good p ognosis, CADM1 p omo e was only me hyla ed in 7% o he samples. Howe e , o he mechanisms esul ing in a loss o gene unc ion, such as dele ion, need o be u he in es iga ed. Clea ly, he p edic i e alue o CADM1 in b ain me as ases o ma ion needs o be es ed on la ge ma ched samples se s in u u e s udies. In addi ion o CADM1, we also analyzed he exp ession o ou o he genes (RECK, SPARC, TNFAIP3 and CXCL14) in p ima y BC and BCBM. All ou genes we e ound down- egula ed in BCBM pa ien s as compa ed o p ima y BC. Fu he s udies will be needed in o de o assess hei biological and clinical ole in b ain me as ases o ma ion. RECK p o ein is capable o inhibi ing ma ix me allop o einases and plays an impo an ole in emb yogenesis and asculogenesis [11]. RECK is o en down- egula ed in a ious p ima y epi helial cance s and gliomas, and he down- egula ion co ela es wi h poo p ognosis ( e iewed in [11, 20]). Recen ly, Hill e al (2011) showed ha RECK down- egula ion by me hyla ion in BC is associa ed wi h elapse and poo su i al [21]. Fu he mo e, Hsu e al. showed ha HER2 can ep ess RECK exp ession in o de o p omo e cell in asion, which is consis en wi h ou esul s showing he lowes exp ession o RECK among HER2 posi i e pa ien s [22]. SPARC/ os eonec in (sec e ed p o ein acidic and ich in cys eine) is a ma icellula calcium-binding glycop o ein in ol ed in cell adhesion, cell-ma ix in e ac ions du ing issue emodelling, cell p oli e a ion, mig a ion, angiogenesis and apop osis [12, 23]. Recen ly, Nagai e al. [24] desc ibed a co ela ion be ween educed SPARC exp ession and ad anced clinical s age and poo ou come in p ima y BC. SPARC exp ession in me as a ic issue has no been epo ed. Bo h CXCL14 (CXC mo i e ligand 14) and TNFAIP3/A20 ( umo nec osis ac o α induced p o ein 3) a e in ol ed in immuno egula o y and in lamma o y p ocesses. CXCL14 is a p o-mig a o y chemokine and plays an impo an ole in umo ecogni ion by he immune sys em [13]. TNFAIP3 is a zinc- inge p o ein in ol ed in he cy okine-media ed immune and in lamma o y esponse mainly h ough he inhibi ion o nuclea ac o NF-κB (NFκB) ac i a ion and umo nec osis ac o (TNF)-media ed apop osis [25]. Down- egula ion o CXCL14 and TNFAIP3 has been desc ibed in a ious p ima y umo en i ies including b eas umo s [13, 14, 26, 27], bu o ou knowledge, hei ole in me as asis has no been epo ed. In conclusion, ou s udy shows a down- egula ion o CADM1, RECK, SPARC, CXCL14 and TNFAIP3 in b ain me as ases samples and especially implies CADM1 as an impo an p ognos ic ac o , which is commonly los in BC sub ypes wi h poo ou come and e e s o an inc eased isk o b ain me as asis o ma ion. The mechanisms leading o Onco a ge 3084 www.impac jou nals.com/onco a ge a loss o CADM1 exp ession a e a leas pa ially induced by p omo e hype me hyla ion, a mechanism po en ially e e sible by he apeu ic in e en ion. As loss o CADM1 seems o ep esen in gene al a mo e agg essi e disease wi h a highe isk o elapse especially o he b ain, hese pa ien s migh need a mo e equen ollow-up including diagnos ics such as egula clinical con ols and ea ly cMRI o he b ain. MATERIALS AND METHODS Cell lines and pa ien ma e ial Fo he me hyla ion sc eening analysis, he human b eas cance cell line MDA-MB-231 (pa en al) and i s bone- (SA) and b ain-seeking (BR) a ian s we e used. MDA-MB-231 was ob ained om ATCC and passaged o a maximum o 6 mon hs. MDA-MB- 231(SA) cells we e ob ained om P o . The esa Guise and we e comp ehensi ely cha ac e ized by compa a i e genomic hyb idiza ion and genome-wide gene exp ession p o iling as desc ibed in Polla i e al. (2011) [28]. The b ain me as a ic MDA-MB-231 a ian was ob ained om P o . Toshiyuki Yoneda and was cha ac e ized in [29]. A non- umo igenic epi helial cell line MCF 10A (ATCC, USA) was used as a con ol (see supplemen a y ma e ial). Au hen ica ions o all cell lines we e conduc ed by sho andem epea (STR) p o iling o exclude c oss- con amina ion be ween he cell lines. Fo he qRT-PCR and me hyla ion speci ic PCR (MSP) analyses, umo issue om 29 ea ly s age p ima y BC pa ien s and 21 BCBMs we e collec ed om esh ozen specimens (see supplemen a y ma e ial) [30]. Pa ien samples we e ob ained a e su gical esec ion a he Uni e si y Medical Cen e , Hambu g-Eppendo (UKE), Ge many. Fo he immunohis ochemis y (IHC) analysis, h ee di e en se s o Tissue Mic o A ays (TMA) we e used. The i s la ge p ognos ic a ay (TMA I) consis ed o 2197 p ima y b eas cance specimens ope a ed in Basel, Swi ze land [31]. Follow-up da a we e a ailable o all pa ien s wi h a median ollow-up ime o 61.0 mon hs ( ange 7-171 mon hs). The second TMA (TMA II) consis ed o 243 in asi e duc al ca cinomas and 243 in asi e lobula ca cinomas ope a ed in Tampe e, Finland, wi h a maximum ollow-up pe iod o 19.8 yea s con aining in o ma ion on he si e and ime o umo ecu ence and o e all su i al [32]. A hi d TMA om UKE (TMA III) con ained 89 cases wi h ma ched p ima y and lymph node me as asis samples [33]. In addi ion, whole sec ion slides om 28 BCBM pa ien s we e also analyzed by IHC (see supplemen a y ma e ial). The REMARK c i e ia we e ollowed o he pa ien and s udy se up, howe e no in o ma ion abou he ea men could be collec ed. All sample dono s om UKE ga e w i en in o med consen o biological esea ch in o hei samples as app o ed by he e hics commi ee o he chambe o physicians, Hambu g, Ge many. The use o he Finnish umo samples and pa ien eco ds was app o ed by he E hics Commi ee o he Pi kanmaa Hospi al Dis ic . All clinical in es iga ions ha e been conduc ed acco ding o he p inciples exp essed in he Decla a ion o Helsinki. 5-Aza-2’-deoxycy idine ea men o cells and isola ion o o al RNA and DNA Di e en concen a ions o 5-Aza-2’-deoxycy idine we e es ed o minimal oxici y (MTT assay) bu ull e e sion o me hyla ion. The concen a ion o 1 µM has been ound as he mos e icien one. Cells we e cul u ed in absence o p esence o 1 µM 5-Aza-CdR (n=3), and he medium was changed daily un il he cells eached a con luence o 90-100%. Expe imen s we e done in iplica es using di e en cell passages (see supplemen a y ma e ial). Tumo issue sec ions we e manually dissec ed o ob ain a umo cell con en o a leas 70% [34]. To al RNA was isola ed wi h he RNeasy Mini Ki (Qiagen, Hilden, Ge many). DNA was ex ac ed using InnuPREP DNA Mic oki (Analy ik Jena, Jena, Ge many) (see supplemen a y ma e ial). The success o 5-Aza-2’- deoxycy idine ea men was e i ied by measu ing he exp ession o me hyla ed MAGE-A1 and RUNX3 genes [35, 36]. Genome-wide exp ession analysis The gene exp ession analyses o he cell lines and BCBM samples we e ca ied ou using he Whole Human Genome Oligo Mic oa ay Ki , 4x44K (Agilen Technologies) (see supplemen a y ma e ial). The non- malignan cell line MCF 10A was used as a con ol o geno oxic s ess esponse o he 5-Aza-2’-deoxycy idine ea men . The MIAME guidelines we e ollowed in sample, a ay and da a p ocessing (see supplemen a y ma e ial). All a ay da a a e a ailable a h p://www.ncbi. nlm.nih.go /geo GSE44354. Po en ially me hyla ed genes in BC cells we e de ined using he ollowing c i e ia: a) 2- old exp ession di e ence be ween he ea ed and non- ea ed MDA- MB-231 BR cells, b) minimum exp ession alue o 100 in one o he cell lines, and c) no exp ession change in he con ol cell line MCF 10A a e he 5-Aza-CdR ea men . The a ay da a om nine BCBM samples we e compa ed o 32 un ea ed p ima y b eas umo s wi hou elapse p esen in he GEO Da aSe GSE21974 [37]. Di e en ially exp essed genes we e selec ed using he signi icance analysis o mic oa ays (SAM) algo i hm wi h a alse disco e y a e o 5%. Genes wi h exp ession alues abo e 100, which we e a leas 2- old down- egula ed in