Integrated RNA-seq and DNase-seq analyses identify phenotype-specific BMP4 signaling in breast cancer

RESEARCH ARTICLE Open Access In eg a ed RNA-seq and DNase-seq analyses iden i y pheno ype-speci ic BMP4 signaling in b eas cance M. Ampuja 1,2*† , T. Ran ape o 1† , A. Rod iguez-Ma inez 1,2† , M. Palm o h 1 , E. L. Ala mo 1 , M. Nyk e 1 and A. Kallioniemi 1,2 Abs ac Backg ound: Bone mo phogene ic p o ein 4 (BMP4) plays an impo an ole in cance pa hogenesis. In b eas cance , i educes p oli e a ion and inc eases mig a ion in a cell line-dependen manne . To cha ac e ize he ansc ip ional media o s o hese pheno ypes, we pe o med RNA-seq and DNase-seq analyses a e BMP4 ea men in MDA-MB-231 and T-47D b eas cance cells ha espond o BMP4 wi h enhanced mig a ion and dec eased cell g ow h, espec i ely. Resul s: The RNA-seq da a e ealed gene exp ession changes ha we e consis en wi h he in i o pheno ypes o he cell lines, pa icula ly in MDA-MB-231, whe e mig a ion- ela ed p ocesses we e en iched. These esul s we e con i med when en ichmen o BMP4-induced open ch oma in egions was analyzed. In e es ingly, he ch oma in in ansc ip ion s a si es o di e en ially exp essed genes was al eady open in uns imula ed cells, hus enabling apid ec ui men o ansc ip ion ac o s o he p omo e s as a esponse o s imula ion. Fu he analysis and unc ional alida ion iden i ied MBD2, CBFB, and HIF1A as downs eam egula o s o BMP4 signaling. Silencing o hese ansc ip ion ac o s e ealed ha MBD2 was a consis en ac i a o o a ge genes in bo h cell lines, CBFB an ac i a o in cells wi h educed p oli e a ion pheno ype, and HIF1A a ep esso in cells wi h induced mig a ion pheno ype. Conclusions: In eg a ing RNA-seq and DNase-seq da a showed ha he pheno ypic esponses o BMP4 in b eas cance cell lines a e e lec ed in ansc ip omic and ch oma in le els. We iden i ied and expe imen ally alida ed downs eam egula o s o BMP4 signaling ha ela e o he di e en in i o pheno ypes and hus demons a e ha he downs eam BMP4 esponse is egula ed in a cell ype-speci ic manne . Keywo ds: Bone mo phogene ic p o ein, B eas cance , NGS, RNA-seq, DNase-seq, T ansc ip ion ac o Backg ound Despi e many ad ances in diagnos ics and he apeu ics, b eas cance emains he leading cause o cance dea h in women [1]. Bone mo phogene ic p o eins (BMPs) a e a g oup o g ow h ac o s ha a e impo an playe s du ing de elopmen [2, 3] bu also con ibu e o cance o ma ion and p og ession [4–6]. As a sub amily o he ans o ming g ow h ac o β(TGF-β) p o ein supe am- ily, BMPs a e ex acellula ligands ha bind as dime s o hei speci ic ansmemb ane ecep o s and ac i a e he in acellula SMAD signaling pa hway leading o phos- pho yla ion o ecep o - egula ed SMADs (SMAD1/5/9). The ac i a ed SMADs bind o SMAD4 and he complex ansloca es o he nucleus whe e i egula es he exp es- sion o BMP a ge genes [7, 8]. Al e na i ely, BMP signals a e also media ed h ough he ac i a ion o ERK, JNK and p38 mi ogen-ac i a ed p o ein kinase pa hways [7, 8]. The unc ional consequences o BMP signaling depend on he BMP ligand and issue ype. We and o he s ha e shown ha BMP4 educes he p oli e a ion o b eas can- ce cell lines, while simul aneously inducing mig a ion and in asion in a subse o cell lines [9–11]. Simila * Co espondence: [email p o ec ed] † Equal con ibu o s 1 BioMediTech, Uni e si y o Tampe e, Tampe e, Finland 2 Fimlab Labo a o ies, Tampe e, Finland © 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. Ampuja e al. BMC Genomics (2017) 18:68 DOI 10.1186/s12864-016-3428-1 dualis ic e ec s upon BMP4 s imula ion ha e also been epo ed in o he umo ypes [12]. Conco dan ly, da a om b eas cance pa ien samples poin o a co ela ion be ween ele a ed BMP4 le els and educed p oli e a ion as well as an inc eased isk o ecu ence [13]. These BMP4- ela ed e ec s ha seem ei he de imen al ( educed cell g ow h) o bene icial (inc eased mobili y) o he cance cells a e likely o be media ed by speci ic BMP4 a ge genes. The iden i ica ion o such a ge genes is hus impo an since i may allow gene a ion o e ec i e cance he apies a ge ing each pheno ype independen ly. We ha e p e iously sea ched o BMP4 a ge genes in a se o b eas cance cell lines ha p edominan ly e- spond o BMP4 ea men by educ ion o p oli e a ion [14]. He e, we used nex -gene a ion sequencing (NGS) echnologies (RNA-seq and DNase-seq) o unco e BMP4-media ed ansc ip ional e en s wi h a speci ic ocus on compa ing cells in which BMP4 has opposing e - ec s, namely an ip oli e a i e and p omig a o y. Ou o he nine b eas cance cell lines we ha e p e iously s ud- ied, T-47D shows one o he mos p ominen g ow h educ ions and MDA-MB-231 cells display he mos o e induc ion o mig a ion [9, 10], and we e hus selec ed o his s udy. RNA-seqme hodquan i ies hele elo geneexp es- sion ac oss he genome [15] while DNase-seq allows iden i ica ion o open ch oma in egions ha a e sensi- i e o diges ion by he DNase I endonuclease [16]. Open ch oma in egions a e conside ed as si es whe e ansc ip ional egula ion can ake place since hey a e accessible o egula o y molecules o bind and exe hei unc ion. By combining da a om RNA-seq and DNase-seq, and using addi ional da a analysis ools, i was possible o iden i y candida e ansc ip ion ac o s in ol ed in he obse ed ansc ip ional esponses. This app oach hus p o ides he means o be e unde s and he ansc ip ional e en s ha link BMP4 signaling and i s esul ing pheno ypes. Resul s We pe o med RNA-seq and DNase-seq analyses in wo b eas cance cell lines, T-47D and MDA-MB-231. The cell lines we e ea ed wi h BMP4 and ehicle con ol o 3 h, hus allowing us o speci ically ocus on ea ly e- sponse e en s. Bo h ehicle- and BMP4- ea ed cell lines we e sequenced (see me hods). BMP4-elici ed ansc ip ional egula ion is highly di e gen in he wo b eas cance cell lines wi h di e en unc ional esponses o BMP4 Sequencing eads om RNA-seq and DNase-seq we e aligned o he human genome and u he analyzed as de- sc ibed in he me hods. To con i m ha he wo da ase s we e consis en , we compa ed he ch oma in openness as de e mined by DNase-seq signal a he ansc ip ion s a si e (TSS) o he exp ession le el o he gene as de e - mined by RNA-seq. As expec ed, we ound ha he in- c eased openness o TSS globally co ela ed wi h inc eased gene exp ession (Addi ional ile 1: Figu e S1, Panels A and B). Howe e , he a iance is high, indica ing ha he di e ences in he ch oma in s a e only pa ly ex- plain gene exp ession pa e ns. Nex we compa ed he exp ession le els om RNA-seq be ween he ehicle- and BMP4- ea ed cells. This analysis iden i ied 91 di e en ially exp essed genes (DEGs) in MDA-MB-231, o which 58 we e up egula ed and 33 down egula ed (Addi ional ile 2: Table S1). In T-47D, he e we e 203 DEGs, o which 160 we e up egula ed and 43 we e down egula ed (Addi ional ile 3: Table S2). In o al, 10 DEGs (ATOH8,BDKRB2,BMF,GS1-124 K5.4, ID1,ID2,ID3,SKIL,SMAD6,andSMAD9)we esha ed by he wo cell lines and all o hem we e up egula ed ex- cep GS1-124 K5.4 which was down egula ed in bo h cell lines. To illus a e ha BMP4 induces ma kedly di e gen ansc ip ional esponses in hese wo cell lines, we gene - a ed a hea map o show he exp ession le els o he p o ein-coding DEGs (Fig. 1a). Using he DNase-seq da a, we examined he ch oma in s a us a he ansc ip ion s a si es (TSSs) o hese p o ein-coding DEGs. Fo he majo i y o he cases he ch oma in was open a he TSS be o e BMP4 s imula ion (app oxima ely 86% o all DEGs in bo h cell lines) (Addi ional ile 1: Figu e S1, Panels C and D). Fo he emaining DEGs, we obse ed ei he opening o closing o he TSS a e s imula ion o no change in he closed ch oma in s a us (Fig. 1a). These da a indica e ha , a his ea ly ime poin , he BMP4-induced di e en ial exp ession mainly in ol es genes whose an- sc ip ion does no equi e changes in he ch oma in s a us a TSS. The DEG lis s included a numbe o genes in ol ed in he canonical BMP pa hway. As expec ed, ID1, ID2 and ID3, known BMP4 a ge genes, we e up egula ed in bo h cell lines (Fig. 1b). Simila ly, he ecep o - egula ed SMAD9 was up egula ed in bo h cell lines whe eas no signi ican di e ence in he o he ecep o - egula ed SMADs o SMAD4 exp ession was obse ed. Among he inhibi o y SMADs, SMAD6wasup egu- la ed in bo h cell lines and SMAD7 in T-47D. In addi ion, he BMP ype I ecep o BMPR1A and nega- i e egula o s o BMP signaling, NOG and BAMBI, we e up egula ed in T-47D while in MDA-MB-231 hei exp ession was no signi ican ly changed (Fig. 1b). Thus BMP4 s imula ion leads o exp ession changes ha ing cha ac e is ics o bo h eedback and eed o wa d loops. We hen e alua ed whe he he di e en ially exp essed genes pa icipa e in speci ic biological p ocesses and Ampuja e al. BMC Genomics (2017) 18:68 Page 2 o 15 A B Fig. 1 The RNA-seq and DNAse-seq da a e eal cell line-speci ic esponses o BMP4. aGene exp ession le els o di e en ially exp essed p o ein-coding genes con e ed o log2 scale a e shown o bo h cell lines and ea men s, up egula ed genes on he le and down egula ed genes on he igh . The s a us column deno es he cell line in which he gene is di e en ially exp essed. The igh mos columns indica e he s a us o he ch oma in a ansc ip ion s a si es (TSS) o he DEGs as measu ed by DNAse-seq. bIllus a ion o he di e en ially exp essed componen s o he BMP signaling pa hway upon BMP4 ea men Ampuja e al. BMC Genomics (2017) 18:68 Page 3 o 15 especially assessed whe he he non-common DEGs ha e di e ing unc ions. To his end we used DAVID o sea ch o GO e ms en iched in he se s o non- common p o ein-coding DEGs. In MDA-MB-231, mos o he en iched e ms we e ela ed o cell mig a ion whe eas o gan de elopmen and mo phogenesis as well as in acellula signaling we e he mos signi ican GO e ms in T-47D (Table 1). These indings imply ha he ansc ip ional changes a e indeed likely o explain he dissimila i ies in he pheno ypic esponses o hese wo cell lines o BMP4 ea men . The ea e , we also wan ed o in es iga e whe he he exp ession le els o DEGs could be linked wi h su i al in b eas cance pa ien s. Fo his pu pose, we used he da a publicly a ailable in he TCGA da abase. The e- sul s showed ha 20 DEGs in he MDA-MB-231 and 46 DEGs in he T-47D cells associa ed wi h ei he good o poo p ognosis (Addi ional ile 4: Tables S3 and S4). O he nine sha ed p o ein-coding DEGs, ou (ATOH8, ID3,SMAD6 and SMAD9) we e co ela ed wi h su i al, all being associa ed wi h poo p ognosis. To alida e he esul s o he RNA-seq analysis and o ex end he scope o he s udy beyond he 3 h ime poin in wo cell lines, qRT-PCR was used o s udy he exp ession le els o 15 selec ed DEGs in MDA-MB-231 and T-47D cells as well as in i e addi ional b eas cance cell lines (BT-474, HCC-1954, MCF-7, MDA- MB-361, and MDA-MB-436) and one no mal b eas epi helial cell line (MCF-10A) ea ed wi h BMP4 and ehicle o 3, 6 and 24 h. The genes we e selec ed based on hei exp ession le els and epo ed cance associ- a ion in he li e a u e, and i e o hese we e up egu- la ed acco ding o he RNA-seq in bo h MDA-MB-231 and T-47D. The exp ession pa e ns o he majo i y o he genes showed simila i ies ac oss he cell line panel and ime poin s wi h he clea excep ion o MDA-MB- 436, in which he exp ession changes we e e y limi ed (Fig. 2). Pa icula ly he i e sha ed genes (ATOH8,ID2, SKIL,SMAD6 and SMAD9)aswellasDLX3 we e con- sis en ly up egula ed upon BMP4 ea men h oughou he ime se ies hus con i ming ha hey ep esen common BMP4 a ge genes. The emaining genes showed mo e a iabili y wi h al e ed exp ession ypic- ally in only wo o h ee cell lines, sugges ing ha hei exp ession is likely o be in luenced by ac o s ha a e cell line-speci ic. Ch oma in landscape and dynamics ollowing BMP4 ea men To gain mo e insigh in o he changes o ch oma in s uc- u e du ing BMP4 ea men , we pe o med peak de ec- ion in a genome-wide manne o iden i y he a eas o open ch oma in. The peak de ec ion app oach was bench- ma ked by compa ison o publicly a ailable DNase-seq da a o uns imula ed T-47D cell line om ENCODE (see me hods), showing ha mos o he peaks iden i ied in ou da a a e p esen also in ENCODE samples (Addi ional ile 5: Table S5). A e il e ing p ocedu es (see me hods), he numbe s o iden i ied DNase hype sensi i e si es (DHSs) in he MDA-MB-231 cell line we e 89,830 and 97,349 in e- hicle- and BMP4- ea ed samples, espec i ely. In T- 47D, he co esponding numbe s we e 68,000 and 73,881. To ob ain a uni ied se o peaks o bo h condi- ions, he o e lapping DHSs we e me ged esul ing in a o al o 106,154 DHSs in MDA-MB-231 and 110,028 in T-47D. A e he me ging, he ac ion o sha ed DHSs be ween BMP4 and ehicle con ol in MDA-MB-231 sam- ples was 75% while he ac ion o unique DHSs in he e- hicle was 9% and co espondingly in he BMP4 sample 16% (Addi ional ile 6: Figu e S2). In he T-47D cell line, Table 1 Gene on ology analysis Cell line GO accession GO e m Numbe o genes Adjus ed p- alue MDA-MB-231 GO:0030334 egula ion o cell mig a ion 5 2.0 × 10 −2 GO:0030335 posi i e egula ion o cell mig a ion 4 2.3 × 10 −2 GO:2000145 egula ion o cell mo ili y 5 2.4 × 10 −2 GO:2000147 posi i e egula ion o cell mo ili y 4 2.5 × 10 −2 GO:0051272 posi i e egula ion o cellula componen mo emen 4 2.7 × 10 −2 T-47D GO:0048513 animal o gan de elopmen 45 2.6 × 10 −8 GO:0035556 in acellula signal ansduc ion 41 4.5 × 10 −8 GO:0009887 o gan mo phogenesis 22 4.0 × 10 −7 GO:0009966 egula ion o signal ansduc ion 36 4.5 × 10 −6 GO:0007166 cell su ace ecep o signaling pa hway 34 9.5 × 10 −5 The DAVID Func ional Anno a ion Tools was used o e eal signi ican ly en iched GO ca ego ies among he di e en ially exp essed p o ein-coding genes. The analysis was done independen ly o each cell line and sha ed di e en ially exp essed genes we e omi ed. The op i e biological unc ion GO e ms a e shown Ampuja e al. BMC Genomics (2017) 18:68 Page 4 o 15 he ac ion o sha ed DHSs be ween he wo condi- ionswas27%whe eas he ac iono uniqueDHSsin he ehicle was 34% and in he BMP4 sample 39% (Addi ional ile 6: Figu e S2). Anno a ion o he me ged DHSs o genomic ea u es e ealed a simila dis ibu ion in he wo cell lines in he ehicle- ea ed condi ion, wi h he la ges ac ion (>30%) o DHSs loca ing in in ons (Fig. 3a). When compa ing he dis ibu ions o he BMP4-induced DHSs be ween he cell lines appa en esemblances we e also obse ed. In bo h cell lines, he p opo ion o DHSs associa ed wi h in onic and in e genic egions inc eased a e BMP4 s imula ion wi h a co esponding dec ease a o he gen- omic loca ions, including he p omo e egions (Fig. 3b). Fig. 2 Exp ession le els o selec ed BMP4 a ge genes by qRT-PCR in a b eas cance cell line panel. The exp ession le els o 15 DEGs we e measu ed a e 3, 6 and 24 h o BMP4 ea men in he indica ed cell lines. The colo code illus a es he ela i e exp ession le els in he BMP4- ea ed sample as compa ed o he co esponding ehicle con ol. FC = Fold change, n.a. = mRNA le el oo low o allow eliable measu emen A B Fig. 3 Dis ibu ion o open ch oma in egions. Anno a ion o open ch oma in egions in MDA-MB-231 and T-47D a e (a) ehicle ea men (basal openness) and (b) BMP4 ea men (consis ing only o he ch oma in ha opened a e BMP4 ea men ) Ampuja e al. BMC Genomics (2017) 18:68 Page 5 o 15 To assess he unc ional impac o he BMP4-induced global changes in he ch oma in s uc u e we conduc ed an en ichmen analysis using GREAT [17] which maps he DHSs o pu a i e egula o y egions o genes and conduc s a gene on ology en ichmen analysis. The e- sul s highligh ed e.g. cell mo ili y and o gan mo phogen- esis as en iched biological unc ions o MDA-MB-231 and T-47D, espec i ely (Addi ional ile 7: Tables S6 and S7). These esul s a e consis en wi h hose ob ained by en ichmen analysis o he di e en ially exp essed genes om RNA-seq (Table 1) and he eby sugges ha , o- ge he wi h speci ic a ge genes, BMP4-induced changes a ch oma in le el may con ibu e o he eme gence o he di e en BMP4-media ed pheno ypes. T ansc ip ion ac o binding si e en ichmen analysis in open ch oma in egions o p omo e s e eals ansc ip ion ac o s in ol ed in BMP4 signaling egula ion Based on ou TSS openness analysis (Fig. 1a), a domin- an ea u e o ou da a is ha he ch oma in o he pu a- i e BMP4 a ge genes (iden i ied by RNA-seq) is open al eady in ehicle- ea ed cells. This is u he suppo ed by ou genome-wide peak analysis, whe e he p omo e egions we e no o e ep esen ed a e he ea men (Fig. 3b). The e o e, he al e a ions in he ch oma in s a e only pa ially explain gene exp ession di e ences induced by he BMP4 ea men . Howe e , di e en ial ansc ip ion ac o binding o open p omo e s may ex- plain he di e en esponses in he cell lines. The e o e we pe o med ansc ip ion ac o (TF) mo i binding analysis. To assess which TFs migh be egula o s o he BMP4 esponse, he sequences o open ch oma in si es in he p oximal p omo e s o up egula ed genes we e ana- lyzed wi h a o al o 426 posi ion weigh ma ixes (PWMs), ep esen ing 401 indi idual TFs o TF-complexes (see me hods). Fo each TF we calcula ed an en ichmen sco e (see me hods) o he numbe o binding si es in ei he MDA-MB-231 o T-47D cells. This analysis led o he iden i ica ion o candida e egula o TFs, including mul iple membe s o he SMAD amily o TFs, as expec ed, as well as a numbe o sha ed common egula o TFs. To ocus on biologically ele an candida es, we il e ed ou hose TFs ha we e no exp essed based on ou RNA-seq da a. In addi ion, we included only hose TFs whose binding si es (TFBSs) in open ch oma in egions o he p omo e s o DEGs we e en iched in one and deple ed in he o he cell line. The op 15 TFs ha a e exp essed in bo h cell lines bu ha e a high en ichmen sco e only in one o he cell lines a e lis ed in Tables 2 and 3. Examples o a ge gene p o- mo e s wi h binding mo i s o p edic ed TFs a e shown in Fig. 4a. Fo mo e in-dep h unc ional analysis we selec ed pa icula TFs om he op en iched candida es using he ollowing c i e ia: 1) a binding mo i wi h a quali y ca ego y o A-C in he HOCOMOCO da abase, 2) ele- ance in he con ex o ou model based on li e a u e, 3) no a highly common egula o o pa o a la ge TF amily, and 4) high exp ession le el o he TF (>1000 eads) in a leas one cell line and di e en ial exp es- sion be ween cell lines acco ding o he RNA-seq. The Table 2 Top 15 ansc ip ion ac o s en iched in MDA-MB-231 cells TF name Mo i Selec ion by: TF binding si es Re . si es Expec ed si es in e . Ra io o en ichmen Mean ead coun MYBL2 MYBB_ 1 2, 3, 4 12 2930 6.2 1.92 2197 BACH1 BACH1_si 1, 2, 3 15 3904 8.3 1.81 531 MYC MYC_ 1 1, 2, 4 10 2698 5.7 1.74 3044 MAFK MAFK_si 2, 3 16 4428 9.4 1.70 688 RELA TF65_ 2 1, 2, 4 19 5467 11.6 1.63 1398 PPARA PPARA_ 1 1, 2, 3 9 2747 5.8 1.54 185 NFIA/B/C/X a 1, 2, 3 15 4669 9.9 1.51 b NFIL3 NFIL3_si 1, 2, 3 11 3494 7.4 1.48 474 FOXA2 FOXA2_ 1 1, 2, 3 36 11477 24.4 1.47 434 REL REL_do 1, 2, 3 17 5422 11.5 1.47 69 ZFHX3 ZFHX3_ 1 2, 3 46 14683 31.2 1.47 66 RXRB RXRB_ 1 1, 2, 4 20 6414 13.6 1.47 1015 SMARCC1 SMRC1_ 1 1, 4 20 6443 13.7 1.46 1478 ETV5 ETV5_ 1 2, 3 16 5199 11.1 1.45 641 NR3C1 GCR_si 1, 2, 4 15 4910 10.4 1.44 1087 The a io o en ichmen is he esul o di iding he numbe o TF binding si es by he numbe o expec ed si es. Mo i s a e de i ed om he HOCOMOCO da abase. a NFIA + NFIB + NFIC + NFIX_ 2, b Read coun ange (51, 148, 748, 444, espec i ely). Re . e e ence Ampuja e al. BMC Genomics (2017) 18:68 Page 6 o 15 las c i e ia was used o ensu e me hodological success in subsequen unc ional assays. Wi h he c i e ia de- sc ibed abo e CBFB, HIF1A, and MBD2 we e selec ed o u he s udy. O hese, MBD2 had a la ge numbe o binding si es in he p omo e s o ou DEGs while binding si es o he o he wo TFs we e less widesp ead. Inaddi ion,SMAD4wasusedasaposi i econ ol. As SMAD4 is a known egula o o BMP signaling, we pe o med co-occu ence analysis o he binding si es be ween ou h ee candida e TFs and he SMAD mo i s. We ound ha he MBD2 mo i was signi i- can ly co-localized wi h he GC- ich SMAD4 consensus mo i s CGCC (P= 1.1e-9), GCCGnCGC (P=1.3e-14), and GGCGCC (P= 2e-10). As binding si es o CBFB o HIF1A we e less equen ac oss DEGs, s a is ical signi i- cance o co-localiza ion wi h SMAD mo i s could no be eliably e alua ed. Howe e , we did ind se e al p omo e s whe e SMAD binding si es co-localized wi h hese ac o s. Silencing o selec ed TFs (SMAD4, CBFB, HIF1A, and MBD2) was hen used o u he e alua e hei impac on BMP4 signaling. A e 48 h o silencing, he cells we e ea ed wi h BMP4 o 24 h and he mRNA le els o he alida ed DEGs we e measu ed o assess whe he he silencing in luences BMP4 a ge gene exp ession (Fig. 4b and Addi ional ile 8: Figu e S3). Down egula- ion o SMAD4 was able o e e se he BMP4-media ed change in he exp ession o all he es ed a ge genes in bo h MDA-MB-231 and T-47D cells (Fig. 4c) indica ing ha hese exp ession changes a e indeed ansmi ed ia he canonical BMP pa hway. Fo mos o he a ge genes, MBD2 silencing led o ab oga ion o he BMP4- media ed induc ion in gene exp ession in bo h cell lines. In T-47D cells, simila da a was also ob ained o mos o he genes upon CBFB (9/10) and HIF1A deple ion (6/10). Howe e in MDA-MB-231, silencing o HIF1A esul ed exclusi ely in up egula ion o he a ge genes and bo h enhanced and diminished exp ession was seen a e CBFB down egula ion. O no e, silencing o all o he TFs in T-47D cells led o he enhanced exp ession o he DLL1 gene, which was consis en wi h i being down egula ed upon BMP4 ea men . These da a imply ha he TFs may unc ion as ei he ep esso s o enhance s o BMP4 a ge gene exp ession in a con ex -dependen manne . Discussion We ha e p e iously cha ac e ized ansc ip ional esponses o b eas cance cell lines o BMP4 by using mic oa ay echnology[14].Howe e ,in ha s udywe ocusedonly on cells ha espond o BMP4 by educed p oli e a ion. E o s by o he s o examine BMP signaling a ge genes ha e concen a ed exclusi ely on non-cance ous cells [18–20]. He e we se ou o unco e he ansc ip ional esponses o b eas cance cell lines wi h di e en phe- no ypes by using one cell line ha esponds o BMP4 by educed p oli e a ion (T-47D) and ano he ha e- ac s wi h inc eased mig a ion (MDA-MB-231). Being able o unco e he mechanisms o hese wo di e en esponses is essen ial o he unde s anding o he ole o BMP4 in b eas cance pa hogenesis. To his end, we used a subs an ially new app oach o combining DNase- seq, RNA-seq and unc ional expe imen s. In o de o ind he ea ly media o s o BMP4 esponse, we ea ed he cells wi h BMP4 o ehicle con ol o Table 3 Top 15 ansc ip ion ac o s en iched in T-47D cells TF name Mo i Selec ion by: TF binding si es Re . si es Expec ed si es in e . Ra io o en ichmen Mean ead coun MBD2 MBD2_si 1, 2, 3, 4 101 6664 39.6 2.55 571 TFAP2A AP2A_ 2 1, 2, 3 115 10363 61.6 1.87 941 E4F1 E4F1_ 1 2, 3 18 1750 10.4 1.73 310 SP1 SP1_ 1 1, 2 392 41453 246.3 1.59 838 CUX1 CUX1_ 1 1, 2, 3 13 1462 8.7 1.50 141 E2F2 E2F2_ 1 1, 2 17 1941 11.5 1.47 215 AHR AHR_si 1, 2, 3 9 1030 6.1 1.47 791 SP2 SP2_si 1, 2 140 16512 98.1 1.43 672 CREB1 CREB1_ 1 1, 2, 3 23 2720 16.2 1.42 177 CBFB PEBB_ 1 1, 2, 3, 4 46 5461 32.4 1.42 457 ZIC2 ZIC2_ 1 1, 2, 3 46 5487 32.6 1.41 118 ZFX ZFX_ 1 1, 2, 3 127 15650 93.0 1.37 287 HIF1A HIF1A_si 1, 2, 3, 4 15 1890 11.2 1.34 1847 E2F3 E2F3_si 1, 2, 3 16 2019 12.0 1.33 322 XBP1 XBP1_ 1 1, 3, 4 12 1545 9.2 1.31 22744 The a io o en ichmen is he esul o di iding he numbe o TF binding si es by he numbe o expec ed si es. Mo i s a e de i ed om he HOCOMOCO da abase. Re . e e ence Ampuja e al. BMC Genomics (2017) 18:68 Page 7 o 15 3 h. A his ime poin , he canonical BMP pa hway h ough SMAD1/5/9 is al eady ac i a ed [9]. The esul s o RNA-seq e ealed ha he cell lines esponded o BMP4 by up egula ing o down egula ing a se o genes ha we e mos ly cell line-speci ic, wi h only en com- mon DEGs iden i ied. Consis en wi h he sequencing da a, alida ion wi h qRT-PCR ac oss mul iple ime poin s (3, 6, and 24 h) and i e addi ional cell lines u he con- i med in a wide con ex he exis ence o common BMP4 a ge genes as well as cell line-speci ic exp ession pa - e ns. O he en sha ed DEGs, h ee we e known BMP4 a ge genes (ID1-3) and wo membe s o he BMP A BC Fig. 4 Examples o p edic ed TFBSs and he impac o ansc ip ion ac o s on BMP4 a ge gene exp ession. aThe p edic ed binding si es o ansc ip ion ac o s MBD2, HIF1A and CBFB a e depic ed a he p omo e s o NOG,SMAD7 and ID1 genes, espec i ely. In addi ion, known BMP- esponse elemen s (BRE) loca ed nea he binding si es a e illus a ed. bThe TFs we e silenced and he cells we e ea ed wi h BMP4 o ehicle con ol ollowed by measu emen o a ge gene exp ession by qRT-PCR. Examples o ela i e exp ession le els o SKIL a e HIF1A silencing in MDA-MB-231 cells ( op panel) and NOG exp ession a e MBD2 silencing in T-47D cells (bo om panel) a e shown. cG aphical summa y o he TF silencing expe imen s. The o de o he genes is iden ical o ha in Fig. 2. Blue colo (dec eased a ge gene exp ession) deno es TFs ha we e essen ial o a ge gene exp ession and ed colo hose whose silencing led o enhanced a ge gene exp ession. No applicable indica es cases whe e BMP4 did no al e he baseline gene exp ession. Da a on he DLL1 gene, which is down egula ed in T-47D upon BMP4 ea men , a e highligh ed wi h a bold line Ampuja e al. BMC Genomics (2017) 18:68 Page 8 o 15 signaling pa hway (SMAD6,SMAD9) [3]. The ac i a- ion o he inhibi o y SMAD6 indica es a nega i e eedback loop, which in T-47D is ein o ced by he up egula ion o BMP an agonis NOG and he pseudo - ecep o BAMBI. On he o he hand, ac i a ion o he ecep o - egula ed SMAD9 seems o poin o a posi i e eedback loop, as alongside o he R-SMADs, SMAD9 has been ound o enhance BMP signaling [21, 22]. Howe e , one s udy indica ed ha SMAD9 may ha e an inhibi o y ole in BMP signaling [23]. In any case, up egula ion o SMAD9 due o BMP4 ea men has also been ecen ly epo ed in a ious cell ypes, o example in p ima y ib oblas s, hepa ocellula ca cin- oma and melanoma cells [24]. To unde s and he unc ion o he cell line-speci ic DEGs, we used GO analysis o seg ega e he DEGs in o biological p ocess ca ego ies and disco e ed ha he e- sul s e lec ed he esponse o he cell lines o BMP4. P ocesses ela ed o mig a ion we e en iched in he MDA-MB-231 cells, whe eas mo e di e se esponses we e ound in T-47D, including ca ego ies comp ised o signaling, de elopmen and mo phogenesis. These ind- ings we e co obo a ed by he DNase-seq da a, whe e we ound ha BMP4-induced global open ch oma in si es we e en iched wi h he same biological ca ego ies ha we e ound wi h RNA-seq da a. While ca ego ies associa ed wi h signaling we e obse ed in bo h cell lines, in MDA-MB-231 hose ela ed o mig a ion we e en iched. These da a ex end ou p e ious esul s show- ing en ichmen o di e en ially exp essed genes in GO ca ego ies ha we e associa ed wi h he BMP4-induced dec ease in p oli e a ion [14]. Taken oge he , he di e - en esponses o he cell lines o BMP4 a e e lec ed bo h a he ansc ip ional and ch oma in le els. In he analysis o TSS ch oma in s a e we could obse e changes in only a ew o he genes ha we e di - e en ially egula ed by BMP4. This migh be due o he ac ha he 3 h s imula ion o BMP4 is oo sho o mos o he TSSs o change hei ch oma in s a us. Mo eo e , we could obse e ha in many cases he ch oma in was al eady open a he TSS, in which case u he changes a e no needed o enhance he ansc ip- ional ac i i y. Toge he wi h he obse a ion ha he e is a la ge a ia ion be ween he ch oma in s a us and gene exp ession when we ex end he analysis o he whole se o p o ein-coding genes, i can be concluded ha he ch oma in s a e o TSSs explains he obse ed exp ession pa e ns only o a small ex en . This esul was no unexpec ed, as gene exp ession is also com- monly egula ed om egions loca ed a om he TSS, such as enhance s [25, 26]. Wi h genome-wide de ec ion o open ch oma in a eas we no iced ha BMP4 s imula ion induces opening o he ch oma in mos ly in he in onic and in e genic egions. This is consis en wi h he ac ha changes in he TSS and p omo e egions we e obse ed wi h only a ew o he di e en ially exp essed genes. Opening o he in- onic sequences may indica e inc eased le el o RNA polyme ase ac i i y a gene bodies. Ch oma in opening a in e genic egions migh sugges ha addi ional egula o y con ol is being a ained in la ge ex en h ough dis al egula o y elemen s such as enhance s and silence s. Thus, al eady a he ea ly 3-h ime poin we a e able o obse e con o ma ional changes ha cells may u ilize in mo e de ailed egula ion o he BMP esponse. Un o una ely, based on his analysis we we e no able o de ine a speci ic ansc ip ion ac o ch oma in signa u e ha could be used o de ine BMP- speci ic egula o y sequences. Hence de ailed analysis o he pu a i e enhance egions would equi e mo e speci ic measu emen da a abou he ch oma in in e - ac ions in hese cells. To u he cha ac e ize he egula ion o BMP4 a ge genes, we analyzed ansc ip ion ac o binding si es (TFBSs) in he open ch oma in egions loca ed on gene p omo e s. Among he op 15 en iched TFs, he e we e a ew which had p e iously been linked o BMP a ge gene egula ion. Fo example, XBP1 and RELA ha e been shown o be ep esso s o BMP a ge genes X en -2 and Id1, espec i ely [27, 28]. Using en ichmen o he TFBSs be ween cell lines as well as o he c i e ia, we selec ed h ee TFs (CBFB, HIF1A, and MBD2) o unc ional cha ac e iza ion and silenced hem in he wo cell lines. In addi ion, we used SMAD4, a key componen o he ca- nonical BMP pa hway, as ou posi i e con ol and indeed SMAD4 was equi ed o ansc ip ional egula ion o all he BMP4 a ge genes in he assay. Al hough BMPs can signal h ough al e na i e pa hways [7, 8], his esul poin s o egula ion h ough he canonical pa hway. In con as , he esponse o o he ansc ip ion ac o s was mo e a iable and cell line-speci ic. MBD2 is a me hyl-CpG-binding ansc ip ion ac o ha plays a ole in de elopmen [29, 30]. Se e al s udies ha e shown ha MBD2 ac s as a ansc ip ional ep es- so by ec ui ing co- ep esso complexes o p omo e s, which in u n leads o o ma ion o ep essi e ch oma in h ough ch oma in emodelling [31, 32]. Howe e , he e is also e idence ha MBD2 can ac i a e ansc ip ion by emo ing me hyla ion om CpG islands loca ed in p o- mo e s [33]. In bo h cell lines, MBD2 seemed o ac mainly as an ac i a o o ansc ip ion, al hough i s ole was mo e p ominen in MDA-MB-231. In ou analysis, MBD2 had a la ge numbe o binding si es ac oss DEGs and i was highly exp essed in bo h cell lines, consis en wi h he obse ed beha io in he silencing expe imen . The key ole o MBD2 in con olling he BMP4 esponse sugges s ha DNA me hyla ion may be in ol ed in BMP4 signaling. Ampuja e al. BMC Genomics (2017) 18:68 Page 9 o 15