Hypomethylation of HOXA4 promoter is common in Silver-Russell syndrome and growth restriction and associates with stature in healthy children

1 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 www.na u e.com/scien i ic epo s Hypome hyla ion o HOXA4 p omo e is common in Sil e - Russell synd ome and g ow h es ic ion and associa es wi h s a u e in heal hy child en Ma i Muu inen1, Ka a iina Hannula-Jouppi1,2, Lo isa E. Reinius3, Cilla Söde häll3,4, Simon Kebede Me id5, Anna Be gs öm5,7, E ik Melén5,6,7, Gö an Pe shagen5,7, Ma i a Lipsanen- Nyman8, Da io G eco9,10 & Juha Ke e 1,3,11 Sil e -Russell synd ome (SRS) is a g ow h e a da ion synd ome in which loss o me hyla ion on ch omosome 11p15 (11p15 LOM) and ma e nal unipa en al disomy o ch omosome 7 [UPD(7) ma ] explain 20–60% and 10% o he synd ome, espec i ely. To sea ch o a molecula cause o he emaining SRS cases, and o ind a possible common epigene ic change, we s udied DNA me hyla ion pa e n o mo e han 450 000 CpG si es in 44 SRS pa ien s. Common o all h ee SRS subg oups, we ound a hypome hyla ed egion a he p omo e egion o HOXA4 in 55% o he pa ien s. We hen es ed 39 pa ien s wi h se e e g ow h es ic ion o unknown e iology, and ound hypome hyla ion o HOXA4 in 44% o he pa ien s. Finally, we ound ha me hyla ion a mul iple CpG si es in he HOXA4 p omo e egion was associa ed wi h heigh in a coho o 227 heal hy child en, sugges ing ha HOXA4 may play a ole in egula ing human g ow h by epigene ic mechanisms. Sil e -Russell synd ome is an imp in ing diso de cha ac e ized by se e e p e-and pos na al g ow h e a da ion and ypical clinical ea u es, including ela i e mac ocephaly, p o uding o ehead, body asymme y and eeding di icul ies1. Imp in ing diso de s a ec he unc ion o genes ha a e no mally exp essed p edominan ly o solely om one pa en al ch omosome. Ch omosomal abno mali ies o changes in DNA me hyla ion le el can al e imp in ed gene dosage and lead o pheno ypes a ec ing g ow h o neu ological unc ioning2. Opposi e changes in he same imp in ed genomic egions can lead o opposi e g ow h pheno ypes, such as g ow h es ic ion in SRS and o e g ow h in Beckwi h-Wiedemann synd ome (BWS), caused by hypo- and hype me hyla ion, espec- i ely, o he imp in ed H19-IGF2 egion in 11p15 (H19/IGF2:IG-DMR) in a p opo ion o he cases3. SRS is a he e ogeneous synd ome bo h in pheno ype and in molecula e iology, bu se e e g ow h e a da ion is a cen al ea u e sha ed by he pa ien s. SRS pa ien s can be di ided in o h ee subg oups based on hei molecula e iology: ma e nal unipa en al disomy o ch omosome 7 (UPD(7)ma ), loss o me hyla ion on ch omosome 11p15, (11p15 LOM; also known as hypome hyla ion o H19/IGF2:IG-DMR o H19 hypome hyla ion), and clinical SRS wi hou any o hese 1Folkhälsan Ins i u e o Gene ics, Helsinki, and Resea ch P og ams Uni , Molecula Neu ology, Uni e si y o Helsinki, Helsinki, Finland. 2Depa men o De ma ology, Skin and Alle gy Hospi al, Uni e si y o Helsinki and Helsinki Uni e si y Hospi al, Helsinki, Finland. 3Depa men o Biosciences and Nu i ion, Ka olinska Ins i u e , Huddinge, Sweden. 4Depa men o Women’s and Child en’s Heal h, Ka olinska Ins i u e , S ockholm, Sweden. 5Ins i u e o En i onmen al Medicine, Ka olinska Ins i u e , S ockholm, Sweden. 6Sachs’ Child en’s Hospi al, Söde sjukhuse , S ockholm, Sweden. 7Cen e o Occupa ional and En i onmen al Medicine, S ockholm Coun y Council, S ockholm, Sweden. 8Child en’s Hospi al, Uni e si y o Helsinki and Helsinki Uni e si y Hospi al, Helsinki, Finland. 9Ins i u e o Bio echnology, Uni e si y o Helsinki, Helsinki, Finland. 10Facul y o Medicine and Li e Sciences & Ins i u e o Biosciences and Medical Technology, Uni e si y o Tampe e, Tampe e, Finland. 11School o Basic and Medical Biosciences, King’s College London, Guy’s Hospi al, London, UK. Co espondence and eques s o ma e ials should be add essed o J.K. (email: juha.k[email p o ec ed]) Recei ed: 8 Feb ua y 2017 Accep ed: 7 No embe 2017 Published: xx xx xxxx OPEN www.na u e.com/scien i ic epo s/ 2 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 molecula indings. App oxima ely 10% o SRS pa ien s ha e UPD(7)ma and 20–60% o SRS pa ien s ha e 11p15 LOM4–10. Duplica ions and dele ions a 11p15 a ec <1% o SRS pa ien s and e en o he molecula changes, such as UPD o ch omosome 11 ha e been de ec ed in single cases3. The emaining cases can be designa ed clinical SRS, which make up 30–70% o he SRS pa ien s. As all SRS pa ien s sha e he same clinical diagnosis, while a subs an ial amoun o pa ien s es nega i e o molecula diagnosis, we hypo hesized ha he e migh occu u he epigene ic changes ha all subg oups o SRS sha e in common. To sea ch o such genes o egions, we analyzed 44 SRS pa ien s o me hyla ion changes by he genome-wide Illumina In inium HumanMe hyla ion450K BeadChip assay ha p o ides quan i a i e me hyl- a ion da a a mo e han 450 000 single CpG si es. In addi ion, we e alua ed associa ions be ween he ound me h- yla ion changes and se e e g ow h es ic ion o unknown e iology (SGR), as well as g ow h in heal hy child en. Resul s Genome-wide me hyla ion analysis pinpoin ed HOXA4 hypome hyla ion as a common epi- gene ic change among SRS subg oups. We used he In inium HumanMe hyla ion450K BeadChip (Illumina) o iden i y sha ed me hyla ion changes in he h ee di e en SRS subg oups. We used a il e ing app oach (Fig.1) o sea ch o a eas ha we e signi ican ly di e en ially me hyla ed in all subg oups compa ed o con ols o e a s e ch o consecu i e CpGs. In he i s s age o il e ing ( il e 1), al oge he 13,226 CpGs we e ound, in which he di e en ial me hyla ion be ween each o he h ee SRS g oups compa ed o con ols eached he empi ical Bayes signi icance (nominal p- alue < 0.05). We hen chose egions ha had a leas h ee consecu i e signi ican CpGs o all subg oups and disca ded CpGs in which he me hyla ion le el o di e en g oups a ied o opposi e di ec ions in compa ison o he con ols ( il e 2). This il e ing p ocess subs an ially na owed down he po en ial a eas and esul ed in 92 CpGs loca ed in 26 genes o in e genic loci on 13 di e en ch omosomes (Supplemen a y Table1). O hese CpGs, 66 showed only small di e ences in median Be a- alue (<0.02) be ween each o he subg oups and con ols. Al oge he 16 CpGs showed median Be a- alue di e ences >0.05 (5% absolu e me hyla ion le el) o all h ee subg oups. Two o he 16 CpGs we e loca ed in an in e genic egion on 5p15.33 and 14 we e loca ed in he HOXA4 gene egion on 7p15.2. HOXA4 was he only a ea ha showed a leas 5% median me hyla ion di e ence o all subg oups s. con ols o a leas h ee consecu i e signi ican CpGs. In o al, HOXA4 showed 12 consecu i e p obes wi h such di e ences. The di e ences o he indi idual g oups in hese 12 CpGs we e also la ge han wi h any o he CpGs, anging om 6.8–25% in UPD(7) ma , 7.1–21% in clinical SRS and 5.0–11% in 11p15 LOM pa ien s. As a esul o he il e ing p ocess o genome- wide me hyla ion da a, HOXA4 eme ged as a speci ic candida e o u he s udy. HOXA4 me hyla ion le els di e among subg oups and a e lowes nea he HOXA4 ansc ip- ion s a si e. All h ee SRS subg oups showed hypome hyla ion h oughou he HOXA4 egion iden i ied by il e ing when compa ing median me hyla ion le els ela i e o con ols (Fig.2). This di e en ially me hyla ed a ea included wo s e ches o consecu i e CpGs: om cg11532431 o cg14359292 (ch 7:27169674–27170892 in GRCh37/hg19 assembly), and om cg24169822 o cg11908057 (ch 7:27170994–27171154). Localized be ween hese s e ches, cg25952581 (a ch 7:27170961) did no each he empi ical Bayes signi icance o di e en ial me hyla ion be ween con ols and 11p15 LOM g oup (p = 0.082), bu clinical SRS and UPD(7)ma had signi i- can me hyla ion di e ences (p = 0.013 and p < 0.001 espec i ely), and he me hyla ion pa e n was simila o he pa e n o o he CpGs in he egion. P- alues o di e en ial me hyla ion be ween subg oups and con ols Figu e 1. Fil e ing o common di e en ially me hyla ed egions in SRS pa ien s. www.na u e.com/scien i ic epo s/ 3 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 o all HOXA4 CpGs included in ou il e ing a e shown in Table1. Fo he majo i y o he CpGs, he median me hyla ion le el was lowes in he UPD(7)ma g oup, ollowed by clinical SRS, 11p15 LOM, and con ols in ascending o de . The me hyla ion le el o he single UPD(7)pa sample was highe han con ols o mos o he CpGs. Figu e3 shows examples o me hyla ion le els a ou indi idual CpG si es, whe e me hyla ion le el o Figu e 2. HOXA4. Median me hyla ion le els o SRS subg oups ela i e o con ols (1). Bounda ies o he egion in which HOXA4 CpGs a e hypome hyla ed in SRS g oups ela i e o con ols a e ma ked wi h wo dashed e ical lines. Coo dina es in hg19 (ch 7) CpG ID Regula o y CGI 11p15 LOM clinical SRS UPD(7) ma 27168609 cg02022102 N_Sho e 0,1337 0,3379 0,0345 27168688 cg03982801 N_Sho e 0,4355 0,2119 0,8381 27168780 cg02706018 N_Sho e 0,9538 0,3056 0,5738 27168962 cg09574499 N_Sho e 0,8489 0,2095 0,4154 27169208 cg14042889 N_Sho e 0,6536 0,8850 0,6749 27169674 cg11532431 CELL TYPE Island 0,0011 0,0013 8,2E-06 27169740 cg00562553 Island 0,0052 0,0037 0,0002 27170241 cg22997113 PROMOTER Island 0,0006 0,0001 1,5E-06 27170313 cg04317399 PROMOTER Island 0,0309 0,0030 4,1E-05 27170388 cg07317062 PROMOTER Island 0,0073 0,0014 8,4E-06 27170394 cg19142026 PROMOTER Island 0,0061 0,0005 7,3E-06 27170412 cg11410718 PROMOTER Island 0,0037 0,0004 3,3E-06 27170717 cg17457637 PROMOTER S_Sho e 0,0349 0,0011 1,4E-05 27170819 cg06942814 PROMOTER S_Sho e 0,0116 0,0004 1,9E-05 27170832 cg08657492 PROMOTER S_Sho e 0,0063 0,0001 4,4E-06 27170880 cg04321618 PROMOTER S_Sho e 0,0172 0,0009 9,7E-06 27170892 cg14359292 PROMOTER S_Sho e 0,0125 0,0001 9,8E-07 27170961 cg25952581 PROMOTER S_Sho e 0,0818 0,0125 0,0002 27170994 cg24169822 PROMOTER S_Sho e 0,0120 0,0005 5,4E-06 27171051 cg17591595 PROMOTER S_Sho e 0,0371 0,0006 0,0033 27171154 cg11908057 S_Sho e 0,0108 0,0001 0,0031 27171203 cg25967031 S_Sho e 0,0804 0,0035 0,0167 27171213 cg15196806 S_Sho e 0,0526 0,0173 0,0674 27171391 cg15624376 S_Sho e 0,8751 0,8147 0,2402 27171401 cg20161965 S_Sho e 0,0634 0,2399 0,0240 Table 1. P- alues o di e en ial me hyla ion o SRS subg oups (11p15 LOM, clinical SRS, UPD(7)ma ) compa ed o con ols o all HOXA4 CpGs ha passed quali y con ol and we e included in he il e ing p ocess o Illumina 450 K BeadChip assay da a. Regula o y ea u e g oup as indica ed by Illumina is shown in column “Regula o y”: “PROMOTER” = P omo e _Associa ed, “CELL TYPE” = Unclassi ied_Cell_ ype_speci ic. Rela ion o UCSC CpG island ch 7:27169572-27170638 is shown in column “CGI”. www.na u e.com/scien i ic epo s/ 4 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 mos SRS pa ien s ell below -2 SD and UPD(7)pa emained wi hin he ange o he con ol g oup’s me hyla ion le el. Some SRS pa ien s showed ex emely low hypome hyla ion pa e ns and we e hypome hyla ed e en as low as 8 SD below he mean o he con ol g oup a cg22997113. The iden i ied di e en ially me hyla ed a ea o HOXA4 was loca ed in he p omo e egion and co-localized wi h he CpG island a ch 7:27169573–27170638. App oxima ely hal o he di e en ially me hyla ed CpGs we e loca ed wi hin he CpG island and he es wi hin he p oximal CpG island sho e. Acco ding o Swi chGea Genomics ansc ip ion s a si es, he HOXA4 TSS is a ch 7:27170399 and acco ding o FANTOM5, a ch 7:27170364–27170377 in he hg19 build (GRCh37/hg19 assembly). Cg19142026 a posi ion ch 7:27170394, he CpG wi h lowes ela i e median me hyla ion o he HOXA4 di e en ially me hyla ed a ea o all SRS sub- g oups, was wi hin 5-17 bp, espec i ely, o he TSS. HOXA4 hypome hyla ion equency was highes o UPD(7)ma , ollowed by clinical SRS and 11p15 LOM pa ien s. While g oup esul s showed he end ha UPD(7)ma pa ien s had he lowes median le el o me hyla ion o CpGs in he HOXA4 di e en ially me hyla ed a ea, ollowed by he clinical SRS and 11p15 LOM g oups, he me hyla ion le el a ied be ween indi iduals wi hin each subg oup. Figu e4(a) shows hypome hyla ion o indi idual SRS pa ien s o CpGs in he HOXA4 di e en ially me hyla ed a ea. The e- quency o HOXA4 hypome hyla ion was highes o UPD(7)ma pa ien s a 80%, ollowed by clinical SRS a 62% and 11p15 LOM g oup a 38%, when hypome hyla ion o an indi idual pa ien was de ined by a leas wo- hi ds o he CpGs being hypome hyla ed (<−2 SD). In addi ion o he pa ien s ha we e hypome hyla ed in majo i y o he CpGs in he HOXA4 a ea, h ee 11p15 LOM pa ien s and wo clinical SRS pa ien s had hypome hyla ion in a leas one- hi d o he HOXA4 CpGs, demons a ing a pa ial de ia ion om no mal me hyla ion s a us in he a ea. Sha ed molecula e iology o SRS pa ien s. Analysis o he me hyla ion da a o HOXA4 egion s ud- ied by he Illumina assay e ealed ha al oge he 24 (55%) o he SRS pa ien s we e hypome hyla ed in a leas wo- hi ds o he CpG si es a HOXA4 di e en ially me hyla ed a ea. O en bo h UPD(7)ma and HOXA4 hypo- me hyla ion, o HOXA4 hypome hyla ion and 11p15 LOM occu ed in he same indi idual, as seen in Fig.4(b). Ei he 11p15 LOM, UPD(7)ma , HOXA4 hypome hyla ion o a combina ion o wo changes was ound in 39 ou o 44 SRS pa ien s (89%). Only 5 pa ien s (11%) we e le wi hou a molecula change de ined by he chosen c i- e ia. In wo o hese pa ien s, we obse ed pa ial hypome hyla ion in he HOXA4 di e en ially me hyla ed a ea. F equency o 11p15 LOM was highe han p e iously es ed. A e ini ial analysis o ou 44 SRS pa ien s, we obse ed ha mo e indi iduals had 11p15 LOM han had p e iously been iden i ied. 15 o ou 21 Figu e 3. Me hyla ion le el a HOXA4 cg22997113 (a), cg11410718 (b), cg08657492 (c) and cg04321618 (d). Indi iduals wi hin each g oup a e ep esen ed in ascending o de o me hyla ion. Dashed line ep esen s -2 s anda d de ia ions o he a i hme ic mean o con ol g oup me hyla ion le el. www.na u e.com/scien i ic epo s/ 5 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 11p15 LOM SRS pa ien s had been p e iously ound hypome hyla ed wi h a es ic ion si e-speci ic me hyla ion me hod11. O he six pa ien s ha had p e iously no been ound hypome hyla ed, wo had no been es ed and ou had been es ed bu ound nega i e. Supplemen a y Table2 summa izes ou esul , indica ing ha 21 o ou pa ien s had clea me hyla ion changes in his a ea wi h a leas wo- hi ds o he 36 CpGs hypome hyla ed (below -2 SD o he con ols) in he egion. EpiTYPER me hyla ion assay e ealed HOXA4 hypome hyla ion in pa ien s wi h se e e g ow h es ic ion o unknown e iology (SGR). We used he Sequenom EpiTYPER me hod o assay he me hyla ion le el o he HOXA4 p omo e egion in o de o alida e esul s om he Illumina genome-wide me hyla ion s udy and o s udy a g oup o child en wi h SGR. The pa icipan s included 44 SRS pa ien s, 16 con ols and 39 child en wi h SGR. The EpiTYPER assay spanned ch 7:27170191–27170313, o e lapping he a ea wi h signi ican me hyla ion di e ences in he genome-wide me hyla ion s udy. One CpG, cg04317399 a ch 7:27170313, was co e ed by bo h Illumina 450 K and EpiTYPER, he e named HOXA_F6_CpG_16. Analysis showed a s ong co ela ion be ween he wo me hods, R2 = 0.86. (Supplemen a y Figu e1). Me hyla ion le els o ou di e en CpG si es (HOXA4F6CpG1, HOXA4F6CpG12.13.14, HOXA4F6CpG15, HOXA4F6CpG16) we e measu ed o all pa icipan s. S a is ically signi ican di e ences be ween he g oups we e de e mined by one-way ANOVA (HOXA4F6CpG1: F(4,92) = 10.831, p < 0.001; HOXA4F6CpG12.13.14: F(4, 91) = 11.549, p < 0.001; HOXA4F6CpG15: F(4,92) = 9.268, p < 0.001; HOXA4F6CpG16: F(4, 92) = 7.556, p < 0.001). Pos hoc compa isons using 2-sided Dunne - es s indica ed ha he mean me hyla ion le els o 11p15 LOM, UPD(7) ma , clinical SRS and SGR g oups we e s a is ically signi ican ly di e en om he mean o he con ols a all ou s udied si es (Fig.5). The la ges mean di e ence (I-J = −0.249, p = 3.389 × 10−7) was obse ed be ween clinical SRS (I) and con ols (J) a HOXA4F6CpG12.13.14. Also, an analysis o indi idual SGR pa ien s showed ha 44% (17/39) o he pa ien s we e hypome hyla ed o a leas wo- hi ds o he CpGs in he HOXA4 p omo e egion a ch 7: 27170191–27170313. Supplemen a y Table3 shows indi idual me hyla ion s a us o all SRS and SGR pa ien s s udied wi h he EpiTYPER assay. Thus, HOXA4 hypome hyla ion was no limi ed o SRS pa ien s, bu also implica ed in pa ien s wi h SGR. Me hyla ion le el o HOXA4 cg11908057 co ela es wi h heigh in heal hy child en. We s ud- ied co ela ion o me hyla ion le els o HOXA4 and heigh o al oge he 227 heal hy child en om he BAMSE (Swedish abb e ia ion o Child en, Alle gy, Milieu, S ockhom, Epidemiology) coho , a bi h coho o chil- d en bo n be ween 1994 and 1996 in S ockholm, Sweden. Me hyla ion le els we e measu ed wi h In inium HumanMe hyla ion450K BeadChip (Illumina)12. Six een CpGs o he HOXA4 di e en ially me hyla ed a ea we e s udied. We obse ed a s a is ically signi ican co ela ion be ween he me hyla ion le el o one o he HOXA4 di e en ially me hyla ed CpGs, cg11908057, and heigh a ou yea s o age ( = 0.142, n = 227, p = 0.033), eigh yea s o age ( = 0.148, n = 225 p = 0.026) and six een yea s o age ( = 0.164, n = 200, p = 0.020) by compu - ing Pea son co ela ion coe icien . As me hyla ion da a o some o he CpGs did no mee c i e ia o no mal Figu e 4. (a) HOXA4 CpG me hyla ion s a us ep esen ed by indi iduals in each SRS g oup. Each column ep esen s one pa ien . G een colo indica es me hyla ion le el below -2SD o he con ols. (b) Molecula changes o SRS pa ien s, n = 44. HOXA4 hypome hyla ion is p esen in 55% o he pa ien s. www.na u e.com/scien i ic epo s/ 6 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 dis ibu ion, addi ional analysis using all CpGs by compu ing Spea man co ela ions e ealed s a is ically sig- ni ican posi i e co ela ion be ween me hyla ion le el and heigh a six een yea s o age o 5 addi ional CpGs: cg04317399 ( = 0.146, n = 200, p = 0.040), cg19142026 ( = 0.139, n = 200, p = 0.050), cg04321618 ( = 0.149, n = 200, p = 0.036), cg14359292 ( = 0.140, n = 200, p = 0.047) and cg25952581 ( = 0.145, n = 200, p = 0.040). Cg11908057 demons a ed simila esul s as abo e: posi i e co ela ion was obse ed be ween me hyla ion le el and heigh a eigh yea s o age ( = 0.140, n = 225, p = 0.036) and six een yea s o age ( = 0.147, n = 200, p = 0.038). Figu e6 demons a es ha he me hyla ion le el o cg11908057 in heal hy child en a eigh yea s o age anges om 0.69 o 0.91. This a ia ion in me hyla ion co esponds o a di e ence o 0.75 SD (o 4.5 cm) Figu e 5. Me hyla ion le els measu ed a ou EpiTYPER si es o 11p15 LOM (n = 21), UPD(7)ma (n = 10), clinical SRS (n = 13), and SGR (n = 38, o HOXA4_F6_CpG_12.13.14 n = 37) pa ien s compa ed o con ols (n = 15). Compa isons made by 2-sided Dunne - es s. Figu e 6. Heigh measu emen s (in SD scale) o BAMSE indi iduals (n = 225) a age 8 plo ed agains mea hyla ion le el o cg11908057 measu ed by Illumina 450 K BeadChip assay. www.na u e.com/scien i ic epo s/ 7 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 in heigh . In he same indi iduals, we s udied co ela ion o heigh a age eigh wi h some o he mos s ongly heigh -associa ed SNPs13,14 (Supplemen a y Table4). In he es ed SNPs, he e was no signi ican associa ion be ween heigh and geno ype (Supplemen a y Figu e2). The esul s indica e ha in ou da ase o heal hy-school aged child en om he BAMSE coho , he me hyla ion le el o cg11908057 a HOXA4 was be e in explaining heigh a ia ion a age eigh han some o he mos s ongly heigh -associa ed SNPs. We also used egulomeDB15 o s udy whe he cg11908057 is localized in a po en ial egula o y p o ein-binding si e. The e was some e idence o binding o an enhance o zes e homolog 2 (EZH2), a me hyl ans e ase enzyme encoded by he EZH2 gene on 7q36.1. Mu a ions o EZH2 cause Wea e synd ome16, a p e- and pos na al o e g ow h synd ome esul ing in all adul s a u e. Me hyla ion le el o HOXA4 p omo e egion co ela es wi h exp ession le el o HOXA4 in blood. We used da a p e iously published in Hannula-Jouppi e al.17, o s udy he co ela ion o HOXA4 p o- mo e egion CpG me hyla ion le el and exp ession le el o HOXA4 in blood in nine UPD(7)ma pa ien s, one indi idual wi h UPD(7)pa and en con ols. The a e age me hyla ion le el o he HOXA4 egion in SD scale co ela ed nega i ely wi h he exp ession le el o HOXA4 in blood, as demons a ed in Fig.7. In all di e en- ially me hyla ed CpGs o HOXA4, we obse ed a s a is ically signi ican nega i e co ela ion be ween exp ession and me hyla ion le el by Spea man co ela ion analysis. The nega i e co ela ion was s onges o cg22997113 ( = −0.81, n = 20, p = 1.40 × 10−5). Da a o o he CpGs is lis ed in Supplemen a y Table5. Discussion We conduc ed a genome-wide me hyla ion s udy, de e mining he me hyla ion s a us o mo e han 450,000 CpG si es o UPD(7)ma , 11p15 LOM and clinical SRS pa ien s, and aimed he da a analysis a inding common me hyla ion changes among hese h ee g oups. Ini ial analysis showed such changes in 26 genes o in e genic loci, and u he analysis pinpoin ed HOXA4 p omo e hypome hyla ion as a majo de ec in SRS. HOXA4 hypo- me hyla ion occu ed in 55% o SRS pa ien s, and was hus mo e common in ou coho han 11p15 LOM (48%) o UPD(7)ma (23%), and speci ically, was ound in eigh 11p15 LOM (38%), eigh UPD(7)ma (80%), and eigh pa ien s wi hou any p io molecula diagnosis (62%). Fu he mo e, a ge ed me hyla ion analysis e ealed ha HOXA4 hypome hyla ion was also p esen in child en wi h se e e g ow h- es ic ion o unknown e iology. Fu he analysis o he HOXA4 egion iden i ied mul iple CpG si es whe e lowe me hyla ion le el was associa ed wi h lowe heigh in heal hy school-aged child en. HOXA4 belongs o he clus e o HOXA genes loca ed on 7p15.2. The HOXA4 hypome hyla ed egion was loca ed a ch 7:27169674–27171154, wi h mos CpGs classi ied as p omo e -associa ed, and hal o he CpGs loca ed wi hin he o e lapping CpG island and he es in he p oximal island sho e. CpG islands in soma ic cells a e usually unme hyla ed, and me hyla ed p omo e CpG islands a e in ol ed in long- e m ep ession o , o example, imp in ed genes18. Me hyla ed CGIs a e o en obse ed nea de elopmen ally impo an genes, such as he HOX genes, and i is hough ha he la ge numbe s o po en ial CGI p omo e s wi hin HOX loci may con ibu e o hei egula ion19. Rela i e HOXA4 me hyla ion le el was lowes o all he SRS subg oups a cg19142026, in close p oximi y o he unc ionally ele an ansc ip ion s a si e. Lowe me hyla ion le el is o en ela ed o ansc ip ional ac i a ion o genes, while hype me hyla ion can lead o ansc ip ional ep es- sion16,20. We obse ed ha hypome hyla ion o he HOXA4 p omo e egion was co ela ed wi h inc eased exp ession o HOXA4 in blood, when using exp ession da a om UPD(7)ma , UPD(7)pa and con ols om ou p e iously published s udy17. O he epo s ha e shown ha hype me hyla ion o HOXA4 p omo e egion p omo es inac i a ion o gene exp ession21,22. Al e ed me hyla ion le els o he HOXA4 p omo e he e o e seem o ha e unc ional ele ance by a ec ing gene exp ession. Figu e 7. HOXA4 exp ession le el in blood measu ed by qPCR s. a e age me hyla ion le el o HOXA4 CpGs in SD scale. A e age me hyla ion was calcula ed om all HOXA4 CpGs ha showed signi ican di e en ial me hyla ion be ween all SRS g oups and con ols wi h Illumina 450 K analysis. www.na u e.com/scien i ic epo s/ 8 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 Se e al genes wi hin he HOXA gene clus e (HOXA2, HOXA3, HOXA4, HOXA5, HOXA6 and HOXA11) a e p edic ed as ma e nally exp essed imp in ed genes23. In ou p e ious s udy, HOXA4 was monoallelically exp essed and he exp ession o HOXA4 was inc eased in whole blood o he UPD(7)ma g oup ela i e o con- ols17, suppo ing he p edic ion o HOXA4 as a ma e nally exp essed imp in ed gene. In UPD(7)ma pa ien s, HOXA4 hypome hyla ion can he e o e be explained by he occu ence o wo ma e nal copies o ch omosome 7, al hough he hypome hyla ion was no obse ed in all o he UPD(7)ma pa ien s in ou s udy. In 11p15 LOM pa ien s, HOXA4 hypome hyla ion could po en ially be due o mul i-locus imp in ing dis u bance (MLID). MLID is epo ed o be p esen in 15–38% o 11p15 LOM pa ien s1. The HOX/Hox genes in human and o he species a e highly in e es ing in he con ex o g ow h diso de s. They encode a amily o ansc ip ion ac o s ha egula e ea ly de elopmen al mo phogene ic p ocesses and a e impo an egula o s o an e io -pos e io axis de elopmen 24. Ten o he 39 HOX genes ha e been associa ed wi h human diso de s25, such as HOXA13 wi h Hand- oo -geni al synd ome (HFGS). Human pheno ypes o he loss-o - unc ion mu a ions in genes o he HOXA clus e include acial dysmo phisms, limb anomalies, ca diac de ec s and u ogeni al mal o ma ions. As he Hox genes a e highly conse ed, human pheno ypes a e e y simila o hei mouse o hologs. Hox gene knockou s in mouse a ec he skele on, p ima y e eb al column and sligh ly less equen ly he bones o he o elimb and hindlimb. Mouse Hoxa4 loss-o - unc ion homozygo e pheno ype showed skele al abno mali ies including pa ial an e io ans o ma ion o C3, pos e io ans o ma ion o C7 o T1, p esence o a C7 ce ical ib and mal o ma ion o he s e num26,27. The epo ed Hox gene de ec s ha o en a ec bone mo phology sugges ha dis u bance in he no mal pa e n o HOXA4 me hyla ion may also ha e ele ance conside ing skele al mal o ma ions in SRS pa ien s. In e es ingly, in 3-M synd ome, a g ow h diso de wi h a pheno ype e y simila o SRS, down egula ion o IGF2 and up egula ion o se e al HOX genes has been epo ed28. Ou s udy also included wo a e samples, a UPD(7)pa and a UPD(7q31-q e )ma , in o de o enable u he in es iga ion o he po en ial indings29. The me hyla ion le el o he UPD(7)pa sample was highe han he median me hyla ion le el o con ols in mos di e en ially me hyla ed CpGs o HOXA4, bu s ill wi hin no mal ange (+/−2SD). I is in e es ing o no e ha he e is no g ow h abno mali y in his indi idual o o he epo ed UPD(7)pa cases29–31. The segmen al UPD(7q31-q e )ma was hypome hyla ed a 10 o he HOXA4 CpGs, e en hough HOXA4 (a 7p15.2) is loca ed ou side o he segmen al UPD(7q31-q e )ma egion. Fu he s udies a e equi ed o be e unde s and his obse a ion. Unexpec edly, we ound ha some o he pa ien s ha had been es ed nega i e o 11p15 LOM p e iously, we e ac ually posi i e acco ding o ou esul s om he genome-wide me hyla ion s udy. P icke e al. epo ed simila indings, whe e me hyla ion-sensi i e RFLP PCR me hod ailed o de ec 11p15 LOM pa ien s while 11p15 LOM was ound by he In inium Human Me hyla ion 450 K BeadChip a ay32. These indings sugges ha he genome-wide me hyla ion assay may be mo e sensi i e a de ec ing hypome hyla ed pa ien s han some o he me hods. Me hyla ion le els may a y in di e en issues and ime in de elopmen . Ou esul s a e limi ed o me hyla- ion le els p esen a he ime o sampling in accessible whole blood issue. Va ia ion in he me hyla ion le el o he di e en whi e blood cell popula ions in whole blood can complica e in e p e a ion o whole blood me hyl- a ion p o iles33. In he HOXA4 egion, such a ia ion be ween blood cell ypes was no obse ed, and he e o e po en ial di e ences in blood cell popula ions o he pa icipan s should no a ec ou esul s. Also genome-wide me hyla ion s udies o human blood ha e epo ed age- ela ed modi ica ions in speci ic CpGs du ing ea ly child- hood34, adul hood35 and om newbo n o elde ly in a me a-analysis36. Ou pa ien s included bo h child en and adul s, while con ol samples we e all adul s. CpGs in he HOXA4 egion we e, howe e , no among he epo ed age-modi ied loci, and he e o e age o he pa icipan s is no expec ed o signi ican ly a ec ou esul s. Ou genome-wide s udy o SRS pa ien s ound hypome hyla ion a mul iple adjacen CpGs in he p omo e egion o he de elopmen ally impo an HOXA4 gene. Ou esul s indica ed ha HOXA4 hypome hyla ion was no speci ic o SRS, bu was also p esen in child en wi h se e e g ow h es ic ion o unknown e iology. Ou s udy in heal hy child en indica ed an e ec in he same di ec ion, showing he associa ion o hypome hyla- ion o mul iple HOXA4 CpGs wi h sho s a u e. I is also wo h men ioning ha in HOXA3, a gene adjacen o HOXA4, an associa ion be ween bi h weigh and me hyla ion has been obse ed in wo CpGs37, demon- s a ing ha he egion may be impo an in egula ing body size. We hypo hesize ha HOXA4 plays a ole in g ow h- egula ing pa hways a ec ing SRS, ela ed g ow h diso de s, and also s a u e in gene al. Rema kably, he e ec size o HOXA4 me hyla ion in ou heal hy child coho was la ge han ound o any o nine highly signi i- can ly heigh -associa ed SNPs. As g ow h diso de s sha e many diagnos ic ea u es, i is plausible ha unde lying gene ic o epigene ic de ec s a e sha ed ac oss closely ela ed pheno ypes. Ou esul s sugges a new candida e egion po en ially ele an in he pa hogenesis o SRS, o he g ow h diso de s and g ow h in gene al. Me hods Pa ien s and con ols. We s udied 44 SRS pa ien s. The pa ien s included h ee subg oups based on molec- ula indings: 21 pa ien s wi h 11p15 LOM, 10 pa ien s wi h UPD(7)ma , and 13 pa ien s nega i e o bo h 11p15 LOM and UPD(7)ma , assigned as clinical SRS. One pa e nal unipa en al disomy 7 (UPD(7)pa )27 sample wi h- ou g ow h e a da ion was also included o compa ison. One o he pa ien s in UPD(7)ma had segmen al UPD(7q31-q e )ma . SRS diagnoses we e made based on c i e ia o SRS diagnosis in use a he ime o e alua- ion. In o ma ion o bi h weigh , bi h leng h, pos na al g ow h, body asymme y, p o uding o ehead, ela i e mac ocephaly and eeding di icul ies a e lis ed in Supplemen a y Table6. Bi h leng h and weigh o ges a ional age in SD scale a e acco ding o Finnish g ow h e e ences38. Ten con ol blood samples we e ob ained om adul olun ee s o no mal heigh . Six addi ional adul con ol samples we e om heal hy blood dono s31. 39 pa ien s wi h se e e g ow h es ic ion o unknown e iology (SGR) we e included in a a ge ed analysis o HOXA4 hypo- me hyla ion. SGR pa ien s had se e e g ow h es ic ion bu did no ul ill c i e ia o SRS diagnosis a he ime www.na u e.com/scien i ic epo s/ 9 SCIENTIFIC REPoR S | 7: 15693 | DOI:10.1038/s41598-017-16070-5 o e alua ion. In o ma ion on bi h leng h and bi h weigh o ges a ional age, and pos na al g ow h is lis ed in Supplemen a y Table7. The majo i y o he SGR pa ien s we e bo n small o ges a ional age (SGA) and all o hem had pos na al g ow h es ic ion. Me hyla ion da a om he BAMSE s udy12 we e used o es o co e- la ion wi h heigh in heal hy child en. The s udy was app o ed by he E hical Re iew Boa d o he Hospi al o Child en and Adolescen s, Helsinki Uni e si y Cen al Hospi al, Helsinki, Finland (SRS and SGR pa ien s and con ols) and he E hical Re iew Boa d No h a Ka olinska Ins i u e (BAMSE, blood dono con ols). In o med consen was ob ained om all s udy pa icipan s. Clinical in es iga ions ha e been conduc ed acco ding o he Decla a ion o Helsinki. Genome-wide me hyla ion analysis wi h Illumina 450K BeadChip assay. Me hyla ion analysis was ca ied ou on 44 SRS pa ien s, one indi idual wi h UPD(7)pa , and 10 con ol samples. DNA om EDTA blood samples was ex ac ed using he FlexiGene DNA Ki (Qiagen) acco ding o manu ac u e ’s ins uc ions. 500 ng o DNA was bisul i e con e ed wi h he EZ-96 Me hyla ion Ki (Zymo esea ch co po a ion) acco ding o he manu ac u e ’s ins uc ions. Samples om 27 pa ien s, UPD(7)pa and 10 con ol samples we e analyzed a he Ka olinska Ins i u e Bioin o ma ic and Exp ession Analysis (BEA) Co e Facili y, and addi ional 17 samples a he Mu a ion Analysis Co e Facili y (MAF, www.ma .ki.se). Bisul i e- ea ed DNA was ampli ied, agmen ed and hyb idized o he HumanMe hyla ion450 BeadChip (Illumina) acco ding o s anda d Illumina p o ocol and imaging was pe o med using Illumina iScan scanne . Addi ional da a om simila ly pe o med me hyla ion analysis o six whole blood con ol samples31 p oduced a MAF we e in eg a ed in o he analysis. Da a p oduced a di e en acili ies we e compa able (da a no shown). Me hyla ion da a o he BAMSE coho had also been p oduced a MAF. Bioin o ma ics analysis o he Illumina 450K BeadChip assay o SRS pa ien s and con- ols. The.ida iles we e impo ed in R and analyzed using he min i package39. A e quali y check, he da a we e no malized using he subse -quan ile wi hin a ay no maliza ion (SWAN) me hod40. The p obes o e lap- ping wi h known SNPs as well as p obes p one o c oss-hyb idiza ion p oblems we e emo ed, ende ing 361,948 CpG si es o u he analysis. The M- alues we e ex ac ed and he ba ch e ec s emo ed using he ComBa me hod41. Di e en ially me hyla ed CpG si es we e ob ained using linea models (y~subg oup + gende ) and pai wise compa isons wi h empi ical Bayes as implemen ed in he limma package42. The di e en ially me hyla ed CpG si es wi h p- alue < 0.05 a e Benjamini and Hochbe g co ec ion o mul iple es ing we e conside ed o u he analysis. Fil e ing p ocess o common me hyla ion di e ences in genome-wide me hyla ion da a. We analyzed 361,948 CpGs o signi ican di e en ial me hyla ion o he UPD(7)ma , 11p15 LOM, and clinical SRS g oups in compa ison o con ols. A he i s s age o il e ing ( il e 1), we selec ed CpGs ha showed signi ican di e en ial me hyla ion wi h a nominal p- alue < 0.05 (empi ical Bayes signi icance) o each o he h ee pa ien g oups when compa ed o con ols. Ou o hese common di e en ially me hyla ed CpGs we selec ed hose ha ul illed wo c i e ia: 1) he e we e a leas h ee consecu i e CpGs p obes wi h a signi ican di e ence, 2) he median me hyla ion le el o all o he h ee subg oups was ei he lowe o highe han ha o he con ols ( il e 2). Consecu i e CpGs we e de ined as occu ing nex o each o he based on hei genomic coo dina es; hey may no be ep esen a i e o ue consecu i e CpGs in he genome, as hose CpGs no included in he a ay as well as hose disca ded in he quali y con ol s eps we e no conside ed. The il e ing p ocess is shown in Fig.1. The ou pu lis o signi ican CpGs ha passed he il e ing p ocess was subjec ed o u he selec ion by com- pa ing he median me hyla ion le el o each o he subg oups o con ols (Supplemen a y Table1). We se a h eshold o 5% me hyla ion di e ence o u he na ow down signi ican p obes. The HOXA4 di e en ially me hyla ed a ea was de ined by he CpGs ha passed he il e ing p ocess, comp ising al oge he 12 + 3 consec- u i e CpG si es. DNA me hyla ion analysis o HOXA4 egion wi h EpiTYPER. We measu ed he me hyla ion le el o CpGs in he HOXA4 di e en ially me hyla ed a ea using Sequenom EpiTYPER (Sequenom, San Diego, CA, USA). The analysis was ca ied ou on 44 SRS pa ien s, one indi idual wi h UPD(7)pa , 39 SGR pa ien s, and 16 con ols. Bisul i e con e sion wi h he EZ-96 Me hyla ion Ki (Zymo esea ch co po a ion), subsequen sample p ocessing, and quali y con ol we e ca ied ou a he Ka olinska Ins i u e Mu a ion Analysis Co e Facili y. All assays we e pe o med in echnical duplica es, and mean o he duplica e measu emen s was used in he analyses. I one o he duplica e samples ailed, hen da a om he one in o ma i e sample only was used. One con ol sam- ple, one SGR sample and he duplica e o one 11p15 LOM SRS sample we e excluded due o expe imen al ailu e. No mali y o he da a was es ed using Shapi o-Wilk es o no mali y and homogenei y o a iances by Le ene’s s a is ic es . Da a we e u he analyzed by one-way ANOVA in SPSS. Dunne - es s, which compa e all pa ien g oups agains he con ol g oup, we e pe o med pos hoc. Indi idual me hyla ion le el analysis and hypome hyla ed s a us. Fo each CpG in he HOXA4 di e en ially me hyla ed a ea, and each EpiTYPER si e, s anda d de ia ion was calcula ed om he me hyla ion le el o he 16 con ol samples. No mal dis ibu ion o he con ol g oup da a o each CpG (o a combina ion o CpG si es in Epi ype si e HOXA4_F6_CpG_12.13.14) was es ed by Shapi o-Wilk no mali y es . Indi idual SD alues o he SRS pa ien s we e calcula ed based on he s anda d de ia ion o he con ols. A CpG si e was conside ed hypome hyla ed when he me hyla ion le el was below -2 SD. We conside ed an indi idual pa ien gene ally hypome hyla ed a HOXA4 when he me hyla ion le el was below -2 SD o a leas wo- hi ds o he CpG si es. Fo he Illumina assay, he eshold was 10 ou o 15 CpG si es (67%), and o he EpiTYPER assay 3 ou o 4 si es (75%).