A comprehensive repertoire of tRNA-derived fragments in prostate cancer

Onco a ge 24766 www.impac jou nals.com/onco a ge www.impac jou nals.com/onco a ge / Onco a ge , Vol. 7, No. 17 A comp ehensi e epe oi e o RNA-de i ed agmen s in p os a e cance Michael Ol edy1,4,5,*, Mau o Sca a illi2,3,*, You i Hoogs a e1, Tapio Visako pi2,3,#, Guido Jens e 1,#, Elena S. Ma ens-Uzuno a1,# 1Depa men o U ology, E asmus MC, Ro e dam, The Ne he lands 2Ins i u e o Biosciences and Medical Technology-BioMediTech, Uni e si y o Tampe e, Tampe e, Finland 3Fimlab Labo a o ies, Tampe e Uni e si y Hospi al, Tampe e, Finland 4 Cu en add ess: VIB Cen e o he Biology o Disease, KU Leu en, Leu en, Belgium 5Cu en add ess: Cen e o Human Gene ics, KULeu en, Leu en, Belgium *These au ho s con ibu ed equally o his wo k #Th ee las au ho s con ibu ed equally o his wo k Co espondence o: Elena S. Ma ens-Uzuno a, e-mail: [email p o ec ed] Keywo ds: RNA-de i ed agmen s ( RFs), p os a e cance (PCa), RNA-sequencing, non-coding RNA, bioma ke Recei ed: Sep embe 04, 2015 Accep ed: Ma ch 02, 2016 Published: Ma ch 23, 2016 ABSTRACT P os a e cance (PCa) is he mos common cance among men in de eloped coun ies. Al hough i s gene ic backg ound is ho oughly in es iga ed, a he li le is known abou he ole o small non-coding RNAs (sncRNA) in his disease. RNA- de i ed agmen s ( RFs) ep esen a new class o sncRNAs, which a e p esen in a b oad ange o species and ha e been epo ed o play a ole in se e al cellula p ocesses. He e, we analyzed he exp ession o RFs in esh ozen pa ien samples de i ed om no mal adjacen p os a e and di e en s ages o PCa by RNA-sequencing. We iden i ied 598 unique RFs, many o which a e de egula ed in cance samples when compa ed o no mal adjacen issue. Mos o he iden i ied RFs a e de i ed om he 5’- and 3’-ends o ma u e cy osolic RNAs, bu we also ound RFs p oduced om o he pa s o RNAs, including p e- RNA aile s and leade s, as well as RFs om mi ochond ial RNAs. The 5’-de i ed RFs comp ise he mos abundan class o RFs in gene al and ep esen he majo class among up egula ed RFs. The 3’-de i ed RFs ypes a e dominan among down egula ed RFs in PCa. We alida ed he exp ession o h ee RFs using qPCR. The a io o RFs de i ed om RNALysCTT and RNAPheGAA eme ged as a good indica o o p og ession- ee su i al and a candida e p ognos ic ma ke . This s udy p o ides a sys ema ic ca alogue o RFs and hei dys egula ion in PCa and can se e as he basis o u he esea ch on he bioma ke po en ial and unc ional oles o RFs in his disease. INTRODUCTION P os a e cance (PCa) is he second mos common cance in men wo ldwide [1]. The ea men o PCa is hampe ed by he lack o eliable ma ke s o disease ou come p edic ion leading o inco ec pa ien s a i ica ion, o e ea men and consequen side e ec s om p os a ec omy and adia ion he apy [2]. A be e unde s anding o he molecula mechanisms behind he onse and p og ession o PCa is needed in o de o disco e be e ma ke s and de elop new he apeu ic s a egies. The ole o small non-coding RNAs (sncRNAs) o he han mic oRNAs (miRNAs) in PCa is poo ly unde s ood. The apid p og ess and popula i y o high h oughpu sequencing led o he disco e y o a no el class o sncRNAs de i ed om RNAs and named RNA- de i ed agmen s ( RFs) [3–5]. RFs a e p esen ac oss all domains o li e [6–8]. While ini ially conside ed andom p oduc s o RNA u no e , hei abundance and ubiqui ous exp ession sugges ha RFs a e ac ual biological en i ies [6, 7]. RFs can be gene a ed by endonucleases such as ibonuclease T2 (Rny1p) in yeas and angiogenin o dice 1 in human. Based on size, hey a e di ided in o wo g oups. Onco a ge 24767 www.impac jou nals.com/onco a ge The i s g oup consis s o RFs wi h a size o 30 o 35 n , which a e gene ally e e ed o as RNA hal es o s ess- induced RFs. RNA hal es a e p oduced by endonucleoly ic clea age a he an icodon loop o he ull-leng h RNA. The second g oup consis s o RFs wi h a size o abou 20 n and can be u he di ided in o 5’- and 3’-de i ed RFs, o igina ing om he 5′- and 3′-pa s o ma u e RNAs, espec i ely [4, 9, 10]. The small RNAs de i ed om he 5’-leade and 3’- aile sequences o he p ecu so RNAs (p e- RNAs) a e also classi ied as RFs [5, 11, 12]. Exp ession o RFs is de ec ed in di e en cance cell lines, including he PCa cell lines LNCaP and C4-2 [4, 5, 13–15]. In a p e ious s udy, we epo ed he disco e y and di e en ial exp ession o RFs in clinical samples o PCa [16]. This sugges s ha RFs migh play an impo an ole in he pa hogenesis o cance . The mechanism behind he unc ion o RFs appea s o be di e se. Se e al epo s demons a e ha RF le els a e ele a ed by cellula s ess condi ions and pa icula ly unde oxida i e s ess such as hypoxia [10, 13, 15, 17]. RFs a e also in ol ed in pos - ansc ip ional egula ion o gene exp ession ia di ec inhibi ion o p o ein syn hesis by displacing he eIF4G ansla ion ini ia ion ac o om mRNA [18–20]. Mo eo e , a 3′-de i ed RF iden i ied in B-cell lymphoma cells possesses he unc ional cha ac e is ics o a guide RNA ha supp esses p oli e a ion and modula es esponse o DNA damage in a miRNA- ashion [21]. I has also been shown ha RFs can compe e o he binding si es o he RNA-binding p o ein YBX1, which is in ol ed in he s abiliza ion o oncogenic ansc ip s supp essing cell g ow h and in asion [15]. In his way, RFs an agonize he ac i i y o YBX1 and ac as umo supp esso s. Taken oge he , hese indings s ongly sugges a unc ional ole o RFs in umo igenesis. Ve y ecen ly, i was p oposed ha al hough RFs a e de ined biological en i ies, hei composi ion and abundance in he ansc ip ome is dependen on gende , issue, disease and e en disease sub ype [22]. This sugges s ha RFs can be explo ed as no el sensi i e bioma ke s o disease. Ye , s udies p o iding sys ema ic insigh in o he composi ion and exp ession o he RF ansc ip ome h oughou a ious disease s ages a e s ill missing. He e, we analyze RF exp ession in an ex ended coho o clinical samples ep esen ing p og essing s ages o PCa. We cons uc a da abase o RFs exp essed ac oss PCa samples and iden i y he mos di e en ially exp essed RFs. Finally, we pe o m a qPCR quan i ica ion in wo coho s o clinical samples o alida e he di e en ial exp ession o selec ed RFs. RESULTS In en o y o RFs exp essed in PCa In o de o ob ain a global o e iew o he RF epe oi e in PCa, we analyzed RFs ac oss no mal adjacen p os a e (NAP), benign-p os a e hype plasia (BPH), PCa om adical p os a ec omies, ans-u e h al esec ed issue om cas a ion esis an PCa (TURP_ PCa), and lymph node me as asis (LN_PCa) using nex - gene a ion RNA sequencing (Table 1). All 21 cy osolic RNA iso ypes (including selenocys ein RNAs) we e ound o p oduce RFs in a iable amoun s (Figu e 1A). RNAAla and RNALys showed he highes numbe s o mapped RFs, while he leas RFs we e p oduced om RNA Ile and RNA Asp . The aw sum o RFs weakly co ela ed wi h he numbe o RNA genes pe iso ype o an icodon, as well as wi h he pe cen age o codon usage (Supplemen a y Figu e 1; codon usage om h p://g nadb.ucsc.edu/Hsapi19/ Hsapi19-summa y-codon.h ml). RFs de i ed om 15 ou o 20 mi ochond ial RNAs (m RNAs) we e also de ec ed (Figu e 1A). We could no de ec RFs co esponding o he mi ochond ial RNA iso ypes m RNAGln, m RNAGlu, m RNALys, m RNAT p and m RNAVal. Wi h he excep ion o m RNA Phe , mos m RNA iso ypes had a lowe numbe o mapped RFs, compa ed o cy osolic RNAs. The ead coun o m RNAPhe in he NAP g oup was 83- old highe han he a e age o all o he m RNA ead coun s. In o de o quan i y he exp ession o RFs, we assembled a PCa RF-da abase using he agmen de ec ion algo i hm FlaiMappe [23]. The ead-co e age o ma u e cy oplasmic RNAs ac oss all g oups was analyzed using CLC-Bio Genomics Wo kbench. Ini ially, 1175 RFs we e iden i ied and mapped o 386 unique cy osolic RNAs [24]. Howe e , since RNA sequences a e highly conse ed wi hin RNA iso ypes, some RFs we e mapped o mo e han one unique RNA (Supplemen a y Figu e 2) and he o al ead-coun in he ini ial mapping was equally di ided ac oss hem. Upon u he examina ion, we no iced ha his causes unde ep esen a ion o sequence coun s o RFs ha had iden ical sequence bu could be mapped o mul iple RNA iso ypes. The e o e, RFs wi h iden ical sequences we e me ged in o single en ies, e en i hey could be de i ed om di e en RNAs, and hei co esponding eads we e summed. A e his co ec ion, a o al o 598 unique RFs we e iden i ied (Supplemen a y Table 1). Mul iple agmen a ion pa e ns, in combina ion wi h low ead-coun , caused low eliabili y in he au oma ed p edic ion o RFs de i ed om m RNAs (Supplemen a y Figu e 3). The e o e, hese RFs we e omi ed om u he analysis. Based on hei size, RNA-de i ed agmen s can be gene ally sepa a ed in o wo majo ca ego ies: RNA hal es, wi h a size o 30-35 n and small RNA agmen s ( RFs), wi h a size o app oxima ely 20 n . In ou da ase , small RFs we e p edominan and hei sizes anged om 15 o 23 n (Figu e 1B). The mos abundan RFs, howe e , we e be ween 18 o 21 n , while 40% o RFs we e 19 n long (Figu e 1B). A g oup o longe RFs, wi h sizes be ween 25 and 29 n , was also iden i ied. In addi ion o RFs de i ed om ma u e RNAs, we we e also able o de ec agmen s co esponding o he 5’-p e- RNA leade (5’U- RFs) and 3’-p e- RNA aile Onco a ge 24768 www.impac jou nals.com/onco a ge (3’U- RFs) sequences o a ious RNAs (Supplemen a y Table 2). The leng h o 5’U- RFs and 3’U- RFs a ied be ween 15 and 25 n . Mos 5’U- RFs we e 17 n long and mos 3’U- RFs we e 18 n long (Supplemen a y Figu e 4A). In e es ingly, mo e han 54% o 3’U- RFs and 30% o 5’U- RFs we e de i ed om sequences igh nex o o 1 n away om he ma u e RNA sequence (Supplemen a y Figu e 4B-4C), sugges ing ha hey a e p oduced du ing he no mal p ocessing o p e- RNA. Bo h 5’U- RFs and 3’U- RFs showed o e all low exp ession alues (da a no shown), wi h he excep ion o RF-1001/cand45. This agmen was p e iously de ec ed in PCa cell lines, as well as in human colon ca cinoma and human emb yonic kidney cells [5, 11]. In ou lib a ies RF-1001/cand45 showed ead coun s om 40 000 in he NAP and PCa (a e age) g oups o 110 000 in he LN_PCa g oup. RFs de i ed om he 5’-end a e dominan in PCa The majo i y o RFs iden i ied in ou samples o igina e om he 5′- and he 3′-end o RNAs (Figu e 1C). This is in conco dance wi h p e ious s udies collec i ely epo ing on he exis ence o sho RFs de i ed om he 5′- and he 3′-end o ma u e RNAs [5, 25, 26]. To analyze he ela i e abundance o each RF class in ou da ase we examined he s a and end posi ions o all unique RFs on hei p ecu so RNAs. All he agmen s wi h a 3’-end nucleo ide a posi ion ≤40 on he ma u e RNA sequence we e conside ed as 5′-de i ed, whe eas all agmen s wi h he i s 5’-nucleo ide a posi ion ≥30 on he ma u e RNA sequence we e conside ed as 3′-de i ed. Based on agmen uniqueness, we ound compa able a es o RF ypes, i.e. 51.7% co esponded o 5′-de i ed RFs and 44.2% o 3′-de i ed RFs. Ne e heless, when ela i e agmen abundance was aken in o accoun a s ong bias owa ds he 5′-de i ed (84.7%) s. he 3’-de i ed RFs was obse ed. To ge a mo e p ecise o e iew o he localiza ion o RFs, we also analyzed hei s a - and end- posi ion equencies. In e es ingly, mo e han 26% o all unique RFs, which in he e ms o abundance accoun o o e 80% o all RFs, we e ound o s a a posi ion 1 on he ma u e RNA sequence (Figu e 1D–1E). Mos o hese Table 1: Clinical pa ame e s o he samples used o he RNA-sequencing G oup name Pa ien samples TMPRSS2_ ERG usion ETV1 abe a ions Cance cells (%) Gleason sco e P eope a i e PSA Mean (Min.- Max.) S a us a e adical p os a ec omy* NAP 4 0 (0%) 0 0 N/A 5.8 (2.5 - 11) N/A BPH 4N/A N/A 0N/A N/A N/A PCa6_cu 4 4 (100%) 0 70-90 3+3 3.8 (2.0 - 7.2) Cu ed PCa6_ no usion 4 0 (0%) 0 70-90 3+3 10.8 (0.5 - 23.7) Recu en PCa6_ TERG 4 4 (100%) 0 80-90 3+3 7.3 (2.0 - 10.5) Recu en PCa6_ ecu 4 4 (100%) 0 80-90 3+3 11.3 (6.0 - 64.3) Recu en PCa7_ ecu 4 2 (50%) 1 ( usion) 80-100 4+3 25.2 (6.5 - 64.3) Recu en PCa8_ ecu 3 1 (25%) 1 (o e exp essed) 90-100 4+4 (5) 32.4 (31.9 - 32.9) Recu en TURP_PCa (cas a ion esis an ) 4 1 (25%) 1 ( usion) 90-100 (3+4) o (5+4) Unknown Recu en LN_PCa 4 3 (75%) 1 ( usion) 100 4+4(5)† 154.6 (80.2 - 252.5) N/A * Pa ien s we e conside ed cu ed i he e was no biochemical elapse o de ec ion o me as asis a e adical p os a ec omy † Gleason sco e o he p ima y umo G oup abb e ia ions: NAP, no mal adjacen p os a e; BPH, benign p os a ic hype plasia; PCa, o gan-con ined p os a e cance ; cu / ecu , cu ed/ ecu en a e adical p os a ec omy; PCa6_no usion, PCa Gleason sco e 3+3 wi h no TMPRSS2- ERG usion o ETV abno mali ies; PCa6_TERG, PCa Gleason sco e 3+3 wi h TMPRSS2-ERG usion; TURP_PCa, ans- u e h al esec ion o he p os a e, cas a ion esis an PCa; LN_PCa, lymph node me as asis. Onco a ge 24769 www.impac jou nals.com/onco a ge RFs ha e he end a posi ion 19 on he ma u e RNA. Based on he peaks gene a ed by he s a posi ions o all unique agmen s (Figu e 1D), we obse ed ha he RF pool cons i u es o se e al dis inc classes (no e he peak appea ing be o e 20 n , ano he a a ound 40 n and ano he be o e 60 n o he ma u e RNA). While ca ego izing RFs in o 5′- o 3′-de i ed RFs is e y common, we ound ha a leas 5 di e en classes a e p esen ac oss ou samples. The e o e, we classi ied RFs in o (i) 5e- RFs wi h a s a posi ion in he i s nucleo ide o he 5’-end o he RNA (“e” s ands o “end”); (ii) D- RFs wi h a s a posi ion be ween nucleo ides 12-23 and o e lapping he D-loop o he p ecu so - RNA; (iii) A- RFs s a ing be ween nucleo ides 31-39 and o e lapping wi h he an icodon loop; (i ) V- RFs wi h a s a be ween nucleo ides 45-49 and o e lapping he a iable loop; and inally, ( ) 3e- RFs s a ing be ween nucleo ides 50-60 and o e lapping he T loop (Figu e 1F). While 5e- RFs ep esen he mos abundan class o RFs (app oxima ely 75%), o he classes o RFs appea o ha e e y simila exp ession (<10% abundance) compa ed o each o he (Figu e 1G). In e es ingly, simila RF ypes ha e been de ec ed in he Figu e 1: RF ypes in p os a e cance . A. Hea map showing he o al ead coun s mapped o indi idual RNA iso ypes in h ee s udy g oups: NAP, no mal adjacen p os a e; PCa, p os a e cance g oup (consis ing o 6 di e en sample pools, he a e age alue is shown); LN_PCa, lymph node me as asis. The colo and i s co esponding alue in log10 scale a e depic ed on he igh . B. RF leng h as based on he ead abundance (o ange ba s) and uniqueness (do ed line). C. G aph depic ing he loca ions o mapped RFs on he sequences o ma u e RNAs. Full-leng h RNA sequences a e aligned o he middle using he an icodon posi ion. RFs mapped o hese RNAs a e depic ed as g ey ba s which ela i e abundance pe pa icula RNA is e lec ed by he colo in ensi y (ligh g ey, low abundance; black, high abundance). RNAs wi h only one mapped RF a e clus e ed a he op, RNAs wi h wo mapped RFs in he middle and RNAs wi h mul iple mapped RFs a e a he bo om. D-E. S a (blue line) and end (o ange line) posi ions o RFs on he ma u e RNA sequence. Rela i e abundance o each RF s a and end based on he uniqueness (D) o abundance (E) is shown. F. An illus a ion o a ious RF ypes and hei app oxima e loca ion on he seconda y s uc u e o RNA. G. Rela i e p opo ions o each RF ypes in ou da ase as based on he uniqueness (% o unique independen eads) o abundance (% o o al numbe o eads). Onco a ge 24770 www.impac jou nals.com/onco a ge lowe euka yo e Te ahymena he mophile, sugges ing he exis ence o an e olu iona y conse ed RNA p ocessing mechanisms [27]. Mo eo e , he posi ion o hese peaks was ound o o e lap wi h all RNA loops, indica ing ha endonucleoly ic clea age occu s in he single-s anded loop egions o RNAs. Se e al RFs a e de egula ed in PCa To in es iga e whe he RF p oduc ion is dys egula ed in PCa, we compa ed he exp ession le els o RFs in no mal issue and in samples om di e en clinical s ages ep esen ing p og essing disease (Table 1). While exp ession le els o o he ypes o sncRNAs co ela ed well be ween he wo lib a ies ep esen ing non-malignan issue, i.e. NAP and BPH (Pea son =0.89, P- alue <0.0001; median old-change -0.002), RFs showed lowe co ela ion and a e y high one-di ec ional de ia ion owa ds inc eased exp ession in he BPH lib a y (Pea son =0.81; P- alue <0.0001; median old-change 0.758; Supplemen a y Figu e 5). These esul s indica e ha RFs, as opposed o o he sncRNAs, migh be di e en ially exp essed in benign p os a e hype plasia. This di e ence can be explained by he di e en ana omical o igin o he BPH and NAP/PCa samples. While BPH occu s exclusi ely in he ansi ion zone o he p os a e, p os a e umo s a e p edominan ly localized in he pe iphe al zone. Bo h zones a e cha ac e ized by dis inc exp ession p o iles indica ing di e en ial egula ion o a la ge numbe o genes [28]. Fo his eason, BPH was excluded as a con ol sample om u he analyses. Only NAP issue, con i med o con ain 0% umo cells by wo independen pa hologis s, was used as a con ol sample. I should be acknowledged, ha adjacen issue could be al e ed in gene/ncRNA exp ession due o umo -s oma pa ac ine in luences [29]. Ne e heless, as much as s oma/epi helium in e ac ions a e una oidable in expe imen s we e samples a e de i ed by mac o-dissec ion, much ca e has been aken o minimize he in luence o he s omal compa men on he exp ession p o iles o small RNAs by selec ing issue sec ions o a leas 70% epi helial cells [24]. We ound se e al RFs o be signi ican ly di e en ially exp essed in PCa when compa ed o NAP (Kal’s Z- es wi h Bon e oni co ec ion, p- alue < 0.05) (Figu e 2 and Supplemen a y Table 3-4). The numbe o di e en ially exp essed RFs a ied sligh ly be ween he s ages o PCa, wi h a minimum o 27 di e en ially exp essed RFs in PCa6_ ecu g oup and a maximum o 61 di e en ially exp essed RFs in he LN_PCa g oup (Figu e 2, Supplemen a y Table 4). We iden i ied 12 RFs o be commonly di e en ially exp essed be ween ecu en PCa g oups wi h Gleason g ade 6, 7, o 8 (Supplemen a y Table 5). O hese, 5 we e up egula ed, 6 down egula ed and 1 was down egula ed in PCa6 g oup bu up egula ed in PCa7 and PCa8 g oups (Supplemen a y Table 5). This esul indica es ha a small subse o di e en ially exp essed RFs can be ound ac oss inc easing g ades o PCa. In summa y, we ound 110 di e en ially exp essed RFs ac oss ou da ase , ou o which 72 we e up egula ed, 24 down egula ed and 13 ha we e up egula ed in one bu down egula ed in o he g oup. RFs de egula ed in PCa belong o dis inc classes I has been p oposed ha 5′- bu no 3′-de i ed RFs, play a ole in s ess g anule assembly o inhibi ion o p o ein syn hesis in i o [19, 30]. On he o he hand, some 3′-de i ed RFs a e able o ep ess hei mRNA a ge s in a miRNA-like ashion and may exe umo supp essi e unc ions [21, 31]. In e es ingly, ou esul s indica e ha he de egula ion o 5’-de i ed RFs di e s om ha o 3’-de i ed RFs (Figu e 2). In o de o s udy which RF ypes a e p esen among he down egula ed and up egula ed RFs in PCa we compa ed he pe cen age o di e en RF ypes among ou g oups o up egula ed and down egula ed RFs (Figu e 3A–3B). We no iced majo di e ences in he abundance o RF ypes in bo h lis s. Mos o he up egula ed RFs we e 5e- RFs (50%) and mos down egula ed we e 3e- RFs (50%). We selec ed RFs o igina ing om 6 di e en RNAs o u he analysis and qPCR alida ion. All o hem we e commonly di e en ially egula ed in ecu en PCa g oups wi h Gleason g ade 6, 7, o 8 (Supplemen a y Table 5). Ou o hese, 4 RFs ( h ee 5e- RFs and one D- RF) we e up egula ed in PCa (Figu e 3C–3F), and 2 RFs (bo h om he 3e- RF class) we e down egula ed (Figu e 3G–3H). Speci ic RF signa u es can se e as p ognos ic ma ke o ecu en p os a e cance The exp ession le els o RFs selec ed o alida ion by qPCR we e s udied in a coho o clinical samples ob ained om E asmus MC, Ro e dam (coho 1) and a coho o samples om Tampe e Uni e si y Hospi al, Tampe e (coho 2). The NAP samples we e iden ical o bo h coho s and we e p ocessed independen ly in coho 1 and coho 2 o accoun o echnical di e ences in sample ea men . Using cus om designed p ime s, we could de ec h ee RFs (Figu e 4A–4C). RF-544 (de i ed om RNA PheGAA ) was signi ican ly down egula ed in he ecu en PCa compa ed o NAP o cu ed PCa in coho 1 (Figu e 4A). In coho 2, RF-544 was down egula ed in PCa wi h Gleason sco e highe han 7 o in PCa wi h pa hological s age 3 sugges ing associa ion wi h agg essi e o la e s age disease. The di e en ial exp ession o his RF was also con i med in a second deep sequencing analysis o a sub-se o PCa samples om Tampe e Uni e si y Hospi al (unpublished da a). RF-315 (de i ed om RNALysCTT) was signi ican ly up egula ed in all PCa g oups o coho 2 (Figu e 4B). We could no de ec s a is ically signi ican di e ence in he exp ession o RF-315 in he smalle coho 1. Ne e heless, he e was a clea end o RF-315 up egula ion in he PCa samples. Onco a ge 24771 www.impac jou nals.com/onco a ge RF-562 (de i ed om RNAGlyTCC) was signi ican ly down egula ed in PCa ecu en s. NAP g oup in he coho 1 and in he PCa pT3 s. NAP g oup in he coho 2 (Figu e 4C). In e es ingly, RF-544 was consis en ly down egula ed in samples om pa ien s ha de eloped ecu en disease compa ed o samples om pa ien s ha we e cu ed by adical p os a ec omy in bo h coho s. Fu he mo e, RF-544 exp ession was lowe in high- (Gleason sco e ≥7) compa ed o low-g ade (Gleason sco e <7) umo s (Figu e 4A). Vice e sa, RF-315 demons a ed a clea end o up egula ion in ecu en disease and i s exp ession was highe in high-g ade umo s (Figu e 4B). The e o e, we easoned ha he exp ession o hese wo RFs migh be p ognos ic o agg essi e umo g ow h and disease ecu ence a e adical p os a ec omy. We ook ad an age o he opposing exp ession pa e ns o hese wo RFs and calcula ed he exp ession a io RF- 315/ RF-544 o bo h coho s (Figu e 4D). The RF-315/ RF-544 a io showed signi ican di e ences, clea ly dis inguishing high om low g ade PCa and cu ed om ecu en disease. Mo eo e , high exp ession a io was signi ican ly associa ed wi h poo e p og ession- ee su i al and sho e pe iod o disease elapse (Figu e 4E), sugges ing ha he RF-315/ RF-544 a io migh ep esen a help ul clinical bioma ke o disease p og ession. Figu e 2: Di e en ially exp essed RFs in p os a e cance . Plo ed a e no malized ead-coun alues o each RF in he no mal adjacen p os a e e sus a ious s ages o p os a e cance . The baseline alue o RFs ha a e no exp essed is 1. Full lines ep esen 4- old change bo de lines. Colo ed poin s ep esen signi ican ly changed RFs (Kal’s Z- es on p opo ions, Bon e oni co ec ed p- alues, p < 0.05) labeled as 5’-de i ed (magen a) and 3’-de i ed (g een) RFs. RFs wi h he 3’-nucleo ide a a posi ion ≤40 on he p ecu so RNA sequence a e conside ed as 5′-de i ed. RFs wi h he s a nucleo ide a a posi ion ≥30 on he p ecu so RNA sequence a e conside ed as 3′-de i ed. RFs ha do no all in o ei he o hese wo ca ego ies a e shown in blue. Posi ions o RF-310, RF-315, and RF-389 a e indica ed as an example o h ee di e en ially exp essed RFs. The g aph a he bo om igh co ne summa izes he o al numbe o di e en ially exp essed RFs pe g oup. The amoun o RFs wi h ≥4- old di e en ial exp ession a e indica ed in da k o ange (up egula ed) o da k blue (down egula ed) colo . Onco a ge 24772 www.impac jou nals.com/onco a ge DISCUSSION The echnical p og ess in sequencing echnologies and he apid inc ease in he numbe o s udies on sncRNA led o he disco e ies o no el small RNA classes including RFs. Since hei ini ial iden i ica ion, RFs ha e been desc ibed in a ple ho a o species and knowledge abou hei unc ion in he cell is s a ing o accumula e. Al hough se e al s udies desc ibe exp ession o RFs in human cell lines, hei ac ual epe oi e in human issues emains la gely unknown [5, 15, 16, 22]. He e, we s udied he composi ion and exp ession o RFs in clinical PCa samples ep esen ing p og essing disease s ages. We ound ha all cy osolic RNAs p oduced RFs in he size ange o 18-21 n , ep esen ing he small class o RFs. The longe RNA hal es we e no as common, which is a consequence o he size selec ion (~15-35 n ) applied o he isola ion o sncRNAs ac ion in ou s udy. We ound a signi ican bu weak co ela ion be ween he exp ession o RFs pe RNA and he codon usage o RNAs, sugges ing ha al hough RF exp ession is dependen on he exp ession le els o hei p ecu so s, mos likely addi ional mechanisms con ol RF le els in he cell. The accu a e quan i ica ion o agmen s de i ed om small RNAs in RNA sequencing da a equi es a p ecise anno a ion o he exac posi ion o he agmen on i s p ecu so ansc ip . To p edic he loca ions o RFs and quan i y hei exp ession we used he p og ams FlaiMappe and CLC-Bio Genomics Wo kbench [23]. We Figu e 3: F equency o RF ypes among di e en ially exp essed RFs. A-B. S a (blue line) and end (o ange line) posi ions o RFs on he ma u e RNA sequence and he p opo ions o each RF ype o 72 up egula ed (A) and 24 down egula ed (B) RFs. C-H. G aphs showing he exac posi ions o 6 selec ed RFs (shown in colo ) and o he RFs (shown in g ey) on hei RNA p ecu so s. The x-axis ep esen s he posi ion on he p ecu so sequence in nucleo ides. The y-axis ep esen s summed ead-coun s pe sample g oup. The exp ession le el pe g oup is indica ed by di e en colo (see he legend a he bo om). Exp ession le els o o he RFs a e shown as means ac oss PCa g oups. Onco a ge 24773 www.impac jou nals.com/onco a ge iden i ied 598 unique RFs de i ed om ma u e RNAs. Based on he pa o ma u e RNA om which agmen s o igina e, we could dis inguish 5 di e en RF classes. Ou o hese, he 5e- RFs class was he mos abundan o all and con ained he highes numbe o unique RFs. This inding is in ag eemen wi h o he epo s showing highe abundance o 5’-end de i ed RFs [4, 5, 22, 32– 34]. Gi en he ole o 5’-de i ed RFs in he inhibi ion o p o eosyn hesis and hei ole in he assembly o s ess g anules, a ype o s ess-induced cy oplasma ic oci wi h high concen a ion o un ansla ed mRNPs [10, 19, 30], i would be in e es ing o es hei po en ial o inhibi ansla ion and induce he assembly o s ess g anules in i o in PCa cell lines using he se o up egula ed 5’-de i ed RFs iden i ied in ou s udy. The impo ance o RFs in s ess g anule assembly becomes e en mo e in iguing hanks o he la es indica ions ha s ess g anules migh play an impo an ole in cance ia he nega i e egula ion o mTORC1-hype ac i a ion- induced apop osis [35]. This sugges s ha up egula ion o RFs migh be indi ec ly linked wi h he supp ession o apop osis in cance cells. Ou disco e y coho included pa ien -de i ed PCa samples wi h di e en clinico-pa hological cha ac e is ics. The majo di e ence in RF exp ession (a leas 110 unique di e en ially exp essed RFs) was obse ed be ween NAP and PCa issue indica ing ha global up egula ion o RF p oduc ion is associa ed wi h malignan ans o ma ion. In e es ingly, 5e- RFs we e he p edominan class up egula ed in PCa. Recen ly, 5’- RFs we e ound o induce ansla ional inhibi ion in siRNA-independen way [36]. I was shown ha he ep essing ac i i y o 5’-de i ed Figu e 4: qPCR alida ion o RF-544, RF-315, and RF-562. A-C. RNA exp ession o RF-544 (A), RF-315 (B), and RF-562 (C) in coho s o clinical samples ob ained om E asmus MC (coho 1) and Tampe e Uni e si y Hospi al (coho 2). Mean alues a e indica ed by a ed line. D. Ra io o RF-315 (de i ed om RNALysCTT) o RF-544 (de i ed om RNAPheGAA). E. P og ession- ee su i al cu es o coho 1 and coho 2 based on he RF-315/ RF544 a io. Legend: NAP, no mal adjacen p os a e; PCa, p os a e cance ; PCa cu (PCa cu ed), pa ien s wi h no disease ecu ence a e adical p os a ec omy; PCa ecu (PCa ecu en ), pa ien s wi h biochemical ecu ence o me as a ic p og ession a e adical p os a ec omy; Gl <7, Gleason sco e <7; Gl ≥7, Gleason sco e 7, 8 o 9; pT2, pa hological s age 2; pT3, pa hological s age 3; *P- alue ≤0.05; **P- alue ≤0.01; ***P- alue ≤0.001. Onco a ge 24774 www.impac jou nals.com/onco a ge RFs was dependen on he p esence o a conse ed “GG” dinucleo ide a hei 3’-end, which is a common ea u e o ~75% o he up egula ed 5e- RFs desc ibed in his s udy. Compa ing ou da a se wi h an ex e nal RF da a se o PCa cell lines gene a ed by Lee e al. [5] demons a ed ha all RFs o igina ing om 3’-p e- RNA aile s and 32 ou o 36 5’- RFs desc ibed by Lee e al. we e de ec ed in ou s udy. This sugges s ha RFs in p os a e (cance ) issue and cell lines a e common and disc e e biological en i ies p oduced by de ined molecula mechanisms. Fo 3’-de i ed RFs we ound a small o e lap o only 6 ou o 77 RFs. A possible eason o ha could be ha 3’-de i ed RFs ep esen a class o RFs wi h a less s able exp ession. On he o he hand, ou esul s demons a e ha mos o he down egula ed RFs a e 3e- RFs, which migh be a gene al ea u e o PCa and PCa cell lines. I ha is he case, he limi ed o e lap o 3’-de i ed RFs be ween bo h da a se s migh be caused by he less eliable de ec ion o low exp essed ansc ip s. Down egula ion o 3’-de i ed RFs migh be an impo an e en in he onse o cance [21]. Possible unc ional implica ions o 3’- RF down egula ion in PCa can be de i ed om he demons a ed in ol emen o he 3′-de i ed RF CU1276 in B-cell lymphoma cells, which supp esses p oli e a ion and modula es he esponse o DNA damage [21]. Ano he ecen s udy in b eas cance desc ibed se e al RFs wi h umo supp esso unc ion ha o igina e om RNAGlyTCC, RNAGluYTC, RNAAspGTC, and RNATy GTA [15]. Upon induc ion in b eas cance cells, hese RFs supp ess he s abili y o mul iple oncogenic ansc ip s by sequence speci ic displacemen o hei 3′-UTRs om he RNA-binding p o ein YBX1. I can be assumed ha down egula ion o such RFs would lead o cance p og ession. In e es ingly, some o he down- egula ed 3’- RFs (e.g. RF-562 and RF-542) in ou sequencing lib a ies o igina e om he same RNAs. Fu u e in es iga ions should add ess he unc ional implica ions on gene egula ion in PCa caused by down egula ion o 3’-de i ed RFs and in pa icula RF-544. Due o high conse a ion o RNAs we we e unable o iden i y speci ic sequences ha would se e as a ecogni ion si e o RNA nucleases ha disc imina e and p e e ably clea e pa icula RNAs. Recen ly, i was p oposed ha ce ain RNAs swi ch om canonical o al e na i e olding and he abili y o do so migh cause he speci ic up egula ion o hei RFs. Fo example, besides he canonical clo e lea s uc u e, RNAIle has he po en ial o o m a long hai pin [37]. RNAAsp also adop s an al e na i e olding in o de o bind o he Alu elemen inse ion in he 3’-UTR o he mRNA o i s own aminoacyl- RNA syn he ase [38]. Since nucleo ide modi ica ions a e known o a ec hyb idiza ion, i is emp ing o specula e o wha ex end hey a ec he al e na i e olding o RNAs [39]. Finally, q-PCR analysis o RFs di e en ially exp essed in di e en g ade PCa demons a ed ha he exp ession a io RF-544, de i ed om RNA PheGAA and RF- 315 de i ed om RNALysCTT e ec i ely disc imina es high om low g ade p os a e umo s and cu ed om ecu en disease. This es ablishes RFs as no el candida e bioma ke s o he ea ly de ec ion o ecu en agg essi e PCa. In conclusion, ou s udy p o ides a comp ehensi e ca alogue o RFs exp essed in a ious s ages o PCa and p o ides leads o he u he in es iga ion o biological ole and ma ke po en ial o hese no el RNA en i ies in p os a e cance . MATERIALS AND METHODS Sample coho s and p ocessing The disco e y se used in his s udy consis s o 10 sequencing lib a ies gene a ed as p e iously desc ibed [24]. B ie ly, each lib a y was cons uc ed om a o al RNA pool p epa ed om ou indi idual pa ien samples wi h simila pa hological o gene ic cha ac e is ics [40]. Di e en g oups ep esen : no mal adjacen p os a e issue (NAP), p os a e umo s wi h Gleason sco e 6, 7, o 8 (PCa6, PCa7, PCa8), me as a ic lymph nodes (LN_PCa), all ob ained by adical p os a ec omy; benign p os a e hype plasia issue (BPH) ob ained by cys op os a ec omy; and cas a ion esis an p os a e umo s ob ained by ans- u e h al esec ion o he p os a e (TURP_PCa) [24]. The clinical pa ame e s o each g oup a e summa ized in he Table 1. PCa g oups wi h Gleason sco e 6 we e di ided in o cu ed and ecu en disease g oups o in o g oups wi h o wi hou TMPRSS2-ERG usion o ETV abno mali ies. Sample ma e ial was ob ained om he issue banks o he E asmus Uni e si y Medical Cen e , Ro e dam, The Ne he lands (E asmus MC, Ro e dam, The Ne he lands) and Tampe e Uni e si y Hospi al (TAUH, Tampe e, Finland). Collec ion and use o pa ien ma e ial was pe o med acco ding o he na ional legisla ions conce ning e hical equi emen s and app o ed by he E asmus MC Medical E hics Commi ee acco ding o he Medical Resea ch In ol ing Human Subjec s Ac (MEC- 2004- 261), and he E hical Commi ee o he Tampe e Uni e si y Hospi al. Samples we e snap ozen and s o ed in liquid ni ogen. Gleason sco e and he pe cen age o no mal and cance epi helial cells we e e alua ed om his ological sec ions by wo pa hologis s. Only samples wi h mo e han 70% o umo cells we e used o sequencing lib a y p epa a ion. All samples ha we e used o he no mal p os a e pool con ained 0% o umo cells. To al RNA was ex ac ed using RNABee eagen (Camp o Scien i ic, GmbH, Be lin, Ge many) acco ding o he manu ac u e ’s p o ocol. qPCR alida ion was pe o med in wo sepa a e coho s. Clinical pa ame e s a e p o ided in Supplemen a y Tables 6-7). The i s coho (coho 1) consis s o 65 samples ob ained om E asmus MC. The samples we e collec ed, handled and e alua ed as men ioned in he p e ious pa ag aph. The second coho (coho 2) consis s o 104 ho monally un ea ed p ima y p os a e umo s om adical p os a ec omy specimens