Dynamic signal processing by ribozyme-mediated RNA circuits to control gene expression

Nucleic Acids Resea ch, 2015 1 doi: 10.1093/na /gk 287 Dynamic signal p ocessing by ibozyme-media ed RNA ci cui s o con ol gene exp ession Shensi Shen1,†, Guille mo Rod igo1,†, Sa ya P akash3, Esz e Maje 2, Thomas E. Land ain1, Bo is Ki o 1, Jos´ e-An onio Da ` os2and Al onso Ja amillo1,3,* 1Ins i u e o Sys ems and Syn he ic Biology, Uni e si ´ ed’´ E y-Val-d’Essonne, CNRS, F-91000 ´ E y, F ance, 2Ins i u o de Biolog´ ıa Molecula y Celula de Plan as, CSIC – Uni e sidad Poli ´ ecnica de Valencia, 46022 Valencia, Spain and 3School o Li e Sciences, Uni e si y o Wa wick, Co en y CV4 7AL, UK Recei ed June 19, 2014; Re ised Ma ch 23, 2015; Accep ed Ma ch 24, 2015 ABSTRACT O ganisms ha e di e en ci cui ies ha allow con- e ing signal molecule le els o changes in gene exp ession. An impo an challenge in syn he ic bi- ology in ol es he de no o design o RNA mod- ules enabling dynamic signal p ocessing in li e cells. This equi es a scalable me hodology o sensing, ansmission, and ac ua ion, which could be assembled in o la ge signaling ne wo ks. He e, we p esen a biochemical s a egy o design RNA- media edsignal ansduc ioncascadesable osense small molecules and small RNAs. We design swi ch- able unc ional RNA domains by using s and- displacemen echniques. We expe imen ally cha ac- e ize he molecula mechanism unde lying ou syn- he ic RNA signaling cascades, show he abili y o egula e gene exp ession wi h ansduced RNA sig- nals, and desc ibe he signal p ocessing esponse o ou sys ems o pe iodic o cing in single li e cells. Theenginee edsys emsin eg a eRNA–RNAin e ac- ion wi h a ailable ibozyme and ap ame elemen s, p o iding new ways o enginee a bi a y complex gene ci cui s. INTRODUCTION Na u al signal ansduc ion sys ems allow o ganisms o adap o luc ua ing en i onmen s, o en by exploi ing sub- cellula localiza ion, molecula cascades and p o ein al- los e ici y (1,2). A majo challenge in syn he ic biology in- ol es he enginee ing o no el signaling sys ems ha sense, p ocess and ansmi in o ma ion. Mos enginee ing e o s ha e elied on he ansla ional usion o known p o ein domains wi h speci ic in e ac ion o ca aly ic unc ionali- ies (2). Howe e , his app oach is limi ed by he a ailabil- i y o known na u al in e ac ion domains ha a e speci ic enough o a oid c oss- alk wi h o he molecules in he cellu- la con ex . Al e na i ely, he use o RNA as p og ammable molecules would allow enginee ing an unlimi ed numbe o in e ac ion pa ne s (3,4). This way, we p opose o enginee syn he ic signal ansduc ion sys ems elying on RNA by using a ansc ip ional usion s a egy, exploi ing sequence agmen s wi h de ini e in e ac ion and ca aly ic p ope ies. In p o ein-based signaling, localized olding domains acil- i a e he enginee ing (o e-enginee ing) o mul iple unc- ions (5,6). Simila ly, he e a e well-known RNA olding s uc u es ha a e s able and capable o in e ac speci ically wi h signaling molecules (ap ame s) o o ca alyze eac ions ( ibozymes) (4). In addi ion, he use o compu a ional ools allows he p edic ion o con o ma ional changes in many cases, opening he doo o he enginee ing o signal ans- duc ion sys ems based on RNA (7). As a p oo o concep , we he e de elop a sys em ( o con ol gene exp ession wi h a molecula signal) ha consis s in he usion o an ap azyme, ac ing as a molecula sensing elemen , wi h a ibo egula- o , ac ing as a signal media o . To simpli y he e minol- ogy, in he ollowing we e e o his mul i unc ional RNA molecule as egazyme. In his di ec ion, pionee ing wo k in syn he ic biology inse ed known ap ame domains in o 5un ansla ed e- gions (UTRs) o messenge RNAs (mRNAs) o sense small molecules (10), and also exploi ed ibo egula ion in com- bina ion wi h small-molecule- esponsi e p omo e s o con- ol gene ne wo ks and me abolic pa hways (8,9). Mo e e- cen ly, impo an s eps owa ds RNA-based sensing ha e been ca ied ou by enginee ing ap azymes in he 5o 3UTRs o sense bo h small molecules (11,12) and small RNAs (sRNAs) (13). Mo eo e , p e ious wo k has com- bined ap ame s wi h ibo egula o s o c ea e no el sens- ing de ices (13–15). Those wo ks exploi he p og amma- bili y o RNA unc ion h ough s and-displacemen e- ac ions and induced con o ma ional changes. He e, ou s a egy allows enginee ing a one- o- wo-componen sig- nal ansduc ion sys em, whe e eme ging RNA unc ion is *To whom co espondence should be add essed. Tel: +44 24 765 73432; Email: Al onso.J[email p o ec ed] †These au ho s con ibu ed equally o he pape as i s au ho s. C The Au ho (s) 2015. Published by Ox o d Uni e si y P ess on behal o Nucleic Acids Resea ch. This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed euse, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. Nucleic Acids Resea ch Ad ance Access published Ap il 27, 2015 a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om 2Nucleic Acids Resea ch, 2015 achie ed by inco po a ing sel -clea age abili y in o a ans- ac ing ibo egula o . This co esponds o a ou -molecule sys em, whe e he i s one is he signal molecule, ei he a small molecule o sRNA, and he las one is a cis- egula ed mRNA as sys em’s eadou . The o he wo molecules ( wo componen s) co espond o he senso and media o , which can be swi ched ON/OFF in p esence/absence o he signal molecule, espec i ely. The de ised sys em sha es p ope ies wi h na u al signal- ing sys ems (1). On he one hand, i is a one-componen sys- em om he inpu iewpoin . Thus, i has he ad an age o subcellula localiza ion independence. On he o he hand, i is a wo-componen sys em om he ou pu pe spec i e ( he senso and media o a e di e en molecules a e clea age). Thanks o he modula i y o e ed by he independence o he senso and media o domains, we could ha e a pale e o domains wi h al e na i e unc ionali ies. Compa ed o endogenous senso s (e.g. ecep o s), ou senso is no lim- i ed o he cell memb ane, meanwhile he media o (i.e., i- bo egula o ) wo ks like a phospho yla ed ansc ip ion ac- o bu a pos - ansc ip ional le el (in e ac ing wi h a 5 UTR a he han wi h a p omo e ). Ou app oach o en- ginee a signal ansduc ion sys em combines he usion o unc ional RNA elemen s oge he wi h he compu a ional p edic ion o each con o ma ional s a e. This is also possi- ble wi h a p o ein-based sys em (16,17), al hough i could become much ha de , equi ing expe imen al sc eening o- wa ds app op ia e ansduc ion p ope ies (5). In his wo k, we show ha RNA s uc u e is p edic able enough o allow a compu a ional design s a egy. In gen- e al, he emendous size o he sys em’s sequence space p e- en s he de no o design wi hou au oma ion. We ha e p e i- ously demons a ed ha an au oma ed design me hodology is able o gene a e de no o ibo egula ion in li e cells (18). The e o e, we he e p opose o gene alize such me hodology o design RNA-media ed signal ansduc ion sys ems. Fo ha , we assume ha any in e ac ion be ween wo RNAs is igge ed by a seed (o oehold) sequence (18). In he case o a egazyme, he signal molecule induces a ca aly ic p o- cess ha eleases a ibo egula o , which in u n induces a con o ma ional change in he 5UTR ha ini ia es in e ac- ion wi h he 16S ibosomal uni (18,19)inEsche ichia coli. This way, we en o ce a hie a chical mode o ac ion consis - ing in swi ching ON each unc ional module, which is ini- ially OFF. In he ollowing, we will p o ide a de ailed desc ip ion o he compu a ional me hodology o design a hie a chical sys em wi h unc ional RNA modules ha couples molec- ula signals in he cell (ei he om he en i onmen o om ups eam biological sys ems) wi h gene exp ession. We will i s desc ibe he de elopmen o a me hodology o nu- cleo ide sequence design, and hen we will p esen a mech- anis ic cha ac e iza ion o assess he sel -clea age ac i i y o he egazyme. Subsequen ly, we will show esul s assess- ing signal ansduc ion wi h ime-dependen induc ion in bac e ial cells, allowing he cha ac e iza ion o he dynamic egula o y p ope ies a bo h popula ion and single cell le - els. MATERIALS AND METHODS Sequence design We de eloped a Mon e Ca lo Simula ed Annealing (20) op- imiza ion algo i hm o design he ansduce modules o egazymes p o ided he sequences o gi en ap azymes (o sRNA-induced ibozymes) and ibo egula o s (Supplemen- a y Figu e S6). Fo ha , we cons uc ed a basic ene gy model ha in ol ed h ee a iables ( o be minimized): he ene gy o ac i a ion co esponding o he ca aly ic ac i i y o he ap azyme, he deg ee o accessibili y o he ibo egu- la o seed be o e clea age, and he deg ee o obs uc ion o he seed a e clea age. The exposu e o obs uc ion o he ibo egula o seed is go e ned by seconda y s uc u e, bu he ap azyme in ol es e ia y con ac s. We he e simpli ied he p oblem by only conside ing he seconda y s uc u e o he ap ame o calcula e he ene gy o ac i a ion o clea - age. Rounds o andom mu a ions ( eplacemen s, addi ions o dele ions) we e applied and selec ed wi h he ene gy- based objec i e unc ion. We used he Vienna RNA package (21) o ene gy and s uc u e calcula ion (see u he de ails in Supplemen a y Ma e ials and Me hods). The sequences o he enginee ed egazymes in his wo k a e shown in Sup- plemen a y Tables S1–S3. Plasmid cons uc ion The di e en RNA de ices we e chemically syn hesized and cloned in plasmid pSyn h (pUC eplica ion o igin, ampi- cillin esis ance) and hen subcloned in o plasmids pSTC1 o pSTC2. These wo plasmids con ain a pSC101m epli- ca ion o igin (a mu a ed pSC101 o i gi ing a high copy numbe ) and a kanamycin esis ance ma ke (Supplemen- a y Figu es S1 and S2). The pSTC2 ec o is based on ou p e iously epo ed ec o pSTC1 (18) by emo ing he mRFP coding sequence and agging he ca boxyl e minus o he supe olde GFP (s GFP) (22) wi h he ss Adeg a- da ion ag (23). Dys unc ional egazymes we e cons uc ed by PCR-based si e-di ec ed mu agenesis (see Supplemen- a y Ma e ials and Me hods). S ains and plasmids used in his s udy a e lis ed in Supplemen a y Table S6. In acellula ca aly ic ac i i y P ocessing ex en o egazyme a di e en ime poin s (0, 2, 4, 8, 16 and 32 min) was analyzed by no he n blo hyb idiza ion using a complemen a y [32P]-labeled RNA p obe a e sepa a ing he di e en RNA samples by de- na u ing polyac ylamide gel elec opho esis (PAGE). RNA p epa a ions we e mixed wi h o mamide loading bu e o dena u a ion, ollowed by PAGE sepa a ion in 5% poly- ac ylamide gels including 8 M u ea and TBE bu e . Gels we e s ained wi h e hidium b omide. Memb anes we e hy- b idized o e nigh , imaged by au o adiog aphy, and hen hyb idiza ion signals quan i ied by phospho ime y (Fuji- ilm FLA-5100). See mo e de ails in Supplemen a y Ma e- ials and Me hods. Fluo escence quan i ica ion Cells we e g own o e nigh in LB medium, and we e e- eshed in cul u e ubes wi h LB medium in o de o each a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om Nucleic Acids Resea ch, 2015 3 s a iona y phase. Cells we e hen dilu ed 1:200 in 200 ␮l o M9 minimal medium in each well o he pla e (Cus om Co ning Cos a ). The pla e was incuba ed in an In ini e F500 mul i-well luo ome e (TECAN) a 37◦C wi h shak- ing. I was assayed wi h an au oma ic epea ing p o ocol o abso bance measu emen s (600 nm abso bance il e ) and luo escence measu emen s (480/20 nm exci a ion il e – 530/25 nm emission il e o s GFP) e e y 15 min. All sam- ples we e p esen in iplica e on he pla e (see u he de ails in Supplemen a y Ma e ials and Me hods). Single cell mic o luidics analysis The design o ou mic o luidics de ice (Supplemen a y Fig- u e S16) (24), which was pe o med in AUTOCAD (AU- TODESK), was adap ed om he p e ious one epo ed by Has y e al. (25). All images we e acqui ed using Zeiss Axio Obse e Z1 mic oscopy (Zeiss). The mic oscope esolu ion was 0.24 ␮m wi h Op o a ia ion 1.6×, esul ing o al mag- ni ica ion 1600× o bo h b igh ield and luo escen im- ages. Images we e analyzed wi h MATLAB (Ma hWo ks). Cells we e acked by de ining a cell- o-cell dis ance ma ix and he cell lineages we e econs uc ed. Finally, he luo- escence le el o each cell in each luo escence ame was ex- ac ed (see u he de ails in Supplemen a y Ma e ials and Me hods). RESULTS Compu a ional design o RNA-media ed signal ansduc ion sys ems o con ol gene exp ession Ou modula s a egy consis s in designing swi chable unc- ional RNA domains, which is implemen ed by exploi ing s and-displacemen p inciples oge he wi h he enginee - ing o allos e ic con o ma ional s a es. This way, we can de- sign chains o se e al domains ha a e ac i a ed in cascade. Wi hou loss o gene ali y, we conside ed a sys em com- posed o wo ansc ip ional uni s: egazyme and mRNA o a epo e gene (e.g. a gene coding o a g een luo es- cen p o ein −GFP), bu ou me hodology could be gene - alized o an a bi a y numbe o ansc ip ional uni s con- aining swi chable unc ional elemen s. To enginee such a syn he ic RNA sys em implemen ing he ansduc ion o molecula signals in o changes in gene exp ession, we ook ad an age o a s anda d physicochemical model (based on Wa son-C ick and wobble pai ing) p edic ing RNA sec- onda y s uc u e and ee ene gy (26) o be used in an op i- miza ion algo i hm o selec o he hie a chical ac i a ion o unc ional RNA modules in he cascade (7). In pa icula , ou sys em co esponds o a cascade o h ee modules: senso (ap azyme designed o speci ically e- spond o a gi en ligand), media o ( ibo egula o designed o speci ically ac i a e a cis- ep essed ibosome-binding si e −RBS), and ac ua o (mRNA wi h cis- ep essed RBS) (Figu e 1a). To c ea e he egazyme, we used an ap azyme elemen , ac ing as a molecula sensing de ice, wi h a i- bo egula o , ac ing as a signal media o , in o he same an- sc ip ional uni . This usion is pe o med wi h lanking se- quences ha o m a s em and unc ion as a ansduce mod- ule, in he same way as when designing allos e ic ap ame s (27). The senso domain (ap azyme) is ini ially in a s a e OFF (ca aly ically inac i e) and is swi ched o ON (ca aly - ically ac i e) only when i acqui es i s unc ional con o ma- ion, which is induced by he signal molecule. The media o domain ( ibo egula o ) will be in a s a e ON when i s seed sequence is exposed o he sol en . We designed he ans- duce module o ensu e ha he senso and media o we e ON/OFF in p esence/absence o he signal molecule. This p e en s any p ema u e elease o he media o o any di ec ac i a ion o gene exp ession by he egazyme. A e wa d, he inpu signal p oduces a s abiliza ion o an al e na i e con o ma ion whe e he ap azyme is ac i e. Once he ap- azyme is ac i e, i will sel -clea e eleasing he media o do- main, which is hen swi ched on (i.e. he ibo egula o seed sequence becomes exposed). Once he media o domain is ac i e, i will di use owa ds i s a ge genes (in pa icula , o in e ac wi h 5UTRs), simila ly o phospho yla ed an- sc ip ion ac o s in he con en ional wo-componen sys- ems (1). To be no ed, he independence be ween he sen- so and media o domains a o s expanding he unc ional epe oi e, which allows hem o be exchanged wi h al e na- i e domains. We he e p opose a new me hodology o enginee one- o- wo-componen signal ansduc ion, which combines he ad an ages o subcellula independence o one-componen sys ems and o modula i y o wo-componen sys ems. We cons uc ed a combina o ial op imiza ion p oblem o ex- plo e he sequence space o he ansduce module (Fig- u e 1b), whe e a nucleo ide-le el ene gy model conside - ing he con o ma ional s a es (unclea ed and clea ed) o he egazyme was used o e alua e he pe o mance o he gene a ed sequences. Fo each s a e, he model accoun s o i s ee ene gy and i s seconda y s uc u e. As objec i es o be op imized (compu ed as Hamming dis ances), he algo- i hm conside s he ene gy o ac i a ion co esponding o he ca aly ic ac i i y o he ap azyme (which we assume de- pends on he co ec o ma ion o he ap ame in he un- clea ed s a e), and he deg ee o exposu e o he sol en o he ibo egula o seed be o e and a e clea age (28). Figu e 1c illus a es he ene gy landscape associa ed o he molecula mechanism o he egazyme, epo ing he di e - en con o ma ional s a es and hei co esponding ee en- e gy le els (see also Supplemen a y Figu e S4). The eac ion coo dina e was de ined he e as he numbe o in e molecu- la hyd ogen bonds, on one side, be ween he ligand and he ap azyme and, on he o he side, be ween he ibo egula o and he 5UTR (in e ms o base-pai s). In absence o sig- nal molecule, he p og ession o he eac ion is limi ed by he p esence o a high-ene gy in e media e ha p e en s he in e ac ion be ween he egazyme and he 5UTR. How- e e , when he signal molecule is a su icien concen a ion, a clea age is p oduced and hen he ac i a ion ene gy o he esul ing ibo egula o y elemen is lowe ed, which speeds up he eac ion (29). As shown by a andom sampling o 1000 sequences (Fig- u e 1d), an op imal sco e (ze o, as ou sco e is conside ed as a penal y) is e y unlikely o be ob ained a bi a ily. This means ha his is a di icul design p oblem o a manual app oach, equi ing au oma ed compu a ion o e icien sequence design. Ou algo i hm designs by op imiza ion he sequences implemen ing he in ended signal ansduc- ion acco ding o he objec i e unc ion. E en hough dis- a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om 4Nucleic Acids Resea ch, 2015 Figu e 1. Compu a ional design o he egazyme signaling pa hway. (a) Scheme o he modula sys em, whe e a signal molecule (ei he a small molecule o a sRNA) induces a ca aly ic eac ion ha eleases a ibo egula o able o ac i a e gene exp ession. Each signaling pa hway is iewed as a wi e ca ying in o ma ion. (b) Scheme o he op imiza ion loop, whe e a egazyme sequence is i e a i ely mu a ed and e alua ed acco ding o an objec i e unc ion. (c) Ene gy landscape o he signaling pa hway showing he di e en con o ma ional s a es (in a- and in e molecula ), oge he wi h he h ee ee ene gy e ms o he objec i e unc ion, in e ms o a eac ion coo dina e. Solid line illus a es he ajec o y co esponding o he ligand-induced clea age o he egazyme and subsequen binding o he ibo egula o o he mRNA. Dashed line co esponds o he ajec o y whe e he unclea ed egazyme binds o he mRNA. (d) Compu a ion o he ee ene gy e ms o he objec i e unc ion o 1,000 andom sequences o e alua e hei dis ibu ion. (e) Alignmen o di e en op imized sequences wi h ap azyme heoHHAz and ibo egula o RAJ12. Highly conse ed nucleo ides a e highligh ed in ed o blue, and he consensus sequence is shown. inc solu ions can be equally good compu a ionally (i.e., acco ding o he objec i e unc ion), expe imen s could dis- ill di e ences in pe o mance among hem. We obse ed, o he sampled sequences, ha he ap ame (in he un- clea ed s a e) is co ec ly o med only o a small subse o sequences (Supplemen a y Figu e S5). Mo eo e , he esul - ing dis ibu ion is appa en ly bimodal (Sa le’s bimodali y coe icien BC =0.630 >5/9) (30), which may be explained by an all-o -none o ma ion o he unc ional s uc u e o he ap azyme. Such p e-o ganized con o ma ions will a- o ligand binding and subsequen clea age, whe eas s uc- u es equi ing conside able ea angemen s will be o side due o a gi en ee ene gy ba ie (29). The dis ibu ion o sco e alues along he axis ep esen ing he seed exposu e in he unclea ed s a e is mo e homogeneous (BC =0.467 <5/9), whe eas he dis ibu ion in he clea ed s a e shows subs an ial he e ogenei y (BC =0.639 >5/9). This may be explained by an in e ac ion o he seed egion wi h pa o he 5end a e clea age (see, o example, Supplemen a y Figu e S8). Modula i y in he design o egazymes In his wo k, we conside ed h ee possible senso domains, wo sensing a small molecule ( heophylline −Theo − a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om Nucleic Acids Resea ch, 2015 5 and hiamine py ophospha e −TPP, Figu e 2and Supple- men a y Figu e S7), and ano he sensing a speci ic sRNA (B eak1, Figu e 3). This sRNA is induced wi h anhyd o e- acycline (aTc) in ou sys em. Mo e speci ically, each sen- so is composed o a binding domain (e.g., an ap ame ) and a ca aly ic domain (e.g. a hamme head ibozyme). Ou ligand-induced ibozymes (ap azymes) a e heoHHAz and ppHHAz o sensing small molecules (11,31), and b eakHHRz o sensing sRNA (32) (Supplemen a y Fig- u e S3). Fo he media o domain, we conside ed h ee syn he ic ibo egula o s known o ac i a e he ini ia ion o ansla ion, wo enginee ed in Rod igo e al. (RAJ11 and RAJ12) (18) and one in Isaacs e al. (RR12) (19) (Supplemen a y Figu e S8). We hen designed he es o he egazyme sequence acco ding o he speci ica ions e- qui ed o gene a e he RNA signaling cascade. Exploi ing he modula i y o his sys em, we enginee ed he ollowing egazymes: heoHHAzRAJ11, heoHHAzRAJ12, heoH- HAzRR12, ppHHAzRAJ12 and b eakHHAzRAJ12. The egazyme p oduces a media o molecule ( ibo egula o ) ha is independen o he signal and senso molecules. In he ollowing, we in es iga e, on he one hand, how di e en signal molecules (Theo, TPP and B eak1) ac i a e a com- mon media o (RAJ12), and, on he o he hand, how di e - en implemen a ions o he wi e (RAJ11, RAJ12 and RR12) ansduce he in o ma ion om a common signal molecule (Theo). Ou compu a ional app oach allowed us o in es iga e he designabili y (de ined as he numbe o sequences ha ha e he desi ed biochemical unc ion) o he solu ion space o a pa icula couple o ap azyme and ibo egula o . We expec , ne e heless, a highe designabili y when sequences a e allowed o a y in leng h, as i is he case o ou algo- i hm. I is ins uc i e o align mul iple solu ions o e eal conse ed nucleo ide posi ions. Figu e 1e shows, o di e - en designs, he consensus sequence o a gi en choice (ap- azyme heoHHAz and ibo egula o RAJ12). Sequences (co esponding o he 5and 3 egions o he ap azyme, see Supplemen a y Figu e S6a) we e aligned by using he an i- seed consensus sequence (CYC in his case; no e ha he seed sequence is GGG) as e e ence. In addi ion, his mod- ula i y would allow a hie a chical design o he egazyme molecule. To c ea e a sui able pipeline, we can exploi com- pu a ional algo i hms o: (i) design he binding domain (ap- ame in pa icula ) o a speci ic signal molecule (33), (ii) design he ibo egula o and cogna e 5UTR (18), and (iii) apply he me hodology de eloped in his wo k o design he app op ia e ansduce module. Expe imen al sc eenings o di ec ed e olu ion echniques (34) could also be applied, es- pecially o link he binding and ca aly ic domains (see ’Dis- cussion’ sec ion). Molecula cha ac e iza ion o RNA-media ed signal ans- duc ion To analyze he mechanism o he signaling pa hway, we i s ca ied ou a kine ic and dose-dependen s udy o he ca aly ic ac i i y. The p edic ed seconda y s uc u e o he small-molecule-sensing egazyme heoHHAzRAJ12 in he unclea ed s a e (Figu e 2a) shows, as designed, ha he ap- ame is al eady a anged o Theo sensing, and ha he seed egion o he ibo egula o is blocked by he ansduce module. A e clea age a he CC dinucleo ide si e be ween he ansduce module and he ibozyme co e (Figu e 2a, ma ked by an a ow), he seed egion is eleased, which al- lows he ibo egula o o in e ac downs eam wi h he cog- na e 5UTR o he epo e gene. An analogous seed-based s uc u al mechanism is illus a ed o he sRNA-sensing egazyme b eakHHRzRAJ12 (Figu e 3a). In his case, he binding domain is only pa ially pai ed o allow an e icien in e ac ion wi h he signal sRNA (B eak1), and he clea - age is done a he GA dinucleo ide si e (Figu e 3a, ma ked by an a ow). Indeed, he e is a seed-media ed in e ac ion be ween B eak1 and he egazyme, simila o he in e ac- ion be ween he ibo egula o and he 5UTR. O no e, ou egazyme b eakHHRzRAJ12 implemen s o he i s ime an RNA cascade in li e cells (independen o any p o ein- based machine y). To moni o he dynamic RNA p ocessing o he sys- em, we pe o med a gel assay om cellula RNA ex- ac s. Cells exp essing egazyme heoHHAzRAJ12 o b eakHHRzRAJ12 we e induced wi h di e en concen a- ions o Theo o aTc and lysed a se e al ime poin s. The gel assays in bo h cases showed as dynamic RNA p ocess- ing, eaching s eady s a es in almos 16 min. The obse ed clea age a e ( i ed wi h a model o exponen ial decay wi h p oduc ion) is 0.15 min−1 o heoHHAzRAJ12 wi h 4 mM Theo, al hough he model does no cap u e inely he expe - imen al end (Figu e 2b and Supplemen a y Figu e S9a). In his case, he band co esponding o he eleased ibo eg- ula o (o 114 n , accoun ing o he e mina o ) mig a es as e han expec ed. This band was no obse ed wi hou Theo, indica ing ha indeed i is a p oduc o he clea - age eac ion. Mo eo e , we no e ha ou p obe did no de- ec he 5 agmen a e clea age, sugges ing a as deg a- da ion o his new species. In hose condi ions, he maxi- mal clea age ac ion is ∼80%, which is >2.5- old inc ease wi h espec o he basal s a e (∼30%). Wi h 0.4 mM Theo, he obse ed clea age a e is 1.5 min−1wi h a mo e accu- a e i ing (no e ha he disc epancy be ween he a es a 4 and 0.4 mM is indeed due o he model i ing; a 4 min he ac ion clea ed is ∼60% in bo h cases). Acco ding o p e ious expe imen al esul s in i o wi hou RNA p o- duc ion and deg ada ion (11), he obse ed clea age a es o heoHHAz in absence and p esence (4 mM) o Theo a e 1.3 and 3.6 min−1, espec i ely, wi h a maximal clea - age ac ion o 90%. Ce ainly, he dynamic esponse in i o aces addi ional challenges due o he balance be ween p o- duc ion and deg ada ion. Fo b eakHHRzRAJ12, he ob- se ed clea age a e is 0.17 min−1wi h 100 ng/ml aTc, bu no ac i i y is epo ed o lowe concen a ions o his in- duce (Figu e 3b and Supplemen a y Figu e S9b). He e, he band co esponding o he eleased ibo egula o (o 112 n , also accoun ing o he e mina o ) mig a es slowe han expec ed, al hough i was no obse ed (as be o e) wi h- ou aTc, indica ing ha indeed i is a p oduc o he clea - age eac ion. These anomalous mig a ions could be due o a di e ence in expec ed leng h (e.g. unp edic ed an- sc ip ion e mina ion, as mos o he clea age occu s co- ansc ip ionally) o o a esidual s uc u e in he eleased ibo egula o e en a e using 8 M u ea in he gel, among o he possibili ies. In case o b eakHHRzRAJ12, p e ious a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om 6Nucleic Acids Resea ch, 2015 Figu e 2. Molecula cha ac e iza ion o small-molecule-sensing egazyme. (a) Sequence and s uc u e o he egazyme heoHHAzRAJ12. A small molecule (Theo) binds o he egazyme o econs i u e he ac i e con o ma ion o he ibozyme and hen p oduce he clea age. An a ow ma ks he clea age si e, be ween he ansduce module and he ibozyme co e. The seed o he ibo egula o is pai ed in he unclea ed s a e. (b) Time-dependen elec opho e ic analysis o cellula RNA ex ac s aken a di e en ime poin s; gel shown o 4 mM Theo. Quan i ica ion o dynamic RNA p ocessing o di e en concen a ions o he signal molecule (Theo). Da a i ed wi h a gene alized exponen ial decay model wi h p oduc ion, whe e he empo al ac o is (1 − exp(−λ ))m, wi h m≈1. E o ba s ep esen s anda d de ia ions o e eplica es. assays in i o (32) e ealed a a e o 0.11 min−1wi h single- s anded DNA as ligand (3 ␮M). The maximal clea age ac ion epo ed he e is almos 25% (∼1% o he basal s a e), which shows a big disc epancy wi h hose p e ious in i o esul s (∼75%). One possible explana ion is ha he exp ession o he sRNA wi h 100 ng/ml aTc does no sa - u a e he sys em, because he egazyme is exp essed om a s ong cons i u i e p omo e , and also because o a high e ec i e dissocia ion cons an . O ele ance, his egazyme has much lowe leakage (1% e sus 30%, al hough main- aining simila old-changes), which could be impo an in case o sensi i e sys ems. Mo eo e , we obse ed highe he - e ogenei y in he dynamic esponse ( om cell o cell) o his sRNA-sensing egazyme, which could be a esul o a he e ogeneous exp ession o he sRNA o e en o a ce ain he e ogeneous sRNA– egazyme in e ac ion. As a esul , by p edic ing RNA s uc u es and quan i ying cellula RNA ex ac s, we ha e shown he p ecise signal sensing and sub- sequen clea age o elease a unc ional ibo egula o . To u he con i m ha ou de ices beha e as ex- pec ed, we pe o med in i o ansc ip ion o sys ems heo- HHAzRAJ12 and b eakHHRzRAJ12. The expe imen s showed simila clea age ac ions (wi h espec o he in i o assays) a e 30 min o eac ion (Supplemen a y Fig- u e S10a). Fo heoHHAzRAJ12, 70% o he molecules we e clea ed in i o, whe eas 80% we e in i o.Fo b eakHHRzRAJ12, 25% o he molecules we e clea ed bo h in i o and in i o. We also obse ed ha heoH- HAzRAJ12 was clea ed in highe ex en in absence o lig- and. Because in i o we can neglec deg ada ion, he clea - age ac ion is expec ed o inc ease wi h ime, in p esence o ligand and also in absence o i due o he basal ac i i y o he ibozyme. We u he pe o med a ime-cou se assay o s udy he clea age o he egazymes. As shown in ou expe - imen al esul s, he ac ion o clea ed p oduc s o heoH- HAzRAJ12 in es ubes was only ∼10% la ge when Theo was p esen (Supplemen a y Figu e S10b). Howe e , when he same RNA was moni o ed in bac e ial cells, he appa - en le el o induc ion o clea age ac i i y by heophylline was >2.5- old. The di e ence in clea age in i o in case o b eakHHRzRAJ12 wi h espec o he p esence o no o B eak1 (in oduced as DNA oligo) was mo e ema kable (Supplemen a y Figu e S10c). The e is ce ain numbe o easons why an RNA migh exhibi di e en beha io s in a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om Nucleic Acids Resea ch, 2015 7 Figu e 3. Molecula cha ac e iza ion o sRNA-sensing egazyme. (a) Sequence and s uc u e o he egazyme b eakHHRzRAJ12. A sRNA binds o he egazyme o econs i u e he ac i e con o ma ion o he ibozyme and hen p oduce he clea age. An a ow ma ks he clea age si e, be ween he ansduce module and he ibozyme co e. The seed o he ibo egula o is pai ed in he unclea ed s a e. (b) Time-dependen elec opho e ic analysis o cellula RNA ex ac s aken a di e en ime poin s; gel shown o 100 ng/ml aTc. Quan i ica ion o dynamic RNA p ocessing o di e en concen a ions o he signal molecule (aTc). Da a i ed wi h a gene alized exponen ial decay model wi h p oduc ion, whe e he empo al ac o is (1 −exp(-λ ))m, wi h m≈2. E o ba s ep esen s anda d de ia ions o e eplica es. i o han in i o. Fo example, in i o he egazyme migh exis in he modynamic equilib ium wi h i s ligand, esul - ing in a di e en e ec i e dissocia ion a e, o di e sligh ly in leng h om he s ands in i o. In addi ion, in i o we used T3 polyme ase o ansc ip ion (wi hou e mina o s) ins ead o Esche ichia coli polyme ase and a highe Mg2+ concen a ion han in i o, which migh esul in di e ences in olding and clea age kine ics o he egazyme. In he ol- lowing, we p esen he ne e ec o egazyme clea age and ibo egula o elease on GFP exp ession in i o (bo h a he popula ion and single cell le els) wi h and wi hou he lig- and, showing a egula o y beha io as designed. Regula ion o gene exp ession in li e cells wi h ansduced RNA signal To cha ac e ize he dynamic ange o ou enginee ed sys- ems, we placed he ansc ip ional uni s co esponding o he egazyme and mRNA o he GFP epo e gene un- de he con ol o unable p omo e s (35). These p omo e s can be induced wi h isop opyl-␤-D- hiogalac opy anoside (IPTG) and aTc in E. coli cells exp essing cons i u i ely he ep esso s LacI and Te R. Thus, ou sys ems imple- men mul i-inpu AND logic ci cui s (Figu e 4a). Mo e- o e , a con ol sys em was implemen ed by using a dys- unc ional mu a ed egazyme (Figu e 4b). In he implemen- a ion o small-molecule signaling, aTc and IPTG con- ol he exp ession o he egazyme and he mRNA, and Theo is he signal molecule ha induces he clea age o he egazyme o elease he ibo egula o . Figu e 4cshows he luo escence esul s o he sys em based on egazyme heoHHAzRAJ12 o all possible combina ions o induc- e s (IPTG, aTc and Theo). The obse ed weak ac i a ion o luo escence in absence o he signal molecule, bu in p esense o IPTG and aTc, can be explained by he leak- age o sel -clea age o he egazyme (see also Supplemen- a y Figu e S11). The dys unc ional egazyme, ob ained by a wo-nucleo ide mu a ion in he ap ame domain ha abolishes ligand binding (Supplemen a y Figu e S7a), was shown o signi ican ly dec ease GFP exp ession (Figu e 4c). Fu he mo e, a single-nucleo ide mu a ion in he ibozyme ca aly ic co e ( heoHHAzRAJ12AGm, A o G in Supple- men a y Figu e S7a) (46) ha inhibi s he sel -clea age ac- i i y showed dec eased GFP exp ession (Supplemen a y Figu e S11c). An addi ional inac i a ing poin mu a ion ( heoHHAzRAJ12Cm and heoHHAzRR12Cm, U o G a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om 8Nucleic Acids Resea ch, 2015 Figu e 4. Func ional cha ac e iza ion o small-molecule-sensing egazymes. (a,b) Schemes o he enginee ed RNA-based ci cui o sense a small molecule and i s co esponding con ol. (c,d) Digi al scheme, associa ed T u h able, and luo escence esul s o egazymes heoHHAzRAJ12 and heoHHAzRR12 (g ay ba s), and o hei dys unc ional mu an s (whi e ba s) o all possible combina ions o induce s. E o ba s ep esen s anda d de ia ions o e eplica es. in Supplemen a y Figu e S7a) (11) also e ealed dec eased GFP exp ession wi h espec o he na i e sequence (Sup- plemen a y Figu e S11a,b). We also enginee ed and cha ac- e ized sys ems based on ibo egula o s RR12 (Figu e 4d) and RAJ11 (Supplemen a y Figu e S11d), al hough he i- bo egula o y ac i i y o heoHHAzRAJ11 wi h espec o i s dys unc ional mu an was mo e mode a e. Fu he mo e, hese h ee egazymes o Theo-signaling ha e esponsi e- ness in a dose-dependen manne (Supplemen a y Figu e S14) wi h an e ec i e dissocia ion cons an o abou 1 mM. Ano he enginee ed sys em o TPP signaling ( egazyme p- pHHAzRAJ12) showed no signi ican ibo egula o y ac i - i y (Supplemen a y Figu e S15). In he implemen a ion o sRNA signaling, aTc and IPTG con ol he exp ession o he sRNA wo king as sig- nal molecule and he mRNA, whe eas he egazyme is ex- p essed om a s ong cons i u i e p omo e (Figu e 5a). In his case, he clea age o he egazyme is induced by ha sRNA. Figu e 5c shows he luo escence esul s o he sys- em based on egazyme b eakHHRzRAJ12 o all possi- ble combina ions o induce s (IPTG and aTc). This logic ci cui could u he be expanded o in eg a e mo e inpu s by eplacing he cons i u i e p omo e o he egazyme o a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om Nucleic Acids Resea ch, 2015 9 Figu e 5. Func ional cha ac e iza ion o sRNA-sensing egazyme. (a,b) Schemes o he enginee ed RNA-based ci cui o sense a sRNA and i s co e- sponding con ol. (c) Digi al scheme, associa ed T u h able, and luo escence esul s o egazyme b eakHHRzRAJ12 (g ay ba s), and o i s dys unc ional mu an (whi e ba s) o all possible combina ions o induce s. E o ba s ep esen s anda d de ia ions o e eplica es. o he unable p omo e . To exclude he possibili y ha he sRNA B eak1 could di ec ly ac i a e he cis- ep essed e- po e gene, we gene a ed a con ol sys em (b eakRAJ12) by using a dys unc ional mu an emo ing he egazyme el- emen (Figu e 5b and Supplemen a y Figu e S13), e eal- ing no signi ican ibo egula o y ac i i y in his case (Figu e 5c). All oge he , hese esul s demons a e he modula i y o ou designs: (i) he same senso module can ansduce he signal o di e en media o s ( ibo egula o s), and (ii) he same ibo egula o (RAJ12 in his case) can be coupled wi h di e en senso modules. We also analyzed expe imen ally he o hogonali y be ween egazymes. To his end, we con- s uc ed new gene ic sys ems based on non-cogna e pai s (be ween ibo egula o s and 5UTRs), in o de o es in i o he e en ual c oss- alk in egula ion o gene exp es- sion (36). Compu a ional p edic ions showed no in e e - ence be ween he ibo egula o s RAJ11, RAJ12 and RR12 (Supplemen a y Figu e S22a), and p e ious expe imen al wo k e ealed no appa en ac i a ion o sRNA o sys em RAJ11 on he 5UTR o sys em RAJ12 (18). Howe e , as shown in Supplemen a y Figu e S22b, signaling c oss- alk h ough egazymes can appea (e.g. be ween RAJ12 and RR12 ibo egula o y sys ems), p obably, as a consequence o non-Wa son-C ick pai ing no co e ed in he physico- chemical model. A u he compu a ional design me hod- ology will accoun o RNA 3D models o be e p edic RNA-RNA in e ac ion (7) and hen enginee RNA ci cui s wi h mul iple wi es. Fo Ndi e en senso modules and Mo hogonal ibo egula o s, we could gene a e, in heo y, NM egazymes. Impo an ly, as he ou pu o one egazyme can be he inpu o ano he egazyme, we could ha e a mos (NM)Pdi e en implemen a ions o ci cui s wi h P egazymes, including cascades and eedback loops (Supple- men a y Figu e S23). Time-dependen RNA-media ed signal ansduc ion in single cells To cha ac e ize he dynamic esponse o he designed egazymes a he single cell le el, we cons uc ed mic o lu- idics de ices acco ding o p e ious wo k (36,37). The e, sin- gle cells we e moni o ed du ing dozens o cell di isions, us- ing app op ia e de ice geome ies o main ain a single laye o cells wi hin he mic oscope ocal plane and a con inu- ous cell g ow h in exponen ial phase (Figu e 6a, Supple- men a y Figu e S16). Bac e ial cells exp essing he designed egazymes we e loaded in o he de ice, and he composi- a UPVA on Ap il 28, 2015h p://na .ox o djou nals.o g/Downloaded om