RNA Folding and Hydrolysis Terms Explain ATP Independence of RNA Interference in Human Systems

Size: px

Start display at page:

Download "RNA Folding and Hydrolysis Terms Explain ATP Independence of RNA Interference in Human Systems"

Brittany Mathews
5 years ago
Views:

1 RNA Folding and Hydrolysis Terms Explain ATP Independence of RNA Interference in Hman Systems The Harvard commnity has made this article openly available. Please share how this access benefits yo. Yor story matters. Citation Pblished Version Accessed Citable Link Terms of Use Ali, Nicole, and Vinothan N. Manoharan. 29. RNA folding and hydrolysis terms explain ATP-independence of RNA interference in hman systems. Oligoncleotides 19(4): doi:1.189/oli Agst 23, 218 2:25: AM EDT This article was downloaded from Harvard University's DASH repository, and is made available nder the terms and conditions applicable to Open Access Policy Articles, as set forth at (Article begins on next page)

2 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli RNA folding and hydrolysis terms explain ATP-independence of RNA interference in hman systems Nicole Ali 1, Vinothan N. Manoharan 1,2* Abstract Althogh RNA interference (RNAi) has emerged as an important tool for stdying the effects of gene knockdown, it is still difficlt to predict the sccess of RNAi effectors in hman systems. By examining the basic thermodynamic eqations for RNA interactions in RNAi, we demonstrate how the free energies of RNA folding and phosphoester bond hydrolysis can drive RNAi withot ATP. Or calclations of RNAi efficiency are close to actal vales obtained from in vitro experimental data from two previos stdies, for both silencing complex formation (2.5 vs. 2.4 for relative efficiency of RISC formation) and mrna cleavage (.5 vs.56 for proportion cleaved). Or calclations are also in agreement with previos observations that dplex nwinding and target site folding are major energy barriers to RNAi. 1 Department of Physics, Harvard University, Cambridge MA USA Present address: Yale School of Medicine, New Haven, CT 2 School of Engineering and Applied Sciences, Harvard University, Cambridge MA USA * To whom correspondence shold be addressed: vnm@seas.harvard.ed 1

3 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli Introdction The reglation of gene expression via the RNA interference (RNAi) process reqires a doble-stranded RNA molecle with a seqence that is complementary to the targeted transcript (Fire, 1998). MicroRNAs (mirnas) are short endogenos reglatory RNAs that act throgh RNAi. mirnas are incorporated into the mirna loading complex (mirlc), forming the RNA indced silencing complex (RISC), which binds to complementary mrna and cleaves at the target site or indces silencing by other mechanisms inclding translational repression and RNA destabilization. The basic mechanism modeled in this paper is smmarized in Figre 1. We focs on the RISCbased cleavage mechanism of RNAi, and note that while other mechanisms of RNAi may be more prevalent in celllar systems, RISC formation is necessary for most modes of RNAi and the mechanism described in this paper is important for design of synthetic RNAi effectors (Shabalina et al, 26). Althogh many synthetic constrcts have been made to mimic mirna fnction (Kim and Rossi, 27), the design of sch constrcts is typically based on a complex set of empirical rles rather than more fndamental RNA thermodynamic data. Fndamental models of RNAi based on thermodynamic calclations cold lead to more accrate and efficient rational design strategies for synthetic mirnas. Here we demonstrate a necessary first step toward that goal: we otline the basic thermodynamics of RNAi, and we show that a model based on RNA thermodynamics can accrately reprodce data for RNAi efficiency in in vitro hman systems. An interesting characteristic of hman RNAi is that, in contrast with RNAi in Drosophila, both RISC formation and mltiple ronds of mrna cleavage do not seem to reqire energy from ATP hydrolysis (Maniataki and Morelatos, 25; Gregory, 25; Provost, 22; Zhang, 22). Some hypotheses are that copling with the phosphoester bond-breaking steps and binding with Ago2 help drive dplex-nwinding (Maniataki and Morelatos, 25; Gregory et al., 25). Here we se existing data to verify these hypotheses. While previos stdies (Shabalina et al, 26; Taxman et al, 26; Ui-Tei et al, 26) have examined a wide range of thermodynamic featres in order to find correlations with RNAi effector efficiency, little work has been done to model RNAi fnction from first principles. In vitro systems provide closed, thermodynamically controlled systems, and the apparent lack of ATP dependence makes absolte thermodynamic modeling possible. We examine data from two in vitro-based papers in which detailed qantitative information is available on concentrations of protein, RNAi effectors, target mrna, and levels of knockdown (Maniataki and Morelatos, 25; Gregory et al., 25). Using an eqilibrim statistical thermodynamic model based on 2

Pblished in Oligoncleotides, 19(4), 341-346 (29). doi:1.189/oli.29.186 nearest-neighbor interactions, we accrately reprodce patterns in ATP-independent knockdown levels measred previosly in vitro.

4 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli nearest-neighbor interactions, we accrately reprodce patterns in ATP-independent knockdown levels measred previosly in vitro. Or calclations spport the hypothesis that phosphoester bond-breaking steps are sfficient to drive RNAi throgh the RISC cleavage mechanism. Figre 1. The RNA interference mechanism modeled in the Methods and Calclations section. Steps 1 to 3 correspond to the methods section Calclation of in vitro RISC formation levels and comparison to data from Gregory et al (25), with the net reaction S + R R + L + P. Steps 4 to 6 correspond to the methods section Calclation of in vitro mrna cleavage levels and comparison to data from Maniataki and Morelatos (25), with the net reaction M R F1+ F2. The energies involved in the labeled steps are: (1) Gbind, (2) Ghydrolysis, (3) GS and G L, P, (4) & (5) GM, (6) G F1, 2 and G hydrolysis. Note that in steps 4 to 6, the RISC binding and detaching terms cancel and are not inclded in the calclations. 3

5 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli Methods and Calclations Energy Calclations In a closed, non-celllar in vitro system withot ATP, the behavior of the system depends on the distribtion of energy states of the participating molecles. To briefly review, the probability that a system is in any particlar microstate s is given by 1 E( s) N P( s) = exp, Z RT where E is the energy of the state, N is the nmber of molecles, and Z is the partition fnction, Z = s E( s) N exp. RT The partition fnction for the system is a central qantity, and it is sed to derive a nmber of important energies as described below. Changes in ensemble energies of the system determine the final otcomes of the associated reactions. The stabilities of RNA strctres can be determined sing the nearest-neighbor model, in which the stability of a particlar base pair depends on the adjacent base pair (Markham and Zker, 25). The DINAMelt webserver (Markham and Zker, 25) was sed for RNA folding and hybridization calclations by the nearest-neighbor model at 37 C. In the calclations we assme that there is no energy inpt from ATP, consistent with the experimental stdies. DINAMelt takes the ensemble of nfolded and nhybridized strands as its reference state (Dmitrov and Zker, 24), so the energy change reqired to fold and hybridize a mixtre of RNAs can be taken as the ensemble energy given by DINAMelt for that mixtre at 37 C at the concentration entered. To smmarize Dmitrov and Zker, given an initial amont A chemical potentials for A and self-hybridized AA are N A µ A = RT ln[ Z ( N A ) + RT ln N A and N of molecles A, with canonical partition fnction Z ( ) N [ ( ) ( ) AA µ AA = RT ln Z N AA + RT ln. 2 N A Taking the free energy of nfolded states to be zero, the free energy F = µ AN A+ µ AAN AA of the ensemble is the energy change from the ensemble of molecles A in completely nfolded states to the final folded ensemble. Likewise, the energy N A, the 4

6 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli reqired to nfold and nhybridize the mixtre is given by the negative of the (folded) ensemble energy given by DINAMelt. Calclations were done at 37 C ( R gas T = kcal/mol) and at the RNA concentrations as indicated from Maniataki and Morelatos (25) and Gregory et al (25). DINAMelt does not accont for variations in ion concentrations. Hydrolysis energies were taken from previos experiments (Alberty, 27), bt we estimate that discrepancies p to 5% may arise since the systems in the experiments were at 25 C and possibly different ionic conditions. From Table 7 of Alberty (27), the energy of hydrolysis of CMP, UMP, and TMP is kj/mol, or approximately 3.1 kcal/mol per bond. Two phosphoester bonds are hydrolyzed in the processing of premirna, and ths we take this energy of hydrolysis to be 6.2 kcal/mol. The first step of RISC formation is the binding of the pre-mirna with Argonate 2 (Ago2) in the mirlc (Kim and Rossi, 27). Yan et al (23) measred a dissociation constant K d 1µ M for binding of the Argonate PAZ domain with short single-stranded RNA. In the experiments examined in this paper, the concentrations of RISC and gide RNA were all near or mch lower than K d, so the free energy change for binding G bind. Ths, for ensing calclations, we took G bind as a first order approximation. Calclation of in vitro RISC formation levels and comparison to data from Gregory et al (25) To compare RISC formation sing pre-mirna vs. dsrna, we compared energies sing the let-7 pre-mirna and dsrna seqences from Gregory et al (25). All calclations were done at.5 nm RNA in accordance with the experiments. Let R = inactivated mirlc, R = activated RISC, S = complete pre-mirna, L = loop fragment, and P = passenger strand. The net reaction for RISC formation is given by: S + R R + L + P Let the sbscript f stand for folded and stand for nfolded, and let A stand for the antisense or gide strand of the pre-mirna. For the prposes of calclating the free energy of formation, the total reaction can be thoght of as the sm of for simpler reactions, each with an associated free energy change G : S S GS S A + L + P Ghydrolysis A + R R Gbind L + P L + P G f f L, P 5

7 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli G, G S and L P can be calclated sing DINAMelt. For the let-7 pre-mirna, which ndergoes the reaction above, inpting the seqence yields a GS of kcal/mol, and inpting the cleaved fragments gives G L, P of kcal/mol. For the let-7 premirna, Ghydrolysis is the energy change de to cleavage of 2 phosphoester bonds to release three new fragments, and can be approximated as 6.2 kcal/mol (Alberty, 27). As mentioned in the previos section, we take G bind as a first order approximation. The total energy change for RISC formation is given by G RISC = GS + Ghydrolysis + Gbind + GL, P. The corresponding eqilibrim constant is given as: K = exp G R T (1) RISC ( ) RISC gas This eqilibrim constant comes ot to.2476 for the let-7 pre-mirna. Using the initial concentration of pre-mirna [ S as the reference concentration for RNA folding and sbseqent reactions, if r is the fraction of S converted to RISC according to experimental data, the eqilibrim can be written as: K RISC [ R [ L [ P [ S [ S [ S [ R [ S [ S [ S r[ S r[ S r[ S [ S [ S [ S ( r)[ S ( 1 r)[ S [ S [ S 3 r = = = (2) 1 r ( 1 ) 2 Combining eqations (1) and (2), it is possible to solve for r given GRISC or vice versa. Solving for r gives the vale.242 for the let-7 pre-mirna. A similar reaction occrs if doble-stranded RNA (dsrna, no loop) is sed. Letting D stand for the dsrna, the reaction can be written as: D f A + P A + R P P f R G G G D bind P For the let-7 dsrna, GD is kcal/mol (energy to separate the two strands), there is no hydrolysis energy (no loop to hydrolyze), and GP is the nfolding energy of the passenger strand, which is kcal/mol. Again, we are taking G bind as a first order approximation. The RISC formation energy is given by GRISC, D = GD + Gbind + GP, which sms to for the let-7 dsrna, and the corresponding eqilibrim constant is given as: K ( G R T ), = exp gas (3) RISC D RISC, D 6

8 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli For the let-7 dsrna, the eqilibrim constant comes ot to.127. If r D is the fraction of D converted to R, the eqilibrim constant can also be written as: K RISC, D 2 D r = (4) ( r ) 2 1 D Combining (3) and (4), it is possible to solve for r D given G RISC, D. For the let-7 dsrna, this method gives a vale of.11 for r D. Assming the reaction rate is roghly proportional to enzyme concentration (Li, 24), r rd approximates the difference in RNAi efficiency for pre-mirna verss dsrna. In this case, r r =.242/.11 which is approximately 2.4. D r rd can also be fond by measring actal RNAi efficiencies in vitro. We sed data from Gregory et al Figre 4, which had experiments performed at.5 nm RNA concentrations. In Figre 4B, the % conversion to cleaved prodct for RISC loaded let-7 dsrna withot ATP is given as 6, and in Figre 4C, the % conversion for let-7 pre-mrna withot ATP is 15. Taking this to be eqal to r rd, we obtain a vale of 15/6 = 2.5, close to or calclated prediction of 2.4. By inptting the let-7 mrna target seqence and the seqences of the cleaved prodct into DINAMelt, we fond the mrna target site nfolding energy, and mrna fragment strands ensemble energy, which are also inclded in Table 1. Calclation of in vitro mrna cleavage levels and comparison to data from Maniataki and Morelatos (25) We next attempt to predict absolte mrna knockdown levels sing the GL3 target and GL3 pre-mirna seqences from Maniataki and Morelatos (25) and compare to their experimental data. Calclations are at 1 nm concentrations in accordance with the experiments. Let R stand for the active RISC complex, let M stand for the target mrna, and let F1 and F2 stand for the two fragments reslting from cleavage of M. Then the cleavage reaction is given by: M R F1+ F2 As before, this reaction can be thoght of as the sm of some simpler reactions: M M f F1 M + F2 F1 + F2 F1 f + F2 f G G G M hydrolysis F1,2 7

9 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli G M and GF can be calclated with DINAMelt, and Ghydrolysis can be approximated as 3.1 kcal/mol (Alberty, 27). GM for the GL3 target seqence was kcal/mol. G F for the GL3 target seqence was The total energy change for the cleavage reaction is given by Gcleavage = GM + Ghydrolysis + GF1, 2, which comes to for GL3, and the corresponding eqilibrim constant is given as: K cleavage = exp ( G R T ) cleavage gas (5) Ths, or calclated eqilibrim constant comes to Using the initial concentration [ M of mrna as the reference concentration for both RNA folding and the total reaction, if c is the cleaved fraction of mrna at eqilibrim, then the eqilibrim constant can also be written as: K cleavage = [ F1 [ F2 [ M [ M [ M [ M [ M c[ M [ M [ M ( 1 c)[ M [ M c 2 c = = (6) 1 c Combining eqations (5) and (6), it is possible to calclate c given Gcleavage. For the GL3 target, c is.56. The calclated fraction c can then be compared to the experimental fraction of mrna cleaved in vitro. From Figre 6 of Maniataki and Morelatos (25), cleaved mrna reaches an eqilibrim fraction of approximately.5, with GL3 target mrna at 1 nm. In addition, we calclated the pre-mirna nfolding and loop/passenger strand ensemble energies by feeding the GL3 pre-mirna and loop/passenger seqences into DINAMelt. These vales, along with the -6.2 kcal/mol hydrolysis vale (from Alberty, 27, as before) are also inclded in Table 1. Seqences for in vitro calclations Seqences taken from Gregory et al (25) are let-7 dsrna (UGAGGUAGUAGGUUGUAUAGU + UAUACAAUGUGCUAGCUUUCU), let-7 pre-mirna (UGAGGUAGUAGGUUGUAU AGUAGUAAUUACACAUCAUACUAUACAAUGUGCUAGCUUUCU) and let- 7 mrna target (UAUACAACCUACUACCUCAUU). Seqences taken from Maniataki and Morelatos (25) are GL3 pre-mirna (UCGAAGUACUCAGCGUAAGUGGCUGUGAAGCCACAGAUGGGCCACUUACG 8

10 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli GAGUACUUUGAGC) and GL3 target (UGGACAUCACUUACGCUGAGUACUUCGAAAUG). Reslts and Discssion Previos work in the literatre has focsed on trends in energy dependence for RNA interference, drawing from large bodies of in vivo data (Shabalina et al, 26; Taxman et al, 26; Ui-Tei et al, 26). In contrast, we attempt to predict absolte levels of mrna knockdown or RNAi effector efficiency based on first principles. We provide a proof of principle by comparing thermodynamic calclations to a limited data pool. In Table 1 we show calclated vales from or model for the free energy changes in varios steps of the RNAi mechanism. Details for the derivation of each vale are given in the Methods. Important energy contribtions inclde folding energies of the premirna and the mrna target site (Table 1), in line with previos observations that premirna nwinding and mrna target site nfolding provide the greatest energy barriers to RNAi (Shabalina et al, 26). With these vales we first examine the difference in efficiency between premirna and dsrna against the same target site, then we calclate the absolte fractions of mrna that wold be cleaved in vitro by a given RNAi effecter. Overall, nmerical reslts from the model agree well with the experimentally measred levels of RISC formation and mrna cleavage in vitro, with no adjstable parameters. Calclated RISC formation levels and comparison to data from Gregory et al (25) If r is the fraction of pre-mirna converted to RISC and r D is the fraction of dsrna converted to RISC, the vale r shold correspond roghly to the ratios of mrna r D cleaved for reactions sing pre-mirna verss dsrna. For RNAi effecters against let-7 mrna targets sed in Gregory et al (25), we calclate a theoretical r of 2.4, close to the experimental vale of 2.5. The hypothesis in the original experimental work was that energy for the dplexnwinding step of RISC is provided by the cleavage of two phosphoester bonds to release the loop fragment of the pre-mirna. The agreement between the reslts from or model and the experimental data spports this hypothesis. While it is well established that stemloop strctres are better RNAi effectors than dsrna, or calclations sggest that the primary reason for the difference in efficiency is that pre-mirna has a loop that can be cleaved whereas dsrna does not. The extra 6.2 kcal/mol liberated by the cleavage reaction is a significant fraction of the total G. r D 9

11 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli Experiment Energy Terms for RISC formation pre-mirna or dsrna nfolding G S G D pre-mirna or dsrna hydrolysis 2 G hydrolysis pre-mirna or dsrna loop/ passenger strand ensemble energy Energy terms for mrna cleavage mrna target site nfolding G M mrna hydrolysis G hydrolysis mrna fragment strands ensemble energy G F Gregory et al (25), prelet-7-mirna Gregory et al (25), ds-let- 7-miRNA Maniataki and Morelatos (25), PML- GL3 G, L P G P Table 1. Calclated energy terms, in kcal/mol, for varios steps in the RNAi mechanism. Reference concentrations were set at.5 nm for the Gregory paper and 1 nm for the Maniataki paper, in concordance with their respective experimental concentrations. Details on how each vale was obtained are inclded in the methods section. Calclated in vitro mrna cleavage levels and comparison to data from Maniataki and Morelatos (25) Maniataki and Morelatos (25) had previosly performed experiments with prified hman minimal RISC (Dicer, Ago2, and the RNA binding protein TRBP), monitoring cleavage of GL3 target mrna (1 nm) for 9 mintes. Assming that the cleavage reaction had reached eqilibrim by that time, or calclation for the fraction c of cleaved mrna shold match the observed fraction. Calclating folding energies with DINAMelt and solving for c as otlined in the methods, we calclate a fraction c of.56, close to the measred vale,.5. The calclation sggests that energy changes de to RNA folding and phosphoester bond hydrolysis are sfficient to explain the levels of mrna cleavage by minimal RISC in vitro. Ths the model shows qantitative spport for the notion that RNA bond cleavage energies drive both RISC formation and mrna cleavage. Bt we also note that energy changes de to binding to RISC proteins were not inclded in or model, and were in fact not necessary to explain the in vitro RNAi reslts. A key implication is that direct modeling from basic physical principles and RNA thermodynamic data may be sfficient 1

12 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli to accrately describe hman RNAi systems; that is, the protein binding energies do not seem to make a sbstantial contribtion to the total free energy, at least for the seqences stdied in the in vitro experiments. However, this may not be the case for dynamic living systems. Ftre experiments measring binding constants of hman Ago2 with a variety of singlestranded RNAs cold help make better estimates of the relevant energy changes in living hman cells. We note also that the simple eqilibrim thermodynamic model presented in this paper does not take into accont many of the complexities of the celllar system, where longer seqences are involved, additional proteins may participate, discarded RNA fragments are qickly cleaved by celllar machinery, additional mechanisms of RNAi are prevalent, and rates of mrna or mirna prodction are of tmost importance. Nonetheless, or calclations show that nearest neighbor thermodynamic model can provide sefl qantitative insight into RNA interference. More thorogh characterization of RISC formation and activity in live hman cells shold help develop more accrate kinetic models of RNA interference. 11

13 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli References ALBERTY, A. (27). Thermodynamic properties of enzyme-catalyzed reactions involving cytosine, racil, thymine, and their ncleosides and ncleotides. Biophysical Chemistry 127, DIMITROV, R.A. and ZUKER, M. (24) Prediction of hybridization and melting for doble-stranded ncleic acids. Bipohys. J. 87, FIRE, A., XU, S., MONTGOMERY, M.K., KOSTAS, S.A., DRIVER, S.E., AND MELLO, C.C. (1998). Potent and specific genetic interference by doble-stranded RNA in Caenorhabditis elegans. Natre. 391(6669), GREGORY, R., CHENDRIMADA, T., COOCH, N., and SHIEKHATTAR, R. (25). Hman RISC Coples MicroRNA Biogenesis and Posttranscriptional Gene Silencing. Cell 123, KIM, D.H., and ROSSI, J.J. (27). Strategies for silencing hman disease sing RNA interference. Natre Rev. Genetics 8, LIU, J., CARMELL, M., RIVAS, F., MARSDEN, C., THOMSON, J., SONG, J.J., HAMMOND, S., JOSHUA-TOR, L., and HANNON, G. (24). Argonate2 Is the Catalytic Engine of Mammalian RNAi. Science 35, MARKHAM, N.R., and ZUKER, M. (25). DINAMelt web server for ncleic acid melting prediction. Ncleic Acids Res. 33, W577-W581. MANIATAKI, E., AND MOURELATOS, Z. (25). A hman, ATP-independent, RISC assembly machine feled by pre-mirna. Genes & Dev. 19, PROVOST, P., DISHART, D., DOUCET, J, FRENDEWEY, D., SAMUELSSON, B., and RADMARK, O. (22). Ribonclease activity and RNA binding of recombinant hman Dicer. The EMBO Jornal 21, SHABALINA, S., SPRIDONOV, A., and OGURTXOV, A. (26). Comptational models with thermodynamic and composition featres improve sirna design. BMC Bioinformatics 7:65. 12

14 Pblished in Oligoncleotides, 19(4), (29). doi:1.189/oli TAXMAN, D., LIVINGSTON, L, ZHANG, J., CONTI, B., IOCCA, H., WILLIAMS, K., LICH, J., TING, J., and REED, W. Criteria for effective design, constrction, and gene knockdown by shrna vectors. (26). BMC Biotech. 6:7. UI-TEI, K., NAITO, Y., and SAIGO, K. Essential Notes Regarding the Design of Fnctional sirnas for Efficient Mammalian RNAi. (26). J. of Biomed. and Biotech. 26, Article ID 6552, 8 pages. YAN, K., YAN, S., FAROOQ, A., HAN, A., ZENG, L., and ZHOU, M.M. (23). Strctre and conserved RNA binding of the PAZ domain. Natre 426:27, ZHANG, H., KOLB, F., BRONDANI, V., BILLY, E., and WITOLD, F. (22). Hman Dicer preferentially cleaves dsrnas at their termini withot a reqirement for ATP. The EMBO Jornal 21,

High-Resolution Quantitative Methylation Profiling with EpiTYPER and the MassARRAY System

High-Resolution Quantitative Methylation Profiling with EpiTYPER and the MassARRAY System High-Resoltion Qantitative Methylation Profiling with EpiTYPER and the MassARRAY System A complete system for discriminating methylated vs. non-methylated DNA; incldes software, reagents, and MALDI-TOF