Isoalloxazine derivatives promote photocleavage of natural RNAs at G·U base pairs embedded within helices
Isoalloxazine derivatives promote photocleavage of natural RNAs at G·U base pairs embedded within helicesPetra Burgstaller+, Thomas Hermann1, Christian Huber, Eric Westhof1 and Michael Famulok*
Institut für Biochemie der LMU München-Genzentrum, Würmtalstraße 221, 81375 München, Germany and 1Institut de Biologie Moléculaire et Cellulaire, CNRS UPR SMBMR, 15 Rue Rene Descartes, 67084 Strasbourg Cedex, France
Received July 16, 1997;Revised and Accepted August 26, 1997
ABSTRACT
We have recently shown that isoalloxazine derivatives are able to photocleave RNA specifically at G·U base pairs embedded within a helical stack. The reaction involves the selective molecular recognition of G·U base pairs by the isoalloxazine ring and the removal of one nucleoside downstream of the uracil residue. Divalent metal ions are absolutely required for cleavage. Here we extend our studies to complex natural RNA molecules with known secondary and tertiary structures, such as tRNAs and a group I intron (td). G·U pairs were cleaved in accordance with the phylogenetically and experimentally derived secondary and tertiary structures. Tandem G·U pairs or certain G·U pairs located at a helix extremity were not affected. These new cleavage data, together with the RNA crystal structure, allowed us to perform molecular dynamics simulations to provide a structural basis for the observed specificity. We present a stable structural model for the ternary complex of the G·U-containing helical stack, the isoalloxazine molecule and a metal ion. This model provides significant new insight into several aspects of the cleavage phenomenon, mechanism and specificity for G·U pairs. Our study shows that in large natural RNAs a secondary structure motif made of an unusual base pair can be recognized and cleaved with high specificity by a low molecular weight molecule. This photocleavage reaction thus opens up the possibility of probing the accessibility of G·U base pairs, which are endowed with specific structural and functional roles in numerous structured and catalytic RNAs and interactions of RNA with proteins, in folded RNAs.
INTRODUCTION
Certain organic compounds of low molecular weight, designated photosensitizers, can damage biomolecules upon irradiation with light. Endogenous photosensitizers such as porphyrins and flavins might be involved in photocarcinogenesis because their reaction products may lead to misreplication of DNA, mutations and cancer (1 ). Natural and synthetic photosensitizers have attracted considerable interest because of their utilization in photochemotherapy, the treatment of diseases with photosensitizing drugs plus light (2 ). Another important application of photochemotherapy is the treatment of tumors with photosensitizers like hematophorphyrin (3 ). Photosensitizers have also been applied to the inactivation of viruses in contaminated medical samples such as blood plasma (4 ).
The isoalloxazine derivative riboflavin can promote photooxidative strand scission in DNA after treatment with piperidine by generating 8-hydroxydeoxyguanosine (5 -7 ). An artificial restriction endonuclease was obtained as a synthetic netropsin-flavin hybrid molecule which cleaves DNA at a specific A:T-rich locus determined by the sequence specificity of the netropsin moiety (8 ). In contrast to DNA, however, little is known about the effect of photosensitizers on RNA.
Recently we isolated an RNA aptamer by in vitro selection that specifically recognizes the isoalloxazine moiety of FMN or FAD in solution (9 ). During the structural characterization of this aptamer we observed that isoalloxazine derivatives (Fig. 1 ) can induce strand breakage 3' of the uracil of G·U wobble base pairs by a photooxidative cleavage mechanism (10 ). Strand breakage also occurred with high specificity in a variant of this RNA aptamer lacking the FMN binding site while still containing the G·U pairs embedded within a helix. Thus, the photocleavage reaction did not require a high affinity FMN or isoalloxazine binding site in the RNA. Furthermore, no bias for the removed nucleoside exists, as cleavage occurs independently of the residue located 3' of the uracil involved in G·U base pairing (10 ).
MATERIALS AND METHODS
Materials
Riboflavin, FMN, FAD, lumiflavin and lumichrome were purchased from Fluka, [[gamma]-32P]ATP and [5'-32P]pCp from Amersham. Yeast tRNAPhe, tRNAfMet and tRNAVal were obtained from Sigma, yeast tRNAAsp was a generous gift from Dr G.Keith (Strasbourg). Plasmid td [Delta]P6-2T containing 100 nt of the 5' exon, the 265 nt td intron and 56 nt of the 3' exon were generous gifts from Dr R.Schroeder (Biocenter Vienna) (27 ). T7 RNA polymerase was purified from the overproducing strain BL21/pAR1219, following the purification protocol provided by F.W.Studier (28 ). DNase I (RNase-free) was from Boehringer Mannheim, Taq polymerase from Eurogentec and T4 polynucleotide kinase, T4 RNA ligase and calf intestinal alkaline phosphatase from New England Biolabs. Ultrapure, unlabeled NTPs and dNTPs were obtained from Boehringer Mannheim. Primers and synthetic oligonucleotides used in PCR amplification reactions were synthesized on a Millipore Expedite oligonucleotide synthesizer using standard phosphoramidite chemistry. Oligonucleotides were purified as described previously (29 ) and concentrations were determined by absorbance measurements at 260 and 280 nm.
Preparation of DNA and RNA
The RNAs used for this study were transcribed from DNA templates containing a T7 promotor. DNA templates were generated by PCR amplification of synthetic oligonucleotides. PCR reactions were performed in PCR buffer (10 mM Tris-HCl, pH 8.3, 50 mM KCl, 0.001% gelatin, 1.5 mM MgCl2, 0.3% Tween 20, 0.2 mM dNTPs) in the presence of 3 mM primer and 2 U Taq polymerase. The precursor RNA of the T4 phage-derived thymidylate synthase (td) intron was in vitro transcribed from plasmid td [Delta]P6-2T containing 100 nt of the 5' exon, the 265 nt intron and 56 nt of the 3'-exon (30 ). Aliquots of 250 nM precursor RNA were pre-incubated and renatured as described previously (27 ,31 ).
RNA labeling
For 5'-end-labeling the transcribed RNA was dissolved in CIP buffer (50 mM Tris, pH 8.5, 0.1 mM EDTA, 0.1 mg/ml BSA) and treated with 0.05 U/pmol RNA calf intestinal alkaline phosphatase for 30 min at 37oC. After purification by preparative gel electrophoresis on polyacrylamide-8.3 M urea gels, ~10 pmol eluted RNA was redissolved in kinase buffer (70 mM Tris-HCl, pH 7.6, 10 mM MgCl2, 5 mM DTT), 5'-end-labeled using 10 U T4 polynucleotide kinase and 30 mCi [[gamma]-32P]ATP for 30 min at 37oC and again purified on polyacrylamide-8.3 M urea gels. For 3'-labeling 30 pmol RNA was incubated with 30 mCi [5'-32P]pCp in 50 mM HEPES, pH 7.5, 20 mM MgCl2 in the presence of 6 U T4 RNA ligase at 4oC for 12-16 h followed by gel purification.
Flavin cleavage experiments
Reactions were performed in Eppendorf tubes in a total volume of 20 µl. Aliquots of 250 nM 5'-32P-end-labeled RNA (~10 000 c.p.m.) were denatured in 200 mM NaCl, 50 mM Tris-HCl, pH 7.6, 2 mM EDTA for 3 min at 95oC and subsequently renatured for 10 min at room temperature. After adding MgCl2 to a final concentration of 12 mM, the RNA was irradidated in the presence of 200 µM FMN or riboflavin (100 µM for lumiflavin; see Fig. 5 ) at ambient temperature for up to 4 h using incident light from a polychromatic lamp (20 J/s) at a distance of between 25 and 30 cm from the RNA sample. Light irradiation had no detectable effect on the temperature of the RNA samples. The reaction was stopped by precipitation by addition of 60 µl ethanol. The RNA was redissolved in H2O and analyzed on polyacrylamide-8.3 M urea gels. For calibration of gel band positions 5'-labeled RNA was cleaved at G residues by digestion with T1 ribonuclease or was subjected to alkaline hydrolysis, as described elsewhere (32 ). For hydroxyl radical cleavage the RNA was incubated in 1 mM Fe(NH4)2(SO4)2·6H2O, 2 mM EDTA, 0.05% H2O2, 5 mM DTT at 25oC for 10 min (33 ,34 ).
Flavin cleavage in the absence of oxygen
All buffers and stock solutions used for cleavage experiments in the absence of oxygen were degassed for 1 h by ultrasonication in the presence of argon. Subsequently, a stream of argon was bubbled through the solvents and buffers for 15 min. These two steps were repeated at least once. Bottles were sealed with rubber caps and solutions removed with tight sealing, argon-rinsed syringes in a stream of argon. Reactions were performed in sealed tubes which were rinsed with argon for at least 15 min. During pipetting of the solutions and during incubation, care was taken to avoid oxygen contact by applying an argon blanket. Reactions were performed in the same way as described above.
Molecular modeling and simulations
Coordinates for the RNA were taken from the crystal structure of yeast tRNAAsp (nt U1-A7 paired to nt U66-A72; 15 ), coordinates for FMN were from a molecular model constructed with the Insight program (Biosym Technologies, San Diego). Docking of the FMN molecule to the RNA helix was done manually, accounting for the probable hydrogen bonding interactions of the isoalloxazine ring with the amino group of the G of the G·U pair and the 2'-OH of the U. The conformation of the FMN exocyclic sugar moiety was chosen in such a way that no interaction of the sugar with the RNA occurred. A Mg2+ ion was placed manually, bridging an exocyclic carbonyl oxygen of the isoalloxazine moiety and the phosphate of the nucleotide 3' of the U of the G·U pair.
For molecular dynamics simulations the AMBER 4.1 (35 ) package was used. Forcefield parameters for FMN were derived as described previously (36 ). Parameters for Mg2+ were from Åqvist (37 ). The RNA was placed in a rectangular box of SPC/E water containing about 2500 solvent molecules. Eleven Na+ counterions were placed according to the electrostatic potential around the solute such that no ion was closer than 4.5 Å to any solute atom. The simulations were run with a time step of 2 fs at a constant temperature of 298 K and a constant pressure of 1 atm. The SHAKE algorithm was used to constrain the X-H bond lengths. Van der Waals interactions were truncated at 9.0 Å, while no cut-off was applied to the electrostatic term. The electrostatic interactions were calculated by the Particle Mesh Ewald method with a charge grid spacing close to 1.0 Å. The equilibration protocol was similar to those used in preceding work (38 ,39 ).
RESULTS AND DISCUSSION
Cleavage is specific for G·U pairs
Natural RNAs with a known secondary and tertiary structure and without known specific affinity for flavins were chosen in order to demonstrate that photoinduced cleavage of RNA by isoalloxazines is a general phenomenon. We first tested a series of transfer RNAs: tRNAPhe, tRNAfMet, tRNAVal and tRNAAsp. Yeast tRNAPhe contains a single G·U wobble base pair in the acceptor stem, G4·U69 (Fig. 2 A). Accordingly, with 5'-end-labeled tRNAPhe a single cleavage signal at U69 was obtained (10 ). Analogously, tRNAfMet and tRNAVal, which contain, respectively, a U51·G65 and G50·U64 pair in the T stem, showed a corresponding single cleavage signal at U51 and U64 (Fig. 2 B and C). No other position was affected by the photosensitizer in these tRNA molecules.
No preference for the removed nucleoside
In the tRNAs described so far the nucleoside removed downstream of the uracil of the cleaved G·U base pair was either a guanosine or a cytosine. We therefore analyzed several synthetic RNA constructs with different residues located 3' of the uracil to test whether a bias for the nucleoside removed exists. In addition, RNA constructs containing tandem G·U pairs or a U·U mismatch adjacent to the G·U pairs were tested. As summarized in Table 1 , only the RNA which contained two G·U pairs in tandem alignment was not cleaved (Table 1 , row 6). In all other constructs cleavage occurred independently of the residue located 3' of the uracil involved in G·U base pairing.
. FMN-dependent cleavage of different synthetic and natural RNAs
RNA
Length (nt)
G·U-containing stem
Residue 3' of U
Intensitya
Rowa
FMN-2
109
5'-CCGACUGUGGU
G
++
1
3'-GGCUGGCACCA
PB-5
107
5'-UGCUC
A
+
2
3'-AUGAG
PB-7
113
5'-CUUC
C
++
3
3'-GAGG
PB-13
110
5'-CUUCACU-GCAUU1CC
U1
+
4
3'-GAAGUGU2 UGUGA-GG
U2
++
PB-9
74
5'-CGCGCC
G
++
5
3'-GCGUGG
PB-5/2
107
5'-GGAGCC
U
-
6
3'-AUUUCGG
A
-
a++, strong cut; +, medium to weak cut; -, no cut. bRow 1, example of a strong cut when G is removed; row 2, example of a cut when A is removed; row 3, example of a strong cut when C is removed; row 4, example of a cut when U1 is removed and example of a cut when U2 belonging to a U-U mismatch is removed; row 5, example of a strong cut when G is removed; row 6, no cleavage detectable at tandem G·U pairs.
The finding that no bias for the nucleoside removed exists and that cleavage site selection is directed by a G·U base pair significantly differs from the sequence specificity observed in photosensitized DNA cleavage and indicates that it is the ribose moiety rather than the base which is attacked during cleavage. Treatment of dsDNA with photosensitizers, including Ru(III) intercalators (40 ,41 ), were found to affect mainly guanine residues. With flavins DNA cleavage was observed only after incubation with piperidine (7 ) and mechanisms in which the photosensitizer directly or indirectly destroys the base have been suggested (5 ). A non-endogenous synthetic system consisting of an isoalloxazine ring covalently attached to either netropsin (8 ) or distamycin (42 ) resulted in a single-strand break in dsDNA upon irradiation with visible light. In this cleavage mechanism an attack at the deoxyribose induced by the irradiated isoalloxazine moiety has been discussed. Sequence specificity for A:T-rich regions was observed which resulted, however, from the attached groove binders.
Photocleavage requires isoalloxazine and divalent metal ions
To compare the activity of different isoalloxazine derivatives in the cleavage reaction, RNAs were incubated with FMN, riboflavin, lumiflavin (Fig. 3 A), FAD or lumichrome. In addition, to test for a requirement for divalent metal ions, the reaction was performed in the presence and absence of magnesium. As expected, the non-photosensitizing FAD and lumichrome were inactive in strand scission (data not shown; FAD is non-photosensitizing because at pH 7.6 the adenosine is stacked onto the flavin ring and quenches the excited triplet state; 43 ). With the other isoalloxazine derivatives no significant difference in activity was observed, although isoalloxazine exhibited a slightly higher activity than FMN and riboflavin. Quantification of the band at the G68·U5 base pair revealed >10% cleavage of the total input RNA after 1 h.
G·U base pairs affected in the td intron
We used our new structure probing technique based on the photooxidative targeting of G·U base pairs to map G·U base pairs in the precursor RNA of the T4 phage-derived thymidylate synthase (td) intron (16 ; Fig. 4 ). The 265 nt group I intron contains several G·U base pairs located within stems or as closing base pairs of loops, as shown in the secondary structure model in Figure 4 A. This intron belongs to subgroup IA2, as does the T4 nrdD intron, which was previously characterized physicochemically and for which a 3-dimensional model exists (17 ). Gel separation of the 5'-labeled intron reveals a major site of cleavage 3' of U102, which forms a base pair with G90, and two minor cleavage sites downstream of U72·G53 and U93·G99 (Fig. 4 B). Two other strong bands are visible in the region close to the 3'-end which were resolved by analyzing the 3'-end-labeled intron RNA. Analysis confirmed that these two additional cleavages occur at G141·U152 and G235·U253, located in stems P7.2 and P9.2 respectively. The second G233·U255 pair of stem 9.2, located at a helix extremity, was not affected. Furthermore, no cleavage could be detected in stem P1, which contains two sets of tandem G·U pairs (Fig. 4 C). As shown in Figure 4 C, a low percentage of spontaneous magnesium-dependent hydrolysis at the 5'-splice site of the td intron is detected under the reaction conditions used in this study, indicating that P1 is correctly positioned on the catalytic core under our experimental conditions and that the intron is folded in an active conformation. An experimental indication of correct docking of stem P1 is site-specific hydrolysis at the 5'-splice site by Mg2+, resulting in products similar to guanosine-dependent splicing (31 ,46 ). The cleavage pattern for the td intron is summarized in Table 2 .
Minor cleavages at other sites
While the overwhelming majority of cuts observed in the RNAs used in this study occur at G·U base pairs, very weak cuts at positions which do not correspond to a G·U base pair were detected in two instances: G50·C63 in tRNAAsp and at the second G in a GGGA loop of a hairpin construct (data not shown). Presently, on the basis of the available data, we cannot explain these weak cuts. The only structural property common to the regions where these cuts occur is that the sugar-phosphate backbone folds back on itself with a sharp turn bringing two or more phosphate groups close to each other. Accordingly, a possible explanation might be that in such negatively charged and open pockets magnesium binding is facilitated while, at the same time, another type of stacking of the isoalloxazine ring, such as intercalation, would be permitted. Further experiments aimed at a better understanding of these weak cuts are under way.
A model for G·U recognition
The specificity of the cleavage reaction for G·U base pairs suggests that the isoalloxazine ring specifically recognizes geometrical and structural features resulting from the presence of a G·U pair within a helical region. It is likely that the isoalloxazine enters its G·U recognition site from the shallow groove of the helix. Indeed, the 3'-end of the paired U is readily accessible since in G·U wobble pairs the uracil base points into the deep groove of the RNA helix, forming a hollow surface within the shallow groove. Such cavities are often a site of specific hydration in RNAs, with the occupying water molecule either bridging the two bases of the G·U pair or the O2(U) and O2'(U) (39 ,48 ). A structural model illustrating a possible complex between an RNA helix containing a G·U pair, FMN and a magnesium ion is presented in Figure 5 A. The model is based on coordinates (15 ) of the acceptor stem helix of yeast tRNAAsp. The binding of FMN exploits the shallow groove asymmetry of G·U pairs by inserting the FMN exocyclic O2 atom in the space left by the presence of the uracil instead of a cytosine. Two hydrogen bonds from FMN to the RNA are suggested by the model, namely to the amino group of the guanine and to the 2'-OH of the uridine within the G·U pair (Fig. 5 A). In order to account for the magnesium requirement in isoalloxazine cleavage, a Mg2+ ion was intially placed as a bridge between O4 of FMN and the phosphate group 3' of the uracil residue. However, molecular dynamics simulations (Fig. 5 B; see also below) performed on the model of the complex suggested that there is an additional water molecule involved in this bridging interaction. The aromatic isoalloxazine ring stacks on the flat surface constituted by the sugar 3' of the uridine. Thus, the N5 atom of FMN, which carries the excited triplet orbital, is positioned close above the H1' and H4' atoms of the cleaved residue, either of which could be attacked during the cleavage reaction (10 ). Sugar ring opening might result from hydrogen abstraction, from either the C4' or the C1' atom (49 ). For steric and accessibility reasons the C4' atom is more susceptible to attack than C1'. Abstraction of H4' would be in accordance with oxidative cleavage mechanisms commonly found in DNA cleavage reactions (49 ) induced, for example, by copper-phenanthroline (50 ), enediines (51 ), bleomycin (52 ), Fe-EDTA and Fe-MPE (53 ). It should be noted, however, that the cleavage mechanism might involve an electron transfer step rather than hydrogen abstraction.
Conclusion
Our results show that RNA cleavage by the isoalloxazine ring depends on a molecular recognition event relying on the geometry and structure at G·U base pairs. Based on molecular dynamics simulations, we present an energetically stable 3-dimensional model illustrating the ternary complex of the G·U-containing helical stack, the isoalloxazine molecule and the metal ion which rationalizes the cleavage specificity along with the experimental data. As it is well established that RNA structures contain various different non-canonical base pairs (55 ,56 ), many of which are proven to also be present in the RNAs tested in this study, the specificity of cleavage site selection by the photosensitizer isoalloxazine for an individual class of non-canonical base pairs is striking. This photocleavage reaction thus opens up the interesting possibility of probing, in folded RNAs, the accessibility of G·U base pairs embedded within helices, which are known to be endowed with specific structural and functional roles in numerous structured and catalytic RNAs and interactions of RNA with proteins.
Remarkably little is known about specificity and even activity of photosensitizer-induced RNA cleavage. The only known example showing that RNAs are affected by photosensitizers is psoralen-induced photocrosslinking (57 ) of various positions in rRNAs, tRNAs and the spliceosome. The photoreaction of psoralens with RNA, however, occurs via a completely different mechanism than the cleavage mechanism described here. In addition to its potential application as a specific structural probe, the endogenous biological cofactor riboflavin may be a cheap and innocuous candidate to be tested for biomedical applications such as photochemical inactivation of RNA viruses in contaminated blood plasma and vaccines.
NOTE ADDED IN PROOF
The following paper came to our attention in which a similar effect was observed with Tris(4,7-diphenyl-1,10-phenanthroline) rhodium(III) using UV photocleavage: Chow,C.S. and Barton, J.K. (1992) Biochemistry, 31, 5423-5429.
ACKNOWLEDGEMENTS
We thank R.Schroeder for the generous gift of the plasmid td [Delta]P6-2T, G.Keith for samples of tRNAAsp, F.Michel, W.Kolanus and D.Faulhammer for helpful discussions and E.-L.Winnacker for support. This study was supported in part by grants from the Deutsche Forschungsgemeinschaft to M.F. and an EC grant (Bio2CT930345) to M.F. and E.W. T.H. is supported by an EMBO post-doctoral fellowship.
REFERENCES
1 Shibutani,S., Takeshita,M. and Grollman,A.P. (1991) Nature, 349, 431-434.MEDLINE Abstract
2 Diwu,Z. and Lown,J.W. (1994) Pharmacol. Ther., 63, 1-35.MEDLINE Abstract
3 Diamond,I., McDonagh,A.F., Wilson,C.B., Granelli,S.G., Nielsen,S. and Jaenicke,R. (1972) Lancet, ii, 1175-1177.
4 Hanson,C.V. (1983) Medical Virology. Elsevier, Amsterdam, The Netherlands, Vol. 2, pp. 45-79.
24 Moore,M.J., Query,C.C. and Sharp,P.A. (1993) In Gesteland,R.F. and Atkins,J.F. (eds), The RNA World. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 303-357.
55 Wyatt,J.R. and Tinoco,J.I. (eds) (1993) In Gesteland,R.F. and Atkins,J.F. (eds), The RNA World. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 465-496.
56 Nagai,K. and Mattaj,I.W. (1994) RNA-Protein Interactions. Oxford University Press, Oxford, UK.
+Present address: Department of Chemistry, University of Florida, PO Box 117200, Gainesville, FL 32611-7200, USA
*To whom correspondence should be addressed. Tel: +49 89 74017 410; Fax: +49 89 74017 448; Email: Famulok@lmb.uni-muenchen.de
