BIOINFORMATICS
Vol. 17 no. 11 2001 Pages 1058–1061
Concepts to automate the theoretical design of effective antisense oligonucleotides Rosel Kretschmer-Kazemi Far 1, Wolfgang Nedbal 2 and Georg Sczakiel 1 1 Medizinische
¨ zu Lubeck, Universitat ¨ Institut fur ¨ Molekulare Medizin, Ratzeburger Allee 160, D-23538 Lubeck, ¨ Germany and 2 A3D GmbH - Antisense Design & Drug Development, Waldhofer Straße 100, D-69123 Heidelberg, Germany
Received on January 11, 2001; revised on August 14, 2001; accepted on August 16, 2001
ABSTRACT Among the large number of possible antisense species against a given target RNA, only a small number shows effective suppression of the target gene in living cells. In the case of short-chain antisense oligonucleotides (asON) which usually comprise less than approximately 25 nucleotides, local structures of the target RNA seem to be of particular importance for the extent of gene suppression. Experimental approaches to identify promising local target sequences and, hence, complementary asON sequences, have provided tools to define asON that are biologically active at higher than statistical probability. However, experimental protocols are expensive, time consuming, and are associated with intrinsic basic and technical limitations. As insights into the structure–function relationship of asON as well as the role of sequence motifs increase, it becomes feasible to consider computer-based theoretical approaches for the design of effective asON. In the following we describe how individual steps of the theoretical design of asON may be automated by establishing and implementing suitable algorithms. Contact:
[email protected]
INTRODUCTION Antisense oligonucleotides (asON) have become a major tool in basic sciences, developmental biology, and molecular medicine to study and to suppress aberrant gene expression and viral functions (for review see Stein and Krieg, 1998). The successful use of asON to suppress gene expression is somewhat limited since only a small portion of all possible antisense species against a given target sequence shows efficacy in the respective test system (e.g. Peyman et al., 1995; Monia et al., 1996). The number of reports on the application of asON is extremely large and many studies have addressed the mode of action of asON in living cells. This bulk of descriptive work, however, does not seem to be reflected by the knowledge on the biophysical and biochemical level of the action of 1058
asON nor by the knowledge about the rules that govern the relationship between specific sequences of asON, the influence of the target structure, the annealing in vitro, and the efficacy in vivo. Currently, it seems to be generally accepted that the effectiveness of asON is strongly dependent on local target RNA structures, on chemical properties and sequences of the asON species, and on the characteristics of the biological system of interest including metabolic properties of the target RNA and the gene product, respectively. While the latter point can only be investigated experimentally, the first two points can, in principle, be automated by establishing and applying computer algorithms. Here, concepts will be described that may substantially support a ‘high throughput’ design of asON species that work at a higher than statistical rate of success in living cells.
ADVANTAGES AND DISADVANTAGES OF THEORETICAL APPROACHES Theoretical approaches to design asON are mainly based on computational analyses of DNA–RNA duplex stabilities or on predictions of secondary structures of a given target RNA sequence in order to search for local sites that are accessible for the invasion of a complementary asON strand (e.g. Amarzguioui et al., 2000; Toschi, 2000; Vickers et al., 2000; Ding and Lawrence, 2001). The major disadvantage of theoretical approaches is related to the lack of reliability of RNA structure predictions and the lack of mechanistic insights into structural requirements for efficient RNA–DNA annealing. Conversely, the basic advantages which favor computational approaches over experimental ones include the possibility of generating and investigating the complete relevant antisense sequence space (Sczakiel, 2000). This means that all possible antisense species can be defined and they can be searched for those which meet specific criteria that may be related to increased efficacy. Similary, large numbers of relevant local target structures along a given target RNA c Oxford University Press 2001
Automation of effective antisense oligonucleotides
sequence can be identified, recorded and analyzed computationally for their suitability to serve as favorable entry sites for antisense strands. In principle, rules that govern the annealing between long-chain RNA and oligomeric complementary DNA may be included into the criteria for which computer algorithms may search in the antisense sequence space and the target space, respectively. However, automated theoretical approaches may only reach a level of confidence that is determined by the reliability of RNA structure prediction and the knowledge on the structure–function relationship of asON. This kind of knowledge may be substantially increased by combining theoretical ‘high throughput’ algorithms for the design of effective asON with experimental analyses. It should be noted that the structure–function relationship of asON is not trivial since, for example, very few base exchanges may convert efficient antisense species into inefficient ones and vice versa. This cannot be simply explained by the base composition nor by the thermodynamic DNA/RNA duplex stability or by characteristics of the target structures (e.g. Milner et al., 1997; Lima et al., 1997; Ho et al., 1998; Patzel et al., 1999).
PARAMETERS THAT INFLUENCE THE EFFECTIVENESS OF asON IN LIVING SYSTEMS Properties of the asON species and the RNA target sequence, respectively, that are directly related to their primary sequence or secondary structure may substantially affect double-strand formation and, hence, the efficacy of asON in living cells (Figure 1). These parameters can be subdivided into major classes: firstly, those which are related to intrinsic properties of the specific nucleotide sequence of the asON such as the potential for dimer formation, for intramolecular folding, or the presence of specific sequence motifs that may be positively or negatively related to efficacy. Secondly, the kind of chemical modification of asON may strongly influence duplex formation and stability as well as their biological properties (Seeberger and Caruthers, 1998). Thirdly, local structural characteristics of the target RNA may favor or hinder the binding of asON strands and, fourthly, the kinetics of the formation of the DNA–RNA duplex and its stability may be related to the phenotypic effectiveness of asON. Regarding the duplex formation it has been shown that kinetic as well as thermodynamic parameters are related to the extent of inhibition of asON in living cells. For example, fast association of complementary nucleic acids which is thought to directly reflect favorable local target RNA structures is directly linked to efficacy of asON in living systems. On the other hand, the potential of asON strands to form intramolecular structures and dimers or oligomers is regarded to be negatively related to their annealing properties. Finally, it is reasonable to
asON dimer formation intramolecular folding specific nucleotide motifs chemistry
target RNA accessibility
asON/RNA duplex annealing kinetics
duplex stability
Fig. 1. Schematic representation of double-strand formation between a given target RNA and an asON strand and the parameters which are known to influence this process.
assume that the thermodynamic stability of the DNA– RNA duplex influences the extent of asON-mediated inhibition provided that a certain threshold value for sufficient stability is not reached. The search for sequence motifs that are common among successful asON revealed some interesting statistical correlations (Smetsers et al., 1996; Matveeva et al., 2000). For example, Tu et al. (1998) found that a GGGA motif in the target sequence of the tumor necrosis factor-α (TNF-α) and TCCC in the asON, respectively, occurs at a statistically increased frequency in case of effective antisense ON. Ratmeyer et al. (1994) suggested that stretches of ribopurines within the target can stabilize DNA–RNA heteroduplexes and concluded that these represent favorable local targets for asON. Meanwhile, more systematic analyses by Matveeva et al. (2000) reveal a number of sequence elements that may be positively (e.g. CCAC, TCCC) or negatively (e.g. ACTG, GGGG) related with effective asON molecules. Biological characteristics of asON that may be related to specific sequence motifs include the cellular uptake, the subcellular localization, immunostimulation, and the properties to serve as a substrate for RNaseH when bound to the target RNA. It is noteworthy that almost all parameters summarized here can be calculated, recorded and analyzed by appropriate computer algorithms. Similarly, the potential of the formation of dimers and of intramolecular structures of asON sequences may be computer-calculated by a number of existing programmes which include for example ‘Oligo’ (Rychlik and Rhoads, 1989). To calculate duplex stabilities between asON and target RNA, Mathews et al. (1999a) have developed the computer algorithm ‘oligo walk’ that scans along a given target RNA sequence and calculates thermodynamic stabilities of the relevant double-strands based on thermodynamic parameters described by Sugimoto et al. (1995). 1059
R.K.-K.Far et al.
FLOW CHART OF A COMPUTER-AIDED DESIGN OF asON The theoretical design of effective asON comprises the following steps which may be improved or executed by computer algorithms: (1) Computer calculation of RNA secondary structure of the target: several computer programmes already exist and are continously updated (e.g. Mathews et al., 1999b; Zuker, 2000). (2) Identification of favorable local RNA secondary structure motifs. (3) Definition of asON sequences complementary to favorable local target segments. (4) Consideration of specific sequence motifs: — exclusion of negatively related motifs;
Fig. 2. A local target motif along a long-chain target RNA (upper panel) may adopt secondary structures (lower panel) that are favorable for the invasion of a complementary asON molecule.
— inclusion of favorable motifs. (5) Exclusion of asON sequences with a high potential for intramolecular and intermolecular folding.
CONSIDERATION OF THERMODYNAMIC PARAMETERS, DNA–RNA DUPLEX STABILITY To look at the role of target RNA structures and to test the potential use of computer-based algorithms we recently tested a theoretical approach to design effective asON species against ICAM-1 target sequences (Patzel et al., 1999; Kretschmer-Kazemi Far et al., unpublished). Extensive sets of secondary structures were generated for overlapping segments of target sequences using the program mfold (Zuker, 2000) and families of lowest energy structures were recorded and compared. Local target structures which were consistently predicted among overlapping target segments, and which were found in most of the lowest energy structures were chosen to design complementary asON. Local secondary structures which were regarded to be favorable for intervening asON sequences include terminal and internal loops, joint sequences, and bulges of 10 or more consecutive nucleotides (Figure 2). This strategy gave rise to a clearly higher than statistical rate of success: a number of 17 out of 34 ICAM-1-directed asON species showed significant inhibition (>50%) of ICAM-1 expression in the cell line ECV304 which is significantly increased when compared to a usual average rate of success of asON of approximately 5 to 10% (Patzel et al., 1999). As indicated in Figure 2 favorable local target motifs do not necessarily have to adopt a single defined structure. It is important, that the local target segment of interest is not involved in intramolecular folding. If this is the case (see lower panel of Figure 2) one is able to design effective asON species. It is important to note, 1060
however, that the characteristic parameters used in this approach affects its reliability. For example, the length of target sequence segments used for computer-aided folding, the step width of overlapping segments, the number of lowest energy structures that are considered, the frequency of the occurrence of calculated target motifs, and the ‘stringency’ of the selection have been shown to influence the design of asON. Further, the specific algorithms used to calculate secondary structures of RNA may affect the results of this theoretical design of asON. It does not influence, however, the basic considerations described here. From the inclusion of substantially more automated steps, we expect to analyze and reveal the role of the above mentioned parameters.
REFERENCES Amarzguioui,M., Brede,G., Babaie,E., Grotli,M., Sproat,B. and Prydz,H. (2000) Secondary structure prediction and in vitro accessibility of mRNA as tools in the selection of target sites for ribozymes. Nucleic Acids Res., 28, 4113–4124. Ding,Y. and Lawrence,C.E. (2001) Statistical prediction of singlestranded regions in RNA secondary structure and application to predicting effective antisense target sites and beyond. Nucleic Acids Res., 29, 1034–1046. Ho,S.P., Bao,Y., Lesher,T., Malhotra,R., Ma,L.Y., Fluharty,S.J. and Sakai,R.R. (1998) Mapping of RNA accessible sites for antisense experiments with oligonucleotide libraries. Nature Biotechnol., 16, 59–63. Lima,W.F., Brown-Driver,V., Fox,M., Hanecak,R. and Bruice,T.W. (1997) Combinatorial screening and rational optimization for hybridization to folded hepatitis C virus RNA of oligonucleotides with biological antisense activity. J. Biol. Chem., 272, 626–638. Mathews,D.H., Burkard,M.E., Freier,S.M., Wyatt,J.R. and Turner,D.H. (1999a) Predicting oligonucleotide affinity to nucleic acid targets. RNA, 5, 1458–1469.
Automation of effective antisense oligonucleotides
Mathews,D.H., Sabina,J., Zuker,M. and Turner,D.H. (1999b) Expanded sequence dependence of thermodynamic parameters improves prediction of RNA secondary structure. J. Mol. Biol., 288, 911–940. Matveeva,O.V., Tsodikov,A.D., Giddings,M., Freier,S.M., Wyatt,J.R., Spiridinov,A.M., Shabalina,S.A., Gesteland,R.F. and Atkins,J.F. (2000) Identification of sequence motifs in oligonucleotides whose presence is correlated with antisense activity. Nucleic Acids Res., 28, 2862–2865. Milner,N., Mir,K.U. and Southern,E.M. (1997) Selecting effective antisense reagents on combinatorial oligonucleotide arrays. Nature Biotechnol., 15, 537–541. Monia,B.P., Johnston,J.F., Geiger,T., M¨uller,M. and Fabbro,D. (1996) Antitumor activity of a phosphorothioate antisense oligodeoxynucleotide targeted against C-raf kinase. Nature Med., 2, 668–675. Patzel,V., Steidl,U., Kronenwett,R., Haas,R. and Sczakiel,G. (1999) A theoretical approach to select effective antisense oligodeoxyribonucleotides at high statistical probability. Nucleic Acids Res., 27, 4328–4334. Peyman,A., Helsberg,M., Kretzschmar,G., Mag,M., Grabley,S. and Uhlmann,E. (1995) Inhibition of viral growth by antisense oligonucleotides directed against the IE110 and the UL30 mRNA of herpes simplex virus type-1. Biol. Chem. Hoppe-Seyler, 367, 195–198. Ratmeyer,L., Vinayak,R., Zhong,Y.Y., Zon,G. and Wilson,W.D. (1994) Sequence specific thermodynamic and structural properties for DNA.RNA duplexes. Biochemistry, 33, 5298–5304. Rychlik,W. and Rhoads,R.E. (1989) A computer program for choosing optimal oligonucleotides for filter hybridization, sequenceing
and in vitro amplification of DNA. Nucleic Acids Res., 21, 8543– 8551. Sczakiel,G. (2000) Theoretical and experimental approaches to design effective antisense oligonucleotides. Front. Biosci., 5, 194–201. Seeberger,P.H. and Caruthers,M.H. (1998) A modified oligodeoxynucleotides as antisense therapeutics. In Stein,C.A. and Krieg,M. (eds), Applied Antisense Oligonucleotide Technologies. Wiley–Liss, New York, pp. 51–71. Smetsers,T.F., Boezeman,J.B. and Mensink,E.J. (1996) Bias in nucleotide composition of antisense oligonucleotides. Antisense Nucleic Acid Drug Dev., 6, 63–67. Stein,C.A. and Krieg,M. (eds) (1998) Applied Antisense Oligonucleotide Technologies. Wiley–Liss, New York. Sugimoto,N., Nakano,S., Katoh,A., Makamura,H., Ohmichi,T., Yoneyama,M. and Sasaki,M. (1995) Thermodynamic parameters to predict the stability of RNA/DNA hybrid duplexes. Biochemistry, 34, 11 211–11 216. Toschi,N. (2000) Influence of mRNA self-structure on hybridization: computational tools for antisense sequence selection. Methods, 22, 261–269. Tu,G.C., Cao,Q.N., Zhou,F. and Israel,Y. (1998) Tetranucleotide GGGA motif in primary RNA transcripts. Novel target site for antisense design. J. Biol. Chem., 273, 25 125–25 131. Vickers,T.A., Wyatt,J.R. and Freier,S.M. (2000) Effects of RNA secondary structure on cellular antisense activity. Nucleic Acids Res., 28, 1340–1347. Zuker,M. (2000) Calculating nucleic acid secondary structure. Curr. Opin. Struct. Biol., 10, 303–310.
1061
