WO2004099411A1

WO2004099411A1 - A method for dissecting the rna secondary structure dependence of rna-ligand interactions

Info

Publication number: WO2004099411A1
Application number: PCT/EP2004/004909
Authority: WO
Inventors: Manfred Auer; Joerg Hackermueller; Markus Jaritz; Nicole-Claudia Meisner
Original assignee: Novartis Ag; Novartis Pharma Gmbh
Priority date: 2003-05-07
Filing date: 2004-05-07
Publication date: 2004-11-18

Abstract

A method for detecting the influence of a secondary structure element of an RNA molecule to the binding of a ligand, its application for selecting RNA-ligand interactions as pharmaceutical target and its pharmaceutical use.

Description

A method for dissecting the RNA secondary structure dependence of RNA-ligand interactions

The present invention relates to a method for dissecting the RNA secondary structure dependence of RNA-ligand interactions, its application for selecting RNA-ligand interactions as pharmaceutical target and its pharmaceutical use.

RNA-ligand interactions are known to be of central importance in the regulation of eukaryotic gene expression. Tightly controlled processes, particularly in posttranscriptional regulation are essentially dependent on the recognition of cis acting RNA elements by ligands, e.g. protein mediators. Beside generic interactions like cap- or poly(A) binding, such interactions occur in a highly specific and controlled manner. Several examples have been described wherein recognition relies not only on pure primary structure but also involves higher order sequence properties, e.g. RNA secondary structure. Many disease relevant genes are regulated predominantly at the posttranscriptional level, including regulatory processes such as such as pre-mRNA splicing, nuclear export, message degradation and translatability. Hence, targeting regulatory RNA-ligand interactions through RNA secondary structure manipulation may serve as a novel strategy for therapeutic intervention.

The complex formation of an RNA, e.g. an ARE-containing mRNA, with a ligand, e.g. a protein, e.g. an HuR protein (ELAVL1) may control the expression of various disease causing/mediating substances, e.g. inflammatory acting substances, such as cytokines, growth factors, proto-oncogenes or viral proteins. Agents which inhibit such a complex formation may thus prevent the expression of such substances, e.g. such agent may prevent or reduce the expression of inflammation mediating substance. Therefore such agents (inhibitors) may be used in the treatment of various diseases, e.g. diseases mediated by cytokines, growth factors, proto-oncogenes or viral proteins.

We have developed a method for dissecting the RNA secondary structure dependence of RNA-ligand interactions by combining RNA secondary structure prediction and affinity measurements with a new computational algorithm. The method allows to (/) determine whether and which RNA secondary structure element is required for the recognition by a ligand, (ii) to quantitatively predict RNA ligand affinities and (Hi) to design RNA sequences with a particular ligand affinity. In consequence, this method provides a means for selection of RNA secondary structure dependent pharmaceutical targets and for the rational design of RNA secondary structure based, target specific RNA-ligand interaction assays.

According to the present invention - RNA as used herein includes all kinds of RNAs, e.g. ribosomal RNA (rRNA), transfer RNA (tRNA) or messenger RNA (mRNA), e.g. of a cytokine, growth factor, interleukin, cyclin, apoptotic protein, hormone, differentiation factor or a viral protein, preferably a mRNA, e.g. including ERGs. For example, the mRNA is an ARE-containing mRNA, e.g. of IL-2 or TNF-α. - a sequence binding motif M is a sequence pattern which is compatible with the RNA sequence s and which is required for the binding to the ligand L. a secondary structure element (M, σ ) of an RNA is a sequence motif M in a particular conformation σ , which is defined as a set of base pairs (positions i, j) where both or either of i, j are within the sequence binding motif M of the RNA sequence. Such base pairs must fulfill the non-pseudoknot condition. This implies that, if the bases are numbered from 5' to 3' end, any pair of base pairs ((i, j), (k, I)) are either successive (i<j<k<l) or nested (i<k<l<j) but must not cross (i<k<j<l) (Figure 2). An RNA-binding ligand L, e.g. a protein, is understood to include any ligand which binds via a binding motif M to said RNA, e.g. including an ELAV1 (=HuR)-protein, which binds via a binding motif M to said RNA, e.g. mRNA, wherein the binding is dependent on a secondary structure element of said mRNA. recognition by a ligand L of the present invention requires that motif M adopts an active conformation σf . A change in this conformation has an impact on the recognition of said element by a ligand, e.g. the interaction of a ligand with the RNA may be lowered (silenced).

In one aspect the present invention provides a method for detecting an influence of a secondary structure element of an RNA molecule on the binding of a ligand comprising (a) providing an RNA-binding ligand L, (b) providing at least 3 different RNAs of sequences Si s₂, and s₃, each of s-i s₂, and s₃ having the same sequence binding motif M for the ligand L, but each differing in the secondary structure, (c) calculating each p(s,M^s ,σf ) for each s-i s₂, and s₃ which is the probability that a structure contains at least one match of M in a particular test conformation σ , e.g. calculating each p(s,M^s ,σ according to the equation [7], as set out below,

(d) measuring each macroscopic affinity constant Kd_app for the binding of L to each of s^ s₂, and s₃,

(e) determining for any test conformation σ^M whether the correlation between each of Kdg_pp and p(s,M^s ,σf ) for each of s-,, s₂ and s₃ follows the dependency

Kd_app ∞ \lp(s ,M^s,σf ) and (f) choosing an RNA secondary structure element (M^s,σ^) which follows said dependency.

An RNA molecule of sequence s will adopt many secondary structures. The set of all secondary structures is called ensemble, ε(s), and the frequency of each individual structure σ_i in the ensemble is determined by its stability, i.e. its energy E(s,σ P) . The frequency of a particular structure σ . can be calculated using Boltzmann's law:

Q_s is called the partition function of the secondary structure ensemble of sequence s, β = \l RT The probability of a particular structure element (M,σ ) , e.g. the probability to find NNUUNNUUU in single stranded conformation, in the ensemble is the probability of a set of structures. As the probabilities of individual structures are roughly independent, the probability of the structure element (M,σ ) is simply the sum of the probabilities of all structures containing element (M, σf ) ,

∑exp(-E(σ_;,s)/?) p(s,M_i ^s,σ ) = ^{σ {M σ}' ⁾'^s) [2]

where ε[(M_f ,σf),s) denotes the ensemble of structures of sequence s constrained to structures containing (M,σJ) and Q ._M, _σγ is the respective partition function.

Equation 2 is valid for structure elements which have only one match of motif M per sequence, i.e the structure element may occur only once. Most elements of interest may occur repeatedly and possibly overlap. For a correlation with affinity data, only the discrimination accessible and not-accessible will be of interest. Consequently the definition of p is extended to p(s,M^s,σf ) , the probability of structures containing at least one element ( ,σ_; ) . A computation via the probability of its complement is not possible due to limitations for constrained folding. As the probabilities of substructures that may occur together are not independent, the sum of the individual probabilities has to be corrected for joint occurrences of (M,σ ) :

{s,(M_k ^s,σf¹ )A(M_l ^s,σf))

+ ∑p(s,(M_k ^s,σ?) Λ (MΪ,σ?)Λ (M ,σf '))-... [4] k<l<m<n where

(M_t ^s ,σ ) is the probability of structures which adopt conformation <ή for the k^th and I^th match of motif M, p{s,(M_k ^s,σf ) Λ (Mf,σf ) A (M_m ^s ,σ )) denotes the probability of structures which adopt conformation σ for the k^th, I^th and m^th match of M, etc.

Analogously to equation 3, the corrective terms in equation 4 can be expressed by partition functions of multi-constrained ensembles: _y(s M^s σ^M ₎ = Y Q'Wl^{j )} y Q>MrfMi*f _[ y Q*W fWntfW.J^}) k<n zis k<l<n i s k<l<m<n \£s where Q „„ _{W I w}, denotes the partition function of the secondary structure ensemble of

s constrained to structures which adopt conformation σf for the k^th and I^th match of motif M, Q ,_x.χ « _»,v _{M ι}„ _M, is the partition function of structures which adopt conformation

σf for the k^th, I^th and m^th match of M, etc. To minimize numerical errors, the probabilities are calculated via the ensemble free energies W_s of the ensemble of sequence s.

W_s = -RTΪnQ_s [6]

W_s „ _σM denotes the free energy of the ensemble constrained to structures containing the

k^h match of motif M in conformation σ . Multiple constraints are denoted analogously to the constrained partition functions as described for equation 5.

Ensemble free energies are computed using the Perl module of RNAfold which is part of the Vienna RNA Package (www.tbi.univie.ac.at/RNA/). The calculation of W_s is straightforward;

W_{s s σ}u is calculated by constraining the sequence positions corresponding to (M- ,σ ) to the particular structure.

Pseudocode for the calculation of p: calc_p (sequence s, motif_position_list, motif_struct) wg = calc_enemle_free_energy (s) p = 0 for I (1 .. number_of_motifs) w_c = calc_ensemle_free_energy (s, motif I, motif_position_list, motif_struct) p⁺ = exp ((w_g - w_c)/kT) push (constrainjist, i) p⁺ = recurse (constrainjist, w_g) next return p recurse (constrainjist, w_g) pi = 0 for j (last_element (constrainjist) .. number_of_motifs) push (constrainjist, j) w_c = calculate_ensemle_free_energy (s, constarinjist, motif_positionJist, motif_struct) if (numer_elements (constrainjist)%2=0) sign = +1 else sign = -1 pj+ = sign * exp ((w_g - w_c)/kT) if ( <number_of_motifs) pj + = recurse (constrainjist, w_g) next return pj

Macroscopic affinity constants Kd_app for the binding of a ligand L to a given RNA may be determined according to known methods, e.g. methods as conventional, such as e.g. EMSA (= electrophoretic mobility shift assay), filter binding assays, ELISA (= enzyme linked immunosorbant assay), fluorescence spectroscopy with a particular focus on applications with single molecule sensitivity e.g. Fluorescence Correlation Spectroscopy (FCS), Fluorescence Intensity Distribution Analysis (FIDA), or applications based on the determination of Fluorescence Anisotropy or Fluorescence Resonance Energy Transfer (FRET), e.g. as described in Kask P. et al, Biophys. J. (2000) 78 (4), 1703-1713.

RNA molecules of equal primary sequence Sj may fold into many different secondary structures s-i, s₂, s₃, etc., the frequency of each structure is dependent on its stability, i.e. its energy. If we anticipate that a ligand binds only to RNA molecules which adopt a specific secondary structure element and assume a 1:1 binding mechanism, the interaction can be described as Kd_fu_πCj wherein Kd u_nd is expressed in the following equation

Kd_fu_n = [RNA_acc] ■ [ligand_free] / [complex] [I]

In equation I, [ligand], [complex] denote the respective equilibrium concentrations, and [RNA_ac_c] is the concentration of free RNA molecules adopting the fold required for binding; that means RNA molecules having a secondary structure element required for binding to a ligand L.

If the secondary structure equilibrium is reached sufficiently fast, the concentration [RNA_aCc] is expressed in the following equation

[RNAacc] = [RNA_free] . p(s,M^s, σ^M ) [II] In equation II, [RNA_fr_Θe] is the concentration of unbound RNA, that means RNA having no structure element required for binding to a ligand, and p(s,M^s,σ ) is the thermodynamic probability of structures containing at least one sequence binding motif M in a particular test conformation σf as defined before.

An experimental determination of the binding affinity between RNA and ligand does not take into account a possible discrepancy between [RNA_acc] and [RNA_free] and will consequently overestimate the dissociation constant (underestimate the affinity) Kd_apPι which Kd_app may be expressed by the following equation:

Kd_app= [RNA_free] * [ligand_free] / [complex] [III] wherein [RNA_free], [ligand_free] and [complex] are as defined above.

If [RNA_free] is substituted according to equation II by [RNA_acc] / p(s,M" ,σf ) and inserted in equation I, then Kd_app can be expressed by the following equation:

Kd_app = Kd _ϋ / p(s,M^s, σ^f ) [IV] We have confirmed that, if a specific RNA secondary structure element is required for the binding to the ligand L, a correlation between measured macroscopic affinity data and computed thermodynamic probabilities of a putatively required secondary structure element will follow the deduced reciprocal relationship expressed in equation IV

We have found, that, if the dependency as set out in equation [IV] is fulfilled, the secondary structure element (M,σf ) determined in (e) is required for the binding process with a ligand L.

The method according to the present invention is thus appropriate to find out -whether a secondary structure element is of importance in the binding process of an RNA to a ligand L and

-which secondary structure elements are of importance in the binding process.

The knowledge that a particular secondary structure element is responsible for recognition by a regulatory ligand is e.g. of immediate relevance for assay development.

In another aspect the present invention provides a method of the present invention comprising the steps (a) to (e) and further comprising the steps (g) determining the fundamental affinity constant Kd_fund of L to M in an active conformation σf based on the equation Kd_app = Kd_fund / p(s,M^s ,σ ) , and (h) (optionally) providing, e.g. novel, sequences with a defined Kd_app for the ligand L.

Kd_fUnd, which describes the microscopic affinity constant to the binding motif M in the active conformation σ„ is often not experimentally ascertainable. Using the method of the present invention, Kd_fund can be determined by curve fitting of apparent macroscopic binding data (Kdgp_p) to RNA molecules of different affinity.

Provided that Kd_fund is known, e.g. determined by a method as described above, Kd_fund may be used as a basis to design novel sequences with a pre-defined Kd_app for the ligand L, e.g. a protein, i.e. by iteratively calculating p(s,M^s,σ^)for novel sequences, e.g. random sequences, with stepwise modification, e.g. random modification. This method thus provides a handle for dissecting the RNA recognition mechanism by a ligand, e.g. a protein, e.g. HuR.

In another aspect the present invention provides a process according to the present invention comprising the steps (a) to (e) and further comprising step (i) screening e.g. computationally, for additional RNAs bound by the ligand L, e.g. a ligand known to bind desease relevant RNAs, based on the occurrence of M in an active conformation σ_a M

In a further aspect the present invention provides the use of a method of the present invention for the identification of a disease-relevant target.

The potential to predict affinities to a certain ligand and to dissect how the affinity can be manipulated is of prime significance for selecting an RNA-ligand interaction as a pharmaceutical target. Disease relevant targets (or mediators) include e.g. proto-oncogenes, inflammatory cytokines and viral proteins.

In another aspect the present invention provides the use of a method of the present invention for the manipulation of an RNA.

Manipulation includes e.g. inducing a change in the conformation of an RNA, e.g forming or disrupting a secondary structure element, by a nucleic acid manipulator, e.g. an aptamer or e.g. an antisense, to enhance or decrease binding to a given ligand, e.g. protein.

In another aspect the present invention provides the use of a method of the present invention for the identification of an agent for the manipulation of an RNA.

In a further aspect the present invention provides an agent as described above for pharmaceutical use. In another aspect the present invention provides a pharmaceutical composition comprising an agent of the present invention which agent is identified by a method of the present invention, in association with at least one pharmaceutical excipient.

In another aspect the present invention provides the use of an agent of the present invention or of a pharmaceutical composition according to the present invention for the treatment of a disorder having an etiology associated with the production of a substance, e.g. an inflammatory acting (causing/enhancing) substance, selected from the group consisting of cytokines, growth factors, proto-oncogenes or viral proteins.

Preferably said substance is selected from the group encompassing the products of the genes BMP6, CCL11 , CSF2, IL1b, IL2, IL3, IL4, 1L6, IL8, MYOD1, MYOG, NF1 , PITX2, TNFa, VEGF, CCNA2, CCNB1 , CCND1 , CCND2, CDKN1A, CDKN1 B, DEK, FOS, HLF, JUN, MYC, MYCN, TP53, HDAC2, MMP9, NDUFB6, NOS2A, PLAU, PTGS2, REN, SERPINB2, UBE2N, ADRB1 , ADRB2, AR, CALCR, CDH2, GAP43, SLC2A1 , PLAUR, SLC5A1 , TNFSF5, ACTG1 , CTNNB1 , MARCKS, MTA1 , PITX2, SLC7A1.

For use of an agent of the present invention in the treatment of a disease, an agent may include one or more agents, e.g. a combination of agents of the present invention. The pharmaceutical compositions according to the present invention may be useful for the treatment of a disorder having an etiology associated with the production of a substance, e.g. an inflammatory acting (causing/enhancing) substance, selected from the group consisting of cytokines, growth factors, proto-oncogenes or viral proteins. Preferably said substances are those as described above. Treatment includes treatment and prophylaxis.

Description of the figures:

Figure 1: Correlation between accessibility of NNUUNNUUU within AREs

{p(s,M^s ,σ^ ) ) and macroscopic Kd (Kd_app).

Experimentally determined affinities of HuR to the AREs of Cox-2 (1), lL-1 β (2), IL-2 (3), IL-4 (4), IL-8 (5), to (AUUU)₃A (10), (AUUUy\ (11), (AUUU)₅A (12), (CUUU)₄C (13) (black circle), to the 3'UTRs of IL-2 and TNFα (14 and 15, crosses) and to the ARE of TNFα (6) as well as the strategically designed variants TNFα'(WT) (7), TNFα'(NZW) (8), and TNFα'mut (9) (black diamond) and their predicted values (open diamond) are plotted against the corresponding accessibilities of the HuR binding motif (M=NNUUNNUUU, σ = all single stranded), p(ssNNUUNNUUU)= p(s,M^s ,σ ) , which is the thermodynamic probability of structures in the ensemble which contain at least NNUUNNUUU sequence match in single stranded conformation. HuR binds to NNUUNNUUU elements with an almost invariable Kd, Kd_fUnd. The correlation between measured affinities Kd_app and calculated accessibilities p(ssNNUUNNUUU) follows the derived reciprocal dependence Kd_app = Kd_fuπd I p(ssNNUUNNUUU). The dotted line depicts the resulting graph for approximation Kd_fundby the experimentally determined value of 0.96 (±0.48) nM (Fig.1). The solid line depicts the corresponding graph obtained by non-linear curve fitting of the data

0.49 +0.09 1 p).

Figure 2: Base pairs in secondary structure elements must not cross. Base pairs must fulfill the non-pseudoknot condition. This implies that, if the bases are numbered from 5' to 3' end, any pair of base pairs ((i, j), (k, I)) are either successive (i<j<k<l) (blue) or nested (i<k<l<j) (green) but must not cross (i<k<j<l) (red, dotted line).

In the following examples all temperatures are given in degree Celsius and are uncorrected.

The following ABBREVIATIONS are used:

IPTG isopropyl-β-D-thiogalaktopyranoside

ORN oligoribonucleotide

RT room temperature RP-HPLC Reversed-Phase High Performance Liquid Chromatography

SDS-PAGE sodium dodecylsulfate polyacrylamide gel electrophoresis

TFA trifluoracetic acid

THF tetrahydrofuran TEAAc triethylammonium acetate

TMR carboxytetramethylrhodamine

s an RNA sequence M an RNA sequence motif

M- i^th occurrence of M in s σ . secondary structure (of an RNA sequence) ε(s) secondary structure ensemble of sequence s σ secondary structure corresponding to motif M

oy a particular (active) conformation corresponding to M necessary for recognition by a ligand E(σ_j,s) free energy of structure σ on sequence s

(M,σ ) = A secondary structure element (motif M in conformation σ )

(M,σ^) = A_a secondary structure element (motif M in conformation c , e.g. NNUUNNUUU in single stranded conformation)

(M ,σ ) = A. the i^th occurrence of M in sequence s is in conformation σ p(s,σ_j) thermodynamic probability of structure σ_;. on sequence s p(s,M. ,σf ) = p(s,A_a ^s _i) the thermodynamic probability of the i^th occurrence of A_a in s p(s,M^s,σ ) = p(s,A_a) the thermodynamic probability of at least one occurrence of A_a in s , i.e. at least one occurrence of motif M is in conformation σ^M

EXA PLE 1: Secondary structure dependence of HuR-RNA recognition mRNAs from many disease-relevant early-response genes (ERGs) are targeted for specific degradation by the presence of AREs in the 3'-untranslated region (UTR), mediated by trans-acting factors or proteins binding to them (see e.g. Chen C.Y et al., Trends Biochem. Sci. 1995, 20(11):465-70). To date, several cytoplasmic mRNA-binding proteins have been identified to specifically interact with the ARE, whereas their binding shows either stabilizing, destabilizing or shuttling effects. Among these, the human ELAV (embryonic- lethal/abnormal-vision)-protein HuR (Hu-Antigen R) is proposed to be the central mRNA stabilizing protein involved in ARE-mediated mRNA degradation pathways (see e.g. Peng S.S. et al., EMBO J. 1998, 17(12):3461-70).

We have experimentally verified the method of the present invention using affinity data of the interaction between the mRNA stability regulator HuR and several natural and artificial target RNAs. This regulatory system is of particular relevance as HuR is a very specific positive regulator of potentially several thousand genes, many of them being disease related. A correlation between measured affinity data and predicted probability clearly shows reciprocal dependence and hence indicates that mRNA secondary structure dominates the target recognition by HuR. The experimental verification of affinities predicted for designed sequences further substantiate the role of secondary structure (elements) in this interaction.

a) Providing an (m)RNA binding ligand L: Preparation of recombinant human HuR:

Recombinant full-length human HuR is prepared using the IMPACT™ -CN system (New England Biolabs) as follows: the nucleotide sequence encompassing amino acids 1-326 of the CDS [RefSeq accession: NP_001410] is PCR-amplified from cDNA prepared from activated human T-Lymphocytes and cloned directionally into the Ndel and Sapl sites of the vector pTXBI , allowing C-terminal fusion with an intein- CBD (Chitin binding fomain) tag without insertion of any additional amino acid.

These constructs are transformed into the host strain E.coli ER2566 (New England Biolabs) for protein expression. The DNA sequence of the constructs recovered from positive clones is verified by automated DNA sequencing on an ABI 310 instrument, according to the manufacturer's protocol.

Expression of the fusion proteins is induced by addition of 1 mM IPTG to a bacterial culture grown to late-logarithmic phase in LB (Lurea Bertani) broth and allowed to proceed for 6 hours at 28 °. The bacterial cells are lysed by successive freezing/thawing cycles in a buffer of 20 mM Tris/CI (Tris(hydroxymethyl)aminomethane) pH 8.0, 800 mM NaCI, 1 mM EDTA (N,N,N',N'-Ethylenediaminetetraacetic) and 0.2 % Pluronic F-127 (Molecular Probes). After DNA digestion, the bacterial lysates are cleared by ultracentrifugation and the fusion protein is captured onto Chitin agarose beads (New England Biolabs) via the CBD. After extensive washing of the beads with lysis buffer, the recombinant protein is recovered by thiol-induced on-column self-splicing of the intein tag by addition of Na-2-MESNA (2-mercaptoethane- sulfonic acid, sodium salt) to a final concentration of 50 mM and incubation for 12 hours at 4° (see e.g. New England Biolabs, 2002). Any co-eluted intein tag and uncleaved fusion protein are removed from the eluate in a second, subtractive affinity step. Finally, the protein is transferred into the appropriate storage buffer by elution through a gel filtration column (DG-10, Bio-Rad) previously equilibrated with the target buffer (25 mM Na₂HPO₄/NaH₂PO₄ pH 7.2, 800 mM NaCI, 0.2 % (w/v) Pluronic F-127), shock-frozen in small aliquots in liquid nitrogen and stored at -80 °.

The accurate concentration of HuR is determined as follows: a sample containing ~200 μg HuR is purified on a VYDAC C₁₈ RP-HPLC column and eluted in a gradient of 100 % eluent A - 100 % eluent B in 40 minutes (A: 5 % CH₃CN, 95 % H₂O, 0.1 % (w/v) TFA; B: 80 % CH₃CN, 20 % H₂O, 0.1 % (w/v) TFA) with UV-detection at 280 nm and fluorescence detection (280 nm excitation/340 nm emission). The eluted fractions are pooled, lyophilised and dissolved at a concentration of ~ 20 μM in PBS (phosphate buffered saline; 1.5 mM KH₂PO₄, 8.1 mM Na₂HPO₄, 136 NaCI, 2.7 KCI, pH 7.2) containing 6M guanidinium ■ HCI. After determination of the concentration by UV-spectroscopy using the theoretical molar extinction coefficient for HuR at 280 nm in 6M guanidinium hydrochloride, this stock solution serves as external standard for a subsequent concentration determination of HuR samples by conventional RP-HPLC quantification. The quality of the purified protein is controlled by denaturing SDS-PAGE, UV-spectroscopy, analytical Size Exclusion Chromatography, RP- HPLC analysis, N-terminal sequencing, LC/ESI-MS (Liquid Chromatography/Electrospray lonization-Mass Spectrometry) analysis and CD- (Circular Dichroism-) spectroscopy and western blotting with mouse monoclonal anti-HuR 19F12 (Molecular Probes), following standard protocols. According to LC/ESI-MS and RP-HPLC analysis, all HuR preparations are >98% pure. N-terminal sequencing revealed the correct N-terminus but quantitatively missing Met-i. Results from Analytical Size Exclusion Chromatography further indicated that the protein is in a soluble state without presence of higher aggregation states. (b) Providing RNA sequences: Preparation of fluorescently labeled RNA:

RNA synthesis is performed on an Applied Biosystems 394A synthesizer using either 5'-O- dimethoxytrityl-2'O-triisopropyloxymethyl (TOM)- protected β'-cyanoethyl-(N,N-diisopropyl-) nucleotide phosphoramidites (0.1 M solutions in anhydrous acetonitrile; Glen Research) and the appropriate 2'TOM-protected nucleotides immobilised on CPG (controlled pore glass), adopting published procedures (see e.g. Chaix C. et al., Nucleic Acids Symp. Ser. 45, 1989 or Scaringe S.A. et al., Nucleic Acids Res. 18, 5433, 1990) and manufacturer's protocols. In the final step of each synthesis, an aminolinker, (6-(4-monomethoxytrityIamino)hexyl-(2- cyanoethyl)-(N,N-diisopropyl)), is attached to the 5'-OH group of the ORN in order to allow the coupling with a range of different dyes in a post-synthesis reaction. The synthesized ORNs are cleaved from the support, base-and phosphate-deprotected with ammonia-saturated ethanol for 17 hours at 40°. The 2' silyl-protection groups are removed by reaction in 1.1 M tetrabutylammonium fluoride/THF for 15 hours at RT. The deprotected ORNs are purified by denaturing polyacrylamide gel electrophoresis following standard protocols.

For the determination of the concentration, the exact molar extinction coefficient at 260 nm is determined according to e.g. Gray D.M. et al., Methods in Enzymology 246, 19, 1995 and concentrations are calculated based on the measured UV-absorption at 260 nm, according to Beer's Law. The purity and quality of the fragments is controlled by analytical RP-HPLC on a VYDAC C₁₈ column with elution in a gradient of 0 to 100 % B in 45 minutes (A: 5% 0.1 M TEAAc pH 7.0, 95 % CH₃CN; B: 50 % 0.1 M TEAAc pH 7.0, 50 % CH₃CN) and UV- detection at 260 nm. TMR as one of the best characterized and photostable dyes is attached to the aminogroup introduced with the 5-'terminal linker by a standard reaction of the primary amine with the succinimidylester-activated fluorophore leading to the formation of a stable carboxamide. Briefly, the RNA is reacted with a 20-30-fold molar surplus of TMR-NHS (5-carboxy- tetramethylrhodamine-N-hydroxy-succinimidylester; Amersham Pharmacia Biotech) in 100 - 250 mM Na₂CO₃/NaHCO₃, pH 8.3 for at least 2 hours at RT and protected from light. Unreacted dye is hydrolyzed by addition of 1.5 M hydroxylamine-hydrochloride to a 50-fold molar excess and incubation for further 30 minutes. After labeling, the RNA is separated from the free dye by gel filtration and further purified by preparative RP-HPLC as described above. Peaks containing labeled RNA are collected and any remaining TEAAc is removed by gel filtration. The concentration of the labeled RNA is determined by UV absorption spectroscopy as described above but with correction for the dye absorption at 260 nm.

Deduction of the HuR sequence binding motif M: Based on our previous observation that polyU is bound by HuR with high affinity, the effect of elongation of U₈ is tested. Individual RNA fragments are synthesized and the affinities (given as Kd_app values) to full length HuR are determined (see TABLE 1). While the simplest variant of U₈ motif (fragment No. 1) is not recognized by HuR, an elongation by one nucleotide to U_g (fragment No.2) shows a sufficient high binding. An influence of the fluorescent dye is excluded by competition experiments with unlabeled RNA fragments. The 9mer fragment (fragment No.3) contains the HuD motif and an additional nucleotide 3'terminally but is not bound by HuR. The high affinity binding to fragment No.4 however indicates that non-U nucleotides are tolerated within HuR binding motif, but at certain positions only. We have found that 9 nucleotides are sufficient for binding of HuR and four different 9mer frames within (AUUU)sA are tested (see fragments No 4a) to 4d) in bold). The exclusive recognition of fragment 4b by HuR within the four corresponding fragments demonstrates that HuR binds to frame 2 within (AUUU)sA. This frame is consistent with the HuD motif, but 5'terminally elongated by one uracil residue, suggesting the preliminary binding motif NN(U/C)UNN(U/C)U(U/C). Fragments 5, 6, 7a-7d and 8a-8c serve to tests tolerance for non-Uracil (exemplified by A=adenine) and C, respectively, at the depicted

(bold and underlined) positions. In consequence we found that HuR sequence binding motif is NNUUNNUUU. This interaction appears to follow an all-or-nothing mechanism: While sequences with single mismatches are not recognized, sequences fulfilling this motif are bound with high affinity and an invariable Kd, Kd_fund, of 0.99 nM. Results are set out in TABLE 1 below.

TABLE 1

These data delivered NNUUNNUUU as the sequence binding motif M. Hence, the NNUUNNUUU matching (m)RNA sequences s-i to s₁₅ as specified in Table 2 were selected for steps (c) to (e).

c) Calculating )p( s, M^s , σf for Si to s₁₅:

An HuR homology model (data not shown) based on the structures of HuD (1 FXL.1G2E) and Sxl (1 B7F) suggests an HuR preference for ssRNA. As the RNA fragments co- crystallized with HuD and Sxl might be too short to form a stable double stranded region, we tested the effect of TNFα-ARE (Table 2) hybridization with its antisense strand. The complete loss of HuR recognition observed, indicates that HuR binding requires single stranded NNUUNNUUU sequences. Based on these observations we selected single stranded NNUUNNUUU as the secondary structure element to be tested. The accessibility p(s,M^s,σf ) , ^* =NNUUNNUUU, σf = all single stranded, is calculated for the sequences Si to s₁₅ as described above (equations 1 to 7) and is shown in Table 2. d) Measuring macroscopic affinities:

HuR - ARE binding assay by 2D-FIDA-anisotropy:

Binding of HuR to fluorescently labeled AREs and ARE-related RNA fragments is monitored by determination of the fluorescence anisotropy with 2dimensional-Fluorescence Intensity Distribution Analysis (2D-FIDA). Direct binding experiments are performed by titration of fluorescently labeled RNA fragments against increasing concentrations of recombinant HuR. For this purpose, the fluorescently labeled RNA is first diluted to a concentration of ~ 5 nM in assay buffer (PBS, 0.1% Pluronic, 5 mM MgCI₂), thermally denatured by incubation for 2 min at 80°C and refolded by slow cooling to RT at -0.13 °C / sec in a thermocyler. Samples corresponding to the individual titration points (usually 11 points in a concentration range of 0.1 - 100 nM HuR and a constant RNA concentration of 0.5 nM) are prepared by mixing the appropriate volumes of assay buffer, RNA and protein solutions, at a final volume of 25 to 100 μl per sample. The 2D-FIDA-r measurements are performed in 96 well glass bottom microtiter plates (Whatman) on an EvotecOAI PickoScreen 3 instrument at ambient temperature (constant at 23.5 °). The Olympus inverted microscope IX70 based instrument is equipped with two fluorescence detectors and a polarisation beam splitter in the fluorescence emission path, adopted to perform anisotropy measurements. A HeNe laser (λ = 543 nm, laser power = 495 μW) is used for fluorescence excitation. To ensure a high degree of polarisation of the fluorescence excitation source, an additional linear polarisation filter is placed in the optical path in front of the sample. The excitation laser light is blocked from the optical detection path by an interference barrier filter with OD (optical density) = 5. 5'-TMR (Molecular Probes) dissolved in assay buffer (c ~ 0.5 nM) is used for the adjustment of the confocal pinhole (70 μm) and for the determination of the G-factor of the instrument (see e.g. Lakowicz J.R, Principles of Fluorescence Spectroscopy, Plenum Publishers, New Yoprk, ed.2, 1999). For the determination of the fluorescence anisotropy from the 2D-FIDA raw data, the Molecular Brightness q is extracted for each channel using the FIDA algorithm (see e.g. Kask P. et al., Biophys.J. 78, 1703, 2000 or Kask P. et al., PNAS 96, 13756, 1999). The anisotropy is then calculated as described in (Lakowicz J.R. cited above). The 2D-FIDA anisotropy signal is averaged for each sample from 10 consecutive measurements (a 5 seconds). After every 11 measured samples, measurements of 5'-TMR are performed for the determination of the G-Factor (calculated using P(_true) TMR = 0.034). For a determination of the equilibrium dissociation constant K , the recorded anisotropy data are fitted based on the exact algebraic solution of the binding equation describing the average steady-state anisotropy signal r in dependence of the degree of complex formation derived from the law of mass action:

[1] wherein

[RNAfree]: equilibrium concentration of free RNA,

[HuR_free]: equilibrium concentration of free HuR,

[RNA-HuR]: equilibrium concentration of RNA-HuR complex,

[RNA₀]: total concentration of RNA, [HuRo]: total concentration of HuR, r_mi_n: .anisotropy of free RNA, rm_ax: anisotropy of RNA-HuR complex, r: average anisotropy for the steady-state equilibrium at the given HuR₀ and RNA₀ concentrations, Q: quenching; for 2D-FIDA-anisotropy measurements, Q = q_tot(max) qtot(min); at q_tot = qn + 2 ^* qj.; qn, qx: molecular brightnesses in parallel and perpendicular polarisation channels.

The equation is compiled in the program GraFit 5.0.3 (Erithacus software, London) and the data are fitted based on least square regression using the Marquardt algorithm. All of the presented data are averages from at least three independent experiments.

For evaluations based on 2D-FIDA-r, optimal results are achieved at an average of ca. < 1 fluorescent particles in the confocal volume (see e.g. Evotec BioSystems AG, 2D-FIDA- Quick Guide, 2001), which corresponds to a fluorophore concentration of ca. 0.5 nM in the described setup. For this reason, the ligand (5'TMR-RNA) is diluted to a concentration of ~0.5 nM for all 2D-FIDA-r measurements. The accurate concentration in each sample is finally calculated based on a determination of the particle number in a parallel FCS (Fluorescence Correlation Spectroscopy) evaluation of the recorded 2D-FIDA-r data and the size of the confocal volume, as determined by the adjustment parameters for the point spread function. The apparent HuR dissociation constants Kd_app for the RNA sequences s^ to s₁₅ determined as described above are given in Table 2. (e) Correlation between Kd_app and p(s,M^s,σ ) :

The experimentally determined Kd_app (step d) are plotted against the calculated p(s,M^s,σ ) values (step c) to evaluate whether Kd_app depends reciprocally on p(s,M^s ,σf) as derived above (Figure 1). The reciprocal dependence is attested visually (Figure 1), but can be supplemented by statistical tests for correlation.

Based on the results illustrated in Figure 1 we conclude that the secondary structure element required for HuR-RNA recognition is the sequence motif NNUUNNUUU in fully single stranded conformation (=σ^ ).

EXAMPLE 2: Designed sequences with predicted HuR affinity:

By iteratively modifying the TNFα ARE sequence (s₆, Table 2) and recalculating p(s,M',σ„ ) , ^* =NNUUNNUUU, σ = all single stranded, as specified in Example 1 (c), we provide the sequences s , s₈, s₉. We predict the apparent HuR affinities to these sequences based on Kd_app = Kd_fun / p(s,M^s ,σ^) . A Kd_fund of 0.49 +/- 0.06 nM is determined by fitting the data from Example 1 (Table 2) to the equation Kd_app = Kd_fu_nd / p(s,M^s,σ ) . Measured Kd_app are in good correlation with the predicted values (sequences, predicted and measured affinities are specified in Table 3).

TABLE 3

Claims

Patent claims

1. A method for detecting an influence of a secondary structure element of an RNA molecule on the binding of a ligand, e.g. a protein comprising (a) providing an RNA-binding ligand L for which the sequence binding motif M is known

(b) providing at least 3 different RNAs of sequences s^ s₂, and s₃, each of s-i s₂, and s₃ having the same sequence binding motif M for the ligand L, but each differing in the secondary structure,

(c) calculating each p(s,M^s ,σf) for each s_1t s₂, and s₃ which is the probability that a structure contains at least one match of M in a particular test conformation σf , e.g. calculating each p(s,M^s,σ ) according to the equation [7], as set out below,

(d) measuring each macroscopic affinity constant Kd_app for the binding of L to each of S_t s₂, and s₃,

(e) determining for any test conformation σ* whether the correlation between each of . Kdgpp and p(s,M^s,σ ) for each of S_! s₂ and s₃ follows the dependency

Kd_app ∞ llp(s ,M^s,σf) and

(f) choosing an RNA secondary structure element (M^s,σ^) which follows said dependency.

2. A method of claim 1 , further comprising the steps of

(g) determining the fundamental affinity constant Kd_fund of L to M in an active conformation <τf based on the equation Kd_app = Kd_fiιnd / p(s_M,M^* ,σ ) and

(h) (optionally) providing, e.g. novel, RNAs, matching motif M, with a defined d_app for the ligand L.

3. A method of claim 1 , further comprising step

(i) screening for additional RNAs bound by the ligand L based on the occurrence of

M in an active conformation σ„ .

4. The use of a method of any one of claims 1 to 3 for the identification of a disease- relevant target.

5. The use of a method of any one of claims 1 to 3 for the manipulation of an RNA.

6. The use of a method of any one of claims 1 to 3 for the identification of an agent for the manipulation of an RNA.

7. An agent of claim 6 for pharmaceutical use.

8. A pharmaceutical composition comprising an agent of claim 7 in association with at least one pharmaceutical excipient.

9. The use of an agent of claim 7 or of a pharmaceutical composition of claim 8 for the treatment of a disorder having an etiology associated with the production of a substance, e.g. an inflammatory acting (causing/enhancing) substance, selected from the group consisting of cytokines, growth factors, proto-oncogenes or viral proteins.

IL/5-May-2004