Formation of a G-tetrad and higher order structures correlates with biological activity of the RelA (NF-[kappa]B p65) `antisense' oligodeoxynucleotide
Formation of a G-tetrad and higher order structures correlates with biological activity of the RelA (NF- [kappa]B p65) `antisense' oligodeoxynucleotideLyuba Benimetskaya, Myriam Berton+, Alexander Kolbanovsky1, Simon Benimetsky1 and C. A. Stein*
Departments of Medicine and Pharmacology, Columbia University, College of Physicians and Surgeons, 630 West 168 Street, New York, NY 10032, USA and 1Department of Chemistry, New York University, 31 Washington Place, New York, NY 10003, USA
Received March 3, 1997;Revised and Accepted May 9, 1997
ABSTRACT
We have examined the behavior of the phosphorothioate antisense Rel A (NF-[kappa]B p65) oligodeoxynucleotide (oligo) and related molecules. Because of the presence of a G-tetrad near its 5' terminus, this molecule is capable of forming tetraplexes and other higher order structures in a temperature and time dependent manner. The G-tetrad in the phosphodiester congener is protected from methylation by dimethylsulfate when the oligomer is 3'-phosphorylated. However, this protection is completely lost when it is 5' phosphorylated, indicating that the formation of at least some higher order structures has been blocked. In addition, we also prevented tetraplex formation by substitution of 7-deazaguanosine (7-DG) for guanosine at several positions within and outside of the tetrad. This substitution retains Watson-Crick base pair hybridization but prevents Hoogsteen base-pair interactions. When murine K-Balb cells were treated with 20 [mu]M antisense RelA oligo, complete blockade of nuclear translocation of RelA was observed. However, this effect was virtually entirely abrogated in most cases by 7-DG substitution within the tetrad, but retained when the substitution was made 3' to the tetrad. The AS RelA-induced downregulation of Sp-1 activity behaved similarly after 7-DG substitution. Thus, the parent phosphorothioate AS RelA molecule cannot be a Watson-Crick antisense agent. However, these conclusions cannot be extrapolated to other G-tetrad containing oligomers and each must be evaluated individually.
INTRODUCTION
Successful application of antisense oligodeoxynucleotide technology relies on the specificity of the Watson-Crick base pair interaction (1 ). This concept has been the motive force behind numerous studies, prominent among them the elegant work of R.Narayanan and co-workers, who examined in detail inhibition of NF-[kappa]B activity in many cell lines (2 -4 ). One of the cell lines which was perhaps the major focus of their work was the K-Balb (K-ras-transformed) murine fibroblast line, to which they targeted a 24mer phosphorothioate oligodeoxynucleotide against the initiation codon region of the RelA (p65) subunit of the murine NF-[kappa]B nuclear transcriptional regulatory factor (2 -4 ). This antisense oligodeoxynucleotide, AS RelA (a phosphodiester or phosphorothioate oligodeoxynucleotide compementary to the mRNA of the p65 subunit of the NF-[kappa]B nuclear transcriptional regulatory factor) was shown to cause a profound block in nuclear translocation of RelA, as well as inhibition of nuclear expression of the Sp-1 transcription factor, a property that was not shared by the sense control (5 ). However, the 5'-region of this oligomer contains four contiguous guanosine residues. Additional experiments employing single base mutated AS RelA oligomers suggested that the presence of four contiguous guanosines was essential for antisense activity (6 ). Furthermore, activity was not dependent only on the presence of this motif, but clearly on its position within the molecule as well.
Other investigators (7 ,8 ) have noted an enhanced non-sequence-specific activity in phosphorothioate oligodeoxynucleotides that contain four contiguous guanosines. Burgess et al. (7 ) examined inhibition of smooth muscle cell proliferation by a phosphorothioate oligodeoxynucleotide targeted to c-myb and noted that most of the activity was a function of this sequence element. Guvakova et al. (8 ) demonstrated increased binding of a phosphorothioate oligomer containing this motif to basic fibroblast growth factor (bFGF). A guanosine-rich phosphodiester oligodeoxynucleotide has recently been demonstrated to be highly nuclease resistant and to be a relatively specific inhibitor of HIV-1 integrase (9 ). Because this and other examples (10 ,11 ) show that G-rich sequences enhance non-sequence specificity of both phosphodiester and phosphorothioate oligomers, we conjectured that the biological activity of AS RelA might not be due to Watson-Crick hybridization, but rather to an effect of the guanosine residues themselves.
Guanine-rich regions in oligonucleotides can form tetrads (also known as quartets), in which the four residues cyclize to form an array held together by hydrogen bonds between N1H and O6 and N2H and N7. The G-rich region on one strand can combine with that on another to form higher order structures, for example in a bimolecular fashion by association of two strands or in a tetramolecular manner by appropriate positioning of four individual strands (12 -17 ). A wide diversity of structures may result, as in some tetraplexes, for example, the strands may be parallel and in others antiparallel. Furthermore, the rates at which these strands associate and dissociate may also be very sensitive to the sequence context, ionic strength and temperature.
In this work we demonstrate that the non-sequence-specific activity of at least some G-rich oligomers is dependent on tetraplex formation only for short oligomers. We then address the question of higher order structure formation in the antisense RelA oligodeoxynucleotide. We demonstrate that this oligomer forms higher order structures, but that 5'-phosphorylation blocks formation of at least some of these structures. We then show that substitution of 7-deaza-2'-deoxyguanosine (which eliminates Hoogsteen but preserves Watson-Crick base pairing) (18 ) within, but not outside, the G-tetrad in the AS-RelA phosphorothioate oligodeoxynucleotide drastically alters its biological activity.
MATERIALS AND METHODS
Synthesis of oligodeoxynucleotides
Phosphodiester oligodeoxynucleotides were synthesized by standard phosphoramidite chemistry on an Applied Biosystems (Foster City, CA) 380B synthesizer. Phosphorothioate oligodeoxynucleotides were also synthesized by standard methods (19 ) and sulfurization was performed using tetraethylthiuram disulfide/acetonitrile (TETD; Applied Biosystems). Following cleavage from the controlled pore glass support, oligodeoxynucleotides were base deblocked in ammonium hydroxide at 60oC for 8 h and purified by reversed phase HPLC [0.1 M triethylammonium bicarbonate (TEAB)/acetonitrile, PRP-1 support]. Oligomers were detritylated in 3% acetic acid and precipitated with 2% lithium perchlorate/acetone, dissolved in sterile water and reprecipitated as the sodium salt from 1 M NaCl/ethanol. Concentrations were determined by UV spectroscopy.
5'-32P-Labeled oligodeoxynucleotides were synthesized using T4 polynucleotide kinase and [[gamma]-32P]ATP by the method of Sambrook et al. (20 ). 3'-End-labeling was performed using terminal deoxynucleotidyltransferase and [[alpha]-32P]dideoxyATP (20 ).
Gel electrophoresis
Higher order structure formation of deoxynucleotides was analyzed in native and 20% polyacrylamide gels (19:1 polyacrylamide/bisacrylamide) after electrophoresis in TBE buffer at room temperature. Gels were run at 10-16 V/cm and then subjected to autoradiography or stained with Stains-All (Sigma).
32P-Labeled oligodeoxynucleotides, after chemical modification with dimethylsulfate (DMS) and cleavage with piperidine, were analyzed by electrophoresis in 20% polyacrylamide-7 M urea sequencing gels and autoradiographed.
Nucleotide base protection assays
DMS and diethylpyrocarbonate (DEPC) cleavage was performed at 25oC for the times shown in the figure legends. 3'- or 5'-32P-labeled oligodeoxynucleotides were dissolved in 5 [mu]l 90 mM Tris-borate, pH 8.3, 1 mM EDTA (TBE), 25 mM KCl and an equal volume of 1% (v/v) DMS (Aldrich) or 10% (v/v) DEPC (Aldrich) in the same buffer. The reaction with DEPC was allowed to proceed for 30 min or 1 h, with vortex mixing at 5-10 min intervals throughout. Reactions were stopped by addition of 1.5 M sodium acetate, pH 7.0, 1 M 2-mercaptoethanol and cooling at 4oC for 5 min. The oligodeoxynucleotides were precipitated with 2% lithium perchlorate/acetone. Chemically modified oligodeoxynucleotides were cleaved by reaction with 20 [mu]l 10% piperidine at 90oC for 45 min. Traces of piperidine were removed by evaporation. Samples were then dissolved in 5 [mu]l formamide loading buffer, heated for 5 min at 90oC, cooled to 4oC and analyzed by electrophoresis in 20% polyacrylamide- 7 M urea sequencing gels.
Cells
K-Balb cells were obtained from the American Type Culture Collection (Rockville, MD). Cells were grown and maintained at 37oC in 95% air/5% CO2 in Dulbecco's modified Eagle's medium (DMEM) (Gibco BRL, Grand Island, NY) containing 10% (v/v) fetal calf serum (Gibco BRL), 1 mM L-glutamine, 100 U/ml penicillin G and 100 [mu]g/ml streptomycin.
Electrophoretic mobility-shift assays (EMSA)
K-Balb cells (25 * 104) were seeded into 100 * 20 mm tissue culture dishes and incubated until adherent and spread. Phosphorothioate oligodeoxynucleotides (20 [mu]M) were then added and cells incubated for 48 h at 37oC. Nuclear extracts were isolated and EMSA were performed as described (3 ). Protein concentrations were determined using the Bradford protein assay (BioRad, Richmond, CA). The NF-[kappa]B HIV-1 consensus probe sequence was 5'-GTA GGG GAC TTT CCG AGC TCG AGA TCC TAT G-3'. The Sp-1 and CK-1 consensus binding sequences were respectively 5'-T CGA ATT CGA TCG GGG CGG GGC GAG C-3' (5 ) and 5'-TGA TAA GGG CCA GGA GAT TCC ACA GTT CAG GTA GTT-3' (4 ). The probes were labeled with [32P]dNTP as described (5 ). Nuclear extracts (7-10 [mu]g) and 32P-labeled probes were co-incubated and complexes were resolved by electrophoresis in a non-denaturing 4% polyacrylamide gel and autoradiographed. Unlabeled probe at 100* concentration was used as a competitive inhibitor to establish gel shift specificity.
Western blot analysis
Aliquots of cell extracts containing 10-15 [mu]g protein were resolved by 12% SDS-PAGE (polyacrylamide gel electrophoresis) and transferred to Hybond ECL filter paper (Amersham, Arlington Heights, IL). Filters were incubated at 25oC for 1 h in Blotto A [5% non-fat milk powder in Tris-buffed saline (TBS); 10 mM Tris-HCl, pH 8.0, 150 mM NaCl] plus 0.05% Tween-20 and then incubated at 25oC for 1 h in Blotto A containing 1:1000 NF-[kappa]B p65 rabbit polyclonal IgG mAb (Santa Cruz Biotechnology, Santa Cruz, CA). After washing in TBS-T (TBS + 0.05% Tween-20), fiilters were incubated for 1 h at 25oC in TBS-T containing a 1:10 000 dilution of peroxidase-conjugated goat anti-rabbit secondary mAb (Amersham). After washing, ECL was performed according to the manufacturer's instructions (Amersham).
Synthesis of alkylating, radioactive phosphodiester oligodeoxynucleotide 5'-N-methyl-N-(2-chloroethyl)aminobenzylamine [32P]OdT12 (ClRNH[32P]OdT12)
This compound, a dodecathymidylate phosphodiester oligodeoxynucleotide derivative with an alkylator moiety:
Competitive inhibition of binding of ClRNH[32P]OdT12 to bFGF by tetraplex-forming oligodeoxynucleotides
This was performed by the method of Yakubov et al. (22 ). Individual G-rich and control oligodeoxynucleotides (Table 1 ) were used at the stated concentrations as competitors of binding of the probe, a 32P-labeled oligodeoxynucleotide to bFGF. The IC50 of competition was determined by visual inspection and values of Kc were calculated by the Cheng-Prusoff equation (23 ):
CD spectra were recorded at room temperature using an AVIV Model 62A DS CD spectropolarimeter (Aviv Associates, Lakewood, NJ). Solutions of oligodeoxynucleotides (1 OD U/ml) were prepared in 0.1 M Tris-HCl, pH 7.6, to which, in some experiments, 140 mM KCl was added. The spectra were recorded at room temperature.
RESULTS
Binding of SdG6Tx to bFGF is a function of oligomer length
We initially wanted to determine the effect of tetraplex formation on the non-sequence-specific binding to protein of phosphorothioate oligodeoxynucleotides. To do this, we synthesized the model series of tetraplex-forming 5'-SdG6Tx-3' oligomers listed in Table 1 . Electophoresis of this series under non-denaturing conditions (with or without 25 mM KCl) demonstrated that the shorter oligomers existed predominately as discrete tetraplexes, as evidenced by their appropriately slow rate of migration (Fig. 1 ). However, for dT >= 8, only smears can be visualized. Moreover, the presence of 100 mM KCl as against 100 mM LiCl did not markedly stabilize the discrete tetraplex form versus the smear. The presence of the smear, which may in part represent `bleeding out' of lower order from higher order structures, may imply that the rate of dissociation of the tetraplexes to the putative dimers and monomers under the electrophoretic conditions is relatively rapid.
Formation of higher order structures by the phosphodiester and phosphorothioate antisense RelA oligodeoxynucleotides
We examined the ability of these oligomers to form higher order structures by native gel electrophoresis. As shown in Figure 3 , at least three distinct bands can be observed. The slowest migrating band (top of the gel) is a tetraplex: we demonstrated this by the method of Wyatt et al. (10 ), by boiling the phosphodiester and phosphorothioate AS RelA oligomer with an equal concentration of a phosphodiester oligomer of half length (5'-GAG GGG AAA CAG-3'). Native gel electrophoresis showed five bands with approximate tetraplex mobility, as would be expected by the number of terms in the binomial expansion (not shown).
The G-tetrad in the RelA phosphodiester oligodeoxy- nucleotide is protected from DMS methylation
Because 5'-phosphorylated AS RelA does not appear to form tetraplexes as easily as non-5'-phosphorylated AS RelA and because of our earlier unpublished observations that certain 5'-fluorescein-labeled 5'-G-rich oligomers are observed largely as monomers on electrophoresis in native gels, we examined the extent of protection at guanosine N7. This was done by alkylation with DMS and piperidine cleavage. The radioactive label was located at the 5'-terminus in some experiments and at the 3'-terminus in others (25 mM K+, 10-15 min reaction time). These conditions produce single hit methylation and the products of piperidine cleavage of the methylated oligomers were resolved on sizing gels. As is seen in Figure 4 , when the 32P label (as ddA) is at the 5'-terminus, G2-G5 are clearly observed and the bands are roughly equal in intensity to those of G5-G9. In contrast, when the oligomer is 3'-phosphorylated, the G2-G5 bands are highly under-represented, indicating that the N7 of these guanosines are protected from methylation by Hoogsteen base pair formation. Weak protection of G6-G7 was also observed. If the 3'-labeled sample was boiled before DMS treatment, G2-G5 were visible and, with the exception of G2, equal in intensity to the other spots. Thus, on the basis of these data, we conclude that the parent AS RelA exists in solution predominately as a relatively labile tetraplex, whose formation can be substantially prevented by 5'-phosphorylation. As a control, we alkylated the adenosine residues with DEPC. Both the 3'- and 5'-32P-labeled phosphodiester oligodeoxynucleotides were treated in 25 mM KCl as in Materials and Methods at 25oC, cleaved with piperidine and electrophoresed. All adenine moieties in both oligomers were equally reactive to DEPC/piperidine, indicating that these bases do not play a significant role in formation of higher order structures.
Biological effects of the modified AS relA phosphorothioate oligodeoxynucleotides
Because of the formation of G-tetrads with the AS RelA oligomer and the subsequent formation of higher order structures, we substituted 7-DG at several positions both inside and outside the G-tetrad (Table 2 ). However, as is shown in Figure 5 , only substitution at certain positions will block formation of higher order structures. For example, under the conditions of electrophoresis, no tetraplexes were observed for the molecule with a 7-DG substitution at G1 (lane 1) and G2 (lane 2), although the G1-substituted molecule should still be able to form a tetraplex. Substitution of 7-DG at G5 (lane 4), however, did not eliminate formation of higher order structures. Murchie and Lilley (18 ) demonstrated that the AGGG motif is also capable of forming tetrads and tetraplexes. Interestingly, it was observed by CD (Fig. 6 ) that the strong positive Cotton effect at 270-280 nm for this molecule (Fig. 6 b) was shifted to a shorter wavelength in the presence of K+. The shift was also observed for the parental AS RelA molecule (Fig. 6 a), but not for molecules with substitutions at G3 (not shown) or G4 or for SdT24, none of which can form tetrads (Fig. 6 c and d).
DISCUSSION
Phosphorothioate oligodeoxynucleotides form tetrads and tetraplexes in the same manner as do phosphodiesters. However, as demonstrated by Wyatt et al. (11 ), isosequential phosphorothioates have less tendency than phosphodiesters to exist in solution as tetraplexes, because of slower association and faster dissociation rates. In addition, as we have shown in a model series of phosphorothioate dG6Tx oligomers, the tendency to form tetraplexes is greatly diminished with increasing oligomer length. This can be seen by the presence of smears rather than discrete tetraplex bands on native gels, where the observed smears represent, at least in part, continuous dissociation in the gel of higher order to intermediate (putative dimeric?) and monomeric species. This increased dissociation rate is probably caused by increased charge repulsion as the strands increase in length and the phosphate groups increase in number. Interestingly, the phosphorothioate dG6Tx tetraplexes are only minimally stabilized by K+ as against Li+. This is not generally the case for phosphodiester tetraplexes, which are stabilized by K+.
On the basis of these data, it would be predicted that the phosphorothioate AS RelA oligomer would form relatively few tetraplexes at physiological saline concentration, a temperature of 37oC and an oligomer concentration of 20 [mu]M. In fact, this is what indeed is observed, in spite of the fact that intermediate forms can still be seen. However, as demonstrated by N7 methylation protection experiments with the phosphodiester congener, the AS RelA oligomer, despite what is seen on the native gels, seems to exist in solution predominately as a tetraplex. This contradiction is more apparent than real, because if the tetraplexes are relatively labile they may dissociate under the electrophoretic conditions, as the more mobile monomers migrate in the electric current. However, the tetrad will not form in the presence of a 5'-phosphate group. This is supported by the lack of protection of G2-G5 after N7 methylation with DMS and piperidine cleavage in the 5'-phosphorylated phosphodiester AS RelA oligomer, as compared with extensive protection when the oligomer was 3'-phosphorylated. Furthermore, there is a relative lack of `antisense' activity (inhibition of NF-[kappa]B nuclear translocation) in K-Balb cells treated with the phosphorothioate 5'-thiophosphate AS RelA oligomer.
Intermediate forms can still be observed with the 5'-phosphorylated phosphodiester AS RelA in native gel electrophoresis: it is possible, but not certain, that these forms are linear homodimers. Tetraplexes form with difficulty, probably due to mutual charge repulsion of the phosphate groups.
Substitution of 7-DG at G5 does not block formation of tetraplexes, as is shown in Figure 5 , most likely because a tetrad can still be formed by the AGGG motif. This correlates with the shift, in 140 mM K+, of the positive Cotton effect in the CD spectrum from 280 to 265 nm, a peak which is seen in molecules that can form all-G-tetrads (11 ,24 ). However, neither phosphorothioate oligomers with 7-DG substitution at G4 or G3 nor SdT24, which may not be able to form intermolecular G tetrads, shift the positive Cotton effect band to shorter wavelength in 140 mM K+. Interestingly, the substitution of 7-DG at G5 produces a molecule that is almost incapable of blocking nuclear translocation of NF-[kappa]B in treated K-Balb cells, in contrast to what is seen with the parent all-G-tetrad AS RelA at equimolar doses. Thus, it appears that the cells can directly recognize the difference between AGGG and GGGG tetraplexes.
The biological activity of the phosphorothioate RelA AS oligomers is drastically altered based on the position of 7-DG substitution. This alteration is observed with both the HIV-1 [kappa]B and CK-1 probes and is also seen on examination of Sp-1 activity. When 7-DG substitution is located in the G-tetrad, intermolecular tetrads (except in the case of G5 substitution, as described above) cannot form and `antisense' activity is partially or totally lost, as assessed by EMSA and Western blotting. However, if 7-DG substitution is made outside the G-tetrad, e.g. at G1 or G7, the intermolecular G-tetrad can, in principle, still form and `antisense' activity is preserved. However, we consider it extremely unlikely that 7-DG single base substitution at any position will affect the value of Tm with complementary mRNA by >0.5oC. (Of interest in this respect is that `antisense' activity is almost completely preserved in the G3 + G4 disubstituted molecule.) Furthermore, it is equally unlikely that cellular uptake will be affected by single base 7-DG substitution. The cellular uptake of phosphorothioate oligodeoxynucleotides is dependent on the processes of fluid phase endocytosis and adsorbtive endocytosis (19 ). The former process will not be influenced by the concentration of higher order structures. The latter process will depend, in part, on the Kd of binding of the oligomers to cell surface heparin binding proteins (8 ). However, as shown above for the phosphorothioate dG6Tx model system, the decrease in Kc of protein binding to oligomer may occur only for short tetraplexes (<14mers).
Thus, on the basis of all of our data, we must conclude that the parent phosphorothioate AS RelA, despite its ability to reduce NF-[kappa]B nuclear translocation in treated K-Balb cells in an apparently sequence-specific manner, cannot be acting solely, if indeed at all, as a Watson-Crick antisense agent. Instead, it appears to exert its remarkable antisense-mimicking effects through a combination of mechanisms, including the following. (i) An aptameric effect of the intermolecular GGGG tetraplex, which as described is different from that of an AGGG tetraplex. The all-G-tetraplex may exert its effect either intracellularly or at the level of the cell membrane. (ii) An intrinsic non-sequence-specific effect of the phosphorothioate oligomer related to its ability to bind to heparin-binding proteins (8 ,22 ). For example, Khaled et al. (24 ) have demonstrated that phosphorothioate AS RelA binds to fibronectin and laminin (on the [alpha]1 chain) and can block binding of laminin to bovine brain sulfatide. This blockade is highly enhanced for the AS RelA versus the sense RelA (the sense sequence with respect to the NF-[kappa]B mRNA) oligomer.
Interestingly, the phosphorothioate oligomer that contains 7-DG disubstitution at G3 + G4, which blocks formation of any tetraplexes and higher order structures, is almost as active as the parental AS RelA at inhibition of NF-[kappa]B nuclear translocation. It also does not, like the parent AS RelA molecule, suppress Sp-1 activity. It is entirely possible that this species represents the Holy Grail for this system, a true antisense effector molecule. If so, the strategy of 7-DG substution provides an effective, though costly, way of circumventing the G-tetrad problem.
However, some caution should be taken not to generalize these results to other well-studied oligodeoxynucleotides that contain a G-tetrad (e.g. the antisense phosphorothioate c-myb codons 2-7 or 2-9 oligomers; 25 ). It cannot be determined a priori, based only on our analysis, whether biological effects observed in other systems with these oligomers are or are not due to an antisense mechanism. Indeed, our preliminary data indicate that the antisense c-myb codons 2-9 oligomer does not form tetraplexes even in the presence of high K+ and at low temperature.
The induction of biological effects by G-rich phosphorothioate oligomers is a highly complex process. Each G-rich oligodeoxynucleotide must be evaluated individually. However, on the basis of our data it appears that 7-DG substitution is not only an appropriate way to approach the problem, but is also one that can provide some reasonably conclusive answers about mechanism.
ACKNOWLEDGEMENTS
C.A.S. was partially funded by NCI grant 60639. C.A.S., L.B. and M.B. were funded by ZW Biomedical AG. We wish to thank Li-Ming Zhang for his expert technical assistance and Prof. R.Letsinger for helpful comments.Note on nomenclature: we have preferred to use the term tetrad instead of quartet (though the two appear to be used interchangably in the literature) as it seems more logical and orthographically esthetic to call the building block of a tetraplex a tetrad, rather than a quartet
REFERENCES
1 Stein,C.A. and Cheng,Y.-C. (1993) Science, 261, 1004-1012.MEDLINE Abstract
2 Higgins,K.A., Perez,J.R., Coleman,T.A., Dorshkind,K., McComan,W.A., Sarmiento,U.M., Rosen,C.A. and Narayanan,R. (1993) Proc. Natl. Acad. Sci. USA, 90, 9901-9905.MEDLINE Abstract
