Towards a general triple helix mediated DNA recognition scheme France This review describes new perspectives offered by the synthesis of non-natural nucleosides to overcome current limitations and extend the triplex-mediated DNA recogni- tion scheme to any sequence. Alternate strand purinebinding, direct pyrimidine recognition, and binding to the whole base-pair are described. The review highlights struc- tural requirements to the design of modified nucleosides, as well as perturbing events such as tautomeric ambiguity and intercalation for the extended heterocyclic bases. 1 Introduction Nucleic acids triple helices were first described in 1957 by Felsenfeld and Rich,' just a few years after Watson and Crick discovered the double-helical nature of DNA.The present interest in triplexes follows the discovery in 1987 by the groups of Hklitne and Dervan that short oligonucleotides can bind in a sequence specific manner to a duplex target in DNA under suitable conditions.2.3 Since binding occurs in the major groove, triplex-forming oligonucleotides (TFOs) have the potential to interfere with regulatory proteins which bind to the same site, hence controlling gene expression. This exciting perspective has stimulated extensive work on triple helical complexes during the past decade (reviewed in refs. 4 and 5). Much effort has been devoted to increasing the stability of triplexes under physiological conditions, as well as overcoming the major drawback of this approach, i.e. the requirement for oligopurine tracts on the target DNA.This limitation arises from the recognition mechanism which involves the purine's two remaining 'Hoogsteen' hydrogen-bonding sites in the major groove of the DNA duplex (Fig. 1); pyrimidines having only one hydrogen-bonding site vacant cannot be efficiently bound. J. P. Behr and S. 0.Doronina This review will mainly focus on new perspectives offered by the synthesis of non-natural nucleosides to overcome current limitations and extend the triplex-mediated DNA recognition scheme to any sequence. Other recent advances to stabilizing triple helices in a sequence-independent manner, such as conjugation of poly- amines or triplex helix-specific intercalating agents, or the widely explored ribose-phosphate backbone modifications5 [peptide nucleic acid (PNA), oligonucleotide N3'-P5' phos-phoramidates and deoxyribonucleic guanidine (DNG)] will not be discussed here.Essentially two families of DNA triple helices have been characterized that differ in their third-strand sequence compo- sition and relative orientation (Fig. 2). In the pyrimidine- *purine-pyrimidine (Py *Pu-Py) family a homopyrimidine (Py) third strand is bound (*) parallel to the purine strand of target duplex (Pu-Py) in the major groove of DNA through iso-morphous T*A-T and G+*G-C base triplets. Within the T*A-T triplex plane a thymine of the TFO interacts with a Watson- Crick A-T base pair by making two Hoogsteen hydrogen bonds with adenine. The requirement of a protonated cytidine for guanosine recognition results in pH-dependent binding with optimal stability much below physiological pH (5.6-6.0).In the purine*purine-pyrimidine (Pu*Pu-Py) family, a purine-rich third strand is bound antiparallel to the purine strand of the target duplex. DNA recognition in this motif involves guanosine binding by reverse-Hoogsteen hydrogen bonding to guanosine of the G-C base pair (G*G-C base triplet), and either adenosine or thymidine binding to adenosine of the A-T base pair (A*A-T or T*A-T). Base triplets within this family are not isomorphous, i.e. location and orientation of the third-strand deoxyribose (dR) is sequence dependent. This leads to helix ~~ Svetlana 0. Doronina was born in Tomsk, Russia. After receiving her Master's degree in Chemistryfi-om Novosibirsk State University in 1985, she became a Junior Researcher in the laboratory of Organic Synthesis at the Novosibirsk Institute of Bioorganic Chemistry.In 1992 she came to France and since 1993 she started her PhD studies dealing with the DNA recognition by modified oligonucleotides under supervision of Dr J. P. Behr at the University Louis Pasteur of Strasbourg. Jean-Paul Behr is Research director at the Centre National de la Recherche Scientifique. He was born in 1947 and graduated from the Chemical Engineering School in Stras- bourg. He joined Jean-Marie Lehn in 1969 for PhD work in physical organic chemistry. After a postdoctoral stay in ShefSield to study molecular motion in biological membranes, he rejoined J.M. Lehn to contribute to the development of artificial receptors, catalysts and carriers. In I989 he moved to the Faculty of Pharmacy to head the Lahoratoire de Chimie Ge'ne'tique. His major interests are now in the development of DNA-binding molecules relevant to bio- technology and gene therapy. Chemical Society Reviews, 1997 63 Major Groove JD A Minor Groove Major Groove n Minor Groove Fig. 1 Watson-Crick hydrogen bond formation (dotted lines) between complementary nucleic bases leaves several H-bonding sites [donor (D) or acceptor (A)] vacant in the major groove of the resulting duplex. Purines (adenine and guanine) have distinct bidentate 'Hoogsteen' sites vacant whereas pyrimidines (cytosine and thymine) have only a single remaining binding site in the major groove.distortion for mixed sequences especially in the case of G*G- C/T*A-T sequences. These triplexes are not pH-dependent but are destabilized by physiological monovalent cation concen- trations. As shown above, DNA recognition by triple helix formation relies on hydrogen bonding interaction within the base triplets. However, the design of new binding motifs should take into account that third-strand binding depends on many factors besides the stability of isolated planar triplets. Indeed, stacking, van der Waals and dipolar interactions between the neighbour- ing heterocyclic bases are widely involved in helix stability. Structural isomorphism of base triplets too is an important concern to avoid ribose-phosphate backbone distortion: triplets are isomorphous if their N-dR bonds are superimposable (e.g.C+*G-C and T*A-T in Fig. 2) and we shall see later that non- isomorphism leads to considerable triplex destabilization for random purine sequences. Cationic heterocycles may be favourable in isolated sites due to their strong H-bonding capacity and to attractive electrostatic interactions in a poly- anionic context. Yet in contiguous sequences electrostatic repulsion between adjacent charges may result in overall destabilization as found when targeting G-stretches with protonated cytidines. Chimeric nucleosides are therefore de- signed to be mostly neutral. Sequence-selective recognition not only means affinity for the target nucleic base, but also discrimination against the three other ones (which is much more stringent).With respect to the latter criterion, tautomerism of heterocyclic bases must seri- ously be considered. Whereas natural bases are a 99.99% in the Watson-Crick tautomeric form shown in Fig, 1, providing both replication fidelity and some possibility of evolution, most polyaza-heterocycles have several tautomeric structures, which may hydrogen bond to more than just one natural base. Ideally, the development of base-modified nucleoside analogues with unambiguous tautomerism would clarify this problem. 2 Current limitations of homopurine triplexes 2.1 Py*Pu-Py triplexes Protonation of cytidine is required in order to establish two hydrogen bonds with guanosine (Fig.2). The pK, of isolated cytidine is 4.3. However, even though its apparent pK, increases when incorporated in oligonucleotides due to the polyanionic environment, optimum C+*G-C triplet stability still requires acidic conditions. This rather limits the potential use of such oligomers in vivo, where intracellular pH is highly regulated at ca. 7.3. The first (and up to now one of the best) solution to provide Py*Pu-Py triplex stability at physiological pH was replacement of cytidine by 5-methylcytidine6 (m5C). This modification enhanced considerably the stability of triple helices at neutral pH and may serve as a reference for G-C base pair recognition for newly designed structures.The influence of cytidine methylation on triplex stability does not appear to result from the 0.2 units pK, increase, but is mainly of entropic origin.7 Recently numerous synthetic nucleosides have been developed that display the hydrogen bonding pattern of protonated cytidine, as we shall see now. 2.1.I Pyrimidine-like analogues of cytidine Pyrimidine-like pseudoisocytidine8 (isOC), pyrazine base9 (pyDDA), 5-methyl-6-oxocytidine10 (m50xC) lead to B*G-C triplets (B = isOC, pyDDA, m5~"C) isomorphous to the canonical T*A-T triplet [Fig. 3(A)]. These heterocycles already have a hydrogen atom at the N3 position which allows them to bind to guanine in a pH-independent fashion. Triplexes containing these modified bases are more stable at neutral pH than cytidine-containing triplexes but generally do not achieve the stability of 5-methylcytidine-containing triple helices.Thus, triple to double helix transition temperatures (T,) of triplexes containing m50xC are still significantly lower at neutral pH than those observed for the corresponding m5C+-containing tri- plexes. This likely reflects the fact that cations are better H-bond donors that show further stabilization on a polyanionic target. To our knowledge, there are no data about the binding specificity of is0C and pyDDA. m50xC has been shown to be very selective, as the only target base pair (besides G-C) that led to detectable triple helix formation was C-G, but with significantly less stability. Both m5oXC and i~oC can adopt several tautomeric structures.X-ray crystallography as well as duplex-DNA melting tem- perature studies suggest that the N3-H tautomer is preferred for m50"C. Indeed m5°xC-containing duplexes are less stable than their natural counterparts. This could be due to steric clash between N3-H of rn50"C and N1-H of G which prevents Wat son-Crick base pairing. isoC has tautomeric structures in which the hydrogen atom can be bound to either N1 or N3 [Fig. 3(B)]. is°C-containing duplexes also show decreased stability, presumably because of steric clash [Fig. 3(C)]. These data lead to the supposition that the tautomeric ambiguity of these nucleoside analogues should not hamper their selectivity of binding. m50xC and isoC seem also to be promising candidates for targeting of G-C stretches.As mentioned previously, such sequences are particularly difficult targets in the Py*Pu-Py motif because of ionic repulsion between adjacent protonated residues when C+ or mT+-containing TFOs are used. Such unfavourable interactions should not be present in adjacent neutral m50xC or is0C nucleotides. Indeed, a triple helix containing the (isoC), sequence was shown to melt at least 22 "C higher at pH 7.5 than the corresponding (m5C+), triplex. Contiguous m5W did not lead to stable triplexes. Undesirable 64 Chemical Society Reviews, 1997 steric effects brought about by the 06 carbonyl atom or poor base stacking could be at the origin of the decreased stability. To illustrate the impact of non-natural base tautomerism on specific nucleic base recognition the example of the pyrido[2,3- dpyrimidine nucleosidel (F) is edifying.This heterocycle can exist in two tautomeric forms F1 and F2 [Fig. 4(A)], which can either bind to guanine or adenine in the Watson-Crick sense to form an antiparallel double helix, Fl-G (3 H-bonds) being considerably more stable than F2-A (2 H-bonds) [Fig. 4(B)]. In Py*Pu-Py triple helixes Fl may bind to G-C or C-G base pairs through a single H-bond, or F2 to A-T through two H-bonds [Fig. 4(C)]. Binding experiments have demonstrated that F recognizes the A-T base pair when incorporated into TFOs, with an affinity similar to that of the canonical T*A-T triplet. This example shows that when tautomers are of comparable stabilities, complementary strand binding selects the tautomers giving the most stable among possible structures, i.e.F1-G (3 H-bonds) and F2*A-T (2 H-bonds). 2.I .2 Purine-like cytidine analogues Some purine-like non-natural nucleotides (shown in Fig. 5) have been described for replacement of protonated cytidine: 3-methyl-5-amino-1H-pyrazolo[4,3-d]pyrimidine-7-one12 (PI), N6-methyl-8-oxoadenosine l3 (8oxA) or N7-guanosine14 (7G). From affinity cleaving analysis, the stability of base triplets containing N7-guanosine decreases in the order 7G*G- C >> 7G*C-G >>7G*A-T = 7G*T-A with affinities compara- ble to those of m5C. Nearly the same results were obtained for the pyrazole analogue PI. Its chemical isomer P2 [deoxyribose is linked to pyrazole N2 for P2 and N1 and PI, see Fig.5(C)], did not show reasonable affinity to any Watson-Crick base pair; this is probably due to energetically unfavourable distortion of the third-strand backbone added to the fact that the methyl group of P2 disfavours the anti conformation. The known tendency of 8-substituted adenosine analogues to be predominantly in the syn conformation led to the use of 8-oxoadenosine as a C+ substitute [Fig. 5(B)]. T, for a given triplex containing 8oXA*G-C was 22 "C at pH 7.0 and 8.0, whereas the corresponding transition in the control triplex (containing canonical C+*G-C) decreased from 28 "C at pH 7.0 to 17 "C at pH 8.0. goxA is also able to form goXA*C-G and goxA*U-A triads, the latter one being stabilized by a hydrogen bond between the N6 exocyclic amino group of 8-oxoadenosine and the O4 of uridine.However, these mismatched triplets are significantly less stable than goxA*G-C. The common drawback of these compounds is the lack of structural isomorphism between B*G-C (B = PI, 7G, goxA) and T*A-T triplets, which prevents complex formation for P2. Although NMR analysisl5 does not reveal any major backbone distortion for PI and 7G, the three orders of magnitude decrease in the affinity of 7G or P1-containing oligonucleotides for targeting an alternated (GA), vs. a contiguous G6 site16 Pv"Pu-Pv Triplexes Pu*Pu-Pv Triplexes Fig. 2 The two families of homopurine-targeted triple helices. In X*Pu-Py, '-' and '*' represent Watson-Crick and Hoogsteen hydrogen bonds, respectively.Half-arrows indicate the relative deoxyribose-phosphate (dR) backbone orientation. Chemical Society Reviews, 1997 65 A dk dk dk C+ isoc PYDDA rnSoXc B dR dk N3-Htautomer N'-H tautomer (a) G -(N3-H) (b) G -(N'-H) Fig. 3 Pyrimidine-like cytidine mimics. ISOC, pyDDA and m5oxC display the hydrogen bonding pattern of protonated cytidine (A). Possible N3-H and N1-H tautomeric forms of pseudoisocytidine (,,oC) (B). N3-H (a) and N1-H (b) tautomers of lS0C in Watson-Crick type I~~C-G base pair (C). confirms the energy penalty for nonisomorphism of adjacent triplets. 2.2 Pu*Pu-Py triplex The obvious advantage of Pu*Pu-Py triplexes is their pH independence: no protonation is required for the G*G-C, A*A- T (or T*A-T) base triplets, which are stabilized by two reverse- Hoogsteen-type hydrogen bonds between bases of the third strand and the purine strand of the Watson-Crick duplex (Fig.2). T*A-T triplets are generally preferred to A*A-T base triplets because the latter, although well characterized for homopolymers, have stabilities critically dependent on the length of the homoA tract and on the presence of divalent cations. Reverse-Hoogsteen T*A-T triplets are able to bind within a G*G-C triple helical structure, but triplexes are far from isomorphism (Fig. 6); A*A-T and G*G-C triplets are not isomorphous either, but more so than T*A-T/G*G-C. The net result is that the stability of Pu*Pu-Py-type triple helices is usually determined by the content of G-residues.2.2.I Inhibition by monovalent cations The development of G-rich TFOs as potential antigene drugs is hampered by the observation that triple helix formation is inhibited by monovalent cations, and especially by potassium which is the predominant intracellular cation. Inhibition is due to the involvement of the TFO into another process. At physiological K+ levels, guanosine-rich oligonucleotides self- associate by stacking of K+/guanine quartets [Fig. 7(A)] which are stabilized by a combination of mutual bidentate H-bonding and coordination of the four 06 atoms of guanosine to K+ located at the centre of the quadruplex. A number of such structures have been identified, including intra- and inter- molecular associations between oligonucleotides.To overcome 66 Chemical Society Reviews, 1997 this problem, guanosine has been replaced by nucleoside analogues that introduce a steric hindrance or disrupt the H-bonding network of the G-tetrad. The larger van der Waals radius and decreased electronegativity of sulfur of 6-thioguan- osine17 (S6-dG), relative to oxygen, was expected to decrease the tetrad stability, but should not directly affect H-bonding of G*G-C triplets (Fig. 2). Experiments showed that S6-dG- containing triple helices were no longer sensitive to potassium, but were also less stable. Complete substitution of S6-dC for G in some model TFOs reduced binding affinity by more than 100-fold.Replacement of guanosine N7 H-bond acceptor by a C-H group (7-deazaguanosinel 8, 7dzaG) or an N-H group (9-deaza- guanosinelg 9dzaG) were supposed to disrupt the H-bonding network of the G-tetrad [Fig. 7(B)], while retaining the hydrogen bond donors and acceptors pattern for formation of dzaG*G-C triplets. Surprisingly, attempts to overcome the binding inhibition in KC1 by substitution of G by 7dzaG or 9dzaG failed. A a 0 dR 0 dR 0 B 0 H F1-G F2-A C H H (a) Fl*G-C (b)Fl*C-G (C) FP*A-T Fig. 4 The tautomers (A) of pyrido[2,3-d]pyrimidine nucleoside (F) bind either single or double stranded DNA. (B) Single-strand binding through FIG and F2-A Watson-Crick type base pairs. (C) Triple helix stabiliza- tion with F1 (a, b) and F2 (c).P1 (ant,) 'G (anti) B dR /CH3 HP1'G-C PS'G-C H Fig. 5 Structures of purine-like cytidine mimics. (A) In contrast to the anti glycosidic bond conformation of most purine nucleosides, *O~Ahas predominantly the syn conformation. Nonisomorphous hydrogen bonding schemes of 7G*G-C and *oxA*G-C (B), and of P1 *G-C and P2*G-C (C) triplets within a Py*Pu-Py triplex. G*G-C A*A-T \J t I , Fig. 6 Non-isomorphism of natural base triplets within a Pu*Pu-Py triple helix. (0)indicates the ribose C1'atom of the third strand. 2.2.2 Towards isomorphous base triplets As mentioned above, another problem within this triplex family is the non-isomorphism of base triplets, especially for the T*A- T/G*G-C triplexes. This results in deformation of the deoxy- ribose-phosphodiester backbone at the transition from a bicyclic purine to a monocyclic pyrimidine, thus decreasing the stability of triple helices.TFOs containing only purines could have a better chance to be isomorphic. 7-Deazaxantosine18 (dzaX) and 2-amin0purine2~ (amp) have been used instead of T (or A) for A-T base pair recognition (Fig. 8). dzaX seems to be very promising: in comparative studies under physiological K+ concentration, the TFO containing G and dzaX showed a greater than 100-fold increase of affinity for the target sequence as compared to the G/T TFO. amP*A-T triplets are isomorphous with G*G-C triplexes in the Hoogsteen orientation [Fig. S(C)] which is different from the typical Pu*Pu-Py reverse-Hoogsteen family [Figs.2,8(B)]. The synthesis of modified TFOs containing amp has been described but no binding experiments have been reported so far. 3 Towards a general triple helix-mediated DNA recognition scheme As discussed previously, triple helix formation is strictly limited to homopurine targets, as a consequence of DNA recognition Fig. 7 (A) G-rich oligonucleotides form G-tetrads in the presence of potassium cations. (B) Putative disruption of the tetrad H-bonding network for 9dzaG. <KAN,HI I dR H dzaX amP ,dR dR H G'G-C dzaX' A-T C ""IK H-N H G'G-C Hoogsteen hydrogen bonds amP'A-T Fig. 8 7-Deazaxantosine (dzaX) and 2-aminopurine (amp) (A) could form isomorphous base triplets with A-T in a G*G-C context, by reverse-Hoogsteen bonding for dzaX (B), or Hoogsteen bonding for amp (C) Chemical Society Reviews, 1997 67 via Hoogsteen-type hydrogen bonds.As further stages towards a general sequence recognition scheme, two types of sequences have been tackled: alternated oligopurine-oligopyrimidine tracts and quasi-homopurine sequences with a single pyrimidine base. 3.1 'Switched' triple helices When the DNA target is made of adjacent oligopurine- oligopyrimidine domains, a continuous oligopurine tract still runs along the helix yet switches strand at each domain junction. Short canonical triple helices of low stability could be formed with each homopurine sequence. However, if all short TFOs are linked via appropriate spacer groups, triplex formation is very much enhanced due to cooperative binding. This approach is called alternate-strand triple helix formation, or 'switched' triple helix.Indeed, short TFOs when joined together form a single third-strand oligomer that zigzags along the major groove, switching from one oligopurine strand to the next one on the other strand. Combining the two canonical triple helix families with opposite strand polarities (Fig. 2) leads to many examples of 'switched' triple helices.21 Non-natural a-ano-meric oligonucleotides,which generally bind in the reverse orientation as compared to p-oligonucleotides, are also attrac- tive blocks to enrich the set of 'switched' TFOS.~~ An alternate-strand recognition scheme exclusively using Py*Pu-Py triplexes is illustrated in Fig.9(A). The alternated purine (R)-pyrimidine (Y) target DNA sequence is R7Y7R7. The third strand is made of three short pyrimidine (Y) TFOs. On both ends, TFOs run in the same direction along the major groove, whereas the central one must have the opposite polarity to stay parallel to its R7 target which is located on the opposite DNA strand. The design of alternate third strands is a complex task requiring linkers adapted to each type of junction crossing the major groove. Molecular modelling is widely used to find optimally rigid linker groups with reduced entropy of binding. Several linkers have been described, such as hexaethylene glycol, propane- 1,3-diol and xylose derivatives.Recently, a direct base-to-base linkage has been described,21 that allows shorter linkers to be used for different types of junctions. Binding tests with a C3'-C3'U-linked oligonucleotide [Fig. 9(B)] as TFO (Py*Pu-Py motif) have shown triple helix formation, but the junction remains to be optimized to increase the complex stability. An as yet unanswered question is the effective participation of the bases close to the junctions in the stability and selectivity of the binding process. A ~~~~~~~~-5'-YYYYYYY 5 YYYYYYY ' 5'-NNNNNRRRRRRR YYYYY RRRRRRRNNN 3'-NNNNNYYYYYYY RRRRRRR YYYYYYYNNN0000000 3'-YYYYYYY-B 5'-YYYYYYYYYYY-dk-3' 3'-dk-YYYYYYYYYYY-5' Fig. 9 (A) Alternate strand recognition in a Py*Pu-Py context.(B) Structure of a short strand-switching pentamethylene C3'-3'U-linker. 68 Chemical Society Reviews, 1997 3.2 Direct pyrimidine base recognition Several solutions have been described to bypass interruption of the target purine stretch by a single pyrimidine. Non-specific stabilization by incorporation of an intercalator at an internal site of the TFO is a simple and efficient solution.23 Alter- natively, pyrazole-, imidazole-, 1,2,4-triazole- and tetrazole- substituted nucleotides have been used in front of a pyrimidine base in an otherwise Pu*Pu-Py triple helix.24 Azole incorpora- tion enhanced TFO binding when compared to that of a natural base. These five-membered heterocycles were not designed for recognition of any particular base but rather to maintain some stacking interaction with neighbouring bases.Furthermore their rather small size reduces steric hindrance, especially when facing T which has a protruding methyl group. The pyrazole derivative seemed to be of particular interest: despite showing only weak preference for T-A over G-C, it discriminated strongly against A-T and C-G base pairs. A fatal limitation to specific recognition of pyrimidine bases in Watson-Crick base pairs is that they possess only a single remaining hydrogen bond-forming site in the major groove (Fig. l), that results in an intrinsically lower stability. Furthermore, thymine binding is obscured by its bulky methyl group. Considerable efforts to solve these problems have been undertaken, yet only limited success achieved.Several non-canonical natural base triplets of intermediate stability, such as G*T-A in a Py*Pu-Py context and T*C-G (Pu*Pu-Py) were discovered following a systematic study.25 Both base triplets involve a single hydrogen bond between the third-strand base (G or T) and the target pyrimidine inversion (T or C) (Fig. 10). GT-A T'C-G Fig. 10 Non-canonical G*T-A and T*C-G base triplets in a Py*Pu-Py and a Pu*Pu-Py context, respectively. In contrast to the canonical purine recognition scheme, only a single hydrogen bond can be drawn between third strand and target bases. Some nucleoside analogues like nebularine26 (N) recognize C-G in a Pu*Pu-Py triple helix. Once again a single hydrogen bond is established, and the N*C-G interaction is weaker than G*G-C, A*A-T or T*A-T. Moreover, nebularine has compara- ble affinities for C and A since it provides a hydrogen bonding ~~~~~ acceptor site to the exocyclic amino group (Fig. 11).Thus solutions found so far have limited destabilization rather than improving pyrimidine recognition. 3.3 Other-strand recognition As stated above, proper pyrimidine recognition with energies comparable to that of purines requires more than one hydrogen bond. Since this is simply not feasible with pyrimidines, an attractive solution would be to bind the facing purine of the opposite strand instead. Along these lines, formycin A (forA) was suggested to form two hydrogen bonds with the guanosine of the other strand at C-G inversions in a Pu*PuaPy triplex,27 as shown in Fig.12(A). To do so however, a 3-5 A translation as well as a rotation around the ribose-phosphate backbone are required, which would result in severe backbone distortion. Incorporation of three forA*C-G instead of G*C-G resulted in a ten-fold increase in binding affinity,27 but this may also coincidentally be due to reduced steric hindrance of formycin relative to guanine. \ H N*T-A N*C-G dR N*A-T Fig. 11 Putative interactions between nebularine (N) and the four Watson- Crick base pairs in an antiparallel Pu*Pu-Py triple helix. Nebularine provides a hydrogen bonding acceptor site to the exocyclic amino groups of C and A. Indeed, when comparing forA*C-G and G*C-G triplets within the Pu*Pu-Py backbone geometry [Figs.12(B) and (C)], formycin escapes the G(N1-H and N2-H)/C(N4-H) clash and may even take advantage of an attractive forA(N1)/C(N4-H) interaction. The future success of this approach will rely on the possibility of crossing the DNA duplex major groove with more extended structures and without backbone distortion. 3.4 Base-pair recognition The most promising way to reconcile affinity and selectivity is to use extended heterocyclic systems that can simultaneously Utilize all major groove hydrogen-bonding sites of a Watson- Crick base-pair for molecular recognition. The first example of A dR such a strategy was the use of 4-(3-benzamidophenyl)imi-dazole28 (D3) expected to hydrogen bond simultaneously C(N4- H) and G(06) of C-G in a Py*Pu-Py triple helix (Fig.13).Other base pairs were supposed to lead to steric clashes with adenine, cytosine and the methyl group of thymine (Fig. 13). Un- fortunately, experimentation showed the following order of stabilities: D3”T-A = D3*C-G > D3*A-T > D3”G-C. The rationale to this was found later, when it was shown that D3 interacts by sequence-specific intercalation, rather than by hydrogen bonding29 When N4-(3-acetamidopropy1)cytidine (adPC) was incorpor- ated in front of C-G inversions, a considerable stabilizing effect was observed relative to third strands with N4-butylcytidine or N4-(3-carboxypropyl)cytidine whose side chains lack hydrogen bond donating groups. The length of the flexible acet-amidopropyl arm of aapC is sufficient to span the duplex major groove and allows the terminal amide to form an additional hydrogen bond with O6 of guanine in the target C-G base pair.To improve binding, a more rigid 6-amino-2-pyridinyl sub- \ H DB*C-G D3*T-A H/ D3*G-C DB*A-T Fig. 13 Best hydrogen bonding fit between the synthetic base analogue D3 and the four base pairs in a Py*Pu-Py triple helix C dR H ‘“A*C-G (“other strand” geometry) ‘“A’C-G (Pu*Pu-Py geometry) G*C-G(Pu*Pu-Py geometry) Fig. 12 (A) For cytosine recognition, forAcould hydrogen bond to the guanine of the opposite DNA strand at the expense of a ca. 5 8,translation (arrow) through the major groove. The dotted line structure indicates the position of guanine in a G*G-C triplex. forA(B)and G (C)facing the same C-G inversion.Chemical Society Reviews, 1997 69 stituent was introduced at the N4 position of cytidine (apyC), thus providing two hydrogen bonding sites for interaction with C-G and three for A-T30 [Fig. 14(A)]. Both triads involve an unusual imino tautomeric form of apyC [Fig. 14(B)] which is supported by 1H NMR and UV spectroscopy of apW. A dR dR amino imino Fig. 14 (A) Possible hydrogen bonding schemes of the irnino tautomer of apyC with C-G and A-T base pairs in a Pu*Pu-Py triple helix. (B) The amino/irnino tautomeric equilibrium of apyC / As shown in Fig. 14(A), the pyridine ring spans the major groove so as to place the 6-amino group of apyC within hydrogen bonding distance of guanine 06 or thymine 04.This interaction is vital and removal of the amino group results in a much decreased affinity or even in no triplex formation at all. Additional stabilization could come from stacking of the pyridine ring with neighbouring bases.Experimentally, stability of the apyC*G-C base triplet was found to be comparable to that of the canonical C+*G-C; unfortunately A-T, and to a lesser extent T-A and G-C base pairs, also recognized. Here again, design [Fig. 14(A)] and experimental facts did not coincide. The apyC-containing oligonucleotide showed two distinct inter-convertible binding modes: hydrogen bonding, as expected, and intercalation as already established for 4-(3-benzamido-pheny1)imidazole (D3, see above).Recently two novel structures for cytidine-guanosine recog- nition have been designed (Fig. 15) and tested in chloroform. Benzaminoimidazole-glycyl31 (BIG) and 2-methyl-8-(N'-butyl- ureido)naphth[ 1,2-d]irnida~ole3~ (UNI) were shown by NMR to bind to a C-G base pair by hydrogen bonding to both bases. Such experiments provide preliminary evidence about com- \ \ H H BIG%-G UNI*C-G Fig. 15 Extended BIG and UNI nucleic base analogues for specific C-G base pair recognition / Fig. 16 A third-strand oligonucleotide centered in the major groove would be able to switch from one DNA strand to the other through the @/a anomerism. 70 Chemical Society Reviews, 1997 plementarity with a given base pair. However they do not take into account base stacking interactions nor isomorphism which are also essential features for triple helix stability. 4 Conclusion A great deal of intellectual effort and organic synthesis has been invested in molecular recognition of double stranded DNA, with the aim of developing antiviral drugs and controlling endogeneous gene expression. Indeed, drugs based on the antigene strategy would have many advantages over those based on the antisense (anti-messenger RNA) strategy.Yet, up to now, the antigene strategy has essentially benefitted from pharmaco- logical advances obtained with antisense oligonucleotides, such as nuclease resistance and intracellular delivery33 due to their common generic structures. Increasing the stability of triple helices in physiological media is just one of the problems related to gene targeting (the adverse corollary being the slow kinetics of triplex formation34).In this review we deliberately focused on ways of extending triplex formation to non-homopurine sequences, as although polyPu-polyPy stretches are not uncommon in the mammalian genome, the conse-quences of making random sequences amenable to targeting are obvious. This remains a challenge for organic chemists.35 The challenge has been taken up as illustrated by the numerous examples described in this review. They highlight, besides unpredictable synthetic difficulties, predictable require- ments for modifying nucleotides, in the area of complementar- ity, triplex isomorphism and base stacking.They also illustrate effects such as intercalation as opposed to hydrogen bonding for extended molecules, and tautomeric ambiguity of heterocyclic bases. Undoubtedly, integration of this knowledge will help chemists to play a pivotal role in the success of drug development based on sequence-selective DNA recognition. Added in proof: Besides this review, our real contribution to a general triple helix mediated DNA recognition scheme has been rather small. Our approach was based on the natural complementarity of primary amides with adenine and of guanidine with guanine. The major difficulty of pyrimidine recognition was circumvented by targeting the opposite purine bases with the same molecular groups, but of the unnatural a-anomer nucleoside.Thus molecular recognition of the four nucleic bases (A, T, G, C) is achieved with only two unnatural bases (X, Z),degeneracy being removed by the anomeric (a, p) pluralism. For this to be possible, and for the four base triplets to be isomorphous, several geometrical constraints apply which severely limit the number of chemically reasonable structures (Fig. 16). Organic chemistry along these lines is in pro- gress.36 5 References 1 G Felsenfeld, D. Davies and A Rich, J Am Chem Soc, 1957, 79, 2023 2 H. E Moser and P. B. Dervan, Science, 1987, 238, 645. 3 T. Le Doan, L. Perrouault, D. Praseuth, N. Habhoub, J. L. Decout, N. T Thuong, J Lhomme and C HClkne, NucZeiL Acids Res ,1987,15, 7749 4 Y.-K Cheng and B.M. Pettitt, Prog Biophys Mol Brol , 1992, 58, 225. 5 J.3. Sun, T. Garestier and C HClkne, Curr Opin Struct Bid , 1996,6, 327 6 J. S. Lee, M. L. Woodsworth, L. J P Latimer and A. R. Morgan, Nucleic Acids Res , 1984, 12, 6603 7 L E. Xodo, G. Manzini, F Quadrifoglio, G. A. van der Marel and J H. van Boom, Nuclerc Acids Res , 1991, 19, 5625 8 A. Ono, P 0. P. Ts’o and L -S Kan, J Am Chem Soc , 1991, 113, 4032. 9 U. von Krosigk and S. A. Benner, J Am Chem Soc, 1995, 117, 5361 10 G. Xiang, R Bogacki and L. W McLaughlin, Nucleic Acids Res , 1996, 24, 1963. 11 A. B. Staubli and P. B. Dervan, Nucleic Acids Res , 1994, 22, 2637 12 J S. Koh and P. B. Dervan, J Am Chem Soc, 1992,114, 1470 13 P S. Miller, P Bhan, C. D Cushman and T L Trapane, Bioc hemistry, 1992,31, 6788 14 J.Hunziker, E. S. Priestley, H. Brunar and P B Dervan, J Am Chem Soc, 1995, 117, 3233 15 1 Radhakrishnan and D. J. Patel, Biochemistry, 1993, 32, 11 228 16 E S Priestley and P B. Dervan, J Am Chem Soc , 1995, 117,4761. 17 T. S. Rao, R. H Durland, D. M Seth, M. A. Myrick, V Bodepudi and G. R Revankar, Biochemistry, 1995, 34, 765 18 J F Milligan, S H. Krawczyk, S Wadwani and M D Matteucci, Nucleic Acids Res , 1993, 21, 327 19 T. S. Rao, A F. Lewis, R. S. Durland and G. R. Revankar, Tetrahedion Lett, 1993, 34, 6709. 20 V Roig, R Kurfurst and N T Thuong, Tetrahedron Lett, 1993, 34, 1601 21 B -W Zhou, C. Marchand, U. Asseline, N T Thuong, J.-S Sun, T. Garestier and C. Helene, Bioconj Chem , 1995, 6, 516 22 S €3 Noonberg, J.-C.Franqois, D. Praseuth, A.-L Guieysse-Peugeot, J. Lacoste, T Garestier and C. Helene, Nucleic Acids Res , 1995, 23, 4042. 23 B -W Zhou, E. Puga, J -S. Sun, T. Garestier and C HClkne, J Am Chem Soc, 1995, 117, 10425. 24 R H Durland, T. S. Rao, V Bodepudi, D. M Seth, K Jayaraman and G. R Revankar, Nucleic Acids Res , 1995, 23, 647. 25 W A Greenberg and P B. Dervan, J Am Cheni Soc , 1995, 117, 5016. 26 H U. Stilz and P B Dervan, Biochemistry, 1993, 32, 2177 27 T. S Rao, M E. Hogan and G. R. Revankar, Nucleosrdes Nucleotides, 1994, 13,95. 28 L. C. Griffin, L. L. Kiessling, P. A. Beal, P. Gillespie and P B. Dervan, J Am Chem Soc , 1992, 114,7976 29 K M. Koshlap, P Gillespie, P. B. Dervan and J. Feigon, J Am Chem Soc, 1993, 115,7908 30 C.-Y Huang, G. Bi and P. S. Miller, Nucleic Acids Res, 1996, 24, 2606 31 S. Sasaki, S. Nakmashima, F. Nagatsugi, Y. Tanaka, M. Hisatome and M. Maeda, Tetrahedron Lett, 1995, 36, 9521. 32 S C. Zimmerman and P. Schmitt, J Am Chem SOL, 1995, 117, 10769 33 0 Boussif, F. Lezoualc’h, M A. Zanta, M. D. Mergny, D Scherman, B. Demeneix and J. P. Behr, Proc Nut1 Acad Scz USA, 1995, 92, 7297 34 S Baillet and J. P Behr, Tetrahedron Lett, 1995, 36, 8981 35 N. Schmid and J. P. Behr, Tetrahedron Lett, 1995, 36, 1447 36 S 0 Doronina, S Blanalt-Feidt and J P. Behr, to be published. Received, 25th October I996 Accepted, 22nd November I996 Chemical Society Reviews, 1997 71