In the previous section of this review we discussed the biological impact of several different DNA-interacting compounds including the nature of their interaction with DNA, mechanism of action, and known anti-cancer activity. In addition to small molecules that interact with specific bases and base sequences, a number of compounds are being developed which target DNA secondary structures such as DNA tetraplexes and quadruplexes, hairpins, and Holliday junctions. Some of these DNA structures have been implicated in regulating numerous nuclear activities, and represent an exciting new area of research into potential anti-cancer targets as well as for treatment of numerous other human diseases including diabetes and neurodegenerative disorders.
DNA quadruplexes (telomeres)
Telomeres are the repetitive DNA sequences (TTAGGG) at the ends of chromosomes that protect the 3' ends from degradation and inappropriate repair activities and interact with a number of different proteins forming the telomeric complex [113-115]. In normal proliferating cells, telomeres are shortened with each round of replication and telomerase expression is negligible. Eventually, telomeres become so short that they are no longer capable of protecting chromosome ends, leading to chromosome fusions and erosion. This results in the induction of "telomere-induced senescence" and loss of cell viability. In many cancer cells, however, a short telomere length is maintained during cell divisions in part because of increased telomerase activity. In fact more than 90% of all human cancers have increased expression of telomerase which is one reason why it has been suggested as a target for anti-cancer drug design efforts. In addition to the telomerase enzyme, the telomeric DNA structure is being examined for its ability to be targeted by anti-cancer treatments[114,116-121].
Telomerase is a multi-component enzyme comprised of protein and nucleic acid. The two main components are the RNA moiety (hTER) and the catalytic subunit (hTERT), although there are several additional regulatory binding proteins (for example: HSP90, p23, and TEP1). There are numerous telomerase inhibitors in development. The main strategies include anti-sense oligonucleotides, peptide nucleic acids and ribozymes targeting the RNA component of telomerase, dominant negative versions of the hTERT subunit, small molecule inhibitors of the enzyme complex, and disruption of the G-quartet (see below)[120,122,123]. While decreases in telomerase activity have been achieved using these approaches, this has not necessarily led to reduced cancer cell viability.
The G-quartets of the telomere (3' G-rich overhang of 150–200 bp that form the DNA secondary structure, the G-quartet) are stacked tetrads arising from planar associations of four guanines in a cyclic Hoogsteen hydrogen-bonding arrangement (Figure 5) . G-quartets can be stabilized by sodium and potassium ions, and this stabilization can inhibit telomerase activity. As such, the ability of small molecules to interact with and presumably stabilize these secondary structures as a means of inhibiting telomerase has been a major drug design effort.
A number of small molecules have been identified that interact with G-quartets. Molecular modelling studies of anthraquinones predicted these compounds would interact with G-quartets by a threading intercalation model . Nuclear magnetic resonance studies have confirmed that the 2,6-diamidoanthraquinone BSU1051  interacts with and stabilizes the G-quartet, and inhibits telomerase activity. A 3,6,9-trisubstituted acridine  was also a potent (IC50 18 nM) inhibitor of telomerase. In A431 human squamous cell carcinoma xenografts it showed a significant additional growth delay compared with paclitaxel alone, with no additional toxicity.
Cationic porphyrins, exemplified by TMPyP4 , are another class of agents that were predicted to bind to G-tetrads by interactive stacking. Two independent research groups showed by a variety of methods (spectroscopy, CD, NMR) that these compounds do interact with parallel and anti-parallel G-quadruplexes[32,128]. The way in which the compounds interact with the DNA is not entirely clear, but most likely involves external stacking of the porphyrins relative to the G-quadruplex. A third class of G-tetrad interacting compound is typified by the perylenetetracarboxylic diimide PIPER  that demonstrates similar binding attributes to the porphyrins. Interestingly these compounds may not merely bind to such DNA structures, but may also induce their formation in cells .
In addition to telomeres, G-quadruplex sequence motifs have been identified in other regions of the genome, particularly in the upstream promoter regions of a number of oncogenes . Within the c-MYC promoter, the nuclease hypersensitive element III1 (NHE), corresponding to bases 2186–2212 in human c-MYC , has been known to play an important role in regulation of c-MYC expression. Insight into the regulatory nature of this region was first demonstrated when synthetic oligonucleotides with sequences complementary to the NHE c-MYC coding were capable of blocking c-MYC expression[132,133]. Further in vitro studies using c-MYC promoter DNA demonstrated this region was capable of forming quadruplex structures (G-rich strand) and i-tetraplexes (C-rich strand) [34-36]. These latter structures are formed based on hemiprotonated cytosine+/cytosine base pairs containing three stabilizing hydrogen bonds between them. The four-stranded structure is composed of two parallel-stranded duplexes zipped together in an anti-parallel configuration [134-136]. In the case of the c-MYC promoter the formation of the G-quadruplex appears to be the biologically relevant structure.
The use of G-quadruplex stabilizing compounds targeted at telomeres prompted the investigation of affects within the c-MYC locus. Using the cationic porphyrin, TMPyP4, Grand et al. (2002) and Siddiqui-Jain et al. (2002) have demonstrated repression of transcriptional activation of c-MYC in cells based on G-quadruplex stabilization. In addition, mutational analysis by replacement of a G to an A within this G-rich region which is predicted to destabilize quadruplex formation results in a 3-fold increase in c-MYC expression also points to a biological role for this secondary structure .
Hairpins and Holliday junctions (mini- & micro-satellites)
A variety of secondary structures (hairpins, cruciforms) have now been detected in the genomic DNA of a number of prokaryotic and eukaryotic species, including humans. These structures are associated with regulation of gene transcription, possibly as recognition binding sites, and may be targets for selectively binding drugs that could either block or enhance transcription. Thus, the potent transcription inhibitor actinomycin D  and analogues bind at least 10-fold more tightly to the hairpin conformation formed from the single-stranded DNA 5'-A7TAGT4A3TAT7-3' than to same strand fully duplexed to its complementary sequence. Similar results have been reported for actinomycin D binding to GC-rich hairpin sequences .
DNA four-way junction structures (Holliday junctions), often created by mismatches, are known to occur during DNA replication, repair and recombination, making them potential targets for the development of novel antiviral and antibacterial agents. Human TOPO IIβ also binds preferentially to four-way junction DNA, suggesting that it might operate via such structures. It has long been known that some intercalating agents show preferential binding to such branched DNA sequences. Recently, a series of crystal structures of both mono-intercalators [24-26,144-146] and dimeric drugs [27,28,144,147], complexed with short oligonucleotides, have shown that the ligands can induce Holliday junction-like DNA structures. It remains unclear whether this reflects the situation in longer DNA.
DNA duplexes of certain sequences can bind a third DNA (or modified DNA) strand, to form a triple helix. The third strand binds to the existing base pairs in a manner called Hoogsteen base pairing; T binds to A (but in a different way to normal) and protonated C binds to G, to form the triplets T.AT and C+.GC. In this way a homopyrimidine third strand can bind to its complementary sequence of duplex DNA. Triple helices are generally less stable than the duplexes, and are thought to be induced naturally in the genome by supercoiling, as triple helix formation can relieve torsional stress. However, some small molecules can stabilize triple helixes by preferentially binding to them. Many such compounds, like those that preferentially bind quadruplexes, tend to be intercalating agents of larger than usual planar area, with side chains in the correct disposition to bind in one or more of the three grooves of the triple helix. One of the best-studied is the benzo [e]pyridoindole BePI , that binds T.AT triplets. The related antiviral indoloquinoxaline 9-OH-B220  provides a very large stabilization of such triplet species, shifting the triplex-to-duplex equilibrium by up to 50°C . The other major class of triplex stabilizers are the aminoglycosides. One of the most effective is neomycin , which binds preferentially in the larger Watson-Hoogsteen groove rather than one of the regular duplex grooves .