Triplex forming oligonucleotides: sequence-specific tools for gene targeting

Abstract

Sequence-specificity is the key to effective genetic targeting. With specificity, targeted genes can be manipulated in multiple ways; without it, gene therapy agents become loose canons within cells. Triplex forming oligonucleotides (TFOs) bind in the major groove of duplex DNA with high specificity and affinity. Because of these characteristics, TFOs have been proposed as homing devices for genetic manipulation in vivo. Here we review work demonstrating the ability of TFOs and related molecules to alter gene expression and mediate genome modification in mammalian cells. Recent studies have established that TFOs can mediate targeted gene knock out in mice, laying the foundation for the potential application of these molecules in human gene therapy.

Received July 6, 2001; Accepted July 16, 2001.

INTRODUCTION

The binding characteristics and sequence-specificity of triplex forming molecules, a diverse class of DNA oligonucleotides and their analogs, give these molecules tremendous potential as gene therapy and molecular biology agents. The ability of nucleic acids to form triple helices was discovered in 1957 when Felsenfeld et al. (1) first reported the formation of such complexes. Third strand binding was found to be weak relative to double stranded Watson–Crick binding, but it was sequence-specific and was stabilized in the presence of divalent cations. The first example of sequence-specific functional use of a third strand was shown by Morgan and Wells in their 1968 paper describing the ability of a bound RNA third strand to inhibit transcription (2).

The potential for specific binding has led to the extensive study of molecules which can form triple helices. Triplex targets have primarily included genomic duplex DNA and single-stranded messenger RNA. Third strand binding to genetic targets can be accomplished by natural and modified DNA oligonucleotides and DNA oligonucleotide mimics such as peptide nucleic acids (PNAs) and polyamides.

DNA triplex forming oligonucleotides (TFOs) bind in the major groove of duplex DNA. They are restricted to binding DNA sites which have runs of purines on one strand and pyrimidines on the other; TFOs and their corresponding targets are typically 10–30 nt in length. Though DNA TFOs can be either polypurine or polypyrimidine molecules, they bind to the purine-rich strand of their target. There is an established binding code for all TFOs, and it is this code that allows sequence-specific targeting (3).

TFOs have been used in a vast array of approaches which share the common goal of altering gene expression. Their chemical nature, binding code, and ability to change gene expression via transcriptional regulation, mutagenesis and recombination will be discussed in this review.

TRIPLEX APPLICATIONS

The sequence-specificity of triplex forming molecules and their ability to compete successfully with other DNA binders gives these molecules great potential as tools to alter gene expression. Through recombination and mutagenesis, permanent heritable genetic changes can be made, and through inhibition of RNA transcription, gene expression can be regulated.

Six key means by which TFOs have been used to directly alter gene expression are illustrated in Figure 1. Figure 1A shows how transcriptional processes can be stopped by two different strategies. Both strategies are based on the binding of the TFO to a target site and the subsequent creation of a physical block to a normal cellular process. Figure 1B highlights both the use of TFOs to deliver a tethered mutagen as well as the ability of a TFO to cause mutagenesis within its binding site even when not conjugated to a DNA reactive molecule. Figure 1C shows two examples of experimental protocols in which TFOs have been used to promote homologous recombination: homologous recombination between repeat sequences which can be induced by TFO binding alone and homologous recombination of a target genome site with a donor fragment that is localized to its target by the TFO.

TRANSCRIPTION

Initial steps were made in the direction of altered gene expression via the inhibition of transcription of extrachromosomal reporter genes. It seems that inhibition is best achieved through the binding of regulatory rather than coding sequences. The binding of TFOs in regulatory regions is believed to compete with the binding of transcription factors, thereby preventing the initiation of transcription (Fig. 1A).

In vitro studies, have shown the following: highly specific TFO binding in physiologic conditions at regulatory sites (4); inhibition of regulatory protein binding secondary to the presence of bound TFOs (5–7); and inhibition of transcription in the presence of bound TFOs (6,8,9). In vivo extrachromosomal studies echoed these results (4,10).

Extension to endogenous chromosomal targets has followed these successes in extrachromosomal transcription inhibition. Studies have shown that TFOs can bind to their genomic target in vitro (11) and in cells within the chromatin structure (12–15). However, some studies show that triplex binding sequences which are tightly associated with nucleosomes cannot be targeted unless the nucleosome association is removed (16,17), and that pre-existing triplexes prevent the association of DNA onto nucleosomes (18). One early success in transcription inhibition was with the c-myc oncogene. Work was done initially in vitro (19) and then in vivo (20). Follow-up work strengthened these results (4,21,22).

Despite early mixed results, there have also been some successes in targeting the coding region rather than the regulatory region of a gene (Fig. 1A). It has been shown that it is possible to block various polymerases, including RNA polymerase II (23–25) and RNA polymerase III (26). Natural DNA oligonucleotides with phosphodiester backbones are unable to arrest elongation without covalent crosslinking of the TFO (23,27–29). However, recent studies using N3′–N5′ phosphoramidate linkages instead of the normal phosphodiester backbone have shown that these TFOs can bind transcribed DNA regions and arrest elongation in vitro (30,31) and in vivo (32,33).

Linear or clamp PNAs also stop RNA polymerases if hybridized to the transcribed strand (34,35). It has been observed that PNA association onto DNA to form a triplex is slow in moderate (physiologic) salt conditions; however, this process can be accelerated if the target DNA is being actively transcribed. The binding of the PNA to an open transcription bubble on the template strand stops further transcription and decreases gene expression (36).

Using TFOs and PNAs, it is also possible to target binding to promoter regions in order to increase transcription. A TFO designed to compete with the binding of a repressor protein increased gene expression in B cell culture lines (37). Also, fusion of a TFO to a transactivator protein domain increased gene expression on a plasmid construct on which a triplex had been formed prior to transformation (38). PNAs are unique among TFOs in that they strand invade and create a single-stranded DNA loop; when this loop is the coding strand of DNA, this structure resembles a transcription bubble and has been shown to stimulate transcription in vitro and in mammalian cells in culture (39,40). In the latter case, the binding of two PNA clamps to nearby sites in the mouse γ-globulin gene promoter lead to increased levels of γ-globulin gene expression. Primer extension analysis indicated that the increased transcription was initiated not only from the site of the PNA induced displaced strand ‘bubble’, but also from the natural transcription start site, downstream of the bound PNAs.

GENOME MODIFICATION

TFOs can also be utilized to create permanent heritable changes in the genome. Our lab and several others have focused upon this aspect of TFO technology in search of methods to harness their sequence-specific properties. A number of studies from our group have shown that DNA TFOs can be used to increase rates of site-specific mutagenesis and site-specific recombination (41–45). Furthermore, PNAs have been used for targeted mutagenesis (46).

Triplex-directed mutagenesis

Initial predictions regarding the capability of TFOs to cause site-directed mutagenesis were based upon their sequence-specificity. TFOs were envisioned to be homing agents which could deliver an agent of choice to a specific sequence according to the binding rules discussed below (Fig. 1B). One such agent was psoralen, a mutagen that intercalates and binds covalently to duplex DNA when activated with long wavelength UVA irradiation. Psoralen-linked TFOs were able to site-specifically increase mutagenesis in plasmid targets in vitro (47) and in mammalian cell culture lines (41). It was further shown by Wang et al. (41) that plasmid mutagenesis rates in cells were highly dependent upon binding affinity and delivery concentration. TFOs with K_ds in the range of 10^–9 M were able to induce mutagenesis, whereas those with K_ds of 10^–6 M were not (41). Studies with psoralen-linked TFOs also produced the first evidence that chromosomal sites could be targeted for mutagenesis (44,48). Majumdar et al. (48) used psoralen linked TFOs (which also contained either acridine or pyrene to provide additional binding affinity via intercalation) to mutate a chromosomal copy of hprt in a CHO cell line; 85% of the observed mutations occurred in the triplex target region. In related work, Vasquez et al. (44) used mouse fibroblast cell lines containing multiple chromosomal copies of a λsupFG1 reporter vector. A 6–10-fold induction of site-specific mutagenesis over background was observed, and absolute mutagenesis rates were in the range of 0.1%. Similar effects have also been seen in yeast engineered to carry a triplex target site (49).

Studies carried out by Wang et al. (42) made the surprising discovery that TFOs not linked to any mutagenic agent could still cause increases in mutagenesis in plasmids transfected into mammalian cell lines (Fig. 1B). Using an SV40 vector containing a TFO target site within the supFG1 reporter gene, Wang et al. (42) were able to show specific induction of mutation in the target site. This effect was absent in either xeroderma pigmentosum group A (XPA) and Cockayne’s syndrome group B (CSB) cell lines. XPA individuals are missing the protein XPA which is a key recognition factor in the nucleotide excision repair (NER) pathway. CSB individuals are deficient in the protein CSB which is specifically necessary for the transcription-coupled repair subset of the NER. The TFO-dependent increase in site-specific mutagenesis could be reconstituted in these cell lines if they were co-transfected with XPA or CSB cDNA respectively. This work suggested that triplex structures can provoke DNA metabolism in part via recognition by the NER pathway, and that the processing of these unusual structures can in some circumstances be error-prone.

Recently, Vasquez et al. (45) demonstrated that systemically administered TFOs could induce mutation at specific genomic sites in the somatic cells of adult mice (Fig. 2). In this work, a 30mer purine TFO which contained a 3′ propanolamine group to prevent exonuclease-mediated degradation was injected intraperitoneally for 5 consecutive days. The TFO was targeted to chromosomal copies of a λsupFG1 reporter vector. After an additional 10 days, the mice were killed and tissues were taken for mutation analysis. In general, mice treated with the sequence-specific TFO had a 5-fold elevated mutation rate in the targeted supFG1 gene, but not in the non-targeted cII gene. All tissues tested showed TFO-induced mutagenesis, except the brain which had no mutagenesis over background; this is consistent with TFOs being unable to cross the blood–brain barrier. This work established that site-directed DNA binding molecules, upon systemic administration, can mediate gene targeting in vivo in whole animals.

Triplex-induced recombination

It has also been shown in a variety of systems that TFOs can stimulate increased recombination in both episomal and chromosomal DNA in mammalian cells. Because double-strand DNA breaks are recombinagenic, several labs have successfully pursued the development of TFOs linked to DNA cleaving agents (50). It has also been possible to detect increases in recombination in reporter constructs that have tandem repeats of the same gene carrying inactivating mutations at different positions. This was seen in plasmids (43) and in chromosomal reporter constructs (51). In the case of the unconjugated TFOs, the increased recombination effect is lost in XPA-deficient cells. The effect can be restored by expression of XPA cDNA in such cells (52). This indicates that the ability of a triplex structure to provoke DNA repair via the NER pathway is critical to the mechanism of TFO induced recombination. With the psoralen-conjugated TFOs, the induced recombination is only partially diminished in XPA-deficient cells, suggesting that the psoralen-triplex structures are processed via both NER-dependent and independent pathways (52). This finding is consistent with recent studies of psoralen crosslink repair (53).

In chromosome targeting work, two mutant thymidine kinase (TK) genes were integrated into a single chromosomal site in mouse fibroblasts (Fig. 1C) (51). Transfection (via cationic lipids) of the cells with high-affinity TFOs targeted to a region between the two TK genes stimulated recombination at a frequency of ∼6–7-fold above background. However, when the TFOs were microinjected into the nuclei of the cells (∼72 000 copies/cell), the yield of recombinants increased, up to 3000-fold over background. These experiments established that TFOs could induce recombination at a chromosomal locus in mammalian cells, but also highlighted the need for effective intranuclear delivery to achieve meaningful levels of targeted genome modification. In addition, the design of the recombination substrate was such that the induced recombination was triggered by triplex formation at a position between the genes and as much as 600–700 bp away from either gene. This raises the possibility that more efficient induction might be achieved by targeting a site within one of the genes.

The observation that the third strand binding can provoke DNA repair and stimulate recombination led us to develop a strategy to mediate targeted gene conversion using a TFO linked to a short donor DNA fragment (Fig. 1C). The donor fragment was homologous to the target site, except for the base pair to be corrected (54). In this bi-functional molecule, the TFO domain mediates site-specific binding which directs the donor to the desired gene. This binding is proposed to trigger repair and thereby sensitize the target site to recombination. The tethered homologous donor fragment can then participate in recombination and/or gene conversion.

Using a bi-functional oligomer with a 40mer donor domain and a 30mer TFO domain, correction of a single base pair mutation in a supFG1 reporter gene within an SV40 vector in COS cells was achieved. Correction frequencies were in the range of 0.1–0.5% with the full bi-functional molecule. Use of either domain alone or of either domain substituted with heterologous sequences reduced activity by 10-fold or more. That the donor domain alone did mediate any gene correction would be expected from previous work (55,56). However, there was a clear synergism between the donor and the TFO. In addition, Gamper et al. (57), in an in vitro study, demonstrated strand invasion of a supercoiled plasmid DNA by a bi-functional oligonucleotide consisting of a guiding TFO domain and a strand invading homology domain. The Gamper work provides direct physical evidence that a TFO domain can promote strand invasion. A similar analysis of DNA binding by a bi-functional oligonucleotide was carried out by Fresco and colleagues (58).

We have also examined the extent to which triplex-induced recombination could be detected in human cell-free extracts. Using a plasmid-based reversion assay, a bi-functional oligomer consisting of a 30mer TFO domain linked to a 40mer donor domain was found to mediate gene correction in vitro at a frequency of 4.5 × 10^–4, 20-fold above background and 4-fold greater than the donor segment alone (59). Interestingly, physical linkage of the TFO to the donor was unnecessary, as a co-mixture of separate TFO and donor segments also yielded elevated gene correction frequencies. This latter result indicates that a TFO can promote recombination between the target site and a third, unlinked DNA molecule, if that molecule is present in high enough concentration. Immunodepletion experiments confirmed a role for the XPA protein as well as for the HsRad51 recombinase (59).

BINDING CODE AND CHEMICAL STRUCTURE: DNA OLIGONUCLEOTIDES, PEPTIDE NUCLEIC ACIDS AND POLYAMIDES

The utility of TFOs for gene targeting derives from the existence of a third strand binding code (3). Regardless of the specificity that can be achieved, triplex formation is restricted to regions of DNA with purines on one strand and pyrimidines on the other. TFOs themselves must comply to either a purine or pyrimidine motif. In addition, there exist limitations to TFO binding under physiologic conditions. Accordingly, substantial effort has been expended in the study of the triplex binding code and the means by which to extend it.

Regardless of composition, DNA TFOs bind the purine-rich strand of the target. Though it was initially observed that only the purine oligonucleotides could bind their target under physiologic condition (most notably neutral pH) the advancement of base substitution technology has allowed the engineering of pyrimidine TFOs which can also bind effectively under physiologic conditions. Typically, TFOs must bind with K_ds in the range of 10^–9 to have biologic function (41).

The purine anti-parallel triplex motif utilizes TFOs made of A and G. They bind to the major groove of DNA in an anti-parallel sequence-specific manner by forming reverse Hoogsteen hydrogen bonds with the purine-rich target strand (3,60). Gs within such TFOs will bind the G of a G:C Watson–Crick pair; the A will bind the A of an A:T Watson–Crick pair (Fig. 3) (3,60,61). It is also known that T may be included in these TFOs without a significant binding affinity reduction. In these cases, the T binds to the A of an A:T Watson–Crick pair (19,60).

Though purine TFOs bind well at physiologic pH, physiologic K⁺ concentrations can inhibit binding in particularly G-rich TFOs. This is a result of the G quartets and other secondary structures that form within the oligonucleotides (62). This binding inhibition has been addressed with a variety of chemical substitutions. Replacement of natural bases with modified bases such as 6-thioguanine (63–65) or 7-deazaxanthine (66,67) has reduced K⁺ inhibition of triplex formation. Conversely, triplex formation is dependent on the presence of Mg²⁺ which mediates charge neutralization when the three anionic phosphodiester backbones come together. Modification of the anionic phosphodiester backbone of the TFO can reduce charge repulsions, decrease K⁺ inhibitions, lessen Mg²⁺ dependence and so facilitate triplex formation (17,68,69).

Pyrimidine TFOs also bind in the major groove of duplex DNA targets using Hoogsteen-type pairing; the pyrimidine TFOs bind in a parallel orientation to the target purine strand (70). Protonated Cs of these TFOs bind the G of a G:C Watson–Crick pair; Ts bind the A of an A:T pair (3). It is the requirement for N-3 protonation of C that drives the requirement of low pH (<6) for pyrimidine triplex formation; base modifications such as replacing cytosine with 5-methyl-2′-deoxcytidine allow binding at physiologic pH (71,72). Other successful replacements of cytosine include: 8-oxoadenine (73), 7,8-dihydro-8-oxoadenine (74), N7-2′-deoxyguanosine (75) 8-oxo-2′-deoxyadenosine (76) and 8-aminoguanine (77).

In the pyrimidine motif, certain backbone modifications are being used to improve binding and stability of TFOs under physiologic conditions. As with purine TFOs, changing the natural phosphodiester backbone can reduce charge interactions, reduce magnesium requirements and confer nuclease resistance to TFOs and the triplexes they form (78). Morpholino oligonucleotides and PNAs are both molecules in which the traditional backbone has been replaced by a charge neutral molecule that is still capable of holding the nucleobases in correct alignment (79).

Base or sugar substitutions have also been used to improve the binding properties of pyrimidine TFOs. Replacement of 2′-deoxythymidine with 5-(1-propynyl)-2′-deoxyuridine will enhance binding affinity at physiologic pH and physiologic magnesium concentrations (72,80). In addition, TFOs containing 2′-O-methyl (OMe) have enhanced stability (48,81). Recently, 2′-O-aminoethyl (AE) substituted oligonucleotides were characterized in vitro and in vivo. AE was able to increase binding affinity and stability and increase rates of gene knockout activity at the hprt locus in CHO cells (82).

Because of reduced charge repulsion, it is advantageous to have an oligonucleotide backbone that is uncharged. PNAs have a flexible, neutral backbone that fills this niche well (83,84). Unlike DNA TFOs, PNAs bind duplex DNA or single-stranded RNA by forming Watson–Crick pairs with the target. In the case of PNA triplexes, the second PNA strand binds in the major groove of the DNA/PNA duplex via Hoogsteen interactions. Though PNAs bind best in low or no salt conditions (85,86), there have been some examples of PNAs mediating gene targeting in cells in culture (40,46).

Another class of DNA binding molecules includes polyamides which bind the minor groove of duplex DNA (87–89). Polyamides can be designed to recognize DNA segments containing all four possible Watson–Crick pairs (90–92). Thus, these are very promising molecules for gene targeting; however, their in vivo applicability remains to be established.

Finally, delivery is one of the major barriers to using gene targeted DNA binding molecules. Numerous strategies have been examined in a search to find an efficient means of TFO delivery: cationic lipids (93,94), DNA condensing agents such as polyethyleneimine (95), cell permeabilization agents such as streptolysin or digitonin (25,43,96), oligonucleotide coupling to peptides like antennapedia (97) or hydrophobic residues such as cholesterol (98), microinjection (51) and electroporation, including square wave electroporation (99).

FUTURE DIRECTIONS

In assessing the challenges facing the field, intracellular delivery of TFOs and related molecules remains a major issue. Microinjection, evolution of cation lipid technology and improvements in electroporation are all promising experimental techniques that will improve the efficiency of laboratory studies. However, with the ultimate goal of treating human patients with genetic disease, we must reconsider the delivery issue. Intraperitoneal injection of DNA TFOs with protected ends was successful in mice (45); however, the rates of mutagenesis remained at a fraction of a percent. For most genetic diseases this would not be sufficient to effect a reduction in symptoms or a cure. Recently, techniques for the intracellular production of a single-stranded DNA molecule have been developed (100); our lab is currently investigating the use of these vectors to produce TFOs intracellularly.

Mechanistic questions also remain unanswered. Though pathways have been implicated in the mutagenesis and recombination effects of TFOs, the exact mechanisms are unknown. Ongoing in vitro studies in our lab are aimed at establishing the precise cellular processes that play a role in mediating these genetic changes.

It is also of interest to determine which factors can influence the availability of genomic targets to DNA binding molecules. Transcription levels, cell cycle phase, nucleosome binding and histone state may all play important roles.

A major long term goal of the field is to develop TFOs and other sequence-specific DNA binding molecules as tools for the targeted modification of the genome. To achieve such an end, the above questions regarding efficiency, delivery and genome availability must be addressed. Diseases which require the majority of cells in a tissue to be corrected may not be amenable to triplex technology. Instead, triplex technology is probably applicable to certain hematologic, metabolic and hepatic diseases in which correction of a small percentage of the target cells may ameliorate the phenotype of the disease; sickle-cell anemia is one such example. In addition, in other diseases, such as adenosine deaminase deficiency, corrected cells may acquire a growth advantage that would lead to an expansion of the corrected cell population, thereby amplifying the effect of the targeting molecules.

References