At least three different approaches may be used for gene targeting including: A) gene knockout by homologous recombination; B) employment of synthetic oligonucleotides capable of hybridizing with DNA or RNA, and C) use of polyamides and other natural DNA-bonding molecules called lexitropsins.
Targeting mRNA is attractive because mRNA is more accessible than the corresponding gene. Three basic strategies have emerged for this purpose, the most familiar being to introduce antisense nucleic acids into a cell in the hopes that they will form Watson-Crick base pairs with the targeted gene's mRNA. Duplexed mRNA cannot be translated, and almost certainly initiates processes which lead to its destruction. The antisense nucleic acid can take the form of RNA expressed from a vector which has been transfected into the cell, or take the form of a DNA or RNA oligonucleotide which can be introduced into cells through a variety of means. DNA and RNA oligonucleotides can be modified for stability as well as engineered to contain inherent cleaving activity.
It has also been proven that because RNA and DNA are very similar chemical compounds, DNA molecules with enzymatic activity could also be developed. This assumption proved correct and led to the development of a “general-purpose” RNA-cleaving DNA enzyme. The attraction of DNAzymes over ribozymes is that they are very inexpensive to make and that because they are composed of DNA and not RNA, they are inherently more stable than ribozymes.
Although mRNA targeting is impeccable in theory, many additional considerations must be taken into account in applying these strategies in living cells including mRNA site selection, drug delivery and intracellular localization of the antisense agent. Nevertheless, the ongoing revolution in cell and molecular biology, combined with advances in the emerging disciplines of genomics and informatics, has made the concept of nontoxic, cancer-specific therapies more viable then ever and continues to drive interest in this field.
The notion that gene expression could be modified through use of exogenous nucleic acids derives from studies by Paterson et al. who first used single-stranded DNA to inhibit translation of a complementary RNA in a cell-free system in 1977 . One year later, Zamecnik and Stephenson noted that a short (13nt) DNA oligonucleotide reverse complementary in sequence (antisense) to the Rous sarcoma virus could inhibit viral replication in culture . This observation is credited as being among the first to suggest the therapeutic utility of antisense nucleic acids, a concept which ultimately led to the awarding of a Lasker Prize in Medicine to Dr. Zamecnik. In the mid 1980s, the existence of naturally occurring antisense RNAs and their role in regulating gene expression was demonstrated [3-5]. These observations were particularly important because the fact that naturally occurring antisense nucleic acids played a role in regulating gene expression lent support to the belief that exogenously introduced reverse complementary nucleic acids might be utilized to manipulate gene expression in living cells. These seminal papers, and the literally thousands which have followed, have stimulated the development of technologies employing nucleic acids to manipulate gene expression. Virtually all available methods rely on some type of nucleotide sequence recognition for targeting specificity, but differ where and how they perturb the flow of genetic information . Simply stated, strategies for modulating gene expression may be thought of as being targeted to the gene itself, or to the gene's messenger RNA (mRNA). Since this review will be focused on strategies aimed at disrupting the use of mRNA, antigene strategies will be addressed only briefly and mainly for the sake of completeness.
At least three different approaches may be utilized for direct gene targeting. The “gold standard” is the gene “knock-out” achieved by homologous recombination [7, 8]. This approach results in the actual physical disruption of the targeted gene as a result of crossover events which occur during cell division between the targeting vector and the gene selected for destruction (Fig. 1A). Homologous recombination is extremely powerful, but the technique is hampered by the fact that it remains inherently inefficient, time-consuming, and expensive. While improvement in the efficiency of this process has been achieved [9, 10], this is a method which remains restricted to use in cell lines and animal models, if for no other reason than selection is required to find the cells in which the desired events have taken place. In clinical situations where high efficiency gene disruptions are required, it seems unlikely that this approach will serve as a useful therapeutic modality anytime in the foreseeable future.
A second option for gene targeting employs synthetic oligodeoxynucleotides (ODN) capable of hybridizing with double-stranded DNA [11-13]. Such hybrids are typically formed within the major groove of the helix, though hybridization within the minor groove has also been reported . In either case, a triple-stranded molecule is produced, hence the origin of the term triple helix-forming oligodeoxynucleotide (TFO) (Fig. 1B). TFOs do not destroy a gene but prevent its transcription either by preventing unwinding of the duplex or preventing binding of transcription factors to the gene's promoter. TFO sequence requirements are based on the need for each base comprising the TFO to form two hydrogen bonds (Hoogsteen bonds) with its complementary base in the duplex. This constrains TFOs to hybridization with the purine bases composing polypurine-polypyrimidine tracks within the DNA. The targeting efficiency of TFOs is further constrained by a number of factors, including need for divalent cations, and perhaps most importantly, by access to DNA compacted within the chromosome structure. Recent experiments from Wang et al. and Kochetkova et al. have provided evidence that triple helix formation can occur in living cells, suggesting that these difficulties may ultimately be overcome [15-17]. If shown practical, it has also been postulated that TFOs may prove useful in the treatment of certain genetic disorders such as sickle cell anemia, and hemophilia B, where their ability to trigger repair mechanisms might be used to correct single base pair mutations responsible for the disease [15, 18-20].
Final approaches worth mentioning are the use of specific nucleic acid sequences to act as “decoys” for transcription factors [21, 22], and the use of polyamides and other natural DNA-binding molecules called lexitropsins, that bind to specific bases in the minor groove of DNA [23, 24]. The use of decoy molecules evolves from the knowledge that transcription factor proteins recognize and bind specific DNA sequences. In theory then, it is possible to synthesize nucleic acids which will effectively compete with the native DNA sequences for available transcription factor proteins in vivo. If effective, the rate of transcription of the genes dependent on the particular factor involved will diminish. Unless single gene transcription factors can be identified, it is difficult to conceive how this approach, though potentially effective for controlling cell growth, can be made gene-specific. The polyamide approach may prove feasible since sequence-specific molecules can likely be designed and it appears that molecules of this type can easily access DNA within the chromosomes [23-25].
A gene may be effectively “silenced” by destabilizing its mRNA, thereby preventing synthesis of the protein it encodes. Targeting mRNA, while less favorable than antigene strategies from a stoichiometric point of view, is nonetheless attractive because mRNA is in theory more accessible. Three basic strategies have emerged for this purpose. One employs an oligonucleotide that acts as an alternate binding site, or “decoy,” for protein-stabilizing elements that normally interact with a given mRNA [26, 27]. By attracting away mRNA-stabilizing protein, the decoy induces instability, and ultimately destruction, of the mRNA. A newly developing approach is to affect RNA interference (RNAi) or post-transcriptional gene silencing [28, 29]. RNAi employs a gene-specific double-stranded RNA which, when introduced into a cell, leads to diminution of the targeted mRNA. The actual mechanism whereby this is accomplished is presently unknown but is under intense investigation with several clues being deciphered already [30, 31] including size and necessity for processing of the targeting dsRNA. In C. elegans and Drosophila this is a highly reproducible method for disrupting gene expression. Some reports suggest that this technique can be adapted for use in mammalian cells , but this remains uncertain at the moment. Finally, there is the more familiar, and more widely applied “antisense” strategy. We will focus on the latter.
Antisense (reverse complementary) nucleic acids are introduced into a cell in hopes that they will form Watson-Crick base pairs with the targeted gene's mRNA. As stated above, duplexed mRNA cannot be translated, and almost certainly initiates processes which lead to its destruction. The antisense nucleic acid can take the form of RNA expressed from a vector which has been transfected into the cell , or take the form of a DNA or RNA oligonucleotide which can be introduced into cells through a variety of means. DNA and RNA oligonucleotides can be modified for stability as well as engineered to contain inherent cleaving activity [34, 35]. A number of these issues will be discussed in more detail in the sections below.
Antisense Oligonucleotides (AS-ONs)
AS-ONs are short stretches of nucleotides that are complementary to a region of targeted mRNA and can specifically suppress expression of that particular transcript. The following discussion will focus on the fundamental concepts concerning AS-ONs and their mechanisms of action. Examples of AS-ON use in experimental and clinical settings have been recently reviewed [36-38].
The exact mechanism of AS-ON action remains unclear, but it is known to be different for various types of AS-ONs. Generally, these molecules block gene expression by hybridizing to the target mRNA, resulting in subsequent double-helix formation. This process can occur at any point between the conclusion of transcription and initiation of translation, or even possibly during translation. Disruption of splicing, transport, or translation of the transcripts are all possible mechanisms, as is stability of transcript. Therefore, a major question is whether AS-ONs are most effective in the cytoplasm or nucleus. In the case of antisense oligodeoxyribonucleotides (AS-ODNs), cellular RNase H is able to bind to the DNA-RNA duplex and hydrolyze the RNA, resulting in increased transcript turnover. Any modification to the deoxy moiety at the 2′-sugar position prohibits RNase H action.
Modified AS-ONs or AS-ON analogs are often employed for in vivo antisense applications due to their increased stability and nuclease resistance. A longer serum half-life ensures that the AS-ON has ample time to reach and interact with its target mRNA. Phosphorothioate AS-ODNs are most widely used due to their long serum half-life and the fact that they are a suitable RNase H substrate. However, phosphorothioates display high affinity for various cellular proteins, which can result in sequence-nonspecific effects [39, 40]. Furthermore, high concentrations of phosphorothioates inhibit DNA polymerases and RNase H, which may render them ineffective as antisense agents . Interestingly, many AS-ONs with 2′-modifications with groups such as O-methyl, fluoro, O-propyl, O-allyl, or many others exhibit greater duplex stability with their target mRNA along with antisense effects independent of RNase H (Fig. 1). These modifications create bulk at the 2′ position, causing steric hindrance to play a significant role in increasing nuclease resistance. Nucleotide analogs generally are also nuclease-resistant and often demonstrate superior hybridization properties due to modified backbone charge, although they usually are not acceptable substrates for RNase H. One example is peptide nucleic acid (PNA) where the sugar-phosphate moiety has been replaced by 2-aminoethyl glycine carbonyl units . To these units are attached nucleotide bases spaced equally apart to DNA nucleotide bases. Instead of phosphodiester linkages between nucleotides, peptide bonds join the monomers to create a backbone neutral in charge. Not only do PNA oligomers hybridize to complementary DNA and RNA by Watson-Crick base pairing, they do so more quickly  and with greater affinity [42-44] because of the neutral backbone. In addition, PNAs are better at discriminating between base pair mismatches  and form less nonsequence-specific associations with proteins than phosphorothioate oligonucleotides . Positive charges can also be introduced to backbone structure as in the case of (2-aminoethyl)phosphonates. Increased stability of duplex formation with both RNA and DNA has been reported with hybrid stability being more pH-dependent and less salt-dependent than natural RNA or DNA duplexes .
Some insight into the mechanism of AS-ON action has emerged recently through the work of Baker and colleagues (unpublished). Differences in ability to inhibit gene expression occur when either 2′-modified AS-ONs or 2′-unmodified AS-ONs are targeted to the exon 9 region of interleukin 5 (IL-5). Two forms of IL-5 exist: a soluble IL-5 lacking the exon 9 region, and a membrane-bound form, which contains exon 9. When unmodified AS-ONs are targeted to exon 9 of the IL-5 transcript, the expression of both membrane-bound and soluble IL-5 is inhibited. However, 2′-modified AS-ONs only suppress membrane-bound IL-5 expression. These observations seem to suggest that RNase H-dependent antisense effects are a nuclear event prior to splicing, whereas RNase H-independent oligonucleotides may affect splicing in transcript processing or may suppress gene expression after splicing has taken place. Additional evidence demonstrates that in the absence of RNase H activity, antisense effects may be a result of interference with translational initiation complex formation for certain types of 2′-modified AS-ON such as 2′-O-(2-methoxy) ethyl AS-ONs .
Naturally occurring ribozymes are catalytic RNA molecules that have the ability to cleave phosphodiester linkages without the aid of protein-based enzymes. This property has been exploited to specifically inhibit gene expression by targeting mRNA for catalytic cleavage especially in viral, cancer, and genetic disease therapeutics . Similar to AS-ONs, ribozymes bind to substrate RNA through Watson-Crick base pairing, which offers sequence-specific cleavage of transcripts. Ideally, these agents should trigger enhanced transcript turnover as compared to RNase H-mediated AS-ON degradation of transcripts, considering ribozymes act through bimolecular kinetics (association of ribozyme and target transcript) whereas RNase H-dependent AS-ONs rely on trimolecular kinetics (association of AS-ON, target transcript, and RNase H). Since ribozymes are RNase H-independent, 2′-modifications to increase stability do not diminish antisense effects and experiments have shown some modifications do not attenuate catalytic ability . Unlike AS-ONs, ribozymes can be expressed from a vector, which offers the advantage of continued production of these molecules intracellularly [50, 51]. However, stable transformation of cells in vivo has its own complications and will not be discussed in this review.
If ribozymes are to perform effectively as “enzymes,” they must not only bind substrate RNA but also dissociate from the cleavage product in order to act on additional substrates. Studies suggest that in some cases, dissociation of cleavage product may be the rate-limiting step [52, 53]. Furthermore, some ribozymes require high divalent metal ion concentrations for efficient substrate cleavage, which may limit their use in intracellular environments . All of these concerns need to be addressed and overcome in order for ribozymes to have a future in medical therapy. Two ribozymes, the hammerhead ribozyme and the hairpin ribozyme, have been extensively studied due to their small size and rapid kinetics. Their application has been recently reviewed in several publications [55-59].
The hammerhead ribozyme consists of a highly conserved catalytic core, which will cleave substrate RNA at NUH triplets 3′ to the H, where N is any nucleotide, U is uracile, and H is any nucleotide but guanidine (Fig. 2) . In fact RNA cleavage may be less restricted since recent studies demonstrate exceptions to the “NUH” rule. Investigators have established that cleavage can actually occur 3′ to any NHH triplet . Furthermore, in vitro selection protocols have made it possible to screen for ribozymes with various cleavage specificities including one that cleaves at AUG sites . Thus, the limitations for sequence specificity of triplet-cleavage sites on the target RNA are less than previously thought. In addition to the catalytic core, a particular cleavage site in a target RNA can be specifically recognized by the hammerhead ribozyme arms. By creating complementary sequences in the arms to sequences flanking the cleavage site, the ribozyme will hybridize specifically to the RNA of interest. Subsequent cleavage will then be directed towards that particular position.
The catalytic ability of hammerhead ribozymes is dependent on the presence of divalent metal ions, of which magnesium is most often used in vitro. It is postulated that the ions not only participate in RNA folding but also in the cleavage step itself . As mentioned previously, studies indicate that catalytic activity requires relatively high Mg2+ concentrations compared to the intracellular environment. This characteristic could be problematic in applying the hammerhead ribozyme to an in vivo setting where intracellular Mg2+ concentrations are 5- to 10-fold lower than optimal in vitro conditions.
Much evidence supports diminished mRNA levels and gene products directly due to hammerhead ribozyme delivery. There is also indication from reverse transcriptase-polymerase chain reaction (RT-PCR) and reverse ligation (RL)-PCR protocols of messenger RNA cleavage at the targeted position in cellular RNA extracts [61, 62]. Recently, more potent ribozyme-mediated effect on viral and cancer cell growth compared to noncatalytic RNAs was reported [63, 64]. However, in some instances, hammerhead ribozymes have not proven to be more effective than AS-ON and instead give equal degrees of gene suppression. Likewise, inactive control ribozymes where antisense binding can occur, but catalytic ability has been abolished, give similar levels of gene inhibition when compared to fully catalytic hammerhead ribozymes, suggesting that the catalytic core, in some instances, plays little role in enhancing antisense effects . Only further detailed studies will reveal the true utility of hammerhead ribozymes.
The natural hairpin ribozyme is derived from a negative strand of the tobacco ringspot virus satellite RNA. Work on engineered hairpin ribozymes has resulted in a broader range of cleavage-sequence specificity. In general, a phosphodiester cleavage takes place 5′ to the G in the sequence NGUC where N is any nucleotide , although recent studies have shown even less restriction on sequence requirements for cleavage .
The overall structure of the hairpin ribozyme consists of two domains connected by a hinge section (Fig. 3). One domain binds the substrate RNA to form two helical regions separated by a pair of single-stranded loops. Cleavage occurs within the single-stranded area of the substrate RNA. The other domain is similar in structure except the helixes are formed from the ribozyme folding back onto itself. The most important sequences for cleavage activity are those within the single-stranded regions where almost every nucleotide is conserved, while the helical portions can be almost any sequence as long as there is double-helix formation . The hinge allows the two domains to be flexible relative to one another in space so that the two can dock together in an antiparallel orientation required for cleavage catalysis [68, 69].
Both the hairpin and hammerhead ribozymes require metal ions for cleavage catalysis. In the hammerhead ribozyme, metal ions are believed to be involved directly in the cleavage step [54, 70], whereas metal ions have not been implicated to be directly involved with cleavage for the hairpin ribozyme . The metal ions in hairpin ribozymes may instead play an important role in ribozyme structure . Fluorescence resonance energy transfer (FRET) studies on docking of the two domains show that docking is metal-dependent, but almost any metal will suffice even though they may not support cleavage . In addition, docking is not the rate-limiting step, and since metal ions are not thought to be involved in the chemical cleavage step, it can only be assumed that there is a slower step in between docking and chemical cleavage.
One of the advantages offered by hairpin ribozymes is their unique ion-dependence for catalytic action. One group has shown that aminoglycoside antibiotics with at least four amino groups are able to both support and to inhibit hairpin ribozyme cleavage depending on metal ion conditions . In the presence of magnesium, aminoglycoside antibiotics inhibit ribozyme cleavage with the degree of inhibition depending on the binding affinity of the antibiotic to the ribozyme. However, in the absence of metal ions, aminoglycoside antibiotics prove to assist cleavage with an optimum reaction condition at pH 5.5 and poorer kinetics as the pH is increased, exactly opposite to trends observed for magnesium. In this case, the metal ions are most likely being replaced by the amino groups of these antibiotics.
Polyamines such as spermidine and spermine have also been reported to support hairpin ribozyme cleavage ability. In the absence of magnesium, spermidine allows the cleavage reaction to persist at very slow kinetics compared to magnesium alone . However, spermine alone gives very efficient cleavage of RNA comparable to that of magnesium, and when in the presence of low magnesium concentrations similar to intracellular conditions, spermine displays considerable increase in cleavage rates . The fact that spermine is the major polyamine in eukaryotic cells may explain why the hairpin ribozyme has shown remarkable intracellular cleavage activity in mammalian cells and may make future therapeutic endeavors with the hairpin ribozyme much easier .
While investigating ways to improve the function of ribozymes, Breaker and Joyce made the assumption that because RNA and DNA are very similar chemical compounds, DNA molecules with enzymatic activity could also be developed . This assumption proved correct and led to the development of a “general-purpose” RNA-cleaving DNA enzyme . The molecule was identified from a library of >1,000 different DNA molecules by successive rounds of in vitro selective amplification based on the ability of individual molecules to promote Mg2+-dependent, multiturnover, cleavage of an RNA target.
The selected molecule was named the “10-23 DNA enzyme,” because it was derived from the 23rd clone obtained after the 10th round of selective amplification . The “10-23 DNA enzyme” is composed of a catalytic domain of 15 deoxynucleotides, flanked by 2 substrate-recognition domains of ∼8 nucleotides each (Fig. 4). The recognition domains provide the sequence information required for specific binding to an RNA substrate. They also supply the binding energy required to hold the RNA substrate within the active site of the enzyme. It is straightforward that by appropriately designing the flanking sequences, the DNAzyme can be made to cleave virtually any RNA that contains a purine-pyrimidine junction.
The attraction of DNAzymes over ribozymes is that they are very inexpensive to make and that because they are composed of DNA and not RNA, they are inherently more stable than ribozymes. Nevertheless, DNAzymes must ultimately overcome the same problems faced by ribozymes and oligonucleotides if they are to be effective in cellular systems (see below). These are stability, ability to be targeted to the cell of interest, ability to achieve sufficient intracellular concentration to cleave to the targeted mRNA, ability to hybridize with their mRNA target, and lack of toxicity. In this regard, many of the chemical modifications employed to stabilize ODNs can be incorporated into the 10-23 DNA enzyme without loss of activity. There is a suggestion from recent reports that issues of intracellular concentration and target hybridization may also be solvable [78, 79].
Application of the “Antisense” Strategy
Although antisense interference methods appear impeccable in theory, many additional considerations must be taken into account in applying the strategy in living cells. Since both AS-ONs and ribozymes are considered oligonucleotides, quite often similar solutions can be offered to address the problems encountered. As mentioned earlier, increasing stability of antisense agents can be easily achieved through nucleotide modifications or analogs. However, additional considerations crucial to reliable experimental outcome include mRNA site selection, drug delivery, and intracellular localization of the antisense agent.
mRNA Site Selection
Within living cells, transcripts exist in low energy conformations in which secondary structures dominate in folding the linear polymer. In addition, interactions with cytoplasmic proteins produce further structural properties. The end result is that much of the mRNA sequence is hidden and only partial sequences within the total mRNA length are accessible for hybridization. RNA folding programs that generate three-dimentional folding patterns based on free energy calculations often give an unreliable depiction for in vivo relevance. Therefore, a good empirical method to probe for suitable sites is necessary.
A system to probe for suitable sites in mRNA for AS-ON or ribozyme-targeting has recently been established using RNase H cleavage as an indicator for accessibility of sequences within transcripts . A mixture of ODNs that are complementary to certain regions of a transcript is added to cell extracts and exposed to RNase H. RT-PCR of the transcript can then be used to show which ODNs actually had access to the transcript and hybridized in order to create an RNase H-vulnerable site. Combining this methodology with computer-assisted sequence selection may enhance this approach as well .
Another technique currently being tested is the use of molecular beacons for site selection (Gewirtz et al., unpublished). These molecules are ODNs with the ability to form stem-loops where the loops are targeted to regions of the transcript . The stems have a fluorophore linked to either the 5′ or 3′ end and a quencher molecule is attached to the other so that in the stem-loop configuration, fluorescence is not observed due to the proximity of the quencher molecule to the fluorophore. However, when hybridization proceeds, the act of forming a double helix between the loop and the transcript causes unfolding of the stem-loop and brings the quencher and fluorophore apart in space. Thus, fluorescence should increase as a result of hybridization. Currently, these molecules are being applied to probe for accessible sites within mRNA with very encouraging results (Jen and Gewirtz, unpublished).
One of the major limitations for the therapeutic use of AS-ONs and ribozymes is the problem of delivery. Import of these compounds into cells can be accomplished by exogenous delivery in which presynthesized oligonucleotides come in direct contact with the plasma membrane, resulting in subsequent cellular uptake . Naked oligonucleotides are poorly incorporated into cells in this fashion and often require a vehicle for efficient delivery. In tissue culture, many classes of compounds have been used as delivery vehicles including cationic liposomes, cationic porphyrins, fusogenic peptides, and artificial virosomes. These compounds share the characteristic of forming complexes with oligonucleotides through electrostatic interactions between the negatively charged oligonucleotide phosphate groups and positive charges contained by the vehicles themselves. In addition, some degree of protection from nuclease degradation is conferred to the oligonucleotide when associated with such delivery vehicles. Other strategies including cell permeabilization with streptolysin-O and electroporation have been used  but are restricted in utility for clinical settings. Presently, some success has been achieved in tissue culture, but efficient delivery for in vivo animal studies remains questionable.
Cationic lipids form stable complexes with oligonucleotides, which exhibit improved cellular uptake [85-87]. The result is enhanced antisense activity. Further studies indicate that phosphorothioated ODNs dissociate from cationic lipids before entering the nucleus where it is free to hinder target transcript function . These compounds have proven to be quite effective in cell culture and have been commercialized, but their relatively high cytotoxic properties may restrict their use.
Alternatives to cationic lipids are being explored. Recently, cationic porphyrins have proven to be effective vehicles for AS-ONs in tissue culture [89, 90]. Two cationic porphyrins used by Benimetskaya and colleagues, tetra(4-methylpyridyl) porphyrin (TMP) and tetraanilinium porphyrin (TAP), demonstrate properties important for AS-ON delivery. 5′-fluorescein-labeled phosphorothioates show that both TMP and TAP more efficiently deliver AS-ONs into cells than naked AS-ONs. Nuclear fluorescence is observed after porphyrin/AS-ON complex exposure to cells while fluorescein labeled AS-ONs alone are taken up into vesicular structures. Thus, cationic porphyrins not only help AS-ON delivery into the cell, but they are also able to localize the AS-ON in the nucleus where mRNA and RNase H are present. FRET studies on the ability of cationic porphyrins to quench 5′-fluorescein-labeled phosphorothioates suggest intracellular dissociation of the oligonucleotide from the porphyrin.
Fusogenic peptides form peptide cages around oligonucleotides in order to boost oligonucleotide uptake. Many of these peptides contain polylysine residues, which cause membrane destabilization . Others are derived from viral proteins such as the fusion sequence of HIV gp41  and hemagglutinin envelop protein [93, 94]. Generally, these agents are less cytotoxic than lipids but are still able to achieve similar delivery efficacy. Artificial virosomes are another class of delivery vehicles which take advantage of the natural ability of a virus to gain entry into cells. Reconstituted influenza virus envelopes known as virosomes can fuse with endosomal membranes after internalization through receptor-mediated endocytosis . Recently, cationic lipids have been incorporated into virosome membranes to further aid delivery [96, 97].
Finally, Dheur and colleagues have noted that while oligonucleotides delivered with lipofectins usually do not elicit antisense activity (likely because cationic lipid formulations do not protect unmodified oligonucleotides from nuclease degradation), a cationic polymer, polyethylenimine (PEI) , improves the uptake and antisense activity of antisense phosphodiester oligodeoxynucleotides (PO-ODN) . Interestingly, PEI-phosphorothioate (PS) ODN particles were efficiently taken up by cells but PS-ODN did not dissociate from the carrier. These investigators suggested that the low cost of PEI compared with cytofectins, the increased affinity for target mRNA and decreased affinity for proteins of PO-ODN compared with PS-ODN might make the use of PEI-PO-ODN very attractive.
In order for AS-ONs or ribozymes to suppress gene expression, they must be colocalized to the same intracellular compartment as their target mRNA. Intracellular trafficking seems to play an important role in the fate of these molecules since their spatial distribution does not correspond to simple diffusion. Many factors determine localization patterns of AS-ON and ribozymes including the antisense agent itself, delivery vehicle, and targeted cell type. In addition, evidence for cell cycle-dependent localization patterns has been reported with nuclear localization predominantly in the G2/M phase .
mRNAs can exist in several cellular compartments including the cytoplasm, nucleus, and nucleolus. It remains unclear as to where oligonucleotides should be directed for most efficient antisense activity to occur, although endosomal localization usually predicts ineffective antisense response. The optimal site for mRNA degradation may be dependent on the type of antisense agent used . Recently, ribozymes attached to small nucleolar RNAs (snoRNAs) called snoribozymes exhibited nearly 100% efficiency in cleaving a target RNA also localized to the nucleolus by snoRNA attachment . Even though this particular experiment is based on cleavage of an artificial substrate, the expanding roles associated with the nucleolus may prove the nucleolus to be an important site to target mRNA degradation . In another study, antisense RNA inserted within a variable region of ribosomal RNA (rRNA) proved to heighten ribozyme efficiency and may be due to colocalization of rRNA with mRNA .
Antisense Drug Design
Certain issues to be aware of concerning antisense experimental design are quite important to the consistent and efficacious outcome of inhibiting gene expression. Even when the above considerations regarding the potential problems of antisense experiments are addressed, other factors may come into play especially involving antisense drug design. Only two will be mentioned here: formation of G quartets and chirality of modified oligonucleotides. Purine-rich oligonucleotides, especially ones containing four consecutive guinine residues, have a tendency to form stable tetrameric structures under physiologic conditions . The guinosines of single-stranded oligonucleotides are not restrained in space by rigid double-helix structure and can therefore form various hydrogen bonds not observed in Watson-Crick base pairing. Tetraplexes known as G quartets arise as a result. Dissociation rates of these structures may be quite slow and may prevent hybridization of AS-ONs or ribozymes to their target transcript, rendering them ineffective as antisense agents. However, the absence of G quartet structures at 37°C under cellular salt conditions could mean that G quartet formation is irrelevant at physiologic temperatures . Interestingly, nonsequence-specific gene inhibition by phosphorothioate oligonucleotides containing tetraguinosine tracts prove aptameric properties can play an important role in gene inhibition for some sequences of ONs .
Another important aspect to consider is the issue of chirality for certain oligonucleotides. Unmodified phosphodiester oligonucleotides do not have a chiral center at the phosphorous position. However, when a terminal oxygen of the phosphate is replaced by a sulfur, as in PS-ONs, the phosphorous gains chirality. The digestion kinetics of PS-ONs by 3′-exonucleases display bi-exponential decay with a fast and slow phase of digestion. These phases are due to stereoselectivity of the 3′-exonucleases on the chiral phosphorothioate center . A 25-mer containing a 3′-terminal internucleotide linkage in the S-configuration degrades 300-fold slower than the same 25-mer with an R-configuration phosphorothioate linkage.
The ongoing revolution in cell and molecular biology, combined with advances in the emerging disciplines of genomics and informatics, has made the concept of nontoxic, cancer-specific therapies more viable then ever. The recent development of a relatively specific biochemical inhibitor of the bcr/abl protein tyrosine kinase in patients with chronic myelogenous leukemia is a stunning example of this principle . For therapies focused on direct replacement, repair, or disabling of disease-causing genes, progress has been much slower and a successful equivalent to the biochemical bcr/abl inhibitor has yet to be achieved. In the case of anti-mRNA strategies, it is hoped that the above discussion will have made the reasons for this clearer. Given the state of the art, it is perhaps not surprising that effective and efficient clinical translation of the antisense strategy has proven elusive. While a number of phase I/II trials employing ONs have been reported [109-116], virtually all have been characterized by a lack of toxicity but only modest clinical effects. A recent paper by Waters et al. describing the use of a bcl-2-targeted ON in patients with non-Hodgkin's lymphoma is typical in this regard [117, 118].
The key challenges to this field have been outlined above. It is clear that they will have to be solved if this approach to specific antitumor therapy is to become a useful treatment approach. A large number of diverse and talented groups are working on this problem, and we can all hope that their efforts will help lead to establishment of this promising form of therapy.
This work was supported by grants from the NIH. Dr. A.G. is a distinguished Clinical Scientist of the Doris Duke Charitable Foundation.