1 History of Antisense Oligonucleotides Paul C. Zamecnik 1. Introduction and Early Studies Blologlcal science 1s a rapidly flowing experlmental stream, at times encountering a dam that impedes further progress. At such a pomt, a single crack may induce a maJor breakthrough Dlscovery of the double hehcal struc- ture of DNA in 1953 (I) caused such an event, with flooding of new mforma- tlon into the area now known as molecular biology. At this same time, our laboratory (2,3) developed a cell-free system for the study of protein synthesis, a domain separate from the DNA world. In 1954, James Watson and this author examined his wire model of DNA and puzzled about how the information from the gene became translated mto the sequence of a protein (4). The histochemlcal studies of Brachet (5) and Caspersson (6) had shown that m the pancreas, an organ very actively synthesizing protems for export, the cytoplasm was rich in what became known as rlbonuclelc acid. But how the DNA of the nucleus unwound its double strand and transcribed the RNA, the apparent intermediate m protein synthesis found in the cytoplasm, was unknown (7). The first example of the versatility of nucleic acid base pairing in the flow of mformation from DNA to protein was the discovery of transfer RNA (8-1 J) and perception of its role m translating the language of the gene mto the sequence of protein (12). Deciphering of the genetic code (13,14) next brought to light the precision of the tRNA-mRNA hybridization steps m protein trans- latlon. tRNA (an antlsense or negatively stranded RNA) acts in four dlstin- guishable ways, as follows. 1. By base palring with messenger RNA to initiate translation of the message, 2 By base pairing with messenger RNA to propagate translation of the message; From Methods m Molecular Medicme Anbsense Therapeutm Edlted by S Agrawal Humana Press Inc , Totowa, NJ 1 2 Zamecnik 3 By base pairmg with rlbosomal RNA to position the trmucleotlde antlcodon region for optlmal hybridization with the messenger RNA codon, and 4 By presentmg a terminating antlcodon (which IS an antisense trmucleotlde) to end the nascent protein sequence Puromycin, a natural nucleotide analog, provided an early example of antisense mhibrtton of protein synthesis (15). Further experimentation supported the hypothesis (16) that hybridization of synthetic exogenously added ohgonucleotides can influence cellular metabo- lism at three distinct levels: rephcation (17), transcription (l&19), and transla- tion (20,21). The variety and importance of these steps mvited the thought that natural ohgonucleottdes might play such roles m living cells (22). In the double helix, the DNA strand that carries the genetic message has been designated the sense strand. Its complementary mate, necessary as a template for the synthesis of a new sense strand, has become known as the antisense strand. Antisense polynucleotides have, m fact, for over two decades been known to occur natu- rally in prokaryotes (23), and have recently been found m eukaryotes (24). For some years synthetic oligonucleotides have also been reported to be capable of playing varied antisense inhibitory roles (25-28). Quite separate from the synthetic oligonucleotide field were independent developments from sphcmg larger segments of negatively stranded DNA into the genomes of cells, with the help of plasmids and viruses. These antisense strands of DNA were successfully integrated mto the genomes of relatively few host cells. Nevertheless, by selection processes these antisense sequences were picked out and found to be replicated along with the recipient’s genomic material. Thus, they were capable of blocking or altering the expression of cellular genes m a hereditary way. An example of the success of this technique is the permanent alteration of the color of petunias by antisense interference with synthesis of the flavonoid genes (29). An important difference between these approaches is that the synthetic, relatively short oligomers enter virtually all cells m a tissue culture or living animal (3&32) whereas the plasmids carry- mg much longer antisense polynucleotides generally enter a small percentage of cells, but may nevertheless have a dramatic genetic therapeutic effect (33). In the early 1960s the limits of DNA and RNA synthesis were in general trmucleotides, but by the end of the decade skilled scientists were able to construct oligomers 10-15 U m length, and to ligate such segments together (34). The message encoded in DNA or RNA, however, remained difficult to decipher. In 1965, Holley and colleagues (35) sequenced the primary structure of tRNA,i,, using the new method for scale-up and isolation of the tRNA fam- ily of molecules worked out m our laboratory by Roger Momer (36). In this highly competitive quest, several groups accomplished the sequencing of other particular tRNAs shortly thereafter (3 7-39), getting little credit for their efforts. An tisense Oligonucleo tides 3 In 1970, the discovery of reverse transcriptase (40,42) made it more feasible to sequence oligonucleottdes by synthesizing the primary structure of DNA enzymatically, and then sequencing the nascent DNA so formed. The wander- mg spot-analysis technique of Sanger and Coulsen (42) at that time made it possible to determine the sequence of approx 15-25 monomer umts at the 3’ end of a polynucleotide. The present dtscussion focuses on the role of the synthetic antisense oligo- nucleotides (20-30-mers) as chemotherapeuttc agents, and omits the splicing insertion into genomes of the larger (1 ,OOO-2,000 monomer unit) biologically synthesized polynucleotides. Those of us raised on the principle of Occum’s razor, which advises making explanations as simple as possible, continue to be surprised at the unfolding complexity of this synthetic oligonucleotide approach to chemotherapy. One was prepared for the nuclease sensittvity of unmodified oligodeoxynucleotides m a hvmg cell system (16,43), and the enhancement of therapeutic efficacy by blocking both ends of the ohgodeoxy- nucleotide (26). The effect of ribonuclease H (44) was generally unexpected, however, particularly the maJor role it plays in some antisense mhibitions. As is now known, a stoichiometrically acting oligodeoxynucleotide inhibitor may activate RNase H during its complementary hybridization with mRNA, then dissociate from its complement when the mRNA is hydrolyzed at the double- stranded area, and hybridize with another molecule of mRNA, this repetrtive action resulting in a catalytic effect. By 1976, we were able to sequence 2 1 nucleottdes inside the 3’-polyA tail of the Rous sarcoma vn-us, a terminus similar to that we had previously found on the avian myelobastosis vn-us (45). Rous sarcoma vu-us was the only purified vn-us for which a sufficient quantity was available to make a sequencmg effort feasible At this time, we learned that Maxam and Gilbert (46) had invented a revolutionary new DNA sequencing technique, and had, unbeknown to us, begun to decipher the 5’- end of the same Rous sarcoma vu-us. Astonishingly, both ends of this linear viral genome bore the same primary sequence, and were m the same polarity (47,48). It occurred to us that the new piece of DNA synthestzed by reverse tran- scriptton at the 5’- end of this retrovirus might circularize and hybridize with the 3’ end, like a dog biting its tail. Electron microscopic studies had suggested the presence of a circularized intermediate m the rephcative process of this vu-us. Thus, we considered the possibility of inhibmng viral replication by adding to the rephca- tion system a synthetic piece of DNA to block the circularizatton step (or alterna- tively some other step essential for replication), m the former case by hybridtzmg specifically with the 3’ end of the viral RNA in a competitive way. It was at this time generally believed that oligonucleotides did not penetrate the external membrane of eukaryotic cells, to enter the cytosol and nucleus 4 Zamecmk (49). Clearly, neither did ATP, except under unusual circumstances, nor Ap,A (50). Segments of cellular genomes were currently coaxed mto cell entry by an inefficient calcmm phosphate precipitation procedure. The negative charge of the oligonucleotide was regarded as presenting a major impediment to traverse of an ohgonucleotrde through the eukaryotic external cell wall. Nevertheless, an experiment testing the possibility of synthetic ohgonucleotide cell wall pen- etration was performed. We added a 13-mer synthetic ollgodeoxynucleotide, complementary to the 3’ end of the virus, to the medium of chick Iibroblasts m tissue culture, along with Rous sarcoma vtrus itself. It inhibited the formation of new vu-us, and also prevented transformation of chick fibroblasts mto sar- coma cells-both of these startling observations (16). In a cell-free system, translation of the Rous sarcoma viral message was also dramatically impaired (20). Until 1985, little further progress occurred, for three intertwined reasons: first, there was still widespread disbelief that ollgonucleotides could enter eukaryotic cells; second, tt was difficult to synthesize an oligomer of sufficient length to hybridize well at 37°C and of specifictty requisite to target a chosen segment of genome; and third, there was very little DNA (or RNA) genome sequence available for targeting m this way. This latter reason determmed the choice of the Rous sarcoma vuus, which Haseltme et al. (47) and our labora- tory (48) were sequencmg contemporaneously and whose results were pub- hshed m tandem. 2. Independent Complementary Developments Two important developments in the late 1970s and early 1980s mcreased the feasibility of the synthetic oligonucleotide hybridization mhibition approach. The first were the dramatic improvements m DNA sequencing that came from the Maxam-Gilbert chemical degradation procedure (46), and the more convenient dideoxy enzymatic sequencing technique of Sanger’s labora- tory (52). The second was the solid-phase ohgonucleotide synthetic approach introduced successfully by Letsmger and Lunsford (52) and Caruthers (53). At this ttme, as well, there developed a growing acceptance that oligonucleotides could pass through the eukaryotic cell membrane and enter the cell, and that they could readily be synthesized and purified. Fmally, an abundance of pot- ential DNA sequence targets began to appear like fireworks m a previously darkened genetic sky. The unmodified antisense oligodeoxynucleotide has proven to be the best RNase H activator, provided there are at least four or more contiguous hybrid- izing base pairs. The phosphorothioate modified oligodeoxynucleotides, although not so effective, still activate Rnase H, and are quite nuclease resis- tant (54). These two properties account for the early general preference of the latter m synthetic oligodeoxynucleotide experiments. In contrast, other varied Antisense Oligonucleotides 5 modifications at the mternucleotide bridging phosphate site result m inability to activate RNase H, and thus present a disadvantage. Included in this category are methyl phosphonates, a-oligonucleotides, intemucleotide peptide bonds, and others. Modifications on the ribosyl moiety, such as the 2’-0 methyl group, also fail to activate RNase H. Hybrid and chimeric ohgonucleotides are coming mto increasing usage, smce they combine terminal nuclease-resistant segments of an ohgonucleotide with a central RNase-sensitive portion (see Chapter 14). Furthermore, if only the central portion of the oligomer is phosphorothioate modified, whereas the peripheral 3’ and 5’ segments are, for example, 2’-0 methyl-modified oligomer moieties (55), the nonspecific effects of the totally phosphorothioate oltgomer (56) are to a considerable extent mmimized The selfstabilized snap-back oligomer is particularly advantageous m providmg enhanced nuclease resistance with other desirable properties (57). In addition to the promise of the antisense approach documented m these pages and elsewhere, there are reasons why the competitive oligonucleotide hybridization technique may not be successful in attempts to inhibit noxious genes, wherever they may exist m the animal and plant kingdoms A central reason is failure to find a smgle-stranded segment of genome that is highly conserved and accessible. Secondary and tertiary structure of the genome may prevent hybridization. Protems that have a high association constant with the area of genome targeted, i.e., promoters, enhancers, and modulators, for example, may prove to be barriers to ohgonucleotide hybridizations. The aggregation effect on ohgonucleotides of a G-quartet motif (58,59) is also an impediment. Zon (60) touches on some of these aspects in a historical review. It would be advantageous if hybridization inhibition could be achieved at the transcription level. This would block the amplification step, which results in numerous copies of mRNA for translation. A favorable site for transcription hybridization is the transcription bubble, consisting of 12-17 nucleotides of unwound double-helical DNA. As an example of this approach, we have found m an in vitro transcription system that specific hybridization mhibition can be induced using a linearized plasmid segment of HIV for a template With T7RNA polymerase, a gag RNA of about 640 nucleotides can be synthesized. This synthesis can be inhibited by a complementary 14-mer unmodified ohgodeoxynucleotide, lust downstream of the T7 promoter (19). The most effective inhibitor is a plus-sense oligonucleotide complementary to the nega- tive DNA strand that serves as a template for pre-mRNA synthesis. 3. Examples of Current Disease Targets Let me mention a few current medically related investigations involving our own laboratory that appear to show promise: HIV, influenza, and malaria. A study on HIV (61) shows that target selection in the HIV genome is important 6 Zamecnik for prevention of development of escape mutants Whereas escape mutants appeared after 20 d treatment of chronically infected Molt-3 cells with an antisense phosphorothioate oligomer pmpomtmg a splice acceptor site, contm- ued inhibition without escape over an 84-d experimental period occurred when revl-28 or gag-28 were the targets. A second example is the use of antisense oligomers to inhibit Influenza viral replication (62). At 10 PM concentration, replication of influenza C vnus was inhibited 90% m tissue cultures of MDCK cells by a sense oligophos- phorothioate targeted against the rephcase gene of the negatively stranded virus. In lo-d-old embryonated chick eggs, phosphorothioate ohgomer injec- tion also induced marked inhibition of vu-us production (63). Another example is inhibition of replication of Phsmodzum fuhparum malaria by a phosphorothioate oligodeoxynucleotide targeted against the dihydrofolate reductase-thymidylate synthase gene of the parasite (64). This enzyme is essential as a donor of a methyl group in the conversion of deoxyuridine monophosphate to thymidme monophosphate m the parasite, which must synthesize its own pyrimidmes, being unable to use exogenous thymidme for synthesis of DNA. The adult erythrocyte is one of the rare eukaryotic cells that oligodeoxynucleotides do not penetrate. Fortunately, how- ever, when a malarial parasite pushes its way into a red cell, it creates a permeabihzed erythrocyte membrane plus a parasitophorous duct (65), through either or both of which the oligodeoxynucleotide reaches the parasite inside its protective erythrocyte envelope. A fluorescently labeled oligodeoxynucleotide lights up a circular area inside an erythrocyte m which the P. fulczpurum para- site resides, surrounded by its own membrane, whereas the uninfected red cells fail to show evidence of cell entry (66). The above-mentioned antisense ohgo- mer shows a sequence-specific IDSo for rephcation of the parasite at 2-5 x 1