Discovery of the antiviral activities of pharmacologic cyclin-dependent kinase inhibitors: from basic to applied science

From the early 1960s to the late 1970s, work from many groups, including those of Prusoff, Purifoy and Powell; Roizman; Schaffer; Subak-Sharpe; and Summers (in alphabetical order) demonstrated that herpes simplex virus (HSV) induces unique DNA polymerase and thymidine kinase activities. At the time, inhibitors of human DNA synthesis were being explored as potential anticancer agents. It soon became evident that inhibitors that were more potent for viral than for human DNA synthesis could be useful as antiviral drugs. This approach led to the development of the first successful antiviral drugs, 5´-iodo-2´-deoxyuridine (IDU, which was too toxic for systemic use), adenine-arabinoside (Ara-A or vidarabine, which was safe enough to use systemically), and acycloguanosine (acyclovir, the first truly nontoxic antiviral drug). Since these early successes, antiviral drugs have been designed to target viral proteins. This approach has ensured that antiviral drugs have no major negative effects on host cells and, consequently, has ensured the safety of current antivirals. However, such drugs also have certain limitations, such as their ability to promptly select for drug-resistant viral strains. The number of potential viral targets for such drugs is also limited, and even more so for the smaller viruses (such as human papilloma virus). These drugs tend to be active against only one or a few related viruses. Moreover, the proteins of a new viral pathogen must first be characterized before such drugs can be developed, which results in a significant lag from the discovery of a new viral pathogen to the development of specific drugs. Owing to these, and similar limitations, cellular proteins are beginning to be considered as valid targets for antiviral drugs.

Selected drugs that target cellular proteins have been considered as potential antivirals in the past. However, such drugs were commonly considered only marginal tools, and did not greatly challenge the concept that anti-viral drugs must only target viral proteins. In recent years, a series of unexpected (and on the surface unrelated) discoveries in molecular, marine and cancer biology, plant hormone metabolism, transcription, viral latency and viral pathogenesis, led to the discovery of the unique antiviral activities of the pharmacologic cyclin-dependent kinase (CDK) inhibitors (PCIs). Owing to the strong potential as antivirals exhibited to date by PCIs, cellular proteins are now more often starting to be considered as valid targets for antiviral drugs. In fact, the first clinical trials of PCIs as antivirals are tentatively scheduled to start in 2005.

The discovery of the antiviral activities of PCIs provides a good example of how basic and applied science are intertwined in the development of novel drugs. In this Editorial, I will review how apparently unrelated research projects in three continents led to the discovery of a new family of potential antiviral drugs and a novel antiviral mechanism.

As for any scientific progress, it is impossible to identify all the experimental and conceptual sources for these discoveries, as each source is itself based on previous ones. I will begin the narration at the time when CDKs and their critical cell-cycle regulatory functions were already known, in great part, through the pioneering work of Hartwell, Hunt and Nurse (who as a result were coawarded the Nobel Prize in 2002), as well as Kanatani, Schorderet-Slatkine, Wasserman and many others. Owing to these, and related studies, it became evident that CDKs should be up- or deregulated in cancer cells. Otherwise, cancer cells would arrest following appropriate stimuli, and tumor growth would stop. From this realization, an intense search for specific CDK inhibitors soon developed. Unfortunately, these efforts did not result in the prompt discovery of any such inhibitor (although many useful protein kinase inhibitors commonly used in research today were described in this search). With the objective of screening for specific CDK inhibitors, Meijer, a le Centre National de la Recherche Scientifique (CNRS) researcher at the Station Biologique de Roscoff, France, developed an assay utilizing an immobilized CDK from starfish oocytes, CDK1 [1]. During the development of the assay, he showed that staurosporine, 6-dimethylaminopurine and isopentenyladenine (cytokinin) inhibited CDK1, although rather weakly or nonspecifically [1]. These studies attracted the attention of Veselý, at the Department of Pathological Physiology of the Medical Faculty, Palacky University, in Olomouc, Czech Republic, resulting in collaboration between Veselý, Meijer and Strnad (from the Institute of Experimental Botany, Czech Academy of Science). After intensive and long screening of numerous compounds, their collaborative efforts resulted in the discovery of the first specific, although not too potent, PCI, 6 benzylamino-2-(2-hydroxyethylamino)-9-methylpurine [2]. The compound was then renamed olomoucine in honor of Olomouc. Olomoucine had originally been synthesized in 1986 in Canberra, Australia, by Parker, Entch and Lentham as an inhibitor of cytokinin 7-glucosyl-transferase [3]; a plant enzyme that inactivates the growth-promoting hormone cytokinin. It had been synthesized again by L Havlíèek and J Hanuš (Isotope Laboratory of the Czech Academy of Science) and obtained by Strnad, from whom Veselý took it to screen in Meijer’s lab as a potential PCI (together with many other compounds). It may now appear surprising that Veselý, Meijer and collaborators experienced major difficulties in convincing skeptic reviewers that olomoucine was indeed specific for CDKs. The major concerns were that it bound to the ATP-binding pocket of the kinase and that it was competitive with ATP. The ATP-binding pockets of the protein kinases were then believed to be too conserved to allow for specific binding of a drug to only one or a small subset of protein kinases. Also, intracellular ATP concentrations were considered too high to be outcompeted by attainable drug concentrations. Elegant structure-activity relationship studies led to the discovery of 2-(1-d,l-hydroxymethylpropylamino)-6-benzylamino-9-isopropylpurine (roscovitine, in honor of Roscoff) as a similarly specific, but more potent, PCI [4]. More elegant experiments and the widespread use of these drugs by many other researchers eventually convinced the research community that roscovitine and olomoucine were indeed highly specific, if not very potent, CDK inhibitors. It is now well known that, perhaps counterintuitively, most of the specific protein kinase inhibitors bind to the ATP binding pocket and compete with ATP. Those readers interested in a more detailed description of the discovery of olomoucine and roscovitine as CDK inhibitors are referred to an excellent review by Meijer and Raymond [5].

While olomoucine was being developed through a mix of basic and applied research, another PCI was discovered through strictly applied research. The National Institutes of Health (NIH) has developed a screen for anticancer compounds. A flavo-noid, 4H-1-benzopyran-4-one, 2-(2-chlorophenyl)-5,7-dihydroxy- 8-(3-hydroxy-1-methyl-4-piperidinyl)-, hydrochloride, (-)-cis- (flavopiridol) was submitted to this screen as a potential inhibitor of epithelial growth factor receptor (EGFR) or protein kinase A (PKA). Flavo-piridol is a semisynthetic flavonoid that was derived from rohitukine, a natural alkaloid originally isolated from the leaves and stems of Amoora rohituka. Rohitukine was later obtained from stem bark of Dysoxylum binectiferum. Both A. rohituka and D. binectiferum are tropical trees of the Meliaceae family, native to India, China and other parts of Asia. In 1992, flavo-piridol was found to have potent activities against a variety of cancer cell lines in the NIH screen [6]. Surprisingly, however, the anticancer activity of flavopiridol was far more potent than its activities against EGFR or PKA. This led to the search for the true molecular targets of flavopiridol, which were discovered to be CDKs in the same year in which olomoucine was first reported to be a specific CDK inhibitor (1994) [7]. Thus, through independent and unrelated projects, flavopiridol and olomoucine became the first two specific PCIs.

Soon after their activities against CDKs were characterized, flavopiridol and roscovitine started to be evaluated as potential antiproliferative agents. Flavopiridol was the first to be tested in animal models and clinical trials. Roscovitine followed a few years later. From the very early studies, it appeared that these drugs had only limited negative side effects. The apparent safety was at the time very surprising because the CDK targets of these drugs were then believed to be essential for mammalian cells.

PCIs would probably never have been evaluated as antivirals if it were not for a basic science interest in the cellular functions required for viral replication. In the early to mid-1990s, it was commonly accepted that replication of most viruses with a nuclear DNA replicative intermediate required proteins involved in the cell-cycle. At that time, most herpesviruses were considered to be the exceptions to this model. However, several groups had already published strong evidence challenging the concept that herpesviruses were an exception to the requirement for cell-cycle proteins in viral replication. Owing to these published results, two groups independently decided to use the newly discovered PCIs to test whether herpesvirus replication required CDK activities. Bresnahan, Albrecht and colleagues tested the effects of roscovitine on human cytomegalovirus replication. Schaffer and myself (then a postdoctoral fellow in her laboratory) tested the effects of olomoucine and roscovitine on replication of HSV. In May of 1997, Bresnahan, Albrecht and colleagues published the first paper on the potential of PCIs as antiviral agents and demonstrated that PCIs inhibited viral DNA replication [8]. In July of 1998, we published the first of a series of papers in which we showed that PCIs inhibit HSV replication as a consequence of inhibition of cellular, not viral, proteins [9]. Perhaps more importantly, we further showed that PCIs inhibit viral transcription and that it is difficult to select for roscovitine- or olomoucine-resistant HSV mutants – a finding that has since been corroborated for other viruses.

Bresnahan, Albrecht and collaborators, as well as Schaffer and I, immediately realized that PCIs could potentially be developed as antiviral drugs. PCIs had strong antiviral activities in vitro at concentrations that were proving to apparently be safe for humans in clinical trials against cancer. The antiviral properties of PCIs might not have attracted major attention; however, they had been limited to inhibition of viral DNA replication. A number of antiviral drugs that inhibited this viral function by targeting viral proteins already existed. Thus, a drug that targeted a cellular protein to inhibit viral DNA replication would have inhibited a viral function already targeted by other drugs, while having potentially higher risks of negative side effects (that could result from inhibition of cellular functions). However, we have demonstrated that PCIs also inhibit viral gene expression, a viral function not targeted by any antiviral drug [9–14]. This was a most unexpected finding, since nuclear CDK2 was then believed to be required for nuclear DNA replication but no roscovitine-sensitive CDK was thought to be required for viral transcription (it was soon discovered that roscovitine inhibits a CDK involved in tran-scription, CDK7, and that HIV transcription requires yet another CDK, CDK9). Although PCIs have since been shown to inhibit multiple other viral functions, it is now commonly accepted that viral transcription is the major target of their antiviral activities. We and others have further shown that concentrations of PCIs that inhibit viral transcription do not affect global cellular transcription. Furthermore, we have recently shown that inhibition of viral transcription is specific for viral genomes, rather than for viral promoter sequences [13]. The specificity for viral genomes may, in part, explain the difficulties in selecting for PCI-resistant viral strains. Drugs that act on viral genomes would require no specific viral sequences or proteins that could mutate and result in drug-resistant strains.

With Schaffer and collaborators, we have also showed that the antiviral activities of PCIs are additive to those of conventional antiviral drugs, as expected, since the latter target viral proteins and PCIs target cellular proteins [15]. This expected result opened the possibility of adding PCIs to current combinations of antiviral drugs. Such additions would allow testing the anti-viral effects of PCIs in human beings without discontinuing partially effective antiviral treatments.

Replication of many viruses has long been known to require CDKs. It was thus expected from the original discoveries of the antiviral activities of PCIs that these drugs would be active against a variety of viruses. In 2000, Chao, Price, Sausville, Senderowicz and collaborators demonstrated that flavopiridol was active against HIV [16]. They further demonstrated that flavo-piridol inhibited HIV transcription, and that this activity was mediated by inhibition of CDK9. Kashanchi and colleagues showed 1 year later that roscovitine and related purine-type PCIs also had anti-HIV activity [17]. In 2002, Malim, Meijer, Schaffer, I and collaborators showed that roscovitine and purvalanol efficiently inhibit replication of HIV strains resistant to multiple antiretroviral drugs [15]. This finding has recently been corroborated by Agbottah and colleagues [18]. Perhaps surprisingly, however, the targets of roscovitine and related PCIs that mediate their effects against HIV have yet to be identified.

The unique antiviral properties of PCIs made them attractive as potential antiviral drugs. They inhibit viral replication by mechanisms that are not common to other antiviral drugs, targeting viral transcription in a promoter-independent manner. Viral resistance against PCIs is difficult to select for, and PCIs are active against strains that are already resistant to existing antiviral drugs. PCIs are also active against a variety of un-related viruses, which makes them promising as drugs against emerging viral disease. Lastly, their antiviral activities are at least additive to those of antiviral drugs that target viral proteins. However, PCIs target a family of CDKs that were commonly believed to be essential for mammalian cells, including CDK2 (probably their major antiviral target). Such believed essentiality of their targets restricted commercial or clinical interest in PCIs as antiviral drugs. A combination of basic and applied science resulted in yet another apparently unrelated discovery, which has since overcome these concerns about PCIs.

After CDKs were discovered to be a relatively large family of protein kinases activated by regulatory subunits (cyclins), van den Heuvel and Harlow tested the effects of inhibiting each of the then known CDKs [19]. They mutated an arginine residue in the predicted catalytic site of each CDK, a residue that is highly conserved among protein kinases. Mutant CDKs were then introduced by transfection into wild-type tumor cells, where the inactive transfected CDKs should sequester the cyclins that would otherwise activate the endogenous active forms of the respective CDKs. In this system, the catalytically inactive CDK2 mutant resulted in inhibition of cell-cycle progression at the G1/S transition. These experiments became the basis for the model developed through the years in which CDK2 was required for nuclear DNA replication in S-phase. However, alternative explanations are possible. For example, the cyclins themselves could be required for cell-cycle progression, or inhibition of cell-cycle progression could have resulted from the compound effects of the inactive kinase together with the stress of transfection. Nonetheless, the requirement for CDK2 in cell-cycle progression became so widely accepted that when Ortega, Malumbres, Barbacid and colleagues first obtained viable CDK2 knockout mice they faced strong skepticism [20]. It was only when Tetsu and McCormick obtained independent evidence of the nonessentiality of CDK2 (in cancer cells) [21] and Berthet, Kaldis and colleagues independently reported that CDK2 knockout mice were viable [22] that the nonessentiality of CDK2 became reluctantly accepted.

The newly discovered nonessential nature of CDK2 resulted in a renewed interest in CDK2-specific PCIs as potential anti-viral drugs. Six new groups have since reported studies on the antiviral effects of PCIs against varicella-zoster virus, Epstein–Barr virus, human T-lymphotropic virus, Kaposi’s sarcoma-associated herpes virus, human cytomegalovirus and HIV (recently reviewed [23]). A biotechnology company, Cyclacel, has tentatively scheduled the first clinical trials of PCIs as antiviral agents for 2005 [101]. The demonstration that CDK2 is nonessential for mammalian cells thus played a major role in raising the interest on PCIs as antiviral agents. Moreover, the preclinical and early clinical trials of PCIs against cancer had indicated that concentrations above those required to inhibit viral replication in cultured cells could be attained in vivo without obvious undesirable side effects. However, clinical trials of PCIs as antivirals would probably not have been scheduled in the absence of preclinical data supporting their antiviral potential. In a series of elegant experiments, Nelson and colleagues took the preclinical analyses of PCIs as antivirals one step further.

We had advocated the use of PCIs in clinical trials against viral diseases the pathogenesis of which includes cell proliferation [24,25]. Patients would benefit from the antiproliferative effects of PCIs while their antiviral activities are evaluated. Nelson and colleagues built upon this concept and tested the effects of PCIs on an animal model of HIV-associated nephro-pathy (HIVAN), an HIV-induced proliferative kidney disease. In two recent papers they demonstrated that PCIs have therapeutic activities in a transgenic mice model of HIVAN [26,27]. In these experiments, tg26 mice, which express the nonstructural proteins of HIV from a transgene integrated into the germline, were treated with flavopiridol or rosco-vitine. Both drugs ameliorated HIVAN, as evaluated functionally, histopathologically or genomically. As encouraging and important as these results are, however, pathogenesis in this model involves viral gene expression but no viral replication. It thus remains to be tested whether PCIs inhibit viral replication in vivo. These experiments nonetheless provide strong experimental support for testing PCIs in clinical trials against virus-induced proliferative diseases.

Much progress has been made since the discovery only 8 years ago of the antiviral properties of PCIs in cultured cells. However, much also remains to be done before the true antiviral potential of these drugs is elucidated. As discussed, the antiviral activity of PCIs in vivo at nontoxic doses has yet to be tested. The cellular targets that mediate these antiviral activities have yet to be identified, and the mechanisms whereby PCIs inhibit viral transcription have yet to be fully characterized. Many research groups, including ours, are currently working to address these important, but technically challenging, issues. Moreover, PCIs are tentatively scheduled to enter the first clinical trials as antivirals in 2005 [101]. We can thus expect that the true antiviral potential of PCIs will be known in the coming years. However, the strong antiviral potential exhibited by PCIs in cultured cells and preclinical trials has already resulted in a reconsideration of cellular proteins as valid targets for antiviral drugs. From this point of view, it can be considered that PCIs have already played a role in the development of antiviral drugs.

Acknowledgements

I apologize to all of the colleagues who also made significant contributions to the discovery of the antiviral properties of PCIs, but were not named, as well as to all those who made contributions that were not discussed, due to space constrains.