Successful drug hunters are an elite group within the pharmaceutical industry, particularly given the difficulty of discovering and developing medicines. What distinguishes those who do find drugs that work from those who do not is not easily defined. Here, we track one success through a decade of exhaustive and highly innovative work by a young group of researchers at Pfizer Sandwich Laboratories in the United Kingdom. They discovered and developed maraviroc, a vital addition to the armory of HIV drugs.
Back in 1996, a series of reports published in quick succession in Nature,
Maraviroc is noteworthy for several reasons. Most importantly, it represents a new option for HIV patients with 'CCR5-tropic' virus infection, who are beginning to fail on their existing regimens. It is also the first antiviral drug of any kind to act on a host rather than a viral target. It is the first oral HIV fusion inhibitor to reach the market: Enfuvirtide (Fuzeon), which employs a different mechanism to block the fusion of the HIV envelope with the membrane of T-helper cells, is an injectable drug.
For those who led the development effort, the drug represents a substantial scientific achievement and a major milestone in their careers. "It was a very fast program by industry standards. From idea to launch it took 10 years: the fastest R&D program we have ever conducted in Sandwich," says Dr. Annette Doherty, Vice President of Research, Pfizer Global Research & Development (PGRD) and Site Leader of Pfizer Sandwich Laboratories.
In late 1996, when Pfizer began putting put a project team together at its UK PGRD facility, adjacent to the picturesque town of Sandwich, near the Kent coast, neither HIV nor antiviral drug research were established strengths. The research teams in Sandwich had discovered and developed a number of successful medicines, including fluconazole (sold as Diflucan) and voriconazole (Vfend), two life-saving antifungals, but the scientists who were initially assigned to the project did not yet have a track record of success in drug discovery. "No one had taken a molecule into man before - not even into Phase I," recalls molecular biologist Dr. Manos Perros, now Executive Director and Head of Anti-Viral Research at Pfizer Sandwich, who was the biology program leader and the first recruit to the team. "Despite our efforts and over a decade of investment, the antiviral group had never had a breakthrough."
At the time, many pharmaceutical companies had exited from HIV drug discovery, following the introduction of combination drug regimens, known as highly active antiretroviral therapy (HAART), which offered significant advances on earlier treatment protocols. "There was some convincing to do, to have the organization invest back into HIV," says Perros. What helped was the fact that the CCR5-32 findings were "truly compelling". The mutation was the equivalent of "a knockout experiment done by nature," says Dr. Tony Wood, now Vice President and Head of Chemistry and Experimental Medicinal Sciences at Pfizer Sandwich, who headed up the medicinal chemistry program that led to maraviroc.
In addition to an understanding of the molecular genetics underlying the HIV resistance phenomenon, information on the drug target, CCR5, was also available. The cloning of the receptor, originally designated CC-CKR5, had been reported in the spring of 1996,
During HIV infection, CCR5 acts as a co-receptor. An initial binding event between a CD4 glycoprotein receptor on the surface of a T-helper cell and an HIV envelope protein called gp120 triggers a conformational change in the viral protein that allows it to bind CCR5 or, in the case of certain other types of HIV-1 virus, an alternative co-receptor called CXCR4. That second binding event is the committing step in the fusion of the viral envelope with the membrane of the T-helper cell. It exposes the HIV 'spike' protein gp41, which then penetrates the host cell membrane. Fusion of the two membranes follows, which allows viral infection of the cell to proceed.
Finding a small, orally available molecule that could bind to CCR5 and prevent its interaction with gp120 was the challenge. "From one side of the equation, it was an easy-to-do G-Protein Coupled Receptor (GPCR). From the other side, it was an impossible protein-protein interaction," says Wood. In fact, despite a number of approaches in the pharmaceutical industry, maraviroc remains the only approved drug against this family of receptors.
Establishing an effective screening assay also threw up uncertainties. While the overarching goal of the screening program was to identify drug-like molecules that could disrupt the binding of gp120 to CCR5, at the time there was no direct way of doing so. "Nobody had any assays that could measure that interaction," Perros says. "It would have taken another year or more to develop one."
Instead, the maraviroc team developed an indirect screen, based on measuring the interaction between CCR5 expressed in a human cell line and radiolabelled MIP-1 beta. It was not clear at this stage whether gp120 used the same CCR5 binding site as the receptor's natural ligands. Indeed, subsequent work at Marc Parmentier's lab at the Free University of Brussels in Belgium established that gp120 actually bound at a different location on the receptor. Because small molecules were unlikely to inhibit the interaction directly in any case, the thinking was that potential drug candidates would act via allosteric inhibition, that is by binding at a distal location on the receptor and locking it into a shape that would make gp120 binding impossible.
High-throughput screening of a compound library containing around a million structures yielded dozens of hits. Most were eliminated using standard medicinal chemistry tactics, such as applying Lipinski's 'Rule of Five', which defines several general physical properties shared by most synthetic small molecule drugs, and assessing parameters such as target affinity and ligand efficiency, which measures the potency of a molecule as a function of its molecular weight (the smaller the better is a general rule of thumb). A further compound series was discarded because of potential toxicity concerns. That left four hit series, two of which formed the basis of the program that eventually led to maraviroc - after the synthesis and evaluation of some 965 analogs over two and a half years.
The general optimization goals at this point were to increase affinity for CCR5 while removing inhibition of cytochrome p450 and reducing lipophilicity. "We also had to find antiviral activity, which was the most important thing," says Wood. The two structures that were taken into 'hit-to-lead' studies actually had no detectable effect on the virus.
The original decision to proceed with a high-throughput screening assay that did not directly measure the binding of gp120 to CCR5 did cause sleepless nights, Perros confesses, and it was repeatedly challenged in progress reviews with the site's management when antiviral activity remained elusive. It took nine months to understand why, he says. All the CCR5 inhibitors that had been synthesized bound at the same site, in a pocket deep within the receptor's three-dimensional structure.
However, the resulting change in its shape did not appear to influence the ability of the receptor to bind the viral protein. The introduction of an amide group was the key step in triggering a conformational change that would prevent gp120 from binding. The initial HIV work was performed at a Level 3 containment facility at Northwick Park Hospital in London, as Pfizer had not yet built its own Level 3 labs in Sandwich. Getting the first evidence of antiviral activity back from London was a major highlight. "I remember looking at the data and saying, 'Oh wow - it works!'," says Perros.
Progressive structural tweaking, informed by computer modeling and by theory, helped to optimize the other key parameters. Restricting conformational flexibility via the introduction of a tropane is a common method that can increase potency in some cases. An imidazopyridine group known to be associated with cytrochrome p450 inhibition, which is an important safety consideration, was removed. Cytochromes play a central role in the metabolism of many drugs. Any effect on their activity can lead to adverse drug interactions, such as a toxic accumulation of a second drug, in patients on combination regimens, which are the norm in HIV therapy. Reducing lipophilicity was achieved by increasing the polarity of the structure. Highly lipophilic or 'greasy' molecules can appear to bind to a target protein in aqueous solution, when non-specific hydrophobic effects, rather than target affinity, are the cause of the apparent interaction. Moreover, excessive lipophilicity can have a negative impact on bioavailability. "The body metabolizes greasy things and eliminates them," says Wood.
The resulting species, a benzamide, which demonstrated both strong receptor binding and promising levels of antiviral activity, became the focus of structure-activity relationship (SAR) studies. These probed deeper into the chemical basis underlying the biological effects observed. Those biological effects were followed using a series of novel assays devised by Perros's group.
A key innovation was the development of a fusion inhibition assay that gave a direct indication of a lead structure's antiviral activity. This consisted of two cell lines, one of which expressed the host receptors CD4 and CCR5, while the other expressed the viral proteins gp120 and gp41. "Getting those [viral proteins] expressed in a cell line was the challenging piece," Perros says. To avoid the need for microscopic inspection, to determine whether fusion between the two cell lines took place, the assay incorporated a standard beta-galactosidase reporter gene, which was put under the control of a promoter that required the HIV-1 transcriptional activator Tat. The reporter gene and the gene encoding Tat were expressed separately, in the two distinct cell lines, so beta-galactosidase activity could only occur after cell fusion had taken place.
In addition to the benzamide lead structure, two other promising molecules were synthesized, an isoproplyamide and a cyclobutyl amide. Chiral synthesis allowed the Pfizer team to assign the biological activity of each to a single optical isomer. The S enantiomer of cyclobutyl amide, which was responsible for the antiviral activity, still displayed some cytochrome p450 inhibition. However, this was removed by modifying the molecule's central piperidine core, to produce two isomers of a tropane-derivative. These demonstrated enhanced antiviral activity, as well as an absence of cytochrome P450 inhibition.
The more potent of the two isomers was close to being nominated as a candidate drug at this point. In addition to inhibiting binding of gp120 to CCR5, it demonstrated potent activity against a range of HIV-1 isolates. The compound was starting to look good. And then the whole effort hit a major roadblock. Around this time, drug developers had started to recognize inhibition of the HERG channel, encoded by the extravagantly titled Human Ether-a-go-go Related Gene (HERG), as a safety concern. The HERG channel forms a subunit of a voltage-gated potassium ion channel found in cardiac and nerve tissues. Loss of HERG channel function disrupts the heart's electrical activity (the technical term is QT interval prolongation) and can lead to cardiac arrhythmia, a disturbance in heartbeat that can be fatal. The HERG protein binds many drugs promiscuously, including SCH-C, a Schering-Plough CCR5 inhibitor that was withdrawn from clinical development because of the problem. Before it ever moved into the clinic, Pfizer's most promising compound, inevitably, exhibited HERG channel inhibition as well.
"The question then became: how on earth can we avoid this?," says Wood. "That became essentially the issue for the remaining 18 months of the program." The solution lay partly in a drug that Pfizer had itself developed. Dofetilide, marketed as Tikosyn, is used to treat irregular heartbeat - and it binds to HERG. The team immersed itself in the literature, Wood recalls, in order to learn about how the HERG channel binds compounds like dofetilide. The aim now was to remove the dofetilide-like activity, which could be followed using a dofetilide binding assay, while ensuring that the resulting compounds continued to tick the other essential boxes, in terms of antiviral activity and bioavailability. 'Tuning' the polarity of the compound helped to reduce HERG binding, but the increased polar surface area of the molecule created new problems with respect to oral absorption. "It's nailing a blancmange to a ceiling in some ways," says Wood. The team next looked at improving that parameter by reducing potential hydrogen bond interactions. While one resulting lead, in particular, showed promise, its pharmacokinetic performance was not up to scratch. It was too easily cleared from the liver.
At this point, an analysis of the data helped Wood's group to define an appropriate lipophilicity range or window within which an acceptable level of stability could be attained. A fundamental problem was that permeability and lipophilicity appeared to be linked. It was difficult to increase one without also increasing the other; and increasing lipophilicity resulted in faster clearance. The team decided to take a look back at an earlier triazole that had been discarded because of its lack of antiviral activity. It had lower molecular weight and lipophilicity, however, as well as good affinity for CCR5. All of the learning that had been amassed on the project thus far was applied to optimizing this structure. Antiviral activity was boosted by generating a series of tropane analogues. Modification of an amide group further improved potency, while the addition of a pair of fluorine atoms reduced its metabolic breakdown. The result of these modifications was a 4,4'-difluorocyclohexlyamide, which was initially dubbed UK-427,857. That structure is now known as maraviroc.
Additional in vivo testing against multiple HIV-1 isolates confirmed its promise. It appeared active against 43 geographically diverse isolates that represented several HIV-1 clades or families.
Trials, not Tribulations
At this point, the program started to gain wider visibility within Pfizer. "For every piece of data, people were waiting with baited breath and excitement," says Annette Doherty. The big question now was whether this compound would be both safe and effective in HIV-infected individuals. But several other supplementary issues had to be explored as well. It was necessary to confirm that maraviroc was indeed an antagonist and not an agonist that would actually bind the receptor like a chemokine and be transported to within the cell, with potentially disastrous consequences. "There is evidence that CCR5 agonists activate viral replication intracellularly," says Perros. Moreover, receptor occupancy needed to be highly efficient. "Even a small number of receptors can mediate viral entry," he notes. "You need to occupy the receptor completely."
Perros' group discovered that all the compounds that showed antiviral activity had in common really tight binding and slow offset from CCR5. This is because binding is a dynamic event, and a rapidly reversible binder comes on and off the receptor constantly, leaving the virus a window during which it can bind the receptor. Unlike small molecule binders, once HIV has bound CCR5 it is committed to entry, and subsequent infection. Slow-offset binders of the series from which maraviroc was issued bind in a two-stage process, that includes a key conformational change by the receptor. Once this has occurred, the receptor is "locked" into an inactive configuration, which the virus can no longer recognize. This was a key realization for the program, and indeed, maraviroc's tight binding to CCR5, which was borne out in assays of serum samples taken from clinical trial participants, is in fact a key feature of its efficacy.
As with any antiviral or antibacterial drug in development, the propensity of the target microorganism to develop resistance also had to be assessed. Virologist Dr. Mike Westby joined the project to explore this issue. With HIV-1, the question of drug resistance is a particularly urgent, as it undergoes what's known as 'error-prone' replication. A retrovirus, HIV-1 reproduces by transcribing single-stranded RNA into double-stranded DNA, using reverse transcriptase to catalyze the reaction. Unlike DNA polymerases (which use DNA as a template), reverse transcriptase does not incorporate a 'proof-reading' capability, to filter out replication errors. That, combined with HIV-1's very high rate of replication, results in a rapid accumulation of mutations in its proteins and can lead to the rapid development of drug resistance. It's the reason why combination therapy involving several drug classes has become the standard treatment for HIV infection.
The specific mechanism by which maraviroc operates raised a number of issues. Maraviroc targets a human rather than a viral protein, making it unique amongst anti-HIV agents and meaning there was no precedent for how resistance to this compound might develop. Specifically, it was necessary to establish whether maraviroc could exert a selective pressure in favor of viral strains with mutated gp120 proteins that were capable of binding CCR5 in its altered (compound-bound) conformation and causing infection. In addition, maraviroc electively inhibits forms of HIV-1 that exploit the CCR5 co-receptor to enter immune cells - the most prevalent form of the virus during early infection. However, it does not affect 'CXCR4-tropic' viruses, that is, those that exploit the alternative CXCR4 co-receptor when entering CD4+ T-cells. CXCR4-tropic viruses appear to reproduce less efficiently but emerge in around 50% of late-stage cases of HIV-1 infection and are associated with faster disease progression. A small percentage of strains are dual- or mixed-tropic, capable of using either co-receptor. It was critical that the use of maraviroc did not trigger a switch from one co-receptor to the other. At the time, little was known about whether this could happen. "That was really a big, big question we had to try and address in the clinical program, [in a way] that would put the fewest number of people possible at risk," Westby says. "What we find now is that doesn't happen," Perros says.
In vitro passage experiments, in which a lab-adapted viral strain was sequentially exposed to increasing concentrations of maraviroc, indicated that the development of resistance via gp120 mutations was also unlikely.
The first clinical studies in healthy volunteers were undertaken in October 2001. The first evidence of efficacy was obtained in 2003, from two Phase IIa trials conducted in a total of 63 individuals infected with CCR5-using HIV-1 (10). Patients on the highest dose of maraviroc achieved mean reductions in viral load of 1.6 log10 copies per milliliter of serum (almost 99%). "These results were absolutely fantastic," says physician scientist Dr. Elna van der Ryst, Senior Director Clinical R&D at Pfizer Sandwich. She joined the project at the beginning of February 1999 and acted as exploratory clinical leader. She later became European clinical lead during maraviroc's development phase. Van der Ryst, who is South African, is particularly passionate about her work in HIV research. "I've lost many friends and patients over the years," she says. "I don't think there's anyone in South Africa who can say that HIV does not affect them."
The Phase IIa data were not unexpected. The in vitro data strongly suggested that maraviroc would work in a controlled study of short duration. "What was really exciting was when the Phase III data came back," says Perros. He was checking into a hotel in Ann Arbor, Michigan, around midnight, when an email came through from Dr. Martin Mackay in the UK (then a senior R&D executive and now President of Pfizer Global Research & Development) who wrote that he got a lump in his throat when he saw the data. For a split second Perros thought he meant the therapy had failed. In fact, it was a resounding success. Mike Westby was present in the room in Sandwich for the first readout of the 24 week Phase III data. "I think that will live with me forever," he says.
Two studies following similar protocols were performed, MOTIVATE-1, which enrolled 601 participants in the USA and Canada, and MOTIVATE-2, which recruited 475 patients across Europe, Australia and North America. To be eligible for inclusion, subjects needed to have at least 5,000 copies of HIV-1 RNA per milliliter (ml) of serum.
Patients received either placebo or a 300 mg dose equivalent of maraviroc once or twice daily, along with an individually optimized background regimen. The studies yielded highly consistent results (see MOTIVATE tables). Those on maraviroc were around twice as likely as those in the placebo group to attain a reduction of viral load to less than 400 copies HIV-1 RNA per ml. They were also twice as likely to have undetectable virus, that is less than 50 copies of HIV-1 RNA per ml, which van der Ryst describes as "the holy grail for the treatment of HIV". They also had a two-fold improvement on CD4 cell count.
"When that data came out, it certainly was one of the most exciting things I had ever seen as a scientist," she says. And there was more to come. The patients and clinical investigators involved in MOTIVATE-1 and MOTIVATE-2 remained blinded for another 24 weeks, and the 48-week data demonstrated the durability of the response. Those receiving maraviroc plus background combination therapy were even more likely to have undetectable virus than those on placebo plus background therapy (see bottom table on p. 74). "There was no drop-off," says Van der Ryst. Its favorable safety profile also held up.
However, Maraviroc failed to demonstrate non-inferiority against a comparator drug, the non-nucleoside reverse transcriptase inhibitor efavirenz, in a separate, head-to-head trial in patients who had not previously received therapy. Pfizer continues to explore its potential as a frontline therapy, but for now, it is approved as a treatment for treatment-experienced patients only - and only in patients with CCR5-tropic virus. Although the clinical program demonstrated that maraviroc does not harm individuals infected with CXCR4-tropic HIV strains, the drug does not confer any benefit either, and patients need to be tested to confirm they have CCR5-using virus only before they can start taking the drug.
Maraviroc's approval is part of a recent watershed in HIV therapy. The timing is important, as resistance to HAART regimens was beginning to emerge, leaving an increasing number of late-stage patients without effective treatment. Other recent significant HIV drug approvals include darunavir (Prezista), a new protease inhibitor, and raltegravir, the first-in-class oral integrase inhibitor. All three considerably broaden the range of available options. "The face of HIV therapy has changed - not only because of maraviroc," says Van der Ryst.
For those who brought the project from initial idea to marketed product - upwards of 200 Pfizer staff were involved in total - there is the considerable sense of achievement in being involved in what is a substantial and singular piece of innovation. "This has been an incredible journey," says Mike Westby. There are many CCR5 discovery programs, but still only one drug, maraviroc, Manos Perros quips.