Discovering new drugs is an uncertain journey and can throw up some surprises along the way, writes Dr Sarah Houlton
Drug discovery is not simple - it's a slow and complicated process that combines chemistry, biology and a variety of other scientific disciplines, and a wide range of modern techniques are being applied to try and make it faster and more efficient. Before medicinal chemists can get to work creating lead compounds, they need a starting point, but lead generation is one of the major bottlenecks in the drug discovery process.
"For successful lead finding, and lead generation, you need an array of techniques," Karl Ziegelbauer, vice president, product-related research at Bayer Pharma told CHI's Mastering Medicinal Chemistry conference held in San Francisco earlier this year. "These include target discovery, high throughput screening, libraries, combinatorial chemistry and computational chemistry, not to mention early pharmacokinetics and toxicity studies."
The compound library and high throughput screening (HTS) are the two basic strategic assets in drug discovery, he said. "The quality of the compound collection is of vital interest, and it has to be nurtured continuously. Compound consumption has to be minimised, and this can be done by miniaturisation in HTS. A high quality compound collection is the key to success in lead generation."
Back in 1994, he explained, Bayer had about 115,000 compounds in mg to kg quantities, and 15,000 were scattered around various biology laboratories. Compounds were retrieved manually on request, which was hardly conducive to meeting the demands of HTS. "Since 2001, we have had a compound repository, and it now has 1.85 million compounds and 2.6 million substance containers," Ziegelbauer said. "It's designed to hold 6.5 million containers. To build this high quality compound collection has taken a decade and Euro 100m in investment."
About 30% of these compounds have come from traditional medicinal chemistry, with the remainder from combinatorial chemistry programmes. More than 90% of the library's entire content is not commercially available, and consists of custom made compounds, whether from the in-house med chem programme or sourced from professional compound library companies such as ArQule and Comgenex.
The company's HTS programme has had a number of successes, including the Factor Xa inhibitor BAY 59-5939, now known as rivaroxaban, which is currently in Phase III trials. "It's an old target, but all the early development compounds have had amidines or basic moieties, so they don't have good oral availability," he said.
The initial screening came up with a lead whose IC50 was below 1µM, but they were unable to achieve the desired pharmacokinetic profile. "We went back to the hit file and found a similar structure that had a very poor IC50 of 20µM," he said. "But this led to BAY 59-5939, which has an IC50 of 0.7nM. It's a new chemical class, with a novel binding mode. It was optimised using structure-based design processes, and has a very high oral bioavailability. It is highly selective potent antithrombotic, and is in Phase III trials for two long-term chronic indications, venal thromboembolism and atrial fibrillation.
"High throughput screening is our single major source of drug leads," Paul Leeson director of medicinal chemistry at AstraZeneca, Charnwood told the Advances in Synthetic and Medicinal Chemistry conference in St Petersburg this August. "It provides unique pharmacophores, although sometimes other people do come up with the same one! Drug candidates from HTS have a lower molecular weight and tend to need fewer synthetic steps than those found through other methods. Our anti-inflammatory lead optimisation programmes deliver drug candidates faster than non-HTS projects, and compound collection enhancement is the key to further success in hit to lead programmes."
However, he sounded a note of caution. "It's challenging to get a good lead," he added. "We put a great deal of emphasis on pharmacokinetic data at this stage. But the whole process has got to be fast - we look to find the first lead in one to three months, and take a further six to nine months to go from hit to lead. If there's no promise, it's also important to terminate a project early."
Leeson gave several examples of projects where they have rapidly managed to go form the hit to the lead stages. In the CXCR2 programme for rheumatoid arthritis, for instance, they found three hits. "This was a good thing, as two of them turned out to be no good!" he said. "While the pharmacokinetics of the third were not perfect and the bioavailability was not good, we pursued it anyway, and found it very amenable to parallel synthesis."
They found that small substituents were best on the aromatic ring, but bioavailability was still a struggle. By losing the hydrogen bond donor at the top, they got a molecule with lower potency. However, they could not improve on this. Leeson explained that the key to fixing it was switching from a nitrogen to an oxygen substituent on the left hand side of the molecule. "Adding another chlorine atom gave spectacular bioavailability," he said. "Although it also has some activity at another receptor, CCR2b, this is about 10 times weaker so we were not worried about it. The molecule is as efficacious as the COX-2 inhibitors, and so it was progressed to development."
Another example of a project where the lead compound originated in an HTS programme was described by Mark Norman, director of medicinal chemistry in Amgen's chemical research & development department, at the San Francisco conference. Capsaicin is the pungent component in chillies and has a medical tradition as folk remedy for toothache, as well as being used in Deep Heat type creams to treat aching muscles. The receptor it hits in the body is the transient receptor potential ion channel vanilloid receptor 1, or TRPV1. These receptors are expressed in patients with chronic pain, such as those with post-herpetic neuralgia and diabetic neuropathy. Agonists of this receptor like capsaicin activate the ion channel, causing neuronal firing that leads to desensitisation of the pain receptors. However, an antagonist could also have therapeutic potential, as if the receptor were blocked it would have the same effect as being desensitised by the neuronal firing. Experiments in TRPV1 knock-out mice supported this theory, so Amgen started a programme looking for an antagonist.
The company's compound collection contained more than a million small molecules at that point, and they developed an assay that could identify both agonists and antagonists. Both fire the TRPV1 receptor in the initial assay, and then adding capsaicin enabled them to pinpoint which hit was an antagonist, and which an agonist. Out of this process, they identified a compound with an IC50 of 1200nM that provided a start for their medicinal chemistry programme.
An obvious step was to conformationally restrict the centre of the molecule, and screening a selection of cyclic analogues led to a pyrimidine compound with an IC50 of 120nM - an order of magnitude better than the initial hit. Further optimisation of the linking groups and substitution patterns improved this to 7nm by introducing an oxygen linkage and a CF3 group, and by changing the quinoline A ring gave a selective compound with an IC50 of 7nM. This has a good bioavailability and low clearance, and successfully blocks flinching and capsaicin-induced hypothermia in rats, but ultimately it was unsuccessful because of the excessively long half life of a fortnight in Phase I trials. However, it provides another good example of how high throughput screening combined with clever medicinal chemistry is an excellent method of producing high quality drug candidates.
While HTS is providing the impetus for a growing number of lead discovery programmes, natural products continue to present a number of good starting points for medicinal chemists. Rapamycin, for example, is an immunosuppressant drug in its own right, but analogues also have great potential in cancer therapies. As Wyeth's Jerauld Skotnicki told the St Petersburg conference, it provided the starting point for temsirolimus, its recently approved drug to treat advanced renal cell carcinoma, which also has potential as a cytostatic agent to treat a variety of other tumours, too. Rapamycin itself was initially isolated from Streptomyces hygroscopius in a soil sample from Rapa Nui on Easter Island.
He explained that it hits at least three pathways involved in cancer proliferation - AKT, PTEN and mTOR - which are mutated in a variety of human cancers, including renal, prostate, breast, glioma, endometrial and melanoma. Preclinical activity has been found in human mouse xenograft models, and he claimed that rapamycin analogues have potential for activity in a broad range of tumour indications by virtue of their targeted mechanism.
A number of possible sites for the modification of the rapamycin structure can be envisaged. Potential chemical changes include triene alteration, alcohol functionalisation, carbonyl manipulation, pipecolinate modification and ring opening, contraction and expansion strategies. "While the initial aim was for improved drugs for transplantation, we found it was better for cancer," he told the conference. "We were initially looking for a back-up compound for rapamycin, and were looking to improve stability, solubility, crystallinity and find something that was easier to synthesise. But we were also looking for other therapeutic indications. This was an important strategic decision as we recognised the importance of mTOR in an oncology setting."
All of the points they targeted for modification were chosen for specific purposes, he said. Primary among these were changes at the C-42 centre. "It's away from the macrocycle, and we thought changes here might adjust the physical properties without affecting binding," he claimed. "We also thought it might be more accessible for chemical modification than some other centres. It has a hydroxyl group, so we thought there may be more opportunities to modulate the properties." More than half the compounds they prepared in the medicinal chemistry programme included modifications at this centre and, he added, for the most part this does not adversely affect the core activity, except if the substituents added are too bulky or polar. Many of the derivatives they made were either carbamates or esters and, importantly, they did not want to add another chiral centre as this would lead to further complications with potential the need for a separation of diastereomers.
In all, they made more than 600 different analogues, giving 31 lead compounds plus 40 further analogues. This was narrowed down to three advanced leads, of which two became pre-project compounds and, ultimately, CCI-779 was chosen as the best option to move forward with because of its solubility properties - it is amenable to both oral and intravenous formulation.
A final example of a natural product that has provided a med chem lead was given by Bill Greenlee, vice president, chemistry CV and CNS drug discovery, at the St Petersburg conference. There is a significant need for more effective antiplatelet therapy, as current treatments have side-effect issues and up to a quarter of patients develop resistance to aspirin. Resistance to the main alternative, Sanofi-Aventis's Plavix (clopidogrel), has also been seen. As thrombin is the most potent agonist of platelet activation, an orally active thrombin activator could well prove a useful addition to the therapeutic options for preventing clotting.
The starting point for their programme was the natural product himbacine, a potent non-selective muscarinic antagonist extracted from an Australian rhododendron that ha previously been a lead in the company's Alzheimer's programme. Solubility problems were overcome by introducing a lipophilic substituent, and the first candidate to make it into development, SCH-205,831, included a CF3 unit which gave acceptable oral bioavailability in rats, and the half-life in primates indicated that it may be suitable for once a day dosing. However, when carrying out tests looking for CYP450 enzyme induction in the liver, they found that two of these were significantly induced. As CYP450 induction is a major cause of drug-drug interactions, this candidate was dropped and they went back to the drawing board.
The group felt that if they were to look at the major metabolites of the drug and try them instead, then that might prove a safer strategy. They found three major metabolites, and the minor one, SCH-219,122, which has a hydroxyl group on the C-ring, was clean in the eight-day CYP450 induction assay. It also had improved pharmacokinetics, and they decided to progress the compound into development. However, this one hit the buffers during another required preclinical test - mass balance. In this, animals are dosed with a radiolabelled form of the development candidate, and the amount that is excreted measured. Unfortunately, clearance of the radiolabels was too slow, and there was a persistent metabolite that simply accumulated.
Rather than abandoning the project at this point, as many medicinal chemists might have been tempted to do, they persevered. Several further modification strategies were pursued, including adding substituents to the C-ring, making this ring aromatic or heterocyclic, and replacing the CF3 group with something less lipophilic. Substitution on the C-ring didn't really work, and making it aromatic was a little better, but still had too slow clearance. Adding heteroatoms such as S, N and O was more successful, and in particular a piperidine analogue looked promising, being clean in the enzyme induction assay and having good clearance, but it had solubility issues.
They then had the idea of moving the carbamate out of the piperidine ring, and after further evaluation the ethyl carbamate derivative became the third development candidate, SCH-530,348. It looked good - 1mg/kg gave 100% inhibition of thrombin at 24 hours, and its dissociation half-life was more than 20 hours, with no accumulation of the parent or metabolites, and no CYP induction. In addition, as it is a crystalline-free base, there are no formulation issues. It has recently completed Phase II trials in 1,Bill 000 patients undergoing elective angioplasty, and no increases in bleeding were seen, with a loading dose of 40mg and maintenance doses of 1 or 2.5mg. Reductions in death, major coronary events and myocardial infarctions were almost significantly significant, and this was higher than expected. Greenlee reports that two Phase III studies in 300,000 patients across 700 centres are about to start - a testament to their persistence in the face of all of those adverse preclinical results.