August 1, 2005 (Vol. 25, No. 14)
Structure-Based Libraried and Early Drug Metabolism and Phamacokinetics Studies
The main challenge in the process of taking a hit to a lead in drug discovery is figuring “what questions to ask and what information will best help focus lead optimization,” said Simon Hirst, Ph.D., CEO of Sygnature (Nottingham, U.K.), as he introduced the Hit-to-Lead track of the “World Pharmaceutical Congress,” sponsored by Cambridge Healthtech Institute.
The ultimate goal of hit-to-lead (H2L) is to identify lead compounds that are unlikely to fail in subsequent preclinical and clinical testing, and to do so while “maximizing the efficient use of drug discovery resources,” said Dr. Hirst.
H2L involves the integration of medicinal chemistry and ADME/Tox studies, generating focused libraries around the most promising hits and using those compounds for early and rapid generation of structure activity relationship (SAR) data.
Following high throughput screening (HTS) of large, highly diverse compound libraries to identify actives (compounds that bind to a target), the discovery effort must undergo a philosophical shift from an emphasis on what can be made (large, random libraries) to what should be made (smaller, focused libraries).
Acting as a bridge between HTS and lead optimization, hit-to-lead presents an important opportunity to fail compounds early, before significant resources have been spent on their development.
Representatives of Big Pharma described to the conference attendees their companies’ unique strategies for optimizing the H2L process. Although these strategies varied regarding organizational and infrastructure issues and terminology, common themes emanated from the presentations:
The importance of generating structural data early on and using it to guide lead generation and optimization;
The integration of chemistry, biology, and computational groups to maximize the synergy of their efforts;
The value of carrying out the various components of the H2L process in parallel for greater efficiency; and
The importance of beginning pharmacokinetic and toxicologic studies, both in vitro and in vivo, early to be able to predict problems with bioavailability, potency, selectivity, and toxicity that could ultimately doom a compound if it were to move into later-stage development.
Maximizing HTS Output
Mark Player, Ph.D., a team leader of lead generation at Johnson and Johnson Pharmaceutical Research and Development (J&J; Springhouse, PA), described how the company applies its Thermofluor HTS technology to screen soluble proteins for use in lead generation.
Thermofluor yields information-rich HTS by exploiting thermal effects on protein melting. Ligand binding to a target can change the melting temperature of the bound protein. A fluorescent dye that is quenched in an aqueous environmentwhen the protein is properly foldedbecomes unquenched when the protein unfolds, and it is exposed to a hydrophobic environment.
J&J uses Thermofluor for hit profiling, to identify active compounds that bind to the folded form of a target. Molecules that bind to the unfolded protein tend to be protein destabilizers and can be excluded from the pool of potential leads early on. Thermofluor is amenable to high throughput screening in 384-well plates; it accommodates low volumes (3 L) and requires <200 ng of protein per well.
Dr. Player also described the benefits of Thermofluor for “functional decryption,” as it enables direct detection of target-ligand binding without the need for any knowledge of the binding site or the function of the protein. The results of Thermofluor HTS can be combined with information gained from x-ray crystallography to characterize the target and optimize SAR for lead generation.
Focused Library Synthesis
Craig Lindsley, Ph.D., technology-enabled synthesis group leader at Merck & Co. (Whitehouse Station, NJ), described Merck’s strategy of having a dedicated H2L group within medicinal chemistry. He presented case studies involving two types of allosteric modulators to illustrate how screening paradigms are changing. The first focused on allosteric Akt kinase inhibitors.
Dr. Lindsley identified four key problems plaguing traditional approaches to lead optimization: a plethora of novel targets; few validated targets (lacking proof of concept); limited medicinal chemistry resources; and limited access to high throughput screening resources.
Often biologists would “develop neat targets, but couldn’t get an HTS slot to evaluate them,” said Lindsley. This prompted a paradigm shift in library synthesis strategies intended to expedite the workflow and move programs forward more quickly.
This shift necessitated a change from solid-phase combinatorial chemistry, which is used to produce large compound libraries, to an emphasis on solution-phase, parallel synthesis to generate focused, iterative libraries of 24- to 96-member arrays.
Dr. Lindsley described the one-week process instituted at Merck that takes a screening hit from focused library design through HPLC (yielding >98% pure compound), post-purification sample handling, and biological assay development.
A single chemist can oversee this parallel workflow, taking compounds from lead identification to proof-of-concept. The chemist can rapidly adjust library design and synthesis strategies based on emerging biological and drug metabolism and pharmacokinetics (DMPK) data.
As an example, Dr. Lindsley described a program to identify inhibitors of the serine/threonine kinase Akt, a potential target for cancer therapeutics. The three isozymes of Akt share >85% homology.
The project objective was to use an iterative analog library synthesis approach to develop soluble, low molecular weight Akt inhibitors that could be delivered to the medicinal chemists. Beginning with a central inhibitory building block, iterative library synthesis enables rapid optimization of compounds for potency, selectivity, novelty, solubility, and molecular weight.
In one year, a single chemist was able to produce 63 libraries that led to the development of multiple scaffolds as potential leads and the identification of a 550 molecular weight compound that could be used for in vivo studies to demonstrate proof-of-concept. These studies validated the hypothesis that selective Akt isozyme inhibition was possible and that a dual Akt1/2 inhibitor could induce apoptosis.
DMPK’s Predictive Power
During a subsequent panel discussion, when asked to identify the biggest challenge in transitioning from a hit to lead optimization, the speakers’ responses ranged from organizational issues (the need for dedicated chemistry groups to follow-up on novel compounds), to introducing automation into the chemistry process, as well as the need to push therapeutic groups that identify novel targets to generate the cell lines necessary for cell-based screening, and the challenge of how to use the DPMK and toxicity information obtained during the H2L process.
Another key concern was compound safety and the emerging importance of hERG assays to test for cardiotoxicity early on.
The speakers emphasized the need for better secondary assays with higher throughput. “Always let the biology lead you to which compounds to pursue,” said Andrew Baxter, Ph.D., a principal scientist in medicinal chemistry at AstraZeneca (Charnwood, U.K.).
Start with a good compound collection and filter out the noise, Dr. Baxter recommended, noting that both binding and functional assays are critical in this process.
Dr. Baxter describes hit-to-lead as turning “what comes out of HTS into something worth pursuing,” and “using DMPK data to push hits to leads.” Essential DMPK parameters should include clearance studies in rat hepatocytes (target criteria, <15 L/min/106 cells) and clearance studies in human liver microsomes (<25 L/min/mg), looking for species differences in vitro. These should be followed by in vivo rat IV clearance studies (<35 mL/min/kg; t1/2 >0.5 hr) that demonstrate a reasonably long half-life.
Determining whether the in vivo results agree with the hepatocyte data will suggest whether there is another mechanism for metabolism and clearance of the molecule. Oral bioavailability in the rat (F>10%), solubility (>10 g/mL), potency (IC50>10 M), and molecular weight (<450) analyses are also important components of H2L.
Other lead-profiling studies should focus on demonstrating clear SAR, generating selectivity data, achieving biological validation, and ensuring patentability.
“Oral bioavailability is one of the toughest things to fix,” said Baxter, and the inability to meet the minimum criteria could eliminate a compound from further consideration. Blocking NH groups to make compounds less basic is one strategy for improving oral bioavailability, but you also need to know when to just give up, he recommended. “One definition of success is to recognize a hopeless project early on,” added Dr. Baxter.
Why bother with DMPK?, he went on to ask. Although parallel synthesis will allow the generation of potent lead compounds, those compounds must eventually prove to be orally active and metabolically stable.
“You need to know you can get there in lead optimization by meeting early target parameters and building in DMPK during lead generation.” Setting up the H2L workflow so these studies can be carried out in parallel will shorten the overall discovery timeline and improve efficiency.
Another important component of H2L is metabolite identification and evaluation to assess the loss or gain of potency and metabolic stability if chemical groups are added or removed from a scaffold of interest. This contributes to knowledge about which parts of a molecule to change when preparing iterative, focused libraries.
“A high failure rate in H2L is okay,” concluded Dr. Baxter. The result will be compounds with a reduced chance of failing in the lead-optimization phase of development.
X-ray Crystallography is Main Bottleneck
Michelle Browner, vp of discovery science and technology at Roche (Palo Alto, CA), discussed Roche’s approach to optimizing lead discovery. Integrated, multidisciplinary, lead discovery teams combine medicinal chemistry, chemoinformatics, HTS, biochemical pharmacology, and DMPK.
These teams work closely with each therapeutic program team. They are responsible for generating “actives” (molecules with demonstrated activity against a target, as measured by the IC50; the primary output of screening assays), taking actives to hits (molecules with confirmed chemical structures and biological activity and with expanded characterization), and hits to leads (compounds with desired affinity, selectivity, and distribution properties).
The first step involves multiple lead generation by applying HTS, focused screening, and high throughput protein engineering (domain truncation and trimming, site-directed mutations, and purification tags and cleavage sites) in parallel.
To assess targets, Browner recommended using automated small-scale protein expression in E. coli or baculovirus to determine which proteins are relatively easy to express and purify and which to take forward into large-scale expression for HTS assay development and crystallography.
Random screening of a large compound library to find novel chemical scaffolds should proceed in parallel with focused screens, in which library design is based on the structural characteristics of the target and which will yield higher hit rates.
Roche then applies Spotfire DecisionSite (Spotfire; Somerville, MA) to visualize the screening data, identify clusters of active compounds (“nearest neighbor” determinations) and their common substructures, and select interesting clusters based on percent inhibition and molecular properties such as molecular weight.
These clusters yield the actives selected for follow-up and profiling, which includes IC50 determination, structure confirmation, and purity measurements. The actives selected are also subjected to evaluation with Rodin, Roche’s global web-based query system that incorporates screening-derived biological, chemical, and structural information.
Using this query system researchers can determine whether any of the actives are “frequent hitters” and should be excluded from the pool.
The actives are then prioritized into “hits” based on medicinal chemistry evaluations and the use of secondary, lower throughput assays. Hit profiling involves activity confirmation in focused screens and biophysical characterization.
Evidence of nonspecific binding leads to elimination of a hit from further consideration. Roche also evaluates the binding kinetics of each hit. At the same time, high throughput protein crystallization is initiated. Applying automated crystallography early on in a parallel timeline allows for hit prioritization using protein structure information derived from x-ray crystallography of bound structures.
“The bottleneck in hit profiling is finding crystals,” said Browner. As the co-crystallization studies progress, in vitro DMPK can begin, together with efforts to assess novelty and SAR studies. The outcome of this process is a set of leads that can be passed on to the medicinal chemists.
Browner identified three current gaps in the hit-to-lead process:
Time: about a nine-month timeframe to progress from HTS to profiled hits;
Resources: limited resources and the need for increased automation to drive parallel processing; and
Throughput: limited throughput of secondary assays.
Structure-Based Design
Ray Unwalla, Ph.D., a senior scientist at Wyeth Research (Monmouth Junction, NJ), described Wyeth’s use of structure-based design in a program aimed at identifying highly selective ligands for estrogen receptor (ER), a member of the estrogen nuclear receptor family.
ER was identified as an attractive target for the discovery of synthetic estrogen ligands because the protein had been cloned and the receptor has an appealing tissue distribution. Although its presence is widespread, it is not the dominant estrogen receptor in the uterus or in breast tissue.
The starting point of the program involved working with available structural information on ER and ERa ligands to understand the nature of target/ligand binding for the receptor.
17-estradiol, for example, is a nonselective inhibitor of Era. It displays rigid anchoring in the binding site with little flexibility. This finding led biologists at Wyeth to do mutation studies to understand the differences between 17-estradiol binding and 16a1-estradiol binding and to probe the selectivity of the target-ligand interaction.
The results pointed to a particular residue that impacts selectivity, providing the medicinal chemists with an opportunity to modify chemical groups at that site to optimize selective binding.
They introduced a chemical hinge that would facilitate access to the binding pocket and then experimented with various chemical analogues to identify more selective templates. They succeeded in identifying a selective ligand for Er that would not be expected to cross-react with estrogen receptors found in the uterus, to stimulate breast tissue, or to cause weight gain.
Andreas Steinmeyer, Ph.D., a medicinal chemist and head of a hit-to-lead team at Schering (Berlin), also emphasized the importance of obtaining three-dimensional structural information on the protein-ligand complex as soon after HTS as possible, so this data can be applied to lead generation.
Identifying high-quality structures for lead optimization requires a thorough evaluation of the potential liabilities of lead structures, focusing on activity/selectivity, “leadlikeness,” pharmacologic and ADMET properties, synthetic accessibility and optimization potential, and patentability.
The H2L workflow at Schering begins with large-scale HTS of 700,000 pooled compounds, followed by focused library design and repeat HTS of about 2,000 single compounds.
This reduces the compound pool 10-fold, sending 200 compounds forward for IC50 assaying. The result is 100 IC50 hits.
As these 100 compounds work their way through the hit-to-lead workflow, purity and structural evaluations bring the number down to 50 validated hits in about one month’s time, and subsequent in vitro efficacy, selectivity, and toxicology studies produce 15 clusters (or single compounds) of “qualified hits” by the end of the third month.
At this point in the process, qualified hits must be resynthesized to yield more compound for subsequent in vitro and in vivo evaluations. These evaluations conclude with the identification of 13 lead structures (or clusters) after about a 10-month period, and these structures are then moved forward into the lead optimization process.
Dr. Steinmeyer recommended that a team of medicinal chemists be involved in hit-to-lead early on, contributing to hit selection based on expected optimization potential, synthesis of qualified hits, synthesis of first analogs for preliminary SAR, and synthesis of virtual hits, novel compounds proposed as a result of computational chemistry, and competitor compounds.
Identifying Global Liabilities
The hit-to-lead stage of drug discovery offers an opportunity for early prediction of potential toxicity and adverse side effects. Vito Sasseville, DVM, Ph.D., director of drug safety evaluation at Millennium Pharmaceuticals (Cambridge, MA), described the company’s Discovery-Assay-By-Stage (DABS) paradigm, which is intended for the aggressive evaluation of development-limiting liabilities early in drug discovery.
DABS determines what assays are needed to advance compounds to the next level. It identifies which tissues and specific cell types in these tissues that express the target, defines expression in animal models and toxicology species, describes the short-term and long-term effects of target inhibition, and determines off-target effects, mutagenic potential, and cardiovascular liabilities.
Early assays should focus on expression profiling of the target, knockout mouse phenotyping, and toxicology/pharmacology evaluation.
To optimize the limited compound resources early in drug discovery, “it is important to do toxicology assessments early stage,” said Dr. Sasseville, using small-scale toxicology studies to look for global liabilities (“sentinel tox”), as well as “satellite tox” for more targeted toxicity assessment during in vivo efficacy.
“Genetic toxicology is the nail in the coffin for many programs,” explained Dr. Sasseville, emphasizing the importance of relying on software-based predictions of mutagenicity in H2L and incorporating a variety of foolow-up mutagenicity and clasotgenicity assays, including Ames II, Ames spot test/Mini Ames, in vitro micronucleus assays, screening chromosome aberration assays, and mouse or rat in vivo micronucleus assays.
Safety testing should include hERG assessment to identify the risk of QT (cardiac rhythm) issues. Millennium uses a high throughput hERG binding assay early in the discovery process, with electrophysiology and in vivo tests in late lead optimization.
Dr. Sasseville presented a case study of an NF-kB inhibitor in which the DABS predicted for the on-target and exaggerated pharmacology’ as well as the off-target toxocity.
In the case described, the knockout phenotype predicted the potential for toxicologic liability, while the satellite toxicology studies revealed a proinflammatory response at high doses. Repeat dosing studies confirmed the predictions, defined a safety window, and provided clues to the mechanism of action.