Send to printer »

Feature Articles : Mar 15, 2010 (Vol. 30, No. 6)

New Predictive ADME/Tox Strategies Reduce Attrition

Emerging Methods Now Assess Absorption, Transport, and Bioavailability Sooner
  • Susan Aldridge, Ph.D.

Predictive toxicology tools and safety strategies put in place earlier in the drug-development process will begin to deliver in the future, according to Stephane Dhalluin, Ph.D., director of investigative nonclinical safety at UCB Pharma, who spoke at Mondial Research Group’s “Predictive Human Toxicity and ADME/Tox Studies” conference held recently in Brussels.

As an example, he referenced data gathered between 1991 to 2000 and between 2001 to 2007 that shows a decrease in attrition for toxicology and safety reasons in the latter period. The formation of data-sharing consortia between big pharma companies and academia like eTOX, a program of the EU’s Innovative Medicines Initiative (IMI), has already helped, and is expected to help drive this improvement.

Among the representatives from big pharma sharing their approaches to predictive ADME/Tox at the meeting was Jonathan Moggs, Ph.D., head of molecular toxicology, translational sciences, at the Novartis Institutes for BioMedical Research. He discussed the growing importance of epigenomics in the assessment of drug safety.

“Most drugs have an epigenetic effect, and it is important to assess this in terms of toxicity,” he said. Novartis is looking at epigenetic mechanisms and biomarkers for drug-induced toxicity, particularly in the context of chronic administration. The value is that such epigenetic effects may be the earliest effects in nongenotoxic carcinogenesis.

It is easy to assess DNA methylation, one of the two main epigenetic changes, and researchers at Novartis have already determined a number of tissue-specific DNA methylomes. It is more challenging to determine histone acetylation, the other main epigenetic change, but Dr. Moggs believes the technology will become available in the next year or so.

Novartis is currently integrating its epigenetic profiling with transcript (mRNA and miRNA) profiling in a mouse liver tumor model—an approach that Dr. Moggs said poses some bioinformatic challenges but promises to lead to an understanding of the underlying mechanisms of drug-induced epigenetic perturbations. He concluded that future challenges in epigenomics in drug toxicity include trying to understand interindividual and inter-species differences, dynamics of healthy versus disease epigenomics, and the specificity, sensitivity, and validation of epigenetic biomarkers.

Dr. Moggs also talked about the work of the IMI’s newly launched MARCAR (Biomarkers and Molecular Tumor Classification for Nongenotoxic Carcinogens) consortium that brings together Novartis, UCB Pharma, Bayer Schering Pharma, Lundbeck, Boehringer Ingelheim, and CXR Biosciences, as well as a number of academic partners. The aims of MARCAR are improved drug safety, more efficient drug development, and advancement of alternative preclinical testing methods.

CXR Biosciences will perform miRNA analysis of samples to help identify possible candidate biomarkers that could be used as indicators of nongenotoxic carcinogenesis. With Taconic Artemis, CXR is making its transADMET humanized mouse models available to MARCAR partners.

Tom Shepherd, Ph.D., CEO of CXR Biosciences, explained that tumor findings are common endpoints in preclinical testing, but such in vivo carcinogenesis is rarely genotoxic in origin because directly genotoxic compounds would be excluded at an earlier stage. At present, there are no sufficiently accurate and well-validated short-term assays for identifying the nongenotoxic carcinogens, thus necessitating an expensive two-year rodent bioassay for assessing the carcinogenic risk of such compounds to human.

The MARCAR project also seeks to provide new mechanistic insights and early biomarkers that should enhance the design of more predictive tests for nongenotoxic carcinogenesis.

Genotoxicity

Genotoxicity testing also needs to be improved. Kathleen Böhme, Ph.D., biochemist at Merck Serono, described her Ph.D. work on an in vitro system for the toxicological evaluation of genotoxic compounds. Assessing genotoxicity is an important aspect of both drug development and chemical safety testing (particularly under the REACH program). However, the current in vitro test battery is limited by low specificity and a high rate of false positives, while the two-year rodent assay is expensive, time-consuming, and also beset by false positives.

“There is a need for new validated tools with higher specificity and reduced cost,” Dr. Böhme said. Accordingly, she has been looking at omics and mechanism-based approaches to in vitro genotoxicity testing using HepG2 cells. “We are aiming to identify the gene-expression signature of compounds being tested.”

Using etoposide, actinomycin D, and MMS as test compounds, she found 66 genes that are possible markers of genotoxicity, with a range of functions including apoptosis, DNA damage, transporter activity, immune response, and development. In parallel, Dr. Böhme has used an assay for p53 activation (with Active Motif’s TransAM™) in which the test compounds showed significant p53 activation.

So far, she has concluded that most direct genotoxicants do show characteristic gene-expression signatures. The challenge now is that most compounds of concern are actually promutagens, and it needs to be established whether the HepG2 cells were capable of metabolizing them to reveal their genotoxic properties.

In fact, most CYPs are expressed only weakly in HepG2 cells compared to human hepatocytes; however, both CYP1A and CYP3A4 are inducible. Therefore, a metabolic-activating system may need to be added to the HepG2 cells to reveal some genotoxicants.

Although p53 activation may be a surrogate marker, gene-expression profiling with a range of genes is actually more promising for genotoxicity screening. Going forward, Dr. Böhme will test more compounds, in different classes, with the aim of building a predictive classification model for genotoxicity that will lead to better risk assessment. Currently, these approaches are at the research stage at Merck Serono, except for the CYP gene induction assay, which is used routinely.

Willem Schoonen, Ph.D., senior research scientist at MSD, part of the U.S. pharmaceutical company Merck, described his company’s approach to in vitro genotoxicity, carcinogenicity, and nongenotoxic carcinogenicity screening as deselection and/or ranking tools in lead optimization.

“Multiple measures of toxicity are needed to decrease attrition in preclinical research,” he said. These screens include in silico (DEREK, TopCat, MultiCASE, Mutalert), in vitro (mutagenicity, clastogenicity, nongenotoxicity carcinogenicity, cytotoxicity, nuclear receptor activation, Phase I and Phase II enzymes) and in vivo (biosafety testing, liver toxicogenomics, teratogenicity) testing.

Dr. Schoonen expanded on development work he has carried out on VitoTox, RadarScreen, and HepG2 genotoxicity assays. VitoTox is based on prediction of DNA damage in S. typhimurium with a luciferase readout, while RadarScreen measures DNA damage to yeast through the promoter of the RAD54 (recombinatorial repair) gene with a beta-galactosidase readout. The HepG2 assays are all luciferase-linked promoter assays with the advantage that HepG2 is a human cell line and can metabolize genotoxic compounds.

Dr. Schoonen described toxicogenomic data obtained with these systems, with a range of compounds (genotoxic, cholestatic, necrotic, and apoptotic reference compounds). A selected set of genes was upregulated in these assays including cystatin A, a possible tumor progression marker; XPC; RAD51C (involved in repair of double-stranded breaks); MDM2, which keeps tumor suppressor p53 inactive in the absence of DNA damage; CKDNIA (p21); and TP5313.

He also explained that VitoTox and RadarScreen have good predictive value. However, the four HepG2 assays had lower predictive value than VitoTox and RadarScreen when compared with in vitro Ames mutagenicity data and in vitro clastogenicity data obtained with CHO or lung cells. But, when compared with in vivo rat clastogenicity data the correlation was much more predictive, which was most likely due to overprediction of clastogenicity scores in the hamster cells lines, which contained a mutated p53 gene.

Pattern recognition is being used as an alternative tool to identify new correlations between these datasets. These comparisons show that HepG2 cells may become better predictors for human in vivo genotoxicity than the currently used gold standards—Ames and in vitro clastogenicity testing.

Only the potent genotoxicants gave true positives in all of the tests investigated, Dr. Schoonen said, while true negatives were easily identified and hierarchical clustering segregated compounds into different classes.

Genotoxicity is not the only reason why a compound can fail on patient safety. James Dykens, Ph.D., drug safety R&D at Pfizer, noted that there are more than 2.2 million adverse drug reactions (ADRs) each year in hospitalized patients in the U.S., more than 350,000 in nursing home residents, and the number of ADRs among ambulatory patients is unknown. These figures translate into at least 106,000 deaths per year, making ADRs the fourth leading cause of death. “Clearly the pharmaceutical industry has been missing some important aspects of drug toxicity,” he said.

Drug-induced mitochondrial toxicity is fast emerging as a new model for these idiosyncratic ADRs. Until recently, mitochondrial toxicity has been missed simply because the tools to detect it were not available.

"Mitochondria are complex organelles that can fail in many ways,” Dr. Dykens commented. “It should not be surprising that xenobiotics produced by pharma can inhibit electron transport.” Among the 44 drugs withdrawn from the market since 1960 and those that have received black-box warnings, there is evidence that several cause mitochondrial impairment, which may be indicated by elevated liver enzymes and lactic acidosis.

Many drugs with organ toxicity will have some kind of mitochondrial liability. Whether a drug’s mitochondrial toxicity will actually have deleterious consequences depends upon its potency.

Mitochondrial ADRs also tend to be idiosyncratic, depending upon the organ’s history and patient genetics. One problem is that young, healthy animals, which are less prone to mitochondrial toxicity, are used in preclinical screens. “We are looking at the wrong models for in vivo mitochondrial studies,” Dr. Dykens said.

Pfizer is, therefore, looking at some new assays for mitochondrial toxicity. One is based on measuring mitochondrial respiration in 96-well plates with the Luxcel oxygen probe, which allows detection of compounds that uncouple oxidative phosphorylation. This revealed how all the thiozolidinediones (some of which have black-box warnings for chronic heart failure) and some statins have this effect.

Another screen is the Seahorse Bioscience metabolic profiling technology that Pfizer has used to detect lactic acidosis with the biguanides. MitoSciences’ assay assesses the effects of compounds on isolated mitochondrial electron transport protein complexes I, IV, and V, which can identify their site of action. This shows that troglitazone interacts with just complex IV, while simvastatin inhibits all three.

Dr. Dykens noted that cell culture is usually carried out at high glucose concentration, which inhibits respiration (the so-called Crabtree effect) and does not, therefore, detect mitochondrial inhibition. “Instead, at Pfizer we have grown cells on galactose to show mitochondrial inhibition. Our goal is to do all these screens much earlier.”

Predictive Methods

Dongzhou Liu, Ph.D., medical affairs and clinical development, new products R&D at GlaxoSmithKline (GSK), discussed advances in predictive methods and applications in ADME and biological properties profiling. His main theme was earlier and iterative in vivo modeling to ease the translation from in vitro to in vivo in an attempt to address the fact that over 90% of drugs fail after first-in-man studies. Earlier assessment of drug absorption, transport, and bioavailability is vital, Dr. Liu concluded.

Finally, Vikash Sinha, M.D., clinical pharmacology leader, global clinical pharmacology and pharmacokinetics at J&J Pharmaceutical Research & Development, discussed the application of physiologically based pharmacokinetic modeling (PBPK) in drug development. PK properties of a molecule cover its distribution, clearance, and absorption, and influence how much drug should be administered and how often.

PBPK looks at organs as compartments with physiological, anatomical, biochemical, and physicochemical properties. “Although PBPK is more complex than empirical approaches, it gives mechanistic insights into the compound and a fuller profile,” claimed Dr. Sinha. PBPK also predicts pediatric PK and drug-drug interactions.

At J&J, PBPK is being put to the test with new software including GastroPlus™ and SimCYP, which have a number of predictive features. GastroPlus allows in silico predictions from molecular descriptors, while SimCYP can predict drug-drug interactions and also PK in special patient populations, such as those with renal failure or diabetes.

In one J&J case study, the physicochemical properties of a small, poorly soluble molecule were known and prediction of its human PK properties made—as the dose increased its bioavailability decreased. In a second case study, there was a big mismatch between predicted and actual plasma profile over time, the reason was determined to be extrahepatic metabolism in the lung. This compound was found to be a PGP substrate, something that would not have otherwise been revealed, Dr. Sinha said.

PBPK also addresses questions about the impact of food on the absorption process, whether reducing the particle size of a compound will improve its absorption, or whether the compound is a potential candidate for a controlled-release formulation. PBPK can also improve the design of clinical trials and was used to determine the dose selection of a third J&J compound, Dr. Sinha said. “Industry is still in the learning phase with PBPK, but there is more and more support from management for these approaches as part of our growth.”