April 1, 2015 (Vol. 35, No. 7)
If Only Predicting a New Drug’s Toxicity Was As Easy As Peering into a Crystal Ball
Therapeutics fail during the discovery and development cycles for many reasons, such as bioavailability, efficacy, or toxicity. Predicting low instances of toxicity, which may only emerge late in development, in clinical trials, or even after the compound is on the market, is very challenging.
The need to identify low instances of toxicity earlier, exemplified by examples such as the CCR5 inhibitor RO5657 and the CD28-specific mAb TGN1412, drives advancements in predictive toxicity assays. Thought leaders meet regularly to discuss state-of-the-art research. For example, they recently convened at the International Summit on Toxicology and Applied Pharmacology, which took place in Chicago.
One of the issues taken up at this event was the application of induced pluripotent stem cells (iPSCs). This approach was considered by Roche, in collaboration with Cellular Dynamics International (CDI), after the pharma giant’s CCR5 inhibitor RO5657 was dropped. Although RO5657 progressed through hERG and preclinical cardiotoxicity assessment, it was ultimately dropped due to severe cardiotoxicity in primates.
“iCell Cardiomyocytes have utility at many points in the drug discovery process. A group at the Hospital for Sick Children (SickKids) in Toronto previously identified a new target, integrin-linked kinase, involved in dilated cardiomyopathy and was able to use the cells to validate this target,” commented Carter Cliff, business development, CDI.
“In addition, scientists in Roche’s cardiovascular program cultured the iCell Cardiomyocytes under diabetic conditions and were able to demonstrate a diabetic cardiomyopathy phenotype similar to that seen by cardiomyocytes made from a diabetic genetic background,” Cliff continued. “In a small molecule screen, this model produced novel and chemically diverse leads worthy of additional development.”
Human iPSC-derived cells provide a more relevant human model that may enable better research, reduce drug attrition, and spur cell therapy development. Fundamentally, the question is not whether to use iPSCs or primary cells but about the shift from target-based screening in heterologous cell lines to phenotypic readouts in contextually relevant human models.
iPSCs can be derived from any person in any quantity needed, providing control over the background genotype. Furthermore, new gene-editing tools make it possible to produce PSC models with specific defects, or isogenic models from individuals with known genetic defects. Multiple efforts are underway to develop better cell culture conditions for support of both primary and PSC models.
A cytokine storm, or an instance of cytokine-release syndrome, most commonly occurs as part of the immune response to some viral infections but can also be frequently associated with monoclonal antibodies (mAbs). Some mAbs carry black box warnings about cytokine-release syndrome, which can affect a significant portion of patients to varying degrees.
The mAb-mediated cytokine-release syndrome may be either a pharmacological effect, or a secondary effect, such as immune stimulation mediated via Fc receptors. Other types of therapeutics, such as siRNA, also have the potential to stimulate immune function by interacting with toll-like receptors.
Given the exquisite specificity of mAbs and differences between human and animal immune systems, animal models may fail to adequately predict cytokine-storm potential. A potential treatment for B-cell chronic lymphocytic leukemia, the CD28-specific mAb TGN1412, was intended to activate regulatory T cells and dampen immune response, but a cytokine storm resulted in systemic organ failure in initial human volunteers.
Studies in rodents and nonhuman primates showed no ill effects. Researchers later discovered that differences in CD28 expression on T-cell subsets between cynomolgus monkeys and humans might explain the vastly different outcomes in the two species.
According to Travis Harrison, Ph.D., director, immunology services, SRI Biosciences, in vitro assays typically use peripheral mononuclear cells (PBMCs) or whole blood, not monocyte-derived dendritic cells (DCs). Obtaining blood from a large number of diverse donors spanning the most frequent HLA types and evaluating multiple cell types would make in vitro evaluation more robust.
Working with compounds that may interact with toll-like receptors, the Harrison laboratory anticipated a greater response in DCs than in PBMCs but saw the opposite. Differences between myeloid and plasmacytoid DC subsets may be the reason, and underscores the need for multiple cell type evaluation.
Analyzing Directly in Tissue
Inoviem Scientific’s nematic protein organization technique (NPOT) is based on the Kirkwood-Buff molecular crowding and aggregation theory. NPOT enables the formation and label-free identification of macromolecular complexes involved in physiological or pathological processes, and is particularly effective for identifying the on- and off-target actions of therapeutic molecules.
Protein-protein interactions can be analyzed directly in human tissue, from complex mixtures, without disrupting the native molecular conformation.
Homogenates are prepared from human normal or pathologic tissues in the absence of any detergent, reducing agent, or protease or phosphatase inhibitors. Dilutions and washes are performed in a buffer with equal osmolality, trace elements, vitamins, and salts, in concentrations as close as possible to those of the interstitial medium or cytoplasm.
The label-free ligand of interest, which is unmodified chemically or molecularly, is put in contact with the total tissue material. Then the macromolecular assemblies related to the ligand are separated using a proprietary differential microdialysis system, wherein the macromolecules migrate in the liquid phase based on their physiochemical properties.
In the presence of the drug, the migrating macromolecules gradually grow from nematic crystals to macromolecular heteroassemblies due to the molecular interactions between the drug and its targets. The latter are isolated and identified by mass spectroscopy directly in liquid.
“NPOT has improved our understanding of the biological effects of pharmacologically active molecules at the cellular and molecular levels,” explained Pierre Eftekhari, Ph.D., director, Inoviem Scientific. “In general, it has increased our knowledge about drug-target interactions, and it can predict off-site interactions in regard to the corresponding environment.
“Today’s pharmacological and toxicological methods are obsolete. A pharmacophore pocket or the toxic effect of a pharmacologically active molecule can only be real when the target is studied in its native environment and not in isolated form. NPOT is the beginning of a changing paradigm.”
Emerging Yeast Models
Much human variability to susceptibility to xenobiotics, such as pharmaceutical drugs and genotoxicants, is conferred by polymorphisms in P450 genes and in housekeeping genes involved in basic DNA repair and metabolism. But due to expression of multiple P450 genes, it is difficult to discern which variant gene confers susceptibility.
“Many of the metabolic pathways in yeast are essentially conserved in all eukaryotes, including man,” noted Michael Fasullo, Ph.D., associate professor of nanobioscience, Colleges of Nanoscale Science and Engineering, SUNY Polytechnic Institute. “In comparing yeast with other organisms for toxicology testing, the question needs to be carefully framed.
“If the question is whether a toxin causes DNA damage, then yeast is an excellent choice since multiple genotoxic endpoints can be measured in a high-throughput fashion. However, if the question focuses on whether the toxin can cause organ damage, then mice or another experimental animal is more appropriate.”
As a model xenobiotic, the potent liver carcinogen aflatoxin B1 (AFB1), which is metabolically activated by CYP1A2 and CYP3A4, was studied. Since budding yeast do not express P450 genes that can metabolically activate AFB1, CYP1A2 variants were phenotyped, and the yeast genome profiled for resistance to AFB1. Approximately 500 genes that confer AFB1 resistance were identified, and these genes were grouped according to ontology groups.
An advantage of the yeast system is that individual human P450 genes can be introduced to mimic the metabolic state of different tissues. For example, CYP1A2 and CYP3A4 mimic expression in the liver, while CYP1A1 is an extrahepatic P450 that is expressed in the lung.
“We are also investigating whether specific polymorphic P450 genes associated with specific cancers confer higher levels of genotoxicity to carcinogens when expressed in yeast,” reported Dr. Fasullo. “The toxicity of thousands of chemicals is unknown. The yeast strains we are developing should rapidly identify those that are clearly DNA-damaging agents in high-throughput toxicology studies.”
Advances in High-Content Analysis
Concerns have been raised about nanotechnology and its impact on the environment and human health. To address these concerns, researchers have turned to high-content analysis (HCA), which has already been successful in predicting human toxicity of pharmaceutical small molecules as well as carrier molecules that aid drug delivery. The researchers, in an article recently published in PLOS ONE, validated the use of HCA in identifying the toxic potential and mechanisms of nanoparticles.
In addition to collaborations with many groups seeking to apply HCA in new areas, the laboratory of Peter O’Brien, D.V.M., Ph.D., lecturer, School of Veterinarian Medicine, University College Dublin, has investigated equine large-particle pulmonary toxicity and mechanisms. Inflammatory airway disease (IAD) affects performance and well-being of horses. Diagnosis is primarily reached by bronchoalveolar lavage (BAL) cytology, which is invasive and requires sedation.
Using equine gene expression microarrays to investigate global mRNA expression in circulating leukocytes from healthy, IAD-affected, and low-performing Standardbred and endurance horses, circulating blood-cell gene expression was shown to reflect inflammatory responses in tissues. Although not specific for IAD, whole blood glutathione peroxidase activity appears to be correlated with BAL neutrophil percentage.
Another study evaluated live blood cells for the automated diagnosis and prognosis of lymphoma in a canine model. Numerous cellular measurements, such as nuclei size, shape, and placement, were taken and the cells immunophenotyped. The model demonstrated the ability to distinguish neoplastic cells and malignancy severity using morphometric and fluorescent analysis. Additional validation and relevance to humans remains to be determined.
“The largest application for HCA is in the field of toxicology, and we have been adapting it for clinical application, stated Dr. O’Brien. “Today, discovery scientists are the main drivers. Sometimes even toxic drugs must be moved forward if they are the best solution for treating a disease. With HCA you can follow that toxicity experimentally.”