Changing the paradigm: Expanding High Content Imaging for early cytotoxicity assessments

There is no single solution to achieving these goals; however there are underutilised tools that provide drug discovery efforts with richer data sets for more prudent/intelligent decision making. Tools that have already emerged into the forefront to await impact upon the process include the use of automated patch clamp technologies that identify liabilities associated with QT interval prolongation; various gene arrays opening the field of toxicogenomics and cell-based surrogate indices of cytotoxicity. One such tool that is making its way into the forefront is high content imaging (HCI). The basic components of HCI are: the integration of image acquisition of fields of cells to detect changes in cellular physiology and the process integration of cell-based assays using fluorescent probes and tags that can be quantified on automated imaging equipment with or without preset, pre-optimised algorithms. One current suggestion is to maximise the use of HCI in early drug discovery. HCI can provide cellular in vitro toxicity information alongside traditional toxicity assays, combined with the in silico tools of the chemist to achieve a more complete view of the chemistry.

High Content/High Throughput Imaging is a reality that can be practiced in every laboratory both large and small. In 2005 there is growing evidence from all aspects of the drug discovery process that the automation of in vitro cellular imaging can query a myriad of biological processes. This has all been made possible with the use of fluorophores, antibodies and biosensors that offer the potential to deliver compound profiles not unlike those seen in mass spectroscopy libraries. The path ahead is quite straightforward – as was the deciphering of the human genome into its various genes. The next quest is the understanding of the functional consequences of the activation, movement and regulation of proteins that were translated from those genes. These pursuits are fully in the realm of cellular biology and this quest can occur from multiple points of origin. The emphasis of the academic community has been to further knowledge, understand disease functions and offer strategies and experimentation to unravel cellular networks, interactions and pathways. Alternatively, the pharmaceutical corporations and biotechnology companies have pursued the chemical tools (the compounds) to approach the increasing demand for better drugs that alleviate or modify disease processes¹.

From either direction, these approaches merge the tools and the cells together and HCI offers a ready platform to this systems biology-like arena, be it compound centric or disease centric or protein centric. HCI, as deployed at Roche, is a multifaceted tool that enables a clearer visualisation of the diverse cellular effects of compounds under a broad umbrella of bioapplications. Firstly, the output image data is richly visual and particularly well suited to interpretation by the cell biologist – just as the tissue section of a diseased heart is richly visual to interpretation by the pathologist. But the reality is that one can only manually evaluate so many images per day whilst maintaining accuracy and effectiveness. The output of numerical data from HCI which is typically multi-parameter, offers extensive quantification of detailed cellular measurements that underlie the phenotype observed. These large data sets can be parsed using automated tools to highlight the particular correlating parameters that match a particular cellular function. An example would be the identification of changes in nuclear condensation as quantified by the heterogeneity in Hoechst or Dapi staining during the late phases of apoptosis. The automation of both the image acquisition and the image processing has now permitted us an unprecedented power to query compound effects on cellular functions and cellular phenotypes. This translates into fast, simple, robust answers to support the advancement of chemistry into pre-clinical development. With that in mind, one of the broad spanning purposes for HCI, not yet fully realised for compound centric pursuits, is usually referred to as in vitro toxicity evaluations of compounds. The detailed evaluation of tens to thousands of compounds in dose response curves querying discrete biological functions is a robust approach to accumulating rich compound-centric data sets. Analogously, one may not have imagined prepackaged molecular biology assays to improve so rapidly as to now see kits that can provide metrics for all sorts of transcription/translation events. Each of the automated HCI platforms offer packages that consist of any number of cell-based assays that can be performed on the cells of choice within certain physical/technical constraints (typically optical limitations and confocal or non-confocal confines).

These include:

• Mitotic Index, identifying phosphorylated histone proteins associated with mitotic spindle formation
• Cell Viability, a single cell based analysis of identifying live and dead cells
• Cell Cycle, an analysis that tracks the individual cells in the cycle phases which can also detect specific protein expression
• Micronucleus Detection, which includes identification of micronuclei and numbers of mono and binucleate cells
• Apoptosis, a quantification of actin cytoskeleton integrity, alterations of mitochondrial mass, potential nuclear fragmentation and changes in membrane permeability
• Neurite Outgrowth, a quantification of number and lengths of axis extensions

Our goal is to successfully integrate HCI/HTI as an evaluation tool to differentiate individual components of cytotoxicity into pathways and molecular events that can be addressed singly by Medicinal Chemisty for appropriate compound modifications. Highlighting potential toxicity issues at an early step in the discovery process is by no means a premature cull or ‘chemisty kill’ based solely on efficacy/potency2. This single parameter decision cannot be enacted without the superimposition of additional query tools. In actuality it is a paradigm shift to evaluate the majority of the compounds to separate from a longer list of the potential actives, those experimentally derived compounds with ADME/Tox liabilities. Likewise for later stage chemistry committed projects, HTI can provide a comprehensive SAR evaluation versus extrapolations typically carried out from a minority of samples. In this manner, these additional cellular parameters that translate into considered liabilities serve to segregate the SAR beyond the target efficacy and target function. The importance of highlighting these issues with Flags/Priorization/Rank order of sensitivity/Predictive Tox Profiling can then serve to encourage synthesis of more ‘acceptable’ ADME/Tox candidates. Quantification of cellular homeostasis following compound treatment is a general way to priorise chemical series and trouble-shoot toxicity issues. An advantage of this approach is that it is automated and expandable to hundreds of compound evaluations. Simultaneous measurements of multiple indicators of cell health have previously been shown to be highly valuable for indicating early potential toxicities in numerous literature references3.Indicators of cell health that are measured include effects on the stage of the cell cycle; internucleosomal DNA fragmentation; plasma membrane permeability changes; membrane blebbing; effects on the mitochondrial transmembrane potential and perturbations of microtubules (the polymerisation or depolymerisation state). Other markers discussed as cellular toxicity indicators include quantification of phospholipidosis and cellular oxidative metabolism which also give indications of changes of cellular homeostasis. Abnormal morphology is a main indicator of cytotoxicity and can be phenotypically exhibited by the development of giant cells, multi-nucleated cells, granular or perturbed cellular surfaces, vacuolarisation within the cytoplasm or in the nucleus and/or ragged cell edges. Each aspect can be identified and quantified on the HTI platforms as an additional numerical feature added to the multiparameter data output. Profiling of cells to establish changes due to compound exposure is occurring in the early stages of the discovery process where the efforts to screen are relatively rapid and inexpensive to obtain4. This scenario, combined with emerging technologies such as toxicogenomics, will reduce failure rates by helping select the right compound for development early on, as well as on a continual process by accelerating toxicity testing5. Using HCI the combined affect of side-by-side gene expression profiling, chemistry SAR data, in silico tools and known biological databases, makes the assessment of compound liabilities feasible. One of the limitations to such an approach is the question regarding which cellular background is the right choice for selection to accumulate correlative toxicity data. One might propose that for the very early hits received from HTS screening the typical HTS cells used in primary screening may suffice. Each of these choices contains some caveats. One caveat is that some cells are of non-human origin such as the monkey kidney SV40 transformed cell line, COS-7 and the hamster ovary lineage displaying epithelial morphology, CHO. Others are of human origin: adenovirus transformed human kidney cells which display epithelial morphology, HEK293, as well as another commonly used human cell line, HeLa, derived from an adenocarcinoma. Sceptics may consider the value of toxicological assessment in these cells as less than optimum and at too early a stage to be useful, yet the use of well established and well defined cell lines for toxicity profiling allows for direct comparison between efficacy models and screening of cellular cytotoxicity functions. The window between efficacy of cellular function and cellular toxicity can be thus described. Alternatively, a more comprehensive approach to profiling for in vitro toxicity using HCI can be addressed by expanding screens to include many of the relevant cell lineages now available. These are the many conditionally immortalised cell lines and the unique differentiated cells derived from human embryonic stem cells. These include specific cell types that are used for models in many areas of research: neural cells for spinal cord injury and Parkinson’s disease; cardiomyocytes for heart disease; pancreatic islet ß cells for diabetes; chondrocytes for osteoarthritis and hematopoietic cells for blood diseases and in immunology research. Expanded profiling can be considered as a filtering device or on an as-needed basis for the compound centric characteristics matched to the cell type of interest as the process proceeds towards development. An implementation of this would be to utilise one or more conditionally immortalised human cell lines. These include hepatocytes, microglia, keratinocytes, proximal tubule epithelial cell lines from normal adult human kidney, or human embryonic cells that develop into multiple types of cardiac myocytes. Numerous commercial sites offer these characterised cell lines that are capable of propagation with archived data available for comparison to primary tissue. Vendor sources include, but are not limited to, Cambrex, Cytomyx, Xenotech, ATCC and Multicell Technologies. In vitro toxicity assays can also be applied to primary cells from animals and humans6. To add to this, HCI provides the ability to assess the results by subpopulations of cells. Evaluating the window between toxicity and efficacy of an oncology candidate target over a range of specific cancer lineages will also offer what may be considered as custom multiplexing. Typically, the differences in the responses of compounds to cell lines versus responses to normal cell lines can be dramatic. HTI can quantify these differences, be they dramatic or less obvious. Since primary cells may be available from animal models or human derived samples, they can be used to query identical morphometric and cytotoxic liabilities due to the effects of compounds. In this manner a database may be developed to obtain information as close to physiologically relevant as possible. The early identification of compounds that impinge upon the status of normal cell health is expected to have a significant impact on the outcome of identifying safe compounds as potential lead candidates. This will hold true whether the compound exhibits maximum target potency or maximum selectivity. The benefit from these combined cell-based technologies as complementary models to assess toxicology liabilities at different phases of the drug discovery pipeline are an effort to improve the accurate identification of safer and more efficacious drug candidates that are eventually forwarded into development.

References

1. Boguslavsky. J. Minimizing Risk in “Hits to Leads”. Drug Discovery & Development, 2001 Vol.4(7): 26-30.
2. Koppal, T. Advancing In Vito ADME/Tox. Drug Discovery & Development, 2004 Vol.5(4): 47-50.
3. Haskins JR, Rowse P, Rahbari R, de la Iglesia FA. Thiazolidinedione toxicity to isolated hepatocytes revealed by coherent multiprobe fluorescence microscopy and correlated with multiparameter flow cytometry of peripheral leukocytes. Arch Toxicol 2001,75(7):425-38.
4. Gross, CJ., Kramer, JA. The role of investigative molecular toxicology in early stage drug development. Expert Opin. Drug Saf. 2003 2(2): 147-159.
5. Suter, L., Babiss, LE., Wheeldon, EB. Toxicogenomics in predictive toxicology in drug development. Chem. Biol. 2004 Feb 11(2): 161-71.
6. Jessen, BA., Mullins, JS., de Peyster, A., Stevens, GJ. Assessment of Hepatocytes and Liver Slices as in Vitro Test Systems to Predict in Vivo Gene Expression. Toxicological Sciences, 2003, 75, 208-222.