HCA of Complex Cell Models for Single Agent and Combination Studies of Drug Efficacy

Abstract

Lack of efficacy is a significant cause of late-stage failure in drug discovery, hence, the importance placed on strategies to better predict efficacy earlier. Critical to this objective is the use of pathophysiologically relevant models of disease in preclinical drug discovery and, in particular, highcontent analysis (HCA) as a key enabler to interrogation of complex models of biology. This paper summarizes the current state of HCA, its application to human primary co-culture models of biological processes and the suite of technologies available to drive continued integration. Examples discussed demonstrate that complex biological systems are translatable to higher throughput techniques allowing larger compound sets to be profiled faster, more consistently and in combination. HCA is a significant tool to enable compound profiling in sophisticated models of disease that may, with continued and wider adoption, result in improved efficacy prediction in the early stages of drug discovery.

Introduction

Late-stage attrition of candidate drugs in clinical trials as a result of poor efficacy is a significant factor in the decline of R&D productivity in recent years [1]. In the past five years there have been 50% fewer new molecular entities approved by the major drug administration regulatory bodies around the world compared with the previous five-year period despite higher levels of financial investment and increasing knowledge of disease [2]. In the post-genome era, scientists have access to large volumes of data to support research into the molecular basis of disease. This, coupled with high-throughput, targetdirected drug discovery, has delivered highly selective compounds aimed at previously intractable molecules, however, a leading cause of attrition in Phase IIb clinical trials remains lack of efficacy. Preclinical drug discovery is increasingly focused on identifying ways to better predict drug efficacy for medicines targeting complex diseases characterized by multiple causalities. Alongside strategies such as routine use of preclinical efficacy biomarkers, improvements to animal models, and use of early proof of concept clinical trials, industry is adopting methods to deliver biology-led programs of early drug discovery in parallel with target-directed, high-throughput strategies [2-5]. The concept of drug discovery based on a holistic understanding of complex disease becomes even more relevant in the context of multi-targeted therapeutics. Genetic studies have shown that in mammalian models inactivation of a single gene is often not enough to lead to a discernable phenotype [6,7]. It is therefore not surprising that increasing numbers of diseases are treated clinically using drug combinations to overcome resistance or pathway redundancy and a significant proportion of newly approved drugs are demonstrated to act on more than one target [8]. In order to rationally design such multi-targeted approaches within early drug discovery more complex modelling of disease is required to better recapitulate network biology in vitro.

Technologies are needed which allow scientists to gain a better understanding of complex pathophysiology and compound mode of action early in discovery. Available approaches include use of next-generation sequencing, proteomics, RNAi, systems biology and cellular- and tissue-based imaging. These technologies enable increased understanding of disease at a molecular level and assessment of a phenotypic outcome. One approach, high-content analysis of cell- and tissuebased assays, allows quantitative functional endpoint data to be obtained for a number of features in parallel e.g., reports of receptor activation, protein localization and apoptosis. This approach allows simultaneous assessment of compound mode of action on both known targetrelated and other phenotypic effects.

High-Content Analysis Tools & Application

Following the introduction of automated fluorescent microscopy platforms in the late 1990’s HCA has evolved rapidly; particularly in the last five years. Current platforms include wide-field and confocal, and offer efficient, unbiased methods which image and quantify cellular phenotype with greater data quality and reduced time and cost compared to previous manual efforts. The majority of automated systems are amenable to integration with platebased lab automation equipment. Also, the introduction of imaging cytometers capable of acquiring and analyzing data at low density formats on-the-fly mean that target-based, high-content assays are being applied at the lead identification stage as well as at lead optimization [9,10]. Furthermore, several examples of phenotypic screening across cell lines have been published in which broad and quantitative physiological measurements are used to profile compounds in a targetagnostic manner [11-14]. High-content data is often rich in content, large in size and diverse in format, so centralized HCA laboratories are often multidisciplinary in skill set with teams delivering biology, automated imaging and expertise in creating bespoke image and statistical analysis, data processing, storage and management solutions [15-17]. Vendors are also now encompassing complete solutions for HCA in addition to automated imaging microscopes. High-content technology has matured considerably and instruments are both powerful and versatile. Having proven application to cell-based screening, focus is shifting to novel disease areas and complex modes of disease.

HCA in Complex Pathophysiologically Relevant Systems

The design and quality of a high-content screen is critical; if designed assays cover limited biology, screening may not detect unanticipated effects, leading to protracted primary and secondary screening. Parallel profiling of compounds through assay panels will undoubtedly bring benefit, however, additional value may be gained in designing assays covering broader biology. Models that more closely recapitulate the multicellular in vivo environment can reduce the requirement for large amounts of in vivo preclinical screening, enable early multitargeted therapeutic discovery and deliver more accurate translation to the clinic. For example, cell lines grown in 2D monocultures have been shown to exhibit differences in gene expression when compared with those in a physiological environment [18] and differences have been observed between clinical and standard 2D in vitro conditions [19] suggesting the need for more complex modelling of disease.

HCA has been of particular benefit in the study of angiogenesis, a multistage process targeted for intervention in a number of disease states including cancer, psoriasis and macular degeneration [20]. A range of preclinical screens including measured effects of proliferation, migration and cord formation in endothelial cell lines are available to assess the efficacy of antiangiogenic compounds [21,22], however, with these approaches it is difficult to interpret the effect of compounds in the context of the multistep process and multicellular environment involved. A more complex organotypic model of angiogenesis is available with which it is possible to conduct both high-throughput [23-25] and high-content compound profiling. This model involves the coculture of human primary endothelial cells with either fibroblasts [26] or vascular smooth muscle cells [23] and enables the modelling of both autocrine pathways between endothelial cells and intercellular interactions with stromal cells.

Human primary endothelial and fibroblast cells grown in co-culture for 10 days were fixed and labelled to identify endothelial cells (CD31, green) and nuclei (Hoechst, blue). Cultures were read and analysed for compound effects on endothelial tubule area and cell cytotoxicity by highthroughput imaging cytometry.

Figure 1: HCA of a human primary co-culture model of angiogenesis

Design and optimization of primary cell coculture models for HCA requires consideration of standard decisions common to many cellbased HCA screens, including decisions as to which automated imaging and analysis platform is most appropriate. Assays can be optimized on platforms supporting speed over resolution such as imaging cytometers, which offer rapid, whole well, high-throughput analysis on-the-fly for lead identification (Figure 1). Alternatively, platforms dedicated to higher resolution imaging confer advantages in assessment of subtle changes in morphology for mode of action and phenotypic profiling studies. Off-the-shelf image analysis solutions are often robust enough for assay development and analysis of characterized events such as nuclear translocation in standard cell lines. However, customization is necessary for heterogeneous cell populations or for phenotypic screening where quantification of subtle changes in morphology is desired. In the case of co-culture screens, relatively simple customization of image segmentation enables effects specific to distinct cell types to be quantified separately, especially where well-characterized cell-specific markers are available, such as in the case of CD31 and endothelial cells (Figure 2A). Furthermore, measurement of tubule branchpoints, length, network complexity and effects on cell nuclei allow detailed assessment of compound mode of action particularly when compared with data from well-characterized compound training sets. A range of object-oriented and machinevision image analysis approaches are available for this purpose, indeed, machinevision based image analysis, where scientists train computer algorithms to recognize cells with distinct morphology may be the most appropriate solution where distinct cellular markers are not available. Customized image analysis also enables characterization of previously unmeasured morphologies such as sometimes the case with treatment of new chemical series.

(A) Image segmentation of high-resolution co-culture images for downstream phenotypic analysis. Images are of cells 10-days post-co-culture: phase-contrast (top left); immunolabelled to identify endothelial cells (CD31, red) and nuclei (Hoechst, green) (top right); image segmentation of tubule morphology metrics including area (blue), branchpoints (red) and length (black) (bottom left); image segmentation to quantify endothelial (blue) and fibroblast (pink) cell nuclei area and shape (bottom right). (B) Example interaction graph obtained from factorial experimental design of co-culture assay. The graph demonstrates the impact on endothelial tubule formation of fibroblast (HDF) cell source (black and red) and media supplemented with growth factors.

Figure 2: Example optimization requirements for HCA of complex organotypic culture

Further to standard considerations, there are a number of specific points to be addressed for HCA of co-culture models. First, whilst use of human primary cells confers greater physiological relevance, it is necessary to address concerns over cell supply, variability of source and in the case of co-culture, conditions appropriate for multiple cell types. Cell numbers large enough for screening are available from commercial suppliers for many cell types including primary fibroblast and endothelial cells. Batch testing coupled with matching of donor cells for population doubling times are protocols established to ensure robust screening in primary cells from week to week. Secondly, primary co-culture assays are complex, hence factorial experimental design (FED) is appropriate for rapid optimization of media conditions, cell source, cell number and treatment regimens (Figure 2B). FED confers several advantages over ‘one-factor at a time’ methods including speedier assay development and the ability to optimize interacting factors such as seeding densities of mixed cell populations. Finally, laboratory automation of complex models requiring long-term culture, multiple interventions and specialized plate formats may necessitate optimization beyond that needed for typical assays. Co-culture assays are easily adaptable to 96-well and higher microwell plate formats and do not call for specialized transwell or 3D systems. This means that the main considerations are sterility of longterm culture and operation alongside other assays. Decoupled screening processes, where plates are manually transferred between instruments dedicated to different phases of the assay cycle such as cell plating, immunocytochemistry and imaging supports parallel operation of screens in the context of long-term assays.

Requirements for downstream data quality control, processing, analysis and storage are similar across all phenotypicbased assays irrespective of complexity. Solutions to these requirements include: methods to associate measurements from disparate sources; drill-down to raw and segmented images for quality control; and statistical methods to process multivariate data. If complex assays are used for assessment of multi-targeted therapeutics, then specialized compound dispensing and analysis routines are required [27]. Integration of HCA with more complex models of disease, as described, maximizes the benefits of multicellular models in drug discovery by: enabling screening at highthroughput; more completely describing compound mode of action using multiple measures; enabling preclinical multitargeted drug discovery; and enabling use of efficacy biomarkers that can be translated forward to clinical samples. The approach described is equally amenable to other complex assays, such as in stem cell differentiation and the role of inflammatory cells in pathological disorders such as cancer and lung disease. Application more broadly shall enable systems-level research across disease areas.

Multidimensional HCA for Improved Biological Understanding

In addition to standard HCA of physiologically relevant models of disease, integrated supporting technologies are available that include those enabling improved spatial and temporal resolution. A suite of dedicated long-term, brightfield and fluorescence kinetic imaging instruments are available for lower throughput studies screened in standard plate formats [28]. Greater off-theshelf availability and understanding of fluorescent probes for such live-cell experiments has widened the biology screened to cover diverse application areas. Examples include monitoring of endothelial cell tubule formation using lentiviral-mediated fluorescent labelling technologies and measurement of apoptosis using cleavable substrates for Caspase that fluoresce on activation (Figure 3). In developing such assays, the following specific considerations need to be made: cell maintenance in longterm culture; probes and methods used to quantify biology without perturbation of cell physiology; image co-registration and object tracking between frames; and analysis of multidimensional data with the added time dimension. The enhanced temporal resolution that such instruments offer has utility in selection of the most appropriate timepoint for endpoint studies along with quantification of transient responses, detailed analysis of cell fate following treatment and insight into effects of dose scheduling regimens.

Kinetic monitoring of: (A) caspasemediated apoptosis in a human microvascular endothelial cell line (HMEC-1). Example phase contrast and fluorescent image of cells following treatment with Mycophenolic acid (10 μM) for 24 h. Caspase-positive cells counted by automated fluorescent object count. Graph shows time-dependent increase in Caspase activity after diverse compound treatment. n=3, error bars are standard deviation. (B) Endothelial tubule formation in a human primary coculture model of endothelial (HUVEC) and fibroblast cells. HUVECs labelled with GFP and images acquired every 3 h for 10 days. Images shown are at days 0, 2, 3, and 7 days post co-culture.

Kinetic Imaging Applications

Low-throughput confocal and multiphoton instruments offer high spatial resolution and tissue penetration to support detailed characterization of biology in 3D or ex vivo tissue [29]. In order to better interrogate molecular signalling events in cell systems, confocal instruments are also sometimes FLIM enabled. Fluorescence lifetime imaging microscopy (FLIM) is a technique used to measure bimolecular interactions via fluorescence resonance energy transfer (FRET) and changes in environment sensed by altered lifetimes of fluorophore biosensors. FLIM is independent of FRET donor concentration and is therefore a robust measure of FRET. The technique has been demonstrated to enable direct assessment of the temporal and spatial effects of pathway activation at a molecular level in EGFR signalling [30] and advances in technology now support the use of high-throughput FLIM to drive further and broader application [31,32]. In addition to the described technology improvements, kit-format consumables for 3D culture are also facilitating the speed and ease with which multidimensional 3D imaging can be adopted to derive additional value from developed biology [33,34]. Another instrument useful to analysis of complex 3D cell culture is that designed for dedicated fluorescent slide imaging. These instruments are typically enabled with automated low-resolution, tissue-finding algorithms, extended focus algorithms accounting for tissue thickness and robotic slide loaders. Such equipment most commonly provides unbiased quantification of tissue acquired during clinical and safety studies but is equally amenable to analysis of 3D culture that has been sectioned prior to imaging. A further significant benefit of fluorescent slide imaging is the potential to refine early disease models as a result of calibration against similar markers analyzed in clinical samples.

Summary

HCA is an established tool in drug discovery that uses target-directed and phenotypic approaches to drive decision making in target validation, lead generation and lead optimization. The examples discussed outline the suitability of HCA to deliver further value from pathophysiological models of disease and showcase the suite of technologies that will drive continued developments. An interesting growth area is the application of HCA to regenerative medicine, where rare events, population heterogeneity and phenotypic changes evident in stem cell differentiation can be effectively measured [35]. This has potential to drive therapeutic opportunities for regenerative medicine but also drive development of organotypic models of disease for increased sophistication of technology and drug discovery. HCA of complex models is part of an expanding toolkit to explore biology at a systems level and move beyond the ‘one target, one drug, one disease’ paradigm. Wider adoption of HCA and retrospective analysis will determine its value in candidate drug selection, biomarker development and prediction of efficacy in preclinical drug discovery.

Acknowledgements

The author would like to thank all colleagues at AstraZeneca and in particular the current and previous members of the Advanced Science and Technology Laboratory for all valuable discussions and support in conducting high-content assays in drug discovery.

References

1. Kola I, Landis J. Can the pharmaceutical industry reduce attrition rates? Nat Rev Drug Discov 2004;3(8):711-5.
2. Paul SM, Mytelka DS, Dunwiddie CT, Persinger CC, Munos BH, Lindborg SR, et al. How to improve R&D productivity: The pharmaceutical industry’s grand challenge. Nat Rev Drug Discov 2010;9(3):203-14.
3. Sams-Dodd F. Target-based drug discovery: Is something wrong? Drug Discov Today 2005;10(2):139-47.
4. Butcher EC. Can cell systems biology rescue drug discovery? Nat Rev Drug Discov 2005;4(6):461-7.
5. Feng Y, Mitchison TJ, Bender A, Young DW, Tallarico JA. Multi-parameter phenotypic profiling: Using cellular effects to characterize small-molecule compounds. Nat Rev Drug Discov 2009;8(7):567-78.
6. Stelling J, Sauer U, Szallasi Z, Doyle FJ,3rd, Doyle J. Robustness of cellular functions. Cell 2004;118(6):675-85.
7. Barbaric I, Miller G, Dear TN. Appearances can be deceiving: Phenotypes of knockout mice. Brief Funct Genomic Proteomic 2007;6(2):91-103.
8. Yildirim MA, Goh KI, Cusick ME, Barabasi AL, Vidal M. Drug-target network. Nat Biotechnol 2007;25(10):1119-26.
9. Zock JM. Applications of high content screening in life science research. Comb Chem High Throughput Screen 2009;12(9):870-6.
10. Bickle M. The beautiful cell: High-content screening in drug discovery. Anal Bioanal Chem 2010;398(1):219-26.
11. Perlman ZE, Slack MD, Feng Y, Mitchison TJ, Wu LF, Altschuler SJ. Multidimensional drug profiling by automated microscopy. Science 2004;306(5699):1194-8.
12. Tanaka M, Bateman R, Rauh D, Vaisberg E, Ramachandani S, Zhang C, et al. An unbiased cell morphology-based screen for new, biologically active small molecules. PLoS Biol 2005;3(5):e128.
13. Young DW, Bender A, Hoyt J, McWhinnie E, Chirn GW, Tao CY, et al. Integrating high-content screening and ligandtarget prediction to identify mechanism of action. Nat Chem Biol 2008;4(1):59-68.
14. Caie PD, Walls RE, Ingleston-Orme A, Daya S, Houslay T, Eagle R, et al. High-content phenotypic profiling of drug response signatures across distinct cancer cells. Mol Cancer Ther 2010;9(6):1913-26.
15. Swedlow JR, Goldberg IG, Eliceiri KW, OME Consortium. Bioimage informatics for experimental biology. Annu Rev Biophys 2009;38:327-46.
16. Niederlein A, Meyenhofer F, White D, Bickle M. Image analysis in high-content screening. Comb Chem High Throughput Screen 2009;12(9):899-907.
17. Ainscow E. Statistical techniques for handling high content screening data. European Pharmaceutical Review 2007; (5):
18. Pan C, Kumar C, Bohl S, Klingmueller U, Mann M. Comparative proteomic phenotyping of cell lines and primary cells to assess preservation of cell typespecific functions. Mol Cell Proteomics 2009;8(3):443-50.
19. Serrels A, Timpson P, Canel M, Schwarz JP, Carragher NO, Frame MC, et al. Real-time study of E-cadherin and membrane dynamics in living animals: Implications for disease modeling and drug development. Cancer Res 2009;69(7):2714-9.
20. Carmeliet P. Angiogenesis in health and disease. Nat Med 2003;9(6):653-60.
21. Auerbach R, Lewis R, Shinners B, Kubai L, Akhtar N. Angiogenesis assays: A critical overview. Clin Chem 2003;49(1):32-40.
22. Staton CA, Reed MW, Brown NJ. A critical analysis of current in vitro and in vivo angiogenesis assays. Int J Exp Pathol 2009;90(3):195-221.
23. Evensen L, Micklem DR, Link W, Lorens JB. A novel imaging-based highthroughput screening approach to antiangiogenic drug discovery. Cytometry A 2010;77(1):41-51.
24. Sanz L, Pascual M, Munoz A, Gonzalez MA, Salvador CH, Alvarez-Vallina L. Development of a computer-assisted high-throughput screening platform for anti-angiogenic testing. Microvasc Res 2002;63(3):335-9.
25. Sylvest L, Bendiksen CD, Houen G. Phosphatase inhibitors with antiangiogenic effect in vitro. APMIS 2010;118(1):49-59.
26. Bishop ET, Bell GT, Bloor S, Broom IJ, Hendry NF, Wheatley DN. An in vitro model of angiogenesis: Basic features. Angiogenesis 1999;3(4):335-44.
27. Zimmermann GR, Lehar J, Keith CT. Multitarget therapeutics: When the whole is greater than the sum of the parts. Drug Discov Today 2007;12(1-2):34-42.
28. Ainscow E, Carragher N. Addressing kinetic applications in High Content Screening. European Pharmaceutical Review 2008; (5):44-50.
29. Helmchen F, Denk W. Deep tissue two-photon microscopy. Nat Methods 2005;2(12):932-40.
30. Grant DM, Zhang W, McGhee EJ, Bunney TD, Talbot CB, Kumar S, et al. Multiplexed FRET to image multiple signaling events in live cells. Biophys J 2008;95(10):L69-71.
31. Kumar S, Alibhai D, Margineanu A, Laine R, Kennedy G, McGinty J, et al. FLIM FRET technology for drug discovery: Automated multiwell-plate high-content analysis, multiplexed readouts and application in situ. Chemphyschem 2011;12(3):609-26.
32. Talbot CB, McGinty J, Grant DM, McGhee EJ, Owen DM, Zhang W, et al. High speed unsupervised fluorescence lifetime imaging confocal multiwell plate reader for high content analysis. J Biophotonics 2008;1(6):514-21.
33. Echeverria V, Meyvantsson I, Skoien A, Worzella T, Lamers C, Hayes S. An automated high-content assay for tumor cell migration through 3-dimensional matrices. J Biomol Screen 2010;15(9):1144-51.
34. Gough W, Hulkower KI, Lynch R, McGlynn P, Uhlik M, Yan L, et al. A quantitative, facile, and high-throughput imagebased cell migration method is a robust alternative to the scratch assay. J Biomol Screen 2011;16(2):155-63.
35. Xu Y, Shi Y, Ding S. A chemical approach to stem-cell biology and regenerative medicine. Nature 2008; 453(7193):338-344.

This article was printed in the August/September 2011 issue of International Drug Discovery,6,, 4,. Copyright rests with the publisher.