As could be seen in a number of the example HTS screens described above, the transition from in vitro assays hits to efficacious compounds in more complex disease model assays- in cells, in tissues, and most critically in animal models- can be beset with many unexpected, and even seemingly “mysterious” failures. While many such compound dropouts will be due to fundamental chemical issues such as drug availability, metabolism, safety, toxicity, the important challenge in producing a successful drug lead candidate for further preclinical development and, ultimately, clinical trials, is whether efficacy in an in vitro assay translates with high predictive value into efficacy in a validated whole-animal disease model.
The in vitro to in vivo transition can have a very high, and sometimes abrupt, dropout rate. Over the last decade, increased emphasis has been placed on developing “high throughput biology” models in the mid-development process to help bridge the gap between simplified in vitro assays and the fully complex biologically context and content of a whole-animal disease model. Thus, “high-content” screens are designed to provide assays with more predictive value for in vivo benefit and clinical potential, but are scalable to evaluate much larger numbers of compounds than in vivo animal models. As such, HCS assays are commonly used as secondary screens to help rank order HTS hits for advancement into full preclinical development; they can also be used as primary screens if smaller target compound libraries (typically hundreds to a few thousands) can be identified with particular relevance to the disease state in question.
There are several ways in which the content of drug screening assays can be increased. One obvious, but not always technically easy, approach is to use primary cells instead of cell lines for a given assay. Any cell-based screen arguably has much increased biological content over any in vitro
assay, but the nature of the cell type in question may have a major impact on its clinical relevance. Indeed, the transformed nature of cell lines makes their relevance for many cell and tissue-types questionable, especially in the nervous system where all neurons in the adult brain are post-mitotic; transformed cell lines are commonly used, of course, for their ease, convenience, and reproducibility. It the context of HD, for example, it would be desirable to use primary striatal neuronal cultures instead of the typical non-neuronal cell line (e.g., HEK or CHO cells) in the design of HCS assays. While for many cell types, especially neurons, the establishment of primary cultures and conditions for their efficient molecular genetic perturbation can be difficult to achieve, the increased biological and clinical relevance of primary cell-based HCS is often considered to be worth the extra effort and expense. The biotechnology company Trophos SA, for example, has developed and implemented a primary striatal neuron HCS assay for HD using image-based endpoints [54
The use of more phenotypic and functional endpoint measures is the other major way in which the biological content of assays can be significantly increased. For neurodegenerative diseases, for example, this includes the use of image-based assays in which morphological indicators of neuronal “wellness”, such as changes in neurite extension, cell shape, and the translocation/transformation of intracellular disease markers (such as the Htt protein for HD) are used instead of more blunt measures such as overt cell death and clearance. Moreover, these more subtle phenotypic expressions of disease induction are generally thought to reflect earlier stages of cellular pathogenesis, and can potentially reveal drug targets and mechanisms that could be used to intervene at earlier stages in the disease process.
An emerging development in the HCS area takes a further step towards “in vivo
-ness” by using tissue explants instead of cell lines or primary cell cultures in order to retain, at least partially, the complex milieu of the local extracellular and tissue environment for a given disease-relevant cell type. In the sphere of neurological diseases, much effort has been placed in development of brain slice-based explant cultures for ischemic stroke, given that scores of neuroprotective clinical candidates developed from in vitro
and cell-based assays for stroke have, without exception, failed in the clinic thus far (in ~100 clinical trials to date) [55
]. Thus, even what in past years had seemed to be a relatively straightforward case of cellular oxygen and glucose deprivation, has turned out to involve a complexity of non-cell autonomous processes that cannot be adequately recapitulated even in primary neuronal cultures.
Thus, a number of organoptypic brain slice-based assays have been developed for ischemic stroke with cell or tissue-based endpoint for neuronal cell death such as propidium iodide (PI) staining [56
]. We have recently described a HCS screen for ischemic stroke based on the creation of a “sentinel” population of pyramidal neurons in cortical brain slice explants via
biolistic transfection of Green Fluorescent Protein [58
]. Subsequent to oxygen-glucose deprivation (OGD), these GFP-expressing sentinel neurons provide a living index of the extent and time course of the ischemic damage produced. The majority of cortical neurons degenerate and die by the 3rd
day following OGD (), thus creating a robust assay for the identification of compounds that can provide neuroprotection to neurons in the context of a living, but degenerating, brain tissue explant ().
Fig. (3) High-content screening. Cortical brain slice explant model for ischemic stroke. (A) Cortical brain slices biolistically transfected with an expression plasmid for yellow fluorescent protein under normal conditions (Left) and 24 h after 7.5 min of oxygen-glucose (more ...)
Using this brain slice-based stroke assay, ~5,000 synthetic and natural product compounds, including all FDA-approved drugs, were screened producing ~74 primary and reproducible hits. Intriguingly, one of the strongest hits was the cardiac glycoside neriifolin (see discussion of Piccioni et al
. above [36
]), and subsequent studies showed that other members of the cardiac glycoside family including the FDA-approved drugs digitoxin and digoxin were efficacious as well, albeit with lower potencies. These results suggest that some common pathogenic mechanisms may play broad roles in neuronal degeneration and death in both traumatic and neurodegenerative diseases such as HD.
Analogous tissue explant models for neurodegenerative diseases take advantage of the genetic underpinnings of such diseases, a prime example being that of htt for HD. Studies described above already underscore the complex and context-dependent nature of htt aggregation between in vitro
and cell-based screens; a further consideration is the potential role that tissue context may play in htt aggregation. Organotypic brain slice explants from the R6/2 transgenic mouse model for HD, for example, can be shown to develop htt aggregates suitable for use in an image-based assay over the course of 3-4 weeks in culture [23
We have taken an analogous approach to the Wang et al
. experiments described above to use biolistic transfection to introduce, along with the vital marker Yellow Fluorescent Protein, DNA expression constructs based on mutant htt into medium spiny neurons (MSNs) in cortico-striatal brain slice explants. The sentinel MSNs thus produced are used to track the progressive neurodegeneration produced by htt transfection, thus creating a brain slice-based HCS assay for HD that can be used to screen small molecule compounds as well as DNA-based drug target probes such as siRNAs. This assay was recently employed to aid the development of hits from the ST14A cell viability screen described above [30
]. Using similar approaches, Murphy and Messer [59
] demonstrated the ability of scFv intrabodies to protect against biolistically transfected htt, and Khoshnan et al.
showed that perturbation of the NF-kB pathway can alter the toxicity of mutant htt substantially [60
More biological content still can be achieved by developing HCS assays that are resident within living organisms. To date, those assays with sufficient throughput to be regarded as “screens” (able to evaluate appreciable numbers of compounds per unit time) have been based in invertebrate model organisms, such as nematodes, and in small vertebrates such as zebrafish. In a novel example of a high throughput screen that can measure a very complex phenotype in whole animals, Burns et al
. developed an assay to detect heart rate in a zebrafish model [61
]. They generated a transgenic line that expressed green fluorescent protein (GFP) in the heart muscle of developing early embryos (24 hr) at a time when the zebrafish embryos are still transparent. This allowed visualization of the embryos by automated, microscopic identification of the fluorescent hearts and recording of heart rates. They developed software that extracted raw data of heart rate readings from more than 90% of embryos per plate and automatically converted to a spreadsheet. This assay was validated by testing drugs that are known to affect the human heart rate in dose-response and time-course experiments. This assay is useful for identifying compounds that affect heart rate both as a toxicity test for drug-leads and for HTS screens in cardiac disease models. Similar approaches may be useful for developing whole animal based HTS in zebrafish models of polyglutamine disorders that have been recently developed [62
Thus, cell- and tissue- and even whole-organism based HCS screens, can be used to bridge the long divide between in vitro findings and the desired in vivo efficacy in disease models critical to the drug development process. Critically, with each assay of increasing biological complexity there is a concomitant increase in clinical relevance as assay context and content approach that of the in vivo situation. While throughput generally decreases and expenses rise with HCS assays of increasing complexity, these considerations may be minor compared the potential major upside of much increased likelihood of clinical success of drug candidates developed using such approaches.
The successful development of numerous HTS assays for neurodegenerative disorders, including assays for aggregation and cell death, promises an increasing role for the use of complex phenotypes in HTS approaches in this therapeutic area. Such assays are identifying a growing list of active compounds (), of which several are FDA-approved drugs and could have very short times lines to the clinic [27
]. Though these small-molecules can reverse phenotypes of several complex neurodegenerative disorders in cell culture and some invertebrate animal models, prioritization and testing in more validated mouse models is a serious challenge. More complex HCS assay and testing in primary cell culture models or invertebrate animal models can provide a reasonable means to prioritize these hits [54
]. However, activity in a small range of secondary assays, though useful for prioritizing such active compounds, is by no means a proof that inactive compounds do not merit further study. More extensive testing in a range of secondary assays, where feasible, would be useful for identifying compounds with therapeutic value. These HTS assays also provide scalable core platforms for larger-scale screening campaigns, both for chemical compounds as well as for genome-wide RNAi-based screens that may provide specific drug targets for selective aspects of neurodegenerative diseases such as aggregation and cell death.
Chemical Tools for Understanding Disease Mechanisms
The compounds discovered in the various assays described in this review may also find use as chemical tools for understanding the mechanisms of neurodegenerative diseases, including those that are insufficiently drug-like in other respects (e.g., stability, blood-brain barrier penetration, etc.) for further development as drug molecules themselves. For example, protein aggregates are common neuropathological components of HD and the other major neurodegenerative disorders including Alzheimer's and Parkinson's disease, but their role in disease pathology has been controversial, with both harmful and protective functions being invoked for aggregates [11
]. However, using small molecules that affect aggregates, one can test the various hypotheses and also probe the mechanisms involved in cellular aggregation. In fact, a recent screen using cell culture models of HD and Parkinson's disease found protective effects of enhancing aggregation with a small molecule hit compound [65
]. Thus, small molecule hits from phenotypic screens can be useful tools for testing hypotheses, and for protein target identification. Together, these approaches using small molecule chemical tools can enhance mechanistic understanding of disease pathways and identify targets for disease therapy.