References 61

Historically, the development of animal models for psychiatric disorders such as schizophrenia, anxiety, and depression has proven to be a challenging task. Indeed, it is difficult to ascertain the level of predictability or validation of an animal model to the human clinical condition. The development of animal models for psychiatric disorders was promoted after the introduction of chlorpromazine for the treatment of schizophrenia in 1954, and again after the introduction of chlordiazepoxide and valium for the treatment of anxiety in the 1960s. These turning points, along with the development of behavioral testing technology in experimental psychology after Skinner's publication of Behavior of Organisms in 1938, essentially brought into context the fields of psychology and pharmacology, leading to the emergence of the field of psychopharmacology (Carlton, 1983).

Drug discovery and drug development are long and complex processes within the pharmaceutical industry. Animal models continue to play an important role in this industry for screening drugs or for fully characterizing new molecular entities (NMEs) in the drug discovery stage, as they provide vital information that helps to determine whether an NME is actually useful in the clinic. Based on evaluation of drugs already known to have efficacy in humans for various psychiatric conditions, animal models have been undergoing continual refinement, and newer paradigms are being developed and validated for enhanced sophistication in terms of behavioral and therapeutic specificity.

The proposed predictive validity of animal models has led to their successful implementation in the pharmaceutical industry and, over the years, has resulted in an increasing trend of using behavioral models to screen compounds for efficacy and side-effects/toxicology. Furthermore, the sophistication and advancements in behavioral testing, such as telemetry, implementation of infrared-photocell technology, and digital video-tracking technology (which precisely tracks the movement of an animal in the testing apparatus), have provided the operational means for high-throughput in vivo testing. A combination of several of these methodologies also allows a fine analysis of behavioral specificity (e.g., whether a change in movement is due to specific changes of locomotor activity or to effects on motor coordination and balance, stereotypies, sedation, etc.).

Instrument and methodological advances can now be integrated with insights into behavioral testing and complementary approaches, i.e., testing in various models that provide combined know-how in an effort to collect as much relevant information as feasible considering the multiplicity, heterogeneity, and complex pathophysiology of neuropsychiatry manifestations and disorders. At the same time, refinement in behavioral testing also aids in the determination of specific and nonspecific behavioral effects, depending on the purpose as well as on the microstructure and domain analyses of a behavior via quantitative measures collected in real time by computerized behavioral monitoring systems and standardized methodologies and protocols. This approach can assist in reliability, replicability, and phenotype validation, especially when considering specific domains or segments of a behavior that may be associated with the induction or progression of a pathophysiological condition or response to a therapeutic intervention. Consistency of a specific behavioral profile across models and species, in the right setting and with appropriate interpretation, could build toward validation and understanding of complex behaviors. Overall, automation of behavioral testing has not only increased reliability and throughput of evaluating drugs in vivo for efficacy and safety in the preclinical stage, but has also prompted the refinement and integration of the data analysis process.

This chapter reviews various novel technologies and approaches currently used in drug discovery at the Johnson & Johnson Pharmaceutical Research and Development, LLC. These approaches range from automation of behavioral testing in a recently proposed model of depression, to the search and identification of biomarkers to substantiate models, to the use of high-throughput screening (HTS) robotics, and to the recently proposed novel paradigm for drug discovery known as functional informatics.

Researchers make every effort to automate behavioral tests in the neurosciences to minimize subjective judgment and manual labor and to maximize repeatability of the results between laboratories while increasing the throughput of experimental end points. In recent years, these efforts have accelerated due to (a) the development of many inbred or transgenic mice strains that needed systematic behavioral pheno-typing to find the functions delineated to strain differences or specific genes, and (b) the development of in vitro biochemical high-throughput drug-screening tests, which highlighted similar needs in behavioral neurosciences where the intact organism is a subject of the study. The main concerns of adapting any behavioral test to large-scale screening of either drugs or animal strains include the reproducibility of measurement end points between laboratories; the throughput or capacity of the test; and, importantly, the validity of the test as a surrogate marker for human disease.

The problem of the reproducibility of measurement end points arose when some of the first transgenic animals were screened by similar behavioral methods, producing different and sometimes opposite behavioral results by independent laboratories (Crabbe et al., 1999). As a consequence, the standardization of behavioral tests was proposed, and several test batteries were developed (Crawley, 1999; Crawley and Paylor, 1997; Rogers et al., 1999; Tarantino et al., 2000).

Another approach to overcome differences in the interpretation of behavioral results between laboratories and to facilitate automatic analysis was a proposal to analyze a given behavior as a group of segments. As it is clearly understood, any behavioral response is more complex than a biochemical reaction because it emerges from the organism as a whole and involves multiple parallel and sequential biochemical and physiological pathways. However, even a relatively simple behavior such as locomotor activity is highly structured and consists of several patterns of exploratory or goal-oriented behaviors that can be further subdivided into episodes such as stops and progression segments (Kafkafi, 2003), each involving multiple neural pathways. Software for the exploration of exploration (SEE) was developed using this principle (Drai and Golani, 2001; Kafkafi, 2003). This software can characterize differences between distinct mouse strains in a highly structured manner using mathematical algorithms to analyze subtle patterns in mouse locomotor activity (Kafkafi, 2003).

The approach to solve the problem of test capacity while retaining the validity of the test as a surrogate marker for human disease is exemplified in the development of the "cat walk" test (Hamers et al., 2001; Vrinten and Hamers, 2003). Before the development of this method, mechanical allodynia was assessed by the stimulation of inflamed tissue with a series of von Fray filaments and observation of the withdrawal reaction (Chaplan et al., 1994). Responses were not always clear with this method and were very much observer dependent. The "cat walk" is an automated, computer-based gait-analysis method that enables objective and rapid quantification of several gait parameters such as different phases of the step cycle and the pressure applied during locomotion. With this method, it is possible to distinguish between spinal cord injuries and neuropathic pain expressed as mechanical allodynia (Hamers et al., 2001; Vrinten and Hamers, 2003).

The dominant-submissive reaction model is another approach which was designed to overcome problems of measurement precision and capacity while retaining relative value as a marker for the human condition of depression. Dominance and submissiveness, defined in a competition test and measured as the relative success of two food-restricted rats to gain access to a feeder, form a behavioral paradigm called the dominant-submissive relationship (DSR). This paradigm results in two models sensitive to drugs used to treat mood disorders. Drugs used to treat mania inhibit the dominant behavior of rats taking food at the expense of an opponent (reduction of dominant-behavior model, or RDBM). Antidepressant treatment increases the competitive behavior of submissive rats that lose in such encounters before treatment (reduction of submissive-behavior model, or RSBM). The RSBM belongs to a broader group of tests employed to study antidepressants that are based on social interactions of animals. They can be divided into three groups that measure the same main process — the ability of some animals to be superior (dominant) to others and the acceptance of others to be inferior (submissive) — in different ways. One group consists of the dominant-submissive interactions seen in groups of animals and measured mainly by observation of agonistic and defensive postures used to rank each animal's position in the group (Blanchard et al., 1987; Blanchard and Riley, 1988; Grant and Mackintosh, 1963; Mackintosh and Grant, 1966; Miczek and Barry, 1977). Another is represented by resident-intruder interactions, based on animal territoriality, that result in a defeated state that resembles aspects of depression (Kudryavtseva et al., 1991; Mitchell and Fletcher, 1993; Willner, 1995). A third group consists of winner-loser relations established in competition tests that measure priority of access to a desired resource (Malatynska et al., 2002; Malatynska and Kostowski, 1984; Masur and Benedito, 1974; Masur et al., 1971; Uyeno, 1966; Uyeno, 1967). The complexity of submissive and dominant social behaviors often results in controversy about their precise definition.

Previous efforts to measure dominant-submissive relationships have used descriptive rating scales that are difficult to share among laboratories (Kudryavtseva et al., 1991; Willner, 1995). Observed differences in specific submissive behaviors between mouse strains (Kudryavtseva et al., 1991) would complicate interstrain comparisons using such end points. This problem is addressed in the DSR, for example, by using the simple end point of milk drinking by competing animals. This is facilitated by the apparatus, which allows only one animal to consume milk at a time. Submissive behavior is measured as the relative difference in time spent drinking milk during a 5-minute interval between the dominant and submissive members of paired animals. This end point is unambiguous and eliminates interobserver variability.

The precision of measuring DS behavior is also improved by the criteria applied to pair selection for having a dominant-submissive relationship. A pair is defined as having a dominant-submissive relation when: (a) the difference in time spent on the feeder by each animal from the pair is significantly different (P < .05) by the two-tail t test; (b) the difference in time spent on the feeder by each animal from the pair is 40% or more of the score value for the higher scoring (dominant) animal; and (c) there is no reversal of daily success as expressed by longer and shorter time spent on the feeder by an animal from the pair during the second week of observation time. Dominance measured by this behavioral procedure is a robust effect that is easily distinguished from submissiveness without complex subjective observations.

We have applied a multiple-subject video-tracking system (PanLab Software, San Diego Instruments, CA) to this method that automatically scores time spent by rats at the feeder (Pinhasov et al., 2005a). Using this method, it is possible to observe four pairs of rats during each 5-minute experimental session (one set). We have also used a duplicate parallel set with a second camera that enables immediate switch to the observation of the next four animals, thereby minimizing total experimental time. The multiple video-tracking systems reduced the variability between observations and improved throughput of dominant-submissive pairs by fivefold without increasing labor input. Thus, the number of animals available for drug testing is increased.

In summary, the automation of behavioral testing is a necessary next step in constructing precisely defined CNS diseases and pathophysiology models that will be repeatable between laboratories. The delineation of the model to a certain disease may improve with the segmentation of the given behavior and automatic parallel analysis of individual segments, domains, and subdomains. This would also serve to eliminate subjective scoring by observers. An application of this process was demonstrated in the SEE software and in the "cat walk" model. The requirement for model automation described above is also directly connected to the requirement for precise quantitative measurement, which was demonstrated in the DSR test by introducing a binary (positive or negative) end point. The process of automation of behavioral tests has only just started, and it is far behind the automation level of in vitro drug screening. However, it is possible that, with further understanding of the components of a complex behavior and with future technology improvements, its throughput will no longer be a serious limitation.

Behavioral pharmacology is an integral part of drug discovery and is essential for evaluation of drug activity. However, drug administration can activate labile and indirect mechanisms and thus mislead in the interpretation of effects caused by a drug. To address this question, the use of molecular biomarkers becomes a valuable addition for the evaluation of the mechanism of drug action. Furthermore, a better understanding of the molecular changes associated with the onset and progression of disease or of the temporal responses to drug administration provides a rational basis for the development of new diagnostic and therapeutic tools. For example, molecular markers such as beta-amyloid, prion protein (PrP), tau protein, and alpha-synuclein, which are basic components of specific brain lesions (amyloid plaques, prion plaques, tangles, and Lewy bodies, respectively), have dramatically improved the characterization and classification of numerous neurodegenerative diseases. In Alzheimer's disease, for example, phosphorylated tau and beta-amyloid deposition were prudently considered to be diagnostic markers (Hampel et al., 2003; Lewczuk et al., 2004; Nordberg, 2004). Discovery of molecular biomarkers also leads to the development of animal models, for example transgenic animals with overdeposition of beta-amyloid (Games et al., 1995; Hsiao et al., 1996) or with accumulation of phosphorylated tau (Gotz et al., 1995; Oddo et al., 2003a; Oddo et al., 2003b). Animal models can serve as a tool for revealing molecular biomarkers that can indicate whether certain alterations in physiological pathways will lead to patho-physiological cascades and disease.

We recently showed an example of such an approach by using the dominant-submissive relationship model for mania and depression (Malatynska et al., 2002), which was described above. In a more recent work using TaqMan quantitative reverse-transcriptase-polymerase chain reaction (RT-PCR) analysis (Pinhasov et al., 2004), it has been found that gamma-synuclein mRNA levels were down-regulated in the cerebral cortex of submissive rats (Pinhasov et al., 2005b). Hence, differential expression of gamma-synuclein in this model may provide initial insights into an aspect of the underlying pathophysiology in this condition and could provide novel tools for the development of new therapeutic interventions.

The development of microarray technologies (a) provides the advantage of being able to investigate the expression of thousands of genes simultaneously in treated versus control samples and (b) brings broader approaches to target identification (Palfreyman, 2002; Palfreyman et al., 2002). By using a microarray-based approach, one can compare normal and diseased tissue to obtain a "disease-associated profile or signature" of abnormally expressed genes. This is important, as the nature of neurological diseases is largely polygenic and can involve alteration in several biological pathways. Altar and colleagues studied the effects of electroconvulsive shock (ECS) exposures on gene transcription and identified genes that seem to be associated with the efficacy of chronic electroconvulsive therapy (ECT) (Altar et al., 2004) and found a complex pattern of differential gene expression among the treated and untreated subjects. Such a strategy could be used to identify lead compounds (Palfreyman, 2002; Palfreyman et al., 2002) that mimic the therapeutic response of ECT. Using the same approach, Altar et al. (2004) also investigated the effect of valproic acid on gene expression profiles of human postmortem parietal and prefrontal cortex samples of normal controls and of patients with bipolar disease; they ultimately identified a group of genes that was proposed to be associated with the disease condition. Overall, gene expression profiles can generate unique molecular patterns that direct to specific physiological-biochemical signaling systems associated with disease pathophysiology or cellular responses to drug administration.

Applications of molecular biomarkers can also translate to the use of diverse imaging methods such as single-photon emission computed tomography (SPECT), positron emission tomography (PET), and magnetic resonance imaging (MRI) (Marek and Seibyl, 2000). The combination of these methods with selective markers for various diseases can substantially impact the diagnosis and treatment of neuro-degenerative diseases and also brings new perspectives in finding new drugs and other therapeutic interventions (Klunk et al., 2003). Direct in vivo detection of amyloid deposits in patients diagnosed with Alzheimer's disease and in the brains of transgenic animals of amyloid deposition using brain neuroimaging techniques would be well suited for the early diagnosis of Alzheimer's disease and the development and assessment of new treatment strategies (Nordberg, 2004; Okamura et al., 2004; Suemoto et al., 2004). MRI, which provides views of anatomic or structural brain abnormalities, has also been especially useful for assessing macroscopic neuro-morphological changes in stroke and multiple sclerosis (Baird and Warach, 1999).

Drug administration associated with neuroactivity often involves alteration in neurotransmitter release (Adell and Artigas, 1998). Monitoring of neurotransmitter release is possible through the use of microdialysis methodology. The present phar-macotherapy of depression includes enhancement of central monoaminergic neurotransmission (Blier, 2003), and microdialysis is useful in the investigation of the effects of antidepressant agents on chemical neurotransmission, including in the syn-apsis (Adell and Artigas, 1998). Analysis and monitoring of acetylcholinergic mechanisms in cognitive and transgenic models of Alzheimer's disease also aid in the investigation of potential novel treatment strategies (Hartmann et al., 2004), and the assessment of amyloid-beta peptide in the brain interstitial fluid may offer new insights into amyloid-beta metabolism (Cirrito et al., 2003).

Combinations of different complementary drug-discovery approaches, such as behavioral pharmacology, genomics, and proteomics with molecular imaging and microdialysis, open new perspectives for the discovery of new molecular biomarkers. These integrated approaches to CNS drug discovery also increase the required tests that need to be conducted in concert with automated sample processing and analysis.

An interesting trend in the development of lab automation is that it comes out of a traditional HTS environment and then spreads to all areas of drug discovery and development. Its applications include, but are not limited to, early ADME (absorption, distribution, metabolism, and excretion) and toxicology studies, high-throughput proteomics, high-throughput target validation, and high-throughput biomarker discovery and validation. Instead of fully integrated systems, applications in these areas tend to rely more on integrated-workstation approaches. High-throughput target validation is a combination of traditional molecular biology methodologies with automation. Some examples are the small interfering RNA (siRNA)-based (Xin et al., 2004) and antisense-based approaches in combination with high-content screening. For high-throughput siRNA-based target validation, a fully integrated automation system for HTS was used to take full advantage of existing automation hardware (Xin et al., 2004). The assay has high throughput (2400 siRNAs could be tested in an 8-hour working day by one full-time employee) and great precision. The siRNA plates were tracked by the automation system, and data could be processed and queried automatically using a corporate database. This represents an example of an approach that allows large-scale siRNA-based target validation studies to be conducted in days rather than months, consequently reducing the timeline between target proposal and lead generation.

Was this article helpful?

0 0

Post a comment