GENOMICS IN DRUG DEVELOPMENT
Am J Pharmacogenomics 2002; 2 (4): 263-271 1175-2203/02/0004-0263/$25.00/0 © Adis International Limited. All rights reserved.
Deductive Genomics A Functional Approach to Identify Innovative Drug Targets in the Post-Genome Era Gabriele Stumm, Andreas Russ and Michael Nehls Ingenium Pharmaceuticals AG, Martinsried, Germany
Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Closing the Gap Between Genes and Function: the Concept of Deductive Genomics . . . 2. Model Organisms as a Tool in Pharmaceutical Discovery . . . . . . . . . . . . . . . . . . . . . 3. The Building Blocks of Deductive Genomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Efficient Genome-Covering Mutagenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 High Through-Put Screening for Medically Relevant Biological Activities . . . . . . . . . 3.3 High Through-Put and Rapid Identification of the Underlying Genetic Variation . . . . . 4. Deductive Genomics in Neurobiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 The Unmet Need for Innovative Therapeutic Approaches in Neuropsychiatric Disorders 4.2 Model Organism Based Discovery in Neurobiology . . . . . . . . . . . . . . . . . . . . . . 4.3 Deductive Genomics in Neurobehavioral Research, an Example . . . . . . . . . . . . . 5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Abstract
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
263 263 265 265 266 266 267 267 267 268 269 270
The sequencing of the human genome has generated a drug discovery process that is based on sequence analysis and hypothesis-driven (inductive) prediction of gene function. This approach, which we term inductive genomics, is currently dominating the efforts of the pharmaceutical industry to identify new drug targets. According to recent studies, this sequence-driven discovery process is paradoxically increasing the average cost of drug development, thus falling short of the promise of the Human Genome Project to simplify the creation of much needed novel therapeutics. In the early stages of discovery, the flurry of new gene sequences makes it difficult to pick and prioritize the most promising product candidates for product development, as with existing technologies important decisions have to be based on circumstantial evidence that does not strongly predict therapeutic potential. This is because the physiological function of a potential target cannot be predicted by gene sequence analysis and in vitro technologies alone. In contrast, deductive genomics, or large-scale forward genetics, bridges the gap between sequence and function by providing a function-driven in vivo screen of a highly orthologous mammalian model genome for medically relevant physiological functions and drug targets. This approach allows drug discovery to move beyond the focus on sequence-driven identification of new members of classical drug-able protein families towards the biology-driven identification of innovative targets and biological pathways.
1. Closing the Gap Between Genes and Function: the Concept of Deductive Genomics Genomics-driven drug discovery faces two major challenges: to efficiently identify the most promising candidate drug targets for areas of large unmet clinical need, and to validate and prioritize them as early as possible to minimize expensive failures in clinical development. For pragmatic reasons, the efforts in genomics-driven drug
discovery have, up to now, largely focused on well-characterized, classical families of drug targets. The ability of the medicinal chemist to create a compound that shows the desired biochemical effect on a chosen drug target has been shaped by decades of research on a limited number of protein families. New members of these families, encoded by ‘drug-able genes’, can be identified with high confidence by the application of standard bioinformatics procedures on expressed sequence tag (EST) sequences, and the human genome. As the biochemical properties of these
264
Stumm et al.
Genome/proteome
Deductive genomics
Inductive genomics
Sequence
Function
Drug-able targets
Drug-worthy targets
Novel therapeutics
Fig. 1. Inductive genomics leads to drug-able targets by sequence-driven and hypothesis-based approaches and therefore will miss medically important drug-worthy targets and biochemical pathways (small horizontal arrow). In contrast, deductive genomics using mutagenesis screens enables biology-driven discovery of novel gene functions and physiological pathways and will, therefore, identify and characterize novel therapeutically relevant pathways with drug-worthy targets in addition to those considered by the current drug discovery efforts (large horizontal arrow).
protein classes are well understood, the development of compounds acting on these familiar structures is reasonably straightforward. However, beyond the initial stages of identification and biochemical characterization in vitro, the development of product candidates derived by this sequence driven approach is much more difficult. The key determinant of successful drug development is not only that a compound with the desired biochemical profile can be produced, but also that the drug candidate shows the desired effect in vivo, and that the preclinical and clinical tests efficiently evaluate the desired and adverse effects of the compound. In the sequence driven paradigm, this in vivo validation has to be achieved by forming hypotheses based on the evidence available, which have then to be tested, frequently using reverse genetics techniques. Evidence to predict the organismal, as opposed to biochemical, function of a newly discovered gene is usually scarce, and the validation steps are time consuming and have to be applied to a large number of product candidates. Consequently, validation becomes a significant bottleneck in the © Adis International Limited. All rights reserved.
discovery process. The ability of the pharmaceutical industry to generate early stage compounds acting on novel drug-able genes outstrips the capacity to prioritize them based on hard biological evidence for further development. If decisions to enter expensive preclinical and clinical development stages are based on circumstantial evidence that does not reliably predict the physiological activity of the drug target, costly failures are a frequent result. In a recent study, it was estimated that the attrition rate of drug product candidates based on sequence-derived targets will increase rather than decrease when compared to the pre-Genomics era.[1] This is clearly opposite to the beneficiary effects on the speed and efficiency that the genomics revolution had been anticipated to provide. A general approach to identify and validate the therapeutic potential of a candidate target in vivo is clearly needed. Genome sequencing has illustrated that with current knowledge we can only assign function to less than 50% of the genes predicted.[2] Thus, it is very likely that novel physiological pathways exist which are sufficiently relevant for therapeutic intervention, i.e. drug-worthy (large horizontal arrow, figure 1). Pure sequence-driven approaches would not only ignore many important players in these pathways, if they don’t fall into the typical drug-able families, but since computer-based prediction of members of non-classical gene families is still in its infancy, novel gene classes might in fact not be identified at all. Purely sequence-driven and hypothesis-based inductive thinking behind the selection of drug-able targets (inductive genomics) will, therefore, inadvertently miss medically important drug-worthy targets and biochemical pathways (small horizontal arrow, figure 1). In contrast, mutagenesis screens provide a tool for biology-driven discovery of novel gene functions and physiological pathways. A genome-wide screen can query each individual gene of a model genome for its biological significance. Consequently, it offers the potential to start a drug discovery process based on the physiological function of the chosen target already validated in the context of a mammalian model organism. The deductive genomics approach described here extends the power of classical genetic analysis by the use of cutting edge genomics techniques and economics of scale. It will not miss the therapeutically valuable product candidates that are identified by ‘inductive genomics’, as there is no inherent bias against classical drug-able targets (figure 1). In addition, the deductive route of genome analysis will identify and characterize novel therapeutically relevant pathways with drug-worthy targets that are not considered by the current drug discovery efforts. Am J Pharmacogenomics 2002; 2 (4)
Deductive Genomics
2. Model Organisms as a Tool in Pharmaceutical Discovery Deductive genomics requires the use of an appropriate animal model. Although human genetics is also based on phenotypedriven analysis, it can only observe the effects of spontaneously arising variations. In contrast, genome-wide mutagenesis screens in model organisms can harness the power of efficient mutagens to explore all biological variation possible in the system of choice. Animal models have proven to be of high value for our understanding of the biological processes underlying complex human disorders. Evolutionarily distant organisms, like Drosophila melanogaster and Caenorhabditis elegans, with life cycles and physiologies very dissimilar to humans, have contributed immensely to the dissection of complex human disease processes into discrete biological pathways.[3] As each model system has its unique strengths and weaknesses, the key to their successful application in drug discovery lies in finding the appropriate level of abstraction. Due to the inherent power of the elaborate forward genetic tools available, the invertebrate models are invaluable for the analysis of signal transduction and cellular interactions. Also, most of our knowledge about the mechanisms of embryonic development is based on studies in these animal models. Since the early 1990s, the zebrafish Danio rerio has entered the stage as a system very well suited to study early development and pattern formation in vertebrates.[4] Findings in these organisms can usually be extrapolated to the human system at the level of the prototypical layout of biochemical and cellular pathways, which is usually highly conserved in evolution. Just two examples of seminal insights into disease biology derived from studies at seemingly unrelated topics in invertebrate models are the elucidation of the raspathway[5,6] and the Notch-pathway (see section 4.2). The predictive power of these models is much more limited when it comes to the prediction of the function of a human gene at the level of whole organism biology. It is obvious that physiological functions that are widely divergent between the species cannot be predicted exactly, for example in specific areas of neurobiology, immunology, metabolism, and cancer. Due to the substantial differences in genomic organization, observations in the distantly related organisms can not be translated one-on-one to a human gene of interest. Typically, there is more than one human ortholog for each conserved invertebrate gene, while the zebrafish genome frequently contains more than one potential ortholog for each human gene.[7,8] In contrast, mammalian models, like the mouse, are very closely related to humans, both in terms of physiology and geno© Adis International Limited. All rights reserved.
265
mics. Especially in the areas of immunology and metabolic control, the mouse closely mimics the human system, as exemplified by the recent rapid progress in the understanding of obesity, which has largely been driven by mouse genetics. Although the mouse obviously does not model the cognitive functions of humans, their central nervous system (CNS) and peripheral nervous system (PNS) are highly developed and of similar architecture. Preliminary comparative analyses of mouse and human genomic sequences show that the mouse genome consists of 200–300 segments of conserved synteny to the human genome,[9] and that most human genes seem to have one mouse ortholog. Exceptions from this pattern mainly occur in clustered gene families which can show differential expansion and contraction between the two species. Until recently, the use of the mouse model was dominated by reverse genetic approaches using transgenic and embryonic stem cell technologies, and the analysis of spontaneously arising mutants. A broad variety of such mouse models for medical disorders such as obesity, hypertension, developmental diseases, neurodegeneration, or epilepsy are available for research and drug validation commercially or from academic groups (e.g. http://jaxmice.jax.org/index.shtml, http://lsd.ornl.gov/mouse/; http://www.ornl.gov/TechResources/Trans/hmepg.html/ ). However, for a large number of spontaneous mouse models the underlying genetic mutation, as well as the causative pathophysiology, is still unclear. Models derived from transgenic or knockout experiments often have only been characterized for one special hypothesis-related medical area but not for additional and unexpected pathophysiologies. Furthermore, mouse mutants generated by gene targeting can only address alterations of known biological pathways, as their generation is always hypothesis driven. Therefore, the detection of new pathology and new pathway entries is limited by these models. With the advent of large scale ethyl nitrosourea (ENU) mutagenesis, and recent progress in genome mapping and sequencing technologies, phenotype driven screens in the mammalian system are now available as tools for pharmaceutical discovery. Some examples can be reviewed at http://www.gsf.de/ieg/groups/enu/mutants/index.html and http://www.mgu.har.mrc.ac.uk/mutabase/.
3. The Building Blocks of Deductive Genomics In order to efficiently screen an entire mammalian model genome for the biological activity of most, if not all, potential drug targets and medically relevant pathways, several key technologies have to be combined: • efficient genome-covering mutagenesis Am J Pharmacogenomics 2002; 2 (4)
266
• •
Stumm et al.
high through-put screening for medically relevant biological activities high through-put and fast identification of the underlying genetic variation.
Genome/proteome
Primary screen assays
A
High sensitivity
3.1 Efficient Genome-Covering Mutagenesis
Secondary screen assays B
ENU is a supermutagen that induces point mutations, predominantly A to T transitions. ENU and similar chemicals are widely used as mutagens in genetic screens in all model organisms. Its impressive efficiency as a mutagen in the mouse was described as early as the 1970s[10] and it was successfully used for the mutational analysis of specific genomic regions,[11] usually marked by deletions or inversions, and genome-wide mutagenesis screens for specific individual phenotypes.[12,13] Large scale genome-wide mutagenesis screens for a wide variety of medically relevant phenotypes are performed in a number of academic centres.[14,15] As the production and analysis of dominant mutations is fast and straightforward, requiring only a one-generation breeding protocol, most of the results reported to date describe dominant screens and mutants. Although logistically more demanding, the implementation of screening protocols for recessive mutations is an essential requirement for comprehensive analysis of gene function. Only recessive screens allow the analysis of loss of function situations and phenotypes detrimental to the viability or fertility of the founder animals. According to current estimates, one-fold genome coverage for recessive mutations requires the analysis of 700 micro-pedigrees consisting of 20–30 F3 animals. With the screening capacity of a well-equipped industrial laboratory, this can be achieved within 12 months. 3.2 High Through-Put Screening for Medically Relevant Biological Activities
The implementation of a sophisticated screening protocol (i.e. asking the right questions) is a prerequisite for the identification of phenotypes that are indicative of, and a discovery portal for, a medically relevant novel molecular pathway. A typical phenotyping protocol, as schematically illustrated in figure 2, employs several levels of screening activities. A primary screen (figure 2, level ‘A’) setting is built on parameters and assays that allow the identification of any altered physiology in the areas of interest with primarily high sensitivity. Models that are filtered through this primary screen are subjected to a disease area-dependent composition of secondary assays (level ‘B’) with emphasis on higher physiological specificity. The primary phenotyping procedure includes a detailed observation of litters from newborn survey to weaning in order to © Adis International Limited. All rights reserved.
Deductive genomics
High specificity Indepth functional analysis
C
Drug-worthy targets
D
Fig. 2. The Ingenium phenotyping protocol employs several levels of screening activities. A primary screen setting is built on parameters and assays that allow the identification of altered physiology with high sensitivity. Models that are filtered through this primary screen are subjected to a disease area-dependent composition of secondary assays with emphasis on high specificity. Examples of primary screen tests and secondary assays are described in the text (section 3.2). The identification of the underlying genetic variation is a prerequisite for detection of the primary drug target candidate, and the application of pathway dissection technologies of additional drug target candidates.
address early phenotypes. Besides a comprehensive examination for neurological and morphological variations, certain easily scoreable behavioral alterations are investigated. For example, the neurological and behavioral screen addresses social behavior, spontaneous locomotion and gait, as well as reflex status. In addition, the mouse models are challenged in a series of sensational, balance, and muscular strength tasks. Examples are the hotplate assay, visual placing, acoustic startle, postural and falling reflexes, hanging wire, or stationary beam and coat hanger climbing. These assays allow us to explore specific function in the afferent nervous system such as mechano- and thermo-sensation, as well as, the ability of hearing and vision, motor function and motor coordination. Behavioral challenges such as the classical elevated mazes, open field or dark-light preference are required to address the anxiety – exploration paradigm. Depression, memory and specific sensory motor gating, a central brain function disturbed in schizophrenia, are addressed by validated classical assays such as tail suspension, passive avoidance or pre-pulse inhibition of the acoustic startle. In order to obtain a comprehensive view that allows a primary differential diagnosis, we employ highly sensitive metaboAm J Pharmacogenomics 2002; 2 (4)
Deductive Genomics
lic, hematological, and immunological assay systems, such as blood cell count, a broad variety of clinical blood chemistry parameters addressing lipid and glucose metabolism, liver, and renal function, and many more, as well as a comprehensive fluorescence activated cell sorter (FACS) and enzyme-linked immunosorbent assay (ELISA) analysis of the immunological status of the organism. The sensitivity and specificity of the screen can be further improved by devising and implementing functional challenges that are appropriate to discover new biological mechanisms in the disease area of interest. An important improvement over the existing standards in the analysis of animal models is the collection of a comprehensive data set from all relevant physiological systems in each animal. Based on this comprehensive profile of primary parameters, a working diagnosis can be formulated that gives rise to an even more specialized panel of secondary screening parameters and challenges (figure 2, level ‘B’). Secondary assays range from behavioral training tests for contextual fear or motor coordination over peripheral dual-energy X-ray densitometer (pDEXA) analysis of body lean mass, to challenges such as glucose tolerance tests or immunological stimulation experiments in vivo and ex vivo. Even detailed histopathological analysis, even as a lethal assay, is available as an important source of high valuable information already as a secondary assay in a recessive screen. The combined set of data from primary and secondary analysis allows the researcher to decide if the mutant phenotype is worthwhile to pursue further. The identification of the underlying genetic variation is a prerequisite for detection of the primary drug target candidate (figure 2, level ‘C’), and the application of pathway dissection technologies of additional drug target product candidates (level ‘D’). In order to be able to support an unbiased deductive analysis of the parameter set, we have developed a computational tool that aids the recognition of novel clinical patterns by application of clustering algorithms. This technology will significantly assist the analysis of physiological networks and the discovery of unsuspected biological interdependencies that might lead to the innovative therapeutic principles.
3.3 High Through-Put and Rapid Identification of the Underlying Genetic Variation
In the past, identification of the genetic variation underlying the mutant phenotype by positional cloning has been a time- and labor-consuming research enterprise. As a typical example, the molecular characterization of the immunodeficiency locus in the nude mouse model required an aggregate of over 20 researcher years.[16-18] Several-fold longer efforts were required for human © Adis International Limited. All rights reserved.
267
genetic disorders like cystic fibrosis,[19,20] Huntington disease[21] and ataxia telangiectasia mutated (ATM).[22] The advent of high throughput genomics technologies has recently dramatically increased the speed and efficiency of positional cloning of genetic variants in murine models. Milestones of this process were: • availability of extensive EST data of human and mouse • genetic and physical maps based on sequence tagged sites • use of single-nucleotide polymorphisms to provide a high density of polymorphic markers even between closely related mouse strains • robust assay systems with corresponding high through-put robotics-based phenotyping capabilities • exponential increase in speed and decrease in cost of genomic sequencing • availability of the human genomic sequence • software tools allowing efficient comparative sequence analysis of large genomic regions. The sum of the incremental improvements implemented to date has increased the speed and efficiency of positional cloning in the murine system by two orders of magnitude, allowing the molecular characterization in a matter of months. The availability of a ‘draft’ mouse genomic sequence by the end of 2001 (http://www.ensembl.org/mus_musculus/), and further improvements in genotyping technology, for example by mass spectrometry, will further enhance the process. 4. Deductive Genomics in Neurobiology The Ingenium clinical screen is currently focused on four disease indications, which are immunology, metabolic disorders, cancer, and neurobiology. The great advantage of the combined screening effort and the intense crosstalk between these major disease programs is not only the collection of a comprehensive data set on each animal, but, more important, a combined broad understanding of pathophysiological mechanisms underlying phenotypical alterations. Nevertheless, each disease program has its own rationales and methods. In this review we would like to detail the principles of deductive genomics in neurobiology. 4.1 The Unmet Need for Innovative Therapeutic Approaches in Neuropsychiatric Disorders
Despite the therapeutic efforts made in neuropsychiatry over the last decades, the deleterious effect of neuropsychiatric disorders is still tremendous. Among the top 10 diseases leading to severe disability, and, therefore, to high socioeconomic costs and health expenditures, 5 are neurological and psychiatric disorders Am J Pharmacogenomics 2002; 2 (4)
Stumm et al.
4.2 Model Organism Based Discovery in Neurobiology
In the mid-1960s both Drosophila and C. elegans, respectively, were introduced as valid model systems for neural development and neurobiological studies. A major milestone in Alzheimer’s disease research was achieved by linking diverse fields such as developmental biology © Adis International Limited. All rights reserved.
60 50 40 30 20 10
di so rd C on er an g om en it O alie al st eo s ar th O rit bs Sch is es izo si ve phr e -c om nia di pu so ls rd ive er s
Bi p
ol
ar
C O PD
0
U
(figure 3).[23] Depression, generalized anxiety and obsessivecompulsive disorders as well as schizophrenia are frequent conditions with a lifetime prevalence of 1 to 5% of the population worldwide.[24-27] However, treatment for all of them is long-term and often unsatisfying.[27-31] Whereas novel, more selective drugs reflecting underlying neurochemical disturbances have evolved for affective disorders and schizophrenia, the available treatments for anxiety are still very inadequate. The adverse effects of all existing therapeutics are severe and detract from quality of life. The therapeutic repertoire for neurological disorders is even smaller. The greatest therapeutic success of the past decade has evolved in the treatment of stroke. However, this concerns mainly thrombolysis, influence of brain edema and supportive intensive medical care. No major breakthrough has been achieved so far in neuroprotection, although, in vitro and experimental studies had been promising in respect to their influence on excitatory stress and apoptosis. Similar slow progress has been made in neurodegenerative disorders such as Parkinson’s disease,[32,33] amyotrophic lateral sclerosis (ALS; Lou Gehring’s disease),[34-36] Huntington’s Chorea, and even Alzheimer’s disease. Neuroprotective agents such as free-radical scavengers and neurotrophic growth factors have been thoroughly investigated for their potential benefit in experimental settings; the clinical trials however were disappointing.[37-39] Therapeutics for two other major neurological disorders, multiple sclerosis[40] and epilepsy,[41] are only able to ameliorate the clinical symptoms and slow down, but not prevent, the degenerative process.[42,43] Therefore, they are just postponing the disabling consequences of the underlying pathological processes. Lifelong treatment is most often needed, and patient compliance is often low due to major adverse effects. The high unmet need for therapeutics in neuropsychiatry is paralleled by the need for relevant animal models that help to analyze the underlying pathophysiology and to identify novel therapeutic entry points. Both neurological diseases and psychiatric disorders are believed to represent disequilibrium in the highly complex structural organization of cell-cell interactions within the nervous system. This restricts the possibility to successfully investigate pathway disturbances in cell culture models.
ni Total years lived with disability (million) po de lar pr m Iro ess ajo n- ion r de fic an ien em cy i Ac a Al ci de co ho nt s la dd ic tio n
268
Fig. 3. The global burden of disease. Five of the top 10 diseases leading to severe disability and, therefore, to high socioeconomic costs and health expenditures, are neurological and psychiatric disorders. All of them are either chronic, progressive or leave behind permanent defects. This epidemiological analysis demonstrates the tremendous deleterious effect of these disorders and the lack of a therapeutic breakthrough in these medical areas. COPD = chronic obstrusive pulmonary disease.
and neurobiology through genetic and biochemical studies in these invertebrate model systems. Assays for behavioral characterization in C. elegans include not only chemo-, mechano- and thermosensation, locomotion, eating and learning but also egg laying. Interestingly, it was the phenotypic expression of the sel12 deficiency in the roundworm leading to an egg-laying defect that was successfully used to identify novel members of the Presenilin signaling pathway. Disease causative mutations in the Presenilin 1 and 2 genes were found in a small fraction of patients with Alzheimer’s disease. The biological significance of this finding became clearer when Greenwald et al.[44] demonstrated that sel-12, a suppressor of a gain-of-function mutation in lin-12 gene was the C. elegans homologue of human Presenilin. Furthermore, this finding connected the Presenilin genes to the Notch signaling pathway, as lin-12 itself is the C. elegans homolog of human Notch.[45-48] This suggested the involvement of Presenilins in Notch cleavage and thereby in the fatal processing of the amyloid precursor protein, a key event in Alzheimer’s pathophysiology. This example nicely illustrates the power of genetic pathway dissection. However, it also indicates that the identification of novel entry points is difficult using lower organisms, as it has been impossible to predict that an a priori egg-laying defect in roundworms has implications in the understanding of certain aspects of Alzheimer’s disease. This is due to the fact that biochemAm J Pharmacogenomics 2002; 2 (4)
Deductive Genomics
istry and cell biology is highly conserved but organismal functions are not necessarily. Another, albeit rare, historic example where the phenotype was indeed predictive was the discovery of the spontaneous Drosophila mutant ‘shaker’.[49] The cloning of the underlying defect led to the identification of a potassium channel and initiated a breakthrough in unraveling underlying defects in human familial disorders of neuromuscular excitation and epilepsies.[50-52] The biological conclusions drawn from worm and fruitfly development and adult pathophysiology of spontaneous variants and early x-ray induced mutants were of such high value that systematic mutagenesis by DNA alkylating agents such as ethyl methane sulfonate (EMS), methyl methane sulfonate (MMS), and ENU[53-55] and phenotypical screens were initiated in invertebrates. Although, simple behavioral tests are possible in C. elegans and Drosophila, a comparable complexity in neural organization is only reached in vertebrates. Mice have been the subjects of neurobiological research as long as fruit flies, and spontaneous neurological mutants like ‘lurcher’ or ‘staggerer’ have long served as valuable animal models for the study of brain development. Some other naturally occurring mouse mutants like ‘lethargic’ or ‘stargazer’ presented striking clinical similarities to certain forms of childhood epilepsy.[56-63] Even more exciting was the discovery that, like in the these mouse models, such complex human disorders as absence epilepsy and juvenile myoclonic epilepsy are caused by single mutations in genes encoding subunits of the voltage-dependent calcium channels.[64-67] In the meantime several causative mutations of different forms of epilepsy have been chromosomally mapped, cloned or modeled by transgenic techniques in the mouse and have contributed to a better understanding of epileptic pathophysiology.[68-72] The power of forward genetics in the mammalian system has furthermore been illustrated by the identification of the ‘clock’ mutant. Identified by ENU mutagenesis and a clever phenotypic screen, it provides a molecular entry point into the pathways of mammalian circadian rhythm biology. Interestingly, the molecular elements of the pathway are highly conserved between vertebrate and invertebrate systems, but the layout of the pathways differ substantially.[73,74] These examples show that medical research, using mouse models, have proven advantageous for neuroscience. This is based on the fact that mice share most aspects of mammalian physiology and pathophysiology with humans, and that novel pathway entry points can be identified by their use. Unfortunately, murine models exist only for a few neurological disorders, such as certain rare neurodegenerative conditions and several distinct epileptic or myopathic disease entities so far. These existing mouse lines are biased for easily observable phenotypes, © Adis International Limited. All rights reserved.
269
whereas less obvious disturbances would only be detectable by a set of sensitive functional assays. Therefore, the major disorders such as Alzheimer’s disease, Parkinson’s disease, multiple sclerosis, and the affective disorders still suffer from a lack of valuable mouse models. Only recent advances in genomic technologies made the use of mouse genetics powerful enough to initiate largescale mutagenesis and sensitive phenotype screening programs in order to meet the need of mouse models and new pathway entries in neuropsychiatric disorders.
4.3 Deductive Genomics in Neurobehavioral Research, an Example
A genetic influence has been shown in a variety of neuropsychiatric disorders such as unipolar depression, schizophrenia, Alzheimer’s disease, Huntington’s disease, and most of the primary neuropathies and myopathies. Underlying mutations have been identified for Huntington’s Disease, in rare familial cases of certain neuropathies and fatal myopathies and for familial early onset Alzheimer’s disease. Unfortunately, these diseases only account for a small percentage of all neuropsychiatric disorders. For the majority of deleterious diseases of the CNS, the underlying genetic alteration and sometimes even the pathophysiology still remains to be discovered. As an example, epilepsy consists of a broad spectrum of different disease entities with different clinical signs and inheritance patterns. Seizures can be secondary to other pathophysiological conditions such as metabolic disturbances. As known so far, the primary epilepsies can be caused by alterations of ion channels,[65,67] by dysfunction of neurotransmitter receptors or transporters, by neuro-developmental disturbances or by neurodegenerative and neuro-autoimmunological conditions.[66,75] As mentioned above (see section 4.2), defects in potassium channels have been cloned as underlying mutations for epileptic disorders, as well as calcium channel alterations and dysfunction of glycine or γ-amino butyric acid (GABA) receptors. Some storage disorders like Laforas disease or Ceroid-lipofuscinosis, for which genes are mapped, lead to seizures by their neurodegenerative effect. Overall, about 40 rare familial human conditions presenting with seizures are linked to mutations in particular chromosomal locations, for less than 20 a homologous gene has been cloned in a corresponding mouse model. However, the majority of highly incidental epileptic entities still await the identification of the underlying mutational event and valuable animal models. Applying the neurobehavioral screening procedure described above allows detection of spontaneous as well as exercise-, acoustic- or photo optical-induced seizures. After screening about 20% of the mouse genome for genes that affect neuroAm J Pharmacogenomics 2002; 2 (4)
270
Stumm et al.
behavioral parameters, about 30 independent mouse models with cramping or seizures have been identified. For several epileptic mouse models the underlying gene mutation has been cloned. Among them we detected novel genes or genes so far not associated with epileptic seizures, but also well known genes such as myosin 5A corresponding to the dilute lethal phenotype. The recessive mutation in myosin 5A leads to progeria and severe spontaneous generalized seizures associated with cachexia and early lethality usually until weaning. The strategy for positional cloning in this case, where affected animals are not able to reproduce, is based on outcross breeding of the phenotypically normal heterozygote parents and littermates. The identification of this mutation is a typical example of the advantages of a recessive ENU screen, as severe phenotype and early lethality does not restrict the ability to clone the gene. A different example is a dominant mouse line characterized by exercise-induced cramping underlying a circadian rhythm in heterozygotes, which is reminiscent of the circadian occurrence of seizures in certain epileptic disorders in humans. In addition, muscular weakness and motor coordination are disturbed. The phenotype is complicated by a certain percentage of male infertility and a perinatal lethality of homozygotes. The positional cloning of the underlying mutation detected a gene so far not connected to hyperexcitability or epilepsy, but the pathway in which this gene easily explains its implication in neuroexcitotoxicity and neurodegeneration.
Acknowledgements The authors have provided no information on sources of funding or on conflicts of interest directly relevant to the content of this review.
References 1. 2. 3. 4.
5. 6. 7. 8.
9. 10.
11.
12.
5. Conclusion 13.
Only recent advances in genomic technologies made the use of mouse genetics powerful enough to initiate large scale mutagenesis and sensitive phenotype screening programs. The aim of these efforts is to meet the need for new mouse models and pathophysiological pathway entries in order to find and biologically validate new drug-worthy targets. Genome-wide mutagenesis screens now provide a tool for biology-driven discovery of novel gene functions and physiological pathways. In addition, a phenotype-driven search offers the potential to start a drug discovery process based on the physiological function of already validated targets in the context of a mammalian model organism. The implementation of screening protocols for medically more relevant recessive mutations has been an essential prerequisite for the analysis of loss of function situations, and even those phenotypes that affect viability or fertility. Therefore, the deductive genomics approach described here will identify, in addition to the therapeutically valuable candidates that are identified by inductive genomics, novel therapeutically relevant pathways with drug-worthy targets so far not considered by the current drug discovery efforts. © Adis International Limited. All rights reserved.
14. 15.
16. 17.
18. 19.
20. 21.
22. 23.
Edmunds RC, Ma PC, Tanio CP. Splicing a cost squeeze into the genomics revolution. McKinsey Q 2001; (2): 15-9 Lander ES, Linton LM, Birren B, et al. Initial sequencing and analysis of the human genome. Nature 2001; 409: 860-921 O’Neil NJ, Martin RL, Tomlinson ML, et al. RNA-mediated interference as a tool for identifying drug targets. Am J Pharmacogenomics 2001; 1: 45-53 Haffter P, Granato M, Brand M, et al. The identification of genes with unique and essential functions in the development of the zebrafish, Danio rerio. Development 1996; 123: 1-36 Gaul U, Chang H, Choi T, et al. Identification of ras targets using a genetic approach. Ciba Found Symp 1993; 176: 85-92 Dickson B, Hafen E. Genetics of signal transduction in invertebrates. Curr Opin Genet Dev 1994; 4 (1): 64-70 Woods IG, Kelly PD, Chu F, et al. A comparative map of the zebrafish genome. Genome Res 2000; 10 (12): 1903-14 Postlethwait JH, Woods IG, Ngo-Hazelett P, et al. Zebrafish comparative genomics and the origins of vertebrate chromosomes. Genome Res 2000; 10 (12): 1890-902 Lander ES, Linton LM, Birren B, et al. Initial sequencing and analysis of the human genome. Nature 2001; 409 (6822): 860-921 Russell WL, Kelly EM, Hunsicker PR, et al. Specific-locus test shows ethylnitrosourea to be the most potent mutagen in the mouse. Proc Natl Acad Sci U S A. 1979 Nov; 76 (11): 5818-9 Rinchik EM, Carpenter DA. N-ethyl-N-nitrosourea mutagenesis of a 6- to 11-cM subregion of the Fah- Hbb interval of mouse chromosome 7: completed testing of 4557 gametes and deletion mapping and complementation analysis of 31 mutations. Genetics 1999; 152 (1): 373-83 Vitaterna MH, King DP, Chang AM, et al. Mutagenesis and mapping of a mouse gene, Clock, essential for circadian behavior. Science 1994; 264 (5159): 719-25 Su LK, Kinzler KW, Vogelstein B, et al. Multiple intestinal neoplasia caused by a mutation in the murine homolog of the APC gene. Science 1992; 256 (5057): 668-70 Hrabe de Angelis MH, Flaswinkel H, Fuchs H, et al. Genome-wide, large-scale production of mutant mice by ENU mutagenesis. Nat Genet 2000; 25 (4): 444-7 Nolan PM, Peters J, Strivens M, et al. A systematic, genome-wide, phenotypedriven mutagenesis programme for gene function studies in the mouse. Nat Genet 2000; 25 (4): 440-3 Nehls M, Kyewski B, Messerle M, et al. Two genetically separable steps in the differentiation of thymic epithelium. Science 1996; 272 (5263): 886-9 Schlake T, Schorpp M, Nehls M, et al. The nude gene encodes a sequence-specific DNA binding protein with homologs in organisms that lack an anticipatory immune system. Proc Natl Acad Sci U S A 1997; 94 (8): 3842-7 Nehls M, Pfeifer D, Schorpp M, et al. New member of the winged-helix protein family disrupted in mouse and rat nude mutations. Nature 1994; 372 (6501): 103-7 Riordan JR, Rommens JM, Kerem B, et al. Identification of the cystic fibrosis gene: cloning and characterization of complementary DNA. Science 1989; 245 (4922): 1066-73 Rommens JM, Iannuzzi MC, Kerem B, et al. Identification of the cystic fibrosis gene: chromosome walking and jumping. Science 1989; 245 (4922): 1059-65 A novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington’s disease chromosomes. The Huntington’s Disease Collaborative Research Group. Cell 1993; 72 (6): 971-83 Savitsky K, Bar-Shira A, Gilad S, et al. A single ataxia telangiectasia gene with a product similar to PI-3 kinase. Science 1995; 268 (5218): 1749-53 Lopez AD, Murray CC. The global burden of disease, 1990-2020. Nat Med 1998; 4 (11): 1241-3
Am J Pharmacogenomics 2002; 2 (4)
Deductive Genomics
24. Aalto-Setala T, Marttunen M, Tuulio-Henriksson A, et al. One-month prevalence of depression and other DSM-IV disorders among young adults. Psychol Med 2001; 31 (5): 791-801 25. Gliatto MF. Generalized anxiety disorder. Am Fam Physician 2000; 62 (7): 1591-602 26. Parikh SV, Lam RW. Clinical guidelines for the treatment of depressive disorders: I. definitions, prevalence, and health burden. Can J Psychiatry 2001; 46 Suppl. 1: 13S-20S 27. Krystal JH, D’Souza DC, Sanacora G, et al. Current perspectives on the pathophysiology of schizophrenia, depression, and anxiety disorders. Med Clin North Am 2001; 85 (3): 559-77 28. Allison DB, Casey DE. Antipsychotic-induced weight gain: a review of the literature. J Clin Psychiatry 2001; 62 Suppl. 7: 22-31 29. Bech P. Social aspects of treatment of depression. Int Clin Psychopharmacol 1995; 10 Suppl. 1: 11-4 30. Bourin M, Chue P, Guillon Y. Paroxetine: a review. CNS Drug Rev 2001; 7 (1): 25-47 31. Marder SR. Facilitating compliance with antipsychotic medication. J Clin Psychiatry 1998; 59 Suppl. 3: 21-5 32. Ploix C, Spier AD. Fighting Bax: towards a Parkinson’s disease therapy. Trends Neurosci 2001; 24 (5): 255 33. Hely MA, Fung VS, Morris JG. Treatment of Parkinson’s disease. J Clin Neurosci 2000; 7 (6): 484-94 34. Rowland LP, Shneider NA. Amyotrophic lateral sclerosis. N Engl J Med 2001; 344 (22): 1688-700 35. Bjartmar C, Trapp BD. Axonal and neuronal degeneration in multiple sclerosis: mechanisms and functional consequences. Curr Opin Neurol 2001; 14 (3): 271-8 36. Julien JP. Amyotrophic lateral sclerosis, unfolding the toxicity of the misfolded. Cell 2001; 104 (4): 581-91 37. Alexi T, Borlongan CV, Faull RL, et al. Neuroprotective strategies for basal ganglia degeneration: Parkinson’s and Huntington’s diseases. Prog Neurobiol 2000; 60 (5): 409-70 38. Skaper SD, Walsh FS. Neurotrophic molecules: strategies for designing effective therapeutic molecules in neurodegeneration. Mol Cell Neurosci 1998; 12 (4-5): 179-93 39. Baekelandt V, De Strooper B, Nuttin B, et al. Gene therapeutic strategies for neurodegenerative diseases. Curr Opin Mol Ther 2000; 2 (5): 540-54 40. Wingerchuk DM, Weinshenker BG. Multiple sclerosis: epidemiology, genetics, classification, natural history, and clinical outcome measures. Neuroimaging Clin N Am 2000; 10 (4): 611-24 41. Gardiner RM. Genetic basis of the human epilepsies. Epilepsy Res 1999; 36 (2-3): 91-5 42. Trask PC, Esper P, Riba M, et al. Psychiatric side effects of interferon therapy: prevalence, proposed mechanisms, and future directions. J Clin Oncol 2000; 18 (11): 2316-26 43. Pellock JM. Treatment of epilepsy in the new millennium. Pharmacotherapy 2000; 20 (8 Pt 2): 129S-38S 44. Greenwald IS, Sternberg PW, Horvitz HR. The lin-12 locus specifies cell fates in Caenorhabditis elegans. Cell 1983; 34 (2): 435-44 45. Levitan D, Greenwald I. Effects of SEL-12 presenilin on LIN-12 localization and function in Caenorhabditis elegans. Development 1998; 125 (18): 3599-606 46. Li X, Greenwald I. HOP-1, a Caenorhabditis elegans presenilin, appears to be functionally redundant with SEL-12 presenilin and to facilitate LIN-12 and GLP-1 signaling. Proc Natl Acad Sci U S A 1997; 94 (22): 12204-9 47. Struhl G, Fitzgerald K, Greenwald I. Intrinsic activity of the Lin-12 and Notch intracellular domains in vivo. Cell 1993; 74 (2): 331-45 48. Sisodia SS, St George-Hyslop PH. Gamma-secretase, Notch, Amyloid ß and Alzheimer s disease. Where do the presinilins fit in? Nat Rev Neurosci 2002; 3: 281-90 49. Baker K, Salkoff L. The Drosophila Shaker gene codes for a distinctive K+ current in a subset of neurons. Neuron 1990; 4 (1): 129-40 50. Kamb A, Iverson LE, Tanouye MA. Molecular characterization of Shaker, a Drosophila gene that encodes a potassium channel. Cell 1987; 50 (3): 405-13 51. Felix R. Channelopathies: ion channel defects linked to heritable clinical disorders. J Med Genet 2000; 37 (10): 729-40
© Adis International Limited. All rights reserved.
271
52. Papazian DM, Schwarz TL, Tempel BL, et al. Cloning of genomic and complementary DNA from Shaker, a putative potassium channel gene from Drosophila. Science 1987; 237 (4816): 749-53 53. De Stasio EA, Dorman S. Optimization of ENU mutagenesis of Caenorhabditis elegans. Mutat Res 2001; 495 (1-2): 81-8 54. Nivard MJ, Pastink A, Vogel EW. Molecular analysis of mutations induced in the vermilion gene of Drosophila melanogaster by methyl methanesulfonate. Genetics 1992; 131 (3): 673-82 55. Fossett NG, Arbour-Reily P, Kilroy G, et al. Analysis of ENU-induced mutations at the Adh locus in Drosophila melanogaster. Mutat Res 1990; 231 (1): 73-85 56. Burgess DL, Jones JM, Meisler MH, et al. Mutation of the Ca2+ channel beta subunit gene Cchb4 is associated with ataxia and seizures in the lethargic (lh) mouse. Cell 1997; 88 (3): 385-92 57. Burgess DL, Noebels JL. Single gene defects in mice: the role of voltage-dependent calcium channels in absence models. Epilepsy Res 1999; 36 (2-3): 111-22 58. Noebels JL, Qiao X, Bronson RT, et al. Stargazer: a new neurological mutant on chromosome 15 in the mouse with prolonged cortical seizures. Epilepsy Res 1990; 7 (2): 129-35 59. Escayg A, Jones JM, Kearney JA, et al. Calcium channel beta 4 (CACNB4): human ortholog of the mouse epilepsy gene lethargic. Genomics 1998; 50 (1): 14-22 60. Hosford DA, Lin FH, Wang Y, et al. Studies of the lethargic (lh/lh) mouse model of absence seizures: regulatory mechanisms and identification of the lh gene. Adv Neurol 1999; 79: 239-52 61. Rho JM, Szot P, Tempel BL, et al. Developmental seizure susceptibility of kv1.1 potassium channel knockout mice. Dev Neurosci 1999; 21 (3-5): 320-7 62. Letts VA, Felix R, Biddlecome GH, et al. The mouse stargazer gene encodes a neuronal Ca2+-channel gamma subunit. Nat Genet 1998; 19 (4): 340-7 63. Jin W, Sugaya A, Tsuda T, et al. Relationship between large conductance calcium-activated potassium channel and bursting activity. Brain Res 2000; 860 (1-2): 21-8 64. Hallmann K, Durner M, Sander T, et al. Mutation analysis of the inwardly rectifying K (+) channels KCNJ6 (GIRK2) and KCNJ3 (GIRK1) in juvenile myoclonic epilepsy. Am J Med Genet 2000; 96 (1): 8-11 65. Steinlein OK, Noebels JL. Ion channels and epilepsy in man and mouse. Curr Opin Genet Dev 2000; 10 (3): 286-91 66. Chevassus-au-Louis N, Baraban SC, Gaiarsa JL, et al. Cortical malformations and epilepsy: new insights from animal models. Epilepsia 1999; 40: 811-21 67. Barclay J, Balaguero N, Mione M, et al. Ducky mouse phenotype of epilepsy and ataxia is associated with mutations in the Cacna2d2 gene and decreased calcium channel current in cerebellar Purkinje cells. J Neurosci 2001; 21: 6095-104 68. Kearney JA, Plummer NW, Smith MR, et al. A gain-of-function mutation in the sodium channel gene Scn2a results in seizures and behavioral abnormalities. Neuroscience 2001; 102: 307-17 69. Ferraro TN, Golden GT, Smith GG, et al. Mapping loci for pentylenetetrazol-induced seizure susceptibility in mice. J Neurosci 1999; 19: 6733-9 70. Skradski SL, Clark AM, Jiang H, et al. A nocel gene causing a mendelian audiogenic mouse epilepsy. Neuron 2001; 31: 537-44 71. Hallmann K, Durner M, Sander T, et al. Mutation analysis of the inwardly rectifying K (+) channels KCNJ6 (GIRK2) and KCNJ3 (GIRK1) in juvenile myoclonic epilepsy. Am J Med Genet 2000; 96: 8-11 72. Barclay J, Rees M. Mouse models of spike-wave epilepsy. Epilepsia 1999; 40 Suppl. 3: 17-22 73. Wager-Smith K, Kay SA. Circadian rhythm genetics: from flies to mice to humans. Nat Genet 2000; 26 (1): 23-7 74. Kume K, Zylka MJ, Sriram S, et al. mCRY1 and mCRY2 are essential components of the negative limb of the circadian clock feedback loop. Cell 1999; 98 (2): 193-205 75. Fleck MW, Hirotsune S, Gambello MJ, et al. Hippocampal abnormalities and enhanced excitability in a murine model of human lissencephaly. J Neurosci 2000; 20 (7): 2439-50
Correspondence and offprints: Dr Gabriele Stumm, Ingenium Pharmaceuticals AG, Fraunhofer Str. 13, Martinsried, D-82152, Germany. E-mail:
[email protected]
Am J Pharmacogenomics 2002; 2 (4)