First workshop on murine models of psychiatric disorders

Colloque organisé par Gene S. Fisch et Jonathan Flint du 14 au 19 avril 2003

Participants

Mark Barad, Sarah-Jayne Blakemore, Michèle Carlier, Peter Driscoll, Michael Fanselow, Gene Fisch, Jonathan Flint, Joseph Gogos, Seth Grant, Jeffrey Gray, Linda Hayes, Andrew Holmes, Maria Karayiorgou, Jacques Mallet, Françoise Muscatelli, Nicholas Rawlins, Ian Reid, Pierre Robertoux, Lawrence Wilkinson, David Wolfer

Compte-rendu

Introduction et un peu d’Histoire
Gene S. Fisch, Ph.D.

James Royce wrote “What is the nature of man? And how shall we come to know it?” the issue we hoped to address in our workshop on murine models of psychiatric disorders. Historical roots of these questions can be found in the 19th century Lamarckian notions on the inheritance of acquired characteristics organism’s lifetime, a notion that persisted into the mid-20th century. Lamarck’s model was surpassed by the Darwin-Wallace theory of natural selection, but Darwin also stated that, “not only has the body been inherited by animal ancestors, but there is a continuity in respect to mind between animals and humans.” Galton elaborated on the model of mental inheritance, while the continuity of animal ancestry with humans gave rise to animal experimental psychology, beginning with Romanes.

Romanes distinguished several categories of thought in humans and infrahuman animals. He proposed that simple ideas (sensations and perceptions) were amenable to both humans and lower organisms, but abstract and general thinking were unique to humans. Later animal experimental psychologists such as Thorndike and Pavlov developed quantitative measures of learning and memory. Others developed techniques to examine visual/spatial learning, or large test batteries to measure cognitive ability. Molecular biologists would develop techniques to generate transgenic mice that were expected to model various human disorders, such as Alzheimer’s disease, mental retardation, depression and anxiety. However, are these adequate models of human psychiatric dysfunction?

Redei et al. (2001) stipulated four criteria: [1] strong behavioral similarities; [2] common etiology; [3] similar pathophysiology; [4] common treatment. However, within the animal experimental community itself, there are marked differences in research strategy, study design and data analysis. Many earlier researchers studied single subjects using within-subject designs. Others remained committed to group designs. Consequently, the two types of experimenters approached data analysis differently. Given the lack of consensus on epistemology within the animal experimental psychology community, is it possible to develop a proper animal model of human psychiatric dysfunction?

Behavioural endophenotypes
Dr Lawrence Wilkinson

A complete account of complex psychiatric disorders cannot be achieved in an animal model. However, progress can be made by modeling parts of the brain and behavioural symptoms. Such ‘intermediate traits’ or ‘endophenotypes’ are more amenable to cross-species analysis with humans, particularly the genetics of brain-behaviour relationships. It may be more useful to map genes to intermediate traits rather than to DSM-IV criteria.

Increasingly, mice are used to model genetic contributions to behavioural endophenotypes. But how useful are they? Ethobiology dictates that mouse behaviour will inform us only about what mice do daily. However, behavioural methods using arbitrary combinations of stimuli and outcomes offer substantial advantages in analyzing specific psychological processes, indexing changes in behaviour across species.

Both approaches are valid. We exploited a specific behavioural function in mice, whereby information about the suitability of eating a novel foodstuff is transferred from one mouse to another via social contact, then used later to guide food choices. We show that olfactory memory formed by this social interaction is sensitive to gene expression implicated in familial forms of Alzheimer’s disease (AD) and neuropathological features of AD in mice. Thus we have a mouse-specific behaviour that may be used to understand and develop therapies for AD.

The behavioural repertoire of mice cannot always be exploited so easily, e.g., to assay ‘attention’ and ‘impulsivity’. These behaviors are noted in many psychiatric disorders, but few attempts have been made to model them in mice. One reason is the erroneous assumption that mice are cognitively challenged. In fact, there is ample evidence to challenge this assumption and we now routinely employ mouse operant tasks taxing aspects of visuospatial attention and impulse control to model genetic contributions to disorders such as autism and ADHD.

Behavioural similarity is not evidence of the operation of identical or similar psychological substrates. We must be aware of the particular behavioural pre-dispositions in mice (e.g., in shaping an operant) and of the valid cognitive differences between species that limit what can be modeled. Evolutionary arguments notwithstanding, mouse and human are linked by a common psychological tool kit whose elements may be put together differently to solve different problems. This is why behavioural methods that remove inherent ethobiological noise to reveal the operation of the core psychological functions are so powerful when attempting to model complex psychiatric disorders in mice.

If Only They Could Speak
Linda J. Hayes

Several criteria have been suggested by which to evaluate the adequacy of animal models of psychiatric disorders: 1) strong behavioral similarities; 2) these similarities are attributable to common sources, both environmental and organismic; and 3) the behavioral and/or physiological pathologies are responsive to similar treatments. I do not believe that animal performances share sufficient similarities to human psychopathologies, in form or function, to constitute meaningful representations in humans. Nor do I believe that an adequate understanding of complex events can be achieved by an additive procedure with respect to facts gleaned in the study of simple phenomena.

Two rules govern the validity of extrapolating from mouse models to human psychiatric disorder. First, events known by way of extrapolation must come from the same materials as those from which this knowledge was derived. In the case of extrapolations to human behavior from observations of animal behavior, this condition appears to be met. The second rule is that events known by extrapolation do not contain significant factors that are absent from the source observations. By this criterion, animal behavior is not an adequate source of information about human behavior, as human behavior contains factors not found in animal behavior, i.e., language. This difference is not insignificant. The articulation of psychological experience by way of verbal action is distinctly human. Hence, an understanding of human psychopathology must take verbal activities and their implications into account, and a murine model will fall short.

The difference between animal and human repertoires lies on a continuum with respect to the substitutional functions of stimuli. Processes of stimulus substitution participate in all circumstances of human and animal behavior in which past actions and their context are relevant to current ones, including all circumstances of learning and remembering. To the extent that behavior-environment relations are expressed in the same principles across species, it may be possible to acquire an understanding of at least some aspects of human behavior by extrapolating from observations of animal behavior. To do so, some dependent measures may have greater credibility than others. I have two suggestions. First, classically conditioned responses are better candidates for extrapolation than are operants. Classical conditioning procedures show greater systemization and standardization than operant conditioning. Operant behavior, being multiply controlled, is a more probabilistic datum. Second, stimulus substitution processes would constitute a valuable focus for modeling research. Substitutional functions are readily established under known experimental conditions, among which classical conditioning figure prominently, and these conditions are readily arranged to produce unusual or maladaptive responding. Hence, a focus on these processes would be useful for constructing murine models of human psychopathologies.

A complex event is often assumed to be combined from simple events: with enough facts about simple events, the complex event may be completely understood. This logic sustains the view that investigations of animal behavior will eventually lead us to understand the complexities of human behavior. A less mechanistic view of “causality” argues otherwise. Causal knowledge is sometimes understood as knowing which factors participate in a given event, along with their interrelations and organization. Understood in this way, a psychological event is not made up of a simple stimulus-response relationship, but is comprised of an interdependent relation of stimulating and responding, taking place in an integrated field involving many factors. From this perspective, any change in the factors in the event field change the entire field, including the interdependent relation of responding and stimulating. In other words, the factors making up a psychological event are not strictly additive. Therefore, no collection of facts about simple events will ever sum to an understanding of the complex event in which all are participating.

Scientific expectations derive from philosophical premises. Mine are such that if an understanding of some particular set of events is the goal, then the strategy most likely to be successful in reaching that goal will be to investigate those events directly. Thus an understanding of human psychopathology is more likely to be achieved if the subject of investigation is human psychopathology, than if a murine model is substituted. The risk in doing the latter is that what one may eventually understand little more than how a tormented mouse behaves.

A meta-analysis of behavioural phenotypes measured in the watermaze
D.P. Wolfer

When tested in procedures originally developed and validated for rats, mice may use problem-solving strategies and be affected by experimental manipulations differently than rats. Place navigation studies of rats in the watermaze have become a popular paradigm to test spatial memory in mice. We have conducted a meta-analysis comprising various mouse strains and different genetically engineered lines and find that water maze performance of mice is limited by behavioral flexibility rather than by spatial memory.

Normal mice explore a variety of strategies while gradually improving their efficiency to find the hidden platform. First orienting toward the sidewalls, they soon scan the pool in an increasingly systematic manner until navigation becomes more and more precisely directed toward the goal. While early learning stages depend on behavioral flexibility, processing and remembering spatial information becomes only important at later learning stages. Using a software algorithm, we categorized training trials into six exclusive strategies according to the predominant swimming strategy: circling, floating, wall hugging, random swimming, scanning, chaining, focal searching, direct swims. Analysis of strategy transitions between trials revealed that, early in learning, mice often perseverate on non-spatial strategies that rapidly lead to a relatively unstable use of spatially focused strategies.

Hippocampal lesions disrupt the flexibility of mice to explore different swimming strategies during early learning, limiting their efforts to wall hugging or floating, and preventing progression to stages where processing of spatial information becomes relevant. Most mutations, including those that disrupt signaling pathways involved in synaptic plasticity or model human cognitive deficits, produce either the same syndrome or a milder form in which progression to spatially directed strategies is only delayed. Strikingly, a selective spatial impairment in the absence of inflexibility during early learning was observed only in two models, in Arg3.1/Arc null-mutants and in mice lacking glucocorticoid receptors in the brain.

Psychological models of Schizophrenia and Autism
Dr. Sarah-Jayne Blakemore

Schizophrenia and autism seem to have biological etiologies and it is possible that the symptoms associated with each are associated with different cognitive and neural abnormalities. Frith proposed that delusions of control and auditory hallucinations, signs of schizophrenia, may be related to deficits in “self-monitoring”. Similarly, social communication problems of autism are thought to result from an impaired “theory of mind, ” while the restricted interests and superior skills seen in autism are explained by low “central coherence.” Exploring these hypothesized deficits at a cognitive and/or neural level might lead to the characterisation of heritable endophenotypes.

In forward models, self and externally produced events are distinguished by comparing the sensory feedback from movement with its predicted consequences. We and others proposed that an impaired predictive mechanism could give rise to symptoms experienced in schizophrenia. If self-produced sensations are interpreted as generated by an external source, then thoughts might be interpreted as external voices (auditory hallucinations) and self-produced movements might be interpreted as externally generated (delusions of control).

We have shown that sensations arising from self-produced movement are attenuated perceptually relative to external sensations. Our functional neuroimaging studies demonstrated that the perceptual attenuation of self-produced events are mediated by lower activity in the somatosensory cortex and anterior cingulate cortex. We also noted that the cerebellum is involved in signalling the discrepancy between the predicted and actual sensory consequences of movement. Attenuation of self-produced stimulation shown in normal controls was not found in psychotics with auditory hallucinations and delusions of control.

Theory of mind proposes that individuals with autism are unable to understand that other people have minds. Experiments have shown that normally developing children acquire facility in mentalising by age 5. By contrast, these abilities develop later than expected in high functioning autistic individuals. In healthy people, mentalising tasks activate three key regions of the brain: the medial frontal lobe, the superior temporal sulcus and the temporal poles adjacent to the amygdala. These regions show reduced activity in people with autism or Asperger’s syndrome

Autism is also characterised by symptoms that cannot be explained by a theory of mind deficit. Some people with autism show savant abilities in music, maths, art, or spatial skills. One theory to explain such superior performance is central coherence, defined as a tendency to focus on the whole rather than the parts of any stimulus. People with strong central coherence are good at integrating material. Those who are weak excel at tasks like proof reading and remembering details. Autism has been characterised by ‘weak’ central coherence. People with autism show weak central coherence on a range of tasks in different modalities. In family studies, parents of boys with autism showed greater piece-meal processing on central coherence tests compared to controls. These results suggest that the extended phenotype of autism may be the cognitive style of weak central coherence with information processing advantages.

Latent Inhibition
Nicholas Rawlins

An abnormality of latent inhibition, wherein prior exposure to a signal (a tone or light) inhibits later learning of an association with that signal, has been considered a useful model of schizophrenia. Latent inhibition (LI), like pre-pulse inhibition, is one of a few models developed in both animals and in humans for which there is a substantial evidence for a common psychological process, in this case a mechanism to avoid learning about irrelevant stimuli. LI is assessed by presenting a signal with no consequence and then subsequently presenting the same signal with an association. What makes LI unusual is that it is one of the few tasks where schizophrenics perform better than healthy controls.

In rats, LI is tested by a pre-exposure phase, a conditioning tone and shock, and finally a measure of drinking in presence of the tone. Pharmacological studies show that low doses of amphetamine disrupt LI in the rat, but increasing doses increase LI. Amphetamine works on the pre-exposure component of LI. When fewer exposures or more conditioning trials are given, it becomes possible to see the enhancing effect of dopamine receptor blockade: haloperidol enhances LI, as do sulpiride and clozapine in normal animals. Lesion studies implicate the entorhinal cortex. In lesioned animals pre-exposure does not impair LI and halperidol reverses the effect and restores LI.

In humans, auditory LI is implemented by a pre-exposure of brief white noise, followed by a training phase in which the ability to make an association is assessed. Subjects are presented with a target (a number) on a screen while listening to a tape of white noise intermixed with nonsense syllables. When there is a white noise, the number on the screen increments. Subjects are asked to say when the number will increment. Pre-exposure to white noise makes this task difficult for normal individuals to acquire, but acute schizophrenics perform better. Haloperidol eliminates this difference; however when drug naïve schizophrenics are compared to drug treated, the drug-naïve non pre-exposed learn as well as normals. Drug-naïve pre-exposed also do not learn the task, in this respect behaving like normals.

In a visual LI paradigm, names of colours are presented in different colours so that, for example, the letters of the word red are coloured green. The task is to say the colour of the letters and not read the word. Negative priming, when pink (in yellow letters) precedes green (in pink letters) proves to be a particularly difficult exercise for normal subjects. Performance on visual LI is sensitive to antipsychotics and to personality type, but schizophrenics are not able to carry out visual LI. Haloperidol improves performance; those with low schizotypal personality scores are more sensitive to the effects of haloperidol (a personality dependent dose response

Variants of tyrosine hydroxylase: a potential susceptibility locus for psychosis
Jacques Mallet

Tyrosine hydroxylase (TH) is the rate-limiting enzyme in the synthesis of dopamine and norepinephrine; regulation of tyrosine hydroxylase activity is critical for controlling the synthesis of these important biogenic amines. The gene may be involved in the pathophysiology of a number of psychiatric disorders and positive associations have been reported for TH gene markers in mood disorders.

Four 5-prime variants are known in human, and two in monkeys (none have been reported in rats). The variants have a large effect on expression. Within the gene, there is a tetranucleotide repeat, variants of which have been associated with psychosis. In addition to simple length variation, the repeat contains sequence variants, including a 1 base pair deletion¹. One percent of normal individuals have the perfect repeat compared to 5% of schizophrenics, suggesting that the repeat could be implicated in the predisposition to the disease.

In order to determine if the repeat has a regulatory role, repeat variants were tested in a luciferase assay. It was found that the repeat has an enhancer like effect. A competition gel-migration assay identified c-jun family proteins binding to the repeat.

Again using a luciferase assay, the repeat element activity was tested in a construct containing the TH gene. The assay worked to repress transcription depending on the number of repeats. By testing the position of the repeat in the intron and the promoter it was found that the effect was not dependent on the position of the sequence.

Mouse models of schizophrenia susceptibility genes
Joseph A. Gogos and Maria Karayiorgou

It is currently unclear how mouse models can contribute to our understanding of complex human complex psychiatric disorders, such as schizophrenia. Two major constraints make generation of mouse models of “a psychiatric disease”unlikely. First, the polygenic nature of human psychiatric disorders make a mouse model unlikely to serve as a model for the entire genetic complexity of the disorder. Second, constraints imposed by the magnitude and pattern of change during human brain evolution. Brain morphological variation and divergence is thought to be a product of changes in the spatiotemporal deployment of regulatory genes and the evolution of genetic regulatory networks. A susceptibility gene may be used in different contexts in the human and mouse brain and a mutation in such a gene may have very different impact in the brain function of mice compared to humans. Sensory modalities used by different species to establish and control behavioral patterns may not be equivalent. Social and mating behavior in mice is driven to a large extent by the olfactory system, and mutations affecting these behaviors may exert their effects through olfactory processing unrelated to the complex genetic components underlying human social behavior.

On the other hand, mouse models of “susceptibility genes” identified in humans may help us understand (through biochemical assays, cellular/molecular biology techniques and neurophysiological/ imaging approaches) the gene function in simple cellular pathways or neural circuits. Animal models focusing on probing well-defined neural circuits affected in schizophrenia rather than overt stereotypic behaviors may be more useful in modeling the disease. One example is sensorimotor gating, which refers to the ability of humans to screen away irrelevant stimuli and focus on objects and ideas of interest.

Recent work on a chromosome 22q11 schizophrenia susceptibility locus is an example of the interplay between human genetics and generation of animal models for candidate susceptibility genes. We focused our efforts on identifying susceptibility genes from the 22q11 locus, since deletions of this region are associated with a high risk for schizophrenia.Recently, we completed a detailed coverage of the 27 individual genes mapping in this interval and have identified a set of genes that are significantly associated with the disease.

We generated mouse models to understand how these genes function and how they relate to schizophrenia. These mouse strains are characterized by: a) Behavioral Analysis: We follow a two-tier exploratory approach: (1) we evaluate the mice using tests designed to model schizophrenia-associated endophenotypes i.e., we test sensory gating and aspects of cognition, (working memory/attention and contextual learning) processes known to be affected in patients with schizophrenia; (2) we attempt to reverse any observed behavioral deficit using neuroleptic drugs. b) Cellular and Molecular Neuropathology: one main goal is to assess gray matter reduction and distinguish between neuron vs. neuropil loss. We assess number of neurons, dendritic and synaptic morphology and neuronal apoptosis. We also evaluate the status of the myelinated white matter and measure the total brain content of selected neurotransmitters. c) Quantification of brain mRNA transcripts via oligonucleotide microarrays: We use microarray technology to identify changes in gene expression in response to the proposed gene disruptions in order to identify individual genes, gene pathways and cellular processes that either serve as direct targets, or interact with the disrupted pathways, to increase susceptibility to schizophrenia.

We found mice with a deletion of the susceptibility gene (PRODH) have deficits in sensorimotor gating. More recent studies using microarray technology in the frontal cortex in these mice revealed increases in the levels of the gene encoding for COMT, which also resides in the deleted region and may control the rate of dopamine breakdown. This finding may indicate a previously undetected interaction between these two genes that may underlie the extremely high risk for schizophrenia associated with the 22q11 deletions.

The potential utility of mutant mice for assessing candidate genes for psychiatric disease.
Andrew Holmes, Ph.D.

Studying transgenic mice has provided novel insights into the neural basis of behavior. Mouse models are used to assess the role of a specific molecule (e.g., neurotransmitter receptor subtype), in cases where other research tools (e.g., selective pharmacological ligands) are unavailable. Fewer studies have engineered mouse mutations in candidate genes for psychiatric diseases, such as anxiety disorders, in part because it has been difficult to identify strong candidates. A polymorphism in the regulatory region of the serotonin transporter (5-HTT) gene (serotonin transporter-linked polymorphic region; HTTLPR) is believed to regulate brain 5-HT function and has been associated with individual differences in trait anxiety/dysphoria. To study how genetically-driven variation in 5-HTT function mediates anxiety, mice were generated with a targeted null mutation in the 5-HTT (htt) gene. Loss of serotonin reuptake in 5-HTT -/- mice causes increased extracellular serotonin, and alterations in serotonin neuronal firing and receptor function. Using behavioral assessment techniques, several laboratories have demonstrated abnormal phenotypes in 5-HTT null mutant mice of relevance to the symptomatology of mood and anxiety disorders. For example, 5-HTT null mutant mice show exaggerated neuroendocrine and adrenomedullary responses to stress, increased paradoxical sleep, increased voluntary consumption of ethanol, and reduced aggression. On a battery of tests for anxiety-related behavior, 5-HTT null mutant mice exhibit abnormalities indicative of increased anxiety-like responses and reduced exploratory locomotion. An anxiety-like phenotype in 5-HTT null mutant mice is consistent with the hypothesis that genetically-driven 5-HTT hypofunction contributes to increased anxiety. Engineering mice with mutations in candidate genes cannot model the full complexity of human disease. However, this approach is of potential utility for understanding gene-gene, gene-environment and gene-drug interactions regulating the etiology and treatment of anxiety disorders.

The Roman High and Low Avoidance Rats
Peter Driscoll

The need for a better understanding of the neurobiological mechanisms underlying anxiety-related disorders has led to the development of various animal models, including those which involve selecting those which show an innate predisposition to express high or low levels of anxiety, i.e. trait (vs state) anxiety models, thus more closely approximating the human condition(s). The most prominent selection programs of this type, all of which involve rats, have been the Maudsley reactive/nonreactive strains (MR/MNR – based on defecation in an open field), the Roman and Syracuse high/low avoidance lines/strains (RHA/RLA & SHA/SLA – based on two-way, active shock avoidance) and the high/low anxiety-related behavior lines (HAB/LAB – based on the elevated-plus-maze). Both similarities and dissimilarities have been noted among the end phenotypes of these selection programs. The Swiss sublines of the RHA/Verh and RLA/Verh avoidance rats show differences in anxiety/emotionality characteristics, locomotor activity and novelty/reward seeking. It has been established that RLA/Verh rats show pronounced stress responses, e.g., freezing behavior, ACTH, corticosterone and prolactin secretion, and increased emotionality in many behavioral tests compared to RHA/Verh rats which are more active copers, show attenuated hormonal stress responses, and display impulsive and novelty (sensation) seeking behavior.

Comparing the selection programs, however, some correlates appear to be more important than others. For example, although SLA rats have larger adrenal glands than SHA rats do, no hormonal differences have been noted between them in response to stressors (unlike the Roman rats). In addition, elevated-plus-maze behavior in Roman rats are inconsistent across laboratories, whereas HAB/LAB rats (also unlike RLA/Verh and RHA/Verh rats, respectively) show response patterns which are the opposite to what is to be expected in the acoustic startle test.

The question of dealing with extremes of behavior, as exemplified by the Roman rats, has also illustrated the superfluousness of including « control » rats in these studies since « normal » rats do not exist, any more than « normal mice ». For example, whereas (French)-Wistar rats resemble RHA/Verh rats in that they markedly increased their voluntary consumption of ethanol following chronic exposure to alcohol vapor (as compared to RLA/Verh rats, which avoid consuming ethanol under all conditions), (Delaware)-Wistar rats resembled RLA/Verh rats in that they showed a reduction in visual evoked potentials (VEPs), in strong contrast to RHA/Verh rats which showed augmented VEPs just as human and cat high-sensation seekers do.

Finally, any gene that influences susceptibility to fear in animals should have a consistent pattern of effects across a broad range of behavioral tests for anxiety. The limited test regimes previously used in genetic mapping experiments and lack of suitable methodologies made it impossible to determine whether previously detected quantitative trait loci (QTLs) specifically influence fear-related traits. The combination of an F2 intercross of inbred RHA/Verh and RLA/Verh rats and utilization of an extensive battery of behavioral tests were used to map QTLs for those traits, and multivariate analyses showed the influence of one locus, on rat chromosome 5, on trait-related behavior. If the neural basis of fear is conserved across species, it is possible that this QTL may have relevance to trait anxiety in humans

Behavioural Inhibition
Jeffrey Gray

Behavioural inhibition is one of the outputs of a Behavioural Inhibition System. It involves the resolution of concurrent conflicts between goals by increasing the perceived aversive consequences of events. It thus underlies the construct of anxiety (taken as activity in a system designed to allow approach to danger). We define it by its sensitivity to anxiolytic drugs. In parallel to the Behavioural Inhibition Systems is a Fight, Flight, Freezing System. This involves escape from or avoidance of danger. It thus underlies the construct of fear. The distinction between anxiety and fear, then, is that they deal with opposing directions of response to threat. In addition to this categorical distinction between the systems in terms of defensive direction we map out a hierarchical behavioural and neural organisation based on defensive distance. Defensive distance can be thought of as perceived level of threat. We see both of the two defence systems as divided into 6 distinct levels. Each level of one system is interconnected with the levels above and below it and with the equivalent level of the other system. This leads us to make a sharp distinction between symptoms and syndromes with respect to threat-related disorders. We propose a typology of such disorders based on a mapping of the primary underlying dysfunction to a specific component of the defence systems. We propose a scheme for differential diagnosis based on the assessment of the presumed underlying neural locus of dysfunction. This high level of proposed discrimination between syndromes, however, is complemented by the expectation of high levels of comorbidity at least in those cases that present for psychotherapy.

Prepulse Inhibition
Jeffrey Gray

Prepulse inhibition (PPI) is the reduction in the amplitude of a startle response, elicited by a high-intensity stimulus (the ‘pulse’), observed when the pulse is preceded by a stimulus of lower intensity (the ‘prepulse’). PPI is usually measured as the percentage change in amplitude, in animals, of a whole-body startle response and, in human beings, the eyeblink reflex. There are many detailed similarities in the parameters that govern PPI in rodents, the most frequently studied animal species, and humans. The neural basis of both the startle response and its modulation by PPI have been described in considerable detail in the rat; and genetic studies have commenced with mice. Thus, data on PPI in human beings can readily be related to behavioural, pharmacological, neuroanatomical and neurophysiological findings in rodents. Interpretation of PPI is usually based upon the concept that it serves to protect processing of the prepulse from disruption by the response to the pulse, although this concept is not well buttressed by evidence. Impaired PPI has been described in several human psychopathological conditions, including schizophrenia (much the most widely studied), obsessive-compulsive disorder and Huntington’s disease. PPI is a function of the stimulus onset asynchrony (SOA) between the onset of the prepulse and pulse respectively. SOA is maximal at about 100 ms. Disrupted PPI is observed in both acute and chronic schizophrenics at SOAs of 30, 60 and 120 ms. It is normalised by atypical neuroleptics at all three SOAs by atypical neuroleptics, which affect both dopaminergic and serotonergic transmission; but only at longer SOAs (fully at 120 ms, partially at 60 ms) by typical neuroleptics, which affect only dopaminergic transmission. These interactions between SOA and pharmacological effects on PPI may prove to be a fruitful avenue of investigation in animal studies, where they have so far been neglected.

Modeling fear and anxiety
Michael S. Fanselow

The requirements of an approach to modeling fear and anxiety depend on what is motivating the conceptual goals of the research. Examples of such conceptual motivation are: disease states, drug effects, function of a particular brain region, function of a particular molecule, or characterization of normal behavior. Then there are different ways to use animals that are contingent upon conceptual motivation. Examples of these are: Animal Models, Animal Assays, and Behavioral Systems.

Our approach is to characterize fear and anxiety in terms of normal behavior using a Functional Behavior Systems approach. The 6 basic assumptions of this approach are: (1) Evaluation of threat is a basic function of the nervous system; (2) Fear, panic and anxiety are parts of a Functional Behavior System that evolved to evaluate threat and coordinate defensive behavior with the level of threat; (3) A major selective force that shaped the evolution of fear, panic and anxiety was predation. Failure to defend against predation has the greatest ultimate cost to biological fitness; (4) the ethological study of responses to threat has emphasized flight distance as a critical determinant of the likelihood defense; (5) typically, flight distance has been thought of as a spatial variable but it is better conceived as a psychological variable, that we call Predatory Imminence; (6) Predatory Imminence determines the likelihood, vigor and topography of defensive responses.

The goals of my talk were to: 1) Describe Fear, Panic and Anxiety in terms of a Functional Behavior System that evolved for defense; 2) illustrate the organization of defensive behaviors into different components or modules; 3) indicate what stimuli activate these modules and ; 4) which behaviors belong to each stimulus; 5) describe a smattering of what is known about the neural circuitry of the different modules; 6) evaluate the relationships between these modules and relate them to anxiety disorders in humans. Specific anxiety disorders relate to distortions of particular defensive behavior modules. Finally, 7) show how past trauma may affect the future expression of these modules.

The extinction of fear conditioning
Mark Barad

The extinction of conditioned fear, as measured by freezing, is a form of inhibitory learning. During extinction, the original association is suppressed by new, context-dependent learning, attributable to plasticity at separate synapses from those that mediate fear acquisition.

Both the acquisition and expression of fear are believed to require NMDA-type glutamate receptor (NMDAR) activity in the amygdala, but considerably less is known about the molecular basis of fear extinction. It is possible that fear extinction depends on NMDAR-independent mechanisms, such as long-term potentiation found in synapses between thalamic afferents and neurons in the lateral amygdala, which depend on L-type voltage-gated calcium channels (LVGCCs) (Weisskopf et al., 1999). From our experiments using injections of two LVGCC antagonists (nifedipine and nimodipine), we show that LVGCCs are required for the extinction but not acquisition of conditioned fear.

While blockade of LVGCCs prevents extinction in a dose-related manner, LVGCC inhibitors fail to prevent conditioned fear acquisition. LVGCC inhibitors do not impair context-dependent learning, and their blockade of extinction probably does not prevent associations with the context of extinction. I show that LVGCCs are not required for the expression of conditional fear: nimodipine and nifedipine interfere neither with the detection of the CS or US nor with the expression of conditioned fear, i.e., freezing. Furthermore, using the effects of the drugs during open-field tests, I show that freezing is behaviorally distinct from reduced locomotion and that the effects of LVGCCs on locomotion neither confounded our freezing scores nor accounted for the persistent blockade of extinction.

A mouse model of Down syndrome
Pierre Roubertoux

The incidence of Down syndrome (DS) is about 1:1, 000 live births, and is the major genetic cause of mental retardation (MR). The syndrome includes a wide range of impairments with highly variable expression. MR is variable but is the most striking feature of the syndrome. Lejeune was the first to prove that DS was caused by an extra copy of chromosome 21. Using partial trisomies it was demonstrated that two regions encompassing the 21q22.2 band were associated with the full DS phenotype. Unfortunately, partial trisomies accounting for less than 1% of DS individuals, too rare for identifying the functions of genes located on chromosome 21, leading researchers to turn to murine models of DS.

Mouse models for DS have the advantage of syntenies between the human chromosome 21 (HSA 21) and the mouse chromosome 16 (MMU 16). By sequencing human 21 chromosome, it was demonstrated that about 80% of HSA 21 and MMU 16 were syntenic. As a result, two families of mouse models are widely used for investigating genes mapped on DCR. . The first refers to full or partial trisomy 16. The second is comprised of transgenic mice over-expressing one gene from HSA 21, some of these genes being carried by DCR (Dyrk-1A, SIM-2, etc…), or other 21 chromosomal regions.

Several limitations are encountered when generating full and partial trisomy 16: MMU16 is not only syntenic with HSA 21 but with regions of HSA 3, HSA 8, HSA 10 and HSA 22. Ts16 are therefore trisomic for chromosomal regions not on chromosome 21. Ts65Dn mice seem to be a better for human trisomy 21 than Ts16 as the triplicate chromosomal region is bounded by App and Mix-2. More to the point, these models do not identify individual gene function and their role in DS. Other limits are encountered with the second family. The hypothesis implicit in the use of transgenic mice with single genes from HSA 21 is that one or more genes may have an effect. Transgenics for individual genes produces less marked effects than those obtained from Ts65Dn. To avoid such difficulties, Smith et al selected short fragments of HSA 21 to produce polytransgenic mice with partial trisomies of the DCR-1. Given previous results with partial trisomies in humans, they selected contig chromosomal fragments covering the DSCR and incorporated them in YACs, each fragment containing between 2 and 10 genes.

Two questions arise with modeling DS in mice. First, how does the behavior of transgenic mice tally with the behavior observed in persons with DS?

Cognitive processes have been relatively well documented for DS: MR is the salient feature, with IQ ranging from 30 to 70, averaging around 50, but does not affect individuals or abilities uniformly. Persons with DS perform normally in simple acquisitions tasks. Likewise, transgenic mice containing different fragments of the DSCR do not differ from controls for simple associative learning tasks, e.g., fear conditioning, number of freezing episodes. In learning, DS children succeed with visual cues but not with spatial cues. No difference appeared between these mice and controls when proximal visual cues were used with the visible platform procedure of the Morris water-maze task; whereas, with the hidden platform version, the mouse is forced to use distal spatial memory. One group of transgenics, F7, took longer to reach the hidden platform. Persons with DS have difficulties acquiring new skills, with most difficulties arising from the persistent use of old strategies to solve new problems. The results with the probe test and the reversal test with F7 mice show a similar lack of the behavioral plasticity needed to succeed when the rules were changed in the water-maze.

Persons with DS show dramatic reduction (12 to 29%) in hippocampus, corpus callosum, and cerebellum. Mice carrying human fragments of HSA21 do not show differences for hippocampal volume. F7 had reduced cerebellar and callosal volumes, however, and E8 mice had smaller cerebellum (with less extent than F7).

The extra chromosomal region carried by the F7 line was associated with behavioral and neuro-anatomical abnormalities that are similar to those observed in persons with DS. The F7 line includes four genes Dscrc, Ttc3, Dscrb, and Dirk1a. These genes are expressed in embryos as early as day 10 post conception. It is difficult to believe that the wide range of traits observed in DS might be the direct consequence of the over expression of these three genes. I assume that these four genes are implicated in DS traits via non-DSCR genes (either located on chromosome 21 or on the other chromosomes). To investigate this hypothesis, a wide proteomic screening using microarrays is in progress.

Behavioral similarities between humans and mice in mouse models of mental retardation
Michele Carlier

Knockout mice or targeted mutations are used to discover how some genes are implicated in human cognitive processes, e.g., for fragile X or Williams syndromes. The study design included behavioral assessment of mice of specific genotypes. It is expected that the mutant mice will behave differently than the non-mutant; more precisely, the learning behavior of mutant mice will be worse. Psychomotor development is supposed to be retarded.

Mental retardation in humans means that the mean score of the children is far from the mean score of the normal developing children. Thus, researchers using animal models should expect not only statistical differences between the mean of the mutants and the mean of the non-mutants but large effect size. I have recommended that statistics evaluate the effect size.

The most difficult aspect regarding behavioral assessment is finding procedures eliciting similar processes in humans and mice. Some tests used with mice are widely known but this does not mean that they are accurate for the testing behavioral hypotheses in humans. I have recommended that a team working with mouse models include (at least during the process of designing the research) a specialist on human genetics and psychology.

Molecular Networks and Cognition
Seth G N Grant

In recent years the application of single gene disruption technologies has led to new insights into the basis of cognitive function and its disorders. Surprisingly, an extensive degree of complexity has emerged. This complexity challenges earlier models on simple mechanisms underlying synapse function and the function of synapses in cognition.

We presented a model for the molecular basis of cognition, using proteomic and bioinformatics approaches. We identified 698 proteins from the postsynaptic terminal of mouse central nervous system synapses. The molecular complexity was simplified by mapping of protein-protein interactions that revealed a scale-free network topology. The synapse network topology is evolutionarily conserved from yeast but enriched with signaling proteins associated with the emergence of multicellularity. The network model predicted the molecular complexity and robustness of synaptic plasticity to mutations and drug interference. Moreover, the network predicted the involvement of proteins in rodent and human cognitive functions. We propose that synapse molecular networks with simple design principles be used as a foundation for the study of cognition and its associated disorders.

1. T)CAT)(TCAT [↩]

---

---