*N. Sousa, Life and Health Science Research Institute, University of Minho, Campus de Gualtar CPII Piso 3, 4710-057 Braga, Portugal. E-mail: email@example.com
Genes and environment are both essential and interdependent determinants of behavioral responses. Behavioral genetics focuses on the role of genes on behavior. In this article, we aim to provide a succinct, but comprehensive, overview of the different means through which behavioral analysis may be performed in rodents. We give general recommendations for planning and performing behavioral experiments in rats and mice, followed by brief descriptions of experimental paradigms most commonly used for the analysis of reflexes, sensory function, motor function and exploratory, social, emotional and cognitive behavior. We end with a discussion of some of the shortcomings of current concepts of genetic determinism and argue that the genetic basis of behavior should be analyzed in the context of environmental factors.
A central tenet of behavioral genetics is that genes and environments are both essential and interdependent regulators of behavior. The discipline exploits natural or artificially created variations in an organism's genetics to dissect out those aspects of behavior that can be ascribed to the function of a particular gene. Indeed, an ultimate aim of behavioral genetics is to understand how the activity of a gene in a specific environmental context can result in the manifestation of a complex behavior.
In general, two complementary strategies can be employed in behavioral genetics (Fig. 1). One, which involves the study of different lines of animals derived from selective breeding, inbreeding or mutagenesis screens and showing inheritance of specific phenotype(s) can serve as starting point for identifying candidate genes. The other approach comprises studies of the behavioral impact of candidate genes by modifying the expression of particular genes using either germline approaches (e.g. ‘knock-out’ or transgenic strategies) or somatic approaches (e.g. antisense and RNAi technology, gene transfer).
Identifying genes that might influence a complex behavioral trait, such as anxiety, is not simple. One approach involves quantitative trait locus (QTL) analysis (Flint et al. 2005). QTLs are chromosomal regions that contribute to the range of variation of a continuous behavioral trait. Over the last decade, QTL analysis has helped map a large number of genetic loci that influence behavior (Doebley et al. 1997; Liang et al. 2003), but achieving sufficiently high resolution remains a nagging challenge. Usually, QTL mapping begins by crossing inbred lines. This means that there is only a limited number of generations available in which genomic recombinations can occur. Mapping resolution is constrained by sample size, an obstacle which can be circumvented by increasing sample size (typically 500–1000). In contrast, outbred lines or strains have the advantage that they reflect numerous generations of recombination. However, the problem that outbred lines present is that a given marker allele may be located adjacent to different QTL alleles and vice versa. An elegant and efficient resolution to this hurdle was introduced by Yalcin et al. (2004). Rather than simply analysing the association between phenotype(s) and a single nucleotide polymorphism (SNP), these authors suggested using haplotypes (sets of closely linked SNPs) of commonly used inbred strains to model the origins of the SNP alleles in the outbred population. By analyzing the association between the phenotype and the inferred ‘ancestral’ haplotype, the authors found a highly significant association between anxiety and a region of chromosome containing Rgs2 as well as two other proximal QTLs (Yalcin et al. 2004). As QTL analysis identifies chromosomal regions that contain more than one gene, it is imperative to perform subsequent analysis of the influence of a candidate gene (within that region) on the phenotype.
This article focuses on the assessment of specific behaviors in rodents. The advents of technologies generating transgenic animals as well as mice with ‘knock-outs’ and ‘knock-ins’ of specific genes have made rats and, in particular, mice, the animal models most frequently employed in biomedical research. We start with some general recommendations for designing behavioral experiments, followed by a brief introduction to some of the most commonly used behavioral tests and end with a critique of some common views on the role of genes in the regulation of complex behavioral processes. Space limitations preclude an extensive survey of the literature and allow only a general overview of the principles and applications of behavioral analysis. By referring to review articles cited, the reader will gain more detailed information on specific topics. Novices in behavioral genetics are referred to Baker (2004) for a concise introduction to the field; a deeper understanding of behavioral experiments in rats and mice may be gained by reference to Crusio and Gerlai (1999), Crawley (2000), Eichenbaum and Cohen (2001), Wahlsten (2001) and Whishaw and Kolb (2005).
Behavioral analysis in rodents – general considerations
Careful experimental design (especially with respect to controls) and standardization of housing conditions, place and time of analysis and the test procedures are essential in behavioral research. As compared with many other biological research methods, behavioral testing is a relatively cheap procedure, and data can be easily acquired. However, it is much more difficult to interpret the data, to standardize the test procedures and to train the experimenters, both in practical and in analytical skills. Numerous factors influence a complex biological process such as the generation of a behavioral response (Fig. 2). Broadly, these factors may be designated as trait, state and technical factors (cf. Wotjak 2004). Trait factors relate to the interplay between genetics (e.g. genetic background) and development (e.g. intrauterine position effect, maternal behavior, stress experience, handling, housing conditions and social hierarchy). State factors include the time at which testing is performed (e.g. circadian time), experimenter (e.g. experience, intra- and interrater variability and bias), characteristics of the animals (e.g. gonadal status, including stage of the ovarian cycle, age, health status and pharmacological treatment) and characteristics of the setup (e.g. construction, surface, hygiene, illumination and test environment). Finally, technical factors include the mode of data acquisition (e.g. automated vs. ‘manual’ observation, calibration and thresholds, choice of behavioral parameters) and data analysis (e.g. distribution of the individual data, differentiation between responders and non-responders and normalization of the data). Difficulties in standardizing all of these factors in different laboratories, and often even within the same laboratory, may contribute to difficulties encountered in reproducing observations of a given behavioral phenotype in different laboratories (Crabbe et al. 1999). Therefore, strict adherence to self-defined standard operating procedures, intensive experimenter training and consideration and reporting of the factors that might influence or determine behavioral performance is of paramount importance (Chesler et al. 2002). Furthermore, comparative analysis of the behavior of distinct inbred strains (each, always obtained from the same breeder) allows judgment of the quality of a given setup and comparison of data between different laboratories. For instance, various C57BL/6 mouse strains show excellent spatial learning, whereas DBA/2 strains are impaired in such tasks. Consequently, these strains might be employed as controls in a newly established behavioral setup that intends to analyze hippocampus-dependent learning. The facts that such positive and negative controls are often forgotten and that the application of standard operating procedures is often neglected contribute to inconsistent reports.
Behavioral studies typically start with a testable question or hypothesis, which dictates the research strategy. Increasingly, however, behavioral analyses of genetically modified animals are hypothesis-free, typically posing the question ‘What is wrong with my mouse/rat?’ Standardized protocols for such experiments (e.g. SHIRPA: Crawley 1999, 2000; Rogers et al. 1997) are now available. As frequent handling affects behavioral performance, these protocols involve analysis of different batches of naive mice by hierarchically structured screening batteries, starting with an analysis of developmental features and general neurological characteristics (including reflexes and sensorimotor capabilities) and concluding with the characterization of emotional and cognitive behavior (e.g. Freitag et al. 2003). From a strategic point of view and in light of the restricted availability of mutant mice, behavioral analysis might alternatively start with more complex tasks (e.g. water maze learning and fear conditioning). If these tests suggest that the genotype contributes to any differences observed, subsequent testing on other batches of animals can be conducted to rule out potential confounding factors, e.g. disturbed sensory and motor capabilities vs. ‘true’ cognitive deficits. We will now provide a brief tour de force through some of the basic and most commonly used tests for the behavioral analysis of rats and mice.
Behavioral analysis in rodents – a practical guide
Most of the behavioral tests described in this article were originally established in rats and later adopted for mice. In this context, it should be noted that mice are simply not just ‘smaller rats’ but rather a different species that occupies a different ecological niche, resulting in different social behaviors, impulsivity and stress-coping strategies (Whishaw et al. 2001; Wotjak 2004). Moreover, there are considerable differences in the performance of various mouse strains (e.g. Crawley et al. 1997). These differences should be considered before designing back-crossing strategies for genetically modified mice or selecting strains to be used in pharmacological experiments (e.g. anxiolytic effects are observed best in a highly anxious status). The following paragraphs contain the quintessence of the behavioral tests that apply to both rats and mice, without an emphasis on procedural differences.
Physical condition, reflexes and sensory capabilities
Before including an animal in any behavioral test, it is crucial to assess its physical condition. General health status requires a strict health monitoring of the animal facilities using serological tests. In addition, the health status of the experimental subjects might be evaluated by examination of the appearance of fur (quality and color) and proportions of body parts, eyes, teeth, paws and genitalia. Animals with infections or any other obvious condition should be excluded from the experiments. Monitoring of simple biometric parameters, such as body weight, serve the latter purpose well. Besides providing information on the efficacy of a given treatment (see Sousa et al. 1999; for example) or the recovery from a surgical intervention, these measurements may also indicate greater or lesser access to food between conspecifics. The latter usually signals behavioral dominance or subordination, which plays an apparently innocuous, but important, role in determining the outcome of the actual test administered.
Reflexes and sensory capabilities are important for appropriate behavioral response (Crawley 1999) (Table 1). Reflexes assess different and, in some cases, individual neural subsystems. Postural and righting reflexes are mediated by the visual system, the vestibular system, surface body senses and proprioceptive senses. Recognition of abnormalities in a particular sensory system requires some expertise (Pellis 1996). For example, when placed on a smooth surface, animals typically rotate toward the side of injury; in contrast, when placed on a rough surface, they tend to prefer rotation in a direction contralateral to the injury.
Table 1. Tests commonly used to measure reflexes and sensory function
Measures response to auditory, olfactory, somatosensory, taste, vestibular and visual stimuli
Typically assessed by approach behavior that, however, could be of short duration.
Mediated by the visual system, the vestibular system, surface body senses and proprioceptive senses
Supporting, righting, placing, hopping reactions are used to maintain a quadrupetal posture. When placed on side or back or dropped in a supine or prone position, adjustments are made to regain a quadrupetal position.
Animals usually restrained during startle testing. This may affect neurochemical and endocrine measures.
Prepulse inhibition procedure allows for testing of multiple sensory modalities.
Because prepulse inhibition depends on the ability to perceive the inhibiting stimulus, varying prepulse intensity or frequency can be used to determine auditory thresholds.
Sensory capabilities to odors, tastes, tactile, visual or auditory stimuli are estimated either by measurements of reflexes and exploratory behavior or by direct assessment of sensory-evoked brain potentials.
Motor behavior and co-ordination
Measures of movement which, besides vocalization and defecation, provide the only direct measure of emotionality and cognition are a primary interest in any behavioral test. As general alterations in locomotion and motor behavior can lead to misinterpretations with respect to ‘higher’ brain functions, animals have to be systematically analyzed for their motor skills (Table 2).
Table 2. Tests commonly used to measure motor function and co-ordination
Measures horizontal and/or vertical movements in a test environment (e.g. open field, home cage).
Automated equipment is necessary (e.g. video tracking; infrared beams). Training of animals is not required.
Walking in rodents can be assessed by analysis of footprint (Carter et al. 1999). Walking starts with movements of the diagonal limb couplets (Clarke 1992). Careful observation of walking patterns can provide crucial information on cerebellar function – making it valuable, for instance, when studying animal models of ataxia. Similarly, turning patterns (the incidence and nature of turning) inform on asymmetrical brain function (Miklyaeva et al. 1995) and serve as an index of recovery after exposure to therapeutic drugs (Freed 1983). Here, it should be noted that turning patterns differ between the sexes (Field et al. 1997).
Besides ‘automatic’ locomotor movements, ‘skilled limb movements’ such as those required for bar-pressing, reaching and retrieving food through a slot (this only for rats) or spontaneous food handling of objects can also be analyzed. All of these movements appear to be much more sensitive to cortical lesions and, regardless of species, extrapolations from rodent models to humans are justifiable (Whishaw et al. 1992). Two commonly used methods for assessing skilled movement are (a) the beam-walking and (b) the rotarod test (Carter et al. 1999). The beam-walking test is based on the fact that a rat or mouse normally walks rapidly along a narrow horizontal pole with its feet placed on the dorsal surface of the beam; walking by grasping the edge of the beam with its digits indicates poor motor co-ordination. The rotarod test evaluates motor co-ordination too. It requires the animal to learn to maintain its balance while placed on a rotating rod. The speed of rotation is gradually increased, as the subject acquires better balancing skills (Le Marec et al. 1997).
Exploratory and emotional behavior
Exploratory behavior is typically assessed in an open field (Holmes et al. 2002). The inner conflict of the animals to avoid novel, potentially dangerous environments and to explore new situations with respect to availability of food, escape sites and mating partners determines their locomotion. When performed under strict ethological conditions, exploratory behavior can be difficult to quantify simply because of the volume of data generated; among others, the parameters usually recorded include the number of home bases, number of trips, kinematics of excursions and returns, number of stops, number and duration of rears, incidence of grooming and duration of trips (Golani et al. 1993). Usually, the animal will select a ‘home base’ (e.g. its initial placement in the open field). It will then pause, rear (place its weight on the hindpaws), turn and groom at this location, before exploring the rest of the open field. Initiation of exploratory behavior is considered to occur when the animal rears and stretches its body with the weight still on the hindpaws to allow an immediate retraction (risk assessment). Eventually, the animal undertakes small ‘trips’ away from the home base, usually along the walls of the enclosure (thigmotaxis). Exploration proceeds with brief and slow outward excursions and rapid returns to the home base. Both the duration of and the distances covered in outward excursion increase in magnitude until exploration of the field is completed. Despite similarities in exploratory strategies used by rats and mice (Drai et al. 2001), there are considerable strain differences in the temporal pattern of exploratory behavior (Ohl et al. 2001).
Today, quantification of open-field behavior is largely automated (video tracking or infrared beam systems) and focuses on horizontal and vertical exploration (e.g. distance moved, speed and number of rearings). Often the open field arena is divided into virtual compartments (e.g. border zone and inner zone). Animals that spend more time in the inner zone are typically regarded as less anxious (Prut & Belzung 2003). Analysis of anxiety-related parameters is valid only if the open field has sufficiently large dimensions.
It should be noted that open-field behavior is subjected to habituation. Thus, re-exposures to the test field normally result in reduced activity and a clear shift in behavior, with more time spent in grooming or sitting still. Interestingly, habituation to the open field does not develop in animals with lesions in the frontal cortex or hippocampus (e.g. Kolb 1974).
As anxiety is not a unitary phenomenon, there is a huge number of tests aimed at measuring different aspects of emotional behavior (Belzung & Griebel 2001; Lister 1990; Millan 2003; Rodgers & Dalvi 1997). Some tests are based on physiological responses to stress (e.g. hyperthermia and plasma levels of corticosterone), but the majority are behavioral in nature. Although an oversimplification, one might classify ‘anxiety tests’ (a) as either unconditioned or conditioned tests, (b) as conflict vs. non-conflict tests or (c) as tests with actual exposure vs. potential exposure to aversive stimuli (Millan 2003). One example of such a classification is provided in Fig. 3. Unconditioned tests include ethologically based paradigms that consider the animals' spontaneous or natural reactions and do not explicitly involve pain or discomfort (Belzung & Griebel 2001). Conditioned tests, in contrast, measure the animals' responses to explicit stressful and often painful events (e.g. electric footshocks) and include aspects of emotional memory (Belzung & Griebel 2001; Fig. 3). Exploration tests are based on the animal's inner conflict of approach or avoid novel situations or stimuli. Interaction tests measure various aspects of approach and avoidance behavior in situations with confrontation to conspecifics as well as ultrasonic vocalization of pups on separation from their mother. Tests on unconditioned acute responses to aversive stimuli (e.g. loud tone, aversive environment and bright light) include measurements of freezing behavior, fear-potentiated startle and ultrasonic vocalization. Furthermore, measurements of defensive behavior are assessed in fear/defense batteries. Conditioned tests include conflict procedures, in which the animal is motivated to execute a distinct behavioral performance (e.g. water consumption following periods of water deprivation), although it will be punished by a mild electric shock upon doing so. Finally, non-conflict conditioned procedures measure the animals' response to an aversive stimulus or situation following classical conditioning procedures.
In the following paragraphs, we introduce some of the most common tasks used to measure ‘anxiety’ (Fig. 3 and Table 3 for a more comprehensive overview).
Table 3. Tests commonly used to measure exploratory and emotional behavior
Exploration- and Interaction-based anxiety tests (elevated plus maze test, light/dark test, open field, social interaction)
Measures approach and avoidance behavior of potentially dangerous environments/stimuli (e.g. open, brightly lit, elevated areas; conspecifics).
Tests are based on the inner conflict between exploration and avoidance of novel situations.
The light/dark (black and white) box test is based on the innate aversion of rodents to brightly illuminated areas (Crawley & Goodwin 1980). The test apparatus consists of a small dark (safe) compartment and a large illuminated (aversive) compartment that are often connected by a tunnel. A decrease in the number of visits of and the time spent in the lit compartment can be interpreted as an increased state of anxiety (Bourin & Hascoet 2003).
The elevated plus maze is one of the most popular tests for the evaluation of anxiety-like behavior in rodents (Hogg 1996; Rodgers & Dalvi 1997). The test has undergone extensive validation and is the test-of-choice in screening the anxiolytic properties of potential therapeutic agents. Like the light/dark box test, it exploits the animal's natural aversion to enter open spaces, i.e. conflicting with its drive to explore a new environment. The maze has a simple configuration. It consists of two opposing ‘open’ arms without side and end walls and two opposing ‘closed’ arms with side and end walls. The whole apparatus has the shape of a ‘plus sign’ and is elevated above floor level. Typically, the number of open arm entries (expressed as a percentage or ratio of total arm entries) and the time spent on the open arms (expressed as a percentage or ratio of total time spent on the open and the closed arms) are used to derive the state of anxiety. In addition, the elevated plus maze yields data on locomotor activity, vertical activity and exploratory behavior and provides indices of risk assessment and decision making (Rodgers & Dalvi 1997). Recent studies have shown that age (Bessa et al. 2005), familiarity with the experimenter (van Driel & Talling 2005) and the gradient of luminosity between open and closed arms (rather than the absolute illumination strength) (Pereira et al. 2005) have an important influence on the outcome of the elevated plus maze test.
Novelty-induced suppression of feeding. This test appears to be particularly sensitive to prior stress experiences and to treatment with antidepressant compounds. It typically measures consumption of a familiar food in a novel environment (e.g. an open field) (Rex et al. 1998). It is based on the fact that animals have first to form a concept of safety in the novel environment (i.e. to habituate) before displaying ingestive behavior. A priori differences in food consumption (i.e. metabolism) and food reward (i.e. activity of the mesocorticolimbic system) can, however, confound interpretation of the data obtained in this test.
In the Vogel Conflict Test (punished drinking test), water-deprived animals are allowed access to a water bottle under experimentally controlled conditions but receive a mild electric shock to the tongue after a certain number of licks. The more shocks an animal accepts (i.e. the more it drinks) the less anxious it is considered to be. This test evaluates approach behavior rather than avoidance behavior, an important consideration in light of the hypothesis that anxiety primarily develops (and can be assessed) if an aversive situation has to be approached (Gray & McNaughton 2000).
The burying test is based on the animal's defensive behavior in response to a threatening or noxious object (De Boer & Koolhaas 2003). In its simplest form (Shock-probe test), the test animal is placed in and allowed to accustom to a box with sawdust on its floor. After a brief period (acclimatization), a shock-delivering probe is inserted into the box through a hole in the wall. The animal receives a brief electric shock when it investigates the probe. Its first response is to withdraw from the object and then to investigate it cautiously and finally, to use its forelimbs to bury the probe under the sawdust. Measures with respect to the animal's learning of the probe include the number of investigations of the object, the duration of time spent burying the object and the depth of the sawdust used to cover it. Various modifications have been made to this test, including burying objects that deliver noxious sounds or odors. Interestingly, animals will also bury other objects in the vicinity of the offending object, indicating that its responses can be (secondarily) conditioned to other objects. Defensive burying has been used to examine the effects of aging on emotionality and also to examine the effects of potential antianxiety agents (Pinel & Treit 1983).
Fear conditioning procedures are typically employed for studying cellular and molecular correlates of consolidation and extinction of aversive memories (LeDoux 2000). Pairing of an initially neutral stimulus (such as a light or a tone) with an electric footshock forces the animals to show fear responses (e.g. freezing or startle responses) on re-exposure to that stimulus. Repeated non-reinforced re-exposure eventually leads to a decline in the respective fear reactions (i.e. extinction).
Forced swim test. In the common version of this stress-coping task, animals are placed in cylinders filled with water to a depth such that the animals cannot reach the ground (note that the diameter of the cylinder and the depth and temperature of the water are critical determinants of the behavioral responses). Rats and mice cope with that situation by immobility (floating), swimming or escape behavior (struggling). Repeated testing leads to a shift in the coping strategies from active to passive coping. Long-lasting floating is typically considered as an indicator of behavioral despair. In fact, the prolonged floating on a second exposure seems to be particularly sensitive to a variety of classical antidepressants (Petit-Demouliere et al. 2005; Porsolt 2000). However, despite the apparent face and predictive validity of the forced swim test, the behavioral response of the animals must be carefully interpreted, as it is obviously confounded by learning processes. In fact, if the animals recognize that there is no escape, adoption of passive coping strategies appears to be the appropriate response. Furthermore, the fact that a pharmacological compound can decrease floating, mimicking the effects of serotonin or noradrenaline re-uptake inhibitors, only points to antidepressant-like (but not antidepressive) effects of that compound. The forced swim test is often combined with more elaborated tests (e.g. chronic mild stress and learned helplessness paradigms) aimed at inducing depression-like symptoms.
A word of caution seems appropriate before concluding this synopsis of how emotional behavior may be measured. As mentioned before, most of the tests of basic emotionality measure avoidance of aversive stimuli or situations. State anxiety, however, develops if animals are forced to approach aversive stimuli. If the motivation to approach such a stimulus is small, avoidance behavior can, at best, be interpreted as cautiousness rather than anxiety. Each test of ‘anxiety’ should therefore be validated pharmacologically (i.e. treatment with anxiolytic or anxiogenic compounds and in every rat or mouse strain to be studied in long-term screening). One can only be certain that a distinct behavioral pattern is an expression of anxiety-related behavior if appropriate responses to a ‘prototypic’ pharmacological agent are observed (Griebel et al. 2000). It should be noted that treatment with anxiolytic compounds might simply ‘abolish’ the anxiogenic consequences of the stress associated with administration (e.g. intraperitoneal injection) of the test compound. This confound can be avoided using multidose designs, including vehicle-treated controls, and a group of untreated animals to provide a ‘basal’ reference point.
Species-specific stereotypic behaviors
Most rodents show stereotypic behaviors (Table 4). They can be species-specific with respect to the rules, sequence and duration of occurrence of each episode of behavior, due to their different natural ecological niches. Examples of stereotypic behaviors include grooming (Berridge 1990), nest building (Kinder 1927), play (Pellis & Pellis 1998) and social behavior (Blanchard et al. 1997), sexual behavior (Dewsbury 1973) and maternal behavior (Meaney 2001). Video recording greatly facilitates qualitative and quantitative analysis of these behaviors under laboratory conditions. Nest building (which depends on an intact limbic system and medial frontal cortex; Shipley & Kolb 1977) and grooming (which involves several levels of the neuraxis, including the cortex, striatum and cerebellum; Berridge 1989) both serve as markers of an animals' emotional status (Kalueff & Tuohimaa 2004, 2005).
Table 4. Tests commonly used to measure species-specific behaviors
Documents successive behavioral acts in species-typical behavior. Provides description of their order or syntax
Video recording is highly recommended. Evaluates grooming, foraging and diet selection, eating patterns, sleeping patterns and nest building. Detailed analysis of grooming patterns might be indicative of the anxiety status of the animals
Social behavior (i.e. all activities that influence, or are influenced by, other members of the same species) is an important aspect of a rodent's life. It includes all sexual and reproductive activities as well as aggressive behavior (e.g. Grant 1963) (Table 5). Social behavior is generally considered as a highly complex function, requiring recruitment of, and interactions between, multiple neural circuits, with endocrine hormones and pheromonal cues playing a key role in their co-ordination and execution (e.g. Insel & Fernald 2004; Vanderschuren et al. 1997). The social status of an animal (dominant vs. submissive) is a critical determinant of its behavioral responses in various test situations (Lathe 2004; Strekalova et al. 2004). Aggressive behavior is used to establish social hierarchies and to defend territories. Patterns of aggressive behavior are also usually distinctive in male and female animals, and aggressive behavior in laboratory rodents is widely used as a correlate of human aggression (Blanchard et al. 1989). Aggressive behavior is mediated by ventral limbic circuits in both rodents and primates, and altered aggression is believed to be indicative of morpho-functional disruption of these circuits. Classical tests of social behavior in rodents involve certain forms of free interaction in group-caged animals, often in a cage with several interconnected chambers (e.g. Lubar et al. 1973). A less-natural version of such a test involves the confrontation between two animals in a novel environment (e.g. Latane et al. 1970). The social interaction test, for instance, is based on multiple interactions between conspecifics in a familiar or unfamiliar arena (File & Seth 2003). Observations in such tests include scoring of interanimal contact times (see Grant & Mackintosh 1963), vocalizations (e.g. Francis 1977) and urine marking (e.g. Brown 1975). The duration of mutual exploration serves as a measure of anxiety-related behavior (File & Seth 2003).
Table 5. Tests commonly used to measure social behavior
Social interaction test
Animal is confronted with an unknown conspecific in a novel environment. Measures latencies, frequencies and duration of social behavior (approach, avoidance, exploration, aggression, submission, sexual behavior).
Simple, rapid, requires no training and only minimal equipment. Can be used for measuring anxiety-related behavior. Characterization of the social status might be essential for the interpretation of emotional behavior.
Removal of the pubs from the nest allows measurement of pup retrieval.
Maternal care is another aspect of social behavior with far-reaching implications for the emotionality and stress-susceptibility of the offspring (Levine 1957; Meaney 2001; Wotjak 2004). Therefore, it is an important consideration in the interpretation of data obtained from studies with selective breeding strategies or genetic manipulations. The contribution of epigenetic factors, such as differential maternal care, to the behavioral phenotype can be estimated from the results of cross-fostering studies. One way to maintain the influence of maternal factors at a constant level is to ensure that both the ‘knock-out’ mice and the wild-type controls derive from the same heterozygous breeding pairs.
Cognition, in a broad sense, refers to learning and memory processes. Consequently, tests on cognitive behavior analyze the capability of an animal to form representations of the outer world that lead to altered behavioral performances on subsequent confrontation with distinct stimuli and situations. Typically, animals have to be motivated to participate in a learning task and to display memory-related behavior. In this context, one distinguishes between positive (e.g. provision of food to food-deprived animals) and negative reinforcers (e.g. punishment by a mild electric shock). Tests of recognition memory form a third category, as they are based on the innate motivation of an animal to explore novel objects, odors or conspecifics. Such tests do not require the application of exogenous reinforcers and are, thus, fairly resistant to alterations in reward systems and pain circuits.
In the past 20 years, certain cognitive tests have become leading paradigms for the study of memory processes that rely on distinct brain structures. The hippocampal formation, for instance, co-ordinates many independent sensory features of an episode. This capability is used for learning tasks that require identification of the spatial and temporal relationships between distinct stimuli of different modality, texture or shape (e.g. spatial learning tasks, contextual fear conditioning, trace fear conditioning and identification of conspecifics). Moreover, tests in which animals have to execute (active avoidance) or not make any behavioral response (passive avoidance) also involve the hippocampus formation. Relearning processes that require the suppression of an originally formed memory have meanwhile been shown to be particularly sensitive to alterations in hippocampus function. It should be noted that the hippocampus formation is only critical for memory recall over a limited time period (typically 3–4 weeks). During the process of system reconsolidation, it seems that neocortical structures serve as the permanent stores of memories that were initially hippocampus dependent (Wiltgen et al. 2004).
Of the other brain structures, the amygdala complex is essential for the formation (and, possibly, storage) of emotional memories (LeDoux 2000), and its involvement in learning and memory is typically assessed in fear conditioning paradigms. The cerebellum is implicated in eyelid conditioning as well as in motor learning (Linden 2003), the dorsal striatum in stimulus-response habit learning and in motor learning (Costa et al. 2004), the rhinal cortices in object recognition (Ennaceur & Aggleton 1997) and the prefrontal cortex in tasks requiring an intact working memory and in memory extinction (Dalley et al. 2004).
The following paragraphs are intended to provide a brief introduction to some of the most commonly used learning task tests (Table 6 for a more comprehensive overview):
Table 6. Tests commonly used to measure cognitive behavior
Memory is assessed by the duration of withholding the response following training.
Tests depend on an intact hippocampus formation.
Spatial learning (Morris water maze, Barnes maze, T-maze)
Animals are taught to navigate in a maze to obtain positive reinforcement (e.g. food) or to avoid negative reinforcement (e.g. escape from water). Measures choice/escape latencies, response accuracy and selectivity of searching.
Animals have to be trained over several days. Requires motivational manipulation. Requires prominent sensory and motor capabilities Once learning has occurred, the procedure allows for determination of various forms of memory (e.g. working or reference memory), and ability to perform cued discriminations.
Measures duration and frequency of conditioned responses (e.g. freezing, fear-potentiated startle, eyelid closures in fear conditioning tasks).
Animals learn the associative relationships between discrete elemental or configural stimuli, with one stimulus being initially ‘neutral’ (conditioned stimulus, e.g. tone, light) and the other (unconditioned stimulus, e.g. electric shock, air puff) being able to evoke an unconditioned response.
Could be learned within a single trial (fear conditioning), within a few trials (fear- potentiated startle) or requires more intensive training (eyelid conditioning).
Depends on the amygdala complex (fear conditioning, fear-potentiated startle), the cerebellum (eyelid conditioning) and the hippocampus (contextual fear conditioning, trace fear conditioning).
Might be confounded by non-associative learning processes.
Operant (instrumental, thorndikian) conditioning
Measures the accuracy of performing a conditioned response (e.g. pressing a distinct lever, nose-poking into a hole).
Animals acquire new behavioural patterns, which enable them to alter the frequency of their exposure to stimulus events. Whereas in classical conditioning subjects learn about relations between signal and significant events such as food or danger (stimulus–stimulus association), in instrumental conditioning they learn about relations between their behaviour and those significant events (response-stimulus learning).
Requires intensive training and positive reinforcers.
Animals may acquire alternative strategies (e.g. side bias) that could be assessed by sophisticated statistical tools (e.g. signal-detection theory).
In passive avoidance tasks, animals learn not to leave an elevated platform (step-down avoidance) or not to avoid a mild electric footshock by escaping from a lit compartment into a dark compartment (step-through avoidance). Other than spatial learning tasks, passive avoidance tasks are readily acquired, even within a single trial. This has the advantage that the time-points of initiation of memory acquisition and consolidation are well defined, thus enabling the analysis of memory-related changes in gene and protein expression.
Active avoidance tasks are more laborious, as they require repeated training sessions. In a classical design, animals are placed in a two-compartment setup (e.g. shuttle box). On presentation of a visual or acoustic stimulus in one of the two compartments, animals have to learn to escape to the other compartment within a limited time period to avoid a mild electric shock. This task has been successfully employed for studying categorical learning in gerbils (Ohl et al. 2003).
Spatial learning is typically assessed in relatively complex mazes. In ‘dry’ mazes, animals have to memorize escape routes and the location of refuges to avoid an aversive situation (the hole-board used in the Barnes maze) or to receive a food reward (radial arm maze). Alternatively, ‘wet’ mazes force the animals to locate a hidden platform to escape from a pool of water (water maze).
The radial arm maze consists of a central box or platform with several radially extending arms (Jarrard 1983; Olton et al. 1979). The location of a distinct arm is either fixed in relation to extra maze cues or marked by a cue on the arm (e.g. texture, color and illumination at the end of the arm). One or more of these arms are baited with an accessible food reward. The task requires the subject to learn the location of the food over a number of test days. This evaluates an individual's ability to form a set of different memory categories: (a) procedural memory (knowledge that some of the arms are baited and that food can be obtained from a baited arm only once during each training trial), (b) working memory (the ability to avoid repeated visits to a baited arm during a given training trial) and (c) reference memory (knowledge about the localization of the baited arms over subsequent training trials).
The Morris water maze task exploits the remarkable ability of rats (and less so of mice) to swim. This type of test has gained popularity, mainly because it (a) is easy to perform, (b) avoids the problems of dry mazes with respect to olfactory guidance (e.g. by urine marks) and (c) is not confounded by the need for food deprivation (D'Hooge & De Deyn 2001; Gerlai 2001; Lipp & Wolfer 1998). The water maze consists of a circular pool filled with opaque water. A platform of approximately 10 cm in diameter is placed in the pool such that its surface is either visible (control for visual perception and motivation to escape from the maze) or submerged under the water surface (hippocampus-dependent learning). The location of this platform remains fixed throughout the training and test sessions. The animals have to learn to navigate to the platform from different, pseudo-randomly chosen starting positions, using spatial mapping (or distal cue) strategy. A decrease in escape latency primarily relates to procedural memory (e.g. learning that there is no other escape from the maze than climbing on the platform). Hippocampus-dependent memory can only be assessed by video tracking analysis, if the time spent and distance moved in the target quadrant as well as the selectivity of searching (i.e. number of crossings of the former platform position) are assessed during a probe trial (removal of the platform, free swimming for 30–90 seconds). Alternatively, the platform can be moved to a new location (reversal training) and perseverance in searching at the former platform position and learning of the new platform position are measured. Despite their apparent simplicity, water maze tasks by no means exclusively measure hippocampus-dependent learning. Often, behavioral performance is confounded by altered emotionality and, thus, altered stress-coping strategies. Animals (in particular mice) might regard the task as a repeated forced swim test, consequently adopting passive (i.e. extensive floating) or active coping strategies (escape behavior along the walls; Lipp & Wolfer 1998; Wotjak 2004). Several modifications of the water maze task have been designed to circumvent some of these caveats and therefore to supposedly serve as better measures of hippocampus-dependent learning, as exemplified by one trial place learning (successfully applied in rats; McDonald et al. 2004) or discriminatory place learning (successfully used in mice; Arns et al. 1999). In the latter task, the water tank contains two identical floating platforms which are visible to the animals. One platform is maintained at the same position through all training sessions, whereas the position of the other is changed from trial to another. The latter platform sinks as soon as the animal mounts it, whereas the stable platform is designed to remain at the surface even after the animal climbs on to it. Mice with intact hippocampus formations rapidly develop a preference for the stable platform, whereas those with impaired hippocampus-dependent learning perform on the basis of chance. Importantly, escape latencies in animals with intact and impaired hippocampal function reach an asymptote within 5 seconds. Therefore, the discriminatory water maze task does not only measure hippocampus-dependent learning throughout training (preference score) but also reduce the emotional load of navigation in an unpleasant environment.
Tests of recognition memory are easy to perform. They are based on the prolonged exploration of novel stimuli as compared to that of familiar stimuli (for reviews see Steckler et al. 1998a, 1998b, 1998c). The number of exploratory bouts and the duration of investigation serve as measures of recognition memory. As information of objects or odors are rapidly acquired, social recognition tasks provide more reliable measures (as compared with other stimuli, the investigation of conspecifics can last 10 times longer). The general procedure for performing recognition tasks is the following: during a sampling phase, animals are exposed to a distinct stimulus (toy object) for the first time. After an intertrial interval ranging from minutes (when measuring short-term memory) to hours or days (when measuring long-term memory), animals are re-exposed to the original stimulus as well as a novel stimulus (test phase). This ‘discriminatory’ version of the task, in which animals have to discriminate between a familiar and an unfamiliar stimulus, serves as an internal control for non-specific drug and habituation effects on exploration (because exploration of the two stimuli can be assessed and compared within the same animal at the same time-point). Tests of recognition memory have high ethological relevance, are less stressful to the animals and can be repeatedly applied to the same animal (with different sets of stimuli). However, it should be noted that rats and mice differ remarkably in their memory spans. Moreover, housing conditions (group vs. singly housing) can significantly affect the memory performance (Kogan et al. 2000; Teather et al. 2002).
In classical conditioning tasks, animals learn about relations between a signal and a significant event such as danger or punishment (stimulus–stimulus association). In a typical fear conditioning task, a light or tone of 20–30 seconds duration coterminates with a mild electric footshock of 0.5–2 seconds duration (delayed conditioning protocol). As a consequence, animals form an association not only between the light/tone and the shock (elemental conditioning) but also between the conditioning context and the shock (configural or background contextual conditioning). Both elemental and contextual conditioning depend on the amygdala complex. Contextual conditioning additionally involves the hippocampal formation. The delayed conditioning protocol can be modified by introducing a temporal gap between the end of light/tone presentation and the onset of the shock (trace fear conditioning). Under these circumstances, also elemental conditioning requires an intact hippocampal formation (Shors 2004).
To differentiate between elemental and contextual memory, memory for the tone–shock association is typically assessed by re-exposure to the light/tone in a novel test environment. Contextual memory, in contrast, is measured in the conditioning chamber without light/tone presentation. Several authors recommend the use of discriminatory contextual conditioning paradigms, because these tasks seem to provide more specific measures of hippocampus-dependent learning (for review, see McDonald et al. 2004). Another modification of the contextual conditioning task is based on the conditioning context pre-exposure facilitation effect (for review, see Rudy et al. 2004). Rats and mice typically require some time of context exploration, before a conjunctive representation of the different context features can be formed by the hippocampus. Consequently, animals fail to form a contextual memory if they receive a footshock immediately before their placement in the conditioning chamber (immediate-shock deficit). However, this deficit can be overcome, if the animals were previously exposed to the conditioning context, allowing them to recall the context representation just before presentation of the shock.
It has been recently suggested that, at least in mice, fear conditioning not only leads to the formation of a memory for the stimulus–shock association but also results in non-associative encoding of information about the intensity of the aversive encounter (sensitization). If the conditioned stimulus (above a certain minimal threshold) per se can elicit an innate fear response in naive animals, sensitization is likely to potentiate this fear response following the conditioning procedure (Fig. 4a–c; Kamprath & Wotjak 2004); it is important that shock intensities are sufficiently moderate so as to avoid confounding by non-associative learning processes (Fig. 4e; Kamprath & Wotjak 2004).
Fear conditioning tasks are frequently employed for the analysis of memory extinction. Today, it is commonly accepted that reductions in memory performance upon repeated, non-reinforced stimulus presentation lead to the formation of a stimulus–no punishment association (safety learning) that inhibits the expression of the originally formed stimulus–punishment association (Myers & Davis 2002). However, considering that the conditioning procedure also sensitizes the animals' innate fear responses to certain stimuli (Fig. 4), there are at least three additional ways in which to account for extinction: (a) non-reinforced stimulus presentation might partially reverse plastic changes triggered by the stimulus–punishment association, (b) desensitization of the animals might occur and (c) the stimulus–response circuit responsible for the innate fear response to the stimulus may show habituation. Interestingly, extinction of the freezing response of conditioned mice seems to primarily relate to habituation processes (Fig. 4d) (Kamprath & Wotjak 2004; McSweeney & Swindell 2002). Therefore, future studies on extinction of conditioned fear need to more carefully differentiate between the four different processes of extinction.
Avoiding misconceptions about genetic determinism
Hundreds of studies on genetically modified animals published during the last decade have causally associated specific behavioral responses with mutations of distinct genes. Drawing from such studies, both the scientific and the public media arrived at rash conclusions about the genetic basis for dullness and super-brains, Angsthasen and bravehearts, lady-killers and lone wolfs. However, because a particular behavior results from the orchestration of a multiplicity of motivations and processes, which often oppose each other (e.g. curiosity vs. cautiousness), and because of the interdependence between the genome and the environmental factors, behavioral traits are mostly polygenic; genetic modifications of a single gene at most influence the range of variation of a single behavioral component. Moreover, multiple genes interact with each other to contribute to the range of variations seen in a complex behavioral trait. There are three genotypic components of variability: (a) additive effects of alleles at all loci, (b) effects of dominance at each locus and (c) the result of interaction between loci (see Fuller et al. 2005 for review). Elaborate breeding schemes, together with polymorphism analyses, allow the characterization of different genes or gene ensembles that may be responsible for quantitative differences in continuous behavioral traits (Flint et al. 2005). In light of these considerations, behavioral studies on knock-out mice in fact only analyze the plasticity of the genetic ensemble to cope with a gene deletion rather than the physiological significance of that gene, especially because of the likelihood that compensatory mechanisms come into play after mutation/deletion of a specific gene (Routtenberg 2002).
The complexity of behavioral responses in the context of different motivational or emotional processes is easily appreciated when one considers that two stable behavioral traits can be distinguished in rodents that are confronted with aversive situations: active-coping and passive-coping strategies. These two traits even seem to exist in inbred strains and might, thus, relate to differences in maternal behavior and the social status. Passive coping is characterized by immobility, reduced exploratory drive and avoidance of dangerous situations. Active coping, in turn, is characterized by the opposite phenotype. Genetic modifications might interfere with the adoption of one of these phenotypes, thus skewing the distribution of actively and passively coping animals among the sample mutant population. Importantly, the mutation might alter the behavior in a subset of mutant mice only. One needs to be aware that the most commonly used statistical tests (t-test or analyses of variance), however, assume a normal distribution of the data and may therefore fail to reveal significant genotype differences. Another issue is the preferred presentation of data as group means and standard deviations, rather than the (additional) depiction of individual data. Performance of pretests to measure different behavioral dimensions and to allow ‘preselection’ of different groups of mice (e.g. dominant vs. submissive mice; Strekalova et al. 2004) would add to the interpretational value of published data.
The behavioral performance of a given rat or mouse strain is determined not only by its genetic ensemble (nature) but also by environmental factors (nurture), e.g. pre- and postnatal maternal influences (Francis et al. 2003; Zaharia et al. 1996; for review, see Meaney 2001), phases of food restriction (Cabib et al. 2000) or enriched housing conditions (Feng et al. 2001; Rampon et al. 2000). To estimate the impact of genetic and environmental factors or the interaction between them, we need to test animals of different genotypes under conditions that allow slight, but measurable, modifications of the environment (e.g. different drug concentrations, light intensities or water temperatures). Application of ‘norm of reaction’ graphics (see Fuller et al. 2005) may also help in the interpretation of experiments in which environmental modifications are used to ascertain the role of genes in the expression of a particular phenotype.
Complex genotype–environment interactions clearly play no small part in accounting for the variability of the results obtained from identical experiments conducted in different laboratories (Crabbe et al. 1999). In summary, a distinct phenotype can only be unequivocally ascribed to a particular genetic modification if the phenotype persists under varying environmental conditions within one's own and other colleagues' laboratories. This goal can be achieved by establishing replicable behavioral measures in cross-laboratory studies (Kafkafi et al. 2005).
What is wrong with my animal – a summary
The objective of this review was to provide a short comprehensive guide to behavioral analysis in rats and mice. The importance of standard operating procedures and the experimenter was underscored and the numerous trait, state and technical factors that may influence the behavioral performance and which must therefore be considered in designing behavioral experiments were highlighted. We recommend the validation of newly established setups by performing comparative analyses with inbred strains of a known behavioral phenotype. We suggest that any observations made in one laboratory must be reproducible within and beyond the institution in which the original studies are conducted. For analysis of emotionality, we emphasize the need for applying more than just a single task and the prerequisite validation using pharmacological manipulations. While highlighting some of the most commonly used behavioral tests and standard protocols, the authors have not intended to discount other sophisticated approaches that have helped in, and will continue to contribute to, the quest for a better understanding of the neural basis of behavior.
Hence, what is the final question addressed in behavioral genetics? Unlike the researchers in Douglas Adams'‘The hitchhiker's guide to the galaxy' (1979), we happen to know our ultimate question: rather than asking ‘What is a certain gene good for?’ or ‘What gene is responsible for a certain behavioral trait?’, our question is ‘How do gene ensembles interact with the environment to determine behavior?’ Using the tools of genetics and behavioral biology in rats and mice, we are well geared to understand a little bit more about ourselves.