Genetic independence of mouse measures of some aspects of novelty seeking

Abstract

High novelty seeking is a complex personality attribute correlated with risk for substance abuse. There are many putative mouse models of some aspects of novelty seeking, but little is known of genetic similarities among these models. To assess the genetic coherence of “novelty seeking,” we compared the performance of 14 inbred strains of mice in five tests: activity in a novel environment, novel environment preference, head dipping on a hole-board, object preference, and a two-trial version of the spontaneous alternation task. Differences among strains were observed for all tasks, but performance in any given task was generally not predictive of performance in any other. To evaluate similarities among these tasks further, we selectively bred lines of mice for high or low head dipping on the hole-board. Similar to results from the inbred strain experiments, head dipping was not correlated with performance in the other measures but was genetically correlated with differences in locomotor activity. Using two approaches to estimating common genetic influences across tasks, we have found little evidence that these partial models of novelty seeking reflect the influence of common genes or measure a single, unified construct called novelty seeking. Based on the substantial influence of genetic factors, ease of implementation, and relative independence from general locomotion, head dipping on a hole-board is a good task to use in the domain of novelty seeking, but multiple tasks, including others not tested here, would be needed to capture the full genetic range of the behavioral domain.

The idea that an individual’s preference for new and exciting stimuli describes some core aspect of personality is a central theme in many personality scales. Novelty seeking (1), sensation seeking (2, 3), and impulsivity/behavioral approach have been described as extensions of Eysenck’s personality model (4). They are typically considered as heritable dimensions of human behavior that are associated with an underlying biological substrate such as monoamine level or turnover, which provides a plausible basis for natural selection of the trait. Polymorphisms in human genes involved in monoaminergic, and particularly dopaminergic, signaling are predictive of the degree of novelty seeking (5, 6), although some associations are controversial (7). Whichever personality construct is considered, a well replicated finding is that individuals characterized as high novelty seekers are more likely to abuse or be dependent on drugs of abuse than are low novelty seekers (see ref. 8 for a review).

Animal models of novelty seeking rely on the overt expression of a behavior and are often described by the degree to which the animal either explores a novel stimulus (such as an open field, e.g., ref. 9) or demonstrates a preference for novel as compared with familiar stimuli (such as novel environment preference, e.g., ref. 10). Individual differences in some aspects of novelty seeking are also inferred from tasks assessing neophobia (or preference) for novel objects, food, odors, flavors, drinking solutions, and access to social interaction with conspecifics, among many others. All of the above tasks involve competing tendencies of the animal to explore and avoid. A human temperament and personality inventory (1, 11, 12) distinguishes subscales, including novelty seeking, harm avoidance, and reward dependence. Novelty seeking is uncorrelated with harm avoidance and reward dependence in this scale. Other domains of human novelty seeking such as impulsivity have traditionally been measured in rodents by using operant procedures (e.g., ref. 13). It is unreasonable to expect an animal model for a complex human trait to capture all aspects of that trait (14, 15). Rather, certain key features can be targeted by multiple, partial animal models.

Many rodent tasks conflate a reaction to a novel situation with a tendency to seek novelty. Thus, these simple models of novel environment reactivity do not allow a dissociation of general activity that exists under all situations, including the home cage, from activity that is specific to a task. Furthermore, they are likely confounded with task-specific stress that could interact with novelty seeking tendencies. For example, Piazza et al. (16) demonstrated that rats that show higher activity in a novel environment also have higher levels of corticosterone during the behavioral assessment, but whether corticosterone is a cause or an effect of locomotion cannot be determined. Tasks that depend on a preference for novel over familiar stimuli could be considered a more direct reflection of novelty seeking, because a comparison can be made to chance levels of preference for novel relative to familiar stimuli within an animal. These tasks, though, are hindered by more complicated and, consequently, long-term memory-dependent experimental procedures and often reintroduce the stress/novelty seeking confound during the preference assessment. Interpretation of partial rodent models of “novelty seeking,” therefore, must include consideration of memory (in novelty preference tasks) and stress. In addition, some of the same tasks (e.g., exposure to an open field) are used to model anxiety-like states (17), and induction of an anxious state may underlie neophobia in some circumstances (18).

Mouse genotype is known to exert a strong influence on putative measures of novelty seeking as well as many other tests of emotional-like behaviors. The existence of many inbred mouse strains, each composed of genetically identical individuals, allows comparison of the influence of genes on multiple behavioral endpoints, because naïve mice can be used for each behavioral test. Because all same-sex members of any inbred strain are genetically identical, when animals from multiple strains are tested under controlled environmental conditions any differences among inbred strains reflect those allelic differences, because they are modulated by prenatal and postnatal environmental factors, including maternal behavior. Individual differences within a strain are caused by the environment or by interactions between genes and the environment (19). Furthermore, knowing how inbred strains perform relative to each other in one task can sometimes predict their relative performance in other tasks. For example, inbred strain rank order in reactivity to a modified mirrored chamber is correlated with their rank order reactivity in an elevated plus maze (20). Because of this, the general interpretation of these strain mean correlations is that common genes are influencing performance in the correlated tests (21); hence, the tests are measuring a similar underlying biological construct.

We sought to determine whether common genes influence performance in several mouse tests putatively modeling some aspects of novelty seeking. We examined activity in a novel environment, novel environment preference (e.g., ref. 10), head dipping and object preference on a hole-board (e.g., ref. 22), and spontaneous alternation (23). These choices were based on both their frequency of use in the literature and, in the case of novel environment preference, the degree to which the task could indicate a relative preference for novelty relative to familiarity. Other studies have addressed similar issues targeting the construct “anxiety,” in some cases using measures similar to those we report. Beuzen and Belzung (17) studied eight inbred mouse strains using a light/dark box, a “free exploration” test (similar to our novel environment preference test), and a passive avoidance learning test. Two factors described the variability in responses and differentiated responses in the light/dark and learning tests (factor 1) from those in the exploration test (factor 2). Although most strains did not differ in the exploration task, BALB/c female mice were highly exploratory whereas CBA mice scored lowest (17). A similarly conceived effort compared individual F₁ and F₂ hybrid mice for responses in tests where a novel object was introduced to a familiar environment and in a free exploration test. This study also included light/dark, elevated plus maze, and hole-board tests (24). Among other results, this analysis differentiated hole-board exploration and plus maze responses (factor 2) from novel object and environment preference (factor 1), but it was not designed to detect genetic differences.

The approach we took was to compare the patterns of difference in performance of inbred strains of mice on the five putative measures of novelty seeking. Genetic correlations among the tasks were estimated from the means of 14 inbred strains measured in all tasks. We identified exploratory head dipping on a hole-board as a highly heritable trait and subsequently selectively bred mice from a B6D2F3 starting population for divergent expression of head dipping for several generations. We then tested the selected lines of mice in three of the other novelty seeking tasks at each generation of selection as a second technique to identify genetic correlations among the tasks.

Results

For more information about results, see Tables 1–7 and Figs. 4–6, which are published as supporting information on the PNAS web site.

Differences among the inbred strains for the measures of novelty seeking are described in Supporting Text, which is published as supporting information on the PNAS web site, and shown in Fig. 4 A–F. The strain-effect sizes for all dependent variables are shown in Table 1, and strain-effect sizes within each of the two passes of the experiment are shown in Table 2. Strain-effect sizes were highest for activity measures, preference for the novel environment, and head dipping (R² ≥ 0.55).

Genetic Correlations.

To simplify the analysis, only the principal dependent variable and a measure of activity from each test were included in the correlational analysis; the variables discussed are indicated in Table 3, which shows all pairwise genetic correlations.

As indicated in Fig. 1 and Table 3, three of the putative measures of novelty seeking were moderately to strongly correlated: degree of preference for a novel environment, head dipping, and rearing (all r > |0.46|). However, the correlations with novel environment preference seemed to be driven by strain A/J, which showed a strong aversion to the novel compartment. Removing this strain from the analysis reduced these values (r <|0.39|; data not shown). On the other hand, spontaneous alternation and object preference were generally uncorrelated with the other novelty seeking measures (r < |0.27|), except that the correlation of spontaneous alternation and rears tended to be higher (r = −0.46).

Fig. 1.

Pearson product moment correlations among the inbred strain means of the novelty seeking measures. Each point represents an inbred strain mean, and the lines represent the least squares regression line for each correlation. “Rears” indicates the number of rears during the initial 30-min activity assessment, and “Activity” refers to the distance traveled (in centimeters) during this same test. “Novel Environment Preference” refers to the amount of time spent in the novel compartment, and “Head Dips” is the number of head dips on the hole-board. “Object Preference” indicates the percentage of time spent head dipping for holes relative to the total time spent head dipping, and “Alternation” is the proportion of mice from each strain that chose the novel arm. NOD refers to strain NOD/LTJ and is labeled to indicate this strain’s influence on the correlations with spontaneous alternation, and A/J is similarly shown as an outlier for the correlations with novel environment preference. The r values including all strains are indicated in each scatter plot. For values without the outlier strains, see Results.

A single strain, NOD/LTJ, appeared strongly to influence the magnitude (and perhaps direction) of the correlations with spontaneous alternation. Removing this strain resulted in lower values for these correlations (all r < 0.19; P > 0.54). The degree of preference for objects was not significantly correlated with any other measure (all r < −0.33; P > 0.26). The number of rears in the initial test of activity in a novel environment was relatively strongly associated with the number of head dips in the hole-board task (r = 0.77), and both measures were highly genetically correlated with locomotor activity.

Locomotor activity in the initial test in a novel environment was highly correlated with activity recorded concurrently in all other novelty seeking tests (all r > 0.8 and P < 0.01; see Fig. 5) as well as with the number of head dips and preference for a novel environment. Given this strong relationship between locomotion and most dependent measures, we examined the pattern of correlations that resulted after correcting statistically for the contribution of locomotion by regressing the main dependent measure from each task on locomotion as measured in that task. The result was a residual score for each mouse that reflected the degree to which this individual deviated from the best-fit regression line, and this residual was explicitly uncorrelated with locomotion (r = 0). Strain means were then calculated from individual residual scores, and genetic correlations among the mean residual scores were calculated. Residuals had a heritable component equal to or slightly less than that of the main dependent measure for each task (see Table 1), and the overall pattern of correlations among the remaining dependent measures was unchanged, with only the correlation of the residual score of head dipping and rearing approaching significance (r = 0.53, P < 0.06; data not shown). Thus, a genetically similar, heritable component exists between these two behaviors that cannot be completely explained by locomotor activity.

The pattern of genetic correlations across the inbred strains suggests that object preference and spontaneous alternation are genetically distinct from each other and largely unrelated to any of the other tasks, whereas activity in a novel environment is correlated with most other measures. This pattern is similar to that obtained from correlating the dependent measures of each mouse from all tests; i.e., the phenotypic correlations among dependent measures are similar to the genotypic correlations (see Table 4).

To characterize the relationships among the variables further we also performed a principal-components analysis of the individual phenotypic data from all of the novelty seeking measures. This analysis is described in Supporting Text, and the results are shown in Table 5. Unlike the genetic correlations, the principal-components analysis indicated that, at the phenotypic level, novel environment and object preference are more closely associated with each other than they are with any other trait. Therefore, we wished to test the relationships among the tasks further by using a second genetic approach: selective breeding for high and low head dipping as measured in the hole-board apparatus.

Selected Lines: Response to Selection.

The head dipping scores for the B6D2F3 founding population were normally distributed with a range of 21–116 head dips (data not shown), a range similar to that seen for the inbred strain means in this task. Thus, there was a good deal of variation in the starting population from which to select for high and low head dipping. Small differences between high exploratory behavior (HEB) and low exploratory behavior (LEB) mice were apparent after one generation of selection (see Fig. 2A and B; generation S1, combined R² = 0.03) and were larger after two generations (R² = 0.11). By the fifth selected generation, selected line accounted for 15% of the total variation. The selection parameters at each generation of selection are shown in Table 7. The realized heritability estimate was slightly larger in both upward replicates (h² = 0.14 for HEB-1 and 0.20 for HEB-2) than in the downward direction (h² = 0.11 for LEB-1 and 0.04 for LEB-2). Heritability for the divergence between HEB and LEB mice was estimated as 0.13 for both replicates.

Fig. 2.

Response to selection for divergent expression of head dipping in HEB and LEB mice. (A and B) The head dipping response across five generations of selection. Connected points indicate the line means, and the unconnected points represent the mice selected at each generation as breeders that produced the offspring of the following generation. (C and D) Activity (beam breaks) in HEB and LEB mice across the entire experiment (HEB and LEB mice were not selected for this trait). Means ± SEM are depicted.

Selected Lines: Correlated Responses to Selection.

LEB mice of both replicates moved less during the selection test at the fifth generation of selection (main effect of line at S5, F_1,128 = 39.48; P < 0.0001; see Fig. 2 C and D) as well as during all other correlated response testing (data not shown). The mice selected as breeders for both LEB lines moved substantially less than the population mean and particularly less than the HEB breeding pairs. The HEB mice selected as breeders, however, did not differ substantially from the population mean for locomotion at any generation of selection, indicating that the selection response for head dips in the upward direction, at least, was not driven solely by selection for high activity.

The overall dissociability of locomotion from head dipping is supported by an analysis of covariance, in which the proportion of variation in head dipping that can be accounted for by locomotion is removed before any potential difference in head-dipping due to the selected lines is considered. At generation 5, a significant main effect of selected line on the number of head dips was still present by analysis of covariance (F_1,127 = 8.03; P < 0.01).

At each generation of selection, all mice were tested first on the hole-board and, the next day, in the object preference task. One to 3 days later the mice were tested for novel environment preference, and 1 day later they were tested for spontaneous alternation in a Y-maze. For simplicity, and because of nearly identical results at all generations of selection, only the results from selection generation 5 are shown in Fig. 3. Although strong overall object preference and novel environment preference were observed, there were no differences between the selected lines in the degree of preference (Fs_1,132 < 1.04) and no effect of replicate or line by replicate interaction (Fs_1,132 < 2.70; P > 0.1). There were likewise no differences in the numbers of HEB and LEB mice that chose the novel arm in the spontaneous alternation task (χ² < 1.15; P > 0.28).

Fig. 3.

Object preference (A), novel environment preference (B), and spontaneous alternation (C) in HEB and LEB mice after five generations of selection. The dashed line indicates chance performance in all measures. Above-chance levels were observed in all tests, but no line differences were apparent. (n = 24–44 per genotype.) Means ± SEM are depicted.

Discussion

The present results indicate that there are genetic influences in each of the tests considered, and, in the case of preference for objects, this influence is relatively small. The pattern of genetic correlations from the inbred strain experiments suggests that, whereas most of the behaviors putatively indexing aspects of novelty seeking correlated relatively highly with locomotion, the novelty-related behaviors are at best modestly genetically correlated with each other. This pattern was also seen in the selectively bred HEB and LEB lines, which showed large, correlated differences in locomotor activity in all apparatus but did not consistently differ in performance in other tasks that putatively measured aspects of novelty seeking other than head dipping. Given the very different modalities of responses recorded in each task (locomotion, time spent in an area, choice behavior, etc.), this finding of few correlations may not be unexpected. However, it implies that the trait being measured by the various tasks either is not isomorphic across the behavioral assays or is phenotypically complex, and each task is sensitive to certain different underlying aspects of the putative trait.

Novel environment preference, which seemed to us to be the trait that most clearly reflected the actual preference for novelty, was moderately influenced by genotype (R² = 0.33). In a similar apparatus, but using a different procedure, Belzung and her collaborators (17, 25, 26) observed some genotypic differences in a free exploratory procedure, which generally indicated few strain differences, although more differences among other inbred strains were found in response to predator odor exposure. These results imply that genotype might play a relatively small role in the expression of novel environment preference, although variants of this task appear to be sensitive to different factors, such as age at weaning or other experiences (27–30). Data in Fig. 1 and Table 4 show that spontaneous alternation is largely unrelated to all of the other tasks.

Head dipping on a hole-board is a commonly used task in many types of pharmacological, genetic, or neural lesion experiments as a measure of exploratory and/or anxiety-like behavior (31, 32). In the inbred strain experiment, a strong genotypic correlation between head dipping and locomotion was observed (r = 0.77). If genotype is ignored and individual values for head dipping and locomotion correlated, a similar strong correlation was found (r = 0.63, see Table 4). However, within a given inbred strain the magnitude of the phenotypic correlation varied greatly: some strains showed strong positive correlations, whereas others showed only small positive correlations and two strains had slightly negative correlations between head dipping and locomotion (see Fig. 6). HEB and LEB mice at the fifth generation of selection also exhibited a genetically correlated difference in locomotor activity, and within each line and replicate the phenotypic association was variable (r = 0.22–0.67; data not shown). Collectively, these results indicate a somewhat unpredictable relationship between head dipping and locomotion within a given strain of untreated mice and suggest that the choice of mouse strain used for an experiment with the hole-board can have a large influence on this correlation.

Despite a high heritability estimate from the inbred strain experiment, the response to selection for head dipping proceeded at a slow rate and resulted in a modest difference between HEB and LEB mice of both replicates, with a realized heritability estimate of 0.13 for each replicate after five generations of selection. In retrospect, the modest response is most likely due to the choice of the C57BL/6J and DBA/2J progenitor strains. Although moderate differences were observed between these two strains in our first inbred strain comparison, this difference was not found in the second, which indicated in the aggregate that only modest additive genetic variance relating to head dipping existed in the starting population. Given the slow response to selection, it is probable that many genes, each exerting a small influence, are contributing to the trait. It does not appear that there are a few genes with large effect on head dipping present in the cross, or else a much more rapid divergence would have been expected.

The presence of a genetic correlation between two tasks is typically taken to indicate the action of common genes in influencing performance in the tasks, a situation called pleiotropy. The implication is that a survey of genotypes in two similar tasks should result in a similar rank order of the genotypes; if the rank orders are different or the strain means do not correlate, the tasks could be measuring independent processes or genetically unrelated aspects of the same process (33). The reliability of each task must also be considered. A failure to find a large genetic correlation between two tasks does not necessarily mean there is there is no common genetic influence between the measures, but it could indicate that one or both of the tasks are simply not reliable enough to detect a correlation that might actually exist. Furthermore, detection of a genetic correlation depends on a measurable genetic contribution to each task, and the larger the genetic effect size for each trait, the more likely a genetic association is to be detected.

With these limitations in mind, locomotion and the measures most closely associated with it, rearing and head dipping, are the most heritable and interrelated measures in the current experiments. Selective breeding of HEB and LEB mice resulted in populations of mice that displayed correlated differences in locomotion during the selection test. Object and novel environment preference, which demonstrated much smaller between-strain differences, tended to resemble each other phenotypically more than they did any other measure. Spontaneous alternation likewise tended to be unrelated to the other measures. Similar to the results obtained with the inbred strains, HEB and LEB mice did not show consistent differences in any of these measures, including the phenotypically similar object preference task, further indicating that these measures are largely dissimilar.

Task-specific stress or anxiety-like states could also contribute to the current pattern of results, but the relationship to novelty depends on the circumstances of testing. Handling mice activates the hypothalamic–pituitary–adrenal axis, leading to elevations in circulating corticosterone. In the free exploration test, a task similar to the novel environment preference test used in the current experiments; mice that were habituated to one compartment for 24 h showed no elevations of corticosterone when allowed to explore a novel compartment. However, mice forced to remain in the novel compartment without escape had elevated corticosterone levels, as did mice placed into a different, novel environment (34, 35). In our studies, mice were habituated to the familiar compartment for only 15 min. Strains could prefer novel objects or locations because they are stressful, seeking to reduce stress through familiarity. Alternatively, strains could select a novel environment because the familiar one has become stressful (36).

To explore the role of stress in the HEB/LEB differences, we performed two additional experiments (see Supporting Text). HEB and LEB mice were tested in the hole-board, and corticosterone levels were determined immediately after the test or from untested animals in the room. Untested HEB and LEB mice did not differ in corticosterone levels (HEB, 4.4 ± 1.4 μg/dl; LEB, 4.4 ± 1.4 μg/dl; not significant). Both genotypes showed significant elevations in corticosterone after exposure to the task, but the increases were similar (HEB, 16.6 ± 1.7 μg/dl; LEB, 15.5 ± 2.1 μg/dl; not significant). Thus, although stress clearly accompanies the hole-board exploration test, it cannot explain the genotypic differences in this behavior. We also tested separate groups of HEB and LEB mice in the novel environment exploration task after habituating them for 23–24 h in the familiar compartment, rather than the 15 min we used in the other experiments. In this task there was again no significant genotypic difference in corticosterone levels that resulted from free exploration of the novel compartment (HEB, 8.6 ± 1.6 μ/dl; LEB, 6.4 ± 1.0 μg/dl; not significant). After longer familiarization, these intermediate corticosterone values appeared to reflect the expected reduction in stress (36). Both genotypes showed an aversion to the novel compartment: HEB mice spent 405 ± 116 sec of the possible 1,200 sec in the novel compartment, and LEB spent a comparable 490 ± 75 sec there. These responses are similar to some inbred strains tested in the free exploratory procedure previously reported (17, 25, 26), suggesting that the relative degree of familiarity with an apparatus can have a large effect on the degree of novelty preference observed. Clearly, the behavioral differences between HEB and LEB mice cannot easily be explained on the basis of differences in stress responses to the testing situation and furthermore demonstrate that stress and exposure to novelty are related in most test situations.

None of the behavioral measures used in the current experiments examined aspects of human novelty seeking that reflect self-reported interoceptive states, such as fickleness or extravagance. To the extent that they reflect responses to novelty given the complex interactions of novelty, neophobia, anxiety, and activity, these tasks may most closely correspond to one of four subdimensions named by the Tridimensional Personality Questionnaire (1, 11) and elaborated in the Temperament and Character Inventory (12), called exploratory excitability. However, the traditional indices of rodent exploration, activity in a novel environment and head dipping on a hole-board, measure behaviors qualitatively different from those in the novel environment preference test, which relies on the overt expression of preference for the novel compartment. This last test could reflect some aspects of the impulsivity component of novelty seeking, because the expression of preference will be influenced by latency to enter the novel compartment.

Overall, these data suggest that, although there may be an underlying construct in mice that can be called novelty seeking, this construct is genotypically complex and confounded with other factors (notably general activity, anxiety, and stress), and no single task can claim to measure it. Earlier studies have also tended to show a lack of correlation among multiple tasks putatively assessing exploratory, anxiety-like, and novelty seeking tendencies (17, 24), and our data suggest that this is also true at the level of genetic influences. The current findings highlight the need for a better understanding of what is actually being measured in putative animal models of personality traits, particularly in studies attempting to identify genetic contributions to individual differences.

Materials and Methods

Subjects and Husbandry.

Inbred strain experiments.

Male and female mice of the following inbred strains from The Jackson Laboratory were used: 129S1/SvImJ, A/J, AKR/J, BALB/cByJ, BTBR/J, C3H/HeJ, C57BL/6J, C57L/J, DBA/2J, FVB/NJ, NOD/LTJ, NZB/B1NJ, PERA/EiJ, and PL/J. All mice were received at ≈45 days of age and housed in the animal facilities of the Portland Department of Veterans Affairs Medical Center for a minimum of 2 weeks before beginning the experiments. The experiments were conducted in two passes that consisted of different strains (n = 6 per sex per strain) within each pass, but 12 C57BL/6J and 12 DBA/2J mice were included in each pass to determine any effects of pass.

Selected line experiments.

B6D2F2/J mice were purchased from The Jackson Laboratory. Random mating pairs were established to produce an F₃ hybrid cross, which was subsequently used as the founding population. Each F₃ individual was genetically unique, possessing either C57BL/6J or DBA/2J alleles at each gene locus, and had experienced multiple recombinations of the progenitor strains’ chromosomes. One hundred eleven 55- to 65-day-old naïve B6D2F3 mice (53 male and 58 female) were tested on the hole-board as described for the inbred strain experiments. The 12 male and 12 female mice that showed the most head dipping behavior in the 10-min task were randomly divided into two replicate lines of six breeder pairs each to form the HEB-1 and HEB-2 lines, whereas the lowest-scoring 12 male and 12 female mice were paired to form the LEB-1 and LEB-2 lines. A mass selection procedure was used; mice were randomly assigned mates so long as the selected mate was not a sibling. This procedure was chosen to produce the fastest rate of response to selection at the expense of somewhat higher rates of inbreeding. After being selected from the initial, common F₃ parental population, each line and replicate was a closed breeding population in which all offspring were tested at each generation of selection, with the highest scoring HEB or lowest scoring LEB mice selected as breeders for each subsequent generation of selection. Six breeder pairs were maintained in each line and replicate throughout the experiment.

Procedure.

All inbred mice were tested serially through the following five tests in this order: activity in a novel environment; novel environment preference; head dipping on a hole-board; object preference on the hole-board; and spontaneous alternation in a Y-maze. In the first pass of these experiments, novel object recognition was also evaluated, but, because of problems with the procedure, this task was not used in the second pass, and the data are not shown. At each generation of selection of HEB and LEB mice, the testing procedure was similar to that used for the inbred strain experiments, with the exception that the mice were first tested on the hole-board, followed by object preference, novel environment preference, and spontaneous alternation. The testing order for the selection experiments was changed to measure head dipping, the trait for which the mice would be selectively bred, in experimentally naïve mice. Activity in a novel environment was not included in the report of the selection because of high correlations among locomotor activity scores seen in inbred strains. More details of behavioral testing are given in Supporting Text.

Activity in a Novel Environment.

Mice were placed into one of 12 identical Accuscan automated activity monitors, and locomotion (distance traveled in centimeters) and the number of rears were recorded for 30 min by infrared beam disruptions.

Novel Environment Preference.

Mice were placed into one side of four identical two-compartment boxes that were constructed of acrylic (40 × 40 cm total area) and consisted of one black-walled side with a white floor and a white-walled side with a black floor, both separated by a vertically sliding door. Initial placement into the black or white compartments was counterbalanced within all genotypes. After a 15-min exposure to one of the compartments, the door was opened, and the time spent and activity (beam breaks) in the novel and familiar sides were measured during a 20-min preference test by infrared beam disruptions. Chance performance in this apparatus is indicated by a time of 600 sec spent in the novel compartment.

Head Dipping on the Hole-Board.

The number of head dips and time spent head dipping into each of four 2.9-cm diameter holes, spaced equidistant from each other in the corners of a 40 × 40-cm board, and horizontal activity counts (beam breaks) were recorded via infrared beam disruption during a 10-min test.

Object Preference on the Hole-Board.

The day after the head dipping assessment, mice were placed back onto the same hole-board apparatus for a second 10-min test. Two small objects (centrifuge caps, black binder clips, or marbles; all objects ≈1.5 cm across) were placed into two of the holes at opposite corners. Preference was calculated as the amount of time spent head dipping in holes that had objects divided by the total time spent head dipping (percent preference). Activity was recorded as the number of beam breaks.

Spontaneous Alternation.

A two-trial version of the task was used. In the first trial, the mice were forced to choose one of the two arms of a Y-maze by occluding one arm with an opaque partition. Each arm was 15 cm tall and 30 cm long × 5 cm wide clear acrylic, and all angles between the arms were 120°. The mice were then confined to this arm for ≈30 sec and placed into a holding cage while the apparatus was cleaned and both partitions were removed. The mice were then placed back into the stem of the Y-maze and allowed to choose either the arm they had previously entered or the novel one. An entrance was defined as all four paws being within the arm. Because of very high performance in B6D2F2/J and B6D2F3/J mice (the proportion of mice that chose the novel arm was 90%), a different procedure was used for the selected line experiments. On trial 1, no partition was used to obstruct one of the arms, so the mouse was allowed to enter either arm. The mouse was then restrained in the chosen arm for ≈30 sec, the added partition was removed, and the mouse then entered either the novel arm or the arm it was placed in at the start of the test. The apparatus was cleaned between the testing of individual mice.

Statistics.

Inbred strain experiments.

To assess potential differences across passes, the responses of C57BL/6J and DBA/2J mice were compared on each dependent measure by two-way ANOVA (genotype X pass). Within these two strains, some relatively small main effects of pass were observed for some variables, but, with the exception of head dipping on a hole-board, no significant genotype X pass interactions were observed. Although the difference in head dipping was clearly unexpected, it is likely that this interaction is stochastic in origin considering the large number of dependent variables assessed. Thus, to simplify the analysis and presentation of the data, all correlations presented ignore any effects of pass. However, inbred strain-effect sizes are shown separately for each pass in Table 2 as well as collapsed across pass in Table 1.

Initial analyses for each dependent variable were by two-way ANOVA with strain and sex as independent variables. Narrow sense heritability (h²) was estimated as the percentage of total trait variance accounted for (R²) by inbred strain in a one-way ANOVA with strain as the only factor. To correct for the varying influence of locomotion in the tasks, residual scores were obtained from the regression of the principal dependent variable from each task (number of rears, number of head dips, percent preference for objects, or time spent in the novel compartment) on locomotion as measured in each task. Strain means for this residual score were then correlated. Scores for spontaneous alternation were not corrected because activity data were not collected in this task.

Selected line experiments.

Differences between HEB and LEB mice were estimated similarly to the inbred strain experiments as the percentage of total trait variation in head dipping accounted for by selected line. Correlated responses to selection were analyzed by ANOVA according to Crabbe et al. (37) where the strongest evidence for a genetic correlation is indicated by a significant difference in a trait between both replicates of the selected lines. Nonselected (“control”) lines of mice were not maintained, because these selected lines were designed for short-term use in the experiments reported here and were not intended as a long-enduring model for which nonselected control lines would be appropriate (37). Realized heritability was estimated according to Falconer and Mackay (38) as the ratio of the cumulative response to selection at each of the five generations of selection to the cumulative selection differential. The response to selection was expressed as a difference from the previous generation, and the cumulative selection differential represents the population-normalized deviation of the mice selected as breeders from the rest of the population.

Abbreviations

HEB

high exploratory behavior

LEB

low exploratory behavior.