Behavioral profiles of inbred strains on novel olfactory, spatial and emotional tests for reference memory in mice

Authors


Corresponding author: A. Holmes, Section on Behavioral Genomics, National Institute of Mental Health, Building 10 Room 4D11, Bethesda, MD 20892-1375, USA. E-mail: aholmes@codon.nih.gov

Abstract

Studying the behavior of genetic background strains provides important information for the design and interpretation of cognitive phenotypes in mutant mice. Our experiments examined the performance of three commonly used strains (C57BL/6J, 129S6, DBA/2J) on three behavioral tests for learning and memory that measure very different forms of memory, and for which there is a lack of data on strain differences. In the social transmission of food preference test (STFP) all three strains demonstrated intact memory for an odor-cued food that had been sampled on the breath of a cagemate 24 hours previously. While C57BL/6J and 129S6 mice showed good trace fear conditioning, DBA/2J mice showed a profound deficit on trace fear conditioning. In the Barnes maze test for spatial memory, the 129S6 strain showed poor probe trial performance, relative to C57BL/6J mice. Comparison of strains for open field exploratory activity and anxiety-like behavior suggests that poor Barnes maze performance reflects low exploratory behavior, rather than a true spatial memory deficit, in 129S6 mice. This interpretation is supported by good Morris water maze performance in 129S6 mice. These data support the use of a C57BL/6J background for studying memory deficits in mutant mice using any of these tasks, and the use of a 129S6 background in all but the Barnes maze. A DBA/2J background may be particularly useful for investigating the genetic basis of emotional memory using fear conditioning.

Genetic background is increasingly recognized as a critical consideration in behavioral studies using transgenic and knockout mice. Characterization of the behavior of mouse strains commonly used as genetic backgrounds can greatly strengthen the design and interpretation of studies of targeted gene mutations (Banbury Conference on genetic background in mice 1997; Crawley et al. 1997; Crawley 1999; Nguyen et al. 2000; Wolfer & Lipp 2000). Mutant mouse models have proven to be of great use in understanding normal cognitive processes (Chen & Tonegawa 1997; Crusio 1999; Mayford & Kandel 1999; Tang et al. 1999) and cognitive impairments symptomatic of neurological disease states (Hsiao et al. 1996; Chen et al. 2000). A number of studies have compared mouse strains on tests for learning and memory, and found diverse performance across strains. To date, however, comparisons have centered upon certain widely used behavioral tasks, including the Morris water maze, 8-arm radial maze, cued and contextual fear conditioning, T-maze, and passive avoidance (Wehner & Silva 1996; Crawley et al. 1997).

Some investigators have voiced concerns over the reliance on a limited battery of behavioral tests for cognition, and have developed novel methods to assess cognitive performance in rodents (Bunsey & Eichenbaum 1995; Gerlai & Clayton 1999; Pompl et al. 1999; Chen et al. 2000; Falls et al. 2000; Rondi-Reig et al. 2001). Cued and contextual fear conditioning is commonly used to test emotional memory in rodents (Phillips & LeDoux 1992; Fanselow et al. 1994; Caldarone et al. 1997; Wehner et al. 1997; Holmes & Rodgers 2001). In the standard ‘delay conditioning’ form of this task, mice are conditioned to associate visual or auditory stimuli (CS) with a footshock (US) via repeated pairings in which the US terminates at the same time as the CS. ‘Trace conditioning’ differs from this procedure in that there is a temporal separation between the cessation of the CS and the presentation of the US. There are examples of mutant mice that have shown cognitive phenotypes for trace, but not delay, fear conditioning e.g. GABAAγ2 subunit null mutant (Crestani et al. 1999) and NMDAR2 null mutant (Huerta et al. 2000), suggesting that trace fear conditioning may be an important additional test for emotional memory.

Barnes (1979) first described a dry-land maze test for spatial learning and memory in which rats escape from a brightly lit, exposed circular open field into a darkened box that is hidden beneath one of 18 holes around the perimeter of the open field. Over repeated trials, rodents learn to locate the spatial position of the escape box using distal environmental room cues, and make progressively fewer visits to the nonescape holes. This task resembles the Morris water maze task as a spatial memory test that is believed to rely upon the formation of a cognitive map of the distal visual cues in the test room. The Barnes maze is less physically demanding and probably less stressful than the Morris water maze because it is on dry land (Pompl et al. 1999). CAMKII (Bach et al. 1995; Mayford et al. 1996) and amyloid precursor protein (Pompl et al. 1999) mutant mice show deficits on a spatial, but not cued, version of the Barnes maze. It is likely that this test will be used in future studies to test for spatial memory in mutant mice.

The social transmission of food preference test (STFP) is based on food neophobia in rodents. Rats show a preference for eating a food that is cued with an odor previously experienced on the breath of a conspecific, over a food cued with a novel scent (Galef & Wigmore 1983; Galef & Stein 1985, Galef et al. 1988; Valsecchi & Galef 1989). In the laboratory, STFP is conducted in two stages. First, there is a brief (e.g. 30 min) social interaction session between an ‘observer’ mouse and a ‘demonstrator’ mouse that has just eaten an odor-cued food. The observer samples the odor on the breath of the demonstrator. Following a specified retention interval (e.g. 24 h) the observer mouse is presented with a choice between the food scented with the previously sampled odor, and food scented with a novel odor. The preference for the cued food over the novel food is taken as an index of memory for the previously cued food. Some studies (Winocur 1990; Bunsey & Eichenbaum 1995), but not others (Burton et al. 2000), have found that this form of olfactory memory is impaired by hippocampal lesions in rats. Mice lacking CREBαΔ isoforms (Kogan et al. 1997; Gass et al. 1998), synaptotagmin IV (Ferguson et al. 2000) and A-type K+ channel, Kvβ1.1 subtype (Giese et al. 1998) have all shown impairments on this task. Deficits on STFP parallel deficits in tasks that are known to be hippocampal-dependent in these mutant mice.

To facilitate the use of novel behavioral paradigms to phenotype transgenic and gene knockout mice, it is important that performance of inbred strains in these tests is carefully characterized. The present studies focus on three strains that are commonly used in behavioral genetics and in the generation and breeding of lines of transgenic and gene knockout mice. A background strain that performs poorly on a given task will obfuscate the detection of possible cognitive deficits caused by the molecular manipulation. Furthermore, it may be important to rule out the possible contribution to a cognitive phenotype of ‘hitchhiker’ genes from the parental embryonic stem cells that still flank the mutated gene, even after repeated backcrossing onto a different strain (Gerlai 1996). This can be achieved by comparing the mutant line to the parental strain, when those data exist. The present experiments generated data for three inbred strains, 129/SvEvTac, C57BL/6J and DBA/2J, on the social transmission of food preference test, Barnes maze task and trace fear conditioning. These are tests for long-term/reference memory for which little or no information exists regarding mouse strain differences. Results for these behavioral tests were compared to the Morris water maze task and standard cued and contextual fear conditioning, for which inbred strain distributions have been described (Upchurch & Wehner 1988a, 1988b, 1989; Paylor et al. 1993, 1994). Cognitive phenotypes in mutant mice have also been described in numerous reports using these standard paradigms of long-term/reference memory (Grant et al. 1992; Silva et al. 1992; Deutsch 1993; Bourtchuladze et al. 1994; Conquet et al. 1994; Kogan et al. 1997; Chen et al. 2000; Janus et al. 2000). Social transmission of food preference, Barnes maze and trace fear conditioning measure olfactory, spatial and emotional forms of memory, respectively. In order to determine whether performance deficits on tests for cognition were affected by motor and emotional differences, strains were compared on tests for exploratory locomotion and anxiety-related behaviors.

Materials and Methods

Subjects

Subjects were adult male and female (a) 129S6/SvEvTac mice (abbreviated as 129S6; Festing et al. 1999) obtained from Taconic (Germantown, NY); (b) C57BL/6J mice obtained from the Jackson Laboratory (Bar Habor, ME) and (c) DBA/2J mice obtained from the Jackson Laboratory. Mice were group-housed, five per cage, by strain and gender, in a temperature and humidity controlled vivarium, under a fixed 12 h light/dark cycle (lights on 0600). Food and water were provided ad libitum in the home cage. Mice were at least 3 months old at the commencement of behavioral testing. Subjects were purchased at 8–10 weeks of age in two batches of 60 mice per batch (10 mice/strain/gender) and one batch of 38 male mice (10 C57BL/6J, 28 DBA/2J). All testing was conducted during the light phase of the light/dark cycle. Batch 1 was tested in the following sequence over a four week period; Digiscan open field, elevated plus-maze, and light/dark exploration test. Four weeks after the testing in the light/dark exploration test, 30 mice were tested in the Barnes maze. Four weeks later, these 30 mice were tested in the Morris water maze. Batch 2 was tested in the social transmission of food preference test, and then trace fear conditioning three weeks later. Batch 3 was compared on standard delay fear conditioning vs. trace versions fear conditioning. Within any given test, mice were tested in an order counterbalanced for strain. All experimental procedures were approved by the National Institute of Mental Health Animal Care and Use Committee and followed the National Institutes of Mental Health Guidelines, ‘Using Animals in Intramural Research’.

Social transmission of food preference test

The social transmission of food preference test was conducted as previously described (Galef & Wigmore 1983; Kogan et al. 1997; Steiner et al. 2001). Mice were first shaped to eat powdered mouse chow (Dyets, Bethlehem, PA) from 4 oz. glass food jars (7 cm diameter, 5 cm depth) by placing one jar of powdered chow in the home cage for 1 h. On the day after shaping, one randomly chosen (‘demonstrator’) mouse from each home cage was placed in an individual holding cage and food deprived for 18 h. At the end of the deprivation period, each demonstrator was presented with a food jar containing powdered chow mixed with a novel flavor. Half the demonstrators were given powdered chow flavored with cocoa (2% by weight, Hershey Foods Corp., Hershey, PA). Half the demonstrators were given powdered chow flavored with cinnamon (1% by weight, Super G Inc., Landover, MD). Access was for a period of 1 h. Food jars were weighed at the end of the hour, to confirm that demonstrators had eaten. Each demonstrator was then immediately placed in its home cage to interact with cagemates for 30 min. 24 h after interacting with the demonstrator, observer mice were placed in individual holding cages and food deprived for 18 h. At the end of the deprivation period, each observer was presented with two food jars, one containing cocoa-flavored chow and one containing cinnamon-flavored chow for a period of 1 h. Weight of the food in each jar was measured before and after the 1-h period.

Trace fear conditioning

Experiment 1; pure tone CS

Cued and contextual trace fear conditioning was conducted using methods previously described for delay conditioning (Phillips & LeDoux 1992; Fanselow et al. 1994; Holmes et al. in press), with the exception that there was a temporal separation of the cessation of the CS and the onset of the US. The training chamber was square with clear Plexiglas walls and a grid floor used to deliver an electric shock (Freeze Monitor, San Diego Instruments, San Diego, CA), evenly illuminated by overhead fluorescent white room lighting (∼ 40 lux). On day 1, mice were individually placed inside the test chamber. After 2 min, a pure tone (∼800 Hz, 80 dB) was presented for 30 seconds (conditioned stimulus; CS). 2.5 seconds after the CS ended, a mild foot shock (2 seconds, 0.5 mA) was delivered (unconditioned stimulus; US). The pairing of the CS and US was repeated for a total of 4 times, with an interval of 2 min between pairings. The mouse was removed from the test chamber 2 min after the final foot shock and returned to the home cage. Twenty-four hours later the mouse was placed in a novel test chamber (novel context). The novel context differed from the trained chamber in color (white Plexiglas), shape (triangular), floor texture (smooth), and odor (one wall was scented with 1.0 ml vanilla extract), but was equivalent to the trained chamber in floor area and overall cubic dimensions. Novel context testing was conducted in a different room from the training room. Freezing in the novel context was measured for 6 min, with the CS presented continuously during minutes 4 to 6. On completion of the novel context test, the mouse was returned to the home cage. Twenty-four hours later the mouse was returned to the training chamber and freezing was measured for 5 min (context test). In all tests, freezing was defined as the cessation of all movement except breathing. The presence or absence of freezing was scored by a highly trained observer at 10 second intervals.

Experiment 2; white noise CS

DBA/2J mice exhibit a profound hearing loss to high frequency sounds by early adulthood (Zheng et al. 1999), whereas C57BL/6J and 129 strains show hearing loss at much later ages than presently tested (Henry & Chole 1980; Willott et al. 1982; Zheng et al. 1999; Willott & Turner 2000; Yoshida et al. 2000). To test whether DBA/2J deficits in cued fear conditioning were caused by an inability to hear the high frequency pure tone auditory CS, a second experiment compared cued conditioning between C57BL/6J and DBA/2J mice using a broadband white noise as the auditory CS. On day 1, C57BL/6J6J mice underwent delay fear conditioning, while DBA/2J mice underwent either delay or trace fear conditioning. Trace fear conditioning was conducted using a procedure identical to that used in Experiment 1, with the exception that the auditory CS was an 80-dB broadband white noise (White Noise Generator, San Diego Instruments, San Diego, CA). For delay conditioning, the procedure was identical to that for trace fear conditioning with the exception that mild foot shock (US) was delivered during the final 2 seconds of white noise CS presentation (i.e. CS and US co-terminated). Twenty-four hours later, mice underwent novel context testing as described for Experiment 1, but using the white noise as the auditory CS. Twenty-four hours after novel context testing, mice underwent context testing as described for Experiment 1. The order of training and testing was counterbalanced for strain and form of conditioning.

Barnes maze

The Barnes maze task was employed to test for spatial memory as described previously (Barnes 1979; Poucet et al. 1991; Bach et al. 1995; Fox et al. 1999; McLay et al. 1999; Pompl et al. 1999; Miyakawa et al. in press; Paylor et al. in press). Mice were trained to locate a black Plexiglas escape box hidden underneath one of 12 holes (2.0 cm in diameter) evenly spaced around the perimeter (1.8 cm from the edge) of an elevated (80 cm above the floor) white circular open field (1.0 m in diameter) (San Diego Instruments, San Diego, CA), evenly illuminated by overhead fluorescent white room lighting (∼ 70 lux). The hole above the escape box was designated the target, analogous to the hidden platform in the Morris water maze task. The location of the target was consistent for a given mouse but randomized across mice. To prevent orientation to the target before a trial began, mice were initially placed in the center of the open field under an opaque cylinder (12 cm in diameter). The cylinder was removed after 15 seconds and the trial begun. Mice were given two training trials per day over 8 days. A 30% ethanol solution was used to wipe clean the open field after every trial, and the escape box after each session. To further discourage the use of intra-maze cues, the whole apparatus was periodically rotated in a random manner, whilst maintaining the location of the escape box with respect to spatial extra-maze cues. When all mice had reached a criterion of < 1 search error for two successive training days, mice were given a 60 second probe trial transfer test without the escape box. The probe trial was administered 2–3 h after the final training trial. For training trials and the probe trial, the number of visits to each hole, movement speed, and total distance traveled were automatically recorded. Data acquisition and analysis were conducted using NIH Image Videotracking software developed by Wayne Rasband, NIMH and modified in our laboratory by T. Miyakawa (Miyakawa et al. in press).

Morris water maze task

The Morris water maze task was conducted as previously described (Morris et al. 1982; Steiner et al. 2001; Holmes et al. in press). Mice were trained to find a hidden platform in a circular pool of water (1.0 m in diameter) (Nalge Inc., Rochester, NY), evenly illuminated by overhead fluorescent white room lighting (∼ 70 lux). Water was maintained at 23 °C (± 1) and rendered opaque by the addition of white non-toxic paint. A training trial began by placing a mouse into a quadrant located either left, right, or opposite to the target quadrant, in pseudo-random order. The sequence of start positions was different on each training day. A mouse that failed to find the platform after 60 seconds was guided to the platform. For acquisition of the visible platform task, four trials per day were conducted for 4 days. For acquisition of the hidden platform task, four trials per day were conducted for 5 days. Latency to find the platform, average swim speed (total cm distance traveled/ seconds to reach the platform), and thigmotaxis (per cent time spent in the outer 8 cm annulus at the perimeter of the pool; i.e. 17% of the surface area of the pool) were automatically measured for each training trial. When all strains had reached a criterion of finding the hidden platform in < 10 seconds, mice were given a 60-second probe trial transfer test without the platform. The probe trial was administered 2–3 h after the final training trial. Time spent in each quadrant, the number of crossings across the trained platform location (and corresponding regions in non-trained quadrants), swim speed and thigmotaxis were automatically recorded. Data analysis and acquisition was conducted using NIH Image videotracking software developed by Wayne Rasband, NIMH, and modified in our laboratory by T. Miyakawa (Miyakawa et al. 2001; Steiner et al. 2001).

Open field test

Spontaneous exploratory activity was assessed in a Digiscan automated open field (Accuscan, Colombus, OH). The open field was a square arena (40 × 40 × 35 cm) with clear Plexiglas walls and floor, evenly illuminated by overhead fluorescent white room lighting (∼ 40 lux). Eight photocell beams were located on each of two sides of the arena, at right angles to one another, forming a grid of 64 equally sized squares. To detect vertical movements, a third set of eight photocell beams was located above the square grid. Mice were individually placed in the center of the open field and left to freely explore for 30 min. The apparatus was cleaned with a 30% ethanol solution between subjects. The number of horizontal and vertical beam breaks were taken as measures of horizontal and vertical activity, respectively. Time spent in a central square (20 × 20 cm) of the open field was automatically recorded as center time.

Elevated plus-maze

The elevated plus-maze was conducted as previously described (Holmes & Rodgers 1998, 1999; Holmes et al. 2000, in press). The apparatus (San Diego Instruments, San Diego, CA) comprised two open arms (30 × 5 × 0.3 cm) and two closed arms (30 × 5 × 15 cm) that extended from a common central platform (5 × 5 cm). The apparatus was constructed from polypropylene and Plexiglas (white floor, clear walls), elevated to a height of 38 cm above floor level, and evenly illuminated by overhead fluorescent white room lighting (∼ 40 lux). Mice were individually placed on the center square, facing an open arm, and allowed to freely explore the apparatus for 5 min. The apparatus was cleaned with a 30% ethanol solution between subjects. Behaviors were scored by a highly trained observer using Hindsight software (Hindsight, Scientific Programming Services, Wokingham, U.K). Behaviors scored were (a) the frequencies of open and closed entries (an arm entry was defined as all four paws into an arm) and (b) the time spent in the open arms. These scores were used to calculate the percentage of open arm entries [(open/total) × (100)], and the percentage of the test session spent in the open arms of the maze [(time in open arms/session duration) × 100].

Light/dark exploration test

The light ↔ dark exploration test was conducted as previously described (Crawley 1981; Mathis et al. 1995; Holmes & Rodgers 2001; Holmes et al. in press). The apparatus consisted of a polypropylene cage (44 × 21 × 21 cm) separated into two compartments by a partition, which had a small opening (12 × 5 cm) at floor level. The larger compartment (28 cm long) was open-topped, transparent and brightly illuminated by white light from a 40-watt desk lamp positioned above the lit chamber (100 lux). The smaller compartment (14 cm long), was covered at the top with black Plexiglas and painted black on all sides. Mice were individually placed in the center of the brightly lit compartment, facing away from the partition, and allowed to freely explore the apparatus for 10 min. The apparatus was cleaned with a 30% ethanol solution between subjects. The number of light/dark transitions between the two compartments, and the total time spent in the dark compartment, were automatically recorded via photocells located at the opening between compartments, connected to a data storage device.

Statistics

Strain differences were analyzed using between subjects analysis of variance (ANOVA) and Newman-Keuls posthoc comparisons where appropriate, using StatView (SAS Institute Inc., Cary, NC). The following within-subject data were analyzed by use of repeated measures ANOVA: errors to find the target hole across repeated trials in the Barnes maze; visits to each of the 12 holes on the Barnes maze probe trial; latencies to find the visible or hidden platforms, swim speed and thigmotaxis across repeated trials in the Morris water maze; platform crossings and time spent in each of the four quadrants on the Morris water maze probe trial; within-subjects open field horizontal and vertical activity across time. Consumption of cued food vs. non-cued food in the STFP test was compared by use of paired students t-tests.

Results

Social transmission of food preference

Each of the three strains consumed more of the cued food than the novel food; 129S6 (t = 4.03, df = 31, P < 0.001), C57BL/6J (t = 2.05, df = 31, P= 0.048), and DBA/2J (t = 2.71, df = 22, P= 0.01) (Fig. 1a). There was no significant effect of strain for cue preference (cued/total food consumed) (F2,84 = 0.77, P= 0.50) (Fig. 1b). There was a significant effect of strain for total amount of food consumed (F2,86 = 11.98, P < 0.001); C57BL/6J mice ate significantly more total food than either 129S6 or DBA/2J mice (P < 0.01; Fig. 1c).

Figure 1.

Social transmission of food preference in three inbred strains of mice. (a) All three strains showed good performance as demonstrated by the significant preference for the familiar cued food over the novel uncued food. (b) The three strains showed similar degrees of preference for the cued food ([cued/total food eaten] × 100). (c) Total food consumption was higher in C57BL/6J than either 129S6 or DBA/2J (c). n= 23–32 for each strain. For Figs 1–6 data are presented as means ± SEM, *P < 0.05, **P < 0.01.

Trace fear conditioning

Experiment 1; pure tone CS

There was no significant effect of strain for freezing during the 2 min prior to the first CS-US pairing (F1,26 = 2.80, P= 0.08; Fig. 2a). Freezing differed between strains during the 2 min after the final CS-US pairing (F1,26 = 3.81, P= 0.04), with lower scores in DBA/2J as compared to 129S6 (P < 0.05), but not C57BL/6J (Fig. 2a). There was a significant effect of strain for freezing during context testing (F1,26 = 10.78, P= 0.001), with lower freezing in DBA/2J as compared to either C57BL/6J or 129S6 (P < 0.01; Fig. 2b). Novel context freezing in the absence of the CS differed between strains (F1,26 = 4.66, P= 0.02), with lower freezing in DBA/2J as compared to C57BL/6J (P < 0.05; Fig. 2c). Strains also differed in novel context freezing in the presence of the CS (F1,26 = 26.62, P < 0.001), with significantly lower freezing in DBA/2J as compared to either 129S6 or C57BL/6J (P < 0.01; Fig. 2c).

Figure 2.

Trace fear conditioning in three inbred strains of mice, using a 80-dB ∼800 Hz high frequency, pure tone as the auditory conditioned stimulus (CS). (a) Freezing during the two minutes prior to the first CS-US pairing was similar across strains. Freezing during the 2 min after the final CS-US pairing was lower in DBA/2J than 129S6. (b) Freezing during same context testing was lower in DBA/2J than either 129S6 or C57BL/6J. (c) Freezing during novel context testing in the absence of the CS was lower in DBA/2J than C57BL/6J6J. Freezing during novel context testing in the presence of the auditory CS was lower in DBA/2J than either 129S6 or C57BL/6J. n= 9–10 for each strain. *P < 0.05, **P < 0.01.

Experiment 2; white noise CS

DBA/2J showed higher freezing than C57BL/6J during the 2 min prior to the first CS-US pairing (F1,36 = 10.70, P= 0.002; Fig. 3a), but not during the 2 min after the final CS-US pairing (F1,36 = 2.70, P= 0.11; Fig. 3a). During context testing, delay conditioned C57BL/6J showed higher freezing than either trace or delay conditioned DBA/2J (F1,36 = 28.86, P < 0.001; Fig. 3b). During novel context testing, delay conditioned C57BL/6J showed higher freezing than either trace or delay conditioned DBA/2J, both in the absence (F1,36 = 21.10, P < 0.001) and presence of the CS (F1,36 = 63.26, P < 0.001; Fig. 3c). There were no differences in freezing between the trace and delay conditioned DBA/2J groups before the first CS-US (F1,26 = 1.68, P= 0.21), after the final CS-US (F1,26 = 0.16, P= 0.69), during context testing (F1,26 = 0.14, P= 0.71), during novel context testing in the absence (F1,26 = 1.28, P= 0.27; Fig. 3c), or presence of the CS (F1,26 = 1.11, P= 0.30) (Fig. 3).

Figure 3.

To exclude the effects of hearing loss for high frequency sounds, DBA/2J mice were compared for trace and delay fear conditioning using an 80-dB white noise auditory CS. C57BL/6J mice served as a positive control for delay fear conditioning using the white noise CS. (a) Freezing during the two minutes prior to the first CS-US pairing was similar between trace and delay groups of DBA/2J. Freezing during this period was higher in DBA/2J than C57BL/6J6J. Freezing during the 2 min after the final CS-US pairing was similar between trace and delay groups of DBA/2J. Freezing during this period was similar between DBA/2J and C57BL/6J. (b) Freezing during same context testing was similar between trace and delay groups of DBA/2J. Freezing during same context testing was lower in DBA/2J than C57BL/6J. (c) Freezing during novel context testing in the absence of the CS was similar between trace and delay groups of DBA/2J. Freezing during novel context testing in the absence of the CS was lower in DBA/2J than C57BL/6J. Freezing during novel context testing in the presence of the CS was similar between trace and delay groups of DBA/2J. Freezing during novel context testing in the presence of the auditory CS was lower in DBA/2J than C57BL/6J. n= 10–14 for each strain. CS = conditioned stimulus. **P < 0.01.

Barnes maze

There was no effect of strain for the number of search errors made during acquisition (F2,26 = 0.64, P= 0.53). No interaction between strain and training day for search errors was detected (F7,182 = 0.80, P= 0.67). Each of the three strains showed a significant reduction in errors to reach the hole across training days (P < 0.001; Fig. 4a). During the probe trial, all three strains showed a significant effect of hole location (P = 0.001). However, while C57BL/6J made more visits to the target hole than any of the non-target holes (Fig. 4e), 129S6 did not make more visits to the target hole than any single non-target hole (Fig. 4d). DBA/2J made more visits to the target hole than any non-target hole, except those immediately adjacent to the trained hole (Fig. 4f). There was a significant effect of strain for the target hole preference index (F2,25 = 7.27, P= 0.003); C57BL/6J had a significantly higher preference score than either 129S6 or DBA/2J (P < 0.01; Fig. 4b). The effect of strain on total hole visits during the probe trial approached significance (F2,25 = 3.08, P= 0.06; Fig. 4c).

Figure 4.

Barnes circular maze performance in three inbred strains of mice. (a) All three strains showed good acquisition, as demonstrated by reduced number of errors to find the target hole across eight training days. (b) During the probe trial, C57BL/6J showed a higher target hole preference index (i.e. target hole visits/average non-target visits) than 129S6 or DBA/2J. (c) C57BL/6J mice showed a trend for more total hole visits than 129S6 or DBA/2J. (d) 129S6 failed to show selective search during the probe trial, and did not visit the target hole significantly more times than non-target holes. (e) C57BL/6J showed selective search during the probe trial, as demonstrated by more visits to the target hole than any non-target hole. (f) DBA/2J showed moderately selective search, as demonstrated by more visits to the target hole than non-target holes, except the holes immediately adjacent to the target hole. n= 9–10 for each strain. T = target hole. *P < 0.05, **P < 0.01.

Morris water maze task

There was no effect of strain for latencies to reach the platform during visible platform acquisition (F2,26 = 0.72, P= 0.50; Fig. 5a). No significant interaction between strain and training day was detected (F3,78 = 1.25, P= 0.29). Each of the three strains showed a fall in latencies over the four training days (P < 0.001). There was no effect of strain for platform hidden acquisition (F2,26 = 0.23, P= 0.79; Fig. 5b); each of the three strains showed a fall in latencies over the five training days (P < 0.001). There was a significant interaction between strain and training day (F4,104 = 2.71, P < 0.01); latencies on day 1, but not days 2–5, which were lower in C57BL/6J than either 129S6 or DBA/2J (P < 0.05). There was no effect of strain (F2,26 = 0.09, P= 0.92), training day (F4,104 = 1.18, P= 0.33), and no strain–day interaction (F4,104 = 0.95, P= 0.48) for average swim speed during hidden platform acquisition (Fig. 5c). There was no effect of strain for thigmotaxis (F2,26 = 2.84, P= 0.08; Fig. 5d), which fell over training days (F4,104 = 34.21, P < 0.001). However, there was a significant interaction between thigmotaxis and training day (F4,104 = 3.79, P < 0.001); C57BL/6J showed less thigmotaxis than 129S6 on day 1 (P < 0.05), while thigmotaxis was lower in 129S6 than C57BL/6J or DBA/2J by day 3, and lower than DBA/2J on day 4 (P < 0.05). All three strains showed selective search in the Morris water maze probe trial, as measured by quadrant stay time (P < 0.01) and platform location crossings (P < 0.05; Fig. 5e). 129S6 and DBA/2J spent significantly more time and made more pseudo-platform crossings in the trained quadrant as compared to any other quadrant (P < 0.05). C57BL/6J spent significantly more time and made more pseudo-platform crossings in the trained quadrant as compared to quadrants 1 or 2 (P < 0.05). There was no significant effect of strain for average swim speed during the probe trial (F2,26 = 1.86, P= 0.18). There was a significant effect of strain for thigmotaxis during the probe trial (F2,26 = 9.25, P < 0.001); 129S6 showed lower levels of thigmotaxis than either C57BL/6J or DBA/2J (P < 0.05), data not shown.

Figure 5.

Morris water maze task performance in three inbred strains of mice. All three strains showed good acquisition of (a) the visible platform task and (b) hidden platform task. There were no strain differences for (c) swimming speed or (d) thigmotaxis on the hidden platform task. All three strains showed selective spatial search during the probe trial, as measured by both (e) quadrant stay time and (f) platform crossings. n= 9–10 for each strain. T = trained quadrant. *P < 0.05.

Exploratory locomotion

There was a significant effect of strain (F2,56 = 70.67, P < 0.001) and a significant strain–time interaction (F2,290 = 5.84, P < 0.001), for open field horizontal activity. C57BL/6J were significantly more active than either 129S6 or DBA/2J, while DBA/2J were significantly more active than 129S6 (P < 0.01; Fig. 6a). All three strains showed an habituation of horizontal activity over time (F5,290 = 82.11, P < 0.001). There was a significant effect of strain, and a significant strain–time interaction (F5,280 = 3.70, P < 0.001) for vertical activity (F2,56 = 37.83, P < 0.001). 129S6 were significantly less active than either C57BL/6J or DBA/2J (P < 0.01; Fig. 6b). There was no significant main effect of time for vertical activity (F5,280 = 1.67, P= 0.14). There was a significant effect of strain for time spent in the center of the open field (F2,56 = 24.45, P < 0.001); C57BL/6J spent significantly more time in the center than either 129S6 or DBA/2J (P < 0.01; Fig. 6c). In the light/dark exploration test, there was a significant effect of strain for light/dark transitions (F1,55 = 11.61, P < 0.001), but not time spent in the dark compartment (F1,55 = 2.16, P= 0.13). C57BL/6J made significantly more transitions than either 129S6 or DBA/2J (P < 0.01; Fig. 7a). In the elevated plus-maze, there was a significant effect of strain for the percentage open arm entries (F2,53 = 3.54, P= 0.04), but not percentage open arm time (F2,50 = 0.41, P= 0.67), or closed arm entries (F2,50 = 0.41, P= 0.67; Fig. 7c); 129S6 had significantly lower scores for percentage open entries than DBA/2J (P = 0.05; Fig. 7b), while higher C57BL/6J scores for percentage open entries in comparison to 129S6 approached significance (P = 0.08).

Figure 6.

Exploratory locomotion in three inbred strains of mice. (a) C57BL/6J showed higher levels of open field horizontal activity than either 129S6 or DBA/2J 129S6 showed lower levels of horizontal activity than DBA/2J. (b) 129S6 showed lower open field vertical activity than either C57BL/6J or DBA/2J. (c) C57BL/6J spent more time in the center of the open field than either 129S6 or DBA/2J. n= 19–20 for each strain.

Figure 7.

Anxiety-related behaviors in three inbred strains of mice. (a) In the light/dark exploration test, C57BL/6J showed more transitions between the brightly lit, open compartment and the enclosed dark compartment than either 129S6 or DBA/2J (n = 19–20 for each strain). (b) In the elevated plus-maze, 129S6 showed lower percentage open arm entries than DBA/2J. (c) There were no strain differences for closed arm entries. n = 18–20 for each strain. *P < 0.05, **P < 0.01.

Discussion

Background genes are an important practical issue in transgenic and knockout studies in mice, as a determinant of baseline performance and as a potential false positive influence on a mutant phenotype. Differences in learning and memory performance between inbred mouse strains have been demonstrated on a number of behavioral tasks (Wehner & Silva 1996; Crawley et al. 1997). In this series of experiments we have compared 129S6, C57BL/6J and DBA/2J mice on three relatively novel tests for long-term/reference memory, for which there is a lack of information on strain distributions. There were no differences between strains in cognitive performance in the STFP. In this test, all three strains showed a significant preference for consuming a food that was cued with a flavor which subjects had sampled on the breath of a cagemate 24 h previously, over a food scented with a novel flavor. While C57BL/6J mice did consume more food in total than either 129S6 or DBA/2J mice, the magnitude of the food preference did not differ between strains, as measured by the ratio of cued to non-cued food consumed. These data suggest that 129S6, C57BL/6J6J, and DBA/2J strains would all be suitable genetic backgrounds for investigating gene mutation-induced cognitive deficits using STFP, as wild type control mice would be expected to show intact STFP. As STFP does not involve exposure to aversive stimuli, and does not require the mouse to perform taxing motor responses, this test may be of particular use as a test for cognition in mutant mice with poor motor function or high anxiety-like behavior in novel environments. Conversely, STFP may be unsuitable for genotypes showing abnormal feeding behaviors, olfactory or gustatory processes. As this test also requires animals to show social behaviors during transmission of the food preference, a strain showing minimal social interaction could produce a false positive memory deficit.

The finding that cognitive performance on STFP was equivalent between 129S6 and C57BL/6J strains argues against the generalization that 129 strains are poor learners. Studies have found poor cognitive performance in certain 129 substrains, including 129/J and 129/SvJ (Crawley et al. 1997; Montkowski et al. 1997; Owen et al. 1997; Hengemihle et al. 1999) and shown that a 129 background strain can compromise the detection of a cognitive deficiency in mutant mice (Wolfer et al. 1997; Paradee et al. 1999; Dobkin et al. 2000). In contrast, there are a number of reports of good learning and memory performance in other 129 substrains (Crawley et al. 1997; Montkowski et al. 1997; Owen et al. 1997; Royle et al. 1999; Nguyen et al. 2000). In this context, 129 strains represent a genetically heterogeneous class of inbred strains (Simpson et al. 1997), and the profile of one 129 substrain should not be taken as evidence of performance in another substrain (Wehner & Silva 1996; Crawley et al. 1997; Montkowski et al. 1997; Owen et al. 1997; Balogh et al. 1999).

We have previously reported that mice overexpressing the neuropeptide galanin, show a deficit in STFP under conditions identical to those presently employed (Steiner et al. 2001). Mice lacking synaptotagmin IV (Ferguson et al. 2000) and A-type K+ channel, Kvβ1.1 subtype-deficient mice (Giese et al. 1998) have also shown deficits in STFP. The observation that these mutant mice were also impaired on hippocampal-dependent tasks such as the Morris water maze and contextual fear conditioning suggested that STFP may also be hippocampal-dependent. However, while Kogan et al. (1997) reported STFP deficits in mice lacking the CREBαΔ isoforms, Gass et al. (1998) found that CREB mutant mice fail to show STFP deficits, despite being impaired on the hippocampal-dependent Morris water maze, and contextual fear conditioning. Some lesion studies in rats have demonstrated that large lesions of the hippocampal region are more efficacious at producing STFP deficits than selective lesions of the hippocampus proper (Winocur 1990; Bunsey & Eichenbaum 1995). However, a recent study by Burton et al. (2000) found that excision of the hippocampus and subiculum (or amygdala) was not able to block STFP in rats, whereas spatial memory in a T-maze was impaired. To date, there have been no lesion or pharmacological studies in mice to ascertain whether STFP is sensitive to hippocampal disruptions in this species.

DBA/2J mice showed evidence of impaired contextual fear conditioning in comparison to 129S6 and C57BL/6J mice, in the trace version of this task. Thus, DBA/2J mice showed less freezing than either 129S6 or C57BL/6J strains during exposure to the trained context. This finding is consistent with previous reports that have demonstrated poor contextual fear conditioning in the DBA/2J strain (Paylor et al. 1994; Fordyce et al. 1995; Wehner & Silva 1996; Crawley et al. 1997; Gerlai 1998; Stiedl et al. 1999b; Nguyen et al. 2000) and was confirmed in a second experiment in which DBA/2J showed low contextual freezing following training on either trace or delay versions of fear conditioning. Contextual fear conditioning is believed to be hippocampal-dependent (Kim & Fanselow 1992; Phillips & LeDoux 1992; Maren et al. 1997; Frankland et al. 1998; Cho et al. 1999; Gewirtz et al. 2000) and learning deficits in DBA/2 mice have been associated with deficits in hippocampal long-term potentiation and lower hippocampal protein kinase C activity relative to C57BL/6J6 and 129 substrains (Wehner et al. 1990; Bowers et al. 1995; Fordyce et al. 1995; Matsuyama et al. 1997; Bampton et al. 1999; Nguyen et al. 2000). Present deficits in contextual fear conditioning provide further evidence for a genetic impairment in hippocampal function in DBA/2J mice, and support the use of this strain as a genetic model for studying deficient hippocampal-mediated learning and memory (Paylor et al. 1993; Ammassari-Teule et al. 1997; Logue et al. 1997). A DBA/2J background could also be an appropriate genetic background with which to study a genetic enhancement of contextual memory.

Trace fear conditioning is a comparatively recent adaptation of the standard delay fear conditioning task for use in mice. ‘Delay’ fear conditioning, in which a visual or auditory conditioned stimulus and an unconditioned footshock are presented at the same time, is believed to be primarily mediated by the amygdala (Phillips & LeDoux 1992). In contrast, there is evidence that introducing a temporal separation between the termination of the conditioned stimulus and the presentation of the unconditioned stimulus (i.e. ‘trace conditioning’) requires intact hippocampal function (Solomon et al. 1986; Moyer et al. 1990; Kim et al. 1995; Thompson & Kim 1996; McEchron et al. 1998). Previous studies have demonstrated that certain 129 substrains and C57BL/6J6 mice show good performance in delay fear conditioning paradigms (Paylor et al. 1994; Owen et al. 1997; Stiedl & Spiess 1997; Gerlai 1998; Stiedl et al. 1999a, 1999b). Present findings show that these two strains showed high levels of freezing during presentation of an auditory cue following trace fear conditioning, indicating that either background is appropriate for studies of cognitive impairments in mutant mice using this form of fear conditioning (Crestani et al. 1999). A secondary finding was that C57BL/6J mice displayed relatively high levels of freezing during exposure to the novel context in the absence of the auditory CS. This observation is consistent with previous reports demonstrating that C57BL/6J6 mice tend to show evidence of a generalization of fear responses to novel contexts following fear conditioning (Gerlai 1998; Radulovic et al. 1998; Stiedl et al. 1999b).

In contrast to the intact cued conditioning in the C57BL/6J and 129S6 strains, DBA/2J mice exhibited very low levels of freezing in the presence of the auditory cue in the trace conditioning paradigm. There are previous reports showing that DBA/2 mice are not impaired on cued conditioning when trained under the delay fear conditioning procedure (Paylor et al. 1994; Fordyce et al. 1995; Owen et al. 1997; Gerlai 1998). A dissociation between impaired contextual conditioning and intact cued conditioning in DBA/2 mice has been taken as evidence for a cognitive deficit on hippocampal-dependent, but not hippocampal-independent tasks (Upchurch & Wehner 1988a, 1988b, 1989; Paylor et al. 1993, 1994). However, other studies have found that DBA/2 do show impairments on cued (delay) fear conditioning (Stiedl et al. 1999b; Nguyen et al. 2000). Many factors could explain this variability in the literature, including genetic differences in the same nominal strain from different commercial suppliers, and genetic drift in a strain that has been bred in-house at academic institutions for many generations (Paylor et al. 1994; Montkowski et al. 1997; Owen et al. 1997). In addition, differences in early life experience, including the stress of shipping, nutrition, weaning, previous test experience, experimental procedure and the parameters of the conditioning stimuli, could all contribute to these variable findings for cued conditioning in DBA/2 mice.

It remains unclear from the present experiment whether DBA/2J deficits in cued fear conditioning are specific to the trace procedure, or reflect a general deficit in cued fear conditioning. An artifactual explanation for trace cued conditioning deficits in DBA/2J mice is that they simply fail to perceive the auditory CS during training. DBA/2J mice exhibit a profound hearing loss to high frequency sounds by early adulthood (Zheng et al. 1999). This compares with an age-related hearing loss in C57BL/6J and 129 strains that is more progressive and which occurs much later in life (Henry & Chole 1980; Willott et al. 1982; Zheng et al. 1999; Willott & Turner 2000; Yoshida et al. 2000). Because DBA/2J mice showed a trace cued conditioning deficit using a ∼800 Hz pure tone as the auditory CS, it was unclear whether performance impairments in this strain were due to a cognitive impairment or a hearing deficit. To test these alternatives, a second experiment compared cued trace fear conditioning between C57BL/6J and DBA/2J mice using white noise as the conditioned auditory stimulus. Results showed that DBA/2J mice were again impaired on cued fear conditioning, relative to C57BL/6J, excluding the possibility that the DBA/2J deficit in this task was due to a hearing impairment. Moreover, comparison of DBA/2J mice for delay and trace forms of fear conditioning under otherwise identical conditions, demonstrated that this strain was equally impaired on both forms of the task. Differences between inbred strains in perception of the footshock during training could be another factor contributing to apparent deficits in learning and memory performance in fear conditioning in DBA/2J mice. As a measure of nociception, all three stains exhibited dramatic increases in freezing immediately following training. However, DBA/2J mice showed significantly lower freezing than the 129S6, but not the C57BL/6J, strain. Therefore, we cannot exclude the possibility that genetic differences in nociception affected strain differences in fear conditioning. However, the literature comparing nociceptive responses in inbred mouse strains, including DBA/2, 129 substrains and C57BL/6J6, shows largely normal nociceptive responses in DBA/2 (e.g. Belknap et al. 1990; Mogil & Adhikari 1999; Mogil et al. 1999).

There was no evidence of strain differences during acquisition in the Barnes maze. 129S6, C57BL/6J and DBA/2J mice all showed a progressive reduction in the number of non-target holes searched across training trials, reaching the criterion of 1 error per trial by the 8th training day. Strain differences were observed in the Barnes maze probe trial, in which the escape box was removed and mice searched for 60 seconds. During the probe trial, C57BL/6J mice made more visits to the target hole than any of the non-target holes. By contrast, 129S6 mice showed a similar number of visits to the target hole, the hole immediately adjacent to the target hole and the hole opposite the target. DBA/2J mice showed an intermediate probe trial profile to the other strains, visiting the target hole more frequently than all the non-target holes except the two holes immediately adjacent to the target. Confirming the superior search selectivity in C57BL/6J, this strain showed a higher score for the target hole preference index than either of the other strains (i.e. ratio of target hole/non-target visits). This pattern of search could be interpreted as poor spatial memory for the location of the target hole in 129S6 and, to a lesser extent, DBA/2J mice, as compared to good memory in the C57BL/6J strain. This latter observation is consistent with previous reports describing excellent Barnes maze performance in C57BL/6J6 mice (Bach et al. 1999; Fox et al. 1999; Nguyen et al. 2000).

The observation that 129S6 mice show good Barnes maze acquisition, but poor probe trial performance, is salient to the finding of performance dissociations between intact hidden platform acquisition and impaired probe trial performance in the Morris water maze (Steiner et al. 2001). These findings have been interpreted as evidence for the use of successful, non-spatial search strategies during acquisition (Wolfer & Lipp 2000; Steiner et al. 2001). It is possible that during Barnes maze acquisition mice were able to use non-spatial search strategies, such as intramaze olfactory cues, to locate the escape box, including odor cues from the escape box itself. However, the use of intramaze olfactory cues is unlikely because the surface of the Barnes maze was thoroughly cleaned with ethanol solution after every trial, and the maze was intermittently rotated (while maintaining the location of the escape box with respect to extra maze cues).

An alternative explanation for the strain differences in the Barnes maze probe trial is that performance on this task was affected by differences in exploratory behaviors between the strains. Previously, Fox et al. (1999) and Nguyen et al. (2000) have reported poor Barnes maze performance in 129/SvEms mice relative to C57BL/6J mice. Both studies noted that the 129 substrain exhibited low exploration that appeared to deleteriously affect performance in this task. In contrast to Barnes maze acquisition trials, when the mouse can locate the escape hole and enter the escape box, the absence of the escape box on the probe trial requires the animal to remain on the open field for the duration of the trial. Low exploration during the probe trial could artificially bias hole visit scores, such that a preference for the escape hole would not be clear in mice showing a low number of total hole visits. In support of this interpretation, comparison of total hole visits indicated that C57BL/6J showed a trend, although non-significant, for a greater number of total hole visits than 129S6 during the probe trial. Furthermore, comparison of strains for exploratory activity in the Digiscan open field showed that 129S6 were markedly less active than C57BL/6J mice for both horizontal and vertical activity, while DBA/2J were intermediate between the other strains for horizontal activity. These data are consistent with previous reports that C57BL/6J6 are highly active as compared to 129 and other inbred strains (Logue et al. 1997; Montkowski et al. 1997; Homanics et al. 1999; Rogers et al. 1999; Royle et al. 1999; Bolivar et al. 2000; Koide et al. 2000; Voikar et al. 2001). 129S6 also showed evidence of reduced exploratory locomotion and heightened anxiety-like behavior in comparison to C57BL/6J mice, as measured by the elevated plus-maze, light/dark exploration test and open field center time. Similar differences in anxiety-like behaviors in these tests have been found between C57BL/6J6 and 129 strains (Homanics et al. 1999; Rogers et al. 1999; Voikar et al. 2001; Montkowski et al. 1997).

Taken together, these data strongly support the conclusion that low exploratory behavior and high anxiety-like behavior negatively affected performance of 129S6 mice on the Barnes maze probe trial, and that this strain is unsuitable as a genetic background for studies using this task. In contrast, high exploratory tendencies in C57BL/6J mice make this strain an excellent background for Barnes maze studies, as reported previously (Bach et al. 1999; Fox et al. 1999; Nguyen et al. 2000). Performance deficits in 129S6 due to low exploration and high anxiety-like behavior are likely to be even more prominent in larger diameter Barnes maze apparatuses than that presently used ( Barnes 1979; Bach et al. 1995; Fox et al. 1999; Nguyen et al. 2000). More generally, these findings reinforce the notion that genetic differences affecting motor, sensory, emotional and motivational phenotypes can strongly affect performance on tests for cognition (Wehner & Silva 1996; Owen et al. 1997; Crawley 1999, 2000). As further support for a lack of strain differences in spatial memory on the Barnes maze probe trial, we found no evidence of strain differences when the same mice were compared on the Morris water maze probe trial. Consistent with some previous reports (Montkowski et al. 1997; Owen et al. 1997; Balogh et al. 1999), but not others (Upchurch & Wehner 1988a, 1988b, 1989; Paylor et al. 1993; Nguyen et al. 2000), all three strains had similar acquisition of the hidden platform task, and showed intact spatial search on the probe trial in the Morris water maze. The good spatial memory performance of DBA/2J mice in the Morris water maze was surprising given the deficits shown by this strain in contextual fear conditioning, another hippocampal-dependent task. The relatively small size of the Morris water maze used in the present comparisons may have contributed to this phenotype. There were only minor strain differences for visible platform acquisition, and swim speed and thigmotaxis during hidden platform acquisition, confirming an absence of overriding sensorimotor or emotional phenotypes in any strain.

In summary, the present data provide important information on the behavioral performance of inbred strains commonly used as genetic backgrounds in gene mutant mouse studies on three recently adapted tests for cognition. The data also extend previous literature on strain comparisons for other learning and memory tests (Wehner & Silva 1996; Crawley et al. 1997), and molecular (Zilles et al. 2000), electrophysiological (Bampton et al. 1999; Nguyen et al. 2000), and anatomical (Schwegler & Crusio 1995) correlates of cognition. On the olfactory-based social transmission of food preference test, 129S6, C57BL/6J and DBA/2J all performed well. On both trace and delay forms of fear conditioning, as well as for contextual fear conditioning, 129S6 and C57BL/6J performed well, while DBA/2J showed profound impairments. Excellent Barnes maze performance for spatial memory in C57BL/6J mice contrasted with intermediate performance in DBA/2J and poor performance in 129S6 on the probe trial component of this task. Strain differences in open field exploratory locomotion and anxiety-like behavior suggest that the Barnes maze deficits in 129S6 reflect a confounding behavioral inhibition in this strain in this test context, rather than a true spatial memory impairment. This interpretation was supported by the observation of intact Morris water maze performance in 129S6 mice. Overall, our findings with relatively new tasks, including olfactory, spatial and emotional learning and memory paradigms, confirms good performance of C57BL/6J on all tasks, and the utility of 129S6 and DBA/2J on some of these tasks. The fact that strain differences were not uniform across all tests, but were determined by the behavioral test employed, indicates that the specific tasks to be used are a primary consideration when choosing the background strain for breeding a new mutation.

Acknowledgments

Research supported by the NIMH intramural research program (NIMH-IRP).

Ancillary