Strain differences in delay discounting using inbred rats

Abstract

A heightened aversion to delayed rewards is associated with substance abuse and numerous other neuropsychiatric disorders. Many of these disorders are heritable, raising the possibility that delay aversion may also have a significant genetic or heritable component. To examine this possibility, we compared delay discounting in six inbred strains of rats (Brown Norway, Copenhagen, Lewis, Fischer, Noble and Wistar Furth) using the adjusting amount procedure, which provides a measure of the subjective value of delayed rewards. The subjective value of rewards decreased as the delay to receipt increased for all strains. However, a main effect of strain and a strain × delay interaction indicated that some strains were more sensitive to the imposition of delays than others. Fitting a hyperbolic discount equation showed significant strain differences in sensitivity to delay (k). These data indicate that there are significant strain differences in delay discounting. All strains strongly preferred the 10% sucrose solution (the reinforcer in the delay discounting task) over water and the amount of sucrose consumed was correlated with sensitivity to delay. Locomotor activity was not correlated with delay discounting behavior. Additional research will be required to disentangle genetic influences from maternal effects and to determine how these factors influence the underlying association between heightened delay discounting and neuropsychiatric disorders.

Drug abuse is characterized by heightened impulsivity (for review see Reynolds 2006), as drug abusers are unable to delay gratification. The role of impulsivity in the development of drug abuse remains unclear. Drug taking may lead to neuroadaptations that affect impulsivity. Alternatively, heightened impulsivity may be a pre-existing condition and therefore a risk factor for drug abuse. Evidence exists for a genetic basis for drug abuse (for review see Begleiter et al. 1995), but fewer data are available on the genetic basis of impulsivity. Isles et al. (2004), using inbred mouse strains, estimated that 16% of responding on a delay discounting task was attributable to genetic influences. Anderson and Woolverton (2005) found differences in delay discounting between two inbred rat lines (also see Perry et al. 2007; Wilhelm & Mitchell 2008, which report differences in selectively bred rat lines). Calculation of strain differences requires more than two rat lines, and at this point, systematic studies examining delay discounting in inbred rats are lacking. Such studies are important because mouse and rat models do not always correspond. For example, Wilhelm et al. (2007) found no difference in delay discounting between mice selectively bred for high or low alcohol consumption, but selectively bred high-alcohol-drinking (HAD) rats were more impulsive (delay averse) than selectively bred low-alcohol-drinking (LAD) rats (Wilhelm & Mitchell 2008).

In this study, we examined delay discounting by six inbred rat strains (Brown Norway, Copenhagen, Fischer, Lewis, Noble and Wistar Furth), attempting to replicate and extend the findings reported by Anderson and Woolverton (2005). To the best of our knowledge, the rat strains chosen for this study are unrelated with the exception of Lewis rats, which were derived from a Wistar stock in 1954 (Mashimo et al. 2006, Mouse Genome Informatics, The Jackson Laboratory – Mouse Genome Database (MGD) & Mouse Genome Informatics 2007). We also compared sucrose preference and consumption as differences in the motivational potency of the reinforcer might be critical to the interpretation of delay discounting data. For instance, animals that exhibit a relatively low preference for the sucrose reinforcer may rapidly devalue delayed sucrose because it is subjectively less reinforcing. Thus, these animals may appear more impulsive in a delay discounting task because the reinforcer is less efficacious and therefore more apt to devaluation, and not because the animal is more sensitive to the imposition of a delay.

Lastly, we correlated locomotor activity with other task results to determine if activity differences predict subsequent discounting. Previous studies indicate locomotor activity differences in four of the inbred rat strains used in this study (Bardo MT, personal communication, August 2008). Locomotor activity may correlate with delay discounting in inbred mouse strains, such that more active mice exhibit heightened aversion to delays (Isles et al. 2004). A similar between-strains comparison has not been carried out in rats, but individual differences in locomotor activity do not appear to correlate with delay discounting behavior in rats (Perry et al. 2005, Wilhelm & Mitchell 2008, Winstanley et al. 2003).

Materials and methods

Subjects

Eight subjects from each of six different inbred rat strains (Brown Norway, Copenhagen, Fischer, Lewis, Noble and Wistar Furth) were acquired from Charles River Laboratories (CRL), Inc. Half of the rats (four from each strain) started the experiment in July 2006, while the other half began the experiment in April 2007. The Lewis and Fischer rats were chosen to compare the results of this study with those of Anderson and Woolverton (2005). Other strains were chosen at random to provide a panel of unrelated inbred strains from which to calculate strain differences in delay discounting. Brown Norway, Copenhagen, Lewis and Noble strains were acquired from the CRL, Inc. facility in Portage, MI, USA. Fischer rats were acquired from the CRL, Inc. facility in Kingston, NY, USA and Wistar Furth rats were acquired from the CRL, Inc. facility in Wilmington, MA, USA. Conditions at all CRL facilities include ambient room temperature of 21 ± 1°C, 30–70% humidity and a 12 h on and 12 h off light–dark cycle.

Male rats were 5–7 weeks old when they arrived and were housed in the Department of Comparative Medicine at Oregon Health & Science University on a 12:12 light–dark cycle (lights on at 06:00 h). Average weights (±SEM) prior to food deprivation were: Brown Norway 173 ± 3 g, Copenhagen 152 ± 4 g, Fischer 155 ± 3 g, Lewis 194 ± 2 g, Noble 194 ± 4 g, and Wistar Furth 145 ± 5 g. All procedures were approved by the Institutional Animal Care and Use Committee and adhered to National Institutes of Health Guidelines.

Apparatus

For the delay discounting procedure, we used eight identical (Med Associates, St. Albans, VT, USA) modular rat test chambers housed individually within melamine sound-attenuating cabinets. Chambers had acrylic front and back panels and stainless steel side panels. A fan provided constant ventilation and low-level background noise. A house light was mounted in the center of one of the stainless steel panels, with a response clicker mounted on the outside of this panel. Opposite this panel, were three non-retractable levers mounted directly below circular lights and directly above recesses with nose poke sensors. The levers remained extended at all times. Computer-controlled pumps were used to deliver sucrose reinforcers (10% w/v) to liquid cups located in the outer recesses.

Training

Three days before the initiation of training, animals were food deprived to approximately 90% of their free-feeding weights. Throughout the experiment, animals were maintained at this level of deprivation by providing supplemental lab chow after the completion of each experimental session. A detailed description of the training procedure has been described elsewhere (Wilhelm & Mitchell 2008). Training was divided into three phases. In Phase 1, rats were trained to collect rewards from the right and left nose poke recesses via delivery of a progressively delayed non-contingent reward. Additionally, rats could earn rewards by pressing either the right or left levers. Rats that attained a level of lever pressing, greater than 65 presses in 60 min, moved to Phase 2 of training. In Phase 2, rats learned to first press the middle lever to activate the outer levers. Following a middle lever press, a press on either of the two outer levers resulted in delivery of a reward. If rats chose the same lever on two consecutive trials, then on the following trial, rats were forced to choose the previously unchosen lever (forced-choice trial). Rats were required to complete 80 free-choice trials in 60 min on two consecutive days to move to Phase 3. In Phase 3, one of the outer levers was assigned to be the adjusting reward lever, and the other lever designated as the 150 μl lever (counterbalanced between subjects but constant throughout the experiment). The size of the reward from the adjusting reward lever was initially 75 μl, but adjusted throughout the session. On free-choice trials when the 150 μl lever was chosen, the size of the adjusting reward increased by 10% for the next trial. Choice of the adjusting reward lever decreased the size of the adjusting reward for the next trial by 10%. Forced-choice trials did not affect the size of the adjusting reward. At the beginning of each trial, the rat was given 22 seconds to respond on the middle lever, after which a response on the outer levers was allowed until 24 seconds had elapsed in the trial. If no response was made during this time, the trial was counted as an omission and all lights were extinguished until the next trial began. If the rat responded within the time constraints, the sucrose reinforcer was delivered, the clicker sounded and the lights were extinguished. A variable length inter-trial interval (ITI) was used to make trials 40 seconds long. Sessions lasted until 60 free-choice trials were completed (approximately 50 min). To complete Phase 3 training, rats were required to respond on at least 55 of the 60 possible free-choice trials on two consecutive sessions. One Brown Norway, one Copenhagen and one Wistar Furth rat were unable to complete Phase 3 of training and therefore, were excluded from the experiment.

Delay discounting phase

The adjusting amount procedure was adapted from a procedure described in Richards et al. (1997) to assess delay discounting. Experimental sessions were as described in Phase 3 of training, except that a response on the 150 μl lever resulted in the delivery of a 150 μl sucrose reinforcer delayed by 0, 2, 4, 8 or 16 seconds. The delay remained constant within a session but varied between sessions according to a pseudo-Latin square design, such that each delay was experienced on six occasions. One Wistar Furth rat had fewer than 40 responses for each session at 4 seconds delay and therefore was not included in subsequent analyses.

Two-bottle sucrose preference

After completing the delay discounting assessment, all animals were tested 4 days per week with each concentration of sucrose in the following sequence 0%, 0.1%, 0.5%, 1.0%, 2.5%, 5%, 10%, 20%, 30% and 0% w/v sucrose. On test days, each rat was placed in a rat drinking cage for 50 min (the approximate length of a discounting session) and given access to two bottles; one containing a sucrose solution and the other containing water. To eliminate the potential for side bias, placement of sucrose and water bottles was alternated daily.

The amount of solution consumed per 50-min test session (grams of sucrose) was determined by weighing the bottles before and after each test session. Sucrose preference was calculated as the ratio of the volume of sucrose consumed to the total volume of fluid (water plus sucrose) consumed. Both measures were averaged over each of the four test days.

Locomotor activity

Locomotor activity was assessed after animals had completed the delay discounting and sucrose preference phases of the experiment. Timing of the decision to collect activity data meant that the first squad of animals was unavailable to complete the task. Food was available ad-libitum during this phase of the experiment. We used five Accuscan automated activity monitors (Accuscan Instruments Inc., Columbus, OH, USA). Chambers consisted of a 40 × 40 × 30 cm clear acrylic test cage placed inside a monitoring unit that recorded photocell beam breaks, which were translated into distance traveled (in centimeters). Eight or 16 evenly spaced photocells and receptors were located 2 cm above the chamber floor. Monitors and test cages were housed in black acrylic chambers that were lined with foam to attenuate external noise. A fluorescent light and fan located within the test chamber were on during testing to provide illumination, ventilation and low-level background noise. Each rat was assigned to a specific locomotor activity chamber. Locomotor activity was recorded for 30 min on two consecutive days and software was used to convert beam breaks into horizontal distance traveled (in centimeters).

Data analysis

Analyses were similar to those described in Wilhelm and Mitchell (2008). The main dependent variable was the volume of sucrose delivered from the adjusting reward alternative at the ‘indifference point’, i.e. when subjects select each alternative with roughly equal frequency. Based on previous findings (Richards et al. 1997), animals reach an indifference point after the first 30 trials within a session. Table 1 shows the percentage of responses on the adjusting lever during the final 30 trials of a session and indicates that animals chose each alternative with roughly equal frequency over trials 31–60. The indifference point was calculated as the median size of the adjusting reward on trials 31–60. Medians, rather than means, were used because changes in the adjusting reward amount on successive trials were proportions of the amount on the prior trial, resulting in a skewed distribution. The first session at each delay was discarded and the medians of the subsequent five sessions were averaged to yield the indifference point for each condition. If a rat did not complete at least 40 trials in a session, the data from that session were not included in the analysis (Table 1). Hyperbolic equations were fitted to each animal's indifference points (modified from Mazur 1987) using Microsoft Excel (Microsoft Corporation, Redmond, WA, USA):

Table 1.

Mean ± SEM percentage choice of the adjusting reward lever during the final 30 trials of each session and the percentage of sessions for which animals completed greater than 40 choice trials

Strain	n	Adjusting (%)	Completed sessions (%)
Brown Norway	7	47.0 ± 1.3	87
Copenhagen	7	51.4 ± 1.8	84
Fischer	8	44.2 ± 0.3	95
Lewis	8	46.5 ± 0.7	99
Noble	8	45.4 ± 0.9	85
Wistar Furth	6	46.9 ± 1.5	96

$$V=\frac{bA}{1+kD},$$ (1)

where V represents the volume of the adjusting reward at indifference, A represents the amount of sucrose solution from the 150-μl alternative and D represents the delay to receiving the reinforcer from the 150-μl alternative (0, 2, 4, 8 or 16 seconds). The bias parameter, b, was calculated such that the product of b and A equaled each subject's indifference point at 0 seconds delay. A b-value greater than 1 indicates a relative preference for the delayed lever on sessions with 0 seconds delay. Conversely, b-values less than 1, indicate a relative preference for the immediate lever on sessions with 0 seconds delay. The discount parameter, k, is a fitted parameter and is an index of the rate of reinforcer devaluation because of the delay to reward receipt. Larger values of k indicate steeper discount functions and a stronger aversion to delayed rewards.

Other aspects of discounting performance from trials 31–60 were averaged and analyzed, including response latency (the time from trial onset until middle lever press) and choice latency (the time from middle lever press until either of the two outer choice levers was pressed). The main dependent variables in the two-bottle sucrose-preference procedure were the grams of sucrose consumed per kilogram of body weight and the preference ratio (milliliters of sucrose solution consumed/total milliliters consumed). For locomotor activity, the main dependent variables were the horizontal distance traveled on days 1 and 2 and the difference in horizontal distance traveled from day 1 to day 2 (habituation).

Animals completed procedures in the following order: (1) delay discounting, (2) two-bottle preference and (3) locomotor activity (April 2007 animals only). Animals that were unable to complete Phase 3 of training in the delay discounting procedure were excluded from subsequent procedures.

Analysis of variance (anova), correlational analyses and other statistical tests were carried out using spss version 16.0 (SPSS Inc., Chicago, IL, USA). Huynh–Feldt corrections were applied where appropriate, and adjusted degrees of freedom are provided. Tukey post-hoc comparisons were used when needed to disambiguate significant main effects. To analyze the effects of delay and strain on indifference points, we used an anova with delay as a within-subject factor and strain and start date (July 2006, April 2007) as between-subject factors. Follow-up analyses were carried out at each delay by performing simple anovas with strain as the between-subject factor. The k-values and b-values were analyzed using anova with strain and start date as between-subject factors. Analysis of response and choice latency required anovas with delay and lever choice as within-subject factors and strain and start date as between-subject factors. The two-bottle sucrose-preference test was analyzed by anova with sucrose concentration as a within-subject factor and strain and start date as between-subject factors. Follow-up analysis used a simple anova with strain as a between-subject factor at the 10% sucrose concentration because this is the concentration of sucrose used in the delay discounting experiment. Analyses of locomotor activity examined activity on day 1 or habituation (day 1 activity − day 2 activity) using anovas with strain as a between-subject factor. For all analyses, significant main effects of strain (P < 0.05) were disambiguated with Tukey post-hoc tests.

Heritability was calculated similar to the manner described in Isles et al. (2004), which was based on the components of variance calculation described in Hegmann and Possidente (1981).

$$\text{Heritability}={\sigma }_{\text{s}}^2/({\sigma }_{\text{s}}^2+2{\sigma }^2)$$ (2)

This calculation, however, was designed to extrapolate to an F₂ population. To calculate the observed heritability for the strains tested without extrapolation, we slightly modified the equation to

$$\text{Heritability}={\sigma }_{\text{s}}^2/({\sigma }_{\text{s}}^2+{\sigma }^2),$$ (3)

where σ² is equal to the mean square within strains variation and σ_s² is equal to (σ² − (mean square between-strains variation))/k. Where k is a function of the unequal replicate numbers for each strain and the mean square values are derived from a simple, univariate anova with strain as the between-subject factor. Inbred strains are genetically homogenous; however, heritability as calculated does not provide a measure of purely genetic variability, as differences in maternal care or other early environmental factors (i.e. environmental differences at the vendor) may influence the results. As noted above, although all of the animals used in this study came from the same vendor, the location and the personnel of the breeding facilities varied.

Results

Delay discounting

An anova with delay as a within-subject factor and strain and start date as between-subject factors indicated that indifference points decreased as the time to reward increased for all strains (F_2.7,104.9 = 154.66, P < 0.001) (Table 2). There were also strain differences (F_5,32 = 2.55, P < 0.05) and a significant strain × delay interaction (F_16.4,104.9 = 2.44, P < 0.01), with different strains exhibiting varying sensitivities to delay. There were no other significant main effects or interactions (all F's <1, P's >0.49). Follow-up analyses were carried out by performing simple, univariate anovas at each delay with strain as a between-subject factor coupled with Tukey post-hoc comparisons. Strain differences were most apparent at longer delays, with Noble rats having higher indifference points at 2, 4, 8 and 16 seconds, with these differences reaching statistical significance at 4 and 16 seconds (there was a relatively large amount of variance in indifference points at the 8 seconds delay), supporting the hypothesis that some strains are more averse to delays than others (Table 2). There were also strain differences at the 0-second delay indicating differences in bias (see below).

Table 2.

Mean ± SEM indifference points (microliters) as a function of delay for each of the six inbred rat strains

	Delay (seconds)
Strain	0	2	4	8	16
Brown Norway	177 ± 29	83 ± 15	84 ± 7	44 ± 4	24 ± 2
Copenhagen	113 ± 22	75 ± 12	63 ± 10**	37 ± 6	32 ± 5
Fischer	225 ± 16*	108 ± 11	69 ± 7**	44 ± 3	22 ± 2**
Lewis	168 ± 19	96 ± 7	65 ± 4**	34 ± 2	20 ± 2**
Noble	195 ± 27	129 ± 18	120 ± 22	57 ± 11	40 ± 8
Wistar	191 ± 23	87 ± 21	59 ± 9**	38 ± 5	23 ± 2

A simple, univariate anova was carried out on indifference points at each delay, with post-hoc Tukey comparisons used to compare the indifference points for each strain.

^*Denotes P < 0.05 compared with Copenhagen rats.

^**Denotes P < 0.05 compared with Noble rats.

To further examine strain differences in discounting, the hyperbolic discount function (eqn 1) was fitted to the data for each individual rat, resulting in the generation of a k-value (Fig. 1a) and b-value (Fig. 1b) for each animal. The k- and b-values were analyzed using anovas with strain and start date as between-subject factors. There were significant strain differences in k (F_5,32 = 6.51, P < 0.001). There were no start date (F_1,32 = 0.81, P = 0.38) or start date × strain (F_5,32 = 1.97, P = 0.11) effects. Post-hoc Tukey tests showed Fischer and Wistar Furth rats had higher k-values than Copenhagen and Noble rats (P's <0.01). The k-values derived from Brown Norway and Lewis rats did not differ from any of the other strains tested. The calculated heritability (eqn 3) was 0.40.

Figure 1. Strain differences in k- and b-values (panels a and b).

The figure depicts the mean (±SEM) k-value or b-value for each rat strain. ^★Denotes higher k-value than Copenhagen P < 0.01, ^#denotes higher k-value than Noble P < 0.01, @denotes higher b-value than Copenhagen P < 0.05.

As noted earlier, b, the measure of bias, reflects responding at the 0-second delay. There was a marginal effect of strain on b (F_5,32 = 2.42, P = 0.06). There were no start date (F_1,32 = 0.83, P = 0.37) or start date × strain effects (F_5,32 = 0.71, P = 0.62). Tukey post-hoc analyses comparing strains indicated that Fischer rats had larger b-values than Copenhagen rats (P = 0.02), while all other comparisons showed no significant differences. A one-sample t-test found that bias in the Fischer strain was significantly greater than 1 (P = 0.02), indicating a significant preference for the 150 μl alternative at the 0 seconds delay condition. All other strains exhibited bias values not different from 1 (all P's >0.13), indicating that rats titrated the volume of the adjusting reward to approximate 150 μl (the size of the other alternative) when the delay to the fixed alternative was 0 seconds.

Latency measures

There were significant main effects of strain in response latency (F_5,32 = 9.27, P < 0.001). Response latency increased as the delay to reward increased (F_4,128 = 38.93, P < 0.001), with analyses indicating the presence of linear and quadratic trends (Fig. 2). Higher-order trends were not considered. There were significant delay × strain (F_20,128 = 4.77, P < 0.001) and lever choice × delay (F_4,128 = 5.44, P < 0.001) interactions, but no other interactions were significant (all F's <2.02, P's >0.09). Post-hoc Tukey comparisons indicate that Lewis rats were the quickest to respond (compared with all other strains except Wistar Furth, P's <0.05; comparison to Wistar Furth, P = 0.08), while Brown Norway rats were slower to respond than Lewis (P < 0.01) and Wistar Furth (P < 0.05). Post-hoc analysis showed that the delay × strain and choice × delay interactions were not systematic and therefore will not be discussed further.

Figure 2. Strain differences in response latency.

Response latency was defined as the time from the start of a trial until animals pressed the middle lever (trials on which animals made no responses were not included in the analysis). The figure depicts the mean (±SEM) response latency for each delay when rats chose the adjusting (a) or 150 μl (b) levers for each rat strain.

For choice latency (Fig. 3), there were also significant strain differences in choice latency (F_5,32 = 5.19, P < 0.001), and a delay × strain interaction (F_20,128 = 2.06, P < 0.01). There were differences in choice latency as a function of delay (F_4,128 = 7.40, P < 0.001), which included the presence of a significant quadratic but not linear trend. Responding on the delayed lever took longer than for the adjusting lever (F_1,32 = 9.23, P < 0.01). There were also significant lever choice × delay (F_4,128 = 7.84, P < 0.001) and choice × delay × start date interactions (F_4,128 = 5.08, P < 0.01), but again the effects were not systematic (Fig. 3). There were no other significant interactions (all F's <1.4, P's >0.26). Post-hoc Tukey comparisons found that Lewis rats again responded more quickly than most other strains: Brown Norway (P < 0.01), Fischer (P < 0.05) and Noble (P < 0.001). Wistar Furth rats were significantly quicker to respond than Noble (P < 0.01) and marginally quicker than Brown Norway (P = 0.05) rats. No other significant differences were found.

Figure 3. Strain differences in choice latency.

Choice latency was defined as the time from middle lever press until one of the outer (choice) levers was pressed. The figure depicts the mean (±SEM) choice latency for each delay when rats chose the adjusting (a) or 150 μl (b) levers for each rat strain.

Sucrose solution preference

All strains consumed more sucrose as the concentration of sucrose increased (F_3.6,115.0 = 1362.04, P < 0.001)(Fig. 4a). However, there were also significant strain differences (F_5,32 = 23.95, P < 0.001), with some strains more sensitive to changes in sucrose concentration than others (concentration × strain interaction F_18.0,115.0 = 12.66, P < 0.001). There was also a main effect of start date (F_1,32 = 59.93, P < 0.01), a strain × start date interaction (F_5,32 = 2.80, P < 0.05), a concentration × start date interaction (F_3.60,115.0 = 7.01, P < 0.001), and a concentration × strain × start date interaction (F_17.98,115.0 = 3.24, P < 0.001). The animals that ran in the April 2007 group tended to drink slightly more than the animals that ran in the July 2006 group. This effect was nominal at the 10% sucrose concentration. We focused our post-hoc comparisons on the 10% sucrose concentration (the concentration of sucrose used in the delay discounting experiment) and performed a follow-up univariate anova with strain as a factor. There were significant strain differences in consumption of 10% sucrose (F_5,38 = 10.61, P < 0.001)(Fig. 4b). Post-hoc Tukey comparisons indicate that Noble rats drank more sucrose than all but Copenhagen rats (P = 0.08) at this concentration of sucrose (all other P's ≤0.01). Copenhagen rats also drank more sucrose than Fischer (P < 0.05) and Wistar Furth rats (P < 0.01). Post-hoc tests showed no other significant differences between strains. The calculated heritability (eqn 3) was 0.57.

Figure 4. Consumption (a) and preference (b) for varying concentrations of sucrose.

Rats were given 50 min of access to two bottles, one containing tap water and a second bottle containing sucrose at the concentrations indicated. Consumption, relative to body weight, and preference were averaged over four sessions at each concentration of sucrose.

The relationship between sucrose consumption and other experimental measures is shown in Table 3. Consumption of 10% sucrose was negatively correlated with k-values derived from delay discounting. This indicates that the rats that drank the most sucrose, expressed as a function of their weight, had discount curves with the most gradual slopes (were least sensitive to devaluation induced by delay).

Table 3.

Spearman's rho correlation coefficients (P-values) for experimental measures from the six inbred rat strains

	b	k	Response latency	Choice latency	10% Sucrose preference	Day 1 locomotor activity
k	0.600 (0.21)
Response latency	0.543 (0.27)	−0.029 (0.96)
Choice latency	0.486 (0.33)	−0.314 (0.54)	0.886 (0.02)
10% Sucrose	−0.314 (0.54)	−0.829 (0.04)	−0.029 (0.96)	0.371 (0.47)
Day 1 locomotor*	0.429 (0.40)	−0.314 (0.54)	0.429 (0.40)	0.771 (0.07)	0.600 (0.21)
Locomotor habituation*	0.600 (0.21)	−0.029 (0.96)	0.600 (0.21)	0.771 (0.07)	0.429 (0.40)	0.829 (0.04)

Correlations between performance measures significant at the P < 0.05 level (n = 6) are indicated by bold type. 10% Sucrose is the amount of sucrose consumed in grams per kilogram of body weight. Bias (b) and k-values were derived from the hyperbolic fit equation.

^*Denotes April 2007 animals only.

All strains more strongly preferred the sucrose solution over water as the concentration of sucrose increased (F_6.72,214.97 = 242.36, P < 0.001). There were strain differences in sucrose preference (F_5,32 = 3.39, P < 0.05), as well as concentration × strain (F_33.59,214.97 = 1.61, P < 0.05), and concentration × start date (F_6.72,214.97 = 6.73, P < 0.001) interactions. All other main effects and interactions were not significant. At the 10% sucrose concentration, preferences ranged from 88.1 ± 1.4% (Wistar Furth rats) to 93.7 ± 0.9% (Lewis rats). A simple, univariate anova showed significant strain differences in preference for 10% sucrose (F_5,38 = 2.90, P < 0.05), with post-hoc Tukey comparisons indicating that Lewis and Noble rats more strongly preferred the 10% sucrose concentration than the Wistar Furth rats (P's <0.05). There were no other significant differences in sucrose preference between strains.

Locomotor activity

A one-way anova indicated that there were significant strain differences in locomotor activity on day 1 (F_5,17 = 7.18, P < 0.001) (Table 4) with calculated heritability (eqn 3) equal to 0.62. Tukey post-hoc comparisons of activity indicated that the Wistar Furth and Lewis rats were significantly less active than most other rat lines. A second anova indicated no significant differences in level of habituation (activity on day 1 minus activity on day 2) (F_5,17 = 2.10, P = 0.12). Neither locomotor activity nor habituation correlated with any aspects of delay discounting or sucrose drinking (Table 3).

Table 4.

Mean total horizontal activity ± SEM (cm) during the 30-min session for the six inbred rat strains

Strain	n	Day 1 (cm)	Habituation (cm)
Brown Norway	4	2943 ± 106*	614 ± 542
Copenhagen	4	3091 ± 274*	337 ± 273
Fischer	4	3281 ± 553**,***	896 ± 202
Lewis	4	1897 ± 221	442 ± 325
Noble	4	3560 ± 128**,***	1285 ± 77
Wistar Furth	3	1400 ± 303	−249 ± 494

Tukey post-hoc tests were carried out to examine line differences in activity. Significant differences are denoted as

^*P < 0.05,

^**P < 0.01 line was more active than Wistar Furth rats,

^***P < 0.05 line was more active than Lewis rats. Habituation is the difference in activity from day 1 to day 2 (activity on day 1 − activity on day 2).

Discussion

The primary goal of this study was to determine if there are significant strain differences in aversion to delayed reinforcement in rats. Our results support the presence of strain differences in k. Based on this, we calculated the degree of heritability for this trait, which, although smaller than the heritability for locomotor activity or sucrose consumption, was considerable. This value for heritability is not purely genetic, however, and includes environmental influences, such as differences in maternal care. Although the present study did not cross-foster the inbred strains to a common maternal background, future studies that do so could be compared with this study to examine the impact of maternal care on impulsivity. The design of the present study allows for the generation of F₂ lines by selecting pairs of progenitor strains from the panel of inbred strains tested, which may subsequently allow for the mapping of delay discounting to specific quantitative trait loci and eventually, specific genes. One previous study using a different, but presumably related delay discounting procedure and four inbred mouse strains (C57BL/6J, CBA/Ca, 129/Sv and BALB/c) also supported the presence of a genetic influence on task performance (Isles et al. 2004). Similarly, work by Anderson and Woolverton (2005) found that inbred Lewis rats were more averse to delays than inbred Fischer rats in a delay discounting procedure that employed discrete sets of trials with increasing delays throughout the session (within-sessions procedure; Evenden & Ryan 1996). Our study showed no significant differences in performance between these two rat strains, but this may be because of procedural differences between the within-sessions procedure and the adjusting amount procedure used here. The dependent variable in the within-sessions procedure is the percent choice of the large reinforcer assessed at each delay. By contrast, the dependent variable of the adjusting amount procedure is the indifference point at each delay. Studies have not compared the within-sessions procedures and the adjusting amount procedures to determine if they provide comparable estimates of delay aversion. Strain differences in memory, ability to distinguish the volume of reinforcer and attention may differentially impact responding in these tasks. Furthermore, the delay times used in the two tasks are not the same, with delays as long as 60–100 seconds in the within-sessions task and maximal delays of 16 seconds in the adjusting amount task. This large temporal disparity may impact responding as the mechanisms involved in tolerating a delay of several seconds may differ significantly from mechanisms involved in tolerating a delay on the order of a minute or more. Previous studies also indicate a disparity between the two procedures, with ablation of the nucleus accumbens decreasing delay aversion (Acheson et al. 2006) in the adjusting amount procedure and ablation of the nucleus accumbens core increasing delay aversion (Cardinal et al. 2001) in the within-sessions procedure. Thus, while it is disturbing that the two procedures do not appear to provide equivalent data, it is not without precedent in the literature.

While a considerable amount of work has been carried out examining strain differences between Fischer and Lewis rats (e.g. Anderson & Elcoro 2007), there are few studies comparing behavior of inbred rat lines other than the Fischer and Lewis rats. Rex et al. (1996) found that Brown Norway rats were more likely to consume food in an open field and to explore familiar locations outside of their homecage than Fischer rats, indicating that this strain may have a lower level of anxiety. Ramos et al. (1997) found similar trends in anxiety in examining behavior in an elevated plus maze and a black and white box, with Brown Norway consistently exhibiting lower levels of anxiety (significant or trends) than Wistar Furth, Lewis or Fischer rats. A lower level of anxiety could be related to the tolerance animals have for delay, i.e. strains with lower levels of anxiety may be more confident that they will be alive or be able to collect rewards that are delivered after a delay, while strains with higher levels of anxiety may worry that they will be attacked or eaten before they are able to retrieve delayed rewards, or may worry that the rewards will not be delivered as expected. Consistent with this hypothesis, Brown Norway rats had lower k-values than Fischer, Lewis or Wistar Furth rats; however, this result was not statistically significant.

Equipment limitations meant that the same adjusting amount procedures were applied in two distinct squads, with four rats from each strain per squad. There were no significant squad effects on measures of delay discounting, suggesting that the impact of strain differences was robust and likely not the result of differences in rearing, circa-annual variations etc. Nevertheless, we cannot rule out the possibility that differences in rearing or slight differences in early environment at the vendor impacted the experimental results. It is also important to note that strain differences were only responsible for a portion of the observed variance. Furthermore, the variation in behavior attributable to strain was higher than delay discounting for both sucrose consumption (10%) and locomotor activity.

Neither response latency nor choice latency was correlated with k or consumption of 10% sucrose; however, response latency was positively correlated with choice latency. In addition to level of deprivation, response and choice latency are likely affected by many factors including attention, visual acuity, learning and operant responsivity (the ease with which an animal learns and responds in an operant task). This experiment was not designed to distinguish between these factors, so strain differences in any of these factors may have influenced response latency and choice latency.

Consumption and preference for sucrose was assessed in each of the inbred strains to determine if there were differences in motivation for the primary reinforcer. The results indicated that Lewis and Noble rats had a higher preference for the sucrose solution than the Wistar Furth rats. Although there was a statistically significant difference, the effect size (<5% difference in preference) is marginal and therefore probably not behaviorally relevant. Furthermore, there was no statistically significant difference in delay discounting performance between Wistar Furth and Lewis rats. Consumption of 10% sucrose was negatively correlated with k, but not b. Thus, animals that drank more of the freely available 10% sucrose were less sensitive to devaluation of the large reward caused by imposing a delay to reward delivery. This relationship was also observed in a delay discounting task examining rat strains selectively bred for high or low alcohol consumption (Wilhelm & Mitchell 2008). Noble and Copenhagen rats drank the most 10% sucrose and also had the smallest k-values. Tordoff et al. (2008) found that Noble and Copenhagen rats drank more sucrose (3.4%) in a 48-h two-bottle choice procedure than Lewis, Fischer or Brown Norway rats with no strain differences in preference between sucrose and water. Interestingly, Noble rats also tended to have slower choice latencies, indicating that the underlying relationship between consumption of sucrose and discount rate is not related to level of motivation, as previous studies suggest that response latencies would be quicker with increased motivation (Richards et al. 1997).

Activity in an open field is one measure of an animal's response to novelty (see Bardo et al. 1996 for review). However, inter-strain comparisons appear to produce variable data. Similar to the current study, Bardo et al. (personal communication, August 2008) observed high relative levels of activity in Fischer compared with Wistar Furth rats. However, unlike the current study, they also observed Brown Norway and Lewis rats to have relatively low and high levels of activity respectively. By contrast, Berton et al. (1997) found Brown Norway and Wistar Furth rats to have high relative activity, while Lewis and Fischer rats had relatively low levels of activity. The source of this between study variability is unclear; however, differences in age or prior experience may be responsible for the apparent differences between these studies. Previous work by Isles et al. (2004) found that the genetic component of locomotor activity was negatively correlated with the genetic component of response bias in mice using a within-sessions procedure. The results presented here do not support a correlation between locomotor activity and delay discounting in rats. Procedural differences may explain the disparate findings, but the current study is limited by the relatively small sample size. Thus the lack of relationship between discounting performance and locomotor activity may be as the result of a lack of power and is therefore not conclusive. However, our findings are consistent with those of Perry et al. (2005) who found no relationship between delay discounting performance and locomotor activity among female Wistar Furth rats. Further, Wilhelm and Mitchell (2008) also observed no relationship between locomotor activity and delay discounting among selectively bred rat lines. Other tasks designed to measure different aspects of impulsivity such as behavioral inhibition, for example, the go/no-go, the five choice serial reaction time and the stop tasks, which assess the ability of subjects to inhibit a prepotent motor response, may be more likely to correlate with locomotor activity.

Heightened impulsivity is associated with numerous neuropsychiatric disorders and in some cases, may predate development of these disorders. Our data showing strain differences in delay discounting suggest that this behavior may have a genetic basis. This has significant implications for disorders characterized by heightened aversion to delayed rewards. It is unclear if there will be strain or genetic influences on other behavioral measures of impulsivity like behavioral inhibition tasks (e.g. go/no-go or stop task). Recent work in two independently selected lines of high- or low-alcohol-drinking rats (HAD and LAD rats respectively) suggests that prior to alcohol exposure, HAD rats are more sensitive to delayed rewards than LAD rats (Wilhelm & Mitchell 2008). Understanding the molecular mechanisms that contribute to decision-making in this task may provide significant insight into the mechanisms that lead to these disorders. Early studies suggest that the molecular mechanisms that govern this process may be complex and are likely not mediated by a single gene product in humans (e.g. Eisenberg et al. 2007). Future studies examining specific genes and gene combinations may provide viable pharmacotherapeutic targets for treatment of disorders characterized by heightened impulsivity.