The present studies examine some parameters involved in flavor avoidance learning, using LiCl to induce malaise, in a novel nondeprivation protocol that allows direct comparison between rats and mice. The procedure involves daily presentation of a gelatin dessert that contains carbohydrate (Polycose) and a distinctive food flavor. Regular chow is additionally available at all times. Both rats and mice showed robust intakes of these gels with little change of gram intake as concentration of Polycose was varied in the range 2-30%; at the highest concentration, the caloric yield was ∼7% of normal daily intake in both species. Rats that were injected on three occasions with LiCl (0.75 meq/kg) 1 h after consumption of a flavored gel formed a complete and sustained conditioned flavor avoidance (CFA). In a two-flavor discrimination protocol, in which a second flavor was followed by injections of saline, rats showed complete avoidance of the LiCl-paired flavor and partial avoidance of the saline-paired flavor. Mice injected on three occasions with LiCl (6 meq/kg) 1 h after intake of a flavored gel formed a partial CFA; a more complete CFA was formed when there was no delay between removal of the flavor and the injection. Using this no-delay protocol, mice, like rats, showed avoidance of a saline-paired flavor in a two-flavor discrimination protocol, and the CFA was strong when the dose of LiCl was reduced to that used in rats (0.75 meq/kg). In comparable protocols, mice thus are able to form complete CFAs using low doses of LiCl that are comparable to CFAs observed in rats, but the interval between flavor and sickness over which associative learning can occur may be shorter in mice.
- associative learning
- conditional stimulus
- taste aversion
- caloric density
there is an extensive literature on taste aversion learning, the associative process by which a neutral or a preferred taste becomes rejected or avoided as a result of an unpleasant postingestional outcome after its initial consumption (1, 4, 5, 10). This type of learning, in combination with a naturally occurring conservative sampling strategy of novel foods (neophobia), clearly has adaptive significance because it allows animals to avoid foods that may be poisonous or spoiled. Harmful foods usually will not cause malaise until substantial absorption has occurred, possibly an hour or more later. Thus, unlike other types of associative learning, taste aversion learning occurs even with long delays between the episode of ingestion and its outcome (4). However, similar to other types of avoidance learning, aversion can be strong after only a single pairing of the conditional stimulus (CS: food) and the unconditional stimulus (US: malaise). Most taste aversion studies have been performed in rats, but with a growing focus on genetic analyses, a small number of papers using mice have appeared recently (6, 7, 11). These studies, using several strains of mice, typically have found that mice show less complete avoidance than rats.
The objective of the above-mentioned papers was not comparison of rats and mice; indeed, there has been no systematic examination of the parameters of food avoidance and the underlying associative processes in mice. Our interest in this problem was not, however, founded in a context of associative learning but rather in work on species differences in anorexia and apparent tolerance to the action of an appetite suppressant dexfenfluramine using a nondeprivation or “dessert” protocol (8, 9). As a positive control agent in these studies, we used LiCl, the standard agent used to induce malaise in taste aversion studies. This protocol that we developed is not typical of those used in the majority of taste avoidance studies but offers some advantages. First, current professional and regulatory guidelines for laboratory animal care seek to limit deprivation procedures, so a demonstration of a viable alternative not involving deprivation has some utility. Second, when comparing species differing in size and metabolic rate, it is unlikely that the relative deprivation and/or motivational states are identical after an identical duration of deprivation (e.g., 24 h). Last, naturally occurring malaise-inducing agents most likely occur in foods rather than fluids and so in most cases are associated with calories and smell as well as taste.
In these studies, we use solid foods (gels) containing defined concentrations of a pure but relatively nonsweet oligosaccharide, Polycose, mixed with various food flavors that serve as cues (conditional stimuli, CSs) to form behavioral discriminations. This protocol is a conditioned flavor avoidance (CFA) because it uses flavors that have both odor and taste and because measures of intake provide a behavioral index of avoidance but do not necessarily imply aversive affect. In previous work from our laboratory using rats and either sucrose/milk or Polycose solutions as the caloric carriers for these CS flavors (8), we found almost complete (>90%) avoidance of a flavor (the CS+) after only two occasions or pairings when its consumption was followed by injection of a low dose (0.75 meq/kg) of LiCl.
In contrast, when using C57BL/6J mice and a gel of flavored Polycose, we found that LiCl (3 meq/kg) produced only modest (<50%) reduction in CS+ intake, even after six pairings (8). These and other findings suggest that mice form a CFA less readily than rats. However, the protocols differed between the rats and mice insofar as we used gels contained in glass beakers for mice and solutions in drinking spouts for the rats. It is unlikely that this difference is attributable to CS intensity because mice eat with their snouts above the surface of this gel, so their exposure to the flavor should be at least comparable to a solution in the orifice of a drinking spout. The use of a gel could introduce other differences from a solution, such as the stimulus being a compound of the CS flavor and the texture of the gel and/or taste of Polycose, and the apparent species difference may relate to stimulus compounding rather than CFA. It is thus important to perform parametric studies with the gel diet in comparable studies in rats and mice.
The organization of this manuscript is to present the rat studies first, followed by the results from the mouse studies. Within each species, the first experiments assess acceptability of the gel as calories are varied. The next experiments examine selected parameters of avoidance of a single stimulus, including delay between ingestion and injection of LiCl, number of pairings, dose of LiCl, and extinction of the CFA. The final experiments examine a diet choice protocol in which, during the learning trials, one flavor (the CS+) is followed by injection of LiCl and another flavor (the CS-) is followed by injection of placebo. As there are likely to be strain differences in taste aversion or CFA in mice (7), we elected to use a standard outbred strain that would be expected to give modal results in these studies.
Animals and housing. Male and female adult Sprague-Dawley rats (Harlan, Indianapolis IN) aged between 3 and 9 mo were used in the rat studies. This is a standard strain that has been used in many published studies on taste and flavor avoidance. For at least 1 wk before and during the studies, they were housed individually in stainless steel mesh cages (17 × 17 × 25 cm) suspended above absorbent paper that was changed twice weekly. The vivarium was maintained at 23 ± 1°C with a 12:12-h light-dark cycle (lights on from 0600 to 1800). All tests were performed in the early afternoon. Purina chow pellets (no. 5001; PMI, Brentwood MO) and tap water were available ad libitum, including during dessert tests.
Male and female adult ICR:CD-1 mice (Charles River, Newton, MA) aged between 3 and 9 mo were used in the mouse studies. Before and during these studies, they were housed individually in clear polycarbonate cages (13 × 10 × 10 cm) with stainless steel grill lids. A 1- to 2-cm depth of Sani-Chips bedding (Harlan-Teklad, Madison WI) was present in each cage, and this was changed twice weekly. Mice were housed in a separate vivarium from rats. The details of food and water availability and light cycle were identical to those described above for rats.
Some of the rats and mice had been used beforehand in unrelated studies that did not involve flavors or avoidance learning; no animal was used in more than one flavor avoidance procedure.
Test diet and presentation. For ∼1 wk, animals were adapted to receive a gel diet presented in a glass container (100-ml food jars for rats, 10-ml beakers for mice) inside their home cages for 30 min daily during the middle part of the day. A few animals did not eat the diet when first presented, so for them the diet was left in overnight for the first day only. The containers were held by springs (rats) or custom-made stirrup holders (mice) that were fashioned from flexible metal strap and ring retainers (9). Intakes were determined by weighing the containers before and after each session. Gels were prepared by dissolving partially hydrolyzed corn starch (Polycose, Ross Labs; the concentration is given in each study) in hot water, adding 25 g/l gelatin (Knox brand) and ∼150 μl of a commercial food flavor extract (McCormick), and pouring into the containers. The gel was allowed to cool and solidify overnight in a refrigerator. Diets were kept in the cold for up to 3 days and were allowed to warm to room temperature before presentation.
The intakes on the last 3-4 days of adaptation were averaged to calculate the individual baseline intakes; day-to-day variability was typically 10-20%. In the rare cases that spillage occurred, data were discarded. Test phase data were expressed as percentage of that baseline. Within a given study, animals were assigned after the adaptation phase to groups matched for mean baseline intakes. Some published taste aversion papers have reported sex differences (2, 10), so some of our experiments were designed to compare males and females. Except as noted, sex differences were not significant, and male and female data were combined for analysis. This result led us to test only one sex in some of the experiments, but we expect that these results would not have differed had the other sex been included.
Experiment 1: influence of caloric density on gel intake in rats. Fifteen male rats ∼5 mo of age and weighing 500 g were divided into three squads of five and initially received for 5 days each either 5, 20, or 30% wt/vol pineapple-flavored Polycose gel. These concentrations correspond to approximately 0.2, 0.8, and 1.2 kcal/g (does not include gelatin). In subsequent 4- to 5-day segments, the alternate concentrations were given such that all rats received all three concentrations. In the next phase, the concentration was reduced in all rats to 2%, then 1%, 0.5%, and eventually to 0%. Intakes from all diet phases, including the end of the adaptation phase (10%), were combined for statistical analysis (1-way repeated-measures ANOVA with Newman-Keuls comparisons, α < 0.05; SigmaStat, SPSS).
Experiment 2: flavor avoidance in rats and effect of number of pairings. Eighteen male rats ∼3 mo of age and weighing between 267 and 316 g at the start of the study were assigned to three treatment groups. The gel used in this experiment contained 10% Polycose, in the middle of the range studied in experiment 1, presented as before for 30 min. For half of the rats in each group, the adaptation flavor was coconut, and for the other half it was banana. In the conditioning phase, rats received as their CS+ the alternate flavor from adaptation (i.e., banana-adapted rats received novel coconut, and vice-versa), and trials were conducted every 2-3 days with no dessert presentation on the intervening days. Two groups received intraperitoneal injections of LiCl (Fisher Scientific, Orlando FL), dissolved in distilled water in a concentration of 0.15 M and injected in a dose of 5 ml/kg or 0.75 meq/kg after the CS+. The first group received one flavor-LiCl pairing, with LiCl injected 1 h after the dessert was removed on the first day. On the subsequent six test days, they received the same CS+ but no injection (i.e., an extinction condition). The second group received three flavor-LiCl pairings on the first three test days and no injections on the subsequent four test days. Rats in the third group received intraperitoneal injections of NaCl (5 ml/kg, 0.15 M). Half of these rats received one saline injection, and the other half received three injections. The data from these two control conditions were combined for analysis. Data were expressed as percentage of the baseline for each rat and were analyzed by 2-way ANOVA with group and days as main factors. This was followed up by planned 1-way ANOVAs for each day and Newman-Keuls subsequent tests with a significance criterion of P < 0.05.
This experiment was part of a larger study, the results of which will not be reported in detail. In male rats run concurrently, the effect of a longer delay (5 h) between CS+ and LiCl was examined. In a later study, female rats were run to assess possible sex differences in acquisition or extinction of the CFA. We have omitted these data from full presentation because we do not believe that they are critical to the central theme of this paper, but they will be mentioned briefly in the discussion.
Experiment 3: two-flavor discriminative learning in rats. Two groups of six male rats, initially weighing 348-452 g, were adapted to the 10% Polycose gel diet flavored with pineapple. During the conditioning phase they received, on alternating and consecutive days, two novel flavors. For half of the animals, banana was designated as the CS+ and was presented on days 1, 3, and 5; for these rats, coconut was the CS- and was presented on days 2, 4, and 6. For the other half of the rats, the CS+ was coconut and the CS- was banana. These flavors were chosen on the basis of pilot work because they are distinctive, readily accepted, and did not noticeably bias or influence the results. The control group received intraperitoneal injections of 0.15 M NaCl 1 h after the dessert was removed each day. The experimental group was injected with LiCl (0.75 meq/kg; 5 ml/kg of 0.15 M) after ingestion of the CS+ on days 1, 3, and 5 and NaCl after ingestion of the CS- on days 2, 4, and 6. On day 7, all rats received two jars of dessert simultaneously, one flavored with the CS+ and the other with the CS-. Intakes were expressed as %baselines and analyzed by 2-way ANOVAs followed for the conditioning phase by between-group t-tests and for the preference by within- and between-group t-tests.
Experiment 4: influence of caloric density on gel intake in mice. Thirteen female mice ∼4 mo of age and weighing ∼30 g were adapted for ∼1 wk to pineapple flavored 10% Polycose; intakes from the last 2 days were averaged. They were then divided into three groups of 3-4 animals, approximately matched for baseline intakes, and received 2, 5, and 20% Polycose gels, also flavored pineapple, and presented in counterbalanced order across groups with 3 days at each concentration, and the data were averaged for the last 2 days. The mean individual intakes from all four phases (2, 5, 10, and 20%) were analyzed by repeated-measures ANOVA.
The acceptability of lower concentrations of Polycose was examined in a second study using 11 male and 9 female mice that had served as controls in a previous study with 10% Polycose. After several weeks off study, they were readapted using 2% pineapple-flavored Polycose gel, then received 1% and 0.5% (3 days each); the last 2 days' data were analyzed as above using repeated-measures ANOVA.
Experiment 5: effect of CS-US interval on CFA in mice. In the first study, to complement experiment 2 in rats, the CS-US interval was 1 h. Twelve male and 12 female mice were divided into control and experimental groups of six of each sex. Males weighed a mean of 43 g and females 33 g. They were adapted to 10% Polycose gels flavored with banana or coconut. After baselines were stable, mice in the experimental groups were presented with the alternate (novel) CS+ flavor and 1 h after diet was removed were injected intraperitoneally with LiCl (6 meq/kg; 40 ml/kg of 0.15 M). This dose is eightfold higher than that we used in rats but is based on effective doses in mice determined previously by ourselves and others (6-8, 11). Mice in the control groups received isotonic saline injections (40 ml/kg). Four conditioning trials were conducted at 2-day intervals, followed by five extinction trials at 2- to 3-day intervals, when no injections were given. Intakes were transformed to % of baseline and analyzed as before by 2-way ANOVAs for effects of sex and treatment, day and treatment, with subsequent Newman-Keuls comparisons
In the second study the CS-US interval was shortened; the LiCl was injected immediately when the CS+ was removed. Because no sex difference was evident in the first part of this experiment, only females were used in this part and were assigned to control (n = 6) and experimental (n = 7) groups matched for baseline intake of pineapple-flavored 10% Polycose gel dessert. Starting with the first conditioning day, and on all subsequent tests, all mice received coconut-flavored gel for 30 min as the CS+. Mice in the experimental group were injected intraperitoneally with LiCl (6 meq/kg; 40 ml/kg of 0.15 M) immediately when the CS+ was removed; controls received injections of isotonic saline. Three conditioning trials were conducted at 2-day intervals, followed by five extinction tests at 2- to 3-day intervals. Intakes were analyzed statistically as before.
Experiment 6. Two-flavor discriminative learning in mice: effect of caloric density and dose of LiCl. In a previously published study the focus of which was dexfenfluramine (8), we reported some degree of discriminative CFA in mice using a 10% Polycose diet and a 1-h delay to injection of LiCl (6 meq/kg). However, experiment 5 suggested that a shorter CS-US interval should produce stronger CFA. Given that this may be a more suitable delay, and two-flavor discrimination tasks are usually more sensitive than one-flavor protocols, we used a discriminative task and examined whether the factors of caloric density or amount of LiCl injected would influence the CFA. A much higher dose of LiCl has typically been found necessary in mice compared with rats, but a viable reason for this has not been offered for lack of systematic examination of the parameters of the conditioning procedure.
In the first study, three groups of 12 female mice weighing 28-40 g were adapted to daily presentation of pineapple-flavored gel of either 2, 10, or 27% Polycose content. After intakes stabilized, mice received on the first day of the conditioning phase a novel flavored gel (designated as the CS+) of the same Polycose content as during adaptation. This was followed with no delay after the end of 30 min access by intraperitoneal injection of LiCl (6 meq/kg; 40 ml/kg of 0.15 M) in the experimental groups and 0.15 M NaCl in control groups. Half of the mice in each group received novel banana flavor and half received coconut. Intakes were recorded as before. The next day, all mice received the alternate flavor, designated as the CS-, followed by NaCl injection. The procedure on day 1 was repeated on days 3 and 5, and the procedure on day 2 was repeated on days 4 and 6. Thus mice in the experimental groups received three pairings of CS+ with subsequent LiCl and three pairings of the CS- with saline. On the seventh day, all mice received a two-jar choice between banana- and coconut-flavored gels, and intakes were again recorded. Intakes expressed as %baselines were analyzed by two-way ANOVA for each day of the study.
In the second study, 24 male mice were adapted to daily 30-min presentation of pineapple-flavored 10% Polycose gel, as before. They were then divided into four groups matched for baseline intakes. These groups underwent a discriminative stimulus procedure identical to that in the first study except that the injections were either saline (control group) or 0.75, 2.25, or 6 meq/kg LiCl. The injections were given intraperitoneally as 0.15 M solutions in a volume of either 5, 15, or 40 ml/kg immediately when the dessert was removed. Two mice each in the control group received 5, 15, or 40 ml/kg isotonic saline, and their data were combined for statistical analysis as in the first part of the experiment.
Experiment 1: influence of caloric density on gel intake in rats. Rats readily accepted the dessert and, within a few days, typically consumed a single large meal as soon as it was presented and were resting by the end of the 30-min session. Figure 1 shows that the intake by weight of the dessert was relatively constant across Polycose concentrations in the range 2-30% but fell off at lower concentrations. Some rats did not consume at the lowest two concentrations. Repeated-measures ANOVA showed a highly significant effect of concentration (P < 0.001), and post hoc comparisons showed that intakes at 0, 0.5, and 1% were lower than all other concentrations (P values < 0.05), except that 30 and 1% did not differ. Intakes at the highest five concentrations (viz: 2-30%) did not differ among themselves, and the mean intakes were 1-2% of body weight. In this concentration range the mean energy intake, estimated using a value of 4 kcal/g Polycose and ignoring the small but invariant caloric content of gelatin itself, varied by >10-fold (Fig. 1).
Experiment 2: flavor avoidance in rats and effect of number of pairings. The mean baseline intakes in this experiment ranged from 3.4 to 5.6 g but did not differ significantly between groups. The results are shown in Fig. 2. On the first conditioning trial (day 1) when a novel flavor was presented, but before any agents were injected, all groups of rats consumed ∼100% of their baseline. In saline-injected controls, the intake of that flavor increased by 40-60% on subsequent days.
The two-way ANOVA found a significant main effect of group, with all groups differing from each other, and a group × days interaction (P values < 0.001). As indicated in Fig. 2, the group that received a single LiCl pairing with the CS+ reduced its intake the next day significantly with respect to the controls, and by ∼50% relative to its own baseline, showing the formation of a partial but statistically reliable CFA. On the subsequent days, intake slowly recovered to and exceeded baseline, indicating complete extinction of the CFA. The group that received three pairings of LiCl with the CS+ showed a comparable effect to the previous group after the first pairing, but after three pairings the intake was almost zero, indicating that the CFA became complete. There was no recovery of intake, and so no evidence of extinction of the CFA, over the subsequent 4 days of no injections.
Experiment 3: two-flavor discriminative learning in rats. The mean baseline intakes for the two groups were 9.3 and 8.7 g. The results from the conditioning phase are shown in Fig. 3A. Neither group showed significant neophobia to the novel flavor on the first day. On subsequent days, with alternating flavors, the intakes of the control group stayed near baseline. In contrast, the intakes of the LiCl group were significantly reduced regardless of whether the flavor was CS+ or CS-, indicating that avoidance of the CS+ generalized substantially to the CS-. The results from the choice test are shown in Fig. 3B. ANOVA of these choice data showed a significant group effect (P < 0.001) and an interaction of group with flavor (P < 0.05). The control group consumed statistically similar amounts of each flavor, with a total of 94% baseline. The LiCl group consumed less of both the CS+ and the CS- than the controls but consumed more CS- than CS+, whence the interaction term. This result confirms the generalized avoidance but indicates some discriminative effect in the choice protocol that was not evident in the conditioning single-stimulus trials.
Experiment 4: influence of caloric density on gel intake in mice. The results from the first part of the experiment using the higher concentration range and male mice are shown in Fig. 4A. Intake varied significantly as a function of the concentration of Polycose (overall ANOVA, P < 0.001) and was highest at the 10% adaptation solution (P < 0.05 vs. all other concentrations). However, estimated mean energy intake varied by >10-fold across this range.
In the second part of the experiment using the lower concentrations and both males and females, the overall ANOVA revealed significant differences in intake as a function of both concentration (P < 0.05) and sex (P < 0.01). The results shown in Fig. 4B indicate that females consumed on average ∼50% more dessert than males. Within each sex, intakes were lower (1-way ANOVA) at 1 and 0.5% concentrations than 2%, and the Newman-Keuls contrasts are indicated in Fig. 4. Unlike rats (Fig. 1), all of the mice continued eating substantial amounts at the lowest concentrations.
Experiment 5: effect of CS-US interval on CFA in mice. In the first part of this experiment, there was no difference in the mean baseline intakes of Polycose gel of males (1.81) and females (1.78). There was also no difference between males and females in the results from the conditioning phase so the data are presented with both sexes combined in Fig. 5A. There was a small (15-20%) reduction in intake when the novel flavor was introduced on day 1, but after a few days, the intake of the saline-injected controls returned to baseline. The intakes of mice injected with LiCl 1 h after dessert for the first 4 days showed significantly lower intakes than controls by the third day (i.e., after 2 prior CS-US pairings) and all days thereafter. The two-way ANOVA showed a highly significant main effect of treatment, days, and treatment × days interaction (P values < 0.001). Intake of the LiCl group was a nadir of 31% baseline after four pairings and approximately doubled over the 5 days of extinction but still remained lower than controls at the end of this time.
The failure of mice to completely reduce their intake of the CS+ after three to four pairings is clearly unlike rats (see Fig. 2), but we noted that 4 of 12 LiCl-injected mice had near-zero intakes after four pairings, suggesting that it is possible to produce a complete avoidance in mice using this general protocol. In the second part of this study, we examined whether a shorter CS-US interval might support a stronger CFA.
The results of the second study (no CS-US delay) are shown in Fig. 5B. One mouse in the LiCl group failed to form a flavor avoidance (intake >80% of its baseline on all trials) and was omitted from analysis. The remaining six LiCl-treated mice formed strong flavor avoidance, with an intake of 12.7 ± 5.5% (mean ± SE) of baseline after three conditioning trials. An unplanned comparison between this intake and that of females in the first study (41.3 ± 12.1% after 4 conditioning trials; datum in Fig. 5A) approached statistical significance (t-test: P < 0.06). These data suggest that mice can form a near-complete CFA but require a shorter CS-US interval than is effective in rats.
Experiment 6. Two-flavor discriminative learning in mice: effect of caloric density and dose of LiCl. In the caloric density study, the respective baselines for 2, 10, and 27% Polycose groups were 1.03 ± 0.14, 1.99 ± 0.09, and 1.54 ± 0.11 g (ANOVA P < 0.001), with all three differing from each other (P < 0.05) by Newman-Keuls tests. These results are consistent with those reported in experiment 4. The results from the conditioning phase of the first study are shown in Fig. 6A. The CS+ intake of the groups treated with LiCl were significantly lower than intake of the control group (P values < 0.05) on each day after the first (on this day the intake occurred before any treatment). The groups presented with either 2 or 10% Polycose seemed to form an avoidance more rapidly than the 27% group, with the 2 and 10% groups consuming a lower fraction of baseline than the 27% group on CS+ days 2 and 3. The only interaction between concentration and treatment was on day 6.
At the choice test (Fig. 6B), the total amount consumed relative to baseline and the fraction of intake as CS+ showed strong treatment effects (P values < 0.001) but no significant difference as a function of Polycose concentration. Thus LiCl-treated mice consumed significantly less of the CS+ than the CS-, whereas the controls showed no systematic difference between the two flavors. The total intake of the LiCl-treated mice was only ∼50% of baseline, whereas the controls consumed at or above baseline. The intake of the CS- did not differ between the control and corresponding LiCl groups. Thus mice showed good discriminative learning with little influence of concurrent calories in the CSs on the CFA.
In the second part of the study examining different doses of LiCl, the intakes of CS+ and CS- declined across conditioning trials in all three groups (Fig. 7). At the final choice test, mice from each LiCl group showed a significant avoidance of the CS+. There is evidence for more extensive avoidance of the CS+ and generalization to the CS- in the group receiving the highest dose of LiCl. Thus, on the third presentation of the CS+, six of six mice in the 6 meq/kg group consumed no gel at all; on the subsequent CS- day, three of six mice failed to eat; and at the choice test, four of six mice failed to eat either diet. In contrast, in the 0.75 meq LiCl group, only one of six mice failed to eat at the third presentation of CS+ and at the choice test, and zero of six mice failed to eat on the last CS- day. In the choice test, four of six of the low-dose group failed to eat any CS+, despite the fact that the intakes of CS+ by these mice on the third conditioning trial ranged from 34 to 56% of baseline.
Previous studies in mice found that they formed incomplete taste aversions even with high doses of LiCl (6, 7, 11). This could be because mice cannot form full avoidance and/or are excessively deprived or are relatively insensitive to LiCl. The present experiments addressed these hypotheses by eliminating the deprivation and then optimizing parameters. Because this protocol differs from that used typically in rat taste aversion studies, it was necessary to establish the procedure in rats. We will first summarize our rat findings within the context of the broader taste aversion literature.
In experiment 1, we determined that the gram intake of the Polycose gel dessert was independent of caloric density in the concentration range 2-30%, or 0.08-1.2 kcal/g. Below a concentration of 2%, intake fell and most rats ate minimal amounts of the “zero calorie” version. Rats are not food deprived in this protocol, so this is truly elective intake, and it is interesting that it seems to be controlled primarily by factors other than caloric content (e.g., flavor, texture, volume), perhaps analogous to human snacking. This target food thus seems to have some desirable properties for studying CFA.
In experiment 2, we examined the effect of number of CS-US pairings on CFA. As has been observed in other protocols (5), one pairing formed a transient avoidance while three pairings produced a complete and sustained CFA. In one of the additional studies that were run but not reported, male and female rats were compared with two CS-US pairings in an otherwise identical protocol. Two pairings produced effects intermediate between those for one and three pairings shown in Fig. 2, and there were no significant sex differences except there was a trend to a quicker extinction in females as reported by others (2, 10). In another additional study using males run concurrently with experiment 2, we examined the effects of lengthening the CS-US interval to 5 h. This delay completely prevented formation of a CFA with intakes of the CS never dropping below a mean of 100% of baseline even after three pairings. Some previous studies have reported taste aversions over longer CS-US intervals, and one might expect that the present rats, that have no physiological need to consume the dessert, would readily form such avoidance (4). However, we used a relatively low dose of LiCl, near the threshold previously reported to obtain a full aversion in one trial in a fluid deprivation protocol (5), so it is possible that if we had used a higher dose then stronger conditioning would have occurred at long delays.
In experiment 3, we examined the ability of rats to discriminate between two novel flavors, one of which was paired with LiCl. As expected, a strong avoidance of the CS+ was observed, but almost as robust an avoidance of the CS- occurred. This was evident both during the single-flavor conditioning trails and in the two-flavor preference test. This largely replicates our previous result using a similar protocol but either sweetened milk or Polycose solutions (8). There was no evidence of weight loss during the study, suggesting that intake of regular chow was unaffected. It could be argued that the salience of the flavor in the CS was insufficient, so producing an avoidance based on common properties such as texture, Polycose flavor, or the time and mode of presentation. That is, the CS is a compound of these properties, plus flavor. The fact that every rat in the LiCl group consumed some CS- in the choice test and all consumed more CS- than CS+, suggests the flavor components were discriminable but did not have full stimulus control. This result has the appearance of generalized avoidance and/or prolonged neophobia for a novel CS- (1), and this could in principle be determined in future studies. This task also is well within memory capacity of rats: for example, remembering sequences of 10 or more food-associated odors has been used as a declarative memory span task in rats (3).
Results in mice. As was the case in rats, mice readily consumed the gel desserts. Maximal intakes were ∼5% body weight and, at the highest concentrations, yielded ∼7% of normal daily caloric intake. The dessert intake was relatively unaffected by Polycose concentration. As in rats, the intake changed relatively little as concentration varied in the range 2-20%. Further, in an unpublished study, the mean intake of 27% Polycose gel was 1.29 g or only slightly less than the intake of 20% reported in the present paper. The maximal energy intakes from these gels were ∼7% of normal daily intakes in both rats and mice, roughly the equivalent of one meal.
In a previous paper using female C57BL/6J mice (8), we used a discriminative stimulus procedure and 3 meq/kg LiCl as the US injected with no delay after the dessert was removed. In that study, we found a maximum of ∼50% avoidance of the CS+. We expected that the use of a single stimulus and a higher dose of US would improve the magnitude of the avoidance relative to this previous work. This hypothesis was supported insofar as the mean minimum intake with no delay was only ∼10% of control (Fig. 5B) and was ∼50% (comparable to our previous result) when the delay was 1 h. The results using this nondeprivation protocol can be compared with published studies using fluid restriction. In wild-type C57BL6 × 129/Sv mice, a single pairing of novel 5% sucrose solution (for 30 min) followed immediately by the US injected yielded ∼50% avoidance of sucrose lasting only one trial when the US was 3 meq/kg LiCl (11). When the dose was increased to 6 meq/kg, initial avoidance was ∼75% and extinguished more gradually. In the present work, using no CS-US delay a more complete initial avoidance was obtained (∼90%), but three pairings were used. In another study using fluid restriction and 0.2 M NaCl as the CS, C57BL/6J mice showed a maximum of ∼75% avoidance after four pairings of the CS with 6 meq/kg LiCl (7). No effect was obtained with lower doses of LiCl in this strain of mice, but DBA/2 mice formed a small avoidance at lower doses (6, 7). In that protocol, the fluid was available for 1 h and the injection was given immediately thereafter. Since most of the fluid was most likely consumed within the first few minutes of access, the effective CS-US delay is in fact almost 1 h and so the protocol is more comparable to that reported in Fig. 5A, and the results are quite similar. This 1-h delay was fully effective in rats using lower LiCl doses.
To return to the hypotheses annunciated at the start of this discussion, are mice insensitive to LiCl, or are they less able than rats to withhold ingestive responses? In the final studies, we attempted to answer these questions using the discriminative stimulus protocol in mice. Previously, using C57BL/6 mice and a high concentration (27%) of Polycose, we found relatively minor avoidance of either CS+ or CS- (8). First, we asked whether the high caloric density used in that study might make the animals less able to reject the food. In Fig. 6, it is evident that higher levels of avoidance were obtained using 2% compared with either 10 or 27% gels, although for reasons we do not understand the intake of some control mice became erratic. It is evident that at all three concentrations the total intake on the choice test was only ∼50% of baseline, indicative of a generalized avoidance, although the overwhelming preference for the CS- at that test is indicative of a specific CFA. In the final study, to investigate sensitivity to LiCl, we used the intermediate (10%) Polycose concentration and no delay before US injection. All three LiCl doses produced significant CFAs although, as might be expected (5), higher doses produced somewhat larger effects. This is, to our knowledge, the first report of CFA in mice using a dose of LiCl <3 meq/kg. Our data suggest that mice are able to form a complete CFA, but what distinguishes them from rats is a steeper gradient of delay of the flavor information available to be linked associatively with sickness. Further studies will be needed to assess this hypothesis, as well as possible functional and mechanistic differences between tastes and flavors as conditional stimuli.
In performing species comparisons of the actions of the anorectic agent dexfenfluramine (8), it became clear to us that the basic parameters of flavor or taste avoidance had not been well examined in mice. In particular, this issue of equating motivational states with larger species or even across different strains of mice has not been addressed. We adapted the gel diets that we developed for measuring actions of anorectic agents in mice (9) to study flavor avoidance. These diets have the advantage over traditional solutions used in taste avoidance insofar as they are solid like most real foods, they contain calories that can be manipulated, and there are an almost limitless number of flavors that can be tested. The results using this protocol in rats are very similar to those obtained using pure tastes and fluids: this was expected because the same associative learning mechanisms should be involved. The fact that similar results can be obtained without food or fluid restriction should be of applicability within the context of animal welfare concerns. Likewise, it provides a protocol that can be extended to other species, to other USs (8), to larger memory span protocols (3), while making minimal assumptions about underlying physiological state.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 the American Physiological Society