Circadian regulation of insect olfactory learning

Abstract

Olfactory learning in insects has been used extensively for studies on the neurobiology, genetics, and molecular biology of learning and memory. We show here that the ability of the cockroach Leucophaea maderae to acquire olfactory memories is regulated by the circadian system. We investigated the effect of training and testing at different circadian phases on performance in an odor-discrimination test administered 30 min after training (short-term memory) or 48 h after training (long-term memory). When odor preference was tested by allowing animals to choose between two odors (peppermint and vanilla), untrained cockroaches showed a clear preference for vanilla at all circadian phases, indicating that there was no circadian modulation of initial odor preference or ability to discriminate between odors. After differential conditioning, in which peppermint odor was associated with a positive unconditioned stimulus of sucrose solution and vanilla odor was associated with a negative unconditioned stimulus of saline solution, cockroaches conditioned in the early subjective night showed a strong preference for peppermint and retained the memory for at least 2 days. Animals trained and tested at other circadian phases showed significant deficits in performance for both short- and long-term memory. Performance depended on the circadian time (CT) of training, not the CT of testing, and results indicate that memory acquisition rather than retention or recall is modulated by the circadian system. The data suggest that the circadian system can have profound effects on olfactory learning in insects.

In view of the widespread effects of circadian regulation on behavior and physiology, the possibility that learning and memory processes may be subject to modulation by the circadian system has long been recognized (1). Studies on this problem have suggested that the influence of the circadian system may take several forms. In some instances, the circadian phase may function as a context for learning (time stamping) so that recall and performance are better at 24-h intervals after learning (1–4). In other cases, performance may be modulated by the circadian phase independent of the phase of learning (5), and there have been several recent reports that memory acquisition or consolidation may depend on the circadian phase of training in mollusks (6–8), rodents (9, 10), and humans (11). Finally, there have been reports that disruption of the circadian system by phase shifting (jet lag) can impair memory (12, 13). Although these studies suggest that the circadian system may have widespread effects on various aspects of learning and memory, including acquisition, retention, and recall, there is virtually no information on the mechanisms by which the circadian system regulates these processes.

Olfactory learning is an excellent model for exploration of the neural basis of learning and memory because of the remarkable similarity in the organization of olfactory systems across a phylogenetically diverse population of organisms that includes arthropods and mammals (14, 15). Olfactory learning in insects, in particular, is widespread and has been extensively used as a model for studies on the behavior, neurobiology, genetics, and molecular biology of learning and memory in bees (e.g., 16, 17), fruit flies (15, 18, 19), and moths (20). As a consequence, much is understood about the anatomical substrate and physiological and molecular mechanisms of insect olfactory learning. Circadian systems of insects also are well understood at the physiological and molecular levels (21–24). Thus, olfactory learning in insects could prove to be a useful model for investigation into the mechanisms of the circadian modulation of learning and memory. Efforts to examine the role of the circadian system in insect learning are, thus far, limited to a few studies on Drosophila melanogaster. In the fruit fly, recent work has indicated that one of the clock genes, the period gene, plays a role in long-term memory formation, although mutations in other clock genes did not affect long-term memory (25). The result led to the conclusion that a functional clock is not necessary for memory formation in the fly; however, these experiments did not directly address the question of whether there is a circadian modulation of long-term memory. To date, there are no studies that bear on the question of how the circadian phase might impact olfactory learning in insects.

There is reason to believe that the cockroach may be an excellent model for such studies. It has been shown that, in a differential conditioning paradigm, the cockroach, Periplaneta americana, can be readily trained to associate peppermint odor with a sucrose reward and to retain the memory for >1 week (26, 27). In addition, in previous studies, we have shown that the circadian system regulates both olfactory reception (28) and olfactory-driven mating behavior (29) in the cockroach Leucophaea maderae, indicating that the circadian system has an important regulatory function in both sensory input and behavioral output involving the olfactory system. In the present paper, we present evidence that the circadian system has profound effects on olfactory learning in L. maderae. We show that, with differential olfactory conditioning, Leucophaea can be trained to approach an otherwise aversive odor and that the memory lasts at least 2 days. Further, we show that the acquisition of memory depends on the circadian phase, but the ability of the animals to recall a previously learned task is independent of the circadian system. These results open opportunities for the study of learning and memory and its modulation by the circadian clock.

Results

Does L. Maderae Exhibit the Ability to Form Long-Term Olfactory Memories?

When given a choice between vanilla or peppermint odors and tested at circadian time (CT) 14 for odor preference by allowing the animals an opportunity to approach and contact (visit) one or the other odor sources, naive animals showed a clear preference for vanilla. Odor preference of each animal was quantified by using a peppermint preference index (PPI), which was defined as 100n_p/(n_v + n_p), with n_p being the number of visits to the peppermint odor source and n_v being the number of visits to the vanilla odor source during the odor preference test (27). Naive animals exhibited an average PPI of 20.4 ± 16.2 (mean ± SD; n = 23) (Fig. 1A). When tested again at CT 14, 48 h after training in which peppermint was paired with a reward of sucrose solution and vanilla was paired with a punishment of saline solution, these same animals now exhibited a distinct preference for peppermint with an average PPI of 73.6 ± 17.0, and all 23 animals showed an increase in PPI (Fig. 1B). The change in PPI was highly significant (P < 0.0001, Wilcoxon signed rank test). These results show that L. maderae exhibits excellent olfactory learning when trained and tested in the early subjective night of the circadian cycle, and that the memory lasts at least 2 days.

Fig. 1.

Long-term learning and memory depends on the circadian phase. The graphs plot the distribution of PPI expressed by individuals either before (A and C) or after (B and D) conditioning. Animals were isolated in constant darkness 3 days before the first preference test. (A) Distribution of 23 naive animals tested for preference for peppermint or vanilla at CT 14 on day 3 of constant darkness. The clustering near the origin shows the strong preference for vanilla. These animals underwent conditioning at CT 14 on day 5 of constant darkness and were tested again on day 7. (B) Results of the preference test 48 h after conditioning. There is a strong preference for peppermint, indicating learning and long-term memory. (C and D) Same sequence of testing, training, and retesting at CT 2 for a group of 19 animals. There is no significant change in the preference for vanilla, indicating a profound deficit at this phase of the circadian cycle.

As a control, we also did training trials in which vanilla was paired with the sucrose and peppermint with saline. Before testing, these animals exhibited a PPI of 18.4 ± 16.9. When tested after 48 h, the animals showed no significant change in preference for the vanilla odor (PPI = 20.8 ± 30.9; P > 0.8; n = 10), indicating that odor preference was indeed associated with the unconditioned stimulus (US) and that the change in odor preference observed when peppermint was paired with sucrose was not a nonspecific consequence of training.

Does the Ability to Form and Recall Long-Term Memories Depend on the Circadian Phase?

Leucophaea is nocturnal, and early subjective night (CT 14) represents the approximate phase of peak locomotor activity for these animals. To determine whether the circadian clock regulates the ability of these animals to form memories, we performed the same testing/training/testing protocol on animals at CT 2 in the early subjective morning of the circadian cycle when the normal level of activity is lowest. Naive animals tested at CT 2 showed a similar preference to vanilla as naive animals tested at CT 14 (Fig. 1C; PPI = 14.4 ± 14.7; n = 19). The results indicated that there was no difference between the early subjective day and the early subjective night in the animals' innate odor preference or in their ability to discriminate between the two odors. Remarkably, when we trained these animals at CT 2 (peppermint paired with the sucrose and vanilla with the saline) and tested them again 48 h after training, there was no significant change in odor preference (Fig. 1D; PPI = 7.7 ± 13.8; P > 0.30). The effect of training was an overall slight decline in PPI, and only 4 of 19 animals (21%) exhibited an increase in PPI over pretraining values.

In view of the circadian rhythm of spontaneous locomotor activity, it could be suggested that the lack of performance at CT 2 was simply due to an overall decline in activity. Therefore, we also examined the average number of visits each group of animals made to the odors. For the 23 animals tested at CT 14, the average number of visits was 5.3 ± 2.4 before training and 4.7 ± 2.4 after training. There were only slightly fewer visits for animals tested at CT 2. Before training the average number of visits was 4.3 ± 1.9, whereas after training the number of visits averaged 3.2 ± 2.3 (n = 19). There was no significant effect of the circadian phase on the number of visits to odor sources animals made before training or on the number of visits after training (ANOVA, P > 0.05), indicating that differences in the levels of activity were not a factor in the disparity in performance between animals trained and tested at the two CTs.

In further experiments, we also trained and tested animals at two additional CTs: CT 8, the late subjective day; and CT 20, the late subjective night. At CT 8, 48% (11 of 23) of the animals showed an increase in preference for peppermint, and at CT 20, 43% (9 of 21) of the animals showed an increase (compared with 21% at CT 2 and 100% at CT 14). The results are summarized in Fig. 2. Before training, there was no effect of CT on odor preference or ability to discriminate between odors, and animals consistently preferred vanilla (Fig. 2A). When animals were tested 48 h after training at CT 2, CT 8, CT 14, and CT 20, there was a clear dependence of performance on CT (Fig. 2 B and C) with a peak in performance at CT 14 (P < 0.0001, Kruskal–Wallis one-way ANOVA). These results show that the circadian system regulates long-term olfactory learning in the cockroach.

Fig. 2.

Circadian regulation of long-term olfactory learning and memory in the cockroach. (A) Mean PPI (±SE) for naive animals tested at CT 2 (n = 19), CT 8 (n = 23), CT 14 (n = 23), and CT 20 (n = 21). All groups showed a strong preference for vanilla, indicating that the ability to distinguish between the odors is independent of the circadian phase. (B) Mean PPI when animals were retested 48 h after training. (C) Change in PPI between the pre- and posttraining tests. Only those animals trained and tested at CT 14 showed a significant increase in the preference for peppermint (Kruskal–Wallis one-way ANOVA, P < 0.0001). Significant differences between groups (P < 0.05, Dunn's post hoc analysis) are indicated by differences in letters at the top of each bar. The results indicate that there is a circadian modulation of long-term olfactory learning and memory. (D) Percentage of saline or sucrose consumed at either CT 2 or CT 14. There is no circadian variation in the response to the US.

Is the Circadian Rhythm in Learning and Memory Related to the Circadian Regulation of Sensory Responses to the Conditioned or Unconditioned Stimuli?

Cockroaches exhibit a circadian rhythm in the sensitivity of the olfactory receptors in the antennae with an ≈5-fold higher sensitivity in the early subjective day than in the early subjective night (28). Thus, to control for the possibility that the increase in olfactory sensitivity at CT 2 results in a strong aversion to peppermint, which prevents conditioning, we repeated the experiments at CT 2 by using a 5-fold lower concentration (4 μl) of peppermint for both training and testing. These animals still preferred vanilla over peppermint both before and after training, and there was no improvement in performance. The PPI before training was 12.8 ± 20.4, and 48 h after training it was 10.9 ± 17.4 (n = 18).

The possibility that the deficit in learning at CT 2 was related to a rhythm in the response to the unconditioned stimuli also was examined by testing the animals' response when given a choice between saline and sucrose solutions at both CT 2 and CT 14. Animals that were isolated from food and water for 2 days (n = 12 at each phase) were simultaneously offered 75 μl of 20% NaCl and 75 μl 10% sucrose at either CT 2 or CT 14. Animals would typically touch the saline solution with their mouth parts and then avoid it while consuming the entire bolus of sucrose solution. The response in terms of percentage of each solution ingested was virtually identical at CT 2 and CT 14, and the roaches consumed nearly all of the sucrose solution and little of the saline solution in both phases (Fig. 2D). The results indicate that it is unlikely that a circadian modulation of the response to the unconditioned stimuli is responsible for the learning deficit at CT 2.

Is the Circadian Rhythm in Performance Due to Memory Acquisition/Retention Deficits or Memory Recall Deficits?

The ability to perform in the olfactory learning task depends on the ability of the animals to acquire and consolidate a memory during training and to recall the memory and act on it during testing. To determine which aspect was deficient in animals trained and tested at CT 2, we performed two additional experiments. First, animals that were trained at CT 2 were tested at CT 14 (when it was clear they could recall). Second, we tested animals that were trained at CT 14 (a phase at which it was clear they could learn) at CT 2. In both cases, testing occurred 36 h after training. There was a significant increase in PPI in animals that were trained in the early subjective night (CT 14) and tested in the early subjective day (CT 2) (P < 0.001, Wilcoxon signed rank test), whereas there was essentially no change in odor preference in animals trained in the subjective day and tested in the subjective night (Fig. 3A). These results suggest that the deficit in performance seen in animals trained and tested at CT 2 was because of an inability to form new memories or retain memories formed at this phase and was not because of an inability to recall memories already established. To confirm this conclusion, we trained animals at CT 14 and tested them after 36 h (CT 2) and again after 48 h (CT 14). Before training, animals (n = 15) exhibited a strong preference for vanilla (Fig. 3B). After training (at CT 14), the animals showed a clear preference for peppermint when tested at CT 2. The performance was only slightly better when retested 12 h later at CT 14, and the difference was not significant (P > 0.10). These results show that, once the memory is formed by training at CT 14, animals are able to recall and use this information in the early subjective day (CT 2), as well as the early subjective night (CT 14). Performance only depends on the CT of training and is independent of the time of testing. Thus, the time of training impacts either memory acquisition or retention and not memory recall.

Fig. 3.

The circadian phase of training is more important than the phase of testing. (A) Change in PPI for animals tested at CT 2, trained at CT 14, and retested at CT 2, 36 h after training (n = 19), compared with animals tested at CT 14, trained at CT 2, and retested at CT 14, 36 h after training (n = 24). (B) PPI for animals tested before training at CT 14 and retested 36 h after training (CT 2) and again 48 h after training (CT 14). In both of the posttraining tests, the animals showed a significant increase in preference for peppermint (P < 0.0001), and there no significant difference between the two posttraining test times. The data indicate that recall/performance is independent of CT of testing and only depends on the CT of training.

Is Short-Term Learning and Memory Regulated by the Circadian System?

Long-term memory formation involves distinct processes that distinguish it from the formation of short-term memories. Thus, clues about which aspects of the physiological steps in memory formation are regulated by the circadian system might be obtained from examining circadian regulation of short- as well as long-term memory formation. To determine the effect of CT on short-term memory, we trained animals at CT 2, CT 8, CT 14, and CT 20 and tested them after 30 min. The results, shown in Fig. 4, indicate that short- and long-term memory depends on the circadian phase. As expected, naive animals showed a distinct preference for vanilla at all phases (although, in this instance, there was a slight, but statistically significant, difference between CT 2 and CT 8) (Fig. 4A). After training, performance was clearly modulated by the circadian system (Fig. 4 B and C). Once again, animals trained at CT 14 outperformed those trained at CT 2, whereas performance at CT 8 and CT 20 was intermediate (P < 0.0001, Kruskal–Wallis one-way ANOVA). The results show that short-term memory is regulated by the circadian system.

Fig. 4.

Circadian regulation of short-term olfactory learning in the cockroach. (A) Mean PPI (± SE) for naive animals tested at CT 2 (n = 26), CT 8 (n = 21), CT 14 (n = 22), and CT 20 (n = 22). All groups preferred vanilla. There is a slight but significant difference in the data between CT 2 and CT 8 (P < 0.05, Kruskal–Wallis one-way ANOVA with Dunn's post hoc analysis) that is largely due to a particularly strong preference for vanilla of the animals tested at CT 2. (B) Change in PPI between the pre- and posttraining tests. There was a significant dependence on CT (Kruskal–Wallis one-way ANOVA, P < 0.001). Significant differences between groups (P < 0.05, Dunn's post hoc analysis) are indicated by differences in letters at the top of each bar. (C) Mean PPI when animals were retested 30 min after training. (D) Results of testing at CT 2 and CT 14 before training and 5 min after training. At CT 2, training had no effect, whereas at CT 14, there is a large shift in the PPI (P < 0.001).

The results are open to two interpretations of the cause of the deficit in performance at CT 2. First, as might be suggested from the fact that performance at CT 2 was not different from the performance of a naive animal, it could simply be that animals are unable to acquire new memories at this phase. Alternatively, it might be that acquisition is normal but retention is severely impaired, and animals are simply unable to retain the memories for even 30 min. To distinguish between these two interpretations, we tested animals 5 min after training at either CT 2 or CT 14. The results are shown in Fig. 4D. Animals trained at CT 2 were still unable to perform when tested 5 min after training. The fact that performance 5 min, 30 min, or 48 h after training at CT 2 was not different and, importantly, was not significantly different from the performance of the naive animal strongly suggests that it is memory acquisition, rather than retention, that is under circadian control.

Discussion

Our results show that olfactory learning and memory in the cockroach is modulated by the circadian system, and that the effectiveness of classical conditioning is strongly dependent on the circadian phase. These data raise several questions concerning the underlying pathways and mechanisms by which the circadian system influences the process of learning and memory. Learning and memory involve several conceptually distinct steps that include the processes of acquisition of memory subsequently followed by retention, retrieval, and performance. A priori, given the widespread impact of the circadian system on behavior and physiology, the circadian phase could conceivably have an impact on any one (or more) of these processes.

Memory Acquisition vs. Memory Retrieval.

One central question is whether the deficits observed when animals were trained and tested in the subjective morning were due to deficits in the ability of the animals to retrieve and/or use memories to direct behavior. In this case, the failure to perform could be due to a general, phase-dependent deficit in the ability to recall a learned memory that was completely independent of training time and memory acquisition or retention. We found that animals performed equally well in the early subjective morning and in the late subject night when trained at CT 14. Congruently, the failure to perform when trained at CT 2 also was independent of the circadian phase of testing. Animals trained at CT 2 were unable to perform in both the early subjective morning and the early subjective night. The results indicated that the ability to recall a previously learned task was independent of the circadian phase and further suggested that the CT at training was not an additional contextual cue (time stamping) necessary for performance. Coupled with the fact that animals trained at CT 2 were unable to recall a memory after only 5–30 min, these results suggest that the inability of animals to perform when trained at CT 2 was due to an inability to form new memories at this circadian phase and focus attention on the earliest stages of olfactory memory formation, rather than retention or memory retrieval. This notion is generally consistent with the recent discovery in mice that early long-term potentiation in the hippocampus is regulated by the circadian system (30), but is contrasted in Aplysia, where the circadian system modulates long-term memory, but short-term memory appears to be independent of the circadian phase (6, 8).

Circadian Organization in the Cockroach.

Two known circadian rhythms in cockroaches could indirectly affect memory acquisition or performance in olfactory learning and memory tasks: (i) the circadian rhythm in the activity/rest cycle (31), and (ii) a circadian rhythm in the sensitivity of olfactory receptors in the antennae (28, 29). Curiously, the peak of olfactory sensitivity is in the subjective morning when the animals are least active and are deficient in olfactory learning.

Conceivably, the rhythm in sensory input could affect the animal's ability to detect or discriminate between odors or may alter the perceptual attractiveness or aversiveness of a particular odor. Similarly, the rhythm in the activity rest cycle could have an impact on the animal's ability to behaviorally respond to the conditioned stimuli. One significant advantage of the differential conditioning paradigm we used is the ability to evaluate the naive animals' capacity to detect and discriminate between odors and to exhibit a behavioral preference. Our results suggest that these factors are independent of the circadian phase in the testing protocol that we used. Naive animals were able to discriminate between the two odors equally well at all CTs and exhibited a consistent preference for vanilla over peppermint. Further, the analysis of the animals' propensity to visit a preferred odor showed that there was no significant difference between CT 2 and CT 14, the trough and peak phases, respectively, of spontaneous locomotor activity. Thus, there was no evidence to suggest that the circadian system had any impact on the ability to detect the odors, discriminate between the two conditioned stimuli, or behaviorally respond. Another possibility is that the increase in olfactory sensitivity in the early subjective day results in peppermint becoming so aversive that animals simply could not be conditioned to approach the odor at that phase. However, we showed that neither the avoidance of peppermint by naive animals nor the phase-dependent ability to be conditioned to approach peppermint was affected by a 5-fold reduction in the peppermint stimulus. Further, animals conditioned at CT 14 readily approach peppermint when tested at CT 2, indicating that the level of peppermint is not prohibitive. Finally, we also showed that the behavioral response to the unconditioned stimuli was independent of the circadian phase; animals clearly preferred sucrose and avoided saline at both CT 2 and CT 14. These results indicate it is unlikely that altered perception of the sucrose as a positive reinforcement and saline as a negative reinforcement account for the rhythm. In summary, we were unable to obtain any evidence that would suggest a rhythm in the primary sensory response to either the conditioned or unconditioned stimuli used in this study could account for the circadian rhythm in memory acquisition.

The Circadian Pathway.

The discovery that the circadian system regulates memory acquisition raises questions about the site of generation of the oscillation as well as the pathway by which the oscillator modulates learning. In cockroaches, one well known circadian clock that regulates spontaneous locomotor activity has been localized to the optic lobes (24, 32). Evidence suggests that the optic lobe clock also is involved in the regulation of visual sensitivity of the compound eye (33) and olfactory sensitivity in the antennae (28). Thus, the optic lobe is a prime candidate for the site of the circadian oscillator that regulates olfactory learning. However, there is compelling evidence from other insects (34, 35) and cockroaches (29, 36) that multiple, anatomically distributed oscillators exist, and it is plausible that these oscillators may influence learning and memory processes. In this context, it is interesting to note that recent experiments have shown that in Aplysia the well known circadian pacemaker in the eye is dispensable to the circadian regulation of long-term memory formation (37). Additional experiments to localize the site of the pacemaker for regulation of learning and memory should provide some clues about the pathway by which the circadian system regulates the formation of olfactory memories.

The Olfactory Pathway.

In insects, the olfactory pathway begins with the olfactory receptors, which project to glomeruli of the antennal lobe where they form synaptic connection with both projection neurons and local interneurons. The projection neurons, in turn, send axons to the lateral lobe of the protocerebrum and to the calyces of the mushroom bodies (MBs). Early studies (15, 19) indicated that the MBs play a critical role in olfactory memory and suggested that olfactory memories are formed upstream of the output neurons of the MBs, which determine performance, and that the memory trace probably lies within the MBs. More recently, studies using food or electric shock as a US in a variety of insects, including bees (38), flies (39), and moths (20), have shown that learning-dependent changes in neural activity also occur as early as the antennal lobe, which therefore appears to be the first site in the olfactory pathway for convergence of the conditioned stimuli and US and thus may be a site for the circadian regulation of memory acquisition in the cockroach. US pathways that lead to this convergence are not well understood; however, there are data that suggest that aminergic input may play a pivotal role. Particularly relevant are observations in honey bees that implicate the octopaminergic, ventral unpaired medial neurons in the transmission of the sucrose US. The ventral unpaired medial neurons are located in the subesophageal ganglion and project to the antennal lobes, MBs, and lateral horn and are stimulated by the contact of sucrose with the bee's proboscis (40). Further, Drosophila mutants deficient in a key enzyme in the octopamine synthesis pathway, tyramine β-hydroxylase, exhibit deficient learning when sucrose is used as the US, but normal learning with electric shock as the US (41). The mutant can be rescued by feeding the fly octopamine before training, which restores the fly's ability to learn with sucrose as the US. Within this neural pathway, several molecular signaling cascades have been implicated with a primary focus (in the MBs) on the cAMP second messenger and the activation of cAMP-dependent protein kinase (PKA) and, for the formation of long-term memory, CREB-dependent gene expression. At this point, we cannot restrict the potential site (anatomical or molecular) at which the circadian clock might modulate the ability to acquire memories. However, the extensive data that are available on the neural pathways and molecular mechanisms of insect olfactory learning make this problem tractable.

Adaptive Significance.

A final question of interest is, what is the adaptive value of circadian regulation of learning and memory? The answer to this question is not at all clear. In most instances where circadian control of memory formation has been demonstrated, animals (and humans) perform better with training during the active phase. For example, in the diurnally active A. californica long-term memory was better during the day, whereas in the nocturnal species A. fasciata training was significantly more effective at night, leading to the suggestion that the adaptive advantage of circadian modulation was tied in some way to the concomitant activity/rest cycle (8). One plausible suggestion that emerges is that memories are only profitable when formed in the environmental context in which they will be used. The acquisition of these memories is likely to be an expensive process, both in the energy expenditure required and in the neural tissue and connections devoted to memory storage. Because the environment is distinctly periodic and because cockroaches are almost exclusively active at night and spend the daytime at rest in dark hides, one could argue that there is little to be gained from the formation of memories of the daytime olfactory environment. These memories would modify the animal's foraging behavior based on information obtained at a time and in an environment in which the animal is unlikely to be foraging. Such memories might even interfere with successful foraging in the rather different environmental conditions the animal would experience during its normal foraging in the nighttime hours.

Materials and Methods

Animals.

Colonies of L. maderae were maintained in a light-tight box in a 12-h light/dark cycle (4-W fluorescent lamp, 5 μE/m² per sec). The box was housed in an environmental room maintained at 24.5°C. All experimental procedures were carried out in the environmental room. At the beginning of the experiment, adult male cockroaches were isolated in 12 × 8 × 6-cm square plastic containers 3 days before initial odor preference testing. During this time, the cockroaches were kept on the light/dark cycle and deprived of food and water. Before lights-on on the third day of isolation, the animals were placed in constant darkness.

Training and Testing Schedule.

The method of olfactory training and testing was adapted from Watanabe et al. (27). On the third day after isolating the animals (first day of constant darkness), an odor preference test was performed at estimated CTs of CT 2, CT 8, CT 14, or CT 20. CT estimates were based on the time of lights-off of the prior light cycle (designated as CT 12) and assumed a free-running period of 24 h. Only animals that showed activity within the first 5 min of initial testing were retained for training and further testing.

Immediately after the initial odor preference test, the active animals were placed in plastic tubes for training. Forty-eight hours after the initial odor preference test, training was performed, and animals were placed back into the square plastic containers until they were tested for memory retention. In another orthopteran, Gryllus bimaculatus, olfactory memory for a task similar to that used in the present study was insensitive to inhibition of protein synthesis for 2–4 h after training, but protein synthesis was necessary for animals to retain memories for >5 h after training (42). Thus, we refer to memories that persisted for 24–48 h as long-term memories and memories expressed within 2 h after training as short-term memories. For long-term memory experiments, animals were subjected to memory retention tests either 36 or 48 h after training. For short-term memory experiments, the memory retention test was performed either 5 or 30 min after training. Training was carried out under a dim red light.

Odor Preference Test.

The testing arena was a circular Plexiglas chamber 30 cm in diameter and 9.5 cm tall. The bottom of the arena was lined with a circle of white poster board, with a 51-mm-diameter circle indicating the center. The sides of the arena were smeared with petroleum jelly to prevent the animals from escaping. For each odor source, a 2-cm-square piece of filter paper was placed in the bottom of a 35 × 10-mm polystyrene Petri dish and treated with 20 μl of either pure peppermint extract or pure vanilla extract. Twenty-four small holes were drilled in the cover of each Petri dish. There were four odor sources, two of peppermint and two of vanilla.

At the beginning of testing, the odor sources were evenly spaced around the perimeter of the testing arena. One cockroach was placed in the center of the circle and observed until it had visited one of the odor sources. A visit was recorded when a cockroach probed the source for 2–3 sec with its head (mouthparts) and feet. After each visit, the arrangement of the odor sources was randomly changed, and the cockroach was prodded back to the center of the arena and allowed to continue foraging. The time of each visit and the total number of visits were recorded for each animal for 10 min. During the test, the animals were observed with a CCD camera (Sony XC-77; Sony, Tokyo, Japan) in dim red light. Illumination was provided by a darkroom safelight equipped with a filter that limited wavelengths to >600 nm (Kodak 1A or GBX-2; Eastman Kodak, Rochester, NY). Light intensity was adjusted with a rheostat to a final intensity at the floor of the testing arena <0.1 mE-m⁻²/sec⁻¹ (Li-Cor photometer; Li-Cor, Lincoln, NE).

Differential Conditioning Procedure.

Cockroaches were trained in 10.5-cm-long and 3.0-cm-diameter cylindrical plastic tubes that constrained the animal's movement. The animals were subjected to a conditioning trial in which peppermint odor was paired with 10% sucrose solution, followed by a trial in which vanilla odor was paired with 20% NaCl solution. One-milliliter syringes equipped with a 20-gauge needle were used to apply the solutions. A 6 × 6-mm piece of filter paper was soaked with 15 μl of either peppermint or vanilla extract and attached 1 cm from the tip of the needle. During odor presentation, odor-soaked filter paper was placed ≈1 cm from the animal's head between the antennae. Within 4 sec after odor presentation, one drop of the corresponding solution was placed on the cockroach's mouthparts.

Each animal was subjected to three conditioning trials. Each trial consisted of presentation of the peppermint odor associated with sucrose solution followed, after 5 min, by the presentation of the vanilla odor associated with saline solution. Trials were separated by 5-min intervals.

Response to US.

Animals that were isolated for 48 h without food and water in constant darkness were simultaneously offered 75 μl of 20% NaCl solution and 10% sucrose at either CT 2 or CT 14 and allowed 15 min to consume the solutions (in constant darkness). The amount of each solution consumed was determined by weight.

Data Analysis.

The Wilcoxon signed rank test was used to compare odor before and after training within a group. Comparisons among groups were made by using Kruskal–Wallace ANOVA, followed by Dunn's test; ≈15% of the animals showed no visits to either odor source and were not included in data analyses.

Abbreviations

CT

circadian time

PPI

peppermint preference index

US

unconditioned stimulus.