Genetic dissociation of acquisition and memory strength in the heat-box spatial learning paradigm in Drosophila

Reinforcer/memory level relationship

A critical requirement in addressing genetically the reinforcement intensity/memory strength relationship in Drosophila is a fast robust conditioning paradigm that shows this correlation. That Drosophila is capable of varying memory strength has been demonstrated in both classical olfactory conditioning and operant visual learning (Tully and Quinn 1985; Brembs 2000). These paradigms are not ideal, however, in screening for mutations, as they are labor intensive or require many hundreds of flies. The heat-box spatial learning paradigm, however, was developed at least in part to screen for learning mutations (Wustmann et al. 1996). It is fast and requires relatively few flies, and robust memories are formed that last minutes to hours (Putz and Heisenberg 2002). To determine whether wild-type Canton-S (CS) flies would vary their place memory strength based on reinforcing temperature, they were trained with different temperature reinforcement for increasingly longer periods. With increased training time, wild-type CS flies reach what appears to be a plateau in memory performance, with the increase in performance between 10 and 20 min of training reduced compared with that found early in the training session (i.e., in the first 4 min) (Fig. 2A). To determine whether memory performance after 20 min of training is indeed asymptotic, wild-type flies were trained 15, 20, or 25 min using either 37° or 45°C as negative reinforcement (Fig. 2B). No significant differences between these different training sessions indicate that 20 min of training induces asymptotic memory, at least in the temperature range tested. Furthermore, memory performance displayed by flies is higher when higher temperature reinforcement is used. Thus, flies modulate their memory strength in the heat-box depending on the temperature of reinforcement. This provides a paradigm in which this phenomenon can be addressed genetically in Drosophila.

The white-ABC transporter affects spatial conditioning

Mutations in the white (w) and rutabaga (rut) genes affect avoidance behavior during training and display memory performance deficits in the heat-box learning paradigm. The white mutation was found to be defective in the heat-box spatial learning paradigm as part of a mutant screen (a nonvisual task) (Fig. 3). A pilot screen of ∼100 extant lines used 4 min of training with a 37°C reinforcer. These conditions were chosen as they quickly induce a submaximal yet robust memory. The rut allele rut²⁰⁸⁰ (Levin et al. 1992) was used as a control learning mutant, as it has proven defective in all learning paradigms in which it has been tested. In the heat-box, the rut²⁰⁸⁰ mutant flies have lower avoidance behavior during training and a short-lived avoidance memory (Fig. 3B), consistent with previous findings (Wustmann et al. 1996; Zars et al. 2000b). This poor performance is the specific consequence of a decrease in rut function, as transgenic expression of the wild-type form of rut can rescue this phenotype (Zars et al. 2000b). A more severe avoidance behavior phenotype and low-memory performance was found in white mutant flies (Fig. 3C), a strain initially considered a control for many of the screened lines. Low-avoidance behavior and poor memory performance could be due to a low-memory acquisition rate or altered high-temperature reinforcement processing. Extended training experiments can differentiate between these possibilities (see below).

The white gene is implicated as critical for place conditioning. The white mutant flies identified in this screen have the w¹¹¹⁸ null mutant allele. They are called wCS13, as they have been out-crossed to the wild-type strain CS for 13 generations. As each out-crossing generation involved more than 20 individuals, and this was done 13 times, it is likely that the white mutation causes the conditioning phenotype. Further mapping of the mutant phenotype took advantage of an X-chromosome duplication containing a wild-type form of the white gene attached to the Y-chromosome. This was out-crossed for six generations to the poor learning wCS13 strain. The wCS13/Dp(1:y)w+ flies show wild-type-like avoidance performance and memory (Fig. 3D), making it even more likely that it is a mutation in the white gene that is responsible for the poor conditioning. As the wCS13 flies provide the strongest behavioral phenotype in the heat-box to date and the rut²⁰⁸⁰ mutation affects a classic learning gene (a type-1 adenylyl cyclase) (Han et al. 1992; Levin et al. 1992), they were the focus of continued genetic and phenotypic characterization.

For flies to learn the association of chamber position and high-temperature reinforcement, they must be able to sense and avoid elevated temperatures. This can be tested with a thermosensitivity assay (Zars 2001), where flies are allowed to choose between chamber halves that have different temperatures. To do this, the same chambers are used, but the temperature inside is altered independent of the flies' behavior. The term “reference” temperature corresponds to 24°C, a temperature that wild-type flies prefer over lower and higher temperatures (Sayeed and Benzer 1996). The “probe” temperature is a temperature that is different from the 24°C reference temperature. Individual flies are presented with a chamber that initially is at 24°C on both sides, but then on one side increases to a probe temperature of 27°C and further to 30°, 33°, 37°, 41°, and 45°C, while the other chamber half is kept at 24°C. The chamber half with the lower temperature switches every minute. Flies are tested a total of 7 min. When flies show wild-type-like avoidance behavior, it implies their ability to sense a given temperature and, importantly, the locomotor ability to walk away from a high-temperature source. Interestingly, the white mutant flies are not “normal” at any temperature tested except 41° and 45°C (Fig. 4A). Although the avoidance behavior of wCS13 flies of 41°C was not significantly reduced, it is somewhat lower than wild-type CS flies' avoidance behavior. A second experiment was done to test the ability of these flies to sense and avoid a 41°C probe temperature (Fig. 4B). In this test, a probe temperature of 41°C was paired with 24°C in minutes 2 and 3. The chamber half with the high temperature switched between minutes 2 and 3. Consistently, white mutant flies were not significantly different from wild-type CS flies in avoiding the 41°C chamber half. The reduced avoidance of the second 41°C probe in both genotypes is probably a consequence of all flies being on the former 24°C side when the next test phase began, rather than being distributed randomly as in the first pretest minute. The rut mutant flies show high-temperature avoidance behavior similar to wild-type flies at all temperatures tested. A defect in temperature sensation may describe part of the wCS13 mutant flies' poor memory tested with screening conditions. However, it also suggests temperatures at which wCS13 flies could be tested to determine whether their conditioning deficit uniformly depends on lowered thermosensitivity (i.e., >41°C). Thermosensitivity defects cannot explain the rut²⁰⁸⁰ mutant conditioning phenotype. Additionally, as both mutant genotypes show normal avoidance of a chamber half at the highest temperatures, it indicates that the mutant flies have a sensory-motor system sufficient to avoid high temperatures.

To gain confidence that the white-ABC transporter is critical for place conditioning in the heat-box, flies mutant for a second, temperature-sensitive, white allele (w^blood [w^bl]) were raised at permissive and restrictive temperatures and tested. Using the mutant screening conditions of 4 min of training and 3 min of memory test with a 37°C reinforcer, w^bl mutant flies showed impaired performance during training and memory test when raised at the nonpermissive temperatures of 25° and 29°C compared with wild-type CS flies raised at the same temperatures (Fig. 5). These rearing temperatures led to a lightening of the eye color, from ruby to light yellow (Ephrussi and Herold 1945; Bingham and Chapman 1986). This performance deficit was not evident when w^bl mutant flies were raised at the permissive temperature of 18°C. As the primary change leading to reduced performance was the rearing temperature in w^bl mutant flies, it is highly probable that the decreased White-ABC transporter expression (caused by a “blood” transposable element located in the second intron of the white gene) (Bingham and Chapman 1986) is the reason for the poor training and memory-test phase performance.

Both w^bl mutant and wild-type CS flies were reared at 29°C and tested in the thermosensitivity assay. Comparing these two genotypes showed significantly reduced avoidance of 27° and 30°C in w^bl mutant flies compared with wild-type CS flies, similar to the cantonized white null allele (wCS13) (Figs. 4, 6). However, avoidance of all other temperatures were at levels not significantly different from wild-type levels, including avoidance of 37°C, the temperature used for reinforcement in the conditioning experiments. Thus, the poor performance in the conditioning paradigm in w^bl mutant flies raised at the restrictive temperatures is not a consequence of a poor sensorimotor process.

To further test the relationship between mutations at the white locus and performance in the heat-box, five additional mutant alleles were tested. These alleles were chosen as they span the range of eye-color changes caused by mutations in the white gene. These include eye colors of white (w¹ and w^ec3), yellow-orange to orange (w^a and w^a2), and brownish-orange (w^a3). In tests of thermosensitivity, flies of all but the w^ec3 allele had lower avoidance of the 37°C probe temperatures than wild-type CS flies (Table 1). All five white mutant alleles had wild-type-like avoidance behavior of 41°C. Similar to the white null mutant flies, four of the additional mutant alleles showed significantly lower conditioned memory performance than wild-type flies after training using both 37°C and 41°C reinforcement. Although the w^ec3 allele showed lower memory scores than wild-type flies after conditioning, it did not reach significance in these tests. Curiously, there does not appear to be a correlation between the severity of eye-color change in the white mutant alleles and performance in the heat-box. This likely indicates the systems requiring white-ABC transporter function for conditioning are not similarly sensitive to the mutations as eye coloration. A detailed understanding of the relationship between gene structure and heat-box conditioning must await a high-resolution analysis of the molecular changes in these mutant lines. (The w¹ and w^a mutations have transposable elements in regulatory regions of the gene [Zachar and Bingham 1982; O'Hare et al. 1991]).

Finally, transgenic copies of the white gene were tested for rescue of the conditioning deficit of null white mutant flies. In two cases, wCS13 mutant flies were completely restored to normal performance levels (Fig. 7; data not shown). However, in several other cases, performance was as poor as the null mutant levels, even though eye color was largely restored (data not shown). As the associated transgenes and insertion sites for the ectopic mini-white gene are different in these transgenic lines, it is presumed that the expression levels or spatial domains are not sufficient for rescue of the white mutant phenotype in those latter cases. Alternatively, insertion effects altering local gene expression or unregulated expression of effector genes in those transgenes could give rise to abnormal performance. Nevertheless, as all white mutations tested have similar performance deficits, and those deficits can be rescued by a chromosomal duplication and a mini-white transgene, it is concluded that the altered White-ABC transporter function is the cause of the abnormal performance of mutant flies in the heat-box conditioning paradigm.

Dwelling times indicate wild-type and wCS13 mutant flies typically receive full reinforcement

Even though wCS13 mutant flies can sense and avoid 41°C at wild-type levels, there remains a possibility that the poor performance of wCS13 mutant flies with the 41°C reinforcer is a consequence of poor avoidance of lower temperatures. This could be the case if wild-type flies use lower temperatures to determine which half of the chamber to avoid, but wCS13 mutant flies require the full measure of high-temperature reinforcement. The increased amount of time required for the wCS13 flies to receive the full reinforcement could lead to a decrease in the performance index during conditioning. To gain insights into the typical temperatures the flies are exposed to during training in the heat-box, the “dwelling times” during the 20-min training experiment with 41°C reinforcement were examined. A dwelling is delimited by the entry and exit from one chamber half. This is essentially the same measure used in determining the time in reinforced and nonreinforced quadrants of the visual surround in visual pattern conditioning (Dill et al. 1995).

Both wild-type CS and wCS13 flies are typically exposed to the predetermined maximum chamber temperature after they enter the chamber half associated with reinforcement. In both wild-type CS and wCS13 flies, there is a decrease in the number of dwellings per fly on the reinforcement-associated “hot” side of the chamber (Table 2). This reduction is not as pronounced in wCS13 as in wild-type CS flies and reflects the lower conditioning performance in these flies. The mean duration of a dwelling also tends to decrease in both wild-type CS and wCS13 mutant flies during the conditioning period, with wCS13 reducing their dwelling duration to ∼4.8 sec. Calculation of the average temperature of the chamber after either wild-type or mutant flies enter the reinforcement-associated side of the chamber indicates that both genotypes typically receive the full reinforcement temperature of 41°C. Based on this measure and the normal avoidance of 41°C in the thermosensitivity assay, the cause of the poor conditioning of wCS13 flies can be separated from a defective sensorimotor system.

scarlet, but not brown, alters spatial conditioning

Based on the pigmentation of the retina, the White-ABC transporter forms heterodimers with both the Brown and the Scarlet proteins. To test whether the known White-ABC transporter binding partners have a role in heat-box conditioning, genetically defined null alleles at the brown and scarlet loci (Dreesen et al. 1988; Tearle et al. 1989) were tested. Flies mutant for the brown gene did not significantly alter 3-min place memory performance compared with wild-type CS flies (Table 1). The scarlet mutant flies' performance, however, was significantly lower than wild-type levels. Indeed, memory tests after training for 20 min still found the scarlet mutant flies with significantly lower-place memory performance (CS = 0.59 ± 0.08, n = 67; st¹ = 0.25 ± 0.07, n = 68; Mann-Whitney U-test, Z = 3.48, P < 0.001). Examination of brown and scarlet mutant flies shows that they can sense and avoid high temperatures in the thermosensitivity assay similar to wild-type flies (Table 1). Thus, poor performance in the conditioning paradigm by scarlet flies appears to be independent of a deficit in their ability to sense high temperatures or a failure to walk away from a high-temperature source.

Activity/spatial conditioning relationship

To explore the potential influence of general activity on the pre-disposition of flies of a genotype to perform well in heat-box conditioning, pre-test-phase walking speed (taken as a measure of activity uninfluenced by conditioning) was compared with performance levels in the memory-test phase. Walking speed can be used as a first measure for locomotor ability, as both step length and frequency are incorporated in this compound measure (Strauss and Heisenberg 1993). This analysis might reveal that low activity is associated with poor memory-phase performance, as flies are incapable of walking away from the side of the chamber associated with high temperature or that flies of some genotypes might be walking so quickly that they cannot draw the conclusion that one-half of the chamber is associated with high and the other half with low temperatures. This could be the case as the whole chamber heats when a fly walks to the high-temperature associated half and the whole chamber cools as it walks back. Table 3 lists the average walking speed in the pre-test phase in the genotypes of all tested flies. It is obvious from this table that there are significant differences between wild-type CS flies and nearly every mutant genotype, some with higher and some with lower walking speeds. Indeed, in different experiments, wild-type CS flies can have different activity levels. These are probably dependent on environmental differences between experiments. However, comparison of pre-test walking activity with memory-test phase performance shows no significant relationship (Fig. 8). Thus, at least with the 18 genotypes and experiments tested here, there is not a relationship between pre-test walking activity and ability to form a place memory in the heat-box. It seems that one cannot predict memory-phase performance in the heat-box based on activity levels. However, mutations with more extreme walking speeds might eventually be found that challenge this conclusion.

Genetic dissociation of memory acquisition and strength

The poor performance of mutant flies could be a consequence of poor memory formation or a defect in processing of the high-temperature reinforcing signal. Examining memory performance in wild-type and mutant flies with increased training duration could lead to one of two results. Either there is a decreased acquisition rate, but eventually the same memory level is reached, reflecting a defect in memory formation, or maximal memory levels for a given reinforcement intensity are lower, indicating a defect in reinforcement signal processing. To discriminate between these possibilities, a series of experiments increasing the training length was used to probe memory formation in wild-type CS, rutabaga, and white mutant flies. Different groups of wild-type and mutant flies were continuously trained with a 37°C reinforcer for 1-20 min and then tested for their avoidance memory. As expected, and consistent with the previous results (Fig. 2), wild-type CS flies showed increased memory performance with increased training time (Fig. 9). Interestingly, increased training of rut²⁰⁸⁰ mutant flies lead to wild-type levels of memory, being indistinguishable with 10 and 20 min of training. In contrast, wCS13 mutant flies showed severely reduced asymptotic memory scores using both the 37° and 41°C reinforcement temperatures. Because the wCS13 results are a hallmark of lower intensity reinforcement, i.e., a lower asymptote in memory performance, it indicates that the white mutant flies have a defect in the processing of reinforcement information.

Olfactory conditioning

To determine whether the low memory levels found in white mutant flies was also evident in a second negatively reinforced conditioning paradigm, wild-type CS and white mutant flies were tested in the classical olfactory conditioning paradigm (Tully and Quinn 1985). Flies were trained under conditions previously shown to induce asymptotic memory (12 electric shocks evenly distributed within 1 min of odorant presentation with voltages ≥90 V). It was predicted that varying the intensity of the electric shock under these conditions might reveal differences between the genotypes. In contrast to the results in the heat-box, classical olfactory conditioning was not negatively affected by the wCS13 mutation (Fig. 10). Indeed, performance is significantly higher than wild-type CS flies at the lower voltages tested.

The increased memory performance of white mutant flies could be influenced by the perceived value of the electric-shock reinforcer or odorant intensity. To address this, both wild-type and white mutant flies were tested for their ability to sense and avoid different electric-shock intensities and odorants. At the lowest shock intensities tested, white mutant flies avoid the shocked tube at higher levels than wild-type flies (Fig. 11). This likely indicates that the higher performance in the learning experiment is influenced by this increased shock sensitivity. Olfactory acuity of white mutant flies was also tested. There were no significant differences between wild-type CS and wCS13 mutant flies in avoidance of the odorants used in conditioning (CS: MCH 50.3 ± 1.8, OCT 31.6 ± 1.6; wCS13: MCH 55.2 ± 3.5, OCT 31.8 ± 1.2; F = 1.54, P = N.S. for MCH, F = 0.006, P = N.S. for OCT; n = 6 for each genotype/experiment). Based on this test, there are no effects of changed olfactory acuity on olfactory conditioning in wCS13 mutant flies.

A modified model of operant conditioning

With the addition of a comparator in which the difference between the preferred and actual stimulus state is calculated, the Wolf-Heisenberg model of operant conditioning can be used to conceptualize the place-conditioning results. In the Wolf-Heisenberg model (Wolf and Heisenberg 1991; Fig. 12), flies (1) have a desired state (or goal). To achieve this goal, they (2) initiate different motor programs. A coincidence detector compares the efference copy of a motor program with the difference between the goal and the current sensory input (3). Flies (4) modify the probability of performing a given motor program with significant coincidence between a decreasing difference between desired and actual state and the motor program efference copy. Finally, flies (5) can persist with a motor program if there is consistent control of a sensory stimulus with a given behavior. It is the initiation of one of several possible behaviors to decrease the difference between the goal and the actual state that differentiates this model from other models that indicate the motor program is a response to a stimulus (e.g., Skinner 1950; Killeen 1994; Dragoi and Staddon 1999). This describes the ability of flies (and other animals) to use arbitrary behaviors to solve conditioning tasks. That flies try different motor programs to reduce the difference between the desired and the actual state was reached when flies were shown capable of using totally artificial behaviors such as inverse coupling of yaw torque and lateral force generation from the legs with the angular position of visual land-marks (Heisenberg and Wolf 1984; Wolf and Heisenberg 1986, 1991; Wolf et al. 1992). Addition of a comparator for the desired and the actual stimulus state provides the substrate for the effects of differing reinforcing temperatures and the white mutation.

In the model, the effect of changing reinforcing temperatures on performance level in the heat-box is done initially at the level of the desired/actual state comparator. At least in naive flies, the desired state is presumably a temperature range that is non-deleterious. This inference is supported by temperature preferences on a linear temperature gradient in which flies strongly prefer 24°C over both higher and lower temperatures (Sayeed and Benzer 1996). The deviation from that desired state provides a graded input to the coincidence detector. That is, small differences from a preferred temperature centered at 24°C should lead to a small input into the coincidence detector, larger differences to a larger input. The positive feedback loop between the behavior initiator and the coincidence detector via the efferent and after-effect pathways leads to a graded change in the probability of initiating a given behavior.

Mutations in the white-ABC transporter affect the comparator. As with a change in the reinforcing temperature, the change evident in the wCS13 mutant flies indicates an effect on the comparator. The mutant flies could be altering the levels of the desired or actual state. The poor avoidance behavior of wCS13 flies of some elevated temperatures indicates a decrement in this avoidance behavior. However, as they still prefer the 24°C temperature over all other temperatures tested, their preferred temperature is still 24°C. The defect found in the wCS13 flies is, therefore, in the value of the term subtracted from the desired state.

Performance in the thermosensitivity assay can be dissociated from performance in place conditioning. The wCS13 mutant flies avoid 41°C at levels not significantly different from wild-type flies yet do not avoid a chamber half associated with that temperature to the same extent as wild-type flies. This dissociation was also seen in w^bl flies at the restrictive temperature and four additional white alleles. That avoidance behavior is dissociated from memory forming behavior indicates that the mechanisms underlying these two behaviors are different, and that one function of the White-ABC transporter is to supply temperature value into the reinforcing circuit. Two other examples of this dissociation include the role of the VUMmx1 neuron mediating the reinforcing property of sugar reward but not eliciting the proboscis extension in the honeybee (Hammer 1993), and the conditioned avoidance of odors depending on the Kenyon cells of the mushroom bodies but not the spontaneous avoidance of noxious odorants (de Belle and Heisenberg 1994; Heimbeck et al. 2001). The different effects of the white mutation on avoidance and conditioned behavior might reflect differences in the neural structures mediating the two behaviors (as in the two other examples listed) or different sensitivities of the behaviors within the same neural circuit to mutations in the white gene, or both. Determining the neural structures that require White-ABC transporter function in the two behaviors will differentiate between these possibilities.

The rut²⁰⁸⁰ mutation, in contrast, affects the feedback loop of behavior initiation and coincidence detection. It is within this feedback loop that the effects on sensory input of an initiated behavior are recognized and maintained. The proposed role of the rut-AC in second messenger (i.e., Ca²⁺/calmodulin and G-protein) integration and synaptic plasticity are consistent with the feedback loop being the site of operant learning in this model. Whether the rut-AC function could be limited to a subset of the feedback loop is not clear.

Further implications of behavioral deficits in white mutant flies

The null white allele restricts maximal memory level, indicating a defect in reinforcement signal processing. Interestingly, the difference between wCS13 and wild-type maximal memory performance decreased as reinforcing temperatures increased. This suggests that the wCS13 mutant flies can learn, but that their subjective interpretation of the significance of the reinforcing temperature is altered. As the white allele used in these experiments is a molecular null mutation, the only interpretation for residual reinforcement evaluation is the presence of at least one additional processing mechanism.

The white gene encodes a member of the so-called half-size ABC transporter family. These proteins are critical for the transport of different substrates into cells. Consistent with the view that these proteins transport many different important substrates, mutations in this class of proteins leads to a large number of human diseases, including cystic fibrosis and Tangiers disease (Dean et al. 2001). The white-ABC transporter, together with the dimer binding partners Brown and Scarlet, is important in the translocation of guanine and tryptophan into cells of the retina (Dreesen et al. 1988; Tearle et al. 1989). Assuming the White-ABC transporter is interacting with Brown or Scarlet in the brain, where white transcripts have been detected (Campbell and Nash 2001), the tryptophan and guanine precursors could be used in the synthesis of many intra- and intercellular signaling molecules. One or some of these are presumably critical for normal heat-box learning. Finding that the scarlet mutation limits memory levels with 4 and 20 min of training restricts the candidates for this signal and centers on serotonin. A similar function for white has been suggested in overexpression and anesthesia studies (Zhang and Odenwald 1995; Hing and Carlson 1996; Nilsson et al. 2000; Campbell and Nash 2001). A better understanding of the neural structures in which these genes function will prove useful in determining the processes in which white and scarlet function in memory formation.

Beyond the specific study of memory performance, the white and scarlet mutations could have applications to human mental health. The human genes homologous to the white- and scarlet-ABC transporters (ABCG1 and ABCG4) have been detected in many tissues, including the brain (Chen et al. 1996; Savary et al. 1996; Croop et al. 1997; Annilo et al. 2001; Oldfield et al. 2002). Additionally, the region 21q22.3, the chromosomal location of ABCG1, has been linked to major depression in humans (Straub et al. 1994; Detera-Wadleigh et al. 1996, 1997; Smyth et al. 1997). Indeed, a mutation in the ABCG1 gene has been identified in male patients with depression (Nakamura et al. 1999). Finally, the commonality of function in tryptophan metabolism (e.g., Zhang et al. 2005) and a common low-asymptotic memory phenotype in human depression (Nestler et al. 2002) may indicate that these gene products have application in understanding at least some components of human depression.

Conclusion

The heat-box operant place-learning paradigm provides for the rapid determination of a mutant effect on reinforcement processing or memory acquisition. Mutations in the rut-AC and white-ABC transporter provide principle examples of altering acquisition rate and asymptotic performance in Drosophila.

Genetic manipulations and culture conditions

The wCS10 stock is a “cantonized” white null allele (w¹¹¹⁸) (Hazelrigg et al. 1984; Dura et al. 1993). The wild-type Canton S (CS) stock was used in behavioral tests and for outcrossing. Flies were raised at 25° and 60% relative humidity in a 12:12 h light:dark cycle on our standard food medium (Guo et al. 1996). The additional out-crossing of the x-linked wCS10 mutation to the wild-type CS line involved three outcrossings of two generations each and was renamed wCS13. The Dp(1:Y)w+ chromosome was chosen for cantonization to attempt a rescue of the white mutant memory scores. It contains a duplication of the X chromosome from 2D1 to 3D4 attached to the male Y chromosome (Smith and Konopka 1981). This chromosome was outcrossed to the wCS13 stock for six generations to have ∼3% of the original genome remaining. The white gene is at 3B2-3C2 (Pirotta et al. 1983) and is complemented by the Dp(1:Y)w+ chromosome based on eye color. The mini-white gene in the cantonized tetanus toxin light chain GAL4-UAS effector line CYO34-1 (Scholz et al. 2000) was also tested for behavioral rescue of wCS13 mutant flies. Additionally, the bw¹ and st¹ mutations were outcrossed to wild-type flies for greater than six generations (Dreesen et al. 1988; Tearle et al. 1989). The w¹, w^a, w^a2, w^a3, w^ec3, and temperature-sensitive w^bl flies were not cantonized (Bingham and Judd 1981; Zachar and Bingham 1982; Bingham and Chapman 1986; Birchler and Hiebert 1989). The outcrossing schemes used at least 20 males and females in each generation. Since the Dp(1:Y)w+ duplicates the white gene on the Y-chromosome, only males of this genotype were tested. Males and females of CS and wCS13 were tested separately, but no significant differences were seen, and their scores were pooled, similar to the findings of Putz (2002) on a large sample size of male and female wild-type CS flies.

Behavioral tests

Flies were tested in the heat-box spatial learning paradigm and in the thermosensitivity assays described in Zars et al. (2000b) and Zars (2001). In the learning paradigm, a fly is allowed to run in a small chamber that is heated to a defined temperature within seconds when it crosses an invisible midline; the chamber quickly cools to baseline (24°C) when it returns to the original side. The baseline temperature of 24°C was used, as flies have a strong preference for this temperature over both higher and lower temperatures when given a prolonged choice (Sayeed and Benzer 1996). A performance index is calculated by subtracting the time spent on the side associated with reinforcement from the time spent on the nonreinforced side and dividing this by the total time. Thus, a scale of -1 to 1 is generated with a total preference for the punished side giving a -1, and for the non-punished side, a 1. In all learning experiments, a 30-sec pre-test was followed by training of different lengths (0-20 min) and a 3-min post-test. Reinforcement temperatures were changed as indicated in the figure legends, including 33°, 37°, 41°, and 45°C as measured at the surface of the Peltier elements with a thermocouple. Flies that were exceedingly inactive in the pre-test (<0.15 mm/sec walking speed) or that failed to have at least one experience with the punished side of the chamber during training were discarded from the analysis. Determining the number of dwellings on the reinforcement-associated chamber half and the average dwelling times were calculated using custom software similar to that used in determining dwellings in the flight simulator (Dill et al. 1995; R. Wolf, pers. comm.). A dwelling was defined as the period between entry and exit from one side of the chamber. A dwelling period (in seconds) was assigned to the training session in which a fly exited the reinforcement-associated chamber half.

The thermosensitivity assay used the same chambers, but the temperature inside was altered independently of the flies' behavior. Individual flies were presented with a chamber that initially was at 24°C on both sides but then on one side increased to a probe temperature of 27°C and further to 30°, 33°, 37°, 41°, and 45°C, while the other chamber half was kept at 24°C. The chamber half with the lower temperature switched every minute. Flies were tested a total of 7 min. A performance index was calculated for this assay as in the learning assay. As tests for normal distribution of performance indices gave varying results, more conservative nonparametric tests for significance were used (Putz 2002). Genotypes were compared using a Kruskal-Wallis ANOVA or Mann-Whitney U-test, P values <0.05 were considered significant.

The olfactory conditioning paradigm followed previous protocols (Tully and Quinn 1985; Zars et al. 2000a). Flies were trained to avoid two odorants (4-methylcylohexanol and 3-octanol) by associating one with an electric shock of different voltage. A preference test for one of the two balanced odors followed in a T-maze choice point. A performance index was calculated by subtracting the number of flies preferring the formerly punished odorant from those preferring the formerly unpunished odorant divided by the total number of flies. This index was multiplied by 100 to generate a half-PI. An average of two half-PIs, measuring avoidance of each shock-associated-odorant, was calculated in generating the final PI. The number of experiments with each condition was six. Tests for shock avoidance similarly followed typical protocols. Shock avoidance used two shock tubes at the T-maze choice point, one of which was used to shock the flies with the same protocol as in the learning paradigm. The number of flies avoiding the shocked tube was determined in a half-experiment. A final PI was calculated pairing avoidance of either one of both shock tubes. The number of experiments was six for each voltage. Tests for olfactory acuity were carried out using the odorants and conditions used for conditioning. Statistics used in this assay were parametric ANOVAs as the data are normally distributed, as is typical of this paradigm.