The w¹¹¹⁸ (used here as a “wild type”) and the HS-vri stocks have been described and validated previously (Glossop et al., 2003; Gunawardhana and Hardin, 2017; Gunawardhana et al., 2020). The HS-vri transgene was backcrossed multiple times into the w¹¹¹⁸ background so that they are isogenic. The HS-Clk^jrk transgenic fly was made from the previously described CLK^jrk mutant fly (Allada et al., 1998; Darlington et al., 2000). A plasmid containing the fly CLK^jrk mutant open reading frame was obtained and an EcoRI fragment was subcloned into the pCaSpeR-HS plasmid, and the resulting plasmid was injected into a w¹¹¹⁸ background stock to make transgenic flies (Thummel, 1996). That w¹¹¹⁸ background was used as the control in Figure 2, Expt #7. The iso³¹ (used here as “wild type”) and Pdp1³¹³⁵ mutant stocks have been described and validated previously (Zheng et al., 2009). The Pdp1³¹³⁵ mutation resides in the iso³¹ background.

Young flies (less than 1 week of age) were used for behavior. Flies were collected and kept in groups of 100–150 individuals/vial and entrained to a 12-h lights on:12-h lights off schedule at 20°C for 3d prior to training. Training started between ZT = 14 and ZT = 16, and flies were returned to their dark period and kept at 20°C until testing. Testing always occurred around ZT = 16 to ZT = 18. For experiments involving changes in post-training lighting conditions, flies were incubated after training under standard LD conditions until constant conditions started, at which time the appropriate group was shifted to darkness or light (at 20°C). For the HS-vri flies that underwent post-training heat-shock, flies were placed in food-containing vials and incubated under light:dark control at 28°C for almost two full-days. For pre-training heat-shock, HS-vri flies were induced at 28°C for 3 h of time. For the HS-Clk^jrk flies that underwent post-training induction, flies were transferred to empty food vials, and incubated at 36°C in a water bath for 30’ duration at 32 h and 56 h after the end of training, then returned to regular vials and incubated at 20°C under light:dark control until testing. For HS-Clk^jrk flies that underwent pre-training induction, heat-shock (a 20°C to 36°C temperature shift for 30’) was done in a water bath, and flies were allowed to recover for 90’ prior to training.

Flies were trained in the olfactory avoidance-training paradigm developed by Tully and Quinn and modified to allow for automated training sessions (Tully and Quinn, 1985; Tully et al., 1994). A single-cycle of training consists of 90 s exposure to ambient air; 60 s of electric shock [the unconditioned stimulus which consists of 70 V pulses lasting 1.5 s and administered every 5 s (12 total)] accompanied by simultaneous exposure to 1 odor (the conditioned stimulus, CS+); 45 s of ambient air exposure to clear the first odor; 60 s of exposure to the second odor with no shock (the CS- condition), 45 s of ambient air to clear the second odor. This single training trial takes about 2.6 min. Spaced training consists of 10 single cycles separated by 15-min rest intervals. This training requires about 2 h 40 min of time. Massed training consists of 10 consecutive single cycles of training, and takes 37 min. Testing was done by placing flies in a choice point and giving them 2 min to decide between the CS+ and CS- stimuli. We used 3-octanol and 4-methylcyclohexanol as the odors. The performance index = [the number of flies making the correct choice] − [the number of flies making the incorrect choice]/total number of flies, multiplied by 100. To avoid odor-avoidance biases, we calculate the performance index of every single N by taking an average performance of two groups of flies, one trained with 3-octanol as CS+, and the other with 4-methylcyclohexanol. Flies were trained in a balanced manner, such that the sequence of shock-paired odors alternates, as well as the assignment of left vs. right arm at the choice point during testing. Data is presented as the standard error of the mean, and the Student’s T-test was used to evaluate statistical significance in pairwise comparisons.

Male flies within 3 days of eclosion were collected and entrained to a 12:12 light/dark cycle for 2 days before assaying for sleep. Individual flies were loaded into tubes under CO₂ anesthesia and placed in the Drosophila Activity Monitor (DAM) System and put into an incubator maintained at 20°C under 12 h light:12 h dark cycles (Pfeiffenberger et al., 2010a). Raw activity counts were collected over a 24-h period. An Excel program was developed and used to calculate the various sleep measures and to plot the sleep profiles (Pfeiffenberger et al., 2010b; Cui personal communication). The sleep curves represent the average amount of sleep in each hour and the error bars represent the standard error of the mean. Dead flies (those with two or more consecutive days without locomotor activity) were manually eliminated. Sleep was measured from two groups of flies, one group was maintained in a 20°C incubator for the entire duration of the experiment. The second group was placed in a different incubator kept at 20°C until the time of heat-shock, when the temperature of the incubator was shifted to 28°C.

Male flies within 3 days of eclosion were collected and entrained to a 12:12 light/dark cycle at 20°C for 2 days prior to assaying for circadian activity. Individual flies were anesthetized using CO₂, loaded into tubes, and placed in the DAM system to measure locomotor activity under 12:12 light/dark cycles at 20°C for 3 days (Pfeiffenberger et al., 2010a). The incubator containing the DAM monitors was shifted to constant darkness for 2 days, then shifted to 28°C for the remaining 4 days while maintained under constant darkness. Locomotor activity was monitored for the entire duration of the experiment, and data were analyzed using an Excel program we developed (Pfeiffenberger et al., 2010a; Cui personal communication). Table 2 compiles the average total number of beam crossings per minute (binned across hourly units) for the different flies and treatments during the entire duration of the recording period.

All flies were initially entrained to 12-h light:12-h dark cycles for 3 days prior to behavioral training (10 cycles of spaced training, indicated with the green hatched box below the histograms in Figure 1). We tested the effects of post-training changes in light:dark incubation by comparing the performance of the experimental group relative to control flies that were incubated in standard light:dark conditions after training (shown as the first cartoon at the bottom of Figure 1, and labeled LD). Each different post-training lighting condition was tested individually against the control LD group, and the behavioral data for each experiment (numbered #1-#5) is shown as a separate pair of histograms. The single hatched histogram is the control condition (post-training LD), while the white or black histograms indicate the different incubation conditions being tested. In the cartoon below the Figure, the double white bars indicate constant light, while the double-hatched bars show when D is imposed. All flies were tested for 3-day (3d) memory (indicated with black arrows on the timelines). We initially chose post-training constant light (LL; Expt #1) because it was known that this treatment disrupted the oscillations of the molecular clock (Hardin et al., 1990). However, post-training LL treatment did not significantly affect 3d memory. We then tested the effect of post-training constant darkness (DD). Imposing DD during the first light period after training (D1; Expt #2) did not affect 3d memory. However, imposing constant darkness during D2 (Expt #3), D3 (Expt #4), or both periods (D2+D3; Expt #5) disrupted 3d memory. These results almost exactly parallel those reported by Inami et al. (2020) who used the courtship suppression behavioral assay. We also tested red-eyed flies and found similar effects (see Supplementary Figure 1), suggesting that these effects are not a function of eye color or w¹¹¹⁸-mediated retinal degeneration (Ferreiro et al., 2018).

To determine if specific circadian molecules are involved, and to minimize pleiotropic effects that circadian disruption might have on neuronal development, we use inducible transgenes to limit genetic interventions to post-development periods. The ubiquitously expressed, heat-shock-driven vrille transgene (HS-vri) has been used previously to characterize the involvement of this core circadian transcription factor in the central clock (Glossop et al., 2003). Chronic, low-level induction of the transgene beginning 1-day (1d) after the end of 10 cycles of spaced training (10 × S) affects 3d memory (Figure 2, Expt #6), while heat-shock itself does not have an effect in the isogenic wild-type stock that contains the transgene (Figure 2, Expt #7). This effect is specific for certain “phases” of memory since induction prior to 1 cycle of training does not affect immediate performance (conventionally known as learning; Figure 2, Expt #8), nor does pre-training induction affect 1d memory after 10 × S trials (Figure 2, Expt #9) or 10 cycles of massed training [Figure 2, Expt #10 conventionally known as anesthesia-resistant memory (ARM); Tully et al., 1994]. Induction of the transgene prior to behavioral testing (3d after training) likewise does not affect performance, suggesting no effects of transgene induction on retrieval (Figure 2, Expt #11).

To test the involvement of other circadian molecules (and the oscillatory circadian molecular network in general) we tested two other core circadian molecules for an effect on 3d memory. The Clk gene codes for the core transcriptional activator of E-box mediated transcription, dCLK. We made transgenic flies that express the Clk^jrk ORF under the control of the heat-shock promoter. The Clk^jrk mutant fly contains a truncation in the Clk ORF, resulting in a dominant negative protein, and heat-shock induction of this CLK^jrk-encoding transgenic fly should disrupt the clock (Allada et al., 1998; Darlington et al., 2000). Post-training induction of two independent insertions of the Clk^jrk transgene both disrupt 3d memory (Figure 3, Expt #12 and #13), but induction of one of the transgenes prior to training has no effect on 1d memory (Figure 3, Expt #14). In the wild-type fly in which the transgenes reside, post-training induction does not affect 3d memory (Figure 3, Expt #15). The induction regime that is used for these experiments involves two 30’ shifts in temperature (from 20°C to 36°C) approximately 30 and 54 h after the end of the training, with the flies returning to 20°C after the end of each heat-shock. This regimen differs from the one used for the transgenic vri flies (see Figure 2) and demonstrates that brief “pulses” of Clk^jrk expression are sufficient to inhibit memory consolidation. We used this acute, shorter-lived induction paradigm to insure that our behavioral effects are not due to flies being exposed to a chronic heat-shock (stress) situation. Western analysis of head extracts from uninduced and induced HS-Clk^jrk flies probed with Clk-specific antibody shows induction of a band of the predicted size (see Supplementary Figure 2). Three hours after induction, the transgenic CLK^jrk protein is in excess over the endogenous CLK protein (see Supplementary Figure 2), consistent with its seeming ability to inhibit endogenous CLK activity.

The vri gene encodes a D-box binding repressor, while the Pdp1 gene makes a corresponding activator of transcription (Cyran et al., 2003; Zheng et al., 2009; Gunawardhana et al., 2020). Within the molecular circadian network, VRI and PDP1 are thought to compete for binding to D-box-containing promoters although it is not clear if they necessarily regulate the same ones. In addition to having opposite molecular functions, we chose to test PDP1 because of the existence of the Pdp1ε mutation, Pdp1³¹³⁵. Testing a loss-of-function mutation complements our experiments using transgenic overexpression of proteins (VRI and CLK^jrk). The Pdp1 gene encodes multiple protein isoforms, and Zheng et al. (2009) previously showed that Pdp1ε is the circadian isoform specifically disrupted in the Pdp1³¹³⁵ mutant. Since overexpression of VRI disrupts 3d memory, we expect that decreased PDP1 protein should do likewise. When compared to its isogenic wild-type strain, the Pdp1³¹³⁵ mutant shows aberrant 3d memory (Figure 3, Expt #16). To rule out a mutant effect on the relevant sensory systems involved in odor avoidance behavior, we compared the wild type and mutant stocks for immediate performance after a single training trial. Although their scores are low, there is no difference between the two stocks, suggesting that the mutant flies have comparable sensory capabilities when compared to their isogenic wild-type stock (Figure 3, Expt #17). Taken together with the vri data, these results suggest the involvement of vri/Pdp1ε in 3d memory. Four different disruptions (DD, induced overexpression of VRI, induced overexpression of CLK^jrk, and removal of PDP1ε) all affect the same temporal window (1d-3d memory). None of the three interventions that were tested affected 1d memory. Based on this data, we believe that disruption of the circadian system very likely affects 1d-3d memory.

It is generally accepted that sleep is essential for memory formation in both vertebrates and flies (Walker and Stickgold, 2004, 2006; Ganguly-Fitzgerald et al., 2006; Bushey et al., 2007; Seugnet et al., 2008; Donlea et al., 2011; Gerstner et al., 2011; Berry et al., 2015; Dissel et al., 2015; Dudai et al., 2015; Haynes et al., 2015; Liu et al., 2019; Chouhan et al., 2021). Decreases in the total amount of sleep, and/or deterioration in the quality of sleep are hypothesized to interfere with consolidation. Sleep is believed to be under homeostatic and circadian control (Borbely, 1982). Therefore we examined whether our behavioral deficits might result from disruptions to sleep. Since mild induction of the HS-vri transgene specifically affects 3d memory, we used the same induction protocol (20°C to 28°C chronic temperature shift) as was used for behavior to test for effects on sleep. We measured the effect of induction (Figure 4, indicated with the vertical dotted line) on sleep in the transgenic HS-vri male flies, as well as in isogenic control flies devoid of the transgene. Their hourly average sleep (+/–SEM) was plotted as a function of time, and Figure 4 shows that transgene induction (red trace) has small effects on the “architecture” of sleep. The transitions between active and inactive periods are identical, as are the kinetics (slopes) of “falling asleep or waking up.” Transgenic flies exposed to heat-shock sleep more during the daytime periods. As a comparison, the isogenic “wild type” stock (w¹¹¹⁸) shows very similar effects, although these flies have less saturated sleep during both the pre- and post-induction nighttime.

The quantitative analysis of sleep is summarized in Table 1. For the HS-vri transgenic fly, induction increased the total amount of sleep, and this was attributable to an increase in daytime sleep. There was little effect on the total number of sleep bouts, with a small decrease in the average number of nighttime bouts that was statistically significant. In contrast, the average total length of bouts, especially those during the daytime sleep period, increased. Thus the large effects of induction were to increase the amount of sleep, and most of this occurred through increasing the average daytime length of bouts. Since transgene induction disrupted memory formation, and increasing sleep is usually thought to “enhance” memory formation, it is very unlikely that our behavioral disruption results from effects on sleep.

For comparison, we have also measured sleep changes in the isogenic w¹¹¹⁸ strain that is the background in which the HS-vri transgene resides. In general, the effects of heat-shock trend in the same direction, with increases in the total amount of sleep during both daytime and nighttime periods. The average number of bouts increase (total, daytime and nighttime), while the average bout lengths (total, daytime, and nighttime) show barely significant changes. Regardless of these subtle changes, there is no measurable effect of heat-shock in w¹¹¹⁸ flies on memory formation.

To evaluate the effect of our transgenic manipulations on circadian behavior, we tested the effect of HS-vri transgene induction on locomotor activity using the DAM apparatus across a 9-day period. Figure 5 plots the average number of beam crossings per minute binned hourly across the entire 216 h of elapsed time. This experiment involves two transitions, one from LD to DD (beginning at 60 h of elapsed time), and the other when heat-shock was begun (at 120 h elapsed time). The black bars below the plot show the nighttime periods and the beginning of constant D, and the red horizontal line indicates when chronic heat-shock (a 20°C to 28°C shift) was begun (in flies under constant darkness).

The HS-vri transgenic flies maintained under LD conditions show the classic “U-shaped” activity profile, with peaks at dawn and dusk. Upon the shift to DD, they almost immediately transition to a single peak in activity spread out evenly during the subjective “daytime.” Heat-shock results in an immediate spike in activity, decreases subsequent total activity slightly, but does not change the architecture (shape) of the activity profile. The w¹¹¹⁸ flies have quantitative differences in their activity (when compared to HS-vri) but exhibit a similar sleep profile (shape of the curve). Under LD conditions, they show a dramatic U-shaped curve, with greater total and peak activities than the HS-vri flies. The shift to constant darkness does not change the architecture as quickly, but by the time heat-shock is delivered, they show the inverted U-shaped, single peak that is broadly distributed across the subjective daytime. The w¹¹¹⁸ flies have a noticeably higher activity level than that of the transgenic flies. In terms of our behavioral results, the only manipulation (+HS in the HS-vri flies) that affects memory formation modestly decreases locomotor activity. The effect of heat-shock on locomotor activity in the w¹¹¹⁸ flies is greater but does not affect memory formation. Table 2 presents a more quantitative analysis of the locomotor activity depicted in Figure 5. The white and gray shading represent light and dark periods, while the red box indicates flies that are maintained under DD conditions but at the inducing temperature of 28°C. The top number in each box indicates the total average number of beam crossings (per minute binned in hourly units) in a 24 h period, while the second line in each box breaks this into the daytime and nighttime (shown in gray) periods. Once DD is imposed, the activity is divided between the subjective day and night periods (all shown in gray). The induced HS-vri flies (Days 6–8) are less active than when they were maintained under constant dark only (Days 4–5), and these flies are less active than their w¹¹¹⁸ counterparts under all conditions. Although affected, the decrease in the activity of the induced HS-vri flies (vs. the uninduced flies) is not a large effect, and is unlikely to contribute to their behavioral differences. We do not think the induced flies are just more “lethargic,” and less likely to move correctly at the behavioral choice point since they perform similarly to their uninduced controls after a single cycle of training (Figure 2 Expt #8), or after 10 cycles of spaced (Figure 2, Expt #9) or massed training (Figure 2, Expt #10) tested 1d after training. Their only deficit in performance is when it is measured at 3d after 10 cycles of spaced training (Figure 2, Expt #6).