Antidepressant suppression of REM and spindle sleep impairs hippocampus-dependent learning and memory but fosters striatal-dependent strategies

REM sleep enhances hippocampus-dependent associative memory but has little impact on striatal-dependent procedural learning1-3. Antidepressant medications like desipramine (DMI) inhibit rapid-eye-movement (REM) sleep but it is little understood how antidepressant treatments affect learning. We found that DMI strongly suppressed REM sleep in rats for several hours and impaired reconsolidation of a familiar maze and consolidation of moved baited positions (reversal learning) in a sleep-dependent fashion. Unexpectedly, DMI also reduced the spindle-rich transitionto-REM sleep state (TR) and spatial memory changes were more related to TR than to REM sleep. Working memory was unaffected, but overnight reference memory was significantly impaired and subjects increased reliance on non-hippocampal strategies. Procedural memory performance was positively correlated with increases in nonREM sleep after DMI serving to offset memory declines, partially preserving performance. Our results suggest that familiar memories are reconsolidated during REM sleep, reversal memories consolidated during TR, and procedural memories consolidated during non-REM sleep. N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 3 We tested the involvement of REM sleep in the process by which memories are rendered stable (consolidation) by looking at REM sleep suppression using a commonly prescribed selective norepinephrine reuptake inhibitor and REM sleep suppressing antidepressant, desipramine. Studies have shown that short term REM sleep deprivation following new learning impairs hippocampus dependent spatial learning and memory, depending on the task and intensity of training1-3. Striatal dependent procedural learning is largely unaffected by REM sleep deprivation1-5. Arguments against the role of REM sleep in memory consolidation include the observation that instrumental REM sleep deprivation can be stressful6,7 thereby confounding the effects of sleep manipulation with the negative effects of stress on memory8,9. Additionally, individuals using REM sleep suppressing antidepressants do not report marked reductions in memory function beyond those already associated with depression, although some have been documented10. We chose to address both issues using DMI as a non-instrumental REM sleep deprivation technique that is commonly employed in humans patients to treat depressive symptoms with a relatively low side effect profile11. DMI inhibits of the reuptake of noradrenaline (NA) thus elevating NA levels at all targets of the locus ceoruleus. Elevated levels of NA actively suppress REM sleep by acting in the pons to inactivate REM-ON cholinergic neurons12. Here we report for the first time that REM suppression impairs reconsolidation of a familiar recalled memory, and provide evidence that the often overlooked TR state is involved in the process of reversal learning2 (learning a new response in a familiar context). Both tasks are dependent on normal hippocampal activity13,14. Finally, we contrasted hippocampal learning effects of DMI-induced sleep changes with striatal learning effects using a procedural fixed choice T-maze task. N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 4 RESULTS DMI inhibits REM and TR sleep DMI taken orally reduces REM sleep for 5-8 hours in patients being treated for depression15. A 10 mg/kg DMI dose p.o. ingested by our subjects suppressed REM sleep for 8 hours (Figure 1a, p = 0.0006 repeated measures MANOVA) with no discernable impacts on health or ambulatory activity. Continuous recordings in 8 subjects showed that time spent in REM sleep went from an average of 8.35 ± 1.24 % during the light phase (measured on the second day of baseline conditions) to 3.51 ± 1.22 % after administration of DMI (p = 0.00038, repeated measures ANOVA, baseline vs. DMI treatment day effect, Figure 1a). Within the first 6 hours post-training, i.e., the REM sleep critical window ascertained for this training regimen1, REM sleep was reduced 95.6 +/0.22% (p = 3.84x10-8, Figure 1e) with no diminution across testing days (Figure 1f, p = 0.95, paired t-test on 1st vs. last day and Figure 1d). DMI administration also produced a significant (p = 0.03, paired 2-tailed t-test) decrease in the number of REM episodes relative to the same circadian time under baseline sleep conditions. Insert Figure 1 here DMI treatment does not impair activity or motivation Twenty-four rats were tested under conditions of 10 mg/kg DMI or Control mash (Supplemental Figure 1) after performing a familiar and reversal (novel) 8-box mazes for 30 min each and 15 trials of a fixed choice T-maze task each day for 5 days. Both DMI and Control groups ran a similar number of laps on the 8-box mazes each test day and between days (Novel maze, Control 14.8 vs. DMI 14.2 laps, Wilcoxon SignRank test p = 0.9387, 1st day 14.8 vs. last day 14.3 laps, p = 0.365; Familiar maze, Control 18.8 vs. DMI 18 laps, p = 0.330, 1st day 18.3 vs. last day 18.5 laps, p = 0.461). Controls and DMI treated rats also did not significantly differ in weight (day 5: N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 5 Control = 336 +/14.8 mean grams +/SEM; DMI = 326.8 +/9.23; p = 0.389, t-test) indicating that maze sampling experience and hunger related learning motivation did not differ between groups. DMI impairs novel spatial learning and familiar reconsolidation Under DMI treatment, subjects showed significant performance deficits on novel reversal learning and familiar maze reconsolidation. A general linear model with repeated measures showed a performance effect of drug treatment: DMI performance was worse than controls (p = 0.001). Performance was better on the familiar vs. novel maze (p = 0.000), and DMI had the same negative effect on performance on both mazes as there was no drug by maze interaction (p = 0.759). Within subjects there was a powerful learning effect of training day (p = 0.000) that was dependent on whether the maze was novel (improved daily performance) or familiar (less marked daily improvement), as shown in a day by maze interaction (p = 0.000), and this day effect persisted independent of DMI administration (day by drug interaction p = 0.143) regardless of the maze run (day by drug by maze interaction p = 0.705). Insert Figure 2 here Reversal learning impaired by DMI treatment TR loss For novel spatial learning on the reversal task (Figure 2a), the DMI group never reached criterion, always performing >1 mean error/lap even after 5 days (Figure 2a and Figure 4c). Interestingly, performance deficits under DMI treatment were not correlated with the level of REM sleep suppression on this reversal learning task (Supplemental Figure 2c) but instead were highly correlated with drops in TR under DMI treatment (Figure 2c and Table 1). Consistent with DMI’s impact on REM sleep, DMI also attenuated the amount of TR sleep (Figure 2d) by a range of 13-75% in the first 12 h after treatment. TR and REM sleep can be independently regulated N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 6 (Supplemental Figure 5d,e); TR often occurs without a subsequent REM period and REM sleep can occur with an abbreviated TR state. Within a single running session, both Control and DMI treatment conditions showed significant improvement of, on average, ~1 error/lap (Control 0.92 +/0.36, DMI 0.71 +/0.30, MANOVA, no group effect p = 0.6423, no day effect p = 0.2031, and no day by group interaction p = 0.996; supplemental Figure 3c and d). Thus, working and short term memory systems do not appear to be influenced by the prior day’s DMI treatment. The source of the learning deficit for DMI treated animals on the novel maze was poor retention across the night. Subjects given DMI committed 0.5 +/0.20 more errors on the first 3 laps than on the last 3 laps of the previous day. Controls, however, made 0.32 +/0.20 fewer errors than the last laps of the day prior (MANOVA drug effect p = 0.0217). There was no significant day effect (p = 0.0825) or day by drug interaction (p = 0.1809) in this retention measure, meaning that DMI affected reference memory retention on all days. The DMI treatment condition also increased the utilization of nonhippocampal or procedural strategies to locate the positions of new food boxes, as revealed by a significant increase in errors after a 180 maze rotation between laps 10 and 11. For this rotation, baited boxes remained in the same allocentric positions but the boxes that held the food was changed, pitting non-hippocampus dependent local cue use against hippocampus-dependent allocentric spatial map strategies (Figure 2b; see methods for details). Reversal learning was thus attenuated by DMI treatment in a sleep-dependent fashion, with the most relevant sleep metric being the length of time spent in the transition to REM sleep state during DMI after training each day. Reconsolidation impaired by DMI treatment N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 7 Performance significantly worsened on the familiar maze when 10 mg/kg desipramine was administered after each training session (DMI negative slope curve fit 1st order polynomial R2 = 0.75). By contrast, control performance held steady (Figure 3a, repeated measures ANOVA DMI vs. Control, p = 0.005). There was no performance difference between groups on the first day of the familiar maze (p = 0.34, t-test), but by day 5 the performance difference between groups was a full error per lap (p = 0.001, t-test). The relative preservation of performance was greatest for those DMI treated subjects that showed the least attenuation of REM sleep (Figure 3c). However, the strongest correlation between sleep measures and familiar maze performance was the percent time spent in TR during testing. Unlike the novel maze performance, this correlation was negative. That is, those subjects with the best performance had the least % TR during DMI treatment and those whose performance worsened most over the 5 days were those with the most %TR during DMI (Table 1 and Figure 3d). Thus, the more TR was preserved during treatment the better the consolidation of novel learning (Figure 2c), and the worse reconsolidation of familiar memory. Insert Figure 3 here As on the novel maze task, DMI treated subjects performing on the familiar maze were unable to effectively utilize reference memory to maintain performance overnight as evident by an increase in the number of errors on the first few laps from day 1 to day 5 (Figure 4a vs. 4c and Supplemental Figure 3a vs. 3b). Animals treated with DMI had more trouble retrieving correct reward box locations between days on the familiar maze than when not being treated with DMI (ANOVA p = 0.0001). DMI performance worsened by 1.19 +/0.29 errors in the first lap compared with the last lap of the prior day. Control performance improved by 0.16 +/0.14 errors overnight. Thus reference memory errors contributed to the degraded reconsolidation during DMI treatment. N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 8 Insert Figure 4 here DMI treatment also resulted in more errors after the maze rotation in the familiar maze condition (p = 0.0059, Figure 3b) revealing increased use of nonhippocampal strategies. Animals also committed more errors on lap 11 than lap 10 under Control conditions (p = 0.042) but remained within criterion (<1 error/lap) performance throughout. Under both treatments, animals improved by an average of about an error per lap between the first and final lap of the training session, showing the ability to improve performance based on working memory (DMI improved 1.13 +/0.20 errors/lap, Controls 0.96 +/0.23, mean +/SEM) with no day effect (ANOVA p = 0.7132) and no group effect (ANOVA p = 0.5762). DMI spatial memory related to REM and TR suppression, not QS Together these results indicate that decreases in spatial maze performance under DMI treatment reflect memory consolidation or reconsolidation errors associated with REM and TR sleep changes. Neither familiar nor novel maze performance was related to non-REM sleep (here referred to as quiet sleep, QS). The Pearson correlations (R) of familiar and novel maze improvement vs. QS ranged from -0.46 (p = 0.24) to 0.01 (p = 0.98) for correlations with QS at baseline, during DMI and % change between conditions (data not shown). Consolidation of reversal learning was correlated with preservation of TR and reconsolidation of the familiar maze was associated with the preservation of REM sleep. On both familiar and novel mazes (Figure 4d), Control conditions showed significant improvement from day 1 (Figure 4b vs. 4d) and above criterion performance on all but the first two laps by day 5, whereas under DMI treatment, subjects were unable to achieve criterion performance (Figure 4c). Finally, on both familiar and novel mazes, alternative non-hippocampal performance strategies were utilized more heavily under DMI-induced REM sleep deprivation. N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 9 DMI procedural memory correlated to sleep parameters In sharp contrast with the effects of DMI on sleep and hippocampus dependent learning performance, DMI treatment performance on the procedural T-maze task was significantly correlated to the level of QS demonstrated by each subject during testing (Figure 5e) and independent of changes in REM sleep during the testing period (Figure 5 b,c). Treated subjects with the largest % increases in QS during DMI administration performed well on the T-maze whereas those with declines in QS made far fewer correct choices (Figure 5f). Although REM sleep during testing was not related to procedural task performance during DMI treatment, the percent REM sleep obtained during the baseline recordings was related to later T-maze performance (Figure 5a). More surprising was that the amount of REM sleep during baseline was inversely correlated with the % change in QS during DMI treatment (R = 0.8942, p = 0.0066, Supplemental Figure 5b). That is, the more REM sleep a subject showed during baseline, the higher the percent of QS during DMI and the better the procedural performance. Little relationship has been found between the amounts of REM sleep and non-REM sleep under normal conditions16,17, much less a relationship separated by days and experimental treatments. Insert Figure 5 here The number of correct choices and the latency to trial completion were highly correlated, as expected (Supplemental Figure 5a). Thus, those sleep parameters that were correlated with the number of correct T-maze choices were often also significantly correlated (inversely) with latency to trial completion (Table 1). T-maze performance was not correlated with the amount of waking or TR sleep obtained during baseline or testing periods (Supplemental Figure 4). DMI, then, increased procedural T-maze performance in a sleep dependent fashion according to the amount of REM sleep during baseline and QS during treatment. N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 10 DISCUSSION Overall, our main hypothesis was strongly supported: DMI suppressed REM sleep and inhibited hippocampus-dependent learning during antidepressant treatment. Both spatial reversal learning and reconsolidation of a familiar maze were significantly impaired, although compensatory non-hippocampal strategies partially rescued performance within a session. Surprisingly, we found that impairments in reversal task learning were best correlated with declines in the amount of the TR state rather than decreases in REM sleep. The subjects with the least %reduction in TR following DMI treatment show reversal learning performances most similar to the control condition: improving performance by an error or more per lap across the 5 day period on the reversal learning task (Figure 2c). Subjects with declines in TR under DMI treatment showed substantially slower learning rates on the novel maze (0.5 errors/lap or less improvement). Conversely, reconsolidation of a familiar memory was poorest in those with the highest retention of the TR state during DMI testing (Figure 3d), as though processes during TR which best predicted successful incorporation of novel information were also unsupportive of the maintenance of an older schema. Reconsolidation of the familiar maze was, as predicted, also associated with the degree of REM sleep disturbance during antidepressants treatment (Figure 3c). Procedural learning, however, was enhanced with augmented QS during testing (Figure 5e,f) and, surprisingly, by the amount of pre-test baseline REM sleep (Figure 5a). Together, our findings suggest that the consolidation of novel reversal learning, reconsolidation of familiar memories, and consolidation of procedural memories may involve routines differentially called during different sleep states, with reversal consolidation occurring during TR (spindles), reconsolidation during REM sleep, and striatum-dependent procedural consolidation during QS. N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 11 REM sleep supports reconsolidation In a human fMRI study, REM sleep deprivation is coincident with a decrease in hippocampal function18. Indeed, induced neural activity during REM sleep has been shown to drive changes in synaptic weights19. Reactivation of neurons encoding a memory network occurs during REM sleep20 in a fashion supporting both LTP of new learning and the reversal of LTP for consolidated memories within the novelty encoding pathway21. Thus memories are likely revisited and reshaped during REM sleep outside of the original learning experience22. NA and DMI would impair the production of REM sleep through inhibition of REM-on cells in the pons. The prevention of REM sleep also prevents all REM sleep related processes including hippocampal reactivation, REM theta, and theta phase activity that support depotentiation (DP). These results for learning agree with those seen with REM sleep deprivation effects on hippocampal function in vitro through loss of induction or maintenance of long term potentiation (LTP)23,24 and reductions in excitability24. Without timely opportunity to enter REM sleep and accomplish synapse-specific strengthening or weakening, consolidation and reconsolidation could be impaired. In the closest study to ours with a once daily 10 mg/kg dose testing rats on an 8-arm radial maze similar impairments in learning were found25; presumably REM sleep was also suppressed in their study. TR sleep supports reversal learning but impairs reconsolidation Our data also suggest that consolidation of novel reversal learning and reconsolidation of familiar places are both proportional to the amount of TR sleep, though in opposite directions. Reactivation of hippocampal neurons and interaction with prefrontal cortex is uniquely intense during sleep spindles, possibly supporting memory consolidation26,27. Sleep spindles, which characterize the TR state, do not occur when NA, acetylcholine (ACh) or serotonin is present in the thalamus28. The DMI block of N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 12 NA reuptake during TR could decimate spindles and the hippocampal-forebrain reverberation that accompany them to incorporate novel information into established consolidated memory networks. Increases or a relative preservation of TR may maintain sufficient spindle-related reverberation to support remodeling of the synaptic memory network. Maintenance of familiar memories in the face of daily novel learning suffered in subjects with higher % TR during DMI. Finally, like REM sleep, TR is a state where P-waves are present, and P-wave density during TR and REM sleep have been strongly correlated with learning novel relationships29,30. Taken together, perhaps the strong correlation between novel learning and TR and the negative association with familiar memories through increased novel interference with familiar map memories, is not so surprising. These results also agree with recent findings on the function of sleep spindles31 in humans. However, little is known about this transient sleep state and its interaction with mechanisms of learning22. QS during consolidation supports dorsal striatal learning Procedural learning, was, in our hands, enhanced proportionately with the degree of increase in QS under DMI treatment. Quiet sleep (QS) dependent consolidation in the striatum may depend on other processes such as changes in dopamine or acetylcholine levels during QS. Increases in QS have been shown to correlate with other procedural learning tasks such as soccer playing and tumbling on a trampoline32. Also, acquisition of tasks that depend on the dorsal striatum (e.g. caudate and putamen), like learning to always turn left at a T-maze junction33,34, often benefit when the competing hippocampal strategy is impaired35,36. In fact, striatal dependent learning strategies may be enhanced by REM sleep deprivation because they are enhanced by hippocampal inactivation. Basal REM sleep supports dorsal striatal procedural learning N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 13 Another group also found a positive correlation between the amounts of baseline REM sleep and striatal learning performance, the swim path length on a visible platform Morris water maze. Interestingly, amounts of serotonin in the striatal caudate structure correlated with the amounts of baseline REM sleep and the numbers of 5HT-2 receptors in the caudate correlated with this visible platform water maze task, implicating some function for baseline REM sleep and caudate serotonin levels in the motor task37 just as we found the relationship between baseline REM sleep and our striatal dependent fixed choice T-maze task. It would be interesting to test whether basal amounts of REM sleep correlate with all striatal-dependent learning tasks and whether serotonin is key to that involvement. Unified theory of sleep-dependent memory consolidation These results fit into a physiological conceptualization of sleep consolidation as depicted in Figure 6. The unifying principle underlying the TRand REM–related reversal and reconsolidation learning benefits is based on literature demonstrating that the noradrenergic (NA) cells of the locus coeruleus (LC) are only off during the initialization of the characteristic spindles (10-15 Hz large amplitude waves that last 1-2 seconds) of TR and during REM sleep38. The absence of NA is necessary for synaptic depotentiation (DP), and DP is necessary alongside its opposite, LTP to efficiently incorporate newly learned information into previously established memory networks39-41. Specifically, NA acts at beta adrenergic receptors to block the action of calcineurin. The activation of calcineurin leads to the dephosphorylation of CAMK-II and MAPK, whose phosphorylation is thought to sustain LTP (see 42). When NA blocks calcineurin, it blocks one half of the process of reshaping a synaptic network: the deconstruction side of remodeling, DP. Thus, DMI enhancement of NA levels supports synaptic strengthening (LTP), while preventing depotentiation in the hippocampus43,44. Striatal dependent tasks may be immune to REM and TR sleep N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 14 deprivation because the LC NA system does not project to the dorsal striatum45 and the dorsal striatum does not respond to NA46, and without NA, DP may always be possible. LC targets of the hippocampus and prefrontal cortex, however, may require the LC silence of TR and REM sleep as the only states that support efficient remodeling of memory networks for learning. Insert Figure 6 here DMI sleep effects parallel human antidepressant use These findings support those found by Rasch et al.47 who demonstrated that antidepressants enhanced procedural learning. Their study and ours indicate that the effects of REM suppression on striatum dependent procedural memory are quite different than on hippocampus dependent spatial memory. Some antidepressants, but not others, have been reported to improve memory function48, which may depend on the sleep outcomes of the particular antidepressant and the memory function tested. Antidepressant administration is an effective method of reducing REM sleep49. Stress is unlikely to mediate the changes seen under DMI treatment. Studies showing stress effects of REM sleep deprivation use a movement-restricting instrumental REM sleep deprivation continuously for many days7. This profile is quite unlike the partialday non-instrumental REM deprivation method used in this study and unlike the 4-6 h REM low water under multiple platform restriction method we have used in the past1,2, which likely avoids activation of the HPA axis. We were able to use a dose resulting in transient REM reduction that fully dissipated by 10 hours then returned to normal and produced this effect repeatedly across all 4 days of treatment. Both the pharmokinetics (despiramine’s half-life of 4.6 h) 47 and the general effects on sleep (810 h) architecture based on our results indicate that DMI’s action had ceased many hours before the following day’s run and thus did not likely mediate the change in N at ur e P re ce di ng s : h dl :1 01 01 /n pr e. 20 11 .6 52 4. 1 : P os te d 13 O ct 2 01 1 Watts et al. DMI suppresses REM sleep and impairs learning P a g e | 15 daily performance directly, but rather worked during the post-learning consolidation period when DMI actively suppressed REM sleep. Desipramine suppresses human REM sleep most strongly in the initial 3 weeks, whereafter REM sleep activity begins to increase slightly toward normal levels despite continued chronic administration 15. If the REM sleep state itself is responsible for hippocampal memory consolidation, then consolidation related performance deficits may improve to somewhat several weeks into treatment coincident with the partial restoration of REM sleep: a hypothesis that remains to be tested. Additionally, we determined that although REM sleep suppressing antidepressant treatments impaired hippocampus-dependent learning, alternative strategies were augmented which improve task performance. Patients taking REM-suppressing antidepressants may also benefit from selecting alternative, REM-immune, non-hippocampal strategies to solve the learning problems at hand.