Circadian Glucocorticoid Oscillations Are Critical

Circadian Glucocorticoid Oscillations Are Critical

for Learning-Induced Synaptic Remodeling and Maintenance

Conor Liston,* Joseph M. Cichon, Freddy Jeanneteau, Zhengping Jia,

Moses V. Chao, and Wen-Biao Gan*

*To whom correspondence may be addressed. Email cliston@stanford.edu or

gan@saturn.med.nyu.edu.

Supplementary Fig. 1: Effects of training and habituation on circadian oscillations in plasma corticosterone. In order to test the effects of training on glucocorticoid rhythms and to validate the effect of habituation, we measured plasma CORT in blood samples obtained from mice during the circadian trough or the circadian peak. Mice were either untrained, trained without habituation, or trained after 3 days of habituation (see Online Methods for protocol). Training occurred at the times specified in the figure, and blood samples were collected approximately 45 minute later. There were significant between-group differences in plasma CORT (F(5,26) = 18.3, p < 0.001). In all three conditions, CORT was significantly higher during the circadian peak than during the trough (t = 2.45-4.61, p = 0.002-0.04). Training at either time caused a significant increase in CORT in unhabituated mice relative to untrained controls (t = 4.39-5.44, p = 0.001-0.002). Habituation attenuated this effect: plasma CORT was significantly lower in habituated mice relative to unhabituated mice (t = 2.55-4.29, p = 0.002-0.03), and there was no significant difference between habituated mice and untrained controls (t = 0.96-1.35, p = 0.21-0.36). These results demonstrate that training increased plasma CORT in unhabituated mice, but the circadian rhythm remained intact in all three groups. They also show that the habituation protocol used in all training experiments reported in the main text was effective in attenuating the impact of training on circadian glucocorticoid rhythms. * =

significantly different from both untrained and training+habituation groups (p < 0.05). ** = peak vs. trough levels were significantly different in all three training conditions (p < 0.05).

1

Nature Neuroscience: doi:10.1038/nn.3387Liston et al., Supplementary Material, Page 2

Supplementary Fig. 2: Glucocorticoid effects on learning-related spine turnover in motor cortex were independent of changes in filopodia.

a) There were no significant differences in the rate of filopodium formation across the various training conditions depicted in Fig. 1 (F(6,33) = 0.40, p = 0.87), indicating that neither learning nor corticosterone affected 2-day filopodium formation rates. Filopodia were defined as long, thin protrusions with ratio of length to neck diameter > 3:1. Filopodia formation rates represent the number of new filopodia formed during the 2-day training period, expressed as a percent of the total number of protrusions (spines + filopodia) quantified at baseline.

b) To confirm that corticosterone effects on learning-related spine formation were independent of effects on filopodia, we pooled spines and filopodia and analyzed them together using univariate ANOVA, while controlling for number of filopodia sampled prior to treatment as a covariate. Filopodia were defined as above. The results were comparable to those reported in Fig. 1: training enhanced formation but only when it coincided with a period of elevated glucocorticoid exposure (F(6,33) = 17.1, p < 0.001). Error bars = S.E.M.; * = significantly different from untrained control (p < 0.05).

Nature Neuroscience: doi:10.1038/nn.3387 2

Liston et al., Supplementary Material, Page 3

Supplementary Fig. 3: RotaRod performance after two days of training was moderately correlated with two-day spine formation (r = 0.68, p = 0.001). In comparison, 7-day retention of the learned motor skill (Fig. 1f) was more highly correlated with the 7-day survival of spines formed during the training period (r = 0.86, p < 0.001). Error bars = S.E.M.; * = significantly different from corresponding control (p < 0.05).

Nature Neuroscience: doi:10.1038/nn.3387 3

Liston et al., Supplementary Material, Page 4

Supplementary Fig. 4: The effects of corticosterone exposure on retention of the learned motor skill were not confounded by time of testing on Day 7. There was a significant main effect of corticosterone exposure at the time of testing (F(1, 60) = 23.61, p < 0.001) but no main effect of time of testing on Day 7 (F(1, 60) = 0.17, p = 0.69). There was also no interaction between these two factors (F(1, 60) = 1.25, p = 0.27). Here, we collapsed across groups that were exposed to corticosterone at the time of training versus those that were not. Error bars = S.E.M.; * = significantly different from corresponding control (p < 0.05).

Nature Neuroscience: doi:10.1038/nn.3387 4

Liston et al., Supplementary Material, Page 5

Supplementary Fig. 5: Effects of pre-training habituation on spine formation and RotaRod performance. As noted in the main text, we found that mice that were trained during the trough failed to retain the benefits of training when tested seven days later (Fig. 1e). This observation may be due in part to the habituation procedure for two reasons. First, prior studies indicate that habituation reduces long-term memory by diminishing arousal at the time of learning.29 Second, the habituation procedure (running on the RotaRod at a constant low speed) may have allowed for some RotaRod-related spine formation, as well as some behavioral improvement, prior to obtaining our baseline image on Day 0 and commencing with training on the accelerating RotaRod on Day 1. If so, habituation may account in part for the lack of additional

improvement. To test this possibility, we compared spine formation and RotaRod performance in habituated and unhabituated mice trained during the circadian peak or trough.

a) 2-day spine formation rates increased after training during the circadian peak (~8pm; t = 6.63, p < 0.001) but not during the circadian trough (~8am; t = 1.05, p = 0.32). This effect was due in part to prior habituation. In unhabituated animals, there was a moderate but

significant increase in spine formation rates relative to untrained controls (t = 2.96, p = 0.03). Spine formation rates represent the number of new spines formed during the 2-day training period, expressed as a percent of the total number of spines quantified at baseline. * = Significantly different from untrained control (p<0.05); NS = not significant.

b) Corresponding patterns of RotaRod performance were observed on Day 2 and Day 7. Performance on Day 2 or Day 7 is expressed as a function of each subject’s baseline performance on Day 1. * = performance significantly different from Day 1 baseline (p < 0.05). ✝ = difference from Day 1 baseline approaches significance (p < 0.10). Error bars = SEM.

Nature Neuroscience: doi:10.1038/nn.3387 5

Liston et al., Supplementary Material, Page 6

Supplementary Fig. 6: Effects of glucocorticoid activity at the time of training on learning-associated spine pruning.

a) As in Fig. 1a, subjects were trained during the circadian glucocorticoid peak or trough, and some received exogenous injections of glucocorticoids. We quantified 2- and 7-day elimination rates for spines that existed prior to training.

b) Spine elimination on Day 2 was unaffected by training during the circadian peak (t = 0.22, p = 0.83) or trough (t = 0.44, p = 0.67).

c) However, training caused a delayed increase in spine elimination that was detectable on Day 7, but only if training occurred during a period of elevated glucocorticoid activity. * = significantly different vs. untrained control (p < 0.05). ✝ = difference vs. untrained control approaches significance (p < 0.10). Error bars = SEM.

Nature Neuroscience: doi:10.1038/nn.3387 6

Liston et al., Supplementary Material, Page 7

Supplementary Fig. 7: Glucocorticoid effects on spine formation are mediated by a non-genomic mechanism. To better understand the mechanisms by which glucocorticoids regulate spine formation in vivo, we applied corticosterone (10 µM) directly to the cortex through a small, adjacent craniotomy and quantified spine turnover using time-lapse 2-photon microscopy (see also Fig. 5c in main text).

a) Corticosterone had no effect on spine elimination at 20 minutes. For comparison, spine

formation effects depicted in Fig. 5c are re-plotted here. Both corticosterone (t = 11.2-14.6, p < 0.001) and a membrane-impermeable synthetic glucocorticoid (Cort:BSA; t = 4.83-6.36, p < 0.003) increased spine formation rapidly but not when administered with mifepristone, a selective GR-antagonist (t = 0.60-2.06, p > 0.09). These findings point to a GR-mediated, non-genomic mechanism. We tested this by co-administering corticosterone and

actinomycin D, an inhibitor of transcription. Actinomycin D alone had no effect on either formation (t = 0.23, p = 0.82) or elimination (t = 0.16, p = 0.88).

b) Similar effects of a greater magnitude were observed after 90 minutes. Drug doses were as described in the caption for Fig. 5. * = significantly different from vehicle-treated control.

Nature Neuroscience: doi:10.1038/nn.3387 7

Liston et al., Supplementary Material, Page 8

Supplementary Fig. 8: Glucocorticoid effects on spine elimination are not affected by co-treatment with a GR antagonist.

a) Corticosterone (15 mg/kg I.P.) increased spine elimination rates (main effect of treatment: F(2,11) = 20.7, p < 0.001; t = 6.15, p < 0.001 vs. control), and similar effects were observed after co-treatment with a GR antagonist (mifepristone, 20 mg/kg I.P.; t = 4.76, p < 0.003 vs. control). There was no significant difference between the groups treated with corticosterone vs. corticosterone+mifepristone (t = 0.31, p > 0.76). This suggests that glucocorticoid effects on spine pruning are not mediated by GR signaling.

b) In contrast, the effects of corticosterone on spine formation (main effect of treatment: F(2,11) = 49.4, p < 0.001; t = 7.99, p < 0.001 vs. control) were blocked by co-treatment with a GR antagonist (t = 1.34, p = 0.23 vs. control), which caused a significant reduction in spine formation (t = 9.15, p < 0.001 vs. corticosterone-treated group). This suggests that glucocorticoid effects on spine formation require GR activity.

Nature Neuroscience: doi:10.1038/nn.3387 8

Liston et al., Supplementary Material, Page 9

Supplementary Fig. 9: Glucocorticoids promote spine formation through non-genomic regulation of LIMK1 and cofilin.

a) We obtained cortical biopsies 20 minutes after administering corticosterone directly to the cortex as above. We then quantified expression levels for the phosphorylated forms of GR, LIMK, and cofilin in cortical lysates. Corticosterone caused rapid phosphorylation of the glucocorticoid receptor (GR), a biomarker of GR activity, as well as increased

phosphorylation of LIMK and cofilin (Fig. 6b-c). GR phosophorylation is depicted as the ratio of phospho-GR to total GR.

b) In corticosterone-exposed subjects, increased GR phosphorylation was associated with increased LIMK phosphorylation, though this correlation only approached significance (r = 0.31, p = 0.092).

c) LIMK1 and cofilin phosophorylation were in turn highly correlated (r = 0.79, p < 0.001), consistent with LIMK1’s established role in regulating cofilin activity.

d) Across subjects, GR activity (pGR) was significantly correlated with cofilin phosphorylation (r = 0.48, p = 0.006). Error bars = S.E.M; * = significantly different from vehicle-treated control (p < 0.05).

Nature Neuroscience: doi:10.1038/nn.3387 9

Liston et al., Supplementary Material, Page 10

Supplementary Table 1a: Circadian glucocorticoid peaks promote new spine formation after learning: Effects on 2-Day Spine Formation. “Day2 Formation” = spine formation rate over two days as a percent of the total number of spines quantified prior to training. SEM = standard error of the mean. N = number of mice. T statistics are for planned contrasts with untrained control. Statistics in bold face type are significant at p < 0.05, corrected. Of note, post-hoc tests showed that spine formation in the “Peak+cort after” group was also significantly greater than in the “Peak trained” group (t = 5.16, p < 0.001) and the “Peak+cort before” group (t = 5.46, p < 0.001), indicating a specific effect of glucocorticoids on learning-induced spine formation.

Condition Untrained Peak trained Trough trained Trough+cort Peak+cort after Peak+cort before Peak+Dex

n 5 10 5 5 6 5 4

Day2 Formation 7.04 11.70 8.04 12.42 16.77 11.94 1.83

SEM 0.30 0.64 0.91 0.74 0.69 0.50 0.47

t 6.63 1.05 6.74 12.95 8.42 9.38

F(6,33) 42.2

Supplementary Table 1b: Circadian glucocorticoid peaks promote new spine formation after learning: Effects on 7-Day Survival of Training-Related Spines. “Day7 Formation” = Day 7 survival of spines formed over two days of training, expressed as a percent of the total number of spines quantified prior to training. SEM = standard error of the mean. N = number of mice. T statistics are for planned contrasts with untrained control. Statistics in bold face type are significant at p < 0.05, corrected. Again, stable spine formation in the “Peak+cort after” group was significantly greater than in the “Peak+cort before” group (t = 3.77, p = 0.02).

Condition Untrained Peak trained Trough trained Trough+cort Peak+cort after Peak+cort before Peak+Dex

n 5 9 5 4 5 3 4

Day7 Formation 1.99 5.50 2.42 5.73 7.44 3.27 0.15

SEM 0.20 0.85 0.60 0.82 1.08 0.23 0.15

t 4.03 0.68 4.42 4.95 4.13 7.34

F(6,28) 9.40

Nature Neuroscience: doi:10.1038/nn.3387 10

Liston et al., Supplementary Material, Page 11

Supplementary Table 1c: Circadian glucocorticoid peaks promote new spine formation after learning: Effects on RotaRod Performance on Day 7. “Day7 RotaRod” = percent

change in RotaRod performance (mean RPM over 15 trials) on Day 7 relative to Day 1. SEM = standard error of the mean. N = number of mice. Statistics in bold face type indicate

performance was significantly different vs. baseline performance on Day 1 at p < 0.05, corrected.

Condition Peak trained Trough trained Trough+cort Peak+cort after Peak+cort before Peak+dex

n 15 7 7 17 6 9

Day7 RotaRod 39.07 -3.63 29.60 46.13 18.72 5.65

SEM 10.09 11.13 6.78 12.38 3.93 8.00

t 3.87 0.33 4.37 3.73 4.76 0.71

F(5,55) 2.84

Nature Neuroscience: doi:10.1038/nn.3387 11

Liston et al., Supplementary Material, Page 12

Supplementary Table 2a: Effects of Circadian Trough on Survival of Training-Related New Spines on Day 7. “Day7 Formation” = 7-day survival of spines formed during two days of training, expressed as a percent of the total number of spines quantified prior to training.

Statistics in bold face type are significant at p < 0.05, corrected, relative to untrained control. N = number of mice.

Condition Untrained

Trained during peak

+Cort during trough week 1 +Cort during peak week 1 +Cort during trough week 2

n 5 9 3 3 3

Day7 Formation 1.99 5.50 1.77 4.60 5.70

SEM 0.20 0.85 0.77 0.50 0.23

t 4.03 0.29 5.75 12.1

F(4,18) 5.02

Supplementary Table 2b: Effects of Circadian Trough on RotaRod Performance on Day 7. “Day7 RotaRod” = percent change in RotaRod performance (mean RPM over 15 trials) on Day 7 relative to Day 1. SEM = standard error of the mean. Statistics in bold face type are

significantly different vs. baseline performance on Day 1 at p < 0.05, corrected. N = number of mice.

Condition Vehicle control

Cort during trough week 1 Cort during peak week 1 Cort during trough week 2

n 15 10 10 4

Day7 RotaRod 39.1 -2.7 31.0 29.8

SEM 10.0 4.9 8.6 4.9

t 3.87 0.55 3.62 6.07

F(3,35) 4.24

Nature Neuroscience: doi:10.1038/nn.3387 12

Liston et al., Supplementary Material, Page 13

Supplementary Table 3: Pruning of Pre-training Spines Requires Intact Glucocorticoid Troughs After Training. Training induced a delayed increase in the elimination of pre-training spines that was evident on Day 7 but not on Day 2. This effect required intact glucocorticoid troughs during the week after training. “Day2 Elimination” and “Day7 Elimination” = 2- and 7-day elimination rates for spines that were present prior to training on Day 0, expressed as a percent of the total number of spines quantified prior to training. Statistics in bold face type are significantly different from untrained control at p < 0.05 after Holm-Bonferroni correction for multiple comparisons. Statistics in italics are significant at p < 0.05 uncorrected, p < 0.10 corrected. N = number of mice.

Condition Untrained

Trained during peak

n 5 9

Day2 Elimination 8.67 8.86

SEM 0.97 0.66

t 0.15

F(1,12) 0.03

Condition Untrained

Trained+vehicle week 1 Cort during trough week 1 Cort during peak week 1 Cort during trough week 2

n 5 6 3 3 3

Day7 Elimination 12.40 16.43 12.47 15.50

17.00 (Day 14)

SEM 1.02 1.32 0.60 0.46 0.90

t 2.86 0.06 2.77 3.38

F(4,15) 3.56

Nature Neuroscience: doi:10.1038/nn.3387 13

Liston et al., Supplementary Material, Page 14

Supplementary Table 4a: Chronic glucocorticoid exposure disrupts the survival of

learning-related spines that were present for at least one week. “Day20 Formation” = 20-day survival of spines formed over two days of training, expressed as a percent of total number of spines quantified prior to training. SEM = standard error of the mean. N = number of mice. Statistics in bold face type are significant at p < 0.05, relative to vehicle-treated control.

Condition Vehicle control Chronic cort

n 3 4

Day20 Formation 5.07 0.90

SEM 0.53 0.35

t 7.36

F(1,5) 54.2

Supplementary Table 4b: Chronic glucocorticoid exposure impairs RotaRod Performance on Day 20. “Day20 RotaRod” = percent change in RotaRod performance (mean RPM over 15 trials) on Day 20 relative to Day 1. SEM = standard error of the mean. N = number of mice. Statistics in bold face type are significant at p < 0.05, relative to vehicle-treated control.

Condition Vehicle control Chronic cort

n 7 11

Day20 RotaRod 26.63 -2.54

SEM 4.04 2.51

t 6.50

F(1,16) 42.2

Supplementary Table 4c: Chronic glucocorticoid exposure increases the elimination of spines present prior to training. “Day20 Elimination” = 20-day spine elimination rate

expressed as a percent of total number of spines quantified prior to training. SEM = standard error of the mean. N = number of mice. Statistics in bold face type are significant at p < 0.05, relative to vehicle-treated control.

Condition Vehicle control Chronic cort

n 3 4

Day20 Elimination 15.80 25.13

SEM 0.92 0.72

t 8.16

F(1,5) 66.6

Nature Neuroscience: doi:10.1038/nn.3387 14

Liston et al., Supplementary Material, Page 15

Supplementary Table 4d: Chronic glucocorticoid exposure increases the rate of net spine loss. “Day20 Spine Loss” = 20-day change in spine number relative to the change observed in vehicle-treated controls. For comparison, absolute values for 20-day spine loss as a percent of the total number of spines quantified prior to training are reported in parentheses. SEM =

standard error of the mean. N = number of mice. Statistics in bold face type are significant at p < 0.05.

Condition Vehicle control Chronic cort

n 3 4

Day20 Spine Loss 0.0 (5.2) -10.3 (15.5)

SEM 0.71 0.59

t

11.22

F(1,5) 125.9

Supplementary Table 4e: Learning has no significant effect on total spine number. “Day7 Spine Loss” = 7-day change in spine number relative to the change observed in untrained

controls. For comparison, absolute values for 7-day spine loss as a percent of the total number of spines quantified prior to training are reported in parentheses. SEM = standard error of the mean. N = number of mice. There were no significant differences observed in this analysis.

Condition

Untrained control PM trained AM trained AM+cort PM+cort

N 5 6 4 4 5

Day7 Spine Loss 0.0 (4.5) +1.2 (3.3) +0.4 (4.1) -0.8 (5.3) -1.0 (5.5)

SEM 1.03 0.40 0.68 0.69 0.30

t 1.06 0.31 0.66 0.99

F(4,19) 2.07

Nature Neuroscience: doi:10.1038/nn.3387 15

Liston et al., Supplementary Material, Page 16

Supplementary Table 5a: Effects of corticosterone on spine formation hours after

exposure. Corticosterone increased spine formation rapidly, within one hour after injection. F statistics depict the results of two-factor ANOVA with corticosterone treatment and time as fixed factors. T statistics are for contrast of corticosterone vs. vehicle-treated control at each time point. Statistics in bold face type are significant at p < 0.05, corrected. N = number of mice.

Main Effect of Cort Treatment:

Main Effect of Time:

Interaction:

Vehicle Control Time Formation (+/- SEM) 1 0.5 +/- 0.3 % 2 1.6 +/- 0.5 % 3 2.4 +/- 1.0 % 5 3.6 +/- 0.9 % 24 5.6 +/- 0.5 %

F(1,40) = 124.0, p < 0.001

F(4,40) = 44.4, p < 0.001 F(4,40) = 10.9, p < 0.001

Corticosterone (15 mg/kg I.P.)

N Formation (+/- SEM) N 4 1.6 +/- 0.3 % 4 3 5.1 +/- 0.3 % 4 3 8.0 +/- 0.4 % 4 5 11.9 +/- 0.8 % 5 9 8.7 +/- 0.4 % 9

t 2.63 6.50 5.88 7.17 5.13

Supplementary Table 5b: Effects of corticosterone on spine elimination hours after

exposure. Corticosterone caused a delayed increase in spine elimination that was detectable five hours after injection, but not before, and continued to accumulate over 24 hours. F statistics depict the results of two-factor ANOVA with corticosterone treatment and time as fixed factors. T statistics are for contrast of corticosterone vs. vehicle-treated control at each time point. Statistics in bold face type are significant at p < 0.05 after Holm-Bonferroni correction for multiple comparisons. Statistics in italics are significant at p < 0.05 uncorrected, p < 0.10 corrected. N = number of mice.

Main Effect of Cort Treatment:

Main Effect of Time:

Interaction:

Vehicle Control Time Elim (+/- SEM) 1 0.1 +/- 0.1 % 2 1.0 +/- 0.6 % 3 1.6 +/- 0.5 % 5 3.0 +/- 0.9 % 24 5.8 +/- 0.4 %

F(1,40) = 102.3, p < 0.001

F(4,40) = 129.1 p < 0.001 F(4,40) = 30.4, p < 0.001

Corticosterone (15 mg/kg I.P.)

N Elim (+/- SEM) N 4 0.2 +/- 0.2 % 4 3 1.2 +/- 0.6 % 4 3 3.7 +/- 0.6 % 4 5 11.4 +/- 0.8 % 5 9 14.6 +/- 0.4 % 9

t 0.23 0.21 2.64 6.91 15.5

Nature Neuroscience: doi:10.1038/nn.3387 16

Liston et al., Supplementary Material, Page 17

Supplementary Table 5c: Corticosterone induces rapid spine formation via non-genomic mechanisms. 20- and 90-minute spine formation rates are reported as mean +/- SEM. F statistics depict the results of two-factor ANOVA with glucocorticoid treatment and time as

fixed factors. T statistics are for the contrast of each drug treatment with vehicle-treated control. Statistics in bold face type are significant at p < 0.05, corrected. N = number of mice. “Cort” = corticosterone (10 µM) administered directly to cortex through a craniotomy. “Cort+Actino.” = corticosterone co-administered with actinomycin D (50 µg/mL). “Cort:BSA” = membrane-impermeable corticosterone / bovine serum albumin conjugate (10 µM). “Cort+Mifep.” =

corticosterone (10 µM) co-administered with mifepristone (100 µM), a selective glucocorticoid receptor antagonist.

Main Effect of Drug Treatment:

Main Effect of Time:

Interaction:

Treatment 20-min Formation Vehicle Ctrl 0.8 +/- 0.3 % Cort 6.4 +/- 0.3 % Cort+Actino. 8.3 +/- 0.3 % Cort:BSA 6.1 +/- 1.1 % Cort+Mifep. 0.6 +/- 0.2 %

F(4,27) = 108.5, p < 0.001 F(1,27) = 100.6, p < 0.001 F(4,27) = 14.2, p < 0.001 N t 90-min Formation 5 1.8 +/- 0.5 % 4 14.6 13.7 +/- 0.7 % 3 18.8 16.6 +/- 1.1 % 3 4.83 13.1 +/- 1.5 % 4 0.60 0.6 +/- 0.4 %

N 3 4 3 4 4

t 11.2 12.4 6.36 2.06

Supplementary Table 5d: Glucocorticoids promote learning-related spine formation through a GR-dependent, non-genomic process. To evaluate the role of GR-dependent, non-genomic signaling in learning-dependent spine formation, we trained mice for two days and

administered mifepristone (20 mg/kg I.P.) immediately after training. Statistics in bold face type are significant at p < 0.05, corrected. N = number of mice.

Condition

Trained during peak (ctrl) +GR antagonist

n 10 4

Day 2 Formation 11.7 2.8

SEM 0.6 0.4

t 8.46

F(1,12) 71.5

Nature Neuroscience: doi:10.1038/nn.3387 17

Liston et al., Supplementary Material, Page 18

Supplementary Table 5e: Effects of MR modulators on 24-hour spine elimination are dependent on transcription and translation. Direct cortical application of a selective MR antagonist (spironolactone, 10 µM) reduced 24-hour spine elimination, while a selective MR agonist (aldosterone, 10 µM) increased spine pruning. This effect was blocked by co-treatment with either actinomycin D or anisomycin, consistent with a genomic mechanism of action. T statistics are for contrast of drug vs. vehicle-treated control. Statistics in bold face type are significant at p < 0.05, after Holm-Bonferroni correction for multiple comparisons. Statistics in italics are significant at p < 0.05 uncorrected, p < 0.10 corrected. N = number of mice.

Condition

Vehicle (control) Spironolactone Aldosterone

Aldo + actinomycin D Aldo + anisomycin

n 5 3 4 5 6

24-hr Elimination 4.8 1.6 16.8 5.1 3.4

SEM 0.4 0.4 0.6 0.6 0.4

t 6.19 18.3 0.46 2.66

F(4,18) 151.1

Supplementary Table 5f: Learning-associated spine elimination requires MR-signaling during the circadian trough in the days after training. To evaluate the role of MR-dependent signaling in learning-related spine pruning, we administered spironolactone during the circadian trough on Days 4, 5, and 6 after training and found that 7-day spine elimination was reduced. Statistics in bold face type are significant at p < 0.05 relative to trained control. N = number of mice.

Condition

Trained during peak (ctrl) +MR antag on Days 4-6

n 6 5

7-day Elimination 16.4 9.6

SEM 1.3 0.5

t 4.84

F(1,9) 20.0

Nature Neuroscience: doi:10.1038/nn.3387 18

Liston et al., Supplementary Material, Page 19

Supplementary Table 6a: Corticosterone causes rapid phosphorylation of LIMK1 and cofilin in vivo. “pLIMK1” = phospho-LIMK1 levels, normalized to total LIMK1 levels in cortical lysates obtained immediately after 20 minutes of corticosterone exposure (10 µM).

“pCofilin” = phospho-cofilin levels, normalized to total cofilin levels in the same cortical lysates. SEM = standard error of the mean. T statistics are for planned contrasts with vehicle-treated controls. Statistics in bold face type are significant at p < 0.05. N = number of mice.

Condition Vehicle control

Corticosterone (10 µM) Cort+Mifepristone

Vehicle+Actinomycin D Cort+Actinomycin D

n 15 15 16 11 16 n 15 15 16 11 16

pLIMK1 0.80 1.26 0.87 0.81 1.19 pCofilin 0.74 2.92 1.11 0.88 2.08

SEM 0.08 0.19 0.12 0.05 0.15 SEM 0.08 0.52 0.15 0.09 0.50

t 2.23 0.48 2.42 t 4.44 0.77 2.26

F(4,68) 2.74 F(4,68) 6.95

Condition Vehicle control

Corticosterone (10 µM) Cort+Mifepristone

Vehicle+Actinomycin D Cort+Actinomycin D

Nature Neuroscience: doi:10.1038/nn.3387 19

Liston et al., Supplementary Material, Page 20

Supplementary Table 6b: Corticosterone causes rapid phosphorylation of LIMK1 and cofilin in cortical pyramidal cells in a co-culture system, but not after transfection with an interfering RNA construct specific for GR. “pLIMK1” = phospho-LIMK1

immunofluorescence (arbitrary units) in cortical pyramidal cell dendrites after 20 minutes of corticosterone exposure (10 µM). “pCofilin” = phospho-cofilin immunofluorescence; “pGR” = phosphorylated glucocorticoid receptor immunofluorescence. SEM = standard error of the mean. T statistics are for contrasts of CORT- vs. vehicle-treated control. Statistics in bold face type are significant at p < 0.05. N = number of analyzed cells.

Scrambled shRNA Construct: Condition n pLIMK1 Vehicle control 10 28.9 Corticosterone (10 µM) 11 41.9 Condition n pCofilin Vehicle control 12 57.6 Corticosterone (10 µM) 17 94.6 Condition n pGR Vehicle control 9 70.4 Corticosterone (10 µM) 11 93.1

GR-specific Interfering shRNA Construct: Condition n pLIMK1 Vehicle control 19 35.1 Corticosterone (10 µM) 18 33.6 Condition n pCofilin Vehicle control 10 50.6 Corticosterone (10 µM) 14 52.9 Condition n pGR Vehicle control 13 40.4 Corticosterone (10 µM) 7 36

SEM

0.9 1.2

SEM 2.2 2.8

SEM 2.8 1.5

t 8.53 t 9.69 t 7.52

F(1,19) 73.0

F(1,27) 94.5

F(1,19) 56.4

SEM 0.9 1.0

SEM 2.0 2.2

SEM 2.3 1.5 t 1.13 t 0.74 t 1.31 F(1,35) 0.59

F(1,22) 0.55

F(1,18) 1.72

Nature Neuroscience: doi:10.1038/nn.3387 20

Liston et al., Supplementary Material, Page 21

Supplementary Table 6c: Glucocorticoid effects on spine formation require LIMK1.

Whereas corticosterone promoted rapid spine formation in wild type controls, corticosterone had no effect on 20-minute or 24-hour formation after a 20-minute exposure (10 µM) in LIMK1

knockouts. Spine formation rates are reported as mean +/- SEM. F statistics depict the results of two-factor ANOVA with corticosterone treatment and genotype as fixed factors. T statistics are for contrasts of CORT- vs. vehicle-treated controls for each genotype. T statistics for contrasts between vehicle treated wildtypes and LIMK1 knockouts are reported in parentheses. Statistics in bold face type are significant at p < 0.05, corrected. N = number of mice.

Main Effect of Cort Treatment: Main Effect of Genotype:

Interaction:

Treatment Wild-type Vehicle Ctrl 0.8 +/- 0.3 % Cort 6.4 +/- 0.3 %

20-minute Spine Formation

F(1,11) = 58.7, p < 0.001 F(1,11) = 85.0, p < 0.001 F(1,11) = 71.1, p < 0.001 N t LIMK1 Knockout 5 0.6 +/- 0.6 % 4 14.6 0.3 +/- 0.3 %

N 3 3

t (0.50)

0.42 (1.27)

Main Effect of Cort Treatment: Main Effect of Genotype:

Interaction:

Treatment Wild-type Vehicle Ctrl 5.9 +/- 0.3 % Cort 23.6 +/- 0.5 %

24-hour Spine Formation

F(1,13) = 636.2, p < 0.001 F(1,13) = 1005.5, p < 0.001 F(1,13) = 645.5, p < 0.001 N t LIMK1 Knockout 5 3.7 +/- 0.3 % 4 33.5 3.6 +/- 0.3 %

N 4 4

t (5.34)

0.14 (5.55)

Nature Neuroscience: doi:10.1038/nn.3387 21