Sustained rescue of prefrontal circuit dysfunction by antidepressant-induced spine formation

Abstract

The neurobiological mechanisms underlying the induction and remission of depressive episodes over time are not well understood. Through repeated longitudinal imaging of medial prefrontal microcircuits in the living brain, we found that prefrontal spinogenesis plays a critical role in sustaining specific antidepressant behavioral effects and maintaining long-term behavioral remission. Depression-related behavior was associated with targeted, branch-specific elimination of postsynaptic dendritic spines on prefrontal projection neurons. Antidepressant-dose ketamine reversed these effects by selectively rescuing eliminated spines and restoring coordinated activity in multicellular ensembles that predict motivated escape behavior. Prefrontal spinogenesis was required for the long-term maintenance of antidepressant effects on motivated escape behavior but not for their initial induction.

Fig. 1.. Targeted, branch-specific spine remodeling underlies behavioral state transitions.

(A) 2P spine imaging of YFP-expressing neurons through a microprism implanted across the midline in the mPFC before and after 21 days of chronic CORT exposure and after antidepressant-dose ketamine (post-ket) (10 mg/kg ip). Red and blue arrows indicate eliminated and formed spines, respectively. FrA, frontal association cortex; ACg, anterior cingulate cortex; PL, prelimbic cortex; IL, infralimbic cortex. (B) Spine elimination increased (*P = 0.0043, Wilcoxon W = 15.0) after 21 days of chronic CORT (n = 6 mice) compared with that in controls (n = 5 mice), whereas ketamine had no effect on spine elimination (n = 7 mice) compared with that in vehicle-treated controls (n = 6 mice). Throughout, box plots depict the median, interquartile range, and minimum and maximum excluding outliers, and outliers are plotted individually. NS, not significant. (C) Spine formation decreased (*P = 0.0087, Wilcoxon W = 44.5) after 21 days of chronic CORT, whereas ketamine increased spine formation. (D)Spine elimination per dendritic branch segment was significantly elevated (t = 8.05, df = 297, P = 1.9 × 10⁻¹⁴) and bimodally distributed in the chronic CORT-exposed group (n = 149 branches from 5 mice) compared with that in controls (n = 150 branches from 6 mice), such that 39.6% of dendritic branch segments in the chronic CORT-exposed group exhibited spine elimination rates that were >2 standard deviations above the mean for controls. (E) Spatially clustered spine elimination after chronic CORT exposure. The average distance between an eliminated spine and its nearest eliminated neighbor (2.5 μm) (red circle) was significantly reduced (P < 0.001), relative to what would be expected by chance given the observed rate of spine elimination, in simulations denoted by the box plot (see supplementary materials). (F) Spine formation per dendritic branch segment was significantly elevated (t = 9.48, df = 205, P = 6.7 × 10⁻¹⁸) and bimodally distributed in the ketamine-treated group (n = 103 branches from 7 mice) compared with that in vehicle-treated controls (n = 104 branches from 7 mice), such that 51.5% of dendritic branch segments in the ketamine-treated group exhibited spine formation rates that were highly elevated compared with those in vehicle-treated controls. (G) The average distance between a newly formed spine and its nearest eliminated neighbor (2.8 μm) (blue circle) was significantly reduced (P < 0.001) relative to what would be expected by chance given the observed rate of spine formation, as estimated in simulated data.

Fig. 2.. Ketamine selectively restores spines lost during chronic CORT exposure.

(A) Experimental timeline (identical to that in Fig. 1A). Restored spines (blue) were defined as spines that formed in the 24-hour period after ketamine treatment in a position <2 μm from a previously eliminated spine. (B) Of newly formed spines, 47.7% were located <2 μm from a spine lost during chronic CORT exposure (n = 7 mice), a proportion significantly greater than the 14.5% observed in vehicle-treated controls (n = 6 mice; *P = 0.0012, Wilcoxon W = 21.0). (C) The observed fraction of restored spines (blue circle) was significantly (P < 0.001) greater than expected by chance given the observed rate of spine formation in simulated data in which formed spines were randomly distributed (box plot) (see supplementary materials). (D) The change in spine density from day 1 (baseline) to day 21 (post-chronic CORT) or day 22 (postketamine) was quantified by subtracting the spine elimination rate from the spine formation rate during the specified imaging interval. Ketamine reversed chronic CORT effects on spine density [Kruskal-Wallis analysis of variance (ANOVA), **P = 5.6 × 10⁻⁴, χ² = 14.96]. (E) Of spines lost during chronic CORT, 48.3% were restored after ketamine treatment (n = 7 mice) versus 3.3% in vehicle-treated controls (n = 6 mice; **P = 0.0012, Wilcoxon W = 21.0). (F) Representative confocal image of PSD-95 immuno-fluorescence in the mPFC. Scale bar, 30 μm. (G) PSD-95 density (z-scored with respect to the mean density in controls; N = 18 samples from n = 6 subjects per group) varied by experimental condition (mixed-effects ANOVA, F_2,45 = 10.39, P = 0.0019). In the dorsal mPFC, PSD-95 density was significantly lower in the chronic CORT-treated group than in untreated controls (t₁₀ = 2.43, *P = 0.035). Ketamine reversed this effect (t₁₀= 4.15, **P = 0.002). (For other PFC subregions, see fig. S6.)

Fig. 3.. Ketamine rescues CORT-induced microcircuit dysfunction.

(A) 2P calcium imaging of GCaMP6s-expressing neurons through a microprism implanted across the midline in the mPFCs of head-fixed, awake mice. Calcium imaging occurred before and after 10 days of chronic CORT treatment and 24 hours after ketamine treatment (10 mg/kg ip). (B) Raster plots depicting changes in activity (ΔF/F) over time for a representative neuronal population before and after chronic CORT and 1 day after ketamine treatment. Each row represents a single cell. Green bands below denote statistically significant multicellular ensemble events. (C) Representative traces depicting the proportion of simultaneously active cells over time for the neuronal population depicted in (B).The green dashed line delineates the upper end of a 99.9% confidence interval for the proportion of simultaneously active cells in shuffled data (see supplementary materials). (D) Representative correlation matrices quantifying functional connectivity between each cell and every other cell. (E) Cumulative distribution plot of cell-cell correlations for all cells across all animals in each condition (N = 98,878 correlations at baseline, N = 145,467 after chronic CORT, N = 114,593 after ketamine from n = 3 mice). Mixed-effects ANOVA (subject = random effect) showed that the distribution shifted left (decreased) after chronic CORT (mean correlation coefficient r = 0.084 versus 0.188 at baseline; t = 3.02; P = 0.0025 for the difference between chronic CORT and baseline), and ketamine rescued this effect (mean r = 0.187; t = 0.39; P = 0.698 for the difference between ketamine and baseline). (F) The frequency of multicellular ensemble events (expressed as the probability of an ensemble event occurring in any given time frame) decreased after chronic CORT (median, 8.4% versus 17.4% at baseline) and was rescued after ketamine treatment (chronic CORT + Ket) (median, 14.7%; mixed-effects ANOVA, F_2,15 = 17.1, P = 0.00013 for main effect of group; N = 8 observations per experimental condition). (G) The proportion of cells participating in a multicellular ensemble event also decreased after chronic CORT (mean = 37.7% versus 74.1% at baseline for N = 39 and N = 28 multicellular ensemble events in n = 3 mice) and was rescued after ketamine treatment (mean = 63.8% for N = 35 events; mixed-effects ANOVA, F_2,93 = 10.8, P = 0.024). By contrast, in untreated controls, no significant changes occurred over time in any of the measures in (E) to (G) (fig. S5). *P < 0.05 (significantly different post hoc contrast, Holm-Bonferroni corrected).

Fig. 4.. Ketamine rescues CORT effects on coordinated PFC activity and active avoidance behavior.

(A) (Top) Experimental timeline and schematic showing optical fiber placement in the vmPFC and representative fiber photometry recording of GCaMP6s-expressing neurons during tail suspension. (Bottom) A representative photometry trace in which pink bands indicate periods of struggling to escape. Insets show representative examples of GCaMP6s activity increasing before the onset of struggling. (B) Mean vmPFC activity (ΔF/F) ± SEM (gray band) time locked to the onset of struggling (time = 0), averaged over all bouts of struggling across all mice. Increased vmPFC activity preceded the onset of struggling by ~1 s. (C) Distribution of calcium-transient events during bouts of immobility by latency to the onset of the next bout of struggling. Monte Carlo simulation showed that during bouts of immobility, calcium transients predicted the onset of struggling: The median latency to the onset of struggling (4.0 s) was significantly shorter than would be expected by chance (7.0 s) in randomly shuffled data (P < 0.0001 across 10,000 iterations), and 37.1% of all transients occurred <1 s before the onset of struggling. (D) Cumulative distribution of calcium transients by latency to the onset of struggling, grouped by transient amplitude. Linear mixed-effects modeling showed that latency to the onset of struggling varied with transient amplitude (t₃₀₃ = 3.84, P = 0.0015), such that larger transients predicted shorter latencies. The gray band denotes the 99% confidence interval for the median latency in shuffled data. The median latency for transients in all four groups was less than that expected by chance in shuffled data, with the largest effects occurring for transients in the upper quartile (median latency = 0.69 s). (E) Chronic CORT increased the total duration of immobility during tail suspension, whereas ketamine rescued this effect [n = 8 control mice, n = 10 chronic CORT-treated mice, n = 14 chronic CORT-plus-ketamine (CORT+ket)-treated mice; Kruskal-Wallis ANOVA, *P = 0.0003, χ² = 15.93]. (F) Chronic CORT significantly reduced the frequency of vmPFC calcium transients during tail suspension, and ketamine rescued this effect (n = 8 control mice, n = 10 chronic CORT-treated mice, n = 11 ketamine-treated mice; Kruskal-Wallis ANOVA, *P = 0.0083, χ² = 9.59). (G) Cumulative distribution of calcium transients by latency since the onset of immobility showing the accumulation of transients overtime, sorted by experimental condition. Linear mixed-effects modeling showed that vmPFC activity accumulated more slowly over time after chronic CORT exposure (median latency, 10.2 s for n = 378 transients versus 7.2 s for n = 305 transients in controls), and ketamine rescued this effect (median latency, 6.8 s for n = 378 transients; t₉₈₃ = 2.21).

Fig. 5.. Prefrontal spinogenesis is not required for inducing ketamine’s effects on behavior or circuit function.

(A) Experimental timeline for data in (B) to (D). To avoid habituation in the TST due to repeated testing on the same day, each subject in (B) was tested only once in this assay (between-subjects design; N = 12 subjects per time point per group). For 2P imaging in (C), each subject was assessed repeatedly at each specified time point (within-subjects design; N = 12 subjects per group). (B) Rapid effect of ketamine on immobility in the TST (main effect of treatment, F_1,132 = 149.1, P < 0.0001; main effect of time, F_5,132 = 1.20, P = 0.314; interaction, F_5,132 = 1.24, P = 0.295). *P < 0.002 (significant post hoc linear contrast). (C) Delayed effect of ketamine on spine formation (main effect of treatment, F_1,108 = 145.7, P < 0.0001; main effect of time, F_5,108 = 81.6, P < 0.0001; interaction, F_5,108 = 17.4, P < 0.000l). *P < 0.0006 (significant post hoc linear contrast). NS, not significant. (D) The survival of restored spines (the percentage of spines lost after chronic CORT and restored after ketamine) was significantly correlated with immobility behavior 2 to 7 days after treatment (Spearman’s rho = 0.770, P = 0.0034). (E) Experimental timeline for data in (F) to (K). To test whether ketamine effects on microcircuit activity preceded or followed effects on spines, 2P calcium imaging was performed before and after chronic CORT and 3, 12, and 24 hours after ketamine treatment (within-subjects design). (F) Raster plots depicting changes in activity (ΔF/F) over time for a representative neuronal population before and after chronic CORT and 1 day after ketamine treatment, as in Fig. 3B.(G) Representative traces depicting the proportion of simultaneously active cells over time for the neuronal population depicted in (F). The green dashed line delineates the upper end of a 99.9% confidence interval for the proportion of simultaneously active cells in shuffled data (see supplementary materials). (H) Representative correlation matrices quantifying functional connectivity between each cell and every other cell. (I) There was a reduction in functional connectivity (mean correlation) after chronic CORT, and ketamine rescued this effect within 3 hours of treatment (N = 5 to 8 samples per condition from n = 3 subjects; mixed-effects ANOVA, subject = random effect; main effect of time, F_4,32 = 3.05, P = 0.032). For cumulative distribution functions for all correlation coefficients, see fig. S12. (J) The frequency of multicellular ensemble events (expressed as the probability of an ensemble event occurring in any given time frame) decreased after chronic CORT and was rescued within 3 hours of ketamine treatment (mixed-effects ANOVA; main effect of time, F_4,32 = 9.28, P = 4.27 × 10⁻⁵). (K) The proportion of cells participating in a multicellular ensemble event also decreased after chronic CORT and was rescued within 3 hours of ketamine treatment (mixed-effects ANOVA; main effect of time, F_4,341 = 7.65, P = 6.55 × 10⁻⁶). For (I) to (K), *P < 0.05 (significantly different post hoc contrast, Holm-Bonferroni corrected).

Fig. 6.. Prefrontal spinogenesis sustains antidepressant-induced behavioral recovery over time.

(A) Experimental time course schematic. To control for individual differences in TST behavior as in a recently published report (72), we tested subjects before chronic CORT exposure and again 2 days after ketamine treatment (Rx) to determine whether ketamine reversed CORT effects on immobility and whether maintaining this effect required spine formation. (B) Photoactivating AS-PaRac1 (N = 7 mice) blocked ketamine effects on spine formation (Wilcoxon W = 50.0, *P = 0.0025) (right) and increased spine elimination (Wilcoxon W = 18.0, *P = 0.0177) (left) compared with that in controls not expressing AS-PaRac1 (N = 5 mice). (C) Photoactivating AS-PaRac1 interfered with the maintenance of ketamine’s effects on immobility behavior (F_1,17 = 9.96, *P = 0.0058) and (D) calcium-transient frequency (F₁,₁₇ = 16.6, *P = 0.019) during tail suspension 2 days after treatment. (E) No effect of photoactivating AS-PaRac1 on sucrose preference behavior 2 days after ketamine treatment (N = 8 mice per group; Wilcoxon W = 63.0, P = 0.645). (F) Photoactivating AS-PaRac1 3 hours after ketamine treatment (before the onset of ketamine’s effects on spine formation) had no effect on TST behavior 2 days after treatment (F_1,17 = 0.201, P = 0.660, N = 10 AsPaRac1 mice and N = 9 controls). NS, not significant.