Expression of TrkB in PV+ interneurons is important for fear erasure induced by fluoxetine treatment

We have previously demonstrated that chronic treatment with fluoxetine combined with fear extinction training promotes the erasure of previously acquired fear memory and alters the configuration of PV+ interneurons thereby reducing the proportion of PV+ interneurons expressing PNN [13]. We first tested whether this promoted fear erasure might depend on TrkB expressed in PV+ interneurons and would therefore be blunted in PV-TrkB hCKO mice (Fig. 1a). In a fear-conditioning paradigm (Fig. 1b), all mice were conditioned with a shock paired with a sound cue in context A during the fear-conditioning/acquisition phase, resulting in an increased freezing that was comparable in duration across all groups, although PV-TrkB hCKO mice conditioned faster than wild-type mice (two way ANOVA, Trials, F (4, 270) = 21.94, P < 0.0001; Genotype, F (1, 270) = 4.049, P < 0.0452; post hoc, wild-type vs PV-TrkB hCKO in Trial 3, P = 0.0037) (Fig. 1c). The wild-type and PV-TrkB hCKO mice were then assigned equally and randomly into groups receiving either water or water supplied with 0.08% (w/v) of fluoxetine, both enriched with 0.1 % (w/v) saccharin. Two weeks later, the mice were exposed to the conditioned stimulus (“beep” sound) in context B during 2 days of extinction training. In the wild-type group, two-way ANOVA showed a significant trial effect in both control and fluoxetine-treated mice (Effect of trials, Ext1, F (11, 336) = 1.988, P = 0.0288; Ext2, F (11, 336) = 9.624, P < 0.0001), indicating that the freezing response decreased during the extinction training (Fig. 1d). Moreover, the fluoxetine treatment showed a significantly stronger effect on the extinction training compared to water during both Ext1 and Ext2 (Effect of treatment, Ext1, F (1, 336) = 34.34, P < 0.0001; Ext2, F (1, 336) = 39.68, P < 0.0001) (Fig. 1d). Both control and fluoxetine-treated PV-TrkB hCKO mice showed decreased freezing after the second day of extinction training (Effect of trials, Ext2, F (11, 288) = 0.8726, P < 0.0001), but not after day 1 (Effect of trials, Ext1 F (11, 288) = 0.8743, P = 0.5660). In PV-TrkB hCKO mice, fluoxetine treatment failed to significantly enhance extinction when compared to water-treated mice (Effect of treatment, Ext1, F (1, 288) = 3.866, P = 0.0502; Ext2, F (1, 288) = 3.776, P = 0.0530) (Fig. 1e). Moreover, the effect of fluoxetine treatment on extinction was significantly better in the wild-type mice than in the PV-TrkB hCKO mice during both days of extinction (difference in freezing (delta) between water and fluoxetine-treated mice in each extinction session, t-test, Ext1, F = 1.616, DFn = 11, P = 0.0071; Ext2, F = 1.887, DFn = 11, P = 0.0108) (Fig. 1f), suggesting that in the absence of TrkB in PV neurons, the effects of fluoxetine are significantly reduced. One week later, the fluoxetine-treated wild-type mice showed decreased freezing compared to the water-treated wild-type mice throughout the whole session in context B (spontaneous recovery, SR) (Treatment, F (1, 112) = 14.05, P = 0.0003; Sidak’s post hoc test “water-treated vs Fluoxetine-treated” Trial 1, P = 0.0387) (Fig. 1g), as well as in the first session of this test (Treatment, F (1, 52) = 4.585, P = 0.0370; Sidak’s post hoc test “water-treated vs Fluoxetine-treated” P = 0.0229) (Fig. 1h). The freezing in the first session of SR and FR has been used for estimating cued and contextual memory, respectively [13, 32], as it is the only session that has not been influenced by a possible gradual habituation or extinction-training effect due to the test itself. However, the PV-TrkB hCKO mice failed to show similar effects of the fluoxetine treatment throughout the trials of the test (two-way ANOVA, SR, Treatment, F (1, 96) = 1.876, P = 0.1739) (Fig. 1g), and in the first trial (two-way ANOVA, treatment, F (1, 52) = 4.585, P = 0.0370, Sidak’s post hoc test “water-treated vs Fluoxetine-treated” for PV-TrkB hCKO P = 0.8618) (Fig. 1h). In addition, the treatment significantly reduced the freezing in the fear renewal test (FR) in context A in wild-type mice, especially in the first session of (Treatment, F (1, 112) = 14.56, P = 0.0002; Sidak’s post hoc test “water-treated vs Fluoxetine-treated” Trial 1, P = 0.0159) (Fig. 1i) (Treatment, F (1, 52) = 7.465, P = 0.0086; Sidak’s post hoc test “water-treated vs Fluoxetine-treated” P = 0.0122) (Fig. 1j). Interestingly, the treatment with fluoxetine decreased the overall freezing of PV-TrkB hCKO mice in the fear renewal test (Treatment, F (1, 96) = 16.61, P < 0.0001, Sidak’s post hoc test “water-treated vs Fluoxetine-treated” Trial 1, P = 0.7996, Trial 2, P = 0.0195, Trial 3, P = 0.0243) (Fig. 1i), but there was no difference in the first session (Treatment, F (1, 52) = 7.465, P = 0.0086; Sidak’s post hoc test “water-treated vs Fluoxetine-treated” P = 0.4948) (Fig. 1j). The absence of an effect of Fluoxetine on the PV-TrkB hCKO mice in the SR and a presence of a smaller effect in the FR suggest a role of TrkB expression in PV neurons in the extinction-enhancing effects of fluoxetine in cued fear conditioning, but a less pronounced role in the contextual component of the paradigm (FR).

Expression of TrkB in PV+ interneurons is important for the improvement of reversal spatial learning induced by fluoxetine treatment

The IntelliCage experiments were conducted to test the effect of chronic fluoxetine treatment on spatial learning, as depicted in Fig. 2a. Mice were implanted with transponders and were treated with fluoxetine-containing water for 2 weeks before the experiments. During the adaptation to freely accessible water bottles in the corners (FA), nose pokes (NPA), and drinking sessions (DSA), six mice were excluded because they could not learn the adaptation tasks [Control group (wild-type, 1; PV-TrkB hCKO, 2), fluoxetine-treated group (wild-type, 1; PV-TrkB hCKO, 2)]. In the acquisition phase of the patrolling task, the location of the open corner changed after each visit, and the water-deprived mice had to patrol the corners in a “clockwise” order to receive a water reward (Fig. 2a, left panel). The percentages of error ratios were calculated as the number of visits in the incorrect corner divided by the number of total visits and expressed as an average of each mouse of the 2-h session. The wild-type mice decreased the error ratio during sessions, and there was no effect of fluoxetine treatment in the acquisition phase (Acquisition (days), F (7, 208) = 8,520, P < 0.0001; Treatment, F (1, 208) = 1.021, P = 0.3133) (Fig. 2c–e). The PV TrkB hCKO mice also decreased the error ratio during sessions (Fig. 2f–h), but interestingly the fluoxetine treatment decreased the error ratio faster than in water-treated mice (Acquisition (days), F (7, 183) = 9.462, P < 0.0001; Treatment, F (1, 183) = 16.37, P < 0.0001) (Fig. 2f). The water-treated PV TrkB hCKO mice had significantly higher error ratios compared to wild-type mice treated with control water (Acquisition (days), F (7, 200) = 6.015, P < 0.0001; Genotype, F (1, 200) = 30.09, P < 0.0001) (Supplementary Fig. 1a). These results indicate that PV TrkB hCKO mice have lower spatial learning skills in acquisition compared to wild-type mice, but the fluoxetine treatment recovers them to a level comparable to wild-type mice.

Fig. 2: Chronic treatment with fluoxetine promotes spatial learning and depends on TrkB expression in PV interneurons.

a Scheme of the IntelliCage system during the chronic treatment with fluoxetine. All mice were implanted with transponders, and were treated with fluoxetine in water. Mice adapted gradually to the tasks in the IntelliCage (FA free adaptation, NPA nose poke adaptation, DAA drinking session adaptation), followed by the actual leaning tasks (Patrolling). b Scheme of the patrolling task. Error ratio in acquisition (c–h) and reversal phase (i–n) in wild-type (c–e, i–k) and PV-TrkB-hCKO mice (f–h, l–n) (n = 12–15 per group). c Wild-type mice decreased the error ratio during the acquisition days (two-way ANOVA, Acquisition (days), F (7,208) = 8.520, P < 0.0001), but there was no difference caused by the treatment (two-way ANOVA, treatment, F (1, 208) = 1.021, P = 0.3133). Significant differences were found in pairwise comparisons between the first and last sessions in control mice (pairwise t-test, t = 3,562, df = 14, P = 0.0031) (d) and fluoxetine water-treated mice (t = 6,626, df = 13, P < 0.0001) (e). Fluoxetine treatment reduced the error ratio during sessions in PV-TrkB-hCKO mice (treatment, F (1, 183) = 16.37, P < 0.0001; Acquisition (days), F (7, 183) = 9.462, P < 0.0001). There was a significant difference in the error ratio between the initial and the last sessions in control (pairwise t-test, t = 6,747, df = 10, P < 0.0001) (g) and fluoxetine-treated mice (t = 5,527, df = 12, P = 0.0001) (h). In the reversal phase, wild-type mice showed a significant effect in days (F (7, 208) = 4.212, P = 0.0002) and treatment (F (1, 208) = 6.794, P = 0.0098) (i). There were significant differences between the error ratio in the initial session and the last one in both control (pairwise t-test, t = 2,845, df = 14, P = 0.0130) (j) and fluoxetine-treated mice (t = 4,543, df = 12, P = 0.0007) (k). (l) In PV hTrkB cKO mice, fluoxetine treatment did not have an effect during sessions (two-way ANOVA, Treatment, F (1, 184) = 0.2608, P = 0.6102; Reversal (days), F (7, 184) = 3.550, P = 0.0013). There was a significant difference in error ratios between first and last session in control (pairwise t-test, t = 3,123, df = 11, P = 0.0097) (m) and fluoxetine treatment (t = 4,628, df = 12, P = 0.0006) (n). Error bars designate SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 (in post hoc test: f, I, l; pairwise t-test: d, e, g, h, j, k, m, n).

In the reversal phase, wild-type mice significantly reduced the error ratio during sessions in both control and fluoxetine-treated groups (Reversal (days), F (7, 208) = 4.212, P = 0.0002) (Fig. 2i–k), but the treatment with fluoxetine facilitated the decrease of the error ratio during sessions (Treatment, F (1, 208) = 6.794, P = 0.0098) (Fig. 2i). These results indicate that the fluoxetine treatment improves the reversal learning in wild-type mice. The PV hTrkB CKO mice also improved their performance during sessions (Reversal (days), F (7, 184) = 3.550, P = 0.0013) (Fig. 2l–n). These results suggest that TrkB expression in PV interneurons is important for the effect of fluoxetine on the reversal learning in spatial tasks.

Fluoxetine treatment potentiates hippocampal LTP through expression of TrkB in PV interneurons

In order to understand whether the improved behavioral flexibility after fluoxetine treatment reflects enhanced neural plasticity in the hippocampus, the main region involved in contextual fear and spatial memory [32], we recorded fEPSPs in acute hippocampal slices of wild-type and PV TrkB hCKO mice after chronic fluoxetine treatment (Fig. 3). As previously reported [8, 28] we observed a significant enhancement of LTP at 45 min after tetanic stimulation in wild-type mice treated with fluoxetine compared to mice treated with water (two-way ANOVA, treatment, wild-type, F (1, 414) = 50.20, P < 0.0001). There was, however, no effect of fluoxetine treatment on LTP in hPV-TrkB CKO mice (treatment, F (1, 40) = 0.2726, P = 0.6019) (Fig. 3). These results indicate that the chronic treatment with fluoxetine enhances the expression of LTP in the Shaffer collateral-CA1 synapses of the hippocampus, in a manner dependent on the expression of TrkB in PV-interneurons.

Fig. 3: Chronic treatment with fluoxetine enhances synaptic plasticity in Shaffer collateral-CA1 synapses of the hippocampus.

LTP induction after chronic treatment with fluoxetine. LTP was significantly enhanced 45 min after tetanic stimulation in wild-type mice treated with fluoxetine compared to control (two-way ANOVA, treatment, wild-type, F (1, 414) = 50.20, P < 0.0001) but not in PV-TrkB hCKO mice (treatment, F (1, 40) = 0.2726, P = 0.6019). Bars indicate mean ± SEM.

PV-specific transcriptomic analysis through TRAP

In order to investigate gene expression in PV interneurons after chronic treatment with fluoxetine, we conducted a TRAP analysis, a system that allows the precipitation mRNA bound to ribosomes, to investigate ongoing protein translation specifically in PV+ interneurons (Fig. 4a) (see Material and method, and Supplementary note). After chronic treatment with fluoxetine, the whole hippocampus of mice expressing EGFP-tagged L10a ribosomal subunits specifically in PV interneurons (Fig. 4b) was used for TRAP followed by next-generation sequencing. We found 879 genes that were differentially expressed after chronic treatment with fluoxetine (P < 0.05) (Fig. 4c and Supplementary Table 2), and these were further studied by pathway analysis using Fisher’s exact test for up- and downregulated genes separately. The chronic fluoxetine treatment significantly affected several of these pathways in the hippocampus (P < 0.1) (Supplementary Table 3), and differentially expressed genes from enriched pathways are shown in Fig. 4d. Interestingly, genes in glycosaminoglycan chondroitin sulfate (Fisher’s exact test, P = 0.0268) and heparan biosynthesis (Fisher’s exact test, P = 0.0377) pathways (B4galt7, Extl3, B3gat3 and Chst3) were significantly downregulated. These are associated with chondroitin sulfate proteoglycans, which are an integral part of PNNs [15]. Also, genes related to glycerolipid (Fisher’s exact test, P = 0.0384) and glycerophospholipid metabolism (Fisher’s exact test, P = 0.0013) were downregulated. These pathways are involved in the regulation of lipid composition of the cellular membrane, which is highly related to AD effects [20]. Furthermore, the fluoxetine treatment significantly changed the expression of genes in the GABAergic synapse pathway (Fisher’s exact test, P = 0.0863), including G Protein Alpha Inhibiting Activity Polypeptide 3 (Gnai3), G protein subunit gamma 4, 8, and 13 (Gng4, Gng8, and Gng13), which are coupled with GABA type B receptor, and mediate slow and prolonged inhibitory action [33]. Huntingtin-associated protein 1 (Hap1) directly interacts with GABA type A (GABAA) receptors and influences the recycling of the receptor by inhibiting its degradation [34]. Such modulation of the expression and localization of GABAA receptors are thought to be a plastic event resulting in the maintenance of the excitatory/inhibitory balance [35].

Fig. 4: TRAP analysis of PV interneuron after chronic treatment with fluoxetine.

a Ribosome-tagged transgenic mice were treated with fluoxetine or control water for 2 weeks, and their hippocampi were isolated and lysated. Ribosomes bound to mRNA were immunoprecipitated with beads coated with GFP-antibody and the mRNA was purified for cDNA synthesis followed by next-generation sequencing (NGS). b Immunohistochemistry analysis with anti-PV antibody. Parvalbumin is co-localized with GFP indicating that the cells expressing GFP-tag in ribosomes are PV cells. Scale bars, 50 µm. c Volcano plot showing log2 of fold change of all genes after fluoxetine treatment in x-axis and negative log10 of P value in y-axis. Downregulated genes that had significantly differential expression are marked in blue and upregulated in red. d Heatmap of significant genes and pathways detected by GO analysis. The expression of genes in a sample is scaled to values between −2 and 2, and these correlate with colors in the heatmap according to the panel on the right. GAGs C glycosaminoglycan biosynthesis chondroitin sulfate, GAGs H glycosaminoglycan biosynthesis heparan sulfate, GPL glycerophospholipid metabolism, GL glycerolipid metabolism.

Overall, our TRAP analysis provides insight into the observed phenomena of increased neural plasticity in PV+ interneurons, such as synaptic formation and turnover of PNNs through the regulation of gene expression after fluoxetine treatment.

Decreased intensity of PV and PNN after fluoxetine treatment depending on TrkB expression in PV+ interneurons

TRAP analysis showed decreased expression of genes related to the formation of PNN. In addition, it has been reported that PV configurations in the CA3 region of the hippocampus are dynamically regulated by experiences, such as environmental enrichment and fear conditioning [36]. We used immunohistochemistry to analyze the intensities of PV, and PNNs surrounding PV interneurons as a measure of their expression levels in the hippocampal CA3 region after chronic fluoxetine treatment (Fig. 5a). After fluoxetine treatment the proportion of low-intensity PV cells increased, and the high-intensity PV cells were reduced in wild-type mice, while there was no obvious difference in the proportions of PV intensity in PV-TrkB hCKO mice (Fig. 5b). In addition, the proportion of PV-positive cells among cells positive for PNN was significantly reduced after fluoxetine treatment in wild-type mice as shown previously [13], but not in PV TrkB hCKO mice (two-way ANOVA, Treatment, F (1, 36) = 1.757, P = 0.1934; Genotype, F (1, 36) = 1.203, P = 0.2801; Fisher’s LSD post hoc test, control vs Flx: wild-type, P = 0.0138; CKO, P = 0.7571) (Fig. 5c). Interestingly, when the intensity of PNNs were separately measured in lower and higher PV-expressing PV interneurons, the fluoxetine treatment significantly reduced the intensity of PNN only in high but not in low PV-expressing cells in wild-type mice (two-way ANOVA, interaction between PV intensity and Treatment, F (1, 208) = 4.785, P = 0.0298; PV intensity, F (1, 208) = 16.92, P < 0.0001; Treatment, F (1, 208) = 3.103, P = 0.0796; Fisher’s LSD post hoc test, control vs Flx: Low, P = 0.6365; High, P = 0.0281) (Fig. 5d). However, the treatments showed no effect on the PNN intensity in either low or high PV-expressing cells in PV-TrkB hCKO mice (two-way ANOVA, interaction, F (1, 108) = 1.326, P = 0.2520; PV intensity, F (1, 108) = 28.88, P < 0.0001; Treatment, F (1, 108) = 2.564, P = 0.1122; Fisher’s LSD post hoc test, control vs Flx: Low, P = 0.6722; High, P = 0.1075) (Fig. 5e). These results strongly suggest that chronic fluoxetine treatment shifts the configuration of PV interneurons toward lower PV and PNN expressing cell state through TrkB signaling. Taken together with the TRAP analysis, the decreased gene expressions of the extracellular matrix might be involved in the reduced PNN formation after chronic treatment with fluoxetine.

Fig. 5: Chronic treatment with fluoxetine enhances PV plasticity in the hippocampus.

a–e Image analysis on PV and PNN expression in the dorsal hippocampus of control and fluoxetine-treated wild-type and PV-TrkB hCKO mice. a Representative image of PV and PNN staining. Immunostaining with PV and PNN followed by intensity analysis on PV and PNN. SP stratum pyramidale, SO stratum oriens, SR stratum radiatum. Scale bar, 50 µm. b Intensity analysis of PV expression in PV interneurons. The ratio of high and intermediate-high PV was lower after fluoxetine treatment in wild-type mice, but this difference was not observed in PV-TrkB hCKO mice. c Fluoxetine-treated wild-type mice have significantly lower percentages of PV interneurons also expressing PNNs, but this effect is blunted in PV-TrkB hCKO mice (two-way ANOVA, Treatment, F (1, 36) = 1.757, P = 0.1934; Genotype, F (1, 36) = 1.203, P = 0.2801; Fisher’s LSD post hoc test, control vs Flx: wild-type, P = 0.0138; CKO, P = 0.7571). d, e PNN intensity analysis in cells separated by PV-intensity. Fluoxetine treatment reduces PNN intensities in high (intermediate-high and high) PV-expressing cells only in WT mice (two-way ANOVA, the interaction between PV intensity and Treatment, F (1, 208) = 4.785, P = 0.0298; PV intensity, F (1, 208) = 16.92, P < 0.0001; Treatment, F (1, 208) = 3.103, P = 0.0796; Fisher’s LSD post hoc test, control vs Flx: Low, P = 0.6365; High, P = 0.0281), but not in PV-TrkB hCKO mice (two-way ANOVA, interaction, F (1, 108) = 1.326, P = 0.2520; PV intensity, F (1, 108) = 28.88, P < 0.0001; Treatment, F (1, 108) = 2.564, P = 0.1122; Fisher’s LSD post hoc test, control vs Flx: Low, P = 0.6722; High, P = 0.1075). WT, control, low, number of cells (n) = 88; WT, Flx, low, n = 79; WT, control, high, n = 28; WT, control, high,17; CKO, control, low, n = 32; CKO, Flx, low, n = 48; CKO, control, high, n = 12; CKO, control, high, 20. Bars indicate mean + SEM. *P < 0.05.