Emergence of 5-HT5A signaling in parvalbumin neurons mediates delayed antidepressant action

Abstract

The behavioral response to antidepressants is closely associated with physiological changes in the function of neurons in the hippocampal dentate gyrus (DG). Parvalbumin interneurons are a major class of GABAergic neurons, essential for DG function, and are involved in the pathophysiology of several neuropsychiatric disorders. However, little is known about the role(s) of these neurons in major depressive disorder or in mediating the delayed behavioral response to antidepressants. Here we show, in mice, that hippocampal parvalbumin interneurons express functionally silent serotonin 5A receptors, which translocate to the cell membrane and become active upon chronic, but not acute, treatment with a selective serotonin reuptake inhibitor (SSRI). Activation of these serotonergic receptors in these neurons initiates a signaling cascade through which Gi-protein reduces cAMP levels and attenuates protein kinase A and protein phosphatase 2A activities. This results in increased phosphorylation and inhibition of Kv3.1β channels, and thereby reduces the firing of the parvalbumin neurons. Through the loss of this signaling pathway in these neurons, conditional deletion of the serotonin 5A receptor leads to the loss of the physiological and behavioral responses to chronic antidepressants.

Fig. 1

5-HT5A mediates inhibition in PV DG neurons after fluoxetine. a 5-HT receptor mRNA levels in hippocampal PV cells relative to the unbound input (n = 3 pairs each pulled from 3 mice). Bars represent means + SD. Paired t-test, *P < 0.05. Dashed line represents enrichment threshold. b mRNA levels of four genes in hippocampal PV-, CCK- and GAD2-expressing cells (n = 3, 3 per group). Bars represent means + SD. Numbers represent mRNA level as percent of gpadh. Paired t-test, *P < 0.05, **P < 0.01 vs. unbound. c Representative traces from PV DG neurons in WT and 5-HT5A cKO mice. d Membrane potential changes by 5-HT in Flx-treated (n = 10 neurons, 6 mice), and vehicle-treated (n = 5, 3) WT mice, or in slices preincubated with SB-669,551 (10 µM) (n = 6, 3). 5-HT had no effect on membrane potential in either Veh-treated (n = 3, 2) or Flx-treated (n = 3, 2) cKO mice. One-way ANOVA with post hoc Bonferroni tests, F (4, 22) = 6.29, ***P < 0.001. e TRAP analysis of 5-HT5A mRNA level in hippocampal PV cells after treatment with vehicle (n = 3, 9) or fluoxetine (n = 3, 9). Bars represent means as percent of Veh + SD. Unpaired t-test. f Representative images of 5-HT5A immunolabeling in SGZ PV cells from mice treated with vehicle or fluoxetine. Arrowheads indicate PV + cells; scale bar, 50 µm. g Quantitation of the percentage of SGZ PV cells co-expressing 5-HT5A. Bars represent mean percent from Veh (194 cells from 3 mice) or Flx (131, 3) + SD. Unpaired t-test. h Representative western blot image of 5-HT5A, PSD-95 and β-actin protein levels in the membrane-bound fractions of hippocampal lysates. i The 5-HT5A/β-actin ratio in the membrane bound fraction (n = 6 mice per treatment). Unpaired t-test, *P < 0.05

Fig. 2

5-HT5AR mediates inhibition in Kv current after chronic fluoxetine. a Action potential (AP) firing in DG PV neurons in Flx-treated WT mice in response to 400 pA injected current. b f–I plot showing that 5-HT decreases AP frequency in Flx-treated WT mice, an effect rescued by the consecutive application of SB-669,551 (10 µM) (n = 7, 5). RM one-way ANOVA, *P < 0.05, **P < 0.01. c 5-HT decreased the frequency of APs in Flx-treated WT mice (n = 7, 5), but not in Veh-treated WT (n = 8, 6) or in Veh-(n = 6, 4) or Flx-treated (n = 9, 6) cKO mice. Two-way ANOVA, F genotype X treatment [1,22] = 7.67, **P < 0.01. Each dot represents a neuron. d Representative traces of Kv potassium currents in PV neurons from Flx-treated WT mice. The currents were evoked with 10 mV potential steps from −70 to +50 mV before and after 5-HT or 5-HT with SB-669,551 (10 µM). e I–V plot showing that 5-HT decreased the amplitude of Kv currents in Flx-treated WT mice, an effect partially rescued by SB-669,551 (10 µM) (n = 8, 4). RM One-way ANOVA, *P < 0.05. f 5-HT decreased the amplitude of the Kv current in Flx-treated WT mice (n = 13, 6), but not in Veh-treated WT (n = 5, 3), Veh- treated cKO (n = 4, 3) or Flx-treated cKO mice (n = 7, 7). Two-way ANOVA, F genotype X treatment [1,25] = 4.32, *P < 0.05. Each dot represents a neuron. g Representative images of pKv immunolabeling in SGZ PV cells from WT and cKO mice treated with Veh or Flx. Arrowheads indicate PV+ cells; scale bar, 50 µm. h Dot plot analysis of the percentage of SGZ PV cells coexpressing pKv (Veh:147 cells from 6 WT mice and 169 cells from 7 cKO mice; Flx:170 cells from 7 WT mice and 194 cells from 7 cKO mice). Two-way ANOVA, F Drug [1,23] = 5.30, P = 0.03, *P < 0.05 by post hoc Bonferroni

Fig. 3

5-HT5A signaling in PV neurons. a Bar graph summary of cAMP level in N2A cells transiently transfected with mouse 5-HT5A. The cells were incubated with the indicated amounts of forskolin and 5-HT. One representative experiment out of three is shown. Bars represent means of cAMP level (n = 4 wells per group) as percentage of baseline + SD. One Way ANOVA, *P < 0.05 vs. baseline. b Representative western blot images (top) and bar graph summary (bottom) depicting the levels of Kv3.1β and Ser503 pKv3.1β in transfected HEK293 cells treated as indicated (n = 4 wells per condition). Bars represent mean densities of the pKv normalized to that of the total Kv. Number and arrowhead indicate protein size in kDa. One-way ANOVA. *P < 0.05, **P < 0.01. OA okadaic acid, PMA Phorbol 12-myristate 13-acetate. c Representative traces of Kv potassium currents evoked with 10 mV potential steps from −70 to +50 mV in PV DG neurons from Veh- and Flx-treated WT mice. d PMA (200 nM) decreased the amplitude of KV currents in Flx-treated mice (n = 4, 3) but not Veh-treated WT mice (n = 3, 3). RM one-way ANOVA, *P < 0.05. e Model for serotonergic signaling in PV hippocampal cells. Left: initial exposure to SSRIs does not alter cAMP levels and PKA activity. PP2A activity remains high and Kv3.1β is dephosphorylated. Right: chronic SSRIs increases 5-HT5A surface levels. 5-HT5AR activation reduces cAMP levels and diminishes PKA and PP2A activities to reduce Ser 503 Kv3.1β dephosphorylation

Fig. 4

5-HT5AR mediates behavioral response to chronic SSRIs. a Immobility time in the forced swim test (FST) after drinking saccharine alone (Veh) or fluoxetine/saccharine mixture (Flx) for 18 days in WT or 5-HT5A cKO. Two-way ANOVA, F [1,44] = 4.58, P = 0.038. b Immobility time in the tail suspension test (TST) after Veh or Flx (for 18 days) in WT or cKO. Two-way ANOVA, F [1,42] = 13.83, P = 0.006. c Latency to approach food pellet in the novelty suppressed feeding (NSF) after Veh or Flx (for 18 days) in WT or 5- cKO. Two-way ANOVA, F [1,41] = 18.58, P < 0.0001. d Immobility time in FST 15 min after a single saline (Sal) or fluoxetine (Flx) intraperitoneal injection in WT or cKO. Two-way ANOVA, F genotype X treatment [1,34] = 3.99, P = 0.07; F treatment [1,34] = 37.27, P < 0.0001. e Immobility time in TST 15 min after a single injection of Sal or Flx in WT or cKO. Two-way ANOVA, F genotype X treatment [1,24] = 0.30, P = 0.590; F treatment [1,24] = 17.47, P = 0.0003; F genotype [1,24] = 17.93, P = 0.0003. *P < 0.05, **P < 0.01 ***P < 0.001 by post hoc Bonferroni. f Study design of viral mediated 5-HT5A gene delivery to PV-Cre and 5-HT5A cKO mice. g GFP-positive cells in the SGZ and hilus but not in the CA1 or outside the hippocampus after AAV.Flex-GFP injection to the DG. Scale bar: 100 μm. h GFP in a SGZ PV+ neuron. Scale bar: 50 μm. i, j Colocalization of 5-HT5A in PV neurons after injection of AAV.Flex-GFP (i) or AAV.Flex-5-HT5A (j) to 5-HT5A cKO. Numbers represent the fraction of SGZ PV+ cells immunopositive for 5-HT5A. Scale bars: 50 μm. k Immobility time in the FST after 18-day fluoxetine treatment in PV-Cre (control) injected with AAV.Flex-GFP or AAV.Flex-5-HT5A or in cKO. Two-way ANOVA, F genotype X AAV [1,42] = 4.54, P = 0.039. l Immobility time in TST. Two-way ANOVA, F genotype X AAV [1,44] = 2.37, P = 0.138; F genotype [1,44] = 60.62, P < 0.0001; F AAV [1,44] = 12, P = 0.0012. m Latency in NSF. Two-way ANOVA, F genotype X AAV [1,33] = 5.75, P = 0.022. *P < 0.05, **P < 0.01 by post hoc multiple t-tests with FDR

Fig. 5

Activating cAMP signaling in PV DG cells attenuates the response to chronic fluoxetine. a Chemogenetic manipulation of PV DG cells consisted of bilateral injection of 1 μl rAAV2/hsyn-DIO- expressing one of three constructs: Gs-DREADD, Gi-DREADD or mCherry. b Immunofluorescence image depicting the expression of mCherry following Gs-DREADD delivery. Scale bar: 200 μm. c Immunolabeling confirmed that all mCherry immunopositive cells co-expressed PV. Scale bar: 50 μm. d Representative traces of patch-clamp recordings in hippocampal slices from mice in which PV neurons were infected with Gs-DREADD or Gi-DREADD. The right panel shows the hyperpolarization/depolarization of the membrane potential induced by the application of 2 µM CNO on the PV neurons infected with Gi-DREADD (−6.13 ± 2.4 mV n = 4, 3) or Gs-DREADD (9.03 ± 4.4, n = 3, 3). e Fluoxetine was administered for 18 days in the drinking water after which a single dose of CNO (4 mg/kg, intraperitoneally) was injected 30 min before the test. f Immobility time in the FST in PV-Cre mice injected with mCherry (n = 10 mice), Gi-DREADD (n = 10 mice) or Gs-DREADD (n = 6 mice). One-way ANOVA F [2,24] = 6.22, P = 0.007. *P = 0.023 by post hoc Dunnet