Lithium Restores Inhibitory Function and Neuronal Excitability through GSK-3β Inhibition in a Bipolar Disorder-Associated Ank3 Variant Mouse Model

Abstract

Bipolar disorder (BD) is a prevalent psychiatric condition characterized by mood dysregulation, psychosocial impairment, and an increased risk of suicide. The gene ANK3 has been identified as a risk locus for BD through multiple genome-wide association studies (GWAS). However, the mechanisms by which ANK3 variants influence BD pathophysiology and treatment response remain unclear. ANK3 encodes ankyrin-G, a protein that organizes the axon initial segment (AIS) and nodes of Ranvier by scaffolding ion channels and cell adhesion molecules to the cytoskeleton. Recent studies show that ankyrin-G interacts with the GABA_A receptor-associated protein (GABARAP) to stabilize inhibitory synapses, potentially linking ANK3 variants to inhibitory (GABAergic) signaling deficits associated with BD. We previously demonstrated that the BD-associated variant, ANK3 p.W1989R, disrupts the ankyrin-G/GABARAP interaction, resulting in inhibitory deficits and cortical pyramidal neuron hyperexcitability in mice. In this study, we investigate how lithium, a common BD therapeutic, modulates neuronal excitability in this model. Our findings show that chronic lithium treatment selectively enhances presynaptic GABAergic neurotransmission, reduces neuronal hyperexcitability, and partially rescues AIS length, without altering the density of GABAergic synapses. We also show that the selective glycogen synthase kinase-3 beta (GSK-3β) inhibitor Tideglusib recapitulates the enhancement of presynaptic GABAergic signaling. These findings shed new light on how ANK3 variants may contribute to inhibitory deficits in BD and demonstrate that lithium treatment is able to restore these deficits, likely through GSK-3β inhibition. Furthermore, these findings highlight GSK-3β inhibition as a promising therapeutic strategy for treating BD and other neurological disorders affected by GABAergic dysfunction.

Figure 1.. Lithium treatment does not affect the density of GABAergic synapses on the soma or axon initial segment (AIS) of cortical pyramidal neurons in Ank3 p.W1989R mice.

(A) Representative images of GABAergic synapses on pyramidal neurons in layer II/III somatosensory cortex of wild type (WT), Ank3 p.W1989R (W1989R), and Ank3 p.W1989R mice following 21 days of lithium treatment (W1989R+Li). Coronal brain sections were immunostained with a presynaptic GABAergic marker vGAT (green) and total ankyrin-G (magenta). Scale bars: 10 μm. (B) Quantification of the total number of vGAT-positive clusters per soma above a set intensity threshold from WT (gray), Ank3 p.W1989R (WR) (red), and lithium treated Ank3 p.W1989R (WR+Li) (blue). (C) Quantification of AIS vGAT-positive clusters from WT (gray), WR (red), and WR+Li (blue). One-way ANOVA: ****P < 0.0001, ***P = 0.0004 and P = ns (WT: n=37; WR: n=29; WR+Li: n=34).

Figure 2.. Chronic lithium treatment increases IPSC frequency, while not affecting current amplitude, in Ank3 p.W1989R mice.

(A) Quantification of sIPSC and frequency (Hz) and (B) amplitude (pA) from cortical neurons from brain slices of wild type (WT) (gray), Ank3 p.W1989R (W1989R) (red), and Ank3 p.W1989R (W1989R) after chronic lithium treatment 19–21 days (blue). Kruskal-Wallis, Dunn’s multiple comparisons test ***P < 0.01 and P= ns (WT: n=16; W1989R: n = 15; W1989R/Day 19 + Li: n = 16; W1989R/Day 20 + Li: n = 17 and W1989R/Day 21 + Li n = 20). (C) Representative traces of sIPSC plots (A) and (B). (D) Quantification of mIPSC frequency (Hz) and amplitude (E) of WT (gray), W1989R (red), and W1989R after chronic lithium treatment 19–21 days (blue). Ordinary one-way ANOVA, Tukey’s multiple comparisons tests: ***P < 0.001, and P = ns. (WT: n = 9; W1989R: n = 8; W1989R/Day 19 + Li: n = 10; W1989R/Day 20 + Li: n = 11 and W1989R/Day 21 + Li n = 17). (F) Representative traces of mIPSC in (D) and (E).

Figure 3.. Chronic Tideglusib treatment increases sIPSC and mIPSC frequency while not affecting current amplitude in Ank3 p.W1989R mice.

(A) Quantification of sIPSC frequency (Hz) and (B) amplitude (pA) from cortical neurons from brain slices of Ank3 p.W1989R (W1989R) (red), and Ank3 p.W1989R (W1989R) with Tideglusib treatment (green). Unpaired t-test **P < 0.005 (W1989R + vehicle: n = 9; and W1989R + Tideglusib n = 9). (C) Representative traces of sIPSC in (A) and (B). (D) Quantification of mIPSC frequency (Hz) and (E) amplitude (pA) from cortical neurons from brain slices of Ank3 pW1989R (W1989R) (red), and Ank3 pW1989R (W1989R) chronic Tideglusib treatment (green). Unpaired t-test **P < 0.005 (W1989R + vehicle: n = 9; and W1989R + Tideglusib n = 9) (F) Representative traces of mIPSC plots (D) and (E).

Figure 4.. Pyramidal neuron firing frequency is restored in Ank3 p.W1989R mice following chronic lithium treatment.

(A) Representative traces of APs generated by current injection in wild type (WT) (gray), Ank3 p.W1989R (WR) (red), and Ank3 p.W1989R after lithium treatment (WR + Li) (blue) neurons. (B) Average firing rate (Hz) vs. current injection (pA) from evoked APs in cortical neurons from wild type (WT) (gray), Ank3 p.W1989R (W1989R) (red), and Ank3 p.W1989R after lithium treatment (W1989R + Li) (blue). Two-way ANOVA, Tukey’s multiple comparisons tests: WT vs. W1989R ***P < 0.001, WT vs. W1989R + Lithium P = ns, and W1989R vs. W1989R + Lithium ***P < 0.001. (C) Maximum firing frequency (Hz) graph of pyramidal neurons from WT (gray), Ank3 p.W1989R (WR) (red), and Ank3 p.W1989R after lithium treatment (WR + Li) (blue). Kruskal-Wallis, Dunn’s multiple comparisons tests: WT vs. W1989R ***P < 0.001, WT vs. W1989R + Lithium P = ns, W1989R + Lithium vs W1989R P < 0.05 (WT: n = 7; W1989R: n = 12; W1989R + Li: n = 15).

Figure 5.. Lithium treatment increases the AIS length in Ank3 p.W1989R mice.

(A) Representative images of pyramidal neuron AIS length from layer II/III somatosensory cortex of wild type (WT), Ank3 p.W1989R (WR), and Ank3 p.W1989R following lithium treatment (WR+Li). Immunostaining for total ankyrin-G (red) and NeuN (white). Scale bars: 10 μm. (B) Quantification of AIS length between WT (gray), WR (red), and WR+Li (blue). One-way ANOVA: ****P < 0.0001, ****P < 0.0001 and **P = 0. 0033 (WT: n=55; WR: n=48; WR+Li: n=48).