Modulation of Neuronal Excitability and Plasticity by BHLHE41 Conveys Lithium Non-Responsiveness

Abstract

Many bipolar disorder (BD) patients are non-responsive to lithium. The mechanisms underlying lithium (non-)responsiveness are largely unknown. By using gene-set enrichment analysis methods, we found that core clock gene-sets are significantly associated with lithium response. Among the top hits was BHLHE41, a modulator of the molecular clock and homeostatic sleep. Since BHLHE41 and its paralog BHLHE40 are functionally redundant, we assessed chronic lithium response in double-knockout mutant mice (DKO). We demonstrated that DKOs are non-responsive to lithium's effect in various behavioral tasks. Cellular assays and patch clamp recordings revealed lowered excitability and reduced lithium-response in prefrontal cortical layer 2/3 DKO neurons and on hippocampal long-term potentiation. Single-cell RNA sequencing identified that lithium deregulated mitochondrial respiration, cation channel and postsynapse associated gene-sets specifically in upper layer excitatory neurons. Our findings show that lithium acts in a highly cell-specific way on neuronal metabolism and excitability and modulates synaptic plasticity depending on BHLHE40/41.

Figure 1.. BHLHE41 is a strong contributor to the association between lithium response and the core clock geneset.

(A) MAGMA analyses based on circadian gene-sets in psychiatric traits. Competitive test P-values in the X-axis are −log10 converted. Dashed blue line indicates the Bonferroni significance threshold (P<0.05/8 gene-sets × 9 traits). Dashed red line indicates the nominal (uncorrected) significance threshold P=0.05. (B) In gene-based tests three genes reach nominal significance (unadjusted p-value). (C-E) Local Manhattan plots of these genes (+/− 15 KB) for dichotomous lithium response show the BHLHE41 locus (C) has the strongest association (5 SNPs with r² > 0.8) compared to CSNK1E (D) and RORB (E). ADHD: Attention- Deficit/Hyperactivity Disorder; PTSD: Post-Traumatic Stress Disorder.

Figure 2.. PsyCoP profile reveals resistance of BHLHE40^−/−/41^−/− mice to lithium’s beneficial effect on memory and cognition.

(A) Schematic of the experiment design. Four groups, two from each genotype and sex, were tested after three weeks of pre-treatment via with either Vehicle (Ctrl) or lithium chloride solution (LiCl). Treatment was continued throughout the experiments. (B) Circadian 24h activity plotted as group means of 1-hour bins and ribbons show the standard error of the mean. The light and dark bars visualize the light and dark phase, respectively. Black line indicates time points used to assess peak activity differences revealing significant genotype (G* = p < 0.05) and lithium treatment (T *** p < 0.001) dependent effects, as indicated. (C-E) Boxplots showing lithium-affected phenotypes in DKO animals. Data are shown as box plots with whiskers extending to no more than 1.5-fold IQR. (C) Nocturnality (relative activity differences between L and D phases) revealed a non-significant tendency of Lithium non-response in DKOs. Lithium treatment resistance was found to be significant in DKOs in (D) recent contextual fear memory and (D) Serial Reversal Learning performance, as indicated. (F) Dimension reduction showing the canonical scores of the first two canonical components Can1 and Can2 (explaining 63.5% and 21.5% of the overall canonical correlation, respectively) depicts the aggregated Li-treatment dependent difference in WT but not DKO animals. Datapoints are overlayed with a 75% data ellipse. (G) A heatmap of the weights of single variables in the structure of the CDA for each term of the profile’s linear model. The variables are grouped in RDoC Top Level Domains as assigned a priori. Multivariate ANOVA results are shown on top. Univariate contrasts reaching significance are indicated with asterisks in the respective panel.* p < 0.05; ** p < 0.01; *** p < 0.001; n.s. not significant; p-values are FDR-adjusted and refer to Wilk’s lambda testing two-way ANOVA; simple effects were tested in a similar but univariate and unifactorial ANOVA procedure; WT: wildtype mice; DKO: BHLHE40^−/−/41^−/− mice; Ctrl: vehicle control; LiCl: lithium-treated; OF: Open Field Test; YM: Y-Maze Test; IC: IntelliCage; Act: circadian activity; PcL: Place Learning; SRL: Serial Reversal Learning; ScP: Sucrose Preference Test; PPI: Prepulse Inhibition Test; TST: Tail Suspension Test; FC: Fear Conditioning Test; G: genotype main effect; T: treatment main effect; GxT: interaction effect; simple T: simple effects; DKO-T: DKO simple treatment effect; WT-T: WT simple treatment effect; C: Cognitive Systems Domain; S: Sensorimotor Systems Domain; “+”: Positive Valence Systems Domain; “−“: Negative Valence Systems; A: Arousal and Regulatory Systems Domain

Figure 3:. Primary cortical neuron cultures and hippocampal networks in BHLHE40^−/−/41^−/− mice are resistant to lithium’s stimulation dampening effect.

(A) To measure neuronal network activity in primary cortical neuron cultures, a previously published Firefly luciferase reporter assay was used. This assay is based on a sensor for synapse-to-nucleus signaling named “enhanced synaptic activity response element (ESARE). (B) Ribbon plots showing the mean ESARE activity in response to bicuculline (BIC) stimulation after 24 hours. The ribbons indicate the standard error of the mean. (C) Quantification of the stimulation peak. Data are shown as box plots with whiskers extending to no more than 1.5-fold IQR. Asterisks indicate statistical significance of the interaction effect and simple effects for both genotypes tested in a two-way ANOVA. (D) The regions of interest for electrophysiology in acute slices were the anterior cingulate cortex (ACC) portion of the medial prefrontal cortex (mPFC) and the ventral Hippocampus (vHIP). (E) Shows the exemplary placement of the patch pipette in layer 2/3. Excitatory Postsynaptic Currents (EPSCs) were quantified in mPFC (F-I) and vHIP (J-M). (F, H) miniature EPSC (mEPSC) amplitude (F) and spontaneous EPSC (sEPSC) amplitude (H) in L2/3 mPFC excitatory neurons were significantly dampened in both WT and DKO, but the significant interaction of genotype and treatment (GxT) in both variables indicates a difference in lithium-induced dampening. (G, I) A significantly smaller log2-transformed foldchange (log2FC) reduction in LiCl-treated compared to Ctrl-treated neurons in both mEPSC (G) and sEPSC (I) amplitude supports this finding. (J) Average mEPSC amplitude was not significantly reduced in CA1 principal neurons. (K) However, the nominal reduction in amplitude is significantly stronger in WT compared to DKO. (L) The average sEPSC amplitude is reduced in both WT and DKO CA1 neurons in response to lithium treatment. (K) Normalized to the mean amplitude in Ctrl treated mice, we find a significantly stronger dampening effect in WT neurons compared to DKO. (N) Short-term (STP) and long-term potentiation (LTP) was assessed by high frequency stimulation (HFS) of the Schaffer collaterals and measuring fEPSP in the CA1 region of the hippocampus. Time point of stimulation is marked by an arrow. fEPSP amplitude was normalized to baseline. (O) Magnitude of STP, determined as maximal responses within first 5 min after HFS, is significantly lower in DKO mice. Treatment of lithium further reduced the STP both in WT and DKO mice, but the significant interaction indicates a difference in dampening amplitude. (P) When normalized to mean STP in Ctrl treated mice, log2FC reduction of STP amplitude is significantly larger in wildtypes compared to DKO mice. (Q) Magnitude of LTP, determined as responses between 50 and 60 min after HFS, is significantly lower in DKO mice. Treatment of lithium further reduced the LTP both in WT and DKO mice, but the significant interaction indicates a difference in dampening amplitude. (R) When normalized to the mean LTP in Ctrl treated mice, log2FC reduction of LTP amplitude is significantly larger in wildtypes compared to DKO mice. * p < 0.05; ** p < 0.01; *** p < 0.001; n.s. not significant; p-values refer to univariate two-way ANOVA with Type 2 sum-of-squares; simple effects were tested in a similar but unifactorial ANOVA procedure; WT: wildtype mice; DKO: BHLHE40^−/−/41^−/− mice; Ctrl: vehicle control; LiCl: lithium-treated; G: genotype main effect; T: treatment main effect; GxT: interaction effect; simple T: simple effects; DKO-T: DKO simple treatment effect; WT-T: WT simple treatment effect.

Figure 4.. Lithium modulates transcription of ATP-synthesis, post-synapse, and cation-channel associated gene sets most prominently in Layer 2/3 excitatory neurons of the mPFC as assessed by RNAseq with single cell resolution.

(A) The anterior cingulate cortex (ACC) portion of the medial prefrontal cortex (mPFC) was selected as region of interest. (B) UMAP dimension plot of lithium treated animals identifies 15 major cell populations, shown in distinct colors. Seven subclasses of excitatory cortical neurons, four subtypes of inhibitory neurons and four major glial cell types were identified (black, blue and grey coloured in the legend, respectively). (C) No differences in cell type clusters were detected between WT and DKO samples. (D) Barplot of significantly deregulated gene-sets (corrected q < 0.05) identified with GSEA against the gene ontology cellular component (GO-CC) collection shows most responses in layer 2/3 excitatory cells (49 up- and 27 down-regulated gene-sets) compared to all other cell types. (E) Condensed heatmap of all up- and down-regulated gene-sets (corrected q < 0.05) ordered for q-values shows the highly layer 2/3 specific response architecture with only sparsely overlapping gene sets enriched in other cell types. (F,G) Dotplot of the ten most significantly up- (F) and down-regulated (G) gene-sets ranked by the −log10 transformed q-value showing 6/10 upregulated gene-sets associated with mitochondrial ATP-synthesis and 3/10 with ribosome associated protein translation (F) and 5/10 and 3/10 down-regulated gene-sets associated with the post-synapse and cation-channel activities. L: cortical layer, EN: excitatory neurons, IT: intratelencephalic, PT: pyramidal tract, NP: near-projecting, CT: corticothalamic, IN: inhibitory neurons; Sst: Somatostatin, Pval: Parvalbumin, Vip: Vaso-Intestinal Peptide, OPC: oligodendrocyte precursor cell, NES: nominal enrichment score.

Figure 5.. Enrichment of Lithium modulated genes in synaptic structures in Layer 2/3 excitatory neurons from wildtype and DKOs.

(A) SynGO plot of enrichment of gene sets associated with synaptic structures in wt (left) and DKOs (right). Significantly associated genes are found in the top-layered pre- and postsynapse and branch our most prominently into post-synaptic membrane associated structures. (B) Protein network of lithium deregulated genes forms a highly interconnected cluster (Up- and downregulated gene products are depicted as blue and red circled nodes, respectively. Protein-protein-interactions (PPI) are plotted as greyish lines between nodes. Nodes added by the PPI network to increase connectivity are plotted as smaller sized grey circles. Only 4 genes are not integrated into the main network based on published PPIs (C-E) Synaptic components from the SynGO pre-synapse (C), post-synapse (D) and post-synaptic density (E) subclusters are highlighted in yellow and are enriched at the ‘synaptic core’ of the network.