Transcriptional and functional effects of lithium in bipolar disorder iPSC-derived cortical spheroids

Abstract

Lithium (Li) is recommended for long-term treatment of bipolar disorder (BD). However, its mechanism of action is still poorly understood. Induced pluripotent stem cell (iPSC)-derived brain organoids have emerged as a powerful tool for modeling BD-related disease mechanisms. We studied the effects of 1 mM Li treatment for 1 month in iPSC-derived human cortical spheroids (hCS) from 10 healthy controls (CTRL) and 11 BD patients (6 Li-responders, Li-R, and 5 Li non-treated, Li-N). At day 180 of differentiation, BD hCS showed smaller size, reduced proportion of neurons, decreased neuronal excitability and reduced neural network activity compared to CTRL hCS. Li rescued excitability of BD hCS neurons by exerting an opposite effect in the two diagnostic groups, increasing excitability in BD hCS and decreasing it in CTRL hCS. We identified 132 Li-associated differentially expressed genes (DEGs), which were overrepresented in sodium ion homeostasis and kidney-related pathways. Moreover, Li regulated secretion of pro-inflammatory cytokines and increased mitochondrial reserve capacity in BD hCS. Through long-term Li treatment of a human 3D brain model, this study partly elucidates the functional and transcriptional mechanisms underlying the clinical effects of Li, such as rescue of neuronal excitability and neuroprotection. Our results also underscore the substantial influence of treatment duration in Li studies. Lastly, this study illustrates the potential of patient iPSC-derived 3D brain models for precision medicine in psychiatry.

Fig. 1. Schematic representation of the experimental design for Li treatment of hCS.

Fibroblasts isolated from 10 controls (CTRL) and 11 bipolar disorder (BD) patients, including 5 lithium non-treated (Li-N) and 6 Li-responders (Li-R), were reprogrammed to induced pluripotent stem cells (iPSCs) and differentiated to cortical spheroids (hCS). All 21 hCS lines were treated with 1 mM lithium (Li) and water as vehicle control. Medium-to-long-term effects of drug exposure (1 month) were assessed through transcriptional profiling and functional analyses in 180-day-old hCS.

Fig. 2. Characterization of Li treated hCS.

a Representative images of CTRL and BD hCS from both Li-N and Li-R patients at days 30, 90 and 150 of differentiation. b Growth of CTRL and BD hCS size through differentiation show decreased volume at most time points of Li-N and Li-R hCS compared to CTRL hCS. Individual hCS volumes were estimated from their areas measured with ImageJ. c Li induced growth after 1 month treatment from day 150 until day 180. Both CTRL and Li-R, but not for Li-N hCS, present a significant increase in size. d Representative immunohistochemistry images of CTRL, Li-N and Li-R hCS with and without Li treatment. Fluorescence was measured for cytoplasmic MAP2 (e) and SV2A (f), by counting signal intensity with ImageJ. Results are presented in arbitrary units (a.u.). CTRL (N = 10), Li-N (N = 5), Li-R (N = 6). Data was analyzed by Mann–Whitney U test and Wilcoxon test for unpaired and paired group comparisons respectively. *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 3. Patch-clamp electrophysiology and calcium imaging of hCS neurons.

a Raw traces of whole-cell currents (holding potential Vh = −60 to −70 mV) representative for each condition. b Raw traces of action potentials (AP) representative for each condition. c Current–voltage (IV) curves for each group. Currents of individual IV curves were normalized to maximum peak values and then averaged. d Evoked whole-cell currents peak maximum values (I_max). Li significantly increases I_max in BD hCS (Δ(I_max) = 171.94 ± 76.80 pA, p = 0.01), particularly for Li-R hCS neurons (Δ(I_max) = 209.34 ± 102.42 pA, p = 0.03). e Threshold membrane voltages (V_t) of I_max, showing BD is less excitable than CTRL hCS neurons (Δ(V_t) = 5.32 ± 2.93 mV, p = 0.04). Li treatment increases excitability of BD hCS (via decreasing threshold voltage, Δ(V_t) = 5.14 ± 2.86 mV, p = 0.02), significant for Li-N hCS (Δ(V_t) = 11.84 ± 5.62 mV, p = 0.02). f Membrane resting potentials (RP). g Voltage–current (VI) curves for each group. Individual VI curves were averaged for correspondent group. h Representative trace images of spontaneous CaCl₂ transients in hCS regions of interest (ROIs) over a 20-min period. i Percentage of signaling ROIs after Li treatment. j Average amplitude of Ca²⁺ signal transients (%ΔF/F). I-shaped box charts represents 25–75% of range intervals and median horizontal brackets, overlapped with data; square symbol represent mean value. CTRL (N = 10), Li-N (N = 5), Li-R (N = 6). Data was analyzed by Mann–Whitney U test and Wilcoxon test for unpaired and paired group comparisons respectively. *P < 0.05, **P < 0.01.

Fig. 4. Transcriptomic characterization and spatiotemporal gene expression analysis of hCS.

a Principal component analysis (PCA) showing a strong donor effects and weaker clustering across diagnostic and response categories. b The same PCA plot displaying pairing information between untreated (circles) and treated (triangles) pairs. c Variance partition plot showing the proportion of gene expression variance attributed to different sources. Residuals constitute additional, unknown sources of variation not accounted for. d Estimation of cell type fractions in untreated CTRL and BD hCS. Significant differences detected in mature neurons and interneurons proportions. Data was analyzed by Mann–Whitney U test for group comparisons. e Overlap between in vitro gene expression (hCS) and in vivo human spatial brain expression profiles (BrainSpan). Brain regions nomenclature and abbreviations is the same as in [71]. f Overlap between in vitro gene expression and in vivo human temporal brain expression profiles. Strong correspondence seen between hCS and mid fetal stages (*p < 0.05, **p < 0.01, ANCOVA). CTRL (N = 10), Li-N (N = 5), Li-R (N = 6). OPCs oligodendrocyte precursor cells, RGs radial glia cells, NPCs neural precursor cells.

Fig. 5. Li-associated DEGs and GO enrichment analysis of hCS stratified by diagnostic and response status.

a Volcano plot of all DEGs identified for the untreated (Li-) BD vs. CTRL hCS. b Volcano plot of all DEGs identified for the treated (Li+) vs. untreated (Li−) hCS, independent of diagnosis. c Gene ontology (GO) pathway analysis (biological processes) of the 132 Li-associated DEGs identified in all samples independent of diagnosis. d, e Volcano plots of all Li-associated DEGs identified in CTRL and BD hCS, respectively. f GO pathway analysis of the Li-associated DEGs (74) that were unique to CTRL hCS (gray) and DEGs (8) that were unique to BD hCS (blue). g, h Volcano plots of all Li-associated DEGs in Li-N and Li-R hCS, respectively. i GO pathway analysis of the Li-associated DEGs [29] that were unique to Li-N hCS (dark blue) and DEGs [5] that were unique to Li-R hCS (light blue). CTRL (N = 10), Li-N (N = 5), Li-R (N = 6).

Fig. 6. Cytokine analysis and Seahorse bioenergetics of Li treated hCS.

a, b IL-1β, IL-6 and TNF-α cytokine levels in hCS at different time points. Li treatment effect in CTRL hCS (c), in BD hCS (d), and stratified in Li-N (e) and Li-R (f). Data is presented as mean ± SEM. g Seahorse oxygen consumption rate (OCR) kinetics graph. h Extracellular acidification rate (ECAR) kinetics graph showing glycolysis activity. i OCR parameters and ECAR basal glycolysis after Li treatment for each condition. j OCR and k ECAR kinetic graphs for Li treatment of BD hCS. l OCR parameters and basal glycolysis after Li treatment in BD hCS. All Seahorse experiments were run in quadruplicates, and values were normalized to total protein. Mitochondrial parameters were normalized to total protein content and data is presented as mean ± SEM. CTRL (N = 10), Li-N (N = 5), Li-R (N = 6). Mann–Whitney U test and Wilcoxon test were used for unpaired and paired group comparisons respectively. *p < 0.05, **p < 0.01.