Microglia contribute to bipolar depression through Serinc2-dependent phospholipid synthesis

Significance

Although microglia are implicated in neuropsychiatric disorders, their role in BD remains unclear. In this study, using induced pluripotent stem cell-derived microglia-neuron cocultures from BDII patients, we identified reduced Serinc2 expression in microglia, leading to impaired synaptic pruning. This deficit was linked to disrupted synthesis of serine-related phospholipids in the microglial membrane, contributing to depression-like behaviors in mice. Moreover, Serinc2-dependent lipid alterations impaired CR3 membrane localization, thereby disrupting microglial responses to neuronal pruning signals.

Abstract

Although clinical research has revealed microglia-related inflammatory and immune responses in bipolar disorder (BD) patient brains, it remains unclear how microglia contribute to the pathogenesis of BD. Here, we demonstrated that Serinc2 is associated with susceptibility to BD and showed a reduced expression in BDII patient plasma, which correlated with the disease severity. Using induced pluripotent stem cell (iPSC) models of sporadic and familial BDII patients, we found that Serinc2 expression showed deficits in iPSC-derived microglia-like cells, resulting in decreased synaptic pruning. Further, combining the microglia-specific Serinc2-deficient mouse and iPSC-microglia models, we found that microglial Serinc2 deficits functioned through attenuating the synthesis of serine-related phospholipids in the plasma membrane, thus resulting in depression-like behavioral abnormalities in the animals. Finally, we showed that the Serinc2-dependent lipid deficits diminished microglial membrane CR3 formation to interrupted synaptic pruning signals from neurons. Therefore, our results indicated that Serinc2 deficits in microglia might contribute to the pathogenesis of BD.

Bipolar disorder (BD) is a complex mental disorder characterized by manic and depressive episodes, among which bipolar I disorder (BDI: manic episodes with or without depression) and bipolar II disorder (BDII: major depressive and hypomanic episodes) are the two main common types. Clinical studies have revealed abnormalities in microglia and related immune responses in BD patient brains (1–3). For instance, some postmortem research has reported that microglia in brain tissues of BD patients are immune inactivated (2); inconsistently, others observed microglia activation in the brain of suicidal BD patients (4). It remains poorly understood what type of microglia dysfunctions and how they contribute to clinical symptoms in different BD subtypes.

Microglia are the resident immune cells in the brain. They can engulf excitatory and inhibitory synapses to regulate neuronal structure and activity, thus playing a vital role in synaptic transmission and neuronal development (5). It has been widely accepted that phagocytosis by microglia is a main route for synaptic pruning or elimination (6, 7). Several neurodevelopmental disorders have been found to be associated with abnormal microglial elimination of synapses and neuronal development. For instance, autism spectrum disorders (ASD) are often correlated with down-regulated microglia pruning of synapses and up-regulated spine density (8), whereas schizophrenia (SCZ) patients are found to show enhanced microglial elimination of synapses and decreased spine density (9, 10). It has long been realized that membrane trafficking is substantially regulated by membrane lipid composition. For instance, the lipid composition of the synaptic active zone (AZ) membrane determines the efficiency of neurotransmitter release from presynaptic terminals (11). However, it remains unclear how the plasma membrane of microglia affects its phagocytosis of dysfunctional synapses and whether this process can possibly contribute to BD.

The gene of serine incorporator 2 (Serinc2) was highly expressed in microglia (the Human Protein Atlas; http://v13.proteinatlas.org). Serinc is a highly conserved protein family in eukaryotes containing five paralogous members with 9-11 transmembrane domains. The Serinc proteins function as a carrier to incorporate serine into membranes and facilitate the synthesis of membrane lipids (12). Serinc1 was highly expressed in macrophages and lymphocytes and was essential for the functions of immune cells (13). In humans, Serinc3-5 function as potent inhibitors of HIV-1, and restrict HIV infection by directly binding to HIV-1 virus particles (14). Serinc2 was highly enriched in the brain, particularly the hippocampus and cerebral cortex (15). Serinc2 has been suggested to be associated with alcohol addiction, a clinical complication of BD (16–18), and genes in the NKAIN1-Serinc2 genomic region are associated with numerous alcoholism-related genes (18). Recently, Serinc2 has been identified as a risk gene for BD in a large cohort of sporadic and familial BD patients (19). Moreover, clinical research revealed that patients with the Serinc2 variant showed an increased volume of white matter in the cerebellum (19). However, it remains unknown whether Serinc2 plays a direct role in BD and the molecular mechanisms possibly involved.

In the present study, using whole-genome sequencing (WGS) and genome-wide association studies technologies, we demonstrated that Serinc2 is associated with susceptibility to BD. Moreover, Serinc2 showed a reduced expression in the plasma of 60 BDII patients, which showed a positive correlation with the disease severity of the patients. Using an iPSC-based microglia-neuron coculture system, we found that Serinc2 expression was attenuated in microglia differentiated from the iPSCs of sporadic and familial BDII patients, which resulted in a decreased synaptic pruning in the iPSC-derived hippocampal neurons. Then, combining the microglia-specific Serinc2 knockout (KO) mouse and iPSC-microglia model, we found that Serinc2 deficits in microglia baffled the synthesis of serine-related phospholipids, phosphatidylserine (PS) and sphingomyelin (SM), in the microglial plasma membrane, which directly induced depression-like behavioral abnormalities in the animals. Finally, we showed that the reduction of membrane PS/SM decreased the assembly of complement receptor 3 (CR3) in the microglial plasma membrane, thus interrupting the reception of synaptic pruning signals of microglia from the neurons. Our results indicated that Serinc2 deficits in microglia might contribute to the pathogenesis of BD.

Results

Serinc2 and Bipolar Disorder Susceptibility.

Serinc2 acts as a carrier for serine incorporation into cell membranes and supports membrane lipid synthesis (Fig. 1A) (12). A Serinc2 exonic variant, rs2275434, was recently found to link to BD risk in Chinese familial samples (19). To validate this, we performed WGS on an independent Han Chinese cohort of 415 BD cases and 496 controls. We observed a higher frequency of the rs2275434 homozygous TT genotype in BD cases (8/415) versus controls (5/496), though the odds ratio was modest. We also analyzed rare missense variants (MAF < 0.01) within Serinc2, observing similar burdens in BD cases and controls (12 each). Notably, a rare missense mutation, rs34728687 in exon 9 (affecting the 8th transmembrane region of the encoded protein), was present in 3 BD cases but absent in controls. This result indicated that Serinc2 may contribute to BD susceptibility.

Fig. 1.

Association of Serinc2 with risk of mood disorders. (A) The genomic position and coding consequences of Serinc2 rare missense mutations observed in the BD cases and controls. Variants discovered in patients with BD are plotted above the gene, and those discovered in controls are plotted below. The rs34728687 variant, observed in three BD patients, was in the exon 9, corresponding to the 8th transmembrane region of Serinc2. (B) Serinc2 mRNA level in the plasma of sporadic BDII patients (SBD) and healthy controls (SHC). SHC, n = 41 subjects; SBD, n = 62 subjects. (C) Negative correlation between Serinc2 mRNA level and HAMD score in 62 sporadic BDII patients. r = −0.2576; P = 0.0432. (D and E) Correlation between Serinc2 mRNA level and CGI-1 (D) or SOFAS (E) score in 31 patients that have the scores. (F) Comparison of plasma Serinc2 mRNA expression in different subgroups of patients, including family disease history (with, n = 19; without, n = 43), drug treatment (with, n = 38; without, n = 24), episode of test time (depression, n = 32; mania/mixed, n = 20; remission, n = 9), gender [male, n = 20(HC)/13(BD); female, n = 21(HC)/49(BD)], and age [<30, n = 34(HC)/56(BD); >30, n = 7(HC)/6(BD)]. (B) Student’s t test; (F) Kruskal–Wallis test. *P < 0.05; **P < 0.001; error bars, s.e.m.

We then conducted qRT-PCR analysis in 62 sporadic BDII patients (sporadic disease, SBD) and 41 age-/gender-matched healthy controls (HC for sporadic patients; SHC) to assess Serinc2 expression in plasma. Compared to SHC, BDII patients showed significantly reduced Serinc2 levels (Fig. 1B). Expression levels were moderately negatively correlated with Hamilton Depression Scale (HAMD) scores, indicating higher depression severity (Fig. 1C). Among 31 patients with additional scores of Clinical Global Impressions (CGI-1) and Social and Occupational Functioning Assessment Scale (SOFAS), Serinc2 expression also showed a moderate negative correlation with CGI-1 scores and a moderate positive correlation with SOFAS scores (Fig. 1 D and E).

We further stratified patients by demographic and clinical features (Fig. 1F). Serinc2 expression was significantly reduced in patients experiencing hypomanic, depressive, or mixed episodes, but not in those in remission. Patients without a family history of BD showed a significant reduction, while those with a family history showed a nonsignificant trend. Both treated and untreated patients had similarly reduced Serinc2 levels, but only the treated group reached statistical significance. Gender- and age-based subgroup analyses revealed downward expression trends in both males and females, as well as in patients under 30, though none were statistically significant.

To assess brain Serinc2 expression, we analyzed postmortem forebrain RNA from four BDII patients and six controls (from SMRI). RNA-seq identified 718 differentially expressed genes (105 upregulated, 613 downregulated). SERINC2 showed significant down-regulation, which was confirmed by qRT-PCR (SI Appendix, Fig. S1 A–C). KEGG analysis revealed perturbations in serine metabolism and cell adhesion pathways (SI Appendix, Fig. S1D), which were potentially linked to SERINC2 downregulation and its relevance to BD pathology.

Together, our results indicated that Serinc2 might contribute to BD susceptibility through genetic variation and reduced expression.

BD Patient iPSC-Derived Microglia Show Serinc2 and Phagocytosis Defects.

To assess Serinc2 expression in the human brain, we analyzed single-cell RNA sequencing (scRNA-Seq) datasets from the Human Protein Atlas and CELL by GENE Discover. Meta-analysis of healthy human brain samples revealed that Serinc2 is highly expressed in microglia (Fig. 2 A and B). Similar expression patterns were observed in mouse brain scRNA-seq data, which was confirmed by qRT-PCR (SI Appendix, Fig. S1 E and F). These results indicated a potential functional role for Serinc2 in microglia.

Fig. 2.

Serinc2 deficits in iPSC-derived microglia-like cells. (A) scRNA-Seq data of SERINC2 from the Human Protein Atlas (http://v13.proteinatlas.org), based on the meta-analysis of the literature on different databases that include healthy human samples. (B) scRNA-Seq data of SERINC2 in the human brain, integrated and analyzed by the CELL by GENE Discover (https://cellxgene.cziscience.com). (C) Schematic showing differentiation of microglia and coculture with cultured mouse hippocampal neurons. (D) Efficiency of microglial differentiation. Left, expression of microglial markers TREM2 (green), P2RY12 (red), and Iba1 (cyan). (Scale bar, 20 μm.) Right, the ratio of Iba1⁺ cells/total cells (Right). n = 5 subjects. (E) Immunoblot analysis of Serinc2 protein in the microglia-like cells of sporadic (Left) and familial (Right) BDII patients. (F and G) Microglial Serinc2 protein (F) and mRNA (G) level of sporadic (Left) and familial (Right) patients. HC, n = 5 subjects; BDII, n = 5; BDII+Serinc2, n = 5. (H) Characterization of coculture system with neuronal marker Tuj1 (red) and microglial marker Iba1 (green). (Scale bar, 10 μm.) White arrowheads indicate microglial (green) and neuronal (red) cell bodies. (I) Sample immunostaining images showing colocalization of excitatory postsynaptic marker PSD95 (Left) or presynaptic marker vGLUT1 (Right) with the lysosome marker CD68 within the Iba1⁺ microglia-like cells of FBD/SBD patients. In each panel, Upper, microglia-like cells across neuronal dendrites (Scale bar, 10 μm); Middle, enlarged area in the Upper image (Scale bar, 5 μm); Lower, 3D-reconstruction and segmentation analysis. (J) Quantification of the number of PSD95/vGLUT1 puncta within the lysosome. n = 5 subjects. (K) Sample immunostaining images of the density of PSD95 (Upper) and vGLUT1 (Lower) puncta along MAP2-labeled dendrites in the coculture system. (Scale bar, 5 μm.) (L) Quantification of the density of PSD95/vGLUT1 puncta along dendrites. n = 5 subjects. (M and N) Sample images (M) and quantification (N) of localization of inhibitory postsynaptic gephyrin (Left) or presynaptic vGAT (Right) within CD68-labeled lysosome. Upper, (Scale bar, 10 μm.) Middle and Lower, (Scale bar, 5 μm.) (O and P) Sample images (O) and quantification (P) of the density of gephyrin/vGAT puncta along MAP2-labeled dendrites. (Scale bar, 5 μm.) For all experiments, n = 5 subjects (2 familial and 3 sporadic subjects). (D) Mann–Whitney test. (F and G) One-way ANOVA. (J, L, N, and P) Student’s t test. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

Previous evidence suggests microglial abnormalities in BD patients (2, 4). Our RNA-seq analysis of postmortem samples identified immune response pathways, such as cytokine–receptor interaction and antigen processing (SI Appendix, Fig. S1D). However, the role of microglia in BD remains unclear. To investigate this, we sought to examine the microglia derived from the iPSCs of BD patients. We reprogrammed the fibroblasts of 6 sporadic BD patients (3 BDII SBD, 3 SHC) and 4 familial individuals (2 BDII FBD, 2 FHC) carrying the Serinc2 rs2275434 variant into iPSCs using the Sendai virus method. Fluorescence analysis confirmed iPSC expression of pluripotency markers TRA1-60 and NANOG (SI Appendix, Fig. S2 A–C), alkaline phosphatase activity (SI Appendix, Fig. S2 D–F), and a normal karyotype (SI Appendix, Fig. S2 G–I). RT-PCR verified expression of pluripotency factors and the elimination of the Sendai virus and Mycoplasma (SI Appendix, Fig. S2 J and K). Additionally, differentiation into all three germ layers was confirmed (SI Appendix, Fig. S2L). We then differentiated these iPSCs into microglia-like cells (20) (Fig. 2C), which expressed microglial markers Iba1, P2RY12, and TREM2 (Fig. 2D and SI Appendix, Fig. S2 M and N). RNA-seq, immunoblot, and qRT-PCR analyses showed reduced Serinc2 protein and mRNA expression in microglia-like cells from SBD/FBD groups compared to SHC/FHC controls (Fig. 2 E–G and SI Appendix, Fig. S4 A–C), which were restored via lentiviral Serinc2 reintroduction. Moreover, analysis of a public transcriptomic dataset of mouse brain-xenografted human iPSC-derived microglial precursors confirmed that Serinc2 expression was independent of microglial differentiation quality (SI Appendix, Fig. S4 D and E), indicating that its downregulation in BDII microglia-like cells was not caused by incomplete or inefficient differentiation.

Serinc2 contributes to membrane lipid synthesis by incorporating serine (12), and serine-related lipids are essential for microglial phagocytosis and membrane trafficking (21, 22), suggesting a possible role of Serinc2 in synaptic pruning. To assess pruning function, we cocultured 39 d in vitro (DIV) iPSC-derived microglia-like cells with 14 DIV mouse hippocampal neurons (Fig. 2 C and H). We labeled the microglia with Iba1 antibody and evaluated phagocytosis by analyzing excitatory synaptic puncta (PSD95 and vGLUT1) within the CD68⁺ lysosomal area of Iba1⁺ cells (Fig. 2I). Compared to SHC/FHC controls, SBD/FBD microglia showed reduced phagocytosed PSD95/vGLUT1 puncta (Fig. 2J). Correspondingly, increased PSD95 and vGLUT1 puncta were observed along MAP2-labeled dendrites in the SBD/FBD group (Fig. 2 K and L), indicating impaired microglial pruning activity.

We further assessed inhibitory synapse phagocytosis by analyzing Gephyrin (postsynaptic) and vGAT (presynaptic) puncta (Fig. 2M). The SBD/FBD microglia showed reduced densities of these puncta within lysosomes compared to SHC/FHC (Fig. 2N), accompanied by increased Gephyrin and vGAT puncta along dendrites in cocultures (Fig. 2 O and P).

In addition, to confirm these findings, we performed a pHrodo-red synaptosome assay. The pHrodo (a pH-sensitive dye)-labeled mouse synaptosomes were incubated with iPSC-derived microglia, and fluorescence intensity was used to indicate phagocytosis. The SBD/FBD microglia exhibited lower pHrodo intensity than controls, indicating reduced synaptosome engulfment (SI Appendix, Fig. S3A).

Together, our results indicated phagocytosis deficits in BDII patient-derived microglia.

Role of Serinc2 in BDII iPSC-Microglia-Mediated Synaptic Pruning Deficits.

Given the reduced Serinc2 expression in BDII microglia-like cells (Fig. 2 E–G), we examined its impact on neuron–microglia interactions. We cocultured FBD/FHC microglia with human iPSC-derived hippocampal-like neurons of an extra HC subject (HC-0) (Fig. 3 A and B and SI Appendix, Fig. S2), which showed normal electrophysiological properties (Fig. 3C). Whole-cell patch-clamp recordings revealed increased frequencies of spontaneous excitatory and inhibitory postsynaptic currents (sEPSCs/sIPSCs) in neurons cocultured with FBD microglia compared to FHC; however, the sEPSC/sIPSC amplitudes were unchanged (Fig. 3 D–G and SI Appendix, Fig. S5 A and B). These results indicated elevated synaptic transmission due to impaired synaptic pruning. Notably, lentiviral reintroduction of Serinc2 into FBD microglia 7 d before coculturing (29 DIV) restored sEPSC/sIPSC frequencies of cocultured neurons.

Fig. 3.

Serinc2 deficits in BDII patient iPSC-microglial contributed to abnormal synaptic pruning. (A) Schematic showing differentiation of microglia and coculture with iPSC-derived DG neurons. (B) Characterization of coculture system with neuronal marker Tuj1(red) and microglial marker Iba1 (green). (Scale bar, 10 μm.) (C) Evoked action potentials (up) and Na+/K + currents (down) of DG neurons in the coculture system. (D) Sample traces of sEPSC recorded in HC-0 DG neurons cocultured with iPSC-microglia. (E) Summary of frequency of sEPSCs recorded in HC-0 DG neurons. Each column indicates one cell line of one subject. n = 14 cells for all subjects. Two-way ANOVA was performed, intragroup differences were not significant, and intergroup differences were indicated in the graph. (F) Sample traces of sIPSCs recorded in cocultured HC-0 DG neurons. (G) Summary of sIPSC frequency. Each column indicates one cell line of one subject. FHC, n = 16/14; FBD, n = 14/14; FBD+Serinc2, n = 12/14. Two-way ANOVA was performed and intergroup differences were indicated. (H) Sample immunostaining (Left) and 3D-reconstructed segmentation (Right) images showing pruning of excitatory PSD95/gephyrin puncta by microglia-like cells. (Scale bar, 4 μm.) (I) Quantification of phagocytosed PSD95/gephyrin puncta. Each data point represents one subject. In all groups, n = 5 subjects (2 familial and 3 sporadic subjects). The Kruskal–Wallis test was used for nonparametric analysis and Mann–Whitney test was used for multiple comparisons. (J) Sample immunostaining images showing PSD95/gephyrin puncta along MAP2-labeled dendrites. (Scale bar, 2 μm.) (K) Quantification of the density of PSD95/gephyrin puncta along MAP2-labeled dendrites. Each data point represents one subject. (E, G, and K) One-way ANOVA. (I) Kruskal–Wallis test. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

To assess synapse elimination of cocultured human iPSC-derived neurons by microglia and the role of Serinc2, we performed immunostaining to quantify excitatory (PSD95) and inhibitory (gephyrin) puncta within Iba1⁺ microglia-like cells (Fig. 3H). 3D reconstruction revealed reduced puncta density in SBD/FBD microglia compared to SHC/FHC groups (Fig. 3I and Movie S1), consistent with our observations in CD68 analysis in the mouse neuron coculture system. Notably, lentiviral re-expression of Serinc2 in patient microglia 7 d prior to coculture rescued these deficits, indicating that impaired synaptic pruning was likely due to Serinc2 deficiency.

In cocultured HC-0 neurons, we observed an increased density of dendritic PSD95 and gephyrin puncta in the FBD/SBD groups compared to FHC/SHC groups (Fig. 3 J and K), indicating reduced synaptic pruning. This increase was reversed by lentiviral Serinc2 re-expression in patient microglia-like cells.

Together, our results indicated that Serinc2 was probably crucial for synaptic pruning by BDII iPSC-derived microglia, and its deficiency impaired synapse elimination in human iPSC-derived neurons.

Performances of Mice with Serinc2 Deficiency in the Microglia.

We then investigated the effects of microglia-specific Serinc2 deficiency on mouse behaviors. Due to the lack of commercially available antibodies for adequate labeling of mouse Serinc2 protein, we generated a mouse strain with a 3xHA-tag conjugated to the endogenous Serinc2 protein (Serinc2^HA). Immunostaining analysis with an HA antibody confirmed Serinc2^HA enrichment in Iba1⁺ microglia in the hippocampus (Fig. 4 A and B), consistent with previous scRNA-seq and qRT-PCR findings. To create microglia-specific Serinc2 conditional knockout (cKO) mice, we cross-mated Serinc2 cKO mice—designed to delete exons 3–5—with Cx3cr1^CreERT2 mice (Serinc2 CcKO) (Fig. 4C and SI Appendix, Fig. S6A) (23). We also deleted exon 3–5 in the Serinc2^HA mice and validated Serinc2 deletion using immunoblotting (SI Appendix, Fig. S6 B and C). Further, qRT-PCR analysis confirmed loss of Serinc2 in cultured microglia from Serinc2 CcKO mice (SI Appendix, Fig. S6C), indicating successful microglia-specific removal of Serinc2.

Fig. 4.

Microglial phagocytosis defects of Serinc2 CcKO mice. (A) Colocalization of Iba1 (green) and Serinc2-conjugated HA (purple). White arrows indicate microglia expressing Serinc2^HA. (Scale bar, 10 μm.) (B) Immunoblot analysis of Serinc2^HA expression in microglia, neurons, and astrocytes of the Serinc2^HA mice. (C) Generation of Cx3cr1^CreERT2±/Serinc2^flox/flox (CcKO) mice. (D) Sample immunostaining images showing the morphology of microglia indicated by Iba1 (red). (Scale bar, 10 μm.) (E) Sholl analysis showing the complexity of the Cx3cr1⁺ microglia. (F and G) Summary of total process length (F). interaction number (G: Left), branch number (G: Middle), and segment number (G: Right). n = 17-18 cells from 3 mice. (H) Sample 3D reconstruction images showing colocalization of excitatory PSD95 (Upper) and vGLUT1(Lower) with lysosome-associated CD68 area within the Iba1⁺ microglia in the hippocampus of Serinc2 CcKO mice. (I and J) Quantification of the number of PSD95/vGLUT1/ gephyrin/vGAT puncta within CD68-labeled lysosome. n = 12. (I) Mann–Whitney test. (F, G, and J) Student’s t test was used. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

We next assessed the morphology of Serinc2 CcKO microglia using Iba1 immunostaining (Fig. 4D). Compared to the controls, Serinc2 CcKO microglia showed reduced numbers of intersections, branches, and segments and the filament length (Fig. 4 E–G), indicating morphological abnormalities in these cells.

To evaluate synaptic pruning, we analyzed excitatory and inhibitory synaptic puncta within lysosome-associated CD68 area within hippocampal Iba1⁺ microglia. Serinc2 CcKO mice showed reduced engulfment of excitatory (PSD95/vGLUT1) and inhibitory (gephyrin/vGAT) synaptic puncta (Fig. 4 H–J and SI Appendix, Fig. S9 A and B). Further, using a pHrodo-red bioparticle assay in multiple brain regions, including the hippocampus, cortex, and striatum, we observed significantly reduced synaptic pruning activity in Serinc2 CcKO microglia (SI Appendix, Fig. S9 C and D).

Together, our results indicated that Serinc2-deficient microglia exhibit both morphological and synaptic pruning deficits.

Neuronal and Behavioral Deficits in Mice with Serinc2 Deficiency in Microglia.

We then assessed the neuronal impact of impaired synaptic pruning in Serinc2 CcKO mice. Immunostaining analysis in hippocampal slices revealed increased densities of excitatory (PSD95/vGLUT1) and inhibitory (gephyrin/vGAT) synaptic puncta along MAP2-labeled dendrites compared to controls (Fig. 5 A and B). Golgi staining also revealed an elevated dendritic spine density in the CcKO mice (Fig. 5C). These increases were not due to altered synapse formation, as Serinc2-deficient neurons cultured without microglia or astrocytes showed normal synapse numbers (SI Appendix, Fig. S6D). Furthermore, whole-cell patch-clamp recordings in hippocampal slices confirmed increased frequencies of sEPSCs and sIPSCs in the CcKO neurons, with unchanged amplitudes (Fig. 5 D and E and SI Appendix, Fig. S6E). Hence, Serinc2-deficient microglia led to synaptic accumulation and enhanced neuronal activity due to impaired pruning.

Fig. 5.

Synaptic and behavioral deficits of Serinc2 CcKO mice. (A) Immunostaining and 3D reconstruction of excitatory postsynaptic PSD95 (Upper) and presynaptic vGLUT1 (Middle) puncta along MAP2-labeled dendrites in hippocampal slices of Serinc2 CcKO mice. Lower, quantification of PSD95/vGLUT1 puncta density along dendrites. (Scale bar, 5 μm.) (B) Immunostaining and 3D reconstruction of inhibitory gephyrin (Upper) and vGAT (Lower) puncta along dendrites. Lower, quantification of gephyrin/vGAT puncta density. (Scale bar, 5 μm.) (A and B) Ctrl, n = 8; CcKO, n = 9. (C) Sample Golgi staining images (Left) and quantification (Right) of spine density in the hippocampus of CcKO mice. (Scale bar, 5 μm.) Ctrl, n = 12; CcKO, n = 17. (D and E) Representative traces (D) and frequency (E) of sEPSCs (Right) and sIPSCs (Right) in mouse hippocampal slices. sEPSC: Ctrl, n = 16; CcKO, n = 15. sIPSC: n = 12. (F) Sucrose preference test of Serinc2 CcKO mice. Ctrl, n = 12; CcKO, n = 9. (G) Novelty-suppressed feeding test of mice. n = 7. (H and I) Index of sociability (H) and social novelty (I) of the three-chamber social test. Ctrl, n = 10; CcKO, n = 8. (B and F) Mann–Whitney test. All other experiments, Student’s t test. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

Next, we assessed the behavioral consequences of microglial Serinc2 deficiency. The sucrose preference test (SPT), which indicates the extent of anhedonia, revealed that the Serinc2 CcKO mice showed reduced sucrose preference (Fig. 5F), In the novelty-suppressed feeding test (NSFT), which indicates the anxiety-like behaviors of the mice and the antidepressant effects of a drug, the Serinc2 CcKO mice showed an increased feeding latency (Fig. 5G). In the three-chambered social interaction (TCS) test, the CcKO mice exhibited normal sociability but impaired social novelty (Fig. 5 H and I). Notably, these behavioral deficits were rescued by re-expressing Serinc2 in hippocampal microglia via LV-DIO-mSerinc2 lentivirus injection (SI Appendix, Fig. S6 F and G). We note that other behavioral domains, including locomotion, recognition, and anxiety/autism-like traits, remained unaffected in the CcKO mice (SI Appendix, Fig. S6 H–M).

Together, our results indicated that microglial Serinc2 deficiency leads to synaptic pruning impairments and behavioral abnormalities in mice.

Lipid Deficits in the Plasma Membrane of Serinc2-Deficient Microglia.

Previous studies have suggested that Serinc2 functions as a transporter that incorporates serine into membrane lipids, particularly serine-derived species such as phosphatidylserine (PS) and sphingomyelin (SM) (12). To assess lipid alterations, we first performed lipidomic analysis in the hippocampus of Serinc2 complete KO mice and found reductions in multiple lipid classes, including PS and SM (Fig. 6A). Due to limited cell numbers, we then used enzyme-linked immunosorbent assay (ELISA) to analyze membrane lipid changes in Serinc2 CcKO microglia and similarly observed decreased PS/SM levels (Fig. 6B). Notably, exogenous supplementation of PS/SM in the culture medium restored their membrane levels in Serinc2 CcKO microglia.

Fig. 6.

Behaviors and microglial synaptic pruning rescued by exogenous phosphatidylserine and sphingomyelin. (A) Heatmaps showing mass spectrometry (MS) analysis of lipid expression in WT and Serinc2 complete KO mouse brains. (B) ELISA analysis of membrane phosphatidylserine (PS; Left) and sphingomyelin (SM; Right) concentration in mouse microglia. P/S: PS/SM. Left, n = 6; Right, n = 3. (C) Schematic showing in vitro PS/SM rescue of microglia and transplantation into the hippocampus of CcKO mice without resident microglia. (D) Sample immunostaining images showing transplantation of microglia infected with Lenti-CMV-GFP into the hippocampus. Iba1-labeled microglia (red) were mostly GFP-labeled transplanted cells. (Scale bar, 20 μm.) (E) Sucrose preference of mice. (F) Time latency of novelty suppressed feeding test. Kruskal–Wallis test. (G) Index of sociability (Left) and social novelty (Right) of the three-chamber social test. n = 10. (E–G) n = 10. (H) Sample immunostaining (Left) and 3D reconstruction (Right) images of PSD95 along TUJ1-labeled dendrites in mouse hippocampal slices. (I) Quantification of PSD95/gephyrin puncta density along dendrites. (Scale bar, 4 μm.) n = 12. (J–L) Sample traces (J) and frequencies of sEPSCs (K) and sIPSCs (F and L) Kruskal–Wallis test. (B, E, G, and I–K) One-way ANOVA. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

Next, to test whether microglia contributed to the phenotypes of Serinc2 CcKO mice through PS/SM deficits, we transplanted microglia from WT mice and Serinc2 complete KO mice, and KO microglia pretreated with PS/SM into the hippocampal dentate gyrus (DG) of the Serinc2 CcKO mice following the depletion of resident microglia (Fig. 6C). Before transplantation, microglia were labeled with green fluorescence protein (GFP) via lentiviral infection using the Lenti-CMV-GFP vector. The recipient animals were then injected with AAV2/9-DIO-TaCaspase-3 and treated with tamoxifen to selectively induce apoptosis in Cre-expressing resident microglia at the injection site. The transplanted microglia, which lacked Cre expression, remained unaffected by this depletion strategy. We note that this AAV design might not precisely remove microglia due to the relatively low efficiency of AAV2/9 infection in microglia and the possible leaking expression of DIO in other types of cells (SI Appendix, Fig. S7). 7 d after transplantation, GFP⁺Iba1⁺ cells were detected in the DG (Fig. 6D). Fluorescence-activated cell sorting analysis confirmed that the GFP⁺ cells expressed microglial markers and could secrete cytokines (SI Appendix, Fig. S8 A and B). Moreover, after 7 d, PS/SM-pretreated KO microglia retained elevated PS/SM levels (SI Appendix, Fig. S8C).

We then performed behavioral analyses in the transplanted mice. The Serinc2 CcKO mice receiving KO microglia transplantation showed reduced sucrose preference, increased feeding latency, and impaired social novelty compared to the WT microglia transplantation group, were rescued by transplantation of PS/SM-pretreated KO microglia (Fig. 6 E–G). Further, immunostaining analysis revealed that the densities of PSD95/gephyrin puncta colocalized with Tuj1⁺ neuronal processes were increased Serinc2 CcKO mice transplanted with KO microglia compared to the WT microglia controls, which was restored by transplantation with PS/SM-pretreated KO microglia (Fig. 6 H and I and SI Appendix, Fig. S9E). In addition, electrophysiological recordings revealed an elevated sEPSC/sIPSC frequency but not amplitude in the CcKO mice transplanted with KO microglia, which was rescued by transplantation of PS/SM pretreated KO microglia (Fig. 6 J–L and SI Appendix, Fig. S9 F and G).

In addition, we also transplanted WT and Serinc2 KO microglia into the Serinc2 complete KO mice and performed behavioral assessments. Compared to the KO microglia transplantation group, the Serinc2 KO mice transplanted with WT microglia showed improved sucrose preference and sociability (SI Appendix, Fig. S10), supporting a critical role of Serinc2-dependent lipid regulation in microglial modulation of behaviors.

Together, our results indicated that microglial Serinc2 regulated synaptic function and behaviors of the mice, likely by modulating PS/SM synthesis in the microglial plasma membrane.

PS/SM Deficits in BDII iPSC-Microglia Induce Serinc2-Dependent Dysfunctions.

After identifying PS/SM as the downstream targets of Serinc2 in mice, we next examined whether, in BDII patient iPSC-derived microglia, PS/SM also functioned downstream of Serinc2. Using ELISA analysis, we observed that the FBD/SBD iPSC-derived microglia exhibited reduced levels of PS/SM in the plasma membrane compared to the FHC/SHC groups (Fig. 7A and SI Appendix, Fig. S11). Notably, both lentiviral re-expression of Serinc2 and external supplementation of PS/SM for 72 h could restore the membrane PS/SM levels (Fig. 7A and SI Appendix, Fig. S11).

Fig. 7.

Synaptic pruning defects are rescued by exogenous phosphatidylserine and sphingomyelin in BDII patient iPSC-microglia. (A) ELISA analysis of membrane PS (Left) and SM (Right) concentration in familial and sporadic iPSC-derived microglia-like cells treated with external P/S (PS/SM). (B and C) Immunostaining (Left Top), 3D reconstruction (Left Bottom), and quantification (Right) showing phagocytosis of PSD95 (B) and gephyrin (C) in FBD/SBD microglia. Kruskal–Wallis test. (Scale bar, 5 μm.) (D and E) Samples images (Left) and quantification (Right) of PSD95 (D) and gephyrin (E) puncta along MAP2-labeled dendrites. (Scale bar, 1 μm.) (F–H) Sample traces (F) and frequency of sEPSCs (G) and sIPSCs (H) in HC-0 DG neurons cocultured with microglia derived from FBD and SBD iPSCs. For all experiments, n = 5 subjects (2 familial and 3 sporadic subjects, each data point represents one subject). The Kruskal–Wallis test was used for nonparametric analysis and Mann–Whitney test was used for multiple comparisons. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

We then evaluated whether PS/SM supplementation could rescue synaptic pruning deficits in the coculture system of iPSC-derived microglia-like cells and HC-0 neurons. Patient iPSC-derived microglia were pretreated with PS/SM for 72 h prior to coculture, and the coculture was conducted without additional PS/SM supplementation. We analyzed the number of excitatory PSD95 and inhibitory gephyrin puncta within the Iba1⁺ microglial cytoplasm (Fig. 7 B and C). Consistent with previous results, segmentation analysis showed that the FBD/SBD microglia-like cells displayed reduced pruning of PSD95/gephyrin puncta compared to the FHC/SHC controls. However, this deficit was rescued by PS/SM pretreatment. We further analyzed the density of PSD95/gephyrin puncta along MAP2-labeled dendrites of the cocultured neurons. HC-0 neurons in the FBD/SBD coculture showed an increased density of both excitatory and inhibitory postsynaptic puncta, which was rescued by PS/SM pretreatment of the microglia (Fig. 7 D and E). Finally, we performed whole-cell patch clamp recordings and observed increased sEPSC/sIPSC frequencies, but not amplitudes, in HC-0 neurons cocultured with BDII patient microglia, which was also rescued by PS/SM pretreatment of the microglia (Fig. 7 F–H).

Together, our results indicated that, similar to the findings in mice, the Serinc2-dependent synaptic pruning deficits we observed in BDII patient iPSC-derived microglia were likely caused by insufficient PS/SM synthesis in the microglial plasma membrane.

PS/SM Deficits Attenuate the CR3 Assembly in the Plasma Membrane of Microglia.

As serine-related lipids have been implicated in microglial synaptic pruning and membrane protein trafficking, we hypothesized that Serinc2 may regulate pruning by modulating these processes. To test this, we first examined potential changes in synaptic pruning induced by Serinc2 deficiency. Previous studies have shown that microglia mediate phagocytosis through the classical complement cascade (24). Specifically, C1q, produced by both microglia and neurons, localizes to redundant synapses to initiate C1 complex formation. This complex cleaves C2 and C4 to generate the C3 convertase (C2aC4b), which activates C3 into C3b. C3b then binds to complement receptor 3 (CR3) on the microglial membrane—composed of CD11b and CD18 subunits (25)—to initiate synaptic phagocytosis (26).

To explore the downstream mechanisms of PS/SM, we first examined C1q expression at synapses in the hippocampus. Immunostaining with a C1q antibody, which effectively detected shRNA-mediated C1q knockdown in HEK293 cells (SI Appendix, Fig. S12A), revealed an increased density of C1q puncta along Tuj1-labeled neuronal processes in the Serinc2 CcKO mice compared to the controls (Fig. 8A). We next assessed the membrane localization of CD11b and CD18, the subunits of complement receptor 3 (CR3), in microglia. Immunoblot analysis of extracted membrane proteins showed no significant change in CD18 levels, but a clear reduction in membrane-bound CD11b in the Serinc2 CcKO microglia (Fig. 8 B and C). Total CD11b protein expression remained unchanged, indicating impaired membrane trafficking rather than transcriptional downregulation. To further verify this finding, we performed permeabilization-free immunostaining and confirmed a decrease in both the number and intensity of CD11b puncta on the microglial outer surface in the Serinc2 CcKO mice, while CD18 localization remained unaffected (Fig. 8 D and E and SI Appendix, Fig. S13 A and B). We then employed a Duolink in situ proximity ligation assay (PLA) to quantify CD11b–CD18 complex formation on the microglial membrane. The assay revealed significantly fewer and smaller PLA puncta in Serinc2 CcKO microglia, indicating impaired CR3 assembly due to membrane CD11b deficiency (Fig. 8 F and G). Notably, treatment with exogenous PS/SM for 72 h restored both CD11b localization and CR3 formation, indicating that the impaired CD11b membrane localization and CR3 formation was directly caused by PS/SM deficits in the microglial plasma membrane.

Fig. 8.

Phospholipid synthesis deficits attenuate CR3 assembly in the microglial plasma membrane. (A) Left, immunostaining of C1q (red) along Tuj1-labeled neuronal process (green) in hippocampal slices of Serinc2 CcKO mice. Right, quantification of C1q puncta number. (Scale bar, 2 μm.) Ctrl, n = 20; CcKO, n = 19. (B) Immunoblot (Upper) and quantitation (Lower) of CD18 protein expression in the plasma membrane of mouse microglia. n = 3 mice. (C) Immunoblot (Left) and quantification (Middle and Right) of the plasma membrane and total CD11b protein in CcKO mouse microglia. n = 3. (D and E) Permeabilization-free immunostaining images (D), puncta number (E: Left), and fluorescence intensity (E: Right) of plasma membrane CD11b. (Scale bar, 10 μm.) Ctrl, n = 17; CcKO, n = 22; CcKO+P/S (PS/SM), n = 18. (F) Duolink PLA images of mouse microglial CR3 protein: Upper, cells incubated with CD11b (Left) or CD18 (Right) antibody only; Lower, cells incubated with both CD11b and CD18 antibodies. (Scale bar, 10 μm.) (G) CR3 puncta area and number. Ctrl, n = 15; CcKO, n = 10; Ccko+PS/SM, n = 12. (H) Duolink PLA images of membrane CR3 in BDII iPSC-derived microglia-like cells. Upper, cells incubated with CD11b (Left) or CD18 (Right) antibody only. Lower, cells incubated with both CD11b and CD18 antibodies. (Scale bar, 10 μm.) (I) Normalized puncta area and number of CR3 in microglia derived from BDII iPSCs. n = 5 subjects. The Kruskal–Wallis test was used for nonparametric analysis and Mann–Whitney test was used for multiple comparisons. *P < 0.05, **P < 0.001. Bars, mean ± s.e.m.

Using an Annexin V staining approach, we confirmed that the applied PS to the microglia could be internalized to facilitate CR3 formation, and the reintroduction of Serinc2 in the microglia would not induce changes in the PS of neurons or synapses (SI Appendix, Fig. S14). In addition, we tested the expression of other components of the PS pathway, including Gas6, Stab1, Gpr34, and MER, in the Serinc2-deficient microglia and found that there were no significant differences between the BDII and HC groups (SI Appendix, Fig. S13C).

In addition, we examined CR3 assembly on the outer plasma membrane of BDII patient iPSC-derived microglia. We carried out Duolink PLA analysis and found that similar to our observations in the Serinc2 CcKO mouse microglia, both the number and area of CR3 puncta in the FBD/SBD groups showed a significant reduction compared to the FHC/SHC controls, which were rescued by 72-h PS/SM treatment (Fig. 8 H and I).

Together, our results indicated that PS/SM deficits in microglia impaired synaptic pruning in both Serinc2 CcKO mice and BDII patient iPSC models by disrupting CR3 assembly at the plasma membrane.

Discussion

In the present study, we combined iPSC-derived microglia-neuron coculture and mouse models to study the role of Serinc2 deficits in microglia in BD and the underlying mechanism. Our results indicated that the microglia with Serinc2 expression deficits might contribute to the occurrence of depression.

In recent years, accumulating research has applied the blood mRNA level of genes to indicate molecular and cellular changes in neurodegenerative and neuropsychiatric disorders (27, 28). Furthermore, some studies have provided examples that plasma mRNAs can be detected and used to evaluate pathological changes (29, 30). In our study, we employed plasma to analyze Serinc2 expression because plasma can exclude the interference from the blood cells and thus might better indicate the changes in brain mRNAs compared to the whole blood. However, we acknowledge that the plasma mRNA cannot fully represent mRNA changes in the brain and can only be used as indirect evidence.

In this study, we identified sex-dependent differences in Serinc2 expression in BD patients (Fig. 1G), indicating potential intrinsic, sex-specific regulation of microglial immune function. Estrogen has been shown to modulate microglial cytokine expression under various conditions (31), while testosterone suppresses glial activation (32). These findings align with the known neuroprotective effects of estrogens on microglial inflammation and phospholipid metabolism (33), as well as documented sex differences in microglial development and immune response (34). Thus, Serinc2 expression may reflect differential hormonal regulation of microglial function, with implications for phospholipid synthesis and synaptic pruning. Further studies are needed to dissect the hormonal regulation of Serinc2 and its role in microglia, potentially guiding sex-specific BD therapies.

It has been suggested that microglia regulate neuronal structures by engulfing redundant synapses, thereby playing a critical role in postnatal development (5, 35). Previous studies have reported microglial abnormalities in BD patients (2, 4). Moreover, growing evidence indicates that microglia mediate synaptic pruning not only during neurodevelopment but also in the adult brain. In the mature brain, microglia-mediated synaptic pruning contributes to the regulation of neuronal activity and the remodeling of neural circuits (36). As a result, microglia may contribute to regulating a variety of behaviors in adult animals, including learning and memory, sleep, anxiety-like behaviors, compulsive behaviors, and alcohol intake (37). In addition, evidence based on optogenetic and chemogenetic studies has verified that these behavioral consequences can be instant effects of manipulating microglial function in various brain regions of adult animals, such as the dorsomedial striatum, the medial prefrontal cortex, and the basolateral amygdala and central amygdala, but not a developmental outcome (37–39). Based on this, in the present study, we chose to conduct behavioral assessments in 2- to 3-mo-old Serinc2 CcKO mice. We observed that deficits in microglia-induced synaptic pruning abnormalities may play a direct role in the emergence of behavioral abnormalities and identified microglial phospholipid metabolism as a new mechanism involved in the regulation of synaptic pruning. Our results are in line with the established role of microglia in the adult brain. We employed three behavioral paradigms for our investigations, including the SPT, the NSFT, and the TCS test. The SPT and NSFT assess anhedonia by measuring the motivation for sucrose and food, respectively, and the TCS test evaluates the social interaction ability. These behaviors have been reported to be associated with hippocampal function (40, 41). Hence, these behavioral assays are appropriate for examining the consequences of microglial Serinc2 deletion, particularly in relation to hippocampal function. Nevertheless, although we did not examine immature mice in this study, we do not exclude the possibility that microglia may also contribute to synaptic pruning and behavioral phenotypes during early development stages. In future work, it would be interesting to trace the progression of synaptic pruning and behavioral alterations in Serinc2 CcKO mice throughout development, which could further elucidate the contribution of microglia to the origin of BD symptoms.

It has been suggested that neuronal PS acts as an “eat-me” signal actively presented by neurons to prompt microglia to perform synaptic pruning (42, 43). Upon receiving this initiating signal from neurons, microglia begin the pruning process through pathways including CR3 (26, 44, 45). In the present study, we found that microglial Serinc2 influenced the overall level of PS on the microglial membrane, which in turn affected the membrane expression of CR3. Hence, microglial Serinc2 functioned downstream of neuronal PS externalization. Meanwhile, we conducted an additional experiment demonstrating that supplementing microglia with PS/SM does not result in the indirect “transfer” of PS/SM to the membrane of cocultured neurons (SI Appendix, Fig. S15). Furthermore, we examined primary cultured hippocampal neurons from Serinc2 complete KO mice and found that synaptic density was unaffected by neuronal Serinc2 deletion (SI Appendix, Fig. S6D). Therefore, microglial Serinc2 contributed to synapse elimination by regulating microglial PS, independently of neuronal PS signaling.

C1q is typically expressed in the brain during development and aging (24, 46). However, in a glaucoma mouse model, its expression increases in adults, suggesting aberrant complement-mediated synaptic pruning under pathological conditions (24). Similarly, we found elevated neuronal C1q expression in adult Serinc2 CcKO mice, likely due to microglial homeostatic responses, supporting a CR3-dependent pruning mechanism downstream of Serinc2 deficiency. While our study focused on synaptic pruning, we do not exclude the possibility that Serinc2 may also regulate cytokine production. Future studies could explore whether microglial Serinc2 affects neuronal function and behavior via cytokine pathways.

While reduced synaptic structures are often linked to behavioral deficits, Serinc2 CcKO mice displayed increased synapses alongside behavioral abnormalities. This rise is unlikely due to enhanced synapse formation, as Serinc2-deficient neurons in culture showed normal synapse density (SI Appendix, Fig. S6D). Regardless of synapse number changes, altered excitation–inhibition balance or synaptic plasticity may similarly disrupt behavior. Alternatively, synapse changes may be brain region- or circuit-specific, with certain areas driving behavioral outcomes independent of direction. For example, both increased and decreased excitability in the ventral hippocampus can induce depression-like behaviors (47).

In our iPSC-microglia culture system, we did not observe the mature morphology typical of brain microglia. Differentiating iPSCs into mature microglia remains challenging. While some protocols report mature morphology (48–51), others failed to observe (52). Notably, a recent study found that microglia only exhibit mature features, including P2RY12 expression, after transplantation into the mouse brain (53), highlighting the importance of environmental context. Although microglial dysfunctions in our iPSC-based microglia and coculture systems were similar to our observations in Serinc2-deficient mice, mature morphology was still lacking. Future studies should adopt improved differentiation strategies to validate our observations in the iPSC-based model.

Serinc2 can act as a scaffold protein and aggregate a variety of serine synthetases, thereby promoting the synthesis of serine-associated cell membrane lipids (12). We observed that in microglia, Serinc2 deficiency reduced the contents of PS and SM. Nevertheless, we reiterate that our results could not exclude the possibility that other types of lipids may also contribute to the Serinc2 deficiency-induced phenotypes. As Serinc2 expression showed a substantial reduction in the plasma of the patients, it is likely that Serinc2 has abnormal expression in other types of cells, including neurons. As lipid metabolism plays an essential role in neurotransmission (11, 54), it is likely that Serinc2 deficit-induced neuronal dysfunction is involved in the pathophysiology of BD. Although we have not dissected the role and mechanism of Serinc2 in BD iPSC-derived neurons, it is necessary to investigate these points in the future. In the meanwhile, we need to be aware that although we employed the Serinc2 CcKO mouse model to investigate the effects of Serinc2-deficient microglia on neuronal functions and mouse behaviors, it is not a model of clinical BD. In addition, we would like to reiterate that we have limited our conclusion to suggest that the behavioral deficits observed in Serinc2 CcKO mice may be related to bipolar disorder, but not to unipolar depression or ASD, which can also show anhedonia and sociability deficits. This is because current clinical sequencing analyses support that Serinc2 is associated to the susceptibility of BD, while no such associations have been reported for unipolar depression or ASD. Nevertheless, if future clinical studies provide evidence linking Serinc2 to ASD and/or depression, the behavioral phenotypes we observed may also offer insights into the mechanisms underlying these disorders.

In the present study, there are several limitations and remaining questions. First of all, in the iPSC-derived microglia, we failed to observe a mature-state morphology similar to microglia in the brain, as we have pointed out above. Second, the plasma mRNA cannot fully represent mRNA changes in the brain and can only be used as indirect evidence. Third, due to the limited number of members within this family, we could only test two familial patients and three familial controls. Hence, we expanded our investigation to three sporadic patients and three controls and verified our findings in the familial patient cells. Fourth, we have not examined the phenotypes of female mice. All animals used in our study were male; however, there are likely sex differences, which might be related to symptomatic diversity between male and female BD patients. Fifth, we have not elucidated how neurons with Serinc2 deficits would contribute to the pathogenesis of BD. It is necessary to carry out future studies in BD patient iPSC-derived neurons or brain organoids. Sixth, it remains unclear about the possible role and mechanism of Serinc2 in the occurrence of bipolar I disorder (BDI). Finally, our results could not exclude the possible involvement of other membrane receptors, such as CD11a/c, in the defective phagocytosis in the patient iPSC-derived microglia. Future work should be carried out to address these critical questions.

Materials and Methods

Subjects.

Sporadic and familial BDII patients were diagnosed according to the DSM-IV. All human donor skin and blood samples described in this study were obtained from subjects of the BD family and sporadic BD patients who had given informed consent, and all the experimental procedures were approved by the Ethics Committee of Tsinghua University. All individuals in the BD family were of Han Chinese ethnicity. The characteristics of the subjects are detailed in Dataset S1.

All other methods and materials used in this study are described in detail in SI Appendix, Extended Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Movie S1.

3D reconstruction and segmentation analysis of phagocytosis. Immunostaining images of iPSC-derived microglia-like cells showing co-localization between the excitatory postsynaptic PSD95 puncta (red) and IBA1⁺ microglia (green). Scale bar, 4 μm.

Data, Materials, and Software Availability

RNA-seq data and lipidomics data have been deposited in NCBI BioProject (PRJNA1253883) (55) and MetaboLights (MTBLS12830) (56), respectively. Biologicalmaterials generated in this study are available from the corresponding author. All other data are included in the manuscript and/or supporting information.