Risperidone induces osteoporosis and neuropsychiatric treatment resistance via SMPD1-lysosome-mediated ferroptosis: dual rescue by active vitamin D analog ED-71

Graphical abstract

Risperidone induces osteoporosis and neuropsychiatric treatment resistance through SMPD1-lysosome-mediated ferroptosis, both of which are dually rescued by the active vitamin D analog ED-71.

Highlights

• •In schizophrenia, RIS exacerbates bone loss, inhibits osteogenesis, partially improves psychosis, reduces hippocampal GABA.
• •RIS directly damages osteoblasts/hippocampal GABAergic neurons via abnormal "SMPD1-lysosome axis", inducing ferroptosis.
• •RIS impairs lysosomal function, blocks autophagic flux; disrupting lysosomal/mitochondrial iron levels, inducing ferroptosis.
• •ED-71 blocks RIS-SMPD1 binding, restores SMPD1 activity, reduces lysosomal phospholipids/dysfunction, reverses ferroptosis.
• •ED-71 reverses RIS-induced bone loss, promotes osteogenesis, synergizes with RIS to combat psychosis, restores GABA levels.

Abstract

Introduction

Risperidone (RIS), a second-generation antipsychotic for schizophrenia (SZ), is linked to osteoporosis and suboptimal symptom resolution. The shared cellular mechanisms underlying these cross-tissue toxicities (bone and hippocampus) remain undefined, hindering therapeutic advancement.

Objectives

To determine if RIS induces toxicity in osteoblasts and hippocampal GABAergic neurons via the acid sphingomyelinase (SMPD1)-lysosome axis, and to evaluate the therapeutic potential of eldecalcitol (ED-71), an active vitamin D analog, in mitigating these effects.

Methods

Dizocilpine-induced SZ mice were used to evaluate bone loss and psychiatric symptoms. Primary osteoblasts induced from bone marrow mesenchymal stem cells (BMSCs) and MC3T3-E1 osteoblasts, as well as primary hippocampal neurons and HT22 hippocampal neurons, were treated with RIS to evaluate ferroptosis, lysosomal dysfunction, and SMPD1 activity. Key techniques included micro-CT, histomorphometric staining, behavioral tests, enzyme-linked immunosorbent assay, RNA sequencing, targeted lipidomics analysis, molecular dynamics simulations, and isothermal titration calorimetry assays. Virtual drug screening was used to identify potential RIS-SMPD1 interaction antagonists, with the identified candidate ED-71 further validated thereafter.

Results

RIS targeted lysosomes, causing dysfunction and membrane permeabilization, which drove ferroptosis in osteoblasts and hippocampal GABAergic neurons. Mechanistically, RIS bound to SMPD1 at ARG294, inhibiting its activity and triggering lysosomal phospholipid accumulation and ferroptosis. ED-71 disrupted RIS-SMPD1 interactions, restored lysosomal integrity, mitigated hyperprolactinemia/sympathetic overactivation, and synergized with RIS to enhance antipsychotic efficacy and prevent osteoporosis in SZ models.

Conclusion

This study identifies SMPD1 activation as a therapeutic target to counteract RIS-induced ferroptosis in bone and hippocampus. The dual-action mechanism of ED-71 provides a novel strategy for SZ intervention, simultaneously addressing psychiatric symptoms and osteoporotic complications.

Introduction

Schizophrenia (SZ), a disabling neuropsychiatric disorder, is characterized by neurotransmitter network dysregulation originating from functional decline of hippocampal GABAergic neurons (HipGABA-Ns) [[1], [2], [3]]. Second-generation antipsychotics like risperidone (RIS) alleviate symptoms through dopamine (DA)/serotonin (5-HT) receptor antagonism [4,5], yet their clinical utility is constrained by two unresolved challenges: biphasic effects may involve HipGABA-Ns, ranging from transient neuroprotection to irreversible damage, with controversial subcellular mechanisms [6,7], treatment-associated bone metabolism disorders, manifesting as osteoblast inhibition and osteoclast hyperactivity [[8], [9], [10]]. While RIS-related osteoporosis (OP) traditionally attributed to hyperprolactinemia and sympathoadrenal overactivation[10], however, the shared cell biological basis for cross-system toxicity remains undefined.

The chemical architecture of RIS, characterized as a cationic amphiphilic drug (CAD) [11]—featuring a protonatable piperidine amino group (–NH₃⁺) and dual-aromatic-ring hydrophobic scaffold—may confer unique membrane interaction properties: cationic charge mediates electrostatic binding to negatively charged membranes, while the hydrophobic scaffold facilitates lipid bilayer penetration and enzyme hydrophobic pocket engagement [12]. Emerging evidence suggests lysosomes, as acidic metabolic hubs, serve as critical subcellular sites for RIS accumulation [13,14]. Acid sphingomyelinase (ASM/SMPD1), a key enzyme in lysosomal sphingomyelin (SM) catabolism, harbors a hydrophobic pocket prone to binding lipophilic ligands [15]. Studies show that certain lysosomotropic drugs inhibit SMPD1 activity, causing lysosomal phospholipid overload [16]. Thus we hypothesize that RIS aberrantly binds to SMPD1 via this pocket, suppressing enzymatic function and triggering cross-tissue toxic cascades.

Central to this hypothesis is the shared vulnerability of two functionally specialized cell types: hippocampal neurons, reliant on lysosomes for synaptic vesicle homeostasis, and osteoblasts, dependent on lysosomal regulation of matrix vesicle secretion [17,18]. Phospholipid overload-induced autophagy-lysosome dysfunction causing lysosomal hypertrophy and lysosomal membrane permeabilization (LMP) may disrupt their role as iron reservoirs, releasing iron to exacerbate Fenton reactions. Notably, recent studies identify lysosomal iron dyshomeostasis as a key trigger of ferroptosis—a non-apoptotic cell death pathway [[19], [20]]. Consequently, whether RIS-related neural and skeletal injuries converge on this common toxic pathway warrants investigation.

To address this mechanistic gap, we employed virtual drug screening to prioritize candidates with dual-target potential. Eldecalcitol (ED-71), a novel active vitamin D analog [21], features a lipophilic structure, possibly enabling blood–brain barrier penetration [22], alongside established osteoprotective effects [[23], [24], [25], [26], [27], [28]] and emerging neurosteroid regulatory functions [29,30]. Building on vitamin D’s role in prolactin secretion and sympathoadrenal tone regulation [31,32], we postulate that ED-71 exerts analogous neuroendocrine effects while displacing RIS from the SMPD1 binding pocket, restoring the enzyme activity and interrupting the lysosome-ferroptosis cascade. This dual-action mechanism may simultaneously improve psychiatric symptom control and skeletal safety.

The study aims to validate the hypothesis that RIS induces cross-tissue toxicity by targeting the SMPD1-lysosome-ferroptosis axis, and that ED-71′s dual mechanisms (neuroendocrine modulation and direct pathway intervention) can specifically disrupt this shared pathway. Findings will provide a unified mechanistic explanation for antipsychotic-related neuro-skeletal comorbidities and establish the lysosomal membrane protein-iron metabolism axis as a trans-tissue therapeutic target.

Materials and methods

Establishment of SZ animal model

The specific drugs used for modeling and a detailed flowchart of the entire procedure are illustrated in figure S1A-B. In this study, 8-week-old male and female C57BL/6J mice purchased from Jinan Pengyue Laboratory Animal Breeding Co. Ltd. (Jinan, China) were employed; each gender was divided into 4 experimental groups with 10 mice per group, namely the control (CON) group, SZ group, RIS-monotherapy group, and ED-71 combined with RIS treatment group. The CON group was subjected to daily intraperitoneal injection of phosphate-buffered saline and intragastric administration of glycerin. For the remaining three groups, a mouse model of SZ was induced by consecutive 14-day intraperitoneal injection of dizocilpine (MK-801) at a dosage of 1 mg/kg/d [33]. Subsequently, MK-801 was continuously administered in the same fashion until the end of the modeling process to sustain the manifestation of mental symptoms. In the RIS-monotherapy group, RIS was intraperitoneally injected at a dose of 1 mg/kg/d for 8 weeks [34], along with intragastric administration of glycerin. In the ED-71 combined with RIS treatment group, RIS was intraperitoneally injected at a dose of 1 mg/kg/d, and ED-71 was intragastrically administered at a dose of 50 ng/kg/d [35], both for a duration of 8 weeks. The ED-71 dose in this study is literature-supported for efficacy and safety [36]; converted via FDA/EMA-recommended body surface area method, its Human Equivalent Dose (HED) is 0.29 μg/day—highly consistent with the clinically validated 0.25–1.0 μg/day safe range. Before the completion of the modeling process, the open field test (OFT), the three-chamber social interaction test, and the Morris water maze (MWM) test were carried out to delineate the disparities in psychiatric symptoms among the different groups.

Cell culture

The mouse osteoblast cell line, MC3T3-E1 cells (Shanghai Cell Center), was cultivated in α-MEM. Following the confirmation that the mouse hippocampal neuron cell line, HT22 cells (Shanghai Cell Center), could recapitulate the key characteristics of GABAergic neurons in vitro to a certain extent, these cells were cultured in high-glucose DMEM. The isolation and culture methods for primary mouse bone marrow mesenchymal stem cells (BMSCs)—which were then induced into osteoblasts for subsequent experiments—were performed as described in a previous study [28]. Similarly, the isolation and culture of primary mouse hippocampal neurons followed the protocol outlined in another prior study [37]. The cells were maintained in a humidified incubator with an atmosphere of 5 % CO₂ at a temperature of 37 °C.

Behavioral tests

The OFT, the three-chamber social interaction test, and the MWM test were conducted following standard protocols [38]. In the OFT, the moving distance, the average speed, along with the proportion of stationary time of the tested mice were measured. In the three-chamber social interaction test, after the acclimation and training periods, the test mice were presented with a familiar mouse and an unfamiliar mouse placed in separate outer compartments. This setup allowed for the assessment of the mice's social interaction preferences. In the MWM test, key parameters including the escape latency (the time taken by the mice to reach the hidden platform), the number of entries into the target quadrant, and the time spent in the target quadrant were carefully recorded. All behavioral test results were subjected to appropriate statistical analyses to draw reliable conclusions.

Bone mass measurement

After the dissected bones were fixed, they were reconstructed and analyzed using a micro-CT analysis system (SCANCO Medical AG). The scanning parameters were set as follows: voltage 70 kV, current 200 µA, and layer thickness 10 µm. The paraffin-embedded bone sections were subjected to dewaxing and rehydration processes. Subsequently, hematoxylin and eosin (H&E) staining and Masson staining were carried out in strict accordance with the standard protocol [39]. Digital images of the stained sections were acquired using a digital panoramic camera system (WS-10, WISLEAP).

RNA sequencing (RNA-seq) analysis

Total RNA was extracted from MC3T3-E1 and HT22 cells that had been treated with vehicle or RIS. Subsequently, RNA-seq was carried out on the Illumina NovaSeq 6000 platform. Transcriptome assembly was accomplished using StringTie2, while read mapping was performed with HISAT2. The DESeq2 algorithm was employed to standardize gene expression levels. Differentially expressed genes (DEGs) were identified based on the value of P less than 0.05 and an absolute value of the log2 fold-change (|log2FC|) greater than 1. Hierarchical cluster analysis of the identified DEGs was conducted using R (version 3.2.0). GO, KEGG, WikiPathways, and GSEA were performed using the same R version.

Targeted lipidomics analysis (TLA)

Total lipids were extracted from MC3T3-E1 and HT22 cells that had been treated with vehicle or RIS for TLA. Subsequently, liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis was carried out. This analysis involved ultra-high-performance liquid chromatography (UHPLC) for sample separation and mass spectrometry for compound identification and quantification. The samples were separated using a Nexera X2 LC-30AD ultra-high-pressure liquid chromatography system (Shimadzu). Mass spectrometry analysis was performed using a QTRAP 5500 mass spectrometer (AB SCIEX) in both positive and negative ion modes. Differentially expressed compounds (DECs) were identified based on the value of P less than 0.05 and an absolute value of the log2 fold-change (|log2FC|) greater than 1. KEGG was carried out using R (version 3.2.0).

Molecular dynamics (MD) simulations

We utilized the CHARMM-GUI membrane builder to construct the lipid bilayer, following a contemporary model of the mammalian lysosomal membrane. The positioning of the SMPD1 protein in Model 1 and Model 2 was determined based on a previous study [14]. In the SMPD1-RIS system and the RIS-only system, molecules were randomly incorporated into the system at a density corresponding to a 6 % molecule-to-lipid ratio. For the SMPD1-RIS system, the final frame from the Model 1 simulation was employed as the starting point for inserting the RIS molecule. During the MD simulations, we adhered to the standard CHARMM-GUI membrane builder protocol for energy minimization and equilibration. Subsequently, a 500-ns production run was carried out for all four systems under consideration, with an integration time step of 4 fs. The production runs were performed within the NPT ensemble. The temperature of the system was maintained at 310 K using the Nose-Hoover thermostat, and the pressure was coupled at 1 bar using the Parrinello-Rahman barostat. The LINCS algorithm was applied to constrain the bonds involving heavy atoms. The cutoff distance was set to 12 Å for both van der Waals and Coulombic interactions. To account for long-range interactions, the Particle Mesh Ewald summation method was used. The LipidDyn package was used to calculate the membrane curvature, and the gmx density tool was employed for density profile calculations. All contact analyses were conducted using PyMOL.

Molecular docking and virtual screening

The amino acid composition of the SMPD1 binding pocket was determined using Schrödinger Glide Global Search. The KEGG Drug Database (Japan PMDA-listed, https://www.kegg.jp/kegg-bin/get_htext#D277), which focuses on metabolism-related drug classifications, was selected for virtual screening of compounds against the SMPD1 binding pocket. This database contains 387 small-molecule drugs approved for marketing. Following AutoDock molecular docking, the binding conformations and energy scores of the top 100 small molecules based on binding energy were retrieved, with the binding modes of the top 10 molecules to SMPD1 further visualized. Pymol and LigPlot were utilized to generate graphical representations of molecular docking results, illustrating binding modes such as SMPD1-RIS and SMPD1-ED-71.

Isothermal titration calorimetry (ITC) analysis

To evaluate the interactions between RIS/ED-71 and wild-type (WT) SMPD1 or SMPD1^R294A mutant, calorimetric experiments were performed using a NANO ITC system (TA Instruments, USA) [40]. The initial titration volume of RIS/ED-71 was 50 μL (injected from the syringe), while that of SMPD1, SMPD1^R294A mutant, or SMPD1-RIS complex was 300 μL (added to the sample cell). Titration was repeated 20 times with 2 μL per injection, with an interval of 120 s, at 25 °C, and a stirring speed of 350 rpm. Experimental groups included titration of SMPD1, SMPD1^R294A mutant, or SMPD1-RIS complex with RIS/ED-71; the CON group consisted of titration of blank solvent with RIS/ED-71. Data were analyzed using Origin 7.5 software to calculate parameters including stoichiometry, equilibrium dissociation constant, enthalpy change, and entropy change.

Statistical analyses

Biological replicates were set as follows: 6 replicates for behavioral tests, 5 replicates for other in vivo experiments and key in vitro experiments, and 3 replicates for the remaining in vitro experiments. Sample size estimation was predicated on an initial experiment. The primary outcomes assessed were bone volume/total volume (BV/TV), hippocampal GABA levels, and SMPD1 enzymatic activity. Using G*Power software (v3.1), the required sample size per group was determined, to achieve an 80 % power level and a 5 % two-tailed significance level [41]. Data were analyzed with GraphPad Prism and presented as mean ± SD. Two-group mean comparisons: Normality (Shapiro-Wilk) and homoscedasticity (F-test) were assessed; independent t-test (normal/equal variance), Welch’s t-test (normal/unequal variance), or Mann-Whitney U test (non-normal) was applied. Three or more group mean comparisons: Normality (Shapiro-Wilk) and homoscedasticity (Brown-Forsythe) were assessed; ANOVA + Tukey’s post hoc test (normal/equal variance) or Kruskal-Wallis H + Dunn’s post hoc test (non-normal/unequal variance) was used. The difference threshold was considered statistically significant at P value < 0.05 and indicated by ‘*’, P < 0.01 were indicated by ‘^**’, P < 0.001 were indicated by ‘^***’, P < 0.0001 were indicated by ‘^****’.

Results

RIS exacerbates bone loss in MK-801-induced SZ mice and exhibits incomplete antipsychotic efficacy

To determine the impact of RIS on bone remodeling in MK-801-induced SZ mice, micro-CT analysis revealed that MK-801 reduced femoral bone mass in female mice, an effect further exacerbated by RIS co-administration (Fig. 1A). H&E staining also showed MK-801 decreased trabecular bone volume, number, and thickness with disorganized architecture, all worsened by RIS (Fig. 1B; fig. S1C). Masson staining confirmed suppressed bone regeneration in MK-801-treated mice, with pronounced inhibition in the MK-801 + RIS group (Fig. 1B). RIS-dependent bone loss was more pronounced in female mice (Fig. 1B; fig. S1C) than in male mice (fig. S1D-E), thus we used female animals for subsequent analysis. Molecular analyses demonstrated downregulated osteogenic (RUNX2, ALP) and osteoclastic (TRAP, CTSK) markers in SZ mice, with RIS specifically exacerbating osteogenic suppression and enhancing osteoclastogenesis (Fig. 1C-D). ELISA revealed elevated levels of prolactin (PRL) and norepinephrine (NE) in the femoral tissue of RIS-treated mice (Fig. 1E). In vitro, PRL (50 μM) and NE (100 μM)—at cytotoxic concentrations—reduced ALP activity and mineralized nodule formation in MC3T3-E1 osteoblasts (fig. S1F-H), linking neuroendocrine signaling to RIS-mediated bone loss. Cell viability assays showed dose-dependent growth inhibition of MC3T3-E1 by MK-801 and RIS, with 0.1 μM MK-801 (non-cytotoxic) and 100 μM RIS selected for mechanistic studies (fig. S2A-B). RNA-seq of MC3T3-E1 cells revealed downregulation of cell proliferation and osteogenic differentiation pathways (Fig. 1F), validated by osteogenic induction assays (Fig. 1G-H). Primary BMSCs were also isolated for osteogenic induction (fig. S2C), which further confirmed that RIS directly impairs osteoblast function (Fig. S2D).

Fig. 1.

RIS Exacerbates Bone Loss in MK-801-induced SZ Mice and Exhibits Incomplete Antipsychotic Efficacy. (A) Micro-CT imaging of femurs from female mice (2D and 3D). (B) H&E and Masson staining of femurs from female mice (n = 5). (C-D) Immunohistochemistry (IHC) staining, TRAP staining, and quantitative analysis of RUNX2, ALP, TRAP, and CTSK from female mice (green arrows: IHC positive; red arrows: TRAP positive) (n = 5). (E) ELISA of PRL and NE levels in femoral homogenates (n = 5). (F) GSEA analysis of RNA-seq data showing enrichment in cell proliferation/osteogenesis pathways (MC3T3-E1). (G-H) ALP/ARS staining with quantitative analysis in MC3T3-E1 cells (n = 3). (I) OFT trajectories and quantitative analysis from female mice (total distance, speed, immobility time) (n = 6). (J) Three-chamber test trajectories and social interaction time analysis from female mice (n = 6). (K) MWM trajectories and quantitative analysis from female mice (escape latency, platform crossings) (n = 6). (L) H&E/Nissl staining of hippocampal neurons from female mice (n = 5). (M) GAD65/SYN IF co-staining and colocalization analysis in HipGABA-Ns from female mice (n = 5). (N) ELISA-measured levels of GABA in hippocampus, DA in midbrain, and 5-HT in raphe nuclei from female mice (n = 5). (O) GSEA analysis of RNA-seq data showing enrichment in GABAergic synapse/axon guidance pathways. (P-Q) SYN/Phalloidin IF staining and IF intensity quantification in HT22 cells (n = 3). (R) ELISA of GABA levels in HT22 cells. Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Behavioral tests showed limited efficacy of RIS in reversing MK-801-induced psychotic phenotypes: partial reduction of hyperlocomotion in the OFT (Fig. 1I), failure to restore social interaction in the three-chamber test (Fig. 1J), and no improvement in spatial memory in the MWM test (Fig. 1K). Histopathology revealed that MK-801-induced hippocampal neuronal loss was exacerbated by RIS (Fig. 1L), with reduced synaptic marker SYN expression and reduced colocalization in HipGABA-Ns (Fig. 1M). ELISA showed RIS decreased hippocampal GABA levels while simultaneously increasing DA in midbrain and 5-HT in raphe nuclei (Fig. 1N). RIS-dependent hippocampal neuronal damage and reduced GABA levels were more pronounced in female mice (Fig. 1L, N) than in male mice (fig. S2E-F); thus, female animals were also used for subsequent analysis. In vitro experiments, primary hippocampal neurons were isolated to more closely mimic the in vivo environment (fig. S2G). Meanwhile, the HT22 cell line was selected for HipGABA-N-related studies due to its GABA-secreting capacity [42] and high expression of the GABAergic neuronal marker GAD65 (fig. S2H). Cell viability assays showed RIS inhibited its viability in a dose-dependent manner, with 0.1 μM MK-801/100 μM RIS selected for studies (fig. S2I-J). RNA-seq of HT22 cells identified the suppression of GABAergic synapse pathway and axon guidance pathway in the RIS-treated group (Fig. 1O), confirmed by immunofluorescence (IF) staining of SYN (Fig. 1P-Q) and reduced GABA secretion (Fig. 1R). Reduced GABA levels in primary hippocampal neurons further validated this finding additionally (fig. S2K), demonstrating direct disruption of hippocampal neurotransmission by RIS.

Collectively, these data show that RIS promotes OP and exerts incomplete antipsychotic effects, likely through direct dual disruption of osteoblast differentiation and hippocampal GABAergic neurotransmission.

RIS induces lysosome-dependent ferroptosis in osteoblasts and HipGABA-Ns

To dissect the direct mechanisms by which RIS disrupts osteoblast differentiation and hippocampal GABAergic neurotransmission, caspase-3 activity and TUNEL staining were assessed in femur and hippocampal tissue; results showed that RIS did not upregulate apoptosis levels (fig. S3A-F). Subsequently, transcriptome sequencing was performed, and analysis revealed significant enrichment of ferroptosis-related gene sets in RIS-treated MC3T3-E1 and HT22 cells (Fig. 2A-F). Concomitantly, expression of ACSL4, a key marker of lipid peroxidation, was upregulated in femoral osteoblasts and HipGABA-Ns of MK-801 + RIS mice (Fig. 2G-H). Subsequent WB analysis validated the upregulation of ACSL4 and downregulation of GPX4 in primary and cell line-derived osteoblasts and hippocampal neurons following RIS administration (fig. S3G-N), linking RIS exposure to enhanced lipid peroxidation. Then we performed TLA to further validate pathway enrichment for ferroptosis in MC3T3-E1 and HT22 cells following RIS treatment (Fig. 2I), with pronounced accumulation of unsaturated lipid species—substrates for peroxidation (Fig. 2J, L). Liperfluo staining confirmed ferroptosis-related lipid accumulation in both cell types following RIS administration (fig. S4A-D). Malondialdehyde (MDA) analysis (Fig. 2K, M) and BODIPY 581/591 C11 staining (Fig. 2N-P; fig. S4E-F) verified that RIS administration induced lipid peroxidation of both cell types. Ultrastructural characterization by transmission electron microscopy (TEM) revealed hallmark ferroptotic features in RIS-treated MC3T3-E1 and HT22 cells, including increased mitochondrial membrane electron density (indicative of lipid peroxidation product deposition) and vacuolization (Fig. 2Q-R). To validate the in vitro-to-in vivo mechanistic extrapolation, dose–response assays for ferroptosis were conducted at 0, 3, 30, 100, 150 μM (fig. S5A-L). Results showed a dose-dependent increase in ferroptosis, starting from 3 μM—a concentration close to the clinically equivalent level [43]—and reaching a plateau at 100 μM (fig. S5A-L). This 100 μM concentration is consistent with those used in previous studies [44,45]. These findings indicate that even clinically relevant low concentrations can induce the effect, while 100 μM, as a sensitive saturating concentration, was therefore used in subsequent experiments.

Fig. 2.

RIS Induces Lysosome-Dependent Ferroptosis in Osteoblasts and HipGABA-Ns. (A, D) KEGG and WikiPathway analysis of RNA-seq in MC3T3-E1 and HT22 cells treated with vehicle vs. RIS. (B, E) GSEA analysis of ferroptosis pathway enrichment across cell types. (C, F) Heatmaps of ferroptosis-related gene sets. (G) ALP/ACSL4 IF co-staining and colocalization analysis in femoral osteoblasts (GP: growth plate. TB: bone trabeculae) (n = 5). (H) GAD65/ACSL4 IF co-staining and colocalization analysis in HipGABA-Ns (n = 5). (I, J, L) KEGG analysis of TLA across cell types. (K, M) MDA quantification in MC3T3-E1 and HT22 cells (n = 5). (N-P) BODIPY 581/591 C11 staining and quantification of lipid peroxidation across cell types (n = 5). (Q-R) TEM images of mitochondrial morphology across cell types (arrows: normal, vacuolated, high electron density mitochondria). (S-T) Mito-Tracker/MitoSOX double staining and colocalization analysis for mitochondrial ROS levels (n = 5). (U-V) Mito-Tracker/Mito-FerroGreen double staining and colocalization analysis of mitochondrial iron loading (n = 5). (W-X) Lyso-Tracker/FerroOrange double staining and colocalization analysis for lysosomal iron levels in both cell types (n = 5). (Y-Z) Flow cytometry analysis of lipid peroxidation in MC3T3-E1 and HT22 cells (n = 3). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Considering that lipid peroxidation is driven by an iron-dependent Fenton reaction [46], mitochondrial redox status and iron levels were further analyzed. Functional assays demonstrated that RIS triggered mitochondrial dysfunction, manifesting as elevated reactive oxygen species (ROS) and loss of mitochondrial membrane potential (MMP), indications of Fenton reaction-driven oxidative stress (Fig. 2S-T; fig. S6A-D). Mito-Tracker/Mito-FerroGreen double staining confirmed robust mitochondrial iron overload in the MK-801 + RIS group (Fig. 2U-V; fig. S6E-H). Given that lysosomes as central iron storage organelles [20], we investigated lysosomal iron homeostasis and observed disrupted iron sequestration in RIS-treated cells, evidenced by reduced colocalization of the lysosomal marker Lyso-Tracker and the iron probe FerroOrange (Fig. 2W-X; fig. S6I-L). Critically, treatment with the ferroptosis inhibitor liproxstatin-1 (Lip-1), which inactivates iron in lysosomes, attenuated ferroptotic phenotypes in both cell types (Fig. 2Y-Z), establishing a causal link between lysosomal iron dyshomeostasis and RIS-induced ferroptosis.

Collectively, these findings establish that RIS elicits lysosome-dependent ferroptosis in osteoblasts and HipGABA-Ns, driving impaired osteogenesis and disrupted GABAergic synaptic function. This mechanistic framework highlights ferroptosis as a convergent node in RIS-mediated systemic and neural toxicity.

RIS induces hypertrophic lysosomes and LMP in osteoblasts and HipGABA-Ns

Given that RIS-induced ferroptosis in osteoblasts and HipGABA-Ns exhibits lysosome dependency—a hallmark indicating the lysosomal tropism of RIS—we performed MD simulations. MD analyses demonstrated that RIS penetrates the lysosomal membrane, with interfacial accumulation and progressive membrane insertion observed within 500 ns (Fig. 3A-C). Critically, this tropism-driven membrane insertion may compromise lysosomal integrity subsequently, thereby facilitating lysosomal iron leakage—a pivotal event in lysosome-dependent ferroptosis. To further characterize RIS-mediated lysosomal alterations, we investigated morphological and functional changes in both cell types. Morphological analysis via LAMP1-mCherry transfection revealed significantly enlarged lysosomes in RIS-treated MC3T3-E1 and HT22 cells compared to MK-801 controls (Fig. 3D-E). Transcriptomic profiling uncovered hyperactivation of lysosomal pathways in RIS-exposed cells (Fig. 3F-G), with GSEA identifying enrichment of lysosomal damage signatures, including phosphatidylinositol-4-phosphate binding (Fig. 3H-I)—a pathway recently implicated in lysosomal injury sensing. Functional assays validated these findings: FITC-dextran labeling revealed RIS-induced lysosomal enlargement, alkaline luminal pH shift, and LMP, as evidenced by increased vesicle size, enhanced fluorescence intensity, and diffuse signal distribution (Fig. 3J-K, N, P; fig. S7A-B). Acridine orange (AO) staining further confirmed LMP-associated membrane instability: reduced AO-Red/AO-Green ratios indicated compromised lysosomal integrity in RIS-treated cells (Fig. 3L-M, O, Q; fig. S7C-D). IF co-staining showed elevated expression of Galectin-3, a conserved marker of lysosomal damage, in both femoral osteoblasts and HipGABA-Ns (Fig. 3R-S), linking RIS exposure to cell-type generalized lysosomal damage. Finally, we used the LMP inducer U18666A [47] to verify the effect of LMP on iron leakage. Following treatment with U18666A, the colocalization of Lyso-Tracker and FerroOrange was further reduced compared with RIS administration (fig. S7E-H); meanwhile, double staining with Mito-Tracker/Mito-FerroGreen confirmed a further exacerbation of mitochondrial iron overload (fig. S7I-L), and consistent results were obtained via flow cytometry (Fig. 3T-W). These findings mechanistically confirm the necessity of lysosomal integrity for RIS-mediated ferroptosis.

Fig. 3.

RIS Induces Hypertrophic Lysosomes and LMP in Osteoblasts and HipGABA-Ns. (A-C) Structural formula of RIS and its interaction with the lysosomal membrane system, including a molecular insertion schematic. (D-E) LAMP1-mCherry transfection for lysosomal diameter analysis and quantification (n = 3). (F-G) GSEA/KEGG analysis in MC3T3-E1 and HT22 cells treated with vehicle vs. RIS. (H-I) GSEA analysis of the phosphatidylinositol-4-phosphate binding pathway enrichment in MC3T3-E1 and HT22 cells. (J-K, N, P) FITC-Dextran fluorescence staining for lysosomal function across cell types (n = 5). (L-M, O, Q) AO staining for LMP detection with fluorescence intensity quantification (AO-Red/AO-Green ratio) (n = 5). (R-S) ALP/Galectin-3 and GAD65/Galectin-3 IF co-staining with colocalization analysis (n = 5). (T-W) Flow cytometry analysis of mitochondrial iron (Mito-FerroGreen) in MC3T3-E1 and HT22 cells, with quantification (n = 3). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

These results collectively demonstrate that RIS induces lysosomal hypertrophy and disrupts lysosomal membrane homeostasis, initiating a cascade of iron leakage critical for ferroptotic execution in osteoblasts and HipGABA-Ns.

RIS induces lysophagy dysfunction in osteoblasts and HipGABA-Ns

Lysosomes contain hydrolytic enzymes that degrade intracellular and extracellular materials via lysophagy. Mechanistic investigations linked lysosomal hypertrophy and LMP to pathological activation or catabolic failure of lysophagy: transcriptomic (Fig. 4A-B) and lipidomic profiling (Fig. 4C-D) revealed upregulation of autophagic programs in RIS-treated MC3T3-E1 and HT22 cells. Using the mRFP-GFP-LC3 reporter system, RIS induced persistent colocalization of autophagosomes and autolysosomes in osteoblasts and HipGABA-Ns, indicative of pathological activation of autophagy and late-stage autophagic flux blockade (Fig. 4E-F). LAMP1-mCherry/GFP-LC3 double transfection confirmed sustained overlap in RIS-treated cells, reflecting lysosomal hydrolytic dysfunction and accumulation of undegraded autophagic substrates (Fig. 4G-H). IF co-staining (Fig. 4I-J, M−N) and biochemical analyses (Fig. 4K-L) showed elevated levels of autophagy markers (Beclin1, p62, LC3B-II/I, LAMP2) in osteoblasts and HipGABA-Ns, consistent with excessive autophagy initiation and impaired flux termination. Next, we used methotrexate (MTX)—an agent that induces early autophagy initiation while inhibiting late autophagic degradation [48,49]—to verify the effect of autophagic impairment on ferroptosis. After MTX treatment, the expression of ferroptosis-related proteins was further exacerbated compared with RIS treatment (fig. S8A-D). Additionally, Liperfluo and BODIPY 581/591 C11 staining confirmed its role in exacerbating ferroptosis-related lipid content and lipid peroxidation (fig. S8E-H), suggesting that autophagolysosomal dysfunction contributes to ferroptosis.

Fig. 4.

RIS Induces Lysophagy Dysfunction in Osteoblasts and HipGABA-Ns. (A-B) GSEA analysis of autophagy pathway enrichment in MC3T3-E1 and HT22 cells via RNA-seq. (C-D) KEGG analysis of TLA of vehicle vs. RIS-treated MC3T3-E1 and HT22 cells. (E-F) mRFP-GFP-LC3 transfection with fluorescence colocalization in both cell types (n = 3). (G-H) LAMP1-mCherry/GFP-LC3 co-transfection and colocalization analysis across cell types (n = 3). (I-J) Beclin1/p62 IF co-staining with 3D fluorescence intensity analysis in both cell types (n = 3). (K-L) WB of autophagy markers (LC3B-II/I, p62, LAMP2) with quantitative analysis in MC3T3-E1 and HT22 cells (n = 3). (M−N) ALP/p62 and GAD65/p62 IF co-staining with colocalization in osteoblasts and HipGABA-Ns (n = 5). (O-R) TFEB-eGFP transfection and nuclear/cytoplasmic ratio quantification across cell types (n = 3). (S) LAMP2/mTOR IF co-staining across cell types (n = 3). (T-W) FRAP imaging of LAMP1-mCherry and quantitative analysis in MC3T3-E1 and HT22 cells (n = 3). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

To further analyze the origin of autophagic impairment, we first focused on the regulatory mechanism of autophagic initiation: it is regulated by the mTORC1-TFEB signaling axis via the CLEAR network [50]. In MC3T3-E1 and HT22 cells, TFEB-eGFP transfection revealed that RIS triggered rapid nuclear translocation of TFEB (signaling mTORC1 inactivation) (Fig. 4O-R), accompanied by reduced phosphorylation of the mTORC1 substrate p70S6K (fig. S9A-B). In vivo, RIS-treated mice exhibited decreased p-p70S6K levels in femoral osteoblasts and HipGABA-Ns (fig. S9C-D). IF co-staining showed loss of LAMP2-mTOR colocalization in both cell types following RIS treatment compared to MK-801 controls (Fig. 4S). These findings collectively indicate RIS-mediated inhibition of mTORC1 activity, a key trigger for autophagy initiation activation. We next explored the lysosomal microenvironment underlying mTORC1 regulation. mTORC1 assembles in lysosomal lipid rafts containing sphingomyelin (SM) [51,52], and lipid raft disruption is linked to altered lysosomal membrane fluidity and mTORC1 inhibition [53]. In LAMP1-mCherry-transfected cells, fluorescence recovery after photobleaching (FRAP) assays demonstrated that RIS accelerated lysosomal membrane fluorescence recovery (Fig. 4T-W)—a quantitative indicator of increased membrane fluidity, consistent with perturbed lipid raft composition. Notably, SM—key components of lipid rafts—not only affect mTORC1 inhibition-mediated lysophagy initiation [54], but also serve as critical substrates for lysosomal lipid catabolism [55], with roles in late-stage autophagy regulation [56]. Thus, lysosomal lipid metabolism is inherently linked to lysosomal function, regulating autophagy initiation and late-stage autophagic processes.

These findings point to a potential interdependence between lysosomal lipid raft stability, lipid metabolism, and lysophagy function, hinting at their possible combined role in RIS-mediated lysosomal dysfunction in osteoblasts and HipGABA-Ns.

RIS inhibits SMPD1 activity causing lysosomal phospholipid storage in osteoblasts and HipGABA-Ns

Given the critical role of lysosomal lipids in maintaining lysosomal homeostasis, we investigated the effects of RIS on lysosomal lipid metabolism. TLA in MC3T3-E1 and HT22 cells revealed significant upregulation of glycerophospholipid and sphingolipid metabolic pathways following RIS treatment (Fig. 5A), especially including phosphatidylcholine (PC) and SM (Fig. 5B-C). Lyso-Tracker/HCS LipidTOX double staining confirmed abnormal accumulation of phospholipids within lysosomal compartments in both cell types (Fig. 5D-E). Subsequently, we employed phosphocholine (PCh)—a precursor for phospholipid synthesis [57] —to induce excessive phospholipid accumulation, thereby verifying its effect on LMP. Following RIS administration, PCh further decreased the fluorescence intensity of AO-red/green and increased the fluorescence intensity of FITC in both cell types (fig. S10A-D), confirming that phospholipid accumulation aggravates LMP. To identify the molecular mediator of this phenotype, transcriptomic GSEA highlighted a significant reduction in lysophospholipase pathway enrichment in RIS-treated osteoblasts (Fig. 5F). SMPD1, one of lysophospholipases, is a lysosomal membrane-resident enzyme essential for phospholipid hydrolysis, especially PC and SM (Fig. 5G) [55].

Fig. 5.

RIS Inhibits SMPD1 Activity Causing Lysosomal Phospholipid Storage in Osteoblasts and HipGABA-Ns. (A) KEGG enrichment analysis of TLA in osteoblasts and HipGABA-Ns. (B-C) Analysis of differential glycerophospholipid/sphingolipid metabolites in MC3T3-E1 and HT22 cells. (D-E) Lyso-Tracker/HCS LipidTOX double staining and colocalization across cell types (n = 3). (F) GSEA analysis of lysophospholipase activity pathway enrichment in MC3T3-E1 cells. (G) Flowchart of SMPD1-mediated hydrolysis of PC and SM in lysosomes. (H) SMPD1 activity quantification in both cell types (n = 5). (I-J) eGFP-SMPD1/LAMP1-mCherry co-transfection images in MC3T3-E1 and HT22 cells (n = 3). (K-L) WB of SMPD1 protein with quantitative analysis (n = 3). (M−N) Schematic diagrams of SMPD1-lysosomal membrane binding and RIS-SMPD1 interaction. (O-Q) MD analysis of RIS-SMPD1 system: lipid atom contacts, interaction frequency, membrane curvature, and residue contact heatmap. (R) Protein binding pocket of SMPD1 with RIS; interactions of SMPD1 with RIS in 3D and 2D formats. (S) Fitting curve for integrated enthalpy change during titration of WT SMPD1 with RIS. (T) Fitting curve for integrated enthalpy change during titration of SMPD1^R294A mutant with RIS. (U) SMPD1 activity statistics in both cell types (n = 5). (V-W) Lyso-Tracker/HCS LipidTOX double staining and colocalization across cell types (n = 3). (X-Y) ALP/SMPD1 and GAD65/SMPD1 IF co-staining with colocalization in femoral osteoblasts and HipGABA-Ns (n = 5). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Next, we evaluated the effects of RIS on the enzymatic activity, localization, and expression level of SMPD1. Biochemical analyses showed that RIS treatment significantly inhibited SMPD1 enzymatic activity, directly impairing lysosomal phospholipid catabolism (Fig. 5H; fig. S10E). Furthermore, dose–response experiments confirmed the dose-dependent inhibition of SMPD1 activity and concentration-dependent saturation (fig. S10F). Through eGFP-SMPD1/LAMP1-mCherry co-transfection assays, we found that RIS disrupted the colocalization of SMPD1 with lysosomal membranes in osteoblasts and HipGABA-Ns (Fig. 5I-J). WB analysis further indicated that RIS treatment downregulated SMPD1 protein expression in both cell types (Fig. 5 K-L).

To dissect the structural basis of SMPD1 dysfunction, MD simulations were performed to model interactions between RIS, the lysosomal membrane, and SMPD1. Two membrane association models for SMPD1 were constructed (Model 1: peripheral membrane binding; Model 2: transmembrane helix insertion, fig. S11A-B) to analyze the roles of anchoring sites and catalytic domains in lysosomal membrane interaction and phospholipid contact diversity (Fig. 5M; fig. S11C–H). The RIS-SMPD1-lysosomal system was further analyzed, revealing that RIS insertion into the water–lipid interface of the membrane disrupting key hydrophobic interactions between SMPD1′s transmembrane domain and phospholipid bilayers (Fig. 5N; fig. S11I). This interference disrupted key residue interactions by the 300 ns mark (Fig. 5O), reducing total contact frequency by 9 % (from initial to final 50 ns of simulation; Fig. 5O) and increasing membrane leaflet curvature (Fig. 5P)—changes that collectively suggested SMPD1 dislocation following RIS treatment.

Direct protein–ligand interaction analysis in MD simulations identified high-frequency interaction residues—ARG294, ALA295, TRP209, and HIS211—in SMPD1′s catalytic core (Fig. 5Q), which overlapped with the binding pocket defined by Schrödinger Global Docking (Fig. 5R). Autodock simulations further revealed that RIS interacted with up to nine amino acid residues within this pocket, forming a covalent bond and polar interactions (hydrogen bonds and van der Waals forces) with multiple other residues in SMPD1. These interactions stabilized a non-functional protein conformation that may hinder transmembrane helix dynamics and phospholipid substrate access (Fig. 5R; Table. S1). Molecular docking and MD simulations identified ARG294 as the highest-frequency interaction residue, yet these are only auxiliary hypothesis generation data—whereas ITC assays provide core evidence for directly validating RIS-SMPD1 physical interaction. We thus conducted ITC to measure heat changes during titration of WT SMPD1 and SMPD1^R294A with RIS. Results showed that when WT SMPD1 was titrated with RIS, the fitting curve corresponding to the integrated enthalpy change (as RIS concentration increased) exhibited a saturation trend of first rising and then leveling off (Fig. 5S); in contrast, when SMPD1^R294A mutant was titrated with RIS, the fitting curve showed a linear downward trend without saturation characteristics (Fig. 5T)—this distinct difference directly confirms that ARG294 is a key residue for the interaction between RIS and SMPD1. Furthermore, functional validation of the ARG294 residue was performed by transfecting RIS-treated MC3T3-E1 and HT22 cells with plasmids encoding WT SMPD1 or the SMPD1^R294A mutant. Notably, SMPD1^R294A failed to rescue the enzymatic activity of SMPD1 and lysosomal phospholipid accumulation, phenocopying the effects of the MK-801 + RIS group (Fig. 5U-W), confirming ARG294 as a pivotal site for RIS-SMPD1 interaction. In vivo, we further measured the levels of SMPD1 in femoral osteoblasts and HipGABA-Ns, and significant reductions were also observed after RIS administration (Fig. 5X-Y).

Taken together, our results illustrate that RIS suppresses the activity of SMPD1 by directly interacting with the residue ARG294 in SMPD1, thereby inducing lysosomal phospholipid storage, then causes LMP and iron-mediated ferroptosis in osteoblasts and HipGABA-Ns.

ED-71 reverses RIS-induced lysosomal dysfunction by antagonizing RIS-SMPD1 interaction and regulating phospholipid homeostasis

To identify therapeutic candidates that disrupt interactions at the RIS-binding pocket of SMPD1 and restore lysosomal phospholipid homeostasis, we performed virtual drug screening using a metabolism-focused drug database (PMDA-listed, https://www.kegg.jp/kegg-bin/get_htext#D277), leveraging SMPD1′s lipid-catabolic structural features (Fig. 6A). Following AutoDock molecular docking and binding energy scoring, the top 10 candidates were analyzed for binding interactions with SMPD1 (Fig. 6A; Table S1). ED-71 emerged as the lead compound with the lowest binding energy of − 8.336 kcal/mol—surpassing RIS’s affinity—and engaged fewer SMPD1 residues without forming covalent bonds, thus could avoid catalytically inert enzyme stabilization (Fig. 6B). In addition, we measured the heat changes during the titration of ED-71 with the pre-formed RIS-SMPD1 complex. The results showed that ED-71 still formed a dose-dependent and saturable fitting curve when titrated with this complex (Fig. 6C), confirming that ED-71 has a specific competitive antagonistic effect on the binding between RIS and SMPD1. Further Cell Counting Kit-8 (CCK-8) assay confirmed the improving effect of ED-71 on cell viability in osteoblasts and HipGABA-Ns compared to the MK-801 + RIS group, and determined the optimal concentration (Fig. 6D).

Fig. 6.

ED-71 Reverses RIS-Induced Lysosomal Dysfunction by Antagonizing RIS-SMPD1 Interaction and Regulating Phospholipid Homeostasis. (A) Flowchart of virtual drug screening and the top 10 docking diagrams. (B) Interactions of SMPD1 with ED-71 in 3D and 2D formats. (C) Fitting curve for integrated enthalpy change during titration of RIS-WT SMPD1 with ED-71. (D) Cell viability of MC3T3-E1 and HT22 cells (n = 3). (E) Meta-analysis flow diagram for study selection. (F) Forest plot of 25(OH)D level differences between groups. (G) SMPD1 activity quantification in MC3T3-E1 and HT22 cells (n = 5). (H-I) eGFP-SMPD1/LAMP1-mCherry co-transfection images across cell types (n = 3). (J-K) WB of SMPD1 protein with quantitative analysis (n = 3). (L-M) Lyso-Tracker/HCS LipidTOX double staining and colocalization in both cell types (n = 3). (N-O) ALP/SMPD1 and GAD65/SMPD1 IF co-staining with colocalization in femoral osteoblasts and HipGABA-Ns (n = 5). (P-Q) FRAP analysis of LAMP1-mCherry in MC3T3-E1 and HT22 cells with quantitative data (n = 3). (R) LAMP2/mTOR co-staining in both cell types (n = 3). (S-T) TFEB-eGFP transfection and nuclear/cytoplasmic ratio quantification across cell types (n = 3). (U-X) mRFP-GFP-LC3 and LAMP1-mCherry/GFP-LC3 transfection for lysophagy analysis in MC3T3-E1 and HT22 cells (n = 3). (Y-Z) ALP/p62 and GAD65/p62 IF co-staining with colocalization in femoral osteoblasts and HipGABA-Ns (n = 5). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

As a novel active vitamin D analog, ED-71 is administered to patients with endogenous active vitamin D deficiency to exogenously supplement active vitamin D. As active vitamin D is synthesized from 25(OH)D, we performed a meta-analysis to compare 25(OH)D levels in SZ patients on atypical antipsychotics versus healthy controls. The retrieval process is detailed in Fig. 6E. Funnel plot asymmetry (Egger's P = 0.02, fig. S12A) indicated potential publication bias, while sensitivity analysis confirmed no significant heterogeneity (fig. S12B). Random-effects model forest plots showed significantly lower 25(OH)D levels in the patient cohort compared to controls (Fig. 6F), underscoring the need for ED-71 supplementation.

Furthermore, the protective effect of ED-71 on the interaction between RIS and SMPD1 was investigated. Biochemical profiling revealed that ED-71 upregulated SMPD1 enzymatic activity in both cell types (Fig. 6G; fig. S12C). Subcellular localization assays (eGFP-SMPD1/LAMP1-mCherry co-transfection) demonstrated that ED-71 restored SMPD1-lysosome colocalization in both osteoblasts and HipGABA-Ns (Fig. 6H-I). WB analysis further indicated that ED-71 treatment upregulated SMPD1 protein expression in both cell types (Fig. 6J-K). Concurrently, Lyso-Tracker/HCS LipidTOX double staining confirmed ED-71 administration normalized RIS-induced lysosomal phospholipid accumulation (Fig. 6L-M). SMPD1 knockdown completely abrogated these protective effects, confirming SMPD1-dependency of ED-71′s protective action (Fig. 6G-M). In vivo experiments validated that ED-71 restored the expression levels of SMPD1 in femoral osteoblasts and HipGABA-Ns of SZ model mice administered with RIS (Fig. 6N-O).

Membrane biophysics analyses using FRAP revealed ED-71 significantly attenuated lysosomal membrane recovery kinetics (Fig. 6P-Q), indicative of restored membrane fluidity. IF co-staining showed ED-71 reestablished mTOR-LAMP2 colocalization (Fig. 6R). Mechanistically, ED-71 normalized mTORC1 signaling by suppressing TFEB nuclear translocation and enhancing p70S6K phosphorylation (Fig. 6S-T; fig. S12D-G), thereby may alleviating autophagic hyperactivation and flux blockade in both cell types. Autophagic flux quantification via tandem mRFP-GFP-LC3 (Fig. 6U-V) and LAMP1-mCherry/GFP-LC3 (Fig. 6W-X) transfection demonstrated ED-71 reduced autophagosome accumulation and restored late-stage autophagic processing. Biochemical analyses further showed ED-71 downregulated LC3B-II/I, p62, and LAMP2, hallmarks of resolved lysophagy dysfunction (Fig. 6Y-Z, fig. S12H-K).

Collectively, ED-71 acts as a selective RIS-SMPD1 antagonist, resolving lysosomal phospholipid overload by preventing enzyme mislocalization and catalytic inactivation, which reverses RIS-induced lysosomal dysfunction in osteoblasts and HipGABA-Ns.

ED-71 suppresses ferroptosis via restoring lysosomal membrane integrity in osteoblasts and HipGABA-Ns

To explore ED-71′s protective effects against RIS-induced ferroptosis, we assessed ED-71 pretreatment to reverse RIS-mediated lysosomal and mitochondrial alterations. In osteoblasts and hippocampal neurons, ED-71 pretreatment prior to RIS exposure induced marked improvements in lysosomal morphology and function. LAMP1-mCherry transfection revealed that ED-71 pretreatment reduced RIS-induced lysosomal enlargement (fig. S13A-B). FITC-dextran labeling confirmed lysosomal functional normalization: ED-71 reversed RIS-induced alkaline pH shift and LMP, with restored vesicular integrity and reduced fluorescence diffusion (Fig. 7A-B, E; fig. S13C-D). AO staining showed enhanced lysosomal membrane stability in the MK-801 + RIS + ED-71 group versus the MK-801 + RIS group (Fig. 7C-D, F; fig. S13E-F). IF co-staining demonstrated ED-71 attenuated RIS-induced upregulation of lysosomal damage marker Galectin-3 in osteoblasts and HipGABA-Ns (Fig. 7G-H). Furthermore, ED-71 improved lysosomal iron sequestration, shown by increased Lyso-Tracker/FerroOrange colocalization (Fig. 7I-J; fig. S13G-J).

Fig. 7.

ED-71 Suppresses Ferroptosis via Restoring Lysosomal Membrane Integrity in Osteoblasts and HipGABA-Ns. (A-B) FITC-Dextran staining for lysosomal function in MC3T3-E1 and HT22 cells (n = 5). (C-D) AO staining for LMP detection across cell types (n = 5). (E-F) Quantification of FITC fluorescence intensity and AO-Red/AO-Green ratio in both cell types (n = 5). (G-H) ALP/Galectin-3 and GAD65/Galectin-3 IF co-staining with colocalization in femoral osteoblasts and HipGABA-Ns (n = 5). (I-J) Lyso-Tracker/FerroOrange double staining for lysosomal iron analysis across cell types (n = 5). (K-L) Mito-Tracker/Mito-FerroGreen double staining and colocalization analysis of mitochondrial iron loading in both cell types (n = 5). (M−N) Mito-Tracker/MitoSOX double staining for mitochondrial function assessment with colocalization (n = 3). (O-P) TEM images of mitochondrial morphology in MC3T3-E1 and HT22 cells (arrows: normal, vacuolated, high electron density mitochondria). (Q-R) MDA quantification across cell types (n = 5). (S-T) BODIPY 581/591 C11 staining for lipid peroxidation via reduced-C11/oxidized-C11 fluorescence ratio analysis (n = 5). (U-V) ALP/ACSL4 and GAD65/ACSL4 IF co-staining with colocalization in femoral osteoblasts and HipGABA-Ns (n = 5). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

Mito-Tracker/Mito-FerroGreen double staining indicated reduced mitochondrial iron overload in MK-801 + RIS + ED-71 cells versus MK-801 + RIS-treated cells (Fig. 7K-L; fig. S13K-L). Functional assays revealed ED-71 mitigated RIS-induced mitochondrial ROS overproduction and restored MMP via JC-1 staining (Fig. 7 M−N; fig. S13M-N). Moreover, TEM showed ED-71 reversed RIS-induced mitochondrial ultrastructural abnormalities, including decreased mitochondrial membrane electron density and vacuolization (Fig. 7O-P). Subsequent WB analysis confirmed that combined ED-71 administration alleviated the abnormal changes in ACSL4 and GPX4 levels in primary and cell line-derived osteoblasts and hippocampal neurons under RIS administration (fig. S14A-D). Biochemical analysis demonstrated ED-71 attenuated RIS-induced MDA accumulation, and its effect was reversed by si-SMPD1 transfection (Fig. 7Q-R). Liperfluo staining confirmed the amelioration of ferroptosis-related lipid accumulation in both cell types following combined ED71 administration (fig. S14E-H). BODIPY 581/591 C11 staining confirmed reduced lipid peroxidation in ED-71-treated cells (Fig. 7S-T; fig. S14I-J). IF co-staining further showed ED-71 downregulated ACSL4 in both cell types in vivo (Fig. 7U-V).

Taken together, ED-71 protects against RIS-induced ferroptosis by restoring lysosomal integrity, preventing mitochondrial iron overload, and inhibiting lipid peroxidation, highlighting its therapeutic potential for RIS-associated disorders in osteoblasts and HipGABA-Ns.

ED-71 ameliorates RIS-Induced OP and synergizes with RIS to enhance antipsychotic efficacy

Building on the cellular ferroptosis protective mechanisms of ED-71 in the context of RIS administration, we performed in vivo evaluations of ED-71 on OP and SZ in MK-801-induced SZ mice treated with RIS. For OP, ED-71 restored bone mass in RIS-treated mice of both sexes: H&E staining revealed significantly increased trabecular bone volume, number, and thickness with organized architecture in the ED-71-treated group compared to RIS-administered SZ mice (Fig. 8A; fig. S15A-C). Masson staining confirmed that ED-71 restored bone regeneration in MK-801 + RIS mice (Fig. 8A; fig. S15C). Micro-CT showed significant improvement in femoral bone mass (Fig. 8D). Molecular analyses demonstrated upregulated osteogenic markers (RUNX2, ALP) and downregulated osteoclastic markers (TRAP, CTSK) after ED-71 treatment (Fig. 8B-C). Mechanistically, in the neuroendocrine aspect, the levels of PRL and NE in bone tissue were reduced in ED-71-treated mice (Fig. 8E-F). In direct action, in vitro osteogenic induction experiments showed that ED-71 relieved the osteogenic impairment in the MK-801 + RIS group (fig. S15D), an effect reversed by si-SMPD1 knockdown (Fig. 8G-I).

Fig. 8.

ED-71 Ameliorates RIS-Induced OP and Synergizes with RIS to Enhance Antipsychotic Efficacy. (A) H&E and Masson staining of femurs from female mice (n = 5). (B-C) IHC staining, TRAP staining, and quantitative analysis of RUNX2, ALP, TRAP, and CTSK from female mice (green arrows: IHC positive; red arrows: TRAP positive) (n = 5). (D) Micro-CT imaging of femurs from female mice (2D and 3D). (E-F) ELISA of PRL and NE levels in femoral homogenates (n = 5). (G-I) ALP/ARS staining with quantitative analysis in MC3T3-E1 cells (n = 3). (J-K) OFT trajectories and quantitative analysis from female mice (total distance, speed, immobility time) (n = 6). (L-M) Three-chamber test trajectories and social interaction time analysis from female mice (n = 6). (N-O) MWM trajectories and quantitative analysis from female mice (escape latency, platform crossings) (n = 6). (P) H&E/Nissl staining of hippocampal neurons from female mice (n = 5). (Q) GAD65/SYN IF co-staining and colocalization in HipGABA-Ns from female mice (n = 5). (R-U) ELISA-measured levels of GABA in hippocampus, DA in midbrain, 5-HT in raphe nuclei (all from female mice), and GABA in HT22 cells (n = 5). (V) SYN/Phalloidin IF staining for synaptic activity in HT22 cells with quantification (n = 3). Data are mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, not significant.

For SZ, behavioral tests showed that ED-71 enhanced the therapeutic effects of RIS: combined treatment markedly reduced hyperactivity (OFT, Fig. 8 J-K), notably improved social novelty preference (three-chamber test, Fig. 8L-M), and partially ameliorated spatial memory (MWM, Fig. 8 N-O). Hippocampal protection in the ED-71-treated group of both sexes was manifested as preserved neuronal activity (H&E/Nissl staining, Fig. 8P; fig. S15E), restored colocalization of SYN with HipGABA-Ns (Fig. 8Q), and normalization of neurotransmitter levels—reversing RIS-induced GABA depletion and DA and 5-HT dysregulation (Fig. 8R-T; fig. S15F). In vitro results showed that ED-71 improved the neural function of HipGABA-Ns in the MK-801 + RIS group (Fig. 8U; fig. S15G), an effect abrogated by si-SMPD1 transfection (Fig. 8V).

In conclusion, ED-71 mitigates RIS-induced OP by protecting osteoblasts and enhances antipsychotic effects by restoring hippocampal GABAergic function, highlighting its dual therapeutic potential.

Discussion

The clinical management of SZ has long faced a critical paradox: antipsychotic drugs such as RIS often exhibit insufficient efficacy alongside the adverse effect of OP [58]. This highlights an urgent need to decipher the shared mechanisms underlying RIS-induced cross-tissue toxicity. Our study reveals that RIS induces lysosome-associated ferroptosis in osteoblasts and HipGABA-Ns by targeting the SMPD1-lysosome axis, thereby elucidating for the first time the common cytobiological mechanism of its cross-tissue toxicity. The novel active vitamin D analog ED-71 ameliorates ferroptosis-mediated osteogenic impairment and hippocampal dysfunction by blocking the interaction between RIS and SMPD1, restoring lysosomal phospholipid metabolism, autophagic flux, and lysosomal membrane integrity, while coordinately regulating the neuroendocrine pathway. This provides a dual intervention strategy for addressing both psychosis and OP in clinical practice.

Although RIS is a cornerstone of atypical antipsychotic therapy, its CAD properties have been occasionally mentioned in prior studies but never mechanistically linked to the clinical challenges of OP and incomplete amelioration of psychiatric symptoms [59]. Previous research has primarily attributed RIS-induced OP to hyperprolactinemia or sympathetic overstimulation—both driven by central broad-spectrum neurotransmitter receptor antagonism [10,60]. By comparison, our study verified the partial role of these classical mechanisms but also identified their limitations: on the one hand, hyperprolactinemia’s pathogenic role remains controversial—though some studies link elevated PRL to long-term RIS treatment, others note this effect is transient, and no hypogonadism-associated uterine atrophy was observed [61], suggesting it is not the sole driver of bone damage; on the other hand, this neuroendocrine pathway cannot explain RIS’s direct osteoblast toxicity [62] or the “synchronized bone-hippocampal function deterioration” documented herein. Notably, our dose–response data showed RIS at in vitro concentrations approximating clinical levels induces ferroptosis-related cytotoxicity; conversely, restoring SMPD1 activity or repairing lysosomal function largely abated core RIS-mediated phenotypes (lysosomal iron leakage, mitochondrial damage). Our results collectively reveal a novel paradigm: RIS exerts direct cytotoxicity via the SMPD1-lysosome-ferroptosis axis, independent of traditional systemic hormonal/neural regulation.

For the first time, we demonstrate that RIS, as a CAD, accumulates in acidic lysosomes through electrostatic interactions of its protonated piperidine moiety, forming a high-affinity complex with SMPD1 at the ARG294 residue. This interaction stabilizes SMPD1 in an inactive conformation, blocking hydrolysis of lysosomal SM and PC. Consequently, the resulting phospholipid overload triggers lysosomal membrane fluidization, which in turn induces excessive autophagic initiation and autophagolysosomal degradation impairment. This subsequently leads to LMP and iron leakage, followed by mitochondrial iron accumulation and the generation of hydroxyl radicals (・OH) via the Fenton reaction—ultimately triggering ferroptosis in osteoblasts and HipGABA-Ns. Dose-response experiments showed that the dose-dependent exacerbation of RIS-induced toxicity and the presence of a saturation plateau mirror the kinetic characteristics of phospholipid accumulation, while WT SMPD1 plasmid rescue experiments further validated this mechanism.

Osteoblasts and HipGABA-Ns, despite divergent functions, share a high dependence on lysosomal homeostasis: the former requires lysosomal exocytosis for matrix mineralization [63], while the latter relies on lysosomal recycling for synaptic vesicle turnover [64]. RIS-induced lysosomal iron leakage disrupts iron balance in both cell types. Notably, this “lysosome-iron-ferroptosis” pathway aligns with the findings of a study identifying lysosomal iron as a universal trigger of ferroptosis [20]. In osteoblasts, the ferroptotic cascade inhibits BMP signaling pathway and downregulates collagen-related proteins, thereby impairing osteoblast differentiation and extracellular matrix mineralization [65]. In the hippocampus, induction of ferroptosis triggers a significant reduction in GABA levels, consequently leading to neurocognitive impairment, particularly in learning and memory functions [66]. This “double hit” manifests in animal models as synchronous declines in bone volume fraction and hippocampal GABA levels, mechanistically explaining the comorbidity of suboptimal symptom control and bone loss during RIS therapy.

Guided by the RIS-SMPD1 interaction pocket, we performed virtual drug screening to identify ED-71, a novel competitive antagonist of RIS-SMPD1 binding, and validated its dual regulatory mechanisms based on the pharmacological properties of this novel active vitamin D analog. First, ED-71 directly rescues lysosomal function: virtual screening showed that ED-71 occupies the RIS-binding pocket of SMPD1, forming stronger hydrogen bonds (ΔGbinding = −8.336 kcal/mol vs. RIS’s −6.576 kcal/mol). At 1 nM, ED-71 restores SMPD1 activity, significantly reducing lysosomal SM and PC levels while normalizing lysosomal membrane fluidity. This blocks the “phospholipid overload-LMP-ferroptosis” cascade, improving osteogenic function and hippocampal GABA levels. Additionally, as a specialized active vitamin D analog, ED-71 regulates neuroendocrine pathways: our results show ED-71 decreases PRL levels and sympathetic overstimulation, mitigating hyperprolactinemia- and sympathetic-driven osteogenic impairment. Notably, hyperprolactinemia correlates with cognitive deficits linked to hippocampal structural changes, and hippocampal parvalbumin-positive GABAergic neurons are PRL targets [67]; thus, ED-71′s PRL regulation may further protect GABAergic neurons and restore hippocampal GABA levels—an indirect protective effect warranting future investigation. This interplay between direct cellular rescue and neuroendocrine regulation enables ED-71 to synergize with RIS, achieving dual efficacy in treating psychosis while preventing OP.

To further validate the core SMPD1-lysosome-ferroptosis axis, we conducted parallel experiments in male mice. Results showed BMD, hippocampal neuron count, and hippocampal GABA levels decreased significantly in male MK-801 + RIS mice—mirroring trends in female mice, with only weaker effect magnitudes. Consistent with clinical findings that female patients are more sensitive to RIS-related metabolic and somatic side effects [68], these results confirm the mechanism’s cross-gender universality and validate our initial use of female mice—their higher drug sensitivity clarifies intervention outcomes, and alignment with females’ higher clinical osteoporosis incidence enhances conclusion relevance.

Our meta-analysis revealed widespread 25(OH)D deficiency in SZ patients on antipsychotic therapy, indicating insufficient active vitamin D synthesis and supporting the population applicability of ED-71 supplementation [42,[69], [70], [71], [72], [73]]. Compared to traditional vitamin D, ED-71 offers advantages including a longer half-life, closer integration with VDR, enhanced target tissue effects, and unique mini-modeling osteogenic activity [21,74,75]. In animal models, the combination therapy of ED-71 and RIS significantly restored both psychiatric symptoms and bone mass, demonstrating “psycho-skeletal” dual-protective clinical potential. ED-71′s dual capabilities to activate SMPD1 and regulate iron homeostasis establish it as a prototype for “lysosomal medicine”—a paradigm shift from single-gene targeting to organelle-centered trans-disease pathway intervention.

Existing evidence suggests the SMPD1-related mechanism has broad cross-drug and cross-disease generalizability. In antipsychotics, studies confirm SMPD1′s functional links to multiple agents: haloperidol significantly inhibits SMPD1 [76], with perphenazine and fluphenazine showing similar effects [77]. For common ones like olanzapine and quetiapine, direct SMPD1 interaction research is limited, but recent studies show they alter sphingolipid levels [78]. As a key sphingolipid metabolism rate-limiting enzyme, SMPD1 likely participates in their actions via sphingolipid homeostasis regulation—implying the SMPD1-lysosome axis may be a common target for antipsychotic side effects. In psychiatric disorders, SMPD1′s potential role has been preliminarily explored: cerebral SMPD1 dysregulation is seen in depression rodent models [79]. Meanwhile, abnormal sphingolipid metabolism is closely tied to bipolar disorder (BD) and major depressive disorder pathogenesis [80]. Given this, SMPD1 may play a key role in these disorders’ development. Notably, beyond psychiatry, the SMPD1-lysosome-ferroptosis axis provides new insights into other diseases: lysosomal storage diseases (such as Niemann-Pick disease, caused by sphingomyelin accumulation due to SMPD1 mutations [81]) and neurodegenerative diseases (such as Parkinson’s disease, characterized by lysosomal iron overload [82]). Thus, based on the targeted regulatory properties of ED-71 on the SMPD1-lysosome axis identified in this study, ED-71 may possess broad-spectrum application potential across different drugs and diseases—representing a key direction for in-depth exploration in the future.

However, the present study still has certain limitations: first, it lacks validation with human primary cells or clinical samples. Second, it does not investigate the effects of RIS on other cell types, with studies only focusing on osteoblasts and HipGABA-Ns—limiting comprehensive understanding of RIS’s cross-organ toxicity regulatory network. Third, in vivo validation lacked tissue/cell-specific conditional SMPD1 knockout/knock-in strategies. To address these gaps, future work will supplement human-relevant validation, broaden research scope, and develop cell-lineage-specific SMPD1 models to dissect functional differences—ultimately refining the SMPD1-lysosome-ferroptosis axis network and strengthening clinical translation support.

In summary, this study establishes the SMPD1-lysosome axis as a therapeutic hub for dissociating the efficacy and toxicity of antipsychotic drugs. By integrating direct cellular rescue and systemic regulation, ED-71 provides proof-of-concept for “dual-target” precision medicine, addressing both psychiatric and skeletal complications in SZ. Beyond offering an integrated “symptom control-side effect prevention” strategy for SZ, this dual intervention paradigm pioneers the field of lysosomal medicine, underscoring its potential to transform the treatment of complex multi-organ diseases.

Additional information

Supplementary information Supplementary Materials and Methods; Table S1; figure S1 to S16.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: