Modulating glycosphingolipid metabolism and autophagy improves outcomes in pre-clinical models of myeloma bone disease

Abstract

Patients with multiple myeloma, an incurable malignancy of plasma cells, frequently develop osteolytic bone lesions that severely impact quality of life and clinical outcomes. Eliglustat, a U.S. Food and Drug Administration-approved glucosylceramide synthase inhibitor, reduced osteoclast-driven bone loss in preclinical in vivo models of myeloma. In combination with zoledronic acid, a bisphosphonate that treats myeloma bone disease, eliglustat provided further protection from bone loss. Autophagic degradation of TRAF3, a key step for osteoclast differentiation, was inhibited by eliglustat as evidenced by TRAF3 lysosomal and cytoplasmic accumulation. Eliglustat blocked autophagy by altering glycosphingolipid composition whilst restoration of missing glycosphingolipids rescued autophagy markers and TRAF3 degradation thus restoring osteoclastogenesis in bone marrow cells from myeloma patients. This work delineates both the mechanism by which glucosylceramide synthase inhibition prevents autophagic degradation of TRAF3 to reduce osteoclastogenesis as well as highlighting the clinical translational potential of eliglustat for the treatment of myeloma bone disease.

Fig. 1. Eliglustat increases trabecular bone in healthy mice by inhibiting OC in vivo.

a Representative micro-CT reconstruction images of tibiae from 8-week-old female C57BL/6J mice treated with normal chow (Ctr, n = 6 biologically independent animals) or eliglustat (Elig, n = 5 biologically independent animals) chow for 19 days. b–g Micro-CT analysis of tibiae: BV/TV, BS, Tb.N, Conn.D, BS/BV, and Tb.Sp (Ctr, n = 6 and Elig, n = 5 biologically independent animals). h Representative TRAP/0.2% methyl green stained tibial sections showing OCs (red/purple) and OBs (black arrowhead) on the endocortical bone surface from each group (left); magnified boxed areas in images on the right. Result is representative of two independent experiments. i–k Bone histomorphometry quantification of OCs includes the percentage of OC surface over total bone surface (Oc.S/BS), number of OCs over total bone area (N.Oc/T.Ar), and number of OCs per bone perimeter (N.Oc/B.Pm) (Ctr, n = 6 and Elig, n = 5 biologically independent animals). l–n Bone histomorphometry quantification of OB including percentage of OB surface over total bone surface (Ob.S/BS), number of OBs over total bone area (N.Ob/T.Ar) and number of OBs per bone perimeter (N.Ob/B.Pm). Ctr, n = 6 and Elig, n = 5 biologically independent animals, each dot is one animal. Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using unpaired two-tailed Student’s t-test for b–g and i–n. Source data are provided as a Source Data file.

Fig. 2. Eliglustat ameliorates 5TGM1-GFP MM cell-induced bone disease.

5TGM1-GFP MM cells were injected to 8-week-old female C57BL/KaLwRijHsd mice to generate the MM model. Eliglustat chow was administered from day 4 post tumor injection (until day 23). a Representative micro-CT reconstruction images of tibiae trabecular bones from naive control (Ctr, n = 5 biologically independent animals), MM mice with normal chow (MM, n = 7 biologically independent animals), or MM mice with eliglustat chow (MM + Elig, n = 7 biologically independent animals). b–h Tibiae trabecular bone parameters were assessed: BV/TV, BS, Tb.N, Conn.D, Tb.Th, BS/BV, and Tb.Sp (Ctr, n = 5; MM, n = 7, and MM + Elig, n = 7 biologically independent animals). i, j Tibiae cortical bone reconstruction (i) and the number of cortical bone lesions (j) (Ctr, n = 5; MM, n = 7, and MM + Elig, n = 7 biologically independent animals). k–n Representative TRAP/0.2% methyl green staining showing red OCs of tibial histological sections with original magnification ×40 (k) and the quantification of OCs with Oc.S/BS, N.Oc/T.Ar, and N.Oc/B.Pm (l–n) (Ctr, n = 5; MM, n = 7, and MM + Elig, n = 7 biologically independent animals). The result is representative of two independent experiments. Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using One-way ANOVA. Source data are provided as a Source Data file.

Fig. 3. Eliglustat reduces bone loss in a diet-induced obesity MGUS model.

a Schematic overview and experiment design. Four-week-old female C57BL/6J mice were divided into MM injection with normal diet (MM, n = 4 biologically independent animals), HFD with MM (HFD + MM, n = 10 biologically independent animals), and HFD with MM plus eliglustat (HFD + MM + Elig, n = 10 biologically independent animals). b Body weight (g) was measured over the duration of the experiment. Statistical analysis was performed by comparing with MM group (MM, n = 4; HFD + MM, n = 10, and HFD + MM + Elig, n = 10 biologically independent animals). c Representative micro-CT reconstruction images of tibiae trabecular bones from respective groups. d–k Micro-CT analysis parameters of tibiae: BV/TV, BS, Tb.N, Conn.D, Tb.Th, BS/BV, Tb.Sp, and trabecular BMD (MM, n = 4; HFD + MM, n = 10, and HFD + MM + Elig, n = 10 biologically independent animals). l, m Osmium tetroxide staining to detect regulated BMAT (in red box) by micro-CT in MM, HFD + MM, and HFD + MM + Elig groups (MM, n = 4; HFD + MM, n = 10, and HFD + MM + Elig, n = 10 biologically independent animals). n–q Representative TRAP/0.2% methyl green stained tibial histological sections showing red OCs (n). The result is representative of two independent experiments. Bone histomorphometry parameters including Oc.S/BS (o), N.Oc/T.Ar (p), and N.Oc/B.Pm (q) (MM, n = 4; HFD + MM, n = 10, and HFD + MM + Elig, n = 10 biologically independent animals). Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using One-way ANOVA. Source data are provided as a Source Data file.

Fig. 4. Eliglustat combined with ZA reduces MM bone disease with greater effect than either agent alone.

a Schematic illustrating the timeline and experimental design of eliglustat and ZA impact on bone measurements in 8-week-old C57BL/KaLwRijHsd 5TGM1-GFP MM-bearing male mice. b Representative images of micro-CT reconstruction of proximal tibiae from each group, including naive control (Ctr, n = 6 biologically independent animals), MM mice (MM, n = 7 biologically independent animals), MM mice with eliglustat chow (for 19 days) (MM + Elig, n = 7 biologically independent animals), MM mice with single dose ZA injection (0.01 mg/kg, MM + ZA, n = 7 biologically independent animals), or MM mice with single dose ZA injection (0.01 mg/kg), and eliglustat chow (for 19 days) (MM + ZA + Elig, n = 7 biologically independent animals). c–h Dot plots of BV/TV, BS, BS/BV, Tb.N, Conn.D, and Tb.Sp from each group (Ctr, n = 6; MM, n = 7; MM + Elig, n = 7; MM + ZA, n = 7; MM + ZA + Elig, n = 7 biologically independent animals). i Representative reconstruction images of proximal cortical bone from each group. j Bone lesions (holes) on the cortical bones from each mouse were counted (Ctr, n = 6; MM, n = 7; MM + Elig, n = 7; MM + ZA, n = 7; MM + ZA + Elig, n = 7 biologically independent animals). k Representative TRAP/0.2% methyl green stained tibial sections showing red OCs on the endocortical bone surface from each group. The result is representative of two independent experiments. l–n Bone histomorphometry parameters including Oc.S/BS, N.Oc/T.Ar, and N.Oc/B.Pm (Ctr, n = 6; MM, n = 7; MM + Elig, n = 7; MM + ZA, n = 7; MM + ZA + Elig, n = 7 biologically independent animals). Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using One-way ANOVA for c–h and l–n, and two-tailed t-test for j. Source data are provided as a Source Data file.

Fig. 5. Eliglustat inhibits OC formation in a TRAF3-dependent manner.

a RAW264.7 cells were differentiated into OCs with 50 ng/ml M-CSF and 75 ng/ml RANKL. Different doses of eliglustat (0.1, 1, 10, 25, 50 μM) were present throughout the culture period and OCs were identified by TRAP staining on day 5. Scale bar represents 200 μm. Result is representative of four independent experiments. b–e Number of OCs per well, number of nuclei per TRAP-positive cell, % OC area over total area and Alamar blue viability assay were quantified (n = 3 independent experiments). f Schematic graph illustrating binding of RANKL to its receptor activates the TRAF6-dependent canonical NFκB pathway, leading to proteasome-mediated IκBα degradation and nuclear translocation of p65 and p50. In the non-canonical NFκB pathway, binding downregulates TRAF3 via autophagy/lysosome-mediated degradation, induces nuclear translocation of p52 and RelB. Panel f is adapted from “NF-KB Signaling Pathway” and “Autophagy Process”, by BioRender.com (2022). Retrieved from https://app.biorender.com/biorender-templates. g–i RAW264.7 cells treated with M-CSF, RANKL, and eliglustat for 1 day. TRAF3 protein levels quantified by western blot and mRNA levels quantified by qRT-PCR. Treatment with BafA1 for the last 2 h (n = 6 independent samples). j–m Primary murine BM cells treated with RANKL together with BafA1 or MG132 or eliglustat for 2 h. TRAF6 (j, k) and IκBα (l, m) levels quantified by western blot (n = 8 biologically independent samples). n Lethally irradiated 8-week-old female CD45.1⁺ B6.SJL mice were reconstituted with littermate (WT) or myeloid specific TRAF3 knockout (LysM-Cre + , Traf3 ^fl/fl) BM cells. After 19 days of treatment with eliglustat, bone volume of tibiae was quantified by micro-CT (WT, n = 4; WT + Elig, n = 4; traf3 KO, n = 5; traf3 KO + Elig, n = 5 biologically independent animals). The experiment is representative of two independent experiments. Results from g, j, and i are representative of three independent experiments. Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file.

Fig. 6. Eliglustat is an autophagy inhibitor.

a–d RAW264.7 cells were treated with RANKL and M-CSF with or without Elig for 24 h. 10 nM BafA1 added 2 h before protein harvest. LC3II (a, b) and p62 (c, d) levels were quantified by western blot. (n = 7 independent samples for DMSO control, Elig, BafA1, and Elig with BafA1). e, f BafA1, CQ and Elig increased LC3 puncta in U2OS cell transfected with GFP-LC3 plasmid after 2 h treatment (n = 43 cells for DMSO control, n = 17 cells for BafA1, n = 15 cells for CQ, n = 46 cells for Elig group). g–j Elig increased GFP/RFP ratio (h; n = 46 cells for DMSO control, n = 42 cells for BafA1, n = 38 cells for CQ, and n = 77 cells for Elig group) and GFP + RFP + puncta number (j; n = 31 cells for DMSO control, n = 42 cells for BafA1, n = 39 cells for CQ, and n = 20 cells for Elig group) in RFP-GFP-LC3 transfected U2OS cell after 6 h treatment. k, l Electron microscope images of Elig induced autolysosome (red arrow) formation in RAW264.7 cells after 2 h treatment (25,000×) (n = 23 cells for DMSO control, n = 13 cells for BafA1, n = 19 cells for CQ, and n = 15 cells for Elig group) (l). m, n TRAF3 and LAMP1 western blot for lysosome-enriched layer from the control group (DMSO), BafA1 group (10 nM, 2 h treatment), CQ group (20 μM, 12 h treatment), and Elig group (50 μM, 12 h treatment) (m); TRAF3 protein level was quantified based on the lysosome marker LAMP1, n = 12 independent samples (n). o, p Confocal images of TRAF3 and LAMP2 in Pre-OCs derived from RAW264.7 cells were treated with BafA1, CQ, and Elig for 60 min, Scale bar represents 10 μm, n ≥ 4 (o); the percentage of TRAF3 and LAMP2 co-localization is quantified (n = 6 cells for DMSO control, n = 9 cells for BafA1, n = 9 cells for CQ, and n = 7 cells for Elig group) (p). Results from a, c, e, g, k, m, and o are representative of three independent experiments. Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file.

Fig. 7. Eliglustat blocks autophagy by inhibiting the GlcCer and LacCer in RAW264.7 cells.

a Membrane fluidity during early OC formation. RAW264.7 cells were treated with RANKL and M-CSF together with eliglustat (25 µM or 50 µM) for 12 h and stained with NR12S dye. Representative images from DMSO group and eliglustat group at 50 µM. Histograms of GP distribution illustrate ordered red end (1) and disordered blue end (−1). Scale bar 10 μm. b Quantification for GP value was evaluated (n = 19 control cells, n = 19 RANKL + M-CSF, n = 19 DMSO + RANKL + M-CSF, n = 25 RANKL + M-CSF + 25 μM Elig, n = 28 RANKL + M-CSF + 50 μM Elig group). c Schematic of lipid, GSL and ganglioside metabolism. Eliglustat blocks GlcCer formation from ceramide, subsequently blocking formation of LacCer, Gangliosides and other GSLs (including globosides, isoglobosides, and [neo]lacto-series). d LTQ-ESI-MS glycosphingolipidomics profile of RAW264.7 cell from DMSO group and eliglustat group (50 µM, 12 h) with 1:1 sample mix for change of specific GSL composition (blue dash line indicates DMSO, red dash line indicates eliglustat) (m/z means mass charge ratio). e, f GlcCer (cluster of molecular ion peaks between m/z 807 and 935; including C16, C22, C24) and LacCer (cluster of molecular ion peaks between m/z 1011 and 1150; including C16, C22, C24) aggregate abundance in DMSO, eliglustat and D-PDMP groups (e). Overall GSLs including GlcCer, LacCer and Lc3 (f); n = 3–6 independent samples per group. g, h RAW264.7 cells were treated with 25 µM eliglustat and 1 µM C16 LacCer, 1 µM C16 GlcCer, or 1 µM C24 LacCer for 24 h (g); LC3II levels were quantified by western blot, n = 8 independent samples per group (h). i, j RAW264.7 cells were treated with RANKL and M-CSF in addition to eliglustat and GSLs for 24 h (i). TRAF3 levels were quantified by western blot; RANKL + M-CSF, n = 14; RANKL + M-CSF + 25 µM Elig, n = 11; RANKL + M-CSF + 25 µM Elig + 1 µM C16 LacCer, n = 15; RANKL + M-CSF + 25 µM Elig + 1 µM C16 GlcCer, n = 14; RANKL + M-CSF + 25 µM Elig + 1 µM C24 LacCer, n = 13 independent samples (j). Results from a, d, g, and i are representative of three independent experiments. Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file.

Fig. 8. Eliglustat inhibits OC formation from MM patients while exogenous GSLs reverse eliglustat inhibition.

a BM mononuclear cells from MM patients were seeded at 1.5 × 10⁵ cells/well into 48-well-plate and differentiated to OCs with 50 ng/ml human M-CSF and 75 ng/ml human RANKL in αMEM media. BM mononuclear cells were treated with increasing doses of eliglustat (0.1, 1, 10, 25 μM) and OCs were identified by TRAP staining on day 11. b–d Number of OCs per filed of view, number of nuclei per TRAP + cell and % OC area over total area were quantified. n = 9 patients. e Eliglustat was applied to the BM mononuclear cells from MM patients with or without C16 LacCer, C16 GlcCer, and C24 LacCer during in vitro OC formation process and mature OCs were recognized by TRAP staining. f–h Number of OCs per filed of view, number of nuclei per TRAP + cell and % OC area over total area were quantified. Scale bar represents 200 μm. n = 9 patients, each with 5–7 replicates. i Schematic graph illustrating the overall proposed mechanism. Binding of RANKL to RANK receptor induces the degradation of TRAF3 via autophagy, which allows NF-kB signaling to occur; this is required for OC formation. Eliglustat reduces the amount of GlcCer and LacCer, preventing autophagic flux, thereby reducing TRAF3 degradation and OC formation. Results from a and e are representative of 3 independent experiments with 9 patient samples. Data are presented as mean values ± SEM. Exact p-values are depicted in the figure. Statistical analysis was performed using one-way ANOVA. Source data are provided as a Source Data file.