GRAMD1B is a regulator of lipid homeostasis, autophagic flux and phosphorylated tau

Abstract

Lipid dyshomeostasis and tau pathology are present in frontotemporal lobar degeneration (FTLD) and Alzheimer's disease (AD). However, the relationship between lipid dyshomeostasis and tau pathology remains unclear. We report that GRAM Domain Containing 1B (GRAMD1B), a nonvesicular cholesterol transporter, is increased in excitatory neurons of human neural organoids (HNOs) with the MAPT R406W mutation. Human FTLD, AD cases, and PS19 tau mice also have increased GRAMD1B expression. We show that overexpression of GRAMD1B increases levels of free cholesterol, lipid droplets, and impairs autophagy flux. Modulating GRAMD1B in iPSC-derived neurons also alters key autophagy-related components such as PI3K, phospho-AKT, and p62, as well as phosphorylated tau, and CDK5R1. Blocking GRAMD1B function decreases free cholesterol and lipid droplets. Knocking down GRAMD1B additionally reduces phosphorylated tau, and CDK5R1 expression. Our findings elucidate the role of GRAMD1B in the nervous system and highlight its relevance to FTLD and AD.

Fig. 1. Characterization of HNOs with the MAPT R406W mutation.

a HNOs were cultured using previously reported methods for unguided differentiation. Scale bars: iPSC: 100 µm; EB: 50 µm; HNOs: 330 µm. b HNO growth was assessed by measuring the longest diameter of the HNO (n = 10 HNO for all except n = 9 for Day 70 HOM HNO). Two-way ANOVA with Tukey post hoc test *P = 0.037 (WT vs. HET), **P = 0.009 (WT vs. HOM), ***P < 0.001 (Day 80: WT vs. HOM). WT: CRISPR-Cas9 edited isogenic control; HET: MAPT R406W/R406R heterozygous; HOM: MAPT R406W/R406W homozygous. Data are shown as mean ± SEM. c, d scRNA-seq data of HNOs at D72 and D120. Two independent experiments were performed. Each independent experiment consisted of two individual samples per genotype for a total of six samples per independent experiment (12 samples for both independent experiments). Each individual sample contained single-cell suspensions from 5 human neural organoids. UMAP of (c) broad-cell type annotations and (d) sub-cell type annotations. Ex: Excitatory Neurons; Inhibitory: inhibitory neurons; RG: Radial Glia; IPC: Intermediate Progenitor Cells; OPC: Oligodendrocyte Progenitor Cells; Unc: Unclassified cells. e, f Cell-type abundance plots indicating the percentage of each broad-cell type or sub-cell type at (e) D72 and (f) D120 of maturation. Rep 1: Replicate 1; Rep 2: Replicate 2; Replicate: n = 5 HNOs/replicate. g IF staining of HNOs at D60 vs. D120 indicating changes in cell-type populations corresponding to progenitor cells: SOX2, PAX6, Ki67; Post-mitotic neurons: TBR1; Deep-layer neurons: CTIP2; Upper-layer neurons: SATB2; Neuronal marker: NEUN; Inhibitory neuronal marker: GAD1; Total tau protein: TAU; Astrocyte marker: GFAP; Oligodendrocyte lineage cell marker: OLIG2. Nuclei are stained with Hoechst33342 dye (blue). Scale bars 50 µm, Insets 10 µm. Source data are provided as a Source Data file.

Fig. 2. MAPT R406W HNOs exhibit increased phosphorylated tau.

a–d IF staining of HNOs at (a) D60 and (c) D120 for total tau protein (TauC) and phosphorylated tau (PHF1). Nuclei were stained with Hoechst33342 dye. Scale bars, 50 µm. b Quantitation of (a) n = 3 biological replicates for each group, WT, HET, and HOM HNO. NS: Not statistically significant; One-way ANOVA. d Quantitation of (c) *P = 0.0280 (WT (n = 10 biological replicates) vs. HOM (n = 3 biological replicates)), **P = 0.0041 (WT (n = 10 biological replicates) vs. HET (n = 9 biological replicates)), Mann-Whitney, two-tailed P-value. e–h D60 Protein expression by WB of (e) PHF1, (f) total tau (TAU), and housekeeping protein GAPDH. Each lane corresponds to an individual HNO. TL: tau ladder; MWL: molecular weight ladder. g Quantitation of TauC/GAPDH from panel (f) n = 6 biological replicates for each group, WT, HET, and HOM HNO. ***P = 0.0005 (WT vs. HOM). **P = 0.0013 (HET vs. HOM), NS: Not statistically significant; One-way ANOVA. h Quantitation of the ratio of PHF1 to TAU levels, n = 6 biological replicates for each group, WT, HET, and HOM HNO. ***P = 0.0004 (HET vs. HOM). *P = 0.0121 (WT vs. HOM), NS: Not statistically significant; One-way ANOVA. i–l D120 Protein expression of WT (n = 4 biological replicates), HET (n = 4 biological replicates), or HOM (n = 4 biological replicates) HNOs by WB assay of (i) PHF1, (j) TAU, and housekeeping protein GAPDH. Each lane corresponds to an individual HNO. TL: tau ladder; MWL: molecular weight ladder. k Quantitation of TauC/GAPDH from panel (i) NS: Not statistically significant; One-way ANOVA. l Quantitation of the ratio of PHF1 to TAU levels, *P = 0.03860.05 (WT vs. HET), NS: Not statistically significant; One-way ANOVA. All data in (b, d, g, h, k, l) are shown as mean ± SEM, where each data point represents an individual organoid. Source data are provided as a Source Data file.

Fig. 3. Functional analysis of MAPT R406W HNOs reveals decreased neuronal and burst activity in mutant HNOs.

a Live imaging of HNOs placed on microelectrode array (MEA) chips for the recording of local field potentials (LFPs). b Recorded LFP traces of WT (black) and mutant HET (blue) HNOs at D75, top panel; Spike analysis highlighting detected spikes from traces, bottom panel. c Representative mean waveforms of WT vs. mutant HET HNOs. Electrophysiological (EP) Metrics were calculated for WT (gray) and HET (blue) HNOs from MEA data at Day 75, including (d) mean firing rate (n = 8 WT HNO, and 11 HET HNO), *P = 0.033 (WT vs. HET), Mann-Whitney test, two-tailed P-value. e number of spikes, (n = 8 WT HNO, and 11 HET HNO), *P = 0.016 (WT vs. HET), Mann-Whitney test, two-tailed P-value, and (f) interburst intervals (n = 4 independent experiments for WT HNO and HET HNO) P = 0.200, Mann-Whitney test, two-tailed P-value. g Representative Synchronized Burst Firing Activity of WT and HET HNOs. Arrows indicate a decreased array-wide spiked detection rate (ASDR) in HET HNOs, while double-speared arrows indicate the time between synchronized burst firing is increased in HET HNOs. All data are shown as mean ± SEM. Source data are provided as a Source Data file.

Fig. 4. Differential gene expression analysis of MAPT R406W HNOs reveals lipid related genes are changed in excitatory neuronal population.

a Heat map of top 24 DEGs in the excitatory neuron population. Lipid-related genes are shown highlighted in black. b Heat map of gene set enrichment analysis indicating top pathways changed in mutant HNOs. Lipid-related pathways are shown highlighted in black. c Representative images of RNAscope HiPlex smFISH in the cortical plate region of WT, HET, and HOM HNOs indicating mRNA transcripts of NEUN (green) and lipid-related genes GRAMD1B (purple), VEGFB (red), and NR2F2 (orange). Nuclei are shown in gray as indicated by Hoechst33342 dye. Scale bars 20 µm. d Representative cells and dilated ROI that are selected for quantification of mRNA counts. Scale bars 5 µm. e, f Individual mRNA counts of individual cells in the cortical plate region for (e) GRAMD1B, (n = 4 HNOs per genotype) One-way ANOVA (WT vs. HET (****P < 0.0001); WT vs. HOM (****P < 0.0001), HET vs. HOM (****P < 0.0001)), (f) VEGFB, (n = 4 HNOs per genotype) One-way ANOVA (WT vs. HET (****P < 0.0001); WT vs. HOM (****P < 0.0001), HET vs. HOM (*P = 0.0268)), and (g) NR2F2. (n = 4 HNOs per genotype) One-way ANOVA (WT vs. HET (****P < 0.0001); WT vs. HOM (****P < 0.0001), HET vs. HOM (****P < 0.0001)). e–g Data are shown as mean ± SEM. Source data are provided as a Source Data file.

Fig. 5. MAPT R406W HNOs reveal altered cholesterol and lipid species at early timepoints.

a Experimental setup. Each sample consists of 1 mL conditioned medium from one well containing 3 HNOs. 16 samples were analyzed (n = 4 replicates per group at Day 60 and Day 120). Created in BioRender. Kim, T. (2025) https://BioRender.com/p38a040. b Lipid population as a percentage of total lipids analyzed in HNOs at D60. c, d Volcano plots of lipid populations identified in WT (n = 3 HNO) vs. HET (n = 4 HNO) culture medium at (c) Day 60 and (d) Day 120 with an FDR of 5% (shown as a red line). True discoveries that increase (red) or decrease (blue) in culture medium HET vs WT HNOs are shown. e–i Changes in cholesteryl ester species (CE), free cholesterol (CHL-free), Lysophosphatidylcholine (LPC), phosphatidylcholine (PC), and phosphatidylethanolamine (PE) in WT (gray) and HET HNOs (pink) at D60 and D120 (e–i) *P < 0.05, **P < 0.01, ****P < 0.0001, 2-Way ANOVA. Exact p-values are provided in the Source Data File for each graph. j–n MAPT R406W Day 120 HNOs were probed for changes in (j) GRAMD1B (n = 9 WT HNO, n = 10 HET HNO) P = 0.0172 (WT vs. HET), Mann-Whitney, two-tailed P-value, k free cholesterol by filipin dye (n = 7 HNOs per group) P = 0.0070 (WT vs. HET), Mann-Whitney, two-tailed P-value, (l) lipid droplets by LipidSpot488 (n = 30 cells from 3 HNO, for WT and HET groups) P = 0.0005 (WT vs. HET), Mann-Whitney, two-tailed P-value, m HMGCS1 (n = 5 HNO per group) P = 0.0102 (WT vs. HET), t test, two-tailed P-value, and (n) SREBP2 (n = 4 HNO per group) P = 0.1493 (WT vs. HET), t test, two-tailed P-value. Representative regions are shown on the right. Quantitation is shown on the left. Scale bars: j–l 50 μm, inset 5 μm; m, n 10 μm, inset 2 μm. Nuclei are stained with Draq5 and shown in magenta (k) or Hoechst33342 dye and shown in blue (j, l–n). e–i, j–n Data shown as mean ± SEM. Source data are provided as a Source Data file.

Fig. 6. GRAMD1B is altered in human FTLD and coincides with increased free cholesterol and lipid droplets.

a Representative images of multiplex IF in human post-mortem brain of control (CT) vs. FTLD cases. Nuclei are stained with Hoechst33342 dye (white), Filipin staining for free cholesterol is shown in blue, Oil-Red-O staining for lipid droplets is shown in red, GRAMD1B protein is shown in green, phosphorylated tau (S396/S404) PHF1 is shown in yellow, and excitatory neurons (SATB2 +) are shown in teal. Merged multiplexed images are shown in MERGE. Scale bar 50 μm. White arrows indicate cells with low PHF1 (PHF1-). Yellow arrows indicate cells with high PHF1 (PHF1 + ). b Correlation plots of PHF1 with GRAMD1B in human cases. For each data point, the mean intensity of 10 individual regions of interest were averaged for each human case. (n = 8 human cases; 4 control and 4 FTLD). ** P < 0.01, r = 0.9136, Pearson correlation coefficient. c Quantification of Filipin, Oil-Red-O, and GRAMD1B in human FTLD and CT cases, each data point represents an individual human case (n = 4 cases for each group CT and FTLD) P = 0.0286 (GRAMD1B in CT vs. FTLD; P = 0.0286 (Filipin in CT vs. FTLD); P = 0.0286 (Oil Red O in CT vs. FTLD), Mann-Whitney Test, two-tailed P-value. d Quantitation of Filipin, Oil-Red-O, and GRAMD1B in cells with accumulation of phosphorylated tau (PHF1 +) vs. cells without phosphorylated tau (PHF1-) (n = 20 cells/case x 4 FTLD cases). P < 0.001 (GRAMD1B in PHF1- vs. PHF1 +), P < 0.001 (Filipin in PHF1- vs. PHF1 +), P < 0.01 (Oil Red O in PHF1- vs. PHF1 +), Mann Whitney Test, two-tailed P-value. e Correlation plots of PHF1 with free cholesterol as measured by Filipin, lipid droplets as measured by Oil-Red-O, and GRAMD1B. ****P < 0.0001, Pearson correlation coefficient. f Correlation of GRAMD1B with cholesterol as measured by Filipin and lipid droplets as measured by Oil-Red-O (n = 20 cells/case x 4 FTLD cases), ****P < 0.0001, Pearson correlation coefficient. b, d Data are shown as mean ± SEM. Source data are provided as a Source Data file.

Fig. 7. GRAMD1B modulates autophagic flux and tau phosphorylation in vitro.

a Proposed mechanism in which GRAMD1B can alter autophagy and tau phosphorylation. b Representative images of GRAMD1B (green), PI3K (teal), p-AKT (yellow), p62 (white), CDK5R1 (orange), and 12E8 (red) immunostaining in iPSC-derived neurons treated with control (CONTROL) vs. GRAMD1B overexpression (GRAMD1B OE) vs. GRAMD1B shRNA knockdown (GRAMD1B SH) lentivirus. Nuclei are stained with Hoechst33342 dye (blue). Scale bar: 10 μm. c–h Quantitation of immunostaining shown in (b). c (n = 83 CT cells; 61 OE cells; 82 SH cells), d (n = 67 CT cells; 64 OE cells; 63 SH cells), e (n = 83 CT cells; 113 OE cells; 101 SH cells, f (n = 67 CT cells; 87 OE cells; 95 SH cells) (g) (n = 70 CT cells; 63 OE cells; 67 SH cells), h (n = 83 CT cells; 60 OE cell; 82 SH cells). Three independent experiments were performed. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, Mann-Whitney Test, two-tailed P-value. Exact p-values are provided in the Source Data File. i Representative images of autophagy reporter FUW-mCherry-GFP-LC3 in iPSC-derived neurons with CONTROL, GRAMD1B OE, or GRAMD1B SH in the presence of 400 nM BAFA1 for 12 hrs. Co-staining with p62 is shown in white. Scale bar 5 μm. j, k Quantification of yellow punctate (j) and p62 punctate (k). j n = 44 CT cells; 51 OE cells; 46 SH cells; k n = 42 CT cells; 32 OE cells; 35 SH cells. j, k Three independent experiments were performed. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, One-Way ANOVA. Exact p-values are provided in the Source Data File. l, Ratio of red (autolysosome) to yellow (autophagosome) puncta in n = 24 CT cells; 35 OE cells; 24 SH cells. P = 0.0081 (CT vs. OE); P = 0.0002 (CT vs. SH), Mann-Whitney Test, two-tailed P-value. c–h, j–l Data are shown as mean ± SEM. Source data and exact p-values are provided in the Source Data File.

Fig. 8. AI-3d Inhibitor of GRAMD1B reduces lipid dyshomeostasis, autophagy, and tau phosphorylation.

a, i Representative images of immunostaining of (a) WT and (i) HET HNO treated with DMSO (left panels) or with 3 mM of AI-3d compound (right panels) for one week. HNOs were probed with filipin (blue), lipid droplets (green), and nuclei were stained with Draq5. In addition, HNOs were probed for PI3K (green), p-AKT (green), p62 (magenta), 12E8 (red), and CDK5R1 (red), nuclei are stained with Hoechst33342 dye. Quantitation of HNOs for (b, j) free cholesterol by filipin, c, k lipid droplets by LipidSpot 488, d, l PI3K, e–m p-AKT, f, n p62, (g, o) 12E8, and (h, p) CDK5R1. For WT HNO’s: b, c n = 4 CT; 5 AI-3d HNO; d–f, h n = 4 CT; 4 AI-3d HNO; g n = 5 CT; 5 AI-3d HNO; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, (CT vs. AI-3d), t test. For HET HNO’s: j, l–n, p n = 4 CT; 4 AI-3d HNO, *P < 0.05, **P < 0.01, ***P < 0.001, t test, (CT vs. AI-3d); k n = 4 CT; 4 AI-3d HNO, P = 0.0571 (CT vs. AI-3d), Mann-Whitney, two-tailed P-value; o (n = 3 CT; 3 AI-3d HNO) P = 0.0323 (CT vs. AI-3d), t test, two-tailed P-value. Exact p-values are provided as a source data. a Scale bar, 20 µm; (i) Scale bar, 10 µm. q–t Effect of the Aster inhibitor, AI-3d, on the number of excitatory neurons (CTIP2-positive cells (red, q, s), and SATB2-positive cells (red, r, t)) in HNOs. Scale bar, 50 µm. Nuclei were stained by Hoechst33342 (blue). Quantification was performed with QuPath. (q) n = 6 CT; 5 AI-3d HNO, (r) n = 6 CT; 4 AI-3d HNO, (s) n = 4 CT; 5 AI-3d HNO, (t) n = 4 CT; 4 AI-3d HNO, N.S: not significant, t test, two-tailed P-value. b–h, j–t Data are shown as mean with SEM. Exact p-values and source data are provided as a Source Data file.