INPP5D regulates inflammasome activation in human microglia

Abstract

Microglia and neuroinflammation play an important role in the development and progression of Alzheimer's disease (AD). Inositol polyphosphate-5-phosphatase D (INPP5D/SHIP1) is a myeloid-expressed gene genetically-associated with AD. Through unbiased analyses of RNA and protein profiles in INPP5D-disrupted iPSC-derived human microglia, we find that reduction in INPP5D activity is associated with molecular profiles consistent with disrupted autophagy and inflammasome activation. These findings are validated through targeted pharmacological experiments which demonstrate that reduced INPP5D activity induces the formation of the NLRP3 inflammasome, cleavage of CASP1, and secretion of IL-1β and IL-18. Further, in-depth analyses of human brain tissue across hundreds of individuals using a multi-analytic approach provides evidence that a reduction in function of INPP5D in microglia results in inflammasome activation in AD. These findings provide insights into the molecular mechanisms underlying microglia-mediated processes in AD and highlight the inflammasome as a potential therapeutic target for modulating INPP5D-mediated vulnerability to AD.

Fig. 1. INPP5D expression in the human brain is restricted to microglia in the human brain.

Human brain sections (25 μm) (a) and cultured iMGs (b) were immunostained for IBA1 and INPP5D. Nuclei are visualized with DAPI and preparations were imaged using confocal microscopy. Images representative of 16 human brain samples and over three iMG differentiations analyzed. Scale bars = 50 μm. c–e UMAP plots of sNucRNAseq of iMGs combined with snRNAseq of dorsolateral prefrontal cortex (dlpfc) from 12 human postmortem brain samples using Harmony to integrate across datasets. Single nucleus data from human brain samples and iMGs are depicted separately in (c) and (d), respectively. Relative INPP5D expression (after applying the SCTransformation in the Seurat package) across the harmonized iMG and human brain single nucleus samples shown in (e). Data are from 7121 iMG nuclei, 55,671 nuclei from the postmortem human brain (39,239 glutamatergic neurons, 14,958 astrocytes, 1474 microglia). f–i Data from microglia subcluster from the snRNAseq in c and d were isolated and re-clustered to examine microglial subsets. UMAP plots are shown in (f–h). Relative INPP5D expression (z-score of log-transformed, normalized data) across the subclusters is shown in (h). INPP5D was significantly lower in cluster 13 (adj. p-val = 8.0 × 10⁻²²⁶, Wilcoxon rank-sum test, with FDR correction). A dot plot showing a subset of the genes that define cluster 13 are shown in (i). See also Supplementary Fig. 1 and Supplementary Data 1 showing enrichments in all microglial clusters.

Fig. 2. Complex post-translational regulation of INPP5D in the AD brain.

a–c Representative images of immunostaining of human brain sections for INPP5D, IBA1, and Aβ. DNA is stained with DAPI. Scale bars = 10 μm. b, c Magnified image from white box outlined in (a). d–h Quantifications from brain immunostaining from 8 NCI and 8 AD individuals. d Quantification of intensity of INPP5D immunostaining across human subjects, two-sided t-test, p = 0.0009. e Quantification of INPP5D intensity across all microglia quantified between the 16 human subjects. Two-sided Wilcoxon test, p < 2.2e−16. f Quantification of INPP5D levels in microglia that were either plaque-associated (plaque) or not plaque-associated (none). Two-sided Wilcoxon test, p < 2.2e-16. For d–f, boxes mark the 1st and 3rd quartiles of data (25th and 75th percentiles), surrounding the median. Whiskers extend to furthest values up to 1.5xIQR (inter-quartile range or distance between 1st and 3rd quartiles). Data beyond the whiskers are “outlying” points for the purpose of plotting and extend to the min and max of individual datasets. g Correlations between form factor and INPP5D intensity across categories of microglia; regression lines for individual datasets as well as the r- and p-value statistics, represent the output of linear model calculation in R. Shaded regions represent 95% CI limits for each dataset. h Histogram of the relative frequency of IBA1+ microglia containing diffuse versus punctal INPP5D immunostaining intensity, determined by the difference between the upper quartile of INPP5D intensity and the lowest quartile of INPP5D intensity. Two-way, unpaired t-tests. i Representative WB of TBS extracts from postmortem human brain, and INPP5D WT and KO iMGs for INPP5D and GAPDH. Vertical line shows INPP5D bands quantified. GAPDH portion of the blot containing iMG samples were digitally over-exposed for presentation purposes. j, k Western blot quantification of INPP5D and GAPDH in TBS postmortem brain (prefrontal cortex, PFC) brain extracts using two antibodies to INPP5D. Kruskal–Wallis test with Dunn’s multiple comparisons test. Quantification of peptides mapping to INPP5D (Q92835) in ROSMAP brain samples extracted with urea (l–n) or TBS (o) and quantified using TMT-MS, urea TMT-MS data generated and reported in. For l, n = 99 AD, 121 HP-NCI, and 101 LP-NCI human subjects. For o, n = 10 AD, 8 HP-NCI, and 12 LP-NCI human subjects. Kruskal–Wallis test with Dunn’s multiple comparisons tests performed in (l–o). p. Schematic summarizing data presented in this figure. For all panels: ns = not significant, *p < 0.05, ***p < 0.001, ****p < 0.0001. LP = no neuropathological diagnosis of AD, HP = neuropathological diagnosis of AD, NCI = not cognitively impaired, AD = clinical and neuropathological diagnosis of AD. n = number of human subjects and is represented by dots in j, k, m, n For all graphs, Data are presented as mean values ± SEM. See also Supplementary Data 9–13 and Supplementary Data 2, 3 for full datasets. See also Supplementary Figs. 2–4.

Fig. 3. INPP5D inhibition induces changes in pathways associated with innate immunity.

a, b iMGs were treated with vehicle (ethanol) or 3AC (1.25 μM) for 6 h. a Cells were then lysed, RNA purified, and RNAseq performed, n = 6 per condition. Volcano plot of DEGs comparing vehicle and 3AC treatment conditions. b Cells were lysed in urea and TMT-MS performed. n = 4 per condition. Volcano plot of DEPs; For a, b differential expression was calculated using a linear modeling with empirical Bayesian statistical analysis using the limma package in R. All p-values are adjusted using the Benjamini Hochberg (BH) procedure. c Heatmap of expression of DEGs that showed concordant differential expression at the RNA and protein level. d Heatmap of relative expression of DEPs encoded by LOAD GWAS candidate genes between vehicle and 3AC-treated iMGs. e. Relative RNA and protein levels of CD33 and PTK2B between vehicle and 3AC-treated microglia, as quantified by RNAseq and TMT-MS. Mean ± SEM. Two-sided Welch’s t-test. f, g. A variety of microglial subtypes previously have been defined through single-cell sequencing, with specific subtypes implicated to be altered in AD brain and model systems^–. Volcano plots comparing 3AC vs vehicle treatment for genes defining these subtypes within the transcriptomic and proteomic datasets are shown; significance determined by BH FDR < 0.05. h Heatmap of relative protein-levels of scavenger receptors in vehicle and 3AC-treated iMGs. i, j. Enriched GO terms of biological processes for the differentially expressed genes that are elevated with 3AC treatment (geneontology.com); Fisher’s exact test, Bonferroni multiple comparison’s test, adj p-value as shown by size of circles. j Heatmap of relative expression between vehicle and 3AC treatment of DEGs associated with NFκB signaling and JAK-STAT signaling. k Relative abundance measures of RNA or protein levels of inflammasome related components in vehicle and 3AC-treated iMGs. Mean ± SEM. Two-sided Welch’s t-test. l Representative western blot of iMGs treated with either vehicle (ethanol) or 3AC (1.25 μM) for 6 h showing protein expression of INPP5D and inflammasome related proteins: CASP1, PYCARD/ASC, GSDMD and NLRP3, as well as GAPDH. Similar WB results obtain in 3 separate differentiations. Cell lines used for experiments and full datasets for can be found in Supplementary Data 4 and 5, Supplementary Data 14. For all graphs, the number of biological replicates is represented by dots. For all panels: **p < 0.01, ***p < 0.001, ****p < 0.0001; ns = p < 0.05. See also Supplementary Figs. 5 and 6.

Fig. 4. Acute INPP5D inhibition results in inflammasome activation and an increase in the secretion of IL-1ß and IL-18 in iMG cultures.

a Fold change of secreted cytokines from iMGs treated with LPS (100 ng/mL) or 3AC (1.25 μM) for 6 h across iMGs derived from four iPSC lines, assayed by ELISA. Fold change was calculated compared to vehicle-treated cells. b Levels of IL1B following 6 h treatment with vehicle, LPS (100 ng/mL), or 3AC (1.25 μM, 2.5 μM) treatment as measured by quantitative real-time PCR (qPCR). Data are normalized to GAPDH expression then values were normalized to vehicle treatment within each experiment. n = 3 differentiations with well-level data shown as dots. One-way ANOVA with Dunnett’s T3 multiple comparisons test. c Secreted levels of IL-1ß measured for the same experiments as (b) following 6 h treatment as measured by ELISA. IL-1ß levels were normalized to 1.25 μM 3AC-treated conditions within each experiment. n = 3 differentiations with well-level data shown as dots. d Representative western blot of protein expression in iMGs with biallelic loss-of-function mutations generated with CRISPR targeting. Loss of INPP5D protein observed repeatedly over 3 differentiations. e, f WT INPP5D iMGs and KO INPP5D iMGs were treated with 3AC for 6 h and the levels of IL-1ß and IL-18 were measured by ELISA. Multiple two-sided t-tests, Two-stage step-up (Benjamini, Krieger, and Yekutieli), **q < 0.01; ***q < 0.005; ****q < 0.001; LLOD= lower limit of detection. n = 3 biological replicates. g, h Treatment of iMGs with either vehicle or 3AC (5 μM) with either VX-765 (25 μM) or its vehicle (DMSO). Levels of secreted IL-1ß and IL-18 measured via ELISA following treatments and normalized to 3AC-treated samples in each experiment. One-way ANOVA with Sidak’s multiple comparison test. 3 differentiations, n = 10 per condition. i, j iMGs were treated with either vehicle, primed with LPS (100 ng/mL) for 3 h and then treated with nigericin (10 μM) for 1 h, or treated with 3AC (1.25 μM or 5 μM) for 2 h. Cells were immunostained for ASC and IBA1, DNA is stained with DAPI (i) and imaged using confocal microscopy. Scale bars = 100 μm. j Quantification of ASC specks. Number of cells analyzed: n = 226 (vehicle); 151 (1.25 μM 3AC); 173 (5.0 μM 3AC) over three experiments and 10 images per condition. Kruskal–Wallis test to determine significance. k, l Treatment of iMGs with either vehicle or 3AC (5 μM) with either MCC950 (10 μM) or its vehicle (DMSO). Levels of secreted IL-1ß and IL-18 in the media were measured and normalized to 3AC-treated samples in each experiment. One-way ANOVA with Sidak’s multiple comparison test. n = 4 differentiations with well-level data shown as dots. m A summary figure of experiments performed to interrogate inflammasome activation, created using BioRender.com. For all graphs, data are presented as mean values ± SEM. For b, c, g, h, j, l: ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Cell lines used for experiments are detailed in Supplementary Data 8. See also Supplementary Figs. 7 and 8.

Fig. 5. Generation of INPP5D heterozygous iMGs using CRISPR targeting.

a Table of CRISPR-Cas9 generated INPP5D heterozygous (HET) and monoclonally selected wild-type (WT) lines. b INPP5D RNA levels measured by qPCR, normalized to GAPDH levels. Two-tailed Mann–Whitney test. n = 3 differentiations with replicate wells as shown by dots. c Western blot quantification of INPP5D protein levels in INPP5D WT and HET iMGs normalized to GAPDH levels; Two-tailed Mann–Whitney test. n = 3 differentiations with replicate wells as shown by dots. d Representative western blot (of three differentiations) of INPP5D, GAPDH, and IBA1 protein levels in INPP5D WT and HET iMGs. e INPP5D WT and HET iMGs were lysed in urea and TMT-MS performed, n = 4 WT, n = 4 HET. Relative protein levels of AIF1, P2RY12, CX3CR1, C1QA; Unpaired two-sided t-test. f Representative images (of three independent differentiations) of INPP5D WT and HET cells immunostained for IBA1 and P2RY12. DNA is stained with DAPI. Imaged using confocal microscopy. Scale bars = 200 μm. g Phase contrast images of INPP5D WT and HET iMGs. h Viability was determined by measuring the level of LDH secretion normalized to total measure of LDH following cell lysis. n = 3. Two-sided Mann–Whitney test: ns = no significance (p > 0.05). i. Quantification of the cell yield following differentiation to iMG fate comparing INPP5D WT or HET lines, 2-sided paired t-test, n = 3. For b, c, e: ns not significant p > 0.05, **p < 0.01. For all graphs, data are presented as mean values ± SEM.

Fig. 6. Chronic reduction in INPP5D levels results in inflammasome activation and reduction in autophagic flux.

a–c Volcano plot (a) of adjusted p-values showing 71 DEPs from the 7868 quantified via TMT-MS (two-way t-test with Benjamini–Hochberg multiple comparisons test; FDR < 0.05. Heatmap (b) of relative expression levels of proteins involved in immune signaling that are differentially up and downregulated in INPP5D WT vs HET iMGs. Asterisks indicate proteins that also were differentially up or downregulated (q < 0.05) with 3AC treatment (Fig. 3). GO analysis (c) of DEPs; Fisher’s exact test, Bonferroni multiple comparison’s test, adj p-value as shown by size of circles. d, e Representative western blot and quantification across three differentiations of MRC1, PLA2G7, COLEC12, IBA1, and GAPDH in INPP5D WT and HET iMGs. n = 3 independent experiments, mean ± SEM; two-sided t-test. f Heatmap of TMT-MS data (z-score) for lysosomal DEPs in INPP5D HET vs WT iMGs. g Representative western blot of INPP5D iMGs treated with either vehicle (DMSO) or bafilomycin (baf, 100 nM) for either 24 or 6 h. Western blots are probed for INPP5D and LC3 with GAPDH as a loading control. h Quantification of autophagic flux (baf - veh of LC3-II/GAPDH) in iMGs with treatment of 24 h of baf. n = 6 wells vehicle, n = 6 wells of 6 h baf. Mean ± SEM, two-sided t-test with Welch’s correction. i, j Secreted IL-1ß and IL-18 measured by ELISA of media collected from INPP5D HET and WT iMGs. Media were concentrated tenfold in order to be above the limit of detection and normalized to WT mean. n = 14 wells WT, n = 21 wells HET. Mean ± SEM. Two-sided Mann–Whitney test. k, l Secreted IL-1ß and IL-18 measured by ELISA of media collected from INPP5D HET iMGs treated with either vehicle (DMSO) or MCC950 (10 μM) for 24 h. Media were concentrated tenfold for detection. n = 10 wells vehicle, n = 10 wells MCC950-treated. Mean ± SEM. Mann–Whitney test. For e, h–l *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. See also Supplementary Fig. 9, Supplementary Data 4 and 5 for full datasets.

Fig. 7. Evidence linking INPP5D levels and inflammasome activation in the human brain.

a IL-1ß and IL-18 levels were measured from human TBS brain lysates and normalized to total protein. Spearman correlation coefficients were calculated between IL-1ß and IL-18 levels (normalized to total protein) and INPP5D/GAPDH levels (AD, HP-NCI, LP-NCI, or all). b–d Human brain tissues (DLPFC, BA9; 8 NCI, 8 AD) were fixed, sectioned, and immunostained for ASC, IBA1 and Aβ. These sections were derived from the same batch as used in Fig. 2h–m to quantify INPP5D levels. Example immunostaining of human brain sections (6 microns) for IBA1, ASC, and Aβ is shown in (b). DNA is stained with DAPI. Imaged using confocal microscopy. Images are representative of data obtained from brain sections from 16 individuals. Scale bars = 10 μm. Spearman correlation coefficients were calculated comparing the percent of microglia containing ASC specks and the average intensity of INPP5D immunostaining by human subject, analyses performed separately on AD and NCI subjects (c, d). e WT iMGs were transduced with either an empty lentivirus plus VSV-G virus-like particles (VLPs), a lentivirus encoding an INPP5D overexpression cassette (OE) plus VLP, or not transduced, and 72 h later the cells were lysed, RNA purified and the levels of INPP5D RNA quantified by qPCR with normalization to GAPDH. n = 4 biologically independent samples; data are presented as mean values ± SEM. One-way ANOVA with Dunnett’s multiple comparisons test; **p < 0.01, ***p < 0.001. f, g From the same cells, levels of IL-1ß and IL-18 were measured in conditioned media following 72 h of viral transduction, data are presented as mean values ± SEM. n = 4 biologically independent samples, two-sided t-test. h RNAseq was performed on samples cultured in parallel, volcano plot of DEGs observed with INPP5D overexpression is shown. RNAseq also was performed on iMGs derived from iPSC lines with biallelic INPP5D loss-of-function mutations and their isogenic WT controls. Volcano plot of DEGs is shown in i. Pseudobulk snRNAseq data of microglia^, was analyzed to identify DEGs between microglia from AD and NCI brain tissue. These data were generated from brain tissue (DLPFC, BA9) from 131 AD and 162 NCI ROSMAP participants. Volcano plot comparing AD to NCI brain tissue is shown in (j). GSEA was performed using the REACTOME database from DEG analysis performed in (h–j). k Table of pathways that were enriched in any two of the three comparisons in (h–j). See Supplementary Data 16–19 and Supplementary Data 6,7 for complete DEG and GSEA results. l Only one pathway was enriched in all three datasets. This pathway relates to NFκB signaling and shown is the gene concept network of leading-edge genes in the INPP5D iMG loss-of-function comparison.

Fig. 8. Loss of one copy of INPP5D in microglia has functional consequences on gene expression profiles of co-cultured neurons.

a Co-cultures of iNs with either INPP5D WT or HET iMGs were fixed and immunostained for TUJ1 and INPP5D. Representative images are shown for at least three independent experiments. DNA is stained with DAPI. Imaged using confocal microscopy. Scale bars = 100 μm. b Clusters generated from single cell sequencing of neuron and microglia co-cultures. c Percent composition of each cluster based upon genotype and cell type is shown. d, e Microglial and neuronal clusters were defined by marker expression. f Top DEGs between cell groups were identified using the Wilcoxon Rank Sum test (2-sided) with a multiple comparisons adjusted (FDR) p-value cutoff of 0.05, and LogFC threshold of 0.25. g APOE levels in the media (as measured by ELISA) of iN and iMG co-cultures (left) or of INPP5D WT or HET iMGs, with treatment of het iMGs with NLRP3 inhibitor for 24 h (right). n = 4 biologically independent samples (left graph) and n = 12, 26, 7 (left to right) biologically independent samples (right graph); data are presented as mean values ± SEM. Two-tailed paired t-test (left graph) and Brown-Forsythe and Welch’s ANOVA with Dunnett’s multiple comparisons test (right graph). h DEGs were identified between cell groups using the Wilcoxon Rank Sum test (two-sided) with a multiple comparisons adjusted (FDR) p-value cutoff of 0.05, and LogFC threshold of 0.25. Top DEGs are shown in (h. i). Scatter plots of top six DEGs in iNs co-cultured with WT or HET iMGs. j, k GSEA was performed using the full dataset of iN gene expression. Gene-concept networks of leading-edge genes for the top pathways identified through GSEA are shown. Upregulated pathways in iNs cultured with HET iMGs were related to synaptic function (j) and those downregulated with co-culture with HET iMGs were associated with mRNA splicing and ribonucleoprotein complexes. See also Supplementary Figs. 10–12 and Supplementary Data 21–24 for full datasets. For g: *p < 0.05, ****p < 0.0001.