Plaque-associated human microglia accumulate lipid droplets in a chimeric model of Alzheimer’s disease

In conclusion, our results indicate that human iPSC-derived DAM microglia adopt a transcriptomic signature in AD mice that mimics that of atherosclerotic foam cells and accumulate LDs, which are reduced in TREM2-R47H mutant microglia. Interestingly, in vitro the R47H mutation exacerbates microglial lipid accumulation [37]. However, our data indicate that in vivo, LD accumulation occurs particularly in the DAM-subset and is thus dependent on microglial reactivity and proximity to plaques, which is impaired by the R47H mutation. Moreover, lipid droplet accumulation was also observed in plaque-associated microglia in postmortem AD brains, further supporting the use of this chimeric approach to model human disease. Whether plaques induce LD accumulation in human microglia via a direct (amyloid uptake) or indirect manner (localized plaque-induced toxicity and cell death) or a combination of the two, remains unknown. However, the observation that LD accumulation occurs secondary to plaque proximity in vivo suggests that it will be important to determine whether TREM2 stimulating antibodies or other TREM2 activating approaches that increase microglial association with plaques [43], could in turn affect LD accumulation in xMGs. In this respect, accumulation of LDs in aged microglia impaired their functionality, and in vitro pharmacological inhibition of lipid droplet accumulation was able to restore microglial phagocytic capacity [34]. Altogether, the observation that both TREM2 wild-type and mutant human DAM xMGs accumulate LDs around plaques further demonstrates the importance of microglial lipid metabolism in AD and suggests that it will be of great future interest to determine how inhibition of LD accumulation and/or induction of microglial lipid metabolism can influence Alzheimer’s pathology and perhaps uncover promising new therapeutic avenues.

Further information and requests for resources and reagents should be directed to and will be fulfilled by Mathew Blurton-Jones (mblurton@uci.edu).

All animal procedures were conducted in accordance with the guidelines set forth by the National Institutes of Health and the University of California, Irvine Institutional Animal Care and Use Committee. The M-CSFh mouse line (hCSF1) was purchased form Jackson Laboratories (stock # 017708) and includes deletion of Rag2 and Il2rγ and humanized M-CSFh [44], which are necessary for human microglial engraftment. The 5xFAD-hCSF1 (5X-hCSF1) model was created by backcrossing the M-CSFh mouse with the well-established 5xFAD transgenic model which overexpress co-integrated transgenes for Familial Alzheimer’s Disease (FAD) mutant APP (Swedish, Florida, and London) and mutant FAD PS1 (M146L and L286V) [45].

Maintenance of all iPSC lines involved culturing in feeder-free conditions in complete TeSR™-E8™medium (Stemcell Technologies), in a humidified incubator (5 % CO2, 37 °C), with medium changed daily. Passaging was performed every 7–8 days using ReLeSR (Stemcell Technologies) and cells were plated onto 6-well plates (Corning), coated with growth factor-reduced Matrigel (1 mg/mL; BD Biosciences), in TeSR™-E8™medium, supplemented with 0.5µM Thiazovivin (Stemcell Technologies) for the first 24 h post-passage. The pluripotency, karyotype, and sterility of all iPSC lines was confirmed via trilineage differentiation (Stem Cell Tech.), array comparative genomic hybridization (aCGH performed by Cell Line Genetics), and MycoAlert (Lonza) testing.

The GFP-expressing iPSC line was purchased from Coriell (AICS-0036 GFP line) and was generated by CRISPR modification of the parental WTC11 line to insert mEGFP into the AAVS1 safe harbor locus (chromosomal location 19q13.4-qter) under the control of a CAGG promoter. More information on this line can be found here: https://www.coriell.org/0/PDF/Allen/ipsc/AICS-0036-006_CofA.pdf↗. We used CRISPR-Cas9 modification to then introduce the TREM2-R47H homozygous mutation into the AICS-0036 GFP line (Additional File 2): 2 × 10⁵ induced pluripotent stem cells were isolated following Accutase enzymatic digestion for 3 min at 37 ^oC. Cells were resuspended in 100 µL nucleofection buffer from Human Stem Cell Nucleofector™ Kit 2 (Lonza). The suspension was combined with 2µM ssODN template (IDTDNA) and 50 µg of RNP complex formed by incubating Alt-R® S.p. HiFi Cas9 Nuclease V3 (IDTDNA) with fused crRNA:tracrRNA (IDTDNA) duplex for 15 min at 23 ^oC. The suspension was transferred to the Amaxa Nucleofector cuvette and transfected using program B-016. Cells were plated in TeSR™-E8™ (STEMCELL Technologies) media with 0.25 µM Thiazovivin (STEMCELL Technologies) and CloneR™ (STEMCELL Technologies) supplement overnight to recover. Cells were digested the following day with Accutase and mechanically single-cell plated to 96-well plates in TeSR™-E8™ media with 0.25 µM Thiazovivin and CloneR™ for clonal isolation and expansion. Genomic DNA was extracted using Extracta DNA prep for PCR (Quantabio) from a sample of each clone upon passage and amplified for sequencing using Taq PCR Master Mix (ThermoFisher Scientific) at the cut site. PCR product from promising clones was transformed using TOPO™ TA Cloning™ Kit for Subcloning, with One Shot™ TOP10 (ThermoFisher Scientific) for allele-specific sequencing. Comparison of GFP TREM2-R47H with its isogenic control allowed the analysis of the effect of the homozygous TREM2-R47H mutation independent of possible genetic modifiers.

iPSC-derived Hematopoietic Progenitor Cells (HPCs) were differentiated according to our published protocol [46]. To begin HPC differentiation, iPSCs were passaged in mTeSR-E8 to achieve a density of 40–60 colonies per 6-well. On day 0, cells were transferred to Medium A from the STEMdiff™ Hematopoietic Kit (Stem Cell Technologies). On day 3, flattened endothelial cell colonies were transferred to Medium B and cells remained in medium B for 7 additional days while HPCs began to lift off the colonies. On day 11, non-adherent CD43 + HPCs were collected by removing medium and cells with a pipette. At this point, HPCs can be frozen in Bambanker (Wako) and stored in liquid nitrogen at a concentration of 4 million cells/mL. Cells used for early-postnatal HPC transplantation were thawed and resuspended at 50 K cells/µL in 1X DPBS (low Ca²⁺, low Mg²⁺).

P1 to P2 5X-hCSF1 mice were placed in a clean cage over a heating pad with a nestlet from the home cage to maintain the mother’s scent. Female pups were then placed on ice for 2–3 min to induce hypothermic anesthesia. Free-hand transplantation was performed using a 30-gauge needle affixed to a 10µL Hamilton syringe, mice received 1µL of HPCs suspended in sterile 1X DPBS at 50 K cells/µL at each injection site (8 sites) totaling 400 K cells/pup. Bilateral injections were performed at 2/5th of the distance from the lambda suture to each eye, injecting into the lateral ventricles at 3mm and into the overlying anterior cortex at 1 mm, and into the posterior cortex in line with the forebrain injection sites, and perpendicular to lambda at a 45° angle. Transplanted pups were then returned to their home cages and weaned at P21. For further details and validation of this chimeric approach please see: [38].

All steps were performed on ice or at 4 °C with ice cold reagents and all centrifuge steps were performed for 10 min at 400xg with full brake and acceleration unless otherwise stated. Anesthetized mice were intracardially perfused with 1X DPBS, half brains were dissected, the cerebellum was removed, and tissue was stored in RPMI 1640 until subsequent perfusions were completed. Brains were manually homogenized using a 7mL Dounce homogenizer by adding 4mL of RPMI 1640 and performing 10 strokes with the “loose” pestle followed by 10 strokes with the “tight” pestle. Samples were then run through a pre-soaked 70 μm filter and the filter was washed with 10 mL of RPMI 1640. The sample was pelleted by centrifugation and myelin was removed by resuspension in 30 % Percoll overlaid with 2 mL of 1X DPBS centrifuged at 400xg for 20 min with acceleration and brake set to 0. The myelin band and supernatant were discarded and cell pellets were resuspended in 80 µL MACS buffer (0.5 % BSA in 1X DPBS) + 20 µL Mouse Cell Removal beads (Miltenyi) and incubated at 4 °C for 15 min. Magnetically labelled mouse cells were separated using LS columns and the MidiMACs separator (Miltenyi) while the unlabeled human cells were collected in the flow through. Human cells were then pelleted by centrifugation and dead cells were magnetically removed using the Dead Cell Removal kit, Annexin V (Stem Cell Technologies) by resuspending the pellets in 100 µL of buffer (2 % BSA + 1mM CaCl₂ in 1X PBS) in 5 mL polystyrene round-bottom tubes and the following manufacturer protocol. Live cells were centrifuged, resuspended in 50–100 µL of MACS buffer, and concentrations were determined by counting on a hemocytometer. Final cell concentrations were then adjusted to 900-1,000 cells/µL.

scRNA-seq library preparation was performed according to the 10X Genomics Chromium Single Cell 3’ Reagents kit v3 user guide except that sample volumes containing 25,000 cells were loaded onto the 10X Genomics flow cell in order to capture ~ 10,000 total cells. The 10X Genomics workflow was then followed according to the manufacturer protocol and libraries were pooled at equimolar concentrations for sequencing on an Illumina NovaSeq 6000, targeting ~ 50,000 reads per cell. FASTQ files were aligned to both the human GRCh38 transcriptome (Ensembl release 95; [47]) using the CellRanger v3.0.2 count command, with the expected cells set to 10,000 and no secondary analysis performed.

UMI count tables were read into Seurat (v3) [48] for preprocessing and clustering analysis (Additional File 1). Initial QC was performed by log normalizing and scaling (default settings) each dataset followed by PCA performed using all genes in the dataset. Seurat’s ‘ElbowPlot’ function was used to select principal components (PCs) to be used for clustering along with a resolution parameter of 0.35 and clusters identified as being doublets, gene poor, or dividing were removed from the dataset prior to downstream analysis. Secondary QC cutoffs were then applied to retain only cells with less than 27.5 % ribosomal genes, 12.5 % mitochondrial genes, greater than 500 genes but less than double the median gene count, and greater than 500 UMI but less than double the median UMI count. Additionally, subsequent analysis identified a small cell population (179 cells; Additional File 1) primarily present in only a single sample, which were removed as the cluster did not appear to be biologically relevant.

Cells passing QC for each sample were then merged using Seurat’s ‘merge’ function and datasets were processed using Seurat’s integrated analysis workflow [49]. In short, samples from individual mice were integrated using the ‘FindIntegationAnchors’ and ‘IntegrateData’ commands using dimensions 1:25. Datasets were then scaled and sources of technical variation were regressed out (library size differences, percent ribosomal genes, and percent mitochondrial genes) and PCA was performed using Seurat’s ‘RunPCA’ command. A shared nearest neighbor (SNN) plot was generated using Seurat’s ‘FindNeighbors’ function using PCs 1:15 as input, clustering was performed using the ‘FindClusters’ function and a resolution parameter of 0.3, and dimension reduction was performed using the ‘RunUMAP’ function with the same PCs used for generating the SNN plot. Differentially expressed genes were determined between clusters using the ‘FindAllMarkers’ function, which employs a Wilcoxon Rank Sum Test, with and FDR cutoff of 0.01, an LFC cutoff of 0.25, and the requirement that the gene be expressed in at least 10 % of the cluster and clusters were labeled according to manual curation of the differential gene lists. Differentially expressed genes between genotypes across all cells were identified using the “FindMarkers” function in Seurat 3.2.1, using a log-FC cutoff of 0.001 and adjusted p-value < 0.01 (Fig. 1E, H, I). The y-axis of the violin plots depicted in Fig. 1I feature counts per cell divided by the total counts for the cell and multiplied by a scale factor of 10,000, then natural-log transformed using log1p, using the “NormalizeData” function in Seurat.

Animals were administered Euthasol and monitored for loss of consciousness. Once animals no longer responded to toe pinch, mice were intracardially perfused with ice cold 1X DPBS. If xMGs were being isolated from half brains, the remaining half brain was drop fixed in 4 % (w/v) PFA for 48 hours, otherwise, the mice were intracardially perfused with 4 % PFA and post-fixed for 24 hours. Samples were then cryoprotected in 30 % (w/v) sucrose until the tissue sank in the solution. Brains were then cut coronally at a section thickness of 40um on a sliding microtome cooled to -79°C. Tissue sections were collected as free-floating sections in PBS with 0.05 % sodium azide. For staining, tissue was blocked for 1 hour in 1X PBS, 0.2 % Triton X-100, and 10 % goat serum. Immediately following blocking, sections were stained for amyloid plaques using the UV ThioflavinS analog Amylo-Glo (Biosensis, TR-300-AG, 1:100) for 20’ at RT on a shaker in the dark. Next, sections were placed in primary antibodies diluted in 1X PBS with 1 % goat serum and incubated overnight on a shaker at 4 °C (GFP-ch, Millipore AB16901, 1:500; CD9-ms clone HI9a, Biolegend 312,102, 1:200; APOE-Rb, Thermofisher PA5-27088, 1:1000; LAMP1-Rt, Abcam ab25245, 1:200; PLIN2-Gp, Fitzgerald 20R-AP002; 1:500, HLA-DR-ms (Invitrogen 14-9956-82; 1:200), P2RY12-Rb (Novus Biologicals NBP233870; 1:400, 1 h RT). Sections were then incubated in conjugated secondary antibodies (AlexaFluor Antibodies, Life Technologies, 1:400) for 1 h, before washing and mounting on microscope slides. Immunofluorescent sections were then visualized and captured using an Olympus FV3000 confocal microscope. In some cases, brightness and contrast settings of confocal images were slightly adjusted to reveal fine structures and morphology. Importantly, no adjustments were made to any images used for quantification. Human AD brain tissue from the hippocampus used in this project were provided by the University of California Alzheimer’s Disease Research Center (UCI-ADRC) and the Institute for Memory Impairments and Neurological Disorders. AD brain sections were stained using Iba1-Rb (Wako, 1:200), PLIN2-Gp (Fitzgerald 20R-AP002, 1:200) and Amylo-Glo (Biosensis, TR-300-AG, 1:100).

HPCs derived from isogenic wild-type and TREM2-R47H GFP iPSCs (Coriell, AICS-0036) were transplanted into 5X-hCSF1 female pups. 7-months later, immunohistochemistry was performed, and confocal Z-stacks collected within the retrosplenial granular cortex (proximity to plaques, plaque area, CD9 area, APOE area, LAMP1 area, HLA-DR area and P2RY12 area) and within the dorsal aspect of the subiculum and CA1 (PLIN2 area and CD9 area) at 40x magnification using identical confocal settings between TREM2 genotypes (n = 4 mice (TREM2 xMGs) n = 3 mice (TREM2-R47H xMGs) with 3–6 images per mouse). To study proximity to plaques human microglia locations were detected and quantified through GFP immunofluorescence within 10 μm per plaque (Amylo-Glo, blue) using the Cellsense software on the Olympus FV3000. Ordinary one-way Anova was performed to confirm similar distribution of samples within a given genotype (Additional File 4A-B) in order to combine all values per genotype in a singular distribution and perform Welch’s t test (Fig. 3C). Plaques near the borders of the image were excluded, and for plaques within 10 μm of each other, only the plaque with the highest number of xMGs was included for both genotypes. Plaque area of the plaques included for xMG counts show no significant difference between genotypes (Additional File 4C) and also after normalization of the number of xMGs within 10 μm per plaque to plaque area, the significant reduction in the number of R47H-mutant xMGs remains (Additional File 4D). IMARIS-based quantification of CD9, APOE, LAMP1, PLIN2, HLA-DR, P2RY12 and total plaque load was performed using the “Surfaces” function and were measured by the sum of surfaces for each image. To determine the number of PLIN2⁺ GFP xMGs, the “classification” function of Imaris was used to measure the number of GFP surfaces that overlap with PLIN2 in the vicinity of plaques (Fig. 2G). The number of surfaces that show overlap of CD9 and PLIN2 was determined by using the “classification” function (Fig. 3G).

Unless stated otherwise, data were tested for statistical significance (P < 0.05) through Welch’s t-test using Prism 8.

All statistical analyses were performed using either R programming or Prism 8.

Additional file 1. Sc-RNAseq information table. Data pertaining to the Seurat analysis of the scRNA-seq data contained in Fig. 1. Tables include samples basics and Seurat analysis parameters, cluster barcodes, results of differential gene expression analysis for each cluster, pseudobulk analysis and the foam cell gene signature.Additional file 2. CRISPR/Cas9 editing to create a homozygous TREM2-R47H mutant human iPSC line. The chromatogram demonstrates CRISPR/Cas9 induced targeting of CGC (Arginine (R)) in the wild-type allele of TREM2 exon 2, to CAC (Histidine (H)) (together with the insertion of some additional silent mutations labeled in red) in both alleles, resulting in a TREM2 homozygous R47H mutation in the commercially available GFP iPSC line from Coriell.Additional file 3. TREM2-R47H mutant xMGs reveal no significant increase in the levels of P2RY12, but a decreased trend in the expression of HLA-DR compared to TREM2-WT xMGs in 5X-hCSF1. (A) Quantification of HLA-DR and P2RY12 in GFP TREM2-R47H vs. GFP TREM2 xMGs (green, GFP; red, HLA-DR; purple, P2RY12) surrounding amyloid plaques (blue, Amylo-glo) in 7-month old 5X-hCSF1 reveal (B) a trend towards decrease in total HLA-DR area (P=0.0789) (C) with no significant differences in total plaque area, (D) a trend towards decrease in the percentage of HLA-DR area normalized to plaque area (P=0.1716), and (E) no significant increase in total P2RY12 area (P=0.2848) and (F) percentage of P2RY12 area normalized to plaque area (P=0.3202) in TREM2-R47H vs. TREM2 xMGs. Scale Bar at 10 mm; n=3-4 mice per genotype; 3 images per mouse.Additional file 4. IHC analysis of xMG proximity per Plaque in 5X-hCSF1. Human microglia locations were detected and quantified through GFP immunofluorescence within 10μm per plaque (Amylo-Glo, blue) using the Cellsense software on the Olympus FV3000. (A-B) Ordinary one-way Anova was performed to confirm similar distribution of samples within a given genotype; (C) Plaque area of the plaques included for xMG counts show no significant difference between genotypes; (D) After normalization of the number of xMGs within 10 μm per plaque to plaque area, the significant reduction in the number of R47H-mutant xMGs remains; Unless stated otherwise, data were tested for statistical significance (P<0.05) through Welch’s t-test using Prism 8 (**** P < 0.0001; n= 3-4 mice per genotype; 5-6 images per mouse).