Work Description

Date: 1 December, 2023

Dataset Title: Human neural stem cells restore spatial memory in a transgenic Alzheimer’s disease mouse model by an immunomodulating mechanism - Source data

Dataset Creators: Kevin S. Chen; Mohamed H Noureldein; Lisa M McGinley; John M Hayes; Diana M Rigan; Jaquelin F Kwentus; Shayna N Mason; Faye E Mendelson; Masha G Savelieff; and Eva L Feldman

Dataset Contact: Kevin Chen, kechen@med.umich.edu

Funding: National Institutes of Health (NIH)

Research Overview:
Therapeutic mechanisms of human neural stem cells (hNSCs) were studied in an Alzheimer's disease mouse model (5XFAD). hNSCs restored spatial memory abilities in 5XFAD animals; however, amyloid beta levels were unchanged. Spatial transcriptomics was used to probe mechanisms of hNSCs. Focusing on a subset of plaque-induced genes, gene normalization was seen particularly in microglia, confirmed by PROGENy and Cell Chat analyses. The spatial transcriptomics data from this publication have been deposited in NCBI Gene Expression Omnibus (16) and are accessible through GEO Series accession number GSE209583 ( https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE209583 and enter token gzglogqkvjqrhmt). Additional supporting data are available from the corresponding author upon reasonable request.

Methodology:
-Animal model: All animal procedures were approved by the University of Michigan Institutional Animal Care and Use Committee (Protocol #PRO00010247). The 5XFAD transgenic mouse model was used (catalog # 034848, The Jackson Laboratory, Bar Harbor, ME), expressing five mutations associated with familial AD: the Swedish (K670N/M671L), Florida (I716V), and London (V717I) mutations in the human amyloid-beta precursor protein (APP), as well as human presenilin-1 (PS1) harboring M146L and L286V mutations.

-hNSC transplantation and immunosuppression regimen: An IGF-1 producing hNSC cell line (Palisade Bio, Carlsbad, CA) was cultured as described previously (McGinley et al., 2016;McGinley et al., 2018). On the day of the transplantation procedure, hNSCs were trypsinized, centrifuged, and resuspended in hibernation media (Seneca Biopharma) and stereotactically transplanted as previously described (McGinley et al., 2018). Briefly, animals were anesthetized with 1-2% isoflurane inhaled on a 100% oxygen carrier, positioned within a mouse cranial stereotactic frame in the standard manner, and a powered burr was used to create a skull opening large enough to permit introducing a 34-gauge needle. Six infusions in total of 1 μL each were targeted to three sites bilaterally in the fimbria fornix of the hippocampus at the following coordinates relative to the bregma (millimeters posterior/lateral/ventral): 0.82/0.75/2.5, 1.46/2.3/2.9, and 1.94/2.8/2.9. Each infusion was administered over 60 seconds, followed by a 60 second equilibration period before withdrawing the needle and closing the scalp. For cohorts undergoing behavioral testing and histological/ELISA Aβ analysis, mice were randomly assigned to four treatment groups at 26 weeks of age: (i) untreated non-transgenic wild-type mice (WT) littermates (n=10), (ii) untreated 5XFAD (n=10), (ii) vehicle injected 5XFAD (n=9), and (iii) hNSC transplanted 5XFAD (n=10). Vehicle animals underwent injections of hibernation media. For hNSC intracranial transplantation, 3 bilateral injections of hNSCs at 30,000 cells/µL, 1 µL per injection for a total 180,000 cells were performed per animal. Procedural animals received mycophenolate mofetil (30 mg/kg subcutaneous) daily for 7 days after the procedure, and tacrolimus (3.0 mg/kg subcutaneous) was administered daily starting at 24 weeks of age until study end. Brain sections utilized for spatial transcriptomics were obtained from WT mice, untreated 5XFAD mice, 5XFAD mice receiving immunosuppression alone, and 5XFAD mice undergoing intracranial hNSC transplantation. Here, immunosuppression consisted of monoclonal antibodies against mouse CD4 (clone GK1.5, rat IgG2b,κ; catalog # BE0003-1, Bio X Cell, Lebanon, NH) and CD40L (clone MR-1, Armenian hamster IgG; catalog # BE0017-1, Bio X Cell). Injections (20 mg/kg intraperitoneal for each antibody) began starting 1 day prior to the surgical procedure and occurred daily for 4 days, then weekly thereafter for the duration of the experiment. hNSC were transduced to express GFP and firefly luciferase (McGinley et al., 2022), and transplants were performed in 5XFAD mice at 6 weeks of age (3 bilateral injections of hNSCs at 50,000 cells/µL, 2 µL per injection, for a total of 600,000 cells per animal).

-Morris water maze: The Morris water maze is a well-established method of assessing spatial memory in rodent models (Bromley-Brits et al., 2011). Briefly, at 34 to 35 weeks of age (8 weeks after hNSC or vehicle injection for procedural groups), animals were placed in a pool of opaque water with visual cues placed around the perimeter of the pool. A submerged hidden platform was placed in one quadrant, which provides an escape from water. The mice can deduce the location of the hidden platform from the spatial relationship to the surrounding visual cues. With intact short-term memory, the latency for animals to swim towards and find the hidden platform decreases over repeated trials as animals learn the spatial relationship between the platform and visual cues (4 trials per day for 9 days). A positive control with a visible platform was performed on the 10th day. Subsequently, to test long-term reference memory, animals were reintroduced into the maze 7 days after the last trial, except the platform was removed as a probe trial. Time spent probing each quadrant of the pool was measured. Mice with intact spatial learning and long-term memory will spend a disproportionate amount of time searching for the platform in the quadrant where it was previously placed. On the other hand, mice with impaired spatial memory spend only 25% of the time in the correct quadrant by chance.

-Tissue preparation and immunohistochemistry: After completing behavioral testing, animals are euthanized at 35 weeks of age by phenobarbital overdose and transcardially perfused with phosphate buffered saline (PBS) solution. Brains were dissected from the calvarium and bisected along the midline sagittal plane. One brain hemisphere was fixed in 4% paraformaldehyde for 24 h, cryoprotected in escalating sucrose gradients, embedded, and cryosectioned onto histologic slides (10 μm thickness, coronal plane sections). Hematoxylin/eosin staining was performed on every 10th slide to identify hippocampal structures and the approximate location of hNSC injections. For immunohistochemistry slides were warmed and washed with PBS, then permeabilized/blocked with 0.3% Triton X-100 and 5% bovine serum albumin (BSA). Slides were then incubated in primary rabbit anti-Aβ antibody (1:1500; catalog # 9888, Cell Signaling Technologies, Danvers, MA) diluted in 0.1% Triton X-100 and 5% BSA overnight at 4 °C. Sections were washed with PBS and incubated with fluorescent anti-rabbit secondary antibody at 1:1000 in 0.1% Triton X-100 and 5% BSA for 1 h. After washing in PBS again, slides were incubated with Hoechst nuclear stain (1 mg/mL) for 10 min, washed, and mounted with glass coverslips and Prolong Gold Anti-Fade mountant (Thermo Fisher Scientific, Waltham, MA). Fluorescent images were captured at 20X magnification (Nikon Microphot-FXA, Nikon, Chiyoda, Japan; CellSens Dimension software, Olympus, Shinjuku, Japan). Target regions included images of dentate gyrus, cornu ammonis fields, and fimbria fornix. Five images per section were captured from two sections per animal for a total of 10 images per animal. Images were converted to 8-bit and the image thresholds were standardized in Fiji/ImageJ (Schindelin et al., 2015). Images were then processed using the Analyze Particles function to quantify the percentage area of immunoreactivity in each section (plaque-to-plaque free area).

-Multiplex ELISA: The remaining brain hemisphere from each animal was flash frozen in liquid nitrogen and stored at -80°C. Production of AD-related and immune-related factors was assessed by multiplex ELISA (Eve Technologies, Calgary, Canada). Brain tissue was harvested after Morris water maze testing, thawed on ice and lysed in RIPA buffer (catalog # 89900, Thermo Fisher Scientific) containing a protease inhibitor cocktail (catalog # 11697498001, Roche Diagnostics, Basel, Switzerland). Lysates were collected, briefly sonicated, brought to a volume of 1 mL and centrifuged at 13,000 rpm for 20 min at 4 °C. Supernatants were collected, diluted ten-fold in RIPA buffer and protein concentration was measured and normalized to 400 µg/mL. Samples were flash frozen in liquid nitrogen and shipped on dry ice to Eve Technologies for analysis. Discovery Assay protein arrays and Custom-Plex Assays were used to quantify the following analytes: Human Aβ and Tau 2-Plex Assay (custom panel: Aβ42, Aβ40), Human Supplemental Biomarker 10-Plex Discovery Assay [catalog # HDHSB10: neural cell adhesion molecule (NCAM), soluble vascular cell adhesion molecule 1 (sVCAM-1), soluble intercellular adhesion molecule 1 (sICAM-1), platelet-derived growth factor (PDGF-AA, PDGF-AB), cathepsin D], mouse matrix metalloproteinase (MMP) 5-Plex Discovery Assay (catalog # MDMMP-S,P: MMP-2, MMP-3, MMP-8, MMP-9, MMP- 12), mouse transforming growth factor beta (TGF-β) 3-Plex Discovery Assay (catalog # TGFβ1-3: TGF-β1, TGF-β2, TGF-β3), and Mouse Focused 10-Plex Discovery Assay [catalog # MDF10: interleukins (IL-1β, IL-2, IL-6, IL-10, IL-12), granulocyte-macrophage colony-stimulating factor (GM-CSF), monocyte chemoattractant protein 1 (MCP-1), and tumor necrosis factor alpha (TNF-α)].

-Statistics: Behavioral data was analyzed by two-way analysis of variance (ANOVA), and probe trial was analyzed by one-way ANOVA using Prism (GraphPad, La Jolla, CA). For immunohistochemistry and ELISA data, statistical differences were determined by one-way ANOVA. A p<0.05 was considered statistically significant and all values are presented as mean ± SEM (standard error of the mean). Spatial transcriptomics data were analyzed using R version 3.5.1 and RStudio 1.3.1093.

-Spatial transcriptomics: In a separate experiment for spatial transcriptomics of the brain, we compared hNSC injected 5XFAD animals to 5XFAD mice receiving immunosuppression alone, untreated 5XFAD mice, and control WT mice. At 34 weeks post-transplantation, all animals were euthanized and perfused with PBS on the same day. Brains were dissected, hemisected along the sagittal midline, placed in OCT media, frozen in isopentane cooled with liquid nitrogen, and stored at -80 °C. Test coronal sections, 10 μm thick, were placed on Visium 10X tissue optimization slides (10X Genomics, Pleasanton, CA) to optimize permeabilization conditions. Next, all experimental hemibrains were removed from -80 °C storage and thawed on the same day, such that they were only subjected to a single freeze/thaw cycle. All hemibrains were sectioned to 10 μm thick coronal sections, which were placed on Visium 10X slides and submitted simultaneously to the University of Michigan Advanced Genomics Core for analysis. There were 2 sections from WT brain from 1 animal, 4 sections from 5XFAD brain from 2 animals, 2 sections from immunosuppression alone brain from 1 animal, and 6 sections from hNSC transplanted brain from 3 animals. All Visium slides were processed simultaneously over two days according to the manufacturer's instructions. Briefly, slides were heated at 37 °C for 1 min on a T100 thermal cycler (BioRad, Hercules, CA), transferred to precooled methanol (Sigma Aldrich, St. Louis, MO) for 30 min, and stained by hematoxylin and eosin (H&E) using a standard protocol. Brightfield images of the H&E-stained sections were captured using an AxioObserver (Zeiss, Oberkochen, Germany). Tissue sections were permeabilized for 12 min, as optimized on test brain sections, using the Visium Spatial Tissue Optimization Slide & Reagent kit. Permeabilization releases mRNA from cells, which bind to oligonucleotides on the capture areas followed by reverse transcription, second-strand synthesis, denaturation, and cDNA amplification, as specified in the T100 thermal cycler protocol using the spatial slide adaptor. Amplified cDNA clean-up was performed using SPRIselect (Beckman Coulter, Brea, CA) and the library was prepared according to the manufacturer’s protocol. The generated cDNA libraries were subjected to 151 bp paired-end sequencing, according to the manufacturer’s protocol (NovaSeq 6000, Illumina, San Diego, Ca).

-Data and Image Analysis for Spatial Transcriptomics: For the spatial transcriptomics data, the NovaSeq 6000 sequencer generated “bcl” files, which were converted to de-multiplexed “fastq” files using Bcl2fastq2 Conversion Software (Illumina, San Diego, CA). Space Ranger 1.3.1 (10X Genomics) aligned and filtered reads and counted barcodes and unique molecular identifiers, using the “Fastq” files and H&E image tiff files as inputs. Reads were aligned against the Genome Reference Consortium Mouse Build 38 patch release 6 (GRCm38.p6). Space Ranger correlates the spatial coordinates to gene expression data by reallocating the barcodes in each read to the respective 55 µm feature (i.e., single spots) relative to the slide’s fiducial frame. Spatial transcriptomics data were analyzed using R version 4.1.0 and RStudio 1.4.1717 on the Greatlakes cluster, available at the University of Michigan (https://arc.umich.edu/greatlakes/). Preprocessed sequencing data were analyzed with the Seurat R package (version 4.0.5) (Satija et al., 2015) following the guided tutorial pipeline (available at https://satijalab.org/seurat/articles/spatial_vignette.html, accessed on 5/27/2022). Individual datasets were normalized using the “SCTransform” function in the Seurat package by regularized negative binomial regression (Hafemeister and Satija, 2019). SCTransformation was specifically performed to account for heterogeneity in single-cell RNA-seq due to batch effects and experimental variation without losing biological signal.(Choudhary and Satija, 2022) Individual section datasets were integrated using “FindIntegrationAnchors” on the top 3,000 most variable genes, and the “IntegrateData” function from the Seurat package compared the different conditions (Butler et al., 2018).

-Detecting hNSCs: To detect hNSCs in brain sections spatial transcriptomics data and identify their differentially expressed genes, the raw sequencing “fastq” files were realigned to the human genome. As hNSCs were tagged with emerald GFP (EmGFP; https://www.snapgene.com/resources/plasmid-files/?set=fluorescent_protein_genes_and_plasmids&plasmid=Emerald_GFP), the EmGFP sequence was appended to the “Homo_sapiens.GRCh38.ensembl” human genome using the “mkgtf” function in “Spaceranger” to create a Spaceranger-compatible reference genome file. Subsequently, the raw “fastq” files were aligned against this custom reference genome. hNSCs were identified as cells expressing EmGFP and a count matrix of expressed genes was constructed for these cells. A heatmap of the most expressed 100 genes in hNSCs was generated using the pheatmap R package (v 1.0.12) (Kolde, Raivo, "Package ‘pheatmap’." R package 1, no. 7 (2015): 790). EmGFP sequence: ATGGTGAGCAAGGGCGAGGAGCTGTTCACCGGGGTGGTGCCCATCCTGGTCGAGCTGGACGGCGACGTAAACGGCCACAAGTTCAGCGTGTCCGGCGAGGGCGAGGGCGATGCCACCTACGGCAAGCTGACCCTGAAGTTCATCTGCACCACCGGCAAGCTGCCCGTGCCCTGGCCCACCCTCGTGACCACCTTGACCTACGGCGTGCAGTGCTTCGCCCGCTACCCCGACCACATGAAGCAGCACGACTTCTTCAAGTCCGCCATGCCCGAAGGCTACGTCCAGGAGCGCACCATCTTCTTCAAGGACGACGGCAACTACAAGACCCGCGCCGAGGTGAAGTTCGAGGGCGACACCCTGGTGAACCGCATCGAGCTGAAGGGCATCGACTTCAAGGAGGACGGCAACATCCTGGGGCACAAGCTGGAGTACAACTACAACAGCCACAAGGTCTATATCACCGCCGACAAGCAGAAGAACGGCATCAAGGTGAACTTCAAGACCCGCCACAACATCGAGGACGGCAGCGTGCAGCTCGCCGACCACTACCAGCAGAACACCCCCATCGGCGACGGCCCCGTGCTGCTGCCCGACAACCACTACCTGAGCACCCAGTCCGCCCTGAGCAAAGACCCCAACGAGAAGCGCGATCACATGGTCCTGCTGGAGTTCGTGACCGCCGCCGGGATCACTCTCGGCATGGACGAGCTGTACAAGTAA Few GFP-positive hNSC-containing spots were detected, suggesting aligned transcript reads to the mouse genome were likely not contaminated by transcripts from human cells to any significant extent. To confirm this, hNSC sections were also aligned against a mixed mouse/human genome, which found that only 0.9% to 3.9% of sequencing reads from hNSC sections aligned to the human genome. These reads were distributed mainly to GFP-positive spots. A smaller fraction of reads was present in other spots, which mainly contained mouse-aligned reads. Aligning reads to the mixed mouse/human genome did not significantly impact results compared to aligning to the mouse genome.

-Dimension reduction, clustering, and visualization: Principal component analysis was conducted using the top 3,000 most variable genes. Clusters were detected using the first 30 principal components. The number of principal components was selected based on a percent change in variation in consecutive principal components less than 0.1%. Next, Uniform Manifold Approximation and Projection (UMAP) of the principal components was used to visualize cells. Graph-based clustering was performed on the principal component analysis-reduced data.

-Cell type annotation and differential expression analysis: To assign each cluster a cell type identity, cluster gene markers were identified using the “FindAllMarkers” function in Seurat. Clusters were annotated based on gene expression patterns and the H&E images using the Allen Brain Atlas (Lein et al., 2007). After distinct brain regions were annotated, different glia were identified based on the expression of canonical markers as follows, oligodendrocytes (Mag, Mog, Olig2), astrocytes (Gfap, Slc1a3), microglia (Itgam, Ptprc), stage 1 disease-associated microglia (DAM) (Trem2, Tyrobp), and stage 2 DAM (Cst7, Itgax, Spp1) (Garden and Campbell, 2016;Deczkowska et al., 2018). Wilcoxon test compared proportions of different cell types across different conditions. Non-parametric Wilcoxon rank sum test compared gene expression across different conditions and identified differentially expressed genes (Alves and Higdon, 2013). Differentially expressed genes were identified across conditions in whole brain sections as well as hippocampus. Genes were considered differentially expressed if adjusted p-value<0.05. Pathway enrichment was performed using the Kyoto Encyclopedia of Genes and Genomes (Kanehisa and Goto, 2000) and Gene Ontology (Harris et al., 2004) databases by applying the richR package (v 0.0.19, https://github.com/hurlab/richR, accessed on 5/27/2022). Pathways were considered significant if adjusted p-value<0.05. All hemibrain sections were processed simultaneously on Visium slides for spatial transcriptomics to minimize batch effects. Additionally, sections from each experimental group were distributed across slides. However, to analyze any possible batch effects, a linear mixed effects model (LMM) was performed. LMM adjusts for different slides by treating sections placed on the same slide as dependent observations. The LMM model was constructed with a random intercept (i.e., the slide) and without a random slope using the model formula form ~condition+(1|slide). The LMM was applied using the lmerSeq R package on the variance-stabilizing transformed pseudobulk aggregated counts.(Vestal et al., 2022) The majority of DEGs from LMM (Table S14 for whole brain; Table S15 for hippocampus) overlapped with DEGs by Wilcoxon, suggesting the absence of batch effects and the robustness of the SCTransformation method.

-Cell-to-cell communication: CellChat examined communications across cells (Jin et al., 2021). CellChat leverages network analysis and pattern recognition to predict signaling outputs from cells and signaling inputs to cells. CellChat also analyzes how input and output signals across cells coordinate. First, CellChat identifies significant ligand-receptor pair signaling pathways across all cell clusters. Second, the software predicts incoming signals to and outgoing signals from specific cell clusters. CellChat also predicts global communication patterns using pattern recognition approaches. Similarity measures and manifold learning from topological perspectives organizes signaling pathways. Lastly, CellChat calculates the communication probability of a signaling pathway by summarizing the probabilities of its associated ligand-receptor pairs.

-PROGENy analysis: In addition to richR for pathway enrichment, PROGENy (Pathway RespOnsive GENes) was also used to estimate the activity of relevant signaling pathways based on consensus gene signatures (Schubert et al., 2018). PROGENy leverages publicly available signaling perturbation experiments to yield a common core of 14 pathways comprising genes responsive to the perturbations for human and mouse (Holland et al., 2020). PROGENy has the advantage of overcoming the limitations of conventional pathway enrichment methods, such as overlooking the effects of post-translational modifications on pathway activation. PROGENy scores were computed to infer pathway activities from the spatial transcriptomics dataset using PROGENy R package (v 1.17.3) (https://saezlab.github.io/progeny/index.html, accessed on 5/31/2022).

Instrument and/or Software specifications:
(Nikon Microphot-FXA, Nikon, Chiyoda, Japan; CellSens Dimension software, Olympus, Shinjuku, Japan)
ELISA (Eve Technologies, Calgary, Canada)
R version 3.5.1 and RStudio 1.3.1093
NovaSeq 6000 sequencer
Bcl2fastq2 Conversion Software (Illumina, San Diego, CA)
Space Ranger 1.3.1 (10X Genomics)
Genome Reference Consortium Mouse Build 38 patch release 6 (GRCm38.p6)
R version 4.1.0 and RStudio 1.4.1717
Seurat R package (version 4.0.5)
EmGFP; https://www.snapgene.com/resources/plasmid-files/?set=fluorescent_protein_genes_and_plasmids&plasmid=Emerald_GFP
lmerSeq R packag
richR package (v 0.0.19, https://github.com/hurlab/richR)
CellChat
PROGENy R package (v 1.17.3)

Files contained here:
-LM6 MWM.pzfx
-LM6 Multiplex Analysis.pzfx
-Quantification for AB.xlsx

Related publication(s):
Chen KS, Noureldein MH, McGinley LM, Hayes JM, Rigan DM, Kwentus JF, Mason SN, Mendelson FE, Savelieffd MG, Feldman EL. Human neural stem cells restore spatial memory in a transgenic Alzheimer's disease mouse model by an immunomodulating mechanism. bioRxiv [Preprint]. 2023 Nov 4:2023.11.01.565161. doi: 10.1101/2023.11.01.565161. PMID: 37961246; PMCID: PMC10635057.

Use and Access:
This data set is made available under a Creative Commons Attribution-Noncommercial license (CC-BY-NC 4.0).

To Cite Data:
Chen, K. S., Noureldein, M. H., McGinley, L. M., Hayes, J. M., Rigan, D. M., Kwentus, J. F., Mason, S. N., Mendelson, F. E., Savelieff, M. G., Feldman, E. L. Human neural stem cells restore spatial memory in a transgenic Alzheimer’s disease mouse model by an immunomodulating mechanism - Source data [Data set], University of Michigan - Deep Blue Data. https://doi.org/10.7302/sf6n-7884