Astrocytes display cell autonomous and diverse early reactive states in familial amyotrophic lateral sclerosis

Abstract Amyotrophic lateral sclerosis is a rapidly progressive and fatal disease. Although astrocytes are increasingly recognized contributors to the underlying pathogenesis, the cellular autonomy and uniformity of astrocyte reactive transformation in different genetic forms of amyotrophic lateral sclerosis remain unresolved. Here we systematically examine these issues by using highly enriched and human induced pluripotent stem cell-derived astrocytes from patients with VCP and SOD1 mutations. We show that VCP mutant astrocytes undergo cell-autonomous reactive transformation characterized by increased expression of complement component 3 (C3) in addition to several characteristic gene expression changes. We then demonstrate that isochronic SOD1 mutant astrocytes also undergo a cell-autonomous reactive transformation, but that this is molecularly distinct from VCP mutant astrocytes. This is shown through transcriptome-wide analyses, identifying divergent gene expression profiles and activation of different key transcription factors in SOD1 and VCP mutant human induced pluripotent stem cell-derived astrocytes. Finally, we show functional differences in the basal cytokine secretome between VCP and SOD1 mutant human induced pluripotent stem cell-derived astrocytes. Our data therefore reveal that reactive transformation can occur cell autonomously in human amyotrophic lateral sclerosis astrocytes and with a striking degree of early molecular and functional heterogeneity when comparing different disease-causing mutations. These insights may be important when considering astrocyte reactivity as a putative therapeutic target in familial amyotrophic lateral sclerosis.


Animals, transgenic models and tissue processing
The following transgenic mouse lines were used, and were analyzed as different experimental groups: (i) Female SOD1 G93A mice [B6SJL-Tg(SOD1*G93A)1Gur/J, Jackson Laboratories], postnatal day 100-107 (n = 3 mice). (ii) wild-type C56BL/6-SJL mixed background (Jackson Laboratories) were used as control (n = 3 mice). (iii) Female 9 month old VCP A323E (n=3 mice) and (iv) VCP WT (n=3 mice) C57BL/6 (St. Jude Children's Research Hospital, Memphis, TN, USA) 3 . Mice were bred and maintained at the UCL Institute of Neurology in standard individually ventilated cages with up to three mice per cage, in a temperature and humidity controlled environment with a 12-h light/dark cycle and had access to drinking water and food ad libitum. Cages were checked daily to ensure animal welfare. Body weight was assessed regularly to ensure no weight loss. For tissue collection, animals were injected with terminal anaesthesia (pentobarbital sodium, Euthatal) and were transcardially perfused with 4% paraformaldehyde. The spinal cord was removed and post-fixed with 4% paraformaldehyde for one hour and then cryoprotected overnight with 30% sucrose; 10 or 20 μm serial transverse cryosections were cut for immunofluorescence staining.

Immunolabeling, imaging and image analysis
Samples were fixed in 4% paraformaldehyde and then blocked in 5% bovine serum albumin (BSA) in PBS, 0.3% Triton X-100. Primary antibodies were diluted as follows: rabbit anti C3 (Dako, A0063) 1:1000, chicken anti GFAP (Abcam, ab4674) 1:500, mouse anti Nestin (Millipore, MAB5326) 1:200, rabbit anti NFκB p65 (Cell Signaling, D14E12) 1:400, rabbit anti NFKB p105 (Thermo Fisher, PAS-85292) 1:400 and rabbit anti Vimentin (Cell Signaling, 5741) 1:200. Primary antibody incubation was performed overnight at 4°C in 5% BSA and 0.3% Triton X-100 in PBS, followed by incubation with secondary alexa fluor fluorescent antibodies (Thermo Fisher) for one hour at room temperature. 4′,6-diamidino-2-phenylindole (DAPI) was used as a nuclear counterstain (100 ng/ml). For iPSC cultures, images were acquired using the Opera Phenix High-Content Screening System (Perkin Elmer) with a 40x water objective as confocal z-stacks with zstep of 1µm and were processed as maximum projection. Cells were then analysed on a cell-bycell basis using an automated pipeline on Columbus TM Image Data Storage and Analysis System (Perkin Elmer) software version 2.8.0. To control for cell density in per field analysis, fields with a total number of cells less than 30 or more than 100 cells were removed from the analysis. For mice spinal cord sections, slides were mounted with mounting media (Dako) and were imaged using a 710 Laser Scanning Confocal Microscope (Zeiss). At least 5 z-stack images were taken per mouse (N=3 mice per genotype) from at least two spinal cord sections using the 40x objective and data was analyzed using Fiji.

RNA extraction and qPCR
The Promega Maxwell RSC simplyRNA cells kit including DNase treatment, alongside the Maxwell RSC instrument, was used for RNA extractions. The nanodrop was used to assess RNA concentration and the 260/280 ratio. Reverse transcription was performed using the Revert Aid First Strand cDNA Synthesis Kit (ThermoFisher Scientific) using random hexamers. qPCR was performed using the PowerUP SYBR Green Master Mix (ThermoFisher Scientific) and QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems). Primers for C3 (Forward: GGGCTCGCTGGTGGTAAAAA; Reverse: AGACACCGGCGTAATCCTTC) and GAPDH (Forward: ATGACATCAAGAAGGTGGTG; Reverse: CATACCAGGAAATGAGCTTG) were used. Specific amplification was determined by melt curve analysis and agarose gel electrophoresis of the PCR products. Primer pairs with 90-110% efficiency were used and RT-minus samples were used as negative controls. Data was analysed using the ddCt method and is expressed as fold change over control group.

RNA-seq
Library preparation followed the Illumina TruSeq RNA Access library preparation kit as per the manufacturer's instructions; 100ng total RNA was first fragmented, cDNA generated using random priming during first and second strand synthesis; sequencing adaptors are ligated to the resulting double-stranded cDNA fragments. The coding regions of the transcriptome were then captured from this library using sequence-specific probes before a final round of PCR amplification and second strand digestion occurs to create the final library. This strand-specific protocol was used for library preparation and samples were barcoded and multiplexed prior to sequencing on a HiSeq 2500 platform. Control, VCP, SOD1 and FUS RNA-seq libraries were generated according to this protocol. C9ORF72 mutant astrocyte datasets were accessed from 4 using accession code GSE142730.

Differential Expression
Differential expression analysis was carried out using DESeq2 (v.1.22.1 9,10 ) according to recommended package guidelines. In brief this involved gene-level Wald tests following shrunken estimation of sample dispersions and fold changes. All genes with at least one count in one sample were used in each analysis to satisfy DESeq2's statistical model. FDR threshold for significance was set at 0.01 and determined using the procedure of Benjamini and Hochberg. Summary plots were generated with the R packages ggplot2 (v.2.3.1), VennDiagram (v.1.6.20) and gplots (v.3.0.1). Gene ontology analysis was carried out using the GO enrichment Analysis command from the WGCNA (v.1.69) R package 4,11 . Included terms were filtered to Bonferroni P-values < 0.1, and overlapping terms collapsed using REVIGO 12 .

Gene regulatory networks
High confidence (category A and B), signed human regulons from the DoRothEA gene set resource were used for calculating enriched transcription factor scores 13 . Regulons were curated based on various lines of evidence including literature resources, ChIP-seq peaks, binding site motifs and interactions inferred directly from gene expression. The regulon enrichment within each gene signature was determined using the analytic rank-based enrichment analysis algorithm from the virtual inference of protein-activity by enriched regulon (VIPER) package 14 . A null model based on sample permutations was incorporated to account for both the expected non-uniform distribution of the targets on the gene expression signature and the extensive co-regulation of gene expression. The minimum regulon size was set at 25 target genes, and the differential activity of each transcription factor was determined with a Student's t-test with threshold set at p < 0.05. Network plots were generated using the R package, RedeR (v.1.36). Additional rototation gene set tests to support transcription factor regulon enrichments were carried out using the roast function of edgeR (v.3.26.8) according to vignette guidelines, using 9999 rotations, and with false discovery rate set at 0.1 15,16 . The regulons used were collected from the TRRUST (v.2) database and limited to a minimal size of 25 target genes 17 .

Griess assay
Release of nitric oxide from iPSC-derived astrocytes following exposure to inducers of inflammatory signalling was indirectly measured using a Griess assay. Media from cultures was collected and mixed in a 1:1 ratio with a modified Griess reagent (Sigma-Aldrich). Sodium nitrite standards (NaNO2) were prepared in media, and absorbances measured using a spectrophotometer at 540 nm.

Cytokine array
Supernatants of hiPSC-derived astrocyte cultures were analysed using the Procartaplex™ 39-PLEX, Human Cytokine/Chemokine assay (Thermofisher, Waltham, MA) using the manufacturer's instructions, as previously described 18 . Each experimental condition/technical repeat was analyzed in duplicate (i.e., two samples per experimental cell culture well). Plates were analyzed on a Luminex 200 platform (Luminex Corporation, Austin, TX). Heat maps were produced using the fold increase of each cytokine from either control or untreated hiPSC-derived astrocytes. Statistical analysis was performed using R.

Statistical analysis
Data and graphs are presented as mean ± SEM. Graphpad prism 8/9 (Graph Pad Software) was used to generate graphs and to perform tests for distribution and statistical significance for immunostaining quantification and Greiss assay data. The type of statistical tests with p values are indicated in the figure legends. Any p value below 0.05 was considered to be statistically significant. Non-significant p-values were labelled as "ns" in the text or in figures where relevant.            Representative images of NFKB1 immunofluorescence (red) in control, VCP and SOD1 astrocytes treated with TNF-α, IL-1α and C1q. Quantification of (B) nuclear/cytoplasmic ratio, (C) mean nuclear intensity and (D) mean cell intensity for NFKB1 from 2-3 lines per group. Data is shown as mean ±SEM. Data points are technical repeats (wells) used in this experiment. P-values for (B and C) were obtained from 2-way ANOVA followed by Tukey's multiple comparison tests and for (D) was obtained from 2-way ANOVA followed by Sidak multiple comparison test. For all representative images, scale bar = 20 μm.

Compliance with ethical standards
Figure S12: Increased nuclear RELA translocation in pro-inflammatory factor treated astrocytes. (A) Representative images of immunofluorescence staining for RELA (red) in control, mutant VCP and SOD1 astrocytes treated with TNF-α, IL-1α and C1q. Quantification of (B) nuclear/cytoplasmic ratio and (C) mean cell intensity for RELA from 2-3 lines per group. Data is shown as mean ±SEM. Data points are technical repeats (wells) used in this experiment. P-values were obtained from 2-way ANOVA followed by Tukey's multiple comparison tests. For all representative images, scale bar= 20 μm.