ARTICLE OPEN ACCESS Motor cortex transcriptome reveals microglial key events in amyotrophic lateral sclerosis Oriol Dols-Icardo, PhD, Vı́ctor Montal, MSc, Sònia Sirisi, PhD, Gema López-Pernas, MSc, Correspondence Dr. Clarimón Laura Cervera-Carles, PhD, Marta Querol-Vilaseca, MSc, Laia Muñoz, BSc, Olivia Belbin, PhD, jclarimon@santpau.cat Daniel Alcolea, MD, PhD, Laura Molina-Porcel, MD, PhD, Jordi Pegueroles, MSc, Janina Turón-Sans, MD, or Dr. Dols-Icardo Rafael Blesa, MD, PhD, Alberto Lleó, MD, PhD, Juan Fortea, MD, PhD, Ricard Rojas-Garcı́a, MD, PhD, and odols@santpau.cat Jordi Clarimón, PhD Neurol Neuroimmunol Neuroinflamm 2020;7:e829. doi:10.1212/NXI.0000000000000829 Abstract Objective To identify transcriptomic changes, neuropathologic correlates, and cellular subpopulations in the motor cortex of sporadic amyotrophic lateral sclerosis (ALS). Methods We performed massive RNA sequencing of the motor cortex of patients with ALS (n = 11) and healthy controls (HCs; n = 8) and analyzed gene expression alterations, differential isoform usage, and gene coexpression networks. Furthermore, we used cell type deconvolution algo- rithms with human single-nucleus RNA sequencing data as reference to identify perturbations in cell type composition associated with ALS. We performed immunohistochemical techniques to evaluate neuropathologic changes in this brain region. Results We report extensive RNA expression alterations at gene and isoform levels, characterized by the enrichment of neuroinflammatory and synaptic-related pathways. The assembly of gene coexpression modules confirmed the involvement of these 2 major transcriptomic changes, which also showed opposite directions related to the disease. Cell type deconvolution revealed an overrepresentation of microglial cells in ALS compared with HC. Notably, microgliosis was driven by a subcellular population presenting a gene expression signature overlapping with the recently described disease-associated microglia (DAM). Using immunohistochemistry, we further evidenced that this microglial subpopulation is overrepresented in ALS and that the density of pTDP43 aggregates negatively correlates with the proportion of microglial cells. Conclusions DAM has a central role in microglia-related neuroinflammatory changes in the motor cortex of patients with ALS, and these alterations are coupled with a reduced expression of postsynaptic transcripts. From the Memory Unit (O.D.-I., V.M., S.S., G.L.-P., L.C.-C., M.Q.-V., L.M., O.B., D.A., J.P., R.B., A.L., J.F., J.C.), Neurology Department and Sant Pau Biomedical Research Institute, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona; Network Center for Biomedical Research in Neurodegenerative Diseases (CIBERNED) (O.D.-I., V.M., S.S., G.L.-P., L.C.- C., M.Q.-V., L.M., O.B., D.A., J.P., R.B., A.L., J.F., J.C.), Madrid; Neurological Tissue Bank of the Biobanc-Hospital Clı́nic-IDIBAPS (L.M.-P.), Barcelona; Alzheimer’s Disease and Other Cognitive Disorders Unit, Neurology Department (L.M.-P.), Hospital Clı́nic, Institut d’Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona; Network Center for Biomedical Research in Rare Diseases (CIBERER) (J.T.-S., R.R.-G.), Madrid; and Neuromuscular Disorders Unit (J.T.-S., R.R.-G.), Department of Neurology, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Spain. Go to Neurology.org/NN for full disclosures. Funding information is provided at the end of the article. The Article Processing Charge was funded by the authors. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND), which permits downloading and sharing the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal. Copyright © 2020 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology. 1 Glossary ALS = amyotrophic lateral sclerosis; DAM = disease-associated microglia; FFPE = formalin-fixed paraffin-embedded; GO = gene ontology; HC = healthy control; IHC = immunohistochemistry; KEGG = Kyoto Encyclopedia of Genes and Genomes; MuSiC = Multi-Subject Single Cell deconvolution; NTB = Neurological Tissue Bank; qPCR = quantitative real-time PCR; RIN = RNA integrity; RNAseq = RNA sequencing; ROI = region of interest; ROSMAP = Religious Orders Study and Memory and Aging Project; snRNAseq = single-nucleus RNAseq; TREM2 = triggering receptor expressed on myeloid cells 2; WGCNA = weighted gene coexpression network analyses. Amyotrophic lateral sclerosis (ALS) is a neurodegenerative The brain motor cortex is affected at the most early stages of disease neuropathologically characterized by the aberrant the disease and is one of the most vulnerable regions in ALS. cytoplasmic aggregation and phosphorylation of the 43-kDa However, studies of this critical region both at the transcrip- transactive response DNA-binding protein (pTDP43) in tional and immunohistochemical level are lacking. In the the majority of ALS cases at postmortem evaluation, which present work, we aim to characterize the motor cortex of is known to occur in a sequential manner starting in the sporadic ALS cases through total RNAseq analyses. We also motor cortex.1 Although the causes that lead to nongenetic use cell type deconvolution using human snRNAseq data as forms of ALS (sporadic ALS) have yet to be fully de- reference and immunohistochemical analyses to resolve termined, the presence of neuroinflammation is a consistent a signature of neuropathologic changes associated with ALS in feature in the CNS of affected patients.2 Microglia are this particular brain region. macrophage-like innate immune cells of the CNS that, as a result of disease conditions, change their gene expression profile to adapt and acquire a reactive state.3 Recently, a study using single-cell RNA sequencing from the CNS of Methods Alzheimer disease and ALS mouse models reported that Human samples a specific microglial subpopulation, known as disease- The study included human motor cortex (Brodmann area 4) associated microglia (DAM), was the cell type responsible samples provided by the Neurological Tissue Bank (NTB) of for microgliosis and that triggering receptor expressed on the Biobanc-Hospital Clı́nic-IDIBAPS. None of the brain myeloid cells 2 (TREM2) is a major driver of DAM acti- tissues presented any infarcts in the motor cortex. Diagnosis vation.4 In support of this discovery, some chitinase pro- of ALS complied with the El Escorial criteria during life,14 and teins, which are known markers of microglial activation,5,6 none of the patients had a family history of ALS or dementia, have been consistently shown to be increased in the CSF of nor did show any sign of cognitive impairment. All patients patients with ALS.7,8 presented pTDP43 inclusions in the motor cortex at post- mortem examination. None of the samples carried the The pivotal role of TDP43 and other genes known to cause C9orf72 hexanucleotide repeat expansion or mutations in the ALS (such as FUS, hnRNPA2B1, hnRNPA1, TAF15 or TBK1 gene, the most common genetic causes related to adult- TIA1) in the metabolism of RNA has implicated RNA onset ALS in Spain.15,16 dyshomeostasis as a crucial event in the pathophysiology of the disease.9 An unbiased resource to obtain a compre- RNA extraction and sequencing hensive signature of gene and isoform expression changes Using a mortar and liquid nitrogen, the tissue (60 mg) was associated with the whole tissue response to disease con- grinded to powder and transferred to a solution of 600 μL of ditions is total RNA sequencing (RNAseq) in bulk tissue. TRIzol reagent (Thermo Fisher Scientific, Waltham, MA). Furthermore, the advent of novel sequencing tools such as We used standard recommendations and procedures to ex- single-nucleus RNAseq (snRNAseq) has opened a new tract RNA with TRIzol, the RNeasy Mini Kit (Qiagen, Hilden, window to better explore the transcriptome at a cell level Germany), and the Rnase-Free Dnase Set (Qiagen). Qubit and has unraveled a complex and huge variety of human was used to measure RNA concentration, whereas RNA in- cell types with unique expression profiles in the human tegrity (RIN) was verified on an Agilent 2100 Bioanalyzer brain.10–12 These transcriptomic signatures can be applied (Agilent Technologies, Santa Clara, CA). Only samples with to perform cell type deconvolution of bulk RNAseq data a quantity threshold of 5 μg and RIN ≥6.5 were used for total using novel deconvolution methods. These methods take RNAseq. The final study group included 11 ALS cases and 8 into account cross-subject and cross-cell variability of gene healthy controls (HCs) (table 1). Paired-end sequencing li- expression profiles, without relying on preselected mark- braries were prepared using the TruSeq Stranded Total RNA ers, thereby providing more realistic cell type proportion Library Preparation kit (Illumina, San Diego, CA) and se- estimations and yielding crucial information about the quenced on the Illumina HiSeq 2500 platform by the Centro cellular heterogeneity that is associated with a pathologic Nacional de Análisis Genómicos (CNAG; Barcelona, Spain) condition.13 with 101-bp paired-end reads to achieve at least 100 million 2 Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 Neurology.org/NN Table 1 Demographic, clinical, and quality control data Sample ID Group Age at death, y Sex Age at onset, y Duration, mo Site of onset RIN PMI, h AF5214 ALS 64 Male 62 33 Limb 8.2 7 AF5215 ALS 50 Male 46 58 Limb 8.8 13 AF5216 ALS 60 Female 54 72 Limb 8.3 6 AF5218 ALS 53 Female 50 32 Limb 8.4 9 AF5220 ALS 82 Male 81 1 Bulbar 7.7 16 AF5222 ALS 54 Female 53 18 Limb 9 7 AF5224 ALS 70 Female 68 24 Limb 8.4 14 AF5227 ALS 78 Male 76 24 Limb 8.1 12 AF5228 ALS 77 Female 75 31 Bulbar 7 16 AF5229 ALS 57 Female 54 24 Limb 6.8 5 AF5232 ALS 52 Male 51 12 Limb 7.4 18 AF5219 HC 64 Male — — — 8.2 10 AF5221 HC 78 Male — — — 7.9 6 AF5226 HC 81 Female — — — 7.5 23 AF5231 HC 83 Female — — — 7.3 7 AF5234 HC 58 Male — — — 7.5 4 AF5235 HC 76 Male — — — 7.5 11 AF5236 HC 83 Female — — — 7.7 7 AF5237 HC 68 Female — — — 6.5 13 Abbreviations: ALS = amyotrophic lateral sclerosis; HC = healthy control; PMI = postmortem interval; RIN = RNA integrity number. paired-end reads for each sample (ranging from 103.447.000 Validation of transcriptome changes by to 138.990.000 paired-end reads). quantitative real-time PCR We performed quantitative real-time PCR (qPCR) on the Data processing same brain-derived RNA samples used for total RNAseq (11 We aligned FASTQ files to the Grch38.p12 genome assembly sALS cases and 8 controls). A total of 500 ng of RNA was used using STAR v2.6.1a.17 We then used GATK Best Practices18 to generate cDNA with the RevertAid First Strand cDNA for variant calling and FeatureCounts (within the Subread Synthesis Kit (Thermo Fisher Scientific), as per the manu- v.1.6.2 package19) to assign fragments to each gene feature facturer’s instructions. All qPCRs were conducted using Fast included in the Grch38v94 gene transfer format file. For dif- SYBR Green Master Mix (Thermo Fisher Scientific) and run ferential isoform usage, we used the STAR v.2.6.1 package in on an ABI Prism 7900HT Fast Real-Time PCR System quantMode TranscriptomeSAM mode and used Salmon to (Applied Biosystems, Foster City, CA). All primer pairs used obtain isoform expression values.20 For differential gene ex- are listed in table e-1, links.lww.com/NXI/A280. For relative pression and differential isoform usage analyses, we used quantification, we applied the DDCt method, using RPL13 as DESeq2 v1.24,21 which applies the Benjamini-Hochberg for p an endogenous control to normalize the data. We decided to value adjustment. We used a false discovery rate threshold of use RPL13 as a housekeeping gene as it showed the smallest 5%. To perform gene ontology (GO) and Kyoto Encyclo- coefficient of variation (<3.5%) compared with the 2 other pedia of Genes and Genomes (KEGG) pathway enrichment putative housekeeping genes that we assessed (RPL0 and analyses, we used Metascape 3.0 (metascape.org),22 applying GAPDH). default parameters (minimum overlap = 3, p value cutoff = 0.01, and minimum enrichment = 1.5). For the synaptic en- Cell type deconvolution richment analyses, we used the web server SynGO (syngo- We used the recently developed Multi-Subject Single Cell portal.org/), which provides an expert-curated resource for deconvolution (MuSiC)13 method to perform cell type synaptic function and gene enrichment analysis.23 deconvolution of our bulk RNAseq data. As a reference, we Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 3 used the most comprehensive human snRNAseq data set peroxidase secondary antibody (1:200) for 1 hour at room available. This data set includes the RNAseq from 70,634 cells temperature. Primary antibodies were Iba1 rabbit poly- in the frontal cortex (Brodmann area 10) of 24 Alzheimer clonal (1:500; Wako Chemicals, Osaka, Japan) and ps409/ disease cases and 24 HCs from the Religious Orders Study 410-TDP rabbit polyclonal (1:2000; Cosmo Bio, Tokyo, and Memory and Aging Project (ROSMAP),12 grouped into Japan). Following peroxidase development with 3,39-dia- 8 major cellular populations (excitatory neurons, inhibitory minobenzidine tetrahydrochloride (Dako liquid DAB + neurons, microglia, astrocytes, oligodendrocytes, oligoden- substrate Chromogen System; Dako Denmark, Glostrup, drocyte precursor cells, endothelial cells, and pericytes). To Denmark), sections were stained with hematoxylin (EnVi- check consistency and validate our results, we used 2 in- sion FLEX Haematoxylin; Dako Denmark) and dehydrated dependent snRNAseq repositories as a reference to estimate using increasing concentrations of ethanol and then major cell type proportions. To this aim, we downloaded the mounted with DPX mounting medium (PanReac Appli- medial temporal gyrus snRNAseq data derived from the Allen Chem, ITW Reagents, Chicago, IL). For immunofluores- Brain Atlas, which includes 15,928 cells from 8 human tissue cence, we used 3 ALS and 3 control motor cortex sections, donors. We selected this data set as it contains double the which were incubated with primary antibodies overnight at number of nuclei than the primary visual cortex and the an- 4°C. We used Iba1 rabbit polyclonal antibody as a marker terior cingulate cortex, also contained in the database.10 The for the broad microglial population (1:500; Wako Chem- third data set comprised 10,319 cells derived from the frontal icals) and human HLA-DP, DQ, DR antigen (MHC class II cortex of 4 individuals.11 Cell type deconvolution of cellular markers), mouse clone CR3/43 (1:20; Agilent, subpopulations was performed using the ROSMAP data as M077501–2), as a marker for activated cells. MHCII marker reference. genes are specifically expressed in the Mic1 gene signa- ture.12 Thus, we decided to combine these primary anti- Weighted gene coexpression network analyses bodies to visualize Mic1 cells as performed in the study of We accounted for genes with more than 10 counts across all Mathys et al.12 as well as to estimate the proportion of cells individuals. DESeq2 was then used to normalize the input of this particular subpopulation. Then, we incubated sec- data to construct coexpression networks using the weighted tions for 1 hour at room temperature with secondary anti- gene coexpression network analyses (WGCNA) package.24 bodies (1:1,000), which were conjugated with Alexa Fluor We then constructed the signed weighted correlation net- 488 and Alexa Fluor 555 (Invitrogen, Carlsbad, CA). Cell work, which takes into account both negative and positive nuclei visualized with Hoechst 33258 (Life Technologies). correlations, using the manual function in WGCNA, applying Sections were mounted with Shandon Immu-Mount the biweight midcorrelation, selecting a power of 7 (the (Thermo Fisher Scientific). We acquired all the images lowest possible power term where topology fits a scale free with a Leica TCS SP5 Confocal Laser Scanning Microscope, network), and run in a single block analysis. We defined all with a 63× objective. modules by the hybrid treecutting option with deepsplit pa- rameter = 2, applying a minimal module size of 30 genes and Quantification analysis merging modules with a cutHeight ≤0.25. Module eigengenes We obtained full-section immunohistochemistry (IHC) (first principal component as a summary of each module) images with Pannoramic MIDI II (3DHistech). Blinded to summarized the modules and correlations performed using the neuropathologic diagnoses, we delimited 6 gray matter spearman correlation for continuous traits and Pearson cor- areas of the motor cortex from each case, and for each gray relation for binary traits (bicor and corPvalueStudent, in- matter area, we randomly generated 6 regions of interest cluded in the WGCNA package). We used Cytoscape 3.7.2 to (ROIs) with the same size using the multicrop.m script. We visualize networks. used the Iba1.m script to quantify densities for Iba1 staining in each ROI, which represents a total of 36 images per in- Immunohistochemistry dividual. This algorithm binarizes the random images and We used formalin-fixed paraffin-embedded (FFPE) motor compute densities of protein expression to quantify the cortex tissue sections of 5 μm in 11 patients with ALS and 7 number of immunoreactive objects. Both in-house algo- controls. One control sample (AF5237) was excluded from rithms can be freely accessed at github.com/Memo- the analysis due to lack of availability of FFPE motor cortex ryUnitSantPau. We used MATLAB R2017b software to sections. We dewaxed the FFPE sections by placing them in develop the algorithm (The MathWorks Inc., Natick, MA). xylene and decreasing concentrations of ethanol, followed For pTDP43 quantification, 2 different researchers blinded by hydration 30 minutes in distilled water. We then treated to neuropathologic diagnoses quantified immunoreactive tissue sections with a solution containing 10% methanol, 3% positive objects manually. We evaluated immunofluores- H2O2, and PBS for 10 minutes to inhibit endogenous per- cence using FIJI imaging software. We estimated an auto- oxidase and boiled in citrate 1x solution to mediate antigen mated threshold to create binary images. Following the noise retrieval. Once cooled, we blocked all sections with 5% removal with the despeckle and the remove outliers’ filters, bovine serum albumin for 1 hour at room temperature and we determined the number of Iba1+ microglial cells incubated them overnight with the corresponding primary expressing MHC class II markers using the analyze particles antibody at 4°C and then with the anti-rabbit horse radish function. 4 Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 Neurology.org/NN Statistical analysis measures in these 10 genes (table e-3, links.lww.com/NXI/ We used the Shapiro-Wilk test to test for deviation from A280). GO and KEGG pathway analyses revealed an enriched a standard distribution. For correlation analyses, we de- involvement of the inflammatory response in ALS (figure termined the Pearson or Spearman correlation coefficients 1C). The assessment of changes at the isoform expression using the ggscatter function within the ggpubr package in R. level resulted in 167 upregulated and 40 downregulated iso- We performed mean comparison analyses through the Stu- forms, of which 181 were from unique genes (table e-4, links. dent t test or Student test with Welch correction (using t.test lww.com/NXI/A280). Enrichment analyses revealed the in- function in R), depending on data distribution. Statistical volvement of postsynaptic density and immune response as significance for all tests was set at 5% (α = 0.05), and all dominant pathways (figure e-1, links.lww.com/NXI/A279). statistical tests were 2 sided. We used R version 3.6.2 to To further characterize and confirm the synaptic involvement, perform all statistical analyses. we used the SynGO database, which provides an expert- curated resource for synaptic function and gene enrichment Standard protocol approvals, registrations, analysis.23 This analysis confirmed the enrichment of post- and patient consents synaptic components in our list of differentially expressed The Sant Pau Hospital Ethics Committee approved the study isoforms (table e-5, links.lww.com/NXI/A280). Among the with all experiments with human tissue performed in accor- list of altered genes identified by the differential isoform ex- dance with the Declaration of Helsinki. The NTB of the pression and gene expression models, 32 of them overlapped Biobanc-Hospital Clı́nic-IDIBAPS supplied all tissue follow- between both approaches (table e-6, links.lww.com/NXI/ ing approval from their Scientific Advisory Committee. A280). The involvement of synaptic-related changes was confirmed from the differential isoform usage analysis (figure Data availability statement e-2, links.lww.com/NXI/A279), whereas the enrichment of We have deposited all raw sequencing data (FASTQ files) at inflammatory markers was evidenced in the gene model ap- the European Genome-phenome Archive (EGA), which is proach (figure e-3, links.lww.com/NXI/A279). hosted by the EBI and the CRG under accession number EGAS00001004286. Cell type deconvolution To characterize the variability of cell types and compare cel- lular proportions between groups, we applied the recently Results developed MuSiC algorithm. As reference, we used data from Demographics, clinical features, and the first and most comprehensive human snRNAseq study sample composition performed in brain tissue of individuals with a neurodegen- We performed RNAseq analysis of postmortem motor cor- erative disease, which includes 24 patients with Alzheimer tex samples from 11 ALS cases (6 females) and 8 individuals disease and 24 HCs.12 Cell type deconvolution revealed an without neurologic disease (4 females). Postmortem interval overrepresentation of microglial cells in the ALS motor cortex and RIN did not differ between groups (p = 0.68 and p = compared with HC (p = 0.025, fold change = 1.65) and 0.089, respectively). The mean age at death was lower for a reduction in excitatory neurons (p = 0.101, fold change = ALS cases compared with HC (63.4 years; range: 54–82 -0.85) (figure 2). We confirmed microglial upregulation using years in ALS and 73.9 years; range: 58–83 years in HC, p = 2 additional independent reference data sets.10,11 In both 0.04). Patients had an average disease onset of 60.9 years cases, we found the same pattern of downregulation of ex- (range: 46–81 years) and a disease duration of 29.9 months citatory neurons (figures e-4 and e-5, links.lww.com/NXI/ (range: 1–72 months). Nine of the 11 cases (81.8%) pre- A279). To further gain insight into which microglial sub- sented a limb disease onset (table 1). We used RNAseq data population might be driving this effect, we estimated the to rule out the presence of other ALS-disease causing proportion of each subpopulation identified in the ROSMAP mutations, confirming the presumably sporadic nature of our study. Our results showed that a specific microglial sub- ALS group of cases. population (Mic1, as named in the study performed by Mathys and collaborators and recognized as the human Gene expression alterations and differential DAM12) is disproportionately presented in the ALS motor isoform usage cortex (p = 6.6 × 10−3, fold change = 4.33) (figure 2). Among Differential gene expression between patients with ALS and the 77 marker genes, which characterize the Mic1 subcluster HCs disclosed a total of 108 upregulated and 16 down- as DAM (Mic1 unique genes or those overlapping with regulated genes (adjusted p < 0.05) (figure 1A and table e-2, DAM),12 22 of them (28.6%) are genes differentially links.lww.com/NXI/A280). We validated 9 of the 10 genes expressed in our bulk RNAseq analysis (unadjusted p < 0.05; selected by qPCR, as we did not confirm the CHI3L2 gene (table e-7, links.lww.com/NXI/A280). Of interest, the pro- overexpression in patients with ALS (p = 0.085) (figure 1B). portion of Mic1 cells showed a positive correlation with the Correlation analyses between counts derived from total expression of TREM2 (p = 1.6 × 10−3; R = 0.67; figure e-6, RNAseq data and qPCR expression values indicated a high links.lww.com/NXI/A279), a receptor required for the tran- and significant concordance between both expression sition from the homeostatic microglia to DAM. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 5 Figure 1 Differential regulation of gene expression in ALS and HC (A) Volcano plot displaying differentially expressed genes between the ALS and the HC motor cortex. The vertical axis (y-axis) corresponds to the −log10 adjusted p value, and the horizontal axis (x-axis) represents the log2 fold change value obtained when comparing ALS with HC gene expression. Significantly differential expressed genes are depicted with blue circles (adjusted p value <0.05), whereas gray circles display the nonsignificant genes. Gene names are shown for the 10 genes selected for qPCR validation. (B) Box plot showing the relative RNA expression values in ALS and HC obtained through qPCR for the 10 genes selected for validation. Dashed line corresponds to reference relative gene expression. *p < 0.05; **p < 0.01. (C) Top 5 gene ontology and KEGG pathway enrichment analyses obtained from the list of genes showing a significant differential expression. ALS = amyotrophic lateral sclerosis; HC = healthy control. Gene coexpression network analyses inflammation and synaptic alterations are interconnected in To identify gene clusters with varying coexpression patterns ALS pathologic processes. Furthermore, the proportion of that could behave differently between patients and healthy Mic1 cells showed a strong direct correlation with the 2 in- individuals, and elucidate possible biological mechanisms flammatory gene coexpression modules associated with ALS driving pathologic processes, we performed a gene coex- (MEblack; p = 2.7 × 10−8, R = 0.92; and MEyellow; p = 1.4 × pression network analyses using the WGCNA package. The 10−3, R = 0.68) and a negative correlation with the synaptic analysis disclosed a total of 8 modules. Among them, 3 module (MEpink; p = 2.4 × 10−4, R = −0.75) (figure 3). modules associated with disease status (MEblack, p = 0.003; R Overall, our results from differential gene and isoform ex- = 0.64; MEyellow, p = 0.025; R = 0.57; and MEpink, p = pression, gene coexpression modules, and cell type decon- 0.011; R = −0.51; figure 3). Moreover, MEblack and MEyel- volution comparisons strongly indicate that microglia-related low were enriched for inflammatory responses, whereas inflammatory changes, mainly driven by the Mic1 sub- MEpink had an overrepresentation of genes involved in population (also known as DAM), are central in ALS path- synaptic and neuronal functions (table e-8, links.lww.com/ ophysiology and that these inflammatory processes are closely NXI/A280). The SynGO database confirmed the post- related to the synaptic disturbances that are present in the synaptic signature of this enrichment in MEpink (table e-9, motor cortex. links.lww.com/NXI/A280). Of interest, the 2 inflammatory gene coexpression modules (MEblack and MEyellow) Immunohistochemistry showed a strong negative correlation with the postsynaptic To further investigate the inflammatory response in post- module (MEpink) (p = 8 × 10−5; R = −0.78 and p = 3 × 10−4; mortem brain tissue, we performed IHC analyses using Iba1, R = −0.74, respectively) (figure 3), suggesting that both a protein whose expression is primarily restricted to homeostatic 6 Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 Neurology.org/NN Figure 2 Cell type composition in the ALS and HC motor cortex Box plot showing major cell type and Mic1 proportions for HC and ALS estimated by MuSiC using the ROSMAP human single-nucleus RNAseq data. *p < 0.05; **p < 0.01. ALS = amyotrophic lateral sclerosis; HC = healthy control; MuSiC = Multi-Subject Single Cell deconvolution; ROSMAP = Religious Orders Study and Memory and Aging Project. microglia, to examine the broad population of microglial cells, compared with HCs (p = 0.017) (figure e-9A, links.lww.com/ and MHC class II, a marker responsible for antigen recognition NXI/A279). Of interest, a high degree of variability on the and the activation of the adaptive immune system, which is number of pTDP43 aggregates was noted in ALS and was not expressed in the Mic1 cell subpopulation. Positive immunos- explained by age at death, age at onset, or disease duration. taining was increased in the ALS motor cortex for Iba1; how- The burden of pTDP43 inclusions inversely correlated with ever, differences between patients and controls did not reach the proportion of microglial cells in our group of patients with statistical significance (p = 0.106; figure e-7, links.lww.com/ ALS (p = 0.0026; R = −0.81) (figure e-9B, links.lww.com/ NXI/A279). Notwithstanding, co-immunofluorescence of Iba1 NXI/A279), and the same pattern was found with the and MHC class II markers revealed an increase in the number of microglial subpopulation Mic1 (p = 0.019; R = −0.69). microglial cells expressing MHCII markers in ALS compared with HC (70.1% vs 10%, respectively; p = 0.021) (figure 4 and figure e-8, links.lww.com/NXI/A279), strengthening our results and suggesting that Mic1 is driving neuroinflammation in Discussion the ALS motor cortex. The motor cortex is one of the major vulnerable and early affected regions in ALS and represents a target region to We finally assessed the frequency of pTDP43 immunoreac- disentangle key pathologic processes in this neurodegenera- tive structures in the motor cortex. As expected, patients with tive disorder. To date, few studies have performed a whole ALS showed a higher density of pTDP43 inclusions transcriptomic assessment of the CNS in ALS.25–27 Through Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 7 Figure 3 WGCNA visualization and correlation map constructed with WGCNA module eigengenes and cell type proportions Network diagrams showing the 50 most connected genes for each of the 3 modules associated with ALS (MEblack, MEyellow, and MEpink) are depicted on the top of the figure. Node darkness and size are proportional to the number of connections within the module. Correlation matrix showing the correlation coefficients between cell type proportions is depicted for the 6 major cell types and the Mic1 (DAM) cell subtype. The 3 significant modules (MEblack, MEyellow, and MEpink), each of one sharing unique groups of coexpressed genes that are differentially expressed between patients with ALS and controls, are also included in the correlogram. Only significant correlations are depicted with a colored circle (blue for direct and red for inverse correlations). The size of the circle is proportional to the correlation significance. The plot indicates the high correlation between the 3 modules and how these modules are related to the proportion of specific cell types. ALS = amyotrophic lateral sclerosis; DAM = disease-associated microglia; HC = healthy control; WGGNA = weighted gene coexpression correlation network analyses. 8 Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 Neurology.org/NN Figure 4 Immunohistochemistry of Mic1 markers Immunohistochemistry with anti-IBA1 (red) and anti-MHC class II (green) antibodies in the ALS and HC motor cortex. Scale bar corresponds to 50 μm. ALS = amyotrophic lateral sclerosis; HC = healthy control. an unbiased transcriptomic analysis using high-throughput related to glial activation and inflammation.27 Our results RNA sequencing, we have elucidated gene and isoform ex- clearly reinforce the idea of these inflammatory-related pression alterations, gene coexpression networks, and cell changes in the human motor cortex and emphasize the role type proportions associated with ALS. For the first time in the of DAM as a key factor in this process. First, differential gene ALS field, we have performed cell type deconvolution using and isoform expression data point toward an inflammatory human brain single-nucleus RNA sequencing data from 3 response as a major event that is strongly intensified in the independent sources as reference, thus providing highly reli- ALS motor cortex. Second, the assembly of weighted gene able results that have been reinforced through our immuno- coexpression networks resulted in 2 significant modules as- histochemical analyses. sociated with ALS (MEblack and MEyellow), both highly enriched with genes related to inflammatory functions. Third, We report a list of 124 genes differentially expressed in the cell type deconvolution of the bulk RNAseq data demon- ALS motor cortex, which reflect the RNA expression changes strated an increased proportion of microglial cells compared in this brain region. Among them, chitinase-related genes with the motor cortex of healthy cases. In this context, studies CHI3L1 and CHI3L2 presented 2 of the most prominent in mice have recently evidenced DAM as the microglial sub- shifts in gene expression, whereas CHIT1 showed a nominal population with a more prominent role in ALS.4 Our results significance. Of interest, these chitinases are neuro- show that Mic1, the human microglial subpopulation that inflammatory biomarkers, which have been shown to be harbors the majority of markers found in the previously de- consistently increased in the CSF of patients with ALS scribed DAM transcriptomic signature,12 is the main micro- compared with neurologically healthy individuals.7,8 These glial subpopulation that drives microgliosis in ALS. In fact, results indicate that among the unbiased signature of RNA almost a third of Mic1 (DAM) marker genes are deregulated alterations provided herein, some of them might lead to the in the motor cortex of our group of patients with ALS. Of discovery of novel promising biomarkers. note, Mic1 proportion highly correlated with the expression of TREM2, which is required to enhance the proinflammatory An exacerbated innate immune response with microgliosis has stage of DAM.4 been recently described in the ALS motor cortex.28 A recent study that has investigated the whole transcriptome of the Our immunohistochemical analyses did not show a relevant ALS motor cortex suggested that among ALS cases, a sub- increased density of Iba1 (a widely used marker of microglial group of them is characterized by a molecular signature processes) in ALS-related brain tissue, nor was the expression Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 9 of its coding gene (AIF1) altered in our RNASeq data set. Overall, our study strongly suggests that DAM plays a key role Previous studies have frequently provided contradictory in driving neuroinflammatory changes and synapse loss in the results when using this marker to assess microgliosis.29 This ALS motor cortex. The identification of specific microglial somewhat unexpected result could be explained by the fact populations with well-defined transcriptional signatures will that some marker genes might downregulate on microglial contribute to disentangle new mechanisms and novel thera- activation. Also, whereas cell type deconvolution is performed peutic targets to fight against this devastating disorder. using a complete catalogue of gene expression profiles obtained from brain-derived snRNAseq, immunohistochem- Acknowledgment istry only uses a single marker and does not reflect the com- The authors are indebted to the IDIBAPS Biobank for sample plexity and heterogeneity of this cell type, making it an and data procurement. underpowered method to detect differences in the proportion of cell subpopulations. That being said, we did validate the Study funding increased proportion of Mic1 cells in the ALS motor cortex The authors are indebted to the “Fundación Española para el through co-immunofluorescence of 2 markers (Iba1 and Fomento de la Investigación de la Esclerosis Lateral MHC class II) previously shown to characterize this micro- Amiotrófica (FUNDELA)” for funding the present study. O. glial subpopulation.12 This finding further establishes Mic1 Dols-Icardo is a recipient of a grant by The Association for (the human disease-associated microglia) as the microglial Frontotemporal Degeneration (Clinical Research Post- subpopulation that drives microgliosis in the ALS motor doctoral Fellowship, AFTD 2019–2021). This work was cortex. supported by research grants from Institute of Health Carlos III (ISCIII), Spain PI18/00326 to J.C., PI18/00435 to D.A., We observed a high degree of variability in the density of and INT19/00016 to D.A., and by the Department of Health pTDP43 inclusions across patients, which could not be Generalitat de Catalunya PERIS program SLT006/17/125 to explained by any of the demographic or clinical features D.A. This work was also supported in part by Generalitat de available in this study. Notably, we found a striking inverse Catalunya (2017 SGR 00547) to the “Grup de Recerca en correlation between the proportion of microglial cells and the Demències: Sant Pau.” amount of pTDP43 aggregates. These results are in line with recent in vivo and in vitro studies, suggesting that although Disclosure microgliosis arises as a phagocytic response to pTDP43 The authors report no disclosures relevant to the manuscript. aggregates, at some point, these cells lose their ability to clear Go to Neurology.org/NN for full disclosures. these neuropathologic insults and are downregulated.30,31 Whether biofluid levels of microglial markers, such as Publication history TREM2, could be used as a proxy of pTDP43 density in the Received by Neurology: Neuroimmunology & Neuroinflammation motor cortex is an avenue worth pursuing and would be February 21, 2020. Accepted in final form May 15, 2020. a valuable addition to the biomarker arsenal for use in de- signing clinical trials and assessing therapeutic efficacy. Synaptic dysfunction is an early pathogenic event in ALS.32 Appendix Authors Our data point toward an underrepresentation of post- Name Location Contribution synaptic markers and a decrease of excitatory neurons in patients with ALS. Together, these results are consistent Oriol Dols- Sant Pau Biomedical Major role in the design Icardo, PhD Research Institute, Hospital and conceptualization of with upper motor neuron degeneration occurring in this de la Santa Creu i Sant Pau, the study; drafted the brain region. A limitation of our approach is the lack of an Barcelona, Spain manuscript; and analyzed the data available motor cortex snRNAseq data set, precluding any firm and more detailed conclusion related to the alteration of Vı́ctor Sant Pau Biomedical Analyzed the data and Montal, MSc Research Institute, Hospital participated in manuscript Betz cells, a unique class of motor neurons expressed in this de la Santa Creu i Sant Pau, drafting specific brain area. Recent studies in other neurodegenera- Barcelona, Spain tive diseases have suggested that microglia are a key and early Sònia Sirisi, Sant Pau Biomedical Analyzed the data and mediator of synapse loss through phagocytosis induced by PhD Research Institute, Hospital participated in manuscript the complement cascade.33,34 Our gene expression data de la Santa Creu i Sant Pau, drafting Barcelona, Spain show an increased expression of some key complement cascade-related genes and the most significant enriched Gema Sant Pau Biomedical Analyzed the data and López- Research Institute, Hospital performed a part of the pathway corresponds to the complement and coagulation Pernas, MSc de la Santa Creu i Sant Pau, experiments cascades, reinforcing this hypothesis. Furthermore, our Barcelona, Spain results indicate that the 2 neuroinflammatory-related mod- Laura Sant Pau Biomedical Participated in manuscript ules of gene expression (MEblack and MEyellow) and the Cervera- Research Institute, Hospital drafting and data Carles, PhD de la Santa Creu i Sant Pau, interpretation proportion of Mic1 cells inversely correlate with the pres- Barcelona, Spain ence of synaptic markers. 10 Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 Neurology.org/NN 5. Steinacker P, Verde F, Fang L, et al. Chitotriosidase (CHIT1) is increased in microglia and macrophages in spinal cord of amyotrophic lateral sclerosis and cerebrospinal Appendix (continued) fluid levels correlate with disease severity and progression. J Neurol Neurosurg Psychiatry 2018;89:239–247. Name Location Contribution 6. Thompson AG, Gray E, Thézénas ML, et al. Cerebrospinal fluid macrophage bio- markers in amyotrophic lateral sclerosis. Ann Neurol 2018;83:258–268. Marta Sant Pau Biomedical Participated in manuscript 7. Illán-Gala I, Alcolea D, Montal V, et al. CSF sAPPβ, YKL-40, and NfL along the ALS- Querol- Research Institute, Hospital drafting and data FTD spectrum. Neurology 2018;91:e1619–e1628. Vilaseca, de la Santa Creu i Sant Pau, interpretation 8. Thompson AG, Gray E, Bampton A, Raciborska D, Talbot K, Turner MR. CSF MSc Barcelona, Spain chitinase proteins in amyotrophic lateral sclerosis. J Neurol Neurosurg Psychiatry 2019;90:1215–1220. Laia Muñoz, Sant Pau Biomedical Contributed to sample 9. Ling SC, Polymenidou M, Cleveland DW. Converging mechanisms in ALS and FTD: BSc Research Institute, Hospital preparation and disrupted RNA and protein homeostasis. Neuron 2013;79:416–438. de la Santa Creu i Sant Pau, participated in 10. Hodge RD, Bakken TE, Miller JA, et al. Conserved cell types with divergent features in Barcelona, Spain experimental procedures human versus mouse cortex. Nature 2019;573:61–68. 11. Lake BB, Chen S. Sos BC Integrative single-cell analysis of transcriptional and epi- Olivia Sant Pau Biomedical Participated in the revision genetic states in the human adult brain. Nat Biotechnol 2018;36:70–80. Belbin, PhD Research Institute, Hospital and editing of the 12. Mathys H, Davila-Velderrain J, Peng Z, et al. Single-cell transcriptomic analysis of de la Santa Creu i Sant Pau, manuscript and data Alzheimer’s disease. Nature 2019;570:332–337. Barcelona, Spain interpretation 13. Wang X, Park J, Susztak K, Zhang NR, Li M. Bulk tissue cell type deconvo- lution with multi-subject single-cell expression reference. Nat Commun 2019; Daniel Sant Pau Biomedical Participated in data 10:380. Alcolea, Research Institute, Hospital collection, analysis, and 14. Brooks BR, Miller RG, Swash M, Munsat TL; World Federation of Neurology Re- MD, PhD de la Santa Creu i Sant Pau, data interpretation search Group on Motor Neuron Diseases. El Escorial revisited: revised criteria for the Barcelona, Spain diagnosis of amyotrophic lateral sclerosis. Amyotroph Lateral Scler Other Motor Neuron Disord 2000;1:293–299. Laura Neurological Tissue Bank Contributed to sample 15. Garcı́a-Redondo A, Dols-Icardo O, Rojas-Garcı́a R, et al. Analysis of the C9orf72 gene Molina- of the Biobanc-Hospital preparation and in patients with amyotrophic lateral sclerosis in Spain and different populations Porcel, MD, Clı́nic-IDIBAPS, Barcelona, participated in worldwide. Hum Mutat 2013;34:79–82. PhD Spain experimental procedures 16. Dols-Icardo O, Garcı́a-Redondo A, Rojas-Garcı́a R, et al. Analysis of known amyo- trophic lateral sclerosis and frontotemporal dementia genes reveals a substantial ge- Jordi Sant Pau Biomedical Participated in data netic burden in patients manifesting both diseases not carrying the C9orf72 expansion Pegueroles, Research Institute, Hospital analysis mutation. J Neurol Neurosurg Psychiatry 2018;89:162–168. MSc de la Santa Creu i Sant Pau, 17. Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal RNA-seq aligner. Barcelona, Spain Bioinformatics 2013;29:15–21. 18. McKenna A, Hanna M, Banks E, et al. The Genome Analysis Toolkit: a MapReduce Janina Sant Pau Biomedical Major role in the framework for analyzing next-generation DNA sequencing data. Genome Res 2010; Turón-Sans, Research Institute, Hospital acquisition of samples and 20:1297–1303. MD de la Santa Creu i Sant Pau, data and reviewed and 19. Liao Y, Smyth GK, Shi W. The R package Rsubread is easier, faster, cheaper and better Barcelona, Spain edited the manuscript for alignment and quantification of RNA sequencing reads. Nucleic Acids Res 2019; 47:e47. Rafael Sant Pau Biomedical Major role in the 20. Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and Blesa, MD, Research Institute, Hospital acquisition of samples and bias-aware quantification of transcript expression. Nat Methods 2017;14: PhD de la Santa Creu i Sant Pau, data and reviewed and 417–419. Barcelona, Spain edited the manuscript 21. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 2014;15:550. Alberto Sant Pau Biomedical Major role in the 22. Zhou Y, Zhou B, Pache L, et al. Metascape provides a biologist-oriented resource for Lleó, MD, Research Institute, Hospital acquisition of samples and the analysis of systems-level datasets. Nat Commun 2019;10:1523. PhD de la Santa Creu i Sant Pau, data and reviewed and 23. Koopmans F, van Nierop P, Andres-Alonso M, et al. SynGO: an evidence-based, Barcelona, Spain edited the manuscript expert-curated knowledge base for the synapse. Neuron 2019;103:217–234.e4. 24. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network Juan Fortea, Sant Pau Biomedical Major role in the analysis. BMC Bioinformatics 2008;9:559. MD, PhD Research Institute, Hospital acquisition of samples and 25. Prudencio M, Belzil VV, Batra R, et al. Distinct brain transcriptome profiles in de la Santa Creu i Sant Pau, data and reviewed and C9orf72-associated and sporadic ALS. Nat Neurosci 2015;18:1175–1182. Barcelona, Spain edited the manuscript 26. D’Erchia AM, Gallo A, Manzari C, et al. Massive transcriptome sequencing of human spinal cord tissues provides new insights into motor neuron degeneration in ALS. Sci Ricard Sant Pau Biomedical Major role in the Rep 2017;7:10046. Rojas- Research Institute, Hospital acquisition of samples and 27. Tam OH, Rozhkov NV, Shaw R, et al. Postmortem cortex samples identify distinct Garcı́a, MD, de la Santa Creu i Sant Pau, data and reviewed and molecular subtypes of ALS: retrotransposon activation, oxidative stress, and activated PhD Barcelona, Spain edited the manuscript glia. Cell Rep 2019;29:1164–1177.e5. 28. Jara JH, Genç B, Stanford MJ, et al. Evidence for an early innate immune response in Jordi Sant Pau Biomedical Major role in funding the motor cortex of ALS. J Neuroinflammation 2017;14:129. Clarimón, Research Institute, Hospital acquisition; design and 29. Hopperton KE, Mohammad D, Trépanier MO, Giuliano V, Bazinet RP. Markers of PhD de la Santa Creu i Sant Pau, conceptualization of the microglia in post-mortem brain samples from patients with Alzheimer’s disease: Barcelona, Spain study; drafted the a systematic review. Mol Psychiatry 2018;23:177–198. manuscript; and analyzed 30. Spiller KJ, Restrepo CR, Khan T, et al. Microglia-mediated recovery from ALS- the data relevant motor neuron degeneration in a mouse model of TDP-43 proteinopathy. Nat Neurosci 2018;21:329–340. 31. Svahn AJ, Don EK, Badrock AP, et al. Nucleo-cytoplasmic transport of TDP- 43 studied in real time: impaired microglia function leads to axonal spreading References of TDP-43 in degenerating motor neurons. Acta Neuropathol 2018;136: 1. Brettschneider J, Del Tredici K, Toledo JB, et al. Stages of pTDP-43 pathology in 445–459. amyotrophic lateral sclerosis. Ann Neurol 2013;74:20–38. 32. Henstridge CM, Pickett E, Spires-Jones TL. Synaptic pathology: a shared mechanism 2. Liu J, Wang F. Role of neuroinflammation in amyotrophic lateral sclerosis: cellular in neurological disease. Ageing Res Rev 2016;28:72–84. mechanisms and therapeutic implications. Front Immunol 2017;8:1005. 33. Hong S, Beja-Glasser VF, Nfonoyim BM, et al. Complement and microglia mediate 3. Prinz M, Jung S, Priller J. Microglia biology: one century of evolving concepts. Cell early synapse loss in Alzheimer mouse models. Science 2016;352:712–716. 2019;179:292–311. 34. Dejanovic B, Huntley MA, De Mazière A, et al. Changes in the synaptic proteome in 4. Keren-Shaul H, Spinrad A, Weiner A, et al. A unique microglia type Associated with tauopathy and rescue of tau-induced synapse loss by C1q antibodies. Neuron 2018; restricting development of Alzheimer’s disease. Cell 2017;169:1276–1290. 100:1322–1336.e7. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 7, Number 5 | September 2020 11