Bioinformatics analysis of RNA sequencing and microarray data

Processed files of microarray datasets were obtained from GEO, including Brockington et al. 2013 (GSE40438, [19]), Kaplan et al. 2013 (GSE52118, [20]), and Shadrach et al. 2021 (GSE162028, [23]), together with processed gene count or intensity tables from Perrin et al. [21], Lobsiger et al. [22], Allodi et al. [24], Nizzardo et al. [25], retrieved from the original authors. To minimize technical and biological heterogeneity, datasets were analysed independently according to species, disease stage, and experimental platform (RNA sequencing vs microarray). Direct merging of dataset was avoided. For RNA sequencing datasets, Bandyopadhyay et al. [26] and Mei et al. [27], were mapped to mm39 genome assembly using STAR (version 2.7.0e) [28], and gene level expression was quantified using rpkmforgenes.py with Ensembl gene annotation (mm39). Expression values were normalized prior to downstream analysis. Differential expression analyses were performed within each dataset separately, comparing biologically matched conditions (e.g., ALS versus control within the same species and stage). Where applicable, normalisation procedures and statistical models account for differences in sequencing depth and library composition.

Ethics statement for the use of human tissues

The use of human post mortem tissues was approved by the ethical review board of Sweden (Etikprövningsmyndigheten) and was conducted in compliance with the code of ethics of the World Medical Association (Declaration of Helsinki). Human samples were retrieved from the Netherland’s Brain Bank (NBB, http://www.brainbank.nl), who in turn retrieved samples with the written informed consent from the donors or next of kin (Table 1). Cases classified as controls from the NBB did not present with dementia or any other neurological disease, while ALS cases had a diagnosis of definite ALS.

Ethics statement regarding animals and animal models used

All animal procedures were approved by the Swedish ethical council (Stockholms Norra Djurförsöksetiska nämnd) in compliance with the code of ethics of the World Medical Association (Declaration of Helsinki) and national legislation and institutional guidelines. Mice were housed according to standard conditions, with access to food and water ad libitum and a dark/light cycle of 12 h at the laboratory animal facility, Karolinska Institutet. To investigate the role of β2m in ALS we used the strain B6.129P2-β2m tm1/Unc/J, Jackson Laboratory, Stock Number 002087, congenic on a C57BL/6J background. These mice are homozygous for the B2mtm1Unc targeted mutation (β2m-KO) and fail to express MHC class I protein on the cell surface and are grossly deficient in CD45− CD8+ T cells [29]. To generate experimental ALS-mice with different doses of β2m we crossed β2m−/− females with transgenic males of the SOD1G93A colony (B6.Cg‐Tg(SOD1*G93A)1Gur/J; Jackson Laboratory Stock, Number 004435). These mice overexpress mutant human SOD1 protein leading to an ALS‐like phenotype [30]. Litters of this first cross (F1), always heterozygous for β2m (β2m±), were crossed to generate all possible genotype combinations with the mutant SOD1 background (F2): SOD1G93A;β2m+/+, SOD1G93A;β2m± and SOD1G93A;β2m−/−. Some experimental animals, normally produced at lower frequency, were also generated by crossing: β2m−/− females from the original colony with mutant SOD1-transgenic males from the F1 (SOD1G93A;β2m+/−) or β2m+/+ females with males (SOD1G93A;β2m+/+), both from the F2.

Genotyping of both colonies and experimental animals was carried out by regular PCR following The Jackson Laboratory protocols while qPCR was routinely used to follow transgenic SOD1 copy number of male breeders (also following Jackson’s recommendations).

Behavioural analysis and disease staging

Body weight of experimental animals was measured twice a week from postnatal day (P) 56. From week 8 (~ P56) to week 20 (~ P140), animals were scored weekly for the extension reflex of their hind limbs and challenged with the inverted grid test. In the extension reflex test, the mice were suspended from the tail and their hind limb extension reflex was scored from 3.0 (normal) to 1.0 (ALS condition) with 0.5 interval points during disease progression. In the inverted grid test, the time a mouse holds onto a grid was measured. Prior to any measurements, mice were first habituated to the grid by being placed onto it on two separate days, performing the procedure, without taking any actual measurements. Then for the experimental days, mice were placed on the inverted grid once without taking any measurement and subsequently with measurements. The maximum duration of the assay was 120 seconds (s) (healthy, normal situation after task learning). If the animal did not reach the 120 s, another trial was carried out after 1–2 min of recovery. The longest time measured was selected as the behavioural performance for that week. Some records were excluded from the inverted grid analysis since a few animals learn to jump after just a few seconds of trial. Ultimately, maximum performance of each mouse was set at 100% and decreased performances normalised against this value. For assessment of survival, animals were kept until end-stage. Once the animals got severely impaired hind limbs (typically around ~ P140), they were monitored twice a day and scored using the KI assessment checklist to follow animal health status. From this point water bottles with long drinking spouts were provided to facilitate water access. In addition, animals received soft gel food (Scanbur) on the cage floor. End stage was defined as the inability of the animals to right themselves within 15 s after being placed on either side.

Tissue collection, processing and analysis

For muscle analysis, mice were sacrificed by inhalation of CO2. Lumbrical muscles (from the plantar surface of the hind-paw), soleus and tibialis anterior (TA) were dissected in 0.1 M phosphate buffered saline (PBS) and fixed in 4% paraformaldehyde (PFA) (Sigma-Aldrich) for 30 min for NMJ analysis. Lumbricals were analysed in whole-mount, whereas soleus and TA muscles were sectioned at 30 μm thickness. For CNS immunohistochemistry, animals were anesthetised with avertin (2,2,2-Tribromoethanol; Sigma-Aldrich) and perfused intracardially with PBS followed by 4% PFA. Brains and spinal cords were dissected and postfixed (for 3 and 1 h, respectively), cryoprotected in sucrose and sectioned (30 μm). Tissues were imaged on a Zeiss LSM700 or 800 confocal microscopes and a Zeiss Axio imager M1 microscope.

Immunofluorescent analysis of astrocytes and CD8+ T cells in mouse spinal cord tissue

Spinal cords were cut at the cervical level, mounted in OCT and sectioned at 30 µm-thickness in a cryostat (2–3 animals per genotype and sex). Sections were attached to Superfrost Plus slides (ThermoFisher Scientific) and stored at -20ºC. Blocking of mouse spinal cord sections was performed in 5% donkey serum in 0.3% triton X-100 (Sigma Aldrich) in PBS. The tissue was subsequently incubated with primary antibodies in 5% donkey serum in 0.3% triton X-100 in PBS overnight at 4 °C. The primary antibodies used were mouse anti-GFAP (Sigma, G3893, 1:500), and rat anti-CD8 (AbD Serotec, MCA341R, 1:50). Sections were subsequently washed with PBS and incubated with secondary antibodies Alexa-488 donkey anti-rabbit, Alexa-568 donkey anti-mouse (Invitrogen, 1:2000) and Hoechst (ThermoFisher Scientific, H3570, 1:2500) in 5% donkey serum in 0.3% triton X-100 in PBS for 1 h at room temperature. Next, slides were washed with PBS and mounted using Fluoromount G mounting media (ThermoFisher Scientific, 00-4958-02). Images of spinal cord sections were acquired with an inverted confocal (LSM800-Airy) microscope (Zeiss) using a 20× objective. Images were captured in the ventral horn of the spinal cord where spinal MNs are located. For GFAP quantification, at least three images per sex and genotype were taken and the mean intensity of each image was measured using the ImageJ software.

NMJ innervation in tibialis anterior, lumbricals and soleus muscles

Muscle was permeabilised with 4% triton X-100 for 1 h. Blocking was performed with 10% donkey serum in 0.1% triton X-100 (Sigma Aldrich) in PBS and the tissue was subsequently incubated with primary antibodies in 10% donkey serum and 0.1% triton X-100 in PBS for 48 h. Primary antibodies used were mouse anti-synaptic vesicle protein (DSHB, SV2, 1:100) and mouse anti-neurofilament 165 kDa (DSHB, 2H3; 1:50). Next, slides were washed with PBS and incubated with secondary Alexa-488 donkey anti-mouse antibodies in PBS (1:500; Invitrogen) for 3 h at room temperature. Finally, to visualise endplates, α-bungarotoxin (α-BTX) staining was performed for 15 min using tetramethyl-rhodamine isocyanate-conjugated α-BTX (1:1,000; Invitrogen). Slides were coverslipped using Mowiol 488 mounting media (Sigma-Aldrich). Innervation of the NMJ was determined by analysing a minimum of 50 endplates across the muscle. Each endplate within a field of view was categorised as either fully occupied (the presynaptic terminal completely overlies the endplate), partially occupied (the presynaptic terminal partially covers the endplate) or vacant (no presynaptic staining overlies the endplate). All analyses and quantifications were performed blind to the genetic status of the muscles.

Quantification of GAP-43 expression at the NMJ

For analysis of GAP-43 expression at the NMJ the first and second deep lumbrical muscles from the hind paw of P140 SOD1G93A;β2m+/+ and SOD1G93A;β2m−/− mice were dissected whole mount to preserve the entire innervation pattern. After fixation, tissue was permeabilised in 4% triton-X 100 (Sigma; 0.1%) in 0.1 M PBS for one hour and blocked in 10% donkey serum (Jackson Immuno Research) and 0.1% triton-X 100 in 0.1 M PBS for a further hour at room temperature. Muscles were incubated over two nights at 4 °C in blocking solution with primary antibodies directed against neurofilament (DSHB, 2H3, 1:50) and GAP-43 (Millipore, AB5520 1:250) in order to visualize axons and regenerating neurons, respectively. Muscles were then washed twice for 30 min in 0.1% triton-X 100 and 10% donkey serum in 0.1 M PBS and incubated for 2 h with Alexa Fluor 488 donkey anti-mouse and Alexa Fluor 568 donkey anti-rabbit secondary antibodies in 0.1 M PBS (1:500, Life Technologies). Muscles were washed in 0.1 M PBS for 30 min and then exposed to Alexa Fluor 647 α-BTX (1:1000, Life Technologies) for ten minutes to label post-synaptic acetylcholine receptors. Muscles were then whole-mounted in Mowiol 488 (Sigma-Aldrich) on glass slides and cover-slipped for subsequent imaging. Imaging was performed on laser scanning confocal microscopes (Zeiss LSM700 and LSM800). At least 50 NMJs were individually categorized per muscle per mouse based on the level of GAP-43 expression at each one (minimum of 65 NMJs per muscle). GAP-43 levels were classed as distinct (bright and defined staining overlying the endplate), diffuse (faint and undefined staining, or only partially overlying the endplate) or devoid (no GAP-43 overlying the endplate), as previously described [31]. All analyses and quantifications were performed blind to the genetic status of the muscles.

Motor neuron counts in mouse spinal cords

For quantification of spinal MN somas, end-stage SOD1G93A;β2m+/+, SOD1G93A;β2m+/−, SOD1G93A;β2m−/− and age-matched WT;β2m−/− mice (2–4 animals per genotype and sex) were analysed. Spinal cords were cut at the lumbar level, mounted in OCT and sectioned (30 µm-thickness) in a cryostat. Lumbar sections were attached to Superfrost Plus slides (Thermo Scientific) and stored at −20 °C. Prior to staining, slides were air-dried, fixed in 25% ethanol for 2 min and Nissl-stained using 2% Cresyl Violet acetate (Sigma-Aldrich, in 25% ethanol, pH8) for 2 min. Slides were washed in water for 1 min and then dehydrated in ethanol 25, 50, 70, 95 and 100% for 2 min each. Finally, slides were moved to Xylene for 3 min, quickly air-dried, mounted using Mountex (Histolab) and kept at room temperature. MNs were counted in a minimum of six pairs of ventral horns at the lumbar level of the spinal cord. Cells were counted on a Zeiss Axio Imager M1 Upright microscope using the Q-capture software. In order to selectively count alpha-MNs, only cells located in the ventral horn with a cell body diameter greater than 20 μm and a clear nucleolus were included. To ensure accurate assessment of soma size the projected image in Q-capture was calibrated to match a grid which could be superimposed over the image of the ventral horn so that each box measured precisely 20 μm. This way only cells with a cell body diameter greater than 20 μm were counted.

Immunohistochemistry in human tissues

Paraffin-embedded human sections of 12 μm were incubated at 60 °C for 30 min, then moved to xylene for 10 min at room temperature. They were sequentially rehydrated with 100% ethanol (2 × 10 min), 95% ethanol (2 × 10 min), 70% ethanol (5 min), 50% ethanol (2 min), 25% ethanol (2 min), and finally rinsed with deionised water for 1 min. Slides were heated in citrate buffer (10 mM Sodium Citrate, Sigma-Aldrich S4641; 0.05% Tween 20, Sigma-Aldrich P7949; pH 6.0) at 95 °C for 20 min and washed in PBS for 30 min. Endogenous peroxidases were quenched using a 50% methanol, 3% hydrogen peroxide solution (H202; Sigma Aldrich, 216763) in PBS for 10 min, followed by a 5-min PBS wash. Sections were blocked with 10% serum in PBS with 0.1% Triton X-100 (Sigma Aldrich) for 1 h, and then incubated with primary antibody in PBS with 0.1% Triton X-100 and 10% serum for 72 h at 4 °C. Primary antibodies used for immunohistochemistry in human tissue were rabbit anti-GFAP antibody (Dako, Z0334, 1:200), mouse anti-HLA Class 1 ABC antibody (Clone EMR8-5, Abcam, ab70328, 1:100), mouse anti-pSer409/410 TDP-43 (Clone 11-9 CosmoBio, CAC-TIP-PTD-M01A, 1:5000) and mouse anti-CD68 (Abcam, ab955, 1:100). For HLA staining, both blocking and primary antibody incubation were carried out without Triton X-100. As negative controls for the antibody stainings we also included tissue sections that did not receive any primary antibodies, but only secondary antibodies. After washing three times in PBS, slides were incubated with Biotin-SP affinitypure donkey anti-rabbit IgG (Jackson ImmunoResearch, 711.065.152, 1:100) or anti-mouse IgG secondary antibody (Jackson ImmunoResearch, 715.065.151, 1:100) in PBS overnight at 4 °C. The avidin–biotin complex (ABC) (VECTASTAIN Elite ABC-HRP Kit, Vector Laboratories PK-6100) solution was applied for 1 h at room temperature, followed by washes in PBS (3 × 10 min). DAB staining was performed using the Vector DAB substrate kit (Vector Laboratories, SK-4100), with a final wash in ddH₂O for 10 min. Nuclei were counterstained with Myers hematoxylin (Sigma Aldrich, GHS132) for 2 min, rinsed in ddH₂O, then treated with 70% ethanol containing 36 mM HCl to remove background staining, followed by another ddH₂O rinse. Dehydration was completed with sequential ethanol treatments for 2 min each (25, 70, 75, 95, and 100% ethanol), xylene (2 × 2 min), and mounted using DPX new non-aqueous mounting medium (Sigma Aldrich, 1005790500). Images of spinal cord and brain sections were acquired with a Zeiss Axio Observer 7 microscope under bright‐field conditions using a 20× objective.

Protein level analysis in human tissue

The human post mortem cases and tissues used for quantification of anti-HLA and anti-GFAP stainings are listed in Table 1. For quantification of HLA staining intensity in MNs in spinal cord and midbrain the following procedures were followed. Specificity of the staining was evaluated by performing secondary antibody–only controls in all donors, using mouse secondary antibodies for HLA staining and rabbit secondary antibodies for GFAP staining. No specific staining was observed under these conditions in any tissue. This was determined by assessing the overall staining intensity across entire sections and comparing it to sections from the same patient in which primary antibodies were included. Neurons located in the ventral horn of the spinal cord were manually outlined using the ImageJ software (NIH, Bethesda, MD), and the intensity of the staining was measured in somas with an area of 300 μm2 or greater, which were considered to be MNs based on size and location, and the background staining was subsequently subtracted. OMNs were identified based on their anatomical location in midbrain tissue sections and morphological features, manually outlined in ImageJ, and staining intensity in somas was quantified after background subtraction. The absolute majority of OMN somas had an area ranging from 200 to 1200 μm2 (Supplementary Fig. 1e). The number of donor samples used and MNs quantified for HLA staining are outlined in Supplementary Table S1, as well as the sections used for anti-GFAP and anti-CD-68 staining in Supplementary Table S2.

Table 1 Information on human post mortem tissue samples used

For anti-GFAP and anti-CD-68 stainings around spinal MNs, images were taken in the ventral horn of the spinal cord where MNs could be identified, and the mean intensity of each image was measured using the ImageJ software. For GFAP quantification around OMNs, images were taken in midbrain sections, and staining intensity for each image was calculated as the average of the mean intensity from three equal-sized regions of interest. The number of donor tissue samples, and sections used from these as well as images captured and quantified are listed in Supplementary Table S2.

Statistics

Data was collected and analysed using GraphPad Prism software. Statistical significance was determined as follows. For the MNs size analyses the Kolmogorov–Smirnov test was performed. For the measurement of staining intensities in MNs the Mann–Whitney was performed. For correlations of intensity with MN size, Spearman correlation was performed. For the mRNA levels and NMJ analyses, experimental data were compared by 2-way ANOVA. For behavioural analysis, mice were compared using ANOVA (Kruskal–Wallis test) followed by Dunn–Bonferroni post hoc correction. The survival comparison was performed using Mantel-Cox test. For analyses of MN numbers, one‐way ANOVA followed by a post hoc Tukey was performed.