Glycogen accumulation modulates lifespan in a mouse model of amyotrophic lateral sclerosis

Abstract

Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease characterized by the progressive loss of motor neurons in the spinal cord. Glial cells, including astrocytes and microglia, have been shown to contribute to neurodegeneration in ALS, and metabolic dysfunction plays an important role in the progression of the disease. Glycogen is a soluble polymer of glucose found at low levels in the central nervous system that plays an important role in memory formation, synaptic plasticity and the prevention of seizures. However, its accumulation in astrocytes and/or neurons is associated with pathological conditions and aging. Importantly, glycogen accumulation has been reported in the spinal cord of human ALS patients and mouse models. In the present work, using the SOD1^G93A mouse model of ALS, we show that glycogen accumulates in the spinal cord and brainstem during symptomatic and end stages of the disease and that the accumulated glycogen is associated with reactive astrocytes. To study the contribution of glycogen to ALS progression, we generated SOD1^G93A mice with reduced glycogen synthesis (SOD1^G93A GS^het mice). SOD1^G93A GS^het mice had a significantly longer lifespan than SOD1^G93A mice and showed lower levels of the astrocytic pro-inflammatory cytokine Cxcl10, suggesting that the accumulation of glycogen is associated with an inflammatory response. Supporting this, inducing an increase in glycogen synthesis reduced lifespan in SOD1^G93A mice. Altogether, these results suggest that glycogen in reactive astrocytes contributes to neurotoxicity and disease progression in ALS.

Graphical Abstract

Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease characterized by the progressive loss of motor neurons in the spinal cord. Using the ALS mouse model expressing the G93A mutant of Superoxide Dismutase 1 (SOD1^G93A), we found that glycogen accumulates in the spinal cord during disease progression, and that increased glycogen is associated with reactive astrocytes. Genetic reduction of glycogen levels increases lifespan and alters inflammatory cytokines, while increasing glycogen levels leads to a more rapid disease progression. These results suggest that glycogen in reactive astrocytes contributes to neurotoxicity and disease progression in ALS. Graphical abstract created with BioRender.com.

Introduction

Amyotrophic lateral sclerosis (ALS) is an adult-onset neurodegenerative motor neuron disease with a lifetime risk of about 1 in 400.¹ ALS is characterized by the rapid degeneration of motor neurons, leading to muscle wasting, paralysis and loss of cognitive function. Patients usually survive only 2–5 years after onset.² Mutations in superoxide dismutase 1 (SOD1) are linked to 20% of familial ALS cases³ and aberrant SOD1 has also been implicated in sporadic ALS.⁴ The most widely studied mouse model of ALS expresses a transgene of human SOD1 with the G93A mutation (SOD1^G93A), leading to a toxic gain-of-function.⁴ The SOD1^G93A mouse recapitulates many aspects of the disease, including rapid motor neuron loss, paralysis and shortened lifespan.⁵

There is growing evidence that ALS is a non-cell autonomous disorder. Astrocytes and microglia contribute to neurodegeneration and the progression of the disease.⁶ Astrocyte dysfunction may trigger motor neuron loss through a variety of mechanisms: impaired glutamate uptake and potassium buffering, reduced metabolic support for motor neurons, increased excretion of neurotoxic factors, and production of oxidative and nitrative stress.⁷ Changes in energy metabolism at both the cellular and organismal level are a hallmark of ALS.^8,9 Glucose and lipid metabolism and mitochondrial dysfunction appear to modulate disease progression and are under investigation as therapeutic targets.^9–12

Glycogen is a soluble polymer of glucose that is synthesized in most cells.¹³ Although glycogen is less abundant in the central nervous system than in other organs, brain glycogen plays a dynamic role in cerebral processes such as learning and memory, synaptic plasticity and seizure prevention.^14–17 In astrocytes and neurons, glycogen is synthesized by the muscle isoform of glycogen synthase (GS). Both cell types have an active glycogen metabolism, although basal glycogen levels are much higher in astrocytes compared to neurons.^18,19

Increased glycogen has been reported in the spinal cords of ALS patients and SOD1^G93A mice.^20,21 In fixed spinal sections from ALS patients, periodic acid-Schiff (PAS) staining showed glycogen in motor neurons and glia of the grey matter as well as white matter glia.²¹ In the same report, increased glycogen levels in the spinal cord and brainstem were observed in end stage but not symptomatic SOD1^G93A mice. In another study, glycogen levels in the lumbar spinal cord of SOD1^G93A mice progressively increased from onset to end stage.²⁰

Several observations suggest that this increased glycogen accumulation may contribute to the pathophysiology of ALS. We have shown that excessive glycogen accumulation is pathological in both astrocytes and neurons. Transgenic expression of a constitutively active form of GS specifically in Purkinje neurons leads to cell death,²² and its expression only in astrocytes leads to astrocyte reactivity, microgliosis, and an increase of the expression of pro-inflammatory factors.²³ In fact, glycogen accumulation itself is a hallmark of reactive astrocytes.^24,25 In mouse models of Lafora disease (LD), an inherited, fatal childhood dementia and epilepsy, glycogen accumulates in the form of polyglucosan bodies known as Lafora bodies (LBs) in both astrocytes and neurons.²⁶ We demonstrated that glycogen accumulation underlies all the pathologic traits of LD, including astrogliosis, inflammation, behavioral abnormalities and seizures.²⁷ An astrocyte-specific knockout of GS prevented astrogliosis, microglial activation, and cerebral metabolic changes in a mouse model of LD, but neuronal LBs were still present and the mice were still susceptible to seizures.²³ These results indicate that astrocytic glycogen accumulation underlies neuroinflammation in LD but not epilepsy. Since glycogen accumulation in neurons and astrocytes underlies the pathophysiology of LD, glycogen accumulation may have pathological consequences in other neurological diseases such as ALS.

In the current study, we found elevated glycogen levels in the spinal cord and brainstem of SOD1^G93A mice at symptomatic and end stages of disease. Using a sensitive glycogen-specific antibody we showed that glycogen is found in healthy motor neurons of control spinal cords, but elevated glycogen in the SOD1^G93A spinal cord was primarily associated with reactive astrocytes. We found that SOD1^G93A mice heterozygous for GS (GS^het) lived significantly longer than SOD1^G93A mice with both GS alleles. The pro-inflammatory cytokine Cxcl10 was decreased in 120-day-old SOD1^G93A GS^het mice compared with SOD1^G93A mice. In contrast, constitutive overexpression of protein targeting to glycogen (PTG), an activator of glycogen synthesis, in SOD1^G93A mice led to a further increase in glycogen levels in the spinal cord and a reduced lifespan. These results indicate that glycogen accumulation contributes to ALS progression and suggest glycogen metabolism could be a therapeutic target in ALS.

Methods

Mouse models

All laboratory animal procedures were carried out following European Union (2010/63/EU) and Spanish (BOE 34/11370–421, 2013) regulations. They were approved by the Institutional Animal Care Committee of Universitat de Lleida and Barcelona Science Park’s Animal Experimentation Committee under institutional approval CEA-OH/9532/2. The study was not pre-registered. B6.Cg-Tg(SOD1*G93A)1Gur/J (SOD1^G93A) animals were purchased from Jackson Laboratories (JAX catalog stock number 004435, RRID:IMSR_JAX:004435). GS^het mice^28,29 and mice overexpressing PTG (PTG^OE mice)²⁷ were generated in our laboratory as previously described. Since the PTG^OE cassette is present at the Hprt locus in the X chromosome, we only analyzed PTG^OE male animals to avoid the confounding effects of X chromosome activation.

Animals were maintained in specific pathogen free (SPF) conditions in a 12/12h light-dark cycle, up to five animals per cage, with free access to water and standard chow diet. The mice were weighed and monitored weekly until they started showing signs of paralysis (altered gait, dragging toes). At this point, food pellets were placed on the floor of the cage and the mice were provided with HydroGel (ClearH2O, cat #70–01-5022) and monitored daily. To minimize animal suffering, mice were considered end-stage and humanely euthanized when they exhibited complete hindlimb paralysis and could no longer right themselves within 5 seconds when placed on their side. The method of euthanasia varied by experiment, as described below. The primary endpoint of this study was to determine lifespan, and most animals were euthanized at disease end-stage. Secondary endpoints were to analyze glycogen and the expression of cytokines are earlier disease stages or ages (Figure 1A). For the glycogen quantification (Figure 1B), the GS^het SOD1^G93A lifespan experiment (Figure 3), and qPCR analysis (Figure 4), mice were euthanized by cervical dislocation and decapitated. For glycogen immunostaining (Figure 1C–E and Figure 2), including mice used in the PTG^OE SOD1^G93A lifespan experiment (Figure 5), mice were first anesthetized by intraperitoneal injections of sodium thiopental (Braun, 100mg/kg), followed by transcardial perfusion with 10% neutral buffered formalin (NBF). Sodium thiopental results in rapid unconsciousness and is widely used as an anesthetic prior to euthanasia.

Figure 1. Glycogen accumulates in spinal cord and brainstem in SOD1^G93A mice.

(A) Flowchart of time points and analytical techniques used in this study. IHC, immunocytochemistry; IF, immunofluorescence. Schematic was made with BioRender.com. (B) Total glycogen content of brain, brainstem and whole spinal cord of early symptomatic, late symptomatic, end-stage SOD1^G93A mice compared to control littermates. Data represent a mean ± SEM (n=3–6 mice per group). Statistical significance was determined by Kruskal Wallis test followed by Dunn’s multiple comparisons test: * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, *** p ≤ 0.0001. (B) Representative immunostaining for glycogen in cortical and brainstem sections from end-stage SOD1^G93A mice and age-matched controls. Scale bars indicate 1 mm. (C) Representative glycogen staining in cervical, thoracic and lumbar spinal cord from end-stage SOD1^G93A mice (n=6 mice analyzed) and age-matched controls (n=9 mice analyzed). In the far-right panel, sections were pretreated with diastase prior to staining to determine specificity of the glycogen antibody. Scale bars indicate 500 μm. (D) High magnification of glycogen immunostaining in the ventral horn and ventral white matter −/+ diastase, corresponding to the lumbar sections in (C). Scale bars indicate 50 μm.

Figure 3. Heterozygous loss of GS increases lifespan of SOD1^G93A mice.

(A) Survival curves for male and female SOD1^G93A mice and SOD1^G93A mice constitutively lacking one allele of GS (GS^het). Significance was determined by a log-rank survival analysis: p=0.0025 for males, and p=0.0138 for females (n=9–15 per group). (B) Weekly weight loss in SOD1^G93A and SOD1^G93A GS^het mice compared to control and GS^het littermates. Data represent mean ± SEM (n=16–26 mice per group). Statistical significance was determined by multiple unpaired t-tests.

Figure 4. Heterozygous loss of GS affects pro-inflammatory factors in SOD1^G93A lumbar spinal cord.

Expression of C3 (A), Ccl2 (B), Cxcl10 (C), and Lcn2 (D) in lumbar spinal cord at 120 and 150 days of age. Statistical significance between genotypes was determined by two-way ANOVA followed by Fisher’s LSD test: * p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001. Data represent mean ± SEM (n=4–8 mice per group).

Figure 2. Glycogen granules in the SOD1^G93A spinal cord are associated with reactive astrocytes.

(A) Co-immunofluorescence staining for glial fibrillary acidic protein (GFAP) and glycogen in lumbar spinal cord sections from SOD1^G93A mice and control littermates. Nuclei were stained with DAPI and are shown in blue. (B) An example of thresholding used to define glycogen-positive and GFAP-positive areas in stained sections. Threshold boundaries are shown as white lines. For each whole section, the overlap between both areas is expressed as a percentage of the total glycogen-positive area in (C). Statistical significance was determined by an unpaired t-test: *** p ≤ 0.0001 (n=4–5 mice per group). In (A) and (B), scale bars indicate 50 μm.

Figure 5. Glycogen accumulation driven by overexpression of PTG leads to reduced lifespan in SOD1^G93A mice.

(A) Survival curves for male SOD1^G93A mice and SOD1^G93A mice overexpressing PTG (PTG^OE). Significance was determined by a log-rank survival analysis: p=0.0184 (n=4–8 per group). (B) Weight loss in SOD1^G93A and SOD1^G93A PTG^OE mice. Data represent mean ± SEM (n=4–6 per group). Statistical significance was determined by multiple unpaired t-tests. (C, D) Glycogen immunostaining in lumbar spinal cord of end-stage SOD1^G93A and SOD1^G93A PTG^OE mice. Scale bars indicate 500 μm in (C) and 50 μm in (D).

A total of 230 mice were analyzed in this study (Table 1): 17 mice were used for glycogen extraction (Figure 1B); 24 mice were used for immunostaining (Figures 1B–E, 2, 5B–C); 41 mice were used for RNA extraction and analysis (Figure 4); 164 mice total were used for lifespan analysis, 148 of which were used in the initial analysis (Figure 3) and 16 in experiments with PTG^OE expression (Figure 5A–B). Eight SOD1^G93A mice and seven of the mice expressing PTG^OE were used both in lifespan analysis and for immunostaining (Figure 5). Control, GS^het, and PTG^OE littermate controls were collected at approximately 23 weeks of age (158.8 ± 17.0 days). Disease monitoring and lifespan were performed in the order that the mice became available: we set up continuous breeders, genotyped animals at weaning age, monitored disease progression as they matured, and euthanized at the indicated time points during disease progression or at end-stage. All mice were monitored and euthanized between 9 AM and 4 PM. While monitoring lifespan and disease progression, the experimenters were blind to genotype. Unblinding was performed after disease endpoint was reached and samples were collected. Glycogen assays, qPCR and immunostaining were performed after a sufficient number of animals were obtained per genotype. Immunostaining was performed at the IRB Histopathology Facility, where the experimenters were blind to genotype. Glycogen and qPCR assays and statistical analyses were performed unblinded. No randomization methods were used. No exclusion criteria were pre-determined. We employed a minimal number of mice, according to the local guidelines of ethics and animal experimentation committees. Sample sizes were determined based on previous studies (glycogen quantification³⁶, weight and lifespan³⁷, and qPCR analysis²³).

Table 1.

Total number and gender distribution of mice used in this study.

Glycogen Quantification	Males	Females	Total
Control	2	4	6
SOD1G93A, early symptomatic	1	2	3
SOD1G93A, late symptomatic	0	3	3
SOD1G93A, end stage	2	3	5
All genotypes	5	12	17
IHC (IF)*	Males	Females	Total
Control	5 (2)	4 (3)	9 (5)
SOD1G93A, end stage	6 (2)	2 (2)	8 (4)
PTGOE	4	0	4
SOD1G93A PTGOE, end stage	3	0	3
All genotypes	18 (4)	6 (5)	24 (9)
Lifespan analysis	Males	Females	Total
Control	19	21	40
GShet	17	16	33
SOD1G93A	27	18	45
SOD1G93A GShet	12	26	38
PTGOE	4	0	4
SOD1G93A PTGOE	4	0	4
All genotypes	83	81	164
RNA analysis	Males	Females	Total
Control, 120 days	4	0	4
Control, 150 days	4	0	4
SOD1G93A, 120 days	5	0	5
SOD1G93A, 150 days	7	0	7
GShet, 120 days	4	0	4
GShet, 150 days	5	0	5
SOD1G93A GShet, 120 days	4	0	4
SOD1G93A GShet, 150 days	8	0	8
All genotypes	41	0	41
Total Mice Used	Males	Females	Total
All genotypes	131	99	230

^*Some mice used for immunohistochemistry (IHC) were also used for immunofluorescence (IF) staining. The numbers used for both IHC and IF are in parentheses. Eight SOD1^G93A, four PTG^OE and three SOD1^G93A PTG^OE mice were used in both lifespan analysis and IHC.

Glycogen quantification

Mice were quickly euthanized by cervical dislocation and decapitated. Spinal cords were rapidly removed by hydraulic extrusion³⁰ into cold saline and immediately flash frozen in liquid nitrogen. Whole brains and brainstems were rapidly dissected and flash frozen, and all samples were stored at −80°C. Brains were pulverized over liquid nitrogen. 4 volumes of 30% KOH were added directly to frozen, whole spinal cord, whole brainstem or aliquots of pulverized brain tissue, and samples were immediately boiled for 30 minutes. Glycogen was subsequently isolated by ethanol precipitation as previously described.³¹ Purified glycogen was digested with 0.5 mg/ml amyloglucosidase in 100 mM sodium acetate pH 4.8 and glucose was quantified using a highly sensitive fluorescence-based assay as previously described.^15,18

Immunohistochemistry and immunofluorescence staining

After mice were anesthetized and perfused with 10% NBF, sections of the spinal cord corresponding to the cervical, thoracic and lumbar regions³² were removed and fixed in 10% NBF for approximately 24 hours, and then transferred to phosphate buffered saline (PBS). After fixation, spinal cord segments were extracted from the spinal column, embedded in paraffin, cut into 3 μm sections using a Leica microtome. Sections were air dried and further dried at 60 °C overnight.

For glycogen immunohistochemistry (IHC), sections were dewaxed and treated with alpha-Amylase (AR171, Artisan Dako Agilent) or Envision Flex Wash buffer (K8000, Dako – Agilent) for 15 min at 40°C. Antigen retrieval was performed with citrate buffer pH6 for 20 min at 97°C using a PT Link (Dako – Agilent). Quenching of endogenous peroxidase was performed by 10 min of incubation with Peroxidase-Blocking Solution (S2023, Dako-Agilent). Nonspecific binding was blocked using 5% of goat normal serum (16210064, Life technology) with 2.5% BSA (10735078001, Sigma) for 60 min. Blocking of nonspecific endogenous mouse Ig staining was also performed using Mouse on mouse (M.O.M) Immunodetection Kit (BMK-2202, Vector Laboratories). The anti-glycogen mouse IgM antibody IV58B6^33,34 was used as primary antibody at 1:500 with EnVision FLEX Antibody Diluent (K800621, Dako-Agilent) with overnight incubation at 4°C. The secondary antibody used was polyclonal Goat Anti-Mouse at 1:100 (P0447, Dako, Agilent) incubated 30 min. Antigen–antibody complexes were revealed with 3–3′-diaminobenzidine (K3468, Dako). Sections were counterstained with hematoxylin (Dako, S202084) and mounted with Mounting Medium, Toluene-Free (CS705, Dako) using a Dako CoverStainer.

For immunofluorescence, sections were dewaxed and antigen retrieval was performed with citrate buffer pH6 for 20min at 97°C using a PT Link (Dako – Agilent). Quenching of endogenous peroxidase was performed with a 10 min incubation with Peroxidase-Blocking Solution (S2023, Dako-Agilent). Nonspecific binding was blocked using R.T.U Animal-Free (SP-5035, Vector) for 60min. Blocking of nonspecific endogenous mouse Ig staining was also performed using Mouse on mouse (M.O.M) Immunodetection Kit – (BMK-2202, Vector Laboratories). Sections were incubated overnight at 4°C with the primary antibodies mouse IgG1 anti-GFAP (MAB360, Merck Millipore, RRID:AB_11212597) and the mouse IgM IV58B6 diluted at 1:1000 and 1:500 with EnVision FLEX Antibody Diluent (K800621, Dako-Agilent). Secondary antibodies used were a goat anti-mouse IgG1 Alexa Fluor 568 (A21124, ThermoFisher, RRID:AB_2535766) and a goat anti-Mouse IgM (Heavy chain) Alexa Fluor 488 (A21042, ThermoFisher, RRID:AB_2535711) at 1:500 for 60 min. Samples were stained with DAPI (D9542, Sigma) and mounted with Fluorescence mounting medium (S3023 Dako). Specificity of staining was confirmed by staining with a mouse IgG1, Kappa Monoclonal (NCG01) isotype Control (ab81032, Abcam, RRID:AB_2750592) or a mouse IgM (PFR-03) Isotype Control (A1–10438, Thermo).

Brightfield and fluorescent images were acquired using a NanoZoomer-2.0 HT C9600 scanner (Hamamatsu, Photonics, France) with the 20X objective and coupled to a mercury lamp unit L11600–05 and using NDP.scan2.5 software U10074–03 (Hamamatsu, Photonics, France). Images were visualized with the NDP.view 2 U123888–01 software (Hamamatsu, Photonics, France) and ImageJ³⁵ (version 2.9.0, National Institutes of Health).

Fluorescent images were quantified in ImageJ as follows: sections were outlined manually to create a region of interest (ROI) and only data within the ROI were considered. Thresholding was performed on the glycogen and GFAP channels to create new ROIs corresponding to glycogen-positive or GFAP-positive areas. The overlap between the two ROIs was generated and the area of the overlap expressed as a total of glycogen-positive area per section. Analyses were performed using custom macros with fixed threshold values to avoid error and bias. Three sections from each mouse were used to determine an average value.

RNA extraction and RT-qPCR

For the analysis of cytokine expression by qPCR, we only collected and analyzed samples from male mice to exclude sexual dimorphism. RNA was extracted from lumbar spinal cord using TRI Reagent (Thermo Fisher Scientific, AM9738) following the manufacturer’s instructions. RNA concentrations were measured using a NanoDrop ND-1000 (Thermo Fisher Scientific). One microgram of RNA was used for retrotranscription employing TaqMan Reverse Transcription Reagent using random hexamers (Thermo Fisher Scientific, N8080234).

Briefly, RT-qPCR experiments were performed using a CFX96 instrument (Bio-Rad, Hercules, California, USA) with SYBR Select Master mix for CFX (Thermo Fisher Scientific, 4472937). Each 20 μL of reaction contained 4μL cDNA, 10 μL SYBR Select Master Mix, 0.2 nM of forward primer and 0.2 nM of reverse primer solutions and 4 μL PCR grade water. RT-qPCR run protocol was as follows: 50 °C for 2 minutes and 95 °C for 2 minutes, with the 95 °C for 15 seconds and 60 °C for 1 minute steps repeated for 40 cycles; and a melting curve test from 65°C to 95 °C at a 0.1 °C/s measuring rate. The following mouse-specific SYBR^® Green sets of primers (Sigma) were used: C3 (forward: 5′-TCCTGAACTGGTCAACATGG-3′; reverse: 5′-AAACTGGGCAGCACGTATTC-3′); Ccl2 (forward: 5′-AGGTGTCCCAAAGAAGCTGTAG-3′; reverse: 5′-TCTGGACCCATTCCTTCTTG-3′); Lcn2 (forward: 5′-CAGAAGGCAGCTTTACGATG-3′; reverse: 5′-CCTGGAGCTTGGAACAAATG-3′); Cxcl10 (forward: 5′-CCGTCATTTTCTGCCTCATC-3′; reverse: 5′-CTCGCAGGGATGATTTCAAG-3′) and β-actin (Actb), used as a housekeeping gene (forward: 5′- GTGACGTTGACATCCGTAAAGA-3′; reverse: 5′-GCCGGACTCATCGTACTCC-3′). Samples were run in duplicate to measure the cycle threshold (Ct). ΔCt was calculated as Ct_(Actb) − Ct_{(gene of interest)} for each sample. ΔCt values for control mice were averaged to calculate ΔCt_{(avg control)}. ΔΔCt was calculated as ΔCt_{(avg control)} − ΔCt_(sample). Results are expressed as relative expression (2^−ΔΔCt) as a percent of control.

Statistical analysis

All statistical analyses were performed in Prism 9 (GraphPad). For glycogen content, statistical significance was determined by Kruskal Wallis test followed by Dunn’s multiple comparisons test. For GFAP-glycogen colocalization, significance was determined by unpaired, two-tailed t-test. For cytokine expression measurements via qPCR, statistical significance between genotypes was determined by two-way ANOVA followed by Fisher’s Least Significant Difference (LSD), and gene expression changes over time in all genotypes were analyzed by two-way ANOVA followed by Sidak’s multiple comparisons test. For lifespan, significance was determined by a log-rank survival analysis. For weight changes in mice, statistical significance was determined by multiple nonparametric, unpaired, two-tailed t-tests (multiple Mann-Whitney tests). Statistical reports for all experiments are reported in Table 2. D’Agostino & Pearson tests were used to determine normality, and nonparametic tests were used for analyses that did not pass the normality test. No tests for outliers were performed.

Table 2.

All statistical reports.

Figure 1B : Glycogen content in brain, brainstem, and spinal cord
Kruskal-Wallis Test			Brain		Brainstem		Spinal Cord
p-value			0.0023		<0.0001		<0.0001
H value			10.84		13.13		13.02
degrees of freedom			3		3		3
Dunn’s Multiple Comparisons (p-values)			Brain		Brainstem		Spinal Cord
Control vs. Early Symp.			>0.9999		0.0306		0.0703
Control vs. Late Symp.			>0.9999		0.0124		0.0065
Control vs. End Stage			0.0232		0.901		0.3648
Early Symp. vs. Late Symp.			>0.9999		>0.9999		>0.9999
Early Symp. vs. End Stage.			0.0402		0.7733		>0.9999
Late Symp. vs. End Stage			0.5155		0.441		0.6458
Figure 2 : GFAP-associated glycogen in SOD1G93A spinal cord
Unpaired t-test			SOD1G93A vs. control
p-value			0.0007
t-value			5.708
degrees of freedom			7
Figure 3 : Lifespan and weight of SOD1G93A, GShet mice
Log-rank (Mantel-Cox) Test			Males	Females
p-value			0.0025	0.0138
chi square			9.15	6.067
degrees of freedom			1	1
Multiple Mann-Whitney Tests			minimum p-value	maximum  p-value		minimum q-value	maximum  q-value
Control vs. GShet Males			0.143279	0.969822		0.97952	0.97952
SOD1G93A vs. SOD1G93A GShet Males			0.087548	0.991527		0.434644	>0.999999
Control vs. GShet Females			0.027689	>0.999999		0.32161	>0.999999
SOD1G93A vs. SOD1G93A GShet Females			0.121526	>0.999999		0.791505	>0.999999
Figure 4 : Inflammatory factors in spinal cord
Two-way ANOVA	C3		Ccl2		Lcn2		Cxcl10
Source of Variation	%	p-value	%	p-value	%	p-value	%	p-value
Interaction	4.732	0.1781	7.901	0.0052	6.93	0.3048	12.27	0.0005
Time	4.216	0.0385	8.286	0.0003	16.01	0.0058	10.88	<0.0001
Genotype	49.89	<0.0001	52.73	<0.0001	16	0.0494	48.01	<0.0001
Fisher’s LSD	C3		Ccl2		Lcn2		Cxcl10
(p-values)	120 days	150 days	120 days	150 days	120 days	150 days	120 days	150 days
Control vs. SOD1G93A	0.0742	<0.0001	0.0142	<0.0001	0.0046	0.4466	0.0395	<0.0001
Control vs. GShet	0.5709	0.509	0.5565	0.7457	0.0473	0.9009	0.8232	0.8973
Control vs. SOD1G93A GShet	0.026	<0.0001	0.0108	<0.0001	0.3261	0.9869	0.025	<0.0001
SOD1G93A vs. GShet	0.0199	<0.0001	0.0029	<0.0001	0.3905	0.3399	0.0232	<0.0001
SOD1G93A vs. SOD1G93A GShet	0.5437	0.9458	0.7963	0.1227	0.0548	0.3478	0.7417	0.0144
GShet vs. SOD1G93A GShet	0.0065	<0.0001	0.0023	<0.0001	0.2953	0.8975	0.0148	<0.0001
Šídák's Multiple Comparisons Test (p-values)	C3 120 vs. 150 days		Ccl2 120 vs. 150 days		Lcn2 120 vs. 150 days		Cxcl10 120 vs. 150 days
Control	>0.9999		>0.9999		>0.9999		>0.9999
SOD1G93A	0.0392		<0.0001		0.047		<0.0001
GShet	>0.9999		0.997		0.1077		>0.9999
SOD1G93A GShet	0.1978		0.0106		0.6859		0.0059
Figure 5 : Lifespan and weight of SOD1G93A, PTGOE mice
Log-rank (Mantel-Cox) Test			Males
p-value			4.94
chi square			1
degrees of freedom			0.0262
Multiple Mann-Whitney Tests			SOD1G93A vs. SOD1G93A GShet Males
minimum p-value			0.285714
maximum p-value			>0.999999
minimum q-value			>0.999999
maximum q-value			>0.999999

Results

Glycogen accumulates in the spinal cord of SOD1^G93A mice

To determine when and where glycogen accumulates in the SOD1^G93A model of ALS, we first quantified the glycogen content of brain, brainstem and spinal cord samples from early symptomatic, late symptomatic, and end stage SOD1^G93A mice and littermate controls (Figure 1B). Mice in the early symptomatic stage (152.7 ± 0.6 days) had very mild changes in hindlimb splay or gait, mice in the late symptomatic stage (163.0 ± 3.5 days) had signs of paralysis in at least one limb but were still able to walk, and ES mice (174.5±11.9 days) showed complete hindlimb paralysis and were unable to right themselves within 5 seconds.

In brain, while glycogen levels were unchanged in early and late symptomatic animals compared to controls, a significant reduction in glycogen in end stage SOD1^G93A mice was observed (Figure 1B). In brainstem, a large increase in glycogen was measured at early and late symptomatic stages in SOD1^G93A mice compared to controls (p=0.0306 and p=0.0124, respectively). End stage glycogen levels in the SOD1^G93A brainstems were not significantly different than in control brainstems. In spinal cord, glycogen levels appeared elevated at all stages in SOD1^G93A mice, but significant differences compared to controls were only found in late symptomatic SOD1^G93A mice (Figure 1B, p=0.0065).

To determine the localization of glycogen, samples from control and end stage SOD1^G93A mice were also collected and fixed for immunostaining. We stained sections of brain, brainstem, and spinal cord with the glycogen antibody IV58B6^33,34, which is more specific and sensitive than PAS staining. In control mice, glycogen was observed sparsely in the hippocampus and cortex, but very little was found in the brainstem (Figure 1C). In end stage SOD1^G93A mice, less glycogen staining was observed in coronal brain sections, but much higher glycogen staining was observed in brainstem (Figure 1C). In control spinal cords, glycogen was observed in grey and white matter of cervical, thoracic and lumbar sections (Figure 1D–E). In all regions, large motor neurons of the ventral horn stained strongly for glycogen, and glycogen signal was also observed in the surrounding neuropil of the grey matter and throughout the white matter (Figure 1E). In end stage SOD1^G93A mice, all spinal cord sections showed an increase in glycogen staining throughout both the grey and white matter (Figure 1D), accompanied by the loss of motor neurons and vacuolization (Figure 1E). Staining was completely eliminated with diastase pretreatment, confirming the specificity of the antibody for glycogen (Figure 1D–E).

We next performed co-immunofluorescence for glycogen and glial fibrillary acidic protein (GFAP), a marker for astrocytes, on the spinal cord sections. In control animals, most of the glycogen granules in the lumbar ventral horn were primarily localized to GFAP-negative cells, with the characteristic morphology of lumbar motor neuron somata (Figure 2). In grey matter of control animals, a few glycogen granules were observed associated with GFAP-positive processes of astrocytes. In SOD1^G93A animals, astrogliosis was very evident in both grey and white matter, which also appeared to have more glycogen granules (Figure 2A).

To determine whether the increased glycogen was associated with GFAP-positive astrocytic processes, we quantified the colocalization of glycogen and GFAP immunostaining using image thresholding for each channel (Figure 2B). The area of overlap between the two channels was expressed as a percentage of total glycogen-positive area. SOD1^G93A spinal cords showed a significant increase in colocalization of glycogen and GFAP immunostaining (Figure 2C, p=0.0007), suggesting that the increase in glycogen in SOD1^G93A spinal cords is associated with reactive astrocytes.

Heterozygous loss of GS increases lifespan in SOD1^G93A mice

Though multiple groups have reported increased levels of glycogen in ALS, it is not clear whether glycogen affects ALS progression or is only a secondary effect of cellular dysfunction. To determine whether glycogen contributes to disease progression, SOD1^G93A mice were crossed with mice lacking one allele of glycogen synthase (GS^het), which present roughly 50% reduced levels of GS protein, activity, and glycogen levels in all cells³⁸. We previously demonstrated that deleting one allele of GS in the central nervous system (CNS) significantly decreases glycogen accumulation and ameliorates disease progression in LD mouse models.²⁷ SOD1^G93A mice lacking both GS alleles were not generated because GS knockout leads to perinatal lethality due to abnormal cardiac development and lung collapse during birth.^28,38 We monitored the lifespan and weight of these mice as well as wild-type controls and GS^het littermates. Both male and female SOD1^G93A GS^het mice lived significantly longer than SOD1^G93A mice (p=0.0025 and p=0.0138, respectively) (Figure 3A). However, no significant difference in weight loss was observed between SOD1^G93A GS^het and SOD1^G93A mice (Figure 3B).

Markers of neuroinflammation are altered in SOD1^G93A mice with heterozygous loss of GS

We have previously demonstrated that the pathogenic accumulation of glycogen in astrocytes in the brain induces neuroinflammation in LD, namely increased expression of C3, Ccl2, Cxcl10 and Lcn2.²³ These pro-inflammatory genes are expressed by reactive astrocytes, activate microglia, and have neurotoxic effects in various pathological contexts.^39–41 These genes are also upregulated in ALS rodent models, patient tissues and patient cerebrospinal fluid.^35,38–41 We analyzed the effect of glycogen reduction on the expression of these genes in the lumbar spinal cord of SOD1^G93A mice at 120 days, when previous studies report highly upregulated C3, Ccl2 and Cxcl10 in this mouse model,^42,43 and at 150 days, the approximate age of our early symptomatic mice in which we found significantly elevated glycogen (Figure 4). Two-way ANOVA showed significant differences both between genotypes and the time points analyzed (Figure 4, Table 2). Within each time point, statistical differences are indicated with asterisks in Figure 4. Within each genotype, statistical differences between the 120 and 150 day time points are listed in Table 2.

Compared to controls, SOD1^G93A and SOD1^G93A GS^het mice show increased expression of C3, Ccl2, and Cxcl10 within each time point (Figure 4A, B, and C). Ccl2 and Cxcl10 expression also significantly increased over time in the lumbar spinal cords of SOD1^G93A and SOD1^G93A GS^het mice (Table 2). At 150 days, SOD1^G93A GS^het mice expressed significantly lower levels of Cxcl10 than SOD1^G93A mice (Figure 4C). At 120 days, but not 150 days, Lcn2 was higher in SOD1^G93A mice than in controls, but there was no significant difference in Lcn2 levels between controls and SOD1^G93A GS^het mice at either time point (Figure 4D). Interestingly, SOD1^G93A mice showed a significant reduction in Lcn2 levels at 150 days compared to 120 days (Figure 4D, Table 2). Overall, SOD1^G93A GS^het mice have lower levels of Cxcl10 at 150 days compared to SOD1^G93A mice. These results indicate that both SOD1^G93A mice and SOD1^G93A GS^het mice show an upregulation of inflammatory cytokines and neurotoxic factors released by astrocytes. Partial reduction of glycogen synthesis has a limited effect on this inflammatory profile.

Increased glycogen reduces lifespan in SOD1^G93A mice

We next sought to evaluate the effect of increased glycogen accumulation on ALS progression. To that aim we crossed SOD1^G93A mice with a previously generated mouse line overexpressing PTG (PTG^OE), which has elevated glycogen in all tissues including brain.²⁷ These mice showed a faster disease progression, with an average lifespan of 14 days less than SOD1^G93A littermates (Figure 5A, p=0.0184) but no significant difference in weight loss (Figure 5B). Histological analysis showed increased glycogen levels in the PTG^OE spinal cords, particularly in the motor neuron somata and grey matter neuropil (Figure 5C). SOD1^G93A PTG^OE spinal cords showed very high levels of granular glycogen staining (Figure 5C). These results confirm that increasing glycogen levels accelerates disease progression in ALS.

Discussion

Our discovery that excessive accumulation of glycogen in astrocytes induces neuroinflammation in LD²³ led us to investigate whether this pathologic mechanism also plays a role in other neurodegenerative conditions that course with glycogen accumulation. ALS is a non-cell-autonomous disease in which astrocytes and microglia are known to contribute to pathogenesis ^6,7,44,46. Altered energy metabolism is also a central theme in ALS^8,9,46,47. Therefore, glial cells and metabolic pathways are being actively investigated as therapeutic modulators of the disease.⁴⁶ In the current study, using both biochemical quantification and immunostaining with a glycogen antibody that is more specific than PAS staining used by previous groups^20,21, we reported elevated glycogen throughout the brainstem and spinal cord at all stages of ALS progression in the SOD1^G93A model. This glycogen is associated with reactive astrocytes. We discovered that constitutive reduction of GS and glycogen levels increases lifespan. The reduction of GS in SOD1^G93A mice is also associated with lower levels of the pro-inflammatory cytokine Cxcl10. Furthermore, we showed that increasing glycogen levels through the overexpression of PTG led to a more rapid disease course.

In the healthy spinal cord, glycogen is also abundant in motor neurons. While we cannot exclude a role for aberrant neuronal glycogen metabolism in the early stages of the disease, our results implicate pathogenic effects of glycogen in astrocytes. Our results also demonstrate that elevated glycogen is not just an epiphenomenon in ALS but appears to play a role in ALS pathophysiology. It has been suggested that increased glycogen may slow ALS progression because rapamycin, which lowered glycogen levels in a mouse model of Pompe disease, accelerated ALS progression.²¹ However, our data show that increased glycogen reduces lifespan and exacerbates ALS progression, while reducing its synthesis has a beneficial impact.

Metabolic dysfunction has been widely reported in ALS and other neurodegenerative disorders and likely contributes to neurodegeneration.^46–48 PET imaging, proteomics from patient fibroblasts, and expression profiling indicate a reduction in glucose utilization and glycolysis in the CNS of ALS patients, although it is not known whether these changes occur in motor neurons or glia.⁴⁷ Astrocytes of SOD1^G93A mice have defects in mitochondrial function and increased oxidative stress that promote motor neuron toxicity, but this was prevented by mitochondria-targeted antioxidants.⁴⁹ A recent study showed increased acetate oxidation in the spinal cord of the SOD1^G93A mouse model of ALS, suggesting astrocytic involvement in ALS pathogenesis.⁵⁰ Another study reported a reduction in transcripts related to carbohydrate metabolism, glycolysis, and lactate efflux in astrocytes from SOD1^G93A mice. Lactate supplementation prevented the loss of motor neurons when cultured with SOD1^G93A astrocytes.⁵¹ Although some studies suggest that lactate derived from astrocytic glycogen directly fuels neurons,⁵² this hypothesis is widely debated and other evidence suggests astrocytic glycogen is primarily utilized to fuel astrocytes themselves.⁵³ Nevertheless, defects in astrocytic metabolism seem to contribute to their neurotoxic role in ALS. As an integral player in cellular metabolism, it is reasonable to hypothesize that glycogen also plays a contributing role. Metabolic targets are now being explored to delay onset or progression of ALS in mouse models.⁴⁶

Astrocytic glycogen could accumulate in the spinal cord of SOD1^G93A mice because of reduced glycogen degradation or an increase in synthesis, possibly as a result of impaired glycolysis, or both. Glycogen is converted to glucose by glycogen phosphorylase (GP) in the cytosol and acid α-glucosidase (GAA) in lysosomes. In the spinal cord of symptomatic SOD1^G93A mice, both protein and mRNA levels of the brain isoform of GP are reduced²⁰ and GAA activity decreases with disease progression.²¹ Induced astrocytes and neurons made from fibroblasts of ALS patients also showed reduced metabolic flexibility and reduced levels of GP protein.⁵⁴ Glucose derived from glycogen can also be utilized via the pentose phosphate pathway to produce NADPH, which plays an important role in reducing oxidative stress. Reduced pentose phosphate activity has been observed both in SOD1^G93A mice and SOD1^G93A -expressing motor neuron cells lines.^55,56 Our results suggest that a reduction in glycogen synthesis is beneficial, likely because it has the same net effect on metabolism as increased GP or GAA. In theory, less glycogen may increase the availability of metabolites (to either astrocytes, neurons or both) which can be used for energy production or to combat oxidative stress.

Astrocytes undergo metabolic changes when they become reactive.⁵⁷ For example, exposure to lipopolysaccharide increased glycolytic activity and mitochondrial respiration, while inhibition of glycolysis prevented astrocytic inflammatory cytokine release.⁵⁸ In primates and mice, glycogen accumulates in reactive astrocytes in response to injury or ischemia, and enhancing glycogenolysis mediated neuroprotection in a mouse model of ischemia and reperfusion.^24,25,27 Ectopic accumulation of astrocytic glycogen leads to reactive astrogliosis, characterized by increased GFAP staining, microglial reactivity, and the production of astrocytic pro-inflammatory cytokines.²³ Astrocytic glycogen metabolism is also connected to glutamine metabolism, and reduced astrocyte glutamine synthesis is often observed in neurological diseases, including ALS.⁵⁹ In astrocytes, glycogen supports glutamine synthesis and glycogen accumulates with inhibition of glutamine synthetase.⁶⁰ Although glutamine synthetase levels are unchanged in SOD^G39A spinal cords and astrocytes,⁶¹ SOD1^G93A astrocytes display a decrease in glutamate release⁶² that may be metabolically linked to the elevated astrocytic glycogen that we and others have observed.

CXCL10 is a chemokine ligand secreted by immune cells and reactive astrocytes in response to IFN-γ signaling.^63,64 CXCL10 binds to the receptor CXCR3, which is primarily expressed on leukocytes, but also expressed in neurons,⁶⁵ astrocytes and microglia.⁶⁶ Thus, CXCL10 could promote neurotoxicity, reactive astrogliosis and/or microglial activation in an autocrine or paracrine manner via CXCR3. Lcn2 encodes lipocalin 2 (Lcn2), an inducible factor secreted by a variety of cell types.⁶⁷ In rat models expressing neuronal FUS or TDP-43 mutants, Lcn2 is released by astrocytes and is toxic to primary neurons.³⁹ Our data suggests that this factor is likely more important in the early stages of ALS progression in SOD1^G93A mice. A complete elimination of glycogen in the CNS or in astrocytes could have a stronger effect on inflammatory cytokines in this ALS mouse model and may show an even greater beneficial impact on SOD1^G93A lifespan. Further studies are necessary to determine the mechanism by which glycogen metabolism impacts the release of inflammatory factors and at what stage of ALS progression these factors are most consequential.

This study supports a growing body of evidence that altered metabolism in glial cells plays a role in ALS. Preventing glycogen synthesis or enhancing its degradation in spinal cord could provide a therapeutic benefit for ALS patients, although further studies are required to determine how astrocytic glycogen accumulation affects energy availability and cytokine release. Glycogen-targeting therapeutic strategies have already shown preclinical success in LD mouse models and could be tested in ALS models.^31,68,69 Astrocytic glycogenolysis is also involved in glutamate uptake, potassium buffering, calcium homeostasis.⁷⁰ Since these processes are impaired in ALS,⁷ further investigations are warranted to define the connection between excess glycogen and astrocytic dysfunction.

Abbreviations:

ALS

amyotrophic lateral sclerosis

C3

complement component 3

Ccl2

chemokine ligand 2

CNS

central nervous system

CXCL10

C-X-C motif chemokine ligand 10

CXCR3

C-X-C Motif Chemokine Receptor 3

GAA

acid α-glucosidase

GFAP

glial fibrillary acidic protein

GP

glycogen phosphorylase

GS

glycogen synthase

LB

Lafora bodies

Lcn2

Lipocalin 2

LD

Lafora disease

NADPH

reduced nicotinamide adenine dinucleotide phosphate

NBF

neutral buffered formalin

OE

overexpression

PAS

periodic acid-Schiff

PTG

Protein Targeting to Glycogen

qPCR

quantitative polymerase chain reaction

ROI

region of interest

SOD1

superoxide dismutase 1