Sensitive quantification of α-glucans in mouse tissues, cell cultures, and human cerebrospinal fluid

The soluble α-polyglucan glycogen is a central metabolite enabling transient glucose storage to


Introduction
Glycogen, a branched polymer of glucose, is a central part of cellular metabolism in many animal tissues, such as liver, muscle, brain, kidney, adipose tissue, and heart (1). Though the exact role and quantity of glycogen is tissue-specific, in general its function is to store glucose in an enzymatically accessible macromolecule inside of cells and to release it when metabolically required (2).
De novo synthesis of the α-glucan glycogen is usually initiated in the cytosol by glycogenin (3), which first auto-glycosylates and then elongates the nascent glucan chain forming α1,4 interglucose linkages, using UDP-glucose as a glucosyl donor. Chain elongation is then continued by glycogen synthase (1). As longer unbranched glucan chains form double helices and precipitate (4), glycogen branching enzyme is required to introduce α1,6 linkages (branching points) and maintain solubility. The concerted action of chain elongation and branching results in a relatively compact yet water soluble glycogen molecule which can reach sizes of ~55,000 glucosyl units. Glycogen degradation is mediated in the cytosol by glycogen phosphorylase and debranching enzyme, generating largely glucose 1-phosphate, and in the lysosome by αglucosidase, generating mostly glucose (1).
Genetic defects in enzymes with a role in glycogen metabolism are responsible for glycogen storage diseases (GSD) and emphasize the centrality of a functioning glycogen metabolism. In some GSDs, enzymes directly involved in glycogen synthesis or its degradation, as the ones described above, can be affected. Other GSDs are caused by deficiencies in enzymes that mediate pathways immediately feeding into glycogen synthesis, or by lack or impairment of enzymes required for downstream processing of glycogen degradation products (5). Consequently, most of the GSDs are characterized by an abnormal quantity of glycogen in the affected tissue (2). In some cases, the quality of glycogen is also affected: in various tissues from patients of Andersen disease (6), Lafora disease (LD) (7), AMPK deficiency (8) and phosphofructokinase deficiency (9) an insoluble form of glycogen accumulates in the form of so-called polyglucosan bodies, which appear to be metabolically inert. For LD, caused by mutations in either EPM2A (laforin) or NHLRC1 (EPM2B, malin), it has been shown that such aggregates of insoluble glycogen are the driver of the disease progression (10). Insoluble brain glycogen has also been described as corpora amylacea in association with ageing (11).
Appropriate methods for quantification of soluble and insoluble glycogen are essential tools in the study of glycogen metabolism in health and disease. Recently, soluble and insoluble glycogen was isolated from skeletal muscle of three different GSD mouse models with polyglucosan body accumulation (12). The methodology was designed to enable structural analysis of the glycogen fractions. For sole quantification of soluble and insoluble glycogen, especially in younger mice with less progressed glycogen accumulation, simpler methods requiring less tissue are favorable. Furthermore, sensitive detection of glycogen in cultured cells is prerequisite for the characterization of candidate therapeutics targeting, for instance, glycogen synthase in GSDs with glycogen overproduction (13). The detection of CSF biomarkers in neurodegenerative diseases is common (14). In urine of Pompe disease (GSDII) patients glycogen degradation products have been found abnormally high (15,16). GSD-related changes in brain glycogen, e.g. in LD, suggest the sensitive detection of glycogen or glycogen-derived glucose as a viable biomarker in CSF. The work in this paper describes methodological advances on all these fronts.
Generally, glycogen can be detected by immunohistochemistry using periodic acid-Schiff stain (PAS). The presence of large PAS-positive aggregates in skin biopsies of suspected LD cases has long been decisive in the diagnosis of the disease (10,17). Quantitative comparison between PAS-stained samples requires strictly controlled tissue treatment and staining procedures which can be achieved in animal studies but hardly in a clinical setting.
Immunofluorescence (IF) detection of glycogen utilizes a glycogen-directed antibody (18) or other glycogen-binding proteins that can be labeled by secondary antibodies (19). This methodology is generally restricted to cell culture experiments. Also, IF-based methods mostly focus on the localization of the antigen and are, unless extensively standardized, considered at best semi-quantitative (20).
Methods for more precise glycogen quantification in animal tissue usually comprise the steps 1) tissue lysis and glycogen extraction, 2) glycogen degradation, and 3) glucose determination. Extraction from pulverized tissue in hot 30% potassium hydroxide has been widely used for glycogen quantification (21)(22)(23). Other extraction procedures include homogenization in cold 10% trichloroacetic acid (24)(25)(26) or in Tris-based glycogen isolation buffer (12,27). Extraction in a suitable buffer allows separation of soluble and insoluble glycogen, followed by glycogen quantification without further purification of the glycogen by repeated ethanol precipitations (12), which are otherwise required to remove excess amounts of potassium hydroxide or trichloroacetic acid. Repeated precipitation of glycogen also removes low-molecular compounds such as blood-derived glucose (28), which can interfere with the determination of glycogen-derived glucose. Hydrolysis of virtually all interglucose linkages in glycogen can be achieved by amylolytic treatment using amyloglucosidase, either alone (12,21,27) or in combination with amylase (25). Hydrolysis in hot 0.7-1 M hydrochloric acid or 2 M trifluoroacetic acid (29)(30)(31) can also be utilized to degrade glycogen prior to glucose quantification.
If glucose is quantified in glycogen hydrolysates after several rounds of ethanol precipitation, the detected amount of glucose is considered to represent the glycogen quantity, as the amount of free (i.e. nonglycogen-derived) glucose is negligible (22,23). Beside macromolecular glycogen, tissue extracts that have not undergone repeated ethanol precipitation contain free glucose (e.g. from blood). In such samples it is therefore crucial to determine the amount of free glucose present before glycogen hydrolysis. Logically, the relative contribution of free glucose to posthydrolysis glucose (free and glycogen-bound glucose) is higher in tissues with low glycogen amount, such as brain or heart (22,23). The difference between glucose contents determined with and without glycogen degradation is the total amount of α-glucan, a proxy for glycogen, which includes glycogen but possibly also smaller α-glucan chains that would be excluded by ethanol precipitations (32).
The sensitivity of the entire glycogen quantification procedure is greatly dependent on the sensitivity of the glucose assay used after glycogen degradation. Simple enzymatic assays detect products of enzymatic glucose conversion. For instance, spectrophotometrically detectable NAD(P)H is generated during the conversions of glucose by hexokinase (HK) and glucose 6-phosphate dehydrogenase (G6PDH) in the presence of ATP and NAD(P) (12,23,25). Alternatively, various dyes can be detected after reaction with hydrogen peroxide as generated by glucose oxidase (27,33). Most of these assays are suitable for glycogen determination in many animal tissues when sufficient tissue is available. If the glycogen content is very low or the amount of available tissue is limited, more sensitive methods for glucose determination are advantageous.
ATP can be quantified with very high sensitivity when utilizing luminescence detection during the ATPdependent oxidation of luminol to oxiluminol as mediated by firefly luciferase (34). ATP detection is now widely used for sensitive cell viability analyses (35,36).
In this paper three methodological advancements are described. First, a simple pre-extraction incubation step is demonstrated to reliably reveal an inert (degradation resistant) fraction of glycogen in skeletal muscle (5 months old) and hippocampus (3 months old) of mouse models of LD. Second, a glucose assay was established that detects glucose-dependent ATP depletion utilizing firefly luciferase-mediated generation of luminescence. Third, a method is presented that enables the detection of minute amounts of α-glucan material in commercial samples of human cerebrospinal fluid (CSF).

Quantification of inert α-glucans in muscle of LD mice
We aimed to develop a simple method that allows quantification of the soluble and insoluble portion of glycogen in small amounts of mouse tissue. We tested whether pre-extraction exposure to physiological temperatures of muscle tissue will lead to the removal of soluble glycogen, whereas insoluble will remain. Ground skeletal muscle from 5 month-old WT and malin-deficient mice (Epm2b -/-, mouse model of LD) was pooled from 3 animals each and divided into equal aliquots. These were incubated at 37 °C for various times between 0 and 6 h before they underwent glycogen extraction and quantification as illustrated in Fig. 1A. In both WT and Epm2b -/samples glycogen degraded rapidly, showing reduction of glycogen levels by over 90% within 15 min of pre-extraction incubation. After about 1 h the content of remaining glycogen stabilized in WT and Epm2b -/samples. However, while there was almost no glycogen remaining in WT, the content in the mutant tissue stabilized at a consistently higher level and remained essentially unchanged within the following 5 h of incubation (Fig. 1B). Subsequently, muscle tissue from 6 individual mice per genotype was subjected to 0 and 1.5 h pre-extraction incubation prior to glycogen quantification. In both genotypes the total glycogen content (0 h) was on average between 15 and 16 umol glucosyl units per gram fresh weight. Pre-extraction incubation led to strongly reduced glycogen contents in all mice. The WT and mutant tissues contained on average 0.08 and 1.26 umol glucosyl units per gram fresh weight, respectively. These values account for about 0.5% and 7.6% of the average total glycogen in the respective genotypes (Fig. 1C).
The average total glycogen content measured in muscle of 5 month-old WT and Epm2b -/mice is within the range of what has previously been published for skeletal muscle in WT mice (12,22,23,31,37,38) with the exception of a recent publication which, using GC/MS following a novel extraction protocol, demonstrated muscle glycogen contents less than 10% of all previous findings (26). Glycogen accumulation in muscle of 9-12 month-old LD mice was previously clearly detected, but changes in total muscle glycogen content were mostly undetectable in mice younger than 5 months (22,37). Pre-extraction incubation of muscle tissue as established here reveals the presence of an abnormal kind of glycogen in skeletal muscle of LD mice already at 5 months of age. This abnormal glycogen is absent in WT mice and seems to be inert to the degradation process that removes essentially all WT glycogen during pre-extraction incubation. This degradation is expected to be mediated by endogenous enzymes that reactivate when the tissue sample warms up. The only enzymes known to degrade glycogen in vivo are glycogen phosphorylase and the lysosomal alpha-glucosidase, which are expressed in essentially all glycogen storing tissues. Phosphorylase has been shown stimulated under hypoxic and hypoglycemic conditions (39,40). Both conditions can safely be assumed fulfilled in a tissue piece cut off from blood circulation. During pre-extraction incubation of powdered muscle tissue blood-derived pancreatic amylase and lysosomal alpha-glucosidase could certainly contribute to glycogen degradation, as they may gain access to cytosolic components due to the breaking of cells and compartments during tissue pulverization. Determination of inert muscle glycogen can be done with less than 30 mg of muscle tissue and will provide unprecedented insight when polyglucosan accumulation is studied in models of GSDs at a younger age, such as in investigations into disease onset or early intervention.

Quantification of inert α-glucans in hippocampi of LD mice
Historically, muscle and brain are the tissues mainly studied in LD. In addition to muscle we examined the presence of inert glycogen in brain tissue, in particular the hippocampus. We used the alternative mouse model of LD (Epm2a -/-), which is indistinguishable from Epm2b -/mice with respect to accumulation of insoluble glycogen in the brain (23). Hippocampi of 3 month-old Epm2a -/and WT mice were subjected to pre-extraction incubation and subsequent glycogen quantification. Hippocampi were dissected from each hemisphere and frozen in liquid nitrogen separately. From each mouse one hippocampus was incubated for 30 minutes, the other for 60 min (Fig. 1D). The brain tissue was incubated at 25 °C as we indication that room temperature exposure of brains suffices to cause a substantial decrease in brain glycogen, such as when dissection and freezing of brains was delayed after cervical dislocation of mice (data not shown). After pre-extraction incubation glycogen was extracted by homogenization in a buffer that provided favorable conditions for the subsequent amylolytic glycogen degradation. Glucose was determined using a conventional G6PDH/HK-based glucose assay before (free glucose) and after amylolysis (free glucose and that bound to α-glucans) (Fig. 1A). The amount of free glucose is similar in identically treated WT and mutant hippocampi, but it seems to be consistently reduced when pre-extraction incubation is extended to 60 minutes. The amount of post-amylolysis glucose tends to be greater than that of free glucose in all hippocampi. However, this effect is much more pronounced in Epm2a -/compared to WT, where the post-amylolysis content is mostly driven by the amount of free glucose (Fig. 1E). The differences of pre-and post-amylolysis glucose contents define the samples' α-glucan content, which is clearly increased in the mutant hippocampi but largely unaffected by an extended incubation time (Fig.  1F).
The results demonstrate that Epm2a -/hippocampi contain inert α-glucan material, which is unaffected by the degradation processes that lead to the nearcomplete depletion of α-glucans in WT hippocampi. It is expected that glycogenolysis in intact brain tissue is mostly phosphorylase-driven (see above). The differences between the WT and Epm2a -/amounts are in line with accumulated glycogen previously described in 3 month-old LD mice brains (22,37), further validating that the inert glucans we measured directly correspond with the abnormal glycogen that accumulates in LD. As in muscle, a pre-extraction incubation as devised removes all soluble and non-inert glycogen, revealing a very robust difference between diseased and WT animals already at a young age. As studies in young mice can be completed faster and therefore present a smaller financial burden, this method will be beneficial in studies assessing efficacy of therapeutic approaches for prevention or removal of insoluble brain glycogen, such as in LD (10,31) or adult polyglucosan body disease (41). The sensitivity of the described method is sufficient to determine inert αglucans in pieces of mouse brain weighing less than 20 mg, e.g. one hippocampus.
As the enzymatic processes leading to glycogen degradation are expected to be similar in any dissected tissues, we anticipate that the method described here can be used to quantify 'inert' glycogen in tissues other than brain and skeletal muscle or in other mouse models. However, it is advisable to perform adequate control experiments for confirmation in each tissue and mouse line.

Increased sensitivity of glucose quantification by bioluminescence detection of ATP depletion
A sensitive method for glucose quantification allows the use of very low source material for αglucan/glycogen quantification. Previously described enzymatic assays require a minimum of 500-1000 pmol glucose per assay for reliable detection. ATP can be quantified by detection of bioluminescence generated in the firefly luciferase-mediated reaction of luminol to oxiluminol, the sensitivity of which exceeds the detection for instance of NADPH by more than 3 orders of magnitude (34,42).
The principle of a glucose-dependent ATP depletion assay with bioluminescence detection is illustrated in Fig. 2A. Authentic glucose standards were subjected to a two-step protocol: 1) the hexokinase (HK) reaction was conducted with a concentration of ATP that exceeds the highest glucose standard, 2) after completion of the HK reaction and dilution, an aliquot was added to a commercial ATP determination kit solution, and bioluminescence was detected with a luminescence microplate reader. A strictly linear negative correlation between glucose amount and relative luminescence units was observed in the range of 4 to 140 pmol glucose in the HK reaction mix (Fig.  2B). In a different experiment glucose standards in the range of 1 to 10 pmol were subjected to the assay and also yielded in a strong analyte:signal correlation, which was best described by a second degree polynomial regression (Fig. 2C). In both experiments deviation between technical replicates was extremely small. These data show the capability of the assay to reproducibly and sensitively detect glucose at quantities that are up to 3 orders of magnitude lower than previously described enzymatic assays.

Quantification of α-glucans in HEK293 cells cultured in 96 well format
Monitoring glycogen synthesis in human cell culture is crucial in any therapeutic study targeting central enzymes of glycogen metabolism. For instance, glycogen synthase has been identified a therapeutic target in LD (13,43,44) and adult polyglucosan body disease (41).
An assay devised to quantify α-glucans in HEK293 cells cultured in a medium throughput format (96 well) has been established and employed in an experimental setup as illustrated in Fig. 2D. Cells seeded into 96 well plates, were starved by exposure to glucose-free medium for 24 h. Subsequently the medium glucose concentration was changed back to 25 mM and 'refeeding' occurred for a variable amount of time. After that, cells were resuspended in cold water and aliquots were taken for the quantification of protein and of glucans (utilizing the glucose-dependent ATP depletion assay). The glucan content changes in correlation with the re-feeding time. After 24 h of glucose deprivation cellular glucan content is close to 0, and it increases in a largely linear fashion within 6 h of re-feeding. Glucan contents after 8 h of re-feeding are indistinguishable from that in cells that did not undergo starvation (Fig. 2E). These results are in essence corresponding to previous data shown in Rat-1 fibroblasts (19). However, in comparison to the previous study, the statistical resolution is very high. Changes of cellular glycogen content within as low as one hour are statistically significant attesting to the precision of the assay. Also, the cell number required for glucan quantification derived from a >80-fold smaller growth area. This facilitates the use of multiple well replicates, which is essential to reliably and quantitatively assess potential therapeutic effects on glycogen synthesis. By re-feeding in the presence of potential glycogen synthase inhibitors and α-glucan quantification after a defined amount of time, this assay could be readily adapted to assess the therapeutic potential of small molecule inhibitors with medium throughput efficiency.

Purification and detection of α-glucans in human CSF
Glycogen-derived glucan material could serve as a biomarker in studies of several GSDs. In this study human CSF was obtained from BioIVT, a commercial provider of biological specimen, and analyzed with respect to the presence of α-glucans.
For initial experiments CSF was pooled from several individuals. For that, the samples underwent heat denaturation to avoid degradation of potentially existing α-glucan chains by amylolytic enzymes present in CSF (45) and subsequent concentration by freeze-drying. At first free and glucan-derived (bound) glucose in the pooled CSF sample were determined using a standard enzymatic glucose assay with or without prior amylolysis. The free glucose content in the pooled CSF was around 2 mM, which is consistent with previous findings demonstrating that blood and CSF glucose levels are closely related (46). Amylolytic treatment of the pooled CSF sample did not reveal a detectable increase in glucose content (Fig. 3A), indicating that, if present at all, α-glucan-derived glucose concentrations in the CSF must be negligibly low compared to free glucose.
In an attempt to remove the vast majority of the free glucose, aliquots of the pooled CSF were subjected to repeated precipitation in 90% ethanol. While glycogen is commonly and quantitatively precipitated at 67% ethanol (47,48), a higher concentration was chosen to preserve putative smaller glycogen-derived α-glucan chains. During the repeated precipitation the recovery of glucose in each ethanolic supernatant was monitored, revealing the effective removal of usually more than 90% of glucose in each round of precipitation (Fig. 3B) and resulting in a free glucose content around 50 nM after 4 rounds. After free glucose had been reduced by a factor of at least 4 x 10 5 , amylolytic treatment revealed a noticeable increase in glucose, indicating the presence of 70 nM α-glucanderived glucosyl residues in the pooled CSF sample. When precipitation had been conducted after spiking the pooled CSF with commercial linear α-glucans (also called oligo-maltodextrins), the amylolytic treatment yielded substantially more glucose without affecting the free glucose levels. This indicates that the precipitations preserved linear α-glucans and that they can be detected if present (Fig. 3C).
It has been shown that under the precipitation conditions applied here (90% ethanol, 4 °C) linear αglucans up to a degree of polymerisation (DP, chain length) of 7 remain soluble (32). The chain length pattern of the commercial oligo-maltodextrin preparation has been analyzed by high performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD). A cumulative weight distribution, which assumes equal molar detection of all chain lengths, would explain a loss of about 20% of the glucosyl units in the maltodextrin mixture during the precipitation procedure, i.e. those chains shorter than DP 8 (Fig. 3D). Recovery of maltodextrin has been examined during the precipitations in the absence or presence of pooled CSF or authentic glucose, equivalent to amounts in CSF (Fig. 3E). Precipitation of maltodextrin alone or with authentic glucose led to a recovery of around 60% of maltodextrin glucosyl units. 50% were recovered when precipitated in the presence of pooled CSF. The discrepancy between the theoretically and experimentally determined maltodextrin losses can be explained by the fact that in HPAEC-PAD chromatograms detector response per chain increases with chain length (49,50), which means detection of long chains is more sensitive. This results in a relative underrepresentation of the shorter chains in the cumulative weight distribution and likely to an underestimation of the impact of a potential loss of glucan chains DP 1 to 7. When the cumulative weight distribution of maltodextrin was adjusted for the differential sensitivity of detection of different chain lengths (using the model of Koch et al. (50)) the DPs 1 to 7 account for around 36% of the maltodextrin weight (Fig. 3D), which is in line with the loss we observed in our maltodextrin recovery experiment. The observed loss of maltodextrins during precipitation is therefore considered within the expected range.

Quantification of α-glucans in independent human CSF samples
α-glucan content was subsequently quantified in independent human CSF samples to estimate biological variation. A procedure was established that requires 0.5-1 mL CSF and utilizes the very sensitive ATP depletion assay for glucose determination after amylolytic treatment in the presence or absence of an internal standard of commercial glycogen (Fig. 3F). In each of the individual samples even after 4 rounds of precipitation free glucose accounted for 50-80% of the total glucosyl content (Fig. 3G). However, amylolysis consistently released additional glucose in all samples indicating the presence of α-glucans in all analyzed samples. Spiking of the CSF with a defined amount of commercial glycogen resulted in a uniform increase of glucose released by amylolysis. The concentration of α-glucan in human CSF was calculated from the difference of glucose contents determined with and without amylolysis in 9 individual samples (Fig. 3H). It ranges from 50 to 300 nM, the average being 133 ± 70 nM (SD). Upon passing the Kolmogorov-Smirnov test for normality (p = 0.092) a two-tailed one sample t test was performed and attested that the average is significantly different from 0.0 (p = 0.00046).
These results demonstrate the presence of α-glucan chains in human CSF. The concentration of glucanderived glucose is more than 5 orders of magnitude below that of free CSF glucose. Investigation into the structure and the origin of the α-glucans have not been within the scope of this study. However, the methodology of their discovery defines they are α-1,4 glucan chains, likely consisting of at least 8 glucosyl units, and ultimately deriving from glycogen the only α-glucan described in the animal kingdom. As our methodology systematically excludes glucan chains shorter than DP 8 from quantification we cannot rule out that human CSF also contains those. In diseases that are characterized by an abnormal accumulation of brain glycogen, CSF α-glucan could serve as a biomarker. In particular in LD, where the accumulating insoluble glycogen is the driver of the disease, this biomarker might be a very direct readout for the efficacy of treatments that aim to reduce glycogen accumulation. However, a correlation of brain glycogen content and CSF α-glucan concentration has yet to be demonstrated.

Conclusions
Methodological advancements are described and employed, that will be relevant tools for the investigation of glycogen metabolism and related diseases. First, in mouse models of LD an inert (degradation resistant) fraction of glycogen has been demonstrated and quantified in skeletal muscle (5 months old) and hippocampus (3 months old) by the introduction of a simple pre-extraction incubation. The inert glycogen (or α-glucan) correlates with the presence of insoluble glycogen and can readily be detected when LD-related accumulation of insoluble glycogen is still very low and masked by biological variation of a comparatively high soluble glycogen amount, such as in young mice. Effective detection of molecular disease markers, such as the insoluble glycogen in LD, in young mice greatly facilitates and accelerates pre-clinical research projects. Second, via the proxy of α-glucan, glycogen has been quantified in HEK293 cells using a medium-throughput assay. This enables reliable quantitative cell-based studies of glycogen synthesis, for instance with candidate inhibitors of glycogen synthase. Third, the high sensitivity glucose assay with detection of ATP depletion described in this work enabled detection of αglucan using only a low number of cells and facilitated the discovery of minute amounts of α-glucan material in normal human cerebrospinal fluid (CSF), establishing a novel methodological platform for biomarker analyses in GSDs, such as LD or APBD.

Quantification of total and inert glycogen in mouse skeletal muscle
Aliquots of 30-50 mg of frozen ground skeletal muscle were prepared and their exact weight determined. For quantification of inert glycogen a pre-extraction incubation was conducted at 37 °C in a hybridization oven for 1.5 h or for the indicated time before samples were frozen at -80 °C until further processing. Total glycogen was quantified without any pre-extraction incubation. Glycogen was extracted and quantified largely as previously published (23). After the addition of 300 uL 30% (w/v) potassium hydroxide to the frozen tissue and vigorous mixing, the sample was incubated at 95 °C for 1 h with intermittent vortexing, prior to cooling to room temperature (RT), the addition of 50 uL 0.4 M sodium sulphate, subsequent mixing and the addition of 950 uL 100% ethanol. The sample was mixed again and left precipitating over night at -30 °C. Following centrifugation at 16,000 x g and 4 °C for 25 min and discarding of the supernatant, the glycogen underwent 3 rounds of re-dissolving (350 uL water at 95 °C for 10 min with intermittent mixing, followed by cooling to RT), precipitation (67% ethanol, 15 mM LiCl at -30 °C, overnight), centrifugation as above, and discarding of the supernatant. Subsequently the glycogen pellet was dried at 70 °C and re-dissolved in water as above. The volume of water added to the dry pellet as well as that of re-dissolved sample added to the subsequent enzymatic hydrolysis reaction (in 100 uL, 100 mM sodium acetate buffer pH 4.5, 1.63 U amyloglucosidase [Megazyme], 1h at 55 °C) were 350 and 15 uL, respectively, when total muscle glycogen was quantified. The volumes were adjusted to allow reliable detection in the following glucose assay when samples with low glycogen contents were processed.
After centrifugation (room temperature (RT), 16,000 x g, 20 min), glucose levels were determined in supernatants of the hydrolysates as well as in seven serially diluted D-glucose standards using a previously reported method (53) performed in the 96-well format exactly as recently described (12). Glycogen content is given as glucose residues per gram fresh weight.

Quantification of inert α-glucans in mouse hippocampus
Intact and deep-frozen hippocampi were transferred to weighed and liquid nitrogen-cooled BR-tubes. Tubes were incubated at RT for 30 or 60 min before being frozen at -80 °C until glycogen extraction. 20 min into the RT incubation the tissue weight was determined avoiding interference by condensation water. For glycogen extraction per tube 6 BR-beads and 440 uL of cold 50 mM sodium acetate pH 4.5 were added on ice, prior to securing the tightened cap with Parafilm and homogenization with a Bead Ruptor 24 located in a cold room (4 °C, program: 6 m/s, 2 cycles, 15 s/cycle, 5 s break). After a quick spin (15 s, 4 °C, 500 x g) and pellet resuspension by gentle vortexing, the homogenate was transferred into a fresh tube and heatinactivated (15 min, 95 °C) before cooling to RT. Subsequently 180 uL of the homogenate were used for both enzymatic α-glucan hydrolysis (adding 1.63 U amyloglucosidase (Megazyme) in 10 uL 50 mM sodium acetate pH 4.5, 1 h 55 °C) and a hydrolysis control (adding 10 uL 50 mM sodium acetate pH 4.5, 1 h on ice). Glucose is determined in 85 uL of hydrolysates and controls (each in duplicates) by a slightly modified version of a previously published method (12,53) were added and the absorbance was recorded again before calculation of glucose concentrations as previously described (12). Hippocampus glycogen was measured as α-glucan, i.e. the difference between glucose units measured in enzymatic hydrolysates (free and α-glucan-bound glucose) and hydrolysis controls (free glucose), and normalized to fresh weight.

Glucose quantification using an ATP depletion assay
In a sealed 96-well PCR plates glucose standards and samples were subjected to hexokinase (HK) reaction in 12 uL total volume containing 10 mM Tris pH 7.5, 5 mM MgCl2, 0.36 U HK, and an amount of ATP that exceeds that of the highest glucose standard by about 20%, such as 180 pmol ATP when the highest glucose standard contains 144 pmol per reaction. The reaction was allowed to run for 45 min at RT. Meanwhile the 'standard reaction solution' of a commercial ATPdetection kit (see above) was prepared as described in the manufacturer's instructions. The complete solution was protected from light and left to equilibrate to RT. The HK reaction mixtures were diluted by adding 108 uL of water with careful mixing prior to transferring 10 uL of the diluted HK reactions to a white 96 well halfarea microplate, which subsequently received 90 uL of the 'standard reaction solution' for ATP detection. The white plate was covered with aluminum foil, mixed for 2 min on a plate shaker at 600 rpm. Luminescence was detected after 15 min at 570 nm using a SpectraMax microplate luminescence reader with 750 ms signal integration time. This assay format was used for αglucan determination in cell lysates (see below). It was adjusted as follows when used for the detection of αglucans in human CSF: 1) HK reaction (14 uL total volume, 22.8 mM tricine pH 8, 1.1 mM MgCl2, 0.18 U HK, 13 pmol ATP for detecting maximal 10 pmol glucose, 1 h incubation), 2) 50 uL of water for HK reaction dilution.
Quantification of α-glucans from HEK293 cells 96-well cell culture plates were coated with poly-Dlysine (PDL) by 30 min incubation at RT with 200 µL/cm 2 of 10 µg/mL PDL in PBS, followed by PDL removal, washing with PBS and air-drying of the plates. One day after seeding the cells at a density of 2-3 x 10 4 cells/well, cells were washed with PBS, and regular medium was replaced either by starvation medium (DMEM without glucose and Na-pyruvate, 5% FBS) or by fresh regular medium (for the nonstarved group). 24 hours later, re-feeding began by changing back to regular medium with 5% FBS. After indicated times cells were harvested in a cold room (4 °C).
The medium was removed, followed by washing the cells three to four times with PBS+ (containing 1 mM CaCl2 and 1 mM MgSO4). After removing PBS+ 80 µL of water were added. The cells were detached and resuspended using a multi-channel pipette, the cell suspension subsequently being split into different 96-well PCR plates for glycogen (45 uL) and protein quantification (18 uL) and (in sealed plates) stored at -20 °C until further analysis.
For protein determination 4 µL of concentrated lysis buffer were added to 18 µL cell suspension (samples) or water (blank) reaching a final concentration of 150 mM NaCl, 20 mM Tris, 12.1 mM deoxycholate, 1% triton X-100, 0.1% SDS. Protein extraction was performed on a PCR block set to 4 °C. Samples were thawed, spun down (4 °C, 2000 x g, 3 min), and mixed by pipetting. The plate was tightly sealed, incubated (4 °C, 7.5 min), vortexed, and spun down (4 °C, 2000 x g, 1 min), followed by another round of incubation (see above) and vortexing. After centrifugation (4 °C, 2000 x g, 10 min) protein was quantified in supernatants using the Bio-Rad DC protein assay according to manufacturer's protocol with calibration against BSA standard solutions.
For glycogen content analysis samples were thawed at 4 °C on a PCR block, spun down (4 °C, 2000 x g, 3 min), and incubated at 95 °C for 80 min, with one intermittent cycle of vortexing and spinning down. Airtightness of the plate was achieved by using 8-strip caps, by applying pressure on the lid, and by cooling down first to at least 50 °C before opening the PCR machine. After heat treatment samples were collected by centrifugation (RT, 2000 x g, 3 min) and the pH adjusted to pH 4-5 by adding 5 µL of 100 mM sodium acetate buffer (pH 4.7). 3.97 uL of amyloglucosidase (Sigma) solution (2.17 mU/uL) or water (nonamylolysis controls) was added to aliquots (22.5 uL) of well-resuspended sample. Amylolysis was conducted at 55 °C overnight while controls were stored at -20 °C. Subsequently, to both a mixture of Tris buffer pH 7.5 and 0.5 M KOH was used to adjust the pH to 7.5 at 20 mM final concentration of Tris. Of the final volume (30.77 µL), 6 uL were used alongside glucose standard solutions for glucose quantification using an ATP depletion (described above). The glucan content as normalized to total protein content of the cells (proteinbased glucan content) was calculated as the difference of glucose amounts per well determined with and without amylolysis and division by protein content per well.

Determination of α-glucans in human cerebrospinal fluid
Frozen commercial normal human CSF was directly subjected to heat denaturing for 15 min at 95 °C in the original sample tube. Subsequently a defined volume (around 1 mL) of CSF was transferred to fresh tubes and freeze-dried overnight. After that several samples were dissolved in water and pooled for initial experiments as described, later CSF samples were treated separately. Aliquots of pooled and ≈17fold concentrated CSF were subjected to glucose assay (as previously described (12)) with and without prior amylolysis (in 100 uL 100 mM sodium acetate pH 4.5, with or without 1.63 U amyloglucosidase [Megazyme]), each before (using 4 uL concentrated and pooled CSF) and after (using 117 uL concentrated and pooled CSF) several rounds of ethanol precipitation (1 h rotating, 4 °C, 90% ethanol, 20 mM NaCl) either in the presence or the absence of an oligomaltodextrin (MD) standard (42.75 nmol glc residues). The same amount of MD was subjected to identical precipitation conditions in the absence of CSF, as well as in the presence of 11.2 uM authentic glucose. After each precipitation samples were centrifuged (20 min, 16,000 x g, 4 °C), the supernatant collected in fresh tubes, and the pellet dissolved in water. The supernatant was dried (vacuum centrifuge), redissolved in water and aliquots were subjected to glucose quantification with and without prior amylolysis (as above).
Freeze-dried individual CSF samples were resuspended in 50 uL 200 mM NaCl and heated for 10 min at 95 °C with intermittent vortexing before being cooled to RT. After a short spin 500 uL of 100% ethanol were added followed by vigorous mixing and leaving the samples to precipitate for >1 h at 4 °C on a rotator and subsequent centrifugation (4 °C, 20 min, 16,000 x g) with discarding of the supernatant. Subsequently three rounds of the following steps were conducted: 1) resuspension of the pellet (50 uL water, 5 min 95 °C), 2) precipitation by adding 450 uL 100% ethanol (for 1h, 4 °C), 3) centrifugation (as described above) removal of the supernatant. The pellet was then dried in a vacuum centrifuge and resuspended in 65 uL 20 mM NaOAc pH 4 by vortexing thoroughly. If necessary, the pH was adjusted to 4-5 by the addition of defined volumes of dilute HCl. 5 uL of the sample were used for amylolysis (overnight at 55 °C in 30 uL total volume: 2 mU amyloglucosidase [Sigma], 20 mM NaOAc pH 4.5, with or without spiking by commercial rabbit liver glycogen as internal digestion control) and amylolysis control (as amylolysis except overnight at -20 °C without amyloglucosidase), both being treated alongside CSF-free controls. 6 uL of hydrolysates and controls were used for glucose quantification using the ATP depletion assay described above. CSF α-glucan content was calculated from the difference between glucose contents with and without amylolysis after correcting all glucose contents for the amount of glucose found in the 'no CSF' controls.

Chain length distribution
20 ug of commercial oligo-maltodextrin was applied to high performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD, ThermoFisher, ICS5000). Oligoglucan chains were separated as described previously (23) and chromatograms analyzed with Dionex Chromeleon 7.2. The relative peak areas for each chain length (degree of polymerization, DP) were determined. A cumulative relative weight distribution, which assumes equal detection of all chain lengths, was calculated by multiplying the relative peak area (RA) of each DP with the DP, followed by normalization to 100% for ∑RAn*DPn, and addition of these normalized weighted relative areas for DP1 to DPn. An adjusted cumulative weight distribution was calculated after dividing RAn by ((85.26/DPn) + 16.15) according to the model of Koch et al. (50), which takes into account different detector responses at different chain lengths. Adjusted RAn were normalized to 100% for ∑RAn and added for DP1 to DPn At least 5 technical replicates were analyzed, a representative shown.

Statistical Analyses
Biological replicates were as indicated for analyses a) in mouse skeletal muscle n = 6, b) in mouse hippocampus n = 2, c) in cell-based experiments n = 7-8, d) in human CSF n = 9. If averages were calculated, standard deviation (SD) was used. When averages of two technical replicates were calculated the mean absolute deviation (MAD; |x1-x2|/2) is displayed. Upon calculation of differences between two independently determined quantities MAD was adjusted by calculating the square root of the sum of the square of both independent MADs. Statistical comparison of two groups has been performed by Welch's unpaired twotailed t test. Multiple groups have been compared by one-way ANOVA and Tukey post hoc tests. To determine whether α-glucan concentrations in human CSF are greater than 0, a Kolmogorov-Smirnov test for normality was performed which was followed by a two-tailed one sample t test if the former test was passed. Statistical testing was performed using GraphPad Prism 8.

Data availability
All data referred to is included in this paper.

Figure 1: Determination of inert α-glucans in mouse muscle and hippocampus
A) Schematic, illustrating the procedure of glucan quantification in biological material. Depending on the type of extraction and tissue lysis, ethanol precipitations or additional controls for amylolytic degradation become necessary. The red line indicates at which step during the procedure an additional incubation step was performed in experiments referred to in B to F. B) Glycogen content in aliquots of pooled skeletal muscle tissue from 5 month-old WT and Epm2b -/mice. Aliquots were subjected to pre-extraction incubation at 37 °C for the indicated amount of time and glycogen determination proceeded as indicated in A for skeletal muscle. Error bars represent the mean absolute deviation between two identically treated tissue aliquots. C) Total and inert glycogen in muscle of 5 month-old WT and Epm2b -/mice. Inert glycogen was determined as the residual glycogen present after 1.5 h pre-extraction incubation at 37 °C. Average of 6 individual mice per group, with error bar representing SD and p value derived from Welch's two-tailed T test. D) Schematic, illustrating the use of the hippocampi from both mouse brain hemispheres for the determination of inert α-glucan material in 3 month-old WT and Epm2a -/mice as referred to in E and F. E and F) Determination of inert glucan content in the hippocampi of two WT (A and B) and two Epm2a -/mice (C and D). Hippocampi were subjected to pre-extraction incubation for either 30 min (left hippocampi) or 60 min (right hippocampi), and glucan determination proceeded as indicated in A for hippocampus. Depending on amylolytic treatment, free glucose or both free glucose and glucan-bound glucose were quantified, respectively (E). Glucan-bound glucose was calculated as the difference of values determined with and without amylolytic treatment (F). Averages of independent glucose determinations are depicted with error bars representing mean absolute deviation of two technical replicates (E). Upon calculation of differences, mean absolute deviations were adjusted to account for variances of the two independently determined glucose quantities used for the calculation (F, see Statistical Analysis).

Figure 2: Sensitive detection of α-glucans using an ATP depletion assay
A) Simplified numeric illustration of a glucose assay with detection of ATP depletion. Glucose (Glc) is converted to glucose 6-phoshate (G6P) in the presence of hexokinase and ATP. In the subsequent luciferase reaction remaining ATP correlates positively with relative luminescence (Rel. Lumin.), which itself negatively correlates with the glucose concentration. Note, that the highest detectable glucose concentration is lower than the ATP start concentration. The high sensitivity of the ATP-dependent luminescence detection can require dilution after the hexokinase reaction. B and C) Authentic glucose standards with ATP-depletion detection in two sensitivity ranges. The top axis denotes the amount of glucose added to the hexokinase (HK) reaction (with total reaction volume, ATP start concentration, and HK activity given). The bottom axis denotes amounts of glucose equivalents (i.e. ATP-depletion units) that are subjected to the luciferase reaction after dilution of the HK reaction. Values are averages of relative luminescence in two technical replicates with mean absolute deviation represented by error bars (note, these are too small to be graphically discernible). Linear and/or polynomial regression curves show the correlation of glucose amount subjected to the assays and detected luminescence. D) Schematic of the experimental setup to measure α-glucan material in HEK293 cells in 96-well format during different stages of re-feeding after glucose deprivation using the ATP depletion assay for glucose detection. Glucan material was quantified as defined by 'bound glucose' in Figure 1A, i.e. the portion of total glucosyl content that is amylolytically sensitive. Glucan amounts were normalized to the protein content of the same cell sample. E) HEK293 α-glucans were depleted by glucose deprivation for 24 h and resynthesized within 8 h of glucose re-feeding. Glucan content as normalized to total protein (protein-based glucan content) was determined as described in D. Values represent averages of cells from typically 7 or 8 individual but identically treated wells. Error bars represent SD. Numeric values are select p values of one-way ANOVA with Tukey post hoc tests.

Figure 3: Detection of a-glucans in human cerebrospinal fluid (CSF)
A) Glucose concentrations determined with and without amylolytic treatment of a heat-denatured and concentrated sample of pooled human CSF. Averages and mean absolute deviations (error bars) of two technical replicates are shown. B) Recovery of total glucosyl units in the supernatants of four consecutive precipitations of a heatdenatured and concentrated sample of pooled human CSF in 90% ethanol. Values represent percentages as compared to total glucosyl content in the unprecipitated sample shown in A, e.g. during the first precipitation 89% of total glucosyl units in the sample were recovered in the ethanolic supernatant. Averages and mean absolute deviation (error bars) of two technical replicates are shown. C) Glucose concentrations determined with and without amylolytic treatment of a heat-denatured and concentrated sample of pooled human CSF that underwent four rounds of precipitation in 90% ethanol, either in the absence or in the presence of an authentic maltodextrin standard. Averages and mean absolute deviations (error bars) of two technical replicates are shown. D) Representative distributions of chain length (degree of polymerization vs. relative peak area, left axis) and of cumulative weight (right axis) of the maltodextrin standard used in C. The elimination of chains with seven or less glucosyl units during repeated precipitation is expected to reduce maltodextrin recovery by at least 20% if the cumulative weight distribution is calculated based on the assumption that each chain length is detected with similar molar sensitivity. However, shorter chains are detected less sensitively and usually underrepresented. An adjusted cumulative weight distribution according to the model of Koch et al. (50) predicts a loss of around 36%. E) Recovery of authentic maltodextrin (MD) after four consecutive 90% ethanol precipitations in the absence or presence of an excess of authentic glucose or of pooled human CSF. Before the precipitations authentic glucose and glucose in human CSF was in 260-fold and 110-fold excess of maltodextrinderived glucose, respectively. Averages and mean absolute deviations (error bars) of two technical replicates are shown. F) Schematic illustrating the procedure for the determination of α-glucans in CSF using the sensitive ATP depletion assay for glucose quantification. Results displayed in G and H were obtained using this procedure.
G) Amylolytical treatment releases glucose from human CSF. Heat-inactivated, concentrated, and precipitated human CSF (see F) was subjected to glucose quantification using an ATP depletion assay with or without prior amylolytic treatment in the presence or absence of an internal standard of commercial glycogen. Averages and mean absolute deviation (error bars) of two technical replicates for four individual CSF samples and 'no CSF' controls are displayed. Detected glucose content is expressed as glucose amount in the hexokinase (HK) reaction of the glucose assay with ATP-depletion detection (see Figure 2A). H) Average glucan content and standard deviation over nine individual human CSF samples, each determined as the difference of glucose detected with and without prior amylolytic treatment (shown in G for four samples).