Enhanced lysosomal glycogen breakdown is associated with liver tumorigenesis in glycogen storage disease type III

Abstract

Background & Aims

Glycogen storage disease type III (GSDIII) is a rare metabolic disorder caused by mutations in the glycogen debranching enzyme (AGL), leading to hepatic glycogen accumulation, fibrosis and increased hepatocellular carcinoma (HCC) risk. This study investigates the metabolic mechanisms driving liver tumorigenesis in an Agl^-/- model of GSDIII.

Methods

Liver and tumor samples from 14-month-old Agl^-/- and Agl^+/+ mice, and liver biopsies from patients with GSDIII (n = 4), were analyzed using histological, biochemical and molecular approaches.

Results

Agl^-/- mice recapitulated key features of GSDIII, including a 3.5-fold hepatic glycogen overload (p <0.001), and chronic liver disease. More than 30% of the animals developed liver tumors, associated with a 2.5-fold increase in alpha-fetoprotein levels (p <0.005). Despite marked reductions in glucose (7.5-fold, p <0.0001), glucose-6 phosphate (266-fold, p <0.0001), lactate (8-fold, p <0.005), cholesterol (1.9-fold, p <0.001) and triglyceride levels (6.2-fold, p <0.001) in the liver, glycaemia was maintained at around 87.0 ± 9.6 mg/dl after 6 h of fasting, through activated extrahepatic, but not hepatic, gluconeogenesis. Intriguingly, most tumors exhibited lower glycogen content than surrounding tissue (3.3-fold decrease, p <0.0001), which was associated with increased lysosomal alpha-acid glucosidase activity (19.5 ± 5.5 in tumor vs. 9.9 ± 2.0 mmol/h/mg in Agl^-/- liver; p <0.0005) and the presence of glycophagosomes. PAS-negative staining in HCCs from patients with GSDIII supported these observations. Although YAP nuclear staining varied among tumors, the overall increase in YAP nuclear localization and CTGF expression suggests that inhibition of the Hippo/YAP pathway may contribute to tumorigenesis in GSDIII hepatocytes.

Conclusions

In GSDIII, liver metabolism is characterized by the accumulation of structurally abnormal glycogen and a significant reduction of key energy substrates. In this metabolic context, enhanced lysosomal glycogen degradation may support tumor growth, highlighting a mechanistic link between glycogen metabolism and the development of liver cancer.

Impact and implications

This study provides novel insights into the metabolic dysregulations driving liver tumorigenesis in glycogen storage disease type III (GSDIII). Our findings reveal a potential link between abnormal glycogen accumulation and liver cancer, highlighting the pivotal role of lysosomal glycogen degradation in supporting tumor growth. These results are particularly important for researchers and clinicians working on metabolic liver diseases, as they suggest potential glycogen-targeting therapeutic strategies for GSDIII and other related liver disorders. Practically, they could guide future interventions aimed at modulating glycogen metabolism, offering new treatment avenues for patients with GSDIII at risk of hepatocellular carcinoma, while contributing to the broader understanding of metabolic dysregulation in cancer biology.

Graphical abstract

Highlights

• •Lysosomal degradation of glycogen was activated in liver tumors of GSDIII mice.
• •Glycogen may serve as fuel to promote tumor growth in GSDIII mice and patients.
• •Loss of glycogen debranching enzyme causes a severe energy deficit in the liver.
• •Extrahepatic gluconeogenesis sustains blood glucose levels in GSDIII mice during fasting.

Introduction

As the central regulator of metabolism, the liver plays a crucial role in maintaining whole-body metabolism. One of its main functions is to maintain blood glucose levels between 0.8-1.1 g/L. This allows humans and animals to endure extended periods without food.^1,2 Glucose is indeed the main energy source for all living cells, supplying carbon skeletons for nucleotide, protein, and lipid synthesis required for cell growth and proliferation.

After a meal, the excess of glucose is stored as glycogen, mainly in the liver and skeletal muscles, through a process called glycogenesis. Glycogen, a highly branched polymer of glucose, constitutes a major form of energy storage. During the post-absorptive state, the breakdown of hepatic glycogen (i.e. glycogenolysis) represents the fastest process to restore blood glucose levels. Glycogen can be degraded by two different pathways. The first occurs in the cytoplasm and involves two reactions catalyzed by the glycogen phosphorylase (PYG) and the glycogen debranching enzyme (GDEamylo-alpha-1,6-glucosidase enzyme, encoded by AGL). The second pathway recycles ∼10% of glycogen and is dependent on autophagy. It occurs in lysosomes and involves acid alpha-glucosidase (encoded by GAA).^3,4 In hepatocytes, glucose-6-phosphate (G6P), produced from cytosolic glycogen degradation, is hydrolyzed by glucose-6-phosphatase (G6Pase) to release glucose into the bloodstream. When glycogen stores become depleted, glucose is synthetized from non-carbohydrate substrates via gluconeogenesis in the liver, kidney and intestine.¹

The loss of function of enzymes involved in glycogen metabolism leads to abnormal glycogen accumulation observed in inherited disorders known as glycogen storage diseases (GSDs). Recent studies have highlighted the importance of glycogen metabolism in supplying energy and in maintaining cellular homeostasis.⁵ Large amounts of glycogen have been found in various cancer cell lines, particularly those undergoing neoplastic transformation.⁶ Glycogen supports cell survival under hypoxic conditions[7], [8], [9] and has recently been identified as an oncogenic metabolite that can drive cell transformation and tumor initiation.^10,11 However, the specific role of glycogen in driving tumor progression and growth is still unreported.

Glycogen storage disease type III (GSDIII, OMIM 232400) is an autosomal recessive disorder caused by mutations in the AGL gene.¹² Impaired glycogen breakdown leads to the accumulation of abnormally structured, soluble glycogen, called limit dextrin, within cells. In cardiac and skeletal muscles, glycogen accumulation results in cardiomyopathy and progressive muscle weakness with age. In the liver, this leads to hepatomegaly and hypoglycemia starting in early childhood, alongside elevated plasma transaminase levels reflecting early liver damage. Limit dextrin accumulation is associated with the development of hepatic fibrosis in childhood, which may progress to severe liver complications in adulthood, including cirrhosis, hepatocellular adenoma (HCA) and hepatocellular carcinoma (HCC). In some cases, a liver transplant may be necessary.[13], [14], [15], [16], [17] Due to the typical glycogen accumulation in GSDIII, hepatocytes from affected individuals represent a relevant model to investigate glycogen’s contribution to tumorigenesis. In this study, we aimed to investigate the long-term progression of liver disease in GDSIII, focusing specifically on the relationship between hepatic metabolism and tumor development. Acknowledging the high energy demands of tumor growth and proliferation, we specifically examined the metabolic substrates, particularly glycogen, that could be utilized by GSDIII hepatocytes to sustain their proliferation and drive tumorigenesis. To this end, we used a GSDIII mouse model (Agl^-/- mice) that recapitulates the key pathological features observed in patients, i.e. glycogen accumulation, hypoglycemia, hepatomegaly, tumor development, and muscle weakness.¹⁸ Selected findings were further validated using liver biopsies from patients with GSDIII who developed HCC.

Materials and methods

In vivo studies

Animal model

Agl knockout mice (Agl^-/-) were generated by replacing exons 6–10 of the Agl gene with a neomycin cassette, as described by Vidal et al.,¹⁸ within the IMPC framework. Mice were bred on a mixed BALB/c-C57Bl/6J background and housed in groups of 4–6 males or females at 21-23 °C, with a 12/12 h light/dark cycle, and free access to chow diet (A04-Safe) and water, at CERFE (Généthon, Evry) or ALECS (University of Lyon 1). Blood was collected in 3.8% citrate or EDTA.

Glutamine and pyruvate tolerance tests were performed after 6 h (post-absorptive) or 12 h of fasting (glycogen-depleted), respectively. Mice received intraperitoneal injections of L-glutamine (1.5 g/kg) or pyruvate (2 g/kg), and blood glucose was monitored for 2 h.

At 14 months, mice were euthanized (fed or 6 h-fasted) by cervical dislocation. Livers were freeze-clamped in liquid nitrogen and stored at -80 °C or fixed for histology.

Study approval

All procedures complied with the European directive 2010/63/EU and were approved by local ethics committees and the French Ministry of Research (Apafis: #23334, #37405 and #44716).

Patients

Four French patients with GSDIII were retrospectively reviewed after liver transplantation or partial hepatectomy. Histological analyses were performed on tumors and peritumor livers. Informed consent was obtained in accordance with French legislation.

Histological analyses

Staining

Fresh liver was fixed in 10% formalin for 48 h, paraffin-embedded and sectioned (4 μm). Sections were stained with hematoxylin–phloxine–saffron (HPS), periodic acid-Schiff (PAS), Sirius red, or Oil red O. Immunohistochemical analyses were performed, after heat-mediated antigen retrieval (TintoRetriever Pressure Cooker, Diagomics), peroxidase inhibition (3% H₂O₂, 5 min), blocking, and overnight antibody incubation at 4 °C (CTAT table). Detection was performed using HRP-conjugated anti-rabbit IgG and DAB chromogen, followed by hematoxylin counterstaining. Quantification methods are described in the supplementary information.

Transmission electron microscopy

Liver samples (5 mice/group) were fixed in 2.5% glutaraldehyde, postfixed in 2% osmium tetroxide, dehydrated, and embedded in an epoxy resin (EMBed-812, Electron Microscopy Sciences). Ultrathin sections were stained with uranyl acetate and lead citrate, observed at 80 kV (CM120, Philips Electronics) and photographed (Morada, Olympus). Ten fields per sample were analyzed and the number of autophagosomes was averaged per group.

Biochemical parameters

Blood parameters

Plasma alpha-fetoprotein (AFP) levels were determined by ELISA (R&D Systems). Aspartate aminotransferase/alanine aminotransferase activities were determined by colorimetric assays (Abcam). Blood glucose was measured with an Accu-Check Go glucometer (Roche Diagnostic).

Liver metabolites

Glucose, G6P and glycogen were measured as previously described from a fresh liver piece homogenized in 6% perchloric acid (FastPrep, MP Biomedicals).^18,19 Lactate levels were measured in liver homogenate using a colorimetric kit (Abcam). Hepatic lipids were extracted (Folch method) from 50-100 mg of liver in chloroform/methanol (2:1). Triglycerides and cholesterol were quantified using colorimetric kits (DiaSys Diagnostic Systems GmbH). NADPH levels were measured from liver homogenate after heating at 60 °C for 30 min, using a colorimetric kit (Abcam).

Glucose-6 phosphatase and acid α-glucosidase activities

G6Pase activity (U/g tissue) was assayed at maximal velocity (with 20 mmol/L G6P) at 30 °C, as previously described.²⁰ GAA activity (nmol/h/mg protein) was measured as previously described,²¹ and normalized to protein concentration, quantified by BCA method (Thermo Fisher Scientific).

Gene expression analyses

Total RNA was extracted using Trizol (Thermo Fisher Scientific). DNA contaminants were removed (Free DNA kit, Thermo Fisher Scientific), and total RNA was reverse transcribed (RevertAid, Thermo Fisher Scientific). Expression of mRNA was assessed by reverse-transcription quantitative PCR using a 7900 Fast Real Time thermocycler and Fast SYBR Green assays (Applied Biosystems). Primer sequences are listed in the CTAT table. Data are expressed as -ΔΔCt (the delta between the gene of interest Ct value and the Ct value of the housekeeping gene, normalized to the Agl^+/+ group).

Protein expression analysis

Whole-cell lysates were obtained by liver homogenization (Fast Prep system, MP Biomedicals). Immunoblots were performed with specific antibodies (CTAT table). Detection of eIF2a, caspase 3/7, acetyl-CoA carboxylase (ACC), carnitine palmitoyltransferase 1α (CPT1 α), starch-binding domain-containing protein 1 (STBD1), vinculin was based on immunofluorescence, visualized with Odyssey (LI-COR Biosciences), and densitometric quantification was performed with Image Studio Lite 5.2 (LI-COR Biosciences). The detection of cytosolic phosphoenolpyruvate carboxykinase (PEPCK-c) and G6PC1 proteins was performed by chemiluminescence (Chemidoc™ imager, Bio-Rad) and densitometric quantification was performed using Image Lab™ software (Bio-Rad). Blots were normalized to a housekeeping protein, to the total form for phosphorylated/cleaved proteins, or to total protein for stain-free gels (Bio-Rad)²² (see figure legend).

Statistical analysis

Data are presented as mean ± SD. GraphPad Prism 7.0 was used for statistical analyses. p value <0.05 was considered significant. The statistical tests performed are indicated in Table S1.

The incidence of nodules between males and females was calculated using a two-sided Fisher’s exact test (RStudio 2025.05.0+496).

Results

Liver analysis revealed chronic liver damage and liver fibrosis in 14-month-old Agl^-/- mice

As previously reported,¹⁸ Agl^-/- mice developed hepatomegaly due to glycogen accumulation (Fig. 1A,B) and showed decreased blood glucose levels compared to Agl^+/+ mice, even when fed (Fig. 1C). However, only two 14-month-old Agl^-/- mice exhibited glycemia below 70 mg/dl. HPS staining revealed enlarged, clear ballooning hepatocytes homogeneously distributed throughout all hepatic lobules, and PAS staining confirmed glycogen accumulation in Agl^-/- hepatocytes (Fig. 1D). In addition, Sirius red staining revealed enhanced centrilobular perisinusoidal fibrosis (METAVIR scale F2/F3; Fig. 1E,F). Accordingly, mRNA levels of activated hepatic stellate cell markers, such as actin alpha 2 (Acta2) and type 1 collagen (Col1a1), were increased, while plasminogen activator inhibitor 1 (Pai1) expression remained unchanged (Fig. 1G). In accordance with the presence of fibrosis and hepatocyte damage, plasma aspartate aminotransferase/alanine aminotransferase levels were elevated in 14-month-old Agl^-/- mice compared to Agl^+/+ mice (Fig. 1H).

Fig. 1.

Liver pathology shows inflammation and fibrosis in Agl^-/- mice.

(A) Liver weight normalized by body weight, (B) hepatic glycogen content, and (C) blood glucose in 14-month-old fed male and female Agl^+/+ (black) and Agl^-/- (red) mice. (D-E) Representative images of HPS, PAS and Sirius red staining of liver sections. Scale bars = 200 μm (main), or 100 μm (magnification). (F) Fibrosis quantification (% Sirius red-positive surface/total surface). (G) Hepatic mRNA expression of fibrotic genes. (H) Plasma ALT and AST activities. (I) Hepatic Ccl2 mRNA expression. (J) Representative images and quantification of hepatic immunohistochemistry for MPO, F4/80, and CD3. Scale bar = 100 μm. Statistical analyses are detailed in Table S1. ∗p <0.05, ∗∗p <0.01 and ∗∗∗p <0.001. ALT, alanine aminotransferase; AST, aspartate aminotransferase; HPS, hematoxylin–phloxine–saffron; PAS, periodic acid–Schiff.

Since hepatic inflammation drives liver disease progression and sustains fibrosis,²³ we explored the hepatic inflammatory response in Agl^-/- mice. Monocyte chemoattractant protein-1 (Ccl2) mRNA expression was higher in Agl^-/- compared to Agl^+/+ livers (Fig. 1I). Immunohistochemical analyses confirmed that the number of F4/80-positive cells (Kupffer cells and macrophages) was greatly increased in Agl^-/- livers, compared with controls (Fig. 1J), supporting macrophage recruitment. CD3⁺ and MPO⁺ cells tended to be slightly more abundant in Agl^-/- mice, suggesting increased hepatic T lymphocyte and neutrophil populations (Fig. 1J).

In conclusion, loss of GDE results in extensive glycogen accumulation in the liver, which is associated with chronic liver damage, fibrosis and inflammation.

Hepatic tumorigenesis in Agl^-/- mice

Liver fibrosis can progress to cirrhosis in approximately 40-45% of patients with GSDIII at a median age of 40 years, and development of liver neoplasms (HCA/HCC) can occur in approximately 15% of patients with GSDIII at a median age of 65 years.¹⁷ In Agl^-/- mice, about one third of 14-month-old mice developed at least one macroscopic tumor (>1 mm), ranging from 1 mm to 10 mm (Fig. 2A). Surprisingly, females showed markedly higher tumor incidence than males (55% vs. 8%) (Fig. 2A), a difference statistically confirmed by a two-sided Fisher’s exact test (p <0.0001, odds ratio = 13.57). Importantly, elevated levels of plasma AFP, a biomarker of HCC,²⁴ were observed in Agl^-/- mice with tumors (Fig. 2B), but not significantly in age-matched Agl^-/- mice without macroscopic nodules. A complete histological characterization of several tumors isolated from Agl^-/- mice revealed non-encapsulated tumors of two types (Fig. 2C). The first type was composed of large cells, with atypical nuclear-cytoplasmic ratio, clonal appearance, eosinophilic infiltration, and no vascularization (nodule 1, Fig. 2C). The second type was composed of small and clarified cells, with the aspect of progenitor cells (nodule 2, Fig. 2C).

Fig. 2.

Hepatic tumorigenesis in Agl^-/- mice.

(A) Representative images of liver and nodules and table showing the number of macroscopic nodules (>1 mm) in 14-month-old Agl^+/+ and Agl^-/- mice at necropsy. Scale bar = 1 cm. (B) Plasma AFP in Agl^+/+ mice (black bars) and Agl^-/- mice without (NT, red bars) and with tumors (T, blue bars). (C) Representative images of HPS and Sirius red staining of two hepatic nodules from Agl^-/- mice (1 and 2). Scale bars = 500 μm (main), and 50 μm (magnification). (D) Representative images and quantification of Ki67 immunohistochemistry in livers. Tumor is outlined with hatched line. NT = peritumor tissue. Scale bar = 200 μm. (E) TEM images from non-tumor livers of Agl^+/+ and Agl^-/- mice. Arrows: dilated ER; circles: glycogen. (F) Quantitative analysis of phosphorylated and total eIF2-alpha protein expression in liver. (G) Hepatic Chop mRNA expression. (H) Quantitative analysis of cleaved and total Caspase 7 and 3 protein expression in livers. (I) Representative images of YAP IHC in tumor (T) and peritumor (NT) liver tissue. Scale bars = 100 μm (main), and 50 μm (magnification). (J) Hepatic Ccn2 mRNA expression. Statistical analyses are detailed in Table S1. ∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001. ER, endoplasmic reticulum; HPS, hematoxylin–phloxine–saffron; IHC, immunohistochemistry; PAS, periodic acid–Schiff; TEM, transmission electron microscopy.

Consistent with tumor development, high cell proliferation was detected by immunohistochemistry of Ki67 marker in the tumors, whereas proliferation was only moderated in non-tumoral regions of Agl^-/- liver (Fig. 2D). Specifically, non-tumor Agl^-/- livers exhibited only a modest increase in mRNA expression of the proliferative markers cyclin E1 (Ccne1), Ki67 and DNA topoisomerase 2 α (Topo2α) compared to Agl^+/+ liver (Fig. S1). On the contrary, the expression of cyclin B1 (Ccnb1) was slightly reduced in the livers of Agl^-/- mice, while the expression of cyclin D1 (Ccnd1) remained unchanged (Fig. S1).

Transmission electron microscopy (TEM) images revealed dilated endoplasmic reticulum (ER) in Agl^-/- non-tumor livers, a sign of ER stress²⁵ (Fig. 2E). ER stress activates the unfolded protein response to restore cell homeostasis; however, sustained activation of ER stress drives cell death. Furthermore, ER stress and the unfolded protein response have been clearly implicated in the pathogenesis of several liver diseases.²⁵ In Agl^-/- mice, the phosphorylation at serine 51 of p-eIF2α, a known marker of ER stress induction,²⁵ was increased compared to control mice, confirming the activation of the stress response in Agl^-/- non-tumor livers (Fig. 2F). In addition, we observed increased mRNA expression of C/EBP homologous protein (Chop), a proapoptotic transcription factor triggered by ER stress²⁵ (Fig. 2G). The increase in the apoptotic signaling was confirmed by the significant increase in the downstream effectors cleaved-caspase 7 in the non-tumor Agl^-/- liver, as well as by the trend towards increased levels of cleaved-caspase 3 (Fig. 2H).

To further investigate the mechanisms underlying tumorigenesis in Agl^-/- livers, we analyzed the Hippo/YAP signaling pathway. Recent studies have demonstrated that excessive glycogen accumulation can inhibit this pathway and activate the oncogenic transcriptional co-activator YAP, promoting tumor initiation.¹¹ Consistently, YAP immunostaining revealed positive nuclear staining in some tumor and adjacent non-tumor tissues (Fig. 2I). Moreover, the YAP downstream target CTGF (Ccn2) was upregulated in both tumor and non-tumor livers (Fig. 2J).

In conclusion, chronic ER stress in Agl^-/- livers is associated with increased hepatocyte apoptosis and compensatory proliferation, mechanisms known to support tumorigenesis. This ER stress is likely driven by accumulation of abnormally structured glycogen, given the well-established link between glycogen and ER.²⁶ Moreover, Hippo/YAP pathway inhibition may participate in the early events driving tumorigenesis in GSDIII livers.

Energy deficiency in Agl^-/- non-tumor livers and induction of extrahepatic gluconeogenesis to maintain glycemia in Agl^-/- mice

We next investigated the metabolic dysregulations associated with the loss of GDE activity in Agl^-/- hepatocytes. As expected, intracellular G6P levels, but also glucose levels, were reduced in 6 h-fasted non-tumor Agl^-/- livers (Fig. 3A). This resulted in lower blood glucose levels in Agl^-/- mice compared to Agl^+/+ mice, but without the development of symptomatic hypoglycemic events. After 12 h of fasting, Agl^-/- mice even showed similar blood glucose levels to Agl^+/+ mice, reaching 100 mg/dl (Fig. 3B). This normalization is likely due to the progressive induction of gluconeogenesis by prolonged fasting. However, there was no evidence of gluconeogenesis activation in the livers of 6 h-fasted Agl^-/- mice. Indeed, G6Pase activity decreased and the expression of key gluconeogenic enzymes (G6pc1 mRNA, PEPCK-c protein) remained unchanged (Fig. 3C).

Fig. 3.

Energy deficiency in GSDIII livers and induction of extrahepatic gluconeogenesis to maintain glycemia in GSDIII mice.

(A) G6P and glucose levels in 14-month-old Agl^+/+ and Agl^-/- mice livers. (B) Fasting blood glucose level over time. (C) Quantification of hepatic G6Pase activity, G6pc1 mRNA expression level, and PEPCK protein expression (normalization to total protein shown in Fig. S3A). (D) Pyruvate (12 h fast, left panel) and glutamine (6 h fast, right panel) tolerance tests. (E) Quantitative analysis of G6PC1 and PEPCK-c protein expression (normalization to total protein or tubulin shown in Fig. S3B and C), and G6pc1 and Pck1 mRNA expression in the kidney (left panel) and intestine (right panel). (F-G) Hepatic lactate (fasted) (F) and cholesterol and triglyceride (G) contents. (H) Representative images of ORO staining in liver. Scale bar = 200 μm. (I) Quantitative analysis of G6PD protein expression in liver (normalization to total protein levels shown in Fig. S3D). (J) Hepatic NADPH content. All the analyzed mice were female. Statistical analyses are detailed in Table S1. ∗p <0.05, ∗∗p <0.01 or ∗∗∗p <0.001. G6P, glucose-6-phosphate; G6Pase, glucose-6-phosphatase; G6pc1, glucose-6-phosphatase catalytic subunit 1; G6PD, glucose-6-phosphate dehydrogenase; GSDIII, glycogen storage disease type III; ORO, Oil Red O; Pck1, phosphoenolpyruvate carboxykinase 1; PEPCK, phosphoenolpyruvate carboxykinase; PEPCK-c, cytosolic phosphoenolpyruvate carboxykinase.

To further assess hepatic gluconeogenesis in GSDIII, we performed a pyruvate tolerance test after a 12-hour fast, which is sufficient to deplete glycogen stores in mice.²⁷ While pyruvate is a key substrate for hepatic gluconeogenesis,¹ but not entirely liver-specific, this test remains widely used to estimate hepatic gluconeogenic capacity in mice. In 12-hour fasted Agl^+/+ mice, pyruvate injection markedly increased blood glucose, confirming the induction of hepatic gluconeogenesis, whereas Agl^-/- mice showed only a marginal rise, suggesting reduced hepatic gluconeogenic flux (Fig. 3D). Conversely, following a 6-hour fast, Agl^-/- mice showed increased blood glucose 60 min after injection of glutamine, the main substrate of renal and intestinal gluconeogenesis¹ (Fig. 3D). In Agl^+/+ mice, blood glucose levels remained unaffected following glutamine injection, as they were able to maintain glycemia by utilizing glycogen stores. These data support that extrahepatic gluconeogenesis likely compensates for impaired hepatic glycogenolysis and gluconeogenesis in Agl^-/- mice. Supporting this, the expression of gluconeogenic enzymes (PEPCK-c and G6PC1) was increased in the kidney of 6 h-fasted Agl^-/- mice, compared to Agl^+/+ mice (Fig. 3E). The expression of PEPCK-c was also increased in the intestine of Agl^-/- mice, whereas G6PC1 expression remained unchanged (Fig. 3E).

Given the limited availability of glucose in Agl^-/- hepatocytes, glycolysis flux was also reduced, as evidenced by lower lactate in non-tumor livers (Fig. 3F). Regarding lipid metabolism, triglyceride and cholesterol levels were significantly reduced in Agl^-/- livers compared to controls (Fig. 3G). There was no change in the phosphorylation at Ser79 of ACC, which controls lipid synthesis, between fasted Agl^-/- and control mice, nor in the protein levels of CPT1α, a key protein involved in mitochondrial fatty acid β-oxidation (Fig. S2A). Similarly, mRNA expression of key lipid regulating transcription factors (Pparα, Chrebp, Srebp1c) was unchanged (Fig. S2B) and no steatosis was observed in Agl^-/- livers (Fig. 3H). Plasma triglyceride and cholesterol levels were also comparable to controls (Fig. S2C).

Finally, in line with the low levels of intracellular G6P in Agl^-/- livers, we also observed reduced expression of G6P dehydrogenase (G6PD), the limiting enzyme of the pentose phosphate pathway (Fig. 3I). However, the decrease in this pathway did not induce a change in NADPH levels in the liver of Agl^-/- mice compared to Agl^+/+ mice, suggesting that NADPH is produced by a different biochemical pathway in the Agl^-/- liver (Fig. 3J).

In conclusion, our data indicate that impaired glycogenolysis in the absence of GDE leads to metabolic rewiring, resulting in greatly reduced glucose, lactate and lipid levels in GSDIII non-tumor hepatocytes. In addition, the induction of extrahepatic gluconeogenesis may help compensate for reduced hepatic glucose production in Agl^-/- mice, potentially contributing to the partial maintenance of glycemia during fasting.

Glycophagy is associated with tumor development in GSDIII hepatocytes

Given the substantial energy requirement for tumor proliferation, we investigated which energy substrates GSDIII hepatocytes use to sustain growth and drive tumorigenesis. Tumor cells often adapt their metabolism through the Warburg effect, switching from mitochondrial respiration to glycolysis. However, in Agl^-/- mice, limited glucose and G6P availability and correspondingly low lactate production (Fig. 3F) suggest that a Warburg effect is unlikely. This was confirmed by similar lactate levels in Agl^-/- tumors compared to non-tumor livers (Fig. S4A). Additionally, G6Pase activity was decreased in both non-tumor livers and tumors (Fig. S4B).

To compensate for limited energy substrates, GSDIII tumor cells may use glycogen stores as an energy source. In Agl^-/- mice, we observed a significant reduction in glycogen levels in most tumors compared to non-tumor tissue, with heterogeneity among nodules, confirmed by variable PAS staining, particularly in the two nodule types described in Fig. 2C (Fig. 4A,B). We also assessed the glycogen level in human GSDIII livers from four patients with GSDIII. These patients had a mean age of 55.4 ± 8.5 years at the time of liver transplant or hepatectomy (Table 1). An analysis of native liver histology revealed: in patient 1, cirrhosis with multiple regenerative nodules, but no malignancy; in patient 2, portal fibrosis, one regenerative nodule, and two HCCs with a mixed pattern (trabecular and pseudoacinar); in patient 3, three HCCs and early stage cirrhosis. Patient 4 underwent hepatectomy for malignant nodules and liver transplantation a few years later. Histological analyses showed one large HCC and one large HCA at the time of hepatectomy, and analysis of the native liver identified fibrosis but no cirrhosis, one HCA and two HCCs (Table 1; Fig. 4C). The presence of HCCs was associated with elevated serum AFP levels in patients 2 and 3, but not in patient 4 (Table 1). Interestingly, PAS staining was negative in HCCs (patients 2, 3 and 4), but positive in non-tumor liver and regenerative nodules (patient 1), confirming dramatic glycogen loss with some heterogeneity among nodules (Fig. 4C and S4C).

Fig. 4.

Glycogen is depleted in liver tumors from both GSDIII mice and patients.

(A) Hepatic glycogen content in 14-month-old Agl^+/+ mice (black bars) or Agl^-/- mice in non-tumor livers (NT; red bars) and tumors (T; blue bars). (B) Representative images of HPS and PAS staining in the liver of 14-month-old Agl^-/- mice (41 tumors from 17 mice). Tumors were outlined with hatched line. NT = peritumor tissue. Scale bars = 200 μm. PAS staining from the same sample as in Fig. 2C is included (1: nodule type 1; 2: nodule type 2); scale bars = 500 μm. (C) Representative images of PAS and HES staining on human GSDIII livers from patient 1 with cirrhosis and regenerative nodules and from patients 2-4 with HCCs. NT = peritumor liver. Statistical analyses are detailed in Table S1. ∗∗∗p <0.001. GSDIII, glycogen storage disease type III; HCC, hepatocellular carcinoma; HES, hematoxylin–eosin–saffron; HPS, hematoxylin–phloxine–saffron; PAS, periodic acid–Schiff.

Table 1.

Characteristics of patients with GSDIII at the time of transplantation/hepatectomy.

ID	Sex	Age (years)	Liver weight	Liver pathology	ALT (U/L)	AST (U/L)	AFP (μg/L)
1	M	43	Native liver (2,270 g)	Cirrhosis Multiple regenerative macronodules in all segments	142	158	4.1
2	F	63	Lobectomy of segment II (178 g)	Fibrosis 2 HCCs (4 cm and 0.8 cm) and 1 regenerative nodule (4 cm)	50	69	183.1
3	M	63	Lobectomy of segment IV and VIII	Early-stage cirrhosis with septal fibrosis 3 HCCs (1.5-2.2 cm)	94	90	73
4	M	51	Right hepatectomy	Fibrosis without cirrhosis 1 HCC (10 cm) 1 HCA (11 cm)	nd	nd	nd
		57	Native liver	Fibrosis without cirrhosis 1 HCA (7 cm) 2 HCCs (3.1 cm and 2.2 cm)	60	60	<5

AFP, alpha-fetoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; HCA, hepatocellular adenoma; HCC, hepatocellular carcinoma; LT: liver transplant; nd, not determined.

Since the cytosolic glycogen degradation pathway is impaired in Agl^-/- mice, glycogen degradation may be redirected through the lysosomal glycogen degradation pathway. Consistently, GAA activity was reduced in non-tumor Agl^-/- livers compared to controls but was restored in tumors from Agl^-/- mice (Fig. 5A). Furthermore, TEM revealed an increased number of autophagic vesicles (and a tendency towards an increased number of lysosomes) in tumor tissue compared to non-tumor liver tissue in Agl^-/- mice (Fig. 5B), suggesting enhanced autophagic flux in tumors. On the contrary, the number of autophagic vesicles and lysosomes tended to decrease in non-tumor tissue compared to Agl^+/+ liver (Fig. 5B), concomitantly with decreased lysosome number. Glycophagy is a specialized form of autophagy that enables cells to target glycogen for conversion into glucose in lysosomes.²⁸ To evaluate the glycophagy pathway in Agl^-/- hepatic tumors, we performed GABA_A receptor-associated protein like 1 (GABARAPL1) and STBD1 immunolabelling, two proteins associated with glycophagy,²⁹ to directly quantify the number of glycophagosomes. Importantly, we observed an increased number of glycophagosomes in non-tumor liver and tumors of Agl^-/- mice compared to Agl^+/+ mice (Fig. 5C). Additionally, glycophagosomes were visualized by TEM in Agl^-/- livers (Fig. 5D), supporting the involvement of glycophagy in glycogen breakdown. On the other hand, our data indicate that autophagy was impaired in Agl^-/- tumors, as evidenced by elevated expression of p62 relative to non-tumor liver (Fig. 5E). Moreover, total STBD1 expression was reduced in both non-tumor and tumor tissues of Agl^-/- mice compared to control liver, while GABARAPL1 expression was increased (Fig. 5E,F), suggesting a shift from general autophagy towards specific glycophagy. Of note, elevated levels of Galectin-3, a marker of lysosomal damage, were recently reported in GSDIII mice.³⁰ These results suggest that while GABARALP1 is retained and potentially recycled within phagolysosomes, STBD1 is likely degraded within the glycophagosome lumen. Interestingly, the observed accumulation of p62/autophagosomes, together with decrease STBD1 levels, implies that glycophagy may be upregulated at the expense of general autophagy in GSDIII-associated tumors.

Fig. 5.

Glycophagy is associated with tumor development in GSDIII hepatocytes.

(A) GAA activity in non-tumor/peritumoral (NT; red bar) and tumor (T; blue bar) tissues of 14-month-old Agl^-/- and Agl^+/+ (black bar) mice. (B) Representative TEM images and quantification of lysosomes and autophagic features. Magenta arrows: lysosomes, yellow arrows: autophagosomes. (C) Immunofluorescence of GABARAPL1 and STBD1 proteins in mouse livers. Arrows indicate glycophagosomes (D) Representative TEM images of glycophagosomes in 14-month-old Agl^-/- livers. Scale bars = 500 μm. (E-F) STBD1, p62, and GABARAPL1 protein expression in livers. Statistical analyses are detailed in Table S1. ∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001. GAA, acid alpha-glucosidase; GABARAPL1, GABA type A receptor–associated protein like 1; GSDIII, glycogen storage disease type III; STBD1, starch binding domain containing 1; TEM, transmission electron microscopy.

Taken together, these results indicate that in the absence of GDE, tumor cells in both mice and humans likely enhance their capacity for glycogen degradation via the lysosomal autophagic pathway, to provide glucose to sustain cell proliferation and growth.

Discussion

Liver pathology in GSDIII is characterized by the accumulation of structurally abnormal glycogen (limit dextrin), and early-onset hepatic fibrosis. In 40-45% of adults, fibrosis can progress to cirrhosis and, in approximately 15%, to HCA/HCC, typically after 65 years of age.¹⁷ Advances in dietary management aimed at preventing hypoglycemia and glycogen accumulation have improved life expectancy,¹² making cirrhosis and liver tumors an emerging concern.

The Agl^-/- mouse represents a robust GSDIII model that faithfully recapitulates all hepatic and muscular disease features.¹⁸ About 30% of Agl^-/- mice develop liver tumors by 14 months, a much higher incidence than currently reported in patients. However, the true prevalence of liver tumors in GSDIII is likely underestimated and may approach that seen in mice, given the limited clinical data on patients over 40–50 years old and the typically late onset of tumor development.^14,16,31 This highlights the urgent need to investigate the molecular pathways driving tumorigenesis in GSDIII.

Abnormal soluble glycogen accumulation in GSDIII liver may drive cellular stress, inflammation, and fibrosis. Limit dextrin likely induces ER stress,¹⁰ which promotes fibrosis and HCC.²⁵ Consistently, Agl^-/- livers exhibited ER dilation and chronic ER stress with increased eIF2α phosphorylation at Ser51, which may exert a cytoprotective effect under low glucose.³² ER stress may be worsened by impaired protein glycosylation due to reduced intracellular glucose.³³ The inflammatory microenvironment further contributes to tumor initiation and progression.³⁴ Although detailed immune profiling was not performed, Agl^-/- livers showed increased T lymphocytes and macrophages compared to Agl^+/+ mice. These macrophages likely promote hepatic stellate cell activation.³⁵ Thus, as seen in patients, Agl^-/- mice developed HCC with fibrosis, but without cirrhosis, indicating that HCC can develop independently of cirrhosis, as observed in patients. This finding suggests that cirrhosis is not a prerequisite for tumorigenesis in GSDIII, and that alternative mechanisms may drive malignant transformation in this context.

Loss of GDE activity in Agl^-/- mice caused a significant decrease in blood glucose under both fed or 6 h-fasted conditions vs. controls. However, only a few Agl^-/- mice developed mild hypoglycemia (blood glucose between 50 and 70 mg/dl), less severe than in GSDIa models.¹⁹ Similarly, patients with GSDIII show milder hypoglycemia than those with GSDIa/b, due to preserved gluconeogenesis. In the absence of glycogenolysis, gluconeogenesis can occur in the liver, kidney, or intestine, to maintain euglycemia. While hepatic gluconeogenesis was expected to be preserved, hepatic glucose and G6P levels were reduced in Agl^-/- compared to Agl^+/+ mice and expression of G6PC1 and PEPCK-c was not upregulated. Thus, Agl^-/- mice likely maintain euglycemia via extrahepatic gluconeogenesis, supported by increased kidney G6PC1 and PEPCK-c expression. Similar mechanisms occur during the anhepatic phase of liver transplantation^36,37 and in liver-specific G6pc1 knockout mice (hepatic GSDIa model).^20,38 Interestingly, amino acids, particularly glutamine, are the main substrates for extrahepatic gluconeogenesis.³⁹ Therefore, a high protein diet can help sustain glucose production.¹² Substituting part of the dietary carbohydrate content with protein also helps to reduce the accumulation of glycogen in the liver and muscle.¹²

In contrast to GSDIa, where tumorigenesis involves Warburg-like metabolic reprogramming,⁴⁰ tumors in GSDIII arise under energy depletion. In Agl^-/- hepatocytes, impaired gluconeogenesis and blocked glycogenolysis cause a severe energy deficit, with reduced G6P, glycolysis, lactate, triglyceride and cholesterol levels. These metabolic changes were not associated with any change in lipogenic gene expression, even on high fat/high sucrose diet (data not shown). Despite this, Agl^-/- hepatocytes displayed increased proliferation and tumor formation, implying alternative energy sources support cell growth.

Our results suggest that glycogen may play a dual role in tumor initiation and progression. Excess glycogen may activate YAP by disturbing Hippo/YAP signaling, thereby promoting tumor initiation.¹¹ Since cytosolic degradation is impaired, lysosomal glycogen breakdown via glycophagy could act as a compensatory mechanism. Tumor tissues from Agl^-/- showed increased GAA activity, reduced glycogen content, and glycophagosomes enriched in GABARAPL1 and STBD, indicating active glycogen degradation. Glycophagy, is increasingly recognized as a key physiological pathway in energy homeostasis. While initially recognized for its role in neonatal glycogen degradation to sustain blood glucose levels after birth,⁴¹ glycogen accumulation has recently been identified as an early oncogenic event during liver tumorigenesis.^11,[42], [43], [44], [45] Interestingly, experimental depletion of glycogen stores has been shown to reduce liver cancer incidence, whereas enhanced glycogen storage accelerates tumor development.^10,11 Consistantly, human GSDIII HCCs exhibited reduced glycogen, a finding not observed in regenerative nodules. Despite the heterogeneity in nodule phenotypes and the limited sample size, this suggests that glycogen could be an energy source for tumor growth in both mice and patients with GSDIII. Inhibiting glycogen synthase II by RNAi reduced tumor growth in GSDIII models⁴⁶ and repressing GAA expression could validate this metabolic model. Thus, the Agl^-/- mouse model could be a valuable tool for identifying anti-tumor drugs targeting glycophagy.

In conclusion, this study provides new insights into the metabolic rewiring of the liver in GSDIII, which differs markedly from the one observed in GSDIa.^19,[40], [45] By characterizing glucose and lipid dysregulations, we uncover two distinct molecular mechanisms that contribute to liver tumorigenesis. Our findings support the emerging concept that, in GSDIII, glycogen likely plays dual roles, serving not only as a metabolic substrate but also as a signaling molecule that could drive tumorigenesis. However, tumor heterogeneity suggests that additional pathways likely contribute to liver tumorigenesis. Further research could provide compelling data to shape the current gene therapy strategies for GSDIII under clinical development^47,48 and open new avenues for therapeutic interventions in GSDIII-related hepatocarcinogenesis and potentially other GSDs.

Abbreviations

ACC, acetyl-CoA carboxylase; AFP, alpha-fetoprotein; CPT1α, carnitine palmitoyltransferase 1 alpha; ER, endoplasmic reticulum; G6P, glucose-6-phosphate; G6Pase, glucose-6-phosphatase; GAA, acid alpha-glucosidase; GDE, glycogen debranching enzyme; GSD, glycogen storage disease; GSDIII, glycogen storage disease type III; HCA, hepatocellular adenoma; HCC, hepatocellular carcinoma; HPS, hematoxylin–phloxine–saffron; IHC, immunohistochemistry; PAS, periodic acid-Schiff; TEM, transmission electron microscopy; YAP, Yes-associated protein.

Authors’ contributions

VMR, LJ, GP, FR performed or directed experimental activities, contributed significantly to experimental design and data analysis. VMR, LJ, GR and FR wrote the manuscript. AML, AGS, GM and EM provided insights into the liver metabolism, glycophagy, and TEM technology and analysis. Histological analyses were performed by VP and CM. JC and CT developed the pipeline for quantification of immunostainings. LJ, AG, MA, AAG, AL, CZA, FE, FB, EBF, CG, and LVW contributed to experimental activities. CM, PL, AG, SG, FD, and FM provided the patients' biological data and the histological images of the human biopsies. LJ, AGS, GM, GR, and FR critically edited the manuscript. All authors approved the final manuscript.

Data availability

All data presented in the manuscript are presented in histograms as individual points. There are no large data files to be shared via online resources.

Financial support

This work was supported by Genethon, the ‘’Association Française contre la Myopathie’’, the “Association Francophone des Glycogénoses”, the “Fondation pour la Recherche Médicale” (Equipe FRM EQU202103012572), and the National Research Agency (ANR-17-CE18-0014; ANR-22-PEBI-0012). The salary of AG was supported by public funding from the French National Institute of Health and Medical Research (“poste d’accueil”, INSERM). The salaries of GP and AL were supported by the Gen&Zic association. The salaries of GR, AML & CZA, and FR & GM were supported by Inserm and CNRS, respectively. The salary of AGS was supported by INRAE. The salary of MA was supported by PreciDIAB (ANR-18-IBHU-0001) and Hauts-de-France Regional Council (#22005973).

Conflict of interest

The authors who have taken part in this study declare that they do not have anything to disclose regarding funding or there is no conflict of interest with respect to this manuscript.

Please refer to the accompanying ICMJE disclosure forms for further details.

Supplementary data

The following are the Supplementary data to this article: