An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Noebels JL, Avoli M, Rogawski MA, et al., editors. Jasper's Basic Mechanisms of the Epilepsies. 5th edition. New York: Oxford University Press; 2024. doi: 10.1093/med/9780197549469.003.0055
Lafora disease (LD) is a horrendous progressive myoclonic epilepsy that strikes healthy teenagers and leads to ever-worsening seizures with no relief from antiepileptic drugs; it then transitions to rapid dementia that ends in death typically after 10 years of onset. LD is the result of mutations in either the EPM2A or EPM2B/NHLRC1 gene that encodes the glycogen phosphatase laforin or the E3 ubiquitin ligase malin, respectively. A hallmark of LD is cytoplasmic, aberrant glycogen-like aggregates called Lafora bodies (LBs); thus, LD is also classified as a glycogen storage disease (GSD). Using LD mouse models, multiple laboratories definitively demonstrated that genetic reduction or elimination of glycogen synthesis decreased LB formation and rescued LD epilepsy, neurodegeneration, and brain inflammation. Thus, LBs are the etiological driver of LD. An antibody-enzyme fusion (AEF) was developed that ablates LBs and normalizes both cell signaling and brain metabolism. This AEF is a promising drug for the treatment of LD and other GSDs as well as a putative therapeutic platform for targeting other neurodegenerative diseases.
Lafora disease (LD) is a devastating, inherited fatal disease characterized by progressive myoclonic epilepsy and childhood dementia with death typically occurring 10 years after diagnosis (Gentry et al., 2018; Turnbull et al., 2016). LD patients progress normally during their first decade of life before experiencing an initial generalized seizure and/or myoclonic jerk. During their second decade, patients experience increasingly severe and more frequent epileptic episodes along with rapid cognitive decline, increased aphasia and ataxia, and eventually a vegetative state prior to death from aspiration pneumonitis or a massive seizure (Turnbull et al., 2016). Antiepileptic drugs provide some initial palliative relief, but seizures invariably return.
LD was first described in 1911 by Dr. Gonzalo Rodriguez Lafora, who identified the histological hallmark of LD, Lafora bodies (LBs) (Fig. 55–1; Lafora, 1911; Lafora and Gluck, 1911). Indeed, it is now known that LBs are periodic-acid Schiff-positive (PAS+), aberrant, glycogen-like aggregates that form in the cytoplasm of cells from most tissues, notably in both glial cells and neurons (Duran et al., 2019a, 2020, 2021; Gentry et al., 2009; Rubio-Villena et al., 2018; Auge et al., 2018). Thus, LD is the only progressive myoclonus epilepsy that is also classified as a glycogen storage disease (GSD) with GSDs affecting 1:20,000–40,000 (Adeva-Andany et al., 2016; Gentry et al., 2018; Laforet et al., 2021). The first 90 years of LD research provided insights into the physiology of the disease, including the regional location of brain LBs and patient epileptic activity. Then, genetic analysis allowed identification of the cellular basis for LD, when mutations in one of two genes were shown to be causative of LD in 1999 and 2003 (Chan et al., 2003; Minassian et al., 1998; Serratosa et al., 1999).
A timeline summarizing Lafora disease (LD) research. The disease was first described in 1911. The genetic loci were identified in 1998 and 2003. Over the next 15 years, LD researchers developed cell culture models, LD mouse models, and determined the (more...)
LD is caused by mutations in either Epilepsy Progressive Myoclonus 2A (EPM2A), which encodes the glycogen phosphatase laforin; or EPM2B/NHLRC1, which encodes the E3 ubiquitin ligase malin (Chan et al., 2003; Minassian et al., 1998; Serratosa et al., 1999). Laforin is a bimodular protein comprised of a carbohydrate binding module family 20 domain (CBM) and a dual-specificity phosphatase (DSP) domain. Laforin binds glycogen via an integrated glycogen binding platform involving the CBM and DSP domains and removes phosphate from glycogen via the DSP domain (Nitschke et al., 2013; Raththagala et al., 2015; Tagliabracci et al., 2007, 2008; Worby et al., 2006). Laforin also interacts with a number of proteins involved in glycogen metabolism and likely serves as a scaffolding protein that is critical for glycogen metabolism (Cheng et al., 2007; Roma-Mateo et al., 2012; Solaz-Fuster et al., 2008; Vilchez et al., 2007; Worby et al., 2008). Malin is an E3 ubiquitin ligase that ubiquitinates multiple enzymes involved in glycogen metabolism (Gentry et al., 2018; Sun et al., 2019; Worby et al., 2008). Loss of either laforin or malin function results in LBs that contain increased levels of phosphate, longer glucose chains, and aberrant chain branching compared to glycogen (DePaoli-Roach et al., 2015; Nitschke et al., 2017; Sullivan et al., 2019; Tagliabracci et al., 2007, 2008, 2011).
Taken together, these discoveries allowed significant biochemical and cell biology insights to be gained and LD animal models to be generated. Building on these foundational discoveries, multiple independent laboratories generated preclinical LD models and definitively demonstrated that LBs are the etiological driver of LD epilepsy, neurodegeneration, and brain inflammation (Gentry et al., 2018; Fig. 55–1).
Multiple laboratories generated Epm2a–/– (laforin knockout [KO]) and Epm2b–/– (malin KO) LD mouse models (Criado et al., 2012; DePaoli-Roach et al., 2010; Ganesh et al., 2002; Turnbull et al., 2010, 2011, 2014; Valles-Ortega et al., 2011). LBs are abundant in the brains and other tissues of these LD mouse models. The LBs are accompanied by increased susceptibility to kainate-induced epilepsy, neurophysiological alterations in hippocampal synapse electrophysiological properties, and progressive neuronal cell death (Turnbull et al., 2011, 2014; Duran et al., 2020, 2014, 2021; Rai et al., 2017; Sinha et al., 2021). Thus, the LD mouse models replicate LD with respect to LB formation, an epileptic phenotype, and increased neurodegeneration.
By crossing these LD mouse models with genetic models that ablate glycogen formation, several groups independently and conclusively demonstrated that LBs drive LD epilepsy, neuronal cell death, and inflammation. Malin KO mice were crossed with mice lacking the gene that encodes glycogen synthase (Gys1) in the brain and these double KO mice were devoid of LBs, exhibited no inflammation or neurodegeneration, displayed normal electrophysiology, and showed decreased susceptibility to kainite-induced epilepsy (Duran et al., 2014). Additionally, malin KO mice lacking one Gys1 allele showed a partial rescue of all phenotypes, demonstrating that 50% reduction of glycogen synthesis was corrective (Duran et al., 2014). A similarly striking rescue of the phenotypes and correction of electrophysiology was also observed when Gys1 KO mice were crossed with laforin KO mice, demonstrating that the phenotypes from loss of malin or laforin could be rescued by decreased glycogen synthesis (Pederson et al., 2013). Importantly, the reverse paradigm has also been demonstrated. Expression of a constitutively active form of glycogen synthase in the brain of otherwise wild-type mice and flies drove rapid neurodegeneration (Duran et al., 2012).
An additional line of evidence to support this finding involved the positive regulator of glycogen synthesis called protein targeting to glycogen (PTG). Ptg–/– mice exhibit a 50% reduction of glycogen levels (Turnbull et al., 2011, 2014). A double transgenic mouse model harboring deletions of both Epm2b and Ptg yielded near complete elimination of LBs and animals displayed no increase in neuronal cell death, no myoclonus, and no increased seizure susceptibly to kainite acid (Turnbull et al., 2014). Similarly, laforin KO mice lacking Ptg exhibited dramatically reduced LBs with minimal to no neuronal cell death and no epileptic phenotype (Turnball, 2011). Most recently, malin KO mice were crossed with mice lacking Gys1 in brain astrocytes using a Gfap-driven cre-lox system (Duran et al., 2021). These mice lacked astrocytic LBs while they still have neuronal LBs. The researchers employed a series of electrophysiology studies combined with immunohistochemistry and metabolomics to demonstrate that astrocytic LBs are the main driver of inflammation and neurodegeneration, while neuronal LBs drive seizure susceptibility. Therefore, a therapeutic strategy is needed to stop and/or ablate LBs in both neurons and astrocytes to address the epilepsy, neurodegeneration, and inflammation.
Cumulatively, several groups utilizing multiple LD mouse models definitively demonstrated that the absence of glycogen synthesis ablates LB formation and rescues LD mouse phenotypes. Indeed, the genetic data established that neuronal LBs underlie the LD epileptic phenotype while astrocytic LBs cause neurodegeneration. In LD, glycogen-like aggregates are synthesized and not broken down which is analogous to a faucet that is always running. Genetically decreasing glycogen synthesis to treat LD is analogous to decreasing the flow from a faucet (Fig. 55–2). These studies provided the foundational impetus to design LD therapies targeting glycogen synthesis and LBs. Moving forward, multiple groups are focused on developing novel therapeutic strategies to ameliorate, treat, and cure LD by targeting LBs via multiple complementary modalities. Two major goals are currently being pursued: preventing new LBs from forming and clearing existing LBs. To prevent LBs from forming, antisense oligonucleotides and pharmacological methods to modulate glycogen synthesis and degradation rates are being pursued (Gentry et al., 2020; Markussen et al., 2021). This review focuses on clearing existing LBs with an antibody-enzyme fusion to treat and reverse LD.
In normal brain metabolism, glycogen is synthesized by glycogen synthase (GYS1) and broken down by the combined efforts of brain glycogen phosphorylase (PYGB) and glycogen debranching enzyme (GDE). In Lafora disease (LD), the aberrant glycogen-like Lafora (more...)
Glycogen is the major glucose storage cache and plays a key role in brain homeostasis and metabolism (Brown et al., 2019; Dienel and Carlson, 2019; Dienel and Rothman, 2019). Glycogen contains up to ~55,000 glucose moieties comprised of 13 glucose unit chains, on average, linked by α-1,4-glycosidic linkages with branches joined by α-1,6-glycosidic linkages (Brewer and Gentry, 2019; Roach et al., 2012). Glycogen is synthesized via the combined effort of glycogen synthase, glycogen debranching enzyme, and glycogenin. Synthesis is initiated by formation of a glycogenin dimer that autoglucosylates multiple glucose residues using UDP-glucose as the glucosyl donor. Glycogen synthase binds the glycogenin-glucose substrate and catalyzes α-1,4-glycosidic linkages to increase the length of the glucose chains (Brewer and Gentry, 2019; Roach, 2002; Roach et al., 2012). There are two isozymes of glycogen synthase in humans, GYS1 and GYS2. GYS2 is the liver enzyme and GYS1 is present in skeletal muscle, heart, brain, and other non-liver tissues. To generate the glycogen macromolecule, glycogen branching enzyme cleaves a segment of an α-1,4-glucose chain and transfers it to generate α-1,6-branches that can be extended by glycogen synthase. These three enzymes generate a single glycogen particle that is referred to as a glycogen β-particle. In the liver, these glycogen β-particles can be connected to form larger α-particles (Nawaz et al., 2021). Additionally, glycogen contains covalently attached phosphate at the C2-, C3-, or C-6 hydroxyls of ~1:300–3,000 glucose moieties (Contreras et al., 2016; Nitschke et al., 2013; Tagliabracci et al., 2007, 2008, 2011). Glycogen synthase can transfer the β-phosphate of UDP-glucose into glycogen as a side reaction (Chikwana et al., 2013). This mechanism could account for phosphate at the C2- and C3-positions, but not phosphate at C6. Further, the role of phosphate in glycogen structure is largely unknown. Thus, glycogen phosphate is an area of active research.
Similar to synthesis, glycogen is degraded by concerted effort of three enzymes. Glucose-1-phosphate is liberated from cytoplasmic glycogen by glycogen phosphorylase that releases terminal glucose residues before stopping four residues from an α-1,6-branch. There are three isoforms of phosphorylase named according to the tissues in which they are enriched: brain (PYGB), muscle (PYGM), and liver (PYGL). The four glucose oligosaccharides that remain from phosphorylase are cleaved by glycogen debranching enzyme. Debranching enzyme cleaves the four as a single chain and reforms them as an α-1,4-glucose chain on an existing chain so that glycogen phosphorylase can degrade it (Roach, 2002; Roach et al., 2012).
In addition to cytoplasmic glycogen degradation, glycogen is also degraded in the lysosome. Lysosomal glycogen is degraded by acid-alpha-glucosidase (GAA) that hydrolyzes both α-1,4- and α-1,6-linked chains. The biological role of lysosomal glycogen degradation is still largely unknown, but GAA is a widely expressed enzyme throughout most tissues and mutations in GAA result in the glycogen storage disorder Pompe disease (GSD III).
Glycogen architecture is an emerging concept that has profound biological and disease implications (Brewer and Gentry, 2019). Glycogen architecture encompasses the length of glucose chains, the frequency of branching, particle size, and the amount of phosphate covalently attached to the glucose residues. In the liver, the concept also encompasses the ratio of α- and β-particles. The relationship between chain length, branching frequency, particle size, and phosphorylation is still being defined. However, perturbations in one or more of these parameters is observed in the insoluble glycogen-like aggregates from both LD and adult polyglucosan body disease (APBD) with chain length being aberrant in both (Sullivan et al., 2019). Glucose chain length is key to maintaining normal architecture because glucose chains, like nucleotide chains, can form helical structures. Glycogen remains water-soluble because the glucose chains are short, that is, on average 13 glucose units long, and branching frequency is consistent so that adjacent chains cannot interact. When either chain length is increased and/or branches are closer together, then adjacent chains form helices that essentially exclude water. These changes render the aggregates resistant to degradation by glycogen phosphorylase and glycogen debranching enzyme. Mutations in the glycogen synthesis or degradation enzymes result in GSDs, many of which form aberrant glycogen-like aggregates. Much effort has focused on defining the architecture of LD LBs to devise a strategy to inhibit and/or degrade them. However, the architectural properties of other GSD glycogen-like aggregates have not been entirely defined.
The two major tissues that store glycogen are the liver and skeletal muscle, but the brain is most susceptible to decreases in glucose availability (Dinuzzo et al., 2014; Fryer and Brown, 2014). Brain glycogen levels vary with wake/sleep cycles and mice lacking all glycogen synthase activity show significant deficiency in learning and hippocampal synaptic plasticity (Duran et al., 2013a, 2020; Dinuzzo et al., 2014; Kong et al., 2002). Similarly, genetically engineered mice that are devoid of brain glycogen have defects in short-term memory consolidation (Gibbs and Hutchinson, 2012), long-term memory formation (Duran et al., 2013b), and maintenance of long-term potentiation of synaptic strength (Duran et al., 2019b).
Glycogen was initially observed in glia and embryonic neurons using methods with limited sensitivity (Cataldo and Broadwell, 1986; Koizumi, 1974). Recent work demonstrated that adult neurons express both glycogen synthase and glycogen phosphorylase and that neuronal glycogen plays important roles in hypoxia resistance and in synaptic plasticity (Vilchez et al., 2007; Duran et al., 2013a, 2014, 2019a; Saez et al., 2014; Pfeiffer-Guglielmi et al., 2003). However, glycogen synthesis must be tightly controlled in neurons and glia because overaccumulation induces apoptosis (Valles-Ortega et al., 2011; Vilchez et al., 2007; Duran et al., 2014). Neuronal LBs are larger and amorphous while glial LBs are smaller and spherically shaped (Auge et al., 2018). The neuronal LBs are primarily responsible for the epileptic phenotype and glial LBs are responsible for neurodegeneration and inflammation (Duran et al., 2021).
Brain glycogen has been visualized by electron microscopy, magnetic resonance spectroscopy, PAS staining, and immunohistochemistry (Oe et al., 2016, 2019). Using immunohistochemistry, the brain regions with high glycogen content include the subiculum region immediately adjacent to the neuronal cell layers of the hippocampus, grey matter of the cerebral cortex, and brainstem (Oe et al., 2019). We recently developed a technique employing matrix-assisted laser desorption/ionization high-resolution mass spectrometry imaging (MALDI-MSI) to both quantify and spatially visualize glycogen in situ with high specificity and sensitivity (Sun et al., 2021). Formalin-fixed paraffin-embedded brain sections were analyzed by MALDI-MSI to reveal near universal glycogen accumulation in the mouse brain with highly heterogeneous abundance in different brain regions. Grey matter regions of the frontal cortex and neuronal hippocampal cell layers displayed significantly higher glycogen levels compared to the rest of the brain. These data highlight and support the role of glycogen in neuronal function, signal transduction, and memory formation.
Cumulatively, several groups utilizing multiple LD mouse models conclusively demonstrated that LBs cause the disease phenotype of LD and that genetically preventing LB formation rescues the disease phenotypes in LD mouse models. These studies provided the foundational impetus to design LD therapies targeting LBs. Antisense oligonucleotides and chemical small molecules targeting the glycogen synthase gene or enzyme, respectively, are being developed to prevent LBs from forming (Ahonen et al., 2021; Gentry et al., 2020; Gumusgoz et al., 2021; Markussen et al., 2021; Tang et al., 2020). The second strategy is to clear existing LBs, which is the critical need of symptomatic patients. To achieve this goal, novel enzyme therapies are being developed.
We recently published three papers presenting work that achieved the goal of ablating existing LBs and normalizing brain metabolism, thereby having considerable promise as a first-line therapeutic modality for symptomatic patients (Austin et al., 2019; Brewer et al., 2019; Sun et al., 2021). We developed an antibody-enzyme fusion (AEF) comprised of a cell penetrating antibody fragment fused to pancreatic α-amylase that can degrade LBs in vivo (Brewer et al., 2019). We also established an enzyme-linked immunosorbent assay (ELISA) to quantify the fusion protein biodistribution (Austin et al., 2019). Using the ELISA and immunohistochemistry, we defined the optimal central nervous system (CNS) delivery method. Lastly, we demonstrated that treatment with the fusion protein normalizes brain metabolism and protein glycosylation in an LD mouse model (Brewer et al., 2019; Sun et al., 2021).
LBs are resistant to degradation by glycogen phosphorylase and glycogen debranching enzyme, the enzymes that release glucose from glycogen. However, the digestive system produces salivary and pancreatic amylases that degrade complex carbohydrates. Similarly, plants, fungi, and bacteria produce a host of glycoside hydrolases, transglycosidases, glycosyl transferases, and lytic polysaccharide monooxygenases to degrade recalcitrant carbohydrates.
We recently developed a method to purify LBs from mouse brain, heart, and skeletal muscle tissue that maintains the physiologically relevant architecture of the LBs (Brewer et al., 2019, 2020; Sun et al., 2021). Using these purified LBs, a panel of carbohydrate degrading enzymes were tested and pancreatic α-amylase was identified as the most efficient enzyme to degrade LBs in vitro (Brewer et al., 2019).
A major limitation in drug development is inadequate therapeutic delivery due to insufficient cellular uptake (Rehman et al., 2016). This inefficiency is especially problematic for enzyme therapeutics. Antibodies offer an ideal delivery platform that can be engineered and optimized to overcome this issue and deliver payloads to either specific or systemic tissues as well as specific cell type targeting (Zhou et al., 2019). Initially, antibody-drug conjugates, antibody-directed enzyme prodrug therapy, and antibody-enzyme fusions (AEFs) were developed and employed as chemotherapies due to their ability to concentrate toxic payloads to specific cell types. Antibody-based drugs have made a major impact on the treatment of cancer as well as both hyperinflammatory and autoimmune diseases with worldwide revenues of $115 billion in 2018 with projections of $300 billion by 2025 (Lu et al., 2020). More recently, these modalities are being utilized to deliver enzymes that are deficient in muscular dystrophies, neurodegenerative diseases, and GSDs (Brewer et al., 2019; Layton and Hellinga, 2011; Weisbart et al., 2005).
Antinuclear autoantibodies (ANA) are generated in systemic lupus erythematosus. These ANAs are IgG antibodies that can penetrate cells and localize to multiple cellular compartments (Alarcon-Segovia et al., 1978; Hansen et al., 2007). While many ANAs are pathogenic, some variants are nonpathogenic and have been developed for their potential as cellular delivery platforms.
The 3E10 ANA is the basis for one such platform. 3E10 is nonpathogenic, penetrates cell membranes, and localizes to the nucleus and cytoplasm (Hansen et al., 2005, 2007, 2012; Weisbart et al., 2012, 2015). 3E10 enters cells by the equilibrative nucleoside transporter 2 (ENT2, SLC29A2) that is involved in the nucleoside salvage pathway (Weisbart et al., 2015). While the full mechanism of how ENT2 transports 3E10 has not been fully elucidated, the full-length 3E10, its antigen-binding fragment (Fab), and its variable fragments have all been used to deliver intracellular cargo (Brinkmann and Kontermann, 2017; Chan et al., 2016; Hansen et al., 2005; Weisbart et al., 1998, 2012). Derivatives of 3E10 have been fused to other proteins or cytotoxic agents to yield promising results in cancer models (Hansen et al., 2012; Noble et al., 2015; Turchick et al., 2017, 2019). The 3E10 single-chain variable fragment was fused to microdystrophin and successfully penetrated multiple cell lines to gain cytoplasmic localization, showing promise as a therapeutic platform for dystrophin-deficient muscular dystrophies (Weisbart et al., 2005). A humanized 3E10 Fab fragment was fused to the lipid phosphatase myotubularin and utilized in a mouse model of myotubular myopathy. Intramuscular injections of the 3E10-myotubularin fusion improved contractile function, muscle pathology, and gait (Lawlor et al., 2013). Thus, the 3E10 Fab has delivered both small-molecule and protein fusions into the cytoplasm, delivering payloads up to 155 kDa in size.
To deliver pancreatic α-amylase into the cytoplasm, we generated the AEF designated VAL-0417 by fusing pancreatic α-amylase to the humanized 3E10 IgG1 Fab and coexpressed this fusion with the 3E10 light chain. The secreted 100 kDa VAL-0417 heterodimer fusion was purified from cell media by affinity chromatography. VAL-0417 was incubated with LBs purified from LD mouse brain, heart, and skeletal muscle and the AEF demonstrated robust digestion of LBs in vitro (Fig. 55–3; Brewer et al., 2019). Pancreatic α-amylase releases glucose, maltose, maltotriose, and other short oligosaccharides when it degrades a carbohydrate (Robyt and French, 1967). To assess the products generated by VAL-0417, we isolated the LB degradation products after treatment and found that maltose and glucose were the major degradation products with maltose as the primary product (Brewer et al., 2019).
A. Schematic of Lafora body (LB) purification and treatment. LBs were purified from the brain, heart, and skeletal muscle (Sk. M.) of LD mice and incubated with VAL-0417 in vitro. B. Scanning electron micrographs of purified LBs. C. LBs were (more...)
The 3E10 Fab is transported into cells by the ENT2 transporter that is ubiquitously expressed on cell membranes in human and mouse tissues (Crawford et al., 1998; Hansen et al., 2005, 2007; Lu et al., 2004). To test if VAL-0417 could gain access and remain active in cells, VAL-0417 was incubated with fibroblast-derived Rat1 cells that accumulate normal levels of glycogen and also with HEK293 cells overexpressing the glycogen synthesis activator Ptg that accumulate enhanced glycogen levels. Both cell lines expressed detectable levels of ENT2 and VAL-0417 gained access into cells as judged by Western analysis and ELISA in a dose-dependent manner that corresponded to a dose-dependent decrease in glycogen levels. Thus, VAL-0417 was actively transported into both cell types, and once in the cytoplasm, it remained active and degraded glycogen.
Since LBs are found in most tissues analyzed in LD patients and mouse models, VAL-0417 efficacy was tested by performing intramuscular (i.m.) and intravenous injections. LD mice develop robust LB levels and pathology by 3 months of age. Ten-month-old wild-type (WT) and laforin KO mice received three i.m. injections in the right gastrocnemius with VAL-0417 or PBS over the course of 1 week on days 1, 4, and 7. The mice were euthanized on day 8, and total glycogen and LB levels were assessed. The gastrocnemius of mice treated with VAL-0417 displayed a >50% reduction in combined glycogen and LB levels relative to PBS-treated control animals defined by biochemical analysis (Brewer et al., 2019). VAL-0417-treated animals displayed similar total carbohydrate levels as WT animals. Intriguingly, WT mice treated with VAL-0417 did not show a reduction in glycogen levels. This result was striking since VAL-0417 readily degrades WT glycogen in vitro and in cellulo. This result has since been verified numerous times in multiple tissues, and each time glycogen levels in WT animals are not decreased by treatment with VAL-0417. Subsequent data suggest that WT cells respond to VAL-0417 treatment by regulating enzymes known to control energy metabolism, specifically AMP-activated protein kinase and downstream signaling modules. This signaling response presumably increases glycogen production in response to VAL-0417 treatment so that WT cells maintain appropriate glycogen levels.
Intravenous (i.v.) injections provide rapid and more extensive biodistribution of drug than i.m. Since LD patients and mouse models exhibit LBs in multiple systemic tissues, VAL-0417 was administered to 8-month-old laforin KO mice by four i.v. injections over the course of 2 weeks. LB loads are very high in the heart, and cardiac abnormalities have been reported in both LD mice and patients (Villalba-Orero et al., 2017). The hearts of laforin KO animals treated with VAL-0417 exhibited a 67% decrease in combined glycogen and LB levels as defined by biochemical analysis, bringing levels down to near WT levels (Brewer et al., 2019). Similarly, PAS-stained heart tissues from laforin KO animals i.v. treated with VAL-0417 exhibited marked decreases in LBs. VAL-0417 biodistribution was also assessed in the gastrocnemius, quadricep, heart, and liver, and VAL-0417 levels were similar in all tissues, suggesting that ENT2 levels were not limiting in any tissue. Collectively, these data demonstrated that VAL-0417, delivered by i.m. or i.v. injections, gains access into cells, is enzymatically active in cells, and degrades cytoplasmic LBs in vivo.
While LBs accumulate in most tissues, LBs in the brain drive LD epilepsy, neurodegeneration, and death of the patient. The blood–brain barrier (BBB) presents a challenge to most drugs, especially antibody-based drugs (Woalk and Thorne; 2013). The BBB is comprised of the CNS vasculature that allows selective nutrients and metabolites across the endothelial cells and the brain parenchyma while opposing entrance of other entities via the endothelial cell tight junctions as well as efflux transporters and restricted transcytosis (Segarra et al., 2021). While the BBB is a dynamic platform that changes during sleep/wake cycles, aging, and epilepsy, it is unlikely that a 100 kDa AEF could cross this barrier to the extent necessary for bioactivity. This necessitated assessing the efficacy of direct physical administration into the CNS.
To assess brain biodistribution of VAL-0417, an enzyme-linked immunosorbent assay (ELISA) was developed. The VAL-0417 ELISA employs an anti-3E10 Fab antibody to capture VAL-0417 and an anti-α-amylase with the appropriate secondary antibody that detects VAL-0417 to between 10–50 pg/μL of biofluid (Austin et al., 2019). With the ELISA in place, we tested CNS delivery of VAL-0417 by continuous infusion for 14 days via either intracerebroventricular (i.c.v.) or intrathecal (i.t.) administration. Administration by i.c.v. directly infuses drug into the brain ventricles and is utilized via Ommaya reservoirs in the clinic for adult and pediatric CNS cancer patients, patients with neuronal ceroid lipofuscinosis, and pain management (Anton et al., 2021; Cohen-Pfeffer et al., 2017; Peyrl et al., 2014; Slavc et al., 2018). Alternatively, i.t. infusion provides delivery to the brain via cerebrospinal fluid (CSF) movement through the CNS and is less invasive. Mice were administered the same amount of VAL-0417 via i.c.v. or i.t. for 14 days. The i.t. catheter was intrathecally implanted via surgery in the lumbar region while the i.c.v. canula was implanted in the cerebral lateral ventricle. The catheter and cannula were attached to an osmotic pump delivering continuous VAL-0417 administration. Mice were sacrificed on day 15, and the brains were rapidly dissected and sectioned into six lateral sections moving from rostral to caudal. Both administration methods generated detectable VAL-0417 in the sections with i.t. yielding ~200 pg of VAL-0417/μg of total protein in each section. Alternatively, i.c.v. administration resulted in 400 pg/μg of total protein in the most rostral and caudal sections with a maximum of 1600 pg/μg of total protein in the most central section. Immunohistochemistry (IHC) using an anti-α-amylase antibody revealed intense intracellular staining of VAL-0417 via both i.t. and i.c.v. administration. The diffusion pattern of VAL-0417 was similar to previous antibody diffusion patterns throughout the perivascular and extracellular brain spaces (Pizzo et al., 2018; Wolak et al., 2015).
Drug delivery to the CNS is a significant challenge with both i.t. and i.c.v. injections utilized in the clinic to bypass the BBB, as discussed above (Calias et al., 2014; Wolak and Thorne, 2013). Distribution of all CNS drugs is highly dependent on fluid movement throughout the CNS, and this dependency is even more critical for large biological drugs like VAL-0417. Defining CNS fluid movement is a relatively nascent field of research and detailed biodistribution of biological drugs via i.c.v. and i.t. delivery is ongoing (Abbott et al., 2018; Kumar et al., 2018; Plog and Nedergaard, 2018). While i.c.v. administration is well-tolerated and becoming the standard of care for delivery of cerliponase alfa for children with neuronal ceroid lipofuscinosis type 2 (CLN2) diseases, novel methods are being developed to deliver biologics across the BBB via noninvasive methods. One exciting platform utilizes a modified Fc antibody fragment to interact with BBB transferrin receptors to allow transport via receptor-mediated transcytosis (Kariolis et al., 2020; Ullman et al., 2020). Technologies such as this one and others provide possibilities for future treatment of CNS diseases. Combining these transferrin-based technologies with VAL-0417 would achieve both BBB infiltration via the transferrin approach and cell penetration via the VAL-0417 fusion protein.
Additionally, most drugs injected into the CSF distribute into the blood and ependymal surface of the brain or spinal cord rather than deep into the brain parenchyma (Pardridge, 2016). In our studies, we demonstrated that a 150 kDa AEF is widely distributed throughout the brain after i.t. administration and maximally distributed upon i.c.v. administration. Strikingly, both administration modalities yielded drug deep in the brain parenchyma with strong intracellular localization in mouse models.
Next, we sought to determine if VAL-0417 cellular entry into brain parenchyma is dependent on the 3E10 Fab. Five-month-old laforin KO mice were continuously i.c.v. administered PBS or an equal molar amount of VAL-0417 or recombinant pancreatic α-amylase for 28 days. The mice were euthanized, brains were collected, and each brain was processed for analysis by IHC and ELISA. VAL-0417 displayed robust cytoplasmic staining in all regions of the brain, including the rostral cortex, choroid plexus, medial cortex, and cerebellum (Austin et al., 2019). Conversely, pancreatic α-amylase lacking the 3E10 Fab displayed little to no intracellular staining in any regions. ELISA data corroborated these findings by detecting 700–3000 pg of VAL-0417/μg of total protein in the six brain sections and no detectable pancreatic α-amylase above background in any of the sections (Austin et al., 2019). LB loads were decreased in all six brain sections in mice treated with VAL-0417 with the most significant decreases in the rostral sections (Austin et al., 2019). These data demonstrate that VAL-0417 degrades brain LBs and that the 3E10 platform is necessary to deliver pancreatic α-amylase into cells.
We then performed a second i.c.v. experiment with an increased dose of VAL-0417 for a shorter duration to define initial preclinical parameters. Seven-month-old laforin KO mice received continuous administration of VAL-0417 or PBS for 7 days (Fig. 55–4A). The mice were euthanized on day 8, brains were hemisected with the right hemisphere fixed for PAS staining and the left hemisphere frozen for biochemical analyses. In the PBS-treated laforin KO mice, LBs were most abundant in the hippocampus, cerebellum, and brainstem with a lower density in the thalamus and least abundant in the frontal cortex. Strikingly, VAL-0417 treatment completely ablated PAS+ LBs in every region of the brain, including the cerebral cortex, thalamus, cerebellum, and brainstem, and these data were corroborated by biochemical analysis (Fig. 55–4B; Brewer et al., 2019). This result demonstrates the therapeutic promise of this novel AEF to clear LBs.
A. Schematic of intracerebroventricular (ICV) administration of VAL-0417 into LD mice. B. PAS-stained brain slices of LD mice untreated, PBS treated, VAL-0417 treated, and wild-type (WT) mice. LB deposits appear purple; tissue was counterstained (more...)
Metabolic pathways are frequently more highly conserved between species than behavioral phenotypes (Perlman, 2016). While there are often species-specific physiological responses to cellular perturbations, the underlying metabolic changes are similar. This phenomenon of similar metabolic signatures has been reported between mice and man for Alzheimer disease, Huntington disease, nonalcoholic fatty liver disease, and type 2 diabetes mellitus (Barr et al., 2010; Manna et al., 2014; Salek et al., 2007; Underwood et al., 2006). Additionally, studies from multiple diseases are demonstrating that the metabolic changes in humans and mice are highly correlated (Barr et al., 2010; Salek et al., 2007; Trushina and Mielke, 2014; Underwood et al., 2006). Since LD is a disease of perturbed glucose metabolism, we hypothesized that LD mouse models would exhibit robust defects in metabolism.
Brain metabolism underlies the biochemistry of memory, cognition, and behavior (Gallagher et al., 1956; Hertz and Dienel, 2002; Vannucci and Vannucci, 2000; Cunnane et al., 2011). While the brain only represents 2% of total body weight, it is responsible for 20% of daily glucose consumption (Mergenthaler et al., 2013). Glucose metabolism involves a complex array of interconnected metabolic networks that includes the initial breakdown of glucose to fuel bioenergetics, redox balance, amino acid production, and nucleotide biosynthesis. Each of these pathways is critical for neuronal and glial function to maintain normal signaling and synaptic transmission (Schurr et al., 1988; Magistretti and Pellerin; 1999; Rangaraju et al., 2014). Defective glucose metabolism has been reported in Alzheimer disease, Parkinson disease, refractory epilepsy, and aging (Mosconi, 2005; Oddo et al., 2008; Engel et al., 1982; Kuhl et al., 1984; DeFronzo, 1981; Kalpouzos et al., 2009). Persistent altered glucose metabolism has also been observed in acute brain injuries, potentially continuing for years, and correlating with a patient’s chronic neurological dysfunction (Oddo et al., 2008; Yamaki et al., 2018; Kurtz and Rocha, 2020; Bastian et al., 2019). Glycogen is connected to glucose metabolism via glucose-6-phosphate that is a key energy source for central carbon metabolism pathways, for example, glycolysis, Kreb’s cycle, and the pentose phosphate pathway. Indeed, abnormal glycogen has been reported in many neurological disorders, but it has not been extensively characterized or studied (Abubakr et al., 2005; Streichenberger et al., 2001; Gertz et al., 1985; Mann et al., 1987; Dodge et al., 2013, 2015).
In recent years, targeted metabolomics has gained popularity in biomarker discovery, defining network perturbations, and establishing molecular mechanisms that drive inborn errors of metabolism (Chen et al., 2020; Liu et al., 2018; Tasdogan et al., 2020; Dang et al., 2009; Ward et al., 2010; Kind et al., 2007; Claudino et al., 2007; Sellers et al., 2015). The power of metabolomics is such that a single sample can produce profiles of hundreds of polar metabolites and thousands of lipid species. Metabolomics is a workflow combining a series of methodologies rather than a single technique. A typical metabolomics workflow includes rapid and consistent sample acquisition, sample fixation (e.g., liquid nitrogen fixation), sample extraction, mass spectrometry analysis, data reduction, statistical analysis, and data analysis (Fig. 55–5A). Targeted metabolomics approaches allow the simultaneous detection of hundreds of metabolites covering all major metabolic pathways (Lu et al., 2008; Hiller et al., 2009). Current metabolomics techniques combine chromatography separation with a high-resolution mass spectrometer that allows two-dimensional assignment of metabolites with high sensitivity, accuracy, and confidence utilizing established metabolite libraries (Kind et al., 2009; Fiehn, 2016). We recently adapted biomass normalization to improve coefficient of variance, rigor, and reproducibility of metabolite analyses (Andres et al., 2020).
A. Schematic of metabolomics workflow. B. Two-dimensional principle component (PC) plot for polar metabolites of LD mice (Epm2a–/–) untreated (pink), PBS-treated (yellow), and VAL-0417-treated (blue) as well as wild-type (WT) (more...)
Cancer researchers have pioneered many methodologies utilizing targeted metabolomics to make high-impact mechanistic discoveries that are being rapidly translated to clinical diagnostics and treatments (Chen et al., 2019; Liu et al., 2018; Tasdogan et al., 2020; Dang et al., 2009; Ward et al., 2010). Additionally, metabolomics strategies are being employed to identify biomarkers using human biofluids for monitoring disease progression and to test and confirm target engagement for therapy efficacy (Johnson et al., 2016; Spratlin et al., 2009). These strategies are beginning to be adapted and utilized for other areas of research, including brain-centric metabolism questions (Wen et al., 2016; Jové et al., 2014).
The key to assessing any potential LD therapeutic is the extent to which it allows functional restoration in various stages of disease. To define the treatment efficacy of VAL-0417, we employed cutting-edge metabolomics to assess brain metabolism using an information-rich methodology. Since the most striking cellular LD hallmark is the accumulation of glucose-dense LBs and brain glycogen actively contributes to the available pool of glucose-6-phosphate that is key for central carbon metabolism, we hypothesized that LD mice would have marked perturbations in brain central carbon metabolism. Therefore, we performed targeted metabolomics to assess central carbon metabolites of glycolysis, the TCA cycle, and the pentose phosphate pathway, as well as amino acids, nucleotides, and mono saccharides, disaccharides, and trisaccharides.
We pulverized the frozen brain tissue from the 7-day VAL-0417 continuous administration experiment described above. Polar metabolites were extracted and derivatized for analysis by gas chromatography mass spectrometry (GCMS). Metabolites were assigned using an in-house software utilizing a modified-FiehnLib metabolomics library (Fiehn; 2016; Kind et al., 2009). The metabolites were analyzed via multivariate analyses that included both principal component analysis and heat map clustering.
The brain metabolite profiles of untreated WT and laforin KO mice segregated into two very discrete groups via principal component analysis (Fig. 55–5B). The metabolite profile from the PBS-treated laforin KO mice clustered into a third group (Fig. 55–5B). The metabolite differences between the untreated and PBS-treated animals are likely due to the physiological effects of surgery and PBS infusion, most probably due to increase inflammation. Strikingly, the metabolic profile of laforin KO mice treated with VAL-0417 clustered tightly within the WT group. Thus, after LB degradation by VAL-0417 the brain metabolism of the laforin KO animals returned to normal. We performed additional analyses on these data-rich metabolomics datasets and corroborated the results with clustering heat map analysis. The metabolic profile of untreated laforin KO animals was distinct, while the metabolic profiles of the VAL-0417-treated laforin KO animals were interspersed with the WT animals (Fig. 55–5). As expected, mice treated with VAL-0417 also exhibited an increase in monosaccharides, disaccharides, and trisaccharides, which are products of α-amylase LB degradation. Therefore, VAL-0417 ablates LBs, removes a large unwanted aggregate from the cytoplasm, and allows restoration of normal cellular metabolism. These data demonstrate that treatment with VAL-0417 ameliorates the LBs and this amelioration provides a positive physiological response.
These metabolomics studies have revealed foundational understandings of LD and, more recently, uncovered key novel insights into glycogen biology and metabolism. Textbooks state that glycogen is comprised of glucose. While liver and skeletal muscle are >99% glucose, brain glycogen was surprisingly recently shown to be comprised of 75% glucose and 25% glucosamine (Sun et al., 2021). Glucosamine is a key building block for N-linked protein glycosylation. Glycosylation is an abundant and complex modification that occurs both co-translationally and post-translationally in the endoplasmic reticulum and Golgi. Proper N-linked protein glycosylation is critical for protein folding, protein stability, subcellular localization, enzymatic activity, and protein-protein interactions (Schwarz and Aebi, 2011). Multiple brain-centric processes rely on N-linked glycosylation, including modulation of synaptic plasticity, neurite outgrowth, neuron morphology, and cognitive processes, such as learning and memory formation (Benson et al., 2000; Kleene and Schachner, 2004; Wasser et al., 2014). Additionally, aberrant N-linked glycosylation results in neuroinflammation, neuronal cell death, and gliosis (Patterson, 2015).
In addition to glucose, the LBs from LD mice were found to contain excessive amounts of glucosamine. Furthermore, LD brain tissue was nearly devoid of free glucosamine and free UDP-glucosamine, while these metabolites were readily abundant in WT brain tissue. Due to the glucosamine being sequestered in LBs, the LD mice exhibited a dramatic hypo-glycosylation phenotype as assessed by traditional biochemistry, metabolomics, and MALDI imaging (Sun et al., 2021). We hypothesized that treatment with VAL-0417 could degrade LBs and thereby release the needed sugar building blocks to rescue the glycosylation phenotype. Therefore, we analyzed the spatial distribution of N-linked glycans and glycogen in the brains of laforin KO mice treated with VAL-0417 by MALDI imaging. Strikingly, treatment with VAL-0417 dramatically decreased LB loads, and this decrease corresponded to normalization in N-linked glycosylation in multiple brain regions (Fig. 55–5C; Sun et al., 2021). Thus, VAL-0417 degrades LBs, thereby releasing both glucose and glucosamine that is utilized to normalize central carbon metabolism and N-linked glycosylation. Cumulatively, these preclinical data indicate that VAL-0417 is a very promising therapeutic to treat LD.
Full-scale preclinical studies are needed for the continued development of VAL-0417 toward clinical use. Key next steps include determining the minimum dose of VAL-0417 needed to ablate LBs and defining the optimal dosing frequency for achieving and maintaining near-normal glycogen and glycosylation levels in the brain. To date, the minimum treatment time for laforin KO mice with VAL-0417 has been 7 days. Defining the minimum dose with maximum response and administration time for that dose is a critical preclinical step. While it is expected that VAL-0417 will behave similarly in laforin KO and malin KO mice, a study should be performed to assess VAL-0417 treatment in malin KO mice as well. Using these preclinical LD models, the rate of LB reformation after an initial efficacious treatment could also be defined. Given that LBs are not prominent in mice until 3 months of age, it is likely that once LBs are ablated it would take multiple months to reform. If this timeline is true, then one could envision a VAL-based treatment of LD patients every 1–3 years. All of these results should be defined in the context of reducing and/or eliminating LD epilepsy, neurodegeneration, and brain inflammation in model systems.
Additionally, further testing and engineering of VAL-0417 may provide key insights and future clinical benefits. While i.c.v. delivery provided a higher intracellular dose of VAL-0417 in brain parenchyma, it is possible that i.t. administration would provide the needed levels to ablate LBs. Given that i.t. administration is far less invasive, it may prove the delivery method of choice. Novel BBB-penetrating technologies could also be employed with VAL-0417 (Kariolis et al., 2020; Ullman et al., 2020). Moving forward, these studies form a strong preclinical foundation for VAL-based therapies to be tested for safety profiles in nonhuman primates before transitioning to the clinic. This work represents a key path toward the goal of developing effective therapeutics for the treatment of LD. Additionally, these data support the use of metabolomics and glycomics as biomarkers for both LD progression and response to therapy.
This study originated with funding from Valerion Therapeutic and was supported by National Institute of Health grants R01N070899 (M.S.G.), R35NS116824 (M.S.G.), P01 NS097197 (M.S.G.), and AG066653 (R.C.S.). This research was also supported by funding from the University of Kentucky Markey Cancer Center, the Bio-specimen Procurement & Translational Pathology Shared Resource Facility of the University of Kentucky Markey Cancer Center P30CA177558.
M.S.G. received funds from Valerion Therapeutics to initiate this work. M.S.G. is a consultant for Enable Therapeutics, Glut1-Deficiency Syndrome Foundation, Maze Therapeutics, and Chelsea’s Hope. M.S.G., R.C.S., and C.W.V.K. are founders of Atterogen, LLC.
This is an open access publication, available online and distributed under the terms of a Creative Commons Attribution-Non Commercial-No Derivatives 4.0 International licence (CC BY-NC-ND 4.0), a copy of which is available at https://creativecommons.org/licenses/by-nc-nd/4.0/. Subject to this license, all rights are reserved.
Your browsing activity is empty.
Activity recording is turned off.
See more...
