GM2 Gangliosidosis: Clinical Features and Current Therapies

: GM2 gangliosidosis are a group of pathologies characterized by GM2 ganglioside accumulation into the lysosome due to mutations on the genes encoding for the β -hexosaminidases subunits or the GM2 activator protein. Three GM2 gangliosidosis have been described: Tay-Sachs disease, Sandhoff disease, and AB variant. Central nervous system dysfunction is the main characteristic of GM2 gangliosidosis patients that include neurodevelopment alterations, neuroinflammation, and neuronal apoptosis. Currently, there is not approved therapy for GM2 gangliosidosis, but different therapeutic strategies have been studied including hematopoietic stem cell transplantation, enzyme replacement therapy, substrate reduction therapy, pharmacological chaperones, and gene therapy. The blood-brain barrier represents a challenge for the development of therapeutic agents for these disorders. In this sense, alternative routes of administration (e.g. intrathecal or intracerebroventricular) have been evaluated, as well as the design of fusion peptides that allow the protein transport from the brain capillaries to the central nervous system. In this review, we outline the current knowledge about clinical and physiopathological findings of GM2 gangliosidosis, as well as the ongoing proposals to overcome some limitations of the traditional alternatives by using novel strategies such as molecular Trojan horses or advanced tools of genome editing. of -1,4-N-acetylgalactosaminyl synthase), which the residue to the GM3 structure GM1 and GD1a gangliosides differ from GM2 by the number and type of monosaccharides presents, as well as by the number of sialic acid residues. to untreated animals suggesting a slight therapeutic effect on the CNS. Overall, these results show that although this strategy must be optimized, the knock-in of HexM gene into hepatocytes is an alternative to reach supraphysiological levels of the enzyme, particularly in the brain, where typical strategies such as ERT fails due to the inability to cross the


Introduction
Gangliosides are a group of glycosphingolipids mainly located in the neuronal cell membrane and that are responsible of several pivotal biological functions for the correct functioning of the central nervous system (CNS) [1]. About 5% of all gangliosides into the brain correspond to GM2 gangliosides [2,3]. In normal conditions, the GM2 gangliosides are catabolized by the lysosomal hydrolases known as β-hexosaminidases (Hex, EC 3.2.1.52) through the hydrolysis of the Nacetylgalactosamine residues present on the structure of the GM2 ganglioside [4]. Hex are a subset of isozymes formed by the dimerization of the α and β subunits as follows: HexA (αβ), HexB (ββ) and HexS (αα). In addition, GM2 gangliosides degradation involves the GM2 activator protein (GM2-AP), which present the gangliosides to α subunit of HexA [1]. Mutations in the genes encoding for the α (HEXA), β (HEXB) or GM2-AP (GM2A) proteins, promote an impaired lysosomal degradation of the GM2 ganglioside, as well as other glycolipids, causing their accumulation into the lysosome [4].  [33,37]. New vesicles are sorted to both early endosomes (EE) and to the secretory pathway, where can be uptake by neighbour cells through M6PR [37,44]. Hexosaminidases are dissociated from the M6PR in the EE; which allows the M6PR recycling to the Trans-Golgi network by both clathrin-dependent and independent mechanisms [37]. On the other side, gangliosides are placed in caveolae-rich microdomains (CvRM), and in the turnover of the plasma membrane undergo caveolae-mediated endocytosis (CME) [1,38]. New caveosomes containing gangliosides (CCV-GM) reach the EE and further fusion events result in a late endosome, which can be fused with the lysosome to give rise to the endo-lysosome (EL, pH: 4.5) [37]. Gangliosides degradation starts with the hydrolysis of the galactose of the GM1 ganglioside to generate GM2 ganglioside which are harboured on intralysosomal vesicles (ILV) [1]. GM2 interacts with HexA through a GM2AP-mediated mechanism to removes the N-acetylgalactosamine resides [38]. Additional reactions implied in the ganglioside degradation to glucosylceramide (GluCer) are shown. The enzymes of each reaction are as follow: 1 and 4: β-Galactosidase/GM2AP, 2: β-Hexosaminidase A/GM2AP, and 3: Neuraminidase. LacCer: Lactoceramide. ECC: Extracellular compartment. ICC: Intracellular compartment.

Mutations of β-Hexosaminidases A and B, and GM2 activator protein.
Several mutations in HEXA, HEXB, and GM2A genes lead to the development of TSD (OMIM #272800), SD (OMIM #268800), or GM2-activator protein deficiency (AB variant; OMIM #272750), respectively [45][46][47]. According to The Human Gene Mutation Database, currently, 181, 103, and 9 mutations have been reported for HEXA, HEXB and GM2A genes, respectively, including missense/nonsense, splicing, small deletion and indels, and gross deletions ( Figure 3) [48 -50]. Although the type and frequency of mutations have been linked to the demographic origin of the patients, for HEXA the most representative mutation is the transition c.533G>A that changes arginine by histidine (p.R178H) and affects the catalytic site of the  subunit and alters its function and stability [51][52][53][54]. In the case of HEXB; a frequent mutation is c.445+1G>A, which occurs in a conserved intronic site that promotes a complete loss of a canonical splice donor site [55,56]. Finally, mutations in GM2A are extremely rare and only 9 mutations have been described on 11 patients [46,50]. Figure 3 show some of the most common mutations on HEXA, HEXB, and GM2A genes.   TSD: Tay-Sachs disease SD: Sandhoff disease. *Note: β-Hexosaminidases can hydrolyse molecules with N-acetyl-hexosamines residues such as glycosaminoglycans (GAGs). Accumulation of partially degraded GAGs has also been reported. For more detail please see [62,63].

Clinical presentations and biochemical correlations of GM2 gangliosidosis
Although TSD, SD, and AB variant are a consequence of mutations in different genes, neurological compromise is similar among these three diseases [1]. Classical findings are associated to the clinical onset as follow: acute: seizures, hypotonia, regression in developmental milestones [60], subacute: motor regression, psychotic episodes, intellectual disability [52] and chronic: dysphagia, cerebellar ataxia, muscle weakness and manic depression [6]. In Table 1 we summarized these neurological findings. In addition, individuals with SD, unlike TSD patients, can present systemic manifestations including cardiomegaly, hepatosplenomegaly, macroglossia, and skeletal abnormalities [7,8].

Diagnosis of GM2 gangliosidosis.
The diagnostic approach of GM2 gangliosidosis patients begins with recognition of the clinical characteristics of these entities (Table 1) [7][8][9]61]. Diagnosis is also favored by neuroimaging characterized by hyperdensity of basal ganglia, which can be accompanied by other changes in white matter and sometimes prominent, but non-specific, cerebellar atrophy [9].
The specific diagnosis requires the enzymatic determination of both HexA and HexB isoenzymes by using artificial substrates. For instance, the use of 4-methlyumbelliferyl-Nacetylglucosaminide (MUG) allows to measure the activity of both HexA and HexB isoenzymes, by taking advantage of HexA thermostability, while to specifically measure HexA activity a sulfated substrate (MUGS) is required [9]. It is important to note that while the test can be performed on plasma, amniotic liquid, and dried blood spots (DBS) [10], the diagnosis gold standard is the assay of the catalytic activity in leukocytes, fibroblasts, or chorionic villi in the case of prenatal diagnosis [10]. Enzymatic determination in GM2 gangliosidosis has some limitations, as it does not allow identification of asymptomatic carriers and the diagnosis of patients with deficiency on the GM2 activator protein, who require molecular verification of the gene defect [9]. Molecular diagnosis by sequencing the HEXA, HEXB, and GM2A genes allows to confirm the diagnosis of all GM2 gangliosidosis subtypes.
The heterogeneity of onset correlates inversely with the residual catabolic activity of Hex [1,64]. Whereas patients with the acute presentation have absent or very low (< 5%) enzyme activity, patients with subacute or chronic onset may have enzyme activities between 5 and 10% [4,65,66]. In this sense, it has been proposed that a 10% of wild-type enzyme activity may avoid the disease, which has been based on early studies describing GM2 ganglioside degradation with activities between 10 and, 15% [67], and pseudo-deficiencies reported in healthy individuals [1,4]. Nevertheless, some pathogenic mutations in both TSD and SD may lead enzyme activities around 15% of wild type levels [57,59]. In this sense, it has been proposed that a therapeutic benefit of any of the treatment alternatives, developed for GM2 gangliosidosis, may be obtained with enzyme activities higher than 15% of the wild type activity. Recent findings in this interesting topic will be discussed in detail later.

Physiopathology of GM2 Gangliosidosis
As a consequence of mutations on subunits of Hex or GM2 activator protein described above, GM2 gangliosides are accumulated into lysosomes [1,4]. Although it is not completely understood yet, early evidence using animal models of SD showed the presence of autoantibodies against GM2 gangliosides in serum and CNS [68], suggesting that its accumulation could promote the disruption of the lysosome and the release of GM2 ganglioside, which has been also described for the mucopolysaccharidosis (MPS) [69]. Since the lysosome has a pivotal function in the cell for the degradation of most macromolecules as well as organelles through autophagy [70][71][72], it is just obvious though that its impaired function must affect the general cellular functions with an impact on the global tissue physiology. Here we described current findings, which are summarized in the  . Physiopathological events in the GM2 gangliosidosis. Innate (IIR) and adaptative (AIR) immune response have been described in the GM2 gangliosidosis. Upon the impaired of lysosomal degradation of the GM2 ganglioside, this can be released into the cytoplasm (a) where its sialic acid can interact with the rough endoplasmic reticulum (RER) (b). Sustained stress into RER induces the activation of proapoptotic proteins like CHOP (c) that promotes mitochondrial-mediated apoptosis (d) In early apoptosis states, the phosphatidylserine (PS) is externalized (e), which promotes the recruitment of several proinflammatory cells such as microglia (f). Astrogliosis has also been reported in the GM2 gangliosidosis (g). The release of several astrocyte-derived cytokines (CK), such as CCL2 and CXCL10, increases the recruitment of microglia and the apoptosis in myelinating oligodendrocytes (AOD), which induces neuron demyelination (h). Finally, auto antibodies (Ab) against GM2 ganglioside seem to contribute to the physiopathology of the GM2 gangliosidosis (i), although the precise mechanism of its release has not described. ECC: Extracellular compartment, ICC: Intracellular compartment, AP: Apoptosome, OD: Oligodendrocyte.

Neurodevelopment Process.
The use of a cerebral organoid model of SD has been an interesting approximation to discover novel implications of GM2 accumulation [73]. In this context, recent findings have described that impaired Hex activity promotes an increase in the size of the cerebral organoids generated from patient-derived iPS cells, which is corrected after transduction with adeno-associated virus (AAV) carrying the HEXA and HEXB cDNAs [74]. In concordance to the above and with sophisticated assays using HEXB deficient zebrafish embryos, Kuil et al., 2019 found an increase in the lysosomal speckles in radial glia and a discrete decrease in the microglia, accompanied by abnormal locomotor activity in larvae 5 days postfertilization [75]. Nevertheless, an increase in the apoptosis rate was not reported in these studies [74,75]. Together, these results are evidence of the cellular and functional consequences of GM2 ganglioside accumulation in the maturing nervous system, and support the fact that these early processes could have an impact on the acute symptoms of the SD, that has been previously identified [76].

Neural Death and Neuroinflammation.
Neural death has been proposed as an important mechanism in the physiopathology of the GM2 gangliosidosis [4]. Early studies of brain and spinal-cord from autopsy samples of TSD and SD patients revealed an increase in the in-situ DNA end-labeling, suggesting that apoptosis could contribute to the neurodegenerative process in these patients [77]. Similar findings were later described in SD animal models [78] and TSD mouse model (HEXA -/-/NEU -/-) [12], showing a marked reduction in neuronal density [79]. Although the precise pathway that explains the increase of neural apoptosis has not been completely resolved, recently findings have shown that GM2 ganglioside can induce endoplasmic reticulum stress [5,80]. In addition, it has been proposed that GM2 ganglioside, but not asialo-GM2 ganglioside, promotes neurite atrophy and cell dead trough PERK-mediated apoptosis with downstream CHOP activation [5], which is an inductor of mitochondrial apoptosis [81]. These results suggest that the sialic acid on the GM2 ganglioside may have proapoptotic properties. Nevertheless, the use of a PERK inhibitor did not completely abolish the apoptosis, suggesting that further mechanisms in the neuronal death in GM2 gangliosidosis could be involved [5].
Given the typical mechanism of early apoptosis such as phosphatidylserine externalization on the surface of the neuron [82,83], it has been suggested that the characteristic microgliosis and infiltration observed in patients and animal models, could be the response to the neuronal death [4,78,84]. Early studies conducted by Jeyakumar et al., 2003 using a SD (HEXB -/-) mouse model, showed that as the disease progress and symptoms such as head tremor, motor disfunction, and hind limb paralysis appear, there is an increase in the immunoreactivity for the Mayor Histocompatibility Complex type II in brain stem and thalamus, as well as an increase in the proinflammatory cytokines TNFα and IL1β [85]. These findings suggest that innate immune activation of the CNS could be produced, at least in part, in response to the neuronal death as was previously proposed. Although authors reported similar findings on the TSD (HEXA -/-) mouse model, cytokines levels were not increased, which could be associated to the bypass of neuraminidase allowing the GM2 ganglioside degradation, and avoiding the typical neurological manifestations [11,12].
Despite microglial activation and proliferation have been extensively reported in models of both TSD and SD [84,86], astrogliosis plays also an important role in GM2 gangliosidosis even in asymptomatic states of the disease [87,88]. Astrocytes are complex cells localized in gray and white matter, blood vessels, and wrapping the synaptic space in quiescent, active or reactive states [89]. In this regard, Ogawa et al., 2017 showed that astrocytes are activated by an FcRγ-dependent mechanism [88]. Upon activation, astrocyte secretes the chemokines CCL2 and CXCL10 which is consistent with a robust microglial activation as wells as an invasion of peripheral immune cells. It could allow the degeneration of myelinating oligodendrocytes and its death, promoting active demyelination [89,90], which is frequently observed in infantile forms of GM2 gangliosidosis [4,91,92]. Together, these findings highlight the pivotal interplay between microglia and astrocytes in the inflammatory response observed in the GM2 gangliosidosis, which could be the functional consequence of neuronal injury due to GM2 ganglioside accumulation contributing to the neurodegenerative process.

Current proposals for the treatment of GM2 Gangliosidosis
Several approaches have been tested for the development of specific treatments for GM2 gangliosidosis, which are summarize in Figure 5 [93][94][95]. These strategies range from traditional enzyme replacement therapy alternatives to novel biotechnological tools such as CRISPR/Cas9 and prime editing, with promising results in both in vitro and in vivo models. Although some of these developments have been translated to clinical trials [96][97][98][99], there is not an approved and effective Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 7 July 2020 doi:10.20944/preprints202007.0137.v1 treatment for TSD, SD, or AB variant yet. In this section, we review the current proposals and advances in therapeutics for GM2 gangliosidosis.

Enzyme replacement therapy.
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 7 July 2020 doi:10.20944/preprints202007.0137.v1 Enzyme replacement therapy (ERT) is a therapeutic alternative conceived in 1964 by Christian de Duve in which the lysosomal enzymes can be uptake through endocytosis and end up into lysosomes [14]. The development of ERT for LSDs was boosted by the description of the crosscorrection mechanism in pivotal experiments using co-culture of fibroblast from Hurler (MPS I) and Hunter (MPS II) diseases [4,100]. Currently, it is well known that some lysosomal enzymes, such as Hex, have M6P residues at the terminal end of the N-glycosylations, which allows their cellular uptake through a M6PR-dependent manner [4,33,93]. ERT has been approved for several LSDs including Gaucher, Fabry, and Pompe diseases, late infantile neuronal ceroid lipofuscinosis type II, acid lipase deficiency, alpha-mannosidosis, and MPS type I, II, IVA, VI, and VII [93,101].
In the case of GM2 gangliosidosis, early studies carried out by Johnson et al., 1973 using intravenous administration of HexA in a patient with SD showed that the enzyme was presented in liver but not in the cerebrospinal fluid or brain parenchyma, suggesting that enzyme was unable to cross the BBB [102], due to the absence of M6P receptors in endothelial cells [14]. To overcome these limitations, promising strategies to enable that exogenous lysosomal enzymes can cross the BBB include the development of fusion proteins, also known as molecular Trojan horses [14]. These fused proteins are recombinant chimeric enzymes fused to a monoclonal antibody (MAb) that recognize either the human insulin receptor (HIR) or the transferrin receptor (TfR), and which allow the passage of the BBB through a receptor-mediated endocytosis [103]. In this field, several studies using nonhuman primates have shown that intravenous administration of HIRMAb fused to different lysosomal enzymes as α-iduronidase [104], iduronate-2-sulphatase [105], sulphamidase [106], α-Nacetylglucosaminidase [107] can cross the BBB without side effects. Although these approaches have been evaluated in animal models of GM2 gangliosidosis, recently Boado et al., 2019 showed that an HexA fused to HIRMAb has a similar activity to a non-fused enzyme (2.464 ±109 mU/mg vs 2.557 ±187 mU/mg, respectively), suggesting that use of molecular Trojan horses could be an alternative to treat GM2 gangliosidosis [103].
The use of several routes like direct injection in cerebrospinal fluid and intrathecal or intracerebroventricular (ICV) injections have shown potential therapeutic effects for some LSDs with CNS compromise [14,108]. In this sense, the most common administration route tested in GM2 gangliosidosis has been the ICV. Using this approach, in early studies conducted by Matsuoka et al. in 2010, it was compared the therapeutic effect of two HexA: wild type (WTHexA) and a modified HexA with an additional N-glycan (NgHexA) on SD mice. The study showed that SD mice treated with NgHexA presented an increase of 1.5-and 2.5-fold on the activity levels in brain and cerebellum after 24 h post-injection, respectively. In addition, all brain regions showed a significant increase on the Hex activity 6 days post-injection and the reduction of GM2 gangliosides into the brain of NgHexA (32%) and WTHexA (7%) SD treated mice [109]. Together, these results suggest that the addition of M6P-type-N-glycan into recombinant HexA may have a positive effect on the enzyme uptake, biodistribution, and reduction of substrates throughout the brain. Similarly, Tsuji et al., 2011 found that by increasing the M6P residues on the N-glycans of a recombinant HexA produced in the yeast Ogataea minuta (Om4HexA), improved the therapeutic effect of the modified enzyme on HexB -/-mice compared to the unmodified HexA enzyme (OmHexA), after ICV administration [110]. Treatment with Om4HexA led a higher reduction of accumulated GM2 and GA2, compared with OmHexA treated mice. Interestingly, the authors reported a decrease of MIP-1α chemokine in the hindbrain region (~60%) when animals were treated with Om4HexA. Also, a single administration of both enzymes OmHexA and Om4HexA increase the lifespan by 7.8% and 12.9%, respectively [110].
Meanwhile, in novel experiments, Matsuoka et al., in 2011 designed a chimeric HexB enzyme that contains the β-subunit with six point mutations (p.R312G, p.Q313S, p.N314E, p.K315P, p.D452N, and p.L453R) and a partial sequence from α-subunit, which allow to this enzyme to bind to charged substrates as well as to GM2AP. Using this approach, the authors found that ICV administration of the chimeric enzyme on SD mice allowed a significant reduction of GM2 ganglioside in the brain (77%) and cerebellum (57%), compared to untreated SD mice. Likewise, a 2-fold increase in Hex activity and a marked reduction of GM2 ganglioside was observed in liver; which could have a significant impact on the disease [111], since, unlike to TSD, hepatosplenomegaly can be found in SD patients [92].
Although most of the enzymes used for ERT are produced in Chinese hamster ovary cells (CHO), since they can produce proteins with similar human N-glycosylation patterns [101], several alternatives like O. minuta and Pichia pastoris have been evaluated [110,112,113]. In this sense, recombinant HexA and HexB produced on the methylotrophic yeast Pichia pastoris GS115 have shown a high stability on a wide pH range and in human serum [112]. Recently, it was found that these enzymes can be internalized by both HEK293 cells and healthy fibroblasts in an M6PRdependent mechanism without further modifications to expose M6P residues [114], as was previously reported for the same enzyme but using the yeast O. minuta [110]. Recombinant HexA produced in P. pastoris normalized lipid accumulation in neural stem cells derived from TSD iPSCs, showing the potential of this enzyme, and P. pastoris, in the development of an ERT for GM2 gangliosidosis [115]. Recent efforts to improve the delivery of this recombinant Hex to BBB, involved their conjugation with nanoparticles [116].

Hematopoietic Stem Cell Transplantation.
Since Hex have N-glycans M6P residues [33] and they can be exported to the extracellular space [37, 44], a cross correction could be possible through of its capture by neighbor cells in an M6PRmediated mechanism [4,117]. As a consequence, the administration of a cells source with pluripotential and self-renewal capacities, like hematopoietic stem cell (HSC), could provide sufficient amounts of the deficient enzyme in a natural o engineered-dependent manner [117,118]. Allogenic HSC transplantation (HSCT) can be performed from bone marrow (BM), umbilical cordon (UC), or peripheral blood after a myeloablative regimen [119]. This strategy has been performed in other LSDs, such as MPS I [120,121], MPS II [122], and Gaucher type 1 and 2 [123].
For GM2 gangliosidosis few approaches have been performed and a limited number of patients have been subject to HSCT. In this regard, Jacobs et al., 2005, used allogeneic BM transplantation to treat a 3 years-old asymptomatic child with subacute TSD. The results showed an increase in leukocytes HexA activity but without prevention of neurodegenerative events of the disease [124]. Later, in a single-center study, five children with infantile TSD were enrolled and subjected to unrelated UC transplantation. After treatment, the survival was extended in 2 cases with an arrest on the neurodegenerative process but without improvement in motor skills [125].
Meanwhile, Ornaghi et al., 2020 showed that the transduction of HSC isolated of healthy donors with bicistronic lentiviral vectors carrying both -and -subunits allowed the increase of up to 2-fold in total Hex activity [126]. Similar findings were reported for the enzyme activity on neural stem cells and murine HSC transduced with the LV; whereas physiological functions of the stem cell such as proliferation, self-renewal, or multipotency, remain unchanged, suggesting that ex vivo gene therapy could be an interesting option for the treatment of GM2 gangliosidosis. Despite the human leukocyte antigen (HLA) compatibility is frequently realized at immunogenetics laboratories as a crucial prerequisite for HSCT [127,128], a common challenge in allogeneic HSCT is the graft-versus-host disease (GVHD) [127]. In fact, studies in patients with inborn errors of metabolic have shown up to 10% of acute GVHD [125]. In order to overcome these limitations, recent developments using hypoimmunogenic human stem cells have opened a new horizon for the HSCT by avoiding the events of GVHD through evasion of both cellular and humoral immune response [129,130]. However, this approach has not been applied to GM2 gangliosidosis yet. Likewise, an attractive alternative for the treatment of GM2 gangliosidosis could be the use of autologous HSCT, in which the mutations are corrected thorough different gene therapy approaches in hematopoietic precursors isolated from the patients [118,131].

Pharmacological chaperones.
Pharmacological chaperones (PC) are small molecules that can influence the correct folding and assembly of native proteins with abnormal configurations codified by mutated genes, improving their physical stability as well as intracellular traffic [132]. Misfolded proteins are subject to Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 7 July 2020 doi:10.20944/preprints202007.0137.v1 premature degradation, through a retrotranslocation from ER to the cytoplasm, after ubiquitination in the N-terminal, to be finally degraded into the proteasome [133]. As a consequence, the protein cannot reach its final cellular location, i.e. lysosome, leading to a pathological condition. PC bind with high affinity and selectivity to a misfolded protein in the neutral pH of the ER and promotes the correct folding of the protein [134,135]. The PC-protein interaction is dissociated into the lysosome as a consequence of the acidic pH, as well as by presence of the natural substrate, which competes with the PC for the active site [134,135]. Since PC may act as competitive inhibitors [132], it has been proposed the use of allosteric sites to identify non-inhibitory PC [134]. In addition, PCs have a mutation-dependent activity, which limits the number of patients that may respond to the treatment [134,135]. PC must not be confused with chemical chaperones like dimethyl sulfoxide (DMSO), which can also bind to and stabilize some proteins but with less selectivity and greater toxicity [136].
In a screening of 1040 FDA-approved drugs, Maegawa et al., 2007 found that pyrimethamine (PYR) was the most promising PC for HexA with an IC50 between 5 and 13 μM at pH 4.3 [137]. PYR is a typical drug used in the treatment of cerebral toxoplasmosis and complicated malaria, which targets folic acid synthesis and can cross the BBB [138]. PYR is a competitive inhibitor of HexA (Ki: 13 μM at pH 4.5) that induced an up to 3-fold increase in enzyme activity on TSD fibroblasts [137]. Effectiveness of PYR as a PC depends on the binding to the active site, which is mediated by hydrogen bonds and Van der Waals forces [139]. Consequently, mutations that affect the catalytic active site would not allow the binding of the PC. PYR was evaluated in a phase I/II clinical trial, in which eight late-onset patients were orally treated with scaling doses of PYR up to a maximum dose of 100 mg per day [98]. HexA activity was increased in leukocytes up to 4-fold, compared to baseline levels, with doses lower than 50 mg/day [98]. Patients experienced side effects such as hypersensitivity with doses higher than 75 mg/day; and neurological side effects, which in one case led to seizures after the drug was increased from 50 to 70 mg/day. Osher et al., 2015, in an extended pilot study, evaluated the effect of low doses of PYR (~2.7mg) in four late-onset TSD patients, in which an increase of 2.24fold on the HexA activity was achieved. Although these levels were gradually reduced during the continuous administration of PYR; upon the instauration of cyclic doses of PYR (~33 weeks) the HexA activity was enhanced again [140]. Nevertheless, only one patient remained stable whereas three individuals showed neurological deterioration.
To identify potential PCs for GM2 gangliosidosis, an in silico analysis of iminosugar inhibitors that bind to the active site of HexA was carried out [141]. Throughout molecular docking and dynamics simulations, the pyrrolidine DMDP amide was identified as the strongest competitive inhibitor of HexA. Using TSD fibroblasts patients, DMDP amide improved the intracellular activity of HexA up to 14.8-fold compared to untreated cells in a dose-dependent manner, reaching up to 43% of wild-type levels [141]. These results showed that iminosugars could be an interesting therapeutic alternative for GM2 gangliosidosis, as has been previously described for other LSDs with a neurological compromise like Gaucher and Fabry diseases [142,143].
Another potential strategy to increase the folding of mutant Hex is the use of progranulin (PGRN), which is a glycoprotein that is secreted by epithelial, neuronal, and immune cells, and is involved in a variety of physiological processes and diseases such as early embryogenesis, cell proliferation, immune and neurodegenerative diseases, inflammation processes, wound healing and tissue repair [144,145]. Jian et al., 2016 showed that the heat shock protein 70 (HSP70) associates with the lysosomal enzyme β-glucocerebrosidase (GCase) in the ER/Golgi apparatus by a PGRNdependent manner, avoiding the aggregation of the GCase [145]. Since PGRN could be considered as a co-chaperone, it was evaluated to increase the HexA activity in fibroblasts from TSD patients. The results showed that the G and E domains of the PGRN bind to HexA, increase its enzyme activity and lysosomal delivery, and promotes the GM2 reduction in TSD fibroblasts [144].

Substrate reduction therapy.
Substrate reduction therapy (SRT) is a therapeutic strategy based on the partial inhibition of enzyme involves in the synthesis of the accumulating substrate [146]. One of the molecules that have been evaluated for this purpose is N-butyldeoxynojirimycin (NB-DNJ, also termed Miglustat or Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 7 July 2020 doi:10.20944/preprints202007.0137.v1 Zavesca) [147], an iminosugar that inhibits the glucosylceramide synthase (GCS), which catabolize the first step of glycosphingolipid synthesis like glucosylceramide (accumulated in Gaucher disease [148], sphingomyelin (accumulated Niemann-Pick type C) [149], GM1 (accumulated in Gangliosidosis GM1) [150] and GM2 gangliosides (GM2 gangliosidosis) [151]. Miglustat is an oral drug able to cross the BBB, slowing the accumulation of gangliosides in neurons, and delaying the progression of neurological symptoms [146]. Therefore, miglustat has been evaluated in murine models of SD [152] and TSD [153] with promissory results regarding ganglioside storage reduction, the decrease of neurological symptoms, and the extension of life span. Subsequent studies that evaluated the effect of miglustat in five patients with juvenile GM2 gangliosidosis in advanced disease stage over a period of 24 months, however, they did not show improvement on the neurological impairment [154]. Likewise, Masciullo et al., 2010 showed that a 3-years treatment with miglustat on an adult with chronic SD, did not arrest the neurodegeneration [155]. In this sense, implementation of miglustat in early disease stages should be assessed, to know the therapy efficacy during early intervention. Under the assumption that the administered dose of miglustat was one of the possible causes of its poor satisfactory results, Ashe et al., 2011 evaluated the molecule Genz-529468, which has an IC50 250-fold greater than miglustat [156]. Genz-529468 was administrated to SD mice and the results were compared with miglustat-treated mice. The effect of each drug on brain GM2 levels was opposite. In this sense, whereas Genz-529468 increased GM2 levels to about 120% of untreated mice; miglustat decreased the GM2 levels to about 90% of untreated mice. However, levels of other brain glycosphingolipids, specially GL1, were dramatically increased with both inhibitors. The impact of this increase was not evaluated. The authors also reported similar results for both drugs in terms of the delayed loss of motor function and extended of the lifespan. In addition, mice treated with miglustat and Genz-529468 showed lower microglia activation and astrogliosis, and delay of neuronal apoptosis. These findings suggest that the use of inhibitors of GCS could improve the clinical outcomes due, at least in part, to their an anti-inflammatory properties [4,157].
Finally, Arthur et al 2013 reported the evaluation of EtDO-PIP2, a GCS inhibitor, on juvenile SD mice [158]. A previous study showed that this drug reaches the CNS and reduce glucosylceramide in wild type mice [159]. The intraperitoneal administration of EtDO-PIP2 to SD mice reduced the total content of brain and liver gangliosides, suggesting that it could be an interesting alternative for the treatment of ganglioside storage diseases with CNS manifestations.

Gene therapy.
Since GM2 gangliosidosis is characterized by mutations in HEXA, HEXB, or GM2A genes [4]. As a consequence, delivering a functional gene should correct the genetic defect and lead to a normal physiological development [160]. In recent years, several strategies based on gene therapy have been developed for LSDs, including GM2 gangliosidosis [160][161][162][163], which will be discussed below.
Vectors for the delivery of therapeutic genes have been a wide research topic. Usually, gene therapy uses recombinant viral vectors to deliver transgenes into specific tissues, from which the adeno-associated virus (AAV)-derived vectors have gained great attention during the last years [4]. AAV are small icosahedral virus that contain a single-strand DNA [164]. Unlike retrovirus and lentivirus, AAV vectors remains as episomal DNA structures avoiding insertional mutagenesis [165,166]. To the date, there are several AAV serotypes with tropisms that can change between hosts [164]. Although several serotypes of AAV can be used, AAV8 has shown tropism to CNS with promising evidence for the delivery of Hex in SD cats [167]. Since its biochemical properties mostly of the capsid, AAV can induce an immune response in the host [168]. Therefore, new viral vectors have been engineered to modify the viral capsid and reduce the immune response that can limit the efficacy of the therapy [169,170].
Consistent with the above, several papers describe a variety of studies of gene therapy applied to the brain of small and larger animal models of GM2 gangliosidosis, mostly using AAV vectors [171]. In this regard, Cachón et al., 2012 evaluated the effect of the intracranial coadministration of AAV vectors carrying the human  and β subunits of HexA on a SD mouse model. The results showed a wide distribution of the hexosaminidases throughout the perivascular space into the brain and spinal cord of the SD mice, as well as the rescue of the classical symptoms of GM2 gangliosidosis up to 2 years post-treatment [172]. Similarly, Bradbury et al., 2013 using SD cats showed that a single bilateral thalamic injection of an AAVrh8 vector expressing feline Hex subunits (α and/or β), led to a 2-fold lifespan increase [173]. The treatment shows improvement on motor functions compared to untreated cats as well as a slight immune response against AAVrh8; contrary to the strong humoral and cellular immune response that was observed when AAV1 and cDNA of human subunits were used [173]. The cats who received the AAVrh8 carrying the feline Hex subunits, did not show clinical or histopathological features of toxicity. Together, these results highlighted the importance of speciesspecific cDNAs and support the effectiveness of gene therapy in a SD validated model.
In 2015, McCurdy et al. also reported the successful translation of a therapeutic approach from GM1 mice [174] to GM2 presymptomatic cats. They evaluated the therapeutic effect of bilateral injections in the thalamus and deep cerebellar nuclei (DCN) of AAV vectors expressing feline Hex subunits (α and β), to treat naturally occurring feline GM2 gangliosidosis. The authors observed an increase on Hex activity up to 19-fold in CNS compared to untreated SD cats after 16 weeks posttreatment. Likewise, clearance of GM2 ganglioside was reported in the brain (89-99%) and cervical intumescence of the spinal cord (72%). Improvement on clinical features, such as reduced ataxia and subtle tremors, were observed in the GM2-treated cats too [175].
Bilateral injection into the thalamus alone or in combination with ICV injection of AAVrh8 vectors encoding feline Hex subunits (α and β) has been tested in SD cats. Using these approaches, reduction of GM2 ganglioside was observed in the cortex (96%), cerebellum (89%), thalamus (87%) and cervical spinal cord (95%), as well as a significant increase in the Hex activity (3.7-56.6-fold) [176]. Moreover, treated SD cats showed an increase on the myelin-enriched cerebrosides and a decrease in the microglial activation, suggesting an attenuation of the neuroinflammation after the treatment [84,176] and supporting the amelioration of an important physiopathology feature of GM2 patients [78,85,89]. Similar findings were reported by Gray-Edwards et al., 2018 when used AAVrh8 vectors carrying the α and/or β subunits to evaluate the therapeutic effect of an intracranial injection in a TSD sheep model, which mimics the late-onset infantile phenotype [177,178]. The TSD sheep that received both subunits, achieved wild type or even supraphysiological levels of HexA activity. Decrease in disease progress was observed in all treated sheep. Also, reduction of accumulated gangliosides to normal levels was reached in most areas of the brain, except in cerebellum and spinal cord where the storage of gangliosides did not change. Neuroinflammation was also attenuated after treatment [86].
To extend the promising results described above to larger brain animals, as an important step towards human clinical trials, Golebiowski et al. used intracranial injection to administer, to unaffected non-human primates, AAVrh8 vectors encoding the cynomolgus macaques Hex subunits (α and β). Supraphysiological levels of Hex subunits were achieved in CNS. Most of the animals subjected to treatment showed altered neurological function reflected in ataxia, general weakness, and lethargy as well as histopathological findings suggestive of neurotoxicity associated with eosinophilic inclusions in neurons, white matter loss, and necrotic areas in the thalamus [179]. These findings suggest different responses not only against AAVrh8 but also against supraphysiological levels of Hex in non-human primates, which contrasts to the results observed in other species such as feline or sheep, where slight or absent toxicity has been reported [86,173]. These aspects could be better resolved with similar preclinical testing but using validated non-human primates' models for GM2 gangliosidosis to know the real therapeutic effect on species close to human.
Despite the interesting findings described with AAV vectors, its limited DNA packaging capacity (~4.6 Kb) has suppose the need to develop new AAV-therapeutic approaches that avoid the use of multi-vector injections. In addition, it is also important to develop novel approaches that allows the correction of CNS impairment. In this sense, Tropak et al. designed a self-complementary AAV9.47 vector encoding a hybrid µ subunit (scAAV9.47-HEXM) [180], which combine the αsubunit active site, the stable β-subunit interface, and unique areas of each subunit necessary for the interaction with the GM2AP [41]. Using both in vitro and in vivo models of TSD and SD, it was found that HexM was able to catabolize the GM2 ganglioside in a GM2AP-dependent manner with its Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 7 July 2020 doi:10.20944/preprints202007.0137.v1 subsequent reduction in thalamus, hypothalamus, and hippocampus [180]. This effect correlated with an improvement on the behavioral tests, which reached similar values to that of heterozygotes animals. Finally, novel approaches using bicistronic vectors to reach high levels of Hex in the correct stoichiometric rate have been developed [181]. With this strategy, a proof-of-concept to evaluate the therapeutic potential of the bicistronic ssAAV9-HexBP2A-HexA vector, which has a short P2A linker, the cDNA of HEXA and HEXB genes under the control of a chicken beta-actin promoter [182]. A single administration of the vector into neonatal SD mice allowed a 56% extension of the lifespan compared with untreated animals. In addition, higher enzyme activity values and lower levels of GM2 gangliosides were obtained in brain and serum of treated SD mice compared with untreated mice [182]. Similar observations have been reported for novel bicistronic lentiviral vectors, carrying murine or human cDNAs encoding for the -and -subunits linked by a P2A self-cleaving peptide. This bicistronic lentiviral vectors allowed an increase of Hex activity up to 5-fold of the normal values in murine neurons and human stem progenitor cells, as well as a 30-60% reduction of GM2 storage. Similarly, in treated SD fibroblasts the total Hex and HexA activity increased in a 8:1 ratio, respectively [126].

CRISPR/Cas9-based gene therapy
Genome editing is a promising strategy for correcting mutations that cause a pathology [161,183,184]. Three major genome editing tools have been developed: zinc finger nucleases (ZFN), transcription activator-like effector nucleases (TALEN), and more recently, clustered regularly interspaced short palindrome repeats/Cas9 (CRISPR/Cas9) [183,185]. CRISPR/Cas9 uses an RNAguide nuclease (sgRNA-Cas9) to induce double-strand breaks (DSB) into a specific locus of the genome [186,187]. DSB can be repaired either through non-homologous end joining (NHEJ) or homologous direct repair (HDR) pathways [15]. On the first case, upon the DSB and on absence of a DNA template, the cellular repair machinery recruits several effectors such as Ku70/80, DNA-PKcs, Artemis, and DNA ligase IV that promote the binding of non-homologous ends in an error-prone mechanism. Through this mechanism is possible to induce deletions or insertions leading to insertional inactivation [188]. This approach is used to knock-out target genes, and has been used to generate in vitro and in vivo models of GM2 gangliosidosis [41, 74,75] DSB can be also repaired through the HDR pathway mechanism by using an exogenous DNA fragment as template (donor DNA), to mediates the insertion of the therapeutic DNA fragment on a specific locus ( Figure 6) [185,188,189]. Although the punctual correction of each mutation could be interesting, in pathologies with hundreds of mutations, like GM2 gangliosidosis, this approach is not suitable. As a consequence, the knock-in strategy using safe harbors to introduce the cDNA into the genome draws more attention [190,191]. Ou et al., 2020 recently used the albumin locus for the intravenous administration of two AAV8 vectors: one carrying a promoterless HexM cDNA, and other one carrying the Cas9 gene and the gRNA [192]. This approach allowed the constitutive expression of the HexM under the control of the albumin promoter. Four-months post-treatment, the enzyme activity increased in plasma, heart, liver, spleen, and brain, respect to untreated SD mice. This increase in the HexM activity positively correlated with a decrease of the GM2 ganglioside in liver, heart, and spleen. On the other hand, neither a decrease of GM2 ganglioside in the brain nor positive changes on behavioural tests (fear conditioning and pole test) were observed as a result of the treatment [192]. Despite the above, a significant improvement in the rotarod test, which measure motor function, was reported in SD mice treated respect to untreated animals suggesting a slight therapeutic effect on the CNS. Overall, these results show that although this strategy must be optimized, the knock-in of HexM gene into hepatocytes is an alternative to reach supraphysiological levels of the enzyme, particularly in the brain, where typical strategies such as ERT fails due to the inability to cross the BBB [13]. Figure 6. Approaches for genome editing using CRISPR/Cas9. The upper panel shows the classical strategy of knock-in using a ribonucleoprotein complex (sgRNA-guide Cas9) to guide to Cas9 to the target DNA and cut the double-strand (DSB). After the DSB, repair machinery is activated. In the presence of a donor sequence (HR template), homologous recombination is favored. To promote the recombination event of a gene of interest (GOI), the HR template must be flanked by homologous recombination arms which are complementary to the 5' and 3'-ends of the sequence into the gene that will be subject of edition. Typically, between 100-150bp and 400-800bp are suitable for small (< 50bp) and large (>100) insertions, respectively [193,194]. In the lower panel, Prime Editing (PE) is represented. PE uses a nickase Cas9 (nCas9-H840A) fused to reverse transcriptase (RT) and a guide RNA (pegRNA) which is engineered with a sequence in the 5'-end.20 nucleotides guide to nCas9 to the target DNA and a sequence in the 3-end with a primer-binding site (A) as well as an RT template (B) that could be between 7 to 12 nucleotides [195]. Upon reverse transcription, newly synthesized strand hybridizes to the unedited strand (US) forming a mismatch and a 5'-flap strand which is removed by exonucleases (EN) like EXO1 [196]. The mismatch is resolved with the introduction of a new nCas9 coupled to a simple sgRNA which guide to nCas9 to the edited strand (ES), about 50 bp from the pegRNA-mediated nick, to cut the US and use the sequence of the ES as a template for repair de simple cut [195]. In both cases upper and lower panels, newly edited DNA is successfully obtained with different efficiencies.
Despite HDR is the main approach used in CRISPR/Cas9-mediated genome editing, novel strategy without DBS or donor DNA, termed prime editing, was recently described [96,197]. This tool uses a nickase Cas9 (H840A) fused to reverse transcriptase (RT) and a short engineered RNA sequence (prime editing guide RNA-pegRNA) [195]. The pegRNA is designed to function as a guide of Cas9 to the target DNA and serve as the template for a RT-mediated retrotranscription, avoiding the need of a DNA donor [197,198]. An additional sgRNA-Cas9 is necessary to nicking the unedited strand and promotes the final correction using the edited strand as a template, in a process that occurs only after the edition and avoiding the generation of DSB [195,198] (Figure 6). Targeted insertions, deletions, and all 12 possible base to base conversions are feasible with this novel strategy [195]. (1278+TATC), which is associated with TSD with high efficiency (31%) and low indels (0.8%). Once mutations were generated, the cells were correct using the same strategy of prime editing. For this, 43 pegRNA and three sgRNA to induce nicking of the unedited strand were tested. Nineteen of all pegRNAs showed edition efficiencies higher than 20% and lower indels 0.32% [195], suggesting that this novel strategy could be a new therapeutic approach for the treatment of GM2 gangliosidosis without DSB or donor template.

Conclusions and Perspectives
During the last years significant advances have been done to understand the physiopathology and natural history of GM2 gangliosidosis. For instance, the natural history programs have allowed to identify the shared and specific manifestations of each disease (TSD and SD) and phenotype (infantile/acute, juvenile/subacute, and adult/chronic), which represent valuable information to improve the diagnosis and patients follow up. However, important efforts still need to be done in order to include a wider number of patients with different genetic backgrounds, since most of the studies have been carried out in specific populations (e.g. clinicaltrials.gov NCT01869270, NCT02851862, NCT00668187, NCT03333200, and NCT00029965).
On the other hand, although gangliosides remain as the main compounds responsible for neuron homeostasis alteration, neuroinflammation and demyelination have also shown to play an important role in disease progression. In addition, the presence of anti-ganglioside autoantibodies, progressive accumulation of α-synuclein, and impaired autophagy, are elements that have been also associated with the disease process [4]. In this sense, as observed in other LSDs [199], it is possible that a single therapeutic strategy would not be enough to treat GM2 gangliosidosis and that the co-administration of different therapies may be required. Noteworthy, different therapeutic strategies including ERT, gene therapy, pharmacological chaperones, HSCT, and substrate reduction therapy have shown promising results and some of them have reached clinical phases (Table 2). In addition, the National Tay-Sachs, and Allied Diseases (https://ntsad.org/) announced the beginning of the gene therapy clinical trials for GM2 gangliosidosis. Although the use of CRISPR/Cas9 is still on its initial stages, the first pre-clinical studies have shown the potential of this tool in the design of novel gene therapy strategies.
Finally, epigenetics should be also considered in GM2 gangliosidosis. Although the role of epigenetic mechanisms in lysosomal diseases has not been well stablished, it has been proposed that these mechanisms may contribute to the clinical heterogeneity observed in these disorders [200]. However, to the best of our knowledge, no studies on this field has been performed for GM2 gangliosidosis. The understanding of the epigenetic alterations observed in GM2 gangliosidosis patients may represent an opportunity to the develop of novel treatment alternatives [201].

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the review.