Advancements in AAV-mediated Gene Therapy for Pompe Disease

Pathophysiology of Pompe disease: A focus on neuropathology

Pompe disease (glycogen storage disease type II) results from mutations in the gene encoding acid-α-glucosidase (GAA) which is responsible for lysosomal glycogen degradation. Lack of GAA activity results in pathologic glycogen accumulation creating deposits that ultimately disrupt cardiac, skeletal muscle, and central nervous system architecture and function [1, 2]. The recent initiation of newborn screening for Pompe disease [3, 4] has raised estimates of prevalence to 1 in 8,700 [1]. Severity of the disease varies by age of onset and the rate of progression and is inversely correlated with residual GAA activity [5, 6]. Infantile-onset patients have GAA gene mutations causing absence of functional GAA. These patients are usually diagnosed during the first few months of life and if untreated, will soon die from cardiorespiratory failure [7]. Late-onset patients retain < 20% of normal GAA activity and show progressive motor dysfunction with respiratory failure being the most common cause of mortality [8, 9]. Severe respiratory insufficiency can occur without evidence for significant limb muscle weakness [10, 11]. This feature is unique in Pompe disease compared to muscular dystrophies in which loss of ambulation precedes ventilatory insufficiency.

To date, there is no cure for Pompe Disease. Enzyme replacement therapy (ERT) was approved by the FDA in 2006 and is the only available treatment. This therapy involves bi-weekly infusion of recombinant human GAA (Myozyme^®). Clinical trials show extended survival and some relief of the associated cardiomyopathy [12–18]. The ERT can also provide a transient, stabilizing impact on respiratory function, but many patients still progress to respiratory failure [12, 13, 19, 20]. A particularly salient point is that ERT cannot be expected to effectively treat any existing neuropathology since the recombinant protein does not cross the blood-brain-barrier [21, 22]. Moreover, by temporarily mitigating cardiac and/or skeletal muscle problems, ERT can actually unmask a neurologic phenotype, particularly in infantile-onset patients [23]. The limited effectiveness of ERT, combined with established neuropathology in animal models [24–28] as well as in patients after failed ERT [2, 27] led to the hypothesis that neuropathology can substantially contribute to motor dysfunction in Pompe disease. Much of the evidence for neural dysfunction in Pompe disease, including early reports of neuropathology in patients [29–32] was summarized by Fuller at al. in 2013 [2]. Since then, more than 20 new publications have provided further evidence for neuropathology and neural impairment in Pompe disease. These new studies are summarized in Table 1, and the reader is referred to earlier descriptions of central nervous system involvement in Pompe disease [2, 27]. Collectively, the literature indicates that new treatment approaches capable of simultaneously treating muscle pathology and neuropathology are warranted.

Lysosomes and autophagy in Pompe disease

Lysosomes are defined as membrane-bound organelles with acidic interior containing a collection of specific hydrolytic enzymes capable of breaking down a variety of macromolecules such as proteins, nucleic acids, lipids, and carbohydrates. A common hallmark of a large group of over 70 lysosomal storage diseases (LSDs) is the accumulation of undigested substrates within the lysosomal lumen, leading to lysosomal expansion [33]. For years, progressive disruption of this basic degradative function of the lysosome was considered an adequate explanation of the pathogenesis of LSDs, including Pompe disease – the first recognized storage disorder linked to the lysosome [34]. However, this long-held view of lysosomes as terminal degradation compartments is now a thing of the past. Instead, the lysosome is viewed as a sophisticated cellular center that controls a variety of cellular processes including cell growth, signaling, nutrient sensing, and autophagy [35, 36].

Macroautophagy (commonly referred to as autophagy) is a fundamental, evolutionarily ancient process that mediates the transfer of intracellular materials to lysosomes for degradation. The process involves the de novo formation of double-membrane vesicles, called autophagosomes, that sequester the cargo destined for degradation [37–40]. Autophagosomes fuse with lysosomes where the engulfed portion of cytoplasm is broken down and the resulting building blocks (e.g., amino acids, glucose, nucleotides, fatty acids) are exported back into the cytosol and utilized for energy generation and in biosynthetic pathways [41].

Initially, autophagy was described as a survival mechanism in response to cellular stressors, in particular amino acid starvation; induction of autophagy under nutrient-poor conditions allows the cell to derive new amino acids and energy from the random, non-selective (bulk) degradation of cellular components [42]. This response to environmental signals is mediated by the concerted actions of the mammalian target of rapamycin complex 1 (mTORC1), the master nutrient sensor and growth regulator, and AMP–activated protein kinase (AMPK), which is a key energy sensor. When nutrients are abundant, mTORC1 is recruited and activated at the lysosomal surface [43, 44]; once active, mTORC1 inhibits autophagy by phosphorylating autophagy-initiating kinase Ulk1. In contrast, when nutrients are insufficient, activated AMPK stimulates autophagy indirectly, by inhibiting mTORC1 (through phosphorylation of TSC2), and directly, by phosphorylating Ulk1 on distinct sites [45, 46]. Moreover, under nutritent-poor conditions, the inactive mTORC1 is detached from the lysosome and promotes autophagy by allowing translocation of transcription factors EB and E3 (TFEB and TFE3) to the nucleus where they activate genes involved in lysosomal and autophagosomal biogenesis [36, 47–50].

In addition to starvation-induced autophagy, autophagic machinery functions at low baseline levels to maintain cellular homeostasis by specifically recognizing and eliminating protein aggregates and damaged organelles [51, 52]. Based on the organelle destined for elimination, selective autophagy is called mitophagy (for mitochondria), lysophagy (for lysosomes), lipophagy (for lipid droplets), etc. Autophagic degradation of glycogen, a process termed glycophagy, was shown to have a critical importance in newborns [53–55].

Thus, the autophagy-lysosomal pathway plays a crucial role in the removal of worn-out organelles and toxic components as well as in cellular adaptation to various stresses and starvation. Dysfunctional autophagy has been associated with a range of pathologies including cancer, neurodegeneration, metabolic and cardiac diseases, and not surprisingly, LSDs including Pompe disease [56, 57]. The process is particularly important for the survival and stress adaptation of post-mitotic cells like neurons or muscle cells that are most affected in Pompe disease. Considering the evolving role of lysosomes, it is now amply clear that lysosomal dysfunction in the diseased muscle cells initiates a cascade of events far beyond the progressive glycogen accumulation. Disturbed autophagy and calcium homeostasis, oxidative stress and mitochondrial abnormalities, signaling and metabolic defects, all contribute to gradual muscle destruction in Pompe disease [58].

Autophagic defect in skeletal muscle of GAA knockout mice (KO) [59] morphologically manifests as massive accumulation of cellular debris containing multivesicular bodies and lysosomes, broken lysosomal membranes, double-membrane autophagosomes with undigested cytosolic content, abnormal mitochondria, multilamellar bodies with electron-dense core, and glycogen particles both inside and outside vesicular structures [1]. The autophagic buildup can be visualized by electron microscopy of muscle biopsies and by confocal microscopy of isolated muscle fibers immunostained with lysosomal and autophagosomal markers. In fact, it is the latter technique that allowed us to appreciate the full extent of this pathology, which often spans the entire length of the fiber and occupies up to 40% of its volume in older KO mice [60, 61]. In addition, undegraded autophagic substrates such as protein aggregates and p62/SQSTM1 can also be detected within the buildup areas by immunostaining. The routine analysis of muscle biopsies by Periodic-acid-Schiff staining might miss the buildup areas altogether, but upon closer look, one can often see “holes” in the core of muscle fibers on cross-sectional images (Fig. 1).

Fig.1

PAS-stained sections (shown in black and white) of gastrocnemius muscle from a 7-month-old GAA^-/- mouse after 4 ERT injections. Arrows indicate “holes” in the core of muscle fibers.

A similar picture is observed in muscle biopsies taken from Pompe disease patients, particularly those with late-onset form [60, 61]. The accumulation of undigested cellular debris [62, 63] leads to the disruption of myofibrillar structures and a significant reduction in muscle force, likely exacerbated by the toxicity of accumulated substrates [64]. Experimental evidence indicates that autophagy in the diseased muscle is impaired at both the initial and final stages of the process: more autophagosomes are generated and less are fully resolved by fusion with lysosomes [65, 66]. This imbalance between the production and clearance of autophagosomes leads to progressive accumulation of autophagic debris. The failure to digest lysosomal glycogen to glucose may deprive muscle cells of energy source, thus initiating autophagic response to glucose starvation. Activation of AMPK-TSC pathway and inhibition of mTORC1 activity in cultured GAA-deficient myotubes and in muscle of KO mice support this hypothesis [67, 68]. Alternatively, the diminished mTORC1 activity and ensuing induction of autophagy in the diseased muscle may be the initial insult originating from the inability of the damaged lysosomes to provide a proper platform for activation of this kinase.

Of particular importance is the data showing the link between defective autophagy and skeletal muscle resistance to ERT in Pompe disease. The autophagic buildup has a profound effect on vesicular trafficking and prevents efficient delivery of the therapeutic enzyme in muscle cells [65, 69]. Several attempts have been made to manipulate autophagy in muscle of KO mice by using genetic approaches: 1) genetic suppression of autophagy by inactivation of a critical autophagic gene, Atg7, in muscle of KO mice resulted in a significant reduction of lysosomal glycogen burden, thus serving as a substrate reduction therapy [70]; 2) overexpression of TFEB and TFE3 in GAA-deficient muscle cells and in muscle of KO mice stimulated autophagosomal-lysosomal fusion and exocytosis, reduced lysosomal size and glycogen content, and lessened the amount of cellular debris [65, 66]; 3) activation of mTORC1 in KO mice by knocking down its inhibitor, TSC2, largely rescued autophagic defect and reversed muscle atrophy [67, 68]. The common thread that runs throughout these experiments is that the currently available enzyme replacement therapy works well if skeletal muscle is cleared of autophagic buildup.

Importantly, recent data demonstrated that the reversal of autophagic pathology is feasible by using a more efficient next-generation drug for ERT. This new experimental drug, AT-GAA (Amicus Therapeutics) cleared muscle cells from lysosomal glycogen and reduced the number of myofibers with autophagic buildup in KO mice [71]. A similar reversal of autophagic accumulation in KO muscle has been recently achieved by gene therapy approaches [72, 73]. These results are, no doubt, very promising. It is, however, remains to be seen if the autophagic defect can be fully reversed and at what stage of the disease progression. In KO mice, the therapy was initiated when the animals showed little, if any, clinical manifestations of the disease. So far, there are no data showing the reversibility of this pathology in older symptomatic KO mice. This leaves open a possibility that the resolution of autophagic buildup in Pompe disease patients may require additional intervention. Of note, little is known about the potential role of neuronal autophagy in Pompe disease, although neuronal loss has been shown in animal models.

Choice of Adeno-Associated virus (AAV) vector serotype, promoter and route of administration.

Pompe disease is a complex multi-systemic condition with manifestations in the cardiac, skeletal, and smooth muscles, endothelial cells, and motoneurons [12]. Therefore, targeting GAA expression in the desired target cells is an important aspect of gene therapy and depends on the selection of vector capsid serotype, specific promoter, and route of administration [74, 75].

In this regard, different strategies have been used to evaluate the efficacy of AAV-mediated GAA expression in animal models (Table 2). Many studies address the skeletal and cardiac manifestations of Pompe disease using an intravenous (IV) administration of AAV driven by ubiquitous promoters such as the human cytomegalovirus (CMV) or the hybrid CMV enhancer/chicken β-actin (CBA). Other studies have primarily focused on addressing the neurological manifestations of the disease using CNS-targeted administration including intrathecal (IT), intra-cisterna (ICM) or intracerebroventricular (ICV) routes. Additionally, there is evidence that using AAV to drive liver expression of GAA under the control of liver-specifc promoters may be effective in ameliorating the peripheral manifestations in muscle. However, an important consideration is that the CNS penetration of GAA delivered via cross correction does not cross the blood brain barrier efficiently, much like the current ERT approach.

Aside from identifying the optimal route of administration for the primary tissues involved, choosing the correct promoter is a major step towards achieving successful therapeutic transgene expression. The CMV and CBA promoters provide rapid, robust and durable expression throughout the body except in the liver where these promoters may become inactive [76, 77]. On the other hand, muscle creatine kinase (CK) and hybrid α-myosin heavy chain creatine kinase (MHCK7) promoters are primarily used to boost expression in skeletal and cardiac muscle while exhibiting low expression in non-muscle tissue and decreasing immune response to the transgene [78]. Other promoters include desmin (DES) which shows preferential expression in motor neurons, skeletal and cardiac muscles [77, 79], the neuro-targeted synapsin promoter, the liver-specific promoter (LSP) used to induce tolerance [80], and the liver-muscle promoter (LiMP) that provides expression in non-dividing extra-hepatic tissues [81].

In the upcoming sections, we will discuss the routes of AAV administration that have been used to mitigate both the muscle and neuronal phenotypes in Pompe disease.

Potential Gene Therapy approaches for Infantile Onset Pompe Disease (IOPD): Intravenous (IV) delivery

ERT delays the onset of skeletal muscle weakness and the associated cardiomyopathy with minimal effects on the neurological aspects of the disease as GAA does not cross the blood brain barrier efficiently [82]. Systemic delivery of AAV vectors encoding GAA is an attractive alternative to ERT since cell autonomous correction in myocardium, skeletal muscle and the CNS will have significantly greater efficacy and comes with the potential for cross-correction of neighboring cells via GAA secretion. The need to correct the deficits in both the musculature and nervous system has become more apparent in recent years [83], and several recent preclinical studies have highlighted the feasibility of using rAAV-mediated IV gene delivery to correct the neuronal phenotype (Table 2).

In 2007, Mah et al. showed that a single systemic administration of an rAAV1 vector to restore GAA activity in vivo in a mouse model of Pompe disease resulted in long-term correction of the heart and skeletal muscle [84]. In addition, there was evidence of phenotypic correction of skeletal muscle exhibited through marked improvements in both soleus and diaphragm force mechanics. These results demonstrated the potential for systemic administration of an rAAV1 vector to mediate sustained therapeutic levels of correction in both skeletal and cardiac muscles [84]. Successful systemic administration using rAAV vectors to reduce glycogen content and augment cardiac, respiratory and motor neuron function in GAA^-/- mice was also demonstrated [85].

To compare and contrast the cardiorespiratory therapeutic potential of AAV vs. ERT, Falk et al. administered a single systemic injection of rAAV2/9-DES-hGAA (AAV9-DES) or bi-monthly injections of recombinant human GAA (ERT) to GAA^-/- mice [86]. Three months post-treatment, there was an improvement in weight gain, elongation of the PR interval, increased ejection fraction, and a reduction in left ventricular mass suggesting cardiac correction for both the AAV- and ERT-treated groups. However, AAV9 treated mice showed increased GAA activity as well as reduced glycogen accumulation in cardiac and respiratory muscles, while ERT treated animals did not. Additionally, GAA^-/- mice given rAAV9-DES had significant changes in respiratory function when compared to ERT-treated mice, suggesting partial correction of lower motor neurons and improved respiratory timing [86]. These findings are in accordance with the results reported by others [85, 87].

Lim et al. also demonstrated that IV administration of an rAAV9 variant, rAAV-PHP.B driven by an MCH promoter and expressing GAA, achieved therapeutic levels of GAA activity and reduced glycogen in both the myocardium and skeletal muscle. Furthermore, treated mice exhibited improved gait as well as very little to no peripheral neuropathy suggesting an effect on brain glycogen accumulation [88]. Similarly, Keeler et al. compared the systemic administration of a newly engineered vector with a high affinity for muscle and CNS, AAVB1-DES-hGAA to AAV9-DES-hGAA in 3-month-old GAA^-/- mice at the dose of 1 x 10¹² vg. AAVB1 treated animals displayed robust weight gains, near normal forelimb strength, and increased transduction in the diaphragm, tongue base, and the thoracic spinal cord when compared to AAV9. Both treated groups displayed above average GAA enzyme activity in the heart, diaphragm, tongue, gastrocnemius, and lung, and reduced glycogen accumulation in each of the muscles studied [89].

Although IV delivery is a successful strategy in mitigating the phenotype and underlying neuropathology of the disease, it faces crucial challenges regarding 1) cost and availability of high amount of product and 2) the host immune response against the vector capsid protein and transgene [90–92].

Intramuscular-targeted GAA expression

Preclinical studies in mice have demonstrated increased GAA activity and glycogen clearance following intramuscular injection in the gastrocnemius [75], the tibialis anterior [93], the heart [93], the diaphragm [94] and the tongue [95]. These preclinical studies strongly suggest that IOPD patients could benefit from a targeted intramuscular treatment. For example, the tongue appears to be particularly susceptible to impairments in Pompe disease, and animal studies show that the hypoglossal motoneurons which control the tongue are among the first cells to show histopathology and glycogen accumulation in the neonatal nervous system [95]. Children with Pompe disease receiving ERT have weak and ineffective swallowing [96] and speech disorders [13]. Neurological symptoms that include impaired language development, possibly indicating hypoglossal motor impairments, persist despite ERT [97]. LOPD patients also have lingual weakness, dysphagia and dysarthria despite ERT [98]. Pompe patients also have difficulty with saliva management, secretions and drooling [97]. Difficulties with phonation are also common in advanced Pompe disease [99]. Accordingly, “correction” of pathology in tongue motor units could have profound functional benefits in a clinical setting. A prior study in GAA^-/- mice indicates that intralingual treatment with rAAV-GAA has excellent potential for treating lingual dysfunction in Pompe disease. Specifically, a single intralingual injection of single stranded rAAV-GAA (1X10¹¹vg) drove persistent GAA expression in both lingual myofibers and brainstem XII motoneurons. The GAA expression caused a reduction of glycogen accumulation with reversal of the histopathology.

The diaphragm has also been a muscle of interest for gene therapy delivery. Smith et al conducted a phase I/II clinical trial (NCT00976352) of intradiaphragmatic delivery of rAAV1-CMV-hGAA in five patients with IOPD. Each subject received a dose of either 1x10¹²vg (n = 3) or 5x10¹²vg (n = 2) of rAAV1-hGAA and the patients were followed for 6 months post-treatment. Significant improvements were noted in each subject’s unassisted tidal volume and length of time of unassisted breathing with median increases of 29% and 425%, respectively [100]. A limitation of the intramuscular route is that the transduction is limited to the cells in the small area around the injection sites. A way to overcome this limitation is the application of multiple small volume intramuscular injections to increase transgene expression [101, 102].

Potential gene therapy approaches for late onset Pompe disease (LOPD)

Challenges in Gene Therapy: Production & Immune Response

Immune-responses to AAV capsid and the transgene, and vector re-administration

A critical challenge for the success of gene therapy is the host immune responses to both the vector capsid and transgene product. These immune responses pose ongoing concerns regarding the safety, longevity, and extent of gene expression. Human immune responses to foreign antigens include humoral and cellular responses to the AAV capsid and to the transgene product. The humoral responses to AAV include activation of antigen-specific CD4 + T helper cells [134, 135] and production of antibodies against the viral capsid [136]. Capsid neutralizing antibodies (NAbs) bind to the AAV capsid and may block or reduce the transduction of target cells. In humans, it has been shown consistently that administration of AAV in the dose range required for regional or systemic exposure leads to development of high and sustained anti-AAV antibodies [137–139]. For example, following IM administration of alipogene tiparvovec (Glybera, uniQure N.V.), all patients (n = 27) showed treatment-emergent antibody responses to AAV1 [140]. Mouse, NHP, and human studies have shown that even low NAb titers can completely block transduction by AAV vectors [139, 141–143]. More recently, ongoing clinical programs for the treatment of neuromuscular diseases such as X-linked Myotubular Myopathy (XLMTM), Spinal Muscular Atrophy (SMA), Duchenne Muscular Dystrophy (DMD), Mucopolysaccharidosis (MPS), Giant axonal neuropathy (GAN) and others announced immune-responses to either the AAV capsid or the transgene with variable severity of safety signals. Management of the host’s immune response to both the vector capsid and transgene product remains a critical challenge for the success of gene therapy.

In recent publications by Corti et al [144, 145], circulating anti-AAV1 and anti-GAA antibodies in a group of children with Pompe disease who received bilateral intra-diaphragmatic injection of rAAV1-CMV-hGAA were studied. Three subjects receiving immunomodulation with rituximab (every 6 months) and Sirolimus (daily) to modulate immune reaction to ERT were compared to six subjects who did not receive immunomodulation. The group that did not receive immunomodulation demonstrated a 150-fold increase in anti-AAV1 titer after exposure to AAV1. However, the three subjects, who received immunomodulation had no detectable levels of anti-AAV1 or anti-GAA antibody through day 365 of the study [144, 145]. In another study with a 2-year-old Canavan disease subject, immunomodulation with rituximab and sirolimus prevented formation of anti-AAV9 antibodies and prolonged the detection of AAV vector DNA in the circulation for up to 17 months [146].

Another important challenge related to gene therapy is AAV re-administration. Patients may be restricted from clinical studies or commercial clinical use due to the development of a humoral immune response against the AAV capsid. Such exposure can occur by environmental exposure to wild type AAV, cross-reactivity of related AAV, or by sub-optimal therapeutic doses as part of early phase clinical trials. This is a critical unmet need in the gene therapy field as many subjects may require redosing later in life. Further, it is possible to achieve a more effective total dose by repeated lower doses of AAV over time. This approach would also increase the safety of the treatment by reducing the potential toxicity of a single high dose exposure. However, evidence suggests that, in the absence of immunomodulation, humoral and cellular memory responses to AAV may decrease duration of gene transfer and compromise subsequent use of the same vector [137–139].

Different strategies have been suggested to allow for vector re-administration. Our group has conducted preclinical studies in mice and NHPs demonstrating that immunomodulation strategies allow safe and effective repeat administration of the same AAV serotype (Corti Oral communication and [129]). To be able to use rituximab in mouse models, we crossed GAA^-/- mice with a line expressing the human CD20 (hCD20) antigen and generated hCD20 + / GAA^-/- mice. Mice were pre-treated with rituximab for transient ablation of B cells prior to AAV exposure and sirolimus for modulation of T cell response after dosing. Our results demonstrated that single and repeat administration of rAAV9-DES-hGAA was safe and that the use of this immune suppression regimen can prevent formation of antibodies against GAA and AAV9 capsid [129]. In a similar experiment using NHP, animals received single or repeated doses (four months apart) of rAAV9-DES-hGAA in the tibialis anterior (TA), with or without immunomodulation. Our data confirms that animals that received immunomodulation had higher GAA activity than those who did not receive immunomodulation.

Additionally, an ongoing clinical trial is testing repeat intramuscular (IM) administration of rAAV9-DES-GAA in adult Pompe patients (#NCT02240407). This clinical study is a blinded crossover design to test safety and effectiveness of IM administration and re-administration of AAV9-DES-hGAA into TA muscle using a similar immune modulation strategy. The immune modulation strategy is to ablate B cells using rituximab and sirolimus prior to the initial exposure to AAV9 in one leg and the subsequent exposure of the same vector to the contralateral leg after four months. Side of administration is randomized at first injection. Preliminary data from the first two patients enrolled confirm preclinical results showing that immunomodulation prevents antibody formation against AAV9 and the transgene (Corti Oral communication ASGCT 21^nd Annual Meeting; #NCT02240407).

Our preclinical and clinical observations suggest that pharmacological immunomodulation prior and during AAV-mediated gene therapy is a successful strategy to modulate vector immunogenicity and to enable effective vector re-administration. Similar results were obtained by Meliani et al. [147] who demonstrated that when synthetic vaccine particles encapsulating rapamycin (SVP[Rapa]) are co-administered with AAV vectors, the anti-capsid humoral and cell mediated responses to AAV8-hF.IX and AAV8-GAA are reduced. This approach allowed successful vector re-administration in mice and NHP. However, this strategy would only work for liver gene transfer and not for our program, where we aim for a global gene transfer targeting both neurological and cardiac impairments. Other strategies have been proposed to address AAV re-administration and pre-immunity including plasmapheresis and capsid switching [147, 148]. However, both of these strategies have some limitations that impact their clinical applicability. Plasmapheresis requires multiple sessions in a short period of time and it does not completely eliminate high titer antibody [149]. Capsid switching is influenced by high cross reactivity of anti-AAV antibody [150]. In addition, it adds additional complexity in AAV manufacturing and product characterization.