Molecular genetics of Pompe disease: a comprehensive overview

Abstract

Pompe disease (PD) is an autosomal recessive lysosomal disorder caused by the deficient activity of acid alpha-glucosidase (GAA) enzyme due to mutations in the GAA gene. The enzymatic deficiency leads to the accumulation of glycogen within the lysosomes. Clinically, the disease has been classically classified in infantile and childhood/adult forms. The GAA gene has been localized to chromosome 17q25.2-q25.3 and to date, 582 mutations distributed throughout the whole gene have been reported (HGMD: http://www.hgmd.cf.ac.uk/ac/). All types of mutations have been described; missense variants are the most frequent type followed by small deletions. Most GAA mutations are private or found in a small number of families. However, an exception is represented by the c.-32-13T>G splice mutation that is very common in patients of Caucasian origin affected by the childhood/adult form of the disease, with an allelic frequency ranging from 40% to 70%. In this article, we review the spectrum of GAA mutations, their distribution in different populations, and their classification according to their impact on GAA splicing process, protein expression and activity. In addition, whenever possible, we discuss the phenotype/genotype correlation. The information collected in this review provides an overview of the molecular genetics of PD and can be used to facilitate diagnosis and genetic counseling of families affected by this disorder.

Introduction

Pompe disease (PD-MIM# 232300), is an autosomal recessive lysosomal storage disorder due to mutations in the acid alpha-glucosidase (GAA) gene (MIM#606800) encoding the lysosomal GAA. This genetic defect leads to the deficient activity of GAA resulting in impaired glycogen degradation and accumulation within the lysosomes (1).

Clinically, PD encompasses a highly variable range of phenotypes that differ in the age of onset, extent of organ involvement, and rate of progression (2). Patients with the classic infantile onset form, the most severe and rapidly progressive phenotype, die within the first year of life from cardiorespiratory insufficiency if untreated (3,4). Patients with onset in childhood or adulthood are characterized by the presence of progressive limb-girdle myopathy and respiratory dysfunction. These patients become wheel chair and/or ventilator dependent, and the respiratory insufficiency is the leading cause of death (5-7).

The GAA gene is localized to chromosome 17q25.2-q25.3 (8) and it was cloned and sequenced in 1991 (9). Since then, significant advances have been made in understanding the molecular bases of this disorder, and to date almost 600 mutations have been reported (HGMD: http://www.hgmd.cf.ac.uk/ac/). However, the pathological mechanism by which a mutant enzyme leads to the wide range of phenotypes in patients affected by this disease remains elusive.

The GAA gene is approximately 28 kb long and encompasses 20 exons. The first exon contains the 5' untranslated sequences and is separated from the second one by a large intron of approximately 2.7 kb. The first ATG is located in exon 2, 32 nt downstream from the beginning of the exon. The second and the last exons are quite big (578 and 607 bp, respectively), while the remaining exons length ranges from 85 to 187 bp (9).

The promoter sequence, located upstream of the first GAA exon, has been characterized and, as expected, has the characteristics of housekeeping gene promoters (10).

The GAA cDNA is 3.6 kb long (8) and encodes a precursor peptide of 952 amino acids with a predicted molecular weight of 105 kD. The precursor has an amino-terminal signal peptide for co-translational transport into the lumen of the endoplasmic reticulum, where it is N-glycosylated resulting in a glycosylated precursor with an apparent molecular mass of 110 kDa (11). Seven glycosylation sites at residues 140, 233, 390, 470, 652, 882 and 925 have been predicted (12).

The enzyme is then transferred to the Golgi complex where high-mannose type oligosaccharide side chains are phosphorylated (13-15) and subsequently targeted to the lysosome via the mannose-6-phosphate receptor. Within the late endosomal/lysosomal compartment the enzyme undergoes a series of proteolytic and N-glycan processing events: the 110-kDa precursor is proteolytically processed at the amino terminus, resulting in a 95-kDa intermediate with a sequence beginning at amino acid 122; the 95-kDa intermediate is processed at the carboxy terminal to a 76-kDa form, which is then proteolytically cleaved at the amino terminus to a 70-kDa mature form (13-16). In 2005, Moreland and colleagues demonstrated that the lysosomal human GAA is composed of four different peptides of 70, 19.4, 10.3, and 3.9 kDa, the latter two being disulfide-bonded (17). Recently, the structure of recombinant human GAA (rhGAA) has been solved by X-ray crystallography, using Myozyme^® as the source of protein (18). Interestingly, four disordered surface loops have been identified—notably, the same that were absent in the protein complex studied by Moreland et al.—that hampered the formation of productive protein crystals, and a high-resolution diffraction model was obtained only after their proteolytic removal (PDB entry: 5NN3). Scanning the entire length of GAA protein from N-terminus to C-terminus, a trefoil Type-P domain is separated from the catalytic GH31 (β/α)₈ barrel domain by a β-sheet domain. A proximal and a distal β-sheet domains constitute the C-terminal end of GAA. Two unstructured inserts arising from the catalytic domain and a loop bearing from the N-terminal β-sheet domain delimit the active site, in which the conserved residues R600 and D282 play a pivotal role in the substrate recognition and stabilization, and residues D518 and D616 are essential for the catalysis. A second substrate-binding pocket, shaped within the N-terminal trefoil Type-P domain, has been identified and could potentially boost the enzyme processivity (18). Finally, of the seven glycosylation sites originally predicted (17), only five of them, namely N140, N233, N390, N470 and N 652, have been experimentally validated to bind M6P glycans (18,19).

The GAA mutations

The mutational spectrum of GAA gene is very heterogeneous. To date 582 mutations distributed throughout the whole gene have been listed at HGMD-http://www.hgmd.cf.ac.uk/ac/. All types of mutations have been described. Missense mutations are most frequent followed by small deletions. Indeed, as shown in Figure 1, 297 (51.0%) of reported mutations are missense, 87 (14.9%) are small deletions, 16 of which are in frame, 74 (12.7%) are splicing variants, 51 (8.8%) are nonsense, 35 (6.0%) are small insertions/duplications, 19 (3.3%) are gross insertions/deletions, 13 (2.2%) are small indels, 5 (0.9%) are complex rearrangements. Only one variant in the regulatory region has been described so far.

Figure 1.

Frequency of GAA mutant alleles reported in the HGMD-http://www.hgmd.cf.ac.uk/ac/ classified by mutation type. GAA, acid alpha-glucosidase.

PD is considered a pan-ethnic disease. However, there are some countries in which the prevalence of the disease is very low. This is the case, for instance, of Finland where only one Pompe patient was definitively diagnosed (20,21).

A number of studies have been conducted aimed at depicting the mutational profile of the GAA gene on a national basis, often distinguishing between the classic severe infantile and the late onset childhood/adult cohorts.

Even though most GAA mutations are private or found in a small number of families, the c.-32-13T>G splice mutation is an exception, since it is very common in patients of Caucasian origin affected by the childhood/adult form of the disease, with an allelic frequency ranging from 40% to 70% in different populations (6,22-36).

The presence of this intronic mutation results in the exclusion of exon 2 from a variable proportion of the expressed GAA mRNA. In other words, by affecting the overall splicing efficiency, it changes the balance between the GAA splicing isoforms towards the non-functional, exon 2-skipped species, yet not completely preventing the expression of the normal transcript that can be translated into an enzymatically active GAA protein. Therefore, the resulting variable levels of GAA residual activity may explain the delay of the phenotypic manifestation of the disease in those who carry the c.-32-13T>G mutation (37-40).

Apart from the common intronic c.-32-13T>G mutation, few variants are overrepresented in particular populations. Perhaps the best examples are the c.525delT and the c.2481+102_2646+31del mutations, the so called “Dutch mutations” due to their high recurrence in Dutch population (41-43). Both variants have been classified as severe mutations. Indeed, the deletion of a single nucleotide (c.525delT) or the entire exon 18 (c.2481+102_2646+31del) cause in both cases a shifting of the open reading frame leading to the generation of a premature stop codon. No transcript from the c.525delT allele is detected in patient’s cells, probably due to the synthesis of an unstable mRNA which would be rapidly degraded (44). As expected, these variants are associated with the infantile-onset form of PD and occur in a cohort of infantile Dutch patients with an allelic frequency of 35% and 31%, respectively (41). The allelic frequencies of both mutations are much lower in Dutch patients with the childhood/adult phenotype, representing only 15% and 8% of the alleles, respectively (41). By comparison, the allelic frequency of the c.525delT in a similar cohort of Italian infantile patients is 13.8% (45).

Geographical differences in the allelic frequency of GAA mutations within the same country have been noted. For instance, the most frequent pathogenic mutation encountered in the infantile subset of patients in the North of China is the c.2662G>T (p.Glu888*) mutation, accounting for 23.1% of total mutant alleles (46). By contrast, the most frequent mutation in a cohort of patients from Southern China is the c.1935C>A (p.Asp645Glu) accounting for the 20–25% of total mutant alleles (46,47). Notably, such a geographical partition is not evident in childhood/adult patients from mainland China in which the most frequent pathogenic GAA mutation is the c.2238G>C (p.W746C) accounting for 27.1% of total mutant alleles (48). Also, in Taiwan the c.1935C>A mutation seems to be the most recurrent mutation in patients affected by the infantile onset form of PD, accounting for 12.2% of total mutant variants. This could be explained by the Chinese origin of the Taiwanese population (49). The importance of historical migratory fluxes in the determination of the mutational profile of specific populations is evidenced by some studies carried out in countries that had experienced a marked immigration in their history, such as Argentina, Brazil, Colombia or Canada. Indeed, a very heterogeneous profile of GAA mutations has been reported in these countries (50-53). However, the severe mutation c.2560C>T (p.R854*) was found to be overrepresented (allele frequency of 16.7%) in a Brazilian cohort of infantile patients (54). This is not an unexpected finding considering the massive migratory flux coming from North Africa, where a high frequency of this mutation is very well documented (55). Not surprisingly, the same c.2560C>T mutation is the most common defect in African-Americans with PD (55).

Recently, Fukuhara and colleagues demonstrated that the mutation c.546G>T occurs with high allelic frequency (22.9%) in patients affected by the childhood/adult phenotype in Japan (56).

Mutations associated with GAA pseudodeficiency

Two sequence variants, c.1726G>A (p.Gly576Ser) and c.2065G>A (p.Glu689Lys), often present in cis on the same allele, are associated with a pseudodeficiency of GAA. The first one, effectively reduces both the amount of expressed GAA protein and its catalytic activity (57-59), while the second one has little effect on GAA functionality (60). Therefore, individuals who are homozygous for this allele have very low GAA enzymatic activity but do not develop PD (49,59,61,62).

This complex allele is quite frequent among people of Asian descent. Indeed, about 4% of subjects in Asian populations carry this allele in homozygosity (62).

The functional impact of GAA mutations

While the possible impact of a nonsense mutation or a deletion/insertion leading to a frameshift on the synthesis and function of the GAA enzyme could be foreseeable, it is much more difficult to predict the effect of missense mutations or variants affecting sequences involved in the splicing process. Indeed, sequence variations leading to the generation of premature stop codons (nonsense of frameshift mutations) would necessarily result in the synthesis of truncated non-functional proteins. Furthermore, in many cases transcripts carrying premature stop codons are quite unstable and degraded via nonsense mediated decay (NMD) with the consequent loss of GAA synthesis. Therefore, in general these types of mutations could be considered as severe ones.

Several in silico prediction tools have been developed to analyze the impact of genetic variants on the mRNA splicing processes. However, although in silico analysis became increasingly promising and trustable in predicting the likely pathogenic effect of most of the analyzed variants (63), this tool cannot substitute functional analysis. Indeed, minigene splicing assays or, even better, direct mRNA analysis should be required for linking variants to a putative splicing defect. Furthermore, functional analysis remains the only reliable tool to determine the severity of a given mutation since in many cases the mutation may not completely abrogate the expression of normally spliced transcript leading to the protein expression and retention of some residual activity (64). The best example of this type of variant is the common c.-32-13T>G mutation (see below).

Similarly, several programs have been developed to predict the potential effect of missense mutations. However, analysis of protein expression and activity in patients’ cells and/or by in vitro expression studies of the mutated proteins remain the gold standard to determine their pathogenetic nature and to gain insights on their severity.

It is worth noting that in vitro expression studies are very useful to test the effect of a given variant when mutations are found in compound heterozygous individuals.

A classification of GAA mutations based on a severity rating system organized in six classes has been proposed by Kroos et al. (58). The severity of a particular mutation is defined based on the results of a transient transfection of an expression construct carrying wild-type or mutated GAA cDNA; each sequence variation is compared to the wild-type GAA in terms of quantity and quality of processed and unprocessed GAA protein, and residual enzymatic activity in cells and media (58).

The six severity classes are identified by letters, whereby A stands for “very severe”, B for “potentially less severe”, C for “less severe”, D for “potentially mild”, E for “presumably nonpathogenic”, and finally, F for “nonpathogenic” (58).

Class A mutations are Cross Reactive Immunologic Material (CRIM)-negative, since no molecular forms of GAA can be detected by immunoblotting, and no residual enzymatic activity can be measured. Even though classes B, C and D mutations are CRIM-positive, the levels of processed and unprocessed forms of GAA are lower than normal ones, leading to a residual enzymatic activity ranging from 0 to 30%. Finally, both classes E and F mutations present quantitatively and qualitatively normal levels of GAA forms, but they differ in residual enzymatic activity, which is higher than 30% in class E and higher than 60% in class F (58,65).

The Pompe Disease Mutation Database, available at http://www.pompecenter.nl has been created with the aim of providing a list of GAA variants and to describe their effect in order to facilitate the diagnosis and genetic counseling.

The database, last updated in May 2016 reports 558 GAA sequence variations: class A and B mutations are the most represented groups with 183 and 139 hits, respectively. Thirty mutations are labeled as “less severe”, 22 as “potentially mild”, and only 6 as “presumably nonpathogenic”. Class F is the third largest group, including 93 sequence variations, and finally, the severity of 85 mutations still remains unknown (Figure 2).

Figure 2.

Frequency of GAA mutant alleles reported in the Pompe Disease Mutation Database, available at http://www.pompecenter.nl classified according to their functional impact on GAA. GAA, acid alpha-glucosidase.

Besides being useful for the interpretation of genetic data obtained in PD patients, the knowledge of the functional consequences of a mutant variant might be useful to select patients that could benefit from experimental therapies, such as the use of antisense oligonucleotides able to rescue splicing defects (66,67) or molecular chaperons to improve GAA activity by promoting folding, processing and trafficking to the lysosome of mutated but partially active variants (68).

The common c.-32-13T>G mutation

The c.-32-13T>G variant is the most frequent GAA mutation associated with the childhood/adult phenotype. Indeed, almost 90% of patients affected by this phenotype carry this mutation on at least one allele. This mutation, first described by Huie et al. in a patient affected by the adult onset form of PD (37), is located 13 nucleotides upstream of the canonical acceptor splice site of GAA intron 1 (Figure 3). As supported by studies performed in fibroblasts derived from patients carrying the c.-32-13T>G base change, the main functional consequence of this mutation is the synthesis of different aberrant splicing variants in which the exon 2 is completely or partially spliced out and a limited amount of normally spliced GAA mRNA. Interestingly, the same non-functional splicing isoforms were detected, although in low quantities, in cells from normal subjects (38). These data were then further confirmed in vitro by Raben and colleagues using a minigene assay (39). Therefore, from a clinical point of view, these data strongly support the concept that the production of a certain amount of “normally spliced” wild-type mRNA may represent a general mechanism underlying the delayed symptomatic expression in patients bearing the c.-32-13T>G mutation (39,40).

Figure 3.

GAA mRNA splicing isoforms expressed in cultured fibroblasts. (A) Schematic representation of the 5' region of the GAA gene (exons 1 to 3). The position of the c.-32-13T>G mutation is highlighted in red. The cryptic splice sites, located 35 nt downstream from the normal donor splice site of exon 1 and at 60 nt upstream from the donor site of exon 2, are shown as c1 and c2, respectively. The presence of the c.-32-13T>G mutation abrogates the binding of the U2AF65 splicing factor. (B) Schematic list of the GAA mRNA species expressed in human fibroblasts. Normal spliced GAA mRNA (N) and splicing species SV2 and SV3 are detected both in fibroblasts from patients carrying the c.-32-13T>G variant and healthy controls. In patients vs. controls samples, however, the relative abundance of the various splicing isoforms is different. While N is the main mRNA species expressed in normal cells, SV2 and SV3 are the main species detected in cells from patients bearing the c.-32-13T>G variant. GAA, acid alpha-glucosidase.

From the functional point of view, it has been shown that the mutation interferes with the binding of the splicing factor U2AF65 to the GAA pre-mRNA, almost completely abrogating its interaction with the polypyrimidine tract of exon 2 leading to the general inefficiency of the splicing process (Figure 3). Moreover, it has been demonstrated that the overexpression of specific mRNA binding proteins can modulate the expression of normally spliced GAA mRNA from the c.-32-13T>G mutated allele (40).

Recently, very promising results have been obtained in vitro by targeting a specific silencer located within exon 2 with a combination of antisense oligonucleotides. Indeed, treatment of myotubes of patients carrying the c.-32-13T>G mutation resulted in a significant increase of exon 2 inclusion and GAA activity, and a decrease in lysosomal glycogen accumulation (67). A similar approach supports the possibility of promoting exon 2 inclusion and GAA enzyme activity by targeting inhibitory sequences within intron 1 of GAA (68).

Phenotype-genotype correlation

As for most genetic diseases, a strict correlation between the phenotype and the genotype cannot be established in PD. However, some general conclusions can be drawn from the data obtained in hundreds PD patients.

In general, patients affected by the severe and rapidly progressive infantile form of PD harbour mutations that completely abolish the expression of all forms of GAA protein or lead to a low expression of unprocessed or processed GAA (classified as A or B mutations). In all cases these mutants do not retain significant amount of residual GAA activity (45,69,70). Conversely, the presence of a potentially mild mutation in one allele seems to prevent the occurrence of the severe classic infantile phenotype (22-26,28,30).

Apart from these general considerations, it is difficult to establish a strict correlation between a particular mutation and the clinical presentation and progression of the disease since most GAA mutations are private and found in compound heterozygosity. As mentioned above, the only exception so far is the frequent c.-32-13T>G mutation, which is also often found in compound heterozygosity in patients affected by the childhood/adult phenotype. Therefore, several authors have analyzed the phenotype of patients carrying this mutation in association with a severe “null” variant that is not contributed to GAA activity. Besides the fact that all described patients presented with the childhood/adult form of PD, they demonstrated a wide variability in residual activity, age at onset, and disease progression (21,33,71-73). Furthermore, different phenotypic expression has also been reported in siblings carrying this genotype (74). Considering these observations, it is very likely that secondary factors, genetic and non-genetic, would act as modifiers of the PD phenotype. This hypothesis is further supported by the data collected in patients carrying the c.-32-13T>G mutation in homozygosity who, unexpectedly, presented with the full spectrum of adult PD phenotype (29,75).

In an attempt to explain this wide phenotypic variability, the c.-32-13G>T haplotype was studied in a cohort of 98 compound heterozygous patients carrying the c.-32-13T>G mutation in combination with a null variant. This study failed to demonstrate a correlation between the c.-32-13T>G haplotype and the phenotype. In addition, several authors have explored the possible modifying effect of an angiotensin I-converting enzyme (ACE) polymorphism (76-80). However, the results of these studies do not concur and the role of this polymorphism in the phenotypic expression of PD remains controversial.

Very recently, the c.510C>T variant has been identified as a genetic modifier of the disease onset in compound heterozygous or homozygous for the common c.-32-13T>G variant. Indeed, the c.510C>T when present in cis with the c.-32-13T>G mutation, modulates the splicing pattern of the mutated transcript further reducing the relative amount of correctly spliced mRNA which in turn, resulted in reduced residual GAA activity (81).