An integrative correlation of myopathology, phenotype and genotype in late onset Pompe disease

Pompe disease is caused by pathogenic mutations in the alpha 1,4‐glucosidase (GAA) gene and in patients with late onset Pome disease (LOPD), genotype–phenotype correlations are unpredictable. Skeletal muscle pathology includes glycogen accumulation and altered autophagy of various degrees. A correlation of the muscle morphology with clinical features and the genetic background in GAA may contribute to the understanding of the phenotypic variability.

Aims: Pompe disease is caused by pathogenic mutations in the alpha 1,4-glucosidase (GAA) gene and in patients with late onset Pome disease (LOPD), genotype-phenotype correlations are unpredictable. Skeletal muscle pathology includes glycogen accumulation and altered autophagy of various degrees. A correlation of the muscle morphology with clinical features and the genetic background in GAA may contribute to the understanding of the phenotypic variability. Methods: Muscle biopsies taken before enzyme replacement therapy were analysed from 53 patients with LOPD. On resin sections, glycogen accumulation, fibrosis, autophagic vacuoles and the degree of muscle damage (morphology-score) were analysed and the results were compared with clinical findings. Additional autophagy markers microtubule-associated protein 1A/1B-light chain 3, p62 and Bcl2-associated athanogene 3 were analysed on cryosections from 22 LOPD biopsies. Results: The myopathology showed a high variability with, in most patients, a moderate glycogen accumulation and a low morphology-score. High morphologyscores were associated with increased fibrosis and autophagy highlighting the role of autophagy in severe stages of skeletal muscle damage. The morphologyscore did not correlate with the patient's age at biopsy, disease duration, nor with the residual GAA enzyme activity or creatine-kinase levels. In 37 patients with LOPD, genetic analysis identified the most frequent mutation, c.-32-13T>G, in 95%, most commonly in combination with c.525delT (19%). No significant correlation was found between the different GAA genotypes and muscle morphology type. Conclusions:

Introduction
Pompe disease (OMIM 232300) is an autosomal recessive disease caused by a defect in the lysosomal alpha glucosidase due to bi-allelic pathogenic mutations in the alpha 1,4-glucosidase gene (GAA). GAA is located on chromosome 17q25. 2-25.3 and contains 20 exons resulting in a protein of 952 amino acids [1]. Nowadays, more than 500 variants of mutations in GAA are known, which are listed at the Pompe Disease Mutation Databank (http://www.pompevariantdatabase.nl/) and are associated with a severity ranking by determining GAA enzyme activity [2][3][4]. However, the clinical phenotypes associated with most of the pathogenic variants show a high variability without clear genotypephenotype correlation, which makes prediction of the disease outcome difficult [2][3][4][5][6][7][8][9].
In individuals with late onset Pome disease (LOPD), the pathogenic variant c.32-13T>G is the most common and is associated with a milder phenotype [9,11,12,15]. In German LOPD patients, the c.-32-13T>G mutation is also the most common. The missense mutations c.307T>G and c.877G>A are more common in German patients than in the other European patients [16][17][18]. In individuals with LOPD, treatment with ERT reduces muscle weakness, respiratory insufficiency and mortality [19,20]. The GAA enzyme defect leads to lysosomal and extra-lysosomal accumulation of glycogen in striated muscles with disorganization of myofibrillar architecture and loss of function of the muscle cell [17,[21][22][23]. In addition, disturbed autophagy contributes to the muscle pathology and autophagic vacuoles (AV) are frequently present in muscle biopsies from patients with Pompe disease [21,[24][25][26][27][28][29]. ERT leads to glycogen clearance in skeletal muscle from the individuals with IOPD and LOPD. Nevertheless, severely abnormal microscopic findings in skeletal muscle at the start of therapy are associated with a worse clinical outcome [21,22,29,30]. Also, muscle pathology can progress even on ERT, a phenomenon, which is not well understood. The aim of the present study was to analyse the muscle morphology of 53 patients with LOPD in detail and compare these findings with the clinical data. Histopathological changes were correlated with individual genotypes in 37 patients. To our knowledge, this is the largest cohort of detailed muscle biopsy analysis in LOPD, and individual correlation of detailed muscle morphology with the GAA genotype has not been described before.

Patients
Patients with LOPD who underwent a muscle biopsy were included in this study. In all 53 patients, a muscle biopsy was taken for diagnostic purposes before ERT was started. Muscle biopsy location was chosen based on the clinical presentation of the myopathy (Table S1). The diagnosis of Pompe disease was confirmed by a reduction in GAA enzyme activity and/or genetic testing. We included all available muscle biopsy tissue samples from confirmed Pompe patients though the German Neuromuscular centre network. The patients have been seen and examined at different Pompe centres in Germany and resin embedded muscle tissue and cryosections were provided from the associated laboratories. Clinical data were collected retrospectively. Data on enzyme activity levels were available in 26 patients with samples derived from either dried blood spot or muscle tissue. The GAA enzyme activity levels were stated as the percentage of normal values. The GAA mutations were available for 37 patients and expected effects were categorized according to the Pompe Disease Mutation Database as potentially mild, severe or unknown phenotype (Table 1) [4]. Clinical data on the duration of symptoms prior to biopsy were available in 30 patients. Serum creatine kinase (CK) at the time of biopsy was available in 39 patients (Table S1).

Muscle pathology analysis
Resin sections and transmission electron microscopy Small samples were taken from open muscle biopsies and fixed in 4% glutaraldehyde/ 0.4MPBS and processed according to standard procedures. One to two µm semithin sections were cut from resin embedded tissue and stained with Periodic acid-Schiff (PAS) and 2%-p-phenyldiamine (PPD) [21]. Morphometric analysis of glycogen accumulation, large empty vacuoles (LEV) and pathology score were performed on PAS-stained resin sections using a light microscope equipped with a motorized stage and digital camera with newCAST software as described before [21]. In brief, the section was delineated and a point grid containing 256 crosses was projected to each field of view. For estimating glycogen deposits and LEV, the area below the points were graded as PAS positive, PAS negative or LEV and estimated as percentage of total area [%/muscle fibre (MF)]. For estimating the muscle score, area below the points were graded from 0 to 5 according to the previously described criteria (grade 0 = normal, grade 1 = small PAS + vacuoles, grade 2 = medium PAS + vacuoles, grade 3 = larger PAS + vacuoles with myofibrillar disruption, grade 4 = larger PAS + vacuoles and LEV with myofibrillar disruption, grade 5 = LEV with myofibrillar disruption) (Figure S1A-C). A morpho-score was calculated (x% grade0 9 0) + (x%grade1 9 1) + (x% grade2 9 2) + (x%grade3 9 3) + (x% grade4 9 4) + (x%grade5 9 5)/100 from these findings expressed in grade 1-5 with 0 as lowest grade demonstrating normal morphology [21].
Muscle fibres with AV (vacuoles with dark inclusions) were analysed on PPD stained resin sections as percentage of all MFs. Additionally, the degree of fibrosis was estimated semi-quantitatively on PAS stained resin sections and rated from normal (0) to severe (3) (Table S2). For contrast in transmission electron microscopy (TEM), ultrathin sections were treated with 3% lead citrate-3H 2 0 with a Leica EM AC20 (ultrastain kit II, Leica Microsystems, Wetzlar, Germany) and examined at a Zeiss EM109 TEM (Carl Zeiss Microscopy GmbH, Jena, Germany) equipped with a sharp eye digital camera.
Statistics Correlations between the morphological score and clinical data and autophagy markers were analysed by calculating the Spearman rank correlation coefficient to asses a possible linear association between two variables. To analyse differences between the genetic combination with regard to score, age and GAA enzyme activity Kruskal-Wallis-test was used. Data were analysed using the R software package (version 3.4.3 for Windows; https://CRAN.R-project. org, R Core Team, [51]).

Patients
Muscle biopsy specimens from 29 (55%) male, age at biopsy varying between 16-78 years (median 46 year), and 24 (45%) female patients, age at biopsy between 14-72 year (median 42 year), were analysed. The distribution of gender and age in this study was similar to those in other studies [29]. Patients were 14À78 years of age (n = 53, median 44 year) at time of biopsy, with n = 29 (55%) between 30 and 50 years. Disease onset was at 8-73 years (n = 33; median 39 year) with a diagnostic delay to the time of biopsy and disease duration of 0-26 years (n = 33, median 5 year) ( Figure 1A and Table S1).
The morpho-score significantly correlated with fibrosis (Spearman's q = 0.46; P = 0.0006) showing remodelling of connective tissue in severe stages. The analysis of AV showed a significant positive correlation with the morpho-score (Spearman's q = 0.34; P = 0.0173), highlighting the role of autophagy during muscle pathology progression. The morpho-score significantly correlated with glycogen accumulation (PAS) (Spearman's q = 0.86; P < 0.0001) and LEV (Spearman's q = 0.35; P = 0.01) as was to be expected since both parameters are included in the morpho-score ( Figure 2).
TEM Samples with different degrees of pathology were examined at the ultrastructural level. Muscle biopsies with a low morpho-score mainly showed lysosomal glycogen and normal sarcomeric architecture ( Figure 3A,C,E). Biopsies with high morpho-scores showed massive lysosomal and non-lysosomal glycogen deposits with sarcomere disorganization ( Figure 3B,D, F). Additionally, AV were more prominent in biopsies with high morpho-scores, often accompanied by abnormal mitochondria with increased variation of size and disruption of cristae structure ( Figure 4A-F). In many biopsies with low morpho-scores, in addition to lysosomal glycogen, intermyofibrillar glycogen was present ( Figure 5A,B). In biopsies with higher scores, extra-lysosomal glycogen surrounding lysosomes was often seen in combination with autophagosomes with double layered membranes and autophagic material ( Figure 5C,D).
Enzymatic staining for acid phosphatase and expression of antibodies against autophagy markers p62, LC3 and BAG3 Staining intensity of acid phosphatase and expression of autophagy markers LC3 and p62 in 22 muscle biopsies showed a stronger upregulation in muscle biopsies graded with a higher morpho-score and increased AV. The analysis of the morpho-score showed a significant positive correlation with p62 (Spearman's q = 0.48; P = 0.031) and acid phosphatase (Spearman's q = 0.70; P = 0.0006) compatible with increased autophagy during Pompe disease progression ( Figure S2A,B and Table S3). Analyses of clinical data compared to muscle morphology. Patients analysed are equally distributed in gender (male 55%) with no difference at age at biopsy. The age at symptom onset varies from 8 to 73 years (median 39 year) and age at biopsy from 14 to 78 years (median 44 year) with a duration of symptoms before biopsy of 0-26 years (median 5 year) (A). The majority of patients are between 30-50 years old and have glycogen accumulation < 10 PAS%/muscle fibre (MF) (75%) and a low morpho-score < 1 (68%) without differences in muscle morphology between different age groups (B). The superimposed box and whisker plots have been constructed using the quartiles and medians (solid horizontal line within the boxes). The lines add 1.5 times the interquartile range (IQR) to the 75th percentile or subtract 1.5 times the IQR from the 25th percentile and are expected to include 99.3% of the data. They are surrounded by violin plots showing the density distribution of the values.
Double immunofluorescent staining against Z-band proteins desmin and BAG3 was performed at 22 biopsies. In the majority of biopsies, small BAG3 positive aggregates were present mainly in the core of well-preserved fibres and were increased in biopsies of high p62 and LC3 expression. In comparison, in MFs with large vacuoles (LEV), there was co-expression of desmin and BAG3 in the peripheral area reflecting remnants of Z-band structures and no BAG3 positive inclusions were detected. At TEM, large autophagosomes were mainly seen in the core of MFs with normal sarcomere structures whereas they were not present in MFs with abundant LEV ( Figure S4). These findings suggest that BAG3 plays a role in early, but not late stage in the single MF pathology. Interestingly, one patient (P28) with a high number of AV showed high expression of LC3, p62 and BAG3 consistent with increased autophagy even at a moderate score (0.38) and glycogen  Table S3).

Discussion
Investigation of muscle biopsies gives a great opportunity to analyse disease pathogenesis and may add to the understanding of the broad clinical variability in LOPD. The aim of our study was a detailed analysis of muscle morphology from a large cohort of patients with LOPD before the start of ERT and comparison of these findings to the clinical, laboratory (GAA enzyme activity, CK levels) and pathogenic variants in GAA.
For grading the muscle pathology, we used a morpho-score, which we had already established on resin embedded biopsies from patients with IOPD [21]. The morpho-score includes, in addition to glycogen accumulation, myofibrillar disruption and LEV as a feature of disturbed autophagy. In IOPD patients, a lower score reflecting less pathology is associated with a better response to ERT with recombinant GAA, whereas patients with a high morpho-score have a limited ERT response [21]. The data are in line with other reports showing a worse response to therapy in patients with severely damaged MFs with worse pathology indicating the need for early therapy in patients with minor Figure 6. Correlation matrix of morpho-score, age at biopsy, disease duration, creatine kinase (CK) and alpha 1,4-glucosidase (GAA) residual enzyme activity obtained from the patients. Upper triangular of the matrix shows the Spearman correlation coefficients (Spearman's q) with significance levels associated to a symbol (P-values: ***<0.001, **<0.01, *<0.05). The distribution of each variable is shown on the diagonal (red). Lower triangular part of the matrix shows the bivariate scatterplots (red) with a fitted line (black) and the confidence interval of the line (grey). The morpho-score shows no significant correlation with clinical symptoms and GAA residual enzyme activity, CK levels are significantly negative correlated with the age at biopsy. pathology [30]. But even early therapy does not predict a good clinical outcome in all patients [10,32,33]. However, the underlying reason why the efficiency of therapy is restricted is not fully understood. Here, we translate this morpho-score to the muscle analysis in LOPD. Although, muscle pathology in LOPD patients has been reported earlier, our cohort encompasses the highest number of tissue specimens investigated in detail so far [17,22,23,29]. In our study, LOPD muscle biopsy samples show a high variability in muscle morphology. Interestingly, 75% of the specimens have minor muscle damage with a glycogen accumulation of <10% and a morpho-score below grade 1. On average, the morpho-score is much lower in LOPD compared to IOPD, which shows muscle pathology before the start of ERT with scores up to 4, as expected by a higher residual GAA enzyme activity in LOPD [17,21]. Muscle biopsies from LOPD patients can show a subtle pathology and sometimes are graded as unspecific on PAS stained cryosections [34]. In the present study, although in four patients the morpho-score was low, glycogen containing vacuoles were present in all of them on PAS-stained resin sections. Treatment with recombinant GAA leads to decreased intra-lysosomal glycogen accumulation [35]. In our study even in muscle biopsy specimens with a low morpho-score, extra-lysosomal glycogen was detectable, but without major disruption of the sarcomeric structure. Whereas high morpho-scores were associated with large extra-lysosomal glycogen deposits and severe sarcomeric disorganization. The morpho-score and glycogen accumulation correlated with increased endomysial fibrosis, informing about severe stages of muscle damage with increased connective tissue replacement. Recently, analyses of glycogen clearance in LOPD muscle samples showed that free extra-lysosomal cytoplasmic glycogen persisted in post-treatment biopsy specimens, suggesting that ERT treatment may have been introduced too late, but neither fibrosis nor sarcomeric disorganisation was described [22].
The age and gender distribution of our cohort reveals a median age of 41 years, 55% of patients were male patients and a diagnostic delay of 5 years, consistent with other reports [9,17,29]. Correlating muscle pathology with disease duration shows that patients with longer disease duration do not have more severe muscle tissue damage. These data underscore the clinical heterogeneity and indicate that additional factors may contribute to disease progression and severity. Interestingly, serum CK levels were lower in older patients, possibly related to disease associated muscle atrophy.
In mechanically strained striated MFs, autophagy is an important process for normal cell function, and dysfunction of autophagy plays a major role in disease progression in lysosomal myopathies [24,36]. In Pompe disease, autophagy is impaired with autophagic buildup of various degree in both IOPD and LOPD showing large vacuoles and increase of autophagy markers, LC3 and p62, which might accelerate the muscle damage [37][38][39]. Large vacuoles and large glycogen deposits remain unchanged on ERT treatment, whereas a reduction in glycogen is associated with a reduction in vacuolated fibres [29].
Therapeutic benefits by modifying autophagy in Pompe disease have been demonstrated in several animal models [40][41][42][43]. In our study, buildup of autophagy showed a significant positive correlation with the morpho-score and glycogen accumulation, highlighting the role of autophagy in severe stages of LOPD muscle pathology. The autophagic markers, LC3 and p62 correlated with the lysosomal enzyme acid phosphatase and were increased in muscle biopsies with more severe pathology. Tension induced accumulation of proteins in skeletal MFs are degraded by chaperone-assisted selective autophagy in which the co-chaperone BAG3 plays an important role [44]. Interestingly in our study, small BAG3 positive inclusions were present mostly in the core of preserved MFs suggesting that BAG3 is involved in early events and not at late stage pathology. BAG3 has been found to be located inside autophagosomes and triggers the autophagy of ubiquitinated clients [45], which is in line with large autophagosomes found at ultrastructural analysis in MFs in our study. These observations also fit well with autophagic buildup observed in the core of the MFs in GAA mice [37]. The autophagic buildup seems to disrupt sarcomere structure and leads to loss of muscle force rather than buildup of glycogen filled lysosomes [46]. Further studies are necessary to investigate the role of BAG3 in the autophagic process in Pompe disease in more detail.
Countries with new-born screening for Pompe disease are increasing and in addition, to the early genetic diagnosis of IOPD, a large number of mutations in GAA associated with LOPD are being detected and are of an uncertain clinical outcome [47,48]. Over 500 different pathogenic GAA mutations are known and affected individuals show high phenotypic variability despite sharing the same GAA mutations, even within a family [3,7,32]. Until now, only a few factors such as angiotensin enzyme polymorphisms have been reported to modify the individual Pompe disease phenotype [5]. The c.-32-13T>G mutation is the most common GAA mutation in Europe. In individuals with compound heterozygous mutations, disease manifestation may be highly variable regarding both age of onset and initial symptoms [4,6,7,16,49,50]. In our cohort of 37 German LOPD patients, 95% are compound heterozygous with the c.-32-13T>G mutation on one allele, and show symptom onset at a mean age of 39 years (12-73 year), consistent with published data [4]. The high variability of mutations affecting the second allele, with the most common combination being with c.525delT in 19% in our patients along with several other less common mutations, makes correlative analysis difficult. Our analysis does not show differences in the morphoscore or disease onset between different allele combinations. Therefore, we ranked the phenotype of the mutations according to the Pompe mutation database. We studied different allele combinations to elucidate whether patients with different combinations had distinct clinical features, ages of onset or muscle pathology [2]. The majority of the LOPD patients had a combination of a mild with a severe GAA (49%). However, clinical findings and muscle morphology do not relate to the genetic findings in the investigated patient group. Therefore, a genotype-phenotype correlation cannot be defined.

Conclusion
The detailed analysis of muscle morphology in a large cohort of patients with LOPD reflects high variability of muscle tissue damage. This variability is accompanied by moderate glycogen accumulation and a low morpho-score in most patients. High morpho-scores are associated with increased fibrosis and autophagy. This highlights the role of autophagy in severe stages of muscle tissue damage. Comparing muscle morphology and genetic findings reveals no meaningful genotypephenotype correlations. This emphasises the putative relevance of yet unknown genetic modifiers independent of the GAA mutational spectrum. Transcriptomics, proteomics and metabolomics studies need to be performed to address these open questions.

Ethical statement
This work was in accordance with the ethical standards of the responsible committee on human experimentation (ethical committee of the University of Giessen; AZ07/09) and with the Helsinki Declaration (1964, amended most recently in 2008) of the World Medical Association and written informed consent was obtained from the patients.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Figure S1. For estimating the morpho-score Periodic acid-Schiff (PAS) stained resin sections were graded from 0 to 5: grade 0 = normal, grade 1 = small PAS + vacuoles, grade 2 = medium PAS + vacuoles, grade 3 = larger PAS + vacuoles with myofibrillar disruption, grade 4 = larger PAS + vacuoles and large empty vacuoles (LEV) with myofibrillar disruption, grade 5 = LEV with myofibrillar disruption (A). Muscle sections show muscle fibres with different grading close to each other (B,C). (A-C: PAS stained resin sections, magnification 9 400). Figure S2. Staining intensity of acid phosphatase and expression of autophagy markers microtubule-associated protein 1A/1B-light chain 3 and p62 in patients with different morpho-scores. Muscle biopsies with a higher morpho-scores shows increased expression compared to muscle biopsies with a lower score (A). Correlation matrix of score, Periodic acid-Schiff (PAS), large empty vacuoles (LEV), autophagic vacuoles (AV) and fibrosis obtained from the muscle biopsies. Upper triangular part of the matrix shows the spearman correlation coefficients (Spearman's rho with significance levels associated to a symbol (P-values: ***<0.001, **<0.01, *<0.05). The distribution of each variable is shown on the diagonal (blue). Lower triangular part of the matrix shows the bivariate scatterplots (blue) with a fitted line (black) and the confidence interval of the line (grey) The morpho-score shows a significant positive correlation with p62 and acid phosphatase compatible with increased autophagy and lysosomal activity during muscle pathology progression (B). Figure S3. With double immunofluorescent staining against Z-band proteins desmin (green) and Bcl2-associated athanogene 3 (BAG3) (red) small BAG3 positive inclusions in the core of muscle fibres (MFs) can be detected mostly in well-preserved MFs and are increased in muscle biopsies with high expression of microtubule-associated protein 1A/1B-light chain 3 and p62. In vacuolated fibres sarcomeric remnants show co-expression of desmin and BAG3 (yellow) and no larger BAG3 positive inclusions. Figure S4. Ultrastructural analysis of muscle biopsies in which immunofluorescence staining against Bcl2-associated athanogene 3 (BAG3) was performed: large autophagosomes are located mainly in the muscle fibre (MF) core. Only focal loss of Z-bands is noted at higher magnification accompanied by swollen microtubules, small vacuoles and intra-and extra-lysosomal glycogen (P53). In single MFs with more severe sarcomere disarray or large areas of glycogen accumulation no increased large autophagosomes are present (P33, P41). Muscle fibres with upregulation of microtubuleassociated protein 1A/1B-light chain 3 expression show increased numbers of large autophagosomes (P30). Figure S5. Analyses of alpha 1,4-glucosidase (GAA) gene mutations compared to muscle morphology and clinical data: Boxplots of the individual morpho-score to GAA gene mutation combination. The combinations for each subject are shown as points, and the box and whisker plots show the quartiles and medians (solid horizontal line within the boxes). The whiskers add 1.5 times the interquartile range (IQR) to the 75th percentile or subtract 1.5 times the IQR from the 25th percentile. The global differences between the individual scores to mutation combination did not met statistical significance (P > 0.1). According to the Pompe Disease Mutation Database mutations were ranked as mild (potential less severe mutation); severe, (very severe mutation); unknown, (unknown mutation) and no single nucleotide polymorphism (SNP) (AV, autophagic vacuoles; LEV, large empty vacuoles). No significant differences are detectable analysing the vacuoles (AV or LEV) with respect to the GAA genotype. Table S1. Clinical data and genetic background of the patients with late onset Pome disease. Table S2. Morphological data of the patients with late onset Pome disease. Table S3. Analysis of autophagy markers in patients with late onset Pome disease.