Introduction

Fabry disease (MIM 301500) is an X-linked deficiency of α-galactosidase A (α-GALA, EC 3.2.1.22) leading to widespread lysosomal accumulation of glycosphingolipids with terminal α-galactosyl moieties (mainly globotriaosylceramide, Gb3Cer). The ensuing phenotype at the clinical level is well known, including its organ-limited milder variants in hemizygous males and in females who are mosaic hemizygotes [9, 10]. The dominating therapeutic approach, enzyme replacement therapy (ERT), is based on substitution of the missing enzyme with recombinant α-galactosidases, FabrazymeTM (agalsidase beta, Genzyme) or ReplagalTM (agalsidase alfa, SHGT) [6, 13]. Despite differences in the glycosylation pattern due to different expression systems [14, 41], both enzymes appear to be functionally indistinguishable from each other [3, 23]. The essential requirement for successful ERT of Fabry disease is the delivery of a sufficient dose of the enzyme into the storage lysosomal compartment. The effect on storage has been evaluated either by monitoring the level of storage or the induction of enzyme activity [12, 21, 3, 23, 33]. However, there is not any report demonstrating complete clearance of the storage lysosomal system generally. The effect is mostly restricted to endothelia, but the partial clearance effect has been seen by some authors also in another cells, e.g., in the kidney [15].

The goal of this study was to compare effectiveness of ERT in two cell systems: by cultured mutant fibroblasts on one side, representing dynamic cell population minimally modified by low degree spontaneous storage, and transiently exposed to trace amount of tritiated substrate to follow its catabolic pathway. This in vitro experimental system was compared to the situation in vivo, represented by a series of cell types of various regeneration capacities exposed to prolonged ERT after longstanding disease course. Extreme counterpart to cultured fibroblasts was represented by cardiocytes, postmitotic cells, modified by longstanding spontaneous storage.

This is also the first study focused on delivery of the applied α-GALA into the storage cardiocytes which may contribute to better understanding of the factors responsible for the limited effectiveness of ERT in vivo.

Patients

For details of examined patients, their ERT regime, and the biopsy samples, see Table 1.

Table 1 Patients’ data
Full size table

Materials and methods

The study was approved by the Ethics Committee of the General Teaching Hospital and the First Faculty of Medicine, Charles University (60/2003F), Prague. Cardiac biopsy samples were taken during coronary surgeries. Written informed consent was obtained from patients included in the study.

Cell lines

Four fibroblast cell lines of patients with classic type of Fabry disease (mutations in the GLA gene:p.Q280K, p.R342Q, c.801ins36, c.delEx2 (c.195−?−369+?del)) came from the cell culture depository of the Institute of Inherited Metabolic Diseases. Fibroblasts with mutations c.801ins36 and p.Q280K were from patient 1 and patient 2, respectively, who underwent heart biopsy later on. All cell lines (including p.R342Q and c.delEx2) were derived from skin biopsies taken several years before ERT from patients with classical form of the disease.

Recombinant α-galactosidases A

REPLAGAL (agalsidase alpha, a gift from Shire Human Genetic Therapies, Cambridge, MA, USA) sample no. 001-0229 was used without any further modification (dilution 1mg/ml solution). FABRAZYME (agalsidase beta, a gift from Genzyme, Cambridge, MA, USA) batch no. H0214AM, diluted according to manufacturer’s recommendation to 5mg/ml and then further diluted to 1mg/ml with redistilled water.

Experiments in cell cultures

Cultured skin fibroblasts were grown in accordance with standard protocols in Dulbecco’s modified Eagle’s medium (DMEM) and 10% bovine fetal serum (BFS).

Preparation of substrate

The globoside, Gb4Cer, was isolated from human erythrocyte membranes, characterized, and tritium-labeled at the ceramide moiety (330MBq/mg). The (3H)Gb4Cer was incorporated into liposome particles (a mixture of phosphatidylcholine, α-tocopherol, and phosphatidic acid, Sigma, Czech Republic) and coated with apoprotein E [2].

Feeding assays

DMEM, 3 ml per 25-cm2 flask, with 1% BFS and radiolabeled glycolipid (3.3 × 106dpm/1ml) was applied to nearly confluent Fabry fibroblasts and incubated in a 5% CO2 incubator for 3days [2].

Uptake of enzyme

Three milliliters of the enzyme solution (1.32μg of enzyme per milliliter, final activity 3.8μmol MU mg−1.h−1, the same for both enzymes) were placed in fresh DMEM containing 10% heat-inactivated BFS and applied to cells in 25-cm2 flasks. The flasks were then incubated for different periods of time. At least two 25-cm2 flasks were analyzed for each time period/enzyme combination (0, 12, 24, 48, 96, and 192 h).

Analysis of (3H)Gb3Cer degradation

Three milliliters of the enzyme solution (0.66μg of enzyme per milliliter, final activity 1.9μmol MU mg−1 h−1, the same for both enzymes) were placed in fresh DMEM containing 10% heat-inactivated BFS and conduritol B epoxide (final concentration 100mM, Calbiochem-Novabiochem GmbH, Germany) and applied to cells in 25-cm2 flasks preloaded with (3H)Gb4Cer and incubated for different periods of time. Analyses were performed at least in two 25-cm2 flasks for each time period/enzyme combination (0, 9, 14, 19, and 24 h).

Cell harvesting and homogenization

The fibroblasts were harvested by trypsinization. The cell pellets were homogenized in distilled water in a cuphorn sonifier filled with ice for 3 × 10s (Ultrasonic homogenizer 4710, Cole-Parmer, Chicago, IL, USA, 40% output). Protein concentration [16], enzyme activities, and uptake of the (3H)Gb4Cer tracer were determined in homogenate aliquots.

Enzyme assay

The activity of α-GALA was assayed in culture medium and in cell and tissue homogenates according to the modified method of Mayes et al. [27]. Briefly, the reaction mixture contained 5mM of substrate, 4-methylumbelliferyl(MU)-α-d-galactopyranoside (Sigma, Germany), in citrate buffer (pH 4.5). For cell and tissue homogenates, the reaction also contained 100mM α-N-acetyl-d-galactosamine to inhibit α-N-acetylgalactosaminidase activity. The complete reaction mixture was incubated for 30min at 37°C, and the fluorescence of the released 4-MU was measured at excitation 365nm and at emission 448nm on a luminescence spectrometer (Perkin Elmer LS50B, Wellesley, USA).

Extraction and thin-layer chromatography analysis of (3H)glycosphingolipids

The remaining cell homogenate was extracted with chloroform–methanol (2:1, v/v), and the lipids were separated on thin-layer chromatography (TLC) plates (Silica gel, Merck, Germany) using chloroform–methanol–water (70:30:05, v/v/v) as the solvent system. Chromatograms were evaluated using a TLC linear radioactivity analyzer (Raytest, Germany) [2].

Immunofluorescence labeling

Fibroblasts were grown on glass coverslips to near-confluence using standard methods. The cells were then grown for an additional 24h in DMEM containing 10% of heat-inactivated BFS and 0.66μg/ml of the recombinant enzyme.

For the experiments using mannose-6-phosphate (M-6-P) receptor blocking, the M-6-P (Sigma, Germany) was added to the medium together with the enzymes to yield a final concentration 10mM. Cells were fixed for 10min in ice-cold methanol before immunodetection and non-specific binding was blocked with 5% BFS. Recombinant agalsidases were detected using rabbit polyclonal antibody against human α-GALA TK88 (anti α-GAL, TKT, Cambridge MA, USA) diluted to 9μg/ml, while cathepsin D was detected simultaneously using monoclonal mouse antibodies (Serotec, Oxford, UK) diluted to 20 μg/ml. Gb3Cer was located using mouse monoclonal anti-Gb3 (CD77) antibody (BGR23, code 370680, Seikagaku, Japan) diluted to 6.7μg/ml after 10min fixation with 4% phosphate-buffered paraformaldehyde and blocking with 5% BFS. The primary antibodies were visualized using Alexa Fluor 488-conjugated goat anti-rabbit IgG, Alexa Fluor 488-conjugated goat anti-mouse IgG and Alexa Fluor 568-conjugated goat anti-rabbit IgG (Molecular Probes, Invitrogen, Carlsbad, CA, USA). The samples were examined using two Nikon confocal microscope setups (1) Nikon C1, Eclipse E800 microscope equipped with a laser scanning confocal head (Nikon, Japan) and a Nikon 60X, NA 1.4 oil immersion objective or (2 Nikon C1si LSM head mounted on TE-2000 inverted microscope and with Nikon TIRF NA 1.49 60× oil immersion objective. All channels were recorded separately to avoid cross-talk between channels. The co-localization analysis was performed with representative images using ImageJ software [31] with JaCoP plugin [4] and Huygens image analysis software (SVI, Hilversum, The Netherlands). The degree of co-localization was assessed by coefficients described by Manders et al. [26]. Two independent experiments were done to support the results.

The image acquisition conditions were kept constant in all parameters for samples presented in Fig. 3e–h.

Histochemical and ultrastructural analyses of tissue samples

Patient 1. The sample of superficial part of the left cardiac ventricle biopsied after 6months of ERT was processed for electron microscopy. Specimens of the heart auricle and part of the saphenous vein graft taken after 21months of ERT were transferred to the liquid nitrogen and stored on dry ice.

Patient 2. Samples of the heart auricle, skin, and skeletal muscle taken after 65months of ERT were frozen in liquid nitrogen or fixed with phosphate-buffered paraformaldehyde (Fluka, Buchs, Switzerland). Samples for standard ultrastructural analysis were fixed with 10% paraformaldehyde overnight, post-fixed in phosphate-buffered 1%Os04, dehydrated and embedded into Epon-Araldit mixture. Samples for electron microscopic immunostaining were fixed with 4% paraformaldehyde for 2 h, dehydrated, and embedded into LR White resin (London Resin Company, Reading, England). Skin biopsies taken before ERT start from both patients were processed for standard electron microscopy. Patient 3 (death one year after ERT termination). Postmortem frozen and resin embedded samples from the left cardiac ventricle were included into the study to substitute approximate status before ERT.

Cryostat sections were prepared from the frozen samples and used for immunohistochemical evaluation of the lysosomal storage status (Gb3Cer storage, lipopigment accumulation) for detection of the applied agalsidase protein and for histochemical evaluation of its activity.

Lysosomal storage status

Gb3Cer storage was detected in cryostat sections after pretreatment with cold anhydrous acetone (at 4°C) for removal of possible apolar lipid admixture and for glycolipid fixation. They were briefly post-fixed with 4% buffered paraformaldehyde (for 5min at 4°C). After standard blockade of endogenous peroxidase (with sodium azide and H2O2) and non-specific protein–protein interactions (with 5% BFS), the sections were incubated with monoclonal mouse anti-Gb3Cer antibody diluted 1:100 in 5% BFS (at 37°C for 1h or at 4°C overnight). Detection of bound primary antibody was achieved using EnVision + System-HRP Anti-Mouse (DakoCytomation, Glostrup, Denmark) and the DAB +Chromogen kit (K 3460, DAKO Cytomation, Carpinteria, USA). The staining specificity was checked using total lipid extraction by chloroform (C)–methanol (M)–water (W) 4:8:3, followed by C–M 2:1 (each step for 30min at room temperature). The sections were rinsed in methanol, briefly fixed in paraformaldehyde, and used for the lipid immunodetection. Lysosomal storage status was also studied by birefringence of liquid crystals of the stored polar lipid and by detection of lipopigment autofluorescence. The latter was evaluated by epifluorescence microscopy using the Nikon BV-2A filters blocks (EX 400-440, DM 455, BA 470). The best results were obtained after brief (5min) pre-extraction of the lipids with C–M mixture (1:1, v/v) to prevent moderate non-specific autofluorescence of the stored lipid. The liquid crystalline state of the stored lipid deposits exhibited considerable instability and was best evaluated using Baker’s fixative solution. The results were compared with samples from age-matched controls. Neutral term lipopigment was used throughout the manuscript because of ambiguous terminology [34].

Immunohistochemical detection of the applied enzyme protein

Cryostat sections were fixed with 4% phosphate-buffered paraformaldehyde exposed to standard blocking procedures (see above) and incubated for 1h at 37°C with the primary polyclonal antibody TK 88 diluted to 4.5μg/ml in 5% FCS then washed in PBS and exposed to the detection kit (EnVision + System-HRP Anti-Rabbit, DakoCytomation). An alternative fixation of sections using a mixture of chloroform–acetone (1:1, v/v) for 30min at room temperature followed by paraformaldehyde was used for some of the experiments. Semi-thin 0.5-μm sections embedded into LR White medium were treated with citrate buffer pH6.0 for 40min at 90°C and incubated after the aforementioned blocking procedures with TK88 antibody for 1h at 37°C and thereafter treated with the detection kit as described above.

The sections immunostained either for Gb3Cer or for α-GALA were counterstained with Harris hematoxylin and mounted into the Immu-Mount medium (Shandon, Pittsburgh, PA, USA) which allowed simultaneous evaluation of the presence of the applied enzyme (eventually Gb3Cer), birefringent lipid and autofluorescent lipopigment.

Histochemical detection of in situ enzyme activities

Activity of α-GALA was evaluated using both the post-coupling technique with 6-bromo-2-naphthyl-α-d-galactopyranoside (LaChema, Brno, Czech Republic) and Fast Blue B and the simultaneous azo-coupling technique with 1-naphthyl-α-d-galactopyranoside (Koch Light, Colnbrook Bucks, UK) and hexazonium pararosaniline; both were done on semi-permeable membranes allowing to detect activities of both bound and soluble enzymes according to the published prescriptions [24]. Semi-permeable membrane technique was used for demonstration of activities of β-galactosidase (indigogenic method), β-glucuronidase, and acid phosphatase (simultaneous azo-coupling method using in both cases naphthol ASBI-type substrates and hexazonium pararosaniline). Acid phosphatase activity was also demonstrated using the aqueous medium [24].

Mannoso-6-phosphate receptor (MPR, cation-independent) was detected using monoclonal antibody (2G11 clone, Abcam, Cambridge, UK), diluted 1:400. Polyclonal antibody against cathepsin D was purchased from DAKO and used in dilution 1:200. In both cases, the incubation was overnight at 4°C.

Results

Studies in cultured mutant cells

Uptake of recombinant enzymes into cultured Fabry cells and efficiency of Gb3Cer hydrolysis

To monitor the function of agalsidases in situ, the uptake into Fabry fibroblasts was measured over the course of 192 h (8days). Both enzymes were similarly internalized into the cultured cells, with maximum activity at 24h incubation. The uptake of agalsidase beta measured with MU-substrate was 2.5-fold greater than that of agalsidase alpha, although the enzymes were applied at the same concentration (1.32mg/ml medium). This result was confirmed by repeated experiments on three different Fabry fibroblast cell lines. After 24h, the cellular activities of both enzymes started to decline. However, even after 8days, the values exceeded several times the normal activity of healthy cells (Fig. 1). Activity in the culture medium decreased equally in both agalsidases, falling to 4–6% of its original value (Fig. 1, inner graph).

Fig. 1
figure 1

Internalization of recombinant α-galactosidases A (agalsidases alpha and beta) into Fabry skin fibroblasts as a function of time. Recombinant enzymes (1.32 μg of enzyme protein per milliliter of medium) were applied into cultures of Fabry cells incubated under standard conditions. After 0, 12, 24, 48, 96, and 192 h, medium was removed, cells harvested, and analyzed in vitro with MU-substrate as described in “Materials and methods”. Each value is average of duplicate flasks. Intracellular agalsidase beta activity exceeded 20 times normal levels at the maximum (24 h), while agalsidase alpha activity was eight times the normal value in healthy cells (50.3 ± 13.6 nmol MU mg−1 protein h−1 in controls, see broken line). Insert is a time course of agalsidases stability in the culture medium

Full size image

The correction of the defect in Fabry cells was monitored with natural substrate in mutant cell cultures pre-loaded with tritium-labeled globoside (3H)Gb4Cer, the (3H)Gb3Cer precursor. The quantity of degradation products calculated from TLC-radioactivity scans reflected approximately the uptake tendencies of the respective enzymes (Fig. 2). Despite the differences in uptake (Fig. 1), both loaded agalsidases proved efficient in intra-lysosomal degradation of Gb3Cer.

Fig. 2
figure 2

Degradation of (3H)Gb3Cer by Fabry fibroblasts loaded with (3H)-globoside after application of agalsidases alpha and beta. Illustration of 24-h experiments (example of one experimental set from three is shown). (3H)-globoside (5 × 106 dpm), precursor of (3H)Gb3Cer, was incorporated into the liposomes coated with Apo E and added to the culture medium as described in “Materials and methods”. After 3 days of incubation, the agalsidases were applied at final activity of 1.9 μmol ml−1 h−1. After 0, 9, 14, 19, and 24 h, cells were harvested, the lipids extracted, separated by high performance thin layer liquid chromatography, and the products were quantified by TLC linear radioactivity analyzer (“Materials and methods”). Percentage of degradation products was calculated from TLC scans in relation to total cell radioactivity comprising substrate (3H)Gb3Cer and the sum of its degradation products partially transformed into products of re-synthesis (phospholipids)

Full size image

Confocal microscopic examination of internalized agalsidases with endosomal/lysosomal markers in cultured Fabry cells

Fabry fibroblasts were incubated with recombinant agalsidases for 24h. The intracellular distribution of agalsidase alpha and agalsidase beta in the cells was detected by anti α-GALA antibody (red channel, Fig. 3a,b). The lysosomes were visualized in the perinuclear region by immunodetection of cathepsin D (green channel, Fig. 3a,b). In a similar manner, the localization of intracellular Gb3Cer deposits was correlated with recombinant enzymes internalized into α-GALA deficient fibroblasts (Fig. 3g,h). The merger of the agalsidase and cathepsin D signals suggested lysosomal localization of loaded recombinant enzymes and was further documented by co-localization maps (Fig. 3c,d). Correlation of agalsidase signal with lysosomal transmembrane protein LAMP 2 was also demonstrated (data not shown).

Fig. 3
figure 3

Co-localization of internalized agalsidase alpha (a–i) and agalsidase beta (b–j) with cathepsin D and Gb3Cer in the Fabry fibroblasts after 24-h incubation (a, b) Confocal analysis of cells labeled for α-galactosidase A (α-GALA, red) and intra-lysosomal marker cathepsin D (CatD, green) displayed granular distribution of both signals and their parallel considerable overlap. Mander’s coefficients M GAL and M CatD ranged from 0.209 to 0.441 and 0.53 to 0.849 in the images, respectively. M GAL and M CatD coefficients represent the fraction of co-localizing pixels from the overall number of pixels with intensity values above the given threshold calculated for the particular image channel. Threshold levels to filter background were set to 10% of the pixel intensity range in both channels. The co-localization of both signals is also illustrated by co-localization maps (c, d) employing overlap coefficient values (0–1), color coded by lookup table placed under this panel. Addition of 5 mM of M6P to the culture medium completely blocked the α-GALA transport to the lysosomes (e, f) and changed the signal of α-GALA from granular (intra-lysosomal) pattern to diffuse low-intensity signal virtually inseparable from nonspecific background. The co-labeling of α-GALA and Gb3Cer in the cultured fibroblasts of Fabry patients incubated 24 h with recombinant agalsidases showed very limited overlap (g, h). The distribution of co-localization of both signals is also demonstrated by co-localization map of overlap coefficient (i, j). Comparing images (g, h) and (i, j) demonstrates minimal correlation of α-GALA and Gb3Cer signals in the cells. This observation suggests that the population of lysosomes with strong α-GALA detection has already been cleared off Gb3Cer. Scale bars in f (valid for e, f) and j (valid for a–d and g–j) correspond to 10 μm

Full size image

Endogenous enzyme protein was not detected in the cells of Fabry patients (data not shown). The addition of M-6-P to the culture medium, together with agalsidases, strongly inhibited the uptake of both enzymes (Fig. 3e,f), confirming that M-6-P receptor endocytosis is a major pathway of agalsidase transport into fibroblasts. The double labeling of Gb3Cer (Fig. 3g,h, green channel) and recombinant enzymes (Fig. 3g,h, red channel) did show only a few granular structures containing non-degraded Gb3Cer which were still present. The co-localization map of Gb3Cer and recombinant enzymes (Fig. 3i,j) showed only limited overlap (compare logarithmically color coded overlap coefficient in co-localization maps of Gb3Cer/α-GALA and CatD/α-GALA).

Studies in tissues

In vitro α-GALA activity in the heart samples

α-GALA activity determined in the auricular sample of the patient 1 was 1.53 vs. 10.18 ± 2.32 nmol MU mg−1 protein h−1 in controls (n = 5). This value represented 15% of the normal activity of the enzyme in the auricular part of the heart which implies slight increase from deficient values. There was no representative sample of the heart auricle in patient 2 for biochemical analysis. Residual enzyme activity in the autopsy sample (patient 3) was barely detectable (0.12 nmol MU mg−1 protein h−1, i.e., about 1% of the control mean).

Immunohistochemical and ultrastructural studies

Evaluation of the lysosomal storage status

Before ERT, dermal capillaries (skin biopsy available in patients 1 and 2) displayed lysosomal storage (Fig. 4a). ERT led to complete clearance of the capillary endothelium lysosomes already after 6 months (patient 1) from the stored lipid and to retraction of the originally expanded system to normal dimensions (Fig. 4b,c). Capillary basement membranes were thin and single. Capillary endothelial storage in various degrees was detectable in the postmortem heart specimen of the patient who died 1 year after ERT withdrawal (Fig. 4d).

Fig. 4
figure 4

Ultrastructure of the storage in the heart capillary endothelium. a Capillary endothelium with prominent storage before ERT. Patient 2 (bar 2 μm). b Capillary endothelium free of storage after 65 months of ERT. Persisting storage in pericapillary cells. Patient 2 (bar 4 μm). c Detail of the capillary endothelium after 6 months of ERT with discrete lysosomal system containing unspecific lipopigment. Patient 1 (bar 2 μm). d Lysosomal storage in the capillary endothelium of the heart after 1 year of ERT termination. This picture represents extreme in the wide range of capillary storage levels. Patient 3 (bar 2 μm)

Full size image

The auricular cardiocytes in both patients on ERT displayed variable, but on average high degree, Gb3Cer storage concentrated in the large perinuclear region (Fig. 5a). Lipid storage was paralleled by increase in lipopigment (Fig. 5c). Cardiocytes free of lysosomal lipid storage and containing only lipopigment were rare, and those free of both lipid and lipopigment were exceptional. The postmortem heart specimen displayed massive and uniform lipid storage occupying the bulk of the cardiocyte cytoplasm (Fig. 5b). Lipopigment was present in the form of dispersed tiny granules (Fig. 5d). Ultrastructure of cardiocytes, studied in all three cases, showed numerous lysosomes with varying degrees of aggregation and fusion with deposits of loosely or densely packed lipid membranes often mixed with pleiomorphic lipopigment deposits (Fig. 6a). In patient 2, there were exceptional cardiocytes with discrete lysosomal lipid storage but with numerous small lucent vesicles, suggesting relatively effective lysosomal clearance (Fig. 6b).

Fig. 5
figure 5

Persisting lipid storage in cardiocytes. a, b Immunohistochemical demonstration of Gb3Cer. a Cardiocytes of the heart auricle (patient 2) after 65 months of ERT (bar 100 μm). b Cardiocytes of the left ventricle. Autopsy (patient 3) 1 year after ERT cessation. Insert shows absence of Gb3Cer staining in control cardiocytes (bars 100 μm). Note certain degree of variability of storage, reflecting parallel excessive lipopigment deposition. c, d Increase of the autofluorescent lipopigment demonstrated after brief lipid extraction with C–A. c Large lipopigment clusters in heart auricle cardiocytes (patient 2). Insert: age-matched control (bars 100 μm). d Cardiocytes in the left ventricle (patient 3) with dispersed numerous discrete lipopigment granules (bar 100 μm). Insert: age-matched control showing coarser lipopigment granules restricted to the perinuclear region (bar 50 μm)

Full size image
Fig. 6
figure 6

Ultrastructure of the cardiocyte storage. a Patient 2. Atrial cardiocytes after 65 months of ERT with numerous individual lysosomes accumulating lipid dominating over lipopigment (bar 2 μm). b Exceptional finding in the same patient showing only discrete lipid storage lysosomes. There are numerous small vesicles with lucent content, which most probably represent cleared lysosomes (bar 2 μm)

Full size image

In all cases, irrespective of the ERT, the vascular smooth muscle cells displayed marked lipid storage. Persistent storage was seen in the skin eccrine glands (Fig. 7a) in pericytes and stromal fibroblasts, especially those in the hair bulb matrix (Fig. 7b). Discrete storage was found in the skeletal muscle fibers (not shown here). In the sample of the saphenous vein graft, there was prominent lipid storage in intimal cells and in the smooth muscle cells of the tunica media (Fig. 7c), which also displayed hypertrophy and hyperplasia. Endothelial cells of the tunica intima were free of storage, as were the endothelial cells of the adventitial vessels.

Fig. 7
figure 7

Persisting storage in other cell types. a, b Patient 2 (ERT 65 months). Birefringent stored lipid in eccrine skin glands (a) (bar 100 μm) and in hair follicle matrix fibroblasts (b) (bar 50 μm). c Patient 1 (21 months on ERT). Graft of saphenous vein used for coronary artery reconstruction with massive birefringent deposits of the stored lipid (bar 100 μm)

Full size image

Storage in some cell types was accompanied by increase in lipopigment (mostly in the cardiocytes, skeletal muscle, but almost absent in the smooth muscle cells). Electron microscopy showed densely packed lipid membranes and pleiomorphic structure of the lipopigment in lysosomes.

Evaluation of the intracellular delivery of the applied enzyme

In situ immunodetection of the applied enzyme performed in heart samples of the ERT-treated patients showed relatively strong signal of variable intensity in a majority of cardiocytes (Fig. 8a). Simultaneous detection of both the immunostained agalsidase protein and the stored birefringent lipid/lipopigment in a single section showed the coexistence of both the targeted enzyme and lipid storage in a very close contact in the majority of cardiocytes (Fig. 8c). More intimate contact between the targeted enzyme and the storage compartment was seen after brief extraction of the stored lipid with chloroform–acetone mixture (see “Materials and methods”; Fig. 8d). Unfortunately, the applied enzyme proved to be fixation sensitive, so that our attempts to define its relation to the storage compartment at the ultrastructural level failed. However, in semi-thin sections exposed to intensive antigen retrieval, there were numerous positive granules of the applied enzyme interspersed between autofluorescent lipopigment granules (Figs. 8e,f). Both enzyme histochemical techniques (see “Materials and methods”) showed the presence of active enzyme in the storing cardiocytes (Fig. 9a). There was also a close relation of the enzyme product to the storage lysosomal compartment better seen in simultaneous azo-coupling technique. Strong immunohistochemical signal for the enzyme protein was seen in the vascular and endocardial endothelium and in the epicardial mesothelium in both patients on ERT. Neither the enzyme protein nor its activity was detectable in cardiocytes in the Fabry postmortem heart sample without ERT (patient 3; Figs. 8b, 9b).

Fig. 8
figure 8

Immunohistochemical detection (TK 88 antibody) of α-GALA protein in cardiocytes. a Strong signal for the enzyme protein in atrial cardiocytes in patient 1 (21 months on ERT; bar 100 μm). b Absence of signal in ventricular cardiocytes in patient 3 (1 year after ERT withdrawal). Insert: signal for endogenous α-GALA in control ventricular cardiocytes (bars 100 μm). c Simultaneous immunohistochemical detection of the enzyme and of the birefringent stored lipid showing coexistence of both in same cells. Semi-crossed Nicols (bar 30 μm). d Immunohistochemical detection of the enzyme after brief lipid pre-extraction shows close relation of the enzyme to the storage compartment (bar 50 μm). e Immunohistochemical detection of the enzyme in a semi-thin section of LR-White-embedded specimen (bar 20 μm). f The stained section was simultaneously examined under UV to show lipopigment (unstained) which represents marker of the storage compartment (bar 20 μm). c, d Patient 1; e, f patient 2

Full size image
Fig. 9
figure 9

Histochemical evaluation of α-GALA activity with simultaneous azo-coupling method. a Presence of enzyme activity in storing atrial cardiocytes of patient 1 (21 months on ERT; bar 100 μm). Compare with Fig. 8. b Absence of enzyme activity in ventricular cardiocytes 1 year after cessation of ERT (patient 3). Insert shows enzyme activity in control ventricular cardiocyte (bars 50 μm)

Full size image

Other lysosomal enzymes in the cardiocytes

The storage atrial cardiocytes displayed strong and uniform induction of acid phosphatase activity (Fig. 10a). In contrast, induction of β-galactosidase activity was highly irregular (Fig. 10b). Activity of β-glucuronidase was barely detectable. Activities of the above enzymes in autopsy samples were uniformly low.

Fig. 10
figure 10

Lysosomal enzyme activities and MPR immunostaining in storage atrial cardiocytes. a High and uniform activity of acid phosphatase (aqueous medium). Insert shows activity in controls (bars 50 mm). b Heterogenous activity of β-galactosidase with remarkably low values in some cardiocytes. Insert shows activity in control cardiocytes (bars 50 mm). c Similarly heterogeneous signal for MPR (bar 30 mm)

Full size image

Signals of both MPR (Fig. 10c) and cathepsin D (not shown) in the biopsy samples were remarkably heterogeneous in storage cardiocytes ranging from barely detectable to very strong intensities.

Discussion

Studies in cell culture

Higher uptake of beta- than alpha-agalsidase reflected greater number of phosphorylated mannose moieties per beta enzyme molecule [23, 33] and presumed principal role of MPR-mediated endocytosis [18] in skin fibroblasts which are receptor-rich. The complete inhibition of cellular uptake of both agalsidases by M-6-P is in accord with this conception.

Lysosomal targeting of the applied active enzyme was double-checked. Firstly, successful targeting of enzyme therapeutics to lysosomes has been proven through co-localization with lysosomal marker cathepsin D. Evaluation of images based on calculation of correlation and co-localization coefficients pointed to slightly lower co-localization parameters for the alpha compared to the beta enzyme.

It must be pointed out that enzyme endocytosis and trafficking via early to late endosomes is a dynamic process. Therefore, only partial co-localization with lysosomal markers can be achieved under the condition of continuous enzyme supply. Final evaluation of Gb3Cer and α-GALA signals in Fig. 3i,j strongly suggested that lysosomes in fibroblast population were almost cleared of substrate after 24-h incubation with any of the enzymes.

Secondly, the delivered enzyme was shown to degrade efficiently the natural substrate (3H)Gb3Cer produced directly in lysosomes from its radiolabeled precursor (3H)Gb4Cer. The exact quantification of lysosomal (3H)Gb3Cer degradation in situ confirmed indirect observation that recombinant agalsidases degrade accumulated Gb3Cer in a dose-dependent manner [33].

In situ (3H)Gb3Cer degradation revealed lower differences between the two enzymes than in vitro assay with soluble MU-analog using the whole homogenate of agalsidase-loaded fibroblasts. This may be related to the intracellular transport and unspecified location of agalsidase in in vitro assay and to the availability of Sap B in lysosomes when supply of enzyme is high.

However, despite the different degree of agalsidase endocytosis, the cell-associated activity of both enzymes exceeded by several times the α-GALA activity of normal healthy cells (see Fig. 1), and both were found effective in cleavage of (3H)Gb3Cer. Their stability in the medium was found identical in accordance with findings in blood plasma [33].

Our experiments demonstrated effective targeting of the applied enzyme to fibroblast lysosomes using both co-localization and cleavage of its substrate locally produced from its exogenously added precursor. (3H)Gb4Cer was used for the first time as an intra-lysosomal tracer of function of recombinant agalsidases that does not increase the substrate load and can serve as a model system for evaluation of catabolic lysosomal functions and their defects.

To sum up, the ERT is effective in the cultured fibroblast system. In our view, it is because Fabry fibroblasts represent a short-lived dynamic cell population unmodified by low degree of spontaneous storage. Such cells exposed transiently to a trace amount of labeled substrate can effectively endocytose the recombinant enzyme. The enzyme is delivered to the storage compartment where it exerts its degradative function. The whole process is facilitated by easy access to the applied enzyme via MPR and by the possibility to endocytose it along the entire cell perimeter of unpolarized cell.

Studies in tissues

Our study showed effective clearing of endothelial cells by ERT corresponding to the published results [21, 7, 36]. The design of our study allows us to conclude that endothelial storage is resumed within 1 year after cessation of ERT. In the Fabry disease mouse model, the storage was resumed after several weeks of withdrawal of the enzyme [21].

In contrast to endothelia, the storage persisted in a set of cells despite the continuous ERT, e.g., in cardiocytes, vascular smooth muscle cells, pericytes, some fibroblasts, and perineurial cells, all repeatedly described to resist substantial storage reduction by ERT [12, 40, 7, 39, 43, 32, 33, 36]. We may add skin eccrine glands, hair bulb fibroblasts, and skeletal muscle. The design of the study did not allow detecting possible diminution of the persisting storage mentioned in some of the above-quoted reports. For instance, the cardiocyte storage pattern in both our cases admits partial clearance (Fig. 6b). Persistent storage in cardiocytes was accompanied by a notable increase in lipopigment which should be taken as irreversible. These findings may explain the absence of cardiac improvement in either patient on ERT (see Table 1).

It is not easy to explain the limited ERT effectiveness. There is the undisputed role of the physical barriers, represented by fibrous stroma and basal membranes, especially when multiplied around capillaries [11]. Free access of the applied enzyme to endocytotically active cell perimeter in non-polarized cultured fibroblasts and through fenestration in the liver sinusoidal endothelium to the hepatocyte basolateral pole may contribute significantly to the effective ERT [35] (also M.E. personal observation), similarly as the number of MPR known to be of high density in fibroblasts [42, 18]. However, the density of the latter was not shown to correlate with the level of the applied enzyme uptake [29]. In polarized storage cells (epithelia, e.g., in the sweat glands, respiratory epithelium, etc.), the applied enzyme uptake is restricted to the basolateral pole. In case of renal tubules, both poles may participate to different degree in various parts of the nephron [38]. In the case of endothelial cells, the reproducible effectiveness of ERT may be explained by the barrierless access of the enzyme to the apical pole active in endocytosis [37].

Our findings point to further potentially important facts. Before all, it is the presence of the applied active enzyme inside the storage cardiocytes in patients under ERT (not in samples from patient 3 with ERT terminated). We base this crucial finding on the following data. No enzyme protein was immunohistochemically detectable in cultured mutant fibroblasts of both patients. As the value of enzyme activity in wild-type fibroblasts is much higher (approximately fivefold) than in the heart, it excludes presence of detectable amount of endogenous mutant enzyme in atrial cardiocytes of the Fabry patients studied. This is in accord with negative results in the heart from patient 3 with ERT terminated (see Fig. 8b). The histochemically demonstrable α-GALA activity in cardiocytes of ERT-treated patients (and not in patient 3) fits perfectly with the conclusion (Fig. 9). Similarly, increase in activity was shown by biochemical assay in one auricular sample available. We would also like to stress that in our studies, we used cryostat sections because the enzyme was found to be remarkably fixation-sensitive (see below). Therefore, sensitivity of our detection of the applied enzyme was higher than that in previous reports [20, 29].

The sensitivity of the applied enzyme to aldehyde fixation prevented its localization inside storage lysosomes at the electron microscope level. However, results in both cryostat and LR-White-embedded sections (Fig. 8a,d–f) indicate at least partial successive delivery of the enzyme. Its quantity may play a role, as the heart in Fabry disease may express storage even in cases with relatively high residual activity [30]. We would like to hypothesize that there may be modification of membranes of the endosomal/lysosomal compartment, induced by longstanding storage, especially in cells with low regeneration potential, cardiocytes being an ideal example. Such modification could lead to decreased effectiveness of the contact of the storage compartment with the endocytotic delivery of the applied recombinant enzyme. Decrease in MPR in many of the storage cardiocytes suggests a changed proportion of lysosomes to late endosomes, and this may result in altered fusion, inevitable for delivery of the applied enzyme cargo to the latter [8, 25]. It might also indicate disturbed contact even with the Golgi apparatus, which might explain low activities of β-glucuronidase, β-galactosidase, and cathepsin D in the cardiocyte storage compartment. Further, parallel deposition of lipopigment (see above) points to irreversible modification of storage lysosomes. Any relevant comparison with the mouse model of Fabry disease is difficult, as the mouse cardiocytes do not seem to display storage regularly [21]. Nothing is known about the effective interaction of the applied agalsidase with Sap B in Fabry disease. The disorder, featured by densely packed lysosomal lipid membranes, differs substantially from the well-defined physiologic conditions [22]. Decreased amount of Sap B in Fabry disease [28] may aggravate the situation. In sum, one of the critical factors in ERT of Fabry disease may, thus, be the degree of storage and physical status of the stored Gb3Cer and availability of Sap B. All this raises an important question whether the effective targeting of the enzyme to the storage cell is the only prerequisite for effective ERT.

The last factor awaiting evaluation is the role of storage cell turnover in Fabry disease (and in other lysosomal storage disease). In the case of endothelial cells in Fabry disease, preliminary results (J. Bultas, personal communication) showed its increase. Therefore, the positive effect of ERT might be also explained as a replacement effect of storage cells by the endothelial cell progenitors [1, 5, 17] that may easily endocytose the applied enzyme and thus prevent storage. This mechanism does not come into consideration in the case of post-mitotic storage cells, typically cardiocytes. Taken together, our observations speak in favor of the strategy of early treatment to prevent development of storage and of irreversible modulation of post-mitotic cells or of cells with low regeneration potential (cardiocytes, smooth muscle cells) and of their lysosomes. This would be the most effective prevention of irreversible tissue changes. Studies on mouse model which would compare effectiveness of early ERT (starting immediately after birth) with its delayed application in adulthood are desirable.

Conclusions

We suggest that future studies should focus on defining the differences between the storage cell, the storage compartment and their wild-type counterparts, and on possible replacement of cell populations with higher turnover by their enzyme supplied precursors. Then, enzyme replacement treatment should probably be better conceptualized as an enzyme preventive therapy (see also [19]).