Elevated globotriaosylsphingosine is a hallmark of Fabry disease

The consideration that a pathogenic, potentially vasoactive metabolite might be present in plasma from Fabry patients prompted us to investigate the presence of lyso-Gb3 (deacylated Gb3). Hitherto the existence of lyso-Gb3 has not been documented. However, various deacylated glycosphingolipids have been identified in lysosomal glycolipid storage diseases (see discussion in ref. 18 for an overview). For example, in Gaucher disease, glucosylsphingosine, the deacylated form of the accumulating glucosylceramide, is known to be increased as well (18, 19). The same holds for Krabbe disease in which galactosylsphingosine or psychosine (deacylated galactosylceramide) is elevated (20–23). It has been proposed that such lysosphingolipids play a role in the pathogenesis of these disorders (18–24). lyso-Gb3 is a cationic amphiphile with a large polar sugar moiety, rendering it relatively hydrophilic and water-soluble (see Fig. 1 Inset). Its presence therefore may have escaped attention during lipid analyses focusing on constituents of the organic phase of extractions. By using normal plasma spiked with a standard of pure lyso-Gb3, an appropriate extraction procedure for lyso-Gb3 was developed. lyso-Gb3 was recovered quantitatively from the upper phase of the extraction according to the Bligh and Dyer procedure (25), and it was purified by repeated extraction of the upper phase with 1-butanol (see Materials and Methods). The free amine moiety of lyso-Gb3 was fluorescently labeled with ortho-phthaldialdehyde according to a procedure used earlier for labeling of other lysosphingolipids (26). Labeled lysosphingolipids were separated by high-performance liquid chromatography and quantitatively analyzed (see Materials and Methods). Plasma samples from symptomatic Fabry patients were found to contain large quantities of a compound with a retention time identical to that of authentic lyso-Gb3 in the spiked plasma samples (Fig. 1). This compound could not be detected in plasma from control individuals. The identity of the abnormal structure as lyso-Gb3 was further verified. Incubation of the plasma sample with recombinant α-galactosidase A resulted in a shift to a value coinciding with that shown by lactosylsphingosine. Tandem mass spectrometry analysis confirmed the identity of the abnormal structure as lyso-Gb3 [see supporting information (SI) Text]. Examination of chromatograms of plasma from symptomatic patients pointed to the presence of another, more polar, abnormal compound X with a retention time of 16.2 min (see Fig. 1, Fabry plasma). Its concentration is ≈5-fold lower than that of lyso-Gb3. The shift in its peak retention by 1 min after incubation with α-galactosidase A indicates that the compound also contains a terminal α-galactoside moiety, which suggests that this compound is presumably also a lysoglycolipid and genuine storage product in Fabry disease. Based on its chromatographic behavior, it seems unlikely that the compound is sphingosinedigalactoside, the deacylated form of ceramidedigalactoside. The latter lipid has been found to be increased in kidney of patients with Fabry disease (27). The low abundance of the compound did not yet allow identification of its structure by tandem mass spectrometry. The ratio of lyso-Gb3 to the other abnormal structure in Fabry plasma specimens was remarkably constant in one individual but varied slightly among different patients.

Plasma levels of lyso-Gb3, Gb3, lactosylceramide, glucosylceramide, and ceramide were determined in members of a large Fabry pedigree (see Materials and Methods). The status of all individuals was confirmed by enzyme analysis and genotyping (mutation encoding amino acid substitution D136Y). Table 1 shows that plasma specimens of the unaffected males (father and two normal sons) showed lyso-Gb3 levels below the detection level of 10 nM and normal levels of Gb3 (1.3–1.8 μM). In sharp contrast, plasma from all five affected sons showed markedly elevated lyso-Gb3 (195–407 nM) and elevated Gb3 (4.9–10.0 μM) levels. In female members of the family, no clear increases in plasma Gb3 were noted, independent of the genetic status of individuals. However, lyso-Gb3 was clearly increased in the symptomatic aunt, mother, and daughter (22–76 nM), and it was below the detection limit in the asymptomatic daughter and the unaffected siblings.

Plotting the plasma concentration of lyso-Gb3 against that of Gb3 for individual members of a Dutch Fabry cohort (see SI Fig. 4) indicated that the elevation of plasma lyso-Gb3, like that of Gb3, is most prominent in hemizygotes. There is a strong correlation between the concentration of plasma lyso-Gb3 and Gb3 in hemizygous patients (n = 22, ρ = 0.75, P < 0.001). In sharp contrast to the patients with classic Fabry disease manifestations, individuals carrying the mutation encoding R112H α-galactosidase A, which is known to be associated with an atypical manifestation of symptoms (28), showed normal Gb3 levels and very low lyso-Gb3 levels.

The relationship of individual plasma lyso-Gb3 concentrations with age and the commonly used Mainz Severity Score Index (MSSI) of disease severity (29) was determined (see also SI Fig. 5). In the case of hemizygotes, no correlation was noted between plasma lyso-Gb3 with regard to age or MSSI. Very high lyso-Gb3 concentrations were detected in plasma specimens of all classically affected Fabry hemizygotes. Of note is that specifically young boys with little disease manifestations already showed high plasma lyso-Gb3. Also in the case of a newborn hemizygote, a prominently increased lyso-Gb3 (81 nM) was demonstrable. No significant correlation of plasma lyso-Gb3 in hemizygotes with left ventricular mass or other disease signs was observed.

A clear exception among the hemizygotes were individuals carrying the R112H mutation. Two atypically affected adult male patients with the R112H mutation (one with polyneuropathy as the only symptom and the other with isolated renal disease, possibly also related to hypertension) showed relatively low lyso-Gb3 concentrations. Their MSSI scores were 22 and 38, respectively. Three very mildly affected, middle-aged hemizygotes for the R112H mutation (MSSI of 2, 2, and 4) also showed very low plasma lyso-Gb3 levels.

Examination of female heterozygotes revealed that, with increasing age, lyso-Gb3 tends to be higher, although not significantly (P = 0.08). In females, lyso-Gb3 also tends to be more elevated with increasing MSSI score (Spearman's ρ = 0.49, P = 0.05). A clear exception in this respect was a 13-year-old heterozygote showing relatively high plasma lyso-Gb3 in combination with a low MSSI of 6. The patient, now receiving enzyme-replacement therapy (ERT), suffered from acroparesthesias on exercise and showed a white matter lesion on the MRI of the brain. The latter is relatively rare in young hemizygotes and suggests a rapidly progressive disease manifestation. Analysis of neuropathology and renal and cardiac involvement revealed that only left ventricular mass in Fabry heterozygotes correlated with plasma lyso-Gb3 concentration (Spearman's ρ = 0.68, P < 0.05). In conclusion, in Fabry hemizygotes marked increases in lyso-Gb3 are noted already at young age, with the exception of atypical carriers of the R112H mutation. In the case of female heterozygotes, a larger series of patients has to be analyzed to establish the relation of plasma lyso-Gb3 with age and disease manifestations.

The plasma concentrations of lyso-Gb3 and Gb3 were determined in mice. Table 2 shows that, at 20 weeks of age, plasma lyso-Gb3 concentrations are clearly less elevated in female heterozygous mice compared with female mice homozygous for the α-galactosidase A gene disruption and male hemizygous mice. Plasma lyso-Gb3 was found to be prominently increased at very young age in hemizygous mice, being 476, 510, and 614 nM at 70, 138, and 243 days, respectively.

The presence of lyso-Gb3 in tissues was studied in 20-week-old hemizygous mice. In parallel, concentrations of Gb3 and some other glycosphingolipids were determined. Dramatic increases in Gb3 and lyso-Gb3 concentrations were noted in most tissues of hemizygous mice (Table 3). Very high concentrations of lyso-Gb3 were observed in liver and intestine, clearly exceeding that in plasma. This finding suggests that lyso-Gb3 is either actively formed or preferentially trapped in these organs.

The structure of lyso-Gb3 predicts a high degree of solubility in water. Consistent with this notion, lyso-Gb3 is largely recovered in the methanol/water phase upon Bligh and Dyer extraction (see also Materials and Methods). Given the high concentration of lyso-Gb3 in plasma of symptomatic Fabry patients, we investigated its distribution among lipoproteins. In contrast to Gb3, which is completely lipoprotein-bound, only a very small percentage of lyso-Gb3 was found to be associated with lipoproteins. The vast majority of lyso-Gb3 (>95%), was recovered in the lipoprotein-free fraction, presumably associated to albumin, as is the case with other hydrophilic lipids such as fatty acids. Unfortunately, the possibility that lyso-Gb3 is secreted into urine can presently not be confirmed nor excluded with certainty. The detection of lyso-Gb3 in urine samples with the present procedure is hampered by the presence of interfering substances in this fluid.

Given the structural similarity of lyso-Gb3 to Gb3, its capacity to inhibit α-galactosidase A was investigated. First, the effect of lyso-Gb3 on activity of recombinant α-galactosidase A toward 4-methylumbeliferyl-α-galactoside was determined. Potent competitive inhibition (K_i = 34 μM) was observed to be identical for the two tested recombinant α-galactosidase A preparations. The activity of recombinant α-galactosidase A preparations toward Gb3 as substrate also was pronouncedly inhibited by the presence of lyso-Gb3. IC₅₀ values of ≈20 μM were observed.

Cultured fibroblasts from normal subjects accumulate Gb3 when exposed to lyso-Gb3. An illustrative example of this is depicted in SI Fig. 6. lyso-Gb3 at concentrations ≈100 nM, similar to those encountered in plasma of symptomatic Fabry patients, is sufficient to increase cellular Gb3. Incubation with 500 nM lyso-Gb3 resulted in an increase of cellular Gb3 from 2.03 nmol/mg protein to 3.16 nmol/mg protein within 9 days. This effect is presumed to be caused by inhibition of α-galactosidase A activity, but it cannot be excluded that, after uptake, some lyso-Gb3 is converted to Gb3. An ≈200-fold more potent inhibitory capacity of lyso-Gb3 in vivo compared with in vitro could be explained by trapping of the membrane-permeable lyso-Gb3 in the acidic lysosome. Its pK_a predicts accumulation of lyso-Gb3 in lysosomes by a factor of ≈200, coinciding with the noted difference in apparent IC₅₀ between in vitro and in vivo experiments.

A characteristic feature of Fabry disease is vascular remodeling. This phenomenon is reflected in increased carotid intima-media thickness as the result of proliferation of smooth muscle cells (29, 30). To establish whether lyso-Gb3 exerts itself deleterious effects on the vasculature, the influence of the compound on smooth muscle cell proliferation was studied. Fig. 2 shows that low (50–100 nM) concentrations of lyso-Gb3, similar to those observed in plasma of symptomatic Fabry patients, markedly promote proliferation of smooth muscle cells in culture. In sharp contrast, the structurally related compounds like Gb3 and lactosylsphingosine have no stimulating effect on smooth muscle proliferation.

We did not observe any stimulation by lyso-Gb3 of proliferation of fibroblasts (data not shown). These findings suggests that the elevation of lyso-Gb3 in plasma of Fabry patients may underlie the observed increase in intima-media thickness.

lyso-Gb3 can act as substrate for α-galactosidase A, resulting in its conversion to lactosylsphingosine. Compared with Gb3, the degradation of pure lyso-Gb3 in the test tube by recombinant α-galactosidase A preparations was found to be at least 50-fold less efficient (data not shown). At neutral pH, recombinant α-galactosidases A hardly hydrolyze lyso-Gb3 because of their acid pH optimum (see SI Text).

The effect of prolonged ERT on plasma lyso-Gb3 was studied for a selected subset of patients. Fig. 3 shows representative examples of the observed responses in plasma lyso-Gb3 to ERT with commercially available enzyme preparations agalsidase alfa (Shire) and agalsidase beta (Genzyme). Patients receiving either agalsidase alfa (Fig. 3 Left) or agalsidase beta (Fig. 3 Right), without developing antibodies, showed a reduction of lyso-Gb3 within 1 year, although a complete normalization was not observed. Some of the male hemizygotes lacking α-galactosidase A develop neutralizing antibodies upon ERT (31–33). Fig. 3B show the responses in plasma lyso-Gb3 and Gb3 to ERT for patients that had developed antibodies after receiving agalsidase alfa (Left) and agalsidase beta (Right). Reduction of plasma lyso-Gb3 in these individuals was relatively poor. At present, no statements can be made about the relative efficacy of the two commercial recombinant α-galactosidase A preparations or the influence of enzyme dose and antibody formation. A detailed investigation with carefully matched patients will be required to answer these important questions.

Selection from the cohort of patients with Fabry disease at the Academic Medical Center in Amsterdam was carried out for analysis of (glyco)sphingolipids and disease manifestations before the initiation of ERT. This selection included 17 classically affected hemizygotes (age range 6–45 years), 17 heterozygotes (age range 11–52 years), two newborns (1 hemizygote and 1 heterozygote), and 5 atypical hemizygotes with the R112H mutation. In addition, 10 control subjects (five males and five females) were included. A large Fabry family was part of this study, including an affected mother and her affected sister, a healthy father, five hemizygous sons, three heterozygous individuals, and five healthy siblings. Ten patients (eight hemizygotes and two heterozygotes) received either 0.2 mg/kg agalsidase alfa (Replagal, Shire) or 0.2 mg/kg or 1.0 mg/kg agalsidase beta (Fabrazyme, Genzyme). The diagnosis of Fabry disease was confirmed by demonstration of deficient activity of α-galactosidase A in leukocytes and genotyping. Blood samples were obtained with informed consent for investigations. Skin fibroblasts from Fabry patients (n = 2) and control individuals (n = 2) were cultured as described in ref. 41.

A complete medical history was obtained, and physical examination was performed on all patients. Clinical assessments are described in detail in SI Text. Briefly, as described in ref. 7, renal function was analyzed and brain magnetic resonance imaging, electrocardiograms, cardiac ultrasound investigations, hearing tests, and assessments of pain were carried out. Overall severity of disease was assessed by using the MSSI (9).

Correlations were calculated for hemizygotes and heterozygotes separately, with the exclusion of the atypical cases carrying the R112H mutation and the two newborns. Correlations between variables were described with the use of Spearman rank correlation coefficients. P values of <0.05 were considered significant.

C3HeB/FeJ and B6;129-Gla^tm1Kul (stock number 003535) strains were obtained from The Jackson Laboratory. Animals were fed a commercially available laboratory diet (RMH-B; Hope Farms, Woerden, The Netherlands).

Smooth muscle cells were obtained from Carlie de Vries (Amsterdam). Smooth muscle cells were seeded in 24-well plates at 1–4 × 10⁴ cells per well and reached 60–70% confluence after 24 h. Smooth muscle cells were made quiescent by incubation for 24 h in FBS-free medium and then were stimulated for 24 h with 5% (vol/vol) FBS. Subsequently, cells were labeled for 4 h with 0.25 μCi per well [methyl-³H]thymidine (Amersham Biosciences). Incorporated [³H]thymidine was precipitated for 30 min at 4°C with 10% (wt/vol) trichloroacetic acid, washed twice with 5% (wt/vol) trichloroacetic acid, and dissolved in 0.5 M NaOH (0.5 ml per well), and radioactivity was measured by liquid-scintillation counting.

Agalsidase alfa (Replagal, Shire) and agalsidase beta (Fabrazyme, Genzyme) were purchased from the local hospital pharmacy.

The enzymatic activity of α-galactosidase A was determined with fluorogenic 4-methylumbelliferyl-α-galactoside as substrate (41). Enzyme preparations were incubated at 37°C with 3.5 mM substrate in 100 mM citrate/200 mM phosphate buffer (pH 4.6) at 37°C. Alternatively, enzymatic activity was measured toward Gb3 in the presence of 0.5% (wt/vol) sodium taurocholate in 50 mM sodium acetate buffer (pH 4.2). The formation lactosylceramide was determined by high-performance thin layer chromatography (HPTLC) analysis (12).

Fasted EDTA plasma was obtained from a patient with Fabry disease and an age- and gender-matched control subject. Lipoproteins were separated by gradient ultracentrifugation according to the procedure of Redgrave et al. (42).

Gb3, lactosylceramide, glucosylceramide, and ceramide were measured by using high-performance liquid chromatography as described in ref. 26. The mean Gb3 concentration in plasma from controls was 1.7 (0.5) μmol/liter (n = 40), mean concentration of glucosylceramide in plasma from controls was 6.3 (1.9) μmol/liter (n = 40), and mean concentration of ceramide in plasma from controls was 9.0 (2.3) μmol/liter (n = 40) (26).

To 100 μl of plasma or tissue homogenates, 50 μl of water was added, and the lipids were extracted with 900 μl of chloroform/methanol 1/2 (vol/vol). The extract was centrifuged for 10 min at 14,000 × g, and the supernatant was removed. To the supernatant, 0.3 ml of chloroform and 0.45 ml of water were added, and after mixing, the phases were separated by brief centrifugation. The upper phase was removed, and the lower chloroform phase was extracted once more with 1.2 ml of methanol/water 1/1 (vol/vol). The combined upper phases were dried and taken up in 1 ml of water, and the water phase was extracted twice with 1 ml of water-saturated 1-butanol. lyso-Gb3 was recovered from the butanol phase with an overall recovery of >90%. The butanol phase was dried and dissolved in 120 μl of methanol. Of this solution, 50 μl was taken to dryness, and the lyso-Gb3 was derivatized with 25 μl of o-phtaldialdehyde (OPA) reagent (5 mg of OPA, 0.1 ml of ethanol, 5 μl of 2-mercaptoethanol, and 10 ml of 3% boric acid, pH 10.7). The OPA-derivatized lyso-Gb3 was separated by HPLC and identified by fluorescence detection as described in ref. 26. Quantification was by external standardization after standard addition of authentic lyso-Gb3 (Sigma-Aldrich) to normal plasma so that increasing concentrations from 0 to 1,000 nmol/liter plasma were obtained. The response was strictly linear with the lyso-Gb3 concentration, and the limit of quantification was 10 nmol/liter. Peak identification was by comparison of the retention time with that of authentic lyso-Gb3.