Article Text

Original article
Plasma globotriaosylsphingosine in relation to phenotypes of Fabry disease
  1. Bouwien E Smid1,
  2. Linda van der Tol1,
  3. Marieke Biegstraaten1,
  4. Gabor E Linthorst1,
  5. Carla E M Hollak1,
  6. Ben J H M Poorthuis2
  1. 1Division Endocrinology and Metabolism, Department of Internal Medicine, Amsterdam lysosome center ‘Sphinx’, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands
  2. 2Laboratory of Genetic Metabolic Diseases, Amsterdam lysosome center ‘Sphinx’, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands
  1. Correspondence to Dr Ben J H M Poorthuis, Laboratory of Genetic Metabolic Diseases, Academic Medical Center, Room F0-220, Meibergdreef 9, Amsterdam 1105 AZ, The Netherlands; b.j.poorthuis{at}


Background Fabry disease (FD), a lysosomal storage disorder caused by α-galactosidase A (GLA) gene variants, has a heterogeneous phenotype. GLA variants can lead to classical FD, an attenuated non-classical phenotype, or no disease at all. This study investigates the value of plasma globotriaosylsphingosine (lysoGb3) to distinguish between these groups. This is of particular importance in the diagnosis of individuals with a GLA variant and an uncertain diagnosis of FD, lacking characteristic features of classical FD.

Methods Subjects with GLA variants were grouped as classical, non-classical, uncertain or no FD, using strict phenotypical, biochemical and histological criteria. Plasma lysoGb3 was assessed by LC/MS/MS (normal ≤0.6 nmol/L).

Results 154 subjects were grouped into classical (38 males (M), 66 females (F)), non-classical (13M, 14F), uncertain (5M, 9F) or no FD (6M, 3F). All subjects with a classical phenotype had elevated lysoGb3 values (M: range 45–150, F: 1.5–41.5). LysoGb3 values in patients with a non-classical phenotype (M: 1.3–35.7, F: 0.5–2.0) were different from healthy controls (M: p<0.01, F: p<0.05), but females overlapped with controls. In the no-FD group, lysoGb3 was normal.

Conclusions LysoGb3 is a reliable diagnostic tool to discern classical FD from subjects without FD. This study suggests that the same applies to patients with a non-classical phenotype. LysoGb3 values of female patients overlap with controls. Consequently, in uncertain cases, increased lysoGb3 values are very suggestive for FD, but normal values cannot exclude FD. Confirmation in larger cohorts and data on the specificity of small lysoGb3 increases are necessary.

  • Fabry disease
  • Globotriaosylsphingosine
  • Genetic screening
  • Diagnosis
  • Phenotype

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Fabry disease (OMIM 301500; FD) is an X-linked lysosomal storage disorder caused by variants/mutations in the α-galactosidase A (GLA) gene. To date, more than 700 variants have been reported.1 The classical FD phenotype is characterised by a markedly reduced or absent activity of the lysosomal hydrolase α-galactosidase A (αGalA, E.C. Subsequently, globotriaosylceramide (Gb3), globotriaosylsphingosine (lysoGb3) and other glycosphingolipids with a terminal α-1,4-galactose accumulate in lysosomes. Characteristic features include neuropathic pain with childhood onset, clustered angiokeratoma, cornea verticillata and complications of the heart, kidney and brain. Since enzyme replacement therapy (ERT) became commercially available, screening for FD in newborns2–6 and high-risk groups (patients with chronic kidney disease (CKD), stroke or left ventricular hypertrophy (LVH)7–13), as well as individual case findings,14 has resulted in the identification of individuals with variants in the GLA gene of unknown clinical significance (genetic variant of unknown significance, GVUS).15 It appears that approximately 17% of these individuals identified by screening presents with a classical FD phenotype, while the majority has an uncertain diagnosis of FD.15 In these uncertain cases, characteristic clinical and biochemical features of FD are lacking and the presenting symptoms, such as CKD or LVH, are not specific for FD, but might have a more common cause. These individuals may have GLA variants leading to an attenuated FD phenotype (also known as non-classical FD or late onset variants), but may also have neutral, non-disease-causing GLA variants. Therefore, misdiagnoses are a potential risk of screening and may cause considerable distress for patients and family members, and lead to inappropriate initiation of costly ERT.

To improve the diagnostic process of patients with an uncertain (genetic) diagnosis of FD, the ‘Hamlet study’ was initiated.16 In this study, international FD experts agreed that non-invasive procedures (such as imaging characteristics) are often insufficient to confirm or exclude FD in individuals with an uncertain diagnosis of FD.17–20 In these uncertain cases, a biopsy of an affected organ (ie, kidney or heart) showing characteristic lamellar inclusion bodies on electron microscopy (EM) assessment confirms the diagnosis.18 ,19 However, biopsies are invasive and not feasible in all cases, for example, in individuals presenting with isolated stroke. We therefore investigated whether plasma lysoGb3 could aid in the diagnostic process, potentially avoiding the need for biopsies.

LysoGb3, a deacylated form of Gb3 has been identified as a storage product in FD.21 Interestingly, in addition to lysoGb3, at least six other lysoGb3 analogues with varying base composition have been identified in plasma and urine.22 ,23 While the concentration in urine of these lysoGb3 analogues is often higher than that of lysoGb3, in plasma lysoGb3 is the major isoform present.22 Plasma lysoGb3 is a sensitive marker for FD,24 ,25 and its concentration in plasma is much higher than in urine. Despite its high water solubility, lysoGb3 is hardly cleared by the kidney.26 Although male patients with FD with a classical phenotype can easily be identified by very high plasma lysoGb3 values,21 ,24 ,26 the diagnostic value of smaller increases of plasma lysoGb3 values in the presence of a GVUS in the GLA gene is not yet clear. This is partly caused by the lack of sensitivity of the high-performance liquid chromatography (HPLC) methods originally used to quantify lysoGb3. With more sensitive tandem mass spectrometric (LC/MS/MS) assays, lysoGb3 can now be quantified at very low levels, as observed in women, non-classical patients with FD and healthy controls.26–28 Niemann et al29 reported a cut-off value for plasma lysoGb3 to detect clinically relevant, often classical FD, although confirmation by histology was not used to discern FD from no FD. Other studies report on small increases of plasma lysoGb3 in combination with characteristic storage on EM in heart and/or kidney biopsies14 ,29–36 supporting the hypothesis that an increased lysoGb3 may indicate FD. By contrast, normal values of lysoGb3 were considered to indicate a neutral, non-pathogenic variant in the GLA gene.14 ,37–39

The aim of the current study was to investigate if plasma lysoGb3 is able to distinguish individuals with a non-classical FD phenotype from those without FD. Therefore, we measured plasma lysoGb3 in the Dutch FD cohort comprising individuals with a classical and non-classical FD phenotype, as well as subjects with a GVUS in whom FD was excluded.



This is a retrospective cohort study including all subjects referred to the Academic Medical Center (AMC) in Amsterdam for FD or a possible diagnosis of FD. Data were retrieved as part of regular clinical care. At the AMC, informed consent for collecting clinical data and blood samples for research purposes has been obtained for all subjects with FD. Ethics approval of the study protocol was not required because of the observational nature of the study.

We included all adult subjects (age ≥16 years) with any variant in the GLA gene, defined as all alterations in the GLA gene being either pathogenic, a neutral variant or a GVUS. Subjects were classified into four groups: classical FD, non-classical FD, uncertain or no FD. We excluded subjects with insufficient data to classify their phenotype and those with unavailable plasma lysoGb3 values.

Groups were defined according to previously described diagnostic criteria,15 ,19 (see online supplementary table S1). Subjects were grouped as classical FD when fulfilling the following criteria: αGalA enzyme activity in leucocytes <5% of the mean reference value (men) and a GLA variant and either one or both of the following criteria: ≥1 of the described characteristic features of FD (neuropathic pain, cornea verticillata, clustered angiokeratoma) or a family member with a definite diagnosis of classical FD. Plasma lysoGb3 >50 nmol/L and/or Gb3 >2.9 nmol/mL are part of the original diagnostic criteria, but for the purpose of this study, plasma lysoGb3 and Gb3 values were not used to group subjects. Non-classical FD was defined as subjects presenting with an FD-like sign or symptom that is not specific for FD, such as LVH, who did not fulfil the diagnostic criteria (see above), but in whom a biopsy of an affected organ was performed, demonstrating storage pattern characteristic for FD on EM in the absence of drugs inducing an FD-like storage pattern. Such a biopsy outcome could also have been in a family member who was carrying the same GLA variant. When biopsies were not available, subjects were grouped as uncertain. Subjects in whom a biopsy was performed that did not demonstrate a characteristic storage pattern, as well as subjects with the p.D313Y variant (generally considered as a neutral GLA variant40 ,41), were grouped as no FD.

Clinical assessments consisted of the medical history, physical examination, cardiac ultrasound, cardiac MRI, brain MRI, ECG and plasma creatinine. LVH was defined as an interventricular septum thickness >12 mm in diastole on cardiac ultrasound. Glomerular filtration rate was estimated with the CKD Epidemiology Collaboration formula.42

Biochemical assessments

αGalA activity was measured in leucocytes43 in a minority without inhibition of α-N-acetyl-galactosaminidase.44 For comparison of αGalA activity between laboratories, values were presented as the percentage of the mean of the reference value. If no mean value was available, the middle of the reference range was used.

LysoGb3 was assessed in plasma by LC/MS/MS as previously described by Gold et al.26 Normal reference values (≤0.6 nmol/L) were established from plasma samples obtained from 20 healthy volunteers (21–60 years of age, 10 men and 10 women).24 No differences were found between lysoGb3 concentrations in healthy males and females, and the lysoGb3 values were always ≤0.6 nmol/L. The analytical characteristics of this method, such as intra-assay and interassay variations, limit of detection and limit of quantification have been described in detail by Gold et al.26 Plasma Gb3 (N<2.46 nmol/mL, n=16 healthy subjects) was assessed by HPLC as described earlier by Groener et al.45 Over time, some modifications of the original plasma Gb3 HPLC method were made. To compare values between methods, mixed models regression analysis was applied (with an adequate fit, R2 0.95), and older Gb3 values could be converted using the formula y=−0.0317+0.665334×(older plasma Gb3 value). The most recent value was selected for which corresponding clinical data were available. In case patients were eligible for treatment with ERT, the last lysoGb3 and Gb3 value before initiation of treatment was used. To assess whether lysoGb3 remains stable over time, data from subjects with a classical and non-classical phenotype with at least two baseline samples were analysed.


Statistical analyses were performed using SPSS V.20 (IBM, Chicago). Continuous data are expressed as median (range). Data are stratified by gender and disease groups as described above. For differences between groups, non-parametric tests were used (Mann–Whitney U). Two-tailed p values <0.05 were considered significant.



Records from 185 adults with a GLA variant were available. For seven subjects, data were insufficient to classify them into the predefined groups; another 24 subjects were excluded due to unavailable lysoGb3 data. In total, 154 subjects were included and grouped into classical, non-classical, uncertain or no FD according to the aforementioned criteria (see also online supplementary table S1). Baseline characteristics are presented in table 1 (for details of individual subjects, see online supplementary tables S2 and S3). All men in the classical FD group had an αGalA activity below 5% and ≥1 characteristic sign or symptom (apart from 1 male subject for whom enzyme activity was missing, who had a classical phenotype and multiple family members with a classical FD phenotype, confirmed with the above criteria). Eighty percent of the women in the classical group had ≥1 characteristic sign or symptom, while 20% was grouped as such because of a family member with a classical FD phenotype.

Table 1

Baseline characteristics

A few exceptions were made to the classification criteria. Four subjects (see online supplementary table S3: subject 5,11 (male); 5,12 (male); 5,15 (female) and 6 (male)) in the non-classical group had atypical non-diffuse corneal abnormalities, very different from the diffuse cornea verticillata observed in subjects with a classical phenotype. Because other FD characteristic signs or symptoms such as neuropathic pain and angiokeratoma were absent, these four subjects were grouped as non-classical.

Seven families were included in the non-classical group. These families were referred to the AMC because at least 1 family member had a hypertrophic cardiomyopathy (HCM) (4 families), kidney failure (2 families) or axonal polyneuropathy (1 family14). Eight patients were classified as non-classical on the basis of histological confirmation of the diagnosis in other patients with the same GLA variants (R112H n=3,14 I319T n=5).

Subjects in the uncertain group consisted of six families who were referred to our centre because at least 11 family member experienced kidney failure (one family), HCM (three families), dilated cardiomyopathy (one family) or small-fibre neuropathy (one family).

Finally, the subjects in whom FD was excluded consisted of five families who were referred because of HCM (three families), small-fibre neuropathy (one family) or stroke (one family) (for baseline characteristics of individual subjects, see online supplementary table S3). Table 1 summarises baseline characteristics of all included subjects.

Plasma globotriaosylsphingosine

Figure 1A and table 1 demonstrate that plasma lysoGb3 values differ between FD subjects (both male and female subjects with the classical and non-classical phenotype) and controls (p<0.01 for all separate groups vs controls). There was no overlap in lysoGb3 values between men and women with a classical phenotype or between men with a classical and a non-classical phenotype. All men and women with a classical phenotype and men with a non-classical phenotype had higher plasma lysoGb3 values than controls (ie, there was no overlap). LysoGb3 values of non-classical female subjects showed some overlap with control values: three out of 14 women with a non-classical phenotype had normal lysoGb3 values although they were close to the upper limit of the normal range. Interestingly, lysoGb3 values seem to cluster per GLA variant. Some variants only caused marginally increased lysoGb3 values (figure 1B, C).

Figure 1

(A) Plasma globotriaosylsphingosine (lysoGb3) levels grouped by phenotype and gender on a logarithmic scale. In all graphs, dotted lines refer to the upper limit of normal defined as 2 SDs above the mean in healthy controls (N≤0.6 nmol/L, mean 0.4 nmol/L, SD 0.1, n=20). Male subjects (M) are depicted as black dots, female subjects (F) are depicted in white squares, horizontal line per group represent the median group value, ***p<0.01. Of note, plasma lysoGb3 in classical M and F subjects differed significantly from healthy controls (p<0.01). (B) Clustering of plasma lysoGb3 level per GLA variant in men with a non-classical Fabry disease (FD) phenotype, and (C) in women with a non-classical FD phenotype. This effect was not observed in classical male and female subjects (data not shown).

In the uncertain group, minor increases of lysoGb3 values were found in some subjects with the p.P60L and the p.L106F GLA variants. A normal value was found for patients with the p.A143T and p.T385A variants. Subjects in the no-FD group, with an available biopsy excluding FD, or those with a p.D313Y GLA variant had normal lysoGb3 values.

In 51 subjects (38 classical FD (five men) and 13 non-classical FD (six men)), more than one pretreatment lysoGb3 value was available (median number of samples 2, range 2–6). Over a median follow-up of 1.4 year (range 0.1–12.8) we did not observe an increase of lysoGb3 in untreated adults with a classical or non-classical FD phenotype (data not shown).

Plasma globotriaosylceramide

Plasma Gb3, depicted in figure 2, was elevated in 92% (34/37) of men with a classical phenotype and in 4.5% (3/66) of classically affected women. All other groups had normal plasma Gb3 levels.

Figure 2

Plasma globotriaosylceramide levels divided by phenotype and gender (male (M) and female (F)); horizontal line per group represents the median group value. The dotted line refers to the upper limit of normal defined as the upper limit of the healthy control range 2.46 nmol/mL.


With an ultrasensitive LC/MS/MS assay in a well-characterised cohort of subjects with a GLA variant, we demonstrated that lysoGb3 is a reliable diagnostic tool to discern classical patients with FD, both male and female subjects, from patients without FD. Plasma Gb3 can only reliably identify males with a classical FD phenotype. With these data we confirm previous studies, consistently demonstrating elevations of lysoGb3 in classical FD using an HPLC method21 ,24 and subsequent studies using a LC/MS/MS method.24 ,29 ,34 In the current study, a lysoGb3 value above 45 nmol/L predicts a classical FD phenotype in men. In women, such a cut-off value could not be generated, because values of women with a classical and non-classical phenotype demonstrated an overlap. Additionally, we found that lysoGb3 values in adults did not increase over time, which seems to indicate that differences in lysoGb3 between groups are primarily related to phenotype, and not to age.

In our study, specific attention was given to lysoGb3 values in those patients with a GLA variant of unknown significance who were classified as non-classical, uncertain or as having no FD, based on strict criteria. We found that plasma Gb3 values could not discern these groups. However, moderately increased values of lysoGb3 could discern men with a non-classical phenotype from healthy controls. Although non-classical female subjects had significantly higher lysoGb3 values than healthy controls, some overlap was present.

A distinction between these groups is of importance, since subjects with an uncertain diagnosis of FD are increasingly identified. They often lack characteristic features of classical FD, but demonstrate FD-like findings, such as LVH or CKD, and have a GLA variant of unknown significance. A non-invasive diagnostic tool, such as lysoGb3 measurement, to discern neutral GLA variants from variants that are associated with a non-classical FD phenotype, is therefore warranted. Several other reports support our finding that lysoGb3 is increased in biopsy confirmed non-classical FD.29–34 ,46 Vice versa, all subjects in whom a diagnosis of FD was excluded had normal lysoGb3 values. This confirms previous data that lysoGb3 is normal in individuals with a neutral GLA variant, such as p.A143T, p.E66Q, p.S126G and p.R118C variants.14 ,29 ,34 ,38 ,39 ,47 Negative biopsies in individuals with two of these variants (p.A143T and p.E66Q) support these findings.14 ,38 ,47

The lysoGb3 values in our group of patients with an uncertain diagnosis are difficult to interpret. In this group, the elevated value in the range of non-classical men in patient 9 with the p.L106F mutation combined with the clinical signs, indicates a non-classical FD phenotype, but no confirmation with a biopsy was available. Normal values, as found in patient 8 with the p.A143T mutation, are probably associated with a neutral GLA variant, but again, histological confirmation (excluding FD) was not available. Patients 10.1, 10.2 and 10.3 with the p.P60L mutation had lysoGb3 values slightly above normal, but below those observed in the non-classical patients. These patients had no signs of FD as described in detail elsewhere.14

Two earlier studies investigated whether the phenotype of certain GLA variants can be predicted by the level of lysoGb3. Niemann et al29 studied lysoGb3 in subjects with a classical phenotype and subjects with ‘atypical’ GLA variants. They categorised subjects on clinical grounds and found lysoGb3 to be much higher in subjects with ‘clinically relevant’ FD. They proposed a single cut-off value for lysoGb3 of ≥2.7 nmol/L to separate the group of male and female patients with a ‘clinically relevant’ FD phenotype, from the ‘atypical’ group.29 This atypical group contains subjects with several phenotypes, and these subjects would have been classified as non-classical FD or uncertain FD with the strict criteria as applied in our study. In the second study by Lukas et al, a very high lysoGb3 in male subjects with GLA variants was associated with a classical phenotype. Most women with classical FD had increased lysoGb3 values, while values were normal in individuals with GLA variants described as causing a ‘minor catalytic defect of αGalA’ (eg, p.S126G, p.A143T and p.D313Y).34 Neither of the two studies used histological confirmation for the diagnosis of FD, and the diagnostic accuracy of lysoGb3 to discern those with a non-classical but confirmed FD phenotype from those with a neutral GLA variant remains, therefore, unclear. The present study uses a strict classification of subjects with a GLA variant, which enables us to assess the diagnostic value of lysoGb3 with better accuracy.

Thus, lysoGb3 emerges as a very promising diagnostic tool in individuals with a GLA variant presenting with a non-specific FD sign (such as LVH or CKD) but without characteristic phenotypical or biochemical features of classical FD. Ideally, we would have been able to generate a cut-off value to distinguish patients with non-classical FD from those without FD. Based on our data, a lysoGb3 value within the normal range cannot exclude FD, because values in women with a non-classical FD phenotype demonstrated some overlap with healthy controls. For men, the data are suggestive that FD could be excluded when a normal lysoGb3 is found. A value in the range of the non-classical male group (>1.3 nmol/L) suggests FD in both men and women. The higher the lysoGb3 value, the more likely a diagnosis of FD appears to be.

However, there are some important considerations to make. First, and most importantly, slight increases of lysoGb3 may indicate glycolipid storage, but do not prove that the signs or symptoms of a patient are a consequence of this storage and related to FD. This indicates that in those patients without characteristic features of FD and only a slight increase of lysoGb3, other causes need to be considered, which may require additional testing including histology. This will have to be considered on an individual basis.

Second, the number of subjects in the non-classical and no-FD group was small and contained a limited number of different GLA variants. Therefore, larger studies should confirm that (male) subjects with a non-classical FD phenotype invariably have an increased lysoGb3 value compared with healthy controls and that neutral GLA variants (no-FD group) will not lead to an increase in lysoGb3. Confirmation by data from larger cohorts is necessary before a definite cut-off value to confirm or exclude FD in uncertain cases can be established with confidence. Until then, we feel that in individuals with a GLA variant and an uncertain diagnosis of FD, histological confirmation is still needed to confirm a diagnosis and to assess the extent of the disease.

The specificity of small lysoGb3 increases is currently unknown, but is expected to be very high. One potential problem could be the use of drugs known to cause a drug-induced lipidosis, of which amiodarone is frequently used in patients with LVH. These drugs could possibly cause elevated levels of lysoGb3. This should be carefully investigated before lysoGb3 may be used in subjects who used these drugs at any time during the medical history.

It is of importance to note that different mass spectrometry assays are currently in use to quantify lysoGb3 that employ different internal standards, some of which show little chemical resemblance to lysoGb3.29 ,34 The analytical accuracy of these different methods is often unknown. As a consequence, data cannot be directly compared between centres that employ different methods.29 ,34 Also, although lysoGb3 is the major of seven isoforms identified in plasma, it would be of interest to investigate the relationship between the other isoforms and clinical phenotypes of FD as we did for lysoGb3.

In conclusion, our study demonstrates that lysoGb3 is a reliable diagnostic tool to discern classical FD from no FD. In individuals with a GLA GVUS presenting with a non-specific FD sign (such as LVH or CKD) but without characteristic phenotypical or biochemical features of classical FD, increased lysoGb3 values >1.3 nmol/L are suggestive for a diagnosis of FD. Normal lysoGb3 cannot exclude FD in women but makes a diagnosis of FD in men highly unlikely.


Professor Dr AH Zwinderman is acknowledged for his advice on the statistical aspects of the study.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • BES and LvdT contributed equally.

  • Contributors BES and LvdT: study design, data collection analyses and interpretation, draft of manuscript. MB, GEL, CEMH: data interpretation, revision of manuscript. BJHMP: laboratory assessments, data interpretation, revision of manuscript.

  • Funding This study was performed within the framework of the Dutch Top Institute Pharma (TIPharma, project number T6-504: ‘Fabry or not Fabry: valorisation of clinical and laboratory tools for improved diagnosis of Fabry disease’. TIPharma is a non-profit organisation that catalyses research by founding partnerships between academia and industry. Partners: Genzyme, a Sanofi company; Academic Medical Centre, University of Amsterdam. Subsidising party: Shire HGT.

  • Competing interests BES has received travel support from Shire HGT and Genzyme. LvdT has received travel support and reimbursement of expenses from Actelion, Shire HGT or Genzyme. MB, GEL and CEMH have received travel support, honoraria for consultancies and speakers fees from Actelion, Genzyme, Shire HGT, Protalix or Amicus. All fees are donated to the Gaucher Stichting or the AMC Medical Research for research support.

  • Provenance and peer review Not commissioned; externally peer reviewed.