Article Text

Original article
Rapid, proteomic urine assay for monitoring progressive organ disease in Fabry disease
Free
1. Ivan D Doykov1,
2. Wendy E Heywood1,2,
3. Valeria Nikolaenko1,2,
4. Justyna Śpiewak1,
5. Jenny Hällqvist1,
6. Peter Theodore Clayton1,
7. Philippa Mills1,
8. David G Warnock3,
9. Albina Nowak4,
10. Kevin Mills1,2
1. 1 Centre for Inborn Errors of Metabolism, UCL Institute of Child Health Library, London, UK
2. 2 NIHR Great Ormond Street Biomedical Research Centre, Great Ormond Street Hospital and UCL Great Ormond Street Institute of Child Health, London, UK
3. 3 Division of Nephrology, Department of Medicine, University of Alabama, Birmingham, Alabama, USA
4. 4 Department of Endocrinology and Clinical Nutrition, University Hospital Zurich and University of Zurich, Raemistrasse, Zurich, Switzerland
1. Correspondence to Dr Kevin Mills, Centre for Inborn Errors of Metabolism, University College London Great Ormond Street Institute of Child Health Library, London WC1E 6BT, UK; kevin.mills{at}ucl.ac.uk

## Abstract

Background Fabry disease is a progressive multisystemic disease, which affects the kidney and cardiovascular systems. Various treatments exist but decisions on how and when to treat are contentious. The current marker for monitoring treatment is plasma globotriaosylsphingosine (lyso-Gb3), but it is not informative about the underlying and developing disease pathology.

Methods We have created a urine proteomic assay containing a panel of biomarkers designed to measure disease-related pathology which include the inflammatory system, lysosome, heart, kidney, endothelium and cardiovascular system. Using a targeted proteomic-based approach, a series of 40 proteins for organ systems affected in Fabry disease were multiplexed into a single 10 min multiple reaction monitoring Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS) assay and using only 1 mL of urine.

Results Six urinary proteins were elevated in the early-stage/asymptomatic Fabry group compared with controls including albumin, uromodulin, α1-antitrypsin, glycogen phosphorylase brain form, endothelial protein receptor C and intracellular adhesion molecule 1. Albumin demonstrated an increase in urine and could indicate presymptomatic disease. The only protein elevated in the early-stage/asymptomatic patients that continued to increase with progressive multiorgan involvement was glycogen phosphorylase brain form. Podocalyxin, fibroblast growth factor 23, cubulin and Alpha-1-Microglobulin/Bikunin Precursor (AMBP) were elevated only in disease groups involving kidney disease. Nephrin, a podocyte-specific protein, was elevated in all symptomatic groups. Prosaposin was increased in all symptomatic groups and showed greater specificity (p<0.025–0.0002) according to disease severity.

Conclusion This work indicates that protein biomarkers could be helpful and used in conjunction with plasma lyso-Gb3 for monitoring of therapy or disease progression in patients with Fabry disease.

• biomarker
• stratify
• Fabry disease
• proteomics
• mass spectrometry

## Introduction

Fabry disease (FD; Anderson-Fabry disease; OMIM no 301500) is an X-linked lysosomal storage disorder caused by mutations in the alpha-galactosidase A (GLA) gene, which leads to a deficiency of α-galactosidase A (α-GAL).1 2 The consequent progressive intracellular accumulation of neutral glycosphingolipids, primarily globotriaosylceramide (Gb3, CTH or GL-3), is proposed to contribute to a wide variety of clinical symptoms of FD. In males with the classic phenotype, who have an absent or very low α-GAL activity, early symptoms, including acroparesthesia, episodic pain crises, angiokeratomas, hypohidrosis, gastrointestinal disturbance and corneal opacities, usually develop in early childhood and the disease gradually progresses with age, resulting in renal, cerebral and cardiac involvement.2 In contrast, in males with the later-onset phenotype, who have a significant residual α-GAL activity, renal or cardiac involvement occur in adulthood, without the preceding early symptoms.3–8 Although FD is an X-linked trait, females do exhibit disease symptoms; however, this is generally in a delayed and milder form due to a random X-chromosomal deactivation.9 Life expectancy in females is reported to be reduced by 15 years compared with the unaffected population, with untreated males having a median survival of 50 years.10 11 The incidence of FD is estimated to range from one in 40 000 to one in 117 000.3–8 However, due to the broad spectrum of clinical symptoms, many patients may remain undiagnosed.12

Treatment for FD comprises symptomatic management, chaperone, substrate deprivation and enzyme replacement therapy (ERT). When to begin treatment of patients is often a difficult decision and, especially in children, can be a contentious issue. Additionally, there is also the dilemma of deciding when or whether to start treatment in asymptomatic patients, including some of whom may have family members who are severely affected in later life. In addition, the high expense of ERT (US$150–US$250K per year) needs to be taken into consideration. The effectiveness of the treatment is also a debatable issue with varying studies offering different conclusions in regards to recommendations on when treatment should be started, on what dose levels and the prediction of different organs responding better to treatment than others.13–15 This is compounded by conflicting views on which biomarkers are the best measure of response to treatment. Conventional biomarkers in FD include plasma and urinary Gb3 levels16 and, more recently, plasma and urinary globotriaosylsphingosine (lyso-Gb3) levels.6 17 18 Auray-Blais et al, have suggested that the measurement of analogues of lyso-Gb3 in both plasma and urine may be more informative in monitoring both organ involvement and response to treatment.19–21

Several studies are underway to decide on the definitive body fluid and Gb3 or lyso-Gb3 lipid moiety that determine response to treatment. In this study, we have taken an alternative approach and attempted to measure downstream markers of the disease-related pathology. Using a targeted proteomic-based test, a series of protein biomarkers for organ systems affected in FD were multiplexed into a 10 min multiple reaction monitoring (MRM) Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS) assay using only 1 mL of urine. Biomarkers included in the assay were determined according to previous biomarker discovery experiments in patients with FD.22 23 In addition, further biomarkers were included after discussion with clinical colleagues and reviewing the literature24–28 and those proteins determined to be potentially useful for monitoring cardiovascular, renal, clotting and endothelial function.29–37 (The full list of proteins quantitated in this assay are shown in online supplementary data 2.) The test was evaluated in urine samples from patients with FD entering University Hospital Zurich and who had various stages of the disease progression associated with FD. Here, we discuss the levels of these biomarkers and their potential to be used, alongside lyso-Gb3 and Gb3 for monitoring treatment and disease progression in FD.

## Materials and methods

### Samples and disease grouping

The study was performed on samples provided by patients attending University Hospital Zurich. Informed consent for collecting clinical data, urine and blood samples for biobanking was obtained from all patients. Lyso-Gb3 results were obtained from the blood but all research analyses presented in this work were performed on urine. We recruited 66 consecutive adult patients (males: n=27 (41%)). All patients had a confirmed GLA-mutation diagnosis and presented for routine annual examinations at the University Hospital Zurich.

All patients had a multidisciplinary work-up, including medical history, cardiac, renal and neurological evaluations. Standard transthoracic two-dimensional echocardiography was routinely performed in all patients. Left Ventricular Myocardial Mass Index (LVMMI) was calculated using the Devereux formula.9 The occurrence of stroke or transient ischaemic attack (TIA) was evaluated during annual examinations by asking the patient and/or using the medical records. For the present analyses, all clinical and routine laboratory results were extracted from the patients’ medical notes. For the classification of the disease severity, the categorization ‘stroke’ has been used for patients who fully or partly or did not recovered from stroke.

### Sample collection and classification

Spot urine samples were collected from adult patients with FD (n=66, 27 males and 39 females) and age-matched healthy adults, obtained from laboratory volunteers (five males and five females). All samples were stored at −80°C for a later batch analyses and had undergone one freeze thaw cycle prior to analysis.

Patients with FD were stratified according to symptoms: heart involvement was defined as the presence of diastolic dysfunction and/or left ventricular hypertrophy (LVH) on echocardiography or heart MRI. Diastolic dysfunction was defined as the left atrial volume index >34 cm/sec and the mitral inflow with the E/e <12. LVH was defined as left ventricular thickness of ≥13 mm. Kidney involvement was defined as either having protein/creatinine coefficient >0.015 g/mmol and/or estimated glomerular filtration rate (eGFR) according to Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula of <90 mL/min/1.73 m2. FD patients were defined as having a proteinurina of <0.03 g/mmol creatinine and an eGFR >90. Cerebral involvement was defined as the history of stroke or TIA. Sample information is supplied as online supplementary data table 1.

### Targeted MRM LC-MS/MS assay

Peptide selection and assessment, urine preparation, Ultra Performance liquid chromatography (UPLC) and mass spectrometry (MS) settings were performed as described previously.38 Quantotypic peptides for each protein were selected using the open source online Global Proteome Machine MRM database at www.thegpm.org and each peptide homology validated by Basic Local Alignment Search Tool (BLAST) searching for their unique identity.39 Custom-synthesised peptides (Genscript, USA) were used to optimise peptide detection and determine the retention time and identify unequivocally the correct peptides in a urine matrix. The optimal peptide from the selection of two peptides/protein and two transitions was selected based on detection in urine, their quantitative response to increasing amount spikes into urine as determined by a standard curve and having no interfering peaks when checked with a pooled urine matrix. Parent/product ion m/z and cone/collision energies are shown in online supplementary data table 2 for the 41 proteins analysed in this study.

An intact protein of yeast enolase from Waters MassPREP protein standard mix (429 ng) was used as internal standard (Waters, UK) and added to 1 mL of urine. Samples were digested according to urine volume (1 mL) and not total protein value. Urines were vortexed and filtered using 3 kDa molecular weight cut-off filters (Millipore, UK). Protein digests were analysed using a Waters Aquity UPLC coupled to a Waters Xevo TQ-S mass spectrometer. Peptides were separated prior to MS detection using a CORTECS UPLC C18 +Column, 90 Å, 1.6 µm, 3 mm × 100 mm column attached to a C18 +VanGuard precolumn. Analysis was performed with chromatographic separation of peptides over a 10 min reverse phase gradient. Quality control (QC) runs of pooled urine digest were run in triplicate at the start of the run and then every 10 injections. A coefficient of variation (CV) <±15% for each QC was considered acceptable. Chromatograms were analysed using Waters TargetLynx software. Peptides were quantitated using the unique yeast enolase peptide (NVNDVIAPAFVK) and calibration curves generated against standard peptides for each protein (GenScript, USA). Finally, all proteins were expressed relative to urinary creatinine to correct for the strength of each urine as per normal chemical pathology procedure. Urine creatinine was determined using MS and as described previously by Manwaring et al 19 22 40 and results expressed as microgram of protein/millimole of creatinine. A pooled sample was created using equal volumes from each sample and spiked with 1 ng of synthetic peptides. This pooled QC was injected at equal intervals throughout the run to record the retention time and MS performance. Analyte/internal standard ratio was evaluated at CV <15%.

### Statistical analysis

Data were exported to Excel. Mean values of duplicates were exported to GraphPad Prism 7 for statistical analysis. Analyses included data QC for peptide performance (CV), QC of sample preparation and LC-MS/MS performance (yeast enolase). Standard curves were analysed by linear regression analysis and Pearson’s correlation. Group comparisons were performed using non-parametric Mann-Whitney test.

## Results

### Urinary proteins elevated in patients with early-stage/asymptomatic FD

Using the targeted proteomic test developed, it was possible to identify the statistically significant elevation of six proteins present in the urine of genetically confirmed patients with FD (n=12), patients who demonstrated no prior symptoms of the disease, that is, no renal or cardiovascular involvement as determined by microalbuminuria, GFR or MRI. All biomarkers developed into the test are shown in online supplementary figure 1 and were compared with the ‘gold standard’ plasma lyso-Gb3 biomarker.

### Supplemental material

Figure 1 shows the protein biomarkers α1-antitrypsin and uromodulin which demonstrated the highest significance and fold-elevation compared with age-matched controls (p=0.0005 and 7.7-fold elevation, p=0.007 and 7.7-fold elevation, respectively). Although the three urinary proteins α1-antitrypsin, uromodulin and glycogen phosphorylase brain form demonstrated a higher fold-change difference in patients with FD, they were not as statistically significantly elevated when compared with plasma lyso-Gb3 levels. Interestingly, urinary albumin levels were observed to be 3.1-fold elevated (p=0.0095) compared with the control group and indicated the presence of subtle microalbuminuria that could not be detected using conventional routine testing techniques used in hospital laboratories. A total of 50% (6/12) of the patients demonstrated urinary albumin levels greater than the controls.

Figure 1

Potential early-stage/presymptomatic biomarkers of Fabry disease (n=12) that are present in the urine of patients that demonstrate no clinical symptoms of cardiovascular or renal impairment.

Although, plasma lyso-Gb3 analyses proved to be a slightly better biomarker for detecting patients with FD with little-to-no clinical symptoms than the proteins used in this study, figure 2 shows that plasma lyso-Gb3 could not detect all patients with FD. A single female patient aged 46 years carrying the R301Q mutation, not on therapy and displaying no symptoms of cardiovascular or renal involvement, demonstrated plasma lyso-Gb3 levels within the control range. This mutation has been previously reported to encode for a residual α-GAL activity and therefore to cause the later-onset phenotype.41 However, in contrast to plasma lyso-Gb3 levels, both the urinary uromodulin and albumin levels in this patient were demonstrated to be 22-fold elevated when compared with the highest control value for uromodulin, and 1.6-fold higher than the highest albumin control. Although comprehensive gene testing, enzymology in males and possibly histological analysis of biopsies are probably the most reliable methods to diagnose FD, this finding indicates that urinary uromodulin and albumin levels could be useful in the management of treatment of patients with FD. Both these biomarkers might also be useful in understanding the more subtle disease causing mutations as well as possibly identify the patients with FD who likely to go on to develop kidney disease.

Figure 2

Uromodulin and albumin levels in a patient with Fabry disease who has normal plasma lyso-Gb3 levels and no signs of cardiovascular or renal involvement (male, aged 46 years, R301Q mutation). The dotted line indicates the control patient upper levels. lyso-Gb3, globotriaosylsphingosine.

### Urinary proteins with a potential to monitor lysosomal function/hypertrophy

The multiplexed panel used in the study included urinary proteins that were reported to be of lysosomal origin or associated with the lysosome itself. Although some overlap between the different severities of the disease was observed, in general three of these proteins prosaposin, GM2 activator protein and sortilin (figure 3A, B and C respectively) demonstrated significant changes with the progression of the disease and that potentially reflect lysosomal hypertrophy or substrate accumulation. The cohorts were grouped according to clinical pathology as described previously in the Methods section. Urinary prosaposin levels demonstrated the most marked changes in progression or organ involvement of FD (figure 3A). Significant differences in urinary prosaposin levels were also observed between patients with no symptoms and cardiac involvement only (p=0.0249), patients with renal impairment only (p=0.0049), patients with renal and cardiac involvement (p<0.001) and patients with a history of stroke and renal and cardiac involvement (p=0.0002). Figure 3B and C show the results of the urinary levels of GM2 activator protein and sortilin, respectively. Although the general trend for urinary levels of both proteins showed an increase with the disease severity, observed elevated levels of GM2 activator protein and sortilin did not reach statistical significance until the occurrence of marked kidney damage or when patients had a clinical history of renal, cardiac events and stroke (figure 3A,B). These data indicate that prosaposin is a potentially better biomarker of lysosomal hypertrophy or dysfunction than GM2 activator protein and sortilin.

Figure 3

Analysis of potential lysosomal markers: (A) prosaposin, (B) GM1 activator protein and (C) sortilin in the urine of patients with Fabry disease (n=66) and their levels categorised according to disease progression. The dotted line indicates the control patient upper levels. CNS, central nervous system.

### Urinary protein biomarkers with a potential to monitor renal injury and/or impairment

A total of six proteins known to be of kidney origin, or associated with kidney disease, demonstrated a significant change in urinary levels with progression of the clinical symptoms associated with FD (figure 4). The potential podocyte markers nephrin and podocalyxin both showed significant elevation with disease progression, although this relationship dropped markedly in the more severe patients with FD who exhibited a history of stroke, kidney and cardiac involvement. Podocalyxin, fibroblast growth factor 23 and cubilin appear to be associated specifically with disease groups with kidney disease. A significant difference was observed not just for the ‘kidney involvement only’ and ‘early-stage/asymptomatic Fabry groups’ but also for the ‘cardiovascular manifestations only’ disease group.

Figure 4

Analysis of potential kidney disease markers (A) nephrin, (B) podocalyxin, (C) uromodulin, (D) FGF-23, (E) cubilin and (F) AMBP in the urine of patients with Fabry disease (n=66) and their levels categorised according to disease progression. *p<0.05, **p<0.01, ***p<0.0001 (bold data points indicate a patient receiving ERT therapy). The dotted line indicates the control patient upper levels. AMBP, Alpha-1-Microglobulin/Bikunin Precursor; ERT, enzyme replacement therapy; FGF-23, fibroblast growth factor 23.

The proteins cubilin and Alpha-1-Microglobulin/Bikunin Precursor (AMBP) both showed a general increase in concentration with disease progression (figure 4E and F, respectively). However, these biomarkers did not show the sensitivity of nephrin, podocalyxin and fibroblast growth factor 23 as their increased levels were not significant from both the non-kidney disease groups until the more advanced disease groups, indicating these proteins are only altered in advanced disease.

Uromodulin levels were observed to be significantly lower in patients with FD with both kidney and cardiac involvement (p=0.0045), and in the late-stage patients exhibiting kidney, heart and central nervous system (CNS) involvement (p=0.0049). Urinary albumin levels, although demonstrating statistically significant elevation in the urine of patients with FD relative to controls (figure 2) did not demonstrate any significant elevation relative to the other stages of FD (online supplementary data figure 2). Somewhat surprisingly, with the higher levels of albumin observed in patients with moderate to advanced kidney disease, these data probably arising from the sensitivity of targeted proteomics resulted in a limited linearity range and the failing of the assay to measure the protein accurately at very high amounts (detector saturation and chromatography splitting).

### Urinary protein biomarkers with a potential to monitor cardiac involvement

A total of five proteins were included in the assay for their potential to monitor cardiac disease. Lumican was identified in a previous biomarker discovery experiment as a marker of hypertrophic cardiomyopathy42 while brain natriuretic peptide (BNP), glycogen phosphorylase B (GPB), troponin T and troponin I are known biomarkers of cardiac disease (figure 5A–E). All five potential biomarkers demonstrated a general increase with the disease progression, with GPB and troponin T showing a significant elevation observed in patients presenting with a cardiac phenotype only (p=0.043 and 0.036; figure 5B and C, respectively). None of the other cardiac proteins were significantly elevated in the cardiovascular only group compared with the early-stage/asymptomatic group. However, significantly elevated and higher levels of all the proteins were also observed in patients presenting with kidney disease and no significant cardiac involvement. These data indicate that the cardiac biomarkers are potentially useful for identifying progressive but not early cardiac involvement and their levels are probably confounded by the later involvement of kidney disease. However, these levels could reflect that cardiorenal syndrome which is a very complex scenario and drawing conclusions are really a challenge in the sample set used in this work. Further research on more and well-characterised patients with cardiovascular and kidney disease would need to be undertaken to better understand the significance of these results. Although the numbers were very limited for the early-stage patient group analysed in this study, a Spearman correlation was performed against other cardiac risk factors (arterial hypertension, smoking, low-density lipoprotein (LDL) cholesterol and body mass index (BMI); online supplementary file 5). A positive correlation was observed with plasma LDL-cholesterol and megalin (r=0.62, p=0.02) and also BMI (r=0.57, p=0.04). A significant negative correlation was observed with BMI and Troponin T2 (TNNT2) (r=−0.56, p<0.05).

### Supplemental material

Figure 5

Analysis of potential heart disease markers (A) brain natriuretic peptide, (B) glycogen phosphorylase brain form, (C) troponin T, (D) troponin I and (E) lumican in the urine of patients with Fabry disease (n=66) and their levels categorised according to disease progression. *p<0.05, **p<0.01, ***p<0.0001 (bold data points indicate a patient receiving ERT therapy). The dotted line indicates the control patient upper levels. ERT, enzyme replacement therapy.

### Potential inflammatory protein biomarkers present in the urine

The inflammatory proteins neutrophil collagenase, hemopexin, intracellular adhesion molecule 1 (ICAM1) and α1-antitrypsin (figure 6A–D) demonstrated elevated levels in the majority of patients with FD. However, this was not observed in all cases and some overlap between the different disease stage groups were observed. The inflammatory markers neutrophil collagenase and hemopexin both demonstrated significantly elevated levels that corresponded with disease progression and indicating that inflammation may play an important role in the disease pathogenesis observed with FD (figure 6A,B). ICAM1, although elevated in the cardiac, kidney, cardiac and kidney, and those patients with cardiac, kidney and CNS involvement, the levels in the urine only reached significant elevation in patients with kidney disease only (p=0.0031). Although α1-antitrypsin demonstrated significant and almost eightfold elevated levels between controls and patients with FD exhibiting no symptoms clinically (figure 1A), these levels did not rise significantly with further progression of the clinical symptoms observed in FD (figure 6D).

Figure 6

Analysis of potential inflammatory markers (A) neutrophil collagenase, (B) hemopexin, (C) intracellular cell adhesion molecule 1 and (D) α1-antitrypsin in the urine of patients with Fabry disease and their levels categorised according to disease progression. *p<0.05, **p<0.01, ***p<0.0001 (bold data points indicate a patient receiving ERT therapy). The dotted line indicates the control patient upper levels. ERT, enzyme replacement therapy.

## Discussion

The main aim of this proof-of-principle project was to see if we could develop a non-invasive, multiplexed and targeted proteomic method that could supplement the conventional tests used to aid diagnosis and monitor treatment for FD. In-depth testing of such biomarkers would aid in development of emerging and novel therapies. In total, a method was developed to detect 41 proteins in 1 mL of urine (listed in online supplementary data table 2). All 41 proteins from this panel could be reliably quantitated but only 20 biomarkers showed any potential changes and use for the diagnosis and monitoring of disease progression in FD. The biomarkers could be divided into two groups, those that had potential for aiding the diagnosis of FD and those that could be used potentially for monitoring treatment or studying disease progression.

Surprisingly, the assay proved useful for looking at the early stage/presymptomatic or early stages of disease as opposed to the later stages of FD, when patients develop a more severe phenotype (kidney, cardiac and CNS involvement or stroke). An interesting finding was the presence of specific proteins in the urine of patients who presented with no symptoms of FD using conventional tests (MRI, GFR, microalbuminuria and plasma lyso-Gb3). α1-Antitrypsin and uromodulin both showed significantly elevated levels in the urine of the presymptomatic patients. Although not as useful as plasma lyso-Gb3 for diagnosing FD, both demonstrated a greater fold elevation (~7.7-fold) compared with plasma lyso-Gb3 (~sixfold elevation). The reasons for this elevation of α1-antitrypsin in the urine are unclear but have been reported previously by Maicas et al.43 They postulated that the elevation of α1-antitrypsin maybe an anti-inflammatory or cytoprotective effect while other studies44–46 demonstrated that α1-antitrypsin levels could be used to predict and monitor kidney function in type-2 diabetic nephropathy, glomerulonephritis and monitoring kidney disease, respectively. Similarly, the elevation of uromodulin, or Tamm-Horsfall protein, has been described previously as a good biomarker for quantitating kidney function in FD47 and as a potent immunomodulatory molecule in kidney disease.48 The latter research agrees with our results that the elevation of α1-antitrypsin indicated an inflammatory component, or involvement, in the early stages of FD. Other proteins showing an elevation in the early-stage/-presymptomatic patients included GPB, albumin, endothelial receptor C and podocalyxin. GPB is highly expressed in the brain and the heart and was included in the panel as a potential biomarker of heart disease. Previous studies49 50 have shown that GPB is a good marker of acute myocardial injury or infarction. Although acute cardiac injury is not a hallmark of FD, GPB demonstrated both, a significant fold-change elevation (sixfold) and indicated that it could be a potential and subtle early marker of cardiac disease in early-stage/presymptomatic patients. Indeed, increased levels of GPB were observed in 50% of the patients with no signs of cardiac disease (presymptomatic group). Surprisingly, increased urinary albumin levels indicated that it was a good protein to accompany the diagnosis of FD with almost 50% of the presymptomatic patient group showing an elevation compared with controls. None of the patients were reported as having microalbuminuria using conventional tests and indicating that targeted proteomics is a much more sensitive and accurate methodology for quantitating proteins than conventional chemical laboratory assays. However, one of the limitations of the assay we found was in its inability to reliably quantitate high levels of albumin, as the method was designed to look only low abundant biomarkers. Further refinement of the assay will be necessary if the assay is to be used to quantitate macroalbuminuria.

Although no proteins analysed in this study proved quite as specific as plasma lyso-Gb3, figure 2 shows one patient with a confirmed later-onset mutation in the alpha galactosidase gene (female, R301Q, aged 46 years), who did not have an elevated plasma lyso-Gb3 level but did demonstrate significantly elevated uromodulin and albumin levels in the urine. The presence of these proteins in the urine indicates that they may be very sensitive markers of early-stage/presymptomatic kidney disease. Although only one patient, this finding indicates that although plasma lyso-Gb3 remains helpful for diagnosing FD, it is not perfect and the use of other markers such as urinary uromodulin and albumin may help the diagnosis in difficult cases and reduce the number of false negatives when used in conjunction with plasma lyso-Gb3.

### Analysis of those biomarkers that may reflect disease progression

During the development of the assay, the proteins included in the test were chosen from either preliminary proteomics experiments or thought to reflect certain organ or cellular systems known to be affected in FD (online supplementary data table 2). The lysosomal proteins that could be detected in the urine included prosaposin, GM2 activator protein and sortilin. All three proteins demonstrated a general increase in levels detected in the urine that reflected the progression of the disease and probably resulting from lysosomal hypertrophy due to storage burden. These data correlated with our previous work showing the increased levels of these biomarkers in early-stage/presymptomatic patients.22 The kidney contains very high levels of lysosomes making urine a very good fluid for lysosomal research. This may also explain the higher levels of lysosomal proteins in the urine of patients with FD with kidney disease only compared with those with cardiac manifestations only. However, in those patients with both, kidney and heart involvement, the levels were significantly higher than individual systems alone. The highest levels of all three biomarkers were observed in the urine of patients with cardiac, kidney and some history of stroke. This is indicative of the storage burden and hypertrophy known to occur in the later stages of FD.

Our next panel of proteins were selected to see if it was possible to quantitate kidney function and included markers of podocyte loss,51 52 tubular dysfunction34 and those proteins that act on the function of the kidney. It is interesting to observe that nephrin, a specific podocyte marker, demonstrated significantly elevated levels (figure 4A) even in those patients with FD presenting with cardiac symptoms only. This indicates that nephrin maybe a sensitive marker of early podocyte loss and therefore early-stage/presymptomatic kidney disease in patients with FD who are thought to have only cardiac involvement. Podocalyxin, fibroblast growth factor 23,53 cubilin and AMBP were not elevated in the group with only cardiovascular disease but were elevated in the kidney disease only group indicating these are markers of kidney disease. Podocalyxin levels appear to become very varied and non-specific in the other two groups with more multisystemic disease. The reason for this is unclear but maybe explained by a more extensive loss of podocytes at the early stages of the disease which could be an indication of podocyte loss in these groups where disease is more advanced. Fibroblast growth factor 23 continued to remain elevated in these groups and cubilin appeared to show a progressive increase as the disease involved more organ systems.

Although other kidney disease-associated proteins showed a similar and general pattern of increased levels observed in the urine which corresponded with kidney disease, this was not observed with uromodulin and demonstrated the inverse with FD progression. Uromodulin is secreted by the thick ascending limb of the loop of Henle and thus the lower levels observed in the urine as the disease progresses, probably reflects tubular damage. Uromodulin is continually secreted into the urine and is thought to be an inhibitor of calcium crystallisation but may also have an antibacterial role and protection against urinary tract infections. Diseases associated with uromodulin include familial juvenile hyperuricemic nephropathy 1 and medullary cystic kidney disease type 2.48 Thus, it is highly likely that the loss of uromodulin excretion in the kidney reflects kidney disease associated with FD but may also play a part of the disease mechanism or process observed in the disease.

Interestingly, five proteins were selected for their potential as biomarkers capable of monitoring cardiac function and could be detected in the urine, in both, controls and patients with FD. Some of these proteins have been detected in urine before and explored for their potential utility in cardiac disease.54 55 Like other biomarkers analysed in this study, the levels of all five proteins followed a similar pattern that corresponded with the disease progression associated with FD and, in the case of GPB and troponin T, were significantly elevated in the patients with cardiac involvement compared with the patients with none (figure 5). However, increased levels were also observed in patients with kidney disease involvement only. This indicated that the specificity of these cardiac markers in the urine is not as specific as would have been predicted or that patients with kidney disease allow these proteins to enter the urine more freely and thus compromised the clinical use or relevance of the markers. The reason for this discrepancy is unclear and the origin of these markers should be further elucidated by looking at different acute and chronic cardiac diseases such as myocarditis, heart failure or acute infarct damage.

Finally, the proteins neutrophil collagenase, hemopexin, ICAM1 and α1-antitrypsin, which are known to be associated with inflammation all demonstrated significant elevation and correlated with the disease progression associated with FD. α1-Antitrypsin demonstrating a significant 7.7-fold elevation between controls and those patients with FD who did not show any symptoms of FD. However, the proteins neutrophil collagenase, hemopexin and, to a lesser extent ICAM1, all followed the disease progression. Only α1-antitrypsin did not show as much potential for monitoring inflammation as levels were not observed to follow the progression of disease as closely. Neutrophil collagenase, or MMP8, is a metalloprotease secreted by neutrophils and is involved in the degradation of the extracellular matrix. It is also known to be involved in the remodelling of tissue, a hallmark of Hypertrophic Cardiomyopathy (HCM) and the kidney disease associated with FD. Another potential marker and indicator of an inflammatory component to FD was the protein hemopexin. Hemopexin is a haeme or iron-scavenging protein whose function is to protect the body from the oxidative damage and proinflammatory nature of iron in the circulation. Hemopexin has been shown to play a role in lupus, malaria and also a potent antioxidant role in the heart.56 Although hemopexin levels observed in the urine do reflect the disease progression observed in FD, what exactly they are indicative of is unclear and could reflect changes in iron homeostasis, inflammation, kidney or heart failure. The higher levels of proinflammatory proteins in the urine certainly indicate that oxidative stress or inflammation may play a part in FD.

All patients who were symptomatic in this study were on ERT; therefore, the proper assessment of these markers in response to ERT was not possible. A longitudinal study is required to ascertain this. The effect of ERT could explain the widespread heterogeneity of some of the protein values observed in the disease groups (particularly more so in the later stage disease groups) as there may be some patients responding or not responding to the ERT. Plasma lyso-Gb3 is useful for determining disease progression in male patients with FD 4 but when grouped according to disease symptoms in this cohort, similarly to the protein markers quantitated in this study, there were no significant differences in the lyso-Gb3 values observed according to disease (online supplementary data figure 3). The inclusion of hemizygous females in the cohort who typically have lower Gb3 and lyso-Gb3 levels may explain why a relationship to disease was not observed for lyso-Gb3. We also looked at any possible relationship of lyso-Gb3 with any of the proteins by Spearman multiple correlation analysis (online supplementary data figure 4). No correlation of plasma lyso-Gb3 with any of the proteins was observed, indicating their expression is independent of lyso-Gb3 and is more likely related to the specific organ pathology of the disease. However, a Spearman multiple correlation analysis of patients with FD demonstrated a strong negative correlation between most proteins and uromodulin. In particular, a strong negative correlation between uromodulin and GM2 activator protein, prosaposin, nephrin, troponin I and lumican were observed. All these proteins that are highly expressed in the lysosome, kidney and heart and all organs known to be affected in FD. In contrast, a strong positive correlation was observed between cubilin with sortilin, lumican and GPB, troponin T with GPB, and finally, uromodulin with endothelial protein receptor C.

The limitations of this study are the biomarkers developed into the test described in this work have only been measured in the urine of patients with FD and a significant number of patients were on treatment. The levels of these biomarkers in other patients with Lysosomal Storage Disease (LSD) or other non-lysosomal causes of cardiovascular and kidney diseases might provide further insight into the relevance of these biomarkers for monitoring disease progression and treatment for FD. In addition, a similar follow-up study needs to be conducted with a well-characterised and longitudinal cohort of samples, ideally from patients pretreatment and post-treatment and preferably over several years.

## Conclusion

We believe this rapid multiplex urine assay has much potential and indicates some proteins may be specific disease biomarkers that could have promise for monitoring treatment or disease progression.

## Acknowledgments

The authors greatly acknowledge the support provided by the National Institute for Health Research Great Ormond Street Hospital Biomedical Research Centre (NIHR GOSH BRC), UCL Biological Mass Spectrometry Centre at the UCL Institute of Child Health and the Szeban Peto Foundation for kind donations. We also wish to thank our technician Ernestas Sirka for making up solutions.

## Footnotes

• IDD and WEH are joint first authors.

• AN and KM are joint last authors.

• AN and KM contributed equally.

• Contributors IDD and WEH, the main authors, contributed equally to study design, the design of peptides, development and optimisation of the digestion and clean up of peptides, the method development, the data analyses and writing of the manuscript. This is the PhD work of IDD. VN, JS and JH helped carry out the laboratory work, for example, extraction of the proteins from the urines, digestion and clean up of the peptides for analyses, running of the mass spectrometer and significant help doing blinded data analyses for significant amounts of ‘peak picking’ and calibration lines creation, that is, integrating protein chromatograms and graph drawing. PTC, PM, DGW are our colleagues, consultant medics, nephrologists and biochemists who provided advise and guidance on what proteins to look at that we should look at as biomarkers of endothelium, lysosomes, cardiac cells, immunity and kidney disease. They were all involved in the study design, how to write the paper and then critical assessment of the final manuscript and conclusions. KM and AN were instrumental in study design, idea of the method development and writing of the manuscript. AN was the clinician involved in seeing all the patients, collecting samples, providing clinical data and diagnoses. KM is the head of mass spectrometry group carrying out this work and was involved in all stages of the analyses, from the initial idea to look at a new way in stratifying patients with FD using urine proteomics, advice of all aspects of the method development, discussion of the results, paper structure and in the writing of the manuscript.

• Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

• Disclaimer The views expressed are those of the author(s) and not necessarily those of the National Health Service, the NIHR or the Department of Health.

• Competing interests None declared.

• Patient consent for publication Not required.

• Ethics approval The study was conducted in accordance with the principles of the Declaration of Helsinki and approved by the Ethics committee of the canton of Zurich, Switzerland (KEK-ZH 2017-00386).

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

• Data availability statement Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.

## 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.