Urine samples were obtained from patients suspected of oligosacharidoses, and were stored at -20°C until analysis. Some samples were from patients affected with very rare oligosaccharide disorders, and were collected at the time of diagnosis some 30 years ago. Suspected patients, aged in majority between 1 month to 13 years with some up to 60 years, were suffering from mild or severe phenotypes with clinical signs most frequently including facial dysmorphism, dysostosis multiplex, hepatosplenomegaly, progressive neurodegeneration, deafness and visual impairment.

Ninety-four urine samples were analyzed. Fifty-eight were obtained from patients affected with various oligosaccharidoses. The diagnosis was ascertained in all cases on the basis of an abnormal characteristic TLC oligosaccharide pattern revealed using sulfuric orcinol 25 and of the demonstration of the specific enzymatic deficiency in serum and/or leukocytes and/or cultured skin fibroblasts (Laboratory of Hereditary Diseases of Metabolism, East Center for Biology and Pathology, Civilian Hospices, Lyon, France (57 cases), Biochemistry Laboratory, University Hospital, Nice, France (1 fucosidosis case)). Other urine samples from unaffected subjects (aged-matched to pediatric patients) were from the two laboratories. For mass spectrometry, samples were analyzed in two different sets. The first set was used for identification and characterization of the various MS/MS patterns; the second one was blindly analyzed for validation of the method (see Table 1).

DHB (2,5-dihydroxybenzoic acid) and 3-AQ (3-aminoquinoline) matrices were purchased from Sigma-Aldrich (St. Louis, MO). All other chemicals were obtained from Merck (Darmstadt, Germany).

After thawing in an ice-bath, the urine samples were centrifuged for 5 min at 25.000 g at 4°C and the supernatants were diluted 1:3 in ultra-pure water. 0.5 μl were spotted on a MALDI 384 target plate ground steel with 2 μl of 3-AQ (15 g/l with 70% (v/v) of methanol, 1% (v/v) of NaCl 0.9% (w/v) and 0.2% (v/v) trifluoroacetic acid, TFA) for positive ion mode analysis and with 2 μl of DHB (50 g/l with 50% (v/v) of acetonitrile) for negative ion mode analysis. These preparations were done in triplicate and dried quickly with a hair-dryer.

All measurements were recorded on an Ultraflex III MALDI-TOF/TOF (BrukerDaltonics, Bremen, Germany). External calibration was performed by spotting 0.5 μL peptide calibration standard II (BrukerDaltonics). All mass spectra were generated by summing 1,000 laser shots for reflectron ion mode, and 1,000 laser shots for the parent mass and at least 1,000 laser shots for the fragments in the lift mode. Laser power was adjusted between 15 and 30% of its maximal intensity, using a 200-Hz smartbeam laser. MS spectra were acquired in the reflectron positive ion mode within a mass range from 300 to 2,000 Da and in the reflectron negative ion mode within a mass range from 700 to 2,100 Da. Reflectron ion mode was chosen to obtain high detection sensitivity and resolution. Mass accuracy was about 14 ppm for 1347 Da. The FlexAnalysis version 3.0 and updated 3.4 provided by the manufacturer were applied for data processing. The GlycoWorkbench version 2.1 was used for the process of structure determination from mass spectrometry data and to draw glycan structures.

Our overall aim was to develop an accurate, simple and rapid method for the urinary screening of oligosaccharidoses using MALDI-TOF/TOF mass spectrometry. To characterize the MS and MS/MS profiles of a particular disease and to identify the accumulated oligosaccharides, we used a training set of urine samples from patients affected with different oligosaccharidoses (Table 1).

To set up the screening method, we took advantage of the relative tolerance for salt of MALDI mass spectrometry to directly analyze urine samples without prior treatment. As neutral glycans are able to ionize by capture of one sodium ion, we analyzed our samples in the positive ion mode meaning that the pseudomolecular ions detected correspond to [M + Na]⁺ 33 . Note that in a salty medium, such as urine, oligosaccharides form adducts with sodium, but also with potassium resulting in a weaker supplemental pseudomolecular ion, [M + K]⁺. The latter is detected by a characteristic double peak separated by a 16 atomic mass unit (amu) difference, which corresponds to a potassium mass minus a sodium mass. Therefore, sodium salt was added to the sample matrix to favor sodium adducts. We also worked in the negative ion mode to detect sialylated glycans since the major form of sialic acid residue in the acidic urine pH is negatively charged. In this mode, the pseudomolecular ions observed correspond to molecular ions that have lost one proton, [M-H]⁻.

DHB and 3-AQ matrices are commonly used for glycan analysis by MALDI mass spectrometry 34 35 36 37 . We tested both matrices in positive and negative ion modes to optimize sample-matrix preparation to detect oligosaccharides without purification and derivatization procedures in urine samples. We adjusted our protocol based on a previous study for which we had to detect glycolipids (sialylated gangliosides and globosides) 38 . Both matrices allowed ionization in each mode. However, we obtained better signal intensity and fragmentation in the positive ion mode with 3-AQ matrix and in the negative ion mode with DHB matrix. Thus, the protocol has consisted in preparing each urinary sample in triplicate with 3AQ and DHB matrices. We recorded and analyzed MS data in reflectron positive and negative ion modes, and achieved the MS/MS analyses on intensive peaks for identification.

The MS/MS data do not provide absolute structure information for isomers such as N-acetylgalactosamine (GalNAc) or N-acetylglucosamine (GlcNAc), which is identified as N-acetylhexosamine (HexNAc). This is also the case for glucose, mannose and galactose, which are identified as hexoses. Predicted assignment of these oligosaccharide structures has been represented on MS spectra.

We first analyzed urines from subjects not suffering from oligosaccharidosis (controls). In positive ion mode, we always observed three intensive peaks from pseudomolecular ions at m/z 429.2, 628.6 and 1148.5 as shown on a representative MALDI-TOF MS spectrum (Figure 1A). We performed a MALDI-TOF/TOF (MS/MS) analysis for each of these ions with the goal to identify these compounds. However, no chemical composition could be related (data not shown). In negative ion mode, we constantly found peaks in the low molecular mass region m/z 700 to 1100 including pseudomolecular ions at m/z 728.9, 750.9, 886.8 and 1078.8, but no oligosaccharides could be identified (Figure 1E).

Analysis of urine from three patients affected with fucosidosis revealed a major pseudomolecular ion at m/z 504.2 and a second less intense one at m/z 1079.4 (Figure 1B), for which we were able to deduce the chemical composition. Carbohydrate fragmentation generates several different types of cleavage, and cationization can occur on different atoms 33 39 40 . The MALDI-TOF/TOF analysis of the parent ion at m/z 504.2 revealed a characteristic fragmentation with the more intense fragment ions at m/z 389.2, 358.2 and 289.1 (Figure 1C), corresponding respectively to a loss of an asparaginyl residue, a loss of a fucosyl residue and to a fragmentation inside the HexNAc cyclic form as previously described ( 39 Additional file 1: Figure S2). Thus, the parent ion at m/z 504.2 corresponds to the [M + Na]⁺ ion of the Fuc-HexNAc-Asn (fucosyl-GlcNAc-Asparagine) oligosaccharide excreted in excess in urine of patients suffering from fucosidosis. Similarly, the MS/MS spectrum for the m/z 1079.5 parent ion revealed several specific fragments, among which the most intense one at m/z 933.4 corresponding to a loss of a fucosyl residue (Figure 1D). In negative-ion MS spectrum, we observed a peak at m/z 1518.5 with two weaker ones at m/z 1680.6 and 1883.7 for which we were able to deduce the chemical composition (Figure 1F), notably with the MS/MS analysis for the parent ion at m/z 1518.5 (Figure 1G). These oligosaccharides do not contain sialic acid, however they are detectable in negative mode thanks to the carboxyl group of asparagine.

Analysis of urine from two patients affected with aspartylglucosaminuria revealed a pseudomolecular ion at m/z 358.2 (Figure 2B). The MALDI-TOF/TOF analysis of the m/z 358.2 parent ion revealed a characteristic fragmentation with the more intense fragment ions at m/z 155.2 and 243.0 (Figure 2B), reflecting respectively a loss of an HexNAc residue, and the fragmentation of an asparaginyl residue as described above. This HexNAc-Asn compound predicted to be GlcNAc-Asn is known as the major stored compound in this disease (Figure 2A). We were also able to reproducibly detect lower intensity peaks at m/z 520.2, 811.3 and 885.3, which are expected to be glycoasparagine compounds as shown in Figure 2A. The negative-ion MS profile for urine from aspartylglucosaminuria affected patients showed two intensive peaks at m/z 787.2 and 809.2, corresponding respectively to [M-H]^- and [M-2H + Na]^- forms of the same compound, with some weaker peaks at m/z 1152.4, 1517.5 and 1882.6 (Figure 2C). The MS/MS analysis of the parent ion at m/z 787.2 gave characteristic fragments notably with the loss of a sialic residue identified at m/z 495.9 and the sialic residue at m/z 289.8 (Figure 2D).

Analysis of urine from five GM1 gangliosidosis patients revealed two major pseudomolecular ions found at m/z 933.3 and 1460.5 in addition to lower intensity ions at m/z 1095.4, 1298.5, 1663.6 and 1825.7 (Figure 3A). All these peaks are separated either by a loss of 162 or 203 amu corresponding respectively to the loss of a hexose or an N-acetylhexosamine residue (Figure 3A). Morever, the MS/MS analysis on the more intense parent ions at m/z 933.3 (Figure 3B) and 1460.5 (Figure 3C) shows with confidence the reproducible and characteristic fragmentation profile of the major glycocompounds accumulated in urine from GM1 gangliosidosis patients. In negative ion mode, no characteristic peak was observed on MS spectrum (Figure 3D).

Analysis of urine from four patients affected with Sandhoff disease revealed characteristic pseudomolecular ions at m/z 771.2, 1136.3 and 1339.4 with lower intensity ions at m/z 974.3 and 1542.4. All these peaks are separated from each other by an amu difference corresponding to a hexose or a HexNAc residue (Figure 3E). We were able to perform MS/MS on the three more intense peaks at m/z 771.2 (Figure 3F), 1136.3 (Figure 3G) and 1339.4 (Figure 3H) which showed a signature with fragments typical for the identification of the Sandhoff disease. No characteristic peak was observed in negative-ion MS spectrum for this disorder (Figure 3I).

Analysis of urine from seven patients affected with α-mannosidosis revealed a major peak at m/z 568.2 with additional peaks all separated by an amu difference corresponding to a hexose residue (Figure 4A). The MS/MS spectrum of the m/z 568.2 ion corresponds to a trisaccharide Hex-Hex-HexNAc with the specific peaks at m/z 365.1 agreeing with the loss of the HexNAc residue and at m/z 244.0 to the loss of two hexose residues (Figure 4B). This trisaccharide is probably Man-Man-GlcNac, which is the major compound described to be accumulated in α-mannosidosis patients (2). No characteristic peak was observed in negative-ion MS spectrum (Figure 4C).

Analysis in positive ion mode of urine from four patients affected with sialidosis could reveal two weak characteristic peaks at m/z 933.5 and 1460.6 (Figure 5A). These two peaks are detected with higher intensity in urine from GM1 gangliosidosis affected patients (Figure 3A). Thus, the chemical structure of these ions corresponds to the same oligosaccharides as in GM1 gangliosidosis, i.e. oligosaccharides that have spontaneously lost their sialic acid residues during ionization/fragmentation in positive ion mode. Interestingly, in negative ion mode, the MS spectrum showed a pseudomolecular ion at m/z 1200.4 and less intense ions at m/z 1362.5, 1565.5, 1727.6 (Figure 5B). Among them, those at m/z 1200.4 and 1727.6 correspond to the oligosaccharides found in positive mode, with an additional sialic acid residue. The biantennary-disialylated oligosaccharide revealed three different peaks at m/z 2018.6, 2040.6 and 2056.5 corresponding respectively to [M-H]^-, [M-2H + Na]^-, and [M-2H + K]^- forms of the compound (Figure 5B). The MS/MS profile for the parent ion at m/z 1200.4, corresponding to the major sialylated compound, gave characteristic fragments with the sialic residue at m/z 289.6 (Figure 5C). Thus, the negative ion mode analysis allowed to detect sialylated oligosaccharides, which is not the case with the positive ion mode.

Finally, analysis of urine from six patients affected with MLII and four MLIII revealed MS and MS/MS spectra, in positive and negative ion modes, overall comparable from those obtained with urine from sialidosis affected patients (Figure 5). For some patients, biantennary-disialylated oligosaccharides were less detectable than for sialidosis. No urine from galactosialidosis affected patients could be analyzed, because of the rarity of the disease, but a similar pattern can be anticipated.

The MS and MS/MS analyses performed on the training set of urine samples from patients known to suffer from various oligosaccharidoses enabled to identify characteristic accumulated oligosaccharides in urine and typical patterns that are useful for the diagnosis of new cases. The specificity of the method was then validated by blind analysis of a series of fifty-three samples including urine from control and oligosacharidosis affected patients (validation set, Table 1). Considering the rarity of these disorders, we have on purpose not taken into account the severity of the pathology when urine samples were randomly divided into the training and validation sets. The objective was to develop a specific assay allowing us to determine the predictive value of our protocol for oligosaccharidose screening in any circumstances.

Specific MS and MS/MS profiles of fucosidosis (2 cases), aspartylglucosaminuria (1 case), GM1 gangliosidosis (7 cases), Sandhoff disease (2 cases) and alpha-mannosidosis (4 cases) were identified from the above described characteristic patterns. Seven profiles corresponding to sialidosis/MLII/MLIII profiles were also recognized. After removal of patient anonymity, all the profiles observed were in perfect agreement with the previous diagnosis. None of the thirty control subjects showed any of the peaks characteristic for oligosaccharidosis.

In summary, analysis of ninety-four urine samples, with the training and the validation sets, made it possible to identify fifty-eight characteristic pathological patterns from fifty-eight urine samples previously diagnosed by TLC/enzymatic assays for different oligosaccharidoses.