Kidney Sulfatides in Mouse Models of Inherited Glycosphingolipid Disorders

Sulfatides show structural, and possibly physiological similarities to gangliosides. Kidney dysfunction might be correlated with changes in sulfatides, the major acidic glycosphingolipids in this organ. To elucidate their in vivo metabolic pathway these compounds were analyzed in mice afflicted with inherited glycosphingolipid disorders. The mice under study lacked the genes encoding either β-hexosaminidase α-subunit (Hexa−/−), the β-hexosaminidase β-subunit (Hexb−/−), both β-hexosaminidase α and β-subunits (Hexa−/− and Hexb−/−), GD3 synthase (GD3S−/−), GD3 synthase and GalNAc transferase (GD3S−/− and GalNAcT−/−), GM2 activator protein (Gm2a−/−), or arylsulfatase A (ASA−/−). Quantification of the sulfatides,I3SO 3 − -GalCer (SM4s), II3SO 3 − -LacCer ( SM3 ),II3SO 3 − -Gg3Cer (SM2a), and IV3,II3-(SO 3 − )2-Gg4Cer (SB1a), was performed by nano-electrospray tandem mass spectrometry. We conclude for the in vivo situation in mouse kidneys that: 1) a single enzyme (GalNAc transferase) is responsible for the synthesis of SM2a and GM2 from SM3 and GM3, respectively. 2) In analogy to GD1a, SB1a is degraded via SM2a. 3) SM2a is hydrolyzed to SM3 by β-hexosaminidase S (Hex S) and Hex A, but not Hex B. Both enzymes are supported by GM2-activator protein. 4) Arylsulfatase A is required to degrade SB1a. It is probably the sole sphingolipid-sulfatase cleaving the galactosyl-3-sulfate bond. In addition, a human Tay-Sachs patient's liver was investigated, which showed accumulation of SM2a along with GM2 storage. The different ceramide compositions of both compounds indicated they were probably derived from different cell types. These data demonstrate that in vivo the sulfatides of the ganglio-series follow the same metabolic pathways as the gangliosides with the replacement of sulfotransferases and sulfatases by sialyltransferases and sialidases. Furthermore, a novel neutral GSL, IV6GlcNAcβ-Gb4Cer, was found to accumulate only in Hexa−/− and Hexb−/− mouse kidneys. From this we conclude that Hex S also efficiently cleaves terminal β1–6-linked HexNAc residues from neutral GSLs in vivo.

Sulfatides, (designating all sulfated glycosphingolipids) such as galactosylceramide I 3 -sulfate, occur enriched in the myelin sheets of the central and peripheral nervous system and in glandular epithelial tissues of mammals. Sulfatides of more complex structure have been found in the kidney (1). In the human renal cell carcinoma cell line SMKT-R3 high levels of sulfatides including gangliotriaosylceramide-II 3 sulfate (SM2a) 1 were observed (2) that they may modulate the metastatic potential of these cells (3). In addition, complex sulfatides have been recognized to rank among the strongest ligands for NKR-P1. This membrane protein, with an extracellular Ca 2ϩdependent lectin domain, is expressed on natural killer cells that display innate immunity (4,5). Other proteins involved in innate immunity, properdin and factor H, have also been reported to bind specifically to the sulfatides (6). More recently it has been shown that intracellular sulfation of lactosylceramide suppresses the expression of integrins (7).
In mice, complex kidney sulfatides belong to the ganglioseries glycosphingolipids (GSL) and thus show structural similarity to "brain type"-gangliosides ( Fig. 1). This relationship is further suggested by largely identical carbohydrate substitution positions for sulfate and sialic acid. These mouse kidney sulfatides include lactosylceramide-II 3 sulfate (SM3), SM2a, and gangliotetraosylceramide-II 3, IV 3 bis-sulfate (SB1a) (8). In general, GSLs are synthesized from a ceramide core by modification with glycosyl-and sulfotransferases in the endoplasmic reticulum and Golgi. Degradation of GSL takes place at the surface of intra-lysosomal vesicles by the action of exoglycosi-dases, sulfatases, and sialidases. For several of these degradation steps, the presence of one of the five known lysosomal activator proteins is required (9). Defects in the lysosomal enzymes or activator proteins that degrade GSLs are the cause of severe human inherited diseases such as metachromatic leukodystrophy and the different forms of GM2 gangliosidosis.
Deficiency in arylsulfatase A in metachromatic leukodystrophy leads to lethal demyelination in the central and peripheral nervous systems. This disease is characterized by the lysosomal accumulation of the sulfatides SM4s, SM4g, and SM3 (10,11).
Defects in the ␤-hexosaminidase isozymes (Hex S, (␣/␣); Hex A, (␣/␤); and Hex B, (␤/␤)) or in the GM2 activator protein lead to GM2 gangliosidosis, in which ganglioside GM2 and related glycolipids accumulate in lysosomes mainly of neuronal cells. The GM2 gangliosidoses include Tay-Sachs disease (B-variant), due to mutations in the HEXA gene encoding the ␣-subunit of ␤-hexosaminidase, Sandhoff disease (0-variant), due to mutations in the HEXB gene encoding the ␤-subunit, and the AB-variant characterized by mutations in the GM2A gene. For all three enzyme defects, the severe infantile forms result in rapidly progressing neurodegeneration, culminating in death before age 4 years (12).
The human Tay-Sachs liver material was from a girl that died at the age of 2 years and 10 months. The girl was affected with the classical form of Tay-Sachs and symptoms were apparent from 8 months of age including hyperaccusis. The control liver material was from a healthy 42-year-old male donor who died in an accident.

Methods
Purification of Sulfated GSL from Kidney Tissue-SM3 was extracted from human kidney. SM2a and SB1a were isolated from rat kidney. In general, 100 g of tissue was homogenized on ice in 100 ml of distilled water with an Ultra Turrax T25 basic from IKA Labortechnik (Staufen, Germany) (6 ϫ 2 min of homogenizing at 24,000 rpm with pauses of 2 min in between). The homogenate was freeze-dried and subsequently extracted with acetone. The freeze-dried tissue then was extracted for GSL, 2 times with chloroform/methanol/water (C/M/W) (10/10/1) and once with C/M/W (30/60/8). The combined C/M/W extracts were concentrated and dialyzed against 5 ϫ 5 liters of distilled water. The dialyzed extract was lyophilized, dissolved in C/M/W (30/60/8) and loaded on DEAE A-25 column to separate neutral and acidic lipids. Elution was with a stepwise gradient of 20, 80, 200, 500, and 1000 mM methanolic potassium acetate (KAc). SM3 was eluted with the 200 mM, SM2a with the 80 mM, and SB1a in the 500 mM KAc fraction. The fractions were desalted by dialysis with 5 ϫ 5 liters of distilled water and lyophilized. From the corresponding fractions, the sulfated GSLs were further purified by repeated silica gel flash column chromatography with the appropriate mixtures of n-hexane/isopropyl alcohol/water or C/M/W as running solvent systems.
Quantification of Purified GSL TLC Standards by Anthrone Reaction (25)-Commercially available, or from tissue isolated and purified, GSLs were dissolved in C/M/W (10/10/1). Aliquots in the range of 5-20 nmol were dried in an 1.5-ml test tube with a gentle stream of nitrogen. 100 l of water and 500 l of anthrone reagent were added. Then the cups were sealed and clamped in between two metal plates so that the lids could not open. One of the plates had appropriate holes for the lower part of the cups to fit through. The cups were incubated for 15 min at 100°C and then cooled in a water bath at room temperature for further 20 min. Calibration curves were obtained with samples containing a mixture of the free sugars in the appropriate equimolar ratio equaling the ratio of the individual GSL to be quantified (as an example: for SB1a quantification, Glc, Gal, and GalNAc were mixed in the ratio 1:2:1). 300 l of each sample was placed into a flat-bottomed transparent 96-well microtiter plate. Absorption was measured at 620 nm.
Synthesis of GSL Standards for Nano-ESI-MS/MS-Lyso-SM3, lyso-SM2a, and lyso-SB1a as well as lyso-GM3 and lyso-GM2 were obtained by treatment of the purified compounds with SCDase according to Ref. 26. The crude products were purified by silica gel column flash chromatography using an appropriate mixture of C/M/W as running solvent system. (Prior to activation 2-hydroxy fatty acids were esterified with acetic anhydride as follows: 2-hydroxy-myristinic acid was dissolved in 200 l of anhydrous and alcohol-free chloroform ϩ 50 l of acetic anhydride ϩ 50 l of 0.1% N,N-dimethylaminopyridine in chloroform. The reaction mixture was incubated for 60 min at 37°C. Then 500 l of toluol were added and the sample dried under a gentle stream of nitrogen at 37°C. The sample was dissolved again in 200 l of toluol and dried again as before.) Fatty acids (65 mol) were dissolved in 4 ml of dry tetrahydrofuran under nitrogen gas and activated with 0.82 equivalents of dicyclohexylcarbodiimide and 0.93 equivalents of N-hydroxysuccinimide. Reaction took place overnight at room temperature.
For condensation, about 100 nmol of lyso-GSL was dissolved in 2 ml of dry N,N-dimethylformamide and 4 l of triethylamine. Then 2 ml of the activated fatty acid were added. The reaction mixture was incubated at room temperature for 2-5 days and monitored by TLC. Upon the long incubation some GSL were also acylated at hydroxyl groups. This resulted in a smear on TLC, running faster than the GSL standard. The reaction mixtures were, therefore, treated with 0.1 M methanolic KOH for 2 h at room temperature. This mild base treatment was also necessary to remove the acetate ester from the 2-hydroxy group of SM4s (18:1,h14:0). The crude products were purified by silica gel column flash chromatography.
Quantification of the Synthesized GSL MS Standards-Aliquots of the synthesized GSL MS standards in the range of 0.3-1.0 nmol were spotted on TLC using a Linomat IV from CAMAG (Muttenz, Switzerland). On adjacent lanes corresponding GSL TLC standards were spotted in different concentrations to obtain calibration curves. After development in chloroform, methanol, 0.2% aqueous CaCl 2 (60/35/8) GSL bands were developed with orcinol/sulfuric acid spray reagent at 110°C for 20 min or with 10% CuSO 4 in 8% H 3 PO 4 at 150°C for 20 min. The amount of the GSL compounds was determined by densitometric scanning of each lane at a wave length of 440 nm (Shimadzu CS-9301 TLC scanner). For each GSL the C14, C19, and C27 fatty acid containing compounds were mixed in an equimolar ratio resulting in the ready-touse MS standard of known concentration.
Extraction of GSLS from Murine Kidney for Mass Spectrometric Analysis-Kidneys wet weight was determined, and the kidney homogenized on ice in 5 ml of distilled water with a Ultra Turrax T25 basic (6 ϫ 30 s of homogenizing at 24,000 rpm with pauses of 30 s in between).
GSL MS standards were transferred to glass tubes and the solvent was evaporated with a gentle nitrogen stream. An aliquot of the aqueous kidney homogenate, equal to 20 mg of organ wet weight, was added, and the sample was sonicated for 5 min. Thereafter, the sample was lyophilized and extracted 2 times with 2 ml of acetone. The residual pellet then was extracted twice with 1.5 ml C/M/W (10/10/1) and with 2 ml of C/M/W (30/60/8) for GSL. Neutral and acidic GSL of the combined C/M/W extracts were separated on DEAE A-25 columns (bed volume: 300 l). The flow-through and wash yielded fraction 1, containing the neutral GSL. Acidic GSL, collected as fraction 2, were eluted with 4 ml of 500 mM methanolic potassium acetate. Solvent was evaporated and acid GSL (fraction 2) desalted with RP-18 (200 mg RP-18 material per column) column chromatography. Fraction 1 was dissolved in 100 l of 5 mM methanolic ammonium acetate, and fraction 2 in 100 l of methanol. If necessary, samples were further diluted for nano-ESI-MS/MS.
Extraction of GSLS from Human Liver for Mass Spectrometric Analysis-Human liver GSLs were extracted in analogy to the mouse kidney protocol. Since the Tay-Sachs liver was stored frozen for more than 25 years it lost barely any weight by freeze drying. Therefore GSL concentrations were calculated per mg dry weight. Extraction without MS standards was carried out with 150 mg dry weight, introducing MS standards with 20 mg.
Determination and Characterization of Sulfatides and GSLS by Nano-ESI-MS/MS-All analyses were performed with a triple quadrupole instrument (VG micromass (Cheshire, UK) model Quattro II) equipped with a nano-electrospray source operating at an estimated flow rate of 20 -50 nl/min. Usually, 10 l of a samples, dissolved in methanol or methanolic ammonium acetate (5 mM), was filled into a gold-sputtered capillary. The capillary was positioned at a distance of 1-3 mm in front of the cone. The source temperature was set to 30°C and the spray was started by applying 800 -1200 V to the capillary. For each spectrum 20 -50 scans of 15-30 s duration were averaged. All tandem MS experiments were performed with argon as collision gas at a nominal pressure of 2.2-2.7 ϫ 10 Ϫ3 mbar. The parameters for the cone voltage and the collision energy of the different scan-modes are listed in Table I.
Evaluation of the Nano-ESI-MS/MS Data and Quantification of Lipids-Quantitative spectra were measured with an average mass resolution of 1200 (ion mass/full width half-maximum). Peak height values of the first mono-isotopic peak of each compound were taken for evaluation. From the peak intensities of the corresponding internal standard lipids a linear trend was calculated. The obtained calibration curve represented the intensity of the internal standard molar amount at a given m/z value. In addition, a linear trend for nϩ2 molecular isotopic signal intensities (molecules containing either two 13 C-atoms or one 34 S-atom, and, thereby, shifted by m/z 2 upwards) was calculated from the internal standards. If necessary, signal intensities were corrected first from influence of nϩ2 signal overlap. This overlap appears if lipids, that contain one additional double bond, are present. Then their nϩ2 signal overlaps with the main signal of the lipid without this double bond. From the corrected intensity ratio (sample lipid/internal standard trend) and the amount of internal standard added the quantity of the individual molecular species (e.g. SM4s (18:1, 16:0) or SM4s (18:1, 24:1) etc.) was calculated. From the sum of individual molecular species then the amount of a lipid (SM4s) resulted. Endogenous SB1a, GM3, and GM2 were correlated to the sole corresponding standard.
Extraction of GSLS from Murine Kidney for TLC Analysis-For TLC analysis, 50 mg of kidney wet weight were extracted as above using the appropriate volumes. The neutral and acidic GSL fractions were each taken up in 100 l of C/M/W (10/10/1). Aliquots according to the Figure  legends were spotted on TLC plates with a Linomat IV from CAMAG (Muttenz, CH). A pre-run was performed with chloroform/alcohol (C/A) (1/1). Then the plates were dried and GSL were separated with the running solvent chloroform, methanol, 0.2% aqueous CaCl 2 (60/35/8), if not otherwise noted. Bands were detected with orcinol/sulfuric acid spray reagent at 110°C for 10 -20 min.
Carbohydrate Constituent Analysis-Carbohydrate constituents were released by acid hydrolysis after hydrofuran treatment, converted into their corresponding alditol acetates and analyzed by capillary GC/MS as detailed elsewhere (27).
Carbohydrate Permethylation Analysis-For determination of linkage positions of monosaccharide constituents, glycolipids were permethylated and hydrolyzed (28). Partially methylated alditol acetates obtained after sodium borohydride reduction and peracetylation were analyzed by GC/MS using the instrumentation and microtechniques described previously (29,30).
Each standard solution was quantified by densitometric scanning of the orcinol/sulfuric acid-or CuSO 4 /phosphoric acidstained TLC band. On TLC, standards with C14, C19, or C27 fatty acid migrated sequentially faster as compared with one another (data not shown). For mass spectrometric quantification of each sulfatide, the three respective fatty acid derivative standards were mixed in an equimolar ratio. Since the concentration of the different sulfated GSLs, i.e. SM4s, SM3, SM2a, and SB1a, was not identical in murine kidney (Fig. 3 (Fig. 4). Correlating the endogenous sulfatide signals to those of the corresponding standards levels of kidney sulfatides were quantified as described under "Experimental Procedures."

At Higher Collision Energies 2-Hydroxy Fatty Acid-containing Sulfatides Are Measured in the Precursor Ion Mode (m/z Ϫ97) with the Same Abundance as Sulfatides Containing
Non-hydroxy Fatty Acids-Since sulfatides SM4s with a 2-hydroxy fatty acid give rise to additional product ions (due to a break between the carboxyl-carbon-and the ␣-carbon-atom of the fatty acid), this might affect the relative abundances of the common fragments (e.g. [HSO 4 ] Ϫ used for quantification) (31,32). These additional fragments could be detected for SM4s (2hFA) at collision energies of 50 -60 eV with not more than 7% of the intensity of fragment m/z Ϫ97 ([HSO 4 ] Ϫ ). At collision energies of 90 -115 eV that were used to quantify SM4s in the precursor ion mode, these fragments were not detectable or had an abundance smaller than 0.2%. For SM3 and SM2a additional fragments due to a 2-hydroxy fatty acid could also be detected at collision energies of 65-70 eV with up to 7% abundance relative to m/z Ϫ97. But none of these fragment appeared at collision energies of 90 -115 eV, which were relevant for quantification.
Linearity of the Mass Spectrometric Method in Comparison to TLC Densitometry-To test the linearity of the mass spectrometric method, a constant amount of SM4s standard (272 pmol) was mixed in several samples with different amounts of bovine brain sulfatide (8.5 to 17.7 nmol). The values obtained by mass spectrometry as plotted against the amounts used showed that linearity was achieved from 35 to 8830 pmol (Fig.  5). The average concentration evaluated from the 9 data points in this range differed by 1.7% from the theoretical value with a standard deviation of 8%. Since bovine brain sulfatide is a mixture of sulfatides with different ceramide compositions, values obtained for some representative individual sulfatides FIG. 2. Generation of lyso-SM2a by enzymatic digestion of rat kidney SM2a. SM2a was purified from rat kidney and hydrolyzed for 24 h with the enzyme sphingolipid ceramide N-deacylase (SCDase) from Pseudomonas sp. as described under "Experimental Procedures." A, reaction products of the aqueous phase were taken up in C/M/W (10/10/1, v/v), separated on TLC with running solvent C/M/0.2% aqueous CaCl 2 (45/45/10) and stained for sugars with orcinol/sulfuric acid. Lane 1, purified SM2a from rat kidney; lane 2, SCDase digest of SM2a. Whereas GSL stained purple, taurodesoxycholate (TDC), used in the assay, turned light blue after several hours at room temperature, and no remaining SM2a staining could be observed in the digest. Densitometric quantification of the product bands revealed a ratio of 57: 100 for sphingosyl-lyso-SM2a to phytosphingosyl-lysoSM2a. B, nano-ESI-MS/MS precursor ion m/z 97-spectra of rat kidney SM2a (i) and its products of SCDase digestion (spectrum ii) corresponding to lanes 1 and 2 of A), respectively. m/z Ϫ97 represents the fragment [HSO 4 ] Ϫ produced in the collision chamber. By this scan only compounds bearing a sulfate group (m/z Ϫ97) are detected and plotted in the spectrum. Therefore no TDC m/z Ϫ1019.5, carrying a sulfono-(giving rise to [⅐SO 3 ] Ϫ with m/z 80) but not a sulfate group, is detected. A ratio of 59:100 for lyso-SM2a (18:1) to lyso-SM2a (h18:0) was determined.
were also plotted in this diagram to demonstrate linearity for the individual species. The results indicate that individual signal intensities down to 2.5% and up to 1000% of the standard signal intensities were in the range of linearity.
Kidney of Wild Type Mice Contain the Sulfated GSLS SM4S, SM3, and SB1A-Acidic GSLs were isolated and separated on TLC as described. Staining with orcinol/sulfuric acid revealed GSLs with migration rates comparable to SM4s, SM3 and SB1a and GM3 (Fig. 3A, lane 3). Except for the compound migrating with GM3, the TLC bands also stained with azur A indicating that they are sulfated glycolipids (data not shown). Quantification by nano-ESI-MS/MS revealed SM4s, SM3, and SB1a to make up 83, 10, and 7% of the sulfated GSLs of mouse whole kidney in close agreement with an earlier report by Tadano-Aritomi and co-workers (8) (Fig. 6A; Table II). As compared with their data, however, the present analyses showed an ϳ1.5 times higher concentration of sulfated GSLs. SM2a and SB2, that are present in rat kidney, could not be detected in the kidney of wild type mice. Considering the limiting background noise of the mass spectra obtained, the concentration of SM2a was calculated to be less than 1.4 pmol/mg wet weight corresponding to less than 0.3% of total sulfated GSL.
With regard to mouse kidney sulfatide ceramide composition, C18-sphingosine was the most prominent sphingoid with less than 6% of additional C18-phytosphingosine and 60 -70% fatty acids of C22-and C24-aliphatic chain length. In addition, fatty acids of C-16, C-18, C-20, C21, C-23, C26, and C28 chain length were also detected. More than 75% of the fatty acids were saturated and the amount of 2-hydroxylated fatty acids, ϳ60% of the total, was twice as high for SM4s than for SM3 and SB1a with ϳ30%. 2-Hydroxylation was identified by both, molecular mass in mass spectrometry, as well as, the additional fragments m/z 522, 540, and 568, that appeared in the corresponding product ion spectra of SM4s (data not shown). These fragments have been reported to be characteristic for sulfatide with 2-hydroxy fatty acids (31,32).
Gg 3 Cer, Gb 4 Cer, and IV 6 GlcNAC␤-Gb 4 Cer Accumulate in HexaϪ/Ϫ and HexBϪ/Ϫ Kidney-Neutral and acidic GSLs of double mutant HexaϪ/Ϫ and HexbϪ/Ϫ mice were isolated. As compared with the wild type mouse, TLC of the neutral GSLs revealed two double and one single bands that stained intensely with orcinol/sulfuric acid indicating the accumulation of three glycolipid components (Fig. 7, lane 1). The upper double band had a TLC migration rate corresponding to Gg 3 Cer, and the lower with Gb 4 Cer. Both GSLs are known to accumulate in these mice. The lower single TLC band, designated compound X, showed a migration between Forssman glycolipid and Gg 4 Cer.
Investigating the neutral GSL fraction in nano-ESI-MS/MS with a precursor ion scan of m/z 264 significantly increased signals for neutral GSL with the sequence Cer-Hex-Hex-Hex-NAc (as in Gg 3 Cer) and Cer-Hex-Hex-Hex-HexNAc (as for Gb 4 Cer) were detected, as compared with wild type kidney. m/z ϩ264 represents the protonated and dehydrated C18-sphingosine base, which is obtained as a characteristic fragment of neutral GSLs under these conditions. The ascribed sequence was confirmed from the collision induced fragments obtained from these molecules (data not shown). Scanning for higher neutral GSLs in nano-ESI-MS/MS, we also used a precursor ion scan of m/z ϩ204. m/z ϩ204 represents a protonated and dehydrated HexNAc residue which should be present at the terminus in all storage compounds of this mutant mouse. By both of these scans, signals for a GSL containing 3 Hex and 2 HexNAc residues could be identified that were not present in wild type kidney. Thus, the third accumulating GSL, compound X, contained five sugar residues.
Comparing the collision induced fragments of the protonated storage compound X with that of protonated Forssman glycolipid by nano-ESI-MS/MS indicated that the characteristic fragments were identical (Fig. 8A). From these data the structure for compound X could be assigned as HexNAc-HexNAc-Hex-Hex-Hex-Cer.
Since in Forssman glycolipid the terminal HexNAc residue is ␣-glycosidically linked and not a substrate for ␤-hexosaminidase, it is assumed not to accumulate in the GM2 gangliosidosis mice. In addition, the TLC band of compound X did not co-migrate with the Forssman lipid standard. Compound X was further analyzed by nano-ESI-MS/MS. Comparing the collision induced fragments of the deprotonated compound X and Forssman glycolipid in the negative product ion mode of nano-ESI-MS/MS, distinct differences were observed. First, the storage compound did not yield a fragment of m/z 154 that appeared in Forssman lipid standard from sheep erythrocytes (Fig. 8B, ii) or from chicken heart (data not shown). Second, a fragment with m/z 322, not present in Forssman lipid, appeared with compound X (Fig. 8B, i). This is a terminal fragment produced by ring cleavage between C2-C3 and C5-oxygen ring of the subterminal HexNAc residue. To ensure that this fragment was not due to impurities, the neutral GSL fraction was scanned for compound X using this fragment in a nano-ESI-MS/MS precursor ion mode. The storage compound with the same ceramide pattern (C18-sphingosine combined with 16:0, 22:0, 24:1, and 24:0 fatty acids) as described before with a precursor ion scan of m/z 220, representing the deprotonated terminal HexNAc-residue was again detected (Fig. 8A, iii and  iv). Since Forssman glycolipid from chicken heart had a distinctly different ceramide composition, including ceramide (C18-sph,18:0) and (C18-sph,20:0) (Fig. 9A, ii), it was admixed with the GSL fraction containing compound X. Both compounds could be detected when scanning the mixed sample either in nano-ESI-MS/MS total negative ion mode (Fig. 9B, i), or with collision induced fragments (m/z 405, Fig. 9B, ii, or m/z 220, Fig. 9B, iii) that appear in the product ion scans of both compounds. In contrast, by scanning with the compound X collision-induced fragment the m/z 322 only compound X could be detected; no signals for Forssman glycolipid appeared (Fig.   FIG. 4. Nano-ESI-MS/MS spectrum of a MS standard mixture for sulfatide determination. Synthetic sulfated GSL standards were mixed as follows: C14-, C19-, and C27-SM4s: 4.8 pmol/l each; C14-, C19-, and C27-SM3: 1.74 pmol/l each; C14-, C19-, and C27-SM2a, 1.14 pmol/l each; and C19-SB1a, 1.69 pmol/l. The mixture was scanned by nano-ESI-MS/MS in negative mode using a precursor ion scan with m/z Ϫ97 (corresponding to [HSO 4 ] Ϫ ) specific for sulfated compounds (31,32,40,53,54). Aliquots of this mixture later were added to the kidney GSL samples for quantification.  Fig. 4). Clearly, accumulating SM2a is detected in high amounts in B, whereas in C, no kidney SB1a could be detected (*). 9B, iv). Therefore, compound X with the structure HexNAc-HexNAc-Hex-Hex-Hex-Cer, must be different from Forssman glycolipid. For a further investigation of the nature of compound X, the glycolipid was isolated by preparative TLC. Subsequent carbohydrate constituent analysis by GC/MS revealed the monosaccharides Gal, Glc, GlcNAc, and GalNAc in the ratio (2.0:1.25:1.0:0.9). And additional permethylation analysis identified 4-substituted Glc, 4-substituted Gal, 3-substituted Gal, 6-substituted GalNAc, and terminal GlcNAc (data not shown).
It is known that in mouse kidney a characteristic globo-/ neolacto-series glycolipid occurs, Gal␤1-4(Fuc␣1-3)GlcNAc␤1-6(Gal␤1-3)Gb 4 Cer, which could be detected by nano-ESI-MS/MS in wild type and in mutant kidney samples (data not shown). Therefore, it appears highly likely that compound X is an accumulated degradation product of this glycolipid with its remnant N-acetylhexosamine-terminal core structure IV 6 -GlcNAc␤-Gb 4 Cer.
Besides C18-sphingosine and non-hydroxylated fatty acids of C16 up to C24 aliphatic chain length, hydroxy fatty acids, as well as, phytosphingosine were determined in Gb 4 Cer and Gg 3 Cer. This explains the appearance of TLC double bands for both of these glycolipids (data not shown).
SM2A Accumulates in HexAϪ/Ϫ and HexBϪ/Ϫ Mouse Kidney-Separation of the acid GSL fraction on TLC and stain-ing with orcinol/sulfuric acid revealed a new prominent band running at the level of SM2a that does not appear in wild type kidney. No significant increase of a band at the level of GM2 was observed.
Whereas quantification of SM4s, SM3, or SB1a by nano-ESI-MS/MS showed no significant changes in concentrations, a large amount of SM2a (239 pmol/mg wet weight) was identified in kidney from a 9-week-old mutant mouse. This corresponds to an SM2a increase of at least 172-fold as compared with kidney from a wild type mouse (Table II and Fig. 10). No significant changes in the ceramide compositions of SM4s, SM3, and SB1a, or accumulated SM2a compared with wild type were detected.
SM2A but Not Neutral GSLS Accumulate in HexaϪ/Ϫ Kidney-TLC analysis of the neutral GSLs of HexaϪ/Ϫ mice kidney showed no significant differences as compared with wild type (data not shown). In contrast, the acidic GSL component profile was characterized by a prominent band running at the level of SM2a that was not present in lipids of wild type kidney (Fig. 3, top, lane 5). However, no significant increase of a TLC band at the level of GM2 could be observed. Similar to TLC analysis, quantification of the sulfated GSL by nano-ESI-MS/MS revealed no significant changes in SM4s, SM3, or SB1a concentrations (Fig. 6B). However, in the case of kidney from 19-and 20-week-old HexaϪ/Ϫ mice, large amounts (248 Ϯ 18  pmol/mg wet weight) of SM2a were detected that corresponded to an average increase of at least 180-fold as compared with the wild type (Table II and Fig. 10). No significant changes in the ceramide compositions of sulfated GSL compared with wild type were detected The SM2a pattern was similar to that of HexaϪ/Ϫ and HexbϪ/Ϫ mice.
SM2A Accumulates in HexbϪ/Ϫ Kidney-In the case of the HexbϪ/Ϫ mouse mutant, TLC of the kidney acidic GSLs showed the appearance of a faint band migrating identically to the SM2a standard (Fig. 3, top, lane 6). By nano-ESI-MS/MS, 24.7 Ϯ 0.05 pmol of SM2a per mg wet weight was quantified in a 13-and 18-week-old mutant kidney corresponding to an average increase of at least 18-fold as compared with the wild  (Table II, Fig. 10). No significant changes in SM4s, SM3, or SB1a-concentrations were found. No significant differences in the ceramide compositions of sulfated GSL compared with wild type were detected. The SM2a pattern was similar to that of HexaϪ/Ϫ and HexbϪ/Ϫ mice.
SM2A Accumulates in Gm2aϪ/Ϫ Kidney-TLC analysis of the neutral GSL fraction from Gm2aϪ/Ϫ kidney showed no significant differences as compared with wild type kidney (data not shown). A faint TLC band of the acidic GSLs, not seen in the wild type, co-migrated with SM2a (Fig. 3, top, lane 4). As determined by nano-ESI-MS/MS, kidney from 23-week-old mutant mice contained 7.1 Ϯ 1.8 pmol of SM2a per mg wet weight corresponding to an average increase of at least 5-fold over wild type (Table II and Fig. 10).
No significant changes in the ceramide compositions of sulfated GSL compared with wild type were detected. The SM2a pattern was similar to that of double mutant HexaϪ/Ϫ and HexbϪ/Ϫ mice.
All Sulfatides Accumulate in Arylsulfatase A-deficient Kidney-The neutral GSLs of ASAϪ/Ϫ mouse kidney were not different from the wild type (data not shown). In contrast, TLC of the mutant mouse kidney acidic GSL fraction showed strong increases in bands co-migrating with SM4s, SM3, and SB1a. All were stained by orcinol/sulfuric acid spray reagent (Fig. 3, bottom, lane 3) and with azur A (data not shown).
In a 11-week-old ASAϪ/Ϫ kidney, 11-, 4.4-, and 15-fold accumulation of SM4s, SM3, and SB1a, respectively, was quantified by nano-ESI-MS/MS. Analysis of a kidney from a 1-yearold ASAϪ/Ϫ mouse demonstrated a further increase in the accumulation of SM4s, SM3, and SB1a to about 80-, 40-, and 60-fold, respectively. However, no further increase in the accumulation of these GSLs was seen in a 2-year-old ASAϪ/Ϫ FIG. 9. Mass spectrometric differences between Forssman glycolipids and compound X from Hexa؊/؊ and Hexb؊/؊ kidney. A, compositions of Forssman glycolipid from chicken heart or from sheep erythrocytes, and from compound X, respectively, as measured by precursor ion scanning. According to the fragmentation patterns of Forssman glycolipid and compound X in Fig. 7B, the dominant fragment m/z 405 was taken to measure Forssman glycolipid from sheep erythrocytes (i) and from chicken heart (ii), and the dominant fragments m/z 220 (iii) and 322 (iv) were taken to scan for compound X in the neutral GSL fraction of HexaϪ/Ϫ and HexbϪ/Ϫ kidney.  , ii) with the neutral GSL fraction from HexaϪ/Ϫ and HexbϪ/Ϫ kidney (A, iii or iv). (i) total negative ion spectrum; (ii) precursor ion scan m/z Ϫ405; (iii) precursor ion scan m/z Ϫ220; and (iv) precursor ion scan m/z Ϫ322. Whereas both compounds, Forssman glycolipid and compound X, are detected with the precursor ions of m/z Ϫ220 (iii) and Ϫ405 (ii) (although with different sensitivities), the precursor ion m/z Ϫ322 (iv) is specific for compound X; no Forssman glycolipid (lack of m/z 1457 and 1485) is detected and the intensity pattern in (iv) returns to that of the pure neutral GSL fraction as shown (A, iv).  Table II. For wild type, SM2a detection limit (1.39 pmol/mg wet weight) was set to 100% since no SM2a could be detected in wild type. Age of mice in weeks is indicated on the y axis.
kidney. Sulfatide concentrations were very similar to those of the 1-year-old ASAϪ/Ϫ kidney (Table III and Fig. 11). No significant changes in the ceramide constituent compositions of SM4s, SM3, and SB1a compared with wild type were detected.
Mice Deficient in ␤-GalNac Transferase Lack SB1A in the Kidney-No significant differences, as compared with wild type, were observed by TLC for the neutral kidney GSLs of GD3SϪ/Ϫ and GalNAcTϩ/ϩ, GD3SϪ/Ϫ and GalNAcTϩ/Ϫ, and GD3SϪ/Ϫ and GalNAcTϪ/Ϫ mutants (data not shown). With regard to the acidic GSLs, the TLC profile of the GD3SϪ and GalNAcTϪ/Ϫ kidney was characterized by the disappearance of a band co-migrating with the SB1a standard (Fig. 3,  bottom, lane 7). Quantification of the sulfated GSL in these mutants revealed a 20% decrease of SB1a with a corresponding increase in SM3 in GD3SϪ/Ϫ and GalNAcTϩ/Ϫ kidney as compared with GD3SϪ/Ϫ and GalNAcTϩ/ϩ mutant. Kidneys from GD3SϪ/Ϫ and GalNAcTϪ/Ϫ mutant mice showed an increase in SM3 with SB1a being undetectable (Table IV).
SM2A Accumulates in Addition to GM2 in a Tay-Sachs Patient's Liver-Acidic GSL were extracted with and without internal MS standards from a Tay-Sachs patient's and a control human liver. In both livers comparable amounts (Ϯ10%) of GM3 were detected in good agreement with the data published by Nilsson and Svennerholm (33). In addition, significant amounts of SM2a and GM2 could only be detected in the Tay-Sachs liver ( Table V). The ceramide composition of all three GSLs, GM3, GM2, and SM2a, was different from each other, whereby that of GM3 and GM2 were comparable to the values reported earlier (33) (Figs. 12 and 13). With 57% GM2 containing stearic acid (GM2(18:1,18:0)) was the main GM2 species. In contrast to this, with 31% SM2a containing a C24: 1-fatty acid (SM2a(18:1,24:1)) was the main SM2a species. DISCUSSION The acidic glycolipids, because of their negative charge and often complex and seemingly systematic structures appear as particularly enigmatic with regard to their physiological significance. Naturally occurring genetic mutants that affect their metabolism have so far been observed only for deficiencies in catabolizing enzymes causing glycolipid accumulations. For the study of such inherited sphingolipid storage diseases, and more recently, with a more basic view on the possible elucidation of GSL functions, mutant mice models have been created that lack enzymes for the synthesis and degradation of GSL. It has been shown that most of the introduced mutations severely affect the presence or concentration of gangliosides. Sulfatides have not been systematically investigated. However, in view of the overall similarity between sulfatides and gangliosides, these mutant mouse models may also be expected to display derangement of the cellular sulfatide component profile. Kidney, particularly in the mouse, is the organ in which sulfatides are highly concentrated and have been associated with renal transport and metabolism (1,34). Therefore, an investigation was initiated to characterize alterations of sulfatides in mutant mouse models. The sulfatide components were characterized and quantified by nano-electrospray tandem mass spectrometry. This technique has already been successfully applied to characterize and quantify different sphingolipids (31,32,(35)(36)(37)(38)(39)(40)(41)(42)(43)(44). Internal standards were synthesized from lyso-sulfatides containing fatty acids of unusual chain length. Lyso-sulfatides were obtained by enzymatic cleavage of fatty acids from the parent GSL compounds using sphingolipid ceramide N-deacylase (SCDase) from Pseudomonas sp. (26). In the case of somewhat alkali-sensitive GSLs such as SM2a and SB1a, the enzymatic method of ceramide fatty acid cleavage offers advantage over the chemical cleavage.
Quantification of sulfatide by mass spectrometry was linear over more than 2 orders of magnitude with R 2 greater than 0.995 from 18 to 8830 pmol. In contrast, the conventional quantification of GSLs by densitometric scanning on TLC plates was only linear in the range of 100 to 700 pmol (with R 2 of 0.995). Accordingly, the linear range of quantification by nano-ESI-MS/MS was ϳ70 times greater than that of densitometric scanning. Furthermore, nano-ESI-MS/MS proved to be more sensitive than chemical staining of GSLs on TLC. In addition, mass spectrometric quantification gives more detailed information about the quantitative ceramide compositions of each individual GSL species, which could vary between different samples. No differences beyond the accuracy of the method could be detected when comparing the mass spectrometric sensitivity of 2-hydroxy fatty acid-containing sulfatide SM4s (18:1,h14:0) with sulfatides containing non-hydroxylated fatty acids (collision energy: 90 -115 eV). Nevertheless, at lower collision energies (50 -60 eV) detection of sulfatides SM4s with a 2-hydroxy fatty acid was about 10% less sensitive than for sulfatides without 2-hydroxy fatty acid. At these energies, additional fragments due to a break between the carboxyl-carbonand the ␣-carbon-atom of the fatty acid are produced (31,32), which seem to influence the abundance of the sulfate fragment. At higher energies (90 to 115 eV) these fragments disappear for all complex sulfatides. Therefore extrapolation of the SM4s data onto SM3, SM2a, and SB1a is justified and all sulfatides can be correlated to standards with non-hydroxy fatty acids.
SB1a was absent in ␤-GalNAc transferase-deficient mice as expected from an in vitro study suggesting that a single enzyme synthesizes both GM2 and SM2a (45). In the kidneys of these knockout mice, SM3 accumulated to a level comparable with the combined levels of SM3 and SB1a of wild type mouse kidney. The total concentration of these two sulfatides taken together, was also retained at the same level in GalNAcTϩ/Ϫ kidneys. Therefore, it appears plausible that SM3 is the precursor for SB1a biosynthesis. First SM3 is converted to SM2a by the action of GalNAcT and, possibly, via formation of SM1a to SB1a. This would be in analogy to the synthesis of GD1a in brain (9). The absence of SB1a in GalNAcTϪ/Ϫ mice is consistent with its structural ganglio-series derivation.
Sulfatides like all other GSL are degraded in the cellular lysosomes. For SM4s, the initial step in degradation is the removal of sulfate by the combined action of arylsulfatase A and the activator protein SAP B (10). Thus far this is the only sulfatase known to act on sulfatides (1). Since SM4s, SM3, and SB1a all carry a sulfate group linked to a terminal sugar residue, the cleavage of the sulfate group is expected to be the first step of their biodegradation. It was shown here that in ASAϪ/Ϫ mice that SB1a together with SM4s and SM3 accumulates in the kidney. This finding demonstrates that for the in vivo catabolism of all three compounds the presence of arylsulfatase A is an absolute requirement. Compared with the corresponding wild type levels, the enrichment factors of the individual sulfatide components is roughly the same. It is concluded that there is no other sulfatase that acts on any one of these three sulfatides individually and ASA is probably the only sphingolipid-sulfatase in murine kidney cleaving the galactosyl-3-sulfate bond.
In analogy to the degradation of GM1a, SM1a, the product of SB1a hydrolysis by arylsulfatase A, then is likely to be further degraded to SM2a and SM3 with the lysosomal enzymes ␤-galactosidase and ␤-hexosaminidase, respectively. An overview of the 4 different GM2 gangliosidosis mouse models investigated including remaining enzyme activities and SM2 accumulation is given in Table VI. All mutant mice that were deficient in the ␣-, ␤-, or both subunits of this enzyme accumulated SM2a confirming that SB1a is processed to SM2a, with the latter being degraded further by ␤-hexosaminidase. Highest accumulation occurred in mice lacking all hexosaminidase isomers (HexaϪ/Ϫ and HexbϪ/Ϫ) or lacking the ␣-subunit (HexaϪ/Ϫ). For these two mutant mice, SM2a accumulation was comparable indicating that Hex B, the only relevant active enzyme isomer expressed in HexaϪ/Ϫ mice, did not act on SM2a. This is in good agreement with previous in vitro studies (46,47).
HexbϪ/Ϫ mice, expressing only an intact Hex S isomer, accumulated 10 times less SM2a than HexaϪ/Ϫ or HexaϪ/Ϫ and HexbϪ/Ϫ mutants. It is, therefore, concluded that Hex S plays an important and necessary role in the in vivo degradation of SM2a. In contrast, Hex A is the pivotal enzyme for the degradation of GM2 (47). Nevertheless, Hex A contributes to SM2a degradation in vivo: HexbϪ/Ϫ mice left with only Hex S, but no Hex A and Hex B activity, accumulate SM2a to a lesser degree. These findings are in agreement with in vitro studies demonstrating the ability of both, human Hex S and Hex A, to degrade SM2a in the presence of human GM2 activator protein (47).
For the degradation of GM2 in mice two metabolic pathways have been described. One catabolic sequence, in humans the only significant pathway, is the degradation of GM2 to GM3 by Hex A in presence of GM2AP, and further hydrolysis of GM3 to LacCer. Another mode of degradation is the formation of Gg 3 Cer from GM2, and further processing to LacCer by Hex A or Hex B, with GM2AP as a cofactor. Therefore, HexaϪ/Ϫ mice accumulate 3-6 times less GM2 than HexbϪ/Ϫ or HexaϪ/Ϫ and HexbϪ/Ϫ mice (16,17). Since, in contrast to GM2, HexaϪ/Ϫ and HexaϪ/Ϫ and HexbϪ/Ϫ mice accumulated SM2a to an equal extent, it is sug-  Table III. Age of mice in weeks is indicated on the y axis.  gested that there is no significant degradative pathway from SM2a to Gg 3 Cer in mice. GM2 activator protein stimulates in vivo not only the enzymatic ganglioside hydrolysis but also the degradation of SM2a. This is concluded from the significant accumulation of SM2a in Gm2aϪ/Ϫ mice, which again is in agreement with in vitro results (46,47). However, as compared with hexosaminidasedeficient mutant mice, the accumulation of SM2a was small (roughly 3.5 times lower than in HexbϪ/Ϫ, and 35 times lower than in HexaϪ/Ϫ and HexaϪ/Ϫ and HexbϪ/Ϫ mice). The degradation of SM2a to SM3 must still be operative in the absence of GM2 activator protein in the Gm2aϪ/Ϫ mice. In conclusion, SM2a is degraded by Hex S and Hex A with both enzymatic hydrolases increased in effectiveness by GM2AP.
SM2a accumulation was also observed in a human Tay-Sachs liver. SB1a has only been described in a human hepatic carcinoma cell line (48). We conclude that this lipid is the degradative precursor of the accumulating SM2a in the human Tay-Sachs liver. Thus, it appears that the human pathway for complex sulfatides is very similar to that of mice. Interestingly, the ceramide patterns of the SM2a and GM2 of the human Tay-Sachs liver are very different. Therefore we conclude that SM2a and GM2 are metabolized in different pools, e.g. different cell types.
With regard to neutral GSLs as storage compounds in gangliosidoses, Gg 3 Cer and Gb 4 Cer are known to accumulate in Sandhoff disease and HexaϪ/Ϫ and HexbϪ/Ϫ mice. Only in the HexaϪ/Ϫ and HexbϪ/Ϫ double mutant mice was a third neutral glycolipid found to accumulate. Its structure, IV 6 GlcNAc␤-Gb 4 Cer, was identified by nano-ESI-MS/MS, GC/ MS, and permethylation analysis. It can be distinguished from Forssman glycolipid in the negative ESI-MS/MS product ion scan by an additional ring-cleavage fragment (m/z 322). It is assumed that the likely precursor of IV 6 GlcNAc␤-Gb 4 Cer is the typical mouse kidney octaosylceramide Gal␤1-4(Fuc␣1- Peaks are labeled according to masses derived by gangliosides containing C18-sphingosine, since this is the main sphingoid in these gangliosides according to Nilsson and Svennerholm (33). Nevertheless, gangliosides with C20-sphingosine (4% in GM3 and 21% in GM1 (33)) and a corresponding smaller fatty acid (shorter by a C 2 H 4 -unit) give rise to the identical signals.

TABLE VI
Mouse models for GM2 gangliosidosis and their relative accumulation of SM2a Ranking: ϩϩϩϩ, very strong; ϩϩ, medium; and ϩ, low accumulation.

Mouse model
Knocked out genes Remaining ␤-hexosaminidase (subunit combination)
The knowledge of all storage compounds and the resulting cellular changes due to GSL accumulation are a prerequisite for a more complete understanding of the pathology of the respective human diseases. And, by analogy, knowledge of all GSL that are missing in mutant mice may be important for a correct interpretation of the respective phenotypes. Furthermore, the accumulation of GSL in pathologic tissues, such as SM2a in human Tay-Sachs liver, demonstrates the existence of certain GSL components in particular organs where they may play special functional roles.