Filipin recognizes both GM1 and cholesterol in GM1 gangliosidosis mouse brain.

Filipin is an antibiotic polyene widely used as a histochemical marker for cholesterol. We previously reported cholesterol/filipin-positive staining in brain of β-galactosidase (β-gal) knockout ((-/-)) mice (GM1 gangliosidosis). The content and distribution of cholesterol and gangliosides was analyzed in plasma membrane (PM) and microsomal (MS) fractions from whole-brain tissue of 15 week-old control (β-gal(+/-)) and GM1 gangliosidosis (β-gal(-/-)) mice. Total ganglioside content (μg sialic acid/mg protein) was 3-fold and 7-fold greater in the PM and MS fractions, respectively, in βgal(-/-) mice than in βgal(+/-) mice. GM1 content was 30-fold and 50-fold greater in the PM and MS fractions, respectively. In contrast, unesterified cholesterol content (μg/mg protein) was similar in the PM and the MS fractions of the βgal(-/-) and βgal(+/-) mice. Filipin is known to bind to various sterol derivatives and phospholipids on thin-layer chromatograms. Biochemical evidence is presented showing that filipin also binds to GM1 with an affinity similar to that for cholesterol, with a corresponding fluorescent reaction. Our data suggest that the GM1 storage seen in the β-gal(-/-) mouse contributes to the filipin ultraviolet fluorescence observed in GM1 gangliosidosis brain. The data indicate that in addition to cholesterol, filipin can also be useful for detecting GM1.

Filipin fl uorescence was previously reported in fi broblasts and in brain tissue from a murine model of GM1 gangliosidosis ( 11,13,15 ). These fi ndings suggested that cholesterol was a secondary storage molecule to GM1 in this disease. GM1 gangliosidosis is an autosomal recessive glycosphingolipid storage disorder caused by mutations in lysosomal ␤ -galactosidase activity (18)(19)(20). The loss or reduction of ␤ -galactosidase activity results in a failure to cleave the terminal galactosyl residue of ganglioside GM1, leading to excessive central nervous system storage of GM1 and its asialo derivative, GA1 (21)(22)(23)(24)(25). Secondary storage materials are not uncommon in some lipid storage disorders (26)(27)(28)(29). The presence of excess sphingolipid (sphingomyelin or lactosylceramide) alone is suffi cient to induce cholesterol internalization and "trapping" ( 30,31 ). The internalization of cholesterol in NPC1 knockout cells is dependent on the presence of ganglioside GM3 ( 32 ). Although abnormal cholesterol traffi cking can be visualized in many lipid storage disorders, the colocalization of cholesterol and other storage materials is not clear ( 28,33 ). The lipid raft hypothesis suggests that cholesterol and GM1 are organized with transmembrane proteins in distinct raft domains ( 34,35 ), providing a model for extensive GM1/cholesterol colocalization seen in normal and disease states.
We previously described fi lipin-positive staining in brain tissue of ␤ -galactosidase ( ␤ -gal) knockout ( Ϫ / Ϫ ) mice, suggesting that cholesterol and GM1 were colocalized in Abstract Filipin is an antibiotic polyene widely used as a histochemical marker for cholesterol. We previously reported cholesterol/fi lipin-positive staining in brain of ␤ -galactosidase ( ␤ -gal) knockout ( ؊ / ؊ ) mice (GM1 gangliosidosis). The content and distribution of cholesterol and gangliosides was analyzed in plasma membrane (PM) and microsomal (MS) fractions from whole-brain tissue of 15 week-old control ( ␤ -gal +/ ؊ ) and GM1 gangliosidosis ( ␤ -gal ؊ / ؊ ) mice. Total ganglioside content ( g sialic acid/ mg protein) was 3-fold and 7-fold greater in the PM and MS fractions, respectively, in ␤ gal ؊ / ؊ mice than in ␤ gal +/ ؊ mice. GM1 content was 30-fold and 50-fold greater in the PM and MS fractions, respectively. In contrast, unesterifi ed cholesterol content ( g/mg protein) was similar in the PM and the MS fractions of the ␤ gal ؊ / ؊ and ␤ gal +/ ؊ mice. Filipin is known to bind to various sterol derivatives and phospholipids on thin-layer chromatograms. Biochemical evidence is presented showing that fi lipin also binds to GM1 with an affi nity similar to that for cholesterol, with a corresponding fl uorescent reaction. Our data suggest that the GM1 storage seen in the ␤ -gal ؊ / ؊ mouse contributes to the fi lipin ultraviolet fl uorescence observed in GM1 gangliosidosis brain. The data indicate that in addition to cholesterol, fi lipin can also be useful for detecting GM1. - Filipin is a polyene antifungal antibiotic produced naturally by the bacteria Streptomyces fi lipinensis ( 1-4 ) ( Fig. 1 ). Early reports showed that fi lipin disrupted permeability of sterol-containing membranes in Neurospora crassa , causing cellular leakage ( 5,6 ). Due to its fl uorescence shift upon binding to cholesterol, fi lipin is visible under simple ultra-Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care Committee.

Brain subcellular fractions
Mice were euthanized by cervical dislocation, and excised brain tissue was immediately frozen for analysis. The procedure of Lomnitski et al. ( 39 ) was used to prepare brain subcellular fractions. Briefl y, whole brains were homogenized in 4 ml of centrifugation solution [1 g/ l aprotinin (Sigma Aldrich; St. Louis, MO), 1 mM PMSF, and 1 mM EDTA in 50 mM potassium phosphate buffer]. The homogenates were transferred to 5 ml polyallomer centrifuge tubes (Beckman; Palo Alto, CA) and centrifuged for 10 min at 1,000 g (4°C). The P1 (nuclear membrane) pellet was separated from the subsequent supernatant. The supernatant was collected and centrifuged again at 10,000 g (4°C), resulting in a P2 pellet enriched in PM and mitochondria. The supernatant was further centrifuged for 1 h at 150,000 g (4°C), resulting in a P3 pellet enriched in microsomes. Both the P2 and the P3 pellets for all samples were suspended in 1 ml dH 2 O. A 500 µl aliquot of the whole-brain homogenate was collected before the fi rst spin. From each whole-brain homogenate, P2 PM pellet suspension, and P3 microsomal (MS) pellet suspension from ␤ -gal Ϫ / Ϫ and ␤ -gal+/ Ϫ mice, half of each homogenate was assayed for protein content, and the other half was used for the isolation and purifi cation of lipids.

Protein quantifi cation
Protein content from whole brain and individual membrane fractions were quantifi ed using a Bio-Rad DC Protein Assay Kit (Bio-Rad; Hercules, CA). Serial 1:10 and 1:20 dilutions of each sample were aliquoted onto a Corning 96-well clear assay plate in triplicate. After the assay reagents were added, samples were incubated at room temperature for 15 min, and then read at 750 nm on an M5 SpectraMax Microplate Reader (Molecular Devices; Sunnyvale, CA). Protein concentrations were determined by fi tting the samples to a standard curve using BSA.

Western blotting
The distribution of calnexin, an endoplasmic reticulum marker ( 40 ), among the subcellular fractions was determined by Western blot. Protein concentrations of cell fractions were estimated using the Bio-Rad DC protein assay as described above. Eighteen milligrams of total protein from each sample was loaded onto a 4-12% sodium dodecyl polyacrylamide gel (Invitrogen; Carlsbad, CA) and separated by electrophoresis. Gels were run for 30 min at 100 V, then 50 min at 150 V. Proteins were transferred for 16 h at 30 V to a polyvinylidene difl uoride Immobilon-P membrane (Millipore; Billerica, MA). The membrane was blocked in 5% blotting-grade blocker nonfat dry milk (Bio-Rad) in 1× TBS with Tween 20 (TBS-T) for 3 h at room temperature. Blots were washed in TBS-T three times for 5 min each, and then probed for 16 h at 4°C with primary antibodies against calnexin (Millipore), diluted 1:750, and ␤ -actin (Cell Signaling; Danvers, MA), diluted 1:4,000. Blots were washed again with TBS-T three times for 15 min, and then probed for 90 min at room temperature with secondary antibodies against mouse (Santa Cruz Biotechnology; Santa Cruz, CA), diluted 1:4,000, and rabbit (Cell Signaling), diluted 1:4,000. Protein bands were visualized with Pierce ECL Western blotting substrate (Fisher Scientifi c; Houston, TX).

Total lipid isolation
Total lipids were isolated and purifi ed from cerebral cortex as previously described ( 19,21 ). Frozen cortex samples were homogenized in centrifugation solution, separated into fractions as lysosomes ( 13 ). Moreover, ␤ -galactosidase gene therapy lessened or completely abrogated GM1 and GA1 ganglioside storage and fi lipin staining. However, the quantification of unesterified cholesterol showed no abnormal storage in the brains of ␤ -gal Ϫ / Ϫ mice. No change in cholesterol content was observed after treatment with adenoassociated virus (AAV) gene therapy, which reduced ganglioside storage ( 13 ). If fi lipin staining is a reliable marker for membrane cholesterol content, then changes in fi lipin staining should correlate with measurable changes in the content of membrane cholesterol.
Here we sought to test the hypothesis that GM1 ganglioside storage disrupts cholesterol localization within the plasma membrane (PM) and microsomes of GM1 gangliosidosis mouse brain. We found that even among subcellular (microsomal) fractions enriched in GM1 and GA1, there was no signifi cant corresponding increase in unesterifi ed cholesterol content. Furthermore, we found that the fi lipin complex, which was reported previously to bind weakly to phospholipids ( 16 ), also binds gangliosides, especially GM1. These fi ndings indicate that the fi lipin-positive staining seen in GM1 gangliosidosis mouse brain cannot be solely attributed to an increase in intracellular cholesterol. A preliminary report of these fi ndings was presented at the 2010 meeting of the American Society for Neurochemistry ( 36 ).

Mice
B6/129Sv mice heterozygous for the GM1 ␤ -galactosidase gene ( ␤ -gal+/ Ϫ ) were obtained from Dr. Alessandra d'Azzo of Saint Jude Children's Research Hospital, Nashville, TN. These mice were derived by homologous recombination and embryonic stem cell technology as previously described ( 37 ). Homozygous ␤ -gal Ϫ / Ϫ mouse pups were derived from crossing ␤ -gal Ϫ /+ females with ␤ -gal Ϫ / Ϫ male mice. Genotypes were determined by measuring ␤ -galactosidase-specifi c activity in tail tissue using a modifi cation of the Galjaard procedure ( 21,38 ). All mice were propagated in the Boston College Animal Care Facility and were housed in plastic cages with fi lter tops containing Sani-Chip bedding (P. J. Murphy Forest Products Corp.; Montville, NJ). The room was maintained at 22°C on a 12 h light/12 h dark cycle. Food (PROLAB R/M/H/3000 Lab Chow; Agway, St. Louis, MO) and water were provided ad libitum . Nursing females were provided with cotton nesting pads for the duration of the experiment. All animal experiments were carried out with ethical committee approval in accordance with the National Institutes of

Filipin recognizes GM1 in ␤ -gal
Ϫ / Ϫ brain 1347 spotted for both neutral lipids and gangliosides. HPTLC plates were sprayed with a solution of fi lipin-PBS [6.0 mg fi lipin complex dissolved in 300 ml dimethyl formamide (Sigma Aldrich) added to 49.7 ml PBS]. The fi lipin-stained plates were shielded from light and placed in a Shel Lab General laboratory incubator (Sheldon Manufacturing; Cornelius, OR) at 37°C for 90 min. The plates were then washed three times in dH 2 O, and visualized via UV transillumination as previously described ( 16 ). UV fl uorescence at 365 nm was determined using a Camag TLC Scanner 4 (Camag Scientifi c, Inc.) with winCATS software. After UV visualization, gangliosides were visualized by spraying the dried plate with the resorcinol reagent and heating at 95°C. Cholesterol was visualized by dipping the dried plate into 3% cupric acetate in an 8% phosphoric acid solution, followed by heating the plate to 140°C, as described above ( 41 ).

RESULTS
To determine whether the fl uorescence from the fi lipin complex corresponded to cholesterol content, we evaluated the content and distribution of lipids in whole brain and in subcellular fractions from 15 week-old normal ( ␤ -gal+/ Ϫ ) and GM1 gangliosidosis ( ␤ -gal Ϫ / Ϫ ) mice. Calnexin, an endoplasmic reticulum marker, was enriched in the MS fraction, compared with whole brain and PM ( Fig.  2 ). No signifi cant differences were found between the +/ Ϫ and Ϫ / Ϫ mice for brain weight or protein content in PMs or microsomes ( Table 1 ). The total sialic acid content in ␤ -gal Ϫ / Ϫ mouse brain was signifi cantly higher than in normal mouse brain ( Table 1 ). Sialic acid concentration ( g/mg protein) was similar in microsomes and PM of +/ Ϫ mice ( Table 1 ). Total sialic acid content for Ϫ / Ϫ mice was 2-fold greater, 3-fold greater, and 7-fold greater in whole brain, PM, and MS fractions, respectively, compared with that of +/ Ϫ mice ( Table 1 ). Most of the increase in total sialic acid concentration in the Ϫ / Ϫ mice was due to an increase in GM1 content ( Fig. 3 ). GM1 content was 50-fold greater in the microsomes of Ϫ / Ϫ mice than in those of +/ Ϫ mice ( Fig. 3 , Table 1 ).

Sialic acid quantifi cation
Gangliosides were further separated from acidic phospholipids and desalted as previously described ( 19,21 ). Total ganglioside content was quantifi ed before and after desalting using the resorcinol assay. Three aliquots of each ganglioside sample were dried under vacuum. A resorcinol-dH 2 O, 1:1, v/v solution (resorcinol reagent-HCl-0.2 M resorcinol-dH 2 O-0.1 M CuSO 4 , 40: 5: 5: 0.125, v/v/v/v) was added to each sample, followed by submersion in a boiling water bath for 17 min. After cooling on ice, the reaction was stopped with butyl acetate-N -butanol, 85:15, v/v. Each sample was vortexed and centrifuged at 700 g for 2 min. The absorbance of the upper aqueous layer was recorded at 580 nm using a Shimadzu UV-1601 spectrophotometer (Shimadzu; Torrance, CA). Sialic acid values were fi t to a standard curve using N -acetylneuraminic acid as a standard.

High-performance TLC
Lipids were spotted on 10 × 20 cm silica gel-60 high-performance TLC (HPTLC) plates using a Camag Linomat III auto-TLC spotter (Camag Scientifi c, Inc.; Wilmington, NC) as previously described ( 19,21,41 ). The amount of lipid per lane was equivalent to 1.5 g sialic acid for gangliosides and 10 g protein for neutral lipids/cholesterol. Gangliosides were separated by a single ascending run (90 min) in a solution of CHCl 3 -CH 3 OH-0.02% CaCl 2 (55:45:10; v/v/v), whereas neutral lipids/cholesterol were separated by an initial ascending run (up to 4.5 cm) with a solution of CHCl 3 -CH 3 OH-acetic acid-formic acid-dH 2 O (35:15:6:2:1; v/v/v/v/v), dried for 15 min, and then run to the top with hexane-diisopropyl ether-acetic acid (65:35:2; v/v/v) (Fisher Scientifi c). Gangliosides were visualized by spraying the dried plates with the resorcinol reagent, followed by heating at 95°C. Neutral lipids and cholesterol were visualized by dipping the dried plates into 3% cupric acetate in an 8% phosphoric acid solution, followed by heating the plate to 140°C ( 41 ).

GM1 and cholesterol quantifi cation
A Personal Densitometer SI with ImageQuant software (Molecular Dynamics; Sunnyvale, CA) was used to estimate the concentrations of gangliosides and cholesterol on HPTLC. The total ganglioside content was normalized to 100%, and the percent distribution of each band was used to determine the sialic acid concentration for individual gangliosides. GM1 content was expressed as g/mg protein. For neutral plates, the band density value for cholesterol in each sample lane (equivalent to 10 g of protein) was fi t to a standard curve and used to determine cholesterol concentration expressed as g/mg protein.

Filipin UV visualization
Purifi ed GM1 and cholesterol (from mouse brain) were successively spotted with increasing nanomole amounts on an HPTLC plate and separated by the neutral solvent systems described above. For UV visualization of whole-brain homogenates and subcellular fractions, the equivalent of 20 mg of protein was Fig. 2. Calnexin is enriched in P3 (MS) fractions of normal and GM1 gangliosidosis mouse brain. Equivalent protein aliquots from each subcellular fraction were separated by electrophoresis, transferred to immobilon-P membranes, and incubated with antibodies against calnexin (upper bands) and ␤ -actin (lower bands). Representative samples are from one ␤ -gal+/ Ϫ or ␤ -gal Ϫ / Ϫ brain. ( Fig. 6A ). GA1, cholesterol, and phosphatidylcholine were detectable in MS fractions of Ϫ / Ϫ mice ( Fig. 6A ). There was no difference in the fl uorescent signal (area units/mg protein) of cholesterol in MS fractions in Ϫ / Ϫ and +/ Ϫ mice ( Fig. 6A , Table 2 ). There was no detectable fl uorescent signal from the MS gangliosides of +/= mice. ( Fig. 6B , Table 2). Conversely, there was a prominent fl uorescent signal corresponding to GM1 in the MS gangliosides of Ϫ / Ϫ mice ( Fig. 6B , Table 2).

DISCUSSION
Using fi lipin as a cholesterol marker, Broekman et al. ( 13 ) showed that GM1 accumulation was accompanied by strong fi lipin staining in lysosomal membranes. These fi ndings suggested that cholesterol accumulated in lysosomes along with GM1. In this study, we showed that GM1 accumulation was not associated with signifi cant increases of cholesterol content in MS membranes. Our fi ndings raise concerns in using fi lipin as an accurate marker for quantitative cholesterol content in GM1 gangliosidosis in cholesterol content between +/ Ϫ and Ϫ / Ϫ mice in whole brain, PM, or microsomes ( Table 1 ).
To quantitatively assess the fl uorescent reactivity of fi lipin, we spotted HPTLC plates with equivalent nanomole amounts of both cholesterol and ganglioside GM1 from mouse brain. Cholesterol exhibited strong UV fl uorescence (area units/nmol), as expected, when incubated with fi lipin complex on HPTLC ( Fig. 5A ). A best-fi t line revealed a linear correlation between nanomole cholesterol content and detectable fl uorescence ( Fig. 5B ); no quenching was observed at the concentrations measured ( 16,42 ). We found that GM1 ganglioside also elicited detectable fl uorescence with the fi lipin complex after a 90 min incubation ( Fig. 5A ). UV fl uorescence also increased, corresponding with a linear increase in GM1 content at higher nanomole amounts ( Fig. 5B ). Notably, the fl uorescence observed from fi lipin/GM1 binding was equal to or greater than the fl uorescence observed from fi lipin/cholesterol binding ( Fig. 5B ).
To determine whether our lipid quantifi cation corresponded with fl uorescence, we applied fi lipin to HPTLC of neutral lipids and gangliosides isolated from the subcellular fractions in ␤ -gal+/ Ϫ and ␤ -gal Ϫ / Ϫ mouse brain. Cholesterol and phosphatidylcholine were both visible by UV transillumination after incubation with fi lipin in the neutral lipids purifi ed from MS fractions of +/ Ϫ mice Values are expressed as the mean ± standard error of the mean. N = 3 mice per group. a Protein values were determined using the Bio-Rad assay as described in Materials and Methods. b Sialic acid values were determined using the resorcinol assay as described in Materials and Methods. c GM1 content was determined through densitometric scanning of HPTLC plates as seen in Fig. 3 . d Cholesterol content was determined through densitometric scanning of HPTLC plates as seen in Fig. 4 . e Signifi cantly different from the ␤ -gal +/ Ϫ value in the same fraction at P < 0.01 (using Student's t -test.)

Fig. 3.
High-performance TLC of gangliosides. Gangliosides were isolated and purifi ed for each fraction from control ( ␤ -gal+/ Ϫ ) and GM1 gangliosidosis ( ␤ -gal Ϫ / Ϫ ) mouse brains. 1.5 g sialic acid were spotted per lane. Quantifi cation of GM1 content is displayed in Table 1 .  Table 1 . cholesterol. However, our biochemical lipid analysis showed that GM1 accumulation alone was capable of producing a fi lipin-positive signal in the brain of ␤ -gal Ϫ / Ϫ mice. We did not quantify cholesterol in lysosomes specifically in this study, but the presumed colocalization of GM1 would overstate the presence of cholesterol in any lipid subdomain examined.
brain. The cholesterol-trapping hypothesis proposes that the accumulation of one type of lipid raft molecule in late endosomes will lead to an accumulation of other types of raft lipids, eventually impeding raft lipid turnover and cholesterol traffi cking ( 34 ). If GM1 and cholesterol were tightly associated in raft lipids, it follows that the accumulation of GM1 should directly cause the accumulation of  conclude that the appearance of fi lipin-positive staining in tissues from GM1 gangliosidosis results from the crossreactivity of fi lipin with GM1 ganglioside in addition to binding of fi lipin with cholesterol. Although our fi ndings should not decrease the utility of fi lipin as a histochemical marker for nonesterifi ed cholesterol, they do question the specifi city of fi lipin as a marker solely for cholesterol in certain disease conditions.
Although fi lipin binding is greater for cholesterol than for phosphatidylcholine on HPTLC, fi lipin was considered somewhat promiscuous in its lipid binding capabilities ( 16 ). Our data suggest that the promiscuity of the fi lipin complex can also extend to gangliosides. Our fi ndings also suggest that accumulation of GM1 can produce misinformation on cholesterol content when fi lipin is used as a cholesterol marker. This is an obvious caveat for using fi lipin as a histochemical marker for known lysosomal storage disorders, because gangliosides can accumulate as either primary or secondary storage material in several storage diseases ( 26-29, 43, 44 ).
Our fi ndings might help to resolve several inconsistencies associated with storage of cholesterol and gangliosides seen in GM1 gangliosidosis and Niemann-Pick Type C disease. For example, cyclodextrin treatment abrogates fi lipin signal and corrects cholesterol storage in NPC1 fi broblasts and mouse models (45)(46)(47). However, cyclodextrin has no effect on fi lipin signal in GM1 gangliosidosis fi broblasts or mouse tissue ( 45 ). Earlier studies in human GM1 gangliosidosis brain showed a decrease in cholesterol content, depending on the severity of the disease ( 48 ). Filipin also inhibits the internalization of the Vibrio cholorae cholera toxin, which avidly binds to GM1 ( 49,50 ). This fi nding further suggests a cross-reactivity of fi lipin with GM1.
We did not attempt to evaluate membrane fractions based on their detergent-resistant properties. The analysis of GM1 and cholesterol in detergent-resistant membranes or lipid rafts holds the caveat that different raft preparations can produce disparate results ( 51,52 ). The membrane fractionation performed here yielded results consistent with previous fi ndings regarding ganglioside GM1 and asialo GA1 storage in late endosomes of ␤ -gal Ϫ / Ϫ mice ( 30 ). We also confi rmed the previous fi ndings of Lomnitski and coworkers ( 39 ) that phosphatidylethanolamine and phosphatidylcholine are enriched in PMs and microsomes, respectively. Although the P2 fractions examined here did contain mitochondria, we have previously demonstrated that purifi ed mitochondria from normal mouse brain contain only trace amounts of ganglioside ( 40 ). Further, a report from Sano et al. ( 25 ) showed that the GM1 storage seen in mitochondrial preparations from GM1 gangliosidosis mouse brain is primarily due to its presence in mitochondria-associated ER membranes. We Values are expressed as the mean ± standard error of the mean (SEM). N = 3 mice per group. a Relative fl uorescence at 365 nm was determined using a CAMAG TLC scanner on HPTLC plates as seen in Fig. 6 . b ND, no detectable fl uorescence.