Glycosphingolipids in Cultured Human Skin Fibroblasts

Fibroblasts cultured from human skin or bone marrow biopsies contain the glycosphingolipids associated with human visceral organs, namely, glucosylceramide (GL-la), lactosylceramide (GL-Za), trihexosylceramide (CL-3), globoside (GL-4), GM3 and GD3. The presence of large amounts of GL-3 and GL-4 distinguishes normal human fibroblasts from most other normal mammalian cell lines so far investigated. Neither gangliosides (glycosphingolipids containing both sialic acid and hexosamine) nor sulfatides appeared to be present in these cells (concentration less than 0.01 pmole per g dry weight of cells). The glycosphingolipids constitute about 3% of the total lipid of fibroblasts, an amount comparable to that of normal human tissue. Further, the glycosphingolipid composition and concentration are largely independent of the amount of serum supplement. Confluent fibroblast cultures showed maximum incorporation of isotope from uniformly labeled D-[U-Wlglucose into the hexose, sphingosine, and fatty acid moieties of all six glycosphingolipids after 24 to 48 hours, indicating de novo synthesis of all three moieties. The half-life of 2 to 3 days was consistent with values generally accepted for membrane-bound phospholipids.


SUMMARY
Fibroblasts cultured from human skin or bone marrow biopsies contain the glycosphingolipids associated with human visceral organs, namely, glucosylceramide (GL-la), lactosylceramide (GL-Za), trihexosylceramide (CL-3), globoside (GL-4), GM3 and GD3. The presence of large amounts of GL-3 and GL-4 distinguishes normal human fibroblasts from most other normal mammalian cell lines so far investigated.
Neither gangliosides (glycosphingolipids containing both sialic acid and hexosamine) nor sulfatides appeared to be present in these cells (concentration less than 0.01 pmole per g dry weight of cells).
The glycosphingolipids constitute about 3% of the total lipid of fibroblasts, an amount comparable to that of normal human tissue.
Further, the glycosphingolipid composition and concentration are largely independent of the amount of serum supplement. Confluent fibroblast cultures showed maximum incorporation of isotope from uniformly labeled D-[U-Wlglucose into the hexose, sphingosine, and fatty acid moieties of all six glycosphingolipids after 24 to 48 hours, indicating de novo synthesis of all three moieties.
The half-life of 2 to 3 days was consistent with values generally accepted for membrane-bound phospholipids.
Some insights into their biosynthesis and catabolism have been gained from the study of changes resulting from viral transformation of mammalian cell lines in tissue culture (10, ll), and the pattern of glycosphingolipid accumulation in genetic diseases characterized by mutant lysosomal hydrolases (7,9). It is difficult to carry out in viva studies 011 humans because of restrictions on the amount of radioactive isotope that may be used and the conlplexity of studies with stable isotopes (12). However, glycosphingolipid metabolism can be studied in human skin fibroblasts by a procedure first established for studying the turnover of glycosphingolipids in porcine blood (13). This paper reports the glycosphingolipid composition of normal human skin fibroblasts and the use of n-[U-Wlglucose to study their metabolism.
When grown under these conditions in Falcon 100 mm diameter petri dishes at 37" under a COS tension of 10% (pII fluctuation of 7.2 to 7.4) a confluent monolayer of cells (2.5 x lo7 per plate) was obtained in approximatel\-4 weeks. WI\-coplasma infection was monitored by routinely culturing cells and media in mycoplasma -igar medium (Gibco).
All studies described in this paper were carried out on cells judged by this criterion to be free from mycoplasma. Prior to harvexting, cells were washed with 0.15 M P\'aCl to remove the culture medium and then scraped from the plate with a rubber police-1m11, mashed with 2.0 1111 0.05 M acetate buffer, centrifuged at 600 x g for 10 min, and the pellet lyophilized.
Isolation of Glycosphingolipids-Total cellular lipids were extracted from the lyophilized powder (lo8 cells; approsimatel\ 100 mg, dry weight) by stirring first iu methanol (100 ml) and 5944 5945 then adding chloroform (2 x 100 ml) over a period of 2 hours. Following the Folch (15) procedure, the resulting lower phase was fractionated into neutral, polar, and individual glycosphingolipids on a silicic acid column by the method of Vance and Sweeley (16). The crude glycosphingolipid fraction (acetone-methanol, 9:l) was treated with 0.6 N NaOH in methanol (1.0 ml) in the presence of 1.0 ml of chloroform for 1 hour.
After neutralization with concentrated HCl (0.05 ml) the mixture plus washings (chloroform-methanol, 2: 1) was dialyzed against several changes of distilled water for 24 hours. Detailed studies (17) have shown that there is no loss of glycosphingolipids during dialysis; the dialysis tubing is washed with 2 x 10 ml of chloroform-methanol (2: 1) as a precaution against such loss. Individual glycosphingolipids were resolved by thin layer chromatography in chloroform-methanol-water (11: 4 : 0.6). The upper aqueous phase (from the Folch procedure) was evaporated to dryness, subjected to alkaline methanolysis as described above, and dialyzed exhaustively.
The crude ganglioside fraction was separated by thin layer chromatography using chloroform-methanol-2.5 N NH,OH (60:40:9) (two sequential developments). Individual glycosphingolipids were visualized by iodine and a permanent record of the thin layer chromatography plate was made on diazo projection paper (B. K. Elliott Co., Pittsburgh, Pa.). The diazo paper is placed over the thin layer chromatography plate and illuminated from below on an x-ray viewer for approximately 5 min. Exposure to ammonia vapor gives the blueprint, examples of which are reproduced in Figs. 1 and 2. The glycosphingolipid bands were scraped, transferred to a sealing tube with reduced ends and a Kimflow fritted disc of coarse porosity (Kimble Products, Owens, Ill.) and eluted with 20 ml of chloroform-methanol-water (10:4:0.5). Analysis and Identijkation of Glycosphingolipids-After addition of mannitol (0.04 to 0.2 pmole), glycosphingolipids were converted to methyl glycosides by methanolysis in 1.0 N HCl (16, 18) neutralized with silver carbonate and reX-acetylated if necessary (18).
Quantitative estimation and identification was carried out in a single procedure by gas-liquid chromatography of the trimethylsilylated sugars (16, 17) using a Hewlett-Packard F and M model 402 equipped with a flame ionization detector and a 6 foot x J< inch 3% OV-1 on Supelcoport (100 to 120 mesh) column.
Glycosphingolipids were identified by their RF value and the ratio of the constituent sugars obtained from the gas-liquid chromatography trace. The fatty acid composition was also estimated by gas chromatography of the fatty acid methyl esters resulting from methanolysis, using a 15% EGA (6 foot x s/4 inch external diameter) U-shaped glass column. Since di-and tri-unsaturated fatty acids were minor constituents of fibroblast glycosphingolipids, a 3% OV-1 column was used for routine fatty acid analyses.
The sphingosine base composition was determined by the method of Polito et al. (19). The structures of the major glycosphingolipids were confirmed by direct mass spectrometry using an LKB 9000 combined Gas-Liquid Chromatograph-Mass-Spectrometer (20).

Incorporation of o-[U-14C]Glucose into Fibroblast Glycosphingolipids-D-[UJ4C]Glucose
in 10 ml of modified Eagle's medium (14) was added to 35 100 mm petri dishes (10 PCi per plate, specific activity 270 PCi per pmole) ; each of which contained approximately 2.0 x lo7 cells (three to seven passages) in a confluent monolayer culture.
Studies with skin fibroblasts at a density of 1.25 x 10' cells/100 mm plate showed that the initial level of glucose in the medium (400 mg per liter) is reduced to 300 after 24 hours, 200 after 48 hours, and 100 after 120 hours; medium is replenished every 72 hours for all studies unless stated otherwise.
After 6 hours the medium from five of the plates was discarded, the cells were washed with saline, harvested, and lipid fractions isolated as described above. The hexose, sphingosine, and fatty acid moieties of the glycosphingolipids were counted separately as previously described (13). Since preliminary experiments had shown that maximum incorporation occurred between 24 and 48 hours, cells were harvested at 6, 24, 36, and 48 hours. After 48 hours, fresh media was added to chase the D-[U-14C]glucose from the remaining plates of cells and further cells were harvested at 72, 102, and 168 hours.

Glycosphingolipid
Composition-Fibroblasts cultured from human skin were found to contain GL-la, GL-2a, GL-3, GL-4 ( Fig. l), GM3 and GD3 (Fig. 2); GL-lb was a minor component (Fig. 3). These glycosphingolipids were identified by their cochromatography with authentic standards (isolated from human  3. GAS liquid chromatographic analysis of GL-la, GL-2a, and GL-3 as trimethylsilyl derivatives of their methylglycosides. Separations were achieved on 3% OV-1 with temperature programming from 160" at 2" per min. Three peaks, representing the furanose, CP and &pyranose forms, are obtained for galactose and two peaks (a-and @-pyranose) for glucose. Mannitol is the internal standard. tissue in this laboratory) and by gas liquid chromatography analyses for monsaccharides (Figs. 3 and 4). Other glycosphingolipids, such as sulfatide or the complex sialoglycosphingolipids associated with nervous tissue, were not present in concentrations greater than 0.01 pmole per g, dry weight of cells. Under standard culture conditions, the glycosphingolipid composition showed little variation from one normal strain to another (Table I).
Glycosphingolipid Structure+-The carbohydrate moieties of normal fibroblast glycosphingolipids appeared identical to analogous compounds isolated from human blood and tissue. No precise information as to the nature of the anomeric linkages was obtained but since GL-3 accumulates in fibroblasts from patients with Fabry's disease (21)) the terminal galactose residue must be ~ (1 -+ 4), since GL-2a accumulates in fibroblasts from a patient with lactosylceramidosis (22) the galactose-glucose linkage is probably p- (1 -+ 4), and since GL-4 accumulates in GM2-gangliosidosis type II fibroblasts (23) the linkage of the terminal galNAc is p-(1 -+ 3) to galactose.
Mass spectrometric analysis of GL-4 (Table II) showed it to be identical to human Separations were achieved on 30/, OV-1 with temperature programming from 160' at 2" per min. Two peaks @-and a-pyranose) are obtained for galNAc and one major peak (a-) for NANA. Fatty Acid Composition of Glycosphingolipids-The distribution of glycosphingolipids in the chloroform-methanol-water (Folch) partition was investigated.
Only trace amounts of GL-la, GL-3, and GL-4 were found in the upper phase (Table III).
The only neutral glycosphingolipid which constituted more than 5% of the total was GL-2a, which may have arisen from the degradation of Gh13 during the processing of the upper phase. The major sialoglycosphingolipid, GM3, appeared to be partitioned almost equally between the two phases while GD3 was exclusively in the upper phase. Reextraction of the cells with CHCL-CHIOH, 1:l did not increase the yield of sialoglycosphingolipids.
Neutral and phospholipids of fetal calf and calf serum contained principally Cleco, C18:0, C"18:1, ClsZS, and CZoZ4 fatty acids. In contrast, the major fatty acids in fibroblast glycosphingolipids were CZZZO and &:o, with C&:1 and C24:1 the only unsaturated fatty acids present in significant amounts (minor components of GL-3 and GL-4 as shown in Table IV).
It seems unlikely, therefore, that serum fatty acids represent a direct source of fibroblast glycosphingolipid fatty acids. Both GL-2a and GM3 were resolved into two bands by the thin layer chromatography system. Fatty acid analysis showed that the faster moving band in each case contained more of the shorter chain length fatty acids, although the separation of C&/c18 and C&/C&., was not complete.
The virtual absence of C19:o, C20:0, and C?l:O fatt! acids accounts for the gap between the two bands.
Origin of Fibroblast Glycosphingolipids-Sirlce calf serum was used as a supplement to the medium, the glycosphingolipid composition of both fetal calf serum and calf serum was examined. In contrast to human serum (16) calf serum contained very low levels of glycosphingolipids; fetal calf serum contained virtual13 none ( Table V). The total amount of serum (and therefore glycosphingolipid) used in the growth of 1 g dry weight of cells is actually quite large (Table  V). However, analysis of the culture medium after esposure to normally metabolizing fibroblasts for 48 hours, showed very little change in the concentration of glycosphingolipids.
Increasing the amount of serum supplement from 105; to 307; had little effect on the intracellular glycosphingolipid content and in fact produced a decrease when expressed on a pmole per g dry weight basis (Table VI).
The sialic acid in human fibroblast,s, as judged by gas liquid chromatography (18) following mild acid (0.01 x HCI) hydrolysis or methanolysis for 2 hours, was found to be exclusively of the K acetyl type. One exception was a strain from a patient with Comparison of mass spectra between jbroblast GL-4 and hurnan erythrocyte GL-4 Gas liquid chromatography analysis showed calf serum to con tail1 significant amounts of X-glycolyl-Ghls.

Glycosphingolipids in Other d1ammalian
Cells in Culture-The gl~-cospbillRoliI)id content appears specific for the cell type but independent of the medium. Thus Swiss and I{alb mouse 31'3 fibroblasts, Cl300 mouse neuroblastoma (NB41A and iW2a) and rat RGC-6 glial tumor (27) cell lines, grown in the same medium as human fibroblasts showed a completely different spectrum of gl~cospllill~olipids (Table VII). The analyses of IsIlK 21/U cells (where the major glycosphingolipid is GMYIa) are in agreement with a qualitative report ( The similarity in fatty acid composition between individual glycosphirrgolipids in the 13HK cells supported the concept of direct metabolic interconversion as previously found for human skin fibroblasts (Table IV).
The synthesis of gl?cosphiligolil)itls was studied by using D- [U-Wlglucose. The data in Fig. 5 show that maximum labeling of all six glycosphingolipids occurred between 24 hours alld 4X hours. After this time (even if the medium was not changed) there was a rapid loss of 90% of the 14C label over the nest 24 hours followed by a much more gradual loss of the final 105, the half-life in this slow phase was 2 to 3 days. Although most of the label in the glycosphingolipid was found in the hesose moiety, significant labeling of the sphingosine and fatty acid moiety occurred.
Both therateand estent of loss of label roughly 5949 1)aralleled the loss of label from the hesose moiety. An example of this effect for GL-la is giv-en in Table VIII.
Thus glyco-~1Aingolipids in cultured human skin fibroblasts are metabolized as n-hole units and do not appear to undergo exchange reactions such as acyltrarisferase.
However, the reutilization of ceramide carrot be ruled out without further study. DISCUSSIOS \Ve have shown in this study that human skin fibroblasts synthesize and degrade the six glycosphingolipids (GL-la, GL-2a, N-3, GL-4, GM3, and Gn3) associated with visceral organs such as liver and spleen. Such glycosphingolipids are believed to be a+ociated with the plasma membrarre of cells (29). Of the glycosl)hingolipids associated with nervous tissue, GL-lb was detected in trace arnounts but sulfatide (GL-lbS) was undetectable. In fact, sulfatide has only been reported to occur in one tJ.l)e of cultured cell, neoplastic mast cells (30) and our studies or, various mammalian fibroblast lines such as RHK, and mouse 4'1'3 (Svviss and Balb), did not reveal measurable sulfatide levels (Table VII).
Since it would be desirable to study the ganglioside (GzII aud Galr) storage diseases in tissue culture, exhaustive analytical studies were carried out to detect GM1 and GM2 in 11orma1 fibroblasts.
Human skin fibroblasts do not contain detectable amounts of the gangliosides usually associated with INWYNIS tissue. The cultured human skiu fibroblast glyco-~l)hirlgolipids which corresponded irr thin layer chromatography mobility to GMM% and Gllll from human brain (Fig. 2), do not contain A~acetylgalactosamine (Fig. 4). Brain gangliosides contain ClgZO (95(;/;) as the major fatty acid and are readily resolved OII thin layer chromatography since they migrate as single bands. Iluman skin fibroblast Gna separates irrto two bands (because of fatty acid heterogeneity) which have RF values similar to GM2 :mtl G&,1. For this reason it is necessary to identify gangliosides it1 cell lines by methods other tharr thin layer chromatog-lXph\..
The GL-2a structure is cornrnorr to all glycosphingolipids, with the exception of the galactosylceramides (GL-lb, GL-lbS, a~rd GL-2b) associated with myelin, renal tubules, and intestine. Previous studies with crude particulate enzyme systems (31, 32) have indicated t,hat there are four possible pathways which irrrolve GL-2a as an intermediat,e in the biosynthesis of gangli-o&les and globosides.
Two of these appear to be present in Iru~n:m skin fibroblasts.
Further, human skin fibroblasts can synthesize GL-3, iudicatirrg the presence of a GL-2a-c-galactosyltrausferase; GL-3 is a major component) of visceral tissue but is absent from brain.
There was no chemical evidence that GL-2a acted as a substrate for either P-A-acetylgalactoPaminyltransferase (to form asialo-G& (6) or a P-Av:-acetyl glucosarninyltransferase (to form blood group substances).
Ow analyses suggest the existence of the following metabolic pathway in human skin fibroblasts. /CrL-la-1 The marked difference in glycosphingolipid composition between brain and liver is mirrored in a neuroblastoma cell line and cultured skin fibroblasts.
Further, fatty acid analysis of the fibroblast.
glycosphingolipids revealed the high content of behenate (C,,,,) and lignocerate (C,,,,) characteristic of visceral organs. The similarity in fatty acid composition between the six glycosphingolipids was sufficient to support the concept of metabolic interconversion as indicated above. It is perhaps significant that the major monosialoglycosphingolipid of visceral tissue (G,a) contains predominantly long chain (hy drophobic) fatty acids, whereas that of brain (G,r) contains predominantly stearate (C,,,,) (less hydrophobic). Using o-[U-%]glucose as precursor, it was possible to show that 14C is incorporated into the hexose, fatty acid, and sphin gosine moieties of skin fibroblast glycosphingolipids.
Maximum incorporation occurred betw-een 24 and 48 hours and 900/, of the label was lost between 48 and 72 hours. A semilogarithmic plot of specific activity versus tirne emphasized the biphaaic nature of the decay curve, suggesting that there were two pools of glgcosphingolipid.
The rapid metabolism of r4Clabeled gl,-cosphingolipids in normal skin fibroblasts indicates that catabolic enzyme deficiencies should be readily detectable.
The application of these techniques to the study of such inborn lysosomal hydrolase deficiencies is described iu the subsequent paper (23).
Acknodedgmenfs---We would like to thank Dr. Gordon Sato, Universit\-of California at Snn Diego, for the gift of tumor cell lines; I)r. L. Sachs, Weizmann Institute of Science, Rehovet, Israel, for the gift of mouse 37'3 and hamster I31-IK C13/21 fibroblasts; and Dr. T. Shows, Rosa-e11 Park Memorial Inst., for the Ralb 31  fibroblasts. Mass spectrometric analyses were carried out at the facility operated by T)r. Charles C.