External Labeling of Cell Surface Galactose and Galactosamine in Glycolipid and Glycoprotein of Human Erythrocytes”

the labeling of galactosyl and N-acetylgalactosaminyl residues on external surfaces of cells with tritium (3H). Labeling patterns and specific activities of galactose and galactosamine in glycolipids and glycoproteins were determined after separation with gel electrophoresis and thin layer chromatography. The labeling patterns of normal adult cells differed greatly from fetal cells, and were significantly altered when cell surfaces were modified by proteases and neuraminidase. The results of analysis indicated that


Treatment
of erythrocytes with galactose oxidase (EC 1. I .3.9) followed by reduction with tritiated sodium borohydride (NaB3H,) at pH 7.4 allowed the labeling of galactosyl and N-acetylgalactosaminyl residues on external surfaces of cells with tritium (3H). Labeling patterns and specific activities of galactose and galactosamine in glycolipids and glycoproteins were determined after separation with gel electrophoresis and thin layer chromatography. The labeling patterns of normal adult cells differed greatly from fetal cells, and were significantly altered when cell surfaces were modified by proteases and neuraminidase.
The results of analysis indicated that (a) the carbohydrate moieties of two glycolipids (globoside and ceramide trihexoside) and at least three glycoproteins (molecular weight 9.5, 8.2, and 6.4 X 104) were exposed to the external environment, but not ceramide dihexoside, ceramide monohexoside, or other glycoproteins with higher molecular weights; (b) the specific activities of galactosamine in glycolipid and of galactose in glycoprotein increased after protease treatment, although total activity of glycoprotein did not change; (c) labeling of glycoprotein was greatly enhanced by neuraminidase treatment, while that of glycolipid was enhanced to a lesser degree; (d) "the relative exposures" of glycoprotein and glycolipid differed greatly between normal and fetal erythrocyte surfaces. Glycoproteins of fetal cells had a very low label as compared to glycolipid.
Enriched localization of glycoproteins and glycolipids in surface membranes of cells has been demonstrated by histochemical visualization of complex carbohydrates in the cell coat (2), by the high content of glycolipids (3)(4)(5)(6) and protein-bound fucose and glucosamine (7) in an isolated plasma membrane, and by reactions of cell surfaces with anti-glycolipid antibodies (8,9) and with plant agglutinin (10,11 cell division and intercellular association has been predicted based on the change of chemical and organizational structures of membrane-bound carbohydrates or related enzymes in association with "contact inhibition" (12)(13)(14)(15), cell aggregation (16)) mitotic cell cycle (17), and malignant transformation (1% 20). Surface-exposed carbohydrates of cells are, therefore, of great cell-sociological significance, and it has become increasingly important to elucidate the exposed chemical structures of cell surfaces.
Labeling of cell surface tyrosyl residues has bcrn developed using lactoperoxidase and radioactive iodine (21,22), while cell surface amino groups have been labeled with s5S-labeled formylmethionylsulfone methylphosphate (23) or with 35S-labeled sulfanilic acid diazonium salt (24) ; labeling of specific surface carbohydrates, however, has been awaiting development. Very recently, labeling of surface sialyl residues by periodate and tritiated sodium borohydride was described (25), whereby sialyl residues were converted to a 3-dcoxy-5-acetamidoheptu ionic acid.
Although the majority of cellular glycosphingolipids are found in plasma membranes, direct evidence that the carbohydrate moiety of glycolipids is exposed to the external environment has not been provided.
Also, nothing has been known about the relative exposures of glycoprotein and glycolipid or about possible change in exposure with change of surface function and with modification of cell surfaces by neuraminidasc and proteases.
In order to solve these problems, we have developed a method using galactose oxidase (26) and tritiated sodium borohydride (NaB3H4), which allowed specific labeling of surface galactose and galactosamine residues in glycolipid and glyroprotein.
Application of this method to analyze the organizational state of surface carbohydrates of human erythrorytes is reported in this paper. had no contamination from protease activity, which was determined using "Azo-albumin" Washing by centrifugation in PBS was repeated five times.
Cells treated with neuraminidase and protease and isolated cell ghosts were labeled by the same procedure. No significant lysis of the cells occurred during the labeling procedure.
After labeling was completed, the cells were lysed, and the ghosts isolated and washed three times in PBS.
In other experiments, isolated ghosts were labeled and then washed three times in PBS.
aliquots of the labeled cell ghosts, glycolipids, or glycoprotein fractions obtained from the labeled cells or ghosts were mixed with 0.5 ml of "NCS" solubilizer (Amersham-Searle, Arlington Heights, Illinois) containing 10% H20 and incubated at 50" for 2 hours or more.
The efficiency for tritium counting was 43%.

Glycolipids and Glycoproteins
Glycolipids were extracted by homogenizing at room temperature for 3 min in an "Omnimixer" (Sorvall Instrument Company) with 20 to 30 volumes of chloroform-methanol (2:1, v/v) to 1 volume of cells and left at +4" overnight.
One-third volume of methanol was added and the samples centrifuged to obtain the protein and glycoprotein sediment, which was then washed twice in chloroform-methanol (1 :l, v/v). The glycolipid fraction was prepared from lipid extract by acetylation procedure (27) ; the fraction was deacetylated and separated into components by thin layer chromatography.
Bands corresponding to globoside, ceramide trihexoside, and ceramide dihexoside were scraped, extracted with chloroform-methanol (2: 1)) filtered, and aliquots were counted. times were recorded and the peaks quantified by using known amounts of galactose, N-acetylgalactosamine, and an internal standard. Immediately after the new aliquots were run and during the elution of galactose and N-acetylgalactosamine, the flame was extinguished and the derivatives collected in a capillary pipette cooled with Dry Ice. The flame was then lit again, and the inositol peak emerging last was recorded.
The trimethylsilyl derivatives were eluted with chloroform-methanol (2: 1) into scintillation vials, evaporated at room temperature, and the radioactivity was determined after addition of scintillation fluid.

Gel Electrophoresis
The cell ghosts were dissolved in 1% sodium dodecyl sulfate, and 5% 2-mercaptoethanol and heated to 100" for 2 min. Electrophoresis was performed with bromphenol blue as tracking dye in 7.570 acrylamide gel (30). Bovine serum albumin with a molecular weight of 68,000 (31), ovalbumin with a molecular weight of 44,500 (31), and ribonuclease A with a molecular weight of 13,700 (32) were used as standards. The gels were sliced with a razor blade gel slicer, and the slices were counted in toluene-based scintillation fluid after NCS solubilization at 50" for overnight. ghosts (plasma membranes)z and by the autoradiograph of cells (Fig. 1).

Autoradiography
Total radioactivities of packed intact cells, of cells treated with proteases and neuraminidase, and of labeled lysed ghosts were compared and are list,ed in Table I. Total labeling of erythrocytes was dependent on the quantity of galactose oxidase and was found to be linear up to 100 pg of galactose oxidase protein added per 5 ml of packed cells (Fig. 2), which indicates that most of the label is due to aldehyde groups created by galactose oxidase. 2 Determination of the proportion of the label in surface membranes to that of total erythrocytes was difficult to perform as the total labeled activity of intact cells was impossible to determine due to chemiluminescence caused by hemoglobin. Labeling of isolated cell ghosts resulted in a much greater label than in intact cells, especially in the lipid fraction (Table  I and Fig. 4b).
Nonspeci$c Labeling Nonspecific labeling, i.e. a label occurring without galactose oxidase but only with tritiated sodium borohydride, was observed (Table I) ; the major nonspecific label, although weak, occurred in an unidentified lipid fraction,3 as demonstrated on gel electrophoresis (see Fig. 4e) and by lipid extraction, but some very weak nonspecific labels were also found in protein. 3 Nonspecific labeling was not remarkable when intact erythrocytes were labeled, but was greatly enhanced when cell ghosts were labeled (Table I,  The specific activities (counts per min per nmoles) of labeled galactose and N-acetylgalactosamine in glycoproteins, globoside, and CTH are shown in Table II, and the dependency of those activities on the amount of galactose residue is shown in Fig. 3. The specific activity was higher in the galactosyl residue for proteins and in the N-acetylgalactosaminyl residue for lipids (Fig. 3). A remarkable increase of specific activity of glycoprotein galactose was observed after treatment with proteases, although total labeling increased only slightly. The specific activities of both galactosamine and galactose in globoside increased after cells were treated with Pronase.
Labeling for both galactose and N-acetylgalactosamine in glycoprotein and for N-acetylgalactosamine in globoside increased (10 times) when the amounts of galactose oxidase added 3 Nonspecific label without galactose oxidase in lipid fractions could be aliphatic aldehydes (plasmals), ketosphingosine, and pyridinium compounds plausibly bound to lipids. They are, however, not identified.
Nonspecific label for protein is unknown, but any reducible structure as have been found in collagenous protein in the form of Schiff's base (34) can be considered.
Nothing is as yet identified.  with a terminal N-acetylgalactosamine (globoside) than giycoprotein, while that of galactose was greater in glyconrotein than in alvcolipid (globoside) : (b) the specific activity of galactosamine in &ob&ide'was greatly 'enhanced on increask of galactose oxidase added, while that of galactose in globoside did not increase much.
were increased, while labeling for the galactosyl residue of globoside increased to a lesser degree when the amount of galactose oxidase increased (Fig. 3). No label was found in glucose.

Comparison of Xpec$ic Activities of Individual Glycolipids
Distribution of labels in various glycolipids is shown in Table  III.
The major label was found in globoside, followed by ceramide trihexoside, while no label was present in lactosylceramide. It is noteworthy that the label ratios for individual glycolipids (globoside-CTH-CDH) are nearly constant and are not greatly changed by treatment with neuraminidase and proteases. Some labels for ceramides with a long carbohydrate chain were also found (see footnote to Table III).
The labeling ratio between globoside, CTH, and CDH differed greatly from the actual chemical quantities of these glycolipids present in membranes (see Table  III).
The "degree of exposure," as expressed by counts per min per /.&M amount, was quite high in globoside as compared to other glycolipids.

Relative Radioactivities of Labeled Glycoprofeins and Glycolipids
The labeled glycoproteins and glycolipids were separated by sodium dodecyl sulfate gel electrophoresis (7), and the relative Peaks a, b, and c, corresponding, respectively, to apparent molecular weights of 9.5, 8.2, and 6.4 x 104, and a sharp lipid peak (L) were observed (Fig. 4a). If the cell membranes were first prepared and then labeled, the sodium dodecyl sulfate polyacrylamide gel electrophoresis pat.tern was significantly different.
There was a dominant Peak L and a labeled glycoprotein with high apparent molecular weight (1.5 x lo&), in addition to Peaks a, b, and a very weak c; thus, a highly active extra peak was labeled when membranes were first prepared then labeled (Fig. 4b and "Discussion").

Labeling Pattern of Cells Whose Surjuces
Were ModiJed by Enzymes About 607, of the total label in intact cells was found in protein and 407, was in lipid.
The total label was greatly increased by neuraminidase treatment but only slightly by protease treatment (Table I).
ilfter protease treatment, the label in glycolipid increased more than in glycoprotein (Table I).
With trypsin treatment glycoproteins a and b lost some activity, but activity for the lipid peak (L) intensified somewhat (see Fig. 4~). After neuraminidase treatment an increased label was found mainly in glycoprotein with a smaller increase in glycolipid (compare Fig. 4, a and d) ; thus, about S5yo of the total label was found in glycoprotein after treatment with neuraminidase. Neuraminidase-treated erythrocytes showed remarkable enhancement in a particular glycoprotein peak with an apparent molecular weight 8.5 x 104, which is probably derived from glycoprotein a, and the appearance of a new peak with an apparent molecular weight 3.5 X 104. Also, the label in the lipid peak (L) was enhanced (see Fig. 4d).

Comparison of Labeling Patterns of Normal Adult Erythrocytes and oj Fetal Erythrocytes
Glycolipid of fetal erythrocytes obtained from 3 months gestation period was labeled less efficiently than adult erythrocytes, although the label proportions of individual glycolipids (globoside-CTH-CDH) is not greatly different in adult and fetal erythrocytes (Table I).
A great deal of enhanced agglutinability by antigloboside antisera was demonstrated in fetal cells, however, in agreement with the previous results (9). The most remarkable label difference between normal and fetal erythrocytes was demonstrated in the ratio of activities between protein-bound carbohydrates (glycoprotein) and lipid-bound carbohydrates (glycolipid) ( Table I and Fig. 4f). Fetal erythrocytes showed a very weak activity in the area of glycoproteins c and d. No activity corresponding to Peaks a and b was demonstrated, whereas the activity for lipid peak (L) was remarkably demonstrated. ificity for galactose and N-acetylgalactosamine, whose primary hydroxyl groups are oxidized to aldehyde groups (26). Oxidation by galactose oxidase followed by reduction with t'ritiated sodium borohydride has been used previously to label galactosyl and galactosaminyl residues in glycolipids (37-40) and glycoproteins (41,42). This reaction has now been successfully applied to external labeling of galactosyl and galactosaminyl residues at the cell periphery.
Lactoperoxidase with a molecular weight of 78,000 (22) is known to react exclusively at the red cell surface. Galactose oxidase has a molecular weight of 75,000 (26), which closely approximates that of lactoperoxidase. Therefore, penetration of this enzyme through the cell membrane should not occur.
Surface labeling of erythrocytes was indicated by the autoradiograph of cells as seen in Fig. 1.

DISCUSSION
Cell ghosts, isolated then labeled, demonstrated a highly labeled protein with apparent molecular weight of 150,000, whereas such a protein is not seen in membranes of labeled intact cells (cells labeled then ghosts isolated) (see Fig. 4b). This protein, therefore, could be located on the inner surface of the plasma membrane.
The results indicate that galactose oxidase cannot penetrate the cell membrane to label this protein in the intact cell, In this study galactose oxidase from Dactylium dendroides Nonspecific labeling (see Footnote 3) was not significant in has been used to obtain specific labeling of cell surface glyco-intact cells, but increased to a great extent when ghosts were proteins and glycolipids.
Galactose oxidase shows a strict spec-labeled. Enhanced lipid labeling in ghosts as seen in Fig. 4h by guest on March 24, 2020 http://www.jbc.org/ Downloaded from is largely due to increased nonspecific labeling, i.e. the naturally occurring reducible lipid component increased when cells were lysed.
The reducible materials iu intact membranes can be made more accessible to borohydride by lysis of membrane, either by change of membrane conformation or by exposure of the inner surface.
Although naturally occurring materials reducible by borohydride have not been identified in membrane, this could be an interesting subject in membrane chemistry. This study gives direct evidence that sugars of both glycolipids and glycoproteins are exposed on the membrane surface. Two proteins have been known to be exposed on the outer surface of human erythrocytes (22,23,41). One of these is the major membrane glycoprotein, which carries all demonstrable sialic acid (22-24), Al3 and MN blood group antigens, and receptors for influenza virus and phytoagglutinins (43). There is evidence that this protein traverses the cell membrane to the cytoplasmic surface (44). The region of protein which is exposed to the outside carries the carbohydrate portion, while the COOH-terminal end is enriched in hydrophobic amino acids and possibly serves to attach the protein to the membrane (43). This idea that membrane proteins have a hydrophobic inner end and a protease-sensitive hydrophilic outer segment is also true for cytochrome bg arranged in the microsomal membrane (45)

and Semliki
Forest virus membrane proteins (46), and it possibly has general significance.
The main labeled glycoprotein peak (a) has an apparent molecular weight of 95,000 and no doubt corresponds to the multispecific glycoprotein (22, 23, 43). The major oligosaccharide portion of this protein has been proposed to have the structure NeuNAc LY (2 + 3) Gal /3 (1 + 3) [NeuNAc Q (2 + 6) ) GalNAc (47). This structure could explain the more efficient labeling after neuraminidase treatment.
We have also obtained direct evidence that some glycosphingolipids of membranes are exposed directly to the external environment.
However, surface exposure of globoside has been doubted because of the following two findings: (a) treatment of intact erythrocytes by jack bean fi-N-acetylhexosaminidase does not hydrolyze globoside in membranes, but globoside is readily hydrolyzed when an aqueous solution of globoside is treated with jack bean @N-acetylhexosaminidase;4 and (b) intact human erythrocytes are only slightly reactive to anti-globoside antisera (S, 9). Human adult erythrocytes were not agglutinated nor hemolyaed by antigloboside antisera although globoside is the majol glycolipid of the erythrocyte membrane.
They become reactive to anti-globoside only after treatment with neuraminidase or with proteases (9). Absorption capability of anti-globoside by human erythrocytes is, in fact, increased by enzyme treatment (9). Isecause human fetal erythrocytes were highly reactive to anti-globoside without enzyme treatment, globoside groups were thought to be "cryptic" in human adult erythrocytes and "exposed" in fetal erythrocytes or by enzyme treatments (9). This interpretation should be revised by the fact that globoside is highly labeled by the galactose oxidase method; the labeled activity of adult erythrocytes is not greatly different from fetal erythrocytes and is unchanged after enzyme treatment. in view of a recent observation by Nicolson (11) that trypsinization can cause clustering of some phytoagglutinin reactive sites on cell surfaces.
The relevant explanation for understanding labeling pattern of globoside (and other glycolipids) and glycoprotein is that globoside may be seated directly on the lipid bilayer among "bushes" of protein and glycoprotein (see Fig. 5)) and that globoside is available only to galactose oxidase but not to larger macromolecular reagents, such as immunoglobulin and P-N-acetylhexosaminidase.
In fact, molecular weight of galactosr oxidase (75,000) is smaller than immunoglobulin (mol w-t 180,000) 01 P-N-acetylhexosaminidase (mol wt 100,000). Other factors such as shape and ionic charge should also be considered. 011 thr surface of protease-treated cells, globoside is available not only to galactose oxidase but also to immunoglobulin (see Fig. 5). Lactosylceramide in the membrane was not labeled by this procedure.
However, CIIH alone can be easily labeled in vitro by t,his method (40), and we have shown that after Triton X-100 solubilization of the erythrocyte membranes, the labeling of CDT1 was greatly enhanced and comparable to the label of CTH. This indicated that the carbohydrate chain of CDH dots not protrude far enough to be reached by galactose oxidase in the membrane, in agreement with immunological data that human erythrocytes are not reactive to anti-lactosylceramide even after trypsin digestion (9).
Comparative labeling of glycoprotein and glycolipid has been observed by gel electrophoresis of labeled membranes. Labeling of glycoprotein varied greatly in contrast to a rather constant labeling of glycolipid. Absence of Peaks a or b glycoprotein in fetal erythrocytes is of great interest, which indicates either of the following possibilities: (a) fetal glycoproteins do not have any galactosyl or galactosaminyl residues so that they are not labeled; (b) galactosyl or galactosaminyl residues of fetal glycoprotein could be highly substituted by other sugar residues such as sialyl or fucosyl residues so that they are not labeled; and (c) fetal erythrocytes have smaller number of glycoproteins; glycoproteins a or b are virtually absent.
A tentative model based on the third possibility that a higher agglutinability of fetal erythrocytes to antigloboside can be ascribed to less steric hindrance due to the absence of some surface glycoprotein is shown by Fig. 5. Further estensive study is needed to correlate the surface structure revealed by ext.ernal labeling to the immunological reactivity of cells surfaces. A study in progress (1) showed that this surface labeling procedure for sugars was found to be extremely useful to distinguish the surface properties of various normal and transformed cells as well.
Note Added in Proof-Since this paper was processed, at the stage of proofing, we noticed that T. L. Steck used galactose oxidase for surface labeling of erythrocytes ((1972) in Membrane Research, edited by C. F. Fox, p. 71 Academic Press, New York).