Topographical distribution of complex carbohydrates in the erythrocyte membrane.

Abstract Human red blood cell membranes treated with galactose oxidase were specifically labeled in galactose and N-acetylgalactosamine residues by reduction with tritiated borohydride. Ceramide tri- and tetrahexosides were the most intensely reactive lipid species. Several labeled polypeptide peaks were resolved by polyacrylamide gel electrophoresis in sodium dodecyl sulfate. While some of the radioactivity appeared to correspond to known, principal glycoproteins, a major part of the label was distributed in regions of the gel bearing only weakly stained protein and glycoprotein components. The pattern of labeling was essentially the same for intact erythrocytes and unsealed ghosts. Preparations of sealed, inside-out vesicles showed very little labeling, all of which could be attributed to small amounts of contamination from accessible outer surface. On this basis, all of the glycoprotein and glycolipid sugars reactive with galactose oxidase plus tritiated borohydride can be assigned to the external surface of the membrane.

SUMMARY Human red blood cell membranes treated with galactose oxidase were specifically labeled in galactose and N-acetylgalactosamine residues by reduction with tritiated borohydride.
Ceramide tri-and tetrahexosides were the most intensely reactive lipid species.
Several labeled polypeptide peaks were resolved by polyacrylamide gel electrophoresis in sodium dodecyl sulfate.
While some of the radioactivity appeared to correspond to known, principal glycoproteins, a major part of the label was distributed in regions of the gel bearing only weakly stained protein and glycoprotein components.
The pattern of labeling was essentially the same for intact erythrocytes and unsealed ghosts. Preparations of sealed, inside-out vesicles showed very little labeling, all of which could be attributed to small amounts of contamination from accessible outer surface.
On this basis, all of the glycoprotein and glycolipid sugars reactive with galactose oxidase plus tritiated borohydride can be assigned to the external surface of the membrane. Isolated human erythrocyte membranes ("ghosts") contain approximately 8% carbohydrate (l), both as glycoprotein and glycolipid (2). The glycoproteins show a complex periodic acid-Schiff staining pattern on polyacrylamide gels subjected to electrophoresis in sodium dodecyl sulfate (3). The most conspicuous component is the sialoglycoprotein, termed PAS-l1 (3) or glycophorin (4). Another major membrane polypeptide, Band 3 (3), has recently been isolated and identified as a glycoprotein (5, 6)) although it reacts poorly with periodic acid-Schiff.
The total number of distinct glycoprotein species, their chemical character, and their disposition within the membrane remain to be established.
The intent of this study was to utilize a specific modification technique to enumerate and probe the orientation of sugar-bearing macromolecules in the erythrocyte membrane. Galactose oxidase has been found to oxidize the C-6 position of D-galactose, N-acetyl-o-galactosamine, and related sugars present in oligosaccharides (11). The resultant aldehyde groups have been labeled by reduction with tritiated borohydride in both isolated glycolipids (12) and soluble glycoproteins (13). We reasoned that while BaH4 might readily penetrate even well sealed membranes, galactose oxidase would not, and could therefore be used to probe the two membrane surfaces selectively for reactive sugars. For this purpose, intact erythrocytes or resealed ghosts were used to examine the external membrane face, while sealed inside-out vesicles presented the cytoplasmic membrane surface for unilateral reaction.
Our findings indicated a surprising multiplicity of glycoprotein species not clearly demonstrated by conventional stains. Furthermore, all of the reactive glycoprotein and glycolipid sugars appeared to be available only at the external membrane surface. Since our first communication of these findings (9), Gahmberg and Hakomori (14) have used this technique to confirm certain features of the labeling reaction.
Its specific activity was given as 180 i.u. or 5 X lo5 Kabi units per mg of protein.
Galactose oxidase preparations from General Biochemicals, Inc. and Worthington were used in early experiments. Proteolysis of the membrane proteins, as reflected by degradation of their gel-banding pattern, was observed following incubation of ghosts with all of these enzymes. The protease could be inactivated without significant loss of galactose oxidase activity by a preincubation of the diluted enzyme at 50" for 30 min.
Sodium and potassium B3H4 were purchased from Amersham-Searle; the nominal specific activities of the three batches used were 304, 102, and 590 mCi per mmole. NaBH4 was obtained from Sigma Chemical Co. Reagents for gel electrophoresis and glycolipid analysis were essentially as in Refs. 3 and 15, respectively.

Labeling
Reaction-In the absence of galactose oxidase, BaH4 introduced a small amount of tritium into membrane protein and lipid fractions (Fig. 1). Galactose oxidase usually stimulated the incorporation of radioactivity lo-to 30-fold. The time-dependent labeling of the lipids (Fig. 1) is attributable to the more complex glycosphingolipids, which continue to be stimulated by the presence of galactose oxidase for at least 24 hours (Fig. 2). In contrast, incorporation of tritium into the protein fraction reached a maximum after approximately 1 hour of incubation with galactose oxidase; thereafter, labeling leveled off or even declined, as in Fig. 1. As a consequence, the relative proportion of label in the lipid fraction increased steadily with time, from approximately 10 y. to 35 % of the total incorporated radioactivity.
The decline in protein labeling was apparently not caused by enzyme inactivation, since the lipid fraction showed increasing incorporation during this period. Furthermore, the decline could not be attributed to the nature of the terminal sugar, since both the terminal N-acetylgalactosamine of globoside and the terminal galactose of trihexosylceramide showed increasing radioactivity over a 24-hour period (Fig. 2). An inexplicable, abrupt slowing of ceruloplasmin labeling was also reported by Morel1 et al. (13).
In order to minimize proteolysis and the breakdown of the membrane permeability barrier, brief and mild incubation condi-  Unsealed ghosts were incubated with the designated amount of galactose oxidase plus KB3Hd (11.4 X lo6 cpm) as described under "Experimental Procedures," except that the reaction mixture was scaled to one-half volume.
FIG. 5 (right). Oligosaccharide oxidation by galactose oxidase: and lacto-n-fucopentaose II (Gal-GlcNAc- were incubated with 0.5 mg of galactose oxidase (Worthington) in 3.0 ml of 1 M Tris-HCl (pH 7.0). Oxidation was followed by measuring the generation of Hz02 with a peroxidase nlus 3,3'-dimethoxvbenzidine system by the increase in optical absorbance at 420 nm (cf. 11). -Raffinose, which is completely oxidized within one hour, was used as a reference standard.
tions. Fig. 3 depicts the time course and Fig. 4 the galactose oxidase concentration dependence of the reaction under these conditions.
We estimate that approximately 0.1 to 1% of the membrane galactose plus N-acetylgalactosamine is labeled in such experiments.
Specificity of Labeling Reaction--In keeping with the known 'characteristics of galactose oxidase (II), the presence of 45 mM n-glucose in the medium did not significantly alter the reaction, while 45 mM n-galactose caused a 99% inhibition of membrane labeling.
This experiment also suggested that the Hz02 generated by the oxidase reaction does not stimulate tritium incorporation into membranes.
Similarly, the presence of catalase did not alter the reaction, indicating that Hz02 neither stimulated labeling nor inhibited the enzyme (11) under the conditions employed. Pretreatment of the membranes with an excess of unlabeled NaBH4 prior to the addition of galactose oxidase did not diminish the subsequent labeling reaction, while the presence of 1 mM KCN, a known inhibitor of galactose oxidase (1 I), did.
The reaction of individual sugars was examined by hydrolyzing labeled ghost membranes in 2 N HCl at 98" for 2 hours and chromatographing the digest on paper in butanol-pyridine-water (6:4: Galactose and galactosamine were labeled in roughly equal proportions; the radioactivity recovered in all other sugars amounted to less than 2% of that found in these two species. However, the recovery of total radioactivity in the monosaccharide spots was only about one-third of that predicted. The question of whether only nonreducing terminal galactose could be attacked by this enzyme was investigated using three model oligosaccharides (Fig. 5). It was found that nearly all of the galactose in lactose could be oxidized during a 400-hour incubation with galactose oxidase. However, only about one-half of the galactose in lacto-N-tetraose and lacto-N-fucopentaose II was oxidized.
This suggests that only one of the two galactosyl moieties in each of the latter two compounds was susceptible to the enzyme, presumably the nonreducing terminal species.
Preincubation of ghost membranes with sialidase from Clostridium perfringens (16) led to a mild ( <a-fold) stimulation of labeling by galactose oxidase plus tritiated borohydride (cf. also Ref. 14). Presumably, the effect of sialidase was to release terminal sialic acid moieties, thus exposing the penultimate galactose residues of the membrane sialoglycoproteins (1, 23, 24), as was found for ceruloplasmin by Morel1 et al. (13). This effect lends support to the premise that only galactose and N-acetylgalactosamine in the nonreducing terminal position are attacked by galactose oxidase. In contrast to sialidase, pretreatment with trypsin led to no significant change in the incorporation of tritium into the red cell membrane or the derived glycolipid fraction. Preincubation with pronase caused a mild reduction in both total and glycolipid labeling (but cf. Ref. 14).
Siokdness of Membrane Labeling-A principal aim of this study was to use galactose oxidase as a nonpenetrating probe which would label in turn the carbohydrate present at each membrane surface. The outer membrane face can be reacted by using intact erythrocytes or resealed ghosts. Conversely, the inner (cytoplasmic) membrane surface is unilaterally accessible in sealed inside-out vesicles. Our standard, unsealed ghosts appear freely permeable to proteins and would permit both surfaces to react (cf. Refs. 9, 17, 18).
Intact erythrocyte membranes, resealed and unsealed ghosts were all well labeled by galactose oxidase plus B3Hd. While intact cell membranes were less well labeled than the isolated ghosts, it will be demonstrated later that the labeling pattern of membranes in the intact cell and in isolation were qualit,at,ively the same. In contrast, the reactivity of inside-out vesicles was invariably greatly reduced compared to an equivalent amount of unsealed or resealed ghosts, as illustrated in Table I. Here, the labeling of resealed ghosts was stimulated more than lo-fold by galactose oxidase, while less than a 2-fold increase over background was observed in the inside-out vesicles. The correlation between the relative amount of inside-out vesicle labeling (6.3% that of ghosts) and the fraction of total outer surface acetylcholinesterase accessible (6.5%) has been observed repeatedly, although such precise agreement does not always occur.
Several possible causes for the diminished labeling of inside-out vesicles were considered.
The presence of an inhibitor in these preparations appeared unlikely, because 1: 1 mixtures of reactive ghosts and unreactive vesicles gave essentially additive labeling (Table I, Line 3). The loss of reactive carbohydrate from the membrane during vesicle preparation was ruled out by these observations.
(a) Disrupting the membrane permeability barrier with Triton X-100 enhanced inside-out vesicle labeling lo-fold, to a value nearly equal to that of the ghosts (which were not stimulated by detergent) ( Table 1, Lines 4 and 5). (b) Inside-out vesicles are not diminished in sialic acid (16), periodic acid-schiffreactive glycoproteins (e.g. Fig. 6), glycolipids, or neutral sugars. (c) Vesicles prepared from membranes which had been reacted with galactose oxidase and tritiated borohydride while in the intact cell show little or no loss of specific activity compared to the parent ghosts. Finally, the possibility that the inverted vesicles react very slowly was excluded by the demonstration that overnight incubation with galactose oxidase led to no increase in tritium incorporation (Table I, Line 6). The most likely explanation for the diminished labeling of inside-out vesicle preparations, therefore, was that few, if any, reactive sugars are localized at the cytoplasmic surface of the membrane.
It remained to be established, however, whether all of the small amount of label incorporated into inside-out vesicle preparations could be attributed to contamination by exposed outer surface or whether unique minor constituents were present at the inner surface.
Characterization of Labeled Proteins- Fig.  6 demonstrates the electrophoretic pattern of the membrane polypeptides and glycoproteins, and the profile of tritium incorporation following treatment of membranes with galactose oxidase + B3H4. Since 2138 TABLE I Reactivity of' right side out and inside-out membranes Resealed ghosts were prepared in Buffer A containing 1 mM MgSO, (17,18). Inside-out vesicles were prepared in 0.5 mM sodium phosphate (pH 8) and purified by dextran 110 gradient centrifugation and aqueous partition (18); 6.5% of their total acetylcholinesterase activity was accessible. Resealed ghosts and inside-out vesicles equivalent in total acetylcholinesterase activity to 30 ~1 of packed unsealed ghosts were diluted with Buffer A to 100 ~1. Where indicated, membranes were treated with 0.1% Triton X-100 (1 pg of detergent per pg of lipid) to disrupt their permeability barrier. Zero or 1 i.u. of galactose oxidase in 100 ~1 of 100 mM sodium phosphate (pH 8.5) and 2 X 10' cpm of B3H& in 20 ~1 of 0.01 N NaOH were added. The tubes were incubated at room temperature for 1 hour, except for one 18-hour incubation (Line 6). In this case, the B3H4 was added at the end of that incubation (but at the same time as the other samples). Following the reaction, the specimens were washed with 50 rnM NagHPOb, except for the Triton-treated samples. To those tubes, 0.1 ml of 30y0 trichloroacetic acid was added. After 10 min on ice, the precipitates that formed were washed 3 times with 10 ml of ice-cold 6% trichloroacetic acid. Coomassie blue does not stain the glycoproteins sufficiently (3), they are considered separately in terms of their periodic acid-Schiff profile, and no correspondence is drawn between the two scans. The only peak of label observed in the absence of galactose oxidase occurred in the lipid region, which trails just behind the tracking dye (Fig. 6, Panel A-3). Incubation with galactose oxidase brought about no alteration in the Coomassie blue or periodic acid-Schiff-staining profiles (compare Panels A-l and A-d with B-l and B-2). These scans indicate that no significant proteolysis or cross-linking occurred.
Galactose oxidase caused several closely spaced peaks of radioactivity in the midzone of the gel (Fig. 6, Panel B-S). The slow moving peaks (Slices b7 to 37) corresponded in relative mobility to Bands 3 and PAS-l, both of which have been demonstrated to be major glycoproteins (cf. Refs. 3 to 6). Most of the radioactivity, however, fell in a series of peaks which ran between Bands 4.2 and 5, a region of the gel where only minor components are detected by the protein and glycoprotein stains. The most intense labeling seemed to coincide with the minor glycoprotein zone, termed PAS-4 (5), and the ill-defined Coomassie blue-stained complex, designated 4.5 (Fig.  6, B-l to B-S). It should be noted that the pattern of labeled peaks was not precisely duplicated in relative intensity, position, or number from experiment to experiment.
We suspect that this variation resides in the multiplicity of overlapping component species which may vary slightly in relative mobility and thus sum differently in each experiment (3).
A peak of radioactivity running just behind the tracking dye was also created by galactose oxidase treatment; this is attributable to glycolipids, as discussed below.
In addition, there is frequently a low profile of radioactivity between the gel origin and Band 3. Since the labeling pattern in this region is not consistent and does not correspond to any demonstrable glycoprotein components, we suspect that this low mobility material may represent traces of glycoproteins which have become covalently cross-linked through Schiff bases formed from galactoaldehydes and reduced by B3H4.
The identical labeling procedure was applied to inside-out vesicle preparations (Fig. 6C). The protein profile of these vesicles (Panel C-l) differs characteristically from that of ghosts, in that Bands 1, 2, and 5 are depleted.
The elution of these three components is an inevitable concomitant of the vesiculation process (16, 18), but does not affect the results here since the missing polypeptides bear no label (vide infra).
The Coomassie blue and periodic acid Schiff-staining profiles of galactose oxidase plus B3H4-treated vesicles were indistinguishable from untreated controls (not shown).
These vesicles, which showed a 9.9% accessibility of their outer surface acetylcholinesterase, incorporated 9.2% as much tritium as the equivalent amount of unsealed ghosts (shown in Fig. 6B). Th e radioactivity was not confined to any distinctive electrophoretic peak, but was distributed much as in the unsealed ghosts, only at one-tenth intensity (Fig. 6, Panel C-S). It is noteworthy that the glycolipid region was no better labeled than the glycoproteins.
Additional experiments were performed to substantiate that the tritium was being incorporated only into glycosylated constituents of the outer surface of the red cell membrane.
1. We knew from previous studies (21) that all of the membrane sialic acid, neutral sugar, and glycoproteins remain membrane-bound following treatment with 0.1 N NaOH and other protein perturbants, while the nonglycosylated polypeptide species are eluted.
If the membrane-bound radioactivity was in fact confined only to glycoproteins and glycolipids, it should be recovered quantitatively in the NaOH residue. We therefore performed such an extraction on ghosts labeled by galactose oxidase plus B3Hd treatment.
As predicted, the radioactivity released by NaOH was no greater than that recovered in the "minus enzyme" control.
The NaOH residue contained Bands 3, 4.5, and 7, all of the periodic acid-Schiff peaks, and the complete radioactivity profile. 2. To verify that the complex, high molecular weight labeling pattern seen in Fig. 6 represented only glycoproteins, ghosts (200 pg of protein) labeled by galactose oxidase plus BaH4 treatment were incubated with 2 pg of pronase (Calbiochem) in 0.1% sodium dodecyl sulfate for 1 hour at 37". Gel electrophoresis showed that the Coomassie blue-and periodic acid-Schiff-stained proteins and the corresponding radioactive peaks were extensively degraded, while the periodic acid-Schiff stain and tritium counts in the lipid region remained unchanged.
3. Since no distinctive gel component was labeled at the cytoplasmic surface of the membrane (Fig. 6C), it was postulated that the unilateral action of galactose oxidase on the outer surface of the membrane (i.e. on intact erythrocytes) should generate the same radioactivity profile seen with unsealed ghosts, where both surfaces are accessible. A comparison of Fig. 6B with Fig. 7 affirms this prediction.
4. Finally, we sought to determine whether the galactose oxi- The gels were stained with A; an inside-out vesicle preparation (with 9.9% of total acetyl-periodic acid-Schiff and scanned at 560 nm, stained with Coomassie cholinesterase accessible) was diluted with Buffer A to the same blue, and scanned again at 515 nm, then sliced for counting as membrane concentration in terms of total acetylcholinesterase described under "Experimental Procedures." A, ghosts minus activity.
dase labeling pattern corresponded to the membrane sialoglycoprotein or represented a set of different polypeptides. Ghosts were treated with NaI04 so as to oxidize only sialic acid (25), then reduced with B3H4. The periodate treatment led to a lo-fold stimulation of labeling.
That this incorporation of radioactivity was attributable to sialic acid was shown by the fact that 78% of tritium incorporation was abrogated by prior digestion of the membranes with sialidase and that 90% of the label was released from the ghosts by digestion with 0.1 N HzS04 at 80" for 1 hour. The pattern of radioactivity upon polyacrylamide gel electrophoresis is depicted in Fig. 8. The profile of label closely resembles that of the normal periodic acid-Schiff-staining pattern (Fig. 6 Table II. Approximately 15% of the total radioactivity was recovered in the lipid fractions. In the absence of galactose oxidase, most of the lipid radioactivity was found in the phospholipid and cholesterol fractions, presumably as a result of the reduction of their double bonds. Incubation with galactose oxidase stimulated the labeling of glycolipids Sl-fold in the ghosts and 5.2.fold in the inside-out vesicles. The glycolipid fraction of the inside-out vesicles was thus labeled 17% as well as in the ghosts, i.e. about at the same level as the unfractionated sample. (While the acetylcholinesterase accessibility was initially 9.9 '$J',, it is conceivable that the overnight incubation increased the accessibility to 17 ye.) The apparent mild stimulation of labeling in the phospholipid fractions by galactose oxidase (Line 2) was probably caused by the presence of some contaminating substances such as glycoproteins, since only 29% of this radioactivity could be recovered in the principal phospholipid classes following thin layer chromatography (i.e. phosphatidylethanolamine, 5.8%; phosphatidylserine, 4.8%; phosphatidylcholine, 7.8%; and sphingomyelin, 11.4%).
The remainder was confined to the base-line (30%) and solvent front regions (40%).
In any case, the galactose oxidase . After a 2-hour incubation at room temperature, the cells were washed once in 40 ml of saline, then lysed in 40 ml of Buffer A. The pelleted membranes were resuspended in 100 ~1 of 100 rnM sodium phosphate, pH 8.0, and 10 ~1 of NaBzH, ('2.0 X lo6 cpm in 0.01 N NaOH) were added. After 10 min at room temperature, the membranes were washed and an aliquot equivalent to 10 ~1 of packed ghosts was electrophoresed.
Top, scan of the Coomassie blue-stained gel of the galactose oxidase-reacted membranes at 530 nm. (This wavelength was chosen so that the absorbance scale was comparable to that in Fig. 6.) Middle, radioactivity profile of galactose oxidase-treated sample. Bottom, radioactivity profile of minus galactose oxidase control.
reaction is seen to be quite specific for the glycosphingolipids when we consider that the molar ratio of glycolipid to cholesterol to phospholipid in this membrane is approximately 1: 16 :20 (2). The specificity of glycolipid labeling was further defined by isolating the principal ceramide fractions and analyzing their sugar, fatty acid, and sphingosine moieties (Table II).
In the absence of galactose oxidase there was consistently more label in the sphingosine than in the fatty acid fractions.
This feature may reflect the low degree of unsaturation in the glycosphingolipid fatty acids (2) and the presumed localization of the fatty acid double bonds deep in the membrane hydrophobic core, as compared to the more superficial location of the A4 unsaturation in sphingosine.
This hypothesis is consistent with the aforementioned observation that sphingomyelin was the phospholipid most extensively labeled by B*H4, despite the relative abundance of unsaturated fatty acids in the other phospholipids.
As demonstrated in Fig. 2, the labeling of the sugars in the ghost glucosyl ceramide was negligible, that in lactosyl ceramide was small but significant, whereas the tri-and tetrahexosyl ceramide fractions showed a large stimulation by galactose oxi-SLICE NUMBER FIG. 8. Periodate-stimulated labeling of membrane polypeptides: 50 pl of packed ghosts were diluted with 50 ~1 of Buffer A, and incubated with 100 ~1 of 2 mM NaI04 in 0.1 M sodium acetate buffer (pH 5) in the dark on ice for 10 min. The reaction was terminated by the addition of 10 ml of sodium arsenite (5 mM) in 50 mM NagHP04 (pH 9.2). The membranes were pelleted and resuspended to 200 ~1 in 50 mM NasHPOa; 20 ~1 of NaB3H4 (5.9 X lo7 cpm) were added. After 15 min at room temperature, the membranes were washed and the equivalent of 10 ~1 of ghosts were taken for electrophoretic analysis. dase. The small apparent increase in fatty acid and sphingosine labeling in the higher glycolipids most likely represents their incomplete separation from the heavily labeled hexoses; the apparent labeling of the hematoside fraction may also be attributable to a slight contamination from the trailing edge of the heavily labeled globoside band. The ganglioside fraction was examined for the presence of minor components of high specific activity but negligible amounts of aH-labeled material were found.
A principal objective in performing this analysis was to determine whether the label incorporated into the lipids of inside-out vesicle preparations represented the presence of any particular glycolipid species at the cytoplasmic surface of the membrane. Table II indicates that none of the glycosphingolipids was preferentially reactive in the inside-out vesicle preparation.
That is, each species was labeled in rough proportion to the 6-fold difference in over-all tritium incorporation between the ghosts and vesicles. This difference in radioactivity does not represent preferential losses of vesicle glycolipids during processing, since the specific activity of the sugar moieties in these samples exhibited the same differential as the total radioactivity values given in Table II. DISCUSSION Galactose oxidase oxidizes terminal galactose or N-acetylgalactosamine residues in complex carbohydrates whether presented in soluble form or attached to the outer surface of cells. Some limitations, however, were encountered in applying the galactose oxidase + B3H4 system to the red cell membrane.
(a) A small amount of labeling was observed, particularly in lipids, in the absence of galactose oxidase; however, this effect was negligible in comparison to the specific enhancement of label in the presence of galactose oxidase. (b) Labeling of the intact cell is reduced compared to an equivalent amount of ghosts, presumably because the cytoplasm consumes borohydride.
We have dealt with this problem by isolating the cell membranes after enzyme treatment but before reductive tritiation.
(c) The terminal sugars were not quantitatively labeled. For example, lactosyl ceramide was much less well labeled than trihexosyl ceramide, despite the fact that both have galactose at their termini and are present in roughly equal proportions in this membrane (2). This feature was also noted by Gahmberg and Hakomori (14). Although in some studies a significant level of labeling of lactosyl ceramide was seen after 24 and 48 hours, we elected to restrict the duration and intensity of the reaction so as to avoid perturbing the membrane permeability barrier or architecture, which would have destroyed the discrimination of sidedness. We have also observed that as BH4 concentrations are raised above 1 mM, the gel pattern of the membrane polypeptides undergoes progressive degradation typical of proteolysis.
That borohydride can cleave peptide bonds (albeit under more drastic conditions) was recognized long ago (27, 28); we therefore kept the concentration of this reagent below 1 mM.
(d) All of the various galactose oxidase preparations tested showed protease activity, to which the protein electrophoretic pattern in sodium dodecyl sulfate is exquisitely sensitive (3). We minimized this hazard by using the purer Kabi enzyme and by preincubating this enzyme at 50" for 30 min when necessary.
(e) The galactoaldehyde enzyme product can conceivably generate Schiff bases, which can be reduced by B3H4 to yield labeled, irreversible intra-and intermolecular linkages. We believe this is the source of the anomalous high molecular weight radioactive material occasionally seen on gels (c./. Figs. 6 and 7 and Ref. 14). Because of a potentially high frequency of collisions, proteins in 2142 species could contribute to the periodic acid-Schiff-staining profile migrating ahead of PAS-4 and to the ill-defined Coomassie blue-stained zone designated 4.5. That such glycoproteins are weakly stained by periodic acid-Schiff is not surprising, considering that Band 3, containing about, 10% of the membrane sugar, is so poorly stained by this technique (3,5). The periodic acid-Schiff profile appears primarily to reflect the sialoproteins, as judged by its resemblance to the mild periodate + B3H4 labeling pattern ( Fig. 8 and Ref. 26).
The galactose oxidase plus B3H4 reaction invariably yields a complex labeling pattern in the vicinity of Band 3 (e.g. Fig. 6, Panel B-S and Fig. 7). Preliminary studies have shown that purified Band 3 manifests a similar complex radioactive pattern.2 The ratio of radioactivity to Coomassie blue-staining intensity is highest over the diffuse trailing edge of the Coomassie bluestained band and is relat.ively weak over the sharp and intense leading edge. This finding is compatible with variable or incomplete glycosylation of the Band 3 polypeptide. The extent to which the complexity of the over-all ghost labeling pattern reflects this type of heterogeneity would be important to know.
The galactose oxidase plus B3H4 labeling technique has thus identified and provided a means to study a level of glycoprotein multiplicity heretofore unrecognized in this membrane. The presence of carbohydrate at the outer surface of plasma membranes has been well established (30) ; the state of the cytoplasmic side of the membrane is less clearly understood.
In the case of the human erythrocyte, Eylar et al. (31) demonstrated that all of the sialic acid could be digested from the surface of the intact cell. The finding that the sialic acid is inaccessible in inside-out vesicles derived from this membrane (16) showed that these sugars are fixed asymmetrically at the outer membrane surface. It is also well known that many antibodies and plant agglutinins can bind to carbohydrate at the red cell surface. Nicolson and Singer (32) have provided electron microscopic evidence that at least, some plant agglutinins do not bind at the inner membrane surface. Unfortunately, this powerful technique has not directly identified the receptor molecules, but rather has indicated the sugar specificity of binding.
Using galactose oxidase as a membrane-impermeable probe, we have shown that all reactive glycoproteins and glycolipids were labeled only at the outer membrane surface. What little labeling occurred in inside-out vesicle preparations was accounted for by contaminating outer surface species, as reflected by the accessibility of acetylcholinesterase activity.
No glycoprotein or glycolipid sugar groups, therefore, were detected at the inner surface. Independent support of this conclusion comes from the fact that those polypeptides believed to be confined to the cytoplasmic side of the membrane (cf. 9) are selectively eluted by NaOH without detectable label.
It is striking that no glycolipid labeling appears to occur at the cytoplasmic side of the membrane, even during an overnight incubation.
These data provide no support. for a diffusional equilibrium (flip-flop) of glycolipids across the plane of the membrane, as reported by McConnell and his colleagues for spinlabeled phospholipid analogues in artificial (33) and natural (34) membrane vesicles.
It was heretofore shown that two glycoproteins present at the external surface of the human red cell membrane (Bands 3 and PAS-l) span the membrane asymmetrically (7)(8)(9)(10).
(Furthermore, PAS-2 may be an alternate form of PAS-l (35).) Our present findings suggest that a variety of other glycoproteins populate the outer surface, but whether they also extend through the thickness of the membrane remains to be established.
The unilateral, outer surface localization of sugars supports the widely held hypothesis that plasma membrane carbohydrates bear information important to intercellular interactions.