Characterization of the binding of propionibacterium granulosum to glycosphingolipids adsorbed on surfaces. An apparent recognition of lactose which is dependent on the ceramide structure.

The binding properties of a strain of Propionibacterium granulosum derived from human skin was investigated with reference to glycosphingolipids separated on thin layer chromatograms or coated in microtiter wells using externally (125I) and metabolically [( 35S]methionine) labeled bacteria. Binding was found to lactosylceramide (LacCer; Gal beta 1-4Glc beta 1-Cer) but not to glycolipids lacking the lactose sequence (i.e. Glc beta 1-Cer, Gal beta 1-Cer or Gal alpha 1-4Gal beta 1-Cer). In microtiter wells, binding occurred at 50 ng and became half-maximal at the theoretical value for a monomolecular layer of LacCer (i.e. 100-200 ng/well). The bacteria also bound to glycolipids with various substitutions (e.g. GalNAc beta 1-4, Gal beta 1-3GalNAc beta 1-4, Fuc alpha 1-2Gal beta 1-3GalNAc beta 1-4, Gal alpha 1-3, GlcNAc beta 1-3, Gal beta 1-3GlcNAc beta 1-3, Gal beta 1-4GlcNAc beta 1-3, and Gal beta 1-3(Fuc alpha 1-4)GlcNAc beta 1-3) at the Gal of LacCer, although only those species with GalNAc beta 1-4 or Gal beta 1-3GalNAc beta 1-4 were as active as LacCer itself. Glycolipids with other additions (e.g. Gal alpha 1-4 and NeuAc alpha 2-3) were negative. For unsubstituted LacCer, the binding required either a trihydroxy base or 2-hydroxy fatty acid, specifying the epithelial type of ceramide; LacCer composed of a dihydroxy base and nonhydroxy fatty acid was negative. This is interpreted as indicating that the proper presentation of the binding epitope depends on the ceramide structure. The relevance of this to biological membranes has not yet been established. Neither free lactose (up to 20 mg/ml) nor lactose-bovine serum albumin (5 mg/ml) prevented the binding of bacteria to LacCer, two facts that support the solid-phase binding data demonstrating a low affinity binding and the crucial importance of a particular lactose epitope.

In microtiter wells, binding occurred at 50 ng and became half-maximal at the theoretical value for a monomolecular layer of LacCer (i.e. 100-200 rig/well). The bacteria also bound to glycolipids with various substitutions (e.g. GalNAcj31-4, Galal-3GalNAcj31-4, Fuccul-2Galj31-3GalNAcpl-4, Galal-3, GlcNAc@l-3, GalBl-3GlcNAcBl-3, Gal@-4GlcNAcBl-3, and GalBl-3(Fucal-4)  at the Gal of LacCer, although only those species with  were as active as LacCer itself. Glycolipids with other additions (e.g.  were negative. For unsubstituted LacCer, the binding required either a trihydroxy base or 2hydroxy fatty acid, specifying the epithelial type of ceramide; LacCer composed of a dihydroxy base and nonhydroxy fatty acid was negative. This is interpreted as indicating that the proper presentation of the binding epitope depends on the ceramide structure. The relevance of this to biological membranes has not yet been established. Neither free lactose (up to 20 mg/ml) nor lactose-bovine serum albumin (5 mg/ml) prevented the binding of bacteria to LacCer, two facts that support the solid-phase binding data demonstrating a low affinity binding and the crucial importance of a particular lactose epitope.
The carbohydrate residues of the animal cell surface glycolipids and glycoproteins appear to be involved in a variety of recognition phenomena (l-4), including the attachment of bacteria (2, 3) and viruses (4) and the interaction between cells during embryogenesis (1). Analysis of the capacity of ligands to interact with non-biological surfaces with bound carbohydrates is one way to characterize such carbohydratebased receptors (5). Recently, a method was developed which allows the testing of glycolipids resolved in thin layer chromatograms to mediate attachment of bacteria (6,7). Screening for bacterial receptors using this technique, revealed binding * This work was supported by Grant 3967 from the Swedish Medical Research Council.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
of diverse bacteria to lactosylceramide, LacCer (8,9). The aim of the present study was to describe in detail the binding of Propionibacterium granulosum, typical of skin habitats, to a large number of glycosphingolipids separated on thin layer chromatograms and coated in microtiter wells. One reason for selecting this bacterial species was the possibility of making a detailed comparison with a related species, Propionibacterium freudenreichii, which apparently recognizes a separate binding epitope on lactose (10). Furthermore, P. grunulosum seems to bind very similarly to other LacCer-recognizing bacteria characterized, including Neisseriu gonorrhoeae (11) and other genera (8,9).'

Preparation of Total Glycosphingolipid
Fractions'-Total neutral and acidic glycolipids from the sources given in Table I  to glycolipids separated on thin layer chromatograms was examined using a broad spectrum of reference glycolipids with different carbohydrate and lipophilic portions ( Fig. 1 and Tables I-II). As inferred from these data, the minimal binding requirement on the oligosaccharide was terminally or internally located lactose (GalPI-4Glc). We also found that this recognition was dependent on the ceramide structure (Table II).
When various mixtures of glycolipids were screened (Fig. l), P. granulosum was found to bind selectively to molecular species in the 2-to &sugar interval (e.g. nos 4a, 11, 13, 14, 16, 21a, 31a, 32a). Many glycolipids (i.e. nos. 8, 9, and 12), including monohexosylceramides (nos. 1 and 2), were negative even though present in amounts of 2 wg or more. As documented in Table I, the recognized glycolipids all contained Galpl-4Glc (lactose), and the most potent binder, being active down to about 50 ng, was LacCer with a heterogeneous ceramide composition (no. 4a). Glycolipids lacking the lactose sequence were negative (nos. l-3). Further arguments for lactose being the minimal binding requirement were the inactivity of Glcpl-Cer (no. 2) and various structures with a Galpl-terminus (nos. 1, 10, 16, 25, 30, and 34). We therefore interpret the weaker binding to the 3-, 4-, and 5-sugar glyco-    Lc4b dl&l-16:0/24:0 (+) Human erythrocytes n The structure of the carbohydrate portion of each glycolipid is given in Table I. b According to an earlier recommendation (61), d stands for dihydroxy long-chain base, t for trihydroxy longchain base, and h for 2-hydroxy fatty acid. Whereas the figures before a colon mean paraffin chain length, those after denote the number of double bounds. Varying components mean a mixture of di-and trihydroxy bases and hydroxy and nonhydroxy fatty acids.
' + indicates optimal binding (detection level of about 50 ng), (+) suboptimal binding (detection level of 500 ng or more), -no binding even at a level of 2 fig. See also Table I and text. d To illustrate the characteristic variation in the structure of the ceramide between different tissues (38), the source of each glycolipid is indicated. The structures were isolated and characterized as described in Table I  lipids carrying various sugar extensions on their common LacCer core structure (nos. 11, 13, 14, 16, 21a, 31a, and 32a) to mean that the bacteria also accommodate lactose in the internal position. However, the inactivity of the majority of the complex glycolipids analyzed in Fig. 1  One group of substituents at this position abolished the binding completely.
On the other hand, the GalNAcfll-4 substituent, specifying the ganglio series of glycolipids, allowed optimal binding activity, as found for Gg03 (no. 31a) and Gg04 (no. 32a), both of which were approximately as active as LacCer itself.
SSl to varying amounts of selected glycolipids adsorbed on microtiter wells.
Abbreviations to the right of each curve refer to the structures listed in the tables: 4a, LacCer with more hydroxylated ceramide; 46, LacCer with less hydroxylated ceramide; 8, Gb4a; 23, H-5-11. The binding assay was performed with Y-labeled bacteria, as described under "Materials and Methods." Data are the average value for triplicate determinations.
The results in Table I also reveal that further additions to tolerated proximal sequences may reduce the binding. Consider, for instance, Lc4a (no. 14) and Lc4b (no. 21a) as part of the type 1 and type 2 blood group ABH and Lewis glycolipids, respectively.
Except for , all tested compounds including Leb-6 were negative. Similarly, the addition of GalNAcPl-3 to Gb3b (no. 12), and Fucal-2 to Gg04 (no. 33), resulted in abolished and reduced binding activities, respectively. dihydroxy base and nonhydroxy fatty acid was inactive. This clear-cut dependence on the mode of hydroxylation of the ceramide was further examined using an expanded series of isolated variants of LacCer (Table II). Thus, a hydroxyl function in position 3 of the base (no. 4d), in position 2 of the fatty acid (no. 4c), or in both these positions simultaneously (no. 4e), was a requirement for activity in the case of free LacCer. Interestingly, however, binding to the more complex glycolipids required no particular structure of the ceramide: glycolipids of the ganglio and neolacto series were recognized regardless of the mode of hydroxylation of their ceramides (Table II). The binding of LacCer is restricted to molecular species Binding of P. granulosum to LacCer Coated in Microtiter typical of the epithelium, which have a trihydroxy base and/ Wells and Inhibition Experiments with Lactose and Lactoseor 2-hydroxy fatty acid (no. 4a, lane D in Fig. 1). LacCer of BSA- Fig.  2 shows the result of binding of 'Z51-labeled P. human erythrocytes (no. 4b, lane A in Fig. l), having a granulosum to selected glycolipids adsorbed in microtiter wells by guest on March 25, 2020 http://www.jbc.org/

Downloaded from
Binding of Bacteria to Glycolipids when measured as a function of the amount added to each well. The results obtained are consistent with the binding studies summarized above, both in view of carbohydrate specificity (compare nos. 4a and 23) and ceramide dependence (compare nos. 4a and 4b). In this analysis, binding was also detectable at a level of about 50 ng and became half-maximal close to 200 ng, the theoretical value of a monomolecular layer of glycolipids. When 3H-labeled glycolipid was used, 90% of LacCer was found to remain attached to the wells throughout the assay procedure (see "Material and Methods"). To put the binding data in perspective with a receptor glycolipid for bacteria reported earlier (33-35), we next examined the Galal-4Gal-specific binding of uropathogenic E. coli to Gb4a (8) in this assay: binding occurred then at lo-IOO-fold lower amounts of glycolipid (Fig. 2). Consequently, the two bacteria also differed with regard to the blocking ability of their corresponding free receptor disaccharides. Binding of P. granulosum to the thin layer chromatogram was not affected by preincubation of the bacteria with lactose up to 20 mg/ml or lactose-BSA (-25 mol/mol, 5 mg/ml). On the other hand, free GalLul-4Gal clearly prevented the binding of E. coli to glycolipids separated on thin layer chromatograms (33).
In the present study, the interaction between P. granulosum and glycolipids has been characterized (Figs l-2 and Tables  I-II). The specific importance of terminally or internally located lactose (Galpl-4Glc) with regard to this interaction is supported by the following. Of the numerous glycolipids tested (Table I), virtually only free LacCer (no. 4a) showed optimal binding, indicating an absolute specificity for lactose (compare nos. 1 and 2 with 4a). Since the other active glycolipids had different saccharide termini on their common lactose core, it seemed likely that the bacteria would accommodate lactose in the internal position. Evidence in support of this was the observation that GalNAcP and Galal-3Gal of structures nos. 32a,b and 11, respectively, showed no activity when found in the terminal position of other glycolipids (i.e. nos. 8,12,18, and 24). Our data do not, however, rule out the possibility of additional lectins with absolute specificity for the di-and trisaccharide termini of Gg03 (nos. 31a,b) or , respectively. This could possibly be resolved by binding studies using synthetic glycolipids carrying such termini but lacking the lactose backbone. In this respect, the finding (34) of many pulmonary pathogens that recognize the GalNAcpl-4Gal sequence of the ganglio series of glycolipids is noteworthy.
The maintenance of binding activity when LacCer was extended into the ganglio, lacto, neolacto, and isoglobo series of glycolipids therefore clearly indicates that the bacteria associate with internally located lactose residues, a feature previously documented for the Galcrl-4Gal-specific binding of uropathogenic E. coli (35) and the Shiga toxin (36). This observation is consistent with an adhesin-combining site restricted to the lactose disaccharide, since neither of these extended structures was superior to free LacCer. Specifically, whereas the extensions specifying the ganglio series of glyco-lipids nos. 31a,b and 32a,b,respectively) did not modify the activity of the lactose core, the other extensions either reduced (e.g. nos. 11,13,14,and Zla,b,nos. 7,6,and 5,respectively) the binding activity. This dual effect of substitutions in both C3 (e.g. Gala and NeuAca) and C4 (e.g. Gala and GalNAcP) probably arises in different preferred conformations, where steric effects make the binding epitope on lactose inaccessible.
In addition to sugar extensions on lactose, the ceramide structure modulated the activity, as first indicated when screening glycolipid mixtures containing free LacCer of different ceramide compositions (compare nos. 4a and 4b in Fig.  1). There was, however, no strict specificity for the ceramide since the single hydroxyl function could either be linked to the base or the fatty acid (Table II). The absence of this ceramide dependence in the case of glycolipids with various saccharide extensions on LacCer (Table II) is also consistent with an indirect involvement of the lipophilic portion. The previously noted (10) preferential binding of a strain of P. freudenreichii to less hydroxylated LacCer species (i.e. nos. 4b and 4f) provides additional support to a modulatory effect of the ceramide on the activity of the lactose head group. As discussed in this previous study (lo), it is reasonable to assume that the ceramide structure determines different lactose head group orientations, exposing the actual binding epitope differently as the result of effects either directly from the ceramide or indirectly from the assay surface. Modulatory effects from ceramide hydroxylation mode have been demonstrated by means of glycolipid-directed monoclonal antibodies in intact cells as well (37), and we therefore anticipate the preferences for different types of LacCer to be of biological relevance. Significantly, the ceramide varies in a characteristic way between different tissues (Table II and (Table II), like most other lactose-binding bacteria (8, 9, ll),' bind preferentially to LacCer with the epithelial, more hydroxylated, type of ceramide, thus correlating with the most common site of bacterial colonization.
The necessity of a polyvalent presentation or clustering of lactose residues was first indicated by the relatively large amounts of glycolipid needed for optimal binding in microtiter wells (i.e. 100-200 rig/well, Fig. 2). Free univalent lactose at a very high concentration (20 mg/ml) was incapable of blocking this interaction. In analogy with this situation, other microbial carbohydrate recognition systems, such as sialosyloligosaccharide binding by influenza virus (39) and Galal-Gal binding by the Shiga toxin (36), are low affinity interactions and are not blocked by free saccharides. The finding that lactose-BSA (5 mg/ml, -25 mol/mol), although polyvalent, was not able to inhibit the binding of P. granulosum may be due to insufficient complementarity to the adhesin pattern on the bacterium. Alternatively, the specific ceramide-dependent binding epitope on lactose proposed above may not be accessible in this conjugate. Furthermore, we have not yet excluded the possibility that the Cl-C2 portion of the ceramide may be required for binding.
The present data, as well as earlier reports (6-11, 31, 33, 35), demonstrate the usefulness of glycolipids immobilized in thin layer chromatograms to probe the fine specificity of microbial-carbohydrate interactions. With regard to the recognition of Galcul-4Gal by E. coli and the Shiga toxin, data generated by the thin layer assay are in good general agreement with those obtained in traditional sugar inhibition experiments (33, 36, 40). It remains, however, to be demon-Binding of Bacteria to Clycolipids strated whether the assay surface conditions are fully relevant to the situation in the plasma membrane. In LacCer, the active saccharide is directly linked to the ceramide and, consequently, the ceramide may be expected to affect receptor function, either directly (by influences on conformation) or indirectly (by interacting with the plastic assay surface). Nevertheless, the sophisticated dependence on the structure of the ceramide (Table II), correlating with a distinct ceramide variation in potential target tissues, is provocative and should be subjected to further studies using model membranes and conformation methods. Slight variations in assay conditions may explain incidental negative binding for LacCer despite unaltered activities of the more complex ceramide-independent glycolipids, as discussed earlier (7). Complete lack of activity of some batches of bacteria may be due to lack of expression of the adhesin. Our collaborative efforts with molecular geneticists on N. gonorrhoeae, which also recognizes LacCer, may shed light on this problem (11).
Most of the bacteria hitherto classified as LacCer binders (8, 9, 4l)l display glycolipid-binding patterns very similar to those reported here for P. grunulosum.
On the other hand, our recent studies on N. gonorrhoeae (11) and P. freudenreichii (10) raise the possibility of a variety of LacCer-binding adhesins that differ in their detailed binding specificities. This may hypothetically correspond to only slight changes in the amino acid sequence of the adhesin-binding sites, analogous to receptor variants of influenza virus (42). The finding that several bacteria have selected a binding to LacCer may indicate an important factor in the colonization of mucosal surfaces. In the case of N. gonorrhoeae relevant glycolipids exist in target cells for infection (11). Several of the bacteria in question belong to the normal bacterial flora of the large intestine. It is therefore of interest that recent studies have identified LacCer of the more hydroxylated type in isolated epithelial cells of the human colon (43) but not in those of the human small intestine (44). Furthermore, both in the rat (45) and humans (46) the epithelial type of LacCer also exists in relatively large amounts in feces as a result of bacterial degradation of more complex glycolipids of extruded cells. However, at the membrane, in contrast to mannose, another common receptor for bacteria (2, 3), LacCer is considered cryptic on normal cells and not directly accessible from the outside (30,47,48). This and the fact that lactose is absent in glycoproteins (49) may suggest that LacCer is used to establish a firm attachment after initial binding to other sites followed by a more intimate collision with the membrane after lateral diffusion of masking surface components. Clearly, lactose is a common nutrient for bacteria, but ability to transport or ferment lactose (50) does not correlate with binding ability to LacCer in our assay. Therefore, it is unlikely that the binding detected is based on a surface-located transport site for lactose.
The biological relevance of the present findings from in uitro assays has to be further tested. A possible continuation is to identify the adhesin by genetic cloning studies, as in the current work with N. gonorrhoeae (11). Synthesis of soluble receptor analogues appears to be necessary before adequate experiments can be performed on the inhibition of bacterial binding to target cells.