Versatile biosynthetic engineering of sialic acid in living cells using synthetic sialic acid analogues.

Sialic acids are critical components of many glycoconjugates involved in biologically important ligand-receptor interactions. Quantitative and structural variations of sialic acid residues can profoundly affect specific cell-cell, pathogen-cell, or drug-cell interactions, but manipulation of sialic acids in mammalian cells has been technically limited. We describe the finding of a previously unrecognized and efficient uptake and incorporation of sialic acid analogues in mammalian cells. We added 16 synthetic sialic acid analogues carrying distinct C-1, C-5, or C-9 substitutions individually to cell cultures of which 10 were readily taken up and incorporated. Uptake of C-5- and C-9-substituted sialic acids resulted in the structural modification of up to 95% of sialic acids on the cell surface. Functionally, binding of murine sialic acid-binding immunoglobulin-like lectin-2 (Siglec-2, CD22) to cells increased after N-glycolylneuraminic acid treatment, whereas 9-iodo-N-acetylneuraminic acid abolished binding. Furthermore, susceptibility to infection by the B-lymphotropic papovavirus via a sialylated receptor was markedly enhanced following pretreatment of host cells with selected sialic acid analogues including 9-iodo-N-acetylneuraminic acid. This novel experimental strategy allows for an efficient biosynthetic engineering of surface sialylation in living cells. It is versatile, extending the repertoire of modification sites at least to C-9 and enables detailed structure-function studies of sialic acid-dependent ligand-receptor interactions in their native context.

Sialic acids are the most prevalent terminal monosaccharides on the surface of eukaryotic cells (1), and they perform important functions in biological recognition phenomena, in-cluding cell-cell interactions (2,3) and binding of toxins, viruses, bacteria, and parasites to their cellular receptors (4,5). Similarly, sialic acids can be key determinants for the stability and biologic activity of hormones or enzymes in vivo (6). Over 30 different naturally occurring members of this family of 9-carbon amino sugars have been identified (5). In all mammals, including great apes, the most abundant sialic acids are NeuAc and N-glycolylneuraminic acid (NeuGc). 1 Interestingly, humans are an exception because they do not express NeuGc due to a deletion in the CMP-NeuAc hydroxylase gene (7,8). It has been suggested that the absence of NeuGc residues and the corresponding increase in the relative abundance of NeuAc residues in humans may profoundly alter biological processes as some adhesion molecules recognize glycoconjugates containing NeuGc and NeuAc with different affinities (9).
Furthermore, sialic acid species can determine the host range of microbial pathogens including influenza A (10) or Escherichia coli K99 (11). In pigs, these enterotoxic bacteria with K99 fimbriae employ NeuGc-GM3 as a cellular receptor, whereas NeuAc-GM3 was shown to be non-functional. The expression of N-glycolyl groups seems to be developmentally regulated in pig mucosa and may explain the resistance of adult pigs to infection with this pathogen (11). From a therapeutic perspective, a group of sialic acid analogues, which were rationally designed as high affinity inhibitors of influenza A and B virus neuraminidase (12), has recently been introduced successfully as an anti-flu medication in humans (13,14).
Many human oncofetal antigens carry sialic acid side chain modifications such as 9-O-acetylation and N-glycolyl modifications, although the source of the NeuGc remains to be identified. For example, the unusual gangliosides O-acetyl-GD3 and N-glycolyl-GM3 and the mammary serum antigen were found in breast tumors (15,16) and GM2 containing NeuGc in human colon cancers (17). In contrast, O-acetylation of sialyl Lewis X antigen decreases from normal colonic mucosa to primary carcinomas and their liver metastases (18,19). Despite the diagnostic and therapeutic potential of sialic acid analogues, detailed structure-function analyses have largely been confined to in vitro binding studies or competitive inhibition assays. The effect of synthetic sialic acid analogues as part of sialoglycoconjugate receptors in living cells has been severely limited due to low incorporation efficiencies using exogenous transfer methods (20 -22).
As a partial solution to this technical problem, we and others (23)(24)(25)(26)(27)(28)(29) have demonstrated that biosynthetic modification of the N-acyl group of cellular sialic acids can be achieved through administration of synthetic sialic acid precursor analogues both in vitro and in vivo. These unphysiological D-mannosamines are taken up, metabolized, and incorporated into cell surface sialoglycoconjugates of mammalian cells. So far this approach has only been successfully applied to sialic acids carrying modifications in the C-5 group.
It is a widely accepted view that sialic acids cannot be efficiently taken up from the extracellular space by eukaryotic cells (30). Contrary to this, we have recently provided evidence for the existence of an efficient uptake mechanism for NeuAc in mammalian cells (31). NeuAc medium supplementation rapidly and potently compensated for endogenous hyposialylation in two human hematopoietic cell lines that are deficient in de novo sialic acid biosynthesis (32). Importantly, this still uncharacterized molecular uptake of NeuAc was active in all cell types tested, including primary human cells, regardless of their prior sialylation status. Based on these findings, we also hypothesized that structural analogues of NeuAc may be taken up and incorporated into cell surface glycoconjugates.
In the current study, we demonstrate that synthetic sialic acid analogues carrying C-5 or C-9 side chain modifications of different sizes and chemical properties can be efficiently taken up from the culture medium and incorporated into cellular glycoconjugates of human cells. As a model application, we showed that infection by a primate polyomavirus that depends on a sialylated receptor can be positively affected by pretreatment of host cells with selected sialic acid analogues, but we observed differential effects of metabolically incorporated NeuGc and 9-iodo-NeuAc on CD22 binding. We thus demonstrated that this approach has the potential to facilitate submolecular analyses of sialic acid-dependent ligand-receptor interactions in their native context.
For serum-free conditions cells were cultivated without fetal calf serum but with addition of Nutridoma-HU at the concentration suggested by the manufacturer (Roche Molecular Biochemicals) for at least 7 days prior to an experiment.
In sugar supplementation assays, cells seeded at 5 ϫ 10 5 or 3 ϫ 10 5 cells/ml in culture medium buffered additionally with 40 mM HEPES, pH 7.2, were cultivated for 24 or 48 h, respectively, in the presence or absence of the respective sugar at 5 mM (stocks of 100 mM dissolved in H 2 O and stored at 4°C).
For HPLC analysis (see below) serum-starved HL60-I cells were incubated with either 9-iodo-NeuAc or 5-N-fluoroac-Neu (each 5 mM) for 48 h in the presence of 40 mM HEPES, pH 7.2. To control for contamination of HPLC samples by input of sialic acid analogue, an equimolar concentration of the non-incubated analogue, 5-N-fluoroac-Neu for 9-iodo-NeuAc-treated cells and 9-iodo-NeuAc for 5-N-fluoroac-Neutreated cells, was added directly before harvesting cells for HPLC analysis.
Flow Cytometry-Lyophilized FITC-conjugated lectin Vicia villosa agglutinin (VVA) was obtained from Sigma, dissolved in H 2 O (1 mg/ml), aliquoted, and stored at Ϫ20°C according to the manufacturer's instructions. FITC-conjugated Limax flavus agglutinin (LFA) was from EY Laboratories (San Manteo, CA), and biotin-coupled Tritrichomonas mobilensis lectin (TML) was from Calbiochem. Lectin staining procedure and fluorescence-activated cell scanning on a Becton Dickinson FACScan cytometer using Cellquest II software were carried out as described previously (35,41). Briefly, HL60-I cells (3 ϫ 10 5 ) were washed twice in cold PBS and then incubated in 100 l of PBS, 0.05% NaN 3 , containing either fluorochrome-conjugated lectins VVA (50 g/ ml) or LFA (20 g/ml), or biotinylated TML (20 g/ml) for 45 min on ice in the dark. For biotinylated TML a secondary incubation step with streptavidin-FITC (20 g/ml) (Sigma) for 30 min on ice in the dark was performed. After washing with PBS, cells were resuspended in 300 l of PBS and analyzed by flow cytometry.
Detection of Sialic Acid Analogues in Cellular Glycoproteins by HPLC-Serum-starved BJA-B K20 cells (1 ϫ 10 7 ) were washed twice with PBS, frozen at Ϫ20°C, lysed by hypotonic shock in distilled water, and sonicated (5 min, 4°C). The crude membrane fraction was pelleted by centrifugation at 10,000 ϫ g for 15 min, and the pellet was washed twice with distilled water. The lyophilized pellet was delipidated by chloroform/methanol (2:1, 1:1, and 1:2). After centrifugation at 10,000 ϫ g (15 min, 4°C), the lyophilized pellet was hydrolyzed for 1 h with 200 l of 2 M acetic acid at 80°C (45), and the acetic acid was evaporated after separation. After washing and resuspending in H 2 O, the sample was further purified using a cation-exchange column (AG-50W-X12, H ϩ -form, 100 -200 mesh) (Bio-Rad), lyophilized, and resuspended in 100 l of H 2 O. Sialic acid derivatization was performed according to a method of Hara et al. (46), and samples were analyzed on a reversed phase C18 column as described (24). The relative content of sialic acid analogue and NeuAc was quantified by integration of peak areas.
LPV Infection-After sugar pretreatment for 3 days, BJA-B subclones were infected with the B-lymphotropic papovavirus (LPV) as described previously (31,34). LPV infection was quantified both by indirect immunofluorescence microscopy as percentage of LPV T-antigen-positive cells and by quantification of the concentration of LPV VP1 in cell lysates by enzyme-linked immunosorbent assay (34).

Effects of Synthetic Sialic Acid Analogues on Binding of
Various Sialic Acid-sensitive Lectins-To test whether cells can incorporate synthetic sialic acid analogues following addition to the culture medium, we first performed a metabolic complementation assay on endogenously hyposialylated, UDP-GlcNAc 2-epimerase-deficient HL60-I cells, in principle as described previously for ManNAc (32) and NeuAc (31). UDP-GlcNAc 2-epimerase-deficient cells are ideal for such studies because the absence of this key enzyme of de novo sialic acid biosynthesis results in a severely hyposialylated phenotype of membrane glycoconjugates. This allows for the sensitive and rapid detection of changes in cellular sialylation. HL60-I cells were cultivated in the presence of 1 of 16 different synthetic sialic acid analogues, carrying distinct C-1, C-5, or C-9 modifications (Fig. 1), and subsequently, changes in cell surface sialylation were assessed by flow cytometry using three labeled sialic acid-sensitive lectins. V. villosa agglutinin (VVA) detects Gal-NAc residues (48), and its binding sites can be masked by terminal sialic acid residues, regardless of side chain modifications the latter may carry. In contrast, L. flavus agglutinin (LFA) and T. mobilensis lectin (TML) both detect terminal sialic acid residues glycosidically linked to either Gal or Gal-NAc (49,50). Importantly, in vitro studies have revealed markedly distinct binding affinities of LFA and TML for different synthetic sialic acid analogues indicating that specific side chain modifications may be recognized differentially by these two sialic acid-binding lectins (49,50). This flow cytometrybased analysis is a simple and quantitative screening method for an increase in cell surface sialylation following exposure to sialic acid analogues. Furthermore, it may provide initial evidence for the incorporation of specific sialic acid analogues into cell surface glycoconjugates.
Medium supplementation with physiological NeuAc, which served as a positive control, increased binding of LFA and TML to HL60-I cells 9.5-and 4.8-fold, respectively (Fig. 2). Concomitantly, binding of VVA in NeuAc-treated cells was decreased by 40% compared with untreated controls, confirming an increased masking of penultimate GalNAc residues by metabolically incorporated NeuAc.
Of the 16 synthetic sialic acid analogues tested, 10 decreased VVA-binding to HL60-I cells by more than 25% ( Fig. 1 and Fig.  2; analogues a, f-k, and n-p) after 24 h, and these compounds also increased binding of LFA and of TML although to varying extent.
Overall, the degree of effects on lectin binding induced by synthetic analogues was similar to or less pronounced than that of NeuAc with the exception of 9-iodo-NeuAc (analogue f), which drastically increased binding of both LFA (19.2-fold) and TML (6.4-fold). Several other sialic acid analogues are also noteworthy.
First, 5-N-trifluoroac-Neu (analogue k) had a very pronounced effect on VVA binding, indicating a substantial increase in cell surface sialylation, yet LFA and TML binding were hardly affected. A similar, though less pronounced, staining pattern was seen for 5-N-thioac-Neu (analogue n). These findings were in clear contrast to effects seen for NeuAc, which considerably enhanced binding of both sialic acid-binding lectins. The weak LFA and TML binding might reflect a low affinity of lectins to 5-N-trifluoroac-Neu or 5-N-thioac-Neu. This is in line with in vitro studies (49 -51) that have shown that modifications of the C-5 residue can affect LFA as well as TLM recognition.
Second, differential effects for LFA and TML binding were observed in cells pretreated with either 9-thio-NeuAc (analogue g) or 9-SCH 3 -NeuAc (analogue h), whereas both analogues induced a comparable reduction in VVA binding. 9-Thio-NeuAc resulted in a stronger increase for LFA binding than in TML binding, whereas the opposite was true for 9-SCH 3 -NeuAc (Fig. 2).
Third, the two C-1 analogues employed in this study, NeuAc-Me-ester (analogue o) and NeuAc-Et-ester (analogue p), induced effects on lectin binding very similar to those seen for NeuAc at equimolar concentrations. Presumably, NeuAc is restored from these esters by intracellular esterases (52,53), and unmodified NeuAc is incorporated into glycoconjugates. Esterification of the carboxyl group does not appear to enhance sialic acid incorporation.
Fourth, of the six compounds that had weak or no effect on VVA binding, five carry an additional charge, three of them a positive charge (analogues b, d, and l) and two a negative charge (analogues e and m). Furthermore, the glycyl-(analogues d and l), succinyl-(analogues e and m) and also the uncharged acetamido residue (analogue c) can be considered as bulky substitutions. Thus, steric hindrance or ionic interaction is conceivable at any step in the uptake and metabolic pathway of these analogues. Although in vitro data on enzymatic activation and transfer are not complete, it has been observed that at least activation of 9-amino-NeuAc (54, 55), 5-N-Gly-Neu (56), 2 9-N-Gly-NeuAc, 5-N-Succ-Neu, and 9-N-Succ-NeuAc 2 is reduced, whereas the acetamino residue NeuAc (54,55) has no effect on both steps.
Taken together, these data demonstrate that medium supplementation with sialic acid analogues carrying distinct C-5 or C-9 substitutions can profoundly affect cell surface sialylation in a human cell line. Differential effects on the three sialic acid-sensitive lectins employed provided initial evidence for the incorporation of the synthetic C-5 and C-9 analogues of sialic acid into cellular glycoconjugates.
Incorporation of Specific C-5-and C-9-modified Sialic Acid Analogues in Cellular Glycoconjugates-As a direct and definitive way of assessing the incorporation of synthetic sialic acid analogues into surface glycoconjugates, we analyzed the membrane-associated sialic acid fraction of cells cultivated in the presence of a C-9 or a C-5 analogue by HPLC. In this study we employed endogenously hyposialylated BJA-B K20 cells, a second recently identified UDP-GlcNAc 2-epimerase-deficient human cell line (32,35). BJA-B K20 cells that had been kept under serum-free conditions to maximally deplete their residual sialic acid pool (31,32) were cultivated in the presence of either 9-iodo-NeuAc or 5-N-fluoroac-Neu. After 2 days the cells were extensively washed and then processed for HPLC analysis. Both synthetic sialic acid analogues were unambiguously identified in the glycoprotein-associated sialic acid fraction (Fig. 3, C and D) and were eluted as discrete peaks. The peaks coincided with those of authentic 9-iodo-NeuAc and 5-N-fluoroac-Neu standards (Fig. 3A), respectively. The relative retention coefficients of standard and sample-derived sialic acids differed by less than 0.3%. In contrast, the specific peaks were not seen in chromatograms of untreated cells (Fig. 3B). In K20 cells integration of peak areas of sialic acids in HPLC analysis revealed that relative amount of 95 and 92% of the total membrane-associated sialic acid fractions in K20 cells consisted of 5-N-fluoroac-Neu and 9-iodo-NeuAc, respectively (data not shown).
We also addressed whether contamination by input sialic acid analogues during the purification procedure could account for these results. To test this, one of the two sialic acid analogues, 9-iodo-NeuAc or 5-N-fluoroac-Neu, was added for 48 h, whereas the other substance was added at an equimolar concentration immediately prior to cell harvesting. Neither of the shortly added sialic acid analogues was detectable in the sialic acid fraction from membrane-bound glycoproteins in cells treated for 2 days with the other analogue (Fig. 3). These results validate our experimental procedures and confirm that both sialic acid analogues were efficiently incorporated into cellular glycoconjugates.
A C-9-modified Sialic Acid Analogue, but Not Its C-6-modified Mannosamines Precursor, Restores Sialylation in a Metabolic Complementation Assay-C-5-modified sialic acids can also be biosynthetically incorporated by modified sialic acid 2 R. Brossmer and R. Isecke, unpublished data. precursors such as D-mannosamines. In contrast, C-9-modified sialic acids probably can only be generated from sialic acid analogues but not from precursors, because the corresponding C-6 modification is expected to interfere with the C-6 phosphorylation of the precursor, a necessary step in sialic acid biosynthesis. We tested this hypothesis comparing cellular sialylation after addition of 9-iodo-NeuAc and 6-iodo-ManNAc. The de novo expression of CD75s, a highly sensitive marker of sialylation in BJA-B K20 cells (31,32,35), was followed in serumstarved BJA-B K20 cells using a monoclonal anti-CD75s antibody that recognizes a sialic acid-dependent epitope. The intracellular signal for CD75s was undetectable in serumstarved BJA-B K20 cells at the time of addition of the amino sugars to the culture medium (data not shown), but treatment with the positive controls NeuAc and ManNAc resulted in a restoration of the CD75s signal (Fig. 4). While 9-iodo-NeuAc induced a strong intracellular CD75s signal within 1 h (Fig. 4), 6-iodo-ManNAc had no effect even after 6 h of treatment. This confirms that only sialic acid analogues but not the precursors can biosynthetically introduce C-9-modified sialic acids into cellular glycoconjugates. Furthermore, NeuAc and 9-iodo-NeuAc induced the CD75s signal faster than ManNAc. These experiments also exclude that 9-iodo-NeuAc incorporation was mediated by degradation and uptake of its ManNAc precursor and further highlight the broad spectrum of sialic acid modifications obtainable in living cells with sialic acid analogues, as opposed to D-mannosamine analogues.
In parallel, we studied the effect of these sialic acid analogues on LPV susceptibility in the UDP-GlcNAc 2-epimeraseexpressing BJA-B subclone K88 (32,35). Although NeuAc was unable to enhance LPV susceptibility in these normally sialylated cells, 9-iodo-NeuAc and 5-N-fluoroac-Neu treatment significantly enhanced the susceptibility 2.3-and 1.7-fold, respectively (Fig. 5A, bottom panel). Because the only step in the viral life cycle known to be sialic acid-dependent is at the level of receptor binding and cellular entry, we speculate that sialylated virus receptors carrying 9-iodo-NeuAc and 5-N-fluorac-Neu displayed an increased affinity for the interaction with LPV resulting in a higher percentage of cells becoming infected and possibly also allowing for more infections to occur per cell.
Differential Effects of Sialic Acid Analogues on CD22 Binding-Finally, we used the sialic acid-dependent binding of CD22 as a second model to demonstrate the experimental potential of metabolically incorporated sialic acids in functionally important cellular glycoconjugates applying NeuGc and 9-iodo-NeuAc to BJA-B cells. CD22 (Siglec-2) is a member of the sialic acid-binding immunoglobulin-like lectin (Siglec) family and functions as a B-lymphocyte-specific adhesion and signaling molecule (58). Binding of recombinant soluble CD22 was determined by flow cytometry (Fig. 6). Murine CD22 has been found to bind with higher affinity to specific glycans terminating in NeuGc than in NeuAc, whereas human CD22 apparently does not discriminate these sialic acid variants (59). Although BJA-B cells are of human origin, they probably carry some NeuGc on their surface as they were cultured in 10% fetal calf serum (60). In hyposialylated K20 cells NeuGc pretreatment enhanced murine CD22 binding stronger than NeuAc, thus confirming its reported binding specificity. Interestingly, murine CD22 binding to 9-iodo-NeuAc-treated K20 cells was even lower than to untreated control cells. In normally sialylated K88 cells NeuGc induced stronger binding of murine CD22, whereas NeuAc had no apparent effect. Here again, 9-iodo-NeuAc strongly reduced binding of murine CD22. As expected, binding of human CD22 to K20 cells increased to a similar extent after NeuGc and NeuAc treatment and showed no difference with K88 in comparison to untreated cells (data not shown). Interestingly, 9-iodo-NeuAc also reduced binding of the human form of CD22 (data not shown).
Thus, even these studies with only a limited panel of sialic acid analogues identified residues that are important for both model ligands analyzed. Functional experiments like these may facilitate the mapping of a variety of sialic acid-dependent ligand-receptor interactions in their native cellular context.

DISCUSSION
Quantitative and structural variations of sialic acid residues can profoundly affect specific cell-cell, pathogen-cell, or drugcell interactions, but experimental manipulation of sialic acids in mammalian cells has been technically limited. A first advance was made by biosynthetically modifying cellular sialic acids through synthetic sialic acid precursor analogues, N-acyl-D-mannosamines or D-glucosamines (for review see Ref. 29). This experimental approach, however, provides only a partial solution. The precursors consist of only 6 carbon atoms (C-4 to C-9 of sialic acid), and processing to sialic acids requires additional enzymatic steps. So far, only modifications of the C-5 position of sialic acid have been successfully introduced via precursor analogues (23)(24)(25)(26)(27)(28)(29). Recently, we provided evidence (31) for efficient cellular uptake and incorporation of NeuAc in mammalian cells and hypothesized that also synthetic sialic acid analogues may be taken up by this molecularly still undefined pathway and may also be incorporated into cell surface glycoconjugates. The metabolic incorporation of sialic acid analogues might overcome some of the limitations of the precur-sor analogue approach.
In the current study, we demonstrate that C-5-and C-9substituted analogues of NeuAc added to the culture medium can be efficiently incorporated into cell surface glycoconjugates. Of 16 analogues tested, at least 10 had specific effects on cellular sialylation as monitored by binding of sialic acid-sensitive lectins. Most of these 10 compounds reduced VVA binding to a similar extent, which indicated a comparable increase in cell surface sialylation. In contrast, their effects on LFA and TML binding were different. This allowed us to subdivide the analogues into three groups as follows: first, analogues that enhanced both LFA and TML binding to a similar extent (analogues a, f, g, i, o, and p); second, analogues that had a more pronounced effect on either LFA or TML binding (analogues h and j); and third, analogues that failed to markedly enhance LFA and TML binding (analogues c, k, and n). Thus, the heterogeneous effects observed with these sialic acid-binding lectins provided initial evidence that structurally modified sialic acids were being presented on the cell surface.
For one C-5 analogue (5-N-fluoroac-Neu) and one C-9 analogue (9-iodo-NeuAc) we unambiguously demonstrated their presence in the sialic acid fraction from membrane glycoproteins. HPLC analysis identified the sialic acid analogue in glycosidic linkage in membrane sialoglycoproteins as it had been added to the culture medium 2 days earlier. Remarkably, up to 95% of the total, membrane-associated sialic acid fraction in pretreated serum-starved BJA-B K20 cells consisted of the modified sialic acid. This level of incorporated modified sialic acids is similar to those described recently for N-acyl-D-mannosamine analogues in the same cell system (61).
The endogenously hyposialylated cells employed in most of the current experiments have a Ͼ60% (BJA-B K20) to Ͼ90% (HL60-I) reduction in cell surface sialylation compared with subclones from the same cell lines with an intact sialic acid biosynthesis (32), and cultivation under serum-free conditions can further deplete this residual sialylation. Thus, UDP-Glc-NAc 2-epimerase-deficient human cell lines are an ideal experimental system to introduce high levels of sialic acid analogues into surfaces glycoconjugates, because competition with endogenous NeuAc is minimized. In addition, the generation of stable transfectants in BJA-B cells is relatively easy (32), enabling the expression of desired recombinant sialoglycoproteins in this context. Importantly, the uptake of physiological NeuAc and sialic acid analogues is not restricted to hyposialylated cells. Here we provide evidence that normally sialylated BJA-B K88 cells incorporate 9-iodo-NeuAc, 5-N-fluoroac-Neu, and NeuGc. We had shown previously that physiological NeuAc is taken up by a variety of cell types, including primary cells, regardless of the prior sialylation status of the cells (31). Consequently, it is conceivable that sialic acid analogues may also be incorporated into different cell types, expanding the potential applications of this experimental strategy.
One key advantage of sialic acid analogues over sialic acid precursor analogues for the metabolic engineering of cellular sialic acids is their independence from the multistep biosynthetic pathway. Modifications of sialic acids at position C-9, and theoretically C-1, C-3, C-4, or C-8, can only be metabolically introduced by the respective sialic acid analogues but not by their D-mannosamine precursors. Although a number of N-acyl-modifications of D-mannosamines appear to be well tolerated (32), modifications at sites other than C-2 might interfere with subsequent enzymatic steps of the sialic acid biosynthesis. For example, in the current study, 6-iodo-ManNAc was unable to restore sialylation in BJA-B K20 cells, whereas 9-iodo-NeuAc was rapidly and potently incorporated into sia- loglycoconjugates. The block in metabolism of 6-iodo-ManNAc likely occurred at the essential phosphorylation step at the position C-6. We further conclude that the cytosolic NeuAc aldolase, which at least in vitro is able to synthesize NeuAc from ManNAc and pyruvate (62), cannot provide an alternative pathway to process 6-iodo-ManNAc in living cells, and the NeuAc-9-phosphate synthase (63) cannot use 6-iodo-ManNAc as an alternative substrate.
The range of sialic acid analogues that could possibly be metabolically incorporated is limited by the fact that they have to be substrates for the cellular sialic acid uptake pathway, the nuclear CMP-NeuAc synthase, the CMP-sialic acid transporter in the Golgi membrane, and the sialyltransferases. Indeed we found that analogues carrying either a glycine or succinyl residue in position C-5 or C-9, or amino or acetamino groups in position C-9, had no significant effects on cellular sialylation after treatment for 24 h. Thus, despite the apparently high degree of promiscuity of this pathway for uptake, metabolization, and incorporation of sialic acid analogues into membrane glycoconjugates, it still excluded certain chemical modifications.
As models for the experimental potential to biosynthetically engineer sialic acids in functionally important cellular glycoconjugates, we tested the effect of three synthetic sialic acid analogues (5-N-fluoroac-Neu, 9-iodo-NeuAc, and 9-deoxy-NeuAc) on the sialylated receptor of the primate polyomavirus LPV and of two analogues (NeuGc and 9-iodo-NeuAc) on binding of CD22. Pretreatment with all three sialic acid analogues drastically enhanced LPV susceptibility in low susceptible BJA-B K20 cells. Somewhat unexpectedly, 9-iodo-NeuAc and 5-N-fluoroac-Neu significantly increased LPV susceptibility even in highly susceptible, normally sialylated BJA-B K88 cells, in which physiological NeuAc had no effect. We attribute this to the synthesis of sialylated LPV receptors carrying these specific side chain modifications. Interestingly, this is the first time that sialic acid modifications have been identified that enhance LPV infection; N-propanoyl, N-butanoyl, and N-pentanoyl sialic acids, metabolically introduced by application of corresponding N-acyl D-mannosamine precursors, all reduced LPV susceptibility in BJA-B K88 cells by 90% (24). The sialic acid modifications analyzed so far for their influence on LPV binding are not sufficient to allow structural predictions on the sialic acid-LPV interaction. However, specific features of the two sialic acid analogues that enhance LPV infection might play a role; the iodine residue is similar in size but more hydrophobic than the hydroxyl group that it substitutes at C-9 and the fluorine at C-5 is highly electronegative.
In contrast to the positive effects of 9-iodo-NeuAc on LPV infection, the same analogue proved deleterious to binding of murine and human CD22. This observation is in line with findings of Kelm et al. (59) that the hydroxyl group at C-9 is essential for sialic acid recognition by CD22. Furthermore, we showed that in cells treated with free NeuGc murine CD22 binding increased stronger than after NeuAc treatment which reflects the reported binding specificity of murine CD22. This result also demonstrates that at least in tissue culture free NeuGc can be taken up and incorporated by human cells. It has been suggested that NeuGc found in human cells in the absence of endogenous biosynthesis might originate from alimentary or intestinal microbial sources (64,65). Little is known about the mechanism of sialic acid resorption, but our findings suggest that also in vivo free NeuGc, once present in the extracellular space, could be taken up directly by cells.
Applying biosynthetic sialic acid modification as described here might complement molecular modeling studies. For some sialic acid-dependent ligand-receptor interactions, the three-dimensional structure has been solved (66 -70) enabling the use of molecular modeling approaches for the design of potent inhibitors. The phenotype of sialic acid analogues predicted by such modeling studies could then be experimentally tested in the appropriate cellular context using metabolically incorporated analogues.
In conclusion, the novel experimental strategy described here allows us to generate membrane-bound, and possibly also secreted, natural and recombinant glycoconjugates carrying modified sialic acids. Functionally important sialic acids with modifications at positions inaccessible for precursor analogues may now be metabolically incorporated into living cells. This will aid the generation of glycoconjugates with desired biological and chemical characteristics and functions.