Cross-reactive carbohydrate determinants of plants are essentially a mixture of N-glycans containing β1,2-xylose and core α1,3-fucose, the latter also found in insect glycoproteins. To determine the relative contributions of these two sugar residues to antibody binding, we prepared an array of glycomodified forms of human apo-transferrin. Using core-α1, 3-fucosyltransferase (EC 184.108.40.206) and β1,2-xylosyltransferase (EC 220.127.116.11) recombinantly expressed in Pichia pastoris and suitable glycosidases, glycoforms containing either only fucose (MMF), only xylose (MMX), both (MMXF), or neither (MM) linked to the common pentasaccharide core were generated. Additional glycoforms were obtained by enzymatic removal of the α1,3-linked mannosyl residue. These transferrin glycoforms served to define the binding specificity of antibodies in western blot, ELISA, and inhibition ELISA. Rabbit anti–horseradish peroxidase serum bound to both the fucosylated (MMF) and the xylosylated (MMX) glycoforms. Inhibition studies indicated two independent highly specific populations reacting with either of the two epitopes. In contrast, the monoclonal antibody YZ1/2.23 appears to recognize a larger structure including both the fucosyl and the xylosyl residue. The mannose-deficient glycoform was a poorer inhibitor for both antibodies. Terminal GlcNAc residues prevented antibody binding. Rabbit anti-bee venom serum reacted with fucosylated forms (MMF and MMXF) only. Experiments with sera from allergic patients suggest that glycomodified human transferrin, especially the MMXF glycoform, is a suitable reagent for the detection of antibodies against cross-reactive carbohydrate determinants. Within the panel studied, several sera contained high levels of fucose-reactive IgE but only a few sera showed any binding to MMX-transferrin.
With regard to glycosylation, plants may appear a fairly uninventive group because their glycoproteins carry the same limited set of structures almost regardless of the species (Lerouge et al., 1998; Wilson et al., 2001). However, just because of this and because some of the structural details are foreign to (and hence immunogenic in) mammals, plant N-glycans probably constitute the most frequently occuring set of epitopes to which we are exposed (Wilson et al., 1998). The vast potential for cross-reaction of antibodies with carbohydrate determinants is further extended by the occurrence of the same epitope structures on glycoproteins of insects, molluscs, and parasitic worms (Altmann et al., 1999; Lommerse et al., 1997; Wilson et al., 1998; Wilson, 2001). In detail, plant N-glycans carry a fucose residue in α1,3-linkage to the innermost GlcNAc (Ishihara et al., 1979; Lerouge et al., 1998), and this immunogenic feature also occurs on insect glycoproteins (Altmann et al., 1999; Kubelka et al., 1993) and even on N-glycans from Schistosoma and Haemonchus (Faveeuw et al., 2003; Haslam et al., 2001; Van Die et al., 1999). In addition, a xylose residue, never seen in mammalian N-glycans, is present on plant N-glycans (Ishihara et al., 1979; Lerouge et al., 1998), but again, this feature is found in invertebrate animals such as snails (van Kuik et al., 1985) and parasitic trematodes (Faveeuw et al., 2003; Haslam et al., 2001).
More than 20 years ago, the existence of a cross-reactive carbohydrate determinant (CCD) was proposed based on the promiscuous binding of allergic patients' sera to periodate-sensitive, heat-stable epitopes in a variety of allergens (Aalberse et al., 1981). Although these criteria might well have been misleading, the hypothesis was later confirmed by increasingly sophisticated experiments. In 1988, a similarly cross-reactive rabbit antiserum was described (Faye and Chrispeels, 1988) and in 1991 anti–horseradish peroxidase (HRP) serum was shown to bind to complex type plant N-glycans containing xylose and core α1,3-fucose (Kurosaka et al., 1991).
Not much later, Tretter et al. (1993) reported that due to the presence of core α1,3-fucose, plant N-glycans can bind IgE from many bee venom allergic patients. In this and other studies, bromelain glycopeptides were employed, and the relative lability of the α1,3-fucosyl linkage allowed the preparation of nonfucosylated glycopeptides, which could be tested in antibody-binding assays. Because these defucosylated glycopeptides—still containing xylose—were unable to bind IgE from bee venom allergic individuals, we concluded that the core α1,3-fucose is the key structural element for antibody binding. Identical results were obtained with patients sensitized against tomato, celery, and other allergens (Foetisch et al., 1999, 2001; Petersen et al., 1996; Westphal et al., 2003). However, other important allergens, such as Ara h 1, Ole e 1, or a vicillin-like 48-kDa protein from hazelnut, to which anti-CCD IgE binds, have been found to contain primarily structures with only xylose (Kolarich and Altmann, 2000; Müller et al., 2000; van Ree et al., 2000). Strong evidence in favor of the assumption that xylose also plays a role as (part of) a glyco-epitope came from a report on the fractionation of a polyclonal anti-HRP antiserum into a fucose-specific and a xylose-specific pool using immobilized honeybee phospholipase (Faye et al., 1993). Unfortunately there was no well-defined glycoprotein containing xylose only that would have helped assess the specificity of these fractions or the specificity of patients' sera IgE. Ascorbic acid oxidase from zucchini, occasionally used for that purpose (Batanero et al., 1999), definitely also contains fucose (Altmann, 1998), whereas not necessarily all of the (partially very complex) structures occuring on hemocyanin from Helix pomatia have so far been elucidated (Lommerse et al., 1997).
Thus, although it is clear that CCDs essentially means complex type plant N-glycans (Foetisch and Vieths, 2001), the relative contributions of fucose and xylose or of other structural features to antibody binding are still unclear. Likewise, a sometimes postulated supportive role of the protein backbone for antibody binding has never been proven or disproven.
The biological significance of carbohydrate determinants in food, insect venom, or pollen allergies, however, is to a much higher degree unclear and topic of a current debate (Foetisch and Vieths, 2001; Foetisch et al., 2003; Hemmer et al., 2001; van der Veen et al., 1997). In many cases, anti-CCD IgE antibodies do not appear to trigger clinical symptoms (van Ree et al., 1997; van Ree, 2002). This is contrasted by positive histamine release tests with some patients' sera (Bublin et al., 2003; Foetisch et al., 1999, 2003; Westphal et al., 2003) and by cases of patients with symptoms elicited by foods in the absence of detectable amounts of IgE against peptide epitopes (Foetisch et al., 2003). The various aspects associated with carbohydrate determinants as allergens, for example, the discrepancies between serum tests, histamine release, skin tests, and finally clinical symptoms, have recently been reviewed and need not to be recapitulated here (van Ree, 2002). Clearly, CCDs play a role as a source of false-positive serum test results in allergy diagnosis as the presence of anti-CCD IgE often has no clinical consequences (Hemmer et al., 2001; Mari et al., 1999; van Ree, 2002). In keeping with earlier reports (Foetisch et al., 1999; Tretter et al., 1993), a recent survey indicated 23% of 1831 patients have anti-CCD IgE (Mari, 2002). The criterion for CCD reactivity was a positive enzyme-linked immunosorbent assay (ELISA) and a negative skin prick test against bromelain—a plant glycoprotein with one complex type N-glycan lacking the α1,3-mannose (Ishihara et al., 1979). Although this report impressively demonstrates the dimension of the problem, it also demonstrates the need for structurally defined and more easily applicable test antigens for the detection of antiglycan IgE antibodies.
Here we report on the synthesis and characterization of a glyco-modified human glycoprotein. Various glycoforms of human transferrin (Tf) with or without xylose, fucose, terminal GlcNAc, or 3-linked mannose have been prepared (Figure 1), and their ability to bind anti-CCD IgG and anti-CCD IgE from patients' sera was tested.
Glycomodification of Tf
Tf carries two N-glycans, which are mainly disialylated and diantennary with only a small fraction being triantennary (Spik et al., 1985). It therefore has a comparatively homogeneous glycosylation, and it can be easily converted into a substrate for transferases requiring nonreducing terminal GlcNAc, such as Fuc-T and Xyl-T, by the use of sialidase and galactosidase. This GnGn-Tf was then subject to incubations with either Fuc-T or Xyl-T, which were recombinantly expressed in Pichia pastoris as recently reported (Bencúrová et al., 2003). Though Xyl-T had been purified to homogeneity by Ni-chelate chromatography, Fuc-T was unstable under such conditions, and hence untagged Fuc-T was enriched by dye-affinity chromatography.
Both Xyl-T and Fuc-T could be obtained in a form concentrated enough to allow an essentially complete conversion of several mg of GnGn-Tf to GnGnX-Tf or GnGnF-Tf, respectively (Figures 1 and 2). Initially we had problems to achieve complete fucosylation, and speculations were made as to a restricted access of the transferase to the attachment on the native Tf. However, the problem could be overcome by adding aliquots of the donor sugar GDP-fucose over the course of incubation; thus, the incomplete conversion was due the instability of GDP-fucose rather than a decreasing activity of the enzyme. Although the Pichia strain used (GS115) secretes detectable amounts of protease (Brierley, 1998) the final products were essentially intact with only a small amount of a fragment band at 41.3 kDa (Figure 4D).
To make sure that in following experiments any observed antibody binding could be exclusively attributed to the glycan moieties of Tf, the Tf incubation mixtures were done in duplicate, one with and one without the donor sugars UDP-xylose or GDP-fucose. However, as will be described later, there was absolutely no binding of anti-HRP and several patients' sera to these controls. Thus we believe that in the future it will not be necessary to add Fuc-T or Xyl-T in the absence of the relevant nucleotide sugar to prepare MM-Tf from GnGn-Tf. An aliquot of the xylosylated GnGnX-Tf was subsequently fucosylated to yield GnGnF-Tf.
Recent reports would suggest that the N-glycans recognized by anti-CCD did not contain terminal GlcNAc (Foetisch and Vieths, 2001; van Ree et al., 2000; Wilson et al., 1998). In addition, preliminary results of anti-HRP western blots with fucosylated and xylosylated forms of GnGn-Tf with or without prior hexosaminidase indicated that in fact the terminal GlcNAc residues had a weakening effect on antibody binding. Therefore the primary transferase products were digested with hexosaminidase to finally yield the glyco-variants MM-, MMX-, MMF-, and MMXF-Tf. The success of the various modification steps was finally verified by mass spectrometry (MS) of the glycopeptides around Asn 630 (Figure 3). The glycopeptide 421–433 could not be observed in the matrix-assisted laser desorption ionization (MALDI) MS spectrum. It should be noted that the glycosidases used (except the α-mannosidase) were not of plant origin, thus avoiding any glyco-contamination from this source.
Specificity of two rabbit antisera and a rat monoclonal antibody
With optimized amounts of fully converted Tf and optimized dilutions of antiserum binding to GnGnF-Tf could not be observed (Figure 4A). An inhibition experiment confirmed the attenuating effect of terminal GlcNAc residues (Table I) and thus the further study was conducted with MMF-, MMX-, and MMXF-Tf. We could not detect any cross-reactions of the sera with components from the expression system despite thorough controls, for example, by making transferase incubations without or with the “wrong” nucleotide sugar.
|Antibody||Coat antigen||Inhibitor concentration (µg/ml)||Inhibition (%) effected by|
|Antibody||Coat antigen||Inhibitor concentration (µg/ml)||Inhibition (%) effected by|
Experiments were performed to give an absorbance in the range of 1.0–1.5 for the uninhibited serum, with the exception of YZ1/2.23 reacting with MMX-Tf, where the absorbance was only about half of that obtained with MMF-Tf as coat. The standard error of the inhibition values was estimated to be 5–10%. An inhibitor concentration of 1 µg/ml translates to an N-glycan concentration of ∼25 µmol/L.
A western blot with MMX- and MMF-Tf revealed specific anti-HRP binding to both new glyco-epitopes (Figure 4A). This is in keeping with a report on fractionation by immobilized bee venom phospholipase of anti-HRP serum into a fucose- and a xylose-specific fraction (Faye et al., 1993). Anti–bee venom serum showed binding to MMF-Tf as well as to MMXF-Tf, but not to MMX-Tf (Figure 4B). This result reflects the absence of xylose in honeybee venom (Kolarich and Altmann 2000; Kubelka et al., 1993, 1995).
An ELISA titration of anti-HRP using MMX-, MMF-, and MMXF-Tf as antigens revealed strikingly similar antibody titers against each of the three glycoforms. This could imply that this serum consists of one population of antibodies that although polyclonal all bind with similar affinity to each of the three glycoforms. However, subsequent inhibition studies revealed the opposite. Binding of anti-HRP to MMX-Tf could not be inhibited by MMF-Tf and binding to MMF-Tf not by MMX-Tf (Table I). In contrast, binding of the rat monoclonal antibody YZ1/2.23 to MMX-Tf and MMF-Tf could be mutually cross-inhibited by MMX- and MMF-Tf (Table I). It is noteworthy that the absorbance measured for binding to MMX-Tf was only half that obtained with MMF-Tf. Apparently, YZ1/2.23 bears a paratope that covers both the fucosyl and the xylosyl residues, but with fucose contributing more to binding strength. Thus YZ1/2.23 cannot be used to discriminate between fucosylated or xylosylated glycoproteins (Figure 4C).
In previous work, defucosylation of bromelain glycopeptides led to a drastic decrease in their antibody-binding capacity, indicating negligible antibody binding by the glycopeptide carrying solely xylose (Petersen et al., 1996; Tretter et al., 1993; Wilson et al., 1998). The data presented here appear to contradict these earlier reports, because binding to MMX-Tf was displayed by both anti-HRP and YZ1/2.23 (Figure 4 and Table I). However, in the previous studies, defucosylated bromelain, which has a MUX structure, was used. Comparison of the inhibitory potency of MMX and MUX for anti-HRP binding to MMX-Tf revealed the significance of the α1,3-mannosyl residue (Table I and also Figure 4). In other words, although the MUXF structure of bromelain glycopeptides is suitable for measuring antifucose antibodies (Wilson et al., 1998), it does not appear to be useful for detecting antixylose antibodies.
Up to that point, only Tf glycoforms had been used as coat antigens. A natural glycoprotein might contain additional structural features, and thus we chose oil seed rape pollen (OSR) extract as more natural example to explore the natural history of anti-CCD antibodies. Essentially complete inhibition of anti-HRP binding to OSR could be obtained by MMXF-Tf at a concentration ranging from 1 to 100 µg/ml (Table I). HRP itself, however, inhibited more effectively on a weight per volume basis by a factor of about 30 to 60 (Table I). Considering at least 7 N-glycans per molecule of HRP with a mass of 45 kDa for HRP and 1.6 fucosylated glycans per molecule of Tf with a mass of 70 kDa, the difference shrinks, in molar terms, to between 4 and 9. This still significant difference in inhibitory potency could point to a contribution to antibody binding of the peptide regions in the neighborhood of the N-glycan, which of course differ between transferrin and HRP to which the antibody was actually raised. It could, however, simply reflect the difference in valency of these two glycoproteins, which is known to be a critical factor for binding strength of carbohydrate epitopes (Welply et al., 1994; Yi et al., 1998). Finally it should be mentioned that GnGnXF-Tf had only negligible inhibitory potency for anti-HRP (Table I).
Detection and characterization of IgE against CCDs
Finally, a small panel of allergic patients' sera was subject to measurement of specific IgE binding to the different Tf glycoforms. All of the bee and wasp venom double-positive patients' sera showed reaction with the fucosylated transferrins MMF and MMXF (the latter, however, could not be tested with all sera). Many sera of this group gave especially high readings in ELISA (Figure 5A). Similarly, most of the rape pollen reactive sera also reacted with MMF and MMXF, which corroborates the recent conclusion that OSR reactivity is in many cases due to anti-CCD IgE, whereby rape pollen itself had not necessarily been the elicitor of this immune response (Hemmer et al., 2001). None of the sera in these groups bound with MMX. In the case of the insect venom allergic patients, this is not surprising. However, these sera also bound stronger to MMXF-Tf which was, in addition to being fucosylated, also xylosylated (Figure 5B).
None of the 20 sera from patients monosensitized to birch pollen reacted with any of the glycan-modified Tfs (data not shown), whereas 4 of the 24 sera from patients monosensitized to grass pollen reacted with MMF-Tf and/or MMXF-Tf. The highest prevalence of anti-CCD IgE was found in the sera from patients with multiple pollen allergy with at least 10 out of the 41 sera (24%) being positive (Figure 5C).
To our surprise and disappointment, none of the sera recognized MMX to a significant degree, with the exception of one heavily atopic patient whose serum reacted strongly even with MM-Tf. As this contrasts the results obtained for rabbit anti-HRP, we assume the lack of MMX-binding by human IgE to be the result of our serum selection rather than of a general invisibility of xylose for the human immune system. This is especially obvious for the insect venom group because insect glycoproteins do not contain xylose. Future studies with panels of patients allergic against primarily xylosylated allergens, such as olive pollen, hazelnut, or peanut, may clarify this point.
Recently, human IgG levels against CCD structures have been measured using honeybee venom phospholipase and Helix pomatia hemocyanin as core α1,3-fucosylated probe and as β1,2-xylosylated standards, respectively (Bardor et al., 2002). We compared the results obtained with the glycomodified Tfs with these naturally available probes. The results obtained with phospholipase and MMF- or MMXF-Tf indeed were in agreement. The ELISA readings, however, were generally lower with phospholipase with the exception of a few sera where especially high values for phospholipase suggested the presence of antiprotein IgE (Figure 6A). In other words, because the phospholipase polypeptide is the major allergen of bee venom, it cannot at all be regarded as a reliable probe for the measurement of anti-CCD antibodies. Hemocyanin, like MMX-Tf, was not bound significantly by any of the sera (data not shown).
Using the divalent human glycoprotein Tf and recombinant glycosyltransferases, it was possible for the first time to generate and study plant glyco-epitopes containing either only fucose or only xylose rather than the usual mixture. In contrast to the use of naturally occuring glycoproteins, such as phospholipase A2 from honeybee venom (MMF and other structures), hemocyanin from H. pomatia (MMX and many undefined structures), ascorbic oxidase (erroneously supposed to contain only MMX; Altmann, 1998) or chemically defucosylated bromelain glycopeptides (Wilson et al., 1998) (MUX with possibly some residual MUXF), the approach presented in this article allows the comparison of defined glycoforms all linked to the same carrier protein that had been subject to defined steps of modification. Hence, stringent controls can be introduced, for example, by omission of the nucleotide sugar in the preparation scheme. Although the influence of substitution of the terminal mannose residues by GlcNAc could be studied using GnGnF- and GnGnX-Tf, no natural glycoprotein with a complete substitution of the mannoses is known. Furthermore, degradation of MMX to MUX by mannosidase can be achieved on the whole protein.
Because the glycans on Tf, especially in the GnGn but also in the galactosylated form, are substrate for a variety of yet other glycosyltransferases, the same approach could obviously also be chosen for other glycodeterminants. In addition to antibody binding, the reactivity with animal lectins could be studied with such Tf glycoforms using western blot, ELISA, or immunohistochemistry with Tf-specific antibodies. Indeed, a Drosophila lectin binding core α1,3-fucose has been characterized using MMF-Tf prepared in this laboratory (Bouyain et al., 2002).
In this study the specificity of polyclonal rabbit sera, especially of an anti-HRP serum and of a rodent monoclonal antibody were studied. The latter, YZ1/2.23, exhibited a somewhat complicated epitope structure where both fucose and xylose residues play roles for antibody binding. In contrast, anti-HRP appears to consist of two distinct populations with paratopes either binding fucosylated or xylosylated glycans. Reflecting the absence of xylose in insect glycoproteins, the anti–honeybee serum only bound to core α1,3-fucosylated transferrin.
The results obtained for GnGnF and GnGnXF show that the mere presence of core α1,3-fucose is, however, not the only criterion for binding of these antibodies. The CCD-epitope can be hidden by additional modifications of the N-glycan as was already suggested as an explanation for the low anti-HRP binding of tree pollens, which predominantly contain N-glycans with terminal GlcNAc residues (Wilson et al., 1998, 2001; Wilson and Altmann, 1998).
Steric factors may also influence antibody binding. In western blots, it appeared that the proteolytic fragment generated during enzymatic modifications of Tf binds slightly stronger than the full-length Tf (Figure 3). A similar albeit contrary observation is made with honeybee phospholipase as compared to the much less abundant hyaluronidase (Hemmer et al., 2003). Unfortunately, such effects of steric presentation on antibody binding are difficult to discern experimentally.
Another modulator of binding strength beyond glycan structure is the valency of the glyco-antigen. We suggest that this valency factor explains the difference in inhibitory potency of MMXF-Tf and HRP. Polyvalency is an important factor for the ability to trigger physiological reactions of effector cells in allergy. Because Tf is divalent it can be expected to be effective in biological test systems, such as histamine release by granulocytes.
The analysis of the specificity of patients' sera revealed once again the importance of core α1,3-linked fucose. This had to be expected in the bee and wasp venom reactive group, but it was also observed for patients with OSR and multiple pollen reactivity. Remarkably, however, MMXF-Tf consistently gave higher ELISA readings than MMF-Tf, even though no patient in our test groups exhibited significant binding with merely xylosylated glycans.
It should be emphasized that the biosynthetic glyco-antigens used here are more reliable probes for anti-CCD IgE than, say, honeybee phospholipase, which also contains a number of highly important peptide epitopes. In the near future, larger numbers of patients' sera shall be analyzed using the glycomodified transferrins with methods more suitable for IgE quantitation than ELISA. Apart from the maybe academic question of whether there are xylose-reactive patients, a major issue will be to render allergy diagnosis more reliable by allowing discrimination between IgE binding to peptide or to carbohydrate epitopes. Although the overall prevalence of anti-CCD antibodies in our patients' sera was lower than that reported by others (Mari, 2002), our data suggest that such antibodies may be commonly found in patients with multiple sensitization to many different allergens. However, there is no obvious correlation between anti-CCD titers as measured with, for example, MMXF-Tf and total IgE levels (Figure 6B), which contradicts the view that CCD reactivity is merely a nonspecific phenomenon observed in highly atopic patients.
Especially in the case of insect venom and food allergy, both including a certain risk for fatal or nearly fatal reactions, it appears appropriate to develop tests that can result in reassuring patients where a positive laboratory result is merely caused by CCDs, which supposedly have no clinical significance—at least in the absence of anti-peptide IgE against the respective allergen.
Materials and methods
Enzymes, antibodies, and other materials
Recombinant forms of Arabidopsis thaliana β1,2-xylosyltransferase and core α1,3-fucosyltransferase were expressed in P. pastoris and purified as recently described (Bencúrová et al., 2003). The enzyme unit is defined as transfer of 1 µmol xylose or fucose per min at 16°C or 37°C, respectively, to a dabsylated or dansylated GnGn-tetrapeptide (Calbiochem, San Diego, CA) (Bencúrová et al., 2003). β-Galactosidase from Aspergillus oryzae (Sigma-Aldrich, St. Louis, MO) was purified before use (Zeleny et al., 1997). Neuraminidase from C. perfringens and α-mannosidase from jack bean were also obtained from Sigma-Aldrich. β-N-acetyl-glucosaminidase from Streptococcus pneumoniae (Calbiochem) was used instead of the enzyme from jack bean because the latter contained plant N-glycan structures interfering with the objectives of this work. Honeybee venom phospholipase was prepared as described (Kubelka et al., 1993) and H. pomatia hemocyanin was obtained from Serva (Heidelberg, Germany).
GDP-fucose was purchased from Sigma-Aldrich, and UDP-xylose was obtained from CarboSource Services at the University of Georgia. The monoclonal antibody YZ1/2.23, raised against elderberry abscission tissue, was a gift of Dr. Daphne Osborne (Open University) and David Ashford (University of York). Polyclonal anti-HRP serum from rabbit, anti–bee venom serum from rabbit, and the respective second antibodies have been described before (McManus et al., 1988; Wilson et al., 1998).
Sera were collected from patients undergoing routine allergy testing for inhalant or insect venom allergy. Patients with inhalant allergy were skin-prick tested with common inhalant allergens, including pollens (hazel, alder, birch, grass, rye, ash, plantain, nettle, mugwort, ragweed, OSR, plane tree), house dust mites, animal danders (cat, dog, horse, guinea pig), molds (Cladosporium, Alternaria, Penicillium), and rubber latex (all Soluprick, ALK, Denmark). Insect venom allergy was confirmed by positive radioallergosorbent test and subsequent skin testing.
Three groups of sera were used: (1) Sera from patients suffering from pollen allergy that according to radioallergosorbent and skin-prick test were sensitized to a narrow range of pollen allergens only, for instance, birch and other Fagales pollen only (n = 20) and grass/rye pollen only (n = 24); all patients from this group also had a positive radioallergosorbent test to the respective allergen; 4 of the 44 patients had a weakly positive skin test but negative serology to one of the tested indoor allergens; (2) sera from patients with multiple pollen sensitization, for instance, sensitization to at least 4 different pollen species (n = 41); patients from this group had multiple positive radioallergosorbent test to pollen allergens, although not all allergens reacting positively in the skin test have been tested by serology; 26 of the patients were also sensitized to one or more of the indoor allergens; (3) preselected sera that were assumed to contain anti-CCD IgE, such as those reacting with HMW glycoallergens in OSR (Focke et al., 1988) (n = 9) or those exhibiting cross-reactivity with bee and wasp venom glycoallergens (Hemmer et al., 2001, unpublished data) (n = 7).
Preparation of glycomodified Tfs
Ten milligrams of human apo-Tf (Sigma-Aldrich) was treated for 16 h with 100 mU neuraminidase in 0.5 ml 50 mM sodium acetate buffer at pH 5.0 at 37°C. Subsequently, 4.2 U β-galactosidase from Aspergillus oryzae (Zeleny et al., 1997) was added, and the sample was incubated overnight. The integrity of the resultant GnGn-transferrin was checked by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and the N-glycans were analysed as will be described.
For xylosylation, GnGn-Tf (5 mg in a final volume of 1.6 ml) was incubated with 4 µmol UDP-xylose and 13 mU Xyl-T (in 25 mM 2(N-morpholino)ethanesulfonic acid, pH 7.0) at 16°C for 48 h. For fucosylation, GnGn-Tf (2 mg in a final volume of 0.8 ml containing 20 mM MnCl2) was incubated with 2.8 µmol GDP-fucose (added in three aliquots over time) and 0.6 mU Fuc-T (dissolved in 25 mM 2(N-morpholino)ethanesulfonic acid, pH 6.8) at 25°C for 48 h. A doubly modified glycoform, GnGnXF-Tf, was obtained by fucosylation of GnGnX-Tf.
To remove the terminal N-acetylglucosamine residues, solutions of GnGn-, GnGnX-, GnGnF-, and GnGnXF-Tfs were diluted threefold with 50 mM sodium citrate buffer of pH 5.0 and digested with β-N-acetyl-glucosaminidase (0.5 mU/mg Tf). For the preparation of glycoforms lacking the α1,3-mannosyl residue, 0.45 mg MM- and MMX-Tf were digested for 24 h at 37°C with 20 mU jack bean α-mannosidase in the buffer containing 0.1 mM ZnCl2.
The structures of the N-glycans on the various Tf glycoforms were verified in two ways. (1) In the first approach, 10 µg glycoproteins were digested for 4 h at 37°C with 0.5 µg pepsin (Sigma-Aldrich) in 30 µl 5% formic acid. Then this solvent was evaporated, and the samples were deglycosylated with peptide:N-glyosidase A described (Kolarich and Altmann, 2000). Oligosaccharides were isolated and analysed by MALDI MS on a linear time-of-flight instrument as described (Kolarich and Altmann, 2000). (2) Alternatively, glycoproteins were subjected to SDS–PAGE, and bands were excised, S-alkylated, and digested with described (Katayama et al., 2001; Kolarich and Altmann, 2000). The (glyco-)peptides were analyzed by MALDI MS on a Waters-Micromass Q-TOF GLOBAL system using α-cyano-4-hydroxycinnamic acid as the matrix.
Tf glycoforms (0.1 µg) were separated by SDS–PAGE and electroblotted to a nitrocellulose membrane. The membrane was blocked with 3% nonfat milk in 10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.1% Tween and subsequently incubated for 1 h with rabbit anti-HRP or anti–bee venom serum diluted 1:2000. After washing in Tris-buffered saline, alkaline phosphatase–conjugated anti-rabbit antibody at a dilution of 1:2000 was added, and bands were stained for 30 min with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium.
Tf glycoforms at a concentration of 5 µg/ml or OSR at a protein concentration of 1 µg/ml was used for coating ELISA well plates (Nunc maxisorp) for 1 h at 37°C in 0.1 M sodium carbonate buffer, pH 9.6. After the washing and blocking steps, plates were incubated for 1 h at 37°C with either anti-HRP diluted 1:20,000 or YZ1/2.23 diluted 1:40,000. The subsequent steps were performed as described previously (Wilson et al., 2001).
For inhibition ELISA, sera were preincubated for 1 h at 37°C with different glycoproteins at different concentrations.
Microtiter plates were coated with Tf glycoforms overnight at 4°C but otherwise as described. The plates were washed twice with phosphate-buffered saline (PBS) containing 0.05% Tween 20 and blocked with PBS containing 1% bovine serum albumin for 2.5 h at room temperature. The plates were then incubated overnight at 4°C with allergic patients' sera diluted 1:10 with PBS. After washing the plates five times with Tween-20 containing PBS, anti-human IgE–alkaline phosphatase conjugate (Pharmingen, San Diego) diluted 1:2000 with PBS containing 0.05% bovine serum albumin was added, and the plates were incubated first for 1 h at 37°C and then for 30 min at 4°C. Plates were again washed five times and then stained with p-nitrophenyl phosphate dissolved to 1 mg/ml in 0.1 M diethanolamine of pH 9.7. The optical density at 405 nm was read after 2 h.
We thank Daniel Kolarich for acquiring the peptide spectra on the Q-TOF mass spectrometer, which was funded by the Austrian Science Council. This work was supported by a joint research program of the Austrian Science Fund (project S8803).
2Glycobiology Division, Institute of Chemistry, University of Natural Resources and Applied Life Sciences (Universität für Bodenkultur), Muthgasse 18, A-1190, Vienna, Austria; and 3FAZ-Floridsdorf Allergy Center, Vienna, Austria