The Regulation of Glycan Processing in Glycoproteins THE EFFECT OF AVIDIN ON INDIVIDUAL STEPS IN THE PROCESSING OF BIOTINYLATED GLYCAN DERIVATIVES*

The effect of the protein matrix on glycan processing by rat liver Golgi enzymes has been evaluated by a direct comparison of substratwproducts conversion of a free glycan and of the same glycan linked to a protein. The glycan substrates had the general struc- ture R-glycan where R represented either biotinyl-Asn-GlcNAc,- or 6-(biotinamido)hexanoyl-Asn-Glc- NAc2- and the protein used was avidin; the extension arm in one of the glycan substrates permitted the ad- ditional comparison of two avidin-biotin-glycan complexes. By the use of different glycans as substrates, by the presence or absence of donor substrates (UDP-GlcNAc, UDP-Gal, and CMP-sialic acid (Sia) and/or the inhibitor, swainsonine, it was possible to dissect the individual steps involved in the conversion of R-Mana (or R-Man5) to a biantennary complex glycan, R- Man3-GlcNAcz-Gal,-Siaz or to the hybrid glycan R- Mans-GlcNAc-Gal-Sia. Using fast atom bombardment-mass spectrometry to identify and quantify the sub- strates and products of each parallel incubation of free and avidin-bound substrates, the following observations were made. With the substrate without the exten- sion arm, avidin-binding inhibited

Galz-Sia+R-Man3-GlcNAc2-Galz-Siaz) and to a lesser extent in the hybrid pathway (R-Man6-GlcNAc-Gal+ R-Mans-GlcNAc-Gal-Sia). GlcNAc transferase I1 did not appear to be affected by avidin. Based on the information available on the biotin-binding site in avidin, it is proposed that the short range effect reflects the masking of the core chitobiose unit in the avidinglycan complexes in the absence of the extension arm, but not in the presence of the arm, and that the early Grant GM 31305 and Robert A. Welch Foundation Grant AU 916.

* This work was supported by United States Public Health Service
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "adoertlsement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Visiting Scholar. Permanent address: Dept. of Biochemistry, Shanghai Medical University, Shanghai, China. processing enzymes thus may require a fully exposed chitobiose for full activity.
The long range effect is tentatively assumed to reflect conformational alterations of the glycan caused by protein-glycan interactions away from the biotin-binding site.
In the previous paper (1) we demonstrated that avidinbiotinylglycan complexes represent reasonable models by which the effect of the protein matrix on glycan processing can be studied. Determining the processing efficiency of free blotinylated glycans in direct comparison to the same glycans bound to avidin, two separate types of inhibitory effects of the protein were observed. The first type was defined as a strong inhibition of processing of substrates with the glycan close to the biotin-binding site (biotinyl-Asn-glycans) which was partially or totally eliminated when the glycan was moved further away from the biotin-binding site and the protein surface (6-(biotinamido)hexanoylglycans). The second type of effect was expressed equally strongly for both kinds of substrates and these were interpreted to involve more subtle interactions between the glycan and the protein. The data were obtained entirely by following the incorporation of radioactive sugars into the glycan substrates and it was thus not possible to properly identify individual processing steps involved in the protein effects. The present work was undertaken to attempt to analyze each major enzyme of the processing sequence for its sensitivity to the effects of the protein matrix and to identify by the use of fast atom bombardmentmass spectrometry the major products of each reaction to establish whether the protein effects represent the specificity determinants that govern which type of glycan will be the final product at a given glycosylation site. A good deal of knowledge is available on the analysis, structure, and biosynthesis of the glycan units of glycoproteins (2, 3). Some of the information on the processing pathway is summarized in Fig.  1 below (3). Presumably the blockage of the early steps involving mannosidase I and GlcNAc transferase I would give rise to high-mannose type glycans, the specific blockage of mannosidase I1 would yield hybrid-type glycans, while full activity of all enzymes would yield the typical complex glycans.

EXPERIMENTAL PROCEDURES AND RESULTS'
Portions of this paper (including "Materials and Methods," part of "Results," Footnote 2, Figs. 2 and 3, and Tables I-VIII) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are available from the Journal of Biological Chemistry, 9650 Rockville Pike, Bethesda, MD 20814. Request Document No. 86M-2532, cite the authors, and include a check or money order for $6.40 per set of photocopies. Full size photocopies are also included in the microfilm edition of the Journal that is available from Waverly Press.

DISCUSSION
Before evaluating the results obtained in this work, it is important to assess both the advantages and disadvantages of the system and the experimental design. The main advantage of the noncovalent avidin-glycan neoglycoprotein is the unique feature of permitting a direct comparison of the processing of a given glycan or a glycan mixture free in solution and bound to a protein. As shown previously (I), the complex is stable during the processing and the bound glycan can be completely recovered for analysis at the completion of the processing reaction. By always comparing only the parallel incubations of free and bound glycan it appears that the results allow meaningful interpretations even in cases where the main substrate is only 50% pure (steps 5 and 6) and where a number of secondary reactions take place. By following strictly the established experimental protocol we have found the individual experiments to be remarkably reproducible in terms of the relative extent of processing of free and bound glycan, even if the absolute values may vary from experiment to experiment. This point is particularly important in view of the analytical technique used. The various uncertainties involved in using the molecular ion intensities from the mass spectra as a direct quantitative estimate of the different components in the reaction mixture have been considered under "Results." The concerns for an unknown extent of fragmentation and for the linearity of ion intensities and concentration outside the narrow range tested (Table I) and for glycans not tested at all are all valid and impose serious restrictions on any interpretation based on strict quantitation. The way the experiments were designed, however, absolute quantitation is not essential; the important feature throughout is the relative extent of processing of free and bound substrate. The parallel analysis of each experimental pair of free and bound substrate includes the same set of substrates and products, whose fractional amounts were estimated in the same way from ion intensities. Even if there are significant errors in determining the absolute quantities involved, the direct comparison of the extent of processing should nevertheless be valid for each pair.
The main disadvantages of the system are related to the practical limitations imposed on the kinetic analysis of processing and to the inherent limitations of the system itself in making available only a single type of protein-glycan interaction in the avidin-biotinylglycan neoglycoprotein. The selfimposed experimental limitations are mostly based on availability of substrates and the investment of time for each analysis; to do a proper kinetic analysis of every processing step would be a monumental task. The decision to make each comparison of the extent of processing of free and proteinbound glycan at a single time point was thus a practical one, but the limitations it imposes on the analytical data need to be evaluated. Based on preliminary data, the single time point of 8 h selected for each incubation was one that gave about 90% conversion of R-Man5 to ['4C]GlcNAc-containing products under the standard assay conditions (1); over a period of time we have found that the actual conversion does vary from experiment to experiment; in the experiment shown in Fig. 3, only 73% of the starting substrate was converted to products. Assuming that the experimental range of conversion in 8 h is 75-loo%, it is clear that the sensitivity in detecting minor rate effects is minimal; in fact for single step reactions that may go to completion in even shorter times, fairly large rate effects might be missed by observing complete conversion for both free and bound substrate in 8 h. There is only one set of data for which that possibility may have to be considered, that of a-mannosidase I1 in step 3, in which the extension arm substrate gave essentially complete conversion to product. In all the other steps, enough unreacted substrate remains to suggest that the one time point should permit detection of significant rate differences. The point here is that the processing rates observed in 8 h for free and bound substrate may underestimate the actual rate differences and, in fact totally fail to detect small differences for fast reactions; however, where differences are observed they can be considered to be significant. The other problem associated with the present kinetic analysis is the fact that the actual experimental substrate concentration is ambiguous. The amount of glycan was identical in the parallel samples of free and bound substrate, but while in the case of free glycan the substrate ([glycan]) concentration was 2 X M, in the case of bound glycan, the substrate with 2.5 mol of glycan/mol of avidin tetramer ([a~idin-glycan~.~]) concentration was 8 X M. In the absence of a complete kinetic analysis, it is impossible to properly evaluate this effect. It may be insignificant or it may reduce the rate with the protein-bound substrate by a factor of 2.5, and in that case it could lead to a wrong interpretation of the effect of the protein on processing. In the results presented in Tables II-VIII, most of the observed differences in processing are fortunately too large to reflect this substrate concentration effect, but some differences, notably those of free and bound 6-(biotinamid0)-hexanoylglycans are small enough to simply reflect a substrate concentration effect. In the case of Gal incorporation (Table VII) and Sia3 incorporation (Table VIII) some of the differences are also small enough to give cause for concern; however, the fact that 2 residues are incorporated in both 5A and 6A and that apparently only the incorporation of the second residue is inhibited (the sum of mono-and diglycosylated product relative to substrate is very similar for free and bound substrate) lends support to the interpretation that these are protein effects on processing rather than substrate concentration effects. The fact that GlcNAc transferase I1 gives identical rates for free and bound substrate provides additional evidence that the effective substrate concentration may be identical in the absence and presence of avidin (Table VI). Figure 1 summarizes the results presented in Table II-VI11 in the Miniprint Supplement. At each processing step the letter S (for short range proximity effect) signifies that a substantial decrease in processing rate was observed for the avidin complex of biotinylglycan but not for the avidin complex of 6-(biotinamido)hexanylglycan; the letter L (for long range proximity effect) signifies that an inhibitory effect of avidin was observed for the complexes of both glycan derivatives. An asterisk signifies that for a particular two-step reaction, only one step appeared to be affected. Based on the quantitative limitations discussed above, the results can only be reported in an all or none fashion; but it is clear that the magnitude of the protein effect on processing varies considerably. Only one enzyme, GlcNAc transferase 11, appeared not to be affected by avidin. The short range proximity effect may be visualized as a simple steric masking of the critical residues and bonds from access to processing enzymes in the cavity of the biotin binding pocket. These effects are similar to those observed previously with this and similar neoglycoproteins when they are exposed to isolated jackbean a-mannosidase or endoglycosidase H (4, 16,17). There is one disturbing feature of this model, namely the size of the cavity that could accommodate the majority of the glycan. Green (18)  with the aminohexanoyl arm inserted between biotin and Asn, the entire chltobiose unit should be fully exposed.
To allow for these features, it is proposed that the regulatory signal to the processing enzymes in distinguishing between the substrates with and without extension arm is based entirely on the exposure of the chitobiose unit. This model has attractive features in terms of in vivo regulation; if the activity (i.e. a low K,) for a given enzyme is determined by the chitobiose exposure, very dramatic effects on processing could perhaps result from quite minor alterations in the protein matrix involving the chitobiose environment only. It is interesting to note in this connection that the branch specificity reported for the bovine cholostrum a2-6-sialyltransferase is expressed only if at least one of the core GlcNAc residues is present in the substrate (14), and thus that the proposed regulatory role of the chitobiose unit is not without precedent. Unfortunately, we do not have sufficiently detailed structural information on the biotin-binding site in avidin to select one model over the other. The long range proximity effect was identified on the basis of its apparent independence of the presence of the extension arm in the substrate, the processing of both substrates was inhibited to the same extent in the protein complex as compared to their free forms. It is natural in this case to consider protein-glycan interactions which may affect glycan conformation, but it is difficult on the basis of current knowledge to present a very precise model. The long range effect was observed primarily for late steps and involving only one branch of the biantennary structure, and it is important to note the increasing body of evidence showing that the two branches indeed behave differently. Based on the knowledge that the rotation around the C5-C6 bond in the 1-6 branch adds considerable conformational flexibility to the common \k and rotation of the glycosidic bond and by examination of models (19) it was realized early that while the 1-3 branch is fairly rigidly fixed relative to the chltobiose core, the 1-6 branch may assume any one of several favorable orientations. These general features have been confirmed by a number of experimental approaches including NMR (20-23) and x-ray diffraction analysis (24) and have been evoked to explain branch specificity of several processing enzymes (12,14,25). In the production of the hybrid glycan in this work only the 1-3 branch is involved and the observation that the incorporation of both Gal and Sia is inhibited in the avidin complex (Tables VI1 and VIII) consequently suggests that it is the 1-3 branch that is masked in the complex. This in turn could imply that the strong long range proximity effect observed for the incorporation of the second Sia residue in the biantennary product also involves the 1-3 branch and that the 1-6 branch was fully exposed as substrate. Such an extrapolation does not seem valid, however. The two substrates are clearly different and the presence of the 1-6 mannosyl residue in the hybrid pathway could obviously have unique effects on the transferase.
It appears that the avidin-glycan system provides a valid model system for glycoprotein processing. The proximity effects observed here confirm the general observation that the protein matrix has a significant effect on in uiuo processing (26)(27)(28) and are consistent with the general mechanisms proposed to explain the effect. Evidence that the proximity effects such as those observed here must be involved in in uiuo processing has been obtained in experiments that demonstrate that high-mannose structures at glycosylation sites that are resistant to endoglycosidase H in the native protein structure are not processed to complex glycans, while those that are susceptible to endoglycosidase H (from a mutant unable to make complex oligosaccharides) correspond to the ones that are processed in the wild type (29, 30). Thus, exposure of the core chitobiose unit by the criterion of endoglycosidase H digestion appears to be required for processing, just as was concluded from our observations for the short range proximity effect. It should be noted that the avidin-biotinylglycan complexes have been found to be resistant to endoglycosidase H digestion under conditions that gave complete hydrolysis of the corresponding free glycan (4) and intermediate rates of hydrolysis with 6-(biotinamido)hexanoylglycan. The new feature of the present work may be the strong indication that the proximity effects reflect different types of protein-glycan interaction, the short range ones which may involve the exposure of the core chitobiose and the long range ones, which presumably reflect conformational alterations of the outer chains of the glycan. In glycoprotein biosynthesis the former could presumably represent one mechanism through which the commitment to produce either a high-mannose, hybrid, or complex-type glycan is made, and the latter one mechanism through which specific variations in the nonreducing termini of multiantennary glycans can be achieved.  Vlll The