Purification and properties of a bacteriophage-induced endo-N-acetylneuraminidase specific for poly-alpha-2,8-sialosyl carbohydrate units.

The soluble form of a bacteriophage-induced endo-N-acetylneuraminidase (Endo-N) specific for hydrolyzing oligo- or poly-alpha-2,8-linked sialosyl units in sources as disparate as bacterial and neural membrane glycoconjugates was purified approximately 10,000-fold and characterized. The enzyme appears homogenous by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and has a subunit Mr 105,000. This corresponds to one of the higher Mr phage proteins which comprises 7.5% (by weight) of the total phage protein. The holoenzyme is active at neutral pH and has a Mr by gel filtration of 328,000, suggesting that the active enzyme is a trimer. Endo-N requires a minimum of 5 sialyl residues (DP5, where DP represents degree of polymerization) for activity. The limit digest products from the alpha-2,8-linked polysialic acid capsule of Escherichia coli K1 are DP4 with some DP3 and DP1,2. DP2-4 do not appear to inhibit depolymerization of polysialic acid. Endo-N digestion of the polysialosyl moiety on neural cell adhesion molecules yields sialyl oligomers with DP3 and DP4. The presence of a terminal sialitol changes both the distribution of limit digestion products and the apparent minimum substrate size. Higher Mr alpha-2,8-linked sialyl polymers (approximately DP200) are better substrates (Km 50-70 microM) than sialyl oligomers of approximately DP10-20 (Km 1.2 mM). Endo-N activity is inhibited by DNA and several other poly-anions tested. An examination of the distribution of intermediate products shows that Endo-N binds and cleaves at random sites on the polysialosyl chains, in contrast to initiating cleavage at one end and depolymerizing processively. Endo-N can serve as a specific molecular probe to detect and selectively modify poly-alpha-2,8-sialosyl carbohydrate units which have been implicated in bacterial meningitis and neural cell adhesion.

types has attracted attention because of its association with meningitis in human and animal neonates (6)(7)(8)(9). A structurally identical polysialic acid capsule is present in Neisseria meningitidis serogroup B (9). Although the wide spread occurrence and biological significance of sialic acid has long been recognized (10, l l ) , the nonbacterial occurrences of polysialic acid was unknown until recently. This was due in part because of the lability of poly-a-2,8-sialosyl linkages to pH 5 or to heating (100 "C/5 min) at neutral pH (12) and to the difficulties in readily distinguishing between mono-and polysialic acid. Recently, prokaryotic derived probes have been developed that permit the simple detection of poly-a-2,8-sialosyl carbohydrate units in embryonic neural cell adhesion molecules (N-CAM)' (13). N-CAMS function in neural cell-cell interaction and neural development (14)(15)(16)(17)(18)(19)(20)(21)(22)(23). Evidence that the polysialosyl epitope of N-CAM has an effect on N-CAM-mediated adhesion between living cells and that the amount of this carbohydrate is important for normal development of neural tissue has also been obtained (24). Further studies on the role of polysialic acid in these and other processes and on the molecular characteristics of polysialic acid biosynthesis will be facilitated by the use of specific, well-characterized reagents that permit detection and selective modification of the polysialosyl moiety. Several endoand exoglycosidases have proven to be highly useful reagents for structural analysis of glycoconjugates (25-28). One such enzyme is an endo-N-acetylneuraminidase (Endo-N) associted with bacteriophages that specifically recognize the polysialic acid capsule of E. coli K1 as a receptor (13). Initial studies utilizing Endo-N and some of its properties have been reported (13,24,29,30). Whereas the enzyme used appeared to be free of proteolytic and exoneuraminidase activity, further experiments to detect polysialosyl units possibly associated with neurological disorders and to elucidate the biological effects of desialylated N-CAM on neural development will be more easily interpreted by using an enzyme that has been purified to homogeneity and is well characterized. Two reports

3553
on some of the properties of a bacteriophage-bound endoneuraminidase have appeared (31,32); however, a purified, soluble endoneuraminidase may have a distinct advantage, particularly for in vivo microinjection experiments to study the functional significance of polysialosyl units on N-CAM. The purification of a soluble endoneuraminidase and its specificity for various substrates containing sialic acid were recently reported (33), but no information was presented as to the size of the limit digestion products or the minimal substrate size.
This report describes the purification and properties of the soluble form of a bacteriophage-induced endo-N-acetylneuraminidase that is specific for degrading at neutral pH oligo-or poly-a-2,8-sialic acid units in prokaryotic and eurkaryotic glycoconjugates. Availability of the purified enzyme provides a valuable molecular probe to detect, modify selectively, and study the structure, synthesis, and function of poly-a-2,8-sialosyl carbohydrate units. These results are discussed in relation to the properties of the previously described forms of endoneuraminidases (31)(32)(33) and the minimum sialyl oligomer size required to interact with the substrate-binding site of the enzyme.

Characterization of Bacteriophage KIF Proteins
The CsC1-purified phage particles isolated as described under "Experimental Procedures" were analyzed by SDS-PAGE to characterize the molecular weights and relative abundance of the phage-associated proteins ( Table I). The K1F phage banded at a density between 1.3 and 1.6 g/cm3 in the CsCl gradient, indicating that it was similar to the endoneuraminidase-containing phages studied by Kwiatkowski et al. (31) and Finne and Makela (32). These phage particles were reported to have a bouyant density of 1.47 g/cm3, but in neither case were the proteins characterized. The K1F phage isolated here appeared homogenous in that only a single, high molecular weight DNA band was seen upon agarose electrophoresis after phenol/chloroform extraction and ethanol precipitation. As shown in Table I, the molecular weights of the major K1F-associated proteins were similar to other capsulespecific bacteriophages previously studied. The three capsular depolymerases previously characterized were reported to be spike elements of the phage that existed as a high molecular weight complex of 313,000 ( E . coli K l l ) , 245,000 ( E . coli 429), and 379,000 (Klebsiella aerogenes). The holoenzymes appeared to consist of non-identical subunits, as shown in Table  I. In contrast, Endo-N is an oligomer ( M , 328,000) composed of three identical subunits of M , 105,000 each. The purpose in purifying the phage and characterizing its proteins was 3fold. First, determination of the protein components of the phage served as a determinant of purity of the soluble depolymerase since Endo-N should consist of one or more of these proteins. Second, a comparison of the relative substrate specificities and other enzymological properties of the phageassociated and soluble Endo-N depolymerase might reveal differences due to immobilization of the enzyme in the phage particle. Third, at least in some cases, it might prove more facile to purify the phage-bound enzyme to homogeneity since the enzyme is naturally enriched in the purified phage preparation, i.e. purification of the phage particle could serve as an "affinity" purification step.
Treatment of purified phage K1F with either 4 M guanidinium C1 or 0.1% sodium dodecyl sulfate destroyed Endo-N activity. However, two other procedures: either mild acid treatment, previously shown to be effective in solubilizing depolymerase-containing phages 429 (42) and K11 (43), or treatment with 8 M urea, previously used to solubilize the K . aerogenes capsular polysaccharide depolymerase (45), were effective in quantitatively solubilizing Endo-N from K1F phage particles. However, substantial losses of activity accompanied further attempts at purification, possibly due to varying degrees of reaggregation of the phage proteins during subsequent purification procedures. Therefore, the soluble form of the enzyme in K1F phage lysates of E. coli K1 was purified to homogeneity, as described below.

Purification of Soluble Endo-N from KIF Phage Lysates
The purification procedure described under "Experimental Procedures" and summarized in Table I1 allowed the purification to apparent homogeneity of about 4.4 mg of Endo-N in 3 days. Overnight incubation of purified Endo-N with a number of high and low molecular weight proteins and glycoproteins prior to SDS-PAGE showed no contaminating protease activity. The enzyme was purified greater than 10,000-fold, based on the lysate activity, which, however, might be underestimated. The 2-fold increase in total activity, and the apparent high degree of -fold purification are probably due to the removal of an inhibitor(s) of Endo-N at the final purification step. Various polyanions, including DNA, chondroitin sulfate, and poly-y-D-glutamic acid, inhibit Endo-N, presumably through ionic interactions since Endo-N is bound to heparin-agarose (results not shown). At least part of the inhibition during purification appears to be due to the presence of DNA. The purified glycohydrolase (20 pg ml-') was inhibited >90% by E. coli DNA (0.75 mg ml-') and treatment of fractions from initial stages of the purification with DNase (-1 mg ml-') resulted in a subsequent %fold increase in Endo-N activity. Thus, it is likely that it is the removal of DNA from the phage lysate by the hydroxylapatite column that is responsible for the increase in total activity and is therefore the most important purification step. The -fold purification shown in Table I1 may be an overestimation due to the unknown amount of inhibition of Endo-N activity in the lysate. However, we believe these results are the most accurate presentation of the data in the absence of detailed characterization of inhibitory factors that might be present in the phage lysate.
The ammonium sulfate pellet ( Table 11, Step 2) was reextracted with buffer to enhance recovery of Endo-N activity entrained in the mass of insoluble debris. The main purpose of the high speed centrifugation (Step 3) was to recover K1F phage, which is why both the total activity and the specific activity decreased at this stage. This step can be eliminated For acid treatment of K1F phage particles, the pH of a purified K1F phage suspension was lowered to 3.5 by addition of an equal volume of 0.4 M glycine HCl (pH 3.5). An immediate precipitate formed. After 10 min at pH 3.5, the pH was readjusted to pH 7.4 by the addition of 1.8 M Tris (pH 8.8), and the high molecular weight material was removed by ultracentrifugation (120,000 x g, 1 h). For solubilization with urea, the phage suspension was made 8 M in urea, incubated at 37 "C for 30 min, and then subjected to ultracentrifugation as above. if only the soluble Endo-N activity is desired. Heating the high speed supernatant at 60°C for 25 min (Step 4) and refractionation with 40% ammonium sulfate (Step 5) resulted in a substantial loss of protein and an approximately 7-fold purification. The most significant purification step was obtained by chromatography of the active fractions from the hydrophobic interaction chromatography (Step 6) on a double column of hydroxylapatite connected in series with DEAE-Trisacryl (Step 7). A single, symmetrical peak of Endo-N activity and protein was eluted at about 0.15 M NaC1.4 When this material was analyzed by SDS-PAGE (Fig. 2, lane 9), it was found to consist of one major protein band of greater than 99% homogeneity that co-migrated with the M , 105,000 protein of the K1 phage (Fig. 2, lane IO). The    a-2,gd 0 Endo-N activity was measured by product formation, as described under ''Experimental Procedures," which yields initial reaction rates with all the substrates used. Thus, V, , is expressed as micromoles of NeuNAc-reducing equivalents formed min" mg of protein".
tracking dye had reached the bottom of the tube, the gel was removed and cut into 2-mm slices. The slices were eluted for 2 h with 50 mM Tris buffer (pH 7.5) and then assayed for Endo-N activity. The only gel slice that showed Endo-N activity contained only the M , 105,000 protein, as shown by subsequent SDS-PAGE. Thus, the purification procedure reported here allows the facile purification to homogeneity of sufficient quantities of Endo-N for further enzymological studies and to specifically detect, modify selectively, and study the function of poly-a-2,8-sialosyl units in a variety of biological systems. In the only previous report of the purification of a soluble endoneuraminidase (33), the enzyme was reported to consist of two subunits ( M , 74,000 and 38,500). This difference in size and number of subunits may be due to different phages being used. That enzyme had been purified 238-fold and had a specific activity of 0.95 (pmol of sialic acid released per min/mg of protein). In contrast, the Endo-N reported here was purified -10,000-fold and is markedly more efficient (-4,200 pmol min-l mg-'). It is thus possible that an inhibitor was present and that one of the two reported subunits could be a contaminant or a proteolytic cleavage product.

Properties of Endo-N-acetylneuraminidase
Kinetic Constants-Relatively little is known about the kinetic constants and enzyme mechanisms of endoglycohydrolases, in particular, endo-N-acetylneuraminidases. Here the apparent K,,, and VmaX for both the phage-bound and soluble form of Endo-N were determined for several oligoand polysialosyl substrates, as summarized in Table 111. No significant difference in apparent K,,, values was found between the K1F phage-bound and soluble form of the enzyme for either the low or high molecular weight a-2,8-linked substrates or the a-2,8-a-2,9-mixed linkage polymer (Table 111). The high molecular weight a-2,8-linked sialyl polymer (-DP150-200) appeared to be a substantially better substrate (apparent Km-50-70 p~) than the shorter (DP10-20) sialyl oligomers (apparent Km-1.2-1.6 mM). The large difference in K,,, values found here for Endo-N action on long polymers and short oligomers of sialic acid suggests that a different enzyme mechanism might be involved in processing these two different substrates! However, since the total number of A recent study (48) on the catalytic mechanism of polyphosphate glucokinase found that the apparent K,,, for polyphosphates was markedly different for chain lengths of 32 or -724. An examination of the variation of K,,, with chain length of polyphosphates demonstrated a distinct inflection at a chain length of 100, which appeared to coincide with a change in the mechanism of enzyme action from processive at longer chain lengths to nonprocessive at shorter chain lengths.
potential cleavage sites is about the same at these concentrations of oligo-and polysialic acid, this suggests that the apparent affinity of the enzyme for these two substrates is a function of the concentration of potential cleavage sites and hence is approximately the same. Possible mechanisms of Endo-N cleavage are discussed in more detail below. Unexpected was the observation that the alternating a-2,8-a-2,9ketosidically linked capsular sialyl polymer from E. coli N67 appeared to be a better substrate (apparent K,,, 6.6 PM) than the a-2,8-linked polymer from E. coli K1 ( K , 51 PM). Only the a-2,8 linkages in this polymer, however, were hydrolyzed by Endo-N (see below). The K,,, for polysialic acid of the free form of endoneuraminidase reported by Tomlinson and Taylor (33) was 7.4 mM or 145-fold higher than what was determined here.
Products of Endo-N Limit Digestion-The products of prolonged digestion (limit digestion) of various oligomers and polymers of sialic acid were examined. As shown in Fig. 3a, the primary digestion product of U-14C-labeled, high molecular weight poly-a-2,8-a-2,9-sialic acid was DP8 with a smaller amount of DP4,6,10,12,14. Since Endo-N does not hydrolyze a-2,g-linked polysialic acid (results not shown), this pattern was due to the specific cleavage of the a-2,8 linkages within the mixed linkage polymer, producing only oligomers containing an even number of sialyl units. This is in contrast to the spectrum of sialyl oligomers with an even and odd number of units obtained by the partial and random hydrolysis of polya-2,8-linked sialic acid (see below; Fig. 5 and 7). in the reducing termini after reduction with NaB3H4 (c) were subjected to overnight digestion by Endo-N, and the products were identified by HPLC as described under "Experimental Procedures." but in this case, the mixed linkage polymer appeared to be a poor substrate, and the digestion products were reported to be DP3,4,6,7 and higher residues, indicating that some cleavage of a,2-9 linkages may have occurred (49). As stated above, the primary hydrolysis product of the poly-a-2,8-a-2,9 capsule by Endo-N was DP8. In contrast, the primary digestion product of U-14C-labeled, high molecular weight poly-a-2,8sialic acid was DP4 with a smaller amount of DP3 and DP1,2 (Fig. 3b). These results further confirm that Endo-N was specific for the poly-a-2,8 linkage in the poly-a-2,8-~~-2,9mixed linkage polymer. The molar distribution of products shown in Fig. 3b is consistent with what can be calculated for the random cleavage of a polymer with a minimum substrate size of DP5. Unexpectedly, the major digestion product from colominic acid (DP10-20) that had been reduced with NaB3H4 was DP3 and not DP4 (Fig. 3c), although smaller amounts of DP4 and DP1,2 were present. These results show that sialyl oligomers terminating in sialitol perturbed Endo-N catalysis and yielded different reaction products than oligomers terminating in sialic acid. As described below, reduction also changes the minimum substrate size from DP5 to DP6. No significant difference was seen between K1F phage-bound and -soluble Endo-N with respect to the products of limit digestion (results not shown).
Minimum Substrate Size-The products of limit digestion of poly-a-2,8-sialic acid suggested that the smallest substrate cleaved by Endo-N was DP5, yielding primarily the tetramer, DP4 (Fig. 3b). However, some DP5 remained in the limit digest of colominic acid that had been reduced with NaB3H4, suggesting that the minimum substrate size for a sialyl oligomer terminating in sialitol was DP6. DP5 is not seen in Fig. 3c because it was present in much smaller amounts than the major limit digestion product (DP3) derived from reduced colominic acid. In the example shown, DP5 contained 2370 cpm. T o determine directly the minimum number of sialyl units required to serve as a substrate, U-14C-labeled sialyl oligomers and sialyl oligomers terminating in 3H-labeled sialitol (labeled by reduction with NaB3H4) were purified as described under "Experimental Procedures." Each oligomer of defined length was then incubated overnight at 37 "C with an excess of Endo-N, and the reaction products were analyzed by HPLC. Minus enzyme controls showed no degradation. Under these conditions, reduced DP5-03H ((NeuNAc),-NeuNAc-03H) (Fig. 4a) was not hydrolyzed by Endo-N, whereas reduced DP6-03H and D7-03H were (Fig. 4, b and  c), again giving DP3-03H as the major digestion product. However, when nonreduced U-"C-labeled oligomers were used, different results were obtained. As shown in Fig. 4d, DP4 was not cleaved by Endo-N, but DP5 was, yielding DP4 and DP1 (Fig. 4e). Whereas the K , for DP5 was not determined, the results shown in Fig. 4e suggest that it is a relatively poor substrate since complete hydrolysis did not occur, even after an overnight incubation. Cleavage of DPlO (Fig. 4f) again gave close to the expected molar distribution of products with some DP5 remaining unhydrolyzed. Analysis of the digestion products of DP6-8 gave similar results (not shown). We conclude from these results that the minimum substrate size for Endo-N is DP5. DP5 and higher oligomers were cleaved by Endo-N to give DP4 and sialic acid as major products (Fig. 4e). In contrast, reduced DP5 (DP5-03H) was not cleaved by Endo-N (Fig. 4a), but DP6-03H and higher reduced oligomers were hydrolyzed to DP3-03H (Fig. 4, b and   c). Thus, the presence of a terminal sialitol changed the minimum substrate size from DP5 to DP6 and also appeared to alter the Endo-N cleavage pattern. Confirmation of this conclusion was provided as shown in Fig. 4 (g-i). In these experiments, 14C-labeled DP5, which served as a substrate (Fig. 4e), was reduced with NaBH, to form [14C]DP5-OH and tested for its ability to serve as a substrate for Endo-N. As shown in Fig. 4g, this reduced oligomer no longer served as a substrate. Reduced DP6 ( [14C]DP6-OH) and DP7 ( [14C]DP7-OH) both served as substrates (Fig. 4, h and i ) . Interestingly, the proportion of product found as DP3 with both [14C]DP6-OH and [14C]DP7-OH was again markedly increased over the nonreduced oligomers. Therefore, these results demonstrate that a terminal sialitol residue not only changes the minimum substrate size but also changes the cleavage pattern from that expected for completely random, as seen with nonreduced substrate, to one that is biased toward producing DP3-OH. In addition, an examination of the digestion products of (NeuNAc),-NeuNAc-03H of DP5-16 shows that this bias was not restricted to oligomers that were close to the minimum substrate size (Table IV). Indeed, the relative proportion of observed products appears to be independent of oligomer length, up to at least DP16, with the exception of DP8 where relatively lower amounts of DP3 and higher amounts of DP4 were observed. This latter result suggests that oligomers containing -eight sialyl units may have a different conformation in solution.
The results of Endo-N digestion of different oligomers of polysialic acid clearly show that the miminum requirement for cleavage is c~-2,8-(NeuNAc)~. This is in contrast to previously described phage-bound endoneuraminidases that were reported to require a minimum of eight cu-2,8-linked sialic acid units (31,32). In the present study, comparison of reduced and nonreduced oligomers of sialic acid has demonstrated that reduction of the reducing terminus to sialitol changes the apparent substrate specificity of Endo-N. In a previous study (32), reduced oligomers of a-B&linked sialic acid were used to define the substrate specificity of a bacteriophage (PK1A)-bound endoneuraminidase. The possible influence of reduction of the substrates on enzyme specificity and activity was not addressed. The previous study by Finne and Makela (32) reported that their enzyme required a minimum of 8 sialyl residues and that cleavage produced a DP3, derived from the nonreducing terminus, and a DP5, derived from the reducing end. It was therefore concluded that digestion of brain polysialosyl glycopeptides by endoneuraminidase would leave 5 sialyl residues attached to the glycopeptide moiety (32). However, this conclusion may have to be reconsidered in light of the results presented here.
Kinetics of Endo-N Digestion of /U-'4CJPoly-~-2,8-sialic Acid-Examination of the products produced by the action of a glycohydrolase as a function of time can yield information about the mode of enzymatic action. At low Endo-N to substrate ( [U-14C]poly-cu-2,8-sialic acid, DP150-200) ratios, an apparent lag in the formation of limit digest products (DP2-4) was seen when the reaction products were examined by paper chromatography and radiochromatographic scanning (results not shown). Under these conditions, the increase in number of free reducing termini, measured by thiobarbituric acid, was linear, suggesting that the initial products of Endo-N action were intermediate size oligomers (10 < DP > 100) that were not resolved by paper chromatography. This would imply that the mechanism of Endo-N hydrolysis of polysialic acid was random since a processive mechanism (repetitive hydrolysis from one end) would produce limit digest products during the initial phase of the reaction. To examine the mechanism of Endo-N catalysis, a detailed kinetic analysis of the products of Endo-N hydrolysis of [U-'4C]poly-cu-2,8-sialic acid was carried out (Fig. 5). When Endo-N digestion of substrate was 75% complete (4 min, Fig. 5a), only oligomers of DP3-8 were present. This suggested either that intermediate size oligomers were preferred substrates or that Endo-N catalysis might be processive with intermediate length oligomers (10 < DP > 50). An examination of the relative rates of disappearance of DP5-8 (Fig. 5, 6-e) supports the earlier conclusion that the smaller oligomers were poorer substrates. These results also confirm the conclusion that DP5 was the minimum substrate size for Endo-N since the major product of limit digestion was DP4 (Fig. 5, d and e).
Partial Digestion Products of Different Length Oligomers of (NeuNAc) ,-NeuNAc-03H-To differentiate whether the products of partial digestion of [U-'4C]polysialic acid noted above were due to a processive mechanism or to preferential hydrolysis of intermediate oligomers, the products of partial Endo-N digestion of DP7-03H to DP14-03H were examined (Fig. 6). DP7-03H and DPS-03H were degraded directly to DP3-03H without the formation of intermediate species.
However, the formation of intermediates became more apparent with the longer chain oligomers (DP10-03H to DP14-03H). This shows that for the intermediate size oligomers, the mechanism of Endo-N hydrolysis is random and not processive. Moreover, for the longer oligomers (DP10-03H to DP14-03H), cleavage is primarily on the distal (nonreducing end) of the terminal sialitol, in contrast to the preferential cleavage on the proximal (reducing end) of the terminal sialitol observed with DP7-03H to DP9-03H. This is additional evidence consistent with the conclusion that conformational differences may exist between sialyl oligomers when their size approaches a DP9, 10. The presence of a terminal sialitol influences Endo-N action even with intermediate length oligomers where presumably binding and hydrolysis near the reducing terminus is hindered due to the presence of sialitol. However, the apparent random mechanism of Endo-N hydrolysis under these conditions is not due to the terminal sialitol since partial Endo-N digestion of nonreduced ["C] DP12 also formed intermediate sized oligomers (Fig. 7). Model of Endo-N Catalysis-Several models have been proposed to account for the binding and subsequent hydrolysis of oligomeric substrates by depolymerases (50)(51)(52)(53)(54)(55)(56)(57). In these models, the active site of the enzyme is postulated to consist of a series of subsites, each of which is capable of binding a single carbohydrate moiety. These models can be used to predict time-dependent changes in the distribution of products, for example, amylase action (58, 59), when the subsite binding affinities have been determined by other methods or the subsite binding affinities can be determined by measuring product distribution as a function of time (60,61). Whereas a detailed, quantitative treatment of Endo-N action in this manner is beyond the scope of this study, a schematic model of Endo-N and substrate interaction (Fig. 8) is instructive and allows some qualitative assessments to be made.
In the case of the nonreduced substrate (Fig. Sa), interaction of sialic acid residues with enzyme-binding subsites is approximately equipotential, although occupancy of subsite 1 is preferred to subsite 2 since cleavage of [l4CJDP5 gives primarily DP4 and sialic acid (Fig. 4e). Additional, long range interactions are probably also important since sialic acid oligomers of intermediate length appear to be better substrates than shorter oligomers (Table 111). In the case of oligomers up to DP9 with a terminal sialitol (Fig. 86), the presence of the alcohol causes preferential binding of the terminal sialitol at subsite 7, favoring cleavage predominantly at (NeuNAc),-(NeuNAc),-NeuNAc-OH. Since DP6-OH is cleaved, this could indicate that occupancy of subsite 2, al-

Conclusions
The purification procedure reported here provides sufficient quantities of highly purified (homogenous) soluble Endo-N, free of proteases and exoneuraminidase activities, suitable for future studies on the biosynthesis and biological function of poly-a-2,8-sialosyl-containing glyco-conjugates. The following considerations are relevant to the future use of Endo-N as a probe for polysialic acid. Since Endo-N catalysis is random and long polymers are substantially better substrates than short ones, high enzyme concentrations and long incubation periods may be necessary to achieve complete digestion of oligomers that are only slightly larger than the minimum substrate size (DP5). Since the minimum substrate size is DP5, the release of sialic acid from a homooligomer by Endo-N would indicate the presence of an oligomer of >DP4. However, the lack of release of sialic acid does not necessarily indicate a polymer of <DP5 since Endo-N could be sterically hindered due to an attached reducing end. Indeed, here it was found that merely the presence of a terminal sialitol changes both the apparent minimum substrate size and the cleavage pattern obtained. Experiments using Endo-N to study the biosynthesis of polysialosyl units in neural cell adhesion molecules and to probe the function of polysialic acid in a variety of biological systems are currently in progress in a number of laboratories.