Transfer of glycerol by Endo-beta-N-acetylglucosaminidase F to oligosaccharides during chitobiose core cleavage.

N-Linked oligosaccharides, when hydrolyzed by glycerol-containing preparations of endo-beta-N-acetylglucosaminidase (Endo) F from Flavobacterium meningosepticum were found to have glycerol attached to their reducing ends. The absence of a reducing end was confirmed by high-field 1H NMR spectroscopy, and the incorporated glycerol was verified through mass spectrometry and collisionally activated decomposition fast atom bombardment/mass spectrometry/mass spectrometry techniques. Periodate oxidation of [1(3)-14C]glycerol-labeled oligosaccharides indicated glycerol was glycosidically linked via its 1(3) carbon to the C1 of the reducing end N-acetylglucosamine. In a second, less favored reaction, the glycerol glycoside was hydrolyzed by Endo F using water as the terminal nucleophile, thus regenerating the N-acetylglucosamine reducing end. Glycerol could be removed from Endo F preparations without affecting enzyme stability, and chitobiosyl core hydrolysis in its absence provided intact oligosaccharides with normal N-acetylglucosamine reducing ends. The incorporation of labeled glycerol may provide a useful method for monitoring of Endo F release of oligosaccharides.

PNGase F hydrolyzes the amide bond between asparagine and the di-N-acetylchitobiose unit of both high mannose and complex oligosaccharides to yield a peptide-bound aspartic acid and a liberated 1-aminooligosaccharide. The subsequent nonenzymatic release of NH: yields N-acetylglucosamine at the reducing end of the intact oligosaccharide.
During studies on the substrate specificity of Endo F preparations, we observed that, depending on the reaction conditions and the glycopeptide hydrolyzed, some or all of the released oligosaccharides failed to incorporate tritium on reduction with alkaline NaB3H4. The work reported in this paper provides an explanation for this effect by demonstrating that glycerol, a stabilizing component added to some commercial Endo F preparations, becomes attached to the C1 carbon of the reducing end GlcNAc during oligosaccharide hydrolysis. These results appear to explain the observation by other investigators (4) that Endo F-released oligosaccharides from a variety of sources have no reducing end. This finding may prove useful for the single-step labeling of oligosaccharides during endoglycosidic release from glycopeptides and glycoproteins.

Materials
Endo F' prepared from cultural filtrates of Flavobacterium meningosepticum by the method of Elder and Alexander (1) is a mixture of two endoglycosidase activities; Endo F and ) and by Grant CHE821164 to the Midwest Center for Mass Spectrometry, a National Science Foundation Regional Instrumentation Facility. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence and requests for reprints should be addressed.

Methods
Endoglycosidase Digestions-Hydrolysis of Asn-oligosaccharides was carried out at 25 "C or 37 "C as indicated using Endo F or H at concentrations listed in the legends to the figures and tables. The buffer used was 50 mM sodium acetate, pH 5.5. One milliunit of Endo F or H is defined as that amount of enzyme which will hydrolyze 0.5 mM Man,GlcNAcAsn-dansyl at 1 nmol/min at 37 "C and pH 5.5 Compositional Analyses-The content of mannose in oligosaccharides was determined by a scaled-down version (7) of the phenolsulfuric acid assay (8) using mannose as a standard. N-Acetylglucosamine was determined as glucosamine on the amino acid analyzer (9) in 5-nmol samples of oligosaccharide hydrolyzed in sealed, Nz-flushed vials with 2 N HCl for 16 h in vacuo at 108 "C. Glucosaminitol was present as an internal standard. Asparagine was determined on the amino acid analyzer as aspartic acid in samples (5 nmol) hydrolyzed with 6 N HCl for 22 h in uacw at 108 "C in sealed, Nn-flushed vials. Norleucine was present as an internal standard. The recovery of glucosamine and aspartic acid was greater than 96%.
Chromatography-Oligosaccharide products of endoglycosidase digestions were resolved on a calibrated Bio-Gel P-4 (-400 mesh) column (0.9 x 195 cm) at room temperature with an eluant of 0.1 N acetic acid in 1% 1-butanol; 0.5-ml fractions were collected at a flow rate of 7 ml/h. Paper chromatography for the resolution of cleavage products from dansylated oligosaccharides was performed ascendingly on Whatman No. 3MM paper using 1-butanokethanokwater (2l:l) as the solvent.
Detection Procedures-Asn-oligosaccharides were dansylated as described (10) and desalted on a Sephadex G-10 column (0.9 X 100 cm) using 0.1 N acetic acid as the eluant. Dansylated oligosaccharides and products were located in column profiles and on paper chromatograms with a UV lamp (360 nm). Mannose was located in column profiles by the phenol-sulfuric acid assay described above. The asparagine moiety of Asn-oligosaccharides was located in column profiles as the fluorescent product after reacting with fluorescamine (11). Radioactivity was eluted from excised paper chromatography spots with 1 ml of HzO for 1 h at 37 'C in 20-ml scintillation vials, after which 10 ml of Beckman HP/b scintillation fluor was added. Radioactivity (disintegrations/min) was quantitated on a Beckman LS7500 spectrometer equipped with a dual label (3H/"C) data reduction program.
'H NMR-Oligosaccharides were exchanged twice by rotary evaporation with 99.8% D20, once by lyophilization from 99.96% D20, and stored over P200 for several days. Samples were dissolved at a concentration of 1-2 mM in 99.996% 40, and spectra were recorded at 298 K using the 500 MHz spectrometer at the Northeast Regional NSF-NMR Facility at Yale University. Chemical shifts were compared to equimolar acetone added as an internal marker (2.225 ppm compared to 4,4-dimethyl-4-silapentanesulfonate). Parameters were as described previously (7).
Fast Atom Bombardment/MS-All mass spectra were obtained on a Kratos (Manchester, UK) MS-50 Triple Analyzer (12), which consists of a high resoluion MS-I, with Nier-Johnson forward geometry, and an electrostatic analyzer as MS-11. The fast atom bombardment source is of standard Kratos design equipped with an Ion Tech saddle field atom gun (Teddington, UK). Samples (1-5 pg) in methanol (1 pl) were added to 1 pl of the matrix (dithiothreitol/dithioerythritol for positive ions and triethanolamine for negative ions) on the copper probe tip of a direct insertion probe. Bombardment of the sample with 6-8-keV xenon atoms desorbs the preformed ions. The collisionally activated decomposition MS-MS spectra were obtained by selecting the desired ion with MS-I followed by collison of the selected ion with helium (sufficient to reduce the primary ion beam intensity by 50%) in a collision cell located between MS-I and MS-11. The collisions transform some of the ions' translational energy into internal energy. This additional internal energy initiates fragmentation. The collisionally activated decomposition spectra were acquired by scanning MS-11, and multiple scans were signal-averaged and processed using software written at the Midwest Center for Mass Spectrometry. Full scan fast atom bombardment spectra were acquired using standard Kratos software. The accurate mass measurement was determined by peak matching with glycerol standard at a resolution of 10,000. The tubes were boiled for 5 min and cooled on ice. After centrifugation, the resulting precipitates were extracted into 1 ml of toluene, and duplicate 0.5-ml aliquots were taken for radioactivity measurements.

RESULTS AND DISCUSSION
During preliminary experiments: we observed that ohgosaccharides released from MaGGlcNAc,Asn by Endo F were T. H. Plummer, Jr., unpublished observations. not reducible with NaBH,. Based on our knowledge of the activity of Endo H (10, 13), this was a highly unexpected result and suggested, at first, a possible difference in the mechanism of hydrolysis by the two endoglycosidases. Detailed comparative studies were conducted using the homogeneous asparagine-oligosaccharide, Man5GlcNAczAsn, in order to simplify product identification.
As expected (lo), Endo H quantitatively hydrolyzed Man5GlcNAczAsn to ManBGlcNAc (peak l a ) and AsnGlcNAc (peak ZZa) (Fig. lA). However, the profile of the reaction products from hydrolysis of Man5GlcNAczAsn by Endo F was clearly different, As shown in Fig. lB, it contained about 30% unhydrolyzed MansGlcNAczAsn (peak Zb), the expected products Man,GlcNAc (peak ZZZB) and AsnGlcNAc (peak Zvb), and, unexpectedly, a new peak (peak Zlb) in a position where oligosaccharides with the composition of Man7GlcNAc and Man5GlcNAcz would elute. However, since the parent oligosaccharide could not yield Man7GlcNAc, and Man5GlcNAcz cannot be formed by the PNGase F activity in Endo F (2, 3), we concluded that a new oligosaccharide of larger mass had been generated during the initial cleavage.
To investigate this species further, the content of Asn, GlcNAc, and Man in the four peaks in Fig. 1B was determined (Table I)   eluted with different sizes on the Bio-Gel P-4 column (Fig.  l), were treated with NaB3H4 to test for reducing ends. Peak Ilb failed to incorporate label, while peak IIIb (Man5GlcNAc) was quantitatively reduced to Man5GlcNAc-[3H]ol as anticipated (data not shown).
' H NMR Studies on Peak ZZb and ZZZb Oligosaccharides-Peak IIb and IIIb oligosaccharides were exchanged with DzO and examined by 500 MHz 'H NMR spectroscopy to identify peak IIb through the resonance signatures of the C1 anomeric protons. Fig. 2A shows the anomeric proton region of the spectrum for peak IIIb, which reveals the C1-H chemical shifts characteristic of the Man5GlcNAc structure depicted (14). Fig. 21B shows the spectrum of the oligosaccharide in peak IIb, which is lacking resonances 1 and 2 seen in Fig. 2A.
These are the reporter resonances for the alp conformation of the reducing end N-acetylglucosamine C1-H which is nor-into Released Oligosaccharides mally formed on hydrolysis of the core GlcNAcpl,4GlcNAc linkage by either Endo F or Endo H. This proton appears in Fig. 2B at 4.539 ppm, which represents an upfield shift of 0.176 pprn from its usual position at 4.715 ppm in unhydrolyzed Man,GlcNAc,Asn (14). This suggests that the GlcNAc on the Endo F-released oligosaccharide is still in a glycosidic linkage. Further NMR evidence for an aberrant C1 terminus on the peak IIb oligosaccharide is the failure of the upper arm &linked mannose (residue 7) to fold back and interact with the core GlcNAc in a manner producing the split (alp) resonance normally seen for the C1-H at 5.098/5.077 ppm (7,14). This is reflected in the C2-H of residue 7 as well, by the loss in anomericity at 4.061 ppm. The C2-H on the @&linked core mannose (residue 3) at 4.257 ppm also shows a loss of anomericity on comparing the Endo H and Endo F products (Fig.   2, A and B).

MS Studies on Peaks ZZb and ZZZb-The larger size on Bio-
Gel P-4 of the oligosaccharide in peak IIb relative to that in peak IIIb (Fig. l), coupled with its failure to display a reducing end either by NaB3H4 reduction or by 'H NMR (Fig. 2B), suggested that a glycosidically linked blocking group had been introduced during enzyme cleavage. To analyze for the presence of a blocking group, peak IIb and IIIb oligosaccharides were compared by mass spectrometric techniques. The presence of glycerol on the oligosaccharide's reducing end was surprising, but was consistent with the available data: a glycerol residue would add mass and would be liberated during acid hydrolysis, but it would be detected neither by amino acid analysis nor by the phenol-sulfuric acid assay ( Table I). The available glycerol appeared to be that added to the Endo F mixture as a stabilizing agent. f4C]Glycerol Incorporation Studies-To show that glycerol was being incorporated into the oligosaccharide from the reaction mixture, [1(3)-'4C]glycerol was added to a comparable Endo F incubation and the radiolabeled peak Ilb oligosaccharide was isolated on Bio-Gel P-4 as in Fig. 1 (data not  shown). The specific activity of the oligosaccharide, based on 5 mannoses, was equal to that of the input labeled glycerol (6000 dpm/nmol) indicating that each oligosaccharide had incorporated one glycerol. Since glycerol could be incorporated via a Cl(3) or a C2 linkage, periodate oxidation was used to distinguish between the two possible linkages depicted in Scheme I. Positive ion spectra were generated using a dithiothreitol/dithioerythritol matrix (panels A and C), while negative ion spectra were generated with triethanolamine as the matrix (panels B and D). The spectra show that the peak Ilb oligosaccharide is Man5GlcNAc with an added mass of 74 Da (CaH,O*), which is consistent with glycerol replacing OH on the GlcNAc reducing end.  I, A).
Because the Endo F was known not to be pure, it was important to show that glycerol incorporation occurred during Endo F cleavage of the oligosaccharide's chitobiosyl core, and not by subsequent addition to free MandGlcNAc by a contaminating enzyme activity distinct from Endo F. That this was not the case was shown by incubating the partially purified Endo F mixture with Man5GlcNAc and [1(3)-'4C]glycerol under the same hydrolysis conditions described in Fig. 1. Chromatography of the products revealed no glycerol incorporation and that the ManSGlcNAc was unaltered.
To show that the glycerol-oligosaccharide was due to Endo F, a highly purified preparation of enzyme, which provided a single band on SDS gels and was free of detectable PNGase F activity, was used to hydrolyze both Man5GlcNAc2Asn and ovalbumin glycopeptide peak C, a mixture of species with the composition Man6GlcNAc4Asn (6, 13, 15). By means of highresolution TSK HW-40s column chromatography, we first observed that incorporation of radioactivity was a time-dependent phenomena: glycerol was transferred very early and completely during the reaction, but was removed later from the oligosaccharide. The kinetics of Man6GlcNAc,Asn-[3H] dansyl cleavage (1 mM) by pure Endo F (100 milliunits/ml) and uptake of labeled glycerol (9.2% (w/v) = 1 M) are shown in Fig. 5. In the initial phase of hydrolysis, there is clearly a 1:l molar ratio between oligosaccharide cleaved and glycerol incorporated. As the reaction nears completion, however, glycerol begins to be removed from the oligosaccharide. After the last sample was taken, the chromatogram was developed ascendingiy with 1-butanokethanolwater (2:l:l) as the solvent. Glycerol and dansyl-AsnGlcNAc move with R p of about 0.6, while unhydrolyzed substrate and glycerol-labeled oligosaccharide remain at the origin. The origin spots are excised, and the label is eluted and counted using a 3H/14C dual label disintegrations/min data reduction program. released oligosaccharides. The range of glycerol concentrations tested (1% v/v to 15% v/v) has little effect on the proportion of blocked ends, although at higher levels of glycerol (10-15% v/v) the hydrolysis reaction is slowed somewhat (20%). In the absence of glycerol, hydrolysis produces oligosaccharides with normal reducing C1 termini. Since some commercial Endo F preparations are packaged in 50% (v/v) glycerol, a 150 dilution (= 1% v/v = 1.26% w/v = 0.14 M) is sufficient to produce a high proportion of blocked ends. In fact, the experiment in Fig. 1 contained only 1% (v/v) glycerol.
Clearly, there are two potential problems introduced by the addition of glycerol to the oligosaccharides. The first is the obvious loss of a reducing end. The second is the aberrant increase in apparent size of the oligosaccharide by an amount equal to two mannoses or one N-acetylglucosamine, which would affect the proper evaluation and characterization of metabolically labeled products. Endo F does not require glycerol for stability, and preparations stored for several months at 4 "C retain full activity in its ab~ence.~ Thus, glycerol may be removed from commercial preparations of Endo F mixtures without affecting the enzyme.
The two most significant factors in generating glycerolblocked ends appear to be the level of Endo F added and the structure of the oligosaccharide being used as substrate. Oligosaccharides which are poorer substrates for Endo F (ie., Man6GlcNAc4Asn relative to Man5GlcNAc2Asn) tend to retain a higher level of glycerol-blocked ends. Lowering the amount of Endo F greatly favors the retention of glycerol by the oligosaccharides. It is not clear yet what the reaction mechanism is, but one possibility is that Endo F incorporates glycerol to the oligosaccharide in a transglycosidase type of ' A. L. Tarentino, unpublished observations. reaction during displacement of the AsnGlcNAc. Subsequently, in a less favored reaction, Endo F acts as a hydrolase to remove the glycosidically linked glycerol from the oligosaccharide employing water as a terminal nucleophile. Since the rate of regeneration of the reducing end depends on the level of Endo F and the oligosaccharide, the kinetics of glycerol removal would have to be determined empirically for each set of conditions.
The off-rate for glycerol implied in the experiment in Fig.  5 could be an underestimate because Endo F may continually recycle glycerol onto the oligosaccharide ends in competition with its hydrolysis by water. To test this possibility, 40 nmol of MaQGlcNAczAsn was hydrolyzed with purified Endo F in the presence of [3H]glycerol as in Fig. 5, and the Man6Glc-NA~~-[~H]glycerol was isolated on Bio-Gel P-4.
Parallel reactions containing labeled oligosaccharide and Endo F, fl.1 M glycerol, were monitored by paper chromatography for [3H]glycerol excision. Under these conditions, unlabeled glycerol did not impede Endo F hydrolysis of the [3H]glycerol from the oligosaccharides. Both reactions had identical time courses, revealing loss of 75% of the label by 30 min and 90% by 1 h. Thus, the slow loss of glycerol from oligosaccharides in a complete reaction (Fig. 5) may not be due to cycling of glycerol, but rather to the formation of an enzyme-substrate activation complex which stabilizes the glycerol-oligosaccharide. Alternatively, the released AsnGlcNAc may be a competitive inhibitor of glycerol hydrolysis. Future experiments will address more thoroughly the mechanism of this reaction.
Finally, experiments show that Endo H will also incorporate labeled glycerol into the oligosaccharides, although the efficiency appears much less than observed with Endo F. So far, no more than 65% of the reducing ends have been found to be blocked by Endo H, and this occurs very early in the reaction. Nevertheless, the utilization of glycerol in preference to water during primary cleavage of the GlcNAc-GlcNAc glycosidic bond by both Endo F and Endo H suggests that these enzymes recognize a common structural feature of the original substrate for which glycerol can partially substitute. Such a determinant might be the three carbon (C3-C4-C5 or C4-C5-C6) glycerol sequence of the asparagine-proximal GlcNAc residue.