Methods and Reviews


Associates and sponsors are encouraged to share ideas, applications research, applications experience, and technical problems by contributing to this column. In addition, short reviews covering aspects of biotechnology that are of interest to members are also encouraged. All contributions will be peer-reviewed, and the deadline for submitting articles for the June issue is May 1, 1997. Periodically articles will be solicited from non-members by the Editorial Board. Contributions may be mailed or faxed to the Editor, Clayton Naeve, St. Jude Children's Research Hospital, Center for Biotechnology, 332 N. Lauderdale St., Memphis, TN 38105-2794, Fax: (901) 495-2945, E-mail: clayton.naeve@stjude.org. Contributions to this column do not necessarily reflect the views of the Editorial Board or the Association.


Quantitative Monosaccharide Analysis:

A Multi-Center Study

 

The ABRF Carbohydrate Analysis Research Committee

R. Reid Townsend, Adriana Manzi, Roberta K. Merkle, Michael F. Rohde, Michael Spellman, Alan Smith, and Steven A. Carr

 

Introduction

Glycosylation is a major post-translational modification found on cytoplasmic, secreted, and membrane proteins. Glycosylated proteins are often identified by their appearance as diffuse bands during SDS-PAGE and by changes in their migration after treatment with endoglycosidases or amidases such as PNGase F. This approach for identifying glycoproteins gives negative results when (i) the N-glycans are not removed by the enzymes used, (ii) there is predominantly O-glycosylation, or (iii) only a glycosylphosphatidylinositol anchor is attached. Gel detection methods based on susceptibility of sugar chains to periodate oxidation were originally used to identify glycoproteins in gels, but this required about 100 mg of sample (1). If a protein is modified with oligosaccharides that are recognized by a lectin, the presence of glycosylation and some structural information can be deduced from blotting studies using about 100 ng of sample (2). A more universal and direct approach is to determine the monosaccharide content of a protein sample. Quantitative monosaccharide composition analysis (i) provides the molar ratio of individual sugars to protein, (ii) suggests the presence of oligosaccharide classes (e.g., N- versus O-glycosylation), (iii) provides the first step in designing a structural elucidation strategy, and (iv) can be used as a measure of production consistency for recombinant glycoprotein therapeutics.

 

Since the 1958 report by McInnes et al. (3), a variety of GC methods have been developed to quantify monosac-charides; the most useful involve coupling gas chromatography and mass spectrometry for positive peak identification. By the 1970s, GC-MS was used as a major tool for determining composition, as well as for linkage analysis after permethylation (4). Methyl ethers (5, 6), acetates (7), and isopropylidene derivatives (7) of monosaccharides have been used for GC, but alditol acetates (8, 9) and per O-trimethyl silyl ethers (TMS) (10) have become the most widely used derivatives. By the early 1980s, pre- and post-column derivatization methods to quantify monosac-charides by HPLC had been developed (11). In the early 1990s, electrochemical detection with high-pH anion-exchange chromatography, which does not require derivatization, was in widespread use (12).

 

The goal of the first study of the ABRF Carbohydrate Analysis Research Committee (CARC) was to determine the ability of research laboratories and core facilities to identify and quantify carbohydrates in protein samples. Two test samples, one glycosylated and the other containing only protein, were distributed to more than 300 laboratories for analysis. We requested that study participants use any techniques available to them to ascertain which sample was glycosylated and, if their techniques permitted, to determine the quantitative monosaccharide composition of the glycosylated sample. In addition, the members of CARC undertook iterative analyses of neutral and amino sugars in the glycosylated test sample to pinpoint errors associated with monosaccharide standard preparation, hydrolysis conditions, sample handling, and instrumentation.

 

Materials and Methods

Preparation of test samples Test sample A was ultrapure bovine serum albumin from Calbiochem, and test sample B was bovine fetuin from Gibco, purified by the Spiro method. Fetuin was chosen because it is one of the most well-characterized glycoproteins available. Both test samples were lyophilized from water in 0.5 mg aliquots. The total amount of protein/mg sample was determined by amino acid analysis and provided the basis for calculating the molar ratio of monosaccharides to protein. Test sample B was found to contain 25 nmol of fetuin/nominal mg, based on the average of the values for Val and Ala.

Monosaccharide analysis by CARC using HPAEC-PAD with different instrument configurations Test samples were dissolved in about 1 ml of water, and the exact volume was determined by weight. After initial analyses using hydrolysis conditions practiced in CARC member laboratories (see legend of Table 1), the uniform hydrolysis conditions adopted by the CARC were 2 M TFA for 3 hours at 100°C for neutral and amino sugars and 2 M acetic acid at 80°C for 3 hours for sialic acids. Monosaccharide standards were prepared as previously described (13). Analyses were performed in the CARC member laboratories as described below.

 

Laboratory A: Monosaccharides were analyzed on a Dionex BioLC system equipped with a Dionex CarboPac PA1 column (4 x 250 mm) and a solvent-compatible electrochemical cell (gold working electrode and Ag/AgCl reference electrode). Neutral and amino sugars were analyzed using water as E1, 0.1 M NaOH as E2, and 0.2 M NaOH as E3. The proportions of E1:E2:E3 at different times were as follows: t = 0, 13:87:0; t = 22 min, 13:87:0; t = 23 min, 0:0:100; t = 33 min, 0:0:100; t = 34 min, 13:87:0; and t = 55 min, 13:87:0. The detector response was set at 3 s, and the detector potentials were: E = 0.05V, t1 = 420 ms; E = 0.75V, t2 = 180 ms; E = -0.15 V, t3 = 360 ms. The autosampler was a Thermo-Separations SP8880 equipped with a stainless steel needle.

 

Laboratory B: Monosaccharides were analyzed on a Dionex DX500 system equipped with a Thermo-Separations AS3500 autosampler containing a stainless steel needle and with an ED40 electrochemical detector with a cell containing a thin, 0.1 mm gasket. Analyses were performed using a CarboPac PA1 column (4 x 250 mm) and the following gradient program and eluants (E1, water; E2, 200 mM NaOH; E2 in E1): t = 0, 8% E2; t = 25 min, 8% E2; t = 27 min, 100% E2; t = 37 min, 100% E2; and t = 39 min, 8% E2. Injections were made every 50 min. The pulse potentials on the ED40 detector were: t = 0, E = 0.05V; t = 0.20, E = 0.05V (begin); t = 0.40, E = 0.05V (end); t = 0.41, E = 0.75V; t = 0.60, E = 0.75V; t = 0.61, E = -0.15V; t = 1.00, E = -0.15V. Instrument control and data processing were done using Dionex PeakNet Software. The 50% NaOH solution and HPLC grade, glacial acetic acid were from Fisher Scientific, and sequencing grade trifluoroacetic acid in 1 g ampules was from Pierce. MilliQ water (18.2 mohms) was used to prepare all eluants. Monosaccharide standards were from Sigma Chemical Co., and sialic acids (Neu5Ac and NeuGc) were from Boehringer Mannheim.

 

Laboratory C: The chromatograph was a Dionex GlycoStation equipped with a gradient pump, a PAD-II, and a Thermo-Separations AS3500 autosampler with a stainless steel needle, all under software control. The pulse potentials for the PAD-II were E1 = 0.05 V, t1 = 480 ms (Range 2, position 5); E2 = 0.60 V, t2 = 120 ms (position 2); and E3 = -0.60 V, t3 = 60 ms (position 1). The time constant was set to 3 s. The Dionex "basic" PAD cell, equipped with a thin gasket, was used. Monosaccharides were separated using a CarboPac PA1 column (4 x 250 mm) equipped with a guard column. The monosaccharides were analyzed using the following gradient program and eluants (E1, water; E2, 200 mM NaOH; E2 in E1): t = 0, 8% E2; t = 25 min, 8% E2; t = 27 min, 100% E2; t =37 min, 100% E2; and t = 39 min, 8% E2. Injections were made every 50 min. Monosac-charide standards were obtained from Pfansteil Laboratories, sodium hydroxide (50%) and acetic acid were from Fisher Chemical Co., and trifluoroacetic acid was from Sigma Chemical Co.

 

Laboratory D: The samples were analyzed with a Dionex BioLC 300 equipped with an AS3500 autosampler using a stainless steel needle. The column was a CarboPac MA1, and analytes were eluted isocratically with 500 mM NaOH. The following settings were used on the PAD II: E1 = 0.05 V, t1 = 480 ms; E2 = 0.60 V, t2 =120 ms; and E3 = -0.60 V, t3 = 60 ms.

 

Monosaccharide analysis by CARC using GC, Laboratory E: Test samples A and B were analyzed for monosaccharide composition after methanolysis, re-N-acetylation, and preparation of trimethylsilyl (TMS) derivatives as previously described (14, 15). Gas chromatography-mass spectrometry was performed on a Hewlett Packard 5890 GC coupled to a 5970 MSD instrument. Monosaccharide derivatives were separated using a J&W Scientific DB1 fused-silica capillary column. TMS methyl glycosides were prepared after methanolysis in 1 M methanolic HCl, followed by re-N-acetylation of amino sugars with pyridine and acetic anhydride. Monosaccharides were from Sigma or Pfanstiel, TFA and Tri-Sil were from Pierce, and the methanolic-HCl kit was from Supelco.

 

MALDI-MS of bovine fetuin Analyses were conducted with a Fisons VG TofSpec single reflectron mass spectrometer, using a 337 nm pulsed nitrogen laser and 25 keV accelerating potential. The instrument was externally calibrated in the linear and reflectron modes using bovine serum albumin. Aliquots of fetuin and glycosidase digests were mixed (1:2 v/v) with sinapinic acid (10 mg/ml in 1:1 CH3CN:water), spotted onto stainless steel targets, and allowed to air-dry before analysis.

 

Glycosidase digestions To remove Neu5Ac, about 200 pmol of fetuin was digested overnight at 37°C with 0.1 units of Vibrio cholerae neuraminidase (Calbiochem) in 100 ml of 50 mM NaOAc, 154 mM NaCl, 4 mM CaCl2, pH 5.5. To cleave Gal-GalNAc linked to Ser or Thr residues, 9.2 milliunits (1.5 ml) of O-glycanase (Genzyme Corp.) were added to the neuraminidase-digested sample, the pH was adjusted to about 7 with dilute sodium hydroxide, and the digestion was allowed to proceed overnight at 37°C. To remove the entire N-linked carbohydrate chains (with conversion of attachment sites from Asn to Asp), fetuin was incubated at pH 7.5 at 37°C overnight with PNGase (1.5 ml of a 0.25 mg/ml solution, glycerol-free, from Genzyme Corp.). After each glycosidase treatment, aliquots containing 1-3 pmol of glycoprotein were applied directly to MALDI-MS sample targets for analysis.

 

Results and Discussion

The first goal of this study was to assess the ability of laboratories to determine which test sample was glycosylated and to quantify the constituent monosaccharides. Approxi-mately 300 research laboratories and core facilities received the two test samples, A and B; 29 study participants, other than CARC members, returned their results. All external study participants correctly identified test sample B as the glycoprotein. The methods for this included mass spectrometry (signal width of the undigested sample or mass changes after glycosidase digestion), amino acid analysis (detection of GlcN), and monosaccharide analysis using HPLC, GC, or gel electro-phoresis (detection of GlcN). In 14 cases, test sample B was identified as the glycoprotein by amino acid analysis, alone or in conjunction with amino sugar or colorimetric tests. To identify both test samples, participating external laboratories analyzed their own amino acid compositions further with PROPSEARCH (URL address: http://www.embl-heidel-berg.de/aaa.html) (16). In 13 of the 14 cases, test samples A and B were correctly identified by amino acid compositions as bovine serum albumin (first or second positions in the PROPSEARCH rankings) and bovine fetuin (top three positions), respectively.

 

Monosaccharide analysis by the CARC using HPAEC-PAD The second goal of this study was to assess how well the constituent monosaccharides of test sample B could be quantified. As part of this study, the members of the CARC each determined the neutral and amino sugar content of test sample B using the glycoprotein hydrolysis and HPAEC-PAD conditions currently in use in each laboratory (A-D). Table 1 shows the values obtained among the CARC laboratories varied over a 2-fold range: GalN, 46-96; GlcN, 350-500; Gal, 240-520; and Man, 190-550 nmol/mg protein. To evaluate reproducibility within a single laboratory, one CARC laboratory repetitively analyzed test sample B, without making changes to its standard hydrolysis conditions or HPAEC-PAD instrumentation; the results from two determinations on triplicate hydrolyzates over 6 months were: GalN, 60 ± 6.6%; GlcN, 410 ± 8.6%; Gal, 320 ± 15%; and Man 195 ± 4.6% nmol/mg protein. The error for these latter determinations was 2- to 5-fold less than observed for inter-laboratory determinations (Table 1). At this point in the study, the range of values obtained by the CARC laboratories as a group was attributed to the differences in acid hydrolysis conditions (see legend of Table 1).

Table 1: Neutral and amino sugar analysis of test sample B by the Carbohydrate Analysis Research Committee (CARC) laboratories using their standard methods of hydrolysis and HPLC analysis. A, B, C, and D represent four laboratories of the CARC. The units are nmol sugar/mg protein.

 

Table 1: Neutral and amino sugar analysis of test sample B by the Carbohydrate Analysis Research Committee (CARC) laboratories using their standard methods of hydrolysis and HPLC analysis. A, B, C, and D represent four laboratories of the CARC. The units are nmol sugar/mg protein.

 Sugar

 A (1)

 B (2)

 C (3)

 D (4)

 Mean

 GalN

 65

 46

 96

 63

 60 ± 31%

 GlcN

 370

 350

 500

 480

 390 ± 25%

 Gal

 330

 240

 350

 520

 330 ± 24%

 Man

 225

 190

 220

 550

 270 ± 27%

 1 Determined after hydrolysis in 4 M TFA at 120°C for 1 hour.

 2 Determined after hydrolysis in 2 M TFA at 100°C for 4 hours.

 3 The amino sugars were analyzed after hydrolysis in 4 N HCl at 100°C for 3 hours. The neutral sugars were determined after hydrolysis in 2 M TFA at 100°C for 3 hours.

 4 Determined after hydrolysis in 5 M TFA at 100°C for 2 hours, followed by re-N-acetylation.
group was attributed to the differences in acid hydrolysis conditions (see legend of Table 1).

The analysis of neutral and amino sugars was repeated by the CARC laboratories using uniform hydrolysis conditions: 2 M TFA for 3 hours at 100°C. Trifluoroacetic acid has been used extensively to release glycoprotein monosaccharides (17), and previous studies had shown good agreement between observed and theoretical monosaccharide content in glycoproteins and oligosaccharides (18). Trifluoroacetic acid is completely volatile, leaving minimal contaminants for high sensitivity measurements. At least 3 hours of hydrolysis is required to de-N-acetylate GalNAc and GlcNAc. Removal of N-acetyl groups is necessary because only the free amino sugars completely separate from the neutral sugars (e.g., Gal, Man, and Glc) by HPAEC (18). However, standardizing the hydrolysis conditions had little effect on variability (Table 2). Also, in this analysis the values for Man and Gal from laboratories C and D weresignificantly lower than previously observed (compare Tables 1 and 2), and lower Man values from these laboratories were observed throughout the study. The Committee concluded that the initial glycoprotein hydrolysis conditions (Table 1) gave the same recoveries as the uniform hydrolysis conditions (Table 2) within the experimental error.

Table 2: Neutral and amino sugar analysis of test sample B using uniform hydrolysis conditions (1). A, B, C, and D represent four laboratories of the CARC. The units are nmol sugar/mg protein.

 Sugar

 A

 B

 C

 D

 Mean

 GalN

 60

 52

 87

 95

 74 ± 28% (31%) (2)

 GlcN

 430

 340

 550

 330

 410 ± 25% (25%)

 Gal

 390

 300

 210

 340

 310 ± 25% (24%)

 Man

 260

 220

 130

 220

 210 ± 27% (27%)

 1 Trifluoroacetic acid (2 M, 100°C for 3 hours).

 2 Values in parenthesis are percent standard deviations from Table 1.

The CARC next determined whether inter-laboratory variability was related to the preparation of monosaccharide standards. Unless rigorous protocols are followed for drying carbohydrates to prepare standards by weight (13), inaccuracies may occur. Analysis of neutral and amino sugar standards from each CARC laboratory (A, B, C, D, and E) on the same instrument and on the same day showed only slight differences in area responses (no greater than 15%), indicating that preparation of monosaccharide standards was not a major contributor to variability.

 

In the CARC laboratories, as in most laboratories using HPLC, the external standard method was used for calibration, with either untreated standards or standards that had been subjected to the same acid hydrolysis conditions as the samples. The rationale for using hydrolyzed standards is to compensate, in part, for monosaccharide destruction after release from the glycoprotein. Using the uniform hydrolysis conditions adopted by the CARC for neutral and amino sugars, CARC laboratories A-D found that the area responses for Fuc, Gal, and Man were not significantly different for hydrolyzed and non-hydrolyzed standards (Table 3). Unexpectedly, hydrolysis increased the electrochemical response for amino sugars by a factor of 2 (Table 3). In a subsequent experiment, an equal volume of 4 M TFA was added to the mixture of monosaccharide standards and the sample was injected. The amino sugar response was enhanced the same amount as observed when the standards were treated at 100°C for 3 hours. These results likely explain the higher amino sugar values consistently observed in laboratory C (Tables 1 and 2). This increase in amino sugar response was less in other laboratories. The lower peak areas were attributed to either selective loss of amino sugars through the injection path or reaction of free amino groups with
aldehyde functions of other monosaccharides in the standard mixture. To overcome the problem of lower amino sugar areas, the CARC laboratories used hydrolyzed external monosaccharide standards or N-acetylated external amino sugar standards (which become de-N-acetylated during acid hydrolysis (18)).

Table 3: Effect of acid hydrolysis of monosaccharide standards on HPAEC-PAD analysis in CARC laboratory C.

 Peak Areas (nA-sec x 10-1)

 Sugar

 Non-hydrolyzed

 Hydrolyzed

 Difference (%)

 Fuc

 562

 568

 1

 GalN

 389

 795

 204

 GlcN

 387

 698

 180

 Gal

 692

 656

 5

 Man

 518

 492

 4

Neutral and amino sugar determinations were next repeated on a single hydrolyzate of test sample B. The hydrolyzate was prepared in laboratory C and distributed to other CARC laboratories for analysis of neutral and amino sugars using hydrolyzed external standards prepared at each CARC laboratory. Closer agreement was achieved, except for Gal (Table 4). However, a 2-fold greater error remained between intra- and inter-laboratory determinations, apparently related to differences in instrument configurations or performance.

Table 4: Neutral and amino sugar analysis of test sample B from the same hydrolyzate and performed using hydrolyzed external standards. A, B, C, and D represent four laboratories of the CARC. The units are nmol sugar/mg protein.

 Sugar

 A

 B

 C

 D

 Mean

 GalN

 54

 74

 58

 82

 67 ± 20% (31%) (1)

 GlcN

 420

 480

 410

 390

 430 ± 9.1% (25%)

 Gal

 360

 350

 320

 210

 310 ± 22% (24%)

 Man

 220

 280

 200

 260

 240 ± 15% (27%)

 1 Values in parenthesis are percent standard deviations from Table 1.

Neutral and amino sugar analysis by participating laboratories using HPLC Eleven laboratories, in addition to those of the CARC, reported quantitative monosaccharide analysis of test sample B using HPLC (Table 5). HPAEC-PAD was used in 10 of 11 cases. Laboratory 2803 used HPLC after labeling with 1-phenyl-3-methyl-5-pyrazolone (PMP) (19). The range of values observed were (Table 5): GalN, 27-96; GlcN, 230-480; Gal, 140-520; Man, 80-340; and Fuc, 4.0-8.0 nmol/mg protein. The percent standard deviation (GalN, 54 ± 41%; GlcN, 380 ± 27%; Gal, 290 ± 43%; Man, 200 ± 50%) was significantly greater for the external group's results than for the values obtained by the CARC before or after iterative analysis. The values obtained using the PMP method were within the range of those obtained using HPAEC-PAD. The variousglycoprotein hydrolysis conditions used in the external study were not significantly different from those used by the CARC laboratories in their initial study (Table 1) and so are unlikely to have contributed significantly to the larger error for participating laboratories. Identifying the sources of error for participating laboratories will be the subject of future studies.

 

The reports from laboratories 3568 and 7574 were not included in the ranges cited above. Monosaccharide analysis in laboratory 3568 was performed after hydrazinolysis of test sample B, which apparently resulted in a low yield of oligosaccharides and poor recovery of monosaccharides. Laboratory 7574 used gel electrophoresis after labeling monosaccharides with 2-aminoacridone (20); their results were inexplicably lower than values obtained by HPLC methods and did not vary by a constant amount between individual monosaccharides.

 

CARC laboratories A-D reported that Fuc was not detected in test sample B (Table 5), although they observed small peaks that eluted with retention times similar to Fuc in HPAEC-PAD chromatograms. Because the Committee knew that Fuc was not present in fetuin oligosaccharides, additional studies were not performed (e.g., co-elution studies). The occurrence of artifactual peaks in HPAEC-PAD chromatograms likely accounts for the trace amounts of Fuc reported by laboratories 0231, 3040, 6568, and 8807. The relatively large amount of Fuc reported from laboratory 3568 may be the result of a contaminant from the hydrazinolysis procedure, because samples prepared by hydrazinolysis often show a large peak that elutes near Fuc (21). Apparently the aminoacridone method used in laboratory 7574 and the PMP-labeling method used in laboratory 2803 also result in artifactual Fuc peaks, a result that was also observed in one of the GC determinations (see below).

 

Analysis of sialic acids (Neu5Ac and NeuGc) by CARC and participating laboratories Sialic acids are considerably more acid labile than neutral and amino sugars and so may be more sensitive to changes in hydrolysis conditions. In addition, other sialic acids contain labile substituents that are released by sulfuric and hydrochloric acids. Previously undetected sialic acid species (e.g., 9-O-acetylated) have been found in every vertebrate species studied to date (22). Mild formic acid hydrolysis was initially used to release O-acetylated sialic acids (23). More recently, it has been shown that hydrolysis in 2 M acetic acid at 80°C for 3 hours is optimal for recovering O-acetylated sialic acids from various sources (24-26). While analyzing such labile species is not a concern for bovine fetuin, this hydrolysis would be used for an unknown glycoprotein and was thus evaluated by the CARC.

 

Sialic acid standards (Neu5Ac and NeuGc) were analyzed in a single laboratory (B) after hydrolysis both at the source laboratories (A, C, D, E, and F) and at laboratory B. The response of non-hydrolyzed standards from each of the above laboratories and the same standards hydrolyzed at their respective laboratories was similar for both Neu5Ac and NeuGc (data not shown), with the exception of laboratory C. When hydrolyzed in laboratory C, the Neu5Ac standard showed a 40% lower response than the non-hydrolyzed standard. However, when the same standard was hydrolyzed in laboratory B, a similar response was obtained for hydrolyzed and non-hydrolyzed Neu5Ac. These results suggest that standards or reagents may contain contaminants that destroy sialic acid during hydrolysis.

Table 5: Monosaccharide analysis of test sample B using HPLC (1). The units are nmol monosaccharide/mg protein.

 Monosaccharides (2)

 Labs

 GalN

 GlcN

 Gal

 Man

 Fuc

 Neu5Ac

0231

-- (3)

--

480

300

3.0

--

0770

72

440

360

170

n.d. (4)

 --

1142

53

400

280

210

n.d.

382

1543

41

290

170

110

n.d.

305

3040

77

480

370

270

8.0

371

4931

68

440

430

340

n.d.

490

6211

29

400

140

90

n.d.

330

6568

27

260

150

80

4.3

87

8807

36

240

180

120

4.0

--

2803

87

500

330

280

4.0

370

3568 (5)

n.d.

2.5

20

10

30

51

7574 (1)

26

95

39

87

47

370

CARC (6)

60

390

330

270

n.d.

370

S.D. (6,7)

31

25

24

27

--

9.5

CARC (8)

67

430

310

240

n.d.

--

S.D. (7,8)

20

9.1

22

15

--

--

 1 Laboratory 7574 used gel electrophoresis after fluorophore labeling.

 2 Values expressed as nmol monosaccharide/mg protein.

 3 Not determined.

 4 None detected.

 5 Monosaccharide analysis performed after sample was subjected to hydrazinolysis.

 6 Values from experiments before iterative analyses by the CARC.

 7 Standard deviation, given as a percentage of mean values.

 8 Values from experiments after iterative analyses by the CARC.

The percent standard deviation for all sialic acid analysis performed in the CARC laboratories was less than 10%. Analysis of test sample B for Neu5Ac in CARC laboratories using HPAEC-PAD gave a range of 330-440 nmol/mg protein (370 ± 9.5%) for 10 determinations over approximately a year. The mean value for Neu5Ac was similar between the CARC laboratories and external study participants, with a slightly larger percent standard deviation for the external group (380 ± 17%). The value for Neu5Ac from Laboratory 6568 was about 4-fold less and if included in the average determination would give a value of 334 ± 37%. Trace quantities of NeuGc were reported by the following laboratories: 1142, 10 nmol per mg protein; 1543, 7; 2803, 8; 3040, 9; and 6568, trace. Laboratory 4931 reported that test sample B contained 80 nmol NeuGC per mg protein after analysis by both HPAEC-PAD and GC analysis.

 

Monosaccharide analysis by GC (internal and external study) For GC analysis of monosaccharides, methanolysis has become the preferred release method (14). Methanolysis has the distinct advantage that all types of monosaccharides can be released with minimal destruction, under a single set of conditions. In the presence of excess methanol and with HCl as a catalyst, glycosidic bonds are cleaved and methyl glycosides are formed. Four possible glycosides are produced, a- and b-pyranosides and furanoside s. Thus, after methanolysis of the oligosaccharide chains of a glycoprotein, each different monosaccharide gives at least four peaks by GC whose relative proportions vary with the monosaccharide. For GlcNAc and GalNAc, some de-acetylation occurs (27), and so a re-N-acetylation step is included. An internal standard was included in the analysis to monitor recovery, and external standards treated exactly as the sample were used to determine response factors.

 

Table 6 shows the monosaccharide analysis of the glycosylated test sample using gas chromatography. Three major steps were used: (i) monosaccharide release using methanolysis, (ii) preparation of either per O-trimethyl silyl ethers or alditol acetates, and (iii) GC analysis with peak identification by mass spectrometry. The results from CARC laboratory E were obtained by identifying the TMS derivatives using retention times and mass spectrometry. The base peak ion at m/z 73 was used as a diagnostic for the TMS methyl glycosides. Ions at m/z 204 and m/z 173 were used for positive identification of the neutral and amino sugar-containing peaks, respectively. The values for GlcNAc were significantly lower than the GlcN values obtained by HPLC, 230 ± 36% compared to 380 ± 27%. One contribution to the lower values for GlcNAc may be related to the inefficiency of methanolysis in cleaving sugar-peptide bonds. The error in GalNAc determination was very large using both GC and HPLC methods. The average neutral sugar values were similar for both HPLC and GC determinations; however, none of the laboratories using GC-MS reported unusually low neutral sugar values as was observed for the HPLC determinations (Table 5).

 

Agreement with bovine fetuin carbohydrate structures Bovine fetuin has three N-linked sites that contain bi- (28, 29) and tri-antennary structures (30-32) (Figure 1). In addition three O-linked sites are present (33) that are occupied with sialylated di- and tetra-saccharides (33, 34) (Figure 1). The proportion of these oligosaccharides and the degree of site occupancy (particularly for the O-linked structures) is not known and thus, the precise molar ratio of individual monosaccharides to protein can only be estimated. The theoretical monosaccharide composition was estimated as follows (Table 7). The value of 3 and 9 for GalNAc and Man assumes complete site occupancy of 3 O- and N-linked sites, respectively. The value of 15 for GlcN accounts for the composition of three triantennary structures and the 12 Gal residues for three triantennary structures and three O-linked disaccharides. Complete sialylation of the Gal residues is used to estimate the value of 12 for Neu5Ac.Table 6: Monosaccharide analysis of test sample B using GC1. E represents one of the CARC laboratories. The units are nmol monosaccharide/mg protein.

 

Table 6: Monosaccharide analysis of test sample B using GC (1). E represents one of the CARC laboratories. The units are nmol monosaccharide/mg protein.

 Monosaccharides

 Labs

 GalN

 GlcN

 Gal

 Man

 Fuc

 Neu5Ac

 E, Exp. 1 (2)

29

230

190

150

--

416

 E, Exp. 2

38

130

370

230

--

n.d.

 E, Exp. 3

65

170

501

330

--

370

 0914

38

230

250

200

--

250

 1957

22

240

300

260

--

280

 4424

67

370

450

260

2.5

12 (3)

 Mean

43

230

340

240

330 (3)

 S.D. (4)

 43

36

35

26

23 (3)

 1 Analyzed as TMS derivatives except for 4424, which also prepared alditol acetates.

 2 In experiments 1 and 2, methanolysis was performed for about 16 hours at 80°C with re-N-acetylation for 1 hour in pyridine and acetic acid. Experiments 1 and 2 were performed on different days. In experiment 3, the methanolysis and re-N-acetylation steps were repeated before preparation of the TMS derivatives.

 3 The value for Neu5Ac from Laboratory 4424 was excluded from the average.

 4 Standard deviation, given as a percentage of mean values.

Figure 1: Bovine fetuin asialo-oligosaccharides.

 

The number of Man residues, determined by both HPLC and GC, is in agreement with 3 N-linked sites being fully occupied with either bi- or tri-antennary type structures. The lower value for GalNAc is consistent with partial site occupancy of O-linked sites. Interestingly, MALDI-MS analysis of O-glycanase-treated desialylated test sample B gave a mass shift corresponding to the loss of three disaccharide units (data not shown).

 

The theoretical values for Neu5Ac are based on three fully sialylated triantennary oligosaccharides (3 residues per chain) and three monosialylated O-linked trisaccharides (Figure 1). Greater amounts of sialic acid could be present from tetra-sialylated triantennary structures and the O-linked hexasac-charide (Figure 1).

 

The values obtained by both methods indicated complete sialylation of the Gal residues and was consistent with the presence of the tetra-sialylated triantennary structures and the O-linked tetrasaccharide. MALDI-MS analysis of test sample B resulted in a molecular weight shift corresponding to a loss of 7.6 Neu5Ac residues, possibly due to incomplete enzymatic desialylation.

 

The presence of biantennary structures would reduce the number of GlcNAc residues, and the presence of the tetrasaccharide would increase it. The HPLC value is consistent with the presence of significant amounts of the tetrasaccharide. The GC-MS values for GlcNAc were significantly lower (Table 7).

 

Overall, these results show that quantitative monosac-charide analysis in combination with amino acid analysis, to determine the amount of protein, gives molar ratios of sugars to protein that are in agreement with the oligosaccharide content (given the caveat that the proportion of oligosaccharide species on a glycoproteins is usually not known).

 

Summary and Conclusions

The 1995/1996 study of the Carbohydrate Analysis Research Committee focused on the ability of research laboratories to identify and quantify carbohydrates that were present in one of two unknown protein samples. In addition to the six CARC laboratories, 29 outside laboratories participated in the study. Participating laboratories used a variety of methods for the determination of carbohydrates, including amino acid analysis (14 laboratories), HPAEC-PAD (14 laboratories), GC-MS (four laboratories), mass spectrometry of the glycoprotein (one laboratory), gel electrophoresis after fluorophore labeling (one laboratory), and pre-column derivatization with HPLC (1 laboratory). In all cases, the glycosylated protein was correctly identified, but low levels of carbohydrate were reported by some laboratories in the non-glycosylated test sample. Using amino acid analysis data, most (13 of 14) participating laboratories identified the non-glycosylated test sample as bovine serum albumin and the glycosylated one as bovine fetuin.

Table 7: Monosaccharide content of bovine fetuin (mole sugar/mole protein)

 Observed (1)

 Sugars

 Theoretical

 HPLC (2)

 GC-MS (3)

 Man

9

9.6

9.6

 GalNAc

3

2.7

1.7

 Neu5Ac

12

14.8

13.2

 Gal

12

12.4

13.6

 GlcNAc

 15

17.2

9.2

 1 Quantity of bovine fetuin was 25 nmol (average of Ala and Val content of 1 nominal mg).

 2 Average value after iterative analysis by CARC (Table 5).

 3 Average value from GC analysis by both CARC laboratories and external study participants.

Unexpectedly, considerable variability (± 25-30%) was observed among the neutral and amino sugar analyses of the glycosylated test sample by the CARC laboratories using HPAEC-PAD. Greater variability was observed in the results from the participating laboratories (± 27-50%). The CARC undertook a systematic study to identify sources of variability among CARC laboratories. We first showed the source of variability in HPAEC-PAD analyses was inter- and not intra-laboratoryerrors within a single laboratory were no greater than ±15%. When standards from CARC laboratories were analyzed on a single instrument, the percent standard deviation was no greater than 15% for GalN, GlcN, Gal, Man, Neu5Ac, or NeuGc. A standardized hydrolysis protocol had little effect on decreasing the inter-laboratory error; however, we found that if free amino sugar standards were not injected under acidic conditions, the PAD response was decreased as much as 2-fold. Analysis of the same hydrolyzate for neutral and amino sugars among the CARC laboratories resulted in some improvement in the errors; however, the percent standard deviation remained quite high for GalN and Gal (about 20%). The inter-laboratory errors associated with GC-MS analysis were similarly high (± 20-40%). The analysis of Neu5Ac proved less problematic with values of 370 ± 9.5% and 380 ± 17% nmol/mg protein for the initial CARC laboratories and external study participants, respectively.

In summary, this study demonstrated that most laboratories can correctly determine whether a protein has associated carbohydrate. Quantitative monosaccharide analyses by both HPAEC-PAD and GC-MS methods showed high variability, yet both methods produced values for the average molar ratios of monosaccharide to protein that agreed well with the expected amounts. The CARC laboratories evaluated several potential sources of variability in the HPAEC-PAD analyses. The need to use the same injection conditions for standards and samples was found to be important. Hydrolysis conditions contributed to but did not account for all variability. Future studies will focus on identifying the major sources of error within the CARC and external study participant laboratories.

Abbreviations

CARC, Carbohydrate Analysis Research Committee; HPAEC-PAD, high-pH anion-exchange chromatography with pulsed amperometric detection; MALDI/MS, matrix-assisted laser desorption/ionization mass spectrometry; PMP, 1-phenyl-3-methyl-5-pyrazolone; TFA, trifluoroacetic acid; TMS, trimethylsilyl.

Acknowledgments

The Carbohydrate Analysis Research Committee thanks all participants of the study. The assistance of Delia Matriano (University of California at San Diego), Lorri Reinders (University of California at San Francisco), Dr. Larry Elveback (Complex Carbohydrate Research Center, University of Georgia), Patricia Derby (Amgen, Inc.), and Robin Fischer and Nicki Sandoli (SmithKline Beecham Pharmaceuticals) is gratefully acknowledged.

 

References

 

1. Segrest, J. P. and Jackson, R. L. (1972) Methods in Enzymol. 28, 54-62.

2. Haselbeck, A. and Hösel, W. (1993) in Methods in Mol. Biol., Vol. 14: Glycoprotein Analysis in Biomedicine (E. F. Hounsell ed.) Humana Press. pp. 161-173.

3. McInnes, A. G., Ball, D. H., Cooper, F. P. and Bishop, C. T. (1958) J. Chromatogr. 1, 556-557.

4. Lindberg, B. and Lönngren, J. (1978) Methods in Enzymol. 50, 3-37.

5. Kochetkov, N. K., Wulfson, N. S., Chizhov, O. S. and Zolotarev, B. M. (1963) Tetrahedron, 19, 2209-2224.

6. Kochetkov, N. K. and Chizhov, O. S. (1965) Tetrahedron 21, 2029-2047.

7. DeJongh, D. C. and Biemann, K. (1964) J. Am. Chem. Soc. 86, 67-71.

8. Sawardeker, J. S., Sloneker, J. H. and Jeanes, A. (1965) Anal. Chem. 37, 1602-1604.

9. Björndal, H., Lindbergh, B. and Svensson, S. (1967) Carbohydr. Res. 5, 433-440.

10. Sweeley, C. C., Bentley, R., Makita, M. and Wells, W. W. (1963) J. Am. Chem. Soc. 85, 2497-2507.

11. Townsend, R. R. (1993) in Chromatography in Biotechnology (eds., C. Horvath and L. S. Ettre) American Chemical Society, Washington, D.C. pp. 86-101

12. Townsend, R. R. (1995) in Carbohydrate Analysis: High Performance Liquid Chromatography and Capillary Electrophoresis (ed. Z. El Rassi) pp. 181-209.

13. Hardy, M. R. and Townsend, R. R. (1994) Methods in
Enzymol. 230, 208-225

14. Merkle, R. K. and Poppe, I. (1994) Methods in Enzymol. 230, 1-15.

15. York, W. S., Darvill, A. G., McNeil, M., Stevenson, T. T., and Albersheim, P. (1985) Methods in Enzymol. 118, 3-40.

16. Hobohm, U., Houthaeve, T., and Sander, C. (1994) Anal. Biochem. 222, 202-209.

17. Gottachalk, A. (1972) in Glycoproteins: Their composition, structure and function (ed. A. Gottschalk) 2nd edition, part A, pp. 225-299.

18. Hardy, M. R., Townsend, R. R. and Lee, Y. C. (1988) Anal. Biochem. 170, 54-62.

19. Strydom, D. J. (1994) J. Chromatogr. 678, 17-23.

20. Starr, C. M., Masada, I. R., Hague, C., Skop, E. and Klock, J. C. (1996) J. Chromatogr. 720, 295-321.

21. Lipniunas, P. H., Reinders, L., Townsend, R. R., Bruce, J., Bigge, C. and Parekh, R., "HPLC mapping of oligosaccharides using high-pH anion-exchange chromatography: Improvements from sample prepara-tion, reduction and fluorometric detection" in Techniques in Glycobiology (eds. R. R. Townsend and A. Hotchkiss) Marcel Dekker, New York, in press.

22. Varki, A. (1992) Glycobiology 2, 24-40.

23. Schauer, R. (1982) Advan. Carb. Chem. Biochem. 40, 131-234.

24. Varki, A. and Diaz, S. (1984) Anal. Biochem. 137, 236-247.

25. Manzi, A. E., Diaz, S. L. and Varki, A. (1990) Anal. Biochem. 188, 20-32.

26. Manzi, A. E., Dell, A., Azadi, P. and Varki, A. (1990) J. Biol. Chem. 265, 8094-8107.

27. Jentoft, N. (1985) Anal. Biochem. 148, 424-433.

28. Townsend, R. R., Hardy, M. R., Wong, T. C., and Lee, Y. C. (1986) Biochemistry, 25, 5725-5731

29. Green, E. D., Adelt, G., Baenziger, J. U., Wilson, S. and van Halbeek, H. (1988) J. Biol. Chem. 263, 18253-18268.

30. Spiro, R. G. (1962) J. Biol. Chem. 237, 382-388

31. Bendiak, B., Harris-Brandts, M., Michnick, S. W., Carver, J. P., and Cumming, D. (1989) Biochemistry 28, 6491-6499.

32. Cumming, D. A., Hellerqvist, C. G., Harris-Brandts, M., Michnick, S. W., Carver, J. P., and Bendiak, B. (1989) Biochemistry 28, 6500-6512.

33. Spiro, R. G. and Bhoyroo, V. D. (1974) J. Biol. Chem. 249, 5704-5717.

34. Edge, A. S. B. and Spiro, R. G. (1987) J. Biol. Chem. 262, 16135-16141.

 

 

R. Reid Townsend may be contacted at the University of California, Department of Pharmaceutical Chemistry, San Francisco, CA 94143-0446, Tel: (415) 476-5189, Fax: (415) 502-1655, E-mail: rrtown@itsa.ucsf.edu.

 



Return to the The ABRF Home Page


Created: 1st April 1997
Last modified: 7th April 1997