Electrophoresis 1996. 17, 681-688
Michael G . O’Shea
Matthew K. Morel1
Cooperative Research Centre for
Plant Science, Canberra, ACT,
Australia
Gel electrophoresis of APTS-tagged oligosaccharides
68 1
High resolution slab gel electrophoresis of
8-amino-l,3,6-pyrenetrisulfonicacid (APTS)
tagged oligosaccharides using a DNA sequencer
A novel electrophoretic method for the analysis of oligosaccharides using
DNA sequencer technology is illustrated using malto-oligosaccharide distributions obtained following isoamylase digestion of glycogen, wheat starch and
potato starch. The debranched starches were derivatized at the reducing end
with the charged fluorophore 8-amino-1,3,6-pyrenetrisulfonicacid (APTS).
This highly reproducible method provides baseline resolution of oligomers
from chain lengths of 3 to more than 80 glucose units, and exhibits high sensitivity with detection thresholds of one femtomole per resolved band. In addition, the reductive amination procedure attaches a single fluorophore per oligosaccharide, allowing calculation of the results on either a mass or a molar
basis. The efficacy of the method is illustrated through the determination of
the profile of individual oligosaccharides of chain length with a degree of polymerization (DP) < 80, derived from loading less than 15 ng per analysis of glycogen, wheat and potato starches. While the results obtained were superior in
resolution and sensitivity to previously reported observations using a range of
techniques, they were nonetheless consistent with the overall differences between these polysaccharides. The resolution, sensitivity, reproducibility and
high throughput of the method provides substantial advantages over existing
methods for the analysis of linear oligosaccharide chain length distributions.
1 Introduction
The analysis of the structural features of carbohydrates
and glycoconjugates faces a number of intrinsic difficulties in both separation and detection. The similar structural form, lack of appropriate functional groups, and
low extinction coefficients for UV and fluorescence
detection of carbohydrates renders commonly used techniques such as gas chromatography, reverse-phase HPLC
and electrophoresis virtually impossible without prior
derivatization. Resolving power is required because
much of the information being sought in carbohydrate
analysis is frequently contained in distributions of oligosaccharide chains containing a limited number of monomeric units, connected by a limited number of linkage
types, and varying in length by monomeric intervals.
Detection systems must be highly sensitive because in
the analysis of material from biological systems, the
amounts of substrate available are usually strictly
limited.
influenced by the chain length distribution of the linear
a-1,4 chains as well as the frequency and spacing of the
a-1,6 branches, which link the linear chains into complex
macromolecules. The heterogenous population of molecules, which constitute a given starch, is traditionally
divided into two subpopulations, amylose and amylopectin. The amylose fraction contains molecules that are
lightly branched and have a degree of polymerization
(DP) between 500-5000 and an average chain length
(ACL) of 100-1000. The amylopectin fraction contains
much larger molecules (DP 5000-500000), which are
highly branched and consequently have a much shorter
ACL (typically 10-25).
The fine structure of amylopectin remains a focal point
in research on the structure of starches. The simplest
analysis of amylopectin structure involves the estimation
of an average chain length using colorimetric procedures
to determine the number of reducing ends (usually by a
modified Park-Johnson method [l])liberated by isoamylase debranching as a function of the total carbohydrate
present
(usually by an anthrone-sulfonic acid method
The analysis of the structure of starches provides a parti[2]).
However,
this result provides no information on the
cular challenge to the carbohydrate chemist. The analytical options open to the starch chemist are limited chain length distribution. The analysis of polysaccharide
because starches contain just one monomer (glucose) chain length distributions has traditionally involved the
connected via only two types of linkages (a-1,4 and use of size exclusion separation techniques (both softa-1,6). Yet the properties of starch are profoundly gel low-pressure gel permeation chromatography and
HPLC) with refractive index or light scattering detection
methods [3]. However, such systems provide sub-optimal
Correspondence: Dr. Michael G.O’Shea, Cooperative Research Centre
resolution for fine structure determination, and massfor Plant Science, GPO Box 475, Canberra 2601, ACT, Australia (Tel:
+61-6-279-8367; Fax: +61-6-247-5896; E-mail: michael.osheak2anu.e- based detection methods lack sensitivity. Recent improvements have been made with the development of
du.au)
anion-exchange HPLC systems coupled with pulsed
Nonstandard abbreviations: ACL, average chain length; ANTS, 8-aminoamperometric detection [4],and a fluorescence-assisted
1,3,6-naphthalenetrisulfonicacid; APTS, 8-amino-1,3,6-pyrenetrisul- carbohydrate electrophoresis and image analysis system
fonic acid; DP, degree of polymerization
[5]. However, neither of these systems satisfactorily
resolve both the resolution and quantification problems
Keywords: Slab gel electrophoresis / Oligosaccharide chain length analencountered in the analysis of the range of oligosacchaysis / Laser-induced fluorescence detection / Carbohydrate labeling /
ride chain lengths present in amylopectin.
Starch structure
0 VCH
Verlagsgesellschaft mbH, 69451 Weinheim, 1996
0173-0835/96/0404-0681 $10.00+.25/0
682
Electrophoreris 1996, 17, 681-686
M. G , O’Shea and M. K. Morel1
In this paper we describe a novel technique for the analysis of the polydisperse population of oligosaccharides
generated by the debranching of typical amylopectins.
The procedure involves the reductive amination of the
reducing end of the debranched oligosaccharides with
the charged fluorophore 8-amino-1,3,6-pyrenesulfonic
acid (APTS) and the electrophoretic separation and
detection of the conjugates in polyacrylamide gels using
a DNA sequencer. This method provides substantially
higher resolution and superior sensitivity than traditional
slab gel PAGE techniques. The ability of the method to
quantify subtle, but functionally significant, chain length
variations among oligosaccharide distributions is demonstrated here using the oligosaccharides obtained by
debranching glycogen, wheat starch and potato starch.
2 Materials and methods
2.1 Reagents
2.3 Electrophoresis
Sample volumes of 1-5 pL were loaded into the wells of
uniform polyacrylamide gels in the concentration range
of 5-10°/o (37.5:l ratio of acrylamide to N,W-methylenebisacrylamide as cross-linker) containing 8.3 M urea and
were electrophoresed using a buffer containing 0.089 M
Tris base, 0.089 M boric acid and 0.002 M EDTA for a
period of 15 h at 40°C at a constant current of 30 mA
using an Applied Biosystems (Perkin-Elmer Corporation,
Applied Biosystems Division, Foster City, CA, USA)
373A DNA sequencer. Fluorescence data was collected
by the 373A DNA sequencer and analyzed using 672
Genescan’” software. The detection system of the DNA
sequencer was calibrated by electrophoresing known
amounts of APTS-derivatized maltoheptaose.
3 Results and discussion
3.1 Derivatization by reductive amination
The most common derivatization procedure for carbohydrates possessing a reducing end group is reductive amination, enabling the introduction of a suitable chromophoric or fluorophoric label to increase the sensitivity of
detection. In this reaction, only the reducing end reacts
with the primary amino group of a suitable label to produce a Schiff base, which, in the presence of excess
sodium cyanoborohydride, is reduced to a stable secondary amine. Therefore, only a single fluorophore is
introduced per oligomer. A wide variety of primary
amines have been successfully employed in this reaction
to introduce both chromophores and fluorophores [6]. In
cases where electrophoretic mobility is necessary, the
2.2 Preparation of labeled oligosaccharides
appropriate labels are required to confer both charge and
detectability to the carbohydrate molecule. Most examGlycogen (bovine liver), wheat starch and potato starch ples of this type are polysulfonic acids of polycyclic arowere debranched with isoamylase according to the fol- matic hydrocarbons [5-71. In separate studies, Jackson
[5] and Stack [8] successfully used 8-amino-1,3,6-naphlowing method. The starches (50 mg) were suspended in
0.25 N NaOH (1 mL) and heated at 100°C for 5 min thalene trisulfonic acid (ANTS) as an electrophoretically
before cooling to room temperature. Glacial acetic acid suitable label in high concentration polyacrylamide slab
(32 pL) was added, followed by 0.05 M sodium acetate gel systems, either with photography of UV-illuminated
buffer (4 mL, pH 4.0) and isoamylase (25 WL,5U), and gels or in a CCD detection system. However, from both
the mixture incubated at 37°C at 100°C for 20 min to reports, it is clear from the analysis of maltooligosacchadenature the enzyme. After centrifugation at 14000 g (2 ride ladders that the resolution of oligomers steadily
min, room temperature), the supernatant was crudely decreases as chain length increases. This is a necessary
desalted by the addition of 0.2 g/mL of a mixed-bed ion consequence of acquiring data once the electrophoretic
exchange resin (Bio-Rad AG 501-X8(D), biotechnology
procedure is complete, when the complete range of
grade, 20-50 mesh). The solution was stirred for 5 min, target oligosaccharides must be present on the gel to
decanted, and diluted with distilled water to a total allow data capture. In contrast, the DNA sequencer
volume of 10 mL. Appropriate aliquots (1-50 nmol) avoids this as the labeled oligosaccharides are electrowere determined from a reducing end assay [l], and phoresed past a highly sensitive scanning laser-induced
evaporated to dryness in a centrifugal vacuum evapo- fluorescence detection system which is continually
rator. These aliquots and oligosaccharide standards ( 5 acquiring data, thus eliminating the need to photograph
nmol) were fluorescently labeled by the addition of 5 WL or digitally image the finished gel. This enables the sepaof a solution of 0.2 M APTS in aqueous glacial acetic ration of oligosaccharides without the typical decrease in
acid (15%) and 5 pL of freshly prepared 1 M aqueous resolution as chain length increases. Due to the fact that
sodium cyanoborohydride. The reaction mixture was individual oligomers can be resolved, precise quantificaincubated at 37°C for 15 h and diluted 100- to 1000-fold tion on a molar basis is possible (due to the stoichiomwith an electrophoresis sample buffer consisting of 6 M etry of the reductive amination procedure).
urea in 0.04 M boric acid and 0.04 M tris(hydroxymethy1)aminomethane base buffer (pH 8.6) prior to electro- In order to utilize the capabilities of the DNA sequencer,
phoresis.
a label is required which has the basic physical requireGlycogen (bovine liver) was obtained from Sigma (St.
Louis, MO, USA). Isoamylase was obtained as a suspension in 3.2 M ammonium sulfate (200 U/mL) from
Megazyme (Sydney, Australia). APTS was purchased
from Lambda Fluoreszenztechnologie (Graz, Austria).
Maltohexaose and maltoheptaose were obtained from
Boehringer Mannheim (Mannheim, Germany). Sodium
cyanoborohydride was obtained from Aldrich (Milwaukee, WI, IJSA). Biotechnology grade 20-50 mesh
AG501-X8(D) mixed-bed ion exchange resin was purchased from Bio-Rad (Hercules, CA, USA).
Gel electrophoresis of APTS-tagged oligosaccharides
Electrophoresis 1996, 17, 681-686
ments for labeling via reductive amination (a primary
amine), electrophoresis (a net charge) and excitation and
emission characteristics that closely match the laser excitation and data capture capacities of the DNA sequencer.
Initial results using ANTS were promising (data not
shown), despite the fact that its excitation wavelength
(A,,, = 365 nm) was well removed from the DNA
sequencer laser excitation source of 488 nm; however, as
expected, APTS proved more suitable. Recently, a
detailed study by Evangelista [9] has shown APTS to be
particularly well suited to the laser excitation system of a
capillary electrophoresis instrument, with derivatized
sugars exhibiting substantial excitation at 488 nm, while
the label itself had a relatively low absorption at the
same wavelength. A number of reports describing the
laser-induced fluorescence detection of labeled carbohydrates following capillary electrophoretic analysis
demonstrate that correct matching of the laser source
with the appropriate excitation maximum for the label
can result in high levels of sensitivity being obtained
[ 10-15].
Glycogen (bovine liver), wheat starch and potato starch
were chosen as suitable starting materials, due to the
documented differences in their chain length distributions (for the corresponding amylopectins in the case of
both starches). Previous work has shown glycogen to
contain more branches than amylopectin and to have a
considerably shorter chain length distribution [16, 171.
Conversely, it has been demonstrated that potato amylopectin contains a greater proportion of longer chains
than wheat amylopectin [3, 181. After debranching with
isoamylase [ 191, fluorescent labeling of the linear a-(1,4)glucan reducing ends with APTS was achieved using an
adaptation of the mild conditions described by Chen [20,
211 and Guttman [22] for the same label. Guttman optimized this reaction for a variety of reducing-end saccharides, and achieved a labeling efficiency of 97% in the
process [22]. In our case, reproducibly cleaner oligosaccharide profiles were observed from the crude reaction
products obtained by adding sodium cyanoborohydride
as a freshly prepared aqueous solution as opposed to a
tetrahydrofuran solution.
3.2 Electrophoretic profiles of APTS-labeled
oligosaccharides
As depicted for APTS (Fig. l), reductive amination of
the reducing end enables the introduction of only one
fluorescent label per molecule, thus providing the basis
for equimolar response using fluorescent detection. The
results obtained from electrophoresis on the DNA
sequencer using 10% polyacrylamide gels are shown in
Fig. 2, with the indicated chain lengths deduced upon
comparison with maltohexaose and maltoheptaose standards. The asymmetric appearance of the individual bands
is not reflected in the primary data and becomes
apparent when the software compresses the data to print
the entire profile. Closer examination of the electropherogram (not shown) reveals each band to be symmetrical. The chain length profiles shown in Fig. 2 are reproducibly observed and are representative of those obtained. Multiple repetitions of the electrophoretic procedure on these samples produced less than a 5% variation
683
APTS
8-amino-I ,3,6-pyrenetrisulfonicacid
+H%
(
H
*(
A
2
rGN
“B_H3 H{%rH
I
HO
OH
H
OH
HO
APTS
OH
HO
APTS
OH
Figure I . Fluorescent labeling of carbohydrate reducing ends by
reductive amination using APTS.
in the absolute peak areas per oligomer. Much of this
variation can be ascribed to the difficulties of reproducibly loading aliquots of 1-2 WLinto the wells of the slab
gel. Analysis of samples electrophoresed on uniform gels
ranging from 5 to 10% polyacrylamide demonstrated that
a polyacrylamide concentration of 10% provided the
required level of resolution in the range from DP 5-80.
The electropherograms clearly show the expected differences in distribution, from glycogen with the shortest
chains, through wheat starch to potato starch with the
longest chains. In addition, it can be seen that potato
starch contains an increased number of longer chains
that wheat starch (from DP 40-60). The baseline resolution of oligomers is maintained all the way to chain
lengths of approximately 70 glucose units with no
change in the incremental elution time between oligomers, which allows for their individual quantification.
Other examples of the use of a higher loading of APTSlabeled oligosaccharides have shown that this resolution
is unchanged out to a chain length of approximately 80
glucose units. The detection of still longer chains is
limited only by the physical nature of the debranched
starch samples used in this study, with the potential for
the resolution of longer chains indicated.
3.3 Interpretation of results
If the efficiency of APTS labeling is assumed to be independent of chain length, these results can be characterized on a molar basis, and because the mass of each oligomer is known, characterization on a mass basis is also
possible. In an analogous fashion to the work of Klockow for ANTS labeling of maltooligosaccharides [ 141,
the efficiency of APTS labeling was examined with respect to chain length. A solution containing equimolar
amounts of glucose, maltotriose, maltopentaose and maltoheptaose was subjected to APTS labeling and DNA
sequencer-mediated slab gel electrophoresis under
standard conditions (electropherogram not shown).
Table 1 shows the results as averages of three replicate
labeling reactions, where the APTS-maltopentaose peak
was assigned as 100% labeled and was used as comparator for the remaining peaks. Unlike the results of Klockow [14], these results show a constant peak area obtained per oligomer with high levels of reproducibility
between replicate samples, supporting the use of relative
peak areas as a means of quantifying the distributions
obtained.
Electrophoresis 1996, 17, 681-686
“VI CT O’Shed d n d M K. Morel1
Scan number
800,
400,
-
O-,~L-..
nb
800
400
0
Potato
800
400
0
Wheat
Figure 2. Electropherograms obtained from
DNA sequencer analysis
of APTS-labeled debranched polysaccharides.
Glycogen
Table 1. Relative APTS labeling efficiency of maltooligosaccharides
Labeling efficiency (%)”’
Standard deviation
Glucose
Maltotriose
Maltopen taoseh)
Maltohep taose
104.6
100.0
100.0
103.3
0.3
2.2
0.0
1.0
a ) Average of three replicate APTS labeling reactions, calculated by
comparing relative peak areas.
b) Labeling efficiency arbitrarily assigned as 100% for comparative
purposes.
Due to the fact that data is unavailable for the derivatization of still longer chain lengths, it would be a tenuous
argument to assume that the relative APTS labeling efficiency remains constant up to a chain length of 80 glucose units. However, relative trends can still be examined between different samples. By plotting mass against
chain length (Fig. 3 ) , it can be seen that the distributions
now approximate those reported using size exclusion
techniques where detection is based on mass responses
[3, 121. The molar response data from the DNA
sequencer analysis has been plotted more realistically as
a histogram based on the assumption of complete
labeling, indicating that the distribution with respect to
each oligomer is known. Analysis of these results in the
case of glycogen shows that the mass distribution exhibits a maximum at DP 13, whereas for the molar distribution this maximum is at D P 7, highlighting the fact that
care must be taken when analyzing results depending
upon the detection system used. As can be seen, these
distributions can provide more information than those
generated where detection is available only on a mass
basis.
In addition to the high resolution offered by this technique, an extremely high sensitivity is also obtained,
with the detection of individual oligomers possible down
to a level of one femtomole of labeled carbohydrate per
peak in the electropherogram. To illustrate the sensitivity
of the technique, the amounts of carbohydrate electrophoresed to generate the profiles described here (15 pg
per sample) constitute less starch than is present in a
Electrophoresis 1996, 17. 681-686
Gel electrophoresis of APTS-tagged oligosaccharides
1
--
600
-E
400
a,
300
v
I500
400
-$
0
2
.c
8
Q
300
200
I
5
0)
5
200
100
100
0
0
500
=
-E
400
I400
c
.-
300
::
200
2
L
I
300
-3
200
2
v)
._
8
5
100
100
0
E
v
1
C
l0OC
a,
0
1
.I****t
't,
f
I
8oo
-F
4600
2
.c
3
50C
._
8
a
200
C
10
20
0
50
40
30
60
Degree of Polymerization
Figure 3. Chain length distributions of APTS-labeled debranched
polysaccharides with respect to both molar (histogram) and mass (-+-)
response. (A) Potato starch, 11.5 ng; (B) wheat starch, 7.5 ng; (C) glycogen, 13.7 ng.
0
41
0
, g ,.
,
10
20
,
,
30
.
,
40
.
,
50
.
present in each distribution (Fig. 4). This capability
stems from the ability of the technique to resolve and
quantitate results per individual oligomer. A positive
value on the Y-axis indicates that a greater proportion of
mass is present in those chain lengths in wheat starch
relative to the comparator, while a negative result indicates a corresponding lower proportion. Here the differences between the three samples can be seen in more
detail, with glycogen containing a greater proportion of
mass in shorter chains (below DP 10) than wheat, but
having a lower mass contribution from chains of a
length greater than 20 glucose units. Conversely, potato
starch contains less mass in shorter chain than wheat
starch (for DP < 16), but greater mass in longer chains
from DP 17-28 and from those above a chain length of
approximately DP 40.
4 Concluding remarks
0
--
685
,
60
Degree of Polymerization
Figure 4. Difference comparison of the normalized mass of APTSlabeled oligosaccharides from debranched glycogen (-A-) and potato
starch (-O-) with those of wheat starch. The percentage of the total
mass present in each individual oligosaccharide of potato starch and
glycogen was subtracted from the corresponding value for wheat
starch.
single starch granule 30 pm in diameter. After calculating the percentage of the total mass present per oligomer for each distribution, subtraction of the results of
both glycogen and potato starch from those of wheat
starch allows a direct comparison of the composition of
the proportional quantities (masses) of each oligomer
The novel approach described in this paper to the analysis of oligosaccharide chain lengths provides a substantial advance over existing chain length analysis systems
because of the resolution obtained and the highly sensitive molar-based detection system. The resolution obtained has been demonstrated to separate individual oligosaccharides from DP 5 - 80. The utilization of the
system to analyze chains with lengths in excess of DP 80,
derived from either amylopectin or from amylose, is currently being examined. In this study we have concentrated on the analysis of starch-derived oligosaccharides;
however, there is obvious potential for the application of
this technology to the analysis of other polysaccharides,
providing they possess a reducing terminus and can be
reproducibly and predictably debranched or depolymerized by specific enzymes.
The adaptation of DNA sequencer technology for the
analysis of oligosaccharides required no hardware
changes to the standard Applied Biosystems 373A DNA
sequencer and was greatly facilitated by the use of
Genescan software. This sequencer, or comparable instrumentation, is widely available in research laboratories because of its central importance to modern molecular biology, and the addition of this application to the
range of applications performed by the DNA sequencer
will provide a cost-effective method for many researchers to obtain access to high-resolution oligosaccharide
analysis without requiring the purchase of additional
expensive dedicated instrumentation. The analysis time
per gel using the DNA sequencer parameters described
in this paper was 15 h for the electrophoresis stage; however, 24 samples could be run concurrently, providing
excellent throughput in comparison to HPLC-based
methods.
One feature of the DNA sequencer that was not
explored in this study is the capacity for running multiple samples in each lane, provided that the samples are
derivatized using labels with emission characteristics
that can be unambiguously distinguished by the DNA
sequencer data capture system. This capacity raises several interesting possibilities. First, an internal standard
ladder can be run in each lane, providing a mechanism
686
M. G . O’Shea and M. K. Morel1
for the highly accurate sizing of oligomers. Second, multiple samples can be electrophoresed in the same lane,
providing a highly robust basis for the comparison of
samples, which effectively eliminates errors due to differences in run conditions between gels or lanes. Third,
samples may be differentially labeled at various stages of
treatment (for example, before and after debranching)
providing a method for obtaining additional information
about macromolecular structure from a single electrophoretic analysis.
The authors thank Ian Batey, Rudi Appels, Norm Cheetham
and John Redmond for helpful discussions and Lynette
Preston for providing technical assistance when using the
DNA sequencer and software. The research was supported
by the Australian Grains Research and Development Corporation Grant CPSI.
Received October 24, 1995
5 References
[I] Hizukuri, S., Takeda, Y., Yasuda, M., Suzuki, A., Carbohydr. Res.
1981, 94. 205-213.
[2] Koehler, L. I<., Anal. Chem. 1952, 24, 1576-1579.
[3] Hizukuri, S., Carbohydr. Res. 1985, 141, 295-306.
[4] Koizumi, K., Kubota. Y.. Tanimoto. T.. Okada. Y.. J. Chromatorr.
1989, 464, 305-373
Electrophoresir 1996, 17, 681-686
[5] Jackson, P., Biochem. J. 1990, 270, 705-713.
[6] Lee, K. B., Al-Hakim, A,, Loganathan, D., Linhardt, R. J., Carbohydr. Res. 1991, 214, 155-168.
[7] Chiesa, C. N., O’Neill, R. A., Electrophoresis 1994, 15, 1132-1140.
[ 8 ] Stack, R. J., Sullivan, M. T., Glycobiology 1992, 2, 85-92.
[9] Evangelista, R. A., Lui, M.-S., Chen. F.-T. A,, Anal. Chem. 1995,
67, 2239-2245.
[lo] Chiesa, C., Horvath, C., J . Chromatogr. 1993, 645, 337-352.
[I11 Stefansson, M., Novotny, M., Anal. Chem. 1994, 66, 1134-1140.
[I21 Liu, J., Shirota, O., Wiesler, D., f r o c . Natl. Acad. Sci. USA 1991,
88, 2302.
[I31 Sudor, J., Novotny, M., Proc. Natl. Acad. Sci. USA 1993, 90, 94519455.
[I41 Klockow, A,, Widmer, H. M . , Amado, R., Paulus, A,, Fresenius J.
Anal. Chem. 1994, 350, 415-425.
[15] Zhao, J . Y., Diedrich, P., Zhang, Y. N., Hindsgaul, O., Dovichi, N.
J., J . Chromatogr. B 1994, 657, 307-313.
[16] Rani, M. R. S., Shibanuma, K., Hizukuri, S . , Carbohydr. Res. 1992,
227, 183-194.
[17] Gunja-Smith, Z., Marshall, J . J., Whelan, W. J., FEBS Lett. 1970,
12, 96-100.
[18] Hizukuri, S., Carbohydr. Res. 1986, 222, 342-347.
[19] Akai, H., Yokobayashi, K., Misaki, A., Harada, T., Biochim. Biophys. Acta 1971, 252, 427-431.
[20] Chen, F.-T. A,, Evangelista, R. A., Anal. Biochem. 1995, 230,
273-280.
[211 Evangelista, R. A., Lui, M.-S., Chen, F.-T. A , , Anal. Chem. 1995,
67, 2239-2245.
[22] Guttman, A,, Chen, F.-T. A., Evangelista, R. A,, Cooke, N.. Glycobiology 1995, 5, 717. (Abstract)