Journal of Chromatographic Science. Vol.

22, N o v e m b e r , 1984

An Analysis of the Wood Sugar Assay Using HPLC:

A Comparison with Paper Chromatography

Roger C. Pettersen, Virgil H. Schwandt, and Marilyn J. Effland

Forest Products Laboratory*, Forest Service, U.S. Department of Agriculture, Madison, Wisconsin 53705

Abstract Experimental
Fundamental chemical research concerning wood and Samples and standards
wood-derived products depends on a knowledge of the
Test samples included six woods (four hardwoods and two
materials' carbohydrate composition. Separation and
quantitation of hydrolyzed carbohydrate components of softwoods), four pulps [two Kraft and two neutral sulfite semi-
woods and wood pulps by high performance liquid chro­ chemical (NSSC)], and a series of aqueous standards. These
matography is fast and efficient. Sugars are separated with samples were selected to represent a variety of preparations and
a lead(lI)-loaded Aminex cation-exchange resin. In this were analyzed as received. The woods had been ground to 40
study, six woods and four pulps were analyzed three times mesh in a Wiley mill. The 10 wood and pulp samples were: (a)
by liquid and paper Chromatography (LC and PC). Statis­ Alaskan birch, Betula papyrifera var. neoalaskana (Sarg.) Raup.
tical analysis at the 95% confidencelevel indicatesthe two Extracted with 9: 1 acetone:water; (b) Quaking aspen, Populus
methods are equally accurate tor glucan, mannan, and tremuloides Michx. Extracted with 9: 1 acetone:water followed
galactan. There is a statistical difference for xylan and by extraction with 1:1 toluene:ethanol; (c) White oak, Quercus
arabinan at 95% confidence. The LC precision is better
alba L. Extracted with 9: 1 acetone:water; (d) Southern red oak,
than PC precisionfor glucan and xylan. The precision is
equivalent for arabinan and mannan but not as good for
Quercus falcata Michx. No extraction; (e) Sitka spruce, Picea
galactan. The differential refractive index detector on the sitchensis (Bong) Carr. No extraction; (f) Loblolly pine, Pinus
LC is linear up to a loading of 2 mg of glucose. The op­ taeda L. Extracted with 9:1 acetone:water followed by extraction
timum column operating temperature is 45-55°C. with 1:1 toluene:ethanol;(g) Kraftloblollypine;(h) NSSC loblolly
pine; (i) Kraft southern red oak; (j) NSSC southern red oak.
Each wood and pulp was analyzed three times from June to
August 1983. An aqueous solution of sugar monomers was
prepared from each sample (see next heading). A portion was
Introduction analyzed by liquid chromatography and a portion was analyz­
ed by paper chromatography (3). Integrated peak areas were
Liquid chromatography (LC) is a fast and efficient means of measured on the LC.
separating the five wood sugar residues (glucose, xylose, galac­ Standard solutions were prepared with D(+)-xylose, D-(+)­
tose, arabinose, and mannose) in neutral, aqueous solutions mannose, and L-(+)-arabinose (Gold Label, Aldrich), D-(+)­
(1.2). Separation is achieved with a commercial column pack­ galactose (Aldrich), and D-(+)-glucose (National Bureau of
ed with cation-exchange resin in the lead(II) form. A differen­ Standards). The internal standard, meso-erythritol, was
tial refractive index detector and an electronic integrator pro­ chemically pure (Pfanstiehl).
vide quantitative measurements for each sugar. An automatic
sample injector and a system controller provide continuous Preparation of sugar solutions from woods and pulps
sample and data flow 24 hr/day without operator attention. The woods and pulps were oven-dried at 105°C. The sample
Separation and quantitation for one sample requires 60-70 min. size for each analysis was about 200 mg. The samples were
Although the LC method has been clearly demonstrated (1,2), hydrolyzed for 1 hr at 30°C in 72% H2SO4, then diluted to 3%
there are presently few quantitative data available concerning H2SO4 and hydrolyzed for 1 hr at 120°C. The insoluble lignin
the analysis of wood and pulp samples. The authors present here (Klason lignin) was filtered off and washed with distilled water.
quantitative results from a variety of woods and wood pulps The filtrate and washing liquid were combined and a measured
using both the LC method and the older, established method, amount of erythritol was added as internal standard. The solu­
paper chromatography (PC) (3). tion was neutralized to pH 7 with barium(II) hydroxide. The
barium(II) sulfate precipitate was either centrifuged immediately
*Maintained at Madison, Wisconsin in Cooperation with the University of Wisconsin or allowed to settle overnight in the refrigerator and was filtered

478
Journal of Chromatographic Science, Vol. 22, November, 1984

the next day. The sample solution was concentrated five- to ten­ trol the other factors. The result is that the response factors
fold in an evaporator (3) and filtered through a 0.45-µ mem­ vary and must be remeasured before a new sequence of samples
brane and a reversed-phase cartridge. The authors checked for is injected. Sometimes, these factors are remeasured within a
loss of sugar and internal standard in the evaporator by diluting sequence if samples contain contaminants. Typical response fac­
a test concentrate (approximately 1.6 mg/ml of each compo­ tors (corrected for hydrolysis loss) and elution times for a flow
nent) back to the original volume and comparing LC peak areas rate of 0.5 ml/min and two columns in series are:
before and after evaporation. The average percent difference
before and after concentration and the standard deviation of Response factor Elution time
three determinations were as follows: (µg/µV/sec) × 10 2 (min)
Glucose 1.328 35.75
Standard Xylose 1.462 39.00
% Difference deviation Galactose 1.574 42.25
Glucose - 0.50 0.41 Arabinose 1.479 47.45
Xylose 0.30 1.36 Mannose 1.401 52.39
Galactose 1.86 1.84
Arabinose - 1.34 1.44 The units for the response factor are equivalent to weight per
Mannose 0.47 0.82 unit area, where µV is10-6 volts (ordinate) and time is in seconds
Erythritol 0.06 0.38 (abscissa). Erythritol elutes at nearly 60 min. Elution times vary
with flow rate, column temperature, and pump pressure.
The data indicate no significant loss of sugar or erythritol dur­
ing concentration. Care of the HPX87P column
The lead(II) loaded on the cation-exchange resin forms com­
Chromatographic equipment and materials plexes with the sugars and enables their separation. Care is
The equipment and materials used for paper chromatography necessary to keep the column efficient. Acetic or other organic
have been described previously (3). acids will strip the lead(II) from the resin. When resolution is
The following equipment (Waters Associates) was used for not good, the lead(II) can be regenerated by washing the resin
the liquid chromatography: Model 6000A pump, Model 710 with 0.1 M Pb(NO 3) 2 in 30% acetonitrile:water (pH 3.5). This
WISP automatic sample injector, Model 730 data module and is accomplished by reversing the flow at 0.1 ml/min for 60 hr
integrator, Model 720 system controller, Model 401 refractive at 50°C. After regeneration, the normal flow is resumed with
index detector, Sep-pak C-18 reversed-phasecartridges. The 7.8­ water at 55°C and 0.1 ml/min for 4 hr or until the baseline is
× 300-mm column for LC (Bio-Rad) was packed with 9-µ stable.
spheres of lead(II)-complexed cation-exchange resin (commer­ The authors operated a pair of columns between December
cial code HPX87P.5). Normally, the authors used two of these 15, 1982 and April 7, 1983, when regeneration was necessary.
columns in series to improve resolution. Cation and anion During that time, 222 standard solutions and 404 wood or pulp
micro-guard cartridges (4.6-mm × 30-mm, Rio-Rad) were samples were injected.
plumbed in front of the analytical columns to protect them from Columns may be repacked when the stainless-steelfrits at each
particles and degrading substances. end become plugged. This is apparent when excessive pressure
A temperature control unit and oven (Eldex) maintained col­ (>2,000 psi) is required for eluent flow. To repack, the follow­
umn temperature to 0.1 "C. The differential refractive index de­ ing steps were used:
tector was kept at a constant 32°C with a circulating water bath 1. Remove resin. Clean and regenerate with 0.5M Pb(NO 3) 2
(Model 73, Fisher Scientific). Filtration membranes of 0.45-µ in 30% acetonitrile:water.
pore size (Millipore) were used to prepare the samples and water 2. Clean empty column with a 1-2dilution of concentrated
eluent. HNO 3; rinse with water, then methanol, then water. Sonicate
column frits in same solvents.
Operation of the liquid chromatograph 3. Assemble the column and attach a reservoir. Fill column
Distilled water was degassed and filtered through a 0.45-µ with water.
pore membrane and maintained at 48°C for column elution. 4. Sonicate a 50:50 slurry of resin:water for 5 min to remove
The column was kept at 48°C and the water flow rate set at air.
0.5 ml/min. The injection volume depends on the total sugar 5. Fill the reservoir with the slurry.
content in the sample solution and is usually 50, 75, or 100 µl. 6. Immediately pump the slurry into the column maintain­
The larger volume may be used to measure trace amounts of ing a constant pressure of 2,500 psi.
galactose and/or arabinose. Each sample was injected three 7. Maintain the pressure for 30 min until the flow rate
times in succession. stabilizes.
Detector response factors are determined from a five sugar 8. Relieve the pressure, disconnect the reservoir, and cap.
standard solution. When possible, the authors tried to match
the relative amount of sugars in the standard with that in the
samples. New standards were made up approximately once a
month and were stored in the refrigerator. The response fac­ Results and Discussion
tors vary with flow rate, column resin condition, column
temperature, pump pressure, cleanlinessof the detector cell win­ Linearity
dow, and peak integration parameters set in the data module. Two series of six test solutions containing the five pure sugars
Flow rate, column temperature, and peak integration parameters and internal standard were prepared by successively diluting stock
can be controlled, but it is not always possible to strictly con­ concentrates (Tables I and II; solution #1). The first series (Table

479
 Journal of Chromatpgraphic Science, Vol. 22, November. 1984

1) represents an optimum operating range of sugar concentrations as good for arabinose (Table II). The standard deviation of the
and is typical of wood samples. The second series (Table II) repre­ intercept is calculated according to Youden (4). If the absolute
sents a much higher loading of glucose, xylose, and mannose, and deviation of the intercept from zero is divided by its standard
lesser amounts of galactose and arabinose. This situation may he deviation, the quotient (t) is a statistical measure of the prob­
encountered with pulps or other wood residues, where the pro­ ability the intercept differs from zero. By the null hypothesis,
blem is to assay the minor sugars without overloading the col­ if this value exceeds 2.78 (6 - 2 = 4 degrees of freedom at the
umn with one or more of the other three sugars. As will be 95% level), there is statistical evidence to suggest the intercept
discussed later, the authors consider the data in Table 11 to be does differ from zero. Inspection of the tables shows only man-
an upper limit for column loading. Both tables contain a statistical nose and arabinose at the higher loadings have intercepts
summary at the bottom, based upon a linear regression analysis significantly different from zero at 95% confidence. Even
of the data. The reciprocal of the slope is the response factor. though a statistical difference from zero is indicated for the two
Linearity is excellent for five sugars at the lesser loading (cf. intercepts, the data represent an extreme case of higher loading.
correlation coefficients in Table I). Linearity is good for glucose, Thus, the authors d o not think it is necessary to make any cor­
xylose, galactose, and mannose at the higher loading, but not rections to the integrated peak areas.

Table I. Test Data and Regression Analysis to Measure Linearity of LC Sugar Quantitation (Integrated Peak Areas
Are the Mean of Three Injections)

Table II. Test Data and Regression Analysis to Measure Linearity of LC Sugar Quantitation at Higher Amounts of
lnjected Glucose, Xylose, and Mannose (Integrated Peak Areas Are the Mean of Three Injections)

480
Journal of Chromatographic Science. Vol. 22. November, 1984

Resolution The reverse is true for mannan. The hardwood traces (a-d)
The column temperature affects the resolution of the eluting show an additional small peak between the galactose and
sugars, especially the xylose/galactose and galactose/arabinose arabinose peak. This is likely due to rhamnose, a common
pairs (Table III). This is possibly due to changing equilibrium minor sugar residue present in hardwoods. It is most abundant
constants between the various anomers. Optimum column in the white oak (Figure 1c) and the southern red oak (Figure
resolution was judged to be at 48°C. Resolution was calculated Id). All 10 traces show elution of substances before glucose.
using the formula These materials are either soluble lignins or unhydrolyzed
oligosaccharides.
Eq. 1
Wood and pulp LC and PC analyses
The numbers in Table V are percent of oven-dried sample
where t2 and t1 are retention times and w2 and w1 are baseline and each is the mean of three independent determinations. The
peak widths, respectively, of solutes 2 and 1. The peak widths summary section at the bottom of the table lists pooled stan­
were measured manually from the traces. A resolution of 1.00 dard deviations (line 1) for the carbohydrate components as
is baseline separation. determined by LC and PC. The LC precision is somewhat bet­
ter than that of PC for glucan and xylan, not as good for galac­
Wood and pulp LC chromatograms tan, and about the same for arabinan and mannan. The mean
The chromatographic traces (Figure 1a-j) show the dif­ difference (LC - PC) and the standard deviation of the mean
ference in relative composition between hardwoods, softwoods, difference is computed and listed in lines 2 and 3 under Statistical
and pulps. Xylan content in hardwoods and hardwood pulps Summary. The null hypothesis states that the mean difference
is higher than in softwoods and softwood pulps (Table IV). between two methods should not be significantly different from
zero to establish the equivalence of the methods. The value to
be calculated is
Table III. Retention Time, Response Factor, and
Resolution of Five Sugar Standards + Erythritol Eq. 2
Internal Standard Injected onto the HPX87P Cation-
Exchange Column at Five Temperatures*
(line 4) where N is the number of samples, d is the mean dif­
ference and sd- is the standard deviation of the mean difference.
The value of t is 2.26 for nine degrees of freedom at the 95%
confidence limit. Values oft calculated for each component are
listed in Table V (line 4, bottom) and indicate a statistical dif­
ference for xylan and arabinan. All the components added
together are listed under total Carbohydrate. The corresponding
value oft (3.26) indicates a statistical difference in methods,
i.e., the amount of total carbohydrate determined by LC is
higher than that determined by PC by 1.73%. It is not known
which method is more accurate, because the sample composi­
tion is unknown. The true value may lie somewhere between
the two results.

Test solutions
A series of six solutions containing various proportions of
glucose and xylose was tested by both PC and LC methods.
The results (Table VI) indicate LC values average slightly higher
than PC values for both sugars. The null hypothesis (t = 2.57
for D.F. = 5 at 95% confidence) indicates a statistical difference
for xylose (t = 3.32, Statistical Summary, Table VI), but no dif­
ference for glucose (t = 2.02, Summary, Table VI). If the
calculated values are compared to the respective measured
values, the null hypothesis indicates LC results are equivalent
for glucose (t = 1.02) and xylose (t = 1.01) and that PC results
are equivalent for glucose (t = 2.18), but different for xylose
(t = 4.99). Thus, the authors' current use of paper
chromatography may yield slightly low results for xylose.

Hydrolysis loss
The entire sample preparation scheme (hydrolysis, neutraliza­
tion, filtration, concentration) was checked to determine if the
hydrolysis loss factors are different for LC and PC. The
preparation procedure was run on a solution containing known
amounts of five sugars. The results (Table VII) indicate
hydrolysis losses are equivalent for PC and LC. The authors

48 1
 Journal of Chromatographic Science. Vol. 22, November, 1984

Figure 1. Chromatographic traces of wood and wood pulp hydrolyzates. See Table IV for peak identification.

Table IV. Retention Times (RT, min) and Measured Amounts (MA, µg) of Five Sugars and Erythritol (IS) for 10

482

Journal of Chromatographic Science, VoI. 22, November, 1984

Table V. Carbohydrate Composition of Six Woods and Four Pulps as Determined by Liquid and Paper
Chromatography: A Statistical Comparison of LC and PC for Each Carbohydrate Component (Values Are the Mean of
Three Measurements in Percent Oven-dry Sample, Standard Deviations Are in Parentheses)

Table VI. LC and PC Results on Test Solutions Table VII. Hydrolysis Loss Factors for LC and PC Wood
Containing Glucose and Xylose (mg/ml) Sugar Methods

continue to use the factors determined by the PC method (3),

since they have been measured many times in the past and are
known with greater precision.

Conclusions

Use o f a liquid chromatograph (LC) and a lead(II)-loaded

cation-exchange column for quantitative analysis o f wood and
pulp sugars was investigated. Neutral, aqueous solutions of the
separated sugars were quantitatively measured with a differen­
tial refractive index detector. Chemically pure m-erythritol was
added as an internal standard. The detector response was linear
over a wide range of injected amounts of each sugar. (Column

483
 Journal of Chromatographic Science, Vol. 22, November, 1984

resolution was best at a temperature between 45 and 55°C.) Disclaimer

Statistical tests showed some differences between the LC and
PC methods at the 95% confidence level. The average signifi­ Mention of specific manufacturers does not signify USDA
cant differences (LC-PC) for the 10 wood and pulp samples product endorsement.
were 0.85, -0.34,and 1.73%, respectively, for xylan, arabinan,
and total carbohydrate. The corresponding average differences
for glucan, galactan, and mannan were 0.85, 0.43, and
- 0.05%, respectively, and were not statistically significant at References
95% confidence. Practically, in many cases these differences
in methods may be smaller than the required accuracy. 1. F.E. Wentz, A.D. Marcy, and M.J. Gray. Analysis of wood sugars
The pooled standard deviations for the 10 samples showed in pulp and paper industry samples by HPLC. J. Chromatogr. Sci.
better precision overall for LC. Galactan was a notable excep­ 20: 349-52 (1982).
tion, probably because of the elution position of galactose from 2. M.G. Paice, L. Jurasek, and M. Desrochers. Simplified analysis
of wood sugars. Tappi 65(7): 103-06 (1982).
the LC column.
3. J.F., Saeman, W.E. Moore, R.L. Mitchell, and M.A. Millett. Tech­
niques for the determination of pulp constituents by quantitative
paper chromatography. Tappi 37(8): 336-43 (1954).
4. W.J. Youden. Statistical Methods for Chemists. John Wiley &
Acknowledgment Sons, New York, 1951, pp. 42-3.

The authors thank Necmi Sanyer, retired, formerly Project

Leader of High Yield Non-Polluting Pulping at the Forest Pro­ Manuscript received June 25, 1984;
ducts Laboratory, for the wood and pulp samples. revision received August 31, 1984.

484