Anal. Chem. 1984, 56,497-504 (9) Bannlster, S. J.; Sternson, L. A.; Repta, A. J. J. Chromatogr. 1983, 273, 301-318. (IO) Sternson, L. A.; Marsh, K. C.; Bannister, S. J.; Repta, A. J. Anal. PrOC. 1983, 2 0 , 366-368. (11) Bannister, S. J.; Sternson, L. A.; Repta, A. J. J. Chromatogr. 1979, 173, 333-342. (12) Riley, C. M.; Sternson, L. A.; Repta, A. J. J . Chromatogr. 1981, 277, 405-420. (13) Riley, C. M.; Sternson, L. A.; Repta. A. J. J. Chromatogr. 1981, 279, 235-244. (14) Marsh, K. C., Ph.D. Thesis, The Unlverslty of Kansas, 1983. (15) Hussain, A. A.; Haddadln, M.; Iga, K. J. Pharm. Sci. 1980, 6 9 , 364. (16) Long, D. F., Ph.D. Thesis, The University of Kansas, 1980. (17) Brlstow, P. A.; Brittain, P. N.; Riley, C. M.; Wllllamson, B. F. J. ChromtOgr. 1977, 131, 57-64. (18) Englehardt, H.; Neue, U. D. Chromatographla 1982, 15, 403-408. (19) Schroeter. L. C. J. Pharm. Scl. 1081. 5 0 , 891-901; 1983, 5 2 , 559, 584, 888. (20) Backstrom, H. L. J. J. Am. Chem. Sac. 1927, 49, 1460-1472. (21) Basolo, F.; Pearson, R. G. “Mechanisms of Inorganic Reactions-A Study of Metal Complexes In Solution“, 2nd ed.; Wlley: New York, 1967; pp 351-410. (22) Pearson, R. G.; Gray, H. B.; Basoio, F. J. Am. Chem. SOC.1980, 8 2 , 787-792. (23) Belluco, U.; Cattallnl, L.; Basolo. F.; Pearson, R.; Turco, A. Inorg. Chem. 1985, 4 , 925-929. (24) Pearson, R. G.; Eilgen, P. C. “physical Chemistry-An Advanced Trea-
(25) (26) (27) (28) (29) (30) (31) (32)
497
tise, Volume VII-Reactions in Condensed Phases”; Eyring, H., Ed.; Academic Press: New York, 1975; Chapter 5. Rich, R. L.; Taube, H. J. Am. Chem. SOC. 1954, 76, 2608-2611. Llppard, S. J. Science 1982, 218, 1075-1082. Halpern, J.; Pribanlc, M. J. Am. Chem. SOC.1988, 9 0 , 5942-5943. Skeggs, J. Am. J. Clin. Pathol. 1957, 2 8 , 311. Deelder, R.; Kroll, M.; Beeren, A,; Van den Berg, J. J. Chromatcgr. 1978, 149, 869-682. Deelder, R. S.; Kroli, M. G. F.; Van den Berg, H. H. M. J. Chromatogr. 1978, 125, 307-314. Schlabach, T. D.; Chang, S. H.; Gooding, K. M.; Regnier, F. E. J. Chromatogr. 1977, 734, 91-106. Jonker, K. M.; Poppe, H.; Huber, J. F. K. Chromatographla 1978, 1 I ,
123. (33) Cieare, M. J.; Hydes, P. C.; Malerbi, B. W.; Watkins, D. M. Biochlmie 1978, 60, 835-650.
RECEIVED for review April 25, 1983. Accepted October 11, 1983. Abstracted in part from the Ph.D. thesis of K.C.M. This work was also supported in part by a University of Kansas Honors Fellowship (to K.C.M.) and Grants CH-149 from the American Cancer Society and CA-09242 and CA-24834 from the National Cancer Institute (National Institutes of Health).
Determination of Neutral Sugars in Plankton, Sediments, and Wood by Capillary Gas Chromatography of Equilibrated Isomeric Mixtures Gregory L. Cowie and John I. Hedges* School of Oceanography, University of Washington, WB-10, Seattle, Washington 98195
A reproducible technique is described for extraction and quantitative anaiysls of neutral monosaccharides from a variety of wild natural sample types, requiring as ifflie as 10 mg of total organic matter. Acid hydrolysis yields monomeric sugars which may exist in up to five isomeric forms when in solution. Lithium perchlorate is used to catalytically equiiibrate sugar isomer mixtures in pyridine prior to conversion to their trimethyisiiyi ether derivatives. Analysis is carried out by use of gas-liquid chromatography on fused-silica capillary columns. Quantlflcation on the basis of a single clearly resolved peak for each sugar is made possible by the equiilbration step. Sugar losses and optimal conditions for maximum reproducible sugar recovery are determined for each extraction stage.
Carbohydrates are major structural and storage compounds in both terrestrial and marine organisms, representing the largest fraction of the photosynthetically assimilated carbon in the biosphere. The purpose of this study was to develop a sensitive and reproducible technique for the hydrolytic extraction and quantitative analysis of complex monosaccharide mixtures in a broad range of solid, natural samples including plant tissues, plankton, and sediments. Further objectives were that the technique should employ commonly available instrumentation, involve a minimum of chemical manipulation prior to analysis, and be as straightforward and rapid as possible. Among the major problems associated with published quantitative carbohydrate methods have been the large variety and frequently poor reliability of techniques used for the extraction, identification, and measurement of carbohydrates 0003-2700/84/0356-0497$01.50/0
from natural samples (1). Due to the lack of a standard extraction technique and, in most cases, information on sugar recoveries, direct comparison of the results of different studies has proven difficult or impossible. Gas-liquid chromatography (GLC) has been a commonly used technique for monosaccharide analysis but requires formation of volatile derivatives since free sugars are insufficiently volatile for direct analysis. Both trimethylsilyl (Me,%) ethers (2-7) and trifluoroacetate (TFA) esters (49) have proven useful for analytical purposes. However, when in solution free sugars may exist in as many as five different forms (one acyclic form and two anomers for each of the fiveand six-membered ring forms). This leads to a complex multiplicity of peaks which has proven difficult to resolve. In addition, TFA derivatives have problems of poor sample stability and on-column losses (9). Reduction of monosaccharides to their corresponding alditols followed by the formation of MeaSi ether or acetate derivatives (10-14) avoids the problem of peak multiplicity by removing the carbonyl group which is normally involved in ring formation through internal glycosidic bonds. Peaks are therefore much simpler to resolve since only one peak is obtained for each sugar. This method, however, involves significant chemical manipulation of the sugars and may lead to information-loss because certain sugar pairs (e.g., lyxosearabinose, gulose-glucose) yield the same alcohol as their reduction product. Ketoses and aldoses present in the same sample can also yield the same alcohol (e.g., sorbitol from glucose or fructose). In addition, ketoses always yield more than one alcohol (e.g., fructose produces both sorbitol and mannitol). Conversion of the free sugars to their oximes followed by formation of Me& ether derivatives (2, 15, 16) also reduces the problem of peak multiplicity (two possible 0 1984 American Chemical Society
498
ANALYTICAL CHEMISTRY, VOL. 56,
NO. 3, MARCH 1984
forms per sugar, with only one found for most sugars) but some problems have been reported due to poor sample stability of the sugar oximes. Bethge et al. (4) introduced a quantitative monosaccharide technique by using GLC of Me3& ethers, overcoming problems of peak multiplicity by using a catalyst (0.2% w/v LiC104, 40 “C) to bring free sugars to mutarotation equilibrium in pyridine prior to derivatization. Equilibrium isomer distributions were determined separately for a number of sugars so that total peak areas for individual sugars within a complex sugar mixture could then be calculated on the basis of a single clearly resolved peak for each sugar. Another bifunctional catalyst, 2-hydroxypyridine, can also be used to achieve mutarotation equilibrium (17). Other techniques which have used MeaSi ether derivatives have either remained qualitative (2, 5) or reached mutarotation equilibrium in aqueous solution and assumed that, because mutarotation in pyridine is slow and trimethylsilylation is fast, subsequent compositional changes during dissolution in pyridine or derivatization do not occur (6, 7). Liquid partition chromatography and colorimetric detection are also common methods for quantitative determination of monosaccharides (2419).With this approach the problem of mutarotation and ring isomerism is removed since only one peak is obtained per sugar. A further technique has recently been introduced which involves the use of reverse-phase high-pressure liquid chromatography and fluorimetric detection (20). However, in many laboratories only GLC is available as a routine analytical method. The carbohydrate analytical procedure described here employs a GLC equipped with a conventional hydrogen flameionization detector (FID).I t is applicable to a wide variety of biological material types and extraction and preparation stages have been optimized for maximum reproducible sugar recoveries. Measi ether derivatives have been chosen for their ease and rapidity of formation, excellent chromatographic qualities, and long-term chemical stability. Problems of peak multiplicity are minimized by catalytically equilibrating sugar mixtures in pyridine prior to derivatization and good peak resolution is obtained through the use of a fused-silicacapillary column.
EXPERIMENTAL SECTION Sugar Extraction and Isolation. Plankton and sediment samples are freeze-dried and plant samples are oven-dried overnight at 60 “C. All samples are ground and homogenized in a mill, or with mortar and pestle, to pass a 0.351-mm mesh sieve. Depending on sample type and organic carbon levels, 50-mg (plants) to 500-mg (sediment) samples, equivalent to an organic carbon loading of 10-30 mg, are weighed out and transferred to 50-mL Pyrex tubes with Teflon-lined screw-caps. A volume of 72 w t % H2S04sufficient to fully expose the sample (1-3 mL) is added by pipet, follawed by a magnetic stirring bar. The samples are stirred for 2 h at room temperature after which the acid is diluted to 1.2 M with distilled water. The capped tubes are stirred at 100 “C in a boiling water bath for 3 h after which the hydrolysis is halted by placing the tubes in an ice bath. Adonitol, a five-C alditol, in pyridine solution (1-5 mg/mL), is used as a GLC internal standard. Depending on sample type and organic carbon levels, 50-1000 pL (to match the internal standard with expected sample sugar levels) is added directly to the hydrolysate by micropipet. The solution is then completely homogenized and transferred to a 100-mL glass beaker. Sorbitol, a six-C alditol, is used as an additional internal standard when there are large differences between levels of different sugars within a sample or to determine “relative”and “absolute” sugar recoveries for a given sequence of the analytical scheme (see later discussion). Finely ground Ba(OH)2(approximately 90% of the stoichiometrically required quantity) is then added to help neutralize and desalt the hydrolysate. Stirring and ultrasonic vibration are used to aid BaS04 precipitation and to avoid incomplete reaction due to protective BaS04 films which may form on the Ba(OH)2
crystals. The mixture is then brought to a pH of 6.5 (i0.2) by further addition of Ba(OH)*and transferred to a 50-mL plastic centrifuge tube. The mixture is then centrifuged for about 15 rnin at 2500 rpm (700g) and the supernatant removed by decantation. The neutralized solution is passed through a 15-20 mL column of 1:l mixed cation (Dowex 50W-X8, H+ form, 20-50 dry mesh) and anion (Dowex 1x8-200, formate form converted from Cl- form, 100-200 dry mesh) exchange resins at 1.5-2.0 mL/min (1). The column is then rinsed with at least one bed volume of distilled water. The deionized hydrolysate is transferred to a 50-mL pearshaped flask. A variable speed Buchi rotoevaporator is used to evaporate the solution to dryness in a water bath maintained at 50-60 “C. The typically large total solution volumes (60-70 mL) require a high rotation speed (150+ rpm) and a vacuum of 28 psi or more to speed up this otherwise time-consuming step. Equilibration and Derivatization. The dried sample is dissolved in approximately 0.5 mL of distilled pyridine. At this stage an aliquot may be removed, dried under NO,and stored frozen for later analysis. The remaining fraction is transferred to a l-dram vial. An equal volume of pyridine containing 0.4% w/v LiC10, is added, the vial is sealed with a Teflon-lined cap, and the solution is left to equilibrate in an aluminum heating block at 60 “C for 48 h. Sugar trimethylsilyl derivatives are formed by adding 0.1 to 0.25 mL of Regisil (bis(trimethylsily1)trifluoroacetamide) + 1% trimethylchlorosilane (Regis Chemical Co.) to the solution. A constant temperature of 60 O C is maintained for another 10 min to ensure complete derivatization of the equilibrated sample. Gas Chromatography. A Hewlett-Packard 5700A gas chromatograph fitted with a flame ionization detector (FID) is used for all analyses. The injector is modified for “split” injections into a capillmy column and has an all-glass injection liner partially fiied with silylated glass beads to increase surface area for efficient sample volatilization. Both the injection port and the FID are maintained at a constant temperature of 300 OC. Analyses are made by using a 30 m by 0.25 mm i.d. fused-silica capillary column, coated with SE-30 liquid phase (J&W Scientific Inc.). The injection is split at an approximate 1/50 ratio with an initial column flow of about 1.2 cm3/min of helium. Column temperature is programmed from 140 O C increasing at 4 “C/min after an initial delay of 4 min. Under these conditions all sugars elute within 20 min and at least one major anomer peak of each quantified sugar can be clearly resolved (Figure la). The only exception is ribose when it is present at low levels and the peak used for quantificationis small relative to adjacent peaks (Figure lb). For such mixtures column temperature can be programmed from 130 “C increasing at 2 OC/min with no initial delay, in order to more completely resolve the ribose peak. Under these conditions the full sugar suite elutes within 30 min although the peaks of interest elute in under 15 min. The FID analog signal is recorded and processed by a Hewlett-Packard Model 3390A integrator. Standard sugar mixtures, including the GC internal standard adonitol, are prepared in 10-mL volumetric flasks at approximately 1 mg/mL per sugar using Reagent-grade sugars (Sigma Chemical Co.) dissolved in 0.2% w/v LiClO, pyridine. About 0.25 mL of this standard solution is transferred to a l-dram vial for equilibration and subsequent derivatization. Response factors and retention times for each sugar are routinely determined relative to the GLC internal standard by injection of the standard mixture at the beginning of GLC analyses and every fourth or fifth run thereafter.
RESULTS AND DISCUSSION This technique is based on a combination of the preparation and extraction steps described by Mopper in 1977 ( I ) and the equilibration step described by Bethge et al. in 1966 (4). The present method, however, includes significant modifications to these original procedures arising from recovery and optimization tests carried out on each stage of the procedure and adaptation to high-resolution capillary GLC analysis. Hydrolysis. Sugar values from wood, plankton, and sediment trap samples, with and without the 72 w t % HzS04 pretreatment, were compared in order to determine the effects of this step on individual sugar yields (Table I). Only glucose
ANALYTICAL CHEMISTRY, VOL. 56, NO. 3, MARCH 1984
499
Table I. Concentrated Acid Pretreatment Tests a
pretreated nonpretreated .t % change
LYX
ARA
RHA
1.89 2.05 t 8.5
2.32 2.29 -1.3
0.87 0.83 -4.6
RIB Wood
Plankton 2.23 2.07 -7.2
mgx 10 XYL FUC
MAN
GAL
GLU
8.89 8.21 -7.7
67.9 62.6 -7.8
9.09 9.40 t3.4
166 15.3 -90 4.16 1.58
t 11
0.21 0.23 +9.5
1.45 1.42 -2.1
1.45 1.57 + 12
1.28 1.32 t 3.1
1.67 1.76 t 5.4
7.93 7.50 -5.4
-62
0.48 0.41 -15
1.08 0.99 -1.9
Sediment Trap Material 2.31 1.37 1.77 2.37 1.44 1.82 t 2.6 t 5.1 + 2.8
2.64 2.78 t 5.3
4.78 4.43 -7.3
6.23 6.52 t 4.7
9.44 4.69 -50
pretreated nonpretreated t % change
0.18 0.20
pretreated nonpretreated 2 % change
a Wood, spruce trunk wood, 50 mg, ribose and fucose were not at quantifiable levels. Plankton, “bulk” sample from Dabob Bay, 100 mg. Sediment trap material, collected over 1 month at 30 m depth, Dabob Bay, 200 mg. Refer to Table I1 for sugar abbreviation identifications.
25
9
Y
Y) z
a a
Y)
Y
1
I
I
I
I
1
TEMPERATURE i’CI
Figure 1. Gas chromatographic traces of (a) a standard mixture of Me3Si derivatlzed monosaccharldes and (b) the Me,Si derivatives of monosaccharldes obtained by acid hydrolysis of a recent sediment from
Dabob Bay, WA (1-2 cm depth Interval). Gas chromatographic equipment and condttions are described in the test. Identiflcations (see Table 111): 1 = L,; 2 = A,; 3 = A,; 4 = XI; 5 = R1; 6 = Ri,, L2; 7 = FIX,; 8 = L?, A,, Ri2; 9 = L4, Ri3; 10 = Lg; 11 = Ri,; 12 A,, F,; 13 = R,; 14 = X,; 15 = F,; 16 = F,, R3; 17 = adonitol (Internal standard); 18 = X4; 19 = MI; 20 = 0,;21 = G,; 22 = GI,; 23 = G,; 24 = M,; 25 = sorbitol (internal standard); 26 = GI,. showed differences (outside the reproducibility of the technique) between pretreated and untreated sample pairs. Pretreated glucose yields are increased by a factor of 10.9 in the wood sample, by a factor of 2.6 in the plankton sample, and by a factor of 2.0 in the sediment trap sample. Because glucose is consistently found to be one of the major sugars, the 72 wt % H2S04pretreatment is included for all sample types. Results of a series of tests carried out by Mopper in 1977 (1)indicate that various published carbohydrate techniques ( 5 , 2 1 , 2 2 )have employed hydrolysis conditions which were far from optimal for efficient sugar extraction because of the
acid type or concentration used, or hydrolysis temperature or duration. In order to determine generally optimal hydrolysis times for different sample types, time-series were carried out on a modern sediment sample (a surface mud from Lake Washington, WA), plankton (a “bulk” sample from combined phyto- and zooplankton net-tows from Dabob Bay, WA), a wood sample (spruce trunk wood), and a standard monosaccharide solution. The results of these tests (for the sediment and plankton samples) are shown in Figure 2 and indicate that sugar yields reach a “plateau” between 2 and 4 h after commencing the 1.2 M H2S04hydrolysis at 100 “C. The only exception is galactose in the phytoplankton sample but the increasing yield of this sugar may only be apparent as it is within the reproducibility of the technique. The wood sample showed similar characteristics to the sediment and plankton, as did percent recoveries determined for the standard monosaccharide solution. In some cases significant losses appear to begin after 5 h and, based on these results, an hydrolysis period of 3 h (following the 2 h 72 wt % H2S04 treatment) was selected for all sample types. This time period is somewhat shorter than that found by Mopper (1) to be optimal for a variety of sediments using the same hydrolysis conditions (4-5 h). Figure 3 shows a similar sugar yield time series for a-cellulose. This plot also shows that optimal glucose recovery occurs in the 2 to 4 h region. Although cellulose is a polymer which is relatively difficult to hydrolyze, as indicated by the need for a concentrated acid pretreatment (231,this hydrolysis series should be more representative than the standard monosaccharide solution series of conditions seen by polymeric or “combined” sugars which predominate in most geochemical samples (I). The results of this test suggest the the hydrolysis reaction essentially goes to completion and that sugar extraction efficiency is good, reaching a “plateau” glucose recovery of approximately 80%. The results also give an indication of the accuracy of the technique in terms of the quantitative sugar values as representations of sugar levels in the parent sample. The feasibility of applying percent recoveries to the back-calculation of absolute sugar levels will be discussed below. Hydrolysis-yieldlinearity tests were carried out on a modern surface sediment from Lake Washington, the results of which are presented as correlation coefficients (r2) in Table 11. Sample sizes tested were 50,200,500, and lo00 mg, equivalent to an organic carbon range of 2.45 to 48.90 mg. Good linearity was obtained for all sugars, all correlation coefficients exceeding 0.98 (including total sugar) except for lyxose and
500
ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1984
Table 11. Hydrolysis-Yield Linearity Tests a sugar
corr ( r z )
% TCH,O
sugar
corr ( r 2 )
% TCH,O
lyxose (LYX) 0.9584 1.51 mannose (MAN) 0.9972 15.1 arabinose (ARA) 0.9698 6.67 galactose (GAL) 0.9974 23.7 rhamnose (RHA) 0.9983 12.0 glucose (GLU) 0.9984 24.8 100 xylose (XYL) 0.9998 8.27 TCH,O 0.9995 fucose (FUC) 0.9888 7.93 a Hydrolysis-yield linearity for surface sediment from Lake Washington (50, 200, 500, and 1000 mg samples, 4.83% OC). Correlation coefficients ( r 2 )calculated by linear regression. N o ribose (RIB) quantification was made for these samples. TCH,O = total Carbohydrate.
9
(1
0 1G A L A C T O S E
GLUCOSE = XYLOSE = MANNOSE o = RHAMNOSE FUCOSE =ARABINOSE + = LY XOSE A A
10
7
6t i /!3
-
o;.e&s----?I I
2
3
4
? 5
i 6
H Y D R O L Y S I S T I M E (hours)
(b)
";I
0: G A L A C T O S E m: GLUCOSE A
28
A
XYLOSE MANNOSE
FUCOSE ARABINOSE +: L Y X O S E e : RIBOSE O =
::
o=RHAMNoSE
-A
i
H Y D R O L Y S I S T I M E (hours)
Flgure 2. Monosaccharide recovery as a function of hydrolysis time for (a) plankton (combined zoo- and phytoplankton from Dabob Bay, WA, 250 mg samples) and (b) a recent sediment from Lake Washington (500 mg samples). Hydrolysis in 1.2 M H2S04at 100 "C (following 72 wt % H2S04 pretreatment).
arabinose which were relatively minor sugars and more difficult to quantify. Poorest results were obtained for the 50-mg sample apparently because the tctal carbohydrate content was too low. Yield linearity improved markedly for most sugars with the 200 mg and larger samples and a lower limit organic
0
1
2 3 4 5 HYDROLYSIS T I M E (hours)
6
Flgure 3. Glucose yield from a-cellulose as a function of hydrolysis time; theoretical 100% glucose yield calculated as mg of glucose = l . l ( m g of ceilulose), assuming cellulose = n (CeH,,05); hydrolysis in 1.2 M H$04 at 100 OC (foiiowlng 72 wt% H2SO4pretreatment); data collected on two separate dates.
carbon loading of 10 mg was therefore taken for all sample types. As indicated by the test results, more care should be taken with the quantification of minor sugars. Neutralization and Deionization. Hydrolysates are neutralized and deionized in a single step in the Mopper technique, using combined anion and cation exchange resins. However, larger volumes of 72 wt % H2S04are used in the present method because it is considered important to fully expose the sample. It is found under these conditions that the ion-exchange resins are not able to neutralize the hydrolysates and, as a result, sugar losses at this and subsequent stages are markedly increased. Neutralization with Ba(OH), is therefore carried out immediately after hydrolysis and internal standard addition. Tests with buffered standard solutions (pH 2.0, 7.0, and 10.5) containing amino acids with a wide range of isoelectric points (glutamic acid, valine and lysine) indicate that sugar losses are minimized and amino acid removal maximized when the solution is neutralized prior to passage through ion-exchange resins. Evaporation. A further difference between this technique and that of Mopper (1) is in the sample drying step. Mopper finds that addition of glycerol at this stage is important in reducing sugar losses ( I , 24), apparently by inhibiting sugar condensation reactions (e.g., with glassware, amino acids, or other sugars). Tests carried out in this study, however, show glycerol to be unnecessary in neutral solutions, with both standard solutions and samples. In fact, glycerol increased losses of some 5-C sugars and interfered with equilibration and derivatization stages. Sugar condensation reactions can
ANALYTICAL CHEMISTRY, VOL. 56, NO. 3, MARCH 1984
501
Table 111. Equilibrium Isomer Distributions a sugar LYX
ARA
RHA RIB
peak no. re1 RT 0.74 L,* 0.79 L, L3 0.81 L, 0.83 L, 0.89 A, 0.75 A,* 0.76 A3 0.81 A4 0.87 Rl* 0.78 R, 0.91 R3 0.95 Ri, 0.79 Ri 0.81 Ri, 0.83 Ri,* 0.86
%TA 55.4 6.52 6.11 22.9 9.07 23.5 30.7 33.7 12.1 76.6 21.2 2.18 10.7 17.3 51.8 20.2
%sd 0.11
n
3
sugar XYL
peakno.
x,
X, x3* FUC 1.89
3
0.68
6
0.59
3
MAN GAL GLU
x,
Fl F, Fa* F, M,* M2
G, G,* G3 G1, G1,*
re1 RT 0.77 0.80 0.93 1.08 0.80 0.87 0.94 0.95 1.15 1.34 1.17 1.24 1.33 1.30 1.48
%TA 3.74 2.71 47.6 45.9 14.2 36.3 39.9 9.56 79.0 21.0 18.5 32.1 49.4 47.7 52.3
% sd
n
0.78
6
1.87
3
0.23
3
1.01
3
0.23
9
a re1 RT, retention time relative to internal standard, adonitol. % TA, % total area for individual monosaccharide, (mean). % sd, % standard deviation about mean value for quantified peak. n , number of determinations. *, peak routinely used for
proportional quantification. Refer to Figure 1 for peak location. Table 1V. Recovery Testsa test A %-REC, %-REC, R-A test B %-REC, %-REC, R - A* test C %-REC, %-REC, R-A test D %-REC, %-REC, R-A
Is1
LYX
ARA
RHA
RIB
XYL
FUC
MAN
GAL
GLU
100 99.3 0.7
86.2 85.6 0.6
94.4 93.7 0.7
99.3 98.6 0.5
69.4 68.9 0.6
89.2 88.6 0.7
102 101 0.7
97.5 96.8 0.7
90.8 90.2 0.6
97.5 96.8 0.7
100
85.0 78.6 0.8
94.9 87.8 1.5
95.1 88.0 1.5
70.2 65.0 -0.3
84.9 78.6 0.8
100 92.9 1.9
95.5 88.3 1.5
87.0 80.5 0.9
91.5 84.7
99.3 98.1 1.2
100 99.7
95.8 94.7 1.1
1.2
104 103 1.3
103 101 1.2
101 100 1.4
109
1.2
107 106 1.3
101
98.8 1.2 100
101
103
102
98.5 1.5
99.9 1.5
101
100
1.5
1.5
94.2 92.8 1.4
97.6 96.1 1.6
100 98.9 1.5
101 99.6 1.5
99.1 97.7 1.5
104 102 1.6
92.5 1.9 100
100
1.2
108
1.3
a Key: Isl, first internal standard, adonitol; test A, neutralization-precipitation step; test B, neutralization-precipitation step plus centrifugation step; test C, ion-exchange step; test D, rotoevaporation step; %-REC, , “relative” % recovery, as determined by addition of internal standard before specified step; %-REC, , “absolute” % recovery, as determined by addition of second internal standard, sorbitol, after specified step. R - A, net loss for specified stage, difference between “relative” and “absolute” recoveries. *, R - A values for test B exclude 5.6% loss incurred by centrifugation (determined gravimetrically).
be catalyzed under acidic or basic conditions (25) and it is possible that glycerol was previously found to be important because hydrolysates were not fully neutralized by the ionexchange technique. Glycerol therefore is not included in the present technique. Equilibration. Equilibrium isomeric compositions (in pyridine) of the quantified sugars are presented in Table 111. The conditions used to attain equilibrium (48 h at 60 “ C in a 0.2% w/v LiC104pyridine solution) are different from those of Bethge et al., who used the same catalyst concentration but only 2 h at 40 OC. Equilibration studies using solutions with different combinations of sugars (so as to avoid peak overlaps) show that although standard solutions appear to reach equilibrium isomer distribution for most sugars under the conditions of Bethge et al., this was not true in all test situations. For example, measurable compositional changes were found when solutions were put through various stages of the extraction procedure in the presence of foreign compounds (e.g., glycerol), particularly for those sugars with larger numbers of isomeric forms (e.g., lyxose, arabinose, or xylose). Isomer distributions are also temperature dependent and the equilibration conditions described above were selected
after temperature- and time-dependence studies indicated that a period greater than 24 h at 60 “C was optimal for equilibration of both standard solutions and samples. Measurable sugar losses do not occur during this stage, as indicated by total peak-area comparisons with sugar solutions freshly made up in pyridine. Under these conditions all solutions came to constant peak-area distributions and better reproducibility was found for triplicate sediment samples than under the original equilibration conditions. By equilibrating in pyridine rather than aqueous solution, calibration standards can be made up, equilibrated, and derivatized in pyridine, whereas with aqueous techniques standards must first be made up and equilibrated in water. The latter procedure involves the assumption that the standard solution behaves identically to all samples during evaporation, dissolution in pyridine, and derivatization. In addition to temperature and solvent-type, factors affecting mutarotation equilibria in aqueous systems include pH and metal ion concentrations (26). Thus sample hydrolysates must be perfectly cleaned and neutralized or there is the risk that sample equilibria may vary distinctly from the standard in aqueous systems. Internal Standard a n d Quantitative Analysis. Adonitol is routinely used as a GC internal standard because it does
502
ANALYTICAL CHEMISTRY, VOL. 56, NO. 3, MARCH 1984
Table V. Standard and Spiked Sample Sugar Recoveries a
plankton u nsp ik e d spiked spike % recovery spruce unspiked spiked spike % recovery sediment u nsp ik e d spiked spike % recovery standard (% recovery) cellulose (% recovery)
LYX
ARA
RHA
RIB
mgx 10 XYL FUC
0.51 2.79 3.01 79.3
1.21 5.87 5.95 82.0
1.69 6.50 5.14 95.2
0.89 3.69 4.83 64.6
1.61 4.83 4.86 74.7
2.26 6.43 6.02 77.6
2.57 12.4 11.9 85.4
0.68 11.0 10.3 100
0.22 5.46 9.65 55.3
0.27 1.45 1.20 87.0
1.37 3.43 2.38 91.2
1.79 3.95 2.06 103
79.3
87.0
93.7
MAN
GAL
GLU
1.92 6.50 4.71 98.1
7.53 29.8 25.0
6.20 14.0 10.9
91.5
81.8
52.1 97.2 59.2 87.4
8.23 15.7 9.72 87.3
0.00
63.3 105 50.0 93.1
9.60 27.2 21.8 86.7
156 250
0.60 1.38 1.93 54.5
1.56 2.71 1.94 77.2
1.28 3.12 98.6
3.41 13.9 9.99 104
3.01 6.49 4.36 88.2
5.54 24.8 23.7 84.9
63.1
80.1
92.2
88.9
83.1
89.7
9.55 9.42 98.2
1.88
118
90.9
80.7
Key: standard, standard monosaccharide solution passed through complete extraction procedure; cellulose, cellulose standard (20 mg) passed through complete extraction procedure, glucose recovery was calculated by using a formula €or cellulose = n(C,H,,O,), a
not have a record of natural ocurrence, it elutes as a single peak in an open “window” at the center of the retention time (RT) range of the measured sugars, and because it is chemi d l y similar to these sugars. Alditols, however, do not contain the relatively reactive carbonyl group found in sugars and therefore should act 89 conservative tracers in relation to sugar condensation reactions. The addition of an internal standard at an early stage in the sugar extraction-preparation procedure requires that all subsequent losses be nonpreferential if quantitative analyses are desired. Satisfaction of this condition allows an arbitrary recovery of 100% to be used for the internal standard peak during quantification. An advantage of this approach is that there is no need for subsequent sample transfers to be quantitative, which would be particularly difficult in the neutralization and desalting steps. Testa were carried out on standard solutions to determine losses of individual sugars, including adonitol, at specific stages of the extraction by the addition of internal standards immediately before and after each stage. The results of these tests are presented in Table IV in the form of “relative” and “absolute”sugar losses for the specified extraction step. These were determined in relation to the first (adonitol) and second (sorbitol) internal standard additions, respectively. The results indicate that no major losses occur at the ion-exchange or rotoevaporation stages and that although “absolute” losses do occur at the centrifugation step, these were accounted for by a gravimetric determination of the amount of sample lost with the wet precipitate following centrifugation. Significant “relative” and “absolute” losses do occur for some sugars at the precipitation step, but there appears to be no loss of adonitol a t any stage (other than centrifugation). Therefore, although nonselective physical sample losses may occur, sugar values obtained with an internal standard level arbitrarily fixed at the original spike value should accurately represent sugar levels in the hydrolysis product. It should be noted that significant physical losses do occur during extraction since some sample transfers are not quantitative and the results in Table IV should not, therefore, be taken as total sugar losses for the extraction procedure (following hydrolysis). Tests on sediment and sediment-trap material samples where a second internal standard was added immediately prior to derivatization show, however, that 60 to 75% of the sample
is recovered, with centrifugation being responsible for 10 to 15% of the indicated overall loss. Table V shows results for samples of plankton, wood, and sediment, spiked with a standard sugar solution and put through the complete extraction procedure. Results for corresponding nonspiked samples are also presented as are percent monosaccharide recoveries which were calculated by difference. These values indicate that the sample matrix does not significantly affect sugar recoveries and that recoveries are similar to those obtained with standard solutions of free sugars through the same procedure. It is questionable whether the percent recoveries for individual sugars shown in Table V are applicable to sugar yields from the hydrolysis of natural samples. These recoveries were determined by using standard free-sugar solutions passed through the extraction procedure, either on their own or spiked into natural samples, and therefore do not necessarily represent accurate recoveries of monosaccharides obtained by the hydrolytic breakdown of natural carbohydrate polymers. However, the observations that: (a) there is efficient hydrolytic breakdown of cellulose under the analytical conditions used, (b) glucose recoveries from cellulose and standard solutions are similar, and (c) major matrix effects on monosaccharide recoveries apparently do not occur, indicate that this type of back-calculation to “original levels” may be legitimate. Because the percent recoveries determined by spiking free sugars into natural samples do not include consideration of the extent of the hydrolytic breakdown of all the parent polymers, such back-calculated values might, as a precaution, be best considered as minimum absolute levels. Analytical Sensitivity and Reproducibility. The FID has a sensitivity for individual Measi-monosaccharide derivatives on the order of 0.1 ng. With a minimum sample volume of 250 p L , a maximum injection volume of 5 pL, and a 1/50 split of the injection volume, this sensitivity corresponds to a detection limit of about 0.1 pg of a monosaccharide in a typical sample hydrolysate. Procedural blanks (containing all reagents but no sample) routinely yield less than 1pg for each monosaccharide except glucose which can reach 10 pg/sample levels in some blanks. However, for the prescribed sample sizes this blank value does not exceed 3% of the total glucose yield and, therefore, no blank corrections are made.
ANALYTICAL CHEMISTRY, VOL. 50, NO. 3, MARCH 1984
503
Table VI. Reproducibilitya LYX
ARA
RHA
mg of sugar/100 mg of organic carbon RIB XYL FUC MAN
GAL
GLU
TCH,O
Sediment SMD %SMD
0.16 i 0.02 12.5
1.16 t 0.10 6.62
1.33 t 0.05 3.76
0.37 i: 0.07 18.9
0.02 t 0.00 0.00
0.07 i 0.01 14.3
0.08 t 0.00 0.00
0.30 i: 0.02 6.67
1.01 t 0.12 11.9
1.90 t 0.07 3.68
0.29 t 0.01 3.45
0.04 i: 0.00 0.00
1.18 i 0.06 5.08
0.99 k 0.08 8.08
2.31 t 0.18 1.79
2.38 t 0.15 6.30
4.28 t 0.07 1.64
14.2 k 0.40 2.82
0.22 t 0.01 4.55
0.60 k 0.02 3.33
0.32 k 0.02 6.25
2.06 0.05 2.43
3.75 t 0.12 3.20
0.00 t 0.00 0.00
30.3 t 0.45 1.48
4.13 t 0.06 1.45
85.8 i. 1.43 1.63
128 i. 0.8 0.62
Plankton SMD %SMD
3
SMD %SMD
0.08 t 0.00 0.00
Wood 4.81 t 0.09 1.87
a Sediment, 10-12 cm interval from Dabob Bay box-core, triplicate, 500 mg, 2.54% organic carbon. Plankton, 1.7 mm mesh sieve fraction, combined phyto- and zooplankton tows from Dabob Bay, duplicate, 200 mg, 41.2% organic carbon. Wood, Douglas fir trunk wood, duplicate, 50 mg, 46.8% organic carbon. TCH,O, total carbohydrate. Reproducibilities are sample and percent sample mean deviations (SMD and %SMD).
Replicate analyses were carried out on different sample types to determine analytical reproducibilities (Table VI). The samples presented were selected as being representative of reproducibilities for their types. Sugar values are normalized to organic carbon and thus corrected for differences in sample size. Reproducibility varies with sample type, being generally better for wood and plankton samples than for sediment samples. This result should be expected due to the lower organic carbon and carbohydrate levels in the sediment samples. Sample mean deviations are found to be less than 10% of the mean for all but the most minor component sugars and consistently under 5% for the major sugars (e.g., glucose, xylose, mannose, and galactose). Precision only decreases for the minor sugars (e.g., lyxose and ribose, Figure lb) when the peaks to be quantified approach base line levels or are overshadowed by large adjacent peaks. Further Considerations. Gas chromatography using a standard hydrogen FID was chosen for this technique because it offers sufficient detection sensitivity and because such instrumentation is common to many laboratories. The sugars elute quickly allowing increased numbers of samples to be analyzed, and the capillary column offers excellent peak resolution. In addition, the MeaSi-sugar derivatives show no tendency to produce asymmetrical peaks and exhibit minimal peak distortion due to injection overload. Quantification has been found to be possible (within f5%) with internal standard peaks as much as 60 times larger than the sugar peak to be quantified. Both standards and samples, once derivatized, are stable within analytical reproducibility for a t least 4 months when stored at room temperature in pyridine. None of the chromatography problems noted by Bethge et al. (4) for pyridine as a solvent were found under the present chromatographic conditions. Except for some plankton samples, base lines were found to be clean with few unidentifiable peaks, particularly in the retention time range of interest. It was initially assumed that this technique would apply equally well to ketoses as to aldoses. However, hydrolysis tests show that ketoses (e.g., fructose and sorbose) consistently give low recoveries through this stage of extraction (less than 5% with standard sugar solutions). Also, both frudose and sorbose were slow to equilibrate in the 0.2% w/v LiC104 pyridine solution and their anomeric distribution were found to be particularly unstable in the presence of foreign compounds (e.g., glycerol or amino acids). Quantitative analysis based on a single anomer peak was therefore found to be unsatisfactory for these sugars.
Because lyxose is not a commonly reported hydrolysis product of wood, we have tested for its production via C2 epimerization of xylose. We find 5% conversion (occurring during or prior to the neutralization step). No measurable conversion of any other major sugar was detected. A number of neutral small molecules other than aldoses may be produced from the hydrolysis of natural samples. For example, mannitol has been found as a major component of brown algae (27) and glycerol is a hydrolysis product of fats (28). The analytical method described here involves minimal chemical manipulation and thus preserves the original chemical diversity of the hydrolysate mixtures. For this reason the technique may prove useful in studies of a variety of neutral compounds in addition to sugars.
Registry No. LiC104,7791-03-9; galactose, 59-23-4; glucose, 50-99-7; xylose, 58-86-6;manncse, 3458-28-4; rhamnose, 3615-41-6; fucose, 2438-80-4; arabinose, 10323-20-3; lyxose, 1114-34-7; ribose, 50-69-1; cellulose, 9004-34-6. LITERATURE CITED Mopper, K. Mar. Chem. 1077, 5 , 585-603. SWeeley, C. C.; Bentley, R.; Makita, M.; Wells, W. W. J. Am. Chem. SOC. 1063,85, 2497-2507. Sawardeker, J. S.; Sloneker, J. H. Anal. Chem. 1065, 3 7 , 945-947. Bethge, P. 0.; Holmstrom, C.; Juhlln, S. Sven. PapperstMn. 1066, 6 9 , 60-63 --
Modzeleskl, J. E.; Laurle, W. A.; Nagy, B. Geochim. Cosmochim. Acta 1071, 3 5 , 825-838. Brower, H. E.; Jeffrey, J. E.; Folsom, M. W. Anal. Chem. 1066, 6 9 , 60-63. Oates, M. D. G.; Schrager, J. J. Chromatogr. 1087, 2 8 , 232-245. Zanetta, J. P.; Breckenrldge, W. C.; Vincendon, G. J. Chromafogr. 1072, 6 9 , 291-304. Eklund, G.; Josefsson, 6.; Roos, C. J. Chromafogr. 1077, 142. 575-565. Crowell, E. P.; Burnett, B. B. Anal. Chem. 1067, 39, 121-124. Borchardt, L. G.; Piper, C. V. Tappi 1070, 5 3 , 257-260. Hoizer, G.; Oro, J.; Smith, S. J.; Doctor, V. M. J. Chromatogr. 1080. 194, 410-415, Ochla, M. J. Chromatogr. 1080, 194, 224-227. Sweet, M. S.; Perdue, E. M. Environ. Sci. Techno/. 1082, 16. 692-898 ..- .._
(15) Petersson, G. Carbohyd. Res. 1074, 33, 49-61. (16) Mason, B. S.; Slover, H. T. J. Agric. FoodChem. 1071, 3 , 551-554. (17) Reid, P. E.; Donaldson, 6.; Secret, D. W.; Bradford, B. J. Chromafogr. 1070, 47, 199-208. (18) Larsson, L.;Samuelson, 0. Mlkrochim. Acta 1087, 2 , 328-332. (19) Mopper. K. Ph.D. Thesis, Massachusetts Institute of Technology, 1971. (20) Mopper, K.; Johnson, L. J. Chromafogr. 1083, 256, 27-38. (21) Rogers, M. A. Geochlm. Cosmochlm. Acta 1085. 2 9 , 163-200. (22) Swain, F. M.; Bratt, J. M. I n "Advances in Organic Geochemistry, 1971"; Gaertner, H. R., Wehner H., Eds.; Pergamon Press: Oxford, 1922; pp 415-425.
504
Anal. Chem. 1984, 56,504-510
(23) Saeman, J. J.; Moore, W. E.; Mitchell, R. L.;Mlllett, M. A. Tappl 1854, 37, 336. (24) Dawson, R.; Mopper, K. Anal. Biochem. 1878, 8 4 , 166-190. (25) Hodge, J. E. J . Agric. Food Chem. 1853, I , 926-943. (26) Dutton. 0.G. S. Advan. Carbohyd. Chem. Biochem. 1873, 30, 38-40. (27) Percival, E.; McDowell, R. H. “Chemistry and Enzymology of Marine Algal Polysaccharides”; Academic Press: New York and London, 1967; pp 1-10, 64-96, 190-194. (28) Cowle, G. L.; Hedges, J. I., submltted for publication in Geochlm.
.
Cosmochlm Acta.
RECEIVED for review July 25, 1983. Accepted November 7 , 1983. This research was funded by National Science Foundation Grants OCE-8023970 and OCE-8219294. This is Contribution No. 1353 from the School of Oceanography, University of Washington, Seattle, WA.
Synthesis and Characterization of Polymeric C,*Stationary Phases for Liquid Chromatography Lane C. Sander* and Stephen A. Wise
Organic Analytical Research Division, Center for Analytical Chemistry, National Bureau of Standards, Washington, D.C. 20234
The synthesls of monomerlc, polymeric, and “ollgomerlc” C,( alkyl phases Is described for a serles of wide pore (300 A) siilca substrates. Chrornatographlc propertles of the phases are compared by use of polycyclic aromatic hydrocarbon (PAH) probes. A three-component test mlxture was used to evaluate the relatlve polymerlc nature of a glven phase. On the basis of the elutlon order of the components of thls mixture, the phase type could be classlfled rapldly and the selectlvlty toward more complex PAH mlxtures could be predlcted. Selectlvlty was observed to be related to surface coverage values while absolute retention was found to be more closely related to the total carbon contalned wlthln the column. Although In past work Intentional polymerlzatlon has usually been avoided In the preparations of alkyl-bonded phases, the unique selectlvlty of polymerlc phases makes them an excellent complement to monomeric phases.
T o a large extent, the chromatographic properties of a reversed-phase sorbent are dependent on the conditions of the bonded phase synthesis. Understandably, the preparation and study of alkyl bonded phase materials have received considerable attention in the literature (1-10). A large portion of this work has been concerned with monomeric bonded phases. Monomeric phases result from the reaction of monofunctional silane reagents with silanol sites at the silica surface. Since only one bond is formed per silane molecule, monomeric surface modification results. Di- and trifunctional silane reagents may also be used to produce monomeric phases, if precautions are taken to exclude water. When water is not excluded, silane hydrolysis and polymerization are possible, forming a polymeric bonded phase. Although much research has been performed on monomeric phases, relatively little effort has been expended in the study of polymeric phases. The reluctance of workers to accept polymeric phases is probably the result of difficulties reported on early pellicular materials. Problems have included low column efficiency due to mass transfer limitations (11-13), poor peak shape (14),and difficulties with reproducibility from one synthesis to the next (15). Recently, Verzele and Mussche (16) synthesized a series of CIS polymeric phases on totally porous silica. They concluded that most polymeric phases
with surface coverages less than 3.5 kmol/m2 are actually monomeric in nature, but differences in polarity between the phases do exist. Other kinds of polymeric phases have also been produced for use in size exclusion chromatography (17) and ligand exchange chromatography (18). From a theoretical point of view, polymeric phases represent a more complex system than monomeric phases. Monomeric bonded phases are often described in picturesque terms such as “bristles”, “brushes”, or “molecular fur”. Polymeric phases are difficult to visualize because little is known about the extent of cross-linking and the degree of polymerization of the alkyl chains. The effect of polymerization on retention processes is not well understood, although selectivity differences among monomeric and polymeric alkyl sorbents have been reported (19-21). The aim of this study was to investigate the preparation of polymeric bonded phases and to examine some of the unique properties of these sorbents. A series of CISmonomeric and polymeric phases have been produced, as well as an intermediate class of phases which we designate as “oligomeric phases”. In addition, a simple empirical LC test is described for determining the relative polymeric nature of a CISbonded phase.
EXPERIMENTAL SECTION Reagents. Silane reagents were purchased from Petrarch Systems, Inc. (Bristol, PA), and were used without further purification. Chromatographic grade solvents were used in all syntheses, wash procedures, and LC separations. The following silica materials were used in preparation of the bonded phases: 8-pm Zorbax 300 silica (E.I. du Pont de Nemours and Co., Wilmington, DE), 5-pm Hypersil WP-300 (Shandon Southern Instruments, Sewickley,PA), 10-pmVydac TP silica (Separations Group, Hesperia, CA), 10-pm LiChrospher 300 (MCB, Gillstown, NJ), and 10-pm Protosil300 silica (Whatman Chemical Separations Inc., Clifton, NJ) (see Table I). A 16-component PAH mixture, SRM 1647 (National Bureau of Standards, Washington, DC), was used to evaluate the columns. Phenanthro[3,4-c]phenanthrene and benzo[a]pyrene were obtained from Aldrich Chemical Co. (Milwaukee, WI) and 1,2:3,4:5,6:7&tetrabenzonaphthalene (dibenzo[g,p]chrysene)was obtained from Rutgers (Caatrop-Rauxel,Federal Republic of Germany). Carbon analyses were performed at the Center for Analytical Chemistry using a LECO CS-244 elemental analyzer. Carbon determinations were made on both bonded and unbonded silica substrates and corrections were made for the carbon content of the unbonded silica.
This artlcle not subject to U S . Copyrlght. Published 1964 by the Arnerlcan Chemlcal Society