Quantitative Determination of Glucose and Galactose HANS FISHER’, R. G. HANSEN, University o f Illinois, Urbana, 111.
end
HORACE W. NORTON
to the spectrophotometer tube maintained in an ice bath. Immediately after addition of the sulfuric acid the solution is thoroughly mixed Tyith the aid of a capillary stream of nitrogen. Under these conditions no appreciable color develops without subsequent heating, which can then be best accomplished in a constant temperature bath for a fixed time.
A rapid and sensitive method, based upon the orcinol procedure of Briiclcner, is used for the simultaneous estimation of glucose and galactose in mixtures. Results are usually within 0.03 mg. for glucose and 0.01 mg. for galactose in the concentration range of 0 to 0.150 mg. of each sugar. Two systems for evaluating the data are presented.
SOLUTIOZT S
Sulfuric acid, 3N. Sulfuric acid, 92% (94 ml. of concentrated sulfuiic acid, specific gravity 1.84, and 6 ml. of water). Orcinol, 2 grams of orcinol in 50 ml. of water, to which is added a cooled solution made by miving 20 ml. of concentrated sulfuric acid and 30 ml. of water. Standard glucose and galactose solutions containing 0.1 mg. per ml. in 3147 sulfuric acid.
THE
simultaneous quantitative determination of glucose and galactose in biological materials is of importance in the study of galactose and lactose metabolism in various animals. Galactosemia in infants (1, 12, 25) and the etiology of Gaucher’s disease, where a glucosidic cerebroside is found in addition to the normal galactose derivative ( 7 , 14, 17-19, 21) are specific examples of application of such a procedure. A variety of techniques has been used for qualitative or approximately quantitative determination of these two sugars. Fermentation by selective yeasts has been extensively used (15, 17, 18, 24). The removal of glucose by glucose oxidase has also been employed ( 2 3 ) for the determination of galactose in blood plasma. A number of chromogenic tests have been applied to this problem: carbazole reactions (8, 9, I S ) , o-tolylhydrazine (11), sulfuric acid-orcinol (2-6), and reducing sugar equivalents of the two hexoses obtained by the methods of Folin-Wu and Sumner (7, 10). Finally, paper chromatographic procedures have also been recommended ( 1 , 6 , 2 3 ) . For quantitative estimation of these two sugars, most of the above procedures lack specificity or are too time-consuming for ordinary application. In some preliminary work the orcinol procedure of Bruckner seemed to offer the greatest promise of quantitative adaptation. This report records a modification of Bruckner’s method which permits rapid simultaneous estimation of glucose and galactose.
Table I. Glucose. hIg. 0 0 0 0 0.050
0.100 0.150 0,200 0.050 0.050 0.050 0.100
0.150 0.100
PROCEDURE
An aliquot of the sample containing up to 0.2 mg. of total sugar is placed in a spectrophotometer tube and adjusted to a volume of 2 ml. with water and enough sulfuric acid to bring the solution to 3N. Two milliliters of orcinol solution are next added while the tubes are cooled in a n ice bath. Then 6 ml. of the sulfuric acid reagent are rapidly added and mixing is completed with a stream of nitrogen through a capillary. The capillary is rinsed in 3 X sulfuric acid between immersions. The tubes are then heated exactly 4 minutes in a bath maintained a t 67-68’ C . with good circulation. After 5 minutes’ cooling in an ice bath, color density is estimated at 470 and 560 mp. STANDARD CURVES
A replicated series of tubes containing glucose and galactose individually and in the various combinations in concentrations ranging from 0 to 0.2 mg. per tube was treated as outlined in the procedure. Absorption curves for the individual glucose and galactose solutions are shown in Figure 1. Absorbance readings a t 470 and 560 mp appear to permit the greatest sensitivity of measurement of concentration of the two sugars. The absorbance measurements for a standard series at the specified wave lengths are given in Table I. This standard series was one of several determined a t various times. These series mere in good mutual agreement.
Absorbance of Glucose and Galactose
Galactose, Mg. 0.050 0.100 0.150 0.200 0 0 0
n
6.050 0.100 0 . I50 0,050 0.050 0.100
Absorbance Replicate 1 Replicate 2 470 inp 560 mr 470 mp 560 m p 0.248 0.120 0.240 0.130 0.255 0.480 0.500 0.278 0.685 0.405 0.740 0.373 0,925 0.925 0.525 0.530 0,058 0.065 0.055 0.052 0.128 0.137 0.123 0.110 0.192 0.198 0.178 0.178 0.250 0.264 0,264 0.250 0.180 0.310 0.303 0.188 0.547 0.541 0.313 0.328 0.455 0.765 0.770 0.450 0.238 0.410 0.360 0.268 0.428 0.413 0.302 0.322 0.610 0.640 0.378 0.412
CALCULATION OF RESULTS
Planes passing through the origin fitted by least squares are fi = 1.222 -I-2 . 6 3 ~ ^v = 1.352 4 . 7 3 ~
+
which can be solved for r and y, giving
5
5 = 2 . 1 3 ~- 1.1% = - 0 . 6 1 ~ 0.550
+
where x and y are, respectively, the milligramsAof glucose and galactose in the spectrophotometer aliquot, and u and *v are the estimated absorbances a t 470 and 560 mp. Statistical analysis (Table 11) showed that quadratic surfaces passing through the origin represented the data more satisfactorilv, the fitted functions being
In Bruckner’s method concentrated sulfuric acid is added t o the sample containing orcinol, the heat of solution providing the necessary conditions for color development. As the development of color is rapid and greatly influenced by temperature, adequate mixing and rapid chilling 8 to 20 seconds after mixing are important. RIore prolonged heating periods increase color density but reduce initial differences in the absorption spectra of the two sugars. That the authors were unable to achieve good reproducibility by the suggested procedure was undoubtedly due to slight differences in mixing or length of time before chilling. In order to overcome these objections, all reagents were added
fi 6
= 1.052 = 1.27s
f 2 . 5 7 ~+ 0 . 9 8 ~+~1 . 3 2 ~+~0.32~2
+ 5 . 0 8 ~+ 0.212’
- 0 . 0 6 ~-~2.27~2
The glucose and galactose values given in Table I11 are calculated from these equations. A sample calculation and the confidence limits of the values obtained are given in the discussion of statistical analysis. The error introduced by using the simpler equations never exceeds 0.03 mg., in either x or y, in the region of the standard data, in which the sum of 2 and y (both positive) does not exceed
1 Present address, Poultry Department, Rutgers University, New Brunswick, N. J.
857
ANALYTICAL CHEMISTRY
858
DISCUSSION
Table 11. Multivariate Analysis of Data Given in Table I Source Quadratic surface Deviations
Ervor
I
Degrees of Freedom
Sum of Squares for u
Sum of Products
Sum of Squares for v
5 9 14
2.570681 0.000426 0.002313
4.270116 0.000523 0.002624
6.998016 0.000827 0.003695
I
I
I
8 7
6 w 0 2 5 U m
a 0 4 cn
The nature of the absorption curves for glucose and galactose as shown in Figure 1 serves as an indicator of the usefulness and limitations of the present method. Thus, ribose has an absorption band similar to that of glucose, so that this absorption appears to be relatively nonspecific. On the other hand, galactose has a very specific spectrum and its determination at 470 and 560 mp is extremely sensitive to small changes in the concentration of a substance exhibiting a glucoselike absorption spectrum. In applying the orcinol method to brain extracts for cerebroside sugar determinations, it was found that pentose-containing compounds are present in appreciable quantities R-hich must be extracted. As shown by chromatography, galactose TTas the only sugar present in pentose-free extracts of chicken brain. The orcinol method applied to the same extracts gave small values for glucose besides the expected galactose values, illustrating that other substances besides pentoses absorb in the glucose region. This is not surprising, as the concentrated sulfuric acid employed will produce a slightly yellow color with all organic materials which, depending on their relative concentration with respect to galactose, can influence the glucose values obtained. When applied to pure phrenosine, however, the present method indicated only galactose, as eupected.
m
STATISTICAL ANALYSIS
U
3
The standard data occur in pairs, an absorbance a t 470 mp and a t 560 mp, each pair measured on a single sample. The experimental errors affecting the two data are probably correlated, and multivariate analysis ( 2 2 ) is appropriate. It might be hoped that the data could be represented by two planes passing through the origin, but a priori considerations suggest that curvature is likely to become apparent at the higher absorbances. This was
2 I
0
I
450 480 510 540 570 600 MI1
Figure 1. Absorbance of glucose and galactose measured in Coleman Junior spectrophotometer
0.2 mg. This may suggest that they would often be sufficiently accurate, but errors of 0.01 mg. may be much larger than the experimental errors and so might entirely obscure what would otherwise be easily detected as a statistically significant difference.
1 I
I
e
APPLICATION OF METHOD TO BLOOD FILTRATES
By feeding chicks purified diets containing varying amounts of glucose, galactose, lactose, and corn meal, blood that contained mixtures of sugars ,was available for study. The plasma was deproteinized with zinc and barium hydroxide (Somogyi) and aliquots of the filtrate were taken for sugar determination. Total reducing substance was estimated by a modified Somogyi procedure, in which 0.5 ml. of butanol was layered above the mixture during boiling and the boiling period was extended t o 30 minutes. Separate aliquots of the plasma filtrate were also deionized with a mixture of I R 45 and Dowex 50 (about 10 to 1). They were then concentrated by lyophilization and the aqueous solution was spotted on chromatograms. An ethyl acetate-pyridinewater solvent system (16) was used to develop the chromatograms and aniline oxalate (20) was used to locate the spots. The partitioned glucose and galactose values for the blood by the orcinol determination are given in Table 111, together with the total reducing values by Somogyi's method. The agreement obtained points to the accuracy and usefulness of the method. The reproductions of the chromatograms shown in Figure 2 further validate these results.
i
1. 2. 3. 6. 7.
1
6 1
2 1
I
I
I
i
I
I
Figure 2.
3
I
i
I
1
2
i
I
I
7
i
i
I
I
Reproductions of chromatograms of blood filtrates
+ ++
Glucose galactose diet Standard galactose (upper spot) a n d glucose (lower spot) Glucose diet Glucose lactose diet Corn galactose diet
Table 111.
CHO in Diet Glucose Corn meal
+ +
Determination of Glucose and Galactose in Blood 470
mp
0.14 0.11 0.16 0.15 0.14 0.14 0.44 0.37
Orcinol Method 560 Glump cose" 0.16 0.13 0.16 0.16 0.17 0.17 0.70 0.62
0.116 0.097 0.151 0.134 0.106 0.106 0.108 0.063
Galactosea 0.002
Somogyi Method 515 Reducing mp sugar"0.50 0.098
0.002 -0,007 -0,003 0.007 0.007 0.116 0.111
0.44 0.51 0.49 0.50 0.47 0.80 0.65
0.086 0.100 0 096 0.098
Glucose 30% lactose 0.092 Glucose 15% 0.197b galactose 0.170b Corn meal 15% 0.70 galactose 0.33 0.52 0.098 0.080 0.167) 0,100 0.51 Std. glucose 0.31 0.100 Std. galactose a Milligrams/0.05 ml. of chicken blood. b Calculated on basis of ratio of glucose to galactose found by orcinol method.
+
V O L U M E 27, NO. 5, M A Y 1 9 5 5
859
clearly shown by the data, though the curvature is not great. The multivariate analysis of the standard data appears in Table
11. Quadratic surfaces passing through the origin were fitted, so that five constants had to be evaluated from the data. Hence the analysis shows 5 degrees of freedom for the sums of squares and products attributable to regression, the remaining 9 (among 14 locations) for deviations from regression, and 14 degrees of freedom for experimental error, arising from differences between duplicate determinations at each location. Comparison of the covariance matrix for deviations to that for error shows that the quadratic surfaces fit the data satisfactorily, and the deviation and error matrices map be pooled to provide estimates of error based on 23 degrees of freedom. These data provide an interesting example of the effectiveness of multivariate analysis. I t was found that two planes through the origin did not fit satisfactorily. However, judged separately, each passed as near t o the origin as it should if the true surface \?-ere a plane through the origin. Only when the two planes were tested simultaneously did it appear that the fit was inadequate. In fact, the intercept for one plane was found to be positive and for the other negative, incompatible with the very high correlation in the experimental errors. This was discernible only by multivariate analysis. The fitted quadratic surfaces can be used as indicated above to estimate the glucose and galactose concentrations in an unknown. Furthermore, the precision of the estimates can be expressed in terms of the several derivatives and the precision of the standard data. To a first approximation,
+ B2p(~')] + ( A D + BC) ~ ( u z ' -) A B M ( v ' ) ] = M 2 [ C 2p ( u 2 ) - 2AC ~ ( u v + ) A' p ( ~ ' ) ]
~ ( 2 ' )=
M2[D2p(u2)- 2BD/i(uv)
-CD
p ( ~ y )=
p(y')
p(~')
where ~ ( 2 2 )is the variance (mean squared error) of the estimate of x, and p(xy) is the covariance of the estimate8 of 5 and y, and A, B, C, D, and M are defined below. The variances and covariance of u and v come from the error line of Table 11, or better from pooling deviations znd error. This gives p ( u 2 ) = 0.000119, p(ut) = 0.000137, k(v2) = 0.000197, on 23 degrees of freedom. The approximate variances and covariance of the estimates of x and y will be satisfactory approximations if Jf is several times as large as its sampling error. For the standard data used her?, M exceeds 100 times its sampling error. Given obsei ved values of u and 2: for an unknown, the quadratic equations il = 1 . 0 5 ~ 2 . 5 7 ~ 0 . 9 8 ~ ~1 . 3 2 ~ ~0 32y2 i = 1.272 5.085 0.212' - 0 . 0 6 ~-~2 . 2 7 ~ '
+ +
+ +
+
+
can he solved for x and y. This is most easily done arithdu du dv dv metically. Letting - = A, B, = C, - = D, and
dx
BD - BC
=
& =
xx
dY
1/M, corrections to assumed values of x and y are
Ax = M(Dau - B A V ) Ay = h f ( A A V - CaU) where
Au=u-;L
Av=v-$ This process starts conveniently a t z = y = 0 = j = v^, and needs no more than two repetitions to arrive a t estimates of x and y which are substantially as precise as the data can furnish. For an example of the calculations, take u = 0.140, v = 0.160 (line 1 of Table 111). Startingfromx = y = 0 = 6 = 8 . thefirst values of A , B , C, D, and M are, respectively, 1.05, 2.57, 1.27, 5.08, and 0.483, and Au and Av are 0.140 and 0.160. Hence Ax = 0.145 and Ay = 0.005. Taking x = 0.145, y = 0.005, il = 0.187,; = 0.214, A = 1.34, B = 2.76, C = 1.33, D = 5.05, Ji' = 0.323, and Au = -0.047, Av = -0.054. Hence Ax = -0.029, Ag = -0.003, and x = 0.116, y = 0.002. These values of x and y give u = 0.1404 and 2) = 0.1603, so no further improvement is needed. The new values of A , B , C, D, and M are 1.28, 2.72,
1.32. 5.06. and 0.346. These mav be used to calculate the variancis a n d covariance of 2 and 0.- I n particular, the varianle of 2 is given by p(x2) = [(5.06)2(0.000119)- 2(2.72)(5.06)(0.000137) (2.72)2(0.000197)1(0.346)2,which comes to 0.0000878. Taking t h e square root, the sampling error of P is found to be 0.0094 Multiplying by 2.069, the 5% value of t for 23 degrees of freedom, we find that values within 0.019 of P are acceptable at the 5% level-that is, the 95% confidence interval is 0.097 5 x 5 0.135. The variance of y is 0.00000804, so that y is estimated over ten times as well as x. The covariance is -0.00001215, and may be used to establish confidence limits for x and y simultaneously, or for functions of x and y. For example, limits on the sum or the difference of x and y might be needed. The size of the experimental error may prove to be much affected by interferences. In that event, the error of estimating x and y can be reduced by inferring them from values of u and v which are the averages of replicate determinations. Finally, the extremely high correlation between experimental errors at the two wave lengths suggests that the method is capable of further improvement.
+
CONCLUSIONS
The procedure as modified is sensitive and reproducible for glucose as the only sugar in biological materials and eliminates standards to be run with each determination. It is extremely sensitive and reproducible for galactose as the only sugar in purified materials (phrenosine), but suffers from interference of sugars other than glucose present in crude brain extracts. It is most useful in the quantitative determination of both glucose and galactose, particularly in blood but probably also in other materials. Statistical analysis shows that results should nearly always be within 0.03 mg. for glucose and 0.01 mg. for galactose in the concentration range specified. ACKNOWLEDGMENT
H. E. Carter, Chemistry Department, University of Illinois, kindly supplied the phrenosine. Financial support was received from the U. S. Atomic Energy Commission contract No. AT (11-1) - 67, project KO.10. LITERATURE CITED
(1) Bickel, H., and Hickmans, E. hl., Arch. Disease ChiZdhood, 27, 348 (1952). (2) Bruckner, J., 2. physiol. Chem., 268, 163 (1941). (3) Ibid., p. 251. (4) Ibid.. 275. 73 (1942). Ibid.; 377; lSl'(1943). Chargaff, E., Levine, C., and Green, C., J . B i d . Chem., 175, 67 (1948).
Danielson, I. S.,Hall, C. H., and Everett, 11. R., Proc.
Soc.
ErptZ. B i d . Med., 49, 569 (1942). Dische, Z., Shettles, L. B., and Osnos, bl., Arch. Riochem., 22, 169 (1949). Edman, P. V.,J . BioZ. Chem., 143,219 (1942). Everett, M. R., and Edwards, B. G., Ibid., 100, xlii (1933). Fowweather, F. S.,Biochem. J., 55, 718 (1953). Goldbloom, A., and Brickman, H. F., J . Pediat , 28, 674 (1946). Gurin, S.,and Hood, D. B., J . B i d . Chem., 131, 211 (1939). Halliday, N., Deuel, H. J., Jr., Tragerman, L. J., and Ward. W. E., Ibid., 132, 171 (1940). Harding, V. J., and Grant, G. A.,Ibid., 94,529 (1931). Jermyn. XI. A., and Isherwood, F. 8.,Biochem. J., 44, 402 (1949). Klenk, E., 2. physiol. Chem., 267, 128 (1940). Lieb, H., and Gunther, V., Ibid., 271, 211 (1941). Ottenstein, B., Schmidt, G., and Thannhauser, S.J., Blood, 3 , 1250 (1948). Partridge, S. LI., Biochemical Society Symposia, No. 3, Cambridge University Press, London, 1950. Polonovski, J., Bull. SOC. chim. biol., 25, 44 (1943). Rao, C. R., "Advanced Statistical Methods in Biometric Research," Wiley, New York, 1952. Rutter, W. J., Krichevsky, P., Scott, H. M., and Hansen, R. G., Poultry Sci., 32, 706 (1953). Sie, Hsien-Giels, Faust, C. E., and Williams, H. H., Federation Proc., 7, 161 (1948). Townsend, E. H., Jr., Mason, H. H., and Strong, P. S.,Pediatrics,7, 760 (1951).
RECEIVED for review April 19, 1954. Accepted December
1 0 , 1954.
