Chemiluminescent Enzyme Method for Glucose John P. Awes,’ Steven L. Cook,2 and J. T. Maloy3 Department of Chemistry, West Virginia University, Morganto wn, W. Va. 26506
In this determination, the oxidation of glucose in the presence of glucose oxidase at pH 7 provides a source for H202 which, in turn, reacts with Fe(CN)63- in the presence of luminol (3-aminophthalhydrazide) to produce chemilumlnescence (CL) proportional to the initial glucose concentration. CL emission is observed only in basic solutlon, and an ammonia buffer of high capacity is used to rapidly adjust the pH of the solution from levels of high enzyme activity to levels of detectable CL emission. The integrated CL slgnal obtained after five minutes of enzyme activity was linear with glucose concentration in 10-pl samples over the entire range of ordinary clinical Interest (2-200 mg/dl). Because the method relies on an emissive process rather than absorbance, it may find applicability in the rapid determination of small quantities of glucose in biological tissues.
Since its discovery by Albrecht in 1928, ( I ), the analytical utility of the chemiluminescence (CL) of luminol 0
has been the subject of extensive study. Much of this work has been cited in recent reviews (2, 3 ) which trace the development of CL methods of analysis. As early as 1951, the utility of luminol as a chemiluminescent acid-base indicator was studied by Kenny and Kurtz ( 4 ) . As an indicator, its fluorescent properties have also been used in the titration of opaque substances such as milk and red wine ( 5 ) . Trace metal analyses employing luminol chemiluminescence have been developed for Co(II), Cu(II), Ni(II), and Fe(I1) (6). In these studies, the metal ions are usually regarded as catalysts for the process producing chemiluminescence so that the CL intensity may be related to the metal ion concentration. In this work, a novel analytical technique for the determination of glucose is presented. As a means of measuring the quantity of glucose present, it employs the luminol chemiluminescence t h a t results from reaction with peroxide generated in the p H dependent enzyme catalyzed oxidation of glucose ( 7 ) Glucose
+
0,
+ H,O
Glucose oxidwe
Gluconic acid
+
H,O,
(1)
Thus, it combines the advantages of enzyme specificity with CL emission sensitivity. Luminol chemiluminiscence may be produced through
’
Present address, Alcoa Technical Center, Alcoa Center, Pa. 15;69.
- Present address, Department of Chemistry, Ohio State University, Columbus, Oliio 43210. ’ T o whom correspondence should be addressed. 244
the action of a n oxidizing’agent in basic solution. Over thirty years ago, Stross and Branch (8)reported dim CL of short duration upon mixing luminol and K$’e(CN)6 in basic solution and dim CL of longer duration upon mixing luminol and H202 under the same p H conditions. However, bright CL was observed when both H202 and K3Fe(CN)6 were mixed in the presence of basic luminol. Because of this, it was deemed advantageous to increase the sensitivity of this analysis by carrying it out in the presence to K3Fe(CN)G, even though this could result in background CL due directly to the ferricyanide. Those early authors recognized that the peroxide reduction of ferricyanide in basic solution H20,
+
2Fe(CN),3-
-
2H’ + 2Fe(CN)64-+ 0, (2)
was important in the light producing sequence, but failed to satisfactorily elucidate the mechanism of the reaction. Some twenty years later, White (9) proposed that this reaction of hydrogen peroxide and a metal [complex] served two purposes: it supplied oxygen and it supplied hydroxy or hydroperoxy radicals. These, in turn, according to White’s mechanism, reacted with the monoprotic anion of 1 to produce the aminophthalate ion
along with the concomitant emission of luminescence similar to the fluorescence of 2. The assignment of this mechanism in aqueous solutions was based on a series of experiments conducted in dimethyl sulfoxide-water mixtures in which CL was observed when oxygen was passed through solutions containing 1 and a strong base (9-11). These media are sufficiently basic to permit the existence of the luminol dianion 0
(3)
Product analysis coupled with experiments carried out with I*O enriched oxygen gas revealed that 3 was stoichiometrically converted to 2 in the overall reaction
3
+
0 2
N, + 2 + h~
(3)
as CL was produced. Because 2 has been found to exhibit similar CL and fluorescence emission spectra in aqueous and nonaqueous solutions, the aminophthalate ion has come to be regarded as the emitting species in luminol CL, even though the aqueous fluorescence spectrum of luminol in acid solutions resembles the aqueous CL spectrum in basic solutions.
A N A L Y T I C A L CHEMISTRY, VOL. 47, NO. 2, FEBRUARY 1975
Linschitz has proposed that luminol fluorescence and luminol CL come from the same state in aqueous solution despite this difference in p H dependence (9). (He has pointed out that it is the pK, of the excited state that is important in evaluating the p H dependency of luminescent acid-base systems.) If this observation is correct and if reaction 2 generates appreciable quantities of singlet oxygen, the CL observed could result from direct energy transfer from excited oxygen molecular pairs. Kahn and Kasha have proposed this general theory of energy transfer from dimeric forms of excited oxygen and have produced spectroscopic evidence of a dimeric 2[l2,+] state with sufficient energy (75-83 kcal/mol) to populate the excited state of luminol (12). McKeown and Waters, however, have argued against the formation of singlet oxygen from two single electron transfers (as in the case of reaction 2) on mechanistic grounds (13). Thus, attributing luminol CL to energy transfer from singlet oxygen generated in reaction 2 is only speculative. If this speculation is correct, however, it serves to emphasize the role of reaction 2 in the CL process, just as White’s original mechanism did by implication. Since the extent (and, in a mechanistic sense, rate) of reaction 2 is pH-dependent, this dependency, which is critical to the analysis, is investigated below. T h e reaction of ferricyanide with luminol to produce CL in the absence of H202 has been investigated by Shevlin and Neufeld (14 1. Since this reaction could produce CL in addition to that produced by the glucose-generated peroxide, it might set the limit of detectability for the analysis. As a result of their studies, these investigators proposed a mechanism that begins with the ferricyanide oxidation of 3. This is followed by further reaction with oxygen (in which Fe(CN)G+ is regenerated from the Fe(CN)e4- produced in the first reaction) to produce 2 accompanied by chemiluminescence. It is important to note that the CL intensity was reported to decrease linearly with decreasing OH- concentration; extrapolation of these data points indicate that no CL would be expected below p H 12. Thus, one might propose that careful control of p H is necessary to eliminate this undesirable background CL process. T h e importance of p H control cannot be underestimated in this analysis. First, the peroxide-generating glucose oxidation is pH-dependent. Second, the extent of reaction 2 which may play an essential role in the production of CL depends upon the pH. Finally, the elimination of a n undesirable CL side reaction appears likely in mildly basic solutions. T h e aspect of p H control is investigated extensively below.
EXPERIMENTAL Reagents. All reagents were used as purchased without further purification. Luminol (3-aminophthalhydrazide) samples were obtained from Aldrich Chemical Company and Eastman Organic Chemicals. In basic solutions, the sample obtained from Aldrich exhibit,ed a fluorescent (490 nm) contaminant that was separable from the luminol by thin layer chromatography. This impurity was minimally present in the Eastman sample, too, but not in sufficient quantity to interfere with fluorescence measurements; thus, t,hese determinations were conducted with the Eastman sample. Both samples gave identical CL results. Glucose oxidase containing 30,000 enzyme units per gram and anhydrous D-glucose were purchased from Nutritional Biochemicals Corporation. Reagent grade potassium hydroxide, ammonium nitrate, ammonium hydroxide. sodium hydroxide, and nitric acid were supplied by Fisher Scientific Co. AutoAnalyzer Certified Standards from Technicon, containing 50, 100, 150, and 200 mg/dl glucose were used to construct some calibration curves. Apparatus. Chemiluminescence was detected and measured by an Aminco grating monochromator equipped with a Hamamatsu 1P21 photomultiplier t,ube and an Aminco photomultiplier microphotometer. Chemiluminescent intensity was recorded as a func-
I
I
I
0
I
I
I
20
1
40
TIME (sec) Figure 1. Typical CL intensity signals
Zero time is at the instant of pH jump. Curve a shows t h e time dependence of the CL signal itself: b is the integrated signal tion of time on a Houston Instruments Omnigraphic HR-96T recorder. The CL signal was integrated electronically with a BurrBrown 3292/14 operational amplifier powered by a Burr-Brown 524 Dual 15VDC power supply, It was then recorded on a Varian Model G-14 strip chart recorder. Fluorescence studies of luminol were undertaken with an Aminco-Bowman Spectrophotofluorometer (SPF). The SPF was also used to obtain the spectrum of any CL emission; while no attempt was made to extrapolate these spectra to zero time, manual reverse scans detected no essential differences in the spectral distribution. Luminol and glucose oxidase volumes were reproduced for each determination with a 2-ml Minipet syringe supplied by the Manostat Corporation. Procedure. A 4.5mM stock solution of luminol was prepared. Initially, sufficient luminol was dissolved with the aid of a sma!l quantity of base, but, before the solution was brought to volume, it was adjusted to pH 7. Similarly, a glucose oxidase stock solution containing 150 unitdm1 was adjusted to pH 7 before final dilution. Neither of these solutions was buffered. A stock solution of 0.002M K3Fe(CN)s was prepared in 1M ammonia buffer at pH 95. Aqueous glucose standards were prepared by dilution of a stock solution containing 200 mg/dl glucose. Blood serum samples were obtained from the West Virginia University Medical Center. These samples had been previously analyzed for glucose with a Technicon Sequential Multiple Analyzer (SMA 6/60). In a typical run, a 10-pl sample of glucose solution was introduced into a cuvette containing 1.6 ml of an unbuffered mixture of glucose oxidase and luminol at pH 7. (This sample volume is typically used for glucose determinations in biological tissues.) Following agitation to ensure mixing of the reagents, the cuvette was placed in the detector cell compartment. A delay time of five minutes was allowed to elapse for hydrogen peroxide to build up in the system. While the light output was being monitored, 0.8 ml of 1M ammonia buffer containing K3Fe(CN)e was rapidly injected into the cuvette. The chemiluminescence obtained immediately upon buffer-ferricyanide addition was recorded as a function of time. Simultaneously, the integrated CL intensity was obtained and recorded. Typical results are shown in Figure 1; the integrated CL intensity (curve b ) approached a limiting value at times in excess of 40 seconds.
RESULTS Figure 2 shows the p H dependence of several important variables considered in this analysis; since different variables are shown as a function of p H , a relative ordinate has been used. Curve a shows the enzyme activity of glucose oxidase as reported by Malmstadt and Pardue ( 1 5 ) ;maximum activity occurs between p H 6 and 7, where no luminol chemiluminescence is observed. Some attempts were made to study the rate of reaction 2 in the absence of luminol as a function of pH. These indicated t h a t reaction 2 occurs a t no appreciable rate a t pH 5, occurs slowly a t pH 9, and occurs rapidly a t pH 12. Because these rates also depend upon the concentrations of the reactants, no extensive study of these was undertaken. Instead, to investigate the p H dependence of reslction 2, a
A N A L Y T I C A L C H E M I S T R Y , VOL. 4 7 , NO. 2, F E B R U A R Y 1975
245
T h e results of this calculation, curve b, show t h a t a t low H202 concentration, reaction 2 is not thermodynamically feasible below pH 8. In curve c, a plot of t h e fluorescence intensity of 4.5mM luminol at 432 nm is shown. The decrease of fluorescence in acid solution could be attributed to the protonation of molecular luminol
+I
H,S
(4)
PH Flgure 2.
Dependence
II 0
of experimental parameters upon p H
In all curves, a relative ordinate is used. Curve a: Enzyme activity of glucose oxidase. Curve b: Thermodynamic extent of H202-Fe(CN)e3- reaction. Curves c and d: Fluorescence of luminol; c is for 4.5mM luminol at 432 nm: d is for 0.15mM luminol at 495 nm. Curves e and f: Luminol CL produced by 100 mg/dl sample of glucose in this analysis; f is in the additional presence of equimolar Fe(CN)e4-
while the decrease above pH 6 has been attributed to the formation of the luminol monoanion
T h e fluorescence of what is probably 3 is shown in curve d ; here the p H dependence of the fluorescence of 0.15mM luminol at 495 nm is shown. (In this set of experiments, t h e p H was not determined potentiometrically; rather, it was determined as the quantity log(CB/K,), where CB was the analytical concentration of NaOH employed.) The increase in fluorescence in basic solution may be attributed to the formation of t h e luminol dianion.
5--'Ht+3 I 300
Figure 3.
I
I
I
500 W a v e l e n g t h (nm)
I
701
Luminol emission spectra
Curve a shows the fluorescence of a 4.5mM solution excited at 392 nm: b, the fluorescence of a 0.15mM solution excited at 350 nm. Curve c is the CL spectrum obtained in this analysis.
thermodynamic approach was employed. Using Standard Electrode Potentials, t h e equilibrium constant K for the reaction was determined (16). This was then used to calculate t h e extent of reaction as a function of pH. Curve h in Figure 2 shows [, the thermodynamic extent of reaction 2. This curve was calculated from the equilibrium expression for this reaction
by assuming t h a t the reaction proceeds in an equimolar mixture of Fe(CN)64- and Fe(CN)& t h a t remains constant throughout; Po2 was assumed to be a constant 0.2 atm. T h e initial H202 concentration, [H202]0,required t o be small with respect to t h e ferricyanide concentration by the constraints imposed previously, was arbitrarily taken to be 0.02mM, a typical reaction mixture concentration for glucose in t h e analysis reported below. T h e extent of reaction was determined from these quantities using the relationship
246
(8)
T h e minimum pK2 that may be estimated from curve d is 13.5; this is in agreement with a value determined previously in a potentiometric experiment (8). T h e p H dependence of luminol chemiluminescence a t 473 nm is shown in curves e and f. Curve e was obtained in t h e presence of 7mM ferricyanide; curve f was obtained under identical conditions except that a n equimolar concentration of ferrocyanide was also present. As such, curve f corresponds directly t o the extent of reaction curve. This information supports the following conclusions about the importance of p H control in the CL analysis of glucose: 1) Since maximum enzyme activity occurs in the vicinity of p H 7, the initial oxidation of glucose t o produce H 2 0 ~ should occur in neutral solution. 2) On the basis of thermodynamic arguments, one would not expect reaction 2 to occur extensively below p H 8. Thus, if this reaction is important in t h e CL process, one would not expect luminol below this threshold. T h e fact t h a t CL is not observed below p H 8 supports the contention that reaction 2 is important in the analysis. 3) While one would not expect exact agreement between a thermodynamic curve like curve b and those obtained from a kinetically controlled process like e and f, the fact t h a t the addition of ferrocyanide suppresses the intensity of luminol CL supports the thermodynamic argument given above. Ferrocyanide is a product in the oxygen-releasing reaction. Of course, this effect could also be attributed to ferrocyanide quenching of the excited state. 4) According to t h e arguments presented above, reaction 2 should occur extensively a t p H 8 and it is a t this point that chemiluminescence is observed. However, from curve
A N A L Y T I C A L CHEMISTRY, VOL. 47, NO. 2, FEBRUARY 1975
1
I
I
I
~~~
I
I
~
Table I. Effect of Age of Reagent on Relative Integrated CL Intensitya Fresh luminol
F r e s h glucose oxidase F r e s h K,Fe(CN), 1.oo 0.64 Aged glucose oxidase F r e s h K,Fe(CN) 0.91 0.50 F r e s h glucose oxidase Aged K-Fe(CN) 0.76 0.45 Aged glucose oxidase Aged K,Fe(CN) 0.59 0.28 In all trials, these concentrations were used. luminol, 1 5mM, FelCS),j' , i m M , glucoce oxidase, 50 units/ml Experimental conditions s e r e identical except for age of reagents Aged reagents uere prepared one week prior t o uqe
1.0 0.2
0.4 0.6 0.8
4gedb lummol
1.0
R E L A T I V E CONCENTRATION
Figure 4. Dependence of the integrated analytical CL intensity upon reagent concentration In all cases, the relative concentration is defined as C/C,,,. Curve a: glucose oxidase, with C,,, = 90 unitslml. Curve b: luminol, with C,,, = 2.7mM. Curve c: Fe(CN)e3- with C,, = 12mM
d , t h e minimum value for pKp for luminol would be 13.5. Thus, less than 0.001% of t h e total luminol is present as 3 a t p H 8 where CL is first observed. I n addition, t h e fact t h a t ferrocyanide addition decreases t h e luminol CL at any p H argues against t h e catalytic role of ferricyanide in t h e reaction. Both these observations suggest t h a t t h e mechanism of Shevlin and Neufeld, which requires the existence of 3 and t h e catalytic action of Fe(CN)63-, is inoperative a t t h e p H levels employed in this analysis. Thus, minimization of background CL due to direct reaction of luminol and ferricyanide appears possible through minimization of pH. (This has also been demonstrated by passing air through luminol-ferricyanide mixtures a t different p H levels. At p H 13, CL was detected using this technique in accord with t h e results of Shevlin and Neufeld. At p H 9.5, no CL was detected by merely passing air through t h e mixture. This result suggests t h a t t h e elimination of background CL may be accomplished by using a buffer of minimum p H necessary to induce reaction 2.) 5 ) Since reaction 1 proceeds best a t p H 7, while no lumino1 chemiluminescence is observed below p H 8, t,his analysis may not be conducted a t constant pH. This fact may actually be used to advantage in t h e analysis. Reaction 1 is allowed to proceed for a predetermined time at p H 7; then the solution is rapidly adjusted t o a higher p H where reaction 2 occurs. This pH jump is used t o terminate t h e enzyme reaction while initiating t h e CL reaction. Fluorescent and chemiluminescent emission specira of luminol are shown in Figure 3. Curve a shows t h e fluorescence of a 4.5mM luminol solution excited a t 392 nm. This spectrum, obtained a t p H 6.2, agrees well with t h e reported emission spectrum of luminol ( I ] ) . Its maximum occurs a t 432 nm; the intensity a t this wavelength generally decreases as the p H is increased above 5.0. Curve b shows t h e fluorescence of 0.15mM luminol in 1.0.44 NaOH. This spectrum was obtained a t a n excitation wavelength of 350 nm; the emission maxima occur a t 415 n m a n d 495 nm. T h e peak a t 495 nm generally decreases a t lower NaOH concentrations so t h a t t h e p H dependence exhibited in curve d of Figure 2 is obtained. This emission may be attributed to t h e luminol dianion (3). It is not due t o any sort of luminol
complex with sodium ion; t h e peak at 495 nm is absent in the fluorescence spectrum of 0.15mM luminol in 1.OM NaCl a t p H 10. Curve c shows t h e CL emission spectrum obtained in this method. Some emission is observed at 430 nm, but t h e primary emission is a t 473 nm. This does not agree with t h e aqueous luminol CL spectrum (exhibiting a maximum a t 424 nm) obtained by White (9, 2 1 ) who used Cu(I1) as t h e oxidant. This apparent difference is probably due t o the absorbance of t h e ferricyanide solution a t the wavelength of maximum CL emission. Fluorescence studies conducted on 0.45mM luminol a t p H 7.5 in t h e presence of 1.8mM Fe(CN)s'- produced an emission spectrum t h a t is quite similar to t h e CL emission spectrum shown in curve c Some studies were conducted t o determine the optimum experimental conditions for t h e CL analysis of glucose. Each of t h e following was investigated independently: Effect of Delay Time. Studies of t h e integrated CL intensity as a function of the elapsed time between the instant of glucose addition and t h e instant of p H jump revealed t h a t t h e total CL output increased monotonically with delay time for times u p t o fifteen minutes. cince 9' a readily detectable signal was obtained when five minutes passed between t h e time of glucose addition t o t h e luminolenzyme mixture and the time of buffer-ferricyanide injection, this delay time was used in all subsequent runs. Effect of Reagent Concentration. Figure 4 shows t h e variation in integrated CL intensity as a function of relative reagent concentration; in these experiments, a 10-gl sample of 100 mg/dl glucose was used throughout. Curve a illustrates t h e effect of changing t h e concentration of glucose oxidase. (The maximum concentration of enzyme used in obtaining curve a was 90 units/ml; t h e concentrations of luminol and Fe(CN)G?- were fixed a t 0.9mM and 4.0mM, respectively.) Low level CL emission was recorded when no glucose oxidase had been added to t h e system. (This is discussed below.) As a result of these experiments, t h e glucose oxidase concentration was arbitrarily set a t 50 units/ml in all subsequent analytical work. Curve b shows t h a t a regular increase in integrated CL intensity occurs as the lumino1 concentration approaches the saturation limit. (In curve b, t h e maximum luminol concentration used was 2.7mM; the concentrations of Fe(CN)& and glucose oxidase were held throughout a t 4.0mM and 30 units/ml, respectively.) Because of long-term solubility considerations, 1.5mM luminol was employed in the analysis. Curve c indicates t h a t t h e integrated CL intensity increases with F e ( c N ) & concentration only u p t o 7mM Thereafter, no substantial increase in CL intensity was observed. ( T h e maximum concentration of Fe(CN)63- employed in curve c was 12mM, t h e glucose oxidase concentration was held at
A N A L Y T I C A L CHEMISTRY, VOL. 47, NO.
2, FEBRUARY 1975
247
T a b l e 11. Typical Results of Glucose Determinations in H u m a n Blood Seruma CL analysis
SXL4 6/60 Analysis,b
water
Clinical
standards
standards
standards
95 102 102 135 155
2 00
IO0
0
Glucore-
clinical
8 1 i IC 82 i 3 90 f 1 126 i 3 136 i 3
91 i
1c
92 i 3 101 = 1 141 i 3 152 i 3
Concentrations expressed in mg/dl. Analysis performed at WVU Medical Center; uncertainty unspecified. Average absolute
GLUCOSE CONCENTRATION ( r n g / d l )
deviation,
Figure 5. Calibration curves
Curves a and b were obtained with aqueous glucose standards using different reagents. Curves b and c were obtained with the same reagents, but clinical standards were used for c
30 units/ml while t h a t of luminol was fixed a t 0.9mM.) Since maximum CL emission was obtained with 7mM Fe(CN)e3- in these experiments, this concentration of this reagent was used in t h e glucose determinations reported below. E f f e c t of Age of Reagents. Since some diminution of expected CL intensity was observed in determinations performed with reagent solutions prepared a few days prior to use, studies were conducted to determine the effect of reagent solution age on the integrated CL intensity obtained. T h e results of this study are given in Table I. T h e relative integrated CL intensities shown were obtained using reagents having identical concentrations; aged reagent solutions, however, were prepared one week prior to use. From t h e data presented, it is evident t h a t it is most important t h a t the luminol solution be freshly prepared, although some solution deterioration is observed in each reagent, and maximum CL emission is obtained with fresh reagents. These results indicate t h a t the CL calibration curve should be prepared a t the time of analysis. E f f e c t of Reagent Mixing. T h e analysis was simplified by mixing the luminol solution with the glucose oxidase solution a t p H 7.0 prior to glucose injection; in addition, the K,Fe(CN)e was prepared in ammonia buffer so t h a t this solution also served two purposes. This combination of reagents was judged desirable for the purposes of analysis automation, and these combinations were found to be particularly useful. When luminol was mixed with K2Fe(CN)G or ammonia buffer, a n obvious color change occurred upon standing; this was accompanied by a loss in chemiluminescent activity. These combinations, therefore, were avoided in subsequent attempts to simplify t h e analytical procedure. After these optimum analytical conditions had been established, the method was used t o determine the glucose content of some aqueous glucose solutions and some previously analyzed human serum samples. Calibration curves were constructed using either standard glucose solutions prepared in this laboratory or clinical standards prepared specifically for use with t h e Technicon SMA 6/60 employed in routine glucose analyses a t t h e West Virginia University Medical Center. These calibration curves are shown in Figure 5 . Previously, it was noted t h a t some CL emission was observed in the reaction mixture even in the absence of glucose oxidase (see Figure 4,curve a).This would lead one to suspect t h a t some background CL process could contribute to the overall emission and thereby fix a lower analysis limit. Because no CL was observed a t p H 9.5 when air was 248
ANALYTICAL CHEMISTRY, VOL. 4 7 ,
bubbled through a mixture of Fe(CNjB3- and luminol, it is unlikely that this CL was due to the direct reaction of lumino1 and ferricyanide. This evidence suggests t h a t the glucose may react directly with luminol and ferricyanide to produce low-level CL. The calibration curves obtained support this contention; very little CL was ever observed at zero glucose concentration. I n particular, curve a investigates t h e behavior of solutions containing small amounts of glucose; the slight amount of CL obtained a t zero glucose concentration corresponds to a detection limit of 1-2 mg/ dl. Calibration curves b and c were prepared a t different times using a different set of reagents than were used t o prepare curve a. T h e same relative integrated intensity scale is used in Figure 5 t o give some idea of the absolute reproducibility of the method. Curve b was constructed using aqueous glucose standards prepared in this laboratory; curve c was prepared using AutoAnalyzer Certified Standards. CL intensities obtained using clinical standards were slightly lower than those obtained using simple glucose solutions. Since the clinical standards also contained known amounts of urea for use in another standardization, this could indicate t h a t urea interferes with t h e determination; it could also indicate t h a t the clinical standards adjusted the unbuffered glucose oxidase-luminol solution to a slightly less favorable p H for enzyme activity than the aqueous glucose solutions did. Because the curves did agree within typical clinical uncertainty, no further investigations were conducted to determine t h e exact cause of this slight variation. Finally, to test the clinical applicability of t h e method, human serum samples were analyzed for glucose content using t h e calibration curves shown in Figure 5 ( b and c j. T h e results of these studies are shown in Table 11. In general, the CL method gave slightly lower results than the SMA 6/60, although better agreement was obtained with the clinical data when AutoAnalyzer standards were used. This is not unusual. Since the SMA 6/60 employs t h e reduction of Cu(I1) in its analysis, results obtained clinically with it give a measure of the total reductant in the serum. Thus, this clinical analysis lacks the specificity of t h e CL enzyme method. Lower results are generally expected in an enzyme analysis ( 71.
DISCUSSION Reported herein, then, are the results of the initial experiments performed in the development of a n inexpensive, rapid, enzyme-specific, CL method for glucose. Because the analysis relies on an emissive process rather than absorbance, its sensitivity may be increased to the limit of background emission merely by increasing photomultiplier sensitivity. The results obtained above indicate that this limit
NO. 2, FEBRUARY 1975
corresponds to approximately 2 mg/dl in a 10-pl sample. Because this sample is diluted to more than 1 ml in the analysis, this detection limit is actually less than 1 ppm of the total analysis volume. T h e analysis time is generally less than 10 minutes. A unique feature of the method is the p H jump. This technique allows the enzyme to catalyze the reaction of glucose a t a favorable p H for a known length of time. T h e sudden change of p H to a level favorable to CL emission serves as a timing device for the enzyme reaction. Because of this, the method seems particularly amenable to automation; this aspect is currently under consideration. (Concurrent with this research, Bostick and Hercules (17) have independently developed an automated method for CL glucose analysis employing similar chemical techniques.) In addition, the extension of this technique to other oxidase enzyme systems is anticipated. The method would seem to be applicable to any substrate that yields hydrogen peroxide when it undergoes enzyme catalysis in neutral solution.
LITERATURE CITED H. 0. Albrecht, Z.Phys. Chem., 136, 321 (1928). W. R. Seitz and M. P. Neary, Anal. Chem., 46, 188A (1974). U. lsacsson and G. Wettermark, Anal. Chim. Acta, 68, 339 (1974). F. Kenny and R. Kurtz, Anal. Chem., 23, 331 (1951). L. Erdey, I. Buzas, and K. Vigh, Talanta, 13, 463 (1966). W. R. Seitz and D. M. Hercules, in "Chemiluminescence and Bioluminescence," M. J. Cormier, D. M. Hercules, and J. Lee, Ed., Plenum Press, New York, N.Y., 1973, pp 427-429. P. L. Wolf, D. Williams, T. Tsudaka. and L. Acosta, "Methods and Techniques in Clinical Chemistry," John Wiley and Sons, Inc., New York, N.Y., 1972, pp 186-190. F. H. Stross and G. E. K. Branch, J. Org. Chem., 3, 385 (1938). E. H. White, in "Light and Life," W. D. McElroy and B. Glass, Ed., The Johns Hopkins Press. Baltimore, Md., 1961, pp 183-199. E. H. White, 0. Zafiriou, H. H. Kagi. and J. H. M. Hill, J. Amer. Chem. SOC.,86, 940 (1964). E. H. White and M. M. Bursey, J. Amer. Chem. SOC., 86, 941 (1964). A. U. Khan and M. Kasha, J. Amer. Chem. Sac., 88, 1574 (1966). E. McKeown and W. A. Waters, J. Chem. SOC.,8, 1040 (1966). P. B. Shevlin and H. A . Neufeld, J. Org. Chem., 35, 2178 (1970). H. V. Malmstadt and H. L. Pardue, Anal. Chem., 33, 1040 (1961). A. J. Bard, "Chemical Equilibrium," Harper and Row, New York, N.Y., 1966, pp 88 and 195. D. T. Bostick and D. M. Hercules, Anal. Lett., 7 , 347 (1974).
ACKNOWLEDGMENT
RECEIVEDfor review April 11, 1974. Accepted September
T h e authors thank Vicente Anido who supplied the analyzed human serum samples used in this work. The authors are grateful to D. M. Hercules for supplying a copy of his manuscript prior to publication.
23, 1974. This work was presented in part a t the 25th Annual Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Cleveland, Ohio, March 1974, paper No. 89.
Electrogenerated Chemiluminescence: Determination of the Absolute Luminescence Efficiency in Electrogenerated Chemiluminescence; 9,lO-Diphenylanthracene-Thianthrene and Other Systems Csaba P. Keszthelyi,' Nurhan E. Tokel-Takvoryan,2 and Allen J. Bard3 Department of Chemistry, The University of Texas, Austin, Texas 787 12
The efficiency of electrogenerated chemiluminescence (ECL) (decl)of several systems is reported using both potassium ferrioxalate actinometry and calibrated-photodiode measurements. Experimental methods and necessary corrections in &l-determinations are discussed and ECL in mixed solvent systems ( e.g., acetonitrile (ACN)-benzenevalues were found: toluene) is described. The following deCl 9,lO-diphenylanthracene (DPA) ( 7 . 8 m M ) , thianthrene (TH) ( 1 l . l m M ) (in mixed solvent): 20% peak efficiency, 5 % for several hours; DPA (7.7mM, in mixed solvent) 4 % ; DPA (2.20mM, in mixed solvent) 8 % ; rubrene (in benzonitrile) 1.9 YO.Good agreement between pulsed stationary electrode and rotating ring-disk electrode measurements and between actinometric and photodiode determinations was found.
Electrogenerated chemiluminescence (ECL) involves the production of excited states, and ultimately light, by the
'
Present address, D e p a r t m e n t of Chemistry, L o u s i a n a State University, B a t o n Rouge, L a . 70803. '' Present address, Chemical Studies, Air Correction Div., P.O. B o x 1107. Darien. Conn. 06820. A u t h o r t o w h o m correspondence a n d r e p r i n t requests should be sent. Li
electron transfer reaction between electrogenerated species, most frequently radical ions; a typical reaction sequence is:
A D
A,'
1 .
+ -
e e
D.'
-
-
Am-
(1) (2)
Do' D
f
A*
(3)
The application of ECL to analytical determinations ( 1 - 3 ) and display devices has been considered. An important parameter in these and other applications, as well as in consideration of the fundamental aspects of ECL, is the ECL efficiency, &I, which represents the number of photons emitted per electron transfer reaction. We have previously discussed problems in the definition of efficiency in ECL systems and reviewed previous measurements ( 4 ) .T h e fact that previously reported efficiency values have shown a wide variation even for the same system under similar conditions [see for example values given for the rubrene system which vary from 0.01 to 8.7% (4-9)] testifies to the difficulties in the measurement, the differences in assumptions made in the calculations, and variations in the nature of the environment in which the radical ion electron-transfer reaction occurs. We report here the ECL of the 9,lOdiphenylanthracene (DPA)-thianthrene (TH) system; the high intensity found with this system allowed the first, di-
A N A L Y T I C A L CHEMISTRY, VOL. 47,
NO. 2,
FEBRUARY 1 9 7 5
249