AnImmobilized-EnzymeFlow-EnthalpimetricAnalyzer ... · Table5.InVitroEffectsofReducingSubstances, Hemoglobin,andBilirubinonNormalRange GlucoseDetermination Apparentglucoseconcn Glucose

CLIN. CHEM. 22/9, 1427-1433 (1976)

CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976 1427

An Immobilized-Enzyme Flow-Enthalpimetric Analyzer:Application to Glucose Determination by DirectPhosphorylation Catalyzed by Hexokinase

Larry D. Bowers’ and Peter W. Carr2

A novel flow-enthalpimetric analyzer is described and itsuse demonstrated by an analysis in which glucose is de-termined by its hexokinase-catalyzed phosphorylationreaction. The method depends on measurement of thetemperature differential across a column packed withglass-supported immobilized enzyme. Sample volumesof 120 tl can be used to obtain a calibration curve that islinear up to 25 mmol of glucose per liter. A precision(within-day) of 5% is generally observed in the optimumconcentration range where glucose is quantitativelyphosphorylated. Results by the technique correlate rea-sonably with those by the o-toluidine and the hexoki-nase/glucose-6-phosphate dehydrogenase methods: Othersugars-including fructose, glucosamine, and man-nose-will interfere; galactose does not. The techniqueis amenable to both routine and emergency analyses.

AddItional Keyphrases: calorimetry #{149}thermochemistryintermethod comparison

Thermoanalytical methodologies such as thermo-metric enthalpy titrations, direct-injection enthalpi-metry, and continuous-flow enthalpimetry have re-ceived increased attention in recent years (1-4). Onereason for this interest is the unique universality ofenthalpimetric techniques. The quantitation of aspecies by enthalpimetric analysis is based on the ab-sorption or evolution of heat by a reaction in which theanalyte takes part. For an adiabatic system, the ana-lytical utility of this heat can be seen from the rela-tion

Department of Chemistry, University of Georgia, Athens, Ga.30602.

‘Present address: Department of Clinical Pathology, Universityof Oregon Health Sciences Center, Portland, Ore. 97201.

whom requests for reprints should be addressed.Received May 5, 1976; accepted July 6, 1976.

where T is the observed temperature change, Q is theintegral heat, k is the heat capacity of the reaction vesseland its contents, n is the number of moles of productproduced, and H is the heat of reaction. In our expe-rience, there are very few processes that do not involveat least 5 kcal of heat per mole. Even in reactions thathave no appreciable heat, the heat of protonization ofa buffer can be used as a “chemical amplifier.”

Unfortunately, there are several limitations to theapplication of these methods. First, the species of in-terest must be present in sufficient amounts to producea measurable temperature change. Second, the reactionmust be fast enough to avoid extensive heat losses fromthe reaction vessel. A third problem arises as a directresult of the universality of the techniques; i.e., allspecies that can take part in the reaction(s) occurringin the reaction vessel will contribute to the observedtemperature change. This lack of intrinsic selectivityshould be contrasted with other analytical methodolo-gies, such as spectrophotometry or amperometry, inwhich the transducer can selectively detect the analytebased on a physical characteristic such as absorption oflight at a certain wavelength or the measurement ofcurrent near the redox potential. Thus, specificity inthermochemical methods originates in the judiciouschoice of reagents and reaction conditions. The dif-ficulties encountered in satisfying these criteria haveretarded the evolution of thermoanalytical chemis-try.

Clearly the use of selective biochemical reactions such(1) as those involving enzyme catalysis or antigen/antibody

interactions (5) can convey high specificity. In fact, thecombination of a universal detector and a biospecificreaction is complementary. Thus, it is in principlepossible to determine a substrate directly withoutcoupling the primary reaction to auxiliary enzymes andsubstrates to obtain a measurable phenomena, e.g., a

1428 CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976

change in the absorbance caused by NADH. As an ex-ample of this concept, it should be possible to determinethe triglyceride concentration merely by doing a selec-tive hydrolysis and measuring the accompanying heat.In addition, enthalpimetric techniques convey someclinically relevant advantages over commonly usedtransducers. First, enthalpimetric measurements areindependent of the optical properties of the sample,such as turbidity and hemolysis. Thermal sensors arealso not influenced by the presence of even high con-centrations of protein, as certain electrochemical sensorsare. Finally, enthalpimetric methods, particularlycontinuous-flow enthalpimetry, are readily amenableto automation, making them useful for large numbersof samples. As a result of these features, a number ofanalytical methods have appeared in which enzyme-catalyzed reactions are used in conjunction with thermalsensors (6-11).

Enzymatic analyses can be implemented in two dis-

tinct fashions: kinetic or rate measurements or end-point or equilibrium determinations. When a substrateis measured via its rate of change, its concentration mustbe well below the Michaelis constant for the system.Thus, the calibration curve at high substrate concen-tration frequently is nonlinear. In addition, kineticmethods of analysis are subject to significant errorsbecause of small changes in pH, temperature, and ac-tivator and (or) inhibitor concentrations. Thus, even ifa kinetic method is both accurate and precise understandard conditions, there is no a priori reason to believethat it will produce accurate results on a sample that isin a complex matrix. Although equilibrium or end-pointassays are subject to problems from enzyme inhibitors,they are generally much more amenable to changes inthe other factors. The major drawback of end-pointmethods is that they either require more time to ac-complish the analysis with the same amount of enzymeas the kinetic method or will require a larger quantityof enzyme. Immobilized enzymes offer the distinct ad-vantage of recyclability, so that one can afford to exposea sample to several hundred units of enzyme andthereby implement a rapid equilibrium assay.

The analytical aspects of immobilized enzymes havebeen recently reviewed (12, 13). Aside from the afore-mentioned reusability and the increased stability nor-mally observed in an insolubilized enzyme, an addi-tional advantage of immobilized enzymes is the fact thatonce a batch of enzyme is linked to the support it be-comes a semi-permanent part of the analytical system.The system can thus be calibrated for incomplete con-version or the presence of other enzymes that can causenonspecificity. This is not possible for soluble enzymes,because different lots will have different activities andimpurities.

This system uses a flow-enthalpimetric system thatminimizes many of the analytical problems inherent inclassical calorimetry, including sample size limitationsand analysis rate. The basis of the technique is themeasurement of the temperature change generated bythe passage of the analyte through an “adiabatic” col-

Pnsp j_HaIsonsYstent}1_t[aI1hbroIcr H__F__Hte

Ihelmally Isolated Heal SInk

DpIterenliol TempeloluleMeasu,ement System

S

Data AquisiliUn

L 0er

Fig. 1. Block diagram of a flow enthalpimetric analyzerThe system consists of: A reference buffer reservoir; B, peristaltic pump; C,sample injection unit; 0, thermally isolated heat sink; E, sample and effluentequilibration units; F. reaction column contaIning solid-phase catalyst; G, dif-ferential temperature measurement circuit; and H, the data-acquisition sys-tem

umn packed with catalyst. The method is, in principle,compatible with any heterogeneous catalyst. We haveused immobilized enzymes because of their specificityand their usefulness in clinical analyses. The systemchosen to illustrate the technique is the assay of glucosevia its hexokinase-catalyzed phosphorylation. Althoughmany techniques have been used to measure glucose(14-16), most enzymatic methods have required cou-pling reactions, to allow spectrophotometric measure-ments. Several groups have studied the thermochem-istry of the reaction.

hexokinaseGlucose + Mg(ATP)2 -

glucose-6-phosphate2 + Mg(ADP) + H (2)

Because a hydrogen ion is liberated, a slightly alkalinepH is used to assure that reaction is complete. Underappropriate conditions, the reaction enthalpy is -18kcal/mol (7, 8). Based on the results of McGlothlin andJordan (8), it should be possible to use the primaryenzyme reaction directly to determine the glucosewithout the need to couple hexokinase to glucose-6-phosphate dehydrogenase.

Materials and Methods

Apparatus

Figure 1 shows a block diagram of the sub-systemsthat comprise the enzymatic flow enthalpimeter. Adetailed description of the system, including physicaldimensions of the major constituents, has appearedelsewhere (11). The principle of operation is as follows:samples are admitted into a flowing stream of bufferand the required co-factors and the entire flow streampumped into a large thermally insulated heat sink. Thestream is equilibrated with the heat sink during itspassage through a small heat-exchange element, afterwhich the fluid is perfused through the immobilizedenzyme column. The temperature differential across the“adiabatic” column is measured with a pair of therm-istors connected to a differential ac phase-lockWheatstone bridge circuit whose output is proportionalto the temperature difference across the column.

200

2

25 i0C

Fig. 2. Enthalpigram of a standardglucose assayNumbers above the peaks indicate the glucose concentration of the aqueousstandards in mg/100 ml

CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976 1429

Table 1. Summary of the Routine OperatingConditions and Variables

Bath temperature: 21 #{176}CFlow rate: 1.0 mI/mmTotal system volume from injector to detector: 1.1 ml

Effluent equilibration unit: 0.03 in. i.d. x 1/16 in. o.d. x15cm stainless-steel coil with coil i.d. of 1.5cm

Sample equilibration unit: 0.010 in. i.d. x 1/16 in. o.d. x30 cm stainless-steel coil with coil i.d. of 1.5 cm;residence time, 3.6 s

Column unit: total volume, 1 ml; void volume, 0.8 ml; resi-dence time, 48 s

The columns used in the present study were insulatedunits with the inside tube (5 mm i.d., 5 cm long, 0.6 mmwall thickness) surrounded by a glass jacket. The areabetween the walls was evacuated. The thermistors weremounted in commercially available couplers (Model200-22; Altex Scientific, Inc., Berkeley, Calif. 94710) andconnected to the ends of the column. Table 1 summa-rizes the routine operating conditions and variables.

Reagents

Chemicals of reagent or primary-standard grade wereused throughout. Calibration standards were made withSRM glucose (Standard Reference Material No. 917;National Bureau of Standards, Washington, D. C.20234). All other sugars were obtained from SigmaChemical Co., St. Louis, Mo. 63178.

A stock diluent was prepared that contained sodiumchloride and ethylenediaminetetraacetic acid (EDTA)(9 g/liter and 1 mmol/liter, respectively). Stock bufferfor the analysis was prepared by dissolving sufficienttris(hydroxymethyl)aminomethane (THAM; Sigma)in the stock diluent to make the solution 1 mol/liter. Thephosphorylation mixture used in this study was madeby adding 60.0 ml of stock buffer solution to 3.00 g ofmagnesium chloride (15 mmol) and 7.20 g of adenosine5’-triphosphate (ATP, Grade II; Sigma). The pH of theresulting solution was carefully adjusted to 7.40 ± 0.02with concentrated hydrochloric acid. A 50.0-mi aliquotof the phosphorylation mixture was added to 200.0 mlof the stock diluent and used as the reference buffersolution.

Hexokinase (EC 2.7.1.1; Type F-300, Sigma) wasimmobilized on controlled-porosity glass (GlycophaseG#{174},120-200 mesh, 25-nm pore diameter; Pierce

Chemical Co., Rockford, III. 61105) by a method pre-viously described (17). The extent of phosphorylationof glucose in the column was spectrophotometricallychecked by measuring the amount of glucose-6-phos-phate produced by its reaction with nicotinamide ade-nine dinucleotide phosphate (NADP, sodium salt;Sigma) in the presence of glucose-6-phosphate dehy-drogenase (EC 1.1.1.49; Type IV, Sigma).

“Versatol” (lot No. 267004) and “Calibrate” (lot No.2751084) control sera (Fisher Chemical Co., Pittsburgh,Pa. 15219) were used in the precision and accuracystudies. The serum samples used in the correlationstudy were obtained from Athens General Hospital,Athens, Ga., and frozen until used. The serum pool wascollected at the University Health Service and dialyzedagainst a sodium chioride/THAM solution (9 g/liter and0.1 mol/liter, respectively; p1-I 7.4). Glucose kits basedon the o -toluidine condensation and hexokinase/glu-cose-6-phosphate dehydrogenase reactions were ob-tained from Sigma. “Glucostat” kits based on the glu-cose oxidase procedure were also obtained (Worthing-ton Biochemical Co., Freehold, N. J. 07728). Thepreparation of the Somogyi filtrate reagents (bariumhydroxide zinc sulfate) and their use followed the pro-cedure described by Henry et al. (14).

Analytical Procedure

Samples were prepared for analysis by pipetting a200-id aliquot of standard or unknown and a 50-tl ali-quot of the phosphorylation mixture into a 2-ml testtube with an automatic pipettor (Model P-2000, Pi-petman; Rainin Instr. Co., Brighton, Mass. 02135). Aftermixing, the samples were aspirated into either a syringeor the sample loop of a sample-injection valve (ModelSV-8031; Chromatronix, Berkeley, Calif. 94710) forintroduction into the flow stream.

The reference buffer solution was continuouslypumped through the entire system during operation. Astable baseline was obtained within 5 mm of starting thepump (Model 1201; Harvard Apparatus Co., Inc., Miffis,Mass. 02054). Samples were then introduced into theflow stream and the resulting temperature peaks re-corded as shown in Figure 2. The temperature changeat the peak maximum was obtained by comparison toa reference resistance change and the substitution of theresistive change at the peak maximum into the rela-tion

=aRo

(3)

where i.R is the resistance change, a is the temperaturecoefficient of resistance of the thermistor, and H0 is theresistance of the thermistor at a reference tempera-ture.

Results and DiscussionIn all enthalpimetric methods, the thermal equili-

bration of the sample with the measuring system is animportant factor in determining the limit of detectionof the method. In the flow-enthalpimetric analyzer

C

A B A

1430 CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976

1.0

5 mm m#{176}C

Fig. 3. Ability of the system to equilibrate samplesConditions: l5O-iil sample volume; bath temperature. 24 #{176}C.Samples wereinjected at the arrow. Curve A, temperature of injected water, 70 #{176}C;flow rate.1 mI/mm. Curve B, temperature of injected water, 5 #{176}C;flow rate, 1 mI/mm.Curve C, long-term stability at flow rate of 2 mI/mm

described here, such equilibration was accomplished bythe use of a stainless-steel coil. The ability of the systemto bring the flow stream to the bath temperature isshown in Figure 3A and B. In each case, a sample thatdiffered grossly from the temperature (>20 #{176}C)of thebath was introduced into the system and the tempera-ture deviation observed to be less than 200 /L#{176}C.Anumber of preliminary experiments indicated that toachieve the high degree of thermal equilibration shownin Figure 3, one must use tubing with a high thermalconductance and must shape the tubing into a coil inorder to enhance the rate of heat transfer, which islimited by laminar flow regimen evident in straightlengths of tubing. In theory, any desired degree ofthermal equilibration can be obtained merely by in-creasing the length of the tubing. If the inside diameterof the tubing remains the same, the increase in lengthwill result in increased axial dispersion, which decreasesboth the peak concentration and the sample throughputas it does in chromatography.

The temperature resolution of any thermal systemis inextricably related to both the short-term andlong-term changes in the temperature of the sur-roundings. A signal trace of the stability observed undernormal operating conditions is shown in Figure 3C. Theperiodic fluctuations observed are caused by the use ofa peristaltic pump and can be eliminated by using asyringe drive system. If one assumes plug flow and anoverall H of -10 kcal/mol, a peak height about 100times greater than the noise level indicated in Figure 3Ccan be calculated for a ‘OO-Hl sample of 1 mmol/litersubstrate. If one takes into account the dispersion of thesample in the column, the temperature resolution af-forded by the flow system should permit the quantita-tion of a number of systems of biochemical and clinicalinterest, particularly in cases where the concentrationexceeds about 1 mmol/liter.

Electrolyte Effects

The total absence of selectivity in thermal detectorswas of some concern to us, because both the glass surfaceof the support and the bound protein could serve asnonspecific ion-exchange sites. We have observed peaksin the absence of glucose, caused by the injection of highconcentrations of electrolytes, which greatly resemblethose obtained in chromatographic heat-of-adsorptiondetectors (18). To ascertain the magnitude of the effect

12.5

1m0c

- I

2 miii

B a5JmaC ..,Av_

2 mm

Fig.4. Systemresponseto added electrolyte withan immobilizedhexokinase reactorAll samples added at the anvw. Curve A, dilution of the reference buffer solutionbefore injection. From left to right, the samples contained 0%, 25%, 50%,75%, and 100% (by vol) reference buffer solution, with the remainder beingstock diluent. Curve B, sample solution contains 100 mmol Na2HPO4 per liter.Curve C, sample solution contains 100 mmol KCI per liter

of electrolyte differences between the average flow-stream conditions and the sample, we packed the col-umn with Glycophase G, Glycophase G bound bovineserum albumin, or bound hexokinase, and perfusedsamples containing various amounts of electrolytesthrough the column. We noted that the signal responsedepended on both the nature of the glass surface and themethod used to immobilize the protein. The peaksshown in Figure 4B and 4C are representative, althoughit should be pointed out that the electrolyte differenceschosen to illustrate the phenomena are much greaterthan those observed under operating conditions.

We believed that small differences in the electrolytecomposition of the reference buffer solution and thesample preparation might account for the differentialpeaks that were observed in preliminary experiments.As can be seen from Figure 4A, injection of an aliquotof the reference buffer solution diluted with a salinesolution did not cause large thermal effects. For com-parison, a thermal peak from a normal glucose standardis shown in Figure 2. Sample preparation is, therefore,important but not critical and although the doublepeaks may be aesthetically annoying, they do not seri-ously affect the observed peak height. This is becausethe position of the peak maximum on the time axiscoincides with the zero crossing point between the ad-sorption and desorption peaks. If the signal integral istaken as a measure of the substrate concentration, theheat of the reversible reaction will not contribute to thesignal. We did not use integration techniques, becausebaseline shifts after the peaks made quantitation dif-ficult with an analog integrator. It should be possible toimprove the precision of the method through the use ofdigital integration techniques.

Note the contrast in the peak shapes of Figures 2 and4. The gaussian peak shape observed for the conversionof substrate by the surface-bound enzyme is a result ofthe fact that the heat produced by the reaction movesthrough the remainder of the column at the velocity of

Table 3. Analytical Recovery of GlucoseAdded to Serum”

Apparent glucose concn

CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976 1431

the eluent. The same spreading phenomena that affectliquid-chromatographic processes cause the spreadingof the temperature peak into an approximately gaussianform. The influence of factors other than those dealingwith dispersion gives rise to the tailed appearance evi-dent in Figure 2.

Flow Enthalpimetric Determinationof Serum Glucose

In light of the data available on the hexokinase-cat-alyzed phosphorylation of glucose (see above), wethought that an equilibrium method of analysis couldbe developed based on the use of an immobilized en-zyme to cause complete conversion of substrate toproduct in the flow stream (8). The extent of reactionupon passage of small volumes of glucose (100-150 gil)through the column was verified by collecting 5-mlfractions of the column effluent. Aliquots of this effluentwere allowed to react to equilibrium with a solution ofNADP and glucose-6-phosphate dehydrogenase, andthe NADPH quantitated by ultraviolet photometry.Total conversion (‘-‘99%) was observed up to glucoseconcentrations of 25 mmol/liter (4.5 g/liter). This is wellin excess of the Michaelis constant of the immobilizedhexokinase (Km = 0.6 mmol/liter).

After establishing the conditions under which com-plete (equilibrium) phosphorylation of glucose wasobtained, we prepared a calibration curve. It is linearto at least 25 mmol/liter (4.5 g/liter), with a functionalrelationship y = (1.95 ± 0.3)x + (-0.2 ± 0.5). The de-tection limit of the method is 0.1 g/liter, as determinedby the method of Hubaux and Vos (19).

As shown in Table 2, the precision of the thermalmethod is comparable with that for other glucosemethods (14). From the data obtained with a septuminjection system, it appears that the introduction of thesample by this method is not amenable to high preci-sion. There are two factors involved in this problem.First, the ability of the operator to inject the samplereproducibly is somewhat suspect. Second, the pertur-bation of the system caused by the sudden addition ofthe sample solution is observed as a thermal “event.”It is evident that the use of the sample-injection valvesis essential for maximum precision. The rate of sampleaddition with an injection valve can be quite high, be-cause two samples can be loaded into the valve simul-taneously.

The long-term repeatability and stability of themethod was evaluated by running replicates of a samplepool over a 20-day period with several different columnsin the system. The total variation of the peak height (intemperature units) was 12%, owing to small flow-ratechanges.

The functional relationship between the flow rate andthe temperature change at the peak height is rathercomplex, owing to heat and mass dispersion in the fluidas well as heat transfer to the sensor. We have foundthat for the present system, a fairly broad optimumexists at flow rates from 0.8-2.0 ml/min. Day-to-daychanges in peak height with flow rate can be corrected

Table 2. PMean

recision ofSD

Thermal AnalyRange

sis of GI

C

ucosea,b

CV, %mg/titer

88d 63 862-930 21 7.12050d 93 1840-2160 15 4.63100’ 98 2910-3250 11 3.2

876e 30 860-920 10 3.92030e 12 1940-2100 12 0.6

a Samples were “Versatol,” “Versatol A,” and “Versatol A Alter-

nate.”b Conditions: Flow rate, 1 mI/mm; tris(hydroxymethyl)amino-

methane buffer (0.2 moi/liter, pH 7.90).C No. samples run at various times during a single day.dOffolumn septum injector, 150-M’ sample.e Sampling valve, 120-l sample.

Glucose Initial

added concn

mg/liter

250 502000 502000 1220

Thermal, Recov- Hk/G6PD,b Recov-mg/liter ery, % mg/liter ary, %

340 113 310 1042000 98 2090 1023280 102 3230 100

Conditions same as in Table 1.b Hexok inase/glucose-6-phosphate dehydrogenase spectrophoto-

metric method.

Table 4. Analysis of Commercial Control SeraMethod

Labelglucose

Control sera value” Thermal Ab BCSomogyi-

Nelson

Versitol 84.8 88.0 85.0 87.0 86.8VersitolA 202 205 218 298 210Versitol A 298 310 302 306 300

Alternate

a Method of analysis is not specified.b Hexoki nase/glucose-6-phosphate dehydrogenase.C Glucose oxidase/peroxidase.

by the use of standards. Hence, the maximum differencein glucose concentration was only 5% at 2.0 g/liter.

The results of a recovery study are shown in Table 3.A known amount of SRM glucose was dissolved in di-alyzed serum or added to a serum poo1of known glucoseconcentration and the resulting sample was analyzedfor glucose. As shown, the thermal method analyticallyrecovers essentially all of the added glucose.

Results obtained on assay of several commercial seraare shown in Table 4. Commercial glucose kits that in-volve glucose oxidase/peroxidase and hexokinase/glu-cose-6-phosphate dehydrogenase methodology wereused. The Somogyi-Nelson procedure followed a pub-lished method (20). As can be seen from the table, thethermal method agrees rather well with the more es-tablished techniques.

Table 5. In Vitro Effects of Reducing Substances,Hemoglobin, and Bilirubin on Normal Range

Glucose DeterminationApparent glucose concn

GlucoseConcn present Thermal H/Ga

o-ToIu-

idine

mg/100 ml

0Ea’

I-

0’

a)V.0C-,

(0

50

‘10

30

20

ID

L

SugarsFructose 105 106 228 107 91Mannose 117 111 240 110 2212-Deoxyglucose 114 112 234 112 102Glucosamine 104 103 219 106 108Galactose 124 105 94 107 220

Other substancesUrea 286 123 128 124 129Uricacid 100 120 118 120 123Hemoglobin 750 126 130 127 140Bilirubin 30 120 120 120 126

a Hexoki nase/glucose-6-phosphate dehydrogenase.

ID 20 20 ‘10 50

Glucose, g/l (o-Toluidine)

Fig. 5. Glucose as measured by the present technique vs. theo-toluidine procedureSlope, 0.98 ± 0.05; Intercept, 0.11 ± 0.09; correlatIon coeffIcient, 0.97; Stu-dent’s t. 18.8; n = 24

‘10aEa)

I-

0’

a)0

C-)

(0

I0

1432 CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976

We selected a number of potential carbohydrate in-terferents, based on their maximum reaction rate andMichaelis constant, and added them to known amountsof glucose dissolved in a dialyzed serum pool. Notunexpectedlly, a number of these sugars were completelyphosphorylated and detected by the thermal sensor (seeTable 5). This is not a serious problem, because none ofthe interfering sugars is usually present in significantamounts in normal serum (14). Experiments (not shownhere) with fructose indicated that the calibration curveis linear up to concentrations of 4.0 g/liter (22 mmol/liter).

Several other substances that act either as reducingor simple photometric interferents have no statisticallysignificant effect on results by the thermal method. (Itshould be pointed out that the o -toluidine method isalso nonspecific.)

Correlation with Established Methods

Frozen serum samples were analyzed by the o-tolui-dine, hexokinase/glucose-6-phosphate dehydrogenase,and thermal methods. The tests were run as nearly si-multaneously as possible, because of the instability ofglucose. The results of these comparisons are shown inFigures 5 and 6. The correlation is rather good for bothtechniques, with both slopes dose to unity and no sig-nificant intercept. The slightly higher slope obtainedfrom the spectrophotometric hexokinase/glucose-6-phosphate dehydrogenase method comparison may bedue to the relatively nonspecific behavior of hexokinasein the thermal method. We wondered if the proteinmatrix could be a substantial contributor to the im-precision of the thermal method. To determine if theassay could be improved, we deproteinized the samples.Owing to the electrolyte effects discussed above, pre-cipitation with zinc sulfate/sodium hydroxide was un-acceptable. The Somogyi precipitation method makesuse of barium hydroxide and zinc sulfate as a protein

SD

I I

ID PD 20 ‘ID SO

Glucose, g/l (Enzymatic)

Fig. 6. Glucose as measured by the present method vs. thehexokinase/glucose-6-phosphate dehydrogenase methodSlope, 1.08 ± 0.04; intercept, 0.19 ± 0.07; correlatIon coeffIcient, 0.98; Stu-dent’s t, 26.2; n = 25

precipitant, which was ideal for our purposes, becauseit does not leave high concentrations of excess electro-lyte in the supernatant fluid. The results of the studyyielded the relationship y = (1.02 ± O.4)x + (0.02 ±0.06). The correlation coefficient was 0.993, with aStudent’s t-value of 31.9. The improvement in the slopeis probably due to the removal of some interferingsugars by the separation of the protein (21) along withthe removal of some nonspecific effects caused by thebuildup of protein.

The use of thermoanalytical methods in analyticaland clinical measurements has suffered in the past fromcomplexity and nonspecificity. The flow-enthalpimetricanalyzer described here overcomes many of the prob-lems that affect classical thermal techniques. First, theuse of a differential temperature measurement, alongwith complete conversion of substrate to product,eliminates the need for expensive temperature-controlequipment. Second, the only operation required is the

CLINICAL CHEMISTRY, Vol. 22, No. 9, 1976 1433

injection of the sample, which is simple compared to themanipulation of the calorimeter required in more clas-sical techniques. In addition, the system provides rapidequilibration of the sample with the temperature bath,which has been a problem in other thermal systems.

The differential detection system used provides areasonable limit of detection. The nature of thethermistor sensor does, however, place limitations onthe analysis. The sample throughput, for example, islimited to about 40 samples/hour by the column sizerequired for complete conversion and the system’sability to respond to the temperature changes at highflow rates. Thus, although the system has sufficientsensitivity and linearity for many assays, the detectordesign is undergoing continuing improvements. Atpresent, the system compares very favorably to otherrecently introduced flow analyzers (22,23) as well as themore classical techniques (6-9) in terms of precision andanalysis rate.

The coupling of enzyme specificity with the univer-sality of thermal detection offers a unique analyticalapproach to a large number of measurements withoutthe need for enzymatic or chemical coupling. Becausethe technique involves the use of immobilized enzymes,this method has an economic advantage in that theenzyme can be reused. This is particularly useful be-cause the amount of enzyme necessary for complete andrapid conversion of the substrate can be packed into acolumn, thereby avoiding the problems inherent in ki-netic analyses. The application of the flow-enthalpi-metric analyzer to glucose determinations has provedto be a reasonable clinical technique. The correlationwith results by more classical glucose methodologies isgood. Because of the small sample requirements, speed,simplicity, and low cost, the flow-enthalpimetric ana-lyzer described here can be used for a variety of reac-tions which must be rapidly and (or) routinely deter-mined in the clinical laboratory, by making use of asolid-phase catalyst.

References1. Jordan, J., Thermometric enthalpy titrations. In Treatise onAnalytical Chemistry, 8, part I; I. M. Koithoff and P. J. Elving, Eds.Interscience, New York, N. Y., 1968, p 5175.

2. Hansen, L. D., Izatt, R. M., and Christensen, J. J., Applications of

thermometric titrimetry to analytical chemistry. In Modern Titri-metry, 1; J. Jordan, Ed., Marcel Dekker, New York, N. V., 1974.

3. Carr, P. W., Analytical and measurement aspects of thermometrictitrimetry. Crit. Rev. Anal. C/tern. 2, 519 (1972).

4. Brown, H. D., Biochemical Microcalorimetry. Academic Press,New York, N. V., 1969.

5. Jordan, J., and Jespersen, N. D., Thermochemical methods ofanalysis. Proc. Colloq. mt. CNRS 201 (Thermochimie), 59 (1972).

6. Johannson, A., Matrix bound enzymes studied by flow calorimetry.Protides Biol. Fluids, Proc. Colloq. 20, 567 (1973).

7. Goldberg, R. N., Prosen, E. J., Staples, B. R., et al., Heat mea-surements applied to biochemical analysis: Glucose in human serum.Anal. Biochem. 64,68 (1975).8. McGlothlin, C. D., and Jordan, J., Enzymatic enthalpimetry, a newapproach to clinical analysis; glucose determination by hexokinasecatalyzed phosphorylation. Anal. Chem. 47, 786 (1975).

9. McGlothlin, C. D., and Jordan, J., Thermochemical determinationof glucose in serum, plasma, and whole blood without deproteiniza-tion. Clin. Chem. 21, 741 (1975).

10. Mosbach, K., and Danielsson, B., An enzyme thermistor. Biochim.Biophys. Acta 364, 140 (1974).

11. Bowers, L. D., Canning, L. M., Jr., Sayers, C. N., and Carr, P. W.,Rapid-flow enthalpimetric determination of urea in serum, with useof an immobilized urease reactor. Clin. Chem. 22, 1314 (1976).

12. Weetall, H. H., Immobilized enzymes: Analytical applications.Anal. Chem. 46, 602A (1974).

13. Bowers, L. D., and Carr, P. W., Applications of immobilized en-zymes in analytical chemistry. Anal. Chem. 48, 545A (1976).

14. Henry, R. J., Cannon, D. C., and Winkleman, J. W., ClinicalChemistry: Principles and Techniques, 2nd ed., Hoeber, New York,N. V., 1974, p 1271.

15. Cooper, G. R., Methods for determining the amount of glucosein blood. CRC Crit. Rev. Clin. Lab. Sci., August 1973, p 101.

16. Slein, M. W., ConG. T., and Con, C. F., A comparative study ofhexokinase from yeast and animal tissue. J. Biol. Chem. 186, 763(1950).

17. Bowers, L. D., and Carr, P. W., Preparation and characterizationof hexokinase covalently bound to controlled porosity glass.Biotechnol. Bioeng. (in press).

18. Smuts, T. W., Richter, P. W., and Pretorius, V., An investigationof the response of the micro-adsorption detector by computer simu-lation. J. Chromatogr. Sci. 9, 457 (1971).

19. Hubaux, A., and Vos, G., Decision and detection limits for linearcalibration curves. Anal. Chem. 42, 849 (1970).

20. Faulkner, W. R., and King, J. W., Eds., CRC Manual of ClinicalChemistry. Chemical Rubber Co., Akron, Ohio, 1970.

21. Somogyi,. M., Determination of blood sugar. J. Biol. Chem. 160,69 (1945).

22. Schmidt, H. L., Krisam, G., and Grenner, G., Microcalorimetricmethods for substrate determination in flow systems with immobil-ized enzymes. Biochim. Biophys. Acta 429, 283 (1976).

23. Mattiasson, B., Danielsson, B., and Mosbach, K., Enzymethermistor assay of cholesterol, glucose, lactose, and uric acid instandard solutions as well as in biological samples. Anal. Lett. 9, 217(1976).