Anal. Chem. 1994,66, 4345-4353
Measurement in Vitro of Human Plasma Glycerol with a Hydrogen Peroxide Detecting Microdialysis Enzyme Electrode Lindy J. Murphy* and Peter T. Galley
Unit of Metabolic Medicine, The Mint Wing, St. Mary's Hospital, Paddington, London, W2 INY, UK
Human plasma glycerol was determined with a microdialysis electrode, containing the enzymes glycerol kinase and glycerol phosphate oxidase held stationarywithin the electrode. A microdialysis electrode is essentially a conventional microdialysisprobe, with a plalinum working electrode inserted into the tip of the dialysis fiber and reference and counter electrodes contained in the upper compartment. The linear range of response to glycerol was directly dependent on the concentration of ATP. At 4 mM ATP, the linear range was 0.5-500 pM. A fast response time of 20 s was obtained. Two types of interferences were observed when plasma glycerol was measured direct oxidation of interferents at the electrode and attenuation of response to glycerol by reaction with hydrogen peroxide and/or poisoning of the platinum electrode. Ascorbate, urate, and acetaminophen were removed from plasma samples by a pretreatment step involving peroxidase and catalase. Any remaining interferent current was reduced by electropolymerizing ophenylenediamine onto the platinum electrode. Adsorption of plasma proteins on the dialysis fiber was minimal and was not reduced by the preadsorption of human serum albumin. Very good correlation was obtained between the electrode and the standard spectrophotometric technique for the variation in glycerol concentration with time. The determination of plasma glycerol is clinically desirable in the study of lipolysis in diabetics. Lipolysis is the process by which triglycerides in fatty tissue are broken down into free fatty acids and glycerol. Plasma glycerol concentration is therefore a measure of the rate of lipolysis. In healthy subjects, lipolysis is much more sensitive to low doses of insulin than glucose metabolism. The loss of the sensitivity of lipolysis to insulin results in increased levels of free fatty acids, which are oxidized in preference to g1ucose.l This could be an early indicator of type I1 diabetes (non-insuhdependent diabetes mellitus). A technique called a glycerol clamp, analogous to the euglycaemic clamp2but using much lower insulin levels, has been proposed to measure the insulin resistance of fatty tissue. This technique requires accurate measurement (flpM) of plasma glycerol every 10 min. At present, glycerol is measured routinely in clinical laboratories by enzymatic spectrophotometric techniques, using deproteinized (1) Randle, P. J.; Garland, P. B.; Hales, C. N.; Newsholme, E. A. Lancet 1963, 785-789. (2) Defronzo, R A; Tobin, J. D.; Andres, R Am. J. Physiol. 1979,237, E214E223. 0003-2700/94/0366-4345$04.50/0 0 1994 American Chemical Society
samples of whole blood? The measurement time using these techniques is too slow for the performance of the glycerol clamp. This paper reports the accurate determination of plasma glycerol using a microdialysis electrode, enabling the glycerol clamp to be performed. The microdialysis electrode is a novel configuration of the frequently used hydrogen peroxide detecting enzyme electrode. The measurement of glycerol*and glutamate5.'j with the microdialysis electrode has been reported previously. A microdialysis electrode is shown in Figure 1, and a schematic diagram is given in Figure 2. Enzyme solution can be infused into the dialysis electrode by means of inlet tubing at the top of the electrode which extends into the tip of the dialysis fiber. The lifetime of the electrode can be extended by infusing in fresh enzyme solution when required. The electrode has a fast response time of 20 s. The glycerol microdialysis electrode contains the enzymes glycerol kinase and L-glycerol-%phosphate oxidase within the dialysis fiber, next to the platinum electrode. The enzymes catalyze the reactions
+ ATP glycerol kinase glycerol phosphate + ADP glycerol phosphate glycerol phosphate + 0, oxidase glycerol
glycerone phosphate
+ H,O,
ATP, the cofactor for glycerol kinase, must be added to the bulk solution. The hydrogen peroxide produced by the second reaction is oxidized at the platinum electrode. The nonspecific oxidation of interferents, in particular ascorbate, urate, and acetaminophen, is a problem common to all hydrogen peroxide detecting enzyme electrodes for measurements in plasma? Electropolymerized coats of phenylenediamine have been shown to signiscantly decrease interference currents.8-lO ~~
~~
(3) Harrison,J.; Hodson, A W.; Skillen, A W.; Stappenbeck R; Agius, L;Alberti, K. G. M. M. J. Clin. Chem. Clin. Biochem. 1988,26,141-146. (4) Albery, W. J.; Galley, P. T.; Murphy, L. J.J. Electroanal. Chem. 1993,344, 161-166. (5) Albery, W. J.; Boutelle, M. G.; Galley, P. T. J. Chem. Soc., Chem. Commun. 1992,12,900-901. (6) Galley, P. T.; Errington, M. L.; Bliss,T. V. P. Brain Res. Assoc. 1993,IO, 34.4P. (7) Palleschi, G.; Rahni, M.A N.; Lubrano, G. J.; Ngwainbi, J. N.; Guilbault, G. G. Anal. Biochem. 1986,159, 114-121. (8) Malitesta, C.; Palmisano, F.; Torsi, L.; Zambonin, P. G. Anal. Chem. 1990, 62,2735-2740. (9) Sasso, S.V.; Pierce, R J.; Walla, R; Yacynych, A. M. Anal. Chem. 1990, 62,1111-1117. (10) Moussy, F.;Harrison, D. J.; O'Brien, D. W.; Rajotte, R V.Anal. Chem. 1993, 65, 2072-2077.
Analytical Chemistry, Vol. 66, No. 23, December 1, 1994 4345
Figure 1. Microdialysis electrode.
,’
inlet
1
polythene tubing ( 0.96 mm O.D.-0.58 mm I.D. ) vitreous silica tubing ( 140 pm O.D. 40 pm I.D. )
b-
‘v
outlet
4
I I -I
‘VN
working electrode counter electrode reference electrode
5 cm
1 t
electrode to reduce any interferent current remaining after the pretreatment step. The adsorption of albumin onto the outside of the dialysis fiber is also reported. It is hoped that by blocking the adsorption sites on the outside of the dialysis fiber with albumin, adsorption of plasma proteins will be reduced. Only a small number of enzyme electrodes have been reported for the determination of glycerol. Merchie et all4coimmobilized glycerol kinase and glycerol phosphate oxidase on nylon membrane which was held next to a platinum electrode detecting H202. However, they employed a laborious differential measurement for determining serum glycerol, involving measurement of serum response in the presence and absence of the enzymatic membrane. Glycerol dehydrogenase has been used in both potentiometric15 and amperometric16determinations of glycerol using ferricyanide and 1,l’-dimethylferrocene as mediators, respectively, although the determination of plasma glycerol by either of these methods was not reported. EXPERIMENTAL SECTION
Unfortunately, mixing of hydrogen peroxide and interferents in the internal solution of the dialysis fiber can still take place. Ascorbate has been observed to cause attenuation of the hydrogen peroxide response, either by direct reaction with hydrogen peroxide,” as frequently observed in spectrophotometric enzymatic determination,12or by fouling the electrode surface, thereby blocking it to hydrogen peroxide 0xidati0n.I~A pretreatment step for plasma samples is reported here for the elimination of major interferents before measuring the glycerol concentration. The pretreatment step was necessary because this second type of interference caused significanterror in the measurement of plasma glycerol at the low glycerol concentrationsfound in human plasma. In addition, o-phenylenediaminewas electropolymerized onto the
Chemicals and Solutions. Glycerol kinase from CeZZuZomonas sp. (lyophilized, 45 units/mg), horseradish peroxidase type I1 (185 unitdmg), catalase from bovine liver (1600 units/mg), human serum albumin (fraction V), ATP (magnesium salt from equine muscle), hydrogen peroxide (30% solution), glycerol, ascorbic acid, acetaminophen, uric acid, and L-cysteine were obtained from Sigma Chemical Co. L-Glycerol-3-phosphate oxidase from microorganisms (lyophilized, 52 units/mg) was obtained from Boehringer Mannheim. &Phenylenediamine (99+%) was obtained from Aldrich Chemical Co. All buffer reagents were AnalaR grade. All solutions were made up with deionized water obtained from an Elgastat Option 4 water purifier. All experiments were carried out in pH 7.4 phosphate buffer containing K2HP04 (0.1 M) and KCl (0.15 M). The buffer was balanced with concentrated HCl. All solutionswere prepared with this buffer. Enzyme solutions were prepared by dissolving known weights of each enzyme in the same buffer solution (30 pL). Enzyme solution was 500 units/mL in each enzyme unless otherwise stated. Stock solutions used were 1mM glycerol, 100 mM ATP, 10 mM ascorbate, 10 mM acetaminophen, 10 mM L-cysteine, 3 mM urate, and 100 mg/mL for both peroxidase and catalase. Hydrogen peroxide solution was 100-folddilute (88.2 mM) for the pretreatment of plasma and 1000-fold dilute (8.82 mM) for all other experiments. Apparatus and Procedures. Microdialysis electrodes were purchased from Sycopel International Limited, Boldon, Tyne and Wear, UK. The electrodes were constructed to our design and specifications. Cuprophan dialysis fibers, MWCO 5-10 000, were obtained from a Gambro dialysis cartridge (Gambro Ltd). The electrodes were supplied dry. Immediately prior to use, buffer was infused into the electrode at a flow rate of 0.5 pL/min by use of a Carnegie Medicin syringe drive. When the electrode had been filled, enzyme solution was then infused into the electrode using the same flow rate. The flow was stopped once the electrode was full of enzyme solution. Note that the internal electrode solution was held stationary throughout the experiment. The dialysis fiber of the filled electrode was then completely immersed in buffer solution (4mL) held in a small beaker. The
(11) Lowry, J. P.; O’Neill, R D. Anal. Chem. 1992, 64, 456-459. (12) Nagele, U.; Ziegenhom, J.; Nose, S. In Methods ofEnzymatic Analysis,3rd ed.; Academic Press: New York, 1974; Vol. 7, pp 140-146. (13) Palmisano, P. G.; Zambonin, P. G. Anal. Chem. 1993, 65, 2690-2692.
(14) Merchie, B.; Girard, A; Maisterrena, B.; Michalon, P.; Couturier, R Anal. Chim. Acta 1992,263,85-91. (15) Chen, A. IC; Starzmann, J. A Biotechnol. Bioeng. 1 9 8 2 , 2 4 , 971-975. (16) Davis, G.; Hill, H. A 0. Eur. Pat. Appl. EP 125867, 1984.
1 cm
1
T
exposed silver wire ( 75 pm diameter ) upper glass compartment ( 1 mm O.D. 0.58 mm I.D.) exposed silver wire
-
dialysis fibre ( 208 pm O.D. 200 pm I.D. ) exposed platinum wire ( 50 pm diameter
1-
I I
I I
I. 1:
I I I
1:
I I
m
r t
1 cm
1
shaded area represents epoxy resin
Figure 2. Schematic diagram of the microdialysis electrode (not drawn to scale). The electrodes are formed from lengths of Tefloncoated wire, with 4 mm of Teflon stripped from either end to expose the metal. One end is soldered to a gold contact, while the other end is used as the electrode. The contacts are held at the top of the electrode in a block formed from epoxy resin set in a rectangular mould.
~~~~~~~~~
4346 Analytical Chemistry, Vol. 66, No. 23, December 1, 1994
platinum working electrode was held at +650 mV with respect to the Ag/AgCl reference electrode. The external solution was stirred with a magnetic flea. When not in use, electrodes were stored with the dialysis fiber immersed in buffer at 4 "C. Electrode potentials were held and currents measured by use of modules constructed in-house by Mr. John Hooper. Currenttime transients were measured using a Gould series 60000 XYT chart recorder. Currents are given, rather than current densities, since the precise area of the platinum electrode was unknown. However, since the diameter of the platinum wire was 50 pm, and the length of exposed platinum was approximately 4 mm, the area of the electrode is estimated to be 6 x cm2. Calibration. Stock ATP solution was added to the external buffer to give a concentration of 1mM, unless otherwise stated. Additions of stock glycerol solution were then made to the external buffer. The external solution was replaced by exchanging the beaker containing the buffer. The electrode was polarized while the beaker was exchanged and could be kept out of external solution for some time, the limiting factor being the drying out of the dialysis fiber, which is irreversibly altered on dehydration. Electropolymerization. If a poly (&phenylenediamine) coat was required, the electropolymerized coat was grown directly onto the platinum electrode in situ withii the dialysis fiber, following the method of Malitesta et aL8 Polymer coats were always grown prior to infusing enzyme solution into the electrode, since immobilization of enzyme was not desired. The dialysis fiber of the electrode was immersed in o-phenylenediamine solution (5 mM in buffer) and nitrogen was bubbled through the solution for 15 min. Electropolymerization took place at +650 mV vs Ag/ AgCl for 15 min, while nitrogen was bubbled through the buffer. The electrode was placed in fresh external buffer after electropolymerization, and enzyme solution then infused into the electrode. Adsorption of Albumin. If the albumin coat was required, human serum albumin was adsorbed onto the external wall of the dialysis fiber by dipping the dialysis fiber into a 4% solution of albumin in buffer for 30 min. The electrode was then placed in buffer as usual. If both electropolymerization and adsorption of albumin were required, electropolymerization was always performed first, followed by adsorption. pH Dependence. The electrode was prepared in the usual manner with the electropolymerized poly(o-phenylenediamine) coat and the adsorbed albumin coat. The electrode was initially calibrated in pH 7.4 buffer. The pH dependence was then determined for the response to 10pM glycerol. The buffers used were phosphate (PH 6.0-8.0), tris @H 7.5-10.0), and succinate (PH 6.0-4.0) in that order. All buffers were 0.1 M in the buffer species and 0.15 M in KCl and balanced with either concentrated HCl or KOH where appropriate. Identification of Interferentsin Plasma. The electrodewas prepared as usual, without enzyme solution, the poly(o-phenylenediamine) coat, or adsorbed albumin. The background current was allowed to settle, and approximately 5 mg of solid of a species naturally present in plasma was added to the bulk buffer solution. The response to the species was measured when all the solid had dissolved. The buffer was replaced and the process repeated for all other species. Responses to 10 pM ascorbate, 300 pM urate, and 10 pM acetaminophen were determined by addition of a known volume of stock solution. LongTerm Stability of Poly(o-phenylenediamine). An electrode, filled with buffer, was calibrated for response to
hydrogen peroxide with 8.82 pM aliquots of 8.82 mM hydrogen peroxide stock solution. The responses to 100pM ascorbate, 100 pM acetaminophen, 100pM cysteine, and 300 pM urate were also determined by addition of a single aliquot of stock solution to fresh external buffer solution. The electrode was then placed in fresh external buffer for 10 min, prior to electropolymerization of o-phenylenediamine in the usual manner. The electrodewas then placed in fresh buffer and the background current allowed to settle. The response of the electrode to hydrogen peroxide and interferents was determined as before, immediately after electropolymerization and on subsequent days. The electrode was stored in buffer at 4 "C when not in use. Effect of Interferents on Hydrogen Peroxide Response. The electrode was prepared as usual with the poly(o-phenylenediamine) coat but without the adsorbed albumin coat. The remainder of this experiment was performed with buffer containing 10 mM Na2EDTA. The response to hydrogen peroxide was determined by addition of aliquots of stock solution (8.82 mM) to the bulk solution in steps of 8.82 pM. This calibration plot was repeated five times before the electrode was placed in buffer containing 300 pM urate. The background current was allowed to settle, and the electrode was calibrated for response to hydrogen peroxide in the presence of urate. The electrode was then placed in fresh buffer and the calibration repeated for response in the absence of urate. This process was repeated for buffer containing 100pM creatinine, 100pM acetaminophen,and 100 pM ascorbate. Effect of Exposure to Plasma on Response. The working electrode was not coated with poly@-phenylenediamine). The electrode was calibrated at least five times for response to glycerol. The electrode was then placed in 10-fold-dilutedplasma for 2 h, followed by several further calibrations in fresh buffer. This same procedure was followed for the effect of plasma on an electrode prepared with the adsorbed albumin coat. Measurement of Plasma Glycerol. The working electrode was coated with poly(o-phenylenediamine), and albumin was adsorbed onto the outside of the dialysis fiber. The electrode was prepared the day before the clamp was performed and stored overnight at 4 "C. On the day of the clamp experiment, the electrode was switched on and allowed to settle for 1-2 h before commencing measuring samples. Glycerol clamps were performed on healthy volunteers. Plasma samples were taken and measured every 10 min for 3-4 h. EDTA was used as anticoagulant on fresh whole blood samples (4 mL). The treated blood (1.5 mL) was spun in a microcentrifuge for 20 s. Plasma (0.4 mL) was extracted, and peroxidase solution (10 pL) and hydrogen peroxide solution (35 pL) were added. The sample was mixed with a whirly mixer for 5 s and left to stand for 1 min. Catalase solution (10 pL) was added and the sample mixed as before. The glycerol concentration of the treated plasma sample was then determined as follows: (1) 0.3 mL of treated sample was added to the external buffer solution (4.0 mL); (2) 10 p L of ATP solution (0.25 mM) was added; (3) 10 pL of glycerol solution was added (approximately 5 pM) . The current was allowed to steady after each of the above stages. The electrode was placed in buffer containing 20 pM glycerol after each measurement for 2-3 min. This was necessary to remove any ATP retained inside the dialysis fiber. The electrode was then placed in fresh buffer (4 mL) and left for at Analytical Chemistry, Vol. 66,No. 23,December 1, 1994
4347
100
80
f 60
40
i
20
0
0
200
400 600 [glycerol] / UM
800
Figure 3. Response to glycerol at different ATP concentrations (bare working electrode and albumin not adsorbed). The response to aliquots of glycerol was determined in buffer containing a known concentration of ATP. The calibrations were performed in order of increasing ATP concentrations.
least 1 min before adding the next treated plasma sample. A plasma sample was measured every 10 min. Plasma samples, taken at the same time as those for the electrode, were deproteinized with perchloric acid and were tested with an automatic analyzer using the standard enzymatic spectrophotometric technique? This uses glycerol kinase and glycerol phosphate dehydrogenase and measures the production of NADH at 340 nm. RESULTS AND DISCUSSION
Optimization of the Electrode Response. Initially the electrode response was investigated for variation with ATP and enzyme concentrations and with pH in order to optimize the electrode and determine its suitability for measurement of plasma glycerol during the glycerol clamp. In a previous paper: the electrode response to glycerol was analyzed according to the membrane electrode theory developed by Albery and Bartlett.17 This indicated the unsaturated response of the electrode was determined by the rate of diffusion of glycerol through the dialysis fiber. Dependence on ATP Concentration. The dependence of the electrode response on ATP concentration is shown in Figure 3. At 4 mM ATP and below, the linear range of the unsaturated region of response is dependent on the concentration of ATP, the cofactor for glycerol kinase. This suggests the saturated response is determined by the glycerol kinase kinetics. Glycerol kinase can convert glycerol more efficiently at higher ATP concentrations. Above 4 mM ATP, no further increase in linear range is observed, suggesting that glycerol kinase is now saturated with respect to ATP. The saturated electrode response is now determined by either the saturated enzyme kinetics (either glycerol phosphate oxidase or glycerol kinase) or by the electrode kinetics. An increase in the glycerol kinase or glycerol phosphate oxidase concentrationdid not result in any increase in saturated response (not shown), indicating that the saturated electrode response at (17) Albery, W. J.; Bartlett, P. N. J. Electyoanal. Chem. 1985,194, 211-222.
4348 Analytical Chemistry, Vol. 66, No. 23, December 1, 1994
-0-
wccinate
phopphare 111,
7 9 11 PH Figure 4. pH dependence of the response to 10 pM glycerol (unsaturated response). The electrode was coated with poly(@ phenylenediamine), and albumin was adsorbed onto the outside of the dialysis fiber.
3
5
high ATP concentration is determined by the electrode kinetics. Note that, with 4 mM ATP, the electrode response is linear up to 500 pM glycerol, which is more than adequate for the measurement of plasma glycerol (27-137 pM in normal subjects1* and raised levels in diabeticslg). Dependence on Enzyme Concentration. The dependence of the response on enzyme concentration was determined, with glycerol kinase and glycerol phosphate oxidase in the ratio 1:l by activity. No response could be detected below 10 units/mL, and above 100 units/mL, the response became approximately independent of enzyme concentration. This supports the conclusion that the unsaturated response is controlled by the rate of diffusion of glycerol through the dialysis fiber. A concentration of 500 units/mL was adopted as the standard concentration for experiments. pH Dependence. The pH dependence of the electrode response to 10 pM glycerol (unsaturated response) is shown in Figure 4. The response to glycerol was approximately independent of pH above pH 6.0 for the phosphate buffer. However, although the response was approximately constant from pH 7.5 to 10 in Tris buffer, there was a fall in response of 40% compared to phosphate buffer. The response time remained fast, suggesting that the response was still controlled by diffusion through the dialysis fiber. This is expected since the permeability of the dialysis fiber is not thought to vary with pH. Below pH 6.0, the response decreases since the enzymes are no longer active. Glycerol kinase is stable over the pH range 6.0-9.8, exhibiting maximum activity at pH 9.8,2O while glycerol phosphate oxidase exhibits broad optimum activity over the pH range 6.5-9.0.21 The odd drop in response observed with tris buffer was also observed with borate buffer @H 8.0-10.0), although pyrophos(18)Foster, K J.; Alberti, K G. M. M.; Hinks, L.; Lloyd, B.; Postle, A; Smythe, P.; Tumell, D. C.; Walton, R Clin. Chem. 1978,24, 1568-1572. (19) Skowronski, R; Hollenbeck, B. B.; Varasteh, Y.-D. I. Chen; Reaven, G. M. Diabetes Med. 1991,8,330-333. (20) Hiyashi, S.4.; Lin, E. C. C. J. Biol. Chem. 1967,242, 1030-1035. (21) Esders, T.W.; Michrina, C. A J. Bid. Chem. 1979,254, 2710-2715.
phate buffer (PH 8.0-9.5) gave a slightly higher response compared to phosphate. These effects were observed whether or not the electrode was coated with poly(o-phenylenediamine) and in response to aliquots of hydrogen peroxide rather than glycerol, suggesting that the drop in response is due to some effect of buffer species on hydrogen peroxide oxidation. The pH range of the electrode is more than satisfactory for the measurement of glycerol in plasma, since the pH of plasma varies very little and plasma samples are diluted 13-fold during the glycerol clamp. Effect of Buffer Purity on Response. The response of the electrode during repeated calibrations to glycerol was observed to fall continuously by up to 50%over several hours, although the speed of the response was unchanged and remained fast, indicating that control of the response by diffusion of glycerol through the dialysis fiber was still occumng. However, if a new electrode was calibrated repeatedly using buffer made from AristaR grade reagents, the response remained stable. The concentration of calcium in AristaR buffer is much less than in AnalaR buffer (0.2 vs 25 pM). Calcium and other divalent and polyvalent cations form insoluble phosphate salts, and it is thought that these salts precipitate within the dialysis fiber, blocking the pores to some extent and causing the diffusion of glycerol through the dialysis fiber to be greatly reduced. Calibrations carried out in AnalaR buffer with 10 mM NazEDTA also gave a stable response, supporting the idea that insoluble phosphate salts block the dialysis fiber. Identification of Interferent Species in Plasma. An advantage of using the ATP-dependent glycerol kinase is that the electrode response to a plasma sample can be measured in the absence and presence of ATP, giving the interferent and glycerol response, respectively. However, if the interferent current could be eliminated, plasma glycerol can be measured in one step. The large, nonspecific oxidation current due to interferents in plasma is particularly detrimental to the measurement of glycerol, since glycerol concentrations in plasma are in the micromolar range. The interferent current can be 1 order of magnitude larger than the response due to glycerol (approximately 80-180 nA vs 6-25 nA). Clearly the nonspecific oxidation current must be reduced if plasma glycerol is to be measured accurately. Ascorbate, urate, and acetaminophen are known to be oxidized at hydrogen peroxide detecting platinum electrode^.^ A number of other species present in plasma22were also tested for electrooxidation at the electrode. Species that gave rise to oxidation currents are given in Table 1, along with their normal concentrations in human plasma. As can be seen, a large number of species can be oxidized. However, if the magnitude of the oxidation current is considered in relation to the concentration in plasma for each individual species, then only ascorbate and urate are of major importance, contributing 50-80% of the oxidation current. Acetaminophen is of equal importance to ascorbate and urate if present in plasma. L-Cysteine is of secondary importance compared to ascorbate and urate, contributing 20-40% of the oxidation current (this assumes no cystine present-the relative importance of L-cysteine decreases on storage at room temperature due to dimerization to cystine). The remainder (up to 10%)is made up of the other interferents. Determination of the individual inter(22) Blood and Other Body Fluids, 3rd ed.; Alhan, P. L., Ditbner, D. s.,Eds.; Federation of American Societies for Experimental Biology: Washington, DC. 1971.
Table 1. Species Present in Plasma That Produce a Measurable Oxidation Current at a Platinum Electrode Held at +650 mV vs AgIAgCP
speciesz2 L-alanine
L-arginine L-asparagine L-citrulline L-cysteine L-glutamine glycine L-histidine L-lysine L-methionine L-ornithine L-serine taurine L-tryptophm L-tyrosine imidazole urinary indican uric acid ascorbic acid guanidoacetic acid sialic acid riboflavin thiamine acetaminophen
current per unit concn (nA/mM)
concn in plasmaz2(uM)
0.03 0.13 0.065 0.02 36.0 0.05 0.02 0.04 0.02 1.0 0.23 0.02 0.13 9.7 0.77 0.05 48.4 235 420 0.02 0.04 0.24 0.14
338-419 58-92 41-49 29 89-107' 570 179-230 46-85 137-165 22-29 47-61 96-119 33-66 55 45-80 0-440 9.4-34 160-390 5.7-119 20-24 1.29-2.10 69-98 30-445
280
20030
The working electrode was bare and albumin was not adsorbed onto the dialysis fiber. Assuming no cystine present Note that cystine is not electroactive. (I
ferents is particularly important in the measurement of glycerol, since even if the oxidation currents due to ascorbate, urate, and acetaminophen are totally suppressed, the magnitude of the remaining oxidation current can still be equal in size to the response due to glycerol. Reduction of InterferentResponse with Poly(ephenylenediamine). &Phenylenediaminewas electropolymerized onto the platinum electrode in order to reduce the interference current. It is pleasing that the platinum electrode of the microdialysis electrode can be coated in situ simply by placing the dialysis fiber part of the electrode into a solution of &phenylenediamine. This reduces the possibility of damaging the coat during construction of the microdialysis electrode. Table 2 shows the poly(& phenylenediamine) coat was found to be effective at reducing the oxidation currents of the major interferents. However, Figure 5 shows that the long-term stability of the coat was poor, particularly toward acetaminophen. The response to acetaminophen greatly increased over the period of 3 days, while the response to hydrogen peroxide also increased. This is consistent with deterioration of the coating, with a concomitant increase in the available electrode surface. This deterioration, in particular for acetaminophen, has been observed by other workers.1° The results suggest the poly(o-phenylenediamine) coat is not sdficient to eliminate the interferent current. Attenuation of Response by Interferents. The attenuation of the electrode response by interferents was also investigated. Ascorbate, urate, and creatinine are all known to react with hydrogen peroxide, causing interference in enzymatic spectrophotometric assays.l2 The effect of interferents on the gradients obtained from consecutive calibration runs for hydrogen peroxide is shown in Figure 6. This experiment was performed in buffer containing 10 mM Na2EDTA, since this stabilized the electrode Analytical Chemistry, Vol. 66,No. 23,December 7, 1994
4349
Table 2. Efficiency of Poly(opheny1enediamine)Coat in Preventing Oxidation of Species Present in Plasma.
1.5
calcd interference concn of current? with species % poly(o-phenylenediamine) tested* (mM) reduction coat (nA)
species L-cysteine L-methionine L-ornithine L-tryptophan L-tyrosine imidazole urinary indican uric acid ascorbic acid acetaminophen
0.1 6.5 8.0 3.4 4.7 11.1 4.4 0.3 0.1 0.1
72 38 78 82 75 0 96 95 97 91
0.89-l.ld 0.014-0.018 0.0024-0.0031 0.093 0.009-0.015 0-0.022 0.017-0.063 2.0-4.6 0.06-1.3 4-16
Albumin was not adsorbed onto the dialysis fiber. An aliquot of stock solution was made to the external buffer for Lcysteine, ascorbate, urate, and paracetamol. For the other interferents, approximately 5 mg of solid was added to the external buffer. Calculated from (1 - % reduction/100) multiplied by (current per unit concentration x concentration in plasma given in Table 1). Assuming no cystine present. 150
100
50
Figure 5. Deterioration of the poly(@phenylenediamine)coat with time (albumin not adsorbed). Day 1 is the day of coating the electrode. Percentage oxidation current is given by (icoatedaare) x 100% for each species. Species tested were 100 pM ascorbate (solid), 100 p M acetaminophen (diagonal hatched), 300 p M urate (open), and 8.82 pM hydrogen peroxide (horizontal hatched).
response (see above). This allowed any changes in gradient due to the effect of interferents to be determined more accurately. Similar results were obtained for the effect of interferents on glycerol response. Typical physiological concentrations of creatinine, acetaminophen, urate, and ascorbate caused 6%,12%, 13%, and 18%reduction in hydrogen peroxide response, respectively. A plasma sample containing these concentrations of interferents would cause a 49%drop in glycerol response, although dilution of the plasma sample would result in much smaller attenuation of glycerol response. Note that urate and ascorbate, and to a lesser extent acetaminophen, caused some permanent suppression of the hydrogen peroxide response after the electrode was placed in fresh solution, suggesting these species cause poisoning of the electrode surface in addition to the solution reaction with hydrogen peroxide. 4350 Analytical Chemistry, Vol. 66, No. 23, December 1, 1994
.o
1
0.5
0.0
I
2 3 4 5
h 7 8 9 10111213 ruii number Figure 6. Effect of interferents on gradient of calibration runs. The working electrode was coated with poly(@phenylenediamine),and albumin was not adsorbed. The buffer contained Na2EDTA (10 mM). The gradients of a series of consecutive calibration runs are shown. For each calibration run, the electrode response was measured at 8.82, 17.64, 26.46, 35.28, and 44.10pM hydrogen peroxide. Runs 4, 7, 9, and 12 were in the presence of 300 p M urate, 100 p M creatinine, 100 pM acetaminophen, and 100 p M ascorbate, respectively.
The presence of interferents also caused the glycerol response to drift slightly, making the determination of plasma glycerol concentration less accurate and extending the response time. The effect of the interferents during measurement of plasma glycerol could be reduced by diluting the sample and using the poly(0phenylenediamine) coat. However, the need for an accurate measurement of plasma glycerol (accurate to within 1pM) every 10 min, together with the drift in response, and the instability of the poly@-phenylenediamine) coat, suggest that another method, which can overcome the interferent effects more successfully, is preferred. The effect of the interferents on glycerol response would be eliminated if the interferents could be prevented from entering the internal solution of the dialysis fiber. Diffusion of ascorbate and urate is greatly reduced through anion exclusion membranes made from cellulose acetatez3or Nafion.24 However, this is not suitable for the glycerol electrode because ATP, the cofactor for glycerol kinase, is negatively charged and so would be unable to pass through the dialysis fiber. An alternative is to remove the interferents from the plasma sample prior to sampling for glycerol. Sample Pretreatment. Peroxidase can oxidize certain interferents in the presence of hydrogen peroxide:25
+
-D + 2H20 peroxidase
Peroxidase and excess hydrogen peroxide can be added to the (23) Sittampalam, G.; Wilson, G. S . Anal. Chem. 1983,55, 1608-1610. (24) Harrison, D.J.; Turner, R F.B.;Balks, H.P.Anal. Chem. 1988,60,20022007. (25) Aebi, H. E. Methods ofEnzymatic Analysis, 3rd ed.; Academic Press: New York, 1974; Vol. 2, pp 267-268.
h
1
2 “‘41
u 1 minute
+
m
0-
’
a
Figure 7. Eliminating acetaminophen with peroxidase and hydrogen peroxide (bare working electrode and albumin not adsorbed). Addition of (a) 100pM acetaminophen, (b) 1OpL of peroxidase solution, (c-f) 44.1 pM hydrogen peroxide, and (9) pL of catalase solution.
d
I 0
20 40 60 [ glycerol ] / pM ( standard technique )
Figure 8. Correlation between plasma glycerol concentration obtained by the electrode and by the standard spectrophotometric technique (n = 31). The electrode was coated with poly(&phenylenediamine), and albumin was adsorbed onto the outside of the dialysis fiber. The limit of detection for the electrode technique was 0.1 pM glycerol.
Both peroxidase and catalase are relatively cheap enzymes, with fast kinetics and high activity, and so it is feasible to treat each plasma sample in this manner. Such an approach has been reported previously for the measurement of glucose with an iodide selective electrode,27and a similar method is reported by Maidan and Heller,28who immobilized peroxidase directly onto a glucose sensing electrode. The enzymatic elimination of acetaminophen with peroxidase and hydrogen peroxide, followed by catalase to remove the excess hydrogen peroxide, is shown in Figure 7. Ascorbate and urate could also be eliminated by this method, although Lqsteine, L-tryptophan,and L-tyrosinewere unaffected. A pretreatment step for plasma was developed (see Experimental Section), which gave a great reduction in the oxidation current due to interferents. However, a small oxidation current still remained (3-15 d), which was observed to decrease on leaving the plasma sample at room temperature for 1h. It was concluded that Lqsteine was probably the major cause of the remaining oxidation current, since it was not affected by the enzymatic pretreatment. L-Cysteine is known to dimerize to L-cystine, a nonelectroactive product, on standing. The reduction in interference current due to the dimerization of Lqsteine has been observed previo~sly.~ Note that peroxidase and catalase, at the concentrations used in the pretreatment step, produced a 1-2% reduction in glycerol r e sponse.
Adsorption of Albumin. Exposure of the glycerol electrode to 10-fold-dilutedplasma for 2 h was found to result in a 1-4% decrease in glycerol response, presumably due to adsorption of plasma proteins on the outside of the dialysis fiber. It was hoped that preadsorption of albumin onto the membrane surface could prevent any further adsorption taking place. A similar approach using adsorption of polymer has been reported recently.B However, the reduction in response on exposure to plasma was unchanged. It is possible that the reduction in response is due to the poisoning of the electrode surface by interferents, as observed above. Note that preadsorption of albumin caused a 5% decrease in electrode response, compared to the response before adsorption of albumin. Also, the background current of the electrode took longer to steady after being placed in fresh buffer, compared to an electrode with no adsorbed albumin coat. After 5 min settling time in fresh buffer, the electrode behaved as usual. The response to glycerol was observed to have decreased by 15-20% at the end of a glycerol clamp. While the adsorption of plasma proteins is expected to have contributed 1-4% decrease in response, the most likely cause of the remainder of the decrease is the formation of insoluble phosphate salts in the dialysis fiber as mentioned above. It is hoped that the use of AristaR buffer in future experiments will eliminate this drop in response. Measuring Plasm Glycerol Concentrations. The electrode has been used to successfully carry out glycerol clamps. The electrode was prepared the day before the clamp with the poly(&phenylenediamine) coat and the adsorbed albumin coat and stored in buffer overnight at 4 “C. Plasma samples were pretreated with peroxidase and catalase. Results are shown in Figure 8 for the correlation between plasma glycerol concentrations obtained by the electrode and by the standard Cobas automatic analyzer technique. Higher glycerol
(26) See ref 25, pp 273-286. (27) Nagy, G.; Von Stop, L. H.; Guilbault, G. G. Anal. Chim. Acta 1973,66, 443-455. (28) Maidan, R and Heller, A Anal. Chem. 1992,64, 2889-2896.
(29) Oliveira Brett, A M.; Pereira, J. L. C.; Dejardm, P. Bioelectrochem. Bioenetg. 1993,31, 311-322. (30) Moatti-Siat, D.; Velho, G.; Reach, G. Biosens. Bioelectron. 1992,7, 345352.
plasma sample to ensure that all interferents have been removed. The excess hydrogen peroxide can be removed by the addition of catalase:26 catalase
2H20 + 0
2
Analytical Chemistry, Vol. 66, No. 23, December 1, 1994
4351
'"p
I
60 W
50 40 I
1 minute
30 -
d -~
-50
0
50 t
100
150
/ min
20 a
+
IO
t e
-0
Figure 10. Typical measurement of plasma glycerol concentration. The electrode was coated with poly(ephenylenediamine), and albumin was adsorbed onto the outside of the dialysis fiber. The electrode was initially in buffer containing no ATP or glycerol. The following step were made: (a) addition of 0.3 mL of treated plasma, (b) addition of 10 p L of ATP solution, (c) addition of 10 pL of glycerol solution, (d) electrode placed in buffer containing 20 pM glycerol, and (e) electrode placed in fresh buffer containing neither glycerol nor ATP.
0
-10
-20 -50
0
50 t
100
150
/min
Figure 9. Glycerol concentrations during a glycerol clamp. The electrode was coated with poly(pphenylenediamine),and albumin was adsorbed onto the outside of the dialysis fiber. (a, top) shows the correlation in behavior of plasma glycerol concentrations obtained by the electrode (solid dots) and the standard spectrophotometric technique (open dots). (b, bottom) shows the change in glycerol concentration with time (change in glycerol concentration = glycerol concentration at t 10 min - glycerol concentration at t min).
+
concentrationswere obtained with the electrode technique compared to the spectrophotometrictechnique, with an average ratio of the spectrophotometricto electrode glycerol concentration for the data in Figure 8 of 0.85. The two techniques are expected to give different concentrations,since the electrode technique gives the plasma glycerol concentration,while the spectrophotometric technique gives the whole blood glycerol concentration. This differencebetween the whole blood and plasma glycerol concentrations has been reported previously,18 and the ratio of whole blood to plasma glycerol concentration was 0.86, in good agreement with the results above. The concentration of glycerol in whole blood is lower because although the concentration of glycerol in the water of the cells and plasma is the same, the water content of the cells is lower than plasma (73%vs 93%). The results from a typical glycerol clamp are shown in Figure 9a. The variation of glycerol concentrations with time was very similar for the two techniques. In Figure 9b, the change in glycerol concentrations with time is shown. It is very pleasing that this change in glycerol concentration is not significantly different for the electrode and Cobas techniques (Student's t-test, It1 = 0.0055, P = 0.001, n = 17). A number of steps were employed in the determination of the glycerol concentrationof a plasma sample to ensure the accuracy 4352
Analytical Chemistry, Vol. 66, No. 23, December 1, 1994
of the measurement. A typical measurement of a plasma sample is shown in Figure 10. The electrode was initially placed in bufEer held in a beaker, containing no ATP. The pretreated plasma sample was added to the buffer, giving rise to a small oxidation current ('1 nA), probably due primarily to L-cysteine. ATP was then added, giving rise to an oxidation current, the magnitude of which is proportional to the glycerol concentration in the beaker. Finally, an aliquot of glycerol solution was added to give a calibration measurementfor each sample. The fast response time of the electrode made this calibration possible for each sample. This increased the accuracy of the measurement, which is particularly important for the glycerol clamp, since only small changes in plasma glycerol concentration (a few micromolar) are observed. The electrode was then placed in fresh buffer containing 20 pM glycerol for 2-3 min, to consume any ATP retained with the dialysis fiber, followed by fresh buffer containing no ATP and no glycerol, ready for the next plasma sample. It was necessary to remove any ATP retained within the dialysis fiber, since if the ATP was not removed, some glycerol response was observed in addition to the interferent current in step B in Figure 10, giving a misleadingly low value for plasma glycerol on addition of the ATP aliquot in step b. The amount of ATP retained in the dialysis fiber was quite small, and could be liberated on placing in fresh buffer after 5-10 min, although such a long stabilization time is not possible during the glycerol clamp. An alternative would be to have a low background concentration of glycerol, say 10-20 pM,always present in the buffer. CONCLUSIONS
This paper highlights the difficulties encountered when one attempts to accurately measure analyte in plasma with a hydrogen peroxide detecting enzyme electrode. While the nonspecific
oxidation current due to plasma interferents is well-known, the attenuation of the response by the homogeneous reaction between hydrogen peroxide and interferents, or by poisoning the electrode surface, is not considered so frequently. The nonspecific oxidation current and the attenuation of response most probably cancel each other out to some extent, disguising the true magnitude of each. Although coating an electrode can greatly reduce the oxidation of interferents, the interferents are still free to mingle with hydrogen peroxide generated by the enzyme held in the electrolyte layer next to the electrode. Awareness and consideration of
the attenuation of response by interferents should lead to the formulation of better strategies for overcoming the effect of interferents. ACKNOWLEDGMENT We gratefully acknowledge Glaxo for ftnancial support of L.J.M. Received for review April 22, 1994. Accepted August 31, 1994.a e Abstract
published in Advance ACS Abstracts, October 1, 1994.
Analytical Chemistry, Vol. 66, No. 23, December 1, 1994
4353
