Anal. Chem. 1997, 69, 3570-3577
A Flow Injection Microdialysis Sampling Chemiluminescence System for in Vivo On-Line Monitoring of Glucose in Intravenous and Subcutaneous Tissue Fluid Microdialysates Qun Fang,† Xiao-Tong Shi,‡ Yu-Qing Sun,† and Zhao-Lun Fang*,§
Shenyang Pharmaceutical University, 110015 Shenyang, China, China Medical University, 110001 Shenyang, China, and Flow Injection Analysis Research Center, Chemistry Department, Box 332, Northeastern University, 110006 Shenyang, China
A novel flow injection on-line microdialysis system for in vivo monitoring of glucose in subcutaneous tissue fluid and blood is described. An implantable loop-type microdialysis probe was used for subcutaneous sampling, and a flow-through microdialyzer was used for intravenous sampling by pumping of the blood from the tested rabbit through the microdialyzer located outside the living system at a flow rate of 10 µL/min. The perfusion rate of the dialysate was 20 µL/min. The glucose in the dialysate was detected on-line with a flow injection chemiluminescence system after passing through an immobilized glucose oxidase reactor. The calibration of the detector system (including reactor) and monitoring of baseline drifts were performed simultaneously to improve the reliability of the monitoring process. The dialysate sample volume was 20 µL, and the sample throughput was 28 h-1. The variation of glucose level in subcutaneous tissue fluid and blood of the rabbits was monitored after the administration of glucose or insulin to demonstrate the favorable resolution and reliability of the system for in vivo on-line monitoring. The monitoring of chemical changes in the environment of living cells poses substantial challenges for modern analytical chemistry. Various continuous and automated approaches have been developed in recent years to achieve such goals efficiently, economically, and reliably. The principles of hitherto two major approaches are shown schematically in Figure 1A,B. Technically, the most straightforward approach is the use of implanted biosensors (mode A in Figure 1). Despite the limitation of biosensors for in vivo monitoring reported by some workers,1 the refinement of biosensor technology has been carried on continuously in recent years. Wilson’s group published a series of reports on the successful use of a miniaturized, enzymatic, amperometricbased glucose sensor2 implanted in the subcutaneous tissue of rats,3 dogs,4 and human volunteers.5 Recently, the subcutaneously †
Shenyang Pharmaceutical University. China Medical University. § Northeastern University. (1) Moscone, D.; Pasini, M.; Mascini, M. Talanta 1992, 39, 1039-1044. (2) Bindra, D. S.; Zhang, Y.; Wilson, G. S.; Sternberg, R.; Thevenot, D. R.; Moatti, D.; Reach, G. Anal. Chem. 1991, 63, 1692-1696. (3) Moatti-Sirat, D.; Capron, F.; Poitout, V.; Reach, G.; Bindra, D. S.; Zhang, Y.; Wilson, G. S.; Thevenot, D. R. Diabetologia 1992, 35, 224-230. ‡
3570 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
implanted sensor was used for the development of a hypodermic alarm system.6 A two-point calibration was used to compnsate for sensitivity variations. However, continuous calibration of the sensor sensitivity (including continuous monitoring of baseline drift) during the monitoring process was not possible. In recent years, microdialysis techniques have been exploited extensively for in vivo analysis and continuous monitoring. Originally, the technique was developed mainly for the in vivo sampling of the extracellular fluid in discrete compartments of living systems, such as for the monitoring of neurotransmitter release in the brain.7,8 Recently, microdialysis has been employed extensively for in vivo analysis at other sites of living systems including blood,9,10 liver,10,11 and subcutaneous tissue.1,12,13 When employing such techniques, instead of direct sensing, a microdialyzer is implanted in the living system to obtain a dialysate sample by perfusion (mode B in Figure 1). Small molecules are selectively sampled into the prefusate through the dialysis membrane. The dialysate is transported either directly to a detection system or to a collection vessel for subsequent analysis. The use of an external detection system in mode B allows better device accessibility, simplifies troubleshooting, and facilitates performance check under the controlled conditions of flow, temperature, etc. In this work, a novel calibration approach is developed by coupling microdialysate sampling to the detection system using a flow injection interface (Figure 1, mode C), which allows alternating introduction of standard and dialysate sample to the detector, offering simultaneous monitoring of the analyte, standard, and baseline signal variations. Thus, on-line compensation of variations in the detection system (including related analytical reactions) may be achieved. However, calibration of (4) Poitout, V.; Moatti-Sirat, D.; Reach, G. Biosens. Bioelectron. 1992, 7, 587592. (5) Poitout, V.; Moatti-Sirat, D.; Reach, G.; Zhang, Y.; Wilson, G. S.; Lemonnier, F.; Klein. J. C. Diabetologia 1993, 36, 658-663. (6) Thome-Duret, V.; Reach, G.; Gangnerau, M. N.; Lemonnier, F.; Klein, J. C.; Zhang, Y.; Hu, Y.; Wilson, G. S. Anal. Chem. 1996, 68, 3822-3826. (7) Blakely, R. D.; Wages, S. A.; Justice, J. B.; Herndon, J. G., Jr.; Neill, D. B. Brain Res. 1984, 308, 1-8. (8) Zetterstrom, T.; Ungerstedt, U. Eur. J. Pharmacol. 1984, 97, 29-36. (9) Scott, D. O.; Steele, K. L.; Lunte, C. E. Pharm. Res. 1991, 8, 389-392. (10) Scott, D. O.; Sorensen, L. R.; Lunte, C. E. J. Chromatogr. 1990, 506, 461469. (11) Scott, D. O.; Lunte, C. E. Pharm. Res. 1993, 10, 335-342. (12) Mascini, M.; Moscone, D.; Bernardi, L. Sens. Actuators B 1994, B6, 1418. (13) Reach, G.; Wilson, G. S. Anal. Chem. 1992, 64, 381A-386A. S0003-2700(97)00324-7 CCC: $14.00
© 1997 American Chemical Society
enzymatic oxidation of glucose to D-gluconic acid and H2O2. To the best of our knowledge, the only report using chemiluminescence detection for in vivo monitoring of glucose with microdialysis sampling was that by Naslund et al.;21 however, an offline batch procedure was used for the measurements with a sampling interval of 15 min. The system is, therefore, not an online monitoring one, and the resolution is low.
Figure 1. Modes of different approaches used for continuous online in vivo monitoring.
analyte recovery in the microdialysis probe implanted in a living system and compensating related variations during the monitoring process are more difficult. If the microdialysis probe is implanted in a blood stream, further difficulties could be encountered. Although variations in blood flow were reported not to seriously affect analyte transfer through dialysis membranes,14 irreproducible partial obstruction of membrane surface of a dialyzer probe by the vein walls during implantation and/or during the monitoring process cannot be completely avoided. Errors originating from such sources cannot be easily detected and compensated. In this work, a system is proposed to avoid some of the limitations associated with implantation of microdialyzers in the living system, particularly in blood vessels, while maintaining the advantages of on-site microdialysis. The system is shown as mode D in Figure 1, characterized by pumping of the blood from the test animal through a microdialyzer located outside the living system. The dialysate is then presented to the detection system through a flow injection interface as in mode C. The reliable in vivo on-line monitoring of glucose in tissue and blood is imperative for many important missions, such as in the development of an alarm system for detecting hypo- or hyperglycemia and, in the future, for the development of a wearable artificial pancreas. Needle-shaped on-site glucose biosensors15 and microdialysis-coupled glucose sensors have been reported in in vivo glucose monitoring.1,12,16-20 Hitherto, microdialysis sampling in glucose determination is performed mainly under mode B.1,12,16-18 A modification of this mode was reported by Moscone et al.,1 who coupled a flow injection system to the glucose sensor to monitor variations in sensor sensitivity, but not for dialysate sample introduction. In this work, attempts were made to improve the reliability of in vivo glucose monitoring by using a mode C system for subcutaneous sampling and a mode D system for intravenous sampling with chemiluminescence detection involving (14) Stenken, J. A.; Topp, E. M.; Southard, M. Z.; Lunte, C. E. Anal. Chem. 1993, 65, 2324-2328. (15) Wilson, G. S.; Zhang, Y.; Reach, G.; Moatti-Sirat, D.; Poitout, V.; Thevenot, D. R.; Lemonnier, F.; Klein, J.-C. Clin. Chem. 1992, 38/9, 1613-1617. (16) Zilkha, E.; Koshy, A.; Obrenovitch, T. P.; Bennetto, H. P.; Symon, L. Anal. Lett. 1994, 27, 453-473. (17) Keck, F. S.; Kerner, W. Sens. Actuators B 1993, B16, 435-438. (18) Mascini, M.; Moscone, D.; Bernardi, L. Sens. Actuators B 1992, B6, 143145. (19) Boutelle, M. G.; Fellows, L. K.; Cook, C. Anal. Chem. 1992, 64, 17901794. (20) Amine, A.; Digua, K.; Xie, B.; Danielsson, B. Anal. Lett. 1995, 28, 22752286.
EXPERIMENTAL SECTION Reagents and Standard Solutions. All chemicals were of analytical reagent grade, and demineralized water was used throughout. The luminol solution (reagent 1, 1.0 mM) was prepared by dissolving 10.6 g of Na2CO3 in 200 mL of water in which 0.177 g of luminol (Merck, Schuchardt, Germany) was dissolved and made up to 1 L with water. The potassium hexacyanoferrate solution (reagent 2, 2.5 mM) was prepared by dissolving 8.23 g of the reagent (Shenyang First Reagent Works, Shenyang, China) in 1 L of water. The Ringer’s solution containing 148 mM NaCl, 4.0 mM KCl, and 2.3 mM CaCl2 used as perfusion medium was prepared by dissolving 8.65 g of NaCl, 0.30 g of KCl, and 0.26 g of CaCl2 in 1 L of water. The phosphate buffer (pH ) 7.4, 0.1 M) was prepared by dissolving 11.0 g of sodium hydrogen phosphate and 2.7 g of sodium dihydrogen phosphate in 1 L of water. Ringer’s solution containing 2 mM phosphate buffer was used as carrier solution. Aqueous glucose standards in the range of 2.814.0 mM were prepared by sequential dilution of a stock solution containing 55 mM glucose with Ringer’s solution. Silanized controlled-porosity glass beads CPG 10 (700 Å, 120-200 mesh, donated by Professor Elo Hansen of Technical University of Denmark) and 1000 IU of glucose oxidase (GOx, Sigma, St. Louis) were used to prepare the immobilized GOx beads (CPG-GOx) according to a known procedure.22 The insulin injection (Haerbin First Biochemical Pharmaceutical Works, Haerbin, China), glucose injection (Shenyang First Pharmaceutical Works) and heparin sodium injection (Shanghai First Biochemical Pharmaceutical Works, Shanghai, China) were used for in vivo experiments. Instrumentation. Microdialysis System. A loop-type microdialysis probe was used for subcutaneous sampling, and a flowthrough microdialyzer was used for intravenous sampling. The dialysis systems were produced using regenerated cellulose hollow dialysis fibers (250-µm i.d., 300-µm o.d., with a molecular weight cutoff of 3000; donated by Professor Z.-Q. Li of China Medical University). The two microdialysis systems are shown in Figure 2. The probe (Figure 2A) was produced by inserting both ends of a 15-cm length of the hollow dialysis fiber into and through a fused silica capillary tubing with protective polymer coating (0.5mm i.d., 0.7-mm o.d., 50 mm long, Yongnian Optical Fiber Co., Yongnian, Hebei, China), leaving a 2-cm length of the fiber loop outside. Abrupt bending of the fiber loop at its turning point was avoided by inserting a thin glass rod within the loop and holding it against the turning point while pulling the fiber into the capillary. The dialysis fiber was fixed in the capillary by sealing with epoxy at the inlet end of the capillary (lower end in Figure 2A). The two ends of the hollow fiber were inserted into Micro-Line tubings (21) Naslund, B.; Arner, P.; Bolinder, J.; Hallander, L.; Lundin, A. Anal. Biochem. 1991, 192, 237-242. (22) Petersson, B. A.; Hansen, E. H.; Ruzicka, J. Anal. Lett. 1986, 19, 649-665.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3571
Figure 2. Schematic diagram of the loop-type microdialysis probe (A) and the flow-through microdialyzer (B). A, perfusion solution; T, connected tubing, Micro-Line tubings, 0.35-mm i.d. and 0.8-mm o.d.; E, epoxy; G, fused silica capillary tubing with protective polymer coating, 0.5-mm i.d. and 0.7-mm o.d.; D, hollow dialysis fiber, 250µm i.d. and 280-µm o.d.; S, sample solution; M, outer Micro-Line tubing, 0.5-mm i.d. and 1.6-mm o.d.
(0.35-mm i.d. and 0.8-mm o.d., Thermoplastics, Sterling, NY), and the connections were sealed with epoxy to the outlet end of the capillary. The flow-through dialyzer (Figure 2B) was constructed by first inserting two stainless-steel needles (0.4-mm i.d., 0.8-mm o.d., 20 mm long, cut from the lower end of a standard hypodermic injection needle) as guiding cannulas through the wall of a MicroLine tubing (0.5-mm i.d., 1.6-mm o.d., 8 cm long, Thermoplastics, Sterling, NY) at 45° and 135° angles to the tubing with a distance of 2 cm between the two inserting sites, and with the slanted apertures of the needles pointing to each other. A hollow dialysis fiber was then easily threaded into the Micro-Line tubing from the end of one needle and out from the end of the other needle. The needles were then pulled out from the Micro-Line tubing, leaving the fiber in the tubing. The two ends of the hollow fiber were connected to two Micro-Line tubings (0.35-mm i.d. and 0.8mm o.d.) for extension to other parts of the system, and the connections between the dialysis fiber and the outer and extension Micro-Line tubings were sealed with epoxy as shown in Figure 2B. Prior to use, the fibers were immersed in perfusion medium at 39 °C for at least 24 h. A MD-1001 microdialysis syringe pump (Bioanalytical Systems Inc., West Lafayette, IN) and a variable-speed peristaltic pump (Model C-4V, Alitea, Ventur, Sweden) were used for delivery of perfusates. The microdialysis manifold is shown schematically in Figure 3. PTFE tubing of 0.25-mm i.d. and 0.8-mm o.d. was used for all connections. Detection System. A home-made luminometer consisting of a Hamamatsu photomultiplier with stabilized high-voltage power supply and amplification system housed in a light-tight dark-box was connected to a strip-chart recorder and used for the chemiluminescence measurements. The flow cell was made by coiling a 50-cm length of Micro-Line tubing (0.5-mm i.d., 1.6-mm o.d.) into a spiral disk shape with a diameter of 3 cm and fixing the disk on the surface of a mirror (6 cm × 4 cm) with pressuresensitive glue. The disk was located directly facing the window 3572 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
Figure 3. Schematic diagram of the flow injection on-line microdialysis system for in vivo monitoring of glucose in subcutaneous tissue fluid and blood of rabbits. P, perfusion solution, Ringer’s solution; P1, variable-speed peristaltic pump; U, loop-type microdialysis probe; F, flow-through microdialyzer; (A), subcutaneous sampling; (B), intravenous sampling; C, carrier; St, glucose standard for calibration of detector; R1, reagent 1, luminol solution; R2, reagent 2, potassium hexacyanoferrate solution; P2, 8-channel peristaltic pump; T, knotted reactor (0.5-mm i.d., 100 cm length); L1, sample loop for dialysate from (A) or (B), 20 µL; L2, sample loop for St, 20 µL; W, waste; E, immobilized glucose oxidase reactor; LD, luminometer; V, valve. (I) First stage of valve operation, 80 s. (II) Second stage of valve operation, 50 s.
of the photomultiplier. A Model LZ-1020 eight-channel peristaltic pump and a Model LZ-1020 eight-channel injector valve (Zhaofa Instruments, Shenyang, China) with a Chemifold block furnished with four three-way connectors (Tecator, Hoeganas) were used to construct the FI system. The flow injection manifold is shown schematically in Figure 3. To produce the enzyme reactor, the CPG-GOx beads were packed into a 3-cm length of 5-mm-i.d. Tygon tubing and kept in place with plastic foam. Connections between the reactor and connecting lines were made using a combination of short Tygon tubings with appropriate bore diameters push-fitted into the reactor column ends. The volume of CPG-GOx packed in the column was approximately 100 µL. Sample loops for perfusate (dialysate) samples and glucose standards were made from PTFE tubings (0.5-mm i.d.) with a volume of 20 µL. A knotted reactor (KR), made by tying interlaced knots in 0.7-mm-i.d., 1.5-mm-o.d., 100-cm long PTFE tubing, was connected downstream of the merging point of R1 and R2. Procedures. In Vitro Experiments. Investigations were conducted at an early stage of this work under in vitro conditions to facilitate optimization of the microdialysis systems. The setup used is shown schematically in Figure 3. In these experiments, aqueous standards were used instead of real samples from test
animals and thermostated at 39 °C, which was the body temperature of the test animal. For the loop probe, intended for subcutaneous sampling, microdialysis was performed by immersing the probe in the glucose standard and perfusing the probe with Ringer’s solution at defined flow rates. For the flow-through microdialyzer, intended for monitoring of blood glucose, sampling was performed by drawing the glucose standard through its outer tubing while perfusing the inner hollow dialysis fiber with Ringer’s solution, all at defined flow rates. Pump rates lower than 10 µL/ min were not feasible using the Alitea peristaltic pump; therefore, a Model MD-1001 microdialysis syringe pump (Bioanalytical Systems Inc.) was used for studies in the range 1-50 µL/min. FI chemiluminescence determination of glucose in the sample perfusates following the dialysis was conducted as follows: In the first stage of operation (Figure 3-I) the perfusate was delivered to the sample loop L1 with a loading time sufficient for filling the 20-µL sample loop at the perfusing flow rate used. Simultaneously, a glucose standard for checking possible drifts in sensitivity, which was stored in sample loop L2 during the previous cycle of operation, was injected into the carrier and passed through the immobilized GOx reactor to produce H2O2. The flow was merged with the premixed reagent solutions of luminol (R1) and hexacyanoferrate (R2) and the chemiluminescence signal recorded. In the second stage (Figure 3-II), the glucose standard was loaded into sample loop L2. Simultaneously, the perfusate in sample loop L1 was injected and glucose determined as for the standard. In Vivo Experiments. Rabbits weighing ∼2.5 kg were used in all in vivo experiments without anesthetization. Movement of the test animal was prevented by fixing it on a wooden chassis. Glucose concentration in the subcutaneous tissue fluid and blood of rabbits was monitored using the loop probe and flow-through microdialysis sampling systems, respectively. For subcutaneous sampling, the loop dialysis probe was inserted through a 3-cm length of stainless-steel hypodermic injection needle (0.8-mm i.d., 1.2-mm o.d.), used as a guide so that the bending point of the fiber loop came flush with the needle’s pointed tip. The probe was then fixed in the guide needle by taping at its blunt end. The sterilized needle was then punctured transcutaneously 2 cm on the back of the test animal, and the needle was pulled out after removal of the tape, leaving approximately 2-cm length of the dialysis probe beneath the skin. The outside part of the probe with the transport lines and the withdrawn stainless-steel guide needle were then taped on the back of the test animal. The loop probe was connected with the pump and valve, and the microdialysis sampling in subcutaneous tissue was performed by perfusing the probe with Ringer’s solution at a rate of 20 µL/min. Then the procedure for FI chemiluminescence measurements was followed as described in the in vitro experiment section. For experiments on the effect of insulin and glucose injection on glucose level in the subcutaneous tissue fluid, the glucose content in the perfusates was monitored by continuous sampling. After achieving a stable glucose signal, an intravenous injection of regular insulin or glucose was given via the edge vein of the test animal’s ear, and the changes in subcutaneous tissue fluid glucose concentration were recorded. For the experiments with insulin, the rabbits were fasted but fed with water 12 h before the experiments, and a 1-g glucose injection was given after the experiments in order to regain the normal level of blood glucose.
For intravenous blood sampling, the flow-through microdialyzer was used. Fifteen minutes before the experiment, the test animal was given an intravenous injection of 5000 IU of heparin sodium via the edge vein of the test rabbit’s ear to avoid clogging by blood coagulation during the experiments. A sterilized stainless-steel needle (0.25-mm i.d., 0.5-mm o.d.) was inserted into the edge vein of the rabbit’s other ear and retreated while pressing on the two sides of the pinprick to avoid bleeding. The pinprick in the vein was enlarged using another sterilized largerbore stainless-steel needle (0.8-mm i.d., 1.2-mm o.d.). After the needle was withdrawn, a sterilized heparin-treated silicone rubber tubing (0.7-mm i.d., 0.9-mm o.d., 20 mm long, donated by Professor J.-M. Chen of Shenyang Pharmaceutical University) was inserted through the enlarged pinprick into the vein, and the point of insert was taped onto the skin of the animal. The tube formed a tight seal at the pinprick, and no bleeding was observed. The outer end of the tube was connected to the flow-through microdialyzer, and the microdialysis sampling was performed by drawing out the blood of the test animal at a pump rate of 10 µL/min, passing through the outer tubing of the microdialyzer while perfusing the inner hollow dialysis fiber with Ringer’s solution at 20 µL/min. Further operations dealing with the perfusate and infusion of glucose and insulin were then conducted as in subcutaneous sampling. Calibration of the microdialysis chemiluminescence system for subcutaneous sampling was made before the on-line monitoring using a series of glucose standards by immersing the loop probe dialyzer in the standards and operating as during the on-line monitoring. For blood sampling, calibration was performed by pumping the standards through the flow-through dialyzer as for the blood sample. Disregarding the intercept at zero concentration, which was negligible, compensation for variation of sensitivity during in vivo monitoring was performed using the equation
Cx ) KIxIs1/Is2
where Cx is the glucose concentration in sample x, K is the slope of the calibration curve expressed in concentration per unit peak height, Ix is the peak height of sample x, and Is1 and Is2 are the peak heights of the sensitivity check standard at the beginning of the monitoring when the calibration curve was constructed and immediately after injecting sample x, respectively. RESULTS AND DISCUSSION Considerations in the Design of the FI On-Line Microdialysis System for in Vivo Monitoring of Glucose. The following considerations were taken in designing a reliable, sensitive, efficient, and relatively robust system: (a) Most microdialyzer designs used for in vivo monitoring involve the use of short dialysis fiber membrane lengths of a few millimeters to facilitate direct insertion of dialysis probes into the living systems. Extremely low perfusion rates are therefore required to achieve acceptable recovery. The microdialyzers used in this work, shown in Figure 2, were designed to incorporate relatively long dialysis fibers of about 20 mm to enhance the phase transfer efficiency. This is important for achieving high resolution of the process being monitored, as well as for coupling of the dialysis system to conventional FI equipment. Difficulties in implantation were avoided either by using a guide needle, as in the case for the loop probe dialyzer, or by locating the dialyzer Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3573
outside the living system, as for the flow-through microdialyzer [see (d)]. The insertion of the loop probe through a guide needle was made easy by producing the loop in a single capillary instead of connecting two separate capillaries as reported by other workers.23 (b) Instead of transporting the dialysate directly to the detector, conventional FI equipment was used to act as a sampling interface. The reliability of the monitoring is improved by continuously monitoring the baseline signal and by recalibrating the reactiondetection system for each measurement by alternatively injecting sample and standard during the entire process. (c) In vivo monitoring of blood is usually performed using implanted microdialyzers. In addition to limitations mentioned in the introduction, this approach also calls for specialized operation in the implantation of the probe in the blood vessel under anesthetized conditions, and a period of recovery is required for the incision in the test animal to heal. Assuming a successful operation, to what extent such operations affect the metabolic conditions of test animals is yet questionable. Special precautions also should be taken to avoid clogging of the dialysis fiber during the entire monitoring process, an event which might not be easy to detect before complete blockage occurs. An approach is therefore adopted in this system design where blood was pumped (via a catheter inserted in the vein) at extremely low flow rates into a microdialyzer located outside the living system. Insertion of a catheter in the edge vein of a rabbit’s ear required little expertise, no anesthetizing is required during the operation, and no recovery period is necessary, i.e., the in vivo monitoring can be initiated immediately after the insertion of the catheter. The rate of blood outflow is only 10 µL/min; thus, the sampling should have little effect on the fluid balance of the test animal, even after a few hours of continuous monitoring. (d) Electrochemical sensors and detectors were used most frequently for in vivo glucose monitoring.1,15-19 However, even after dialysis, samples encountered under in vivo conditions were reported to poison the electrode sensing surface fairly rapidly.1,18 The detector may also suffer interferences from other reductive substances in the dialysate sample. The chemiluminescence detector appears to be less susceptible to such interferences and poisoning. However, to the best of our knowledge, only one instance of the use of a chemiluminescence method for in vivo determination of glucose using an off-line batch approach has been reported.21 The potentials of this detection mode for in vivo glucose determination are, therefore, not fully exploited. In this system design, chemiluminescence detection is coupled to the on-line microdialysis system by the FI interface. FI is currently broadly accepted as the ideal sample introduction mode for chemiluminescence;22 the coupling to microdialysis for on-line monitoring should further demonstrate its versatility. Optimization of the FI Microdialysis Sampling Chemiluminescence System. Parameters used in the FI chemiluminescence system for glucose determination were mainly adapted from those reported by Petersson et al.22 with minor modifications. A 100-cm-long knotted reactor (0.5-mm i.d.) located downstream of the merging point of the luminol and hexacyanoferrate reagent solutions was found to decrease the analyte signal as much as 50% while suppressing the background (baseline signal). This is presumably the result of more efficient mixing of the reagents as (23) Linhares, M. C.; Kissinger, P. T. J. Chromatogr. Biomed. Appl. 1992, 116, 157-163.
3574 Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
Figure 4. Effect of sample flow rate passing through the flowthrough microdialyzer on analyte peak height. Sample solution, glucose standard, 5.5 mM; other conditions are as in Figure 3.
well as an increase in their reaction time, which contributes to the decay of background emission of the mixed reagents before the main reaction with peroxide. The stability of the immobilized GOx reactor, which is particularly important in continuous monitoring, was examined under in vitro conditions by repetitive injections of 5.5 mM glucose standard solution for about 5 h at a frequency of 30 h-1, using both a new reactor and one already used for 40 h. The peak height decreased by 54% in 5.5 h with the new reactor and by 15% in 5 h with the used reactor. Presumably, the initial reduction of sensitivity with the number of samples analyzed might be primarily due to loss of insufficiently immobilized enzyme and, to some degree, due to the progressive tightening of packing material, which partially blocks the active sites on the CPG particles. Such effects are therefore more pronounced with newly packed reactors, with which some conditioning is definitely required when they are used for continuous monitoring. Even with used reactors, the magnitude of sensitivity variation is sufficiently large to call for frequent recalibration of the reactiondetection system during the monitoring process. This is even more important under in vivo conditions, where the small molecules in blood or subcutaneous tissue fluid may pass through the dialyzer membrane and interfere with the enzymatic catalysis. Parameters associated with the microdialysis were optimized using the FI chemiluminescence detection system under in vitro conditions. The effect of the effective length of microdialysis fibers in probes on analyte transfer efficiency is well documented.23 A proportional enhancement in peak signal is expected with an increase in the effective fiber length. However, an effective length of 2 cm was used for both the loop probe and and flow-through microdialyzer, purely for the sake of convenience, since long fibers have a stronger tendency to crease during operation, particularly for the loop probe. The effect of perfusion rate on glucose recovery of the microdialysis fiber used in this study was investigated employing the system using the loop-type microdialysis probe with a 2-cm length of the fiber loop exposed in standard solutions. The perfusate (dialysate) was loaded directly into the sample loop of the FI system. The loading time for each perfusion rate studied was long enough to ensure complete filling of the sample loop with 20% extra rinsing time. The results show a decrease in recovery with an increase in perfusion rate, the decrease being most significant in the range 0-10 µL/min. This is in accordance
Figure 5. Chart recordings obtained in in vivo experiment with subcutaneous sampling after the administration of glucose (A) and insulin (B). (A) Glucose standard for detector calibration, 0.20 mM; V, intravenous injection of 0.08 g/kg body weight glucose. (a) Glucose concentration in subcutaneous tissue fluid vs time curve of (A) after correction for sensitivity drifts. (B) Glucose standard for detector calibration, 0.15 mM; V, intravenous injection of 4 IU/kg body weight insulin. (b) Glucose concentration in subcutaneous tissue fluid vs time curve of (B) after correction for sensitivity drifts; the peak marked with an asterisk occurred due to irregular movement of the test animal, and the corresponding data were omitted in (b).
with results reported by most other workers using in vivo microdialysis sampling systems,24 where perfusion rates of 1 µL/ min or lower were usually used. Owing to the low flow rates required in microdialysis, collection times of about 5-30 min are often necessary in order to obtain sufficient sample and/or sample volumes (1-20 µL) for analysis.25 In order to adapt the microdialysis system to FI sample introduction, it appears that either specially produced microvolume injectors should be used, or excessively low sample throughputs must be tolerated, which implies a loss in resolution in on-line monitoring processes. In the FIA systems reported by Amine et al.20 and Deterding et al.,25 two special valves with a sample volume of 1 µL were used to achieve sample throughputs of 30 and 42 h-1, respectively. Instead of using injectors with sample volumes in the microliter range, in this work a conventional injector with the lowest volume sample loop (20 µL) and conventional reactors were used to make the system more universally adaptable. Instead of filling the loop completely, the sample loop was only partially filled, the extent depending on the perfusion rate and loading time. Under such (24) Zhao, Y.; Liang, X.; Lunte, C. E. Anal. Chim. Acta 1995, 316, 403-410. (25) Deterding, L. J.; Dix, K.; Burka, L. T.; Tomer, K. B. Anal. Chem. 1992, 64, 2636-2641.
conditions, some loss of sensitivity through sample dispersion during transport of the dialysate sample to the detector is unavoidable. Larger sample volumes obtained by increasing either the perfusion rate or the loading time reduce the dispersion, however, only at the expense of lower recovery and/or throughput. For practical reasons, compromised conditions for achieving acceptable sensitivity and throughput were pursued in this study. Thus, the sample loading time was fixed at 80 s to maintain a sample throughput of 28 h-1 for achieving a reasonably high level of resolution in monitoring. The analyte response was then studied for different perfusion rates. Different perfusion rates within a large range of 5-50 µL/min produced almost the same peak height with 80-s loading time, apparently owing to compensation effects between simultaneous increases in dispersion and phase transfer. However, the precision of sample loading was degraded at lower perfusion rates. For the best precision, a perfusion rate of 20 µL/min employing a 20-µL sample loop and an 80-s sample loading time was adopted, in which case the sample volume was defined by the sample loop rather than the loading time. The microdialysis recovery was 3.1 ( 0.1% (n ) 6) under static conditions (subcutaneous measurements) and 4.6 ( 0.3% (n ) 6) under dynamic conditions (intravenous measurements), Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3575
Figure 6. Chart recordings obtained in in vivo experiment with intravenous sampling after the administration of glucose (A) and insulin (B). (A) Glucose standard for detector calibration, 0.50 mM; V, intravenous injection of 0.4 g/kg body weight glucose. (a) Glycemia concentration vs time curve of (A) after correction for sensitivity drifts. (B) Glucose standard for detector calibration, 0.30 mM; V, intravenous injection of 6 IU/kg body weight insulin. (b) Glycemia concentration vs time curve of (B) after correction for sensitivity drifts.
estimated by comparison of dialysates with standards without being subjected to microdialysis. Despite the relatively low recovery, sufficient sensitivity could be obtained with good precision using chemiluminescence detection, and high throughputs were then feasible using conventional FI equipment. The effects of sample flow rate on analyte peak height were studied for the flow-through microdialyzer, and the results are shown in Figure 4. The peak signal increased steeply at low flow rates but became much more gradual above 5 µL/min. A sample flow rate of 10 µL/min was employed in in vivo experiments for achieving sufficient sensitivity without significantly disturbing the fluid balance of the test animal, even under extended periods of observation. Performance of the FI On-Line Microdialysis Chemiluminescence System. In Vitro Experiments. Calibration graphs were linear in the range of 0-14.0 mM glucose using both the loop probe and flow-through microdialyzers expressed by the regression equations: I ) 0.998C + 0.009, r ) 0.9995; I ) 1.532C - 0.349, r ) 0.9996, I being the chemiluminescence intensity and C the glucose concentration. The normal level of glucose in blood and subcutaneous tissue fluid of rabbits is in the range of 5.510.0 mM. The detection limit based on 3 times the baseline noise was 0.01 mM glucose. The precisions of the system using the loop probe and flow-through microdialyzer were 0.51% and 0.68% RSD (n ) 11, 5.5 mM glucose), respectively. 3576
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
In Vivo Experiments. The in vivo monitoring of glucose in subcutaneous tissue fluid of rabbits after administration of glucose and insulin was performed using the loop probe microdialyzer with the system in Figure 3. Typical chart recording with calibrated concentration-time curves are shown in Figure 5. Following injection of glucose, the average times for initial rise and achieving maximum were 4.8 ( 0.9 and 15.2 ( 3.8 min (n ) 5), respectively. The average time for observing a decrease in glucose level after the injection of insulin was 12.5 ( 3.5 min (n ) 3). The in vivo monitoring of glucose in blood of rabbits after administration of glucose and insulin was performed using a flowthrough dialyzer with the system in Figure 3 as described in the Experimental Section. Typical chart recordings and concentration-time curves corrected for sensitivity drifts are shown in Figure 6. Following injection of glucose, the average times for initial rise and achieving maximum were 2.2 ( 0.2 and 7.4 ( 0.9 min (n ) 3), respectively. The average time for observing a decrease in glucose level after the injection of insulin was 6.2 ( 0.4 min (n ) 3). The recordings in Figures 5 and 6 show the advantages and necessity of an on-line calibration system which allows continuous monitoring of baseline and frequent recalibration. Since drifts in sensitivity are not always uniform, a calibration only at the beginning and at the end of the in vivo monitoring process can
produce errors by simple interpolation, leading to a distorted picture of the process. Microdialysis sampling in subcutaneous tissues is simpler than blood sampling both in the probe implanting process and in that no special measures are required for avoiding blood clogging in the transport conduits.1 However, at least as observed in our experiments, a motionless state was required for the test animal in order to avoid variations in recovery due to movement of the dialysis probe in the subcutaneous tissue. After the administration of glucose and insulin, retardations of about 2 min for glucose and 6 min for insulin in responding to the change of glucose level were also observed in subcutaneous sampling when compared to blood sampling. Thus, further work will be required before the monitoring of glucose in subcutaneous tissue fluids can replace that in blood, at least as demonstrated by results obtained with rabbits using the system described here. The recovery of the microdialyzer, evaluated before and after the in vivo monitoring using standard solution of glucose, showed a deviation of less than 5% in all experiments. Further calibration of the microdialyzer was not attempted in this work owing to this constancy of recovery. The effective time resolution of the microdialysis system is demonstrated in the chart recordings in Figure 6A, where the administration of glucose is reflected in significant rise in glycemia within 2 min. Despite the relatively low recovery of the microdialysis system, particularly in subcutaneous monitoring, the reproducibility of measurements was not significantly degraded. This was demonstrated by taking 11 successive measurements immediately before administration of glucose in in vivo experiments. The average
RSD for three in vivo experiments was 1.8%. This included drifts in the glucose level of the test animal during the 24 min used for collecting the data. CONCLUSIONS The FI microdialysis chemiluminescence system developed in this work demonstrated considerable potentials for in vivo online monitoring in achieving improved resolution and reliability using relatively simple instrumentation. The system can be readily adapted to other analytes by varying the enzyme reactor. While 2.2 min sampling intervals were used for these experiments, shorter times may be used for achieving higher resolution if required. The concept of on-line recalibration for sensitivity variation may be readily adapted to other detection systems in on-line monitoring to improve reliability, and the system design forms the basis for a computerized ingenious feed-back system which could automatically control drug or metabolite levels in body fluids of living systems. ACKNOWLEDGMENT The authors thank Professor Qing-Hai Yu, Ms. Jing-Hua Xu, and Mr. Shu-Li Zhang of Shenyang Pharmaceutical University for their technical support on in vivo experiments. Received for review March 25, 1997. Accepted June 5, 1997.X AC970324E X
Abstract published in Advance ACS Abstracts, July 15, 1997.
Analytical Chemistry, Vol. 69, No. 17, September 1, 1997
3577
