Biosensors and Bioelectronics 24 (2009) 1417–1423 Contents lists available at ScienceDirect Biosensors and Bioelectronics j our nal homepage: www. el sevi er . com/ l ocat e/ bi os Wireless enzyme sensor system for real-time monitoring of blood glucose levels in fish Hideaki Endo a,∗ , Yuki Yonemori a , Kyoko Hibi a , Huifeng Ren a , Tetsuhito Hayashi a , Wakako Tsugawa b , Koji Sode b a Department of Ocean Sciences, Faculty of Marine Science, Tokyo University of Marine Science and Technology, 4-5-7 Konan, Minato-ku, Tokyo 108-8477, Japan b Department of Biotechnology and Life Science, Faculty of Engineering, Tokyo University of Agriculture and Technology, 2-24-16 Naka-cho, Koganei-shi, Tokyo 184-8588, Japan arti cle i nfo Article history: Received 13 April 2008 Received in revised form 9 August 2008 Accepted 11 August 2008 Available online 2 September 2008 Keywords: Biosensor Enzyme sensor Glucose Blood Wireless Telemetory Real-time Monitoring Fish abstract Periodic checks of fish health and the rapid detection of abnormalities are thus necessary at fish farms. Several studies indicate that blood glucose levels closely correlate to stress levels in fish and represent thestateofrespiratoryornutritionaldisturbance. Wepreparedawirelessenzymesensorsystemto determine blood glucose levels in fish. It can be rapidly and conveniently monitored using the newly developed needle-type enzyme sensor, consisting of a Pt–Ir wire, Ag/AgCl paste, and glucose oxidase. To prevent the effects of interfering anionic species, such as uric acid and ascorbic acid, on the sensor response, the Pt–Ir electrode was coated with Nafion, and then glucose oxidase was immobilized on the coated electrode. The calibration curve of the glucose concentration was linear, from 0.18 to 144mg/dl, and the detection limit was 0.18mg/dl. The sensor was used to wirelessly monitor fish glucose levels. The sensor-calibrated glucose levels and actual blood glucose levels were in excellent agreement. The fluid of the inner sclera of the fish eyeball (EISF) was a suitable site for sensor implantation to obtain glucose sample. There was a close correlation between glucose concentrations in the EISF and those in the blood. Glucose concentrations in fish blood could be monitored in free-swimming fish in an aquariumfor 3 days. © 2008 Elsevier B.V. All rights reserved. 1. Introduction In recent years, mass outbreaks of pathogenic bacterial infec- tionshavekilledlargequantitiesoffish, causinggreatfinancial damage tofishfarms. Major disease outbreaks typically occur when the balance among the fish, the etiologic agent, and the environ- ment is disturbed. To prevent or reduce the severity of epizootics, impaired disease resistance in fish must be detected at the earli- est opportunity. Periodic checks of fish health for early detection of abnormalitiesarethusnecessaryat fishfarms. Further, it is importanttoestablishfarmingtechniqueswithminimaldepen- denceonantibioticsandantimicrobialagentsforfarmingsafety andhigh-qualityculturedfish. Toaddressthisproblem, heme- analysismethodsforfishbloodhavebeenarecentfocusinfish physiology. Several studies have demonstrated that blood glucose levelscloselycorrelatetothestresslevelsinfishandarerep- resentativeof therespiratoryandnutritional state(Carballoet al., 2005;Chowdhuryetal., 2004;Jentoftetal., 2005;Trenzado etal., 2003;VanAnholtetal., 2004). Inaddition, bloodcholes- ∗ Corresponding author. Tel.: +81 3 5463 0616; fax: +81 3 5463 0616. E-mail address:
[email protected] (H. Endo). terol levels are a useful indicator of reduced resistance to bacterial infection(Maitaetal., 1998a,b). Therefore, monitoringof blood glucoseandcholesterol levelsisimportantforascertainingfish health. These tests are currently conducted using clinical labora- tory test kits designed for humans. Furthermore, as each sample needs to be analyzed separately, testing is time- and labor- intensive. For fishhealthtestingat fishfarms, it is necessary totestasmanyfishaspossibleinafastandconvenientman- ner. Recently, researchers have beenactively investigating the devel- opmentofbiosensorsformeasuringbloodcomponentssuchas glucose (Hiratsuka et al., 2001; Ishikawa et al., 1998; Jeong et al., 2003; Quintoet al., 2000; WangandZhang, 2001), cholesterol (Bertrand et al., 1979; Bongiovanni et al., 2001; Karube et al., 1982; MotonakaandFaulkner, 1993;Satohetal., 1997)andlacticacid (Wu et al., 2000). Wang and Zhang (2001) developed a miniature needle-type enzyme sensor suitable for simultaneous amperomet- ric monitoring of glucose and insulin. The sensor was constructed from dual modified carbon-paste working electrodes inserted into a14-Gneedle. Wuetal. (2000)reportedanenzymesensorfor whole blood lactate monitoring. The sensor consisted of a stainless steel needle with the surface modified for enzyme immobilization usingapolymermaterial. Studiesof non-invasiveglucosesen- 0956-5663/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.bios.2008.08.038 1418 H. Endo et al. / Biosensors and Bioelectronics 24 (2009) 1417–1423 sorshavealsobeenundertakenforapplicationinpatientswith diabetes(Caduff etal., 2003;Gourzi etal., 2003, 2005). Caduff et al. (2003) developedaglucosemonitoringsystembasedon impedance spectroscopy. Changes inglucose concentrations canbe monitored by varying the frequency in the radio band over a range optimized to measure the impact of glucose on the impedance pat- tern. Most sensor systems, however, are designed for humans or livestock. We recently developed a biosensor to measure cholesterol lev- els in fish plasma in an attempt to establish rapid and convenient methods for ascertainingfishhealth(Endoet al., 2003). Becausethis method is based on flow injection analysis (FIA), continuous mea- surement is possible and only 5min is required for analysis of each sample, markedly reducing testing time. Even this method, how- ever, requires a certain amount of skill to draw blood, and because the method is based on FIA, the measurement apparatus itself is unwieldy and complicated. Tests can be performed in a laboratory, but testing at fish farms is difficult and inconvenient. To establish a simple and rapid method to determine blood glucose levels in fish, we developed a needle-type enzyme sensor system (Endo et al., 2006). The sensor consists of a needle-type hollow container, immobilizedenzymemembrane, andfiberopticprobecontain- ing a ruthenium complex. The enzyme membrane was prepared fromglucoseoxidase, AWP(azide-unitdependentwater-soluble photopolymer), and ultra-thin dialysis membrane. The fiber optic probewasinsertedintotherolledenzymemembraneplacedin the needle-type hollow container. The calibration curve was linear intherangeof10–180mg/dlglucoseinfishblood(Niletilapia). The sensor was inserted into the caudal vein of the fish to mea- sure blood glucose levels. A very strong correlation was observed between values determined by the sensor and conventional meth- ods, in the range of 48–157mgdl −1 (correlation coefficient: 0.9474, y =0.8452x −3.4018). Asthesensor, spectrometer, andpersonal computer are small; this test system is easily transportable for use at fish farms. It is still necessary, however, to capture fish using a spoon-net from the fish tank to obtain the blood sample for each measurement. Further, it is difficult to measure blood components inreal-timebecausesensor output decreases over timeduetoblood coagulation and protein (e.g., albumin, ␥-globulin) coalescing on the sensor. Glucose biosensor systems have developed remarkably recently in the medical field. Minimed Inc. developed a continuous glucose monitoring system that allows blood glucose to be continuously monitoredfor 72h(Mastrototaro, 2000). Thesystemdoes not measurebloodglucoselevels, but rather, theglucoselevelsin the interstitial fluid (ISF) of subcutaneous tissue surrounding the sensor. Briefly, themethodis basedontheprinciplethat ISF glucoselevelsinsubcutaneoustissuereflectbloodglucoselev- els. Glucose kinetics can be described using a two-compartment model of theseparatedcapillarybloodglucoseandtissuebed (Rebrin and Steil, 2000). Even in fish, the two-compartment model might be applicable for continuous glucose monitoring in subcu- taneous tissue. We focused on the fluid-filled inner eyeball sclera (eyeball sclera ISF; EISF) and hypothesized that if the glucose dif- fusionrateisconstant intheEISFasinthetwo-compartment model, scleral ISFcanbeusedformonitoringglucoselevelsin fish. Thepresent paper describesthestepstowarddevelopinga rapid and convenient method for real-time monitoring blood glu- cose levels in fish, as follows: (1) investigation of the relationship betweenEISFandbloodglucoselevelsinfish, (2) preparation ofneedle-typeenzymesensorusingflexiblewireelectrodesfor implantingintothefishbody, (3)wirelessmonitoringof blood glucose levels under free-swimming conditions in an aquar- ium. 2. Materials and methods 2.1. Reagents Glucoseoxidase(fromAspergillusniger;E.C. 1.1.3.4, typeVII- S;197000unit/g)waspurchasedfromSigma–Aldrich(St. Louis, MO). Bovineserumalbumin(BSA)and2-phenoxyethanolwere purchased from Wako Pure Chemical Industries (Tokyo, Japan) d- glucose was dissolved in 0.1Mphosphate buffered saline (PBS) and allowedto mix by the rotationfor 24hbefore use. All other reagents used for the experiments were commercial- or laboratory-grade. 2.2. Relationship between blood and EISF glucose levels Nile tilapia (Oreochromis niloticus) cultured at Tokyo University of Marine Science & Technology was used as the model fish. A test was performed to confirm whether the temporal change in blood glucose levels is reflected by a change EISF glucose levels. A 50- l fish tank (60cm×30cm×36cm) was used. During the test, the water temperature was 30 ◦ Cwithcontinuous aerationandrecircu- lation through a biologic filter. The photoperiod was set with lights on from 10:00 to 20:00h provided by a fluorescent light. The fish (body weight, ca. 300g; body length, ca. 25cm) were netted from the preserve andanesthetizedwith0.1%2-phenoxy ethanol by bath exposurefor5min. BloodwasthensampledfromtheCuvierian duct(Ikedaetal., 1985)usingaheparinizedsyringefittedwith a 27-G needle. The blood samples (100–500l) were centrifuged (650×g, 10min) to separate the plasma. EISF sampling was per- formed as follows; while slightly pushing on one eye, a blister-like membrane rose up to the surface. A heparinized syringe fitted with a 27-G needle was then inserted into the blister-like membrane. EISF samples (50–100l) were obtained from the ISF of the eye- ball sclera. To reduce measurement error due to the stress induced by handling the fish, the sampling procedure was restricted to less than 5min per fish. EISF samples were transferred to test tubes and stored at −80 ◦ C before use. Blood and EISF samples were collected at the same hour each day using the procedure described above. The test was performed over 27 days and the fish were fed at days 5and16, 4hbeforeglucosemeasurement, toevaluatewhether the change in the blood glucose concentration was continuously reflected by the EISF glucose concentration. Glucose levels were determined using an enzymatic colorimet- ricmethod(C-II glucosetest; WakoPureChemical Industries). When each sample (20l) was mixed with 3ml PBS (pH 7.1) con- taining enzymes (glucose oxidase, horseradish peroxidase) and the color-producing reagent, the glucose was oxidized, producing glu- conolactone and hydrogen peroxide. The hydrogen peroxide was reacted with 4-aminoantipyrine and phenol to produce a red color. Glucose concentrations were measured using a UV/vis spectropho- tometer (JASCO Co., Tokyo, Japan) 2.3. Enzyme sensor preparation A needle-type enzyme sensor was prepared for subcutaneous glucosemonitoringinfish. Fig. 1(a)showstheschematicofthe needle-type enzyme sensor. A working electrode was made using an18mmlengthof Teflon-coatedplatinumiridiumwire. The Teflon wasstrippedatoneendtoexpose1mmofthePt–Irwireasa sensing cavity. Copper wire was wrapped around the Teflon coated surface. An Ag/AgCl paste (BAS, Tokyo, Japan) wire was used as a reference electrode/counter electrode. The tip of the wire (2mm) was sealed with epoxy resin. The sensing cavity was dipped in 5% Nafion solution and dried for 10min. The Nafion-coated electrode wasdippedin6%acetyl-cellulosesolutionanddriedfor10min. Anenzymesolutioncontaining2mg(352unit/ml) glucoseoxi- H. Endo et al. / Biosensors and Bioelectronics 24 (2009) 1417–1423 1419 Fig. 1. Schematic diagramof the wireless monitoring systemfor fish. 1, needle-type enzyme sensor; 2, waterproof sheet; 3, wireless potentiostat; 4, nylon threads; 5, receiver; 6, personal computer; 7, sample fish (Nile tilapia). dase and 20mg BSA dissolved in 1ml 0.1M phosphate buffer (pH 7.8) was freshly prepared. The Nafion and acetyl-cellulose-coated electrodewasdippedintheenzymesolutionandair-driedfor 10min; this procedure was repeated twice. The electrode was held inaverticalpositioninavial, and100-lglutaraldehyde(50%) wasaddedtoinducecross-linkingbetweentheglucoseoxidase and BSA. The electrodes were stored at 30 ◦ C for 6h. The sensors were soaked and stored in PBS (pH 6.5) overnight at 4 ◦ C before use. The prepared sensor was connected to a wireless potentiostat (3102BP, Pinnacle Technology Inc., Lawrence, KS), and the measure- ment was performedusingareceiver (3100RX, PinnacleTechnology Inc.) andapersonal computer(Dimension4700C, DELL, Round Rock, TX). A650-mVpotential(vs. Ag/AgCl)wasappliedbythe potentiostat to the Pt–Ir working electrode for the amperometric glucose measurement. Hydrogenperoxide, producedby anenzyme reaction inducing glucose oxidation, was detected by an electrocat- alytic oxidation reaction at the sensing cavity of the Pt–Ir working electrode, producing an electric output current. 2.4. Measurement of glucose standard solution For the amperometric glucose measurements, the sensor was immersedintoacell consistingof 10ml of 0.1MPBS(pH6.5) at 25 ◦ C and dissolved oxygen was saturated with stirring. The background outputcurrentwasallowedtostabilizebeforemeasurement. A glucosestandardsolutionwas thenaddedtothecell andthe responsetimetoreach90%steadystatelevel (T90) wasmea- sured. When the sensor output had stabilized, the procedure was repeated until the sensor output reached the rate-limiting enzyme phase. 2.5. Sensor implantation and device immobilization Aschematic diagramof thewireless monitoringsystemis shown in Fig. 1(b). The systemwas constructed using a needle-type enzyme sensor, a wireless potentiostat, a receiver, and a personal computer. Sensor implantation and fixation was performed using thefollowingprocedure. Niletilapiawasstarvedfor24hbefore continuous glucose monitoring. First, the fish were anesthetized in a water bath containing 0.1% 2-phenoxy ethanol for 5min. A 22- G catheter consisting of an outer Teflon layer and inner puncture needle (23mm, SerflowTM, Terumo Co, Tokyo, Japan) was inserted into the ISF of the eyeball sclera for sensor implantation. The inner punctureneedlewasremovedandtheexcessexposedcatheter on the skin was removed. The sensor was then inserted into the outerTeflonlayerandfixedinplaceusingbiomedical adhesive (Aronalpha-ASankyo, Toagosei, Tokyo, Japan), whichcontained anethylcyanoacrylatemonomerasthemajoringredient. Water- proofing the device was essential because the wireless potentiostat was not designed to work under water. The potentiostat was cov- eredusingawaterproofpolypropylenesheet(HondaMotorCo. Ltd., Tokyo, Japan) and sealed with a thermo-compression bond- ingdevice(NL-101J, IshizakiElectricMfgCo. Ltd., Tokyo, Japan). The waterproofed wireless potentiostat was attached to the dorsal andpectoral fins of the fishusing nylonthreads andthenconnected to the sensor. 2.6. Monitoring of blood glucose levels in fish Nile tilapia equipped with the sensor systemwas transferred to a 50-l fish water tank (60cm×30cm×36cm) and allowed to swim freely. Dissolved oxygen in the water tank was saturated using an airpump. Atfirst, tocomparebetweendirectlymeasuredblood glucose levels and sensor-estimated glucose levels, blood was col- lected and blood glucose levels were measured using a Glucocard (Arkray KDK Co. GT 1640, Kyoto, Japan). After a few minutes, the fish were placed back in the tank, and the sensor output current wasrecordedviacomputer-controlleddataacquisitionsoftware (Pinnacle Acquisition Laboratory, Pinnacle Technology). The estimation of blood glucose levels was performed using a one-point calibration method or a two-point calibration method in vivo. For the one-point calibration method, the sensor sensitiv- ity “S” was determined from a single blood glucose measurement as the ratio between the sensor output current “I” and the blood glucose level “G”. Glucose levels were then estimated fromcurrent “I” using the equation G(t) =I(t)/S. The sensor output current was first allowed to stabilize after the sensor was implanted into the sclera. Bloodglucosewas thensampledfromtheCuvierianduct and the blood glucose level “G” was measured. These parameters (I, G) wereusedas thecalibrationreference. For thetwo-point calibration method, the sensor sensitivity “S”, expressed in (nA)/(mg/dl), and the intercept “I 0 ” with the sensor output (current, expressed in nA) axis, i.e., the theoretical sensor output that would be observed in the absence of glucose in ISF as I 0 =I 1 −G 1 ×S. These parameters “S” and “I 0 ” were subsequently used to transformthe sensor output I(t) to estimate glucose levels G(t) using the equation G(t) =(I(t) −I 0 )/S. AmethodfordeterminingSandI 0 takesintoaccount, duringa change in blood glucose levels “G” from “G 1 ” to “G 2 ”, the change in the sensor current I from “I 1 ” to “I 2 ”: S =(I 2 −I 1 )/(G 2 −G 1 ), and I 0 =I 1 −(S ×G 1 ). 1420 H. Endo et al. / Biosensors and Bioelectronics 24 (2009) 1417–1423 Fig. 2. Relationship between blood and eyeball ISF (EISF) glucose levels. Nile tilapia wasbredina50-lfishtank. Thewatertemperaturewas30 ◦ Cwithcontinuous aeration and recirculation through a biologic filter. The photoperiod was set with lights on from 10:00 to 20:00h provided by a fluorescent light. 3. Results and discussion 3.1. Relationship between blood and EISF glucose levels As described above, continuous monitoring of the blood com- ponentsisdifficultusingabiosensorbecausethesensoroutput decreases over time due to blood coagulation and coalescing pro- tein on the sensor. Additionally, implantation of enzyme sensors into fish blood vessels needs expertise due to their fine vascular anatomy. Therefore, we focused on EISF and hypothesized that if theglucosediffusionrateisconstantintheEISFasinthetwo- compartment model, thescleral ISFcanbeusedasthesensor implantation site in fish. The relationship between blood and EISF glucose levels was therefore investigated. Continuous changes in glucose levels following feeding were investigated in the blood and EISF. Fig. 2 shows the temporal changes in glucose levels in Nile tilapia. Blood glucose levels were almost equal to EISF glucose lev- els. Additionally, after feeding, glucose levels in the EISF increased along with an increase in the blood glucose levels. These glucose levels were40–50mg/dl higher thannormal glucoselevels. Theglu- cose levels returnedto the normal 30hafter feeding. These findings suggest that blood glucose levels are rapidly reflected by EISF glu- cose levels. The glucose levels ranged from20 to 60mg/dl. Because the average glucose level in Nile tilapia is generally in the range of 50mg/dl (Nolanet al., 1999), these measuredvalues are reasonable. These findings indicated that blood glucose levels can be estimated from the EISF in fish. 3.2. Response curve and calibration curve of the sensor The most common glucose amperometric enzyme sensors are based on the electrochemical detection of hydrogen peroxide pro- ducedbytheenzymaticoxidationofglucose. Withintheliving body, thereisalsoconcomitantelectro-oxidationofmanyinter- fering species, such as uric acid and ascorbic acid. Nafion film is highly selective due to its negatively charged sulfonic functional groups (Harrison et al., 1988; Hu et al., 1994). In this study, to pre- vent the effects of these interfering species on the sensor response, thePt–Ir electrodewas coatedwithNafionandthenglucoseoxidase was immobilized on the coated electrode. Fig. 3(a) shows a typical response curve of the needle-type enzyme sensor equipped with a Nafion-coated membrane for various substrates, such as glucose, ascorbic acid, and uric acid. The sensor was immersed into a cell containing 10ml of 0.1M PBS (pH 6.5) at 25 ◦ C and dissolved oxy- gen was saturated with stirring and then each substrate solution was added to the cell. For the glucose solution (0.1mM; 18mg/dl), the sensor output increased within 30s after adding the sample, and a maximum current was obtained within 60s. When ascorbic acid (0.1mM) or uric acid solution (0.1mM) was added to the cell, the sensor output current did not change. Without Nafion, the sen- sor produced a clear response when ascorbic acid or uric acid was addedtothecell (Fig. 3(b)). Therefore, theNafioncoatingis required for in vivo glucose monitoring. Weinvestigatedtherelationshipbetweensensoroutputand theglucosestandardsolutionconcentration(Fig. 4). Thecali- brationcurveoftheglucoseconcentrationwaslinearfrom0.18 to144mg/dl (y =0.3315x +1.013, R=0.9983), andthe detection limit was 0.18mg/dl. The reproducibility of the sensor output was demonstratedbyusingasingleelectrodetomake20measure- ments of an 18-mg/dl glucose solution (Table 1). Five needle-type enzyme sensors were examined. The sensors were evaluated based Fig. 3. Typicalresponsecurvesofaneedle-typeenzymesensorwithandwith- out Nafion membrane. (a) Enzyme sensor with Nafion; (b) enzyme sensor without Nafion. Sample solution was added in pH 6.5 PBS under a +650-mV potential (vs. Ag/AgCl) for 25 ◦ C. H. Endo et al. / Biosensors and Bioelectronics 24 (2009) 1417–1423 1421 Fig. 4. Relationship between sensor output and the glucose standard solution con- centration. The assay conditions were the same as those described for Fig. 5. on the relative standard deviation because sensor output current varies fromlot to lot. The sensors had good reproducibility for stan- dard glucose measurement. (R.S.D.%=1.93). The low R.S.D. values indicated that these glucose biosensors can be used to determine glucose withsatisfactory precisionandcantherefore provide stable and accurate continuous in vivo glucose measurements. 3.3. Correlation of actual blood glucose levels and glucose levels estimated using the wireless monitoring system Theneedle-typeenzymesensorwasusedforwirelessmon- itoringof bloodglucoselevelsinfish. Tocompareactual blood glucoselevels andsensor-estimatedglucoselevels for calibra- tion, Niletilapiawasanesthetizedwith0.1%2-phenoxyethanol by bath exposure for 5min, and then the wireless sensor system wasattached. Thefishwastransferredtoa50-lfishwatertank containing 1ppmof 2-phenoxy ethanol as a light-anesthetic agent. After 30min, the output current of the sensor was monitored when the fishbeganswimming. After continuous monitoring for 180min, blood samples were collected from fish several times and the fish were placed back into the tank. Fig. 5 shows the response curves for in vivo continuous glucose monitoring over 180min with fish underlight-anesthesia. AsshowninFig. 5(a), initial glucoselev- elswere85mg/dl (point(o)). Theinitial sensoroutputcurrent increased with an increase in blood glucose levels. Sensor-calibratedglucose levels were estimatedusing either the one-point or two-point calibration method. Fig. 5(b) and (c) rep- Table 1 Reproducibility of glucose measurements using needle-type enzyme sensor Electrode no. Standard deviation (S.D.) Relative standard deviation (R.S.D.%) 1 1.31 1.38 2 0.70 2.29 3 0.61 2.77 4 2.97 1.66 5 2.70 1.55 Average 1.66 1.93 resents blood glucose levels and sensor-calibrated glucose levels calculated using both calibration methods. The estimated glucose levels using both calibration methods paralleled the blood glucose levels. In Fig. 5(b), one-point calibration (1) was used as calibra- tion value (G 1 , I 1 ) in which G 1 was the initial blood glucose level of 88mg/dl and I 1 was the subsequently obtained sensor output current(1.59nA) at30min(0minonthemonitoringtime)after sensor implantation (point (o)). The one-point calibration (2) was calculated using as calibration value (G 2 , I 2 ) in which the blood glu- cose level and sensor output current obtained 45min (15min on the monitoring time) after sensor implantation (G 2 , 261mg/dl; I 2 , 7.81nA; point (p)). As shown in the figures, the one-point calibra- tion (2) yielded more accurate glucose monitoring than one-point calibration (1), except at 30min of the monitoring time, possibly due to the saturation of the sensor output current when the actual bloodglucoselevelwas430mg/dlbecausethecalibrationcurve Fig.5. Wirelessmonitoringofglucoselevelsforshort-termunderlightanesthe- sia. (a) Comparison between the change in blood glucose levels and sensor output current; (b) Comparisonbetweenbloodglucose levels andsensor-calibratedglucose levels calculatedusingone-point calibration; (c) Comparisonbetweenbloodglucose levels and sensor-calibrated glucose levels calculated using two-point calibration. Sensor implantation and fixation was performed at −30min. 1422 H. Endo et al. / Biosensors and Bioelectronics 24 (2009) 1417–1423 ofthesensorwaslinearforglucoselevelsrangingfrom0.18to 144mg/dl (Fig. 4). Moreover, in the one-point calibration method, not enough time was allowed for the background output current (I 0 ) to stabilize, resulting in an inaccurate calibration reference (I, G). In animals, approximately 2h is necessary for the background current to stabilize (Rebrin et al., 1999). Therefore, the one-point calibration method should be performed at least 45min after sen- sor implantation. On the other hand, the two-point calibration utilized the first parameter (I 1 ; 1.59nA, G 1 ; 88mg/dl) (point (o)) at 30min and the second parameter (I 2 ; 7.81nA, G 2 ; 261mg/dl) (point (p)) at 45min after sensor implantation (Fig. 5(c)). The sensor-estimated glucose levels and actual blood glucose levels were well-correlated. How- ever, despite its theoretical superiority, the two-point calibration method is a time-consuming procedure and is susceptible to mea- surement errors (Choleau et al.). Moreover, although the two-point calibrationprocedurecanbeusedtodeterminesensorsensitiv- ity(S)andthebackgroundcurrent(I 0 )byobtainingthecurrent responses (I 1 , I 2 ) to two different glucose levels (G 1 , G 2 ), it is very difficult to obtain blood samples from unanesthetized fish under free-swimming conditions and without an anesthetic. Therefore, in this study, the one-point calibration method (the calibration time point was better than 45min after sensor implantation) was used to monitor the glucose levels in fish. 3.4. Wireless monitoring of glucose levels in fish We used the needle-type enzyme sensor to wirelessly monitor glucose levels in free-swimming fish in an aquarium. Nile tilapia was anesthetized with 0.1% 2-phenoxy ethanol, and then the wire- less sensor systemwas attached. The fish were transferred to a 50-l fish water tank containing no anesthetic. The fish became active within 30min. The time courses of the output current of the sen- sor and sensor-calibrated glucose levels are shown in Fig. 6. The calibrationtime point was 2hafter sensor implantationfor the one- pointcalibrationmethod. Initially, theglucoselevelswerehigh, which may have been caused by stress induced by fixation of the sensor or by the sensor itself. The levels decreased gradually and stabilized after 24h. After feeding, glucose levels again increased and the maximum value was 70mg/dl at 48h. Subsequently, the glucose levels decreased gradually and stabilized over 60h. In this experiment, the glucose levels infree-swimmingfishcouldbe mea- sured for 3 days inthe aquarium. Inmore than10 replications of the experiment, the sensor response was similar. Although the sensor was not sterilized prior to implantation into the fish, there was no apparent effect on the active conditions of the fish. However, mon- itoring with the wireless potentiostat could not continue for more Fig.6. Wirelessmonitoringofbloodglucoselevelsinfish. Niletilapiaequipped with the sensor systemwas placed in a 50-l fish water tank without anesthesia and allowed to swim freely. than3days because of insufficient battery power. This newly devel- oped sensor provided rapid and convenient real-time monitoring of glucose levels in fish. 4. Conclusion The present findings confirm that blood glucose levels in fish can be rapidly and conveniently monitored using the newly devel- opedneedle-typeenzymesensor, consistingof aPt–Ir wire, Ag/AgCl paste, andglucoseoxidase. Topreventtheeffectsofinterfering anionicspecies, suchasuricacidandascorbicacid, onthesen- sor response, the Pt–Ir electrode was coated with Nafion, and then glucose oxidase was immobilized on the coated electrode. The cal- ibration curve of the glucose concentration was linear, from 0.18 to 144mg/dl, and the detection limit was 0.18mg/dl. Nafion coat- ing was useful for monitoring the glucose concentration in vivo. The sensor was used to wirelessly monitor fish glucose levels. The sensor-calibratedglucoselevelsandactual bloodglucoselevels were in excellent agreement. The fluid of the inner sclera of the fish eyeball (Nile tilapia) was a suitable sampling site to obtain ISF. Therewas aclosecorrelationbetweenglucoseconcentrations inthe EISF and those in the blood. Glucose concentrations in fish blood couldbemonitoredinfree-swimmingfishinanaquariumfor3 days. In future studies, we aimto solve the rapid battery consump- tion problem and apply the sensor system to various types of fish such as carp, rainbow trout, and eel. Acknowledgments This research was supported in part by a Grant-in-Aid for Scien- tific Research (B) from The Ministry of Education, Culture, Sports, ScienceandTechnology, andaGrant-in-AidfromtheFacultyof Marine Science, Tokyo University of Marine Science and Technol- ogy. References Bertrand, C., Coulet, P.R., Gautheron, D.C., 1979. Anal. Lett. 12, 1477–1488. Bongiovanni, C., Ferri, T., Poscia, A., Varalli, M., Santucci, R., Desideri, A., 2001. Bio- electrochemistry 54, 17–22. Caduff, A., Hirt, E., Feldman, Y., Ali, Z., Heinemann, L., 2003. Biosens. Bioelectron. 19, 209–217. Carballo, M., Jimenez, J.A., de la Torre, A., Roset, J., Munoz, M.J., 2005. Environ. Toxicol. 20, 119–125. Chowdhury, M.J., Pane, E.F., Wood, C.M., 2004. Comp. Biochem. Physiol. C, Toxicol. Pharmacol. 139, 163–173. Endo, H., Maita, M., Takikawa, M., Ren, H., Hayashi, T., Urano, N., Mitsubayashi, K., 2003. Fish. Sci. 69, 1194–1199. Endo, H., Yonemori, Y., Musiya, K., Maita, M., Shibuya, T., Ren, H., Mitsubayashi, K., 2006. Anal. Chim. Acta 573–574, 117–124. Gourzi, M., Rouane, A., Guelaz, R., Nadi, M., Jaspard, F., 2003. J. Med. Eng. Technol. 27, 276–281. Gourzi, M., Rouane, A., Guelaz, R., Alavi, M.S., McHugh, M.B., Nadi, M., Roth, P., 2005. J. Med. Eng. Technol. 29, 22–26. Harrison, D.J., Turner, R.F.B., Baltes, H.P., 1988. Anal. Chem. 60, 2002–2007. Hiratsuka, A., Kojima, K., Suzuki, H., Muguruma, H., Ikebukuro, K., Karube, I., 2001. Analyst 126, 658–663. Hu, Y., Mitchell, K.M., Albahadily, F.N., Michaelis, E.K., Wilson, G.S., 1994. Brain Res. 659, 117–125. Ikeda, M., Tsuda, S., Kurawaka, K., Fukao, R., Shirasu, Y., 1985. J. Soc. Vet. Sci. 47, 951–955. Ishikawa, M., Schmidtke, D.W., Raskin, P., Quinn, C.A., 1998. J. Diab. Comp. 12, 295–301. Jentoft, S., Aastveit, A.H., Torjesen, P.A., Andersen, O., 2005. Comp. Biochem. Physiol. A: Mol. Integr. Physiol. 141, 353–358. Jeong, R.A., Hwang, J.Y., Joo, S., Chung, T.D., Park, S., Kang, S.K., Lee, W.Y., Kim, H.C., 2003. Biosens. Bioelectron. 19, 313–319. Karube, I., Hara, K., Matsuoka, H., Suzuki, S., 1982. Anal. Chim. Acta 139, 127. Maita, M., Aoki, H., Yamagata, Y., Satoh, S., Okamoto, N., Watanabe, T., 1998a. Fish. Pathol. 33, 59–63. Maita, M., Satoh, K., Fukuda, Y., Lee, H., Winton, J.R., Okamoto, N., 1998b. Fish. Pathol. 33, 129–133. Mastrototaro, J.J., 2000. Diabetes Technol. Ther. 2, 13–18. H. Endo et al. / Biosensors and Bioelectronics 24 (2009) 1417–1423 1423 Motonaka, J., Faulkner, L.R., 1993. Anal. Chem. 65, 3258–3261. Nolan, D.T., Op’t Veld, R.L.J.M., Balm, P.H.M., Bonga, S.E.W., 1999. Aquaculture 177, 297. Quinto, M., Losito, I., Palmisano, F., Zambonin, C.G., 2000. Fresenius J. Anal. Chem. 367, 692. Rebrin, K., Steil, G.M., Van Antwerp, W.P.J.J., Mastrototaro, J.J., 1999. Am. J. Physiol. Endocrinol. Metab. 277, 561–571. Rebrin, K., Steil, G.M., 2000. Diabetes, Technol. Ther. 2, 461–472. Satoh, I., Karube, I., Suzuki, S., 1997. Biotechnol. Bioeng. 19, 1095–1099. Trenzado, C.E., Carrick, T.R., Pottinger, T.G., 2003. Gen. Comp. Endocrinol. 133, 332–340. VanAnholt, R.D., Spanings, F.A., Koven, W.M., Bonga, S.E., 2004. Gen. Comp. Endocrinol. 139, 215–226. Wang, J., Zhang, X., 2001. Anal. Chem. 73, 844–847. Wu, M., Lin, Z., Li, Y., Ren, S., 2000. Sens. Actuator B 66, 269–271.