Anal. Chem. 2000, 72, 4185-4192
A Fluorescence Affinity Hollow Fiber Sensor for Continuous Transdermal Glucose Monitoring Ralph Ballerstadt and Jerome S. Schultz*
University of Pittsburgh, Center for Biotechnology and Bioengineering, 300 Technology Dr., Pittsburgh, Pennsylvania 15219
A novel concept of a fluorescence affinity hollow fiber sensor for transdermal glucose monitoring is demonstrated. The glucose-sensing principle is based on the competitive reversible binding of a mobile fluorophorelabeled Concanavalin A (Con A) to immobile pendant glucose moites inside of intensely colored Sephadex beads. The highly porous beads (molecular weight cutoff of 200 kDa) were colored with two red dyes, Safranin O and Pararosanilin, selected to block the excitation and spectrum of the fluorophore Alexa488. The sensor consists of the dyed beads and Alexa488-Con A confined inside a sealed, small segment of a hollow fiber dialysis membrane (diameter 0.5 mm, length 0.5 cm, molecular weight cutoff 10 kDa). In the absence of glucose, the majority of Alexa488-Con A resides inside the colored beads bound to fixed glucose. Thus, excitation light at 490 nm impinging on the sensor is strongly absorbed by the dyes, resulting in a drastically reduced fluorescence emission at 520 nm from the Alexa488-Con A residing within the beads. However, when the hollow fiber sensor is exposed to glucose, glucose diffuses through the membrane into the sensor chamber and competitively displaces Alexa 488-Con A molecules from the glucose residues of the Sephadex beads. Thus, Alexa 488-Con A appears in the void space outside of the beads and is fully exposed to the excitation light, and a strong increase in fluorescence emission at 520 nm is measured. At a medium to high loading degree of Sephadex with Alexa488-Con A (10 mg mL-1 bead suspension), the absolute fluorescence increase due to 20 mM glucose was very large. It exceeded the response of other sensor devices based on FRET by a factor of 50 (Meadows and Schultz Anal. Chim. Acta 1993, 280, 21-30; Russell et al. Anal. Chem. 1999, 71, 3126-3132). The new sensor featured a glucose detection range extending from 0.15 to 100 mM, exhibiting the strongest dynamic signal change from 0.2 to 30 mM. It showed a reasonably fast response time (4-5 min). The combination of all the beneficial sensor features makes this sensor extremely attractive for future in vivo implantation studies for glucose monitoring in subdermal tissue. The Diabetes Control and Complication Trial (DCCT) report supplied strong arguments for an intensive treatment of insulin* To whom correspondence should be addressed. Phone: (412) 383 9700. Fax: (412) 383 9710. E-mail:
[email protected]. 10.1021/ac000215r CCC: $19.00 Published on Web 08/08/2000
© 2000 American Chemical Society
dependent diabetes mellitus.1 The study showed that the development and progression of a variety of complications ranging from diabetic retinopathy, nephropathy, and neuropathy can be significantly reduced by measuring glucose levels several times a day to guide insulin administration to the individual. The use of an implantable glucose-specific biosensor has been widely accepted as a needed tool for such an intensive treatment.2-5 Various prototypes of fluorescence affinity hollow fiber sensors have been developed in recent years that were targeted at in vivo blood glucose monitoring.6-9 They were comprised of a semipermeable membrane chamber (hollow fiber) which ensured that the analytical macromolecular reagents remain confined inside the chamber while glucose from surrounding tissue can diffuse freely through it. The hollow fiber was linked to the distal end of a single optic fiber. This setup allowed one to guide light into the semipermeable membrane chamber and to measure the fluorescence emission as the result of competitive reaction between glucose and dextran for the binding site of Concanavalin A (ConA), which is a glucose and mannose-specific lectin.10 The emitted fluorescence light was coupled back into the optical fiber and detected by means of a photomultiplier. This scheme has been tested successfully in vivo in a canine model.7 A potential disadvantage of this approach is the risk of bacterial infection at the site where the optical fiber penetrates the skin. To avoid this danger and take advantage of the relatively low light absorption of the upper skin tissue, our group has proposed a semi-invasive approach for glucose monitoring.11 The method relies on subcutaneous implantation of a self-contained semipermeable membrane chamber containing the glucose-selective bioreagents, which display a fluorescence change at rising glucose concentrations. The sensor is expected to be placed within 0.5 mm of the skin surface to permit easy penetration of light to and from the sensor (1) Diabetes Control and Complication Trial Research Group N. Engl. J. Med. 1993, 329, 977-986. (2) Gerritsen, M.; Jansen, J. A.; Kros, A.; Nolte, R. J. M.; Lutterman, J. A. J. Invest. Surgery 1998, 11, 163-174. (3) Abel, P.; Muller, A.; Fischer, U. Biomed. Biochim. Acta 1984, 43, 577584. (4) Wilkins, E.; Atanasov, P. Biosens. Bioelectron. 1995, 10, 485-494. (5) Pickup, J. C.; Thevenot, D. R. In Advances in biosensors, supplement 1; JAI Press: London, UK, 1993, 273-288. (6) Schultz, J. S.; Mansouri, S.; Goldstein, I. J. Diabetes Care 1982, 5, 245253. (7) Mansouri, S.; Schultz, J. S. Biotechnology 1984, 2, 885-890. (8) Meadows, D. L.; Schultz, J. S. Anal. Chim. Acta 1993, 280, 21-30. (9) Ballerstadt, R.; Schultz, J. S. Anal. Chim. Acta 1997, 345, 203-212. (10) Sumner, J. B.; Howell, S. F. J. Bacteriol. 1936, 32, 227-237. (11) Brumfield, A.; Ballerstadt, R.; Schultz, J. S.; Schultz, J. S. Fifth World congress on Biosensors; Berlin, Germany, 1998; Elsevier: Amsterdam, 1998; p 48.
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000 4185
element. Glucose monitoring can be achieved continuously by illuminating the region where the dialysis chamber is located. The fraction of emitted fluorescence light returning back through the skin is captured and measured externally by means of a sensitive photocell or photomultiplier. However, the smallness of the sensor (about 300 nanoliters) in conjunction with some loss of light due to scattering and light absorption of the skin, requires a sensor device that yields a substantial change of fluorescence in response to glucose. The novel approach described here is based on the idea of shielding the fluorochrome labeled Con A when it resides within the porous beads due to binding to the glucose residues within the Sephadex. The beads (selected with a porosity that allows the penetration of Con A with a molecular weight of 100 kDa) are made up of a dye-colored dextran matrix, which prevents the excitation light from impinging in the beads (see Figure 1, left picture). The excitation light cannot effectively penetrate through the beads due to the fact that the absorption spectrum of the dyes overlaps with the fluorescence excitation and emission spectra of the fluorochrome. Yet in the presence of glucose, the fluorochrome Con A conjugate is displaced from the glucose residues located inside the dextran matrix, by mass action, and diffuses out of the colored beads into the accessible void space where it is exposed to illumination, thus resulting in a fluorescence increase (Figure 1, right picture). Utilizing beads with a very high binding capacity for fluorochrome-labeled Con A ensures that the absolute and the potential increase of fluorescence is large. The simplified equations that describe this binding process are shown below:
Fluo-Con A + Glucose Matrix a Fluo-Con A/Glucose Matrix (1) Fluo-Con A/Glucose Matrix + Glucose a Fluo-Con A/Glucose + Glucose Matrix (2)
For this study, the fluorochrome Alexa 488 was substituted for fluorescein (that we used previously) to label Con A because of better photostability. Dye-colored Sephadex beads were used as the matrix containing immobilized pendant glucose residues. Both components were encapsulated in a hollow dialysis fiber, and the glucose-induced fluorescence change of the fabricated sensor was studied. Sensor experiments reported here were targeted at the optimization of signal strength, signal stability, and glucose sensitivity range. EXPERIMENTAL SECTION Chemicals and Materials. Divinyl sulfone (DVS), FITClabeled Con-A, Rhodamine-labeled dextran (2000 kDa), and Sephadex G 200 and G150 (20-50 µm) were purchased from Sigma (St.Louis, MO). The dyes Safranin O and Pararosanilin were purchased from Aldrich Chemical Co., Inc. (Milwaukee, WI). Alexa488-Con A was purchased from Molecular Probes (Eugene, OR). All experiments were done in 12 mM phosphate-buffered saline solution (pH 7.1, 0.9% NaCl, 0.1% NaN3). Dialysis hollow fibers were from Kunstseidewerk (Pirna, Germany). Dyeing of Sephadex Beads. For the dyeing procedure, the DVS method previously described by Porath et al.12 was applied. (12) Porath, J.; Laas, T.; Janson, J.-C. J. Chromatogr. 1975, 103, 49-62.
4186 Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
Figure 1. Schematics illustrating the principles of the fluorescence affinity hollow fiber sensor. In the absence of glucose, fluorochromelabeled Concanavalin A is bound to fixed glucose residues inside porous beads (left hand). The beads are colored with dyes that prevent the excitation light from penetrating into them and inducing Con A to fluoresce, thus keeping the fluorescence emission at 520 nm. After diffusion of glucose through the hollow fiber membrane (molecular weight cutoff, 10 kDa), Con A is displaced from the beads and diffuses out of them, and hereby fluorochrome-labeled Con A becomes exposed to excitation light resulting in a strong increase in fluorescence (right hand).
Sephadex (G150, G200, bead diameter 20-50 µm, 250 mg) were preswollen in 20 mL of distilled water overnight. The beads were washed over a sieve with several volumes of distilled water. The bead suspension (12 mL) was then mixed with 12 mL of a 1 M sodium carbonate buffer solution (Na2CO3, pH 11.4) in a beaker. The suspension was intensively stirred on a magnetic stirrer throughout the procedure. DVS (800 uL) was added to the suspension and allowed to react for 1 h. The beads were washed over a sieve with copious amounts of distilled water to remove nonbound DVS and subsequently equilibrated with 0.5 M sodium hydrogen carbonate buffer (NaHCO3, pH 11.4). Safranin O and Pararosanilin (each 30 mg) were dissolved in DMSO (1 mL). This solution was then slowly added to the stirred suspension, and the reaction was allowed to proceed overnight. Then glycine (1 g) was introduced into the mixture to neutralize remaining active DVS groups. After 1 h, the beads were transferred into a 15-mL plastic vial and centrifuged in order to remove nonbound dye molecules. The supernatant was discarded. The beads were resuspended in DMSO, shaken, and centrifuged again. This procedure was repeated several times until the supernatant was color-free. The bead suspension was then equilibrated with PBS and stored in the refrigerator at 4 °C. Determination of Dye Content in Sephadex beads. A small volume of dye-colored Sephadex bead suspension (0.3 mL) was equilibrated with 1 mL of 1 M HCl and incubated for 48 h at 80 °C. The colored solution was then diluted, and the optical density was measured at 520 nm using a spectrophotometer (Turner, model 340, Barnstead, Indiana). The dye content was expressed as molar equivalents of Safranin O per liter of bead suspension.
The extinction coefficient of 5 × 104 M-1 cm-1 for Safranin dissolved in PBS was used. Purification of Alexa488-Con A. Prior to its use, Alexa488Con A was purified by affinity chromatography on Sephadex G150 and subsequently dialyzed twice against 4 L of PBS. The purified lectin was stored in darkness in the refrigerator. Manufacturing the Hollow Fiber Sensor. A small volume of dyed bead suspension (100 µL) was mixed with 1-5 mg of Alexa488-Con A in 100 microliters. After the binding equilibrium was reached, the tube was centrifuged and the supernatant discarded. Hollow fibers were cut to a length of 3-5 cm and glued into the tip of a 10-µL pipet tip with cyanacrylate-based adhesive (Loctite 410, Rocky Hill, CT). This setup served as the loading device. The particle suspension was aspirated into a 100-µL pipet tip by means of an adjustable pipet. The tip was pushed firmly into the opening of the tip of the loading device. The thumb knob of the pipet was slowly turned down, pushing the bead suspension slowly into the loading device until the suspension/air meniscus hit the opening of the hollow fiber. The suspension was slightly forced into the hollow fiber by alternately depressing and releasing the thumb knob of the pipet that produced a net motion of the beads toward the distal end of the hollow fiber. Sections of the hollow fiber that were evenly packed with beads were cut to a length of 1 cm and quickly sealed with adhesive. The prepared fiber was stored in PBS in the refrigerator. Manufacturing a Homogeneous Hollow Fiber Sensor. This procedure was reported in a previous paper.9 Solutions containing concentrations of Alexa488-labeled Con A (0.3 mg/mL) and 2000 kDA Rhodamine-labeled dextran 2000 (0.15 mg/mL) were prepared in 1.5-mL plastic tubes. Hollow fibers were cut in segments to a length of approximately 2 cm and put into this solution. When the solution had completely filled the whole length of that segment by aspiration, the upper end was closed with cyanacrylate-based adhesive. The whole fiber was quickly removed from the tube and cut to a length of 1 cm, and the other end was sealed. The fabricated hollow fiber segment was stored in PBS. Stability Study. A single sensor fiber was glued on a white plastic material and introduced into a translucent tubing (length, 6 cm) whose diameter (0.4 cm) was slightly narrower than the width of the plastic base material. For fluorescence measurements, the sensor assembly was connected with two tubes, of which one was linked to a glucose stock solution and the other to a waste beaker. Flow of solution through the sensor assembly was generated by gravity, and clamps were used to control the flow rate. To measure fluorescence, the head of the optical fiber bundle linked to the fluorescence spectrophotometer was positioned above the sensor assembly with the optical path length of 2 mm. At this distance, the out-coming excitation light covered the full width of the sensor fiber (optical fiber bundle head, 0.4 cm in diameter). During the fluorescence measurements, the entire setup was covered with black felt to eliminate interference by stray light. The sensor fluorescence change in response to freshly prepared solutions of 0 and 20 mM glucose was monitored at various intervals. Between measurements the sensor assemblies were stored in glucose-free PBS at ambient light and room temperature. The 20 mM glucose-induced fluorescence response was expressed as the percentage increase of fluorescence by
Figure 2. Spectra of fluorescence and absorption of the various chromophoric components of the bead-based affinity sensor. (- - -) Excitation spectrum of fluorescein, (‚‚‚) emission spectrum of fluorescein, (s) absorption spectrum of dye-labeled Sephadex beads. For measuring the absorption spectrum of the dye-colored beads, the beads were incubated in 1 M HCl at 60 °C for 4 h and the optical density of the supernatant at different wavelengths was measured.
glucose from the baseline. The initial measurement of a glucosefree solution at the first day was set as 100%, and the subsequent measuring points were normalized. Fluorescence Measurements. Fluorescence measurements were done with the fluorescence spectrophotometer (LS50B, Perkin-Elmer, Beaconsfield, UK). For measuring the fluorescence of Alexa488-labeled Con A, the excitation and emission wavelengths of the spectrophotometer were set at 495 and 520 nm, respectively. Single sensor fibers were introduced into a throughflow quartz cuvette of the fluorescence spectrophotometer and its fluorescence response to glucose was measured. For remote fluorescence measurements, the fiber-optic assembly (PerkinElmer part no. L225 0137) comprising a bundled fiber cable (length 1 m) was used. Fluorescence Microscopy. A single hollow fiber filled with dye-labeled Sephadex G150 beads and Alexa488-Con A was positioned inside a translucent perfusion chamber that was mounted on the moving stage of a microscope (Axiovert 35, Carl Zeiss, Inc., West Germany). At different times, solutions with (50 mM) and without glucose were alternately flushed through the perfusion chamber, and images of the fluorescence response of the sensor were taken at 1-min intervals. A CCD camera (model Xc-77, Hamamatsu Corp., Bridgewater, NJ), controlled by the camera control module C2400 (Hamamatsu), was used for image taking. The software package Oncor-Image 2.05 that ran on a Macintosh computer served to store and analyze the images. RESULTS Characterization and Properties of Dyed Beads. Various Sephadex bead preparations (G150, G200, diameter 20-50 µm) were intensely colored using two dyes, Safranin O and Pararosaniline. The dyes were chemically attached to the beads via divinyl sulfone (DVS), a bis functional linker, which preferentially binds to amino, hydroxyl, and sulfur groups. The dyed beads showed a broad absorption shoulder at the wavelength of 540 nm (see Figure 2). This ensured that the excitation and emission light of Alexa488 (at 495 and 520 nm, respectively) was efficiently absorbed when Alexa488-Con A was bound inside the beads. The fluorochrome Alexa488 has been chosen as a substitute for the Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
4187
Figure 3. Light microscopy picture of a small section of the hollow fiber that was filled with dye-colored Sephadex G150 beads. A bead fraction having a bead diameter of less than 25 µm was used that was obtained by sieving the original material (mesh size of 25 µm).
more commonly used fluorescein because it was shown to be less prone to photo bleaching and quenching due to high fluorochrome concentrations (inner filter effect). The experiments revealed that the color intensity of the beads is a function of the DVS concentration and reaction time. At 10 mg DVS per milligram dry weight of beads and after 20 h of reaction, the colored beads looked red-violet. The Safranin O dye content of the beads was 0.20-0.40 millimoles per liter of bead suspension. Figure 3 shows a light microscopy picture of a hollow fiber segment. As shown, the inner lumen of the fiber is evenly filled up with the amorphouslooking beads. The occupied volume of the beads within the fiber was dependent on the efficacy of the filling procedure. Sensor fibers, which were filled with beads occupying more than 50% of its volume, were found to show a strong and rapid sensor signal. Note that the volume ratio of beads to accessible free space should be kept in adequate proportion to keep the pathway for Con A diffusion short (preferentially less than the bead diameter of 2050 µm), to obtain a fast response time. Fluorescence Microscopy of a Sensor Fiber. A time series of the fluorescence response of a magnified section of the hollow fiber sensor after introduction and removal of 50 mM glucose is displayed in Figure 4. Photo A shows the sensor fiber in the absence of glucose. Only contours of the hollow fiber membrane (see gray line on top and bottom) and an amorphous mass consisting of light and dark spots (presumably colored beads) are distinguishable. The lighter areas may be due to unbound Alexa488-Con A and/or Con A which is bound to the periphery of the beads. Photo A illustrates impressively the highly efficient absorption of the excitation light by the colored beads, which prevented hidden Alexa488-Con A from emitting fluorescence. The introduction of glucose (50 mM) led to a strong increase in fluorescence within the next 4 min as it is shown in pictures A-D. Note the strong increase in brightness. This can be explained by the glucose-induced displacement of Alexa488-Con A from their binding sites inside the beads and subsequent diffusion into the external fluid. The appearance of Alexa488-Con A in free solution allows the excitation of the fluorophore and detection by the 4188
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
optical system. Reversibility of the sensor system is demonstrated with the perfusion with a glucose-free solution. There is a subsequent disappearance of the fluorescence from the interstitial space as the Alexa488-Con A re-enters the bead. Frames E-H were obtained after a 5-min lag to flush out the system with a plain-water wash. Fluorescence Signal Strength. The fluorescence signal from the bead-based sensor was found to be high in absolute and relative terms when compared with previous configurations we have published (e.g., Meadows and Schultz Anal. Chim. Acta 1993). The high level of fluorescence signal is due in part to the high Con A binding capacity of Sephadex beads that results in a high free solution concentration of Alexa488-Con A when glucose is present to displace the fluorophore from the internal volume of the bead. Furthermore, column chromatography experiments (not shown) indicate that only 50% of the void volume within the bead is available to Con A. Thus, as the bead volume fraction in the sensor element is increased, the volume of liquid available to Con A decreases and the concentration of Con A in this fluid is enhanced slightly, due to this excluded volume effect, shown as the calculated dashed line in Figure 5 displaying a slight upward curvature. Measurements of the fluorescence of the supernatant solution at low-bead volumes indicate that it appears to follow the expected pattern (solid line in Figure 5). Because of the fabrication techniques, the bead volume fraction within the hollow fiber sensors is difficult to determine exactly. We estimate that beads occupy approximately 50-80% of the inner volume. The actual fluorescence response of the sensors when exposed to 100 mM glucose is in the range of 2.5 times the background fluorescence of the sensor fibers. These results are in the range predicted by the dashed line in Figure 5, with the predicted fluorescence increase of 250%. To appreciate the large fluorescence change obtained, the response of this novel fiber sensor was compared with that of a well-described homogeneous sensor system that consisted of fluorescein-Con A and 2000 kDa Rhodamine-dextran confined inside a sealed hollow fiber.8 The fluorescence increase due to fluoroscein of this earlier configuration is based on the glucoseinduced dissociation of both binding ligands, resulting in a spatial separation of fluorescein and rhodamine greater than 50 angstroms. Dissociation leads to a simultaneous fluorescence increase due to inhibition of fluorescence resonance energy transfer (FRET).13 The absolute fluorescence change of this homogeneous sensor system (Alexa488-Con A was used instead of fluorescein in the current comparison) toward 20 mM glucose was nearly 50 times lower in absolute terms than that with the bead-based fluorescence sensor (250 vs 6 arbitrary units). The relative change of the homogeneous system was nearly 15 times lower (100 vs 10%) than our new heterogeneous bead system. Various other fluorescence affinity sensor systems using fluorochrome labeled Con A showed relative fluorescence increases induced by 20 mM glucose of 20,14 30,8 16,15 and 11%.16 Again, these numbers are (13) Fo ¨rster, T. Radiat. Res., Suppl. 1960, 2, 326. (14) Rolinski, O. J.; Birch, D. J. S.; McCartney, L. J.; Pickup, J. C. In Proceedings of Advances in fluorescence sensing technology IV.; Lakowicz, J. R., Soper, S., Thompson, R. B., Eds.; SPIE: San Jose, CA, 1999; pp 6-14. (15) Tolosa, L.; Malak, H.; Rao, G.; Lakowicz, J. R. Sens. Actuators, B 1997, 45, 93-99. (16) Russell, R. J.; Pishko, M. V.; Gefrides, C. C.; McShane, M. J.; Cote, G. L. Anal. Chem. 1999, 71, 3126-3132.
Figure 4. Sequence of images showing the fluorescence response of the sensor fiber after introduction (A-E) and removal (E-H) of 50 mM glucose. The photos were taken at 1-min intervals (see Experimental Section for details).
Figure 5. Plot of relative fluorescence vs volume fraction of beads in solution in the presence of saturating amounts of glucose (100 mM). Various volumes of a suspension containing dye-colored Sephadex beads which were loaded with 1 mg/mL Alexa488-Con A, were filled in 1.5-mL tubes, and were mixed with a solution of 100 mM glucose to a total volume of 60 µL. After an incubation time of 15 min, aliquots of the supernatant (20 µL) were diluted in 400 µL of PBS and the fluorescence was measured. The full line represents the best fit of data points (triplicates, SD ) (3%). The dashed line represents extrapolation toward a higher volume fraction. Triangles represent the relative fluorescence response of three different hollow fiber sensors after reaction with 100 mM glucose.
significantly lower than that obtained with the herein-described bead-based fluorescence sensor. However, for practical application purposes, the absolute increase in fluorescence is more important than the relative increase. This parameter is predominately determined by the maximally obtainable concentration of the fluorescence-emitting binding ligand. It was found that the absolute fluorescence signal increases with rising concentrations of Alexa488-Con A. We found that at medium to high Con A loading of the beads (50-80%), the largest fluorescence increase due to 20 mM glucose challenge was obtained. This range of loading translates into a Con A concentration of around 10-15 mg mL-1 of bead suspension. In other sensor systems the maximally allowable concentration of the fluorescence-emitting binding ligand can be limited by various factors, such as unspecific fluorescence quenching due to an inner filter effect, precipitation of Con A/dextran complexes, or others. The precipitation phenomenon was repeatedly reported for the homogeneous sensor systems.8,9 In those studies, very low Con
A concentrations had to be used (less than 200 µg mL-1 FITCCon A). Schultz and co-workers studied the response of a heterogeneous fiber-optic sensor that was based on the displacement of fluorescein-labeled dextran (70 kDa) from Con A that was immobilized at the inner wall of a hollow fiber.6,7 This sensor generated a relative fluorescence signal increase of approximately 200-300% at 30 mM glucose, which was accomplished with 0.3 µM of mobile fluorescein-dextran, the component responsible for fluorescence emission after dissociation from membrane-bound Con A. The relative fluorescence signal is comparable to results obtained in this paper. Yet one has to consider the extremely low concentration of fluorescein-dextran (0.3 µM) used in the latter sensor variant and compare this to the that of Alexa488-Con A used in the herein-described system that was much higher (120 µM). This is due to the high number of fixed glucose-binding sites provided by the dextran beads whose capacity equals approximately 2.5 µmoles Con A mL-1of bead suspension. In contrast, the fiber-optic sensor by Schultz et al.6,7 provided a much lower concentration of available Con A binding sites (approximately 0.1 µmole mL-1), based on a surface coverage of membranebound Con A which was reported to be 21 µg cm-2.17 Response Time. The rates of chemical reaction between Con A and glucose residues are very rapid, with time constants in the order of milliseconds. Thus, the quantitative time-response curve of a sensor fiber shown in Figure 6 with an average response time of around 4 to 5 min is most likely determined by diffusional processes. The response time is mainly affected by diffusion of Alexa488-Con A into and out of the beads. Diffusion of Con A through G200 particles with an average bead radius of 15 µm takes approximately 1 min, according to our own measurements (not shown) and published data.18 This would give an apparent diffusion coefficient (D*) for Con A of approximately 0.4 × 10-7 cm2 s-1. The particles used in the sensor were somewhat larger, causing an increase in the response time, and transport through the dialysis membrane also slows the response as well. Glucose Detection Range. It is noteworthy to mention the broad glucose sensitivity range of this sensor extends to 100 mM (see also Figure 6). The physiological range of blood sugar varies (17) Srinivasan, K. R.; Mansouri, S.; Schultz, J. S. Biotech. Bioeng. 1986, 28, 233-239. (18) Ballerstadt, R.; Ehwald, R. Biosens. Bioelectron. 1994, 9, 557-567.
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
4189
Figure 6. The time response of a hollow fiber sensor filled with colored Sephadex G200 beads. The fiber was positioned inside a flow-through quartz cuvette of the fluorescence spectrophotometer. The fluorescence is given in arbitrary units. The arrows indicate the change in glucose concentration. The insert shows the plot of fluorescence vs glucose concentrations, which were derived from the data obtained from the perfusion experiment.
Figure 8. Normalized fluorescence response of four different hollow fiber sensors over several weeks. Graph in (a) is the baseline fluorescence in the absence of glucose. The dashed line indicates the best fit of sensor signal progression. (b) The relative fluorescence response toward 20 mM is displayed.
Figure 7. Plot of glucose concentration vs fluorescence response. Three measurements of a single sensor fiber were conducted. Note the remarkably low detection sensitivity of less than 0.5 mM glucose with good precision.
from approximately 2 to 30 mM. Within this concentration range the fluorescence sensor signal featured the highest dynamic change which yielded more than 100% increase from 0 to 25 mM glucose. Because of the strong fluorescence signal of the sensor, the system was challenged to measure extremely low glucose concentrations of less than 1 mM. The plot of glucose concentration vs fluorescence, that is displayed in Figure 7, shows the possibility of measuring down to 0.15 mM of glucose, with a detectable fluorescence response and reasonable accuracy (SD less than 1%). The feature is thought to be very valuable for measuring hypoglycemic blood-sugar levels in diabetic patients. Stability. The fluorescence responses of four sensor fibers were studied under ambient light and temperature conditions over a period of 3 months. These sensors were stored under glucosefree conditions (Figure 8a) and periodically exposed to 20 mM of glucose (Figure 8b). Of course, under glucose-free conditions, most of the Alexa488-Con A is within the bead so that the observed variation in Figure 8a of about 30% is at low fluorescence intensity levels. When the sensors were challenged with glucose, the observed fluorescence level increased about 2.5-fold initially, but then in three of the four sensors there was a 40-50% drop in intensity over the 2-3 month period. A number of possible factors may play a role herein, such as hydrolysis of chemical linkage between Con A and Alexa488, denaturation of Con A, photobleaching of dye, irreversible binding of Con A to the bead, 4190
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
declining integrity of the hollow fiber membrane and sealant. This issue will be looked at more closely in the future. Even though the cause of the decrease in signal has not been determined, as yet, we believe that the importance of these results is that there is a significant response of the sensors even after several months, thus providing some confidence that long-term implantable sensors could be developed by improvements in this technology. Mathematical Model for Sensor Response. To formulate a mathematical model for the sensor behavior as a function of Con A, Sephadex, and glucose concentrations, a series of experiments were undertaken to measure the extent of binding of Con A to Sephadex in the absence of glucose. Previous experiments have shown that the binding of Con A to dextran is not a simple monomolecular reaction.19 The binding behavior of Con A was found to be dependent on the molecular weight of dextran. For low molecular weight dextrans (about 3000), the binding was monomolecular in nature. At higher molecular weights, the binding was di- or trivalent. For these circumstances, there is no simple equilibrium expression to describe the binding behavior. Thus, we decided to use the following empirical power law expression to correlate the binding of ConA to Sephadex [CD]
[CD] ) A[Cf]m[VSephadex]n
(3)
[Cf] and [VSephadex] denote the free Con A and the volume fraction of Sephadex beads, respectively. Table 1 shows the values (19) Mansouri, S. Optical glucose sensor based on affinity binding. Ph.D. Thesis, The University of Michigan, 1983.
Table 1. Experimental Data Used for Developing the Empirical Model total vol (Vt)
total Con A added (10-6 M)
free Con A measured (10-6 M)
Sephadex vol (µL)
65 65 65 65 65 65 65 65 65 65 65 65 65 65
1 2 4 8 13.3 150 75 3 1.5 0.3 150 75 3 1.5
0.1 0.445 1.16 2.87 5.78 34 14.8 0.74 0.34 0.1 106 35 1.9 0.8
10 10 10 10 10 20 20 20 20 20 7.5 7.5 7.5 7.5
of free and bound Con A concentrations at various volumes of Sephadex suspension. The data from these experiments were used to determine the constants A, n, and m by a nonlinear least-squares fitting routine. The best estimates were
A ) 78.55 Figure 9. Model glucose calibration curves of various hypothetical sensor fibers at various volume fractions of Sephadex with Con A content of 25 mg (a) and 75 mg (b). Total volume 10 mL. Equilibrium constant, Kglucose 25 103 µM (see text for details).
m ) 0.745 n ) 1.73 It is assumed that the equilibrium binding of glucose to Con A [CG] is given by
[CG] )
[Cf][G] KGlucose
(4)
Using the empirical model for the binding of ConA to Sephadex (eq 3), we can estimate the response of various sensor formulations with various contents of Con A and Sephadex. First we note that a material balance on all forms of ConA [Ct] within the sensor can be written as:
[Ct] ) [Cf] + [CG] + [CD]
(5)
The fluorescence response of the sensor would be given by all forms of Con A outside of Sephadex beads, i.e., the sum of [Cf] and [CG]. [CD] is calculated from eq 3, and [CG] is calculated from eq 4, i.e.
[Ct] ) [Cf] +
[Cf][G] + A[Cf]m[VSephadex]n KGlucose
higher Sephadex volume fractions. In addition, the sensor would have a higher fluorescence response due to the higher content of Con A.
(6)
Figure 9 shows hypothetical response curves for assumed values of [G] (free glucose concentration). As can be seen in Figure 9a, increasing the volume fraction of Sephadex at a constant Con A content increases the dynamic range of the sensor and provides a more linear calibration curve as well. Figure 9b displays the result of a similar series of calculations but at higher (3 times) Con A content. Again a better sensor response is obtained at
CONCLUSIONS AND OUTLOOK This study was intended to optimize the fluorescence signal of a novel variant of a hollow fiber affinity sensor for transdermal fluorescence-based glucose monitoring. An essential prerequisite for that scheme to be successful is, among other things, to ensure that the glucose-induced fluorescence emission is sufficiently high to be measurable through skin tissue. Thus, we focused on producing the highest absolute fluorescence from the smallest volume sensor. First, keeping the sensor volume small should make the surgical procedure for implantation minimally invasive. And second, the possibility of using off-the-shelf optical components which are usually less sensitive than specialized research equipment that is usually cumbersome to carry and expensive. The main achievement of this study is the large absolute glucoseinduced signal change that was mainly accomplished by two factors. One is the utilization of Sephadex beads that provided a high number of glucose residues for Alexa488-Con A binding. The other factor was the efficient dyeing procedure of those beads with chromophores, which prevented the incoming excitation light from penetrating into the bead and generating fluorescence emission of the bound Alexa488-Con A molecules due to highly efficient light absorption. Note that skin tissue shows complex optical behavior in the wavelength range from 450 to 600 nm due to autofluorescence and light absorption.20 This poses no signifi(20) Richards-Kortum, R.; Rava, R. P.; Fitzmaurice, M.; Tong, L. T.; Ratliff, N. B.; Kramer, J. R. IEEE Trans. Biomed. Eng. 1989, 36, 1222-1232.
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
4191
cant problem for this type of sensor, since it requires only selection or development of near-infrared dyes and fluorochromes. In fact, we have obtained preliminary results with a sensor using alkali blue 6b for dyeing the beads and Cy5 (emission wavelength 670 nm) for labeling Con A. To sum it up, the described hollow fiber sensor system excels through the combination of all its features: its minute size (0.5 × 0.2 mm), a strong dynamic signal change in the relevant glucose level range (0 to 30 mM), a reasonably fast response time (4-5
4192
Analytical Chemistry, Vol. 72, No. 17, September 1, 2000
min) for increasing and decreasing glucose levels, and a potential for long-term performance. ACKNOWLEDGMENT This work was supported by a grant from the Juvenile Diabetes Foundation (JDF). Received for review February 18, 2000. Accepted June 16, 2000. AC000215R