Anal. Chem. 2004, 76, 1403-1410
Dual-Labeled Glucose Binding Protein for Ratiometric Measurements of Glucose Xudong Ge, Leah Tolosa,* and Govind Rao*
Department of Chemical and Biochemical Engineering, University of Maryland, Baltimore County, 1000 Hilltop Circle, Baltimore, Maryland 21250
Highly sensitive glucose monitoring has potential applications in conditions where the glucose levels are below the detection limit of currently available technology. Examples include bioprocess monitoring of bacterial cultures and measurement of minute amounts of human interstitial fluid extracted by iontophoresis. Here we describe a ratiometric glucose sensor capable of measuring micromolar levels of glucose. This sensor is based on an E. coli glucose binding protein (GBP) labeled with two fluorophores. The L255C mutant of GBP was labeled with the environment-sensitive fluorophore, acrylodan, at the cysteine mutation and a long-lived metal ligand complex of ruthenium (ruthenium bis(2,2′-bipyridyl)-1, 10-phenanthroline-9-isothiocyanate) at the N-terminal. The acrylodan emission is quenched in the presence of glucose while the ruthenium emission remained constant, thereby serving as a reference. The sensitivity of the sensor is in the micromolar range (Kd ) 0.4-1.4 µM). Thus, it is possible to measure glucose concentrations at micromolar levels and higher (with dilution). Calculations of the fluorescence energy-transfer efficiency between acrylodan and ruthenium gave an approximate distance of 25 Å between the two fluorophores, consistent with X-ray crystallographic data. The effect of temperature on glucose binding was measured and analyzed. Maximum signal changes and apparent binding constants increase with temperature. The enthalpy change for glucose binding as calculated from the apparent binding constants is ∼43.1 kJ/mol. In addition to ratiometric measurements, the presence of the long-lived ruthenium metal ligand complex allows for low-cost modulation-based sensing. Glucose is the major carbon and energy source in cellular metabolism. The lack of glucose in the medium will severely limit cell growth and product yield in industrial bioprocess applications. However, excessive glucose can also be detrimental, leading to lactate formation via the glycolytic pathway.1 Therefore, glucose monitoring and control is important for the healthy growth of cells and maximum product formation in bioprocesses. Additionally, glucose levels in blood and plasma are used as a clinical indicator of diabetes. Diabetes is a chronic disease characterized by the * Corresponding authors. Tel.: 410-455-3432; Fax: 410-455-6500. E-mail:
[email protected] or
[email protected]. (1) Garnier, A.; Cote, J.; Nadeau, I.; Kamen, A.; Massie, B. Cytotechnology 1994, 15, 145-155. 10.1021/ac035063p CCC: $27.50 Published on Web 01/16/2004
© 2004 American Chemical Society
body’s inability to produce or utilize insulin, thereby leading to uncontrolled glucose levels in blood and tissues. This disease affects 6.2% of the total population and 20.1% of those aged 65 years or older in the United States.2 Thus, the cost for the control of this disease and the prevention of its complications can be as high as $132 billion dollars annually.2 For this reason, there is continued interest in new and improved methods of glucose monitoring for clinical purposes. To date, a number of different glucose sensors have been reported in the literature. Most of these sensors utilize glucose oxidase as the biological element.3-12 When glucose is present, the enzyme converts glucose into gluconolactone and hydrogen peroxide, which is then detected at an electrode. As the product of the enzyme reaction is hydrogen peroxide, which tends to degrade the enzyme itself over time, these devices usually suffer from drift. Additionally, these sensors are prone to interferences from other electroactive species, such as ascorbate and ureate. Furthermore, glucose oxidase-based sensors are insensitive to concentrations in the submillimolar range. Another protein that has been extensively investigated as a possible glucose sensor is concanavalin A (ConA).13-19 This sensor (2) National Institute of Diabetes and Digestive and Kidney Diseases. National diabetes statistics fact sheet: general information and national estimates on diabetes in the United States; NIH Publication 03-3892, 2000. (3) Chen, T.; Friedman, K. A.; Lei, I.; Heller, A. Anal. Chem. 2000, 72, 37573763. (4) Couto, C. M. C. M.; Arau´jo, A. N.; Montenegro, M. C. B. S. M.; Rohwedder, J.; Raimundo, I.; Pasquini, C. Talanta 2002, 56, 997-1003. (5) Gouda, M. D.; Kumar, M. A.; Thakur, M. S.; Karanth, N. G. Biosens. Bioelectron. 2002, 7, 503-507. (6) Haouz, A.; Stieg, S. Enzyme. Microb. Technol. 2002, 30, 129-133. (7) Male, K. B.; Gartu, P. O.; Kamen, A. A.; Luong, J. H. T. Biotechnol.. Bioeng. 1997, 55 (3), 497-504. (8) Moscone, D.; Pasini, M. Talanta 1992, 39 (8), 1039-1044. (9) Narang, U.; Prasad, P. N.; Bright, F. V. Anal. Chem. 1994, 66, 3139-3144. (10) Sapre, A. G.; Bedekar, A.; Deshpande, A. V.; Lali, A. M. Biotechnol. Lett. 2000, 22, 569-573. (11) Tatsu, Y.; Yamashita, K.; Yamaguchi, M.; Yamamura, S.; Yamamoto, H.; Yoshikawa, S. Chem. Lett. 1992, 1615-1618. (12) Wang, B.; Li, B.; Deng, Q.; Dong, S. Anal. Chem. 1998, 70, 3170-3174. (13) Ballerstadt, R.; Schultz, J. S. Anal. Chem. 2000, 72, 4185-4192. (14) Meadows, D. L.; Schultz, J. S. Anal. Chim. Acta 1993, 280, 21-30. (15) McCartney, L. J.; Pickup, J. C.; Rolinski, O. J.; Birch, D. J. S. Anal. Biochem. 2001, 292 (2), 216-221. (16) Rolinski, O. J.; Birch, D. J. S.; McCartney, L. J.; Pickup, J. C. Spectrochim. Acta, A 2001, 57 (11), 2245-2254. (17) Russell, R. J.; Pishko, M. V. Anal. Chem. 1999, 71, 3126-3132. (18) Tolosa, L.; Malak, H.; Rao, G.; Lakowicz, J. Sens. Actuators, B 1997, 45, 93-99. (19) Tolosa, L.; Szmacinski, H.; Rao, G.; Lakowicz, J. Anal. Biochem. 1997, 250, 102-108.
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004 1403
utilizes the competition between glucose and a sugar-containing macromolecule such as dextran for binding to ConA. The main method of choice for the ConA competitive assay is fluorescence energy transfer (FRET). Unfortunately, ConA is not specific to glucose and tends to agglutinate and precipitate in a few hours. Additionally, this protein, because of its agglutinating properties is considered toxic. These drawbacks greatly reduce its specificity and applicability. The present authors and other researchers have studied the Escherichia coli glucose binding protein (GBP) for the construction of glucose biosensors.20-24 The binding affinity of GBP in terms of dissociation constants is in the micromolar range. This allows for glucose measurements in low-glucose media, such as the Luria-Bertani (LB) broth used in bacterial fermentations. Although important, these measurements are not routinely done because of the lack of a simple technique that is sensitive at submillimolar glucose concentrations. Another application where micromolar sensitivity may be useful is in measuring glucose in extracted human interstitial fluid, particularly by iontophoresis. In our previous studies, a single-cysteine GBP mutant (Q26C) was prepared and labeled with the environment-sensitive, thiolreactive fluorophore, anilinonaphathalenesulfonate (ANS).23 We showed that this labeled protein exhibits all the properties desirable for a reagentless sensor including reversibility of binding, short response and recovery times, and stable activity lasting several months.20 The use of this single-labeled protein in monitoring glucose levels in cell culture and fermentation was successfully demonstrated.20 However, due to the lack of measurable lifetime changes, ANS-labeled GBP can only be used for intensity-based measurements. Intensity-based measurements are easily affected by fluctuations in the intensity of the excitation source, by leaching or photobleaching of the fluorophore, and the positioning of the sample. Thus, to make the measurements more robust, a dual-emitting GBP was prepared by labeling the singlecysteine mutation with acrylodan, this time at position 255, and the N-terminal with the long-lived environment-insensitive ruthenium bis(2,2′-bipyridyl)-1, 10-phenanthroline-9-isothiocyanate (ruthenium). Upon glucose binding, the polarity-sensitive acrylodan is exposed to the solvent, resulting in a decrease in fluorescence intensity. On the other hand, the emission intensity of ruthenium is unaffected by glucose, thereby acting as a reference. In this paper, the preparation and optical properties of this dual-labeled GBP are described and discussed. MATERIALS AND METHODS Materials. 6-Acryloyl-2-dimethylaminonaphthalene (acrylodan) and tris(2-carboxyethyl)phosphine (TCEP) were purchased from Molecular Probes (Eugene, OR). Fucose, sucrose, glucose, DEAE Sephadex A-50, N,N-dimethylformamide (DMF), NaCl, KH2PO4, Na2HPO4, NaH2PO4, and MgCl2 were purchased from SigmaAldrich. Tryptone and yeast extract were obtained from Becton Dickinson (Sparks, MD). All chemicals were used without further (20) Ge, X.; Tolosa, L.; Simpson, J.; Rao, G. Biotechnol. Bioeng. 2003, 84 (6), 723-731. (21) Marvin, J. S.; Hellinga, H. W. J. Am. Chem. Soc. 1998, 120, 7-11. (22) Salins, L. L. E.; Ware, R. A.; Ensor, C. M.; Daunert, S. Anal. Chem. 2001, 294, 19-26. (23) Tolosa, L.; Gryczynski, I.; Eichhorn, L. R.; Dattelbaum, J. D.; Castellano, F. N.; Rao, G.; Lakowicz, J. R. Anal. Biochem. 1999, 267, 114-120. (24) Ye, K.; Schultz, J. S. Anal. Chem. 2003, 75, 3451-3459.
1404
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
purification. Slide-A-lyzer dialysis cassettes were purchased from Pierce (Rockford, IL). SstI, PstI restriction enzymes were purchased from Invitrogen Life Technologies. The Quick-Change mutagenesis kit was obtained from Stratagene (Cedar Creek, TX). Construction of the Plasmid. The plasmid JL01 encoding for the wild-type GBP was used as the template for the construction of the L255C GBP mutant. The single-cysteine mutation at position 255 was accomplished using the Quick-Change mutagenesis kit from Stratagene. The 5′ primer used for the mutagenesis was GCACTGGCGGGCAC CGTATGCAACGATGCTAACAACC. The 3′ primer was GGTTGTTAGCATCGTT GCATACGGTGCCCGCCAGTGC. Both primers had the desired mutation (underlined). After the PCR, the product was treated with DpnI restriction enzyme to digest the parental supercoiled dsDNA. The mutated plasmid was then transformed into XL-1 Blue supercompetent cells and spread on LB plates with ampicillin. Colonies appeared in 16 h. Four colonies on each plate were selected for making 5-mL overnight cultures, and the plasmids were then extracted using QIAprep Spin Miniprep Kit from Qiagen. The DNA gel (not shown) confirmed the existence of the desired restriction sites: SstI and PstI. All colonies, except one have the insert of correct length. The DNA sequencing data (not shown) verified the presence of the desired mutation (Biopolymer Core Facility, University of Maryland, Baltimore, MD). Protein Expression and Purification. The procedure for the expression, release, and purification of the L255C mutant is similar to that described previously23 except for a few modifications. First, 20 mL of LB medium in a 50-mL tube was inoculated with a single colony on the plate and incubated at 37 °C with shaking at 260 rpm for 8-10 h. The seed culture obtained above was then used to inoculate 500 mL of LB medium supplemented with 1 mM fucose in a 1000-mL shake flask. The culture was then incubated at 37 °C with shaking at 260 rpm for 8 h. To release the periplasmic proteins, the cells in the above culture were first harvested by spinning for 5 min at 13000g and then washed twice with 40 mL of 10 mM Tris-HCl, 30 mM NaCl, pH 7.5. The cells were collected by centrifugation at 13000g for 5 min and resuspended in 40 mL of 33 mM Tris-HCl, pH 7.5. After mixing with 40 mL of 40% sucrose, 0.1 mM EDTA, 33 mM trisHCl, pH 7.5, the suspension is then left at room temperature for 10 min with very slow shaking, followed by centrifugation at 13000g for 5 min. The periplasmic proteins were released by adding 10 mL of ice-cold 0.5 mM MgCl2 and shaking vigorously in an ice bath for 10 min. Finally, the suspension was centrifuged at 13000g for 10 min. The total concentration of the periplasmic proteins was determined using Micro Protein Determination (Sigma Diagnostics, Inc., Louis, MO). A 10-fold excess of TCEP was added to the protein solutions to prevent the oxidation of the thiol groups. The GBP content in the supernatant was estimated by SDS-PAGE to be ∼50% (Figure 1). The crude protein products were then purified on a DEAE Sephadex A-50 column.25 Fluorophore Coupling. The L255C GBP mutant has a singlecysteine mutation at position 255, which allows for the specific labeling of an environmentally sensitive, thiol-reactive fluorophore. To label the protein with such a fluorophore, acrylodan, a 10-fold excess of this dye in DMF was added dropwise to the protein solution, and the conjugation was allowed to occur for 4 h at room (25) Boos, W.; Gordon, A. S. J. Biol. Chem. 1971, 246 (3), 621-628.
Figure 1. SDS-PAGE assay of the expression of GBP in E. coli NM303.
temperature. The removal of unreacted dye was performed by dialyzing overnight in Slide-A-lyzer dialysis cassettes. For labeling of the second fluorophore, a 10-fold excess of ruthenium bis(2,2′bipyridyl)-1, 10-phenanthroline-9-isothiocyanate in DMF was added to the single-labeled protein solution and the mixture was then allowed to sit for 4 h at room temperature. The N-terminal of the protein was selectively labeled by maintaining the pH at 7.5. The unreacted dye was removed by applying the protein solution to a DEAE Sephadex A-50 column.25 The protein-dye conjugate was then eluted from the column using increasing amounts of NaCl (0-0.2 M) in phosphate buffer. The fractions were analyzed by SDS-PAGE, and the fractions containing the desired protein were collected. The total protein concentration was determined, and the labeling efficiency was estimated from the total protein concentration and the concentration of protein-bound fluorophore, which was calculated from its absorbance and extinction coefficient.26 The final product was 0.2-µm filter-sterilized and stored at 4 °C. Fluorescence Measurements. Steady-state fluorescence spectra were recorded on a Varian Cary Eclipse fluorescence spectrophotometer (Varian Instruments, Walnut Creek, CA). Absorbance spectra were recorded on a Hewlett-Packard 8452A diode array spectrophotometer. Frequency-domain intensity decays were measured on an ISS-Koala fluorometer (Champaign, IL) with several modifications. An ultraviolet LED NSHU550E (Nichia America Corp., Lancaster, PA) with a peak wavelength at 375 nm driven by a current source was used as the excitation source. The modulation voltage was applied through bias T.24. The standard radio frequency amplifier for the photomultiplier tubes was replaced with a ZHL-6A (Mini-circuits, Brooklyn, NY) to enhance the low-frequency performance. The excitation light was filtered by 400, 450, 550, and 650 FL07 short-wave pass filters (Andover Corp., Salem, NH). The emission light was filtered by a 450FH90 (26) Haughland, R. P. Handbook of Fluorescent Probes and Research Chemicals; Molecular Probes: Eugene, OR, 1996.
Figure 2. Absorbance spectra of GBP-Acr, Ru-GBP-Acr, ruthenium alone, and GBP-Acr + ruthenium. The total protein concentration is 3.2 µM; the labeling extents of acrylodan and ruthenium are 72 and 94%, respectively.
long-wave pass filter (Andover Corp.). All samples were measured in quartz cuvettes (Starna Cells, Inc. Atascadero, CA). RESULTS Although the mutated plasmid was first transformed into XL-1 Blue supercompetent cells after the mutation was accomplished, this host did not express the new mutant protein very efficiently. To find the best host for protein expression, the mutated plasmid was transformed into several other E. coli strains including NM303, HB101, and JM109. All these cell lines expressed the mutant with comparable efficiency. However, NM303 does not produce the wild-type GBP. Thus, NM303 was chosen as the host for the expression of the L255C mutant. Additionally, unlike the Q26C mutant, the addition of fucose proved to be very critical to the expression of L255C. Without the addition of 1 mM fucose, the L255C mutant was not efficiently expressed. Figure 1 shows the SDS-PAGE gel of the crude periplasmic extract containing L255C GBP. The absorbance spectra of GBP-Acr, Ru-GBP-Acr, Ru alone, and GBP-Acr + Ru are shown in Figure 2. The total protein concentration in all samples is 3.2 µM. The calculated labeling extent of protein-bound acrylodan based on an extinction coefficient26 of 20 000 cm-1 M-1 is ∼72%. The calculated labeling extent of ruthenium based on an extinction coefficient27 of 15 000 cm-1 M-1 is 94%. Figure 3 shows the excitation spectra of GBP-Acr and RuGBP-Acr, respectively. It can be seen that the excitation spectra monitored at the 510-nm emission exhibits a peak at ∼380 nm in both single- and dual-labeled GBP. However, the excitation spectra monitored at 610 nm clearly show a peak at ∼460 nm in the duallabeled GBP, an indication of the presence of ruthenium. The emission spectra of GBP-Acr and Ru-GBP-Acr in the presence of glucose are shown in Figure 4a and b, respectively. To obtain a larger signal, an excitation wavelength of 380 nm was (27) Youn, H. J.; Terpetschnig, E.; Szmacinski, H.; Lakowicz, J. R. Anal. Biochem. 1995, 232, 24-30.
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
1405
Figure 3. Excitation spectra of GBP-Acr and Ru-GBP-Acr at emission wavelengths of 510 and 610 nm.
used for the single-labeled GBP (GBP-Acr). However, for the dual-labeled GBP (Ru-GBP-Acr), an excitation wavelength close to 410 nm should be used in order to obtain comparable fluorescence intensities for acrylodan and ruthenium (Figure 3). Otherwise, the fluorescence intensity of the ruthenium label will be too low to be used as a reference. Figure 4 shows that, for both single-labeled and dual-labeled GBP, the fluorescence intensity of acrylodan (λmax ) 510 nm) decreases with increasing concentration of glucose. Figure 4 also shows that the maximum emission wavelength of acrylodan undergoes a slight red shift of ∼7 nm upon glucose binding. These observations are consistent with the exposure of the fluorophore to the aqueous environment as the protein undergoes conformational changes in the presence of glucose. In Figure 4b, the fluorescence intensity of acrylodan at 510 nm decreases with glucose while the luminescence intensity of Ru at 610 nm remains constant. Thus, the luminescence intensity of Ru can be used as a reference for ratiometric measurements. These measurements are expected to be more impervious to errors due to sample positioning, dye photobleaching, and fluctuations in the excitation light source. Examination of the absorbance and fluorescence spectra of acrylodan and ruthenium indicates that FRET between these two fluorophores is possible. When acrylodan and ruthenium are covalently labeled to the same protein, FRET may occur as the acrylodan (donor) and ruthenium (acceptor) are in proximity. This can be observed as a decrease in fluorescence intensity of acrylodan in the dual-labeled protein. Indeed, in Figure 5, the fluorescence intensity of the acrylodan (Figure 5a) has a 3-fold decrease in emission intensity when ruthenium is labeled to the N-terminal of the same protein (Figure 5c). Comparatively, the fluorescence intensity of the acrylodan label decreases only slightly when the same amount of free ruthenium is added (Figure 5b). On the other hand, the ruthenium intensity in Figure 5b is less than the ruthenium intensity in the dual-labeled (Figure 5c). All these observations indicate that the decrease in the acrylodan emission from the dual-labeled protein is not caused by innerfilter effects but by FRET. Note that FRET between the two fluorophores makes ratiometric measurements possible. As shown 1406
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
Figure 4. Emission spectra of (a) GBP-Acr excited at 380 nm and (b) Ru-GBP-Acr excited at 410 nm in 0-16 µM glucose with maximum fluorescence intensity in 0 µM glucose normalized to unity (solid lines). For comparison, the emission spectra in 16 µM glucose with its maximum intensity normalized to unity are also shown (dotted lines), which indicates that the maximum emission wavelength is slightly red-shifted from 510 to 517 nm upon glucose binding.
in Figure 5b, the fluorescence intensity of ruthenium is too weak to be used as a reference if no FRET occurs. Using the equations described in ref 28, the Fo¨rster distance between the two fluorophores, R0, was estimated to be 29 Å. The energy-transfer efficiency between the two fluorophores, E, was estimated to be 70% according to Figure 5. With both R0 and E being known, the distance between the two fluorophores, r, can then be calculated using the following equation:28
E ) R06/(R06 + r6)
(1)
The result is 25 Å. Since the distance between the two fluoro(28) Lakowicz, J. R. Principles of Fluorescence Spectroscopy, 2nd ed.; Kluwer Academic/Plenum Publishers: New York, 1999.
Figure 5. Emission spectra of (a) GBP-Acr, (b) GBP-Acr + ruthenium, (c) Ru-GBP-Acr, and (d) ruthenium alone at an excitation wavelength of 420 nm. No glucose is added.
Figure 7. Effect of temperature on the calibration curves of RuGBP-Acr. The error bars shown are the standard errors of three repeated experiments. Table 1. Best Nonlinear Fit Results of Maximum Signal Changes (∆Fmax) and Apparent Binding Constants (Kd) at Different Temperatures temp, °C
∆Fmax
Kd, µM
R2
15 20 25 30 35
0.278 ( 0.001 0.335 ( 0.001 0.393 ( 0.001 0.458 ( 0.001 0.545 ( 0.003
0.42 ( 0.04 0.55 ( 0.04 0.71 ( 0.03 0.91 ( 0.02 1.42 ( 0.07
0.9974 0.9982 0.9992 0.9998 0.9991
dynamic quenching of the fluorophore by the polar solvent. Taken together, these factors affect the response of the glucose sensor. The apparent binding constants could be calculated by fitting the experimental results to the binding isotherm:30
∆F ) ∆Fmax [S]/(Kd + [S])
Figure 6. Tertiary structure of GBP. Leucine 255 was mutated to cysteine and labeled with acrylodan (present work). Glutamine 26 was mutated to cysteine and labeled with ANS (refs 20 and 23).
(2)
where ∆F is the normalized signal change at any ligand concentration, ∆Fmax is the normalized signal change at saturating ligand concentration, Kd is the apparent binding constant, and [S] is the concentration of the ligand in free state. Together with the equations for single binding equilibria and for mass balance, the following equation can be derived:
(
)
∆Fmax ∆F [E] ) 1 + Kd/ [S]t ∆F ∆Fmax t
(3)
phores is about half of the diameter of GBP (Figure 6), the size of GBP can then be estimated to be ∼50 Å. This value is in agreement with that reported from X-ray crystallographic data.29 The effect of temperature on the ratios of the fluorescence intensities of the labeled fluorophores is shown in Figure 7. It can be seen that the fluorescence intensity ratios increase with temperature. This temperature effect can be attributed to many factors including changes in the equilibrium of ligand binding and the quantum yields of both fluorophores. Temperature also affects
where [E]t and [S]t represent the total concentrations of GBP and glucose, respectively. The experimental results were analyzed with nonlinear regression and the best results for ∆Fmax and Kd are listed in Table 1. It can be seen that both the maximum signal changes and the apparent binding constants increase with temperature. Figure 8 gives the comparison between the experimental data and the fitted values. The Scatchard plot of [ES]/[S] (ratio of bound glucose concentration to free glucose concentration)
(29) Mowbray, S. L.; Petsko, G. A. J. Biol. Chem. 1983, 258 (13), 7991-7997.
(30) Dattelbaum, J. D.; Lakowicz, J. R. Anal. Biochem. 2001, 291, 89-95.
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
1407
Figure 9. Frequency-domain intensity decay traces of Ru-GBPAcr in increasing glucose concentrations at room temperature.
Figure 8. Nonlinear fit of experimental results to the binding isotherm. Inset: Scatchard plot of [ES]/[S] versus [ES]. As in Figure 7, the error bars shown are the standard errors of three repeated experiments.
versus [ES] is also shown in the inset. This figure shows that the experimental data conform to equations for a single binding site. The binding constant for the single-labeled GBP at 25 °C was estimated with nonlinear regression. A value of 0.84 ( 0.04 µM was obtained, which is close to that of the dual-labeled GBP, 0.71 ( 0.04 µM. The similarity between the two binding constants shows that the Ru label at the N-terminal has a minor effect on the binding activity of the protein. The effect of temperature on equilibrium constant is given by31
d ln Kd ∆H0 ) dT RT2
(4)
where T is absolute temperature. ∆H0 is the enthalpy change of the process. R is the gas constant. Suppose that ∆H0 does not change with temperature; integration of eq 4 gives
-ln Kd )
∆H0 1 +C R T
µM
(6)
Additionally, the maximum signal change and temperature were (31) Kyle, B. G. Chemical and Process Thermodynamics, 2nd ed.; PTR Prentice Hall: Englewood Cliffs, NJ, 1992.
1408
correlated using a linear equation. The result is
∆Fmax ) 0.01314T - 3.516
(7)
(5)
where C is a constant. Using eq 5, the value of ∆H0 was estimated to be 43.1 kJ/mol. As expected, the dissociation process is endothermic, while the opposite process is exothermic. Thus, for any given temperature the apparent binding constant for this sensor can be estimated as follows:
Kd ) exp(17.1 - 5180/T),
Figure 10. Modulation at 1.58 MHz versus glucose concentration.
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
Beyond ratiometric measurements, the presence of the longlived ruthenium metal ligand complex in the acrylodan-labeled GBP also allows for modulation-based sensing at relatively low frequencies. In the absence of ruthenium, changes in the fluorescence lifetime of acrylodan in response to glucose can be detected using a fast and expensive laser excitation source at frequencies close to 1 GHz.30 With ruthenium, the modulation changes are detected between 1 and 10 MHz as shown in Figure 9. Indeed, the frequency decay traces shown here were collected using a low-cost blue light-emitting diode as excitation source. Figure 10 shows the modulation at 1.58 MHz as a function of glucose concentrations. Knowing that modulation measurements can be accurate to (0.01,23 glucose concentrations can be conservatively measured to an accuracy of (0.5 µM from the modulation data.
DISCUSSION Glucose sensing remains as a very active research effort in the field of diagnostics, as well as bioprocess monitoring. This can be attributed to the near-epidemic proportions of diabetes in the United States and the predicted further increase in incidence as the population ages. Additionally, the ever-growing reliance on biotechnology requires more efficient and reliable means of monitoring bioprocesses involving cell culture and fermentation. Alternatives to the current state-of-the-art technology to detect lower levels of glucose that can be adapted to higher concentrations are desirable. The dual-labeled GBP presented here has micromolar sensitivity for both glucose and galactose,21,22 so it can be used for monitoring either of the two analytes in the absence of the other. Relatively, the binding affinity to galactose is slightly lower than to glucose.21,22 In a previous paper, we have demonstrated that this high sensitivity allows for a wider range of glucose concentrations that can be monitored.20 For example, we have shown that the performance of the GBP sensor is comparable to the standard electrochemical enzyme electrode in monitoring glucose consumption in yeast culture at glucose concentrations of 0-100 mM. This was accomplished by diluting minute amounts of samples accordingly. A more important breakthrough was the demonstration of glucose monitoring of E. coli fermentation in microliter volumes of LB media. Such measurements are not feasible with the glucose oxidase electrode because the glucose concentrations (e1mM) are below the detection limit. Additionally, the electrodes require larger sample volumes thereby precluding applications for high-throughput cell culture. At the present time, monitoring of glucose levels in fermentations using LB medium is not routine possibly because of the lack of a sensitive and convenient measuring technique. With the GBP sensor, control of glucose concentrations in LB fermentations can have a profound impact on productivity and the purity of products. Another possible application of the GBP sensor is as an alternative to glucose oxidase-based diagnostics. Of particular interest are nonconventional sampling techniques such as iontophoresis, where interstitial fluid (ISF) is extracted through the skin by the application of a mild electrical current. The collected ISF in this method contains micromolar amounts of glucose32 and may be correlated to blood glucose levels. The previous work with GBP described above utilized the ANSlabeled Q26C mutant. Although functional, as evidenced by our extensive work, improvement of the biosensor is necessary for the next step, the design of simple, reliable, low-cost instrumentation. ANS is a UV-excitable probe that absorbs and emits at wavelengths where some biological components in the media autofluoresce and may interfere with measurements. A similar probe that absorbs and emits at longer wavelengths is more desirable. Thus, we chose acrylodan, which has the added advantage of being 5 times brighter than ANS. Unfortunately acrylodan is nonresponsive when attached to position 26. Based on the reports by Marvin and Hellinga,21 we constructed the L255C mutant labeled with acrylodan, which has a relative signal change of ∼50% at glucose saturation. The acrylodan-labeled L255C GBP presented here exhibits a maximum signal change of ∼35%. (32) Tamada, J. A.; Bohannon, N. J.; Potts, R. O. Nat. Med. 1995, 11 (1), 11981201.
Further purification and 100% labeling efficiency of the protein may allow us to approach the relative signal change of ∼50% as reported.21 Further improvement of the sensor was to convert it to a ratiometric sensor with the possibility of lifetime-based sensing. We have demonstrated this in a previous paper with the glutamine binding protein labeled with acrylodan and ruthenium.33 An additional concern with the ANS label on the Q26C GBP was that it is very close to the N-terminal of the protein (Figure 6). As such its fluorescence intensity was almost totally quenched when ruthenium was attached to the N-terminal. The position at 255 is ∼15 Å further away from the N-terminal and in that sense more suitable. Certainly, even at this distance, quenching of the acrylodan was observed. Nevertheless, this quenching, which is attributed to FRET between acrylodan and ruthenium, was advantageous in that the ruthenium signal was enhanced to allow the ratiometric measurements. Thus, we have succeeded in creating a biosensor with a built-in internal reference. An alternate method for the dual-labeled GBP sensor is lifetimebased modulation sensing at frequencies much lower than those required for the single-labeled biosensor. A similar attempt has been reported with the ANS-26-GBP, but the long-lived ruthenium was simply painted onto the outer surface of the cuvette.23 With the ruthenium directly attached to the GBP-Acr, control of the thickness and concentration of the painted ruthenium is no longer a concern. Additionally, this technique can be improved further by dual-frequency lifetime discrimination. In such a method, the ratio of the modulation at two frequencies is correlated to the analyte concentration. This prevents ambient light from interfering with the measurements as demonstrated previously in a low-cost semiconductor-based ratiometric oxygen-sensing device.34,35 Another essential factor that determines the response of a biosensor is the temperature. This is particularly important for this sensor, which is designed to be used for different applications. The relations between the maximum signal changes, apparent binding constants, and temperature obtained in this paper can be used to predict or correct for the effect of temperature on glucose measurements. One may also “tune” the sensor response range by simply changing the operating temperature if required.
CONCLUSIONS A glucose sensor based on dual-labeled GBP was described in this paper for monitoring micromolar glucose levels or higher. It provides a viable alternative to current technology in that lower levels of glucose can be determined. This sensor will most likely find relevance in biomedical diagnostics and bioprocessing. The process of designing a working protein-based sensor such as this requires the combination of different disciplines from molecular biology, structural biology, probe chemistry, and fluorescence spectroscopy. As a working device is being crafted to make a more practical sensor, we may add electrical engineering to the various disciplines. We foresee that the design of other biosensors will follow the same vein. (33) Tolosa, L.; Ge, X.; Rao, G. Anal. Biochem. 2003, 314, 199-205. (34) Kostov, Y.; Harms, P.; Pilato, R. S.; Rao, G. Analyst 2000, 125, 1175-1178. (35) Kostov, Y.; Harms, P.; Rao, G. Anal. Biochem. 2001, 297, 105-108.
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
1409
ACKNOWLEDGMENT The authors thank Dr. Yordan Kostov for his assistance in frequency-domain measurements and acknowledge the support of the Juvenile Diabetes Research Foundation International through Grant 4-1999-793, the National Institutes of Health through Grant DK062990, NSF through Grant BES 0091705, and unrestricted funding from DuPont, Fluorometrix, Genentech,
1410
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
Merck, and Pfizer. The E. coli NM303 was from cultures originally provided by Dr. W. Boos, University Konstanz, Germany.
Received for review September 11, 2003. Accepted December 9, 2003. AC035063P