Fluorescent Imaging of pH with Optical Sensors Using Time Domain

Anal. Chem. 2001, 73, 4354-4363

Fluorescent Imaging of pH with Optical Sensors Using Time Domain Dual Lifetime Referencing Gregor Liebsch, Ingo Klimant,* Christian Krause, and Otto S. Wolfbeis

Institute of Analytical Chemistry, Chemo- and Biosensors, University of Regensburg, 93040 Regensburg, Germany

We present a referenced scheme for fluorescence intensity measurements that is useful for imaging applications. It is based on the conversion of the fluorescence intensity information into a time-dependent parameter. A phosphorescent dye is added in the form of ∼10-µm particles to the sample containing the pH-sensitive fluorescent indicator. Both the reference dye and the pH probe are excited simultaneously by a blue LED, and an overall luminescence is measured. In the time-resolved imaging method presented here, two images taken at different time gates were recorded using a CCD camera. The first image is recorded during excitation and reflects the luminescence signal of both the fluorophore (pH) and the phosphor (reference). The second image, which is measured after a certain delay (after switching off the light source), is solely caused by the long-lived phosphorescent dye. Because the intensity of the fluorophore contains the information on pH, whereas phosphorescence is pHindependent, the ratio of the images displays a referenced intensity distribution that reflects the pH at each picture element (pixel). The scheme is useful for LED light sources and CCD cameras that can be gated with square pulses in the microsecond range. The fundamentals and potential of this new method, to which we refer as time domain dual lifetime referencing (t-DLR), are demonstrated. Charged coupled device (CCD)-based imaging techniques using luminescent optical sensors have been developed and applied to screening and mapping applications during the past years.1-10 One also notes an enormous improvement in the potential of CCD cameras and image-processing software that are now available for high-resolution detection and processing of luminescence signals. Most significantly, CCD-based imaging (1) Rumsey, W. L.; Vanderkooi, J. M.; Wilson, D. F. Science 1988, 241, 16491651. (2) Lu ¨ bbers, D. W.; Koster, T.; Host, G. A. Adv. Exp. Med. Biol. 1996, 388, 59-68. (3) Glud, R. N.; Ramsing, N. B.; Gundersen, J. K.; Klimant, I. Mar. Ecol.: Prog. Ser. 1996, 140, 217-226. (4) Lakowicz, J. R.; Berendt, K. W. Rev. Sci. Intrum. 1991, 62, 1727-1734. (5) Morgan, C. G.; Mitchel, A. C.; Murray, J. G.; Wall, E. J. J. Fluoresc. 1997, 7, 65-73. (6) Potyrailo, R. A.; Hieftje, G. M. Trends Anal. Chem. 1998, 17, 593-604. (7) Markgren, P.; Hamalainen, M.; Danielson, U. Anal. Biochem. 1998, 265, 340-350. (8) Gauglitz, G. Mikrochim. Acta 1999, 131, 9-17. (9) Sevick-Muraca, E. M.; Paithankar, D. Y. PCT. Appl. WO 9708538, 1997. (10) Brecht, A.; Kraus, G.; Gauglitz, G. Exp. Technol. Phys. 1996, 42, 139-160.

4354 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

accelerates data acquisition and processing to a speed that cannot be achieved using sequential methods. Luminescent sensing schemes are mostly based on monitoring the change of the luminescence intensity in response to an analyte and are widely used in microscopy.11,12 Unfortunately, the accuracy of luminescence intensity measurements is compromised by drifts in the optoelectronic system, leaching and bleaching of the indicator, nonhomogeneous indicator distribution, and turbidity. In imaging, the correction of each single picture element (pixel) for each of these interferences is cumbersome; therefore, selfreferenced methods are sought for the precise measurement of intensity. Thus, ratiometric calibration schemes use the different spectral behaviors of the fluorescence indicator and a calibration standard. Luminescence is measured at a minimum of two wavelengths.16,17 However, scattering and reflection of light are not referenced out by this method, thus limiting the applicability of the scheme. Lifetime-based schemes are superior, because lifetime is not affected by intensity or wavelength-dependent interferences (in a first approximation). A single calibration function can be used for sensors of the same type. The decay time is measured in either the frequency domain15 or the time domain;13,14 however, the complexity and the demands on the components increase with decreasing decay times. We present an alternative concept that uses a fluorophore/ phosphor couple for referencing fluorescence intensity signals and demonstrate its power for imaging poorly defined heterogeneous systems. The scheme makes use of instrumentation recently introduced for time-resolved lifetime imaging of optical sensors not based on time domain dual lifetime referencing (t-DLR).21 In (11) Miller, D. S.; Letcher, S.; Barnes, D. M.; David, M. Am. J. Physiol. 1996, 271, 508-520. (12) Pagliaro, L. Adv. Mol. Cell Biol. 1995, 11, 93-123. (13) Holst, G.; Kohls, O.; Klimant, I.; Ko¨nig, B.; Ku ¨ hl, M.; Richter, T. Sens. Actuators B 1998, 51, 163-170. (14) Hartmann, P.; Ziegler, W.; Holst, G.; Lu ¨bbers, D. W. Sens. Actuators B 1997, 38-39, 110-115. (15) Morgan, C. G.; Mitchel, A. C. Chromosome Res. 1996, 4, 261-263. (16) Zhong, X.; Rollins, A.; Alcala, R.; Marchant, R. J. Biomed. Mater. Res. 1998, 39, 9-15. (17) Parker, J. W.; Laksin, O.; Yu, C.; Lau, M.-L.; Klima, S.; Fisher, R.; Scott, I.; Atwater, B. W. Anal. Chem. 1993, 65, 2329-2334. (18) Klimant, I.; Wolfbeis, O. S. Book of Abstracts, 4th European Conference on Optical Sensors & Biosensors (Europt(r)ode IV), Mu ¨ nster, Germany, 1998, 125-126. (19) Klimant I. Ger. Pat. Appl. DE 198.29.657, 1997. (20) Lakowicz, J. R.; Castellano, F. N.; Dattelbaum, J. D.; Tolosa, L.; Grycynski, L. Anal. Chem. 1998, 67, 3160-3166. (21) Liebsch, G.; Klimant, I.; Frank, B.; Host, G.; Wolfbeis, O. S. Appl. Spectrosc. 2000, 54, 548-559. 10.1021/ac0100852 CCC: $20.00

© 2001 American Chemical Society Published on Web 07/27/2001

luminescence in Aem. Ideally, the long-lived reference luminescence is longer than that of the short-lived indicator by at least a factor of 100. The image Aex represents the sum of the intensities of the short-lived and the long-lived luminescence. In contrast, the image Aem is modulated exclusively by the intensity of the long-lived reference phosphorescence. Following the excitation pulse, both luminescence intensities rise depending on their rise times. While excited, the long-lived reference phosphorescence causes an intensity signal in Aex, referred to as AREF-ex. Assuming the luminescence intensity does not reach saturation, it can be expressed as

AREF-ex )

∫

t2

t1

IREF dt

(1)

where τ is the decay time; IREF the intensity of the reference luminophore; and ∆t, the time of integration. The short-lived indicator fluorescence causes an intensity offset IFLU after reaching saturation (see Figure 1). Integration results in AFLU, which can be expressed as a rectangular integral because after reaching saturation, ∆IFLU ) 0. The decay time of the shortlived indicator has no effect on this signal.

AFLU ) IFLU ∆t

Figure 1. Schematic representation of the time domain DLR (tDLR) scheme. The short-lived indicator fluorescence and the longlived phosphorescence of the inert reference are simultaneously excited and measured in two time gates. The first (Aex) is in the excitation period in which the light source is on and the signal obtained is composed of short-lived fluorescence and long-lived luminescence. The second gate (Aem) is opened in the emission period in which the intensity is exclusively composed of the reference luminescence. The rise and decay of the short-lived indicator fluorescence is excluded from the measurement by means of a delay. Rationing both images results in an intrinsically referenced signal that is not affected by the usual optical interferences.

contrast to the method reported there (where luminescence is monitored after an excitation pulse only), in t-DLR, luminescence is recorded both during and after illumination. It needs to be emphasized, however, that the t-DLR-based scheme requires both a tight spatial proximity and a constant ratio of fluorescent and phosphorescent dye within an imaged surface in order to take full advantage of the method. In this work, t-DLR imaging is demonstrated for mapping pH (a) in microtiterplates and (b) on surfaces in heterogeneous systems. It is clearly demonstrated that nonhomogeneities in the light field and indicator distribution as well as optical heterogeneities of the sample are being referenced out, thus providing a unique method for improving the quality of assays. EXPERIMENTAL SECTION Method. t-DLR imaging is based on the acquisition of two images, one taken in the excitation period (Aex) when the light source is on, the other in the decay period (Aem) when the light source is off. This is schematically shown in Figure 1. Luminescences are measured after a certain delay (100 ns in our case) in order to eliminate interferences by stray light and by short-lived

(2)

Therefore, the image Aex reflects the sum of both luminescences and reveals an intensity information about the analyte concentration in the sample. This is equivalent to intensity-based imaging schemes. Interferences are not referenced out in this step, however.

Aex ) AREF-ex + AFLU

(3)

The second image (Aem) is recorded in the period after the LED has been turned off. After the end of the excitation pulse and after the short-lived luminescence has decayed, luminescence AREF-em is exclusively composed of the long-lived reference luminescence (see Figure 1). It can be expressed as

Aem ) AREF-em )

∫

t4

t3

IREF dt

(4)

whereby both the delay and the length of the integration intervals are identically set in image Aex and Aem (see Figure 1). This makes the integrals of the reference phosphor in Aex and Aem complementary. The short-lived indicator fluorescence has no effect on this signal. The image Aem contains information about the disturbing interferences in Aex if the long-lived luminophore is completely inert to the analyte and if the ratio of long-lived to short-lived luminophores is constant within the imaged area. In other words, Aem displays a possible heterogeneous distribution of the lightfield or the indicators that disturb the intensity but cannot be separated or excluded in Aex. The analyte-sensitive short-lived luminescence has no effect on Aem. As a result, the interferences can be referenced out by rationing both images, Aex and Aem. Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

4355

R)

Aex AREF-ex + AFLU ) Aem AREF-em R ) a + bIFLU

(5) (6)

According to eq 6, a plot of IFLU versus the time-resolved parameter R is linear and exclusively modulated by the ratio IFLU. The apparent dynamic range can be adjusted by varying the concentration of the indicator or the reference luminophore. However, losses in the intensity of the luminophores, for example leaching or bleaching of the indicator, will not be referenced out in this scheme. To make applicable chemical sensor and imaging technology, the concentration of the respective fluorophore has to be optimized with respect to inner filter effects and concentration quenching of the fluorophores. In addition, the intensities have to be adjusted to the dynamic range of the CCD detector, and the signal-to-noise ratio (SNR) has to be optimized for both images. This can be accomplished by either varying the concentrations of the respective indicator or by changing the integration interval ∆t of one image (Aem or Aex). The latter affects the constants in eq 6, but the relation of R versus the IFLU remains linear. Referencation is also achieved if the short-lived luminescence is constant and the long-lived luminescence is sensitive to an analyte; however in this case, the measurement of the decay time in the decay period (Aem), only, is superior. In this context, the aim of the t-DLR method, the mechanism of referencing nonhomogeneous excitation intensity, nonhomogeneous indicator distribution, and filtering effects of the sample can be made clearly evident. The aim of t-DLR is intrinsically referenced imaging of fluorescent indicator dyes while using comparatively simple instrumentation (if compared to systems using laser-based excitation light sources or image intensifiers). For a successful implementation of the t-DLR scheme, it is mandatory that the ratio of both dyes is constant over the sensor foil. However, a constant total amount of dye per pixel can hardly be ensured, even if high-tech processes (e.g., offset printing) instead of standard spreading techniques are employed for manufacturing the sensor foil. As a result, neither the layer thickness nor the indicator distribution within the sensitive layer can be sufficiently controlled, which causes a thickness- and indicator-distribution-dependent luminescence intensity pattern, even if using a completely homogeneous excitation lightfield, which is technically and optically complicated to realize. These effects, however, are referenced out in the t-DLR scheme, because the quantum efficiency of the indicator dye and the reference dye depend on neither their concentration nor the excitation intensity. Therefore, the effect is identical on IREF in eqs 1 and 4 and IFLU in eq 2, and the ratio R in eq 5 remains constant at a given analyte concentration. If one develops this further and assumes the use of an ideal t-DLR sensor (both dyes show identical absorption and emission spectra - discussed under Performance of the t-DLR Method), all influences that cause an identical effect on both dyes are referenced out, no matter if the effect is actually generated within the sample (e.g., filtering effects), in the sensor foil (e.g., nonhomogeneous dye distribution), or by the system components (e.g., nonhomogeneous excitation lightfield). 4356 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

Figure 2. Schematic representation of the imaging system (bottom) and the two types of the targets used: A, microtiterplate (96 wells) with wells containing (on the bottom) spots of a planar pH-sensor; B, strip of a planar pH-sensor foil is placed inside a plastic tube and imaged from outside.

Imaging System. A gated CCD camera and a pulsed LED excitation light source is used for time-resolved monitoring of the emission intensity. The components of the imaging system and the setup were described recently21 and are schematically shown in Figure 2. A fast-pulsable LED array, consisting of 12 bright blue emitting LEDs (λmax ) 470 nm, NSPB, Nichia) served as the light source. The array was covered with a short-pass filter (BG12, Schott) for excluding the red fraction of the LED emission. Blue excitation light hit (a) a microtiterplate containing spots of the pH sensor or (b) a plastic tube equipped with the respective planar sensor. The emission light passed an emission filter (OG 530, Schott) and then was detected by a CCD camera (SensiMod, PCO, Kelheim, Germany). The camera had a black/white CCD chip with 640 × 480 pixels (307 200 pixels, VGA resolution) and a 12bit resolution, equivalent to 4096 Gy-scale values. The CCD chip could be gated directly using a minimal trigger time of 100 ns. Additional image intensification was not required. The camera and the light source were triggered by a pulse generator (DG535, Scientific Instruments), and the image data were transferred to a personal computer. Hardware and image processing were controlled by a software developed in-house.21 The data were acquired in a circuit where the images Aex and Aem, as well as the corresponding background (dark Aex and dark Aem), are recorded one after another (see Figure 3A). The system switches between two phases in which either the software or the hardware are active. In a first (software-active) phase, the

Figure 3. Schematic representation of the data acquisition process. A, time-course of the data acquisition process. The images for the different gates (Aex, dark Aex, Aem, and dark Aem) are acquired one after the other. Following the respective exposure (during the “hardware-active phase”), the image is read-out and new trigger parameters are passed (“software active phase”). The read-out time for an image and parameter passing takes, typically, 100 ms in total. The exposure time (up to 500 ms) and the trigger data (step width, 100 ns) can be individually adjusted and depends on the decay time and the luminescence intensity of the sensor that is used. The figure shows a typical setting, the exciting pulse width being 5 µs, the gate widths for both Aex and Aem being 4.9 µs. The acquisition of 4 images, each with an exposure time of 100 ms, takes 800 ms. B, data acquisition process. Similar to dealing four stacks in a card game, the images are collected until the data acquisition circuit is stopped. Summing and averaging combines the image stocks obtained to give four new images, referred to as Bex, dark Bex, Bem and dark Bem. In the final step, the respective background images are subtracted and two images (Cex and Cem) are obtained containing the basic data for evaluation.

parameters for the gates of Aex were transferred to the pulse generator (75 ms). In the subsequent (hardware-active) phase, the pulse generator, camera, and light source, respectively, were active and image Aex was acquired (up to time-point 500 ms, see Figure 3). The camera was gated (during the total exposure) by an external trigger signal in order to collect the emission in the gate. The repetition frequency and the length of the excitation pulses was adjusted to the decay time of the long-lived lumino-

phore. In our application, a 4.9 µs width for both the excitation pulse gate and the emission gate and a 100 ns delay from the LED trigger pulses were chosen (Figure 3). In other words, recording of one gate takes 10 µs and, hence, the internal repetition frequency for triggering can be adjusted to 100 kHz (10 µs-1). Assuming a total exposure of 100 ms and an internal frequency of 100 kHz, the overall emission signal is acquired 10 000 times to obtain a single image. Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

4357

It has to be noted that using an inadequate excitation power (e.g., if using a laser light source instead of the LED array) affects the efficiency of t-DLR, because then the phosphorescence accumulates in Aex. This may cause regional saturation effects, and an nonhomogeneous lightfield distribution is no longer referenced out; however, intrinsic referencation of nonhomogeneous indicator distribution or layer thickness remains valid. The effect of phosphorescence accumulation can be avoided by shortening the excitation period and lengthening the emission period. In contrast, if using excitation light sources with moderate output (e.g., the LED array), the phosphorescence, if not decayed completely during the emission period, rapidly reaches a steady state without being saturated. The resulting deviation can be neglected, because up to 10 000 cycles are added up to image Aex or Aem. After exposure, the software was activated again to read-out the image data from the memory of the camera chip (typically 25 ms) and to send the new trigger parameters for measurement of dark Aex to the pulse generator (typically 75 ms). The same is true for recording Aem and dark Aem. Thereafter, one measurement cycle is complete (tcircuit ) 800 ms at texposure ) 100 ms) and can be repeated up to 10 times to increase the SNR. The time-course of the total process is schematically given in Figure 3A. An acquisition process comparable to dealing four stacks in a card game is obtained, as shown in Figure 3B. The images in the stacks obtained are added and averaged. The resulting new images are referred to as Bex, dark Bex, Bem and dark Bem. In the next step, the background is subtracted and the information stored in two 16-bit image files (Cex and Cem) which represent the basic data for the evaluation. Slow variations in background luminescence as well as temperature effects of the LEDs are referenced out, because these are equal in all four image stacks. A constant level of background can be tolerated, because it can be easily subtracted by measuring the dark image without illumination. Sensor Materials. These were prepared from ruthenium(II)4,7-diphenyl-1,10-phenanthroline, referred to as Ru(dpp),22 and fluorescein (Merck). The aqueous solution that was used for demonstrating the performance of the t-DLR method was obtained by dropping 50 mL of a 1 mM Ru(dpp) solution in ethanol into 500 mL of an aqueous 0.5% solution of polysulfonic acid (Aldrich) while stirring. After heating and evaporating the alcohol, Ru(dpp) is electrostatically bound to the polysulfonic acid and, therefore, soluble in water. A 200-mg portion of fluorescein was added, and the resulting pH-sensitive solution was ready for use. The t-DLR pH sensor foil was prepared by immobilizing pHsensitive particles consisting of a poly(acrylonitrile)-based core that contains the inert phosphorescent luminophore Ru(dpp) and a hydrogel shell with covalently bound carboxyfluorescein in a polyurethane layer (see Figure 4). The particles were covered with hydrogel via suspension polymerization. The PAN-particles (500 mg) were suspended in 500 mL of water. Then 500 mg of acrylamide (Fluka), 100 mg of N,N-methylene(bis-acrylamide) (Fluka) and 10 mg of 2-aminoethylmethacrylate hydrochloride (Polyscience) were added to the mixture. After solvation of the monomers, the mixture was heated to 70 °C, and polymerization was initiated by UV irradiation. Polymerization was continued for (22) Klimant, I.; Wolfbeis, O. S. Anal. Chem. 1995, 67, 3160-3166.

4358 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

Figure 4. Cross section of the pH sensor membrane. The polyester film serves as an inert and optically transparent support.

2 h. After filtration, the particles were washed extensively with water and ethanol. A 500-mg portion of the coated particles was dispersed in 20 mL of buffer, pH 9. To this suspension, 10 mg of carboxyfluorescein succinimidyl ester (Molecular Probes) was added. After stirring at room temperature for 10 h, the mixture was filtered, and the particles were again washed with copious amounts of water and ethanol. Ru(dpp) and carboxyfluorescein were immobilized to particles (beads) that have a granular size