UV Fluorescence Lifetime Imaging Microscopy: A Label-Free Method

DOI: 10.1016/j.medpho.2014.12.001. STANLEY W. BOTCHWAY, KATHRIN M. SCHERER, STEVE HOOK, CHRISTOPHER D. STUBBS, ELEANOR WESTON, ...
0 downloads 0 Views 568KB Size
Anal. Chem. 2006, 78, 663-669

Accelerated Articles

UV Fluorescence Lifetime Imaging Microscopy: A Label-Free Method for Detection and Quantification of Protein Interactions Mark Schu 1 ttpelz, Christian Mu 1 ller, Hannes Neuweiler, and Markus Sauer*

Applied Laser Physics and Laser Spectroscopy, University of Bielefeld, Universitaetsstrasse 25, 33615 Bielefeld, Germany

Due to the ability to detect multiple parameters simultaneously, protein microarrays have found widespread applications from basic biological research to diagnosis of diseases. Generally, readout of protein microarrays is performed by fluorescence detection using either dyelabeled detector antibodies or direct labeling of the target proteins. We developed a method for the label-free detection and quantification of proteins based on time-gated, wide-field, camera-based UV fluorescence lifetime imaging microscopy to gain lifetime information from each pixel of a sensitive CCD camera. The method relies on differences in the native fluorescence lifetime of proteins and takes advantage of binding-induced lifetime changes for the unequivocal detection and quantification of target proteins. Since fitting of the fluorescence decay for every pixel in an image using a classical exponential decay model is time-consuming and unstable at very low fluorescence intensities, we used a new, very robust and fast alternative method to generate UV fluorescence lifetime images by calculating the average lifetime of the decay for each pixel in the image stack using a model-free average decay time algorithm.To validate the method, we demonstrate the detection and quantification of p53 antibodies, a tumor marker in cancer diagnosis. Using tryptophancontaining capture peptides, we achieved a detection sensitivity for monoclonal antibodies down to the picomolar concentration range. The obtained affinity constant, Ka, of (1.4 ( 0.6) × 109 M-1, represents a typical value for antigen/antibody binding and is in agreement with values determined by traditional binding assays.

* To whom correspondence should be addressed. Telephone: +49-521-1065451. Fax: +49-521-106-2958. E-mail: [email protected]. 10.1021/ac051938j CCC: $33.50 Published on Web 12/23/2005

© 2006 American Chemical Society

Detection and quantification of the entire spectrum of proteins that is expressed by an organism, tissue, or cell at any one time is of fundamental importance for systems biology and proteomics.1,2 Both proteomics and systems biology are thought to open new horizons in many research fields of life sciences. Especially clinical proteomics, i.e., proteomics activities in the field of medicine, is of special interest because it promises to accelerate the discovery of new drug targets and molecular markers for in vitro diagnostic applications.3 Since proteins can be expressed across enormous dynamic ranges, from 1 to 106 in cells, new advanced techniques are urgently required for their analysis. In addition, there is no protein equivalent of the polymerase chain reaction. Thus, the desired techniques should exhibit superior sensitivity. In conventional proteomics, mass spectrometry is used to identify proteins separated on polyacrylamide gels, which are frequently stained with Coomassie, fluorescent dyes, or silver stain.4-7 Alternatively, ordered arrays of peptides and proteins can be used for parallel protein analysis. Microarray technology allows the simultaneous analysis of thousands of parameters within a single experiment. Microspots of capture molecules are immobilized in arrays onto a solid support and exposed to samples containing the corresponding binding molecules. In combination with fluorescence-based readout techniques, protein microarrays (1) Ideker, T.; Thorsson, V.; Ranish, J. A.; Christmas, R.; Buhler, J.; Eng, J. K.; Bungarner, R.; Goodlett, D. R.; Aebersold, R.; Hood, L. Science 2001, 292, 929-934. (2) Weston, A. D.; Hood, L. J. Proteome Res. 2004, 3, 179-196. (3) Vitzthum, F.; Behrens, F.; Anderson, N. L.; Shaw, J. H. J. Proteome Res. 2005, 4, 1086-1097. (4) Patterson, S. D.; Aebersold, R. Electrophoresis 1995, 16, 1791-1814. (5) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Nat. Biotechnol. 1999, 17, 994-999. (6) Gygi, S. P.; Corthals, G. L.; Zhang, Y.; Rochon, Y.; Aebersold, R. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 9390-9395. (7) Fey, S. J.; Larsen, P. M. Curr. Opin. Chem. Biol. 2001, 5, 26-33.

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006 663

achieve extraordinary sensitivity.8-13 The ability to quantify multiple target structures simultaneously has applications in basic biological research, molecular classification and diagnosis of diseases, identification of therapeutic markers and targets, and profiling of response to toxins and pharmaceuticals.14 Ultimately, protein chips might allow scientists to create protein “snapshots” of cells or to identify tumor-associated proteins. However, introduction of fluorescent labels requires chemical modification and additional incubation and washing steps. Furthermore, fluorescence modification might perturb the functionality of ligandreceptor interactions. Finally, selective fluorescence labeling of a small number of molecules present in a tiny volume, for example, in a single cell, is cumbersome or even impossible. In this regard, efforts have been made in direct measurement of the UV absorbance or native fluorescence of proteins to eliminate the problems of staining in two-dimensional gel and microchip electrophoresis.15-22 Generally, proteins exhibit an absorption maximum in the UV region around 280 nm, which mainly arises from the absorption of the three aromatic amino acid residues tryptophan, tyrosine, and phenylalanine.23 Since 99.5% of all human proteins with a mass of more than 10 kDa contain at least one tryptophan or tyrosine residue (phenylalanine shows substantially weaker fluorescence23), native UV fluorescence detection volunteers as an alternative detection technique for proteins.19 Due to differences in fluorescence quantum yield and resonance energy transfer from proximal phenylalanine to tyrosine or tyrosine to tryptophan, fluorescence of proteins is usually dominated by tryptophan fluorescence. Recently,19 Roegener et al. utilized a frequency-tripled Ti:sapphire laser for direct excitation of unlabeled proteins in polyacrylamide gels at 280 nm and achieved a detection sensitivity of 1-5 ng for various protein bands. Alternatively, time-resolved UV fluorescence detection schemes can be applied for the readout of unlabeled protein microarrays. Here, detection of target proteins solely relies on the change in native fluorescence lifetime measured upon binding to capture molecules. For example, Striebel et al. moved the substrate (a protein microarray) step by step through the focus of a pulsed laser beam (8) Abbott, A. Nature 1999, 402, 715-720. (9) MacBeath, G.; Schreiber, S. Science 2000, 289, 1760-1763. (10) Templin, M. F.; Stoll, D.; Schrenk, M.; Traub, P. C.; Vohringer, C. F.; Joos, T. O. Trends Biotechnol. 2002, 20, 160-166. (11) Holt, L. J.; Enever, C.; de Wildt, R. M.; Tomlinson, I. M. Curr. Opin. Biotechnol. 2000, 11, 445-449. (12) de Wildt, R. M.; Mundy, C. R.; Gorick, B. D.; Tomlinson, I. M. Nat. Biotechnol. 2000, 18, 989-994. (13) Pandey, A.; Mann, M. Nature 2000, 405, 837-846. (14) Zhu, H.; Bilgin, M.; Bangham, R.; Hall, D.; Casamayor, A.; Bertone, P.; Lan, N.; Jansen, R.; Bidlingmaier, S.; Houfek, T.; Mitchell, T.; Miller, P.; Dean, R. A.; Gerstein, M.; Snyder, M. Science 2001, 293, 21012105. (15) Hogan, B. L.; Yeung, E. S. Appl. Spectrosc. 1989, 43, 349-350. (16) Yamamoto, H.; Nakatani, M.; Shinya, K.; Kim, B. H.; Kakuno, T. Anal. Biochem. 1990, 191, 58-64. (17) Koutny, L. B.; Yeung, E. S. Anal. Chem. 1993, 65, 148-152. (18) Kazmin, D.; Edwards, R. A.; Turner, R. J.; Larson, E.; Starkey, J. Anal. Biochem. 2002, 301, 91-96. (19) Roegener, J.; Lutter, P.; Reinhardt, R.; Bluggel, M.; Meyer, J. E.; Anselmetti, D. Anal. Chem. 2003, 75, 157-159. (20) Sluszny, C.; Yeung, E. S. Anal. Chem. 2004, 76, 1359-1365. (21) Striebel, H. M.; Schellenberg, P.; Grigaravicius, P.; Greulich, K. O. Proteomics 2004, 4, 1703-1711. (22) Schulze, P.; Ludwig, M.; Kohler, F.; Belder, D. Anal. Chem. 2005, 77, 1325-1329. (23) Lakowicz, J. R. Principles of fluorescence spectroscopy; Plenum Press: New York, 1983.

664

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

and measured the integrated fluorescence decay of each individual fluorescent spot using time-correlated single photon counting (TCSPC).21 Applying a standard biexponential fit model to the decay curves measured with and without binding partner, they were able to detect proteins in the nanomolar concentration range.21 Here, we demonstrate a novel label-free method for the detection and quantification of protein interactions that uses the differences in steady-state and time-resolved fluorescence intensity of unlabeled proteins to detect and quantify binding. We developed a time-gated two-dimensional UV fluorescence lifetime imaging microscope (UV-FLIM) system capable of monitoring changes in the fluorescence lifetime caused by specific binding of proteins to capture molecules immobilized on a nitrocellulose membrane. Frequency-tripled laser pulses from a Ti:sapphire with a wavelength of 280 nm and a repetition rate of 80 MHz were used to uniformly illuminate a sample area of 1 cm2. Time-resolved fluorescence intensity images were obtained using a time-gated image intensifier in combination with a UV-sensitive CCD camera. Besides the avoidance of the problems related to staining, the use of native UV-FLIM based on changes in the fluorescence lifetime offers an additional advantage: the independence of the fluorescence lifetime signal to alterations in protein concentration facilitates the quantification of target proteins. The method enables the simultaneous two-dimensional readout of microarrays (i.e., detection and identification of different antigens and antibodies) solely through their characteristic fluorescence lifetime. Furthermore, we show that UV-FLIM can be used advantageously for the development of heterogeneous label-free assays. Therefore, we investigated binding of p53 antibodies to linear peptide epitopes derived from human p53 protein. p53 antibodies are independent and highly specific tumor markers and can potentially play an important role in early cancer diagnosis.24 We used tryptophancontaining peptide epitopes as immobilized capture molecules to detect and quantify monoclonal model antibodies (BP53-12) down to the picomolar concentration range. The determined association constant corresponds well to those determined for other antigen/antibody systems and demonstrates that the introduced label-free method is generally applicable for investigations of protein interactions starting from principal protein binding studies in molecular biology to diagnosis in medicine and drug screening.

EXPERIMENTAL SECTION Materials. The following proteins and conjugates were purchased from Sigma: bovine serum albumin (Cohn fraction V, A2153) and monoclonal anti-bovine serum albumin antibody produced in mouse (clone BSA-33, B2901), rabbit IgG (I5009), and sheep anti-rabbit IgG F(ab′)2 fragment (C2306). Monoclonal antibodies directed against human p53 (clone BP53-12) were purchased from Dianova. The p53 peptide (sequence: SQETFSDLWKLLPEN) was synthesized by Dr. Pipkorn at the DKFZ (Deutsches Krebsforschungszentrum, Heidelberg, Germany) using standard solid-phase synthesis. All proteins were solved and diluted in phosphate buffered saline (PBS, pH 7.4, 0.25% Tween 20). (24) Lane, D. P. Nature 1992, 358, 15-16.

Figure 1. Time-gated UV fluorescence detection. (a) Principle of time-gated fluorescence lifetime measurements. A short laser pulse from a frequency-tripled Ti:sapphire laser excites the native UV fluorescence at 280 nm (for details see Experimental Section). The fluorescence decay information is obtained by measuring the fluorescence intensity at various delay times with a predetermined exposure and acquisition time and gate width. (b) Time-gated fluorescence intensity measurements of aqueous solutions (PBS, pH 7.4, 0.25% Tween 20) of different proteins, antibodies, and peptides using a gate width of 300 ps and a gate pitch of 900 ps. A stack of 10 images was obtained with a delay ranging from 0 to 8100 ps and a total exposure time of 100 ms for each gate. Fitting of the experimental data (dots) by biexponential fits results in fluorescence decay curves.

Immobilization or Proteins and Peptides. For immobilization, 1 mg/mL p53 peptide was incubated on a nitrocellulose blotting membrane (BioTrace NT; Pall) for 30 min. Excess peptide and unspecific protein-surface interactions were removed by a washing step (PBS, pH 7.4, 0.25% Tween 20). In the second step, p53 antibody (BP53-12) solutions of different concentrations (100 pg/mL up to 10 µg/mL) were incubated and allowed to equilibrate for 4 h at room temperature followed by washing with PBS. Measurements were performed in PBS, pH 7.4, 0.25% Tween 20. Time-Resolved UV Fluorescence Detection. Briefly, the setup consisted of a pulsed laser for excitation and a time-gated detection unit that is highly sensitive in the ultraviolet (UV) region (Figure 1a). A mode-locked Ti:sapphire laser (Tsunami; SpectraPhysics) tuned to 840 nm was pumped by a 532-nm all-solid-state laser (Millennia; Spectra-Physics). The emitted laser light was thereafter frequency tripled using a flexible harmonic generator (GWU-Lasertechnik). Thus, 280-nm laser light pulses with a pulse length of 100 fs (fwhm) and 80-MHz repetition rate were generated and used as excitation light. By means of a holographic diffuser, the beam was expanded to illuminate the sample area of ∼1 cm × 1 cm, resulting in an average laser excitation intensity of 20 mW/cm2. The fluorescence signal of the sample was one-to-one imaged using two UV-transparent objectives (UV Nikkor, f ) 105 mm; Nikon), a time-gated image intensifier (High Rate Imager; Kentech Instruments), and a UV-sensitive CCD camera (Sensicam; PCO). The high rate imager (HRI) was able to acquire time-gated fluorescence intensity images with a gate width tg between 200 ps and 1 ns that were synchronized to the laser pulse repetition rate. The delay between the laser pulse and the start of the gate was triggered by a programmable delay controller. To avoid disturbing reflectance and scattering of the excitation light, an additional fluorescence filter (300-375 nm; Edmund Optics) was used between the two objectives. A mechanical shutter prevented

the proteins from bleaching caused by unnecessary exposure to the excitation light. To measure the fluorescence decay of the sample, the gate width tg of the image intensifier was set to 300 ps with a gate pitch td of 900 ps. A stack of 10 images was obtained with a delay ranging from 0 to 8100 ps and a total exposure time of 100 ms for each gate. After background correction, each pixel was fitted with a multiexponential decay model by use of a nonlinear least-squares curve-fitting algorithm. Alternatively, an “average fluorescence decay time” 〈T〉 was calculated. Intensity images were obtained by adding up the complete series of time gated intensity images. Fluorescence Lifetime Algorithms. Generally, the data structure obtained can be described by a stack of N images. To parametrize the time information, each image is related to a time ti ) itd with I ) 0...N - 1. Therefore, each pixel (x,y) is related to a decay function D(x,y) ) {di(x,y)} ) {d(ti;x,y)}, where di(x,y) represents the intensity at the position (x,y) in the image i. Usually, the decay of the fluorescence signal of a protein is described by a multiexponential function assuming that each aromatic amino acid exhibits a characteristic fluorescence lifetime τn. The fractional contribution of each fluorophore to the total intensity is indicated by An. Therefore, the time-dependent intensity function can be written as

d(ti;x,y) )

∑A (x,y)e

-ti/τn

n

(1)

n

Although this model constitutes the accurate way to describe the measured fluorescence decays, in practice, a simplified biexponential decay model is often routinely used. Here, we propose an alternative technique to describe multiexponential decay functions from complex samples, an average decay time algorithm that is absolutely model free. The function D(x,y) can Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

665

be transposed to the average decay time 〈T(x,y)〉 using the following equation. N-1

∑ t d (t ;x,y) i i

〈T(x,y)〉 )

i

i)0

(2)

N-1

∑ d (t ;x,y) i

i

i)0

Hence, no information about the response function of the high rate imager and the underlying decay model of each aromatic amino acid contributing to the measured decay function is necessary. This property enables simplification of the analysis of multiexponential decay fluorescence signals. Additionally, the algorithm is extremely fast compared to least-squares curve-fitting algorithms and can be easily implemented in every kind of analysis software package, allowing one to generate online fluorescence decay information. RESULTS AND DISCUSSION One serious limitation of native UV fluorescence detection of aromatic amino acids constitutes their low photostability under one- and two-photon excitation conditions, which renders the application of native fluorescence for highly sensitive detection schemes more complicated.25 In addition, the intrinsic fluorescence of tryptophan or tyrosine is weaker compared to conventional fluorescent dyes. On the other hand, many proteins contain more than one tryptophan, which compensates the described limitations and enables the detection of tryptophan-rich proteins at nanomolar concentrations in aqueous solutions.25 The native fluorescence lifetime of proteins in solution strongly depends on amino acid sequence, the tertiary structure, and the local environment, for example, polarity and pH of the solvent, and varies typically between 1 and 7 ns.23 This comparatively long fluorescence lifetime enables their determination using both classical time-domain and frequency-domain techniques.23 Time-domain techniques record the intensity of the fluorescence signal as a function of time following pulsed excitation. For example, TCSPC measures the arrival times of individual photons registered at the detector with respect to the excitation light pulse and sorts the arrival times into channels labeled with the corresponding time.26,27 Frequency-domain techniques, on the other hand, record the phase and the amplitude of a signal as a function of frequency. To detect unlabeled proteins by the native fluorescence of intrinsic tryptophan and tyrosin residues, we used pulsed laser excitation at 280 nm generated from a frequency-tripled Ti: sapphire. In contrast to confocal scanning fluorescence lifetime imaging microscopy (FLIM) where the image is acquired pixel by pixel moving the sample through the excitation focus,21,28-30 (25) Lippitz, M.; Erker, W.; Decker, H.; van Holde, K. E.; Basche, T. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 2772-2777. (26) Becker, W. Advanced time-correlated single photon counting techniques; Springer: Berlin, 2005. (27) O’Connor, D. V.; Phillips, D. Time-correlated single photon counting; Academic Press: London, 1984. (28) So, P. T.; Dong, C. Y.; Masters, B. R.; Berland, K. M. Annu. Rev. Biomed. Eng. 2000, 2, 399-429. (29) Konig, K. J. Microsc. 2000, 200, 83-104. (30) Tinnefeld, P.; Herten, D.; Sauer, M. J. Phys. Chem. A 2001, 105, 79898003.

666 Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

Table 1. Spot-Integrated UV Fluorescence Lifetimes and Corresponding Amplitudes Derived from a Biexponential Fit, τ1, A1, τ2, A2, and the Mean Average Fluorescence Lifetimes 〈T〉 of Different Proteins and a Peptide Epitope Measured in PBS Solution, pH 7.4, 0.25% Tween 20a

anti-Rab IgG Rab IgG anti-BSA BSA BP53-12 p53 epitope

τ1 (ns)

A1

τ2 (ns)

A2

〈T〉 (ns)

5.69 3.20 4.53 5.98 6.07 3.51

0.73 0.27 0.49 0.81 0.82 0.75

1.27 0.97 1.21 1.54 1.33 0.97

0.27 0.73 0.51 0.19 0.18 0.25

3.16 1.77 2.56 3.26 3.28 2.60

a Fluorescence lifetimes were calculated from all photons detected within an area of interest applying a biexponential fit. Mean average fluorescence lifetimes were calculated from the average fluorescence lifetimes of all pixels included in the spot.

we applied wide-field camera-based FLIM31,32 to simultaneously gain the desired lifetime information from each pixel of a UVsensitive CCD camera using a time-gated image intensifier (Figure 1a). Figure 1b shows native UV fluorescence decay curves recorded from aqueous phosphate-buffered solutions of different proteins and a tryptophan-containing peptide (100 mg/mL) using a gate width of 300 ps and a gate pitch of 900 ps. In these experiments, 2 µL of protein or peptide solution was transferred onto the sample stage consisting of pure silica glass to reduce background fluorescence. A series of 10 background-corrected images was taken at different delay times and accumulated to reconstruct false color fluorescence intensity images (Figure 2a,c,e). Fluorescence lifetime information was obtained by applying either a biexponential fit to the time-resolved intensity data or by calculating a model-free average fluorescence lifetime 〈T〉 (Table 1). Spot-integrated fluorescence lifetimes were calculated using all photons detected within an area of interest (a rectangular area with a size of ∼0.2 × 0.2 cm covering the whole spot was chosen in the background-corrected images as shown in Figure 2a,c,e). As can be seen in Table 1, the least-squares curve-fitting algorithm yields lifetimes that vary from approximately 1 to 6 ns, which is in good agreement with previous reports.21 Although native protein fluorescence often exhibits multiexponential character,23 fluorescence decays reconstructed from time-gated intensity measurements with varying delay times could be satisfactorily fit by a simple biexponential fit. The use of additional exponents did not improve the quality of the fit but rather induced numerical instabilities of the fitting algorithm. Since fitting of the fluorescence decay for every pixel in an image using a classical exponential decay model is time-consuming and might become unstable at very low fluorescence intensities, we used a new, very robust and fast alternative method to generate UV fluorescence lifetime images from time-gated fluorescence intensity images. That is, we calculated the average lifetime of the decay for each pixel in the image stack using an average decay time algorithm that is absolutely model free (see Experimental Section). This yields a false color image (Figure 2b,d,f) in which each pixel represents (31) Elson, D.; Webb, S.; Siegel, J.; Suhling, K.; Davis, D.; Lever, J.; Phillips, D.; Wallace, A.; French, P. Opt. Lett. 2002, 27, 1409-1411. (32) Agronskaia, A.; Tertoolen, L.; Gerritsen, H. J. Phys. D 2003, 36, 16551662.

Figure 2. Background-corrected false color native fluorescence intensity images (a,c,e) and corresponding average fluorescence decay time (UV-FLIM) 〈T(x,y)〉 images (b,d,f) of different aqueous protein and peptide solutions deposited on bare silica glass surface. Signals of anti-Rab IgG (a,b), anti-BSA (c,d), and model antibody BP53-12 (e,f) are shown on the left sides of the intensity and lifetime images. Signals of corresponding Rab IgG, BSA, and p53 epitope peptide are shown on the right sides of the images.

the average lifetime 〈T(x,y)〉 at the respective position (x,y). In contrast to fluorescence intensity images, fluorescence lifetime images appear by far more homogeneous. As can be seen in Table 1, the different fluorescence lifetimes of antibodies and corresponding antigens, for example, Rab IgG and anti-Rab IgG with average decay times of 1.77 and 3.16 ns, respectively, differ strongly and can therefore be used for the unequivocal identification of proteins by UV-FLIM (Figure 2b,d,f). Comparison of the fluorescence intensity and UV-FLIM images (Figure 2) demonstrates clearly the advantage of UV-FLIM for the identification of different proteins and peptides by native fluorescence detection. Differences in fluorescence lifetime as small as a few hundred picoseconds are easily resolved. In addition, the quality and reliability of UV-FLIM measurements is not influenced by the concentration of the protein. It solely depends on the specific configuration of the probed aromatic amino acids. Furthermore, the more homogeneous average fluorescence lifetimes facilitate identification and discrimination of the fluorescence from background signal. To demonstrate the potential of UV-FLIM for diagnostic applications, we studied in detail binding of a p53 peptide epitope to the corresponding monoclonal antibody BP53-12 with average fluorescence lifetimes, 〈T〉 of 2.60 and 3.28 ns, respectively, as a realistic antigen/antibody model system. Gene mutations in the tumor suppressor protein p53swhich is also referred to as the “guardian of the genome”sare the most common mutations in human cancer tissues and often trigger an immune response to produce autoantibodies directed against p53. It has been shown that the presence of p53 autoantibodies in human serum validates

a malignant disease with a specificity of 100% if autoimmune diseases can be neglected.33 Lubin et al. have shown that the immune response of cancer patients is not directed against the mutated central region, but against two short linear peptides mainly localized in the transactivation domain.34 We used one of these two short tryptophan-containing peptides denoted here as p53 peptide epitope and B53-12 as model antibody for human p53 autoantibodies. For binding studies, p53 peptide epitopes were immobilized on a nitrocellulose membrane. Nitrocellulose membranes are the most popular membranes for blotting of proteins. Furthermore, nitrocellulose membranes show comparably low background fluorescence in the UV under the applied experimental conditions and exhibit a high protein binding capacity (up to ∼200 µg/cm2) by means of hydrophobic and electrostatic interactions. The p53 peptide epitope (amino acid sequence: SQETFSDLWKLLPEN) was immobilized on nitrocellulose membrane snippets (∼0.2 cm × 0.6 cm pieces). Subsequently, the membrane was incubated with BP53-12 antibodies at concentrations ranging from 6.6 × 10-8 to 6.6 × 10-13 M (corresponding to 10 µg/mL-100 pg/mL). Additionally, one sample was incubated with PBS solution only as a negative control. UV-FLIM measurements were performed immediately after a washing step in order to keep the nitrocellulose wet to avoid unwanted denaturation of (33) Soussi, T. Trends Immunol. Today 1996, 17, 354-356. (34) Lubin R.; Schlichtholz, B.; Bengoufa, D.; Zalcman, G.; Tredaniel, J.; Hirsch, A.; Caron de Fromentel, C.; Preudhomme, C.; Fenaux, P.; Fournier, G.; Mangin, P.; Laurent-Puig, P.; Pelletier, G.; Schlumberger, M.; Desgrandchamps, F.; Leduc, A.; Peyrat, J. P.; Janin, N.; Bressac, B.; Soussi, T. Cancer Res. 1993, 53, 5872-5876.

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

667

Figure 3. False color fluorescence intensity (a) and corresponding UV-FLIM 〈T(x,y)〉 images (b) of different BP53-12 antibody concentrations bound to p53 peptide epitope immobilized on nitrocellulose. Measurements were performed after a washing step in PBS, pH 7.4, 0.25% Tween 20.

the protein and to maintain antibody activity. The average fluorescence lifetimes 〈T(x,y)〉 were calculated from the photons recorded per pixel in 10 images. Fluorescence intensity images were constructed by accumulation of images. Figure 3 shows fluorescence intensity and UV-FLIM images obtained for different antibody concentrations. In strong contrast to the fluorescence intensity images, the UV-FLIM images demonstrate unequivocally binding of antibodies to the peptide epitopes at different concentrations reflected by an increase in average fluorescence lifetimes (compare Figure 3a and b). Compared to the fluorescence lifetimes measured in phosphate-buffered solutions (Table 1), lifetimes of immobilized peptides and bound antibodies appear to be drastically shortened upon immobilization on nitrocellulose membranes from 2.60 to 2.17 ns and from 3.28 to 2.37 ns for the peptide epitope and the antibody, respectively. Whereas the decrease in fluorescence lifetime of the peptide of ∼0.4 ns may well be caused by quenching of the tryptophan residue by nitrocellulose, a reduction of the apparent lifetime of the more buried tryptophan residues in the antibody of ∼1 ns seems to be less plausible. However, the measured fluorescence signal might be dominated by the peptide epitopes that bind at high concentrations to the nitrocellulose membrane. Since the majority of peptide epitopes might be unsuitable for specific binding of antibodies due to steric hindrance, the apparent fluorescence lifetime is strongly dominated by the quenched peptide epitopes. Therefore, apparent native UV fluorescence lifetimes of proteins and peptides in different environments are hardly predictable. Nevertheless, the binding-induced increase in fluorescence lifetime of ∼200 ps is adequate for a definite discrimination between free and bound peptide epitopes. In addition, it has to be pointed out that immobilization of proteins on nitrocellulose membrane results in increased photobleaching as compared to measurements in homogeneous solution under the applied experimental conditions, that is, 280-nm pulsed laser excitation. Rapid photobleaching of aromatic amino acids upon exposure to UV radiation25 seriously limits the detection limit of UV-FLIM because it deteriorates the quality of measured 668 Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

fluorescence decays especially at longer decay times. To minimize photobleaching, sample exposure by UV light was prevented when moving the gate position by the use of a mechanical shutter. Furthermore, photobleaching could be substantially reduced by decreasing the number of images taken per gate and by shortening of the exposure time. Using 10 images per gate with an exposure time of 100-ms photobleaching could be confined to less than 5%; that is, the fluorescence intensity decreased by less than 5% during the measurement. Although this procedure allowed us to obtain reliable and reproducible results, a lack of accuracy concerning weakly appearing decay components that cannot be resolved has to be currently accepted. To quantify binding, a rectangular area of interest (∼0.15 cm × 0.25 cm) was selected from each sample surface to calculate the mean of the average lifetimes 〈T〉 measured for different antibody concentrations. Plotting of the mean lifetime values against the logarithm of antibody concentration results in a sigmoidal binding isotherm with 〈T〉 values varying between 2.17 and 2.37 ns (Figure 4). The lifetime converges against the upper limit for antibody concentrations above 10-6 M and against the lower limit for concentrations below 10-11 M resulting in the lifetime of the unbound peptide epitope. The measured titration curve demonstrates specific binding of the antibody BP53-12 to the peptide epitope immobilized on nitrocellulose membrane and reveals a detection sensitivity in the picomolar concentration range. This intriguing result has been achieved using solely timeresolved fluorescence information applying UV-FLIM. Commonly, the ELISA technique is used to detect tumor markers in a heterogeneous assay format. The commercially available p53autoantibody ELISA is based on the use of immobilized wild-type p53 protein as antigen in combination with secondary detector antibodies. Although sensitive (detection limit in the range of 10-9-10-11 M antibody), the ELISA is slow (several hours) and, due to the requirement of recombinant p53, expensive. Using specific peptide epitopes and model antibodies, UV-FLIM achieves a detection sensitivity and dynamic range comparable to conventional ELISA.

Figure 4. Average fluorescence lifetimes plotted against the logarithm of antibody BP53-12 concentration resulting in a sigmoidal binding isotherm.

From the point of inflection of the titration curve shown in Figure 4, we obtained an affinity constant, Ka, of (1.4 ( 0.6) × 109 M-1, which is substantially smaller than the affinity constant of 6 × 1010 M-1 determined from fluorescence correlation spectroscopy binding studies in homogeneous solution.35-37 Nevertheless, the obtained value and the expected distinct affinity reduction in the heterogeneous assay format are typical for antigen/antibody interactions38 and reflect a high affinity of the monoclonal antibody to the peptide epitope. CONCLUSIONS We developed a label-free detection method based on timegated, two-dimensional UV-FLIM that enables the identification (35) Scheffler, S.; Sauer, M.; Neuweiler, H. Z. Phys. Chem. 2005, 219, 665678. (36) Neuweiler, H.; Schulz, A.; Vaiana, A.; Smith, J.; Kaul, S.; Wolfrum, J.; Sauer, M. Angew. Chem., Int. Ed. 2002, 41, 4769-4773. (37) Neuweiler, H.; Scheffler, S.; Sauer, M. In Diagnostic Optical Spectroscopy in Biomedicine III; Mycek, M.-A., Ed.; 2005; pp 99-106. (38) Friguet, B.; Chaffotte, A. F.; Djavadi-Ohaniance, L.; Goldberg, M. E. J. Immunol. Methods 1985, 77, 305-319.

and quantification of proteins on microarrays with an intriguing limit of detection in the picomolar concentration range. Basically, the method uses differences in the fluorescence lifetime of tryptophan- or tyrosine-containing proteins and peptides for the explicit identification of target binding. Usually, data analysis in fluorescence lifetime imaging microscopy is challenging because a fluorescence decay model has to be fitted to complete TCSPC curves recorded for every pixel. Besides, time-consuming, data fitting by an exponential model involving several decay times may become imprecise and unstable at very low signal intensities. Therefore, we used a fast and robust average fluorescence lifetime 〈T〉 model to fit the measured decay curves. To validate the method, we detected and quantified p53 antibodies, a tumor marker in cancer diagnosis, using a tryptophan-containing peptide epitope as capture molecule. The performance of the new method is remarkable: fluorescence lifetime differences as small as 200 ps are sufficient for the definite detection of small amounts of target proteins (in the nano- to picomolar concentration range) even at low intensity levels. Thus, the presented method achieves a detection sensitivity and dynamic range comparable to conventional ELISA. Consequently, the presented method is potentially useful for the detection of other antigen/antibody interactions and might thus play an important role in future protein microarray technology such as cancer diagnosis, creation of protein snapshots of cells, or systems biology. To apply the developed UV-FLIM method to microarray technology, a reduction of the spot size is necessary that can be easily achieved using, for example, standard ink-jet printing techniques. ACKNOWLEDGMENT This work was supported by Philip Morris USA Inc., Philip Morris International, and the BMBF (grant 13N8027).

Received for review December 3, 2005.

October

30,

2005.

Accepted

AC051938J

Analytical Chemistry, Vol. 78, No. 3, February 1, 2006

669