Anal. Chem. 1996, 68, 1778-1783
Multiphoton-Excited Fluorescence of Fluorogen-Labeled Neurotransmitters Jason B. Shear,* Edward B. Brown, and Watt W. Webb
Applied Physics, Cornell University, Ithaca, New York 14853
Fluorescence detection of fluorogen-labeled neurotransmitters is demonstrated using 100 fs pulses from a titanium-sapphire mode-locked laser to achieve molecular excitation by simultaneous absorption of two and three photons of near-IR radiation. Two-photon excitation spectra are determined for the naphthalene-2,3-dicarboxaldehyde derivative of glycine and the fluorescamine derivative of leucine enkephalin, with the peak excitation cross section (σ2) approximately equal to 1 × 10-50 cm4 s/photon for both species. Three-photon-excited fluorescence is demonstrated for o-phthaldialdehyde-labeled glutamate using excitation wavelengths between 965 and 1012 nm. The three-photon excitation cross section (σ3) remains nearly constant in this wavelength range, with an absolute value of ∼10-84-10-85 cm6 s2/photon2. Rapid cycling of analytes through the fluorescent excited state and detection that is free from background caused by Rayleigh and Raman scatter combine to make multiphoton-excited fluorescence a highly sensitive approach for detecting trace amounts of neurotransmitters. Measurements of two-photon-excited fluorescence of fluorescamine-labeled bradykinin and analysis of multiphoton-excited background reveal the potential of this method to detect fewer than 1000 neurotransmitter molecules. Fluorogenic reagents are used extensively in the assay of natively nonfluorescent trace components in solution.1-3 Unlike highly fluorescent fluorescein or rhodamine labels, these reagents are not intrinsically fluorescent but form fluorescent adducts when reacted with appropriate analytes, such as primary amines. Rapid and efficient labeling of trace components in solution can be accomplished using fluorogenic compounds because a large molar excess of reagent can be used while still maintaining a nominally zero-level fluorescence background. Fluorogenic derivatization previously has been used to achieve fluorescence detection limits of ∼1 amol (105-106 molecules),4,5 an impressive level that is nevertheless insufficient to measure many fundamental properties of single cells. Quantal secretion from neurons, for example, has been estimated to release as few as 104 neurotransmitter molecules into the synaptic cleft.6 Be(1) de Montigny, P.; Riley, C. M.; Sternson, L. A.; Stobaugh, J. F. J. Pharm. Biomed. Anal. 1990, 8, 419-430. (2) Gilman, S. D.; Ewing, A. G. Anal. Chem. 1995, 67, 58-64. (3) Guzman, N. A.; Moschera, J.; Bailey, C. A.; Iqbal, K.; Malick, A. W. J. Chromatogr. 1992, 598, 123-131. (4) Shippy, S. A.; Jankowski, J. A.; Sweedler, J. V. Anal. Chim. Acta 1995, 307, 163-171. (5) Ueda, T.; Mitchell, R.; Kitamura, F.; Metcalf, T.; Kuwana, T.; Nakamoto, A. J. Chromatogr. 1992, 593, 265-274. (6) Bruns, D.; Jahn, R. Nature 1995, 377, 62-65.
1778 Analytical Chemistry, Vol. 68, No. 10, May 15, 1996
cause relatively small changes in this quantal size could accompany neuronal growth or stimulation,7,8 a detection limit of as few as 102-103 molecules may be required to study these processes. Although electrochemical detection can be used to measure natively electroactive transmitters at these levels, fluorogenic derivatization provides a means for measurement of a wider variety of species, including neuroactive amino acids and peptides. The main excitatory and inhibitory neurotransmitters in the brainsglutamate, aspartate, γ-aminobutyric acid, and glycinesare not detectable using standard electrochemical techniques but can be optically measured after fluorogenic derivatization. High-sensitivity detection of fluorogenic derivatives requires an excitation source that can rapidly cycle labeled analytes through a fluorescent excited state and a detection system that efficiently discriminates fluorescence from background. Some fluorogens, such as fluorescamine,9 react to form products optimally excited at wavelengths not available using standard lasers. In other instances, fluorogenic derivatives have relatively small Stokes shifts, making it difficult to discriminate fluorescence from elastic and inelastic scatter without sacrificing a large fraction of the fluorescence signal. Naphthalene-2,3-dicarboxaldehyde10 (NDA) derivatives, for example, are frequently excited with the 442-nm line of the helium-cadmium laser. The fluorescence from NDA derivatives in aqueous solutionscentered at ∼490 nmsmust be distinguished from intense Rayleigh scatter at 442 nm and from H2O Raman bands at ∼520 and 480 nm. In liquid chromatography, the presence of organic solvents complicates Raman spectra further, making background interference more problematic. In this work, we introduce the use of a tunable near-IR modelocked laser to generate multiphoton-excited (MPE) fluorescence from fluorogen-labeled neurotransmitters. Two-photon fluorescence excitation spectra are determined for fluorescamine-labeled leucine enkephalin and for NDA-labeled glycine, and three-photonexcited fluorescence is demonstrated for o-phthaldialdehyde (OPA)-labeled glutamate. In addition, we present measurements of two-photon-excited fluorescence for fluorescamine-labeled bradykinin in deep-well slides and examine components of background. Because the excitation wavelength is far removed from the fluorescence wavelengths, contributions of Rayleigh and Raman scatter to the measured background can be eliminated without sacrificing the fluorescence signal, leaving fluorescence (7) Kuffler, S. W.; Nicholls, J. G.; Martin, A. R. From Neuron to Brain, 2nd ed.; Sinauer Associates: Sunderland, MA, 1984; pp 258-259. (8) Van der Kloot, W. J. Physiol. 1993, 468, 567-589. (9) Udenfriend, S.; Stein, S.; Bo ¨hlen, P.; Dairman, W.; Leimgrubeer, W.; Weigele, M. Science 1972, 178, 871-872. (10) de Montigny, P.; Stobaugh, J. F.; Givens, R. S.; Carlson, R. G.; Srinivasachar, K.; Sternson, L. A.; Higuchi, T. Anal Chem. 1987, 59, 1096-1101. S0003-2700(96)00007-8 CCC: $12.00
© 1995 American Chemical Society
from reagent species as the principal source of background. A prescription is given for the future combination of MPE fluorescence with microcolumn separations, in which the predominate background will be MPE fluorescence from solution impurities. Our measurements predict detection limits below 1000 molecules for multiphoton-excited fluorescence of fluorogen-labeled neurotransmitters used in concert with separation procedures, an improvement of ∼100-fold compared to that possible with onephoton-excited fluorescence. THEORY Two-Photon Excitation. Electronic excitation of materials through the simultaneous absorption of two photons was demonstrated in 1961 by Kaiser and Garret11 and is now used in a wide variety of fields, including gas-phase physical chemistry,12 analytical chemistry,13 and laser-scanning microscopy.14 When molecular excitation is achieved via a two-photon process, the fluorescence emission rate depends on the square of the instantaneous laser intensity, provided that the excitation rate is sufficiently low to avoid ground-state depletion. Consequently, efficient two-photon excitation is achieved using lasers that provide brief (∼100 fs), high-intensity pulses approximately once per several fluorescence lifetimes of the fluorophore. It has been shown previously15 that the measured fluorescence count rate (in the thick sample limit) using pulsed excitation and a Gaussian beam is
〈F(t)〉2 ≈ (1/2)φηCσ2g2
nπλ〈P(t)〉2 h2c2
(1)
In eq 1, φ is the fluorescence-to-counts conversion efficiency, η is the fluorescence quantum yield, C is the concentration of sample (neglecting ground-state depletion caused by excitation, photobleaching, or intersystem crossing), σ2 is the two-photon excitation cross section (in units of distance4 × time/photon), n is the index of refraction of the sample medium, 〈P(t)〉 is the average power at the sample in watts, λ is the excitation wavelength in vacuum, h is Planck’s constant, and c is the speed of light. The coefficient, 1/2, in eq 1 allows η to assume the standard (one-photon) definition of fluorescence quantum yield. The parameter g2 is a measure of the second-order temporal coherence of the laser:
g2 )
〈P2 (t)〉 〈P(t)〉
2
)
gp fτ
(2)
Here, gp is a quantity that reflects the shape of the laser pulse, and fτ is the frequency of the pulse train multiplied by the pulse width (fwhm), or the laser duty cycle. For mode-locked femtosecond Ti-sapphire lasers, g2 ≈ 104-105. Note from eq 1 that the measured two-photon-excited fluorescence rate for a homogeneous solution of dye is independent of focal volume. The selection rules for two-photon absorption are different than for single-photon events; however, the degree of molecular symmetry and vibronic coupling in a chromophore has significant effects on the strength of the allowed transitions in two-photon (11) Kaiser, W.; Garret, C. G. B. Phys. Rev. Lett. 1961, 7, 229-231. (12) Johnson, P. M.; Otis, C. E. Annu. Rev. Phys. Chem. 1981, 32, 139-157. (13) Dinkel, D. M.; Lytle, F. E. Anal. Chim. Acta 1992, 263, 131-136. (14) Denk, W.; Strickler, J. H.; Webb, W. W. Science 1990, 248, 73-76. (15) Xu, C.; Webb, W. J. Opt. Soc. Am. B 1996, 13, 481-491.
spectra.15-17 Accordingly, two-photon fluorescence excitation spectra in general cannot be deduced from one-photon spectra. Three-Photon Excitation. Molecular excitation can be achieved through simultaneous absorption of three photons.18 In this case, the measured fluorescence count rate in the thick sample limit for a Gaussian focus is
〈F(t)〉3 ≈ (1/3)φηCσ3g3
2λ2〈P(t)〉3 3wo2h3c3
(3)
where σ3 is the three-photon excitation cross section (in units of distance6 × time2/photon2), wo is the 1/e2 radius of the focal spot, and g3 ∝ (fτ)-2. The factor, 1/3, is necessary for η to assume the standard definition of fluorescence quantum yield. Because the rate of three-photon-excited fluorescence depends strongly on the measured quantities, 〈P(t)〉, τ, and wo, relatively small errors in the measurement of these values can result in large miscalculations of three-photon excitation cross sections. EXPERIMENTAL SECTION Excitation Source. Two separate Ti-sapphire laser systems were used to accumulate data. In one, the multiline output of a Spectra Physics large-frame argon ion laser (BeamLok) was used to pump a Spectra Physics Ti-sapphire laser (Tsunami) operating at ∼80 MHz. The excitation intensity was controlled by attenuating the laser output with an achromatic half-wave plate/polarizer pair. In the other laser system, a Coherent 8 W argon ion laser (Innova 300) was used to pump a Coherent Ti-sapphire laser (Mira 900) operating at 76 MHz. Excitation intensity and stability were controlled using a Pockels cell (Conoptics, LASS II). For both systems, pulse widths were monitored with an intensity autocorrelator (Femtochrome, FR103) before the objective, and the Ti-sapphire lasing wavelength was determined using a spectrometer with a resolution of e1 nm. The temporal pulse width after the objective (at the sample) was not directly measured. Two different mirror sets were used in the Tsunami: one for measurements performed at wavelengths longer than 950 nm, and another for data taken between 840 and 920 nm. A single mirror set was used in the Mira to collect data between 710 and 810 nm. Sample Excitation and Fluorescence Collection. For twophoton cross section measurements, a low numerical aperture (NA) objective (Zeiss 0.3 NA 10× Plan-Neofluar) was used to focus the Ti-sapphire output into the sample. A high NA objective (Zeiss 1.2 NA 40× water immersion) was used to excite threephoton-excited fluorescence from OPA-labeled glutamate. The diameter of the beam entering the back aperture of the objective was controlled using beam expanders. Laser powers measured at the sample ranged from ∼5 to 200 mW. Homogeneous solutions of fluorogen-labeled analytes were loaded in deep-well slides (∼0.5 mm depth) and were sealed beneath coverslips (no. 1 1/2) using wax. The position of the slide with respect to the objective was controlled by mounting the slide on a micrometer-driven translation stage. With this setup, essentially all two-photon excitation was confined to the solution. (16) McClain, W. M.; Harris, R. A. In Excited States; Lim, E. C., Ed.; Academic Press: New York, 1977; pp 1-56. (17) Xu, C., Cornell University, personal communication, 1995. (18) Singh, S.; Bradley, L. T. Phys. Rev. Lett. 1964, 12, 612-614.
Analytical Chemistry, Vol. 68, No. 10, May 15, 1996
1779
Figure 1. Reactions of (a) fluorescamine, (b) naphthalene-2,3dicarboxaldeyde, and (c) o-phthaldialdehyde with neurotransmitters (RNH2).
The dimensions of the excitation volume produced by a low NA focus could be determined by translating the beam waist from within the coverslip to within a sample well containing fluorescent solution. By assuming a Gaussian focal volume, the Raleigh distance, zr, could be calculated from the observed change in fluorescence versus distance translated; the radius of the focal spot, wo, is set by a direct geometrical relationship between zr and wo. Fluorescence from the sample was collected through the excitation objective and was split off at 90° using a dichroic mirror (Chroma Technology, 650DCXR), which passes wavelengths greater than ∼690 nm (g90% transmission) and reflects light between ∼410 and ∼620 nm (g95% reflectivity). One or two BGG22 blue glass filters (Chroma) were used to further block Rayleigh and Raman scatter, and the signal was measured using a photomultiplier tube (Hamamatsu, HC125-02) connected to a discriminator/photon counter (Stanford Research Systems, SR400). One-photon absorption spectra were measured using a diode array spectrophotometer (Hewlett-Packard, 8451A). Sample Preparation. All reagents and standards were obtained from Sigma except where noted and were used without further purification. Spectroanalyzed acetone (99.7%, Fisher) and HPLC-grade methanol (100.0%, J.T. Baker) were used to prepare fluorogen solutions. 5-(2-Carboxyphenyl)-5-hydroxy-3-phenyl-2-pyrrolin-4-one (CPP) derivatives of peptides (Figure 1a) were generated by adding fluorescamine in acetone (90 µL) to solutions of leucine enkephalin or bradykinin in 10 mM borate buffer (pH 9.2, 210 µL) while rapidly vortexing the peptide solution. Leucine enkephalin was reacted at 90 µM with 1.1 mM fluorescamine and was diluted to 40 µM in borate buffer before measurement. Bradykinin was reacted at 0.72 µM with 0.55 mM fluorescamine and was diluted to 0.36 µM before analysis. The N-substituted 1-cyanobenz[f]isoindole (CBI) derivative of glycine (Figure 1b) was produced by adding 10 mM NaCN (100 µL, Aldrich) to amino acid in 8 mM borate buffer (pH 9.1, 500 µL), followed by 1 mM NDA (400 µL, Research Organics) in methanol. The concentration of glycine in the reaction was 40 µM for samples used to measure cross sections. The reaction was allowed to proceed for ∼0.5 h before measurement. 1780 Analytical Chemistry, Vol. 68, No. 10, May 15, 1996
Figure 2. Experimentally measured two-photon excitation cross sections for (a) CBI-glycine and (b) CPP-enkephalin. For reference, the relevant one-photon absorbance spectra are plotted at twice the actual wavelengths (dotted traces) using an arbitrary y-scale. Estimated errors in cross sections are (15% (relative values) and (50% (absolute values).
OPA derivatization (Figure 1c) was performed by adding OPA reagent (100 µL, Sigma complete reaction solution P-0532) to 2.5 mM glutamate in H2O (100 µL). Reagent and amino acid were allowed to react for 60 s before the reaction was quenched with pH 7.25 phosphate buffer (1 mL, 25 mM). Fresh samples were prepared for each wavelength measurement. For all samples, a blank was prepared by substituting the appropriate buffer solution in the reaction for the peptide or amino acid. RESULTS AND DISCUSSION Cross-Section Measurements. (i) Two-Photon Excitation Spectra. Figure 2, parts a and b, displays the measured twophoton excitation cross sections for CBI-glycine and CPPenkephalin, respectively. In assigning absolute values to determined cross sections, we have assumed that the quantum efficiencies of fluorescence, η, are approximately equal to literature values19,20 for single-photon-excited fluorescence (ηCBI-gly ≈ 0.75, (19) Matuszewski, B. K.; Givens, R. S.; Srinivasachar, K.; Carlson, R. G.; Higuchi, T. Anal. Chem. 1987, 59, 1102-1105. (20) de Bernardo, S.; Weigele, M.; Toome, V.; Manhart, K.; Leimgruber, W.; Bo ¨hlen, P.; Stein, S.; Udenfriend, S. Arch. Biochem. Biophys. 1974, 163, 390-399.
ηCPP-peptide ≈ 0.25) and do not depend on the excitation wavelength. One-photon absorption spectra for the two compounds are plotted at twice the actual wavelength of absorption (dotted lines) and reveal a strong qualitative agreement between the one- and twophoton spectra. Fluorescence from a standard fluorescein slide (pH > 11) was monitored to calibrate the cross section measurements with respect to the established two-photon cross sections of fluorescein.15 This procedure minimizes errors caused by uncertainty in the values of φ and τ at the sample. The fluorescence-to-counts conversion efficiencies for fluorophores were taken into account when calibrating cross section values using the fluorescein standard. The maxima of the two-photon excitation cross sections for both species are ∼1 Go¨ppert-Mayer (1 GM t 10-50 cm4 s/photon), a value similar to the reported peak two-photon cross sections for dansyl hydrazine and Lucifer Yellow.15 Although these cross sections are significantly lower than the peak cross sections of fluorescein (∼40 GM) and rhodamine B (∼300 GM),15 the fluorogenic species examined in these studies could be excited efficiently at moderate excitation powers, even when the laser beam was focused with a low NA objective. Significant levels of fluorophore bleaching were achieved at excitation powers in the 100-200-mW range for a low NA focus. The reproducibility of these data was evaluated by two means. First, after measuring fluorescence from a sample, the sample slide was removed and then returned to the sample holder for repeat measurements. Second, after acquiring data at a wavelength setting, the Ti-sapphire laser was tuned to other wavelengths and then was retuned to the original wavelength for remeasurement of the sample. The determined variability in the results at individual wavelengths was approximately (10%. Verification that the observed fluorescence of these species results from two-photon excitation is shown in Figure 3. For twophoton-excited fluorescence, a plot of log(fluorescence) vs log(excitation power) should generate a line with slope m ) 2, assuming that saturation, photobleaching, and intersystem crossing do not significantly deplete the population of molecules in S0. In Figure 3, the determined slopes for CBI-glycine (844 nm) and CPP-enkephalin (744 nm) are 2.02 and 2.01, respectively. All measurements for CPP-enkephalin yielded slopes in the range 1.97-2.01, and excitation of CBI-glycine produced slopes between 1.98 and 2.10. A slope of 2.27 was measured for excitation of CBI-glycine at 713 nm (cross section data not shown), possibly revealing a combination of two- and three-photon excitation. Although this possibility has not been investigated in detail, a strong one-photon UV absorption peak is observed for NDA derivatives near 250 nm. Three-photon excitation via this UVaccessible state is expected to cause a greater positive deviation of the measured slope when the two-photon cross section is small, as is the case at 713 nm. The uncertainties in relative two-photon cross section values arise from a combination of factors, including measurement reproducibility and deviations from a true power-square law dependence of fluorescence. For all data in Figure 2, the estimated uncertainty in the relative cross section measurements is approximately (15%. The errors in the absolute cross section values depend, in addition, on uncertainties in the published twophoton cross sections for fluorescein and the fluorescence quantum yields and, consequently, are likely (50%. Two-photon
Figure 3. Demonstration that fluorescence from NDA and fluorescamine derivatives is generated by two-photon excitation and that fluorescence from OPA-labeled glycine is generated by three-photon excitation. Log(fluorescence) vs log(excitation power) plots for CBIglycine ( 0, λex ) 844 nm) and CPP-enkephalin (O, λex ) 744 nm) yield slopes of 2.02 and 2.01, respectively. The slope of the OPAglycine plot (], λex ) 994 nm) is 2.99. Because these plots were constructed from raw data (no corrections were made for differences in τ or g2 at the different wavelengths), a comparison of relative twophoton cross section values cannot be made from this graph. For all cross section data for CBI and CPP derivatives shown in Figure 2, the measured slopes of log(fluorescence) vs log(excitation power) plots fall in the range 1.97-2.10. The determined slopes for OPAglycine plots fall in the range, 2.85-3.23 (four excitation wavelengths).
fluorescence action cross sections can be obtained from these data by multiplying the determined excitation cross sections by the assumed quantum yields of fluorescence. (ii) Three-Photon Excitation of OPA-Glutamate. The OPA derivative of glutamate was examined for three-photon-excited fluorescence between 965 and 1012 nm. Because three-photonexcited fluorescence rates depend on the sharpness of focus (as shown by the reciprocal square of the focal radius in eq 3), a 1.2 NA water-immersion objective (with a nearly filled back aperture) was used to focus the Ti-sapphire laser beam into the sample slide (wo < 0.5 µm). In addition to significantly decreasing wo, the use of a high NA objective improved the fluorescence collection efficiency by a factor of ∼15 over that obtained with the 0.3 NA objective. The determined three-photon cross sections were constant in this wavelength range to within a factor of ∼2, with an absolute value, σ3, between 10-84 and 10-85 cm6 s2/photon2. Absolute three-photon cross section values are estimated to be accurate to within ∼1 order of magnitude. The error in these measurements arises primarily from uncertainties in the focal radius and in the temporal pulse width at the sample. Because reference standards with accurately determined three-photon cross sections do not exist, direct measurement of wo and τ will be required to reduce cross section uncertainties. The plot of log(fluorescence) vs log(excitation power) for OPA-glutamate at λex ) 994 nm (Figure 3) yields the expected slope for three-photon-excited fluorescence, m ≈ 3. The deterAnalytical Chemistry, Vol. 68, No. 10, May 15, 1996
1781
mined slopes of log-log plots at three other excitation wavelengths fell in the range 2.85-3.23. Limits of Detection. (i) Fluorescence Measurements Using Deep-Well Samples. Measurements of CPP-bradykinin were conducted to assess the limits of detection of MPE fluorescence detection. Although not directly examined here, issues involved in these studies are applicable to multiphoton excitation of other fluorogenic labels. A moderately low concentration solution of CPP-bradykinin (reacted at 0.72 µM, diluted to 0.36 µM before measurement) was excited by a 754-nm pulse train focused using a 0.3 NA objective (with a nearly filled back aperture). The excitation volume21 with this arrangement was determined to be ∼300 fL, with a beam waist wo ≈ 1.7 µm, and a Rayleigh length along the propagation axis zr ≈ 12 µm. Although higher NA objectives can produce significantly smaller two-photon probe volumes that more effectively limit interference from background,21 these dimensions were considered to be particularly relevant for potential use with microcolumn separations. The probe volume was estimated to contain ∼0.1 amol of CPPbradykinin (6 × 104 molecules), with a mean diffusional refreshing time22 τd ≈ 5 ms. Using this estimation of molecular refreshing time, it is possible to calculate mass sensitivities from the bulk solution fluorescence count rate measurements performed in these experiments. The signal was measured over a range of excitation powers. At low powers, the measured fluorescence rate scaled quadratically with excitation power, as expected for a two-photon process. At the highest incident excitation power used (150 mW), the fluorescence count rate was ∼20% lower than predicted by simple extrapolation from the low-power measurements, probably as a result of photobleaching and intersystem crossing. At this highest power, the net count rate for CPP-bradykinin ([sample count rate] - [blank count rate]) was ∼2.1 × 104 s-1, which corresponds to 100 counts in the diffusonal refreshing time, or ∼0.002 count/ molecule. The fluorescence conversion efficiency, φ, was estimated to be ∼4 × 10-4. The 0.3 NA objective used in these experiments collects a solid angle corresponding to only 1.3% of the total fluorescence (taking into account the index of refraction of the sample). By adopting an alternative configuration in which a separate, 1.3 NA objective is used to collect fluorescence, it would be possible to increase the measured signal ∼20-fold (making φ ≈ 0.01). In this case, the signal would be nearly 0.05 count/molecule. To obtain an estimate of the detection limit, a determination of the magnitude of the background must be made. For fluorescence analysis of fluorogen-labeled derivatives, this background can be classified as one of two varieties: (1) inescapable background inherent to MPE fluorescence measurements and (2) background that can be fractionated from analytes using a separation procedure. The latter category is responsible for most of the background when measuring fluorescence from our deepwell samples but does not inherently influence the detection limits possible with this technique. (ii) Inherent Background. For useful applications, the background contribution of Rayleigh and Raman scatter must be negligible, even at sample concentrations 100-fold lower than those used in this work. For λex ) 750 nm, we estimate (from component specifications) that a detection system composed of a (21) Mertz, J.; Xu, C.; Webb, W. W. Opt. Lett. 1995, 20, 2532-2534. (22) Berland, K. M.; So, P. T. C.; Gratton, E. Biophys. J. 1995, 68, 694-701.
1782
Analytical Chemistry, Vol. 68, No. 10, May 15, 1996
650DCXR dichroic mirror, 5 mm of BGG22 colored glass, and the HC125-02 detector should convert ∼10-20-10-25 of the collected scatter photons into measured counts. Even in the case where all the incident excitation power (∼1018 photons s-1) is backscattered into the objective, the excitation light should produce e0.01 count s-1. In these experiments, 180 mW of 750nm CW light focused into an H2O sample slide resulted in a negligible addition (∼3 counts s-1) to the dark count rate (∼90 counts s-1). This result clearly demonstrates that background caused by scatter does not influence the limit of detection in these measurements. A more significant source of background in these experiments is radiation generated by nonlinear phenomena. Potential sources of background include multiphoton-excited fluorescence from solution impurities, second-harmonic generation by the dichroic mirror,17 and, at extremely high intensities, solvent breakdown23 and white light generation.24 We examined background produced by focusing mode-locked 750-nm light into a slide containing HPLC-grade H2O. The average background count rate using a filled 0.3 NA objective was equal to the fluorescence signal from ∼4000 molecules of CPP-enkephalin (∼20 nM) and scaled approximately as [excitation power]2. Photon shot noise with this background level places an ultimate constraint on the limit of detection of MPE fluorescence. For excitation with 150 mW of 750-nm light focused using a 0.3 NA objective and detection with a separate 1.3 NA objective, the expected background is ∼200 ( 15 counts in the diffusional refreshing time of CPP-bradykinin. This excitation/collection geometry should yield ∼0.05 count/ molecule, and assuming that no additional background exists, ∼300 molecules of CPP-bradykinin are needed to achieve a signal-to-noise ratio of 1 (i.e., a signal of 15 counts). The resulting mass detection limit (∼103 molecules) is ∼100-fold better than those reported for single-photon-excited fluorescence of fluorogenlabeled amines.5 Fluorescent solution impurities appear to be the main source of background in HPLC-grade “pure” water samples. Addition of a 420-nm long-pass filter to the light collection pathway (which blocks second-harmonic radiation) had little effect on the background count rate, and the excitation intensities reached with this configuration are far below those needed to induce solvent breakdown or superbroadening of the laser pulse. Furthermore, the background count rate was larger for sample slides measured on the second day after preparation, suggesting that fluorescent impurities had leached from the cover slip, deep-well slide, or sealing wax. Approaches for minimizing background fluorescence are currently being investigated. (iii) Separable Background. To achieve the lowest possible mass detection limits with MPE fluorescence detection, a separation procedure will be necessary to distinguish analyte fluorescence from reagent fluorescencesthe largest overall background in measurements performed on deep-well samples. In the fluorescamine reaction mixture, for example, the fluorescence rate from unreacted or hydrolyzed reagent species was approximately equal to the fluorescence rate from 300 nM CPP-bradykinin (∼15fold greater than the background produced by HPLC-grade H2O samples), and a similar background level was observed in MPE fluorescence measurements of CBI-glycine. Fortunately, micro(23) Stuart, B. C.; Feit, M. D.; Rubenchik, A. M.; Shore, B. W.; Perry, M. D. Phys. Rev. Lett. 1995, 74, 2248-2251. (24) Smith, W. L.; Liu, P.; Bloembergen, N. Phys. Rev. A 1977, 15, 2396-2403.
column fractionation of components responsible for background fluorescence in the fluorescamine reaction mixture has been demonstrated previously,4 allowing measurement of peptides derivatized at the nanomolar level. Future Utility of MPE Fluorescence Detection. The advantages of MPE fluorescence detection of fluorogenic derivatives are of particular value in analysis of neuronal microenvironments. Olefirowicz and Ewing used a micropositioned 2-µm-i.d. capillary to collect subpicoliter neurotransmitter samples from individual neurons.25 Samples then were electrophoretically separated in the same capillary, and isolated components were analyzed with electrochemical detection. Used in a similar procedure, MPE fluorescence detection may provide a general approach for measurement of individual secretory vesicles and secretion at individual release sites on cultured neurons or neuronal cell lines. Although an additional analysis steps fluorogenic derivatizationswill be necessary to render analytes fluorescent, on-column reaction techniques can be used to achieve fluorescence derivatization of nanoliter and subnanoliter biological samples without prohibitive dilution or sample loss.2,26,27 Furthermore, low-concentration fluorogenic analysis of amino acids, peptides, and proteins in nanoliter volumes has been accomplished by others in combination with microcolumn separations. Jorgenson and Nickerson28 reported a detection limit for OPA-labeled myoglobin of 10-8 M after separation with capillary electrophoresis, and Sweedler and co-workers4 have demonstrated detection limits for electrophoretically separated CPP-neuropeptides in the low nanomolar range. Although MPE fluorescence measurements of fluorogen derivatives will offer the best detection limits when used with a separation procedure, applications that do not rely on separations also should benefit significantly from multiphoton excitation. Microscopic histochemical localization in intact cultured cells, for example, can be performed by labeling cellular components with fluorogens.29,30 In the absence of cellular autofluorescence in overlapping spectral regions, the MPE fluorescence detection limit for such applications is estimated to be
