Effects of Molecular Oxygen on Multiphoton-Excited Photochemical

Anal. Chem. 2000, 72, 3821-3825

Effects of Molecular Oxygen on Multiphoton-Excited Photochemical Analysis of Hydroxyindoles Michael L. Gostkowski, Theodore E. Curey, Eric Okerberg, Tai Jong Kang, David A. Vanden Bout,* and Jason B. Shear*

Department of Chemistry and Biochemistry & The Institute for Cellular and Molecular Biology, University of Texas, Austin, Texas 78712

We have examined the effects of dissolved molecular oxygen on multiphoton-excited (MPE) photochemical derivatization of serotonin (5HT) and related cellular metabolites in various buffer systems and find that oxygen has a profound effect on the formation efficiency of visibleemitting photoproducts. Previously, end-column MPE photoderivatization provided low mass detection limits for capillary electrophoretic analysis of hydroxyindoles, but relied on the use of Good’s buffers to generate highsensitivity visible signal. In the present studies, visible emission from 5HT photoderivatized in different buffers varied by 20-fold under ambient oxygen levels but less than 2-fold in the absence of oxygen; oxygen did not significantly alter the photoproduct excited-state lifetime (∼0.8 ns). These results support a model in which oxygen interferes with formation of visible-emitting photoproducts by quenching a reaction intermediate, an effect that can be suppressed by buffer molecules. Deoxygenation of capillary electrophoresis separation buffers improves mass detection limits for 5-hydroxyindoles fractionated in 600-nm channels by approximately 2-fold to e30 000 molecules and provides new flexibility in identifying separation conditions for resolving 5HT from molecules with similar electrophoretic mobilities, such as the catecholamine neurotransmitters. Analysis of neuronal microenvironments requires the implementation of high-sensitivity techniques for characterizing lowvolume mixtures of neurotransmitters and their metabolites.1-3 Various detection strategies, including amperometry, mass spectrometry, and laser-induced fluorescence, have been adapted for use with microcolumn separation techniques to achieve attomole detection limits of biological components.4-7 Recently, we have (1) Karin, P.; Schroeder, J.; Wightman, R. W. Anal. Chem. 1994, 66, 45324537. (2) Olefirowicz, T. M.; Ewing, A. G. Anal. Chem. 1990, 62, 1872-1876. (3) Lillard, S. J.; Yeung, E. S. Anal. Chem. 1996, 68, 2897-2904. (4) Kennedy, R. T.; Oates, M. D.; Cooper, B. R.; Nickerson, B.; Jorgenson, J. W. Science 1989, 246, 57-63. (5) Valaskovic, G. A.; Kelleher, N. L.; McLafferty, F. W. Science 1996, 273 (5279), 1199-1202. (6) Douglas, G. S.; Ewing, A. G. Anal. Chem. 1995, 67, 58-64. (7) Floyd, P. D.; Moroz, L. L.; Gillette, R.; Sweedler, J. V. Anal. Chem. 1998, 70, 2243-2247. 10.1021/ac000278+ CCC: $19.00 Published on Web 07/19/2000

© 2000 American Chemical Society

explored the use of multiphoton-excited (MPE) fluorescence as a high-sensitivity detection strategy for capillary electrophoresis (CE) and have demonstrated the application of MPE photochemistry as a derivatization scheme to enhance the luminescence properties of serotonin (5-hydroxytryptamine, 5HT) and related 5-hydroxyindoles.8,9 Mass detection limits ranging from ∼70 to ∼130 zmol (42 000 molecules) were achieved using MPE photochemistry to analyze 5HT, 5-hydroxytryptophan (5HTrp) and 5-hydroxyindole-3-acetic acid (5HIAA) after electrophoretic fractionation in submicrometer capillaries. These values are at least severalfold lower than CE detection limits reported for measuring unmodified hydroxyindoles with other detection strategies3,10,11 and approach levels that will be needed to analyze individual synaptic vesicles.12 In this photochemical analysis strategy, multiple near-IR photons are absorbed to initiate the rapid transformation of hydroxyindoles into compounds that emit in the visible spectral region (λmax ≈ 500 nm) when excited with additional near-IR photons. In the initial studies of this process, visible emission was measured as a function of femtosecond pulsed 830-nm laser intensity (I) and was found to scale as I 6 at relatively low laser intensitiessa result that demonstrated a total of six photons are required to generate and probe the 5HT photoproduct.13 To deconvolve photoproduct generation and measurement steps, experiments were performed under diffusional non-steady-state conditions: a several-microsecond train of high-intensity laser pulses generated photoproduct in the vicinity of the focal point, the laser intensity was rapidly switched to a lower level that could excite photoproduct emission without generating additional photoproduct, and visible emission was probed as the photoproduct diffused from the focal region into bulk solution. Signal was measured as functions of both the initial photoproduct-generating intensity and the subsequent probe intensity, revealing that the two independent steps depend on absorption of four photons and (8) Gostkowksi, M. L.; McDoniel, J. B.; Wei, J.; Curey, T. E.; Shear J. B. J. Am. Chem. Soc. 1998, 120, 18-22. (9) Gostkowski, M. L.; Wei, J.; Shear, J. B. Anal. Biochem. 1998, 260, 244250. (10) Wallingford, R. A.; Ewing A. G. Anal. Chem. 1989, 61, 98-100. (11) Timperman, A. T.; Oldenburg, K. E.; Sweedler, J. V. Anal. Chem. 1995, 67, 3421-3426. (12) Bruns, D.; Jahn, R. Nature 1995, 377, 62-65. (13) Shear, J. B.; Xu, C.; Webb, W. W. Photochem. Phtotobiol. 1997, 65, 931936.

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000 3821

two photons, respectively. Power scaling measurements of 5HT UV fluorescence demonstrated that excitation to the lowest excited electronic state can be achieved by absorption of three 830-nm photons (∼1.5-eV photon-1). Because CE does not suffer from most of the band-broadening phenomena that degrade chromatographic techniques, efficiencies of several hundred-thousand theoretical plates often can be achieved. Nevertheless, even CE can face substantial challenges in resolving analytes that are structurally similar or maintain no net charge. By modifying various characteristics of the buffer system, however, it is usually possible to identify conditions sufficient to achieve the desired separation. Borate anions, for example, can alter the relative electrophoretic mobilities of components by forming negatively charged complexes with hydroxylated analytes, including carbohydrates and catechols.10,14,15 Unfortunately, efficient photochemical detection of hydroxyindoles relies on the use of Good’s buffers (e.g., Hepes, MOPS),13 a fact that imposes significant constraints on strategies available for effecting resolution when these species are present in complex mixtures. In the current study, we have examined the effects of various solution parameters on 5HT photoproduct emission, including molecular oxygen content and the presence of different pH buffers. In addition to offering insights into the mechanisms of 5HT photochemistry, the removal of dissolved oxygen has two important practical effects. First, hydroxyindole mass detection limits using optimized buffer conditions improve by factors of 1.5-2.5, with absolute limits at or below 30 000 molecules for all tested species. Second, the importance of oxygen is shown to be highly buffer dependent. For the set of optical parameters used in these studies, deoxygenation of Good’s buffers approximately doubles 5HT visible emission, while removal of oxygen from other common buffer systems can provide much larger improvements. Measurements of the photoproduct excited-state lifetime in ambient and deoxygenated solutions support a model in which the primary effect of oxygen is to modify the formation efficiency of the photoproduct, not its fluorescence quantum yield. The signal enhancements achieved in deoxygenated solutions expand the applicability of MPE photochemistry to high-sensitivity analysis in all buffer systems examined in this work. The advantages of greater buffer flexibility are demonstrated by characterizing solutions containing 5HT and the neurotransmitters epinephrine and dopamine. EXPERIMENTAL SECTION. Chemicals and Materials. Sodium phosphate monohydrate, sodium citrate, sodium acetate, sodium bicarbonate, sodium acetate, and sodium borate decahydrate were purchased from EM scientific (Gibbstown, NJ). All other chemicals, including Hepes free acid and Tris hydrocholoride, were purchased from Sigma Chemical Co. (St. Louis, MO). All chemicals were used as received. Water used to prepare buffers was purified using a Barnstead UV water system, and all buffers used in CE separations were filtered with 0.2-µm pore-size cellulose acetate membranes (Osmonics, Livermore, CA). Separation capillaries were obtained from Polymicro, Inc. (Phoenix, AZ). (14) Susumu, H. J. Chromatogr., A 1996, 720 (1 + 2), 337-351. (15) Kaneta, T.; Tanaka, S.; Yoshida, H. J. Chromatogr. 1991, 538 (2), 385391.

3822

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

Laser Systems. A Coherent Mira 900-B femtosecond modelocked titanium:sapphire (Ti:S) oscillator pumped by a 5-W Coherent Verdi solid-state frequency-doubled Nd:vanadate (Nd: YVO4) laser was used for all studies, except for lifetime measurements. The Mira was operated at 750 ((5) nm, a wavelength region that typically yields pulses with a sech2 temporal shape (pulse width ≈100-150 fs; repetition rate ≈76 MHz). To achieve desired powers, the Mira output was attenuated using a half-wave plate/polarizer pair. Excited-state lifetime measurements were made using a Spectra Physics Tsunami femtosecond mode-locked Ti:S oscillator pumped by a 5-W Spectra Physics Millenia frequency-doubled Nd:YVO4 laser. The Tsunami output characteristics were similar to those of the Mira. Capillary Electrophoresis with Multiphoton Excited Fluorescence. Separations were performed in 600-nm-i.d. or 5-µmi.d. channels using fields of 900-950 V/cm ((30-kV power supply, Spellman Inc.). The overall detection system is similar to that reported earlier.9 In brief, the Ti:S beam is directed through a series of beam splitters and dichroic mirrors into the back aperture of the microscope objective. The objective focuses the light through a cover slip a short distance (∼10-100 µm) into a plastic cuvette that serves as the outlet electrolyte reservoir (held at electrical ground). The separation capillary enters the cuvette through a septum on one side and terminates at the beam focus near the opposing side of the cuvette. The cuvette is secured to a 3-axis (XYZ) translation stage. In addition, the capillary can be positioned relative to the cuvette using a Z-axis (optical axis) translation stage and a gimbal (2-axis) tilt control. For CE and cuvette studies, diffraction-limited focusing is achieved by overfilling the back aperture of the microscope objective (Zeiss Fluar 100×, 1.3 NA oil immersion, infinity corrected). Fluorescence produced by analytes migrating from the channel (or in bulk solution) is collected by the excitation objective and is reflected at 90° from the beam path using dichroic mirrors at two separate locations. The dichroic mirror closest to the objective reflects UV fluorescence in the approximate wavelength range, 280-380 nm, and transmits wavelengths greater than ∼430 nm. Two UV band-pass filters are used to further isolate UV fluorescence, and the signal is measured using a photomultiplier tube (Hamamatsu, model HC-125) connected to a two-channel photon counter (Stanford Research Systems, Sunnyvale, CA, model SR400). Visible fluorescence at wavelengths shorter than 620 nm (transmitted through the UV dichroic) is reflected by a second dichroic mirror. A colored glass filter and 2 cm of 1 M CuSO4 solution are used to isolate visible emission from residual UV fluorescence and laser scatter. Signal from this visible channel is measured with a second photomultiplier connected to the photon counter. The two channels of the photon counter transfer data through a GPIB interface to a Macintosh running LabView-based software. A single electrophoresis separation thus yields two data tracessone for UV emission and one for visible emission. Lifetime Measurements. Excited-state lifetimes were measured using standard time-correlated single-photon counting (TCSPC) techniques to electronically determine the delay between the excitation pulse and emission of a fluorescence photon.16 For these measurements, a Ti:S laser (Spectra Physics 5W Millenia(16) O’Connor, D. O.; Philips, D. Time-Correlated Single-Photon Counting; Academic Press: London, 1984.

Figure 1. Comparison of visible emission produced by multiphoton-excited photochemistry of 3 µM serotonin in common buffer systems. Bars in the front row represent visible emission generated in solutions containing ambient levels of molecular oxygen; visible emission in deoxygenated solutions is shown in the back row. The relative amplifications achieved by deoxygenation (i.e., the ratio of 5HT signal in deoxygenated solution to that in ambient oxygen solution; second-to-back row) reveal that deoxygenation has greater effects in buffer systems that initially support the lowest levels of visible emission. All buffers were prepared at 20 mM, and except where noted, solutions were adjusted to pH 7.0. Measurements were made using ∼100 mW of Ti:S laser light focused to a diffraction limit in cuvette solutions. Hep ) sodium Hepes; Cit ) sodium citrate; Tris ) Tris hydrochloride; Bor ) sodium borate decahydrate; Carb ) sodium bicarbonate; Acet ) sodium acetate; H2O ) unbuffered; Phos ) sodium phosphate monohydrate.

pumped Tsunami) was operated at 780 nm, and the output was attenuated to ∼30 mW to reduce the likelihood that more than one fluorescence photon would be measured per laser pulse. At this power level, the fluorescence count rate was maintained at e2 MHz, or ∼1 count/40 laser pulses. The excitation pulse was detected using a fast photodiode, and fluorescence was measured by an MCP-PMT (Hamamatsu, R1564-07). A PC-Card with TCSPC electronics (Picoquant TimeHarp-100) determined the delay between these two signals. The instrument response function (IRF) was determined to be ∼200 ps by examining laser scatter from nonfluorescent solutions. To determine fluorescence lifetimes, measured decays were compared to convolutions of the IRF with single- and double-exponential decays using standard weighting procedures.17 Diffusional Non-Steady-State Measurements. To assess the power scaling of 5HT photoproduct emission, diffusional nonsteady-state measurements were performed. A Pockel’s cell (Conoptics, Danbury, CT, model 350-50) driven by a delay generator (Stanford Research Systems, model DG 535) was used to rapidly switch the laser excitation intensity to a high (photoderivitization) level for brief periods (25 µs), thus constraining efficient production of photoproducts to well-defined “points” in time. The photon dependence for fluorescence excitation of photoproducts was determined by measuring emission as a function of probe intensity following photoderivitization. A zerodeadtime multichannel scaler (model SR430, Stanford Research (17) Grinvald, A.; Steinberg, I. Z. Anal. Biochem. 1974, 59, 583-598.

Systems) was used to digitize photomultiplier signal and to sum data from multiple photoderivitization cycles. RESULTS AND DISCUSSION We have noted previously that, at millimolar 5HT concentrations, visible emission was almost equally intense in oxygenated Hepes, phosphate, and unbuffered (self-buffered) solutions. At somewhat lower 5HT concentrations, however, the intensity of visible emission was seen to be strongly buffer-dependent: in Hepes solutions; visible emission scaled linearly with 5HT concentration, while in non-Good’s systems signal diminished more rapidly (signal ∼[5HT]1.3 in phosphate). This rapid falloff in signal complicates quantitation and yields poor concentration detection limits, especially in CE analyses of ultralow-volume samples. In the current studies, we have examined 5HT photochemistry in a broad selection of common buffers and have found substantial variation in the capacity of these systems to support visible emission at low (3 µM) 5HT concentration. Serotonin visible emission in Hepes buffer was ∼20-fold greater than in phosphate buffer, the lowest sensitivity system we examined. Hepes, citrate, Tris, acetate, and borate all enhance visible emission relative to unbuffered 5HT solutions, while phosphate solutions appear to mildly suppress visible emission (Figure 1, front row). Remarkably, removal of molecular oxygen enhances visible emission for all low-concentration 5HT solutions, with the largest signal amplifications attained in the buffer systems that initially supported the lowest visible emission (Figure 1, back two rows). Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

3823

Figure 2. Fractionation of hydroxyindole standards in a 600-nmi.d. electrophoresis channel using 15 mM deoxygenated Hepes (pH 7.0) as the separation medium. Measurement of serotonin (5HT), 5-hydroxytryptophan (5HTrp), and 5-hydroxyindole-3-acetic acid (5HIAA) is achieved at the capillary outlet using multiphoton-excited photochemistry; mass detection limits for all three species are e30 000 molecules. The sample concentration of each species is 10 µM, and injection volumes range from ∼200 to 400 fL.

Although the relative signal measured in different buffer systems is strongly dependent on various experimental parameters, including the concentrations of buffer and analyte, the excitation intensity, and the degree to which diffraction-limited focusing is achieved, the trends shown in Figure 1 are representative of those seen for other experimental conditions. Of the buffer systems examined in these studies, Hepes offers the most sensitive hydroxyindole analysis both in ambient oxygen solutions and in a deoxygenated environment. Figure 2 shows a separation of a standard mixture containing 5HT, 5HTrp, and 5HIAA using deoxygenated 15 mM Hepes (pH 7.0) as the separation buffer in a submicrometer electrophoresis channel. Electrokinetic sample injection volumes for these species range from 200 fL for 5HIAA to 400 fL for 5HT. Mass detection limits based on the sample necessary to produce a signal 3 times the rms noise level are 28 000 molecules for 5HT and 30 000 molecules for both 5HTrp and 5HIAA. Compared to our previous studies using photochemically generated visible emission to detect hydroxyindoles after CE fractionation,9 these values represent improvements of ∼1.5-fold for 5HT and 5HTrp and ∼2.5-fold for 5HIAA. Enhancements of 5HT photochemical signal in deoxygenated Hepes are partially offset by a somewhat larger MPE background produced by deoxygenation of buffer. Although the source of this increased background has not been examined in detail, diffusional non-steady-state measurements suggest that visible-emitting photoproducts can be generated from Good’s buffers. Although absolute detection sensitivity is maximized by using deoxygenated Hepes, the large enhancements in visible emission achieved in other deoxygenated buffers can offer increased flexibility in optimizing CE separation conditions. For example, when attempting to characterize mixtures containing 5HT and two catecholamine neurotransmitters (dopamine and epinephrine) in Hepes buffer, the similar electrophoretic mobilities of the analytes result in incomplete resolution (Figure 3a). By replacing Hepes buffer with a deoxygenated mixture of 15 mM borate and 15 mM phosphate (pH 7.0), baseline-resolved separation of 5HT, dopamine, and epinephrine is readily achieved (Figure 3b). In this hybrid 3824 Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

Figure 3. Demonstration of the increased flexibility for effecting separations in deoxygenated buffer solutions. In part a, a standard mixture of 5HT, dopamine (Dop), and epinephrine (Epi) are electrophoretically separated in a 5-µm-i.d. channel using 15 mM Hepes (pH 7.0) as the separation medium. Because only 5HT solutions yield strong MPE visible emission under these conditions, detection is accomplished using three-photon excitation to promote intrinsic UV emission from these species. By using a deoxygenated 15 mM borate/ 15 mM phosphate buffer as the separation medium (b), baseline resolution of all components is achieved. Moreover, in this buffer system, 5HT, Dop, and Epi all produce significant visible emission via multiphoton-excited photochemical reactions.

buffer system, differential complexation of the catecholamines with borate provides a mechanism for separation of these species from 5HT. Although it may be possible to achieve similar resolution of these species by incorporating borate in a Hepes buffer system, elimination of Good’s buffers also can improve the detectability of catecholamines. We found that dopamine and epinephrine in borate/phosphate solution undergo MPE photochemical transformations that yield visible-emitting productssreactions that are strongly suppressed in Hepes solutions. As a consequence, UV fluorescence was used to produce the electropherogram in Figure 3a while photochemically generated visible emission was monitored in Figure 3b. Catecholamine detection limits in the latter case are ∼4-fold better. Substantial challenges exist in elucidating a detailed molecular mechanism for 5HT photochemistry. When using a focused femtosecond Ti:S laser, efficient multiphoton excitation can be achieved only in femtoliter volumes. This size limitation makes it inherently difficult to generate sufficient photoproduct for analysis by traditional molecular identification strategies (e.g., NMR or mass spectrometry) even when irradiating low-volume cuvettes for several hours. In addition to the large dilution factor incurred in such experiments, it appears that the visible-emitting photoproduct is not thermally stable for extended periods. Nevertheless, our results offer important clues regarding possible photochemical pathways. The observation that detection sensitivity is nearly uniform in all deoxygenated systems is consistent with a model in which oxygen interferes with formation of a visible-emitting photoproduct by quenching a reaction intermediatesa process that can be suppressed to varying degrees by different buffers or by increasing the concentration of 5HT itself.

An alternate explanation for these results is that oxygen modifies the emission characteristics of the photoproduct, either by decreasing its fluorescence quantum yield or by compromising its photostability. If the fluorescent state of the photoproduct were quenched by oxygen (thus lowering the fluorescence quantum yield), then removal of oxygen should increase the excited-state lifetime. This result was not observed: serotonin photoproduct emission could be represented as a single-exponential decay with a lifetime, τ ≈ 0.80 ns, regardless of the buffer system or the oxygen state of the solution. Another mechanism by which oxygen could decrease the intensity of visible emission is to enhance the rate of photoproduct degradation (i.e., through photobleaching)s an effect that would decrease the steady-state concentration of visible-emitting photoproduct in the laser focal volume. In such a case, one would observe a deviation from the expected (intensity)2 scaling of photoproduct emission at much lower laser powers when oxygen is present than when it has been purged from solution. In diffusional non-steady-state measurements, we used up to 75 mW of Ti:S light to probe 5HT photoproduct generated by a high-intensity train of laser pulses and found that two-photon scaling of photoproduct fluorescence was maintained in Hepes, phosphate, and deoxygenated phosphate buffers. Thus, differences in steady-state visible emission observed for 5HT in these solutions are not caused by changes in the bleaching rate of the visibleemitting photoproduct. Together, the lifetime and power-scaling measurements indicate that enhancement of visible emission is achieved by changes in the efficiency of 5HT phototransformation and that oxygen likely quenches a relatively long-lived intermediate in the photoreaction pathway. The excited-state lifetime measurements also establish that a single photoproduct is responsible for most (or all) of the measured visible emission and that the observed visible signal represents fluorescence, rather than emission from a triplet state.

In these studies, we have examined the sensitivity of neurotransmitter photoderivatization in a variety of chemical environments and have demonstrated that molecular oxygen and pH buffers play important roles in the generation of hydroxyindole photoproducts. These results do not distinguish between simple collisional deactivation of a reaction intermediate by ground-state molecular oxygen or reactive deactivation by a variety of possible oxygen species (e.g., singlet oxygen). Further studies of the chemical mechanisms responsible for hydroxyindole photoproduct generation are being conducted in our laboratory; by gaining a more detailed understanding of the reaction process, we seek to increase both the sensitivity and the versatility of this powerful detection strategy. The current work extends the range of conditions that can be used to fractionate hydroxyindoles in combination with high-sensitivity end-column photochemical derivatization and identifies conditions necessary to perform analogous MPE photochemical analysis of catecholamines. Mass detection limits for 5HT and related hydroxyindoles are substantially improved by deoxygenation, making photochemical derivatization an increasingly attractive strategy to meet the severe requirements of analysis at the subcellular level. ACKNOWLEDGMENT We gratefully acknowledge support for this work from the Searle Scholars Program, the Texas Advanced Technology Program, the National Science Foundation, and the University of Texas Institute of Cellular and Molecular Biology.

Received for review March 9, 2000. Accepted June 6, 2000. AC000278+

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

3825