Channel Electrophoresis for Kinetic Assays - American Chemical Society

Channel Electrophoresis for Kinetic Assays. Yi-Ming Liu and Jonathan V. ... width dimension is used as a time axis by moving the outlet of the samplin...
1 downloads 0 Views 283KB Size
Anal. Chem. 1996, 68, 2471-2476

Channel Electrophoresis for Kinetic Assays Yi-Ming Liu and Jonathan V. Sweedler*

Department of Chemistry, University of Illinois at UrbanasChampaign, 600 South Mathews Avenue, Urbana, Illinois 61801

A rectangular channel electrophoresis system and a cylindrical sampling capillary combination allows chemical changes in nanoliter-volume samples to be monitored as a function of time. The electrophoretic microseparation is carried out in a rectangular channel with a 7-cm-long, 40-µm × 2.5-cm geometry and is coupled to a 50-µmi.d. cylindrical sample introduction capillary. The channel width dimension is used as a time axis by moving the outlet of the sampling capillary across the entrance of the separation channel. Detection of the separated analyte bands is achieved with laser-induced fluorescence and spatially resolved detection based on a charge-coupled device. The system is characterized with a series of fluorescein thiocarbamyl amino acid derivatives; limits of detection are 2 mA for a 1000-V separation potential. All injections are made by gravity. The sample is pulled into the sampling capillary with a syringe, and the outlet end of the filled capillary is placed into the entrance of the channel. The other end of the capillary is put into a vial containing water, which is outside of the plexiglass box. Materials in the capillary are injected into the separation channel by raising the vial 20 cm over the channel. After sample injection, the vial is lowered and the outlet of the sampling capillary moved away from the entrance of the channel with the use of the two-axis motion controller. Detection System. The fluorescence detection system has been designed to address the spatial resolution, speed and sensitivity constraints of a thin-channel separation system; most notably, it has the ability to collect fluorescence emission at high efficiency and spatially separate the different lanes of information. The excitation source is an Ar/Kr laser (Innova Spectrum 70, Coherent, Palo Alto, CA). This mixed gas Ar/Kr laser allows selection of excitation wavelengths ranging from 350 to 750 nm. The beam is expanded 12-fold with an in-house-constructed Galilean type beam expander, and an interference filter (Model 59335, Oriel, Stratford, CT) spectrally removes the laser tube background. The beam is then focused to a strip ∼0.5 mm wide with a 150-mm-focal-length cylindrical lens (CSX 150AR.14, Newport Corp., Irvine, CA). A custom-shaped aperture is placed between the filter and the cylindrical lens to improve the uniformity of illumination across the separation channel. A scientific cooled charge-coupled device (CCD) camera combined with two commercially available 35-mm camera lenses collects the fluorescence emission at a 45° angle relative to the

excitation beam (and at 90° relative to the channel). Before the first camera lens (105 mm, f/1.8, Nikon, Tokyo, Japan), the appropriate spectral interference filter is used to discriminate against Rayleigh scatter. The lens is positioned backwards so that its back focal point is the channel, and the lens is set to infinite focus, collimating the fluorescence emission from the channel. A second camera lens (50 mm, f/1.2, Nikon) is used to collect this light and image it on the surface of the CCD. The two lenses together produce a system demagnification of 2.1. The ∼2-cmwide detection window produces a 1-cm image across the 1-cmwide CCD sensing surface. Finally, a holographic notch filter (HNF-488-1.0, Kaiser Optical Systems, Inc., Ann Arbor, MI) is located before the CCD for additional spectral discrimination. The imaging detector is a TK512 back-side illuminated CCD (Tektronix, Beaverton, OR). It consists of 512 × 512-pixel elements, each measuring 20 µm × 20 µm. The spatial resolution at the channel corresponds to 42 µm/CCD element. The CCD readout is controlled by a CE 200 camera electronics unit and an AT 200 controller card (Photometrics Ltd., Tucson, AZ). CCD 9000 spectral acquisition software from Photometrics Ltd. is used for data acquisition. The two-dimensional electropherograms are processed by GRAMS/386 V3.0 software from Galactic Industries Corp. (Salem, NH) and IBM Visualization Data Explorer (IBM, Watson Center/Hawthorne, Yorktown Heights, NY) run on an IBM RS6000 workstation. Chemicals. Amino acids, 3-(cyclohexylamino)propanesulfonic acid (CAPS), fluorescein isothiocyanate (FITC), fluorescein, and β-galactosidase are from Sigma (St. Louis, MO). Fluorescein diβ-D-galactopyranoside (FDG) is from Molecular Probes (Eugene, OR). Other chemicals are of analytical grade. Milli-Q (Millipore, Bedford, MA) water is used throughout. Preparation of Thiocarbamyl-Derivatized Amino Acids. A portion (5 µL) of 1 mM aqueous sample solution is mixed with 25 µL of 0.1 M carbonate buffer solution (pH 9.0) and 10 µL of 0.5 mM FITC methanol solution.27 The mixture is vortexed and kept in the dark at room temperature (∼25 °C) for 14 h. The products, the fluorescein thiocarbamyl (FTC) derivatives of the amino acids, are appropriately diluted with the running buffer of the CE separation before injection. Hydrolysis of FDG by β-Galactosidase.28 The enzymatic hydrolysis is carried out by mixing 45 µL of 100 mM sodium phosphate buffer solution (pH 7.5), 3.5 µL of the diluted β-galactosidase solution (100 times dilution with the phosphate buffer), and 3.5 µL of the 2 mM FDG solution prepared with dimethyl sulfoxide (DMSO). The mixture was drawn into the sampling capillary and injected into the separation channel at the specified time. For calculation of the enzyme concentration, it is assumed that all the protein contained in the enzyme vial received from Sigma is active β-galactosidase (Mr ≈ 540 000).29 RESULTS AND DISCUSSION Rectangular Channel Fabrication. An important step in characterizing these rectangular separation channels is to determine how reproducibly they can be manufactured. Using a stereomicroscope equipped with a 35-mm camera, the thicknesses of a series of channels have been examined. Because of the (27) Van Den Beld, C. M. B.; Lingeman, H.; Van Ringen, G. J.; Tjaden, U. R.; Van Der Greef, J. Anal. Chim. Acta 1988, 205, 15-27. (28) Huang, Z. Biochemistry 1991, 30, 8535-8540. (29) Craig, D.; Arriaga, E. A.; Banks, P.; Zhang, Y.; Renborg, A.; Palcic, M. M.; Dovichi, N. J. Anal. Biochem. 1995, 226, 147-153.

transparency of the UV adhesive used for the preparation, individual glass beads can be visualized between the two channel plates. The channel thickness is virtually identical to the diameter of the glass beads used as the spacers. From the micrographs of the channels, the relative standard deviation (RSD) of the thickness is found to be 3.7% (n ) 5) for the 40-µm-thick channels. Another indication of reproducible channel fabrication is that the channel-to-channel variation in FTC-amino acid migration times is less than 5%. The thickness of the channel can be easily changed by using spacer beads of different diameters. The channel is constructed of standard borosilicate glass microscope slides. Thus, the surface flatness is not expected to be as high as those constructed using ground and polished plates.21-24 We have seen little effect arising from surface imperfections. However, the use of borosilicate glass rather than fused silica decreases the chemical resistance and hence the lifetime of the channels; for example, separation efficiency deteriorates with a channel after 3 days of continuous use and does not return after rigorous washing with methanol, 0.1 M HCl, and 0.1 M KOH. An advantage of using microscope slides is that a large number of channels can be manufactured quickly and inexpensively, used for a short time, and discarded. The design of the detection system alleviates the need for realignment of the channel position or optical system between these changes. Characterization of Channel Performance. Two independent efficiency parameters characterize the channel performance: the longitudinal direction corresponding to separation efficiency and the lateral direction corresponding to time resolution. The effective separation length for the channel of 6 cm is shorter than that typically employed in CE. For the separations reported here, 25 mM CAPS buffer yields better results than the corresponding borate buffer. This is likely due to a smaller EO flow with the CAPS buffer, which benefits separations with such a short length. The effect of the voltage applied across the channel critically affects separation efficiency. For the three tested compoundssFITC, FTC-Arg, and FTC-Glysthe optimum voltage is in the range from 800 to 1300 V, with a 1150-V potential used for the rest of the electropherograms shown here. Under the optimum conditions, the numbers of theoretical plates are found to be 11 000 for FTC-Arg and 8000 for both FITC and FTC-Gly. A typical electropherogram for linear separation with this system is shown in Figure 2, for a mixture containing FITC and three FITClabeled amino acids. The sampling capillary is held constant, an injection is made, the capillary is moved several millimeters, and a second injection made. The four components are nearly baseline separated in less than 4 min. As can be seen, the two injections yield indistinguishable electropherograms, suggesting highly uniform sample injections and channel geometry. One of the most important characteristics of the channel system is the effective time resolution obtainable with a single pass of the sampling capillary. The lateral dispersion (in the direction perpendicular to the channel EOF) of the analytes during the injection and separation process is the predominant factor that determines the measurement precision in the injection time dimension. From the data shown in Figure 2, values of the full width at half-maximum (fwhm, 2.3 σ) for the FTC-amino acids and FITC are found to be ∼600 µm, close to the values calculated using the Einstein equation (σ ) (2Dt)1/2) with a migration time of 120 s and a diffusion coefficient 3 × 10-6 cm2/s, in accordance with previous reports.23,24 Thus, it appears that diffusion of the Analytical Chemistry, Vol. 68, No. 15, August 1, 1996

2473

Figure 2. Two-dimensional electropherogram obtained when the sample introduction capillary is held stationary. Two sequential injections are made of a mixture of FITC, FTC-Gly, FTC-glutamic acid, and FTC-aspartic acid (from front to back), each at 1 µM. CCD measurement starts at 90 s after the second injection is made. Running buffer: 25 mM CAPS, pH 10.15. Separation voltage: 1150 V. Channel current: 630 mA. Fluorescence detection: λex ) 488 nm; λem ) 520 nm.

analyte in the channel is the major source of lateral spreading, and it is not dominated by injection artifacts. In an analogous manner to conventional separations, the peak capacity in the time axis can be calculated as L/4σ.30 As the 4σ baseline width is ∼1 mm (see Figure 2) and the channel width is 2 cm, 20 independent separation lanes are obtained for each pass of the sampling capillary. While the number of independent time channels is fixed by diffusion and channel geometry, the actual time resolution is controlled by the sampling capillary movement rate. For the particular motion controller used, the sample introduction capillary needs about 1 s to move across a distance equal to the width of a separation lane (1 mm) in its “jog” motion mode, thus yielding an effective time resolution of 1 s. Actuators are readily available that increase this speed by an order of magnitude. As the imaging optics and the detector combination has a spatial resolution of ∼40 µm and continuous injections are possible, the effective time resolution can be higher than this, as baseline separation is not always required. An important point is that there is no sampling “dead” time using channel electrophoresis, unlike the situation when using conventional separations with fast, repetitive injections.9 Even arbitrarily fast transient events are measured as sampling is occurring continuously. It is the ability to fix an exact time on a particular event that is limited by lateral spreading. The longest effective measurement time for a single capillary pass is controlled not by actuator movement (which can be arbitrarily slow) but rather by practical constraints such as loss of separation efficiency. For slower movement rates, the easiest method of extending the measurement period and time resolution is to have the introduction capillary move back and forth across the rectangular capillary inlet.24 For each pass, the peak capacity is ∼20, and so arbitrarily long observation times and numbers of time channels are obtainable using multiple passes, with the (30) Giddings, J. C. Unified Separation Science; John Wiley & Sons, Inc.: New York, 1991. Chapter 5.

2474 Analytical Chemistry, Vol. 68, No. 15, August 1, 1996

constraint that the analyte bands do not overlap. Even this restriction can be removed using appropriate deconvolution algorithms. Another advantage of the thin rectangular channel geometry is their high heat dissipation ability.14,18 A typical electrophoresis current at an applied voltage of 1150 V is 600 µA, and currents are as high as 2.5 mA during successful separations. Thus, the rectangular capillary is able to dissipate over 2 W of heat because of the large surface area and relatively thin capillary walls without the use of active cooling. The use of a CCD allows considerable flexibility in readout parameters. CCD read options, including detection window size, readout rate, column, and parallel binning, have a large effect on the analytical performance. The entire CCD array views an area of ∼20 mm × 20 mm on the separation channel, which is 35% of the entire separation length. This allows the appropriate detection window size and position on the channel to be controlled electronically rather than by changing optics or physically manipulating the detector or channel. We can dynamically trade off sensitivity versus separation efficiency by changing the observation zone. For example, the peak resolution between FITC and Leu-FTC increases from 0.8 to 2.0 when the detection window size is reduced from 20 mm × 2 mm to 20 mm × 0.5 mm. Of course, the system sensitivity decreases with this separation improvement. For the remainder of this work, 25 rows (each consisting of 512 pixels) of the CCD are binned and read with column binning equal to 1, which corresponds a detection window of ∼20 mm × 0.5 mm, with 512, 42-µm elements obtained across the width of the channel. With these readout parameters and using CCD9000 software, the fastest data acquisition rate is 400 ms. Although this is adequate for most CE work, the readout speed can be increased several-fold by using custom CCD drivers developed in this laboratory for LabView and LabWindows.31 With the arc lamp excitation-based channel electrophoresis system, the LOD was 1 µM for arginine labeled by FITC.25 Laser

Figure 3. Electropherogram of a single CCD column resulting from the inject of 50 nM amino acids labeled by FITC. Separation and detection conditions are as in Figure 2.

excitation improves the LODs considerably. Figure 3 shows an electropherogram of a single column of the CCD obtained from a diluted FTC-amino acid derivative solution. The concentration of each amino acid is 50 nM before derivatization. Assuming 100% derivatization yield, the LOD (S/N ) 3) is 8 nM for each amino acid. This represents a 125-fold improvement as compared with the previous arc lamp-based excitation. To obtain a better estimate of system sensitivity (e.g., eliminating the effect of nonunity derivatization yield), fluorescein is used as a test compound. Figure 4a shows the electropherogram obtained for a 10 nM fluorescein solution prepared with the CE running buffer. In this experiment, the sample introduction capillary moves back and forth across the entrance of the rectangular channel at ∼1 min in each direction. In this manner the fluorescein solution is continuously injected (gravity injection, 40-cm height) into the channel. The uniformity of injection and reproducibility of capillary movement are visible as a series of 1-min-long zones. The continuous line at position 220 is from a stationary scattering center (e.g., dust or a bubble) on the surface of the glass channel. The intensity versus time for the CCD at column 296 is extracted and plotted in Figure 4b. The LOD is 4 nM (S/N ) 3) for this single CCD pixel column. To obtain the injected sample mass, the injection volume under the same injection conditions is determined by measuring the time required for the fluorescein solution to reach the entrance of the channel (as detected by the CCD). This yields an injection rate of 1.25 nL/second. Therefore, 750 amol of fluorescein is injected in each 1 min trace shown in Figure 4a. For the LOD calculation, the amount is divided by 180, because this signal is integrated by 180 CCD columns each minute; thus, for an individual element in Figure 4a (i.e., CCD column 296, as shown in Figure 4b), the LOD is 1.7 amol of fluorescein. This calculation uses the information in a single CCD element column and does not use any smoothing, take into account the multiple elements imaging each band, or consider lateral spreading of a band. Figure 4c shows the same signal trace extracted from the CCD after smoothing the two-dimensional electropherogram (Figure 4a) using a 7 × 7 Gaussian filter.32 As can be seen, a 4-fold improvement in the LOD to 425 zmol/CCD element is obtained without significant loss of spatial resolution. The conversion of (31) Timperman, A. T.; Khatib, K.; Sweedler, J. V. Anal. Chem. 1995, 67, 139144. (32) Ratzlaff, K. L. Introduction to Computer-Assisted Experimentation; John Wiley & Sons, Inc.: New York, 1987; Chapter 11.

Figure 4. (a) Electropherogram when the sample introduction capillary moves back and forth along the entrance of the channel, continuously injecting a 10 nM fluorescein solution. (b) Signal versus time response of CCD column 296 extracted from (a). (c) Signal versus time response of CCD column 296 extracted from (a) after 7 × 7 Gaussian smoothing of the two-dimensional data. The separation and detection conditions are as in Figure 2.

this LOD to the minimum detectable injected amount depends on the desired time resolution and current CCD parameters but can approach 500 zmol with sufficient CCD binning. Optimum extraction of time and intensity information using sophisticated filtering and calibrations are currently under investigation.33 Time-Based Dynamic Electrophoresis. The dynamic channel electrophoresis has application in following dynamic events in biology and chemistry. As an important example, FDG is a widely used fluorescent substrate for enzymological studies and enzyme assays.28,34 The hydrolysis of its two glycosidic bonds by the enzyme β-galactosidase (β-Gal) is a two-step enzymatic reaction. During the hydrolysis, an intermediatesfluorescein mono-β-galactoside (FMG)sis formed: By detecting changes in fluorescence intensity for each of the separated analyte bands, dynamic information on both reaction steps can be obtained. (33) Oldenburg, K. E.; Liu, Y.; Sweedler, J. V. Maximizing signal-to-noise in multidimensional electrophoreses, manuscript in preparation. (34) Berger, C. N.; Tan, S. S.; Sturm, K. S. Cytometry 1994, 17, 216-221.

Analytical Chemistry, Vol. 68, No. 15, August 1, 1996

2475

Studies of FDG hydrolysis by β-gal has been extensively investigated using conventional fluorometry,28 HPLC,35 and CE.29 Fiedler and Hinz 35 used an HPLC method to separate the species and quantitate them individually. Portions (200 µL each) of the hydrolysis solution were taken out from the reaction vial at specified times, and the hydrolysis was stopped by adding 20 µL of 1 M H3PO4. In our work, ∼500 nL of the hydrolysis solution is drawn into the sample introduction capillary immediately after mixing the compounds, and the reaction takes place in the capillary. At a specified time, the sample is injected into the channel, and the individual species are continuously separated and detected to acquire kinetic data. To make a reliable kinetic assay for this multistep enzyme reaction, both the intermediate (FMG) and the final product (fluorescein) must be separated before fluorescence intensities are measured.35 A single electropherogram (column) extracted from the time-resolved two-dimensional electropherogram for a FDG hydrolysis mixture is shown in Figure 5, illustrating the separation of these two compounds. FMG migrates faster than fluorescein under these conditions. Figure 6 shows the peak maxima as a function of time for both FMG and fluorescein. A 5-min continuous injection of the FDG hydrolysis solution starts 5 min after the initiation of the reaction. As can be seen from Figure 6, during this time interval (from 5 to 10 min) the FMG concentration is constant. These results support other recent results29,35 showing that the previously reported intermediate channeling28 does not exist under these conditions. The fluorescein formation versus time data have been fit using linear regression analysis with a regression coefficient of 0.955. The regression line is shown in Figure 6 (the solid line), together with the 95% confidence limits (broken lines). From the slope of this regression line the formation rate of fluorescein under these hydrolysis conditions is 200 nmol L-1 min-1. Additional enzymatic studies are being initiated to study degradation of neuropeptides by a series of peptidases with this channel electrophoresis system. CONCLUSIONS The channel electrophoresis system is based on free zone electrophoresis in thin channels and LIF with spatially resolved detection. The separation of a series of fluorescein and FITClabeled amino acids is achieved in less than 4 min, with LODs below 1 nM. The analyte spreading while in the rectangular (35) Fiedler, F.; Hinz, H. Eur. J. Biochem. 1994, 222, 75-81.

2476 Analytical Chemistry, Vol. 68, No. 15, August 1, 1996

Figure 5. Single extracted trace showing an electropherogram corresponding to a single injection time, illustrating the separation of FMG and fluorescein with the channel electrophoresis system. [FDG] ) 67 µM; [β-Gal] ) 15 nM. Running buffer: 25 mM CAPS, pH 10.12. Separation voltage: 1150 V. Channel current: 600 µA. Fluorescence detection: λex ) 488 nm; λem ) 520 nm.

Figure 6. Maximum peak height versus time for FMG and fluorescein obtained from a FDG hydrolysis solution (data for fluorescein are divided by 10 before display). The experimental conditions are as in Figure 5.

channel is dominated by lateral diffusion, as described by the Einstein equation, and not injection artifacts. Dynamic events such as chemical reactions can be continuously followed with nanoliter volumes of reactants. Improvements in separation efficiency are the emphasis of further study using viscous sample matrices such as linear polymer solutions and gels. In addition to following dynamic events, this system is ideally suited to coupling capillary-based separations. This should provide efficient multidimensional microseparations, such as capillary LC and IEF to free zone CE. ACKNOWLEDGMENT The support of an NSF NYI Award (CHE-92-57024), the National Institutes of Health (NS31609), and the David and Lucile Packard Foundation is gratefully acknowledged. Received for review December 29, 1995. Accepted May 14, 1996.X AC951242Y X

Abstract published in Advance ACS Abstracts, June 15, 1996.