Anal. Chem. 2007, 79, 765-768
Instrumentation for Medium-Throughput Two-Dimensional Capillary Electrophoresis with Laser-Induced Fluorescence Detection Cuiru Zhu, Xinya He, James R. Kraly, Megan R. Jones, Colin D. Whitmore, David G. Gomez, Michael Eggertson,† Wes Quigley,‡ Anna Boardman, and Norman J. Dovichi*
Department of Chemistry, University of Washington, Seattle, Washington 98195-1700
In two-dimensional capillary electrophoresis, a sample undergoes separation in the first dimension capillary by sieving electrophoresis. Fractions are periodically transferred across an interface into a second dimension capillary, where components are further resolved by micellar electrokinetic capillary electrophoresis. Previous instruments employed one pair of capillaries to analyze a single sample. We now report a multiplexed system that allows separation of five samples in parallel. Samples are injected into five first-dimension capillaries, fractions are transferred across an interface to 5 second-dimension capillaries, and analyte is detected by laser-induced fluorescence in a five-capillary sheath-flow cuvette. The instrument produces detection limits of 940 ( 350 yoctomoles for 3-(2-furoyl)quinoline-2-carboxaldehyde labeled trypsin inhibitor in one-dimensional separation; detection limits degrade by a factor of 3.8 for two-dimensional separations. Two-dimensional capillary electrophoresis expression fingerprints were obtained from homogenates prepared from a lung cancer (A549) cell line, on the basis of capillary sieving electrophoresis (CSE) and micellar electrophoresis capillary chromatography (MECC). An average of 131 spots is resolved with signal-to-noise greater than 10. A Gaussian surface was fit to a set of 20 spots in each electropherogram. The mean spot width, expressed as standard deviation of the Gaussian function, was 2.3 ( 0.7 transfers in the CSE dimension and 0.46 ( 0.25 s in the MECC dimension. The standard deviation in spot position was 1.8 ( 1.2 transfers in the CSE dimension and 0.88 ( 0.55 s in the MECC dimension. Spot capacity was 300. Jorgenson’s seminal paper on capillary electrophoresis was published 25 years ago.1 Since that time, the technology has supplanted alternatives for DNA sequencing and for most genotyping applications.2 However, capillary electrophoresis has been less successful for other bioanalysis applications. This lack of success can be attributed to several factors, including the * Corresponding author. E-mail:
[email protected]. † Present address: Department of Biochemistry and Molecular Biology, University of Calgary, Calgary, AB T2N 4N1, Canada. ‡ Present address: Boeing, Inc., Seattle, WA. (1) Jorgenson, J. W.; Lukas, K. D. Anal. Chem. 1981, 53, 1298-1302. (2) Dovichi, N. J.; Zhang, J. Angew. Chem., Int. Ed. 2000, 39, 4463-4468. 10.1021/ac061652u CCC: $37.00 Published on Web 12/10/2006
© 2007 American Chemical Society
requirement of chemical derivatization of low concentration mixtures, the challenges of resolution of highly complex samples, and the need for highly reproducible separations. We have reported a highly reproducible two-dimensional capillary electrophoresis system for the analysis of complex samples.3-6 That system employs the fluorogenic reagent 3-(2furoyl)quinoline-2-carboxaldehyde (FQ) to label primary amines, including low molecular weight biogenic amines and the -amine of lysine residues of proteins. Components are separated using a comprehensive two-dimensional capillary electrophoresis system and detected using postcolumn laser-induced fluorescence in a sheath-flow cuvette. In many applications, high throughput analysis is of value, and there have been several reports of high-throughput DNA sequencers based on multiple capillary electrophoresis.7-20 In this Technical Note, we expand on those previous reports by modifying the instrument described in ref 18 to perform two-dimensional capillary electrophoresis on five samples in parallel. An interface (3) Michels, D. A.; Hu, S.; Schoenherr, R. M.; Dovichi, N. J. Mol. Cell. Proteomics 2002, 1, 69-74. (4) Michels, D. A.; Hu, S.; Dambrowitz, K. A.; Eggertson, M. J.; Lauterbach, K.; Dovichi, N J. Electrophoresis 2004, 25, 3098-3105. (5) Hu, S.; Michels, D.; Fazal, M. A.; Ratisoontorn, C.; Cunningham, M. L.; Dovichi, N. J. Anal. Chem. 2004, 76, 4044-4049. (6) Kraly, J. R.; Jones, M. R.; Gomez, D. G.; Dickerson, J. A.; Harwood, M. M.; Eggertson, M.; Paulson, T.; Sanchez, R.; Feng, Z.; Reid, B.; Dovichi, N. J. Anal. Chem. 2006, 78, 5977-5986. (7) Zagursky, R. J.; McCormick, R. M. Biotechniques 1990, 9, 74-79. (8) Mathies, R. A.; Huang, X. C. Nature 1992, 359, 167-169. (9) Scherer, J. R.; Kheterpal, I.; Radhakrishnan, A.; Ja, W. W.; Mathies, R. A. Electrophoresis 1999, 20, 1508-1517. (10) Hernandez, L.; Escalona, J.; Joshi, N.; Guzman, N. J. Chromatogr. 1991, 559, 183-196. (11) Taylor, J. A.; Yeung, E. A. Anal. Chem. 1993, 65, 956-960. (12) Ueno, K.; Yeung, E. S. Anal. Chem. 1994, 66, 1424-1431. (13) Quesada, M. A.; Zhang, S. Electrophoresis 1996, 17, 1841-1851. (14) Lu, X.; Yeung, E. S. Appl. Spectrosc. 1995, 49, 605-609. (15) Anazawa, T.; Takahashi, S.; Kambara, H. Anal. Chem. 1996, 68, 26992704. (16) Kambara, H.; Takahashi, S. Nature 1993, 361, 565-566. (17) Takahashi, S.; Murakami, K.; Anazawa, T.; Kambara, H. Anal. Chem. 1994, 66, 1021-1026. (18) Zhang, J.; Voss, K. O.; Shaw, D. F.; Roos, K. P.; Lewis, D. F.; Yan, J.; Jiang, R.; Ren, H.; Hou, J. Y.; Fang, Y.; Puyang, X.; Ahmadzadeh, H. and Dovichi, N. J. Nucleic Acids Res. 1999, 27, e36. (19) Crabtree, H. J.; Bay, S. J.; Lewis, D. F.; Coulson, L. D.; Fitzpatrick, G.; Harrison, D. J.; Delinger, S. L.; Zhang, J. Z.; Dovichi, N. J. Electrophoresis 2000, 21, 1329-1335. (20) Zhang, J.; Yang, M.; Puyang, X.; Fang, Y.; Cook, L. M.; Dovichi, N. J. Anal. Chem. 2001, 73, 1234-1239.
Analytical Chemistry, Vol. 79, No. 2, January 15, 2007 765
Figure 1. Instrument design for five-sample, two-dimensional electrophoresis instrument. The detector is based on a system used for DNA sequencing.18 Two high-voltage power supplies are not shown; one is used to apply potential to the samples in the microtiter plate, and the other is used to apply potential to the interface.
has been constructed that allows transfer of sample between capillaries for the comprehensive analysis of complex samples. A five-capillary laser-induced fluorescence detector, originally developed for high-throughput DNA sequencing, is used as a single spectral channel detector in this instrument.18 EXPERIMENTAL SECTION Instrumentation. A block diagram for the instrument is shown in Figure 1. The two-dimensional electrophoresis system consisted of five pairs of capillaries that were joined through an interface, described below. Two 30 kV high-voltage power supplies (Spellman CZE1000R) were used in the experiment. The first was used to apply separation potential to the samples, and the second was connected to the interface through a solution inlet. The detector was held at ground potential. The instrument was configured to operate in one-dimensional separation mode simply by removing the interface and first capillary. The distal ends of the second-dimension capillaries were inserted into a rectangular sheath-flow cuvette; this cuvette formed the detector for a multiple capillary electrophoresis system, and has been described in detail elsewhere.18 This cuvette held the five capillaries in contact, with spacing determined by the outer diameter of the capillaries, typically 150-µm. Fluorescence was excited by a 473 nm beam from a solid-state diode-pumped laser (Lasermate), operated at 12 mW. Fluorescence was collected with a 0.65 NA long working distance objective, passed through a 580 nm long-pass interference filter, and imaged onto a set of fiberoptic coupled avalanche photodiodes operating in photon-counting mode. The five-capillary interface is shown in Figure 2. A polycarbonate block (∼1 cm × 10 cm × 5 cm) was macromachined using a mill to generate a central well. Two sets of grooves were milled to accommodate 1.5 mm o.d. × 0.5 mm i.d. alignment tubes. A fused-silica spacer tube of 0.5 mm o.d. × 0.16 mm i.d. was glued inside of each alignment tube. Finally, 150 µm o.d. × 25 µm i.d. separation capillaries were threaded through the spacer tubes and glued with their faces separated by about 40 µm. A microscope slide was glued to the top of the polycarbonate, forming a seal. A solution inlet was provided to allow introduction of buffers to the interface between runs and to provide electrical contact during runs. The solution inlet was connected to a microcentrifuge tube with a piece of Tygon tubing, and high voltage from the second power supply was applied to the microcentrifuge tube with 766 Analytical Chemistry, Vol. 79, No. 2, January 15, 2007
Figure 2. Interface between five pairs of capillaries for twodimensional electrophoresis. The top shows a photograph of the interface; the bottom shows the schematic of the design.
a platinum electrode. A connection to waste was used to flush the interface between runs; it was stoppered during runs. Capillary Electrophoresis. Separations employed 25 cm long × 25 µm i.d. × 150 µm o.d. fused-silica capillaries. The polyamide coating was burned from the capillary ends with a gentle flame, and the tip was cleaned with an ethanol-soaked tissue. EOTrol LN (Target Discoveries) was used as a dynamic coating to reduce electroosmosis for the 2D separations. The first dimension separation was performed by capillary sieving electrophoresis in a 100 mM Ches, 100 mM Tris, 3.5 mM SDS, 5% (w/v) dextran (500 kDa), and 5% EOTrol buffer. The second dimension separation was performed by micellar electrokinetic capillary chromatography in a 100 mM Ches, 100 mM Tris, 15 mM SDS buffer, and 5% (v/v) EOTrol (pH 8.9). The voltage program for 2D separations has been reported previously.6 The cuvette was held at ground potential. Samples were injected by applying 5 kV for 5 s across the first dimension capillary, introducing ∼30 pL of sample. A preliminary separation was then performed by applying -25 kV to the sample and -1 kV to the interface for 122 s, at which point the fastest moving analyte approaches the end of the first dimension capillary. A series of steps was then performed. First, a plug of analyte was transferred from the first to the second capillary by applying -26 kV to the sample and -16 kV to the interface for 1 s. Second, -18 kV was applied to both the sample and the interface for 19 s. This set of voltages produced zero potential across the first capillary so that components were stationary in that capillary; the voltage drop applied across the second capillary drove the second dimension separation. Up to 250 of these transfer and second dimension separations were performed under computer control. Sample Preparation. Samples were labeled with 3-(2-furoyl)quinoline-2-carboxaldehyde (FQ) and KCN using the procedure described elsewhere.6. A homogenate was prepared from the A 549 cell line and labeled with FQ before analysis following the cell culture procedure and labeling reaction as described earlier.6 Data Collection and Processing. Data were collected with a computer using software written in LabView. Data processing was performed using software written in Matlab.3 RESULTS AND DISCUSSION Limit of Detection. A series of injections was performed of 45 zmol of FQ-labeled trypsin inhibitor with the instrument
Figure 3. Five two-dimensional capillary electrophoresis separations of an A549 cell-line homogenate. One component with inconsistent migration time in the MECC dimension is circled. The CSE fraction number and mobility are indicated on the left and right axes, respectively.
configured for one-dimensional MECC separation. The data were treated with a 5-point median filter to remove noise spikes due to particles and bubbles passing through the laser beam. The data were then convoluted with a 236 point Gaussian filter with a 29point standard deviation. Eleven sets of five-capillary runs were generated. Noise was estimated from the standard deviation of the baseline across a 33 s period. The mean detection limit (3σ) determined from the 55 electropherograms was 940 ( 350 ymol (560 ( 200 copies) injected onto the capillary. This detection limit is about three times poorer that that reported for fluorescein in this detector,18 which reflects the poorer spectral properties of FQ-labeled proteins. The relatively tight standard deviation in detection limit reflects uniform sensitivity across the five capillaries. These detection limits degrade for two-dimensional separations. Three replicate five-capillary analyses of the FQ-labeled trypsin inhibitor were performed with the instrument configured for two-dimensional separation, generating 15 electropherograms. A blank was also analyzed. The data were first treated with a 3 × 3 median filter to remove spikes from bubbles. Next, a morphological opening procedure over a 17 pixel radius was used to estimate the background, which was subtracted from the data. The data were then treated with a two-dimensional convolution filter based on a Gaussian surface with a standard deviation of 0.5 transfers in the capillary sieving dimension and 0.1 s in the micellar electrokinetic chromatography dimension. Detection limit was the amount of injected analyte that would generate a spot whose amplitude was three times the noise in the blank. The average detection limit was 3.5 ( 1.6 zmol. Our detection limit for the two-dimensional separation was 3.8 times poorer than that obtained for the one-dimensional separation. A degradation of detection limit is inherent in the twodimensional separation. While all of the analyte molecules are present in a single peak in a one-dimensional separation, they are spread across the area of the two-dimensional spot in a twodimensional separation. As shown below, a Gaussian surface fit to the typical two-dimensional spot has a standard deviation of 2.3 transfers in the capillary sieving dimension, which corresponds to a full-width at half-height of about 5.4 transfers. As a result, the analyte molecules are divided among ∼5 fractions, decreasing the maximum signal in proportion. The background signal is unaffected by the two-dimensional separation, and the signal-to-
noise ratio will decrease in a two-dimensional separation compared to a one-dimensional separation. For the raw data, the signal-tonoise ratio will decrease in proportion to the number of transfers the analyte spot spans; smoothing across two-dimensions will recover some of that lost signal-to-noise ratio. 2D-5Cap CE Separations of Lung Cancer A549 Cell Homogenates. The five-capillary instrument was used to analyze a homogenate prepared from the A549 cell line, Figure 3. A set of spots is observed consisting of proteins and biogenic amines. The data are plotted as a gel image where the optical density is proportional to the logarithm of fluorescent intensity; the image is overexposed to highlight some of the lower amplitude components. The horizontal axis represents the MECC time while the vertical axis represents the fraction number transferred from the CSE capillary. We also present the mobility of components in the CSE dimension. To characterize the separations, an unsupervised routine was used to select the local maxima in the five electropherograms. There were 131 ( 8 spots identified that were 10 times the standard deviation of the background signal. The positions of the spots in the two dimensions were uncorrelated; the average linear correlation coefficient for spot position in the two dimensions was -0.01 ( 0.03 for the five capillaries. A nonlinear least-square routine was used to fit Gaussian surfaces to a set of 20 common spots selected from the five electropherograms. The mean spot width, expressed as standard deviation of the Gaussian function, was 2.3 ( 0.7 transfers in the CSE dimension and 0.46 ( 0.25 s in the MECC dimension. The peak capacity was 28 in the CSE dimension and 11 in the MECC dimension. The spot capacity for the two-dimensional separation is given by the product of the peak capacities in the two dimensions, or roughly 300 for these separations. The spot position was quite reproducible in the CSE dimension. The standard deviation of the transfer number was determined for each of the 20 spots across the five capillaries, and the average equaled 1.8 ( 1.2 transfers. The uncertainty in spot position is slightly smaller than the spot width in the CSE dimension; the spot width dominates reproducibility in the spot position. The spot migration time reproducibility is not as good in the MECC dimension. The average standard deviation in the MECC spot position was 0.88 ( 0.55 s, which is roughly twice the spot Analytical Chemistry, Vol. 79, No. 2, January 15, 2007
767
width in this dimension. The major source of poor reproducibility is drift in the position of one component, circled. This component was tentatively identified as histidine by spiking.6 This component’s mobility appears to be particularly sensitive to slight changes in buffer composition and temperature. CONCLUSIONS This paper addresses throughput in two-dimensional capillary electrophoresis of proteins and biogenic amines. This twodimensional separation required 85 min to complete. Only a few minutes are required to recondition the capillaries for a subsequent run, and the system is capable of generating over 25 twodimensional separations in an 8 h day. There are two obvious extensions of this work. One is to develop higher throughput instruments, ultimately based on 96 capillary-pairs operating in parallel.19-20 The other is to extend the instrument to single cell analysis. While this system has sufficient sensitivity to detect the content of a single eukaryote
768
Analytical Chemistry, Vol. 79, No. 2, January 15, 2007
cell, simultaneous injection of cells into the instrument remains a technological hurdle. Nevertheless, such a system would be invaluable in chemical cytometry to generate sufficient data for statistically useful conclusions to be drawn on the heterogeneity of a cellular population. ACKNOWLEDGMENT This work was supported by Grant R01GM071666 from the National Institute of Health. D.G.G. is grateful to La Consejeria de Infraestructuras y Desarrollo Tecnologico de la Junta de Extremadura y al Fondo Social Europeo for a postdoctoral fellowship.
Received for review September 1, 2006. Accepted October 31, 2006. AC061652U
