Anal. Chem. 1999, 71, 1293-1297
Ultrahigh-Voltage Capillary Zone Electrophoresis Katariina Maria Hutterer and James W. Jorgenson*
Department of Chemistry, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3290
An ultrahigh-voltage capillary electrophoresis system was built to demonstrate the possibility of extending the applied potential and thus the separation power of capillary electrophoresis. A commercial 30-kV power supply was extensively modified in order to provide electrical potentials up to 120 kV. A unique electrical shielding system was developed to prevent capillary breakdown and corona or spark discharges. Electrophoretic studies using a mixture of peptide standards, as well as a complex mixture of peptides obtained from a protein digest, showed that the numbers of theoretical plates achieved increase linearly with applied voltage. Theoretical plate counts ranging from 2.7 to 6.1 million plates were obtained for peptides in a separation done at 120 kV. Resolution also increased with the square root of applied voltage, as predicted by theory. Capillary zone electrophoresis (CZE) is usually done with electrical potentials of up to 30 kV. Separation efficiencies as high as 500 000 theoretical plates can be obtained for low-molecularweight compounds in under 30 min using such potentials.1 In a few instances, potentials as high as 45 kV have been used to do analyses.2-10 Electrical breakdown of the fused-silica wall and increased risk of electric discharge and shock, however, have prevented much exploitation of potentials beyond 30 kV. The use of large potential differences is desirable in CZE because separation efficiency is predicted to increase linearly with the applied potential:1
N ) [(µep + µosm)Vl]/2DL
(1)
where N is the number of theoretical plates attained, µep is the electrophoretic mobility of the analyte, µosm is the electroosmotic mobility due to electroosmotic flow, l is the length of the capillary to the detection window, L is the total length of the capillary, V is the applied potential difference, and D is the diffusion coefficient of the analyte. Resolution (Rs) also increases with applied potential, but with a square root dependence: (1) Lukacs, K. Ph.D. Dissertation, UNC Chapel Hill, 1983. (2) Gross, L.; Yeung, E. S. Anal. Chem. 1990, 62, 427-431. (3) Altria, K. D.; Simpson, C. F. Chromatographia 1987, 24, 527-532. (4) Kuhr, W. G.; Yeung, E. S. Anal. Chem. 1988, 60, 2642-2646. (5) Kuhr, W. G.; Yeung, E. S. Anal. Chem. 1988, 60, 1832-1834. (6) Altria, K. D.; Simpson, C. F. Anal. Proc. 1986, 23, 453-454. (7) Bruin, G. J. M.; Chang, J. P.; Kuhman, R. H.; Zegers, K.; Kraak, J. C.; Poppe, H. J. Chromatogr. 1989, 471, 429-436. (8) Belder, D.; Schomburg, G. J. Chromatogr., A 1994, 666, 351-365. (9) Tsuda, T.; Nomura, K.; Nakagawa, G. J. Chromatagr. 1983, 264, 385-392. (10) Kraft, E. M.S. Thesis, UNC Chapel Hill, 1996. 10.1021/ac981221e CCC: $18.00 Published on Web 02/27/1999
© 1999 American Chemical Society
(
RS ) 0.25 N1/2
)
∆µ ) µ j ep + µosm 0.177(µ1 - µ2)
(
V D h (µ j ep + µosm)
)
1/2
(2)
where µj ep is the average electrophoretic mobility of the analytes and D h is the average diffusion coefficient of the analytes. Clearly there is an advantage both in resolution and in separation efficiency in going to manyfold higher potentials (i.e., >100 kV). However, the problems associated with large electric potentials and fields must be overcome. Applying a potential difference across the capillary induces electrical fields, both axially (down the length of the capillary) and radially (out through the capillary walls). The axial electric field drives the separation and is responsible for Joule heating. To reduce the problems associated with Joule heating, the resistance of the capillary can be increased to reduce the current, thus reducing the power that must be dissipated. Through appropriate choice of capillary length, inner diameter, and electrolyte conductivity, power dissipation problems associated with the increase in the electrical potential drop can be avoided. The radial electric field strength increases from the grounded end of the capillary to the high-voltage end as the potential inside the capillary increases. The radial field is responsible for stress on the capillary wall, and at very high potentials, the radial field causes eventual dielectric breakdown of the fused-silica capillary wall. The dielectric breakdown of the fused-silica capillary wall is evident when current through the capillary drops to zero and never recovers. Physical examination of the capillary shows either a hairline crack or complete fragmentation of the fused-silica wall. Early attempts at using voltages greatly in excess of 30 kV in our laboratory demonstrated that the capillary lasts only briefly when no attempt at electrical shielding is used. With simple shielding methods, such as surrounding the capillary with a weakly conducting fluid, the capillary would last for a day or two.10 This paper describes the design and separation performance of an ultrahigh-voltage CE system. The system is designed to permit operation of CE capillaries at potentials to 120 kV without capillary breakdown. This is achieved using a voltage multiplier of the Cockroft-Walton type, along with a unique electrical shielding system designed to reduce electrical stress on the voltage multiplier circuitry and the capillary. EXPERIMENTAL SECTION CE System. A commercial 30-kV power supply (Spellman RHR series, Plainview, NY) was modified by removing the internal Analytical Chemistry, Vol. 71, No. 7, April 1, 1999 1293
Figure 1. Schematic of the experimental setup for ultrahigh-voltage capillary zone electrophoresis.
rectification/multiplication stack and using the original voltage regulation, voltage reference, series control, and power rf oscillator to drive a new rectification/multiplication stack. The original multiplier was a 4-fold multiplier of the Zimmermann-Wittka type. The new multiplier is a 26-fold multiplier of the Cockroft-Walton type (shown in Figure 1). Ceramic capacitors (2400 pF, rated to 20 kV dc) were purchased from Newark Electronics (Chicago, IL) and diodes (ED2139, 25 kV maximum reverse operating voltage, 40 mA maximum forward current) were purchased from Electronic Devices, Inc. (Yonkers, NY). As can be seen in Figure 1, diodes are used to direct the charging of two sets of capacitors in series. The capacitors of the ac input series all have a significant ac component to their potential, while the capacitors of the ground series have a negligible ac component. When the circuit reaches a steady state (the capacitors are fully charged), the first ac capacitor has a potential of Vp + VAC, where Vp is the potential difference between the peak maximum of the ac input and ground and VAC is the ac input. Therefore, the potential of the first dc capacitor is 2Vp. Each successive capacitor has a dc component of its signal that is 2 times the peak ac voltage higher than the preceding capacitor. The voltage regulation circuitry monitors the voltage present at the quadrupling stage (second dc capacitor) and uses this in a feed-back mode to control the ac input to the rectification/multiplication stack. The capillary was protected from dielectric breakdown by means of a metal shielding system consisting of a set of aluminum cylinders, as shown in Figure 1. The protection that such shields afford the capillary (as well as the voltage multiplier circuitry) can easily be understood by modeling the experimental setup as a cylindrical capacitor. Consider the solution inside the capillary to be one plate of a capacitor, and the nearest conductive object the other plate. The electric field strength at the capillary inner wall is then given by
E ) (V1 - V2)/ln(b/a)κea
(3)
where E is the electric field strength, V1 is the voltage inside the capillary, V2 is the voltage of the nearby conductor, b is the distance from the center of the capillary to that conductor, κe is the dielectric constant of the material between the plates of the capacitor, and a is the inner radius of the capillary. Electrical stress on the capillary wall depends on the potential difference between the buffer inside the capillary and some nearby conductor outside the capillary. If one were to coat the outside of the capillary with 1294 Analytical Chemistry, Vol. 71, No. 7, April 1, 1999
a uniform coating of some weakly conductive material, the potential on the outside the capillary would drop along the length of the capillary in the same manner as the potential drops in the solution on the inside of the capillary. Thus, the potential on the outside would match the potential on the inside. Under such circumstances, the capillary wall would endure negligible electrical stress. This weakly conductive coating approximates a set of infinitesimally thin rings around the capillary, with each biased at the same potential of the capillary slice it surrounds. Rather than actually attempting to coat the capillary, one can imagine grouping these infinitesimally thin rings into cylinders of finite width. The use of cylinders, each biased to the approximate average potential of the capillary it surrounds, keeps the potential difference between the two “plates” of the capacitor small. This protects the capillary from experiencing electric fields large enough to cause the fused-silica wall to breakdown. Each aluminum shielding cylinder was ∼15 cm long, ∼15 cm in diameter, with a 1-cm wall thickness. The ends of the cylinder walls were rounded off to a 0.5-cm radius of curvature, and the external cylinder surfaces were polished. Each of the cylinders was electrically biased by tapping-off of the multiplication stack at the intervals depicted in Figure 1. The highest-voltage shield cylinder (the injection housing) was biased at the full run potential. The entire setup, including the shielding cylinders, multiplication stack, and capillary, were immersed in transformer oil (Diala AX, Shell Corp., Houston, TX, dielectric strength of 280 kV/cm) contained in a covered clear polycarbonate plastic tank. The grounded end of the capillary exits from the top of the tank, and the detection is performed away from the oil, in the conventional manner. To prevent flashover (arcing) between the individual shielding cylinders, the electrical field strength versus separation distance between the shielding cylinders was modeled. To do this, the parallel rims of the cylinders were modeled as parallel rods. The relation of separation distance versus electric field strength is then given by11
EM )
Vx(D2 - 4r2)
[ x(
D + 2r(D - 2r) ln 2r
D 2 -1 2r
)
]
(4)
where EM is the maximum electric field strength, r is the radius of curvature of the rods, D is the separation distance, and V is (11) Alston, L. High-voltage Technology; Oxford University Press: London, 1968.
Figure 2. Side view of injection device.
the potential difference. The minimum separation distance before flashover was predicted to occur was determined to be less than 1 cm for a 45-kV potential difference between each cylinder. Typical separation distances between one shield and the next were thus set to 3 cm, to ensure a margin of safety. The final shielding cylinder (the injection housing) is a 30cm-long metal cylinder made from the same aluminum stock as the other shielding cylinders. At one end, the edges of the housing are rounded to a 0.5-cm radius of curvature, as with the other shielding cylinders. At the other end, it is capped with a hollow hemisphere of aluminum. Toward the capped end, a 8.9-cmdiameter hole and a metal guide allow the insertion of an airtight aluminum sample chamber. This chamber functions to isolate the sample, buffer, and end of the capillary from the transformer oil. The chamber contains a spring-loaded rotating sample tray which autoindexes to each sample or buffer reservoir by means of a central shaft which terminates in a handle. This mechanism allows the operator to safely change between samples and buffers without exposure to high voltage or exposure of the solutions to transformer oil. The capillary is fed into the sample chamber through a Swagelock fitting from the inside of the shielding cylinder and then is guided into the indexed reservoir. The sample chamber, which bayonet locks into the injection housing, can easily be lifted out from the shielding cylinder, and its contents can then be accessed by simply unscrewing the rounded lid of the chamber. The process is slightly messy but can be completed in a few seconds. The entire assembly is shown schematically in Figure 2. Laser-Induced Fluorescence Detector. Laser-induced fluorescence was achieved with the use of a 442-nm HeCd laser (LiCONiX, Santa Clara, CA) focused onto the capillary with a 10× microscope objective (Melles Griot, Irvine, CA). The induced fluorescence was measured by collecting the light through a 60× microscope objective (Edmund Scientific, Barrington, NJ), passing the light through a band-pass filter with a 530-nm center wavelength and 30-nm band-pass (Omega Optical, Brattleboro, VT) to remove scattered light, and focusing the light onto a photomultiplier tube (PMT) (Hamamatsu model R1477, Bridgewater, NJ). The PMT voltage was set to 1 kV. The PMT current was sent to an amplifier (Stanford Research System, Sunnyvale, CA) set at a gain of 2 µA/V and with a low-pass filter cutoff setting of 10 Hz (-3 dB point) and a 12 dB/octave rolloff. Data were acquired at a rate of 20 Hz on a 16-bit ADC board using an in-
house data acquisition program written in LabView (National Instruments, Austin, TX). Capillary Column and Sample Introduction. The fusedsilica capillary (Polymicro Technologies, Pheonix, AZ) was 31µm i.d., 360-µm o.d., and 394 cm long with a 384-cm length to the detector. The capillary inner surface was not treated or coated in any manner. Injections were performed hydrodynamically, with a 10-s misleveling of 4 cm. The variance produced by injection was calculated to contribute less than 1/20th of the total variance of the sharpest peaks. Model Peptide Mixture. A mixture of five model peptides, tagged with fluorescein isothiocyanate (FITC) (Molecular Probes, Portland, OR) was run on the ultrahigh-voltage instrument at 120 kV and at 28 kV. The FITC tagging reaction was accomplished by adding a 10-fold molar excess of FITC (dissolved in DMSO) to each ∼10 mg/mL peptide solution (in 10 mM borate buffer at pH ) 9.0) and allowing the mixture to react in the dark for 3 h. The peptide samples were then combined and diluted in 10 mM borate buffer at pH ) 9.0 to final concentrations of 1.3 mg/mL tryptophan, 1.1 mg/mL leucine enkephlin, 1.3 mg/mL GlyLeuTyr, 1.6 mg/mL AlaGln, and 1.3 mg/mL GlyAsn. Tryptic Digest. A tryptic digest of horse heart myoglobin at 1.0 mg/mL, tagged with an excess of FITC, was prepared by adding 10 mg of myoglobin and 1 mg of trypsin to 1 mL of 10 mM borate buffer at pH ) 9.0. This solution was allowed to digest at 35 °C for 12 h. The resulting solution was then tagged with a 10-fold molar excess of FITC to tryptic fragment, as with the model peptides. RESULTS AND DISCUSSION The linearity of the voltage output for the ultrahigh-voltage power supply was tested by measuring the electrophoretic current through a capillary. Current through the capillary as a function of applied voltage was found to be linear (coefficient of determination, R2, of 0.997), indicating that Ohm’s law is obeyed and that the voltage achieved is proportional to the voltage expected. The actual voltage achieved was calculated by determining the current-voltage curve for the capillary with an unmodified power supply of known output voltage. The slope of this current-voltage curve is the conductance (reciprocal of resistance) of the capillary. This resistance value can then be multiplied by the measured current in the capillary to obtain the run voltage under ultrahighvoltage conditions. Although others have observed nonlinear increases in current with applied voltage in CZE, this is attributable to Joule heating effects.6,7,9 Joule heating is negligible in our system, as evidenced by the fact that no nonlinear current-voltage effects have been observed on this system. Analysis of Model Peptides. Table 1 shows the results obtained from electrophoresis of the four FITC-labeled peptide standards and tryptophan at 28 and 120 kV. Theoretical plates were calculated by fitting each peak to a Gaussian curve using in-house software written in LabView 4.1. Since the voltage ratio between the 120-kV run and the 28-kV run is a factor of 4.3, the theoretical plate count is expected to be a factor of 4.3 times greater in the 120-kV run. The observed ratio ranged from a low value of 2.3 to a high value of 4.4 with an average for the five compounds of 3.6 (standard deviation of 0.9). This indicates generally good agreement with the expected increase in plate Analytical Chemistry, Vol. 71, No. 7, April 1, 1999
1295
by fitting to Gaussians, as with the peptide standards. Resolution of each pair of neighboring peaks was calculated by the equation
Table 1. Efficiency and Migration Time Data for Peptide Standards at 120 and 28 kV migration time migration plate ratio time ratio (min) 120 kV/ 28 kV/ 120 kV 28 kV 28 kV 28 kV 120 kV 120 kV millions of plates
LeuEnk GlyLeuTyr AlaGln GlyAsn Trp av std dev expected
3.4 4.4 4.0 3.2 2.2
1.5 1.1 0.9 0.7 0.7
2.3 4.0 4.3 4.4 3.1
267.9 283.6 310.6 318.2 345.2
59.88 63.80 70.61 72.60 79.69
3.6 0.9 4.3
4.47 4.45 4.40 4.38 4.33 4.41 0.06 4.30
counts. Similarly, migration time should decrease from the lowvoltage run to the high-voltage run by a factor of 4.3. The actual observed factor was 4.4, in very good agreement with theory. Analysis of a Tryptic Digest. Electropherograms for the FITC-labeled tryptic digest of horse heart myoglobin run at 28 and 120 kV are shown in Figure 3. These electropherograms clearly demonstrate the dramatic increase in resolution achieved for this complex sample in the 120-kV run. The theoretical plate counts for all the electrophoretic peaks that had a signal-to-noise ratio of 3 or greater in both runs, and that were sufficiently well resolved in both runs to be fit to Gaussian curves, were considered for purposes of data analysis. Theoretical plates were calculated
Rs ) 2(tM2 - tM1)/(w1 + w2)
(5)
where tM1 and tM2 are the migration times of the first and second peaks, respectively, and w1 and w2 are the widths of the peaks at the “base” (4σ width), as found by the Gaussian fit. In all cases where a poor signal-to-noise ratio or poor resolution prevented inclusion of a peak in the data set, it was due to deficiencies in the data from the 28-kV run. As can be seen in Figure 3, the peaks in the 120-kV run generally yielded twice the detector response (absorbance) as those in the 28-kV run. This is exactly as would be expected, as the peaks should be approximately twice as narrow spatially and, thus, twice as concentrated in the 120-kV run. Observed theoretical plate counts ranged from 2.7 to 6.1 million plates for these FITC-labeled tryptic peptides in the 120-kV run. The average ratio in plate counts between the 120- and 28-kV runs was a factor of 3.8 ( 1.3, in fair agreement with the expected factor of 4.3. The average improvement in resolution between each neighboring pair of peaks was a factor of 2.1 ( 0.2, in excellent agreement with the expected factor of 2.1. Finally, the ratio of migration times between the two runs was 3.9 ( 0.04, in fair agreement with the expected ratio of 4.3. This discrepancy between the expected and measured ratios of migration times is
Figure 3. Electropherograms of myoglobin digest (a) on UHVCZE instrument at 28 kV and (b) on UHVCZE instrument at 120 kV. Peak numbers correspond to peak numbers in Table 2 (in Supporting Information). 1296 Analytical Chemistry, Vol. 71, No. 7, April 1, 1999
Figure 4. Expanded view of electropherograms of myoglobin digest (a) on UHVCZE instument at 28 kV and (b) on UHVCZE instrument at 120 kV. Peak numbers correspond to peak numbers in Figure 3 and Table 2 (Supporting Information).
due to a change in the electroosmotic flow between these two runs. The slower-than-predicted electroosmotic flow in the 120kV run also explains why the ratio of theoretical plate counts between the 120- and 28-kV runs is not quite as high as theory would predict, and yet the ratio of resolutions observed in the two runs is exactly as theory predicts. A table containing detailed information on each peak is available as supplemental data. An enlargement of the sections of both electropherograms around the peaks 12-20 is shown in Figure 4. This expanded section of both electropherograms serves to illustrate a potential source of measurement bias between the two data sets. The higher resolution of the 120-kV data set provides information that allows us to extract data from the 28-kV data set that would otherwise be difficult to obtain. The peak pairs 13 and 14, and 15 and 16, for example, are clearly seen to be doublets in the 120-kV run. This knowledge permits us to accurately fit two Gaussians to each of the poorly resolved pairs of peaks in the 28-kV run. Without this information, the peaks in the 28-kV run might have been mistaken for poorly shaped “fronting” or “tailing” single peaks. And yet there is no comparable higher resolution information that permits us to do the same thing for peaks in the 120-kV data set. Peak 18, for example, appears to be a singlet in both runs. Upon closer inspection of the 120-kV run, however, peak 18 appears to be
unusually broad, flat-topped, and may have a poorly resolved shoulder on its right side. In the absence of information confirming this suspicion, we have simply fit a single Gaussian to this peak. If this actually is a pair of peaks just verging on resolution, the effect of fitting a single Gaussian to this peak is more damaging, in a relative sense, to the plate count in the 120-kV run than in the 28-kV run. In this way, a bias will be introduced which will tend to underestimate the plate counts for some of the peaks in the higher resolution 120-kV run relative to the lower resolution 28-kV run. ACKNOWLEDGMENT This work was funded by NSF Grants CHE-9215320 and CHE9727505. SUPPORTING INFORMATION AVAILABLE Table listing the efficiency, resolution, and migration times of a myoglobin digest at 120 and 20 kV. Supporting Information is available free of charge via the Internet at http://pubs.acs.org. Received for review November 9, 1998. Accepted January 26, 1999. AC981221E
Analytical Chemistry, Vol. 71, No. 7, April 1, 1999
1297
