Anal. Chem. 1994,66, 955-962
Flow Counterbalanced Capillary Electrophoresis Christopher T. Culbertson and James W. Jorgenson’ The Department of Chemistry, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
A new approach for increasing the efficiency and resolving power of capillary electrophoresisis presented. This approach utilizes a pressure-inducedcounterflow to actively retard, halt, or reverse an analyte’s electrokinetic migation through a capillary. By retarding, halting, or moving the analytes back and forth across the detection window, one effectively keeps the analytes of interest in the separation field much longer than under normal separation conditions, thereby increasing both the efficiency and the resolving power of the separation. With this system, separationefficiencies of 17.3 millionplates have been achieved with fluorescently labeled amino acids in 8 h. Separation efficiencies of this magnitude allow one to resolve compounds with electrophoretic mobility differences of 1 X 10-7 cm*/s. This is 1 order of magnitude lower than with conventional capillary zone electrophoresis. In our laboratory we have been performing two-dimensional
(2-D) liquid chromatography-capillary electrophoresis (LCCE) separations for several years on complex biological samples.l4 Two-dimensionalseparations are especially useful for complex samples because the peak capacity of a 2-D separation is the product of the peak capacities for each dimension of the separation, if the two separation methods are orthogonal to one another. The peak capacity of CE alone is insufficient to resolve all of the components in these complex samples. There are, however, samples which are much less complex and do not overwhelm the peak capacity of the CE separation system but which contain components that are difficult to separate because their structures and electrophoretic mobilities are so similar. For such compounds, a two-dimensional separation system will probably not aid in the separation very much. Rather a method to extend the resolving power of a one-dimensional separation system is needed, and in CE there are two feasible ways in which one may attempt do this. One approach is to increase the applied potential (v), because resolution (Rs) increases with the square root of the applied voltage:
where P I and pz are the electrophoretic mobilities of the two analytes involved, p is their average electrophoretic mobility, (1) Bushey, M. M. Ph.D. Thesis, University of North Carolina-Chapcl Hill, 1990. (2) Bushey, M. M.;Jorgcnson, J. W. AMI. Chem. 1990,62, 978-984. (3) Bushey, M. M.; Jorgenson, J. W. A m / . Chem. 1990,62, 161-167. (4) Larmann, J. P.; Lemmo, A. V.; Moore, A. W. Electrophoresis1993,14,439447. (5) Lcmmo, A. V.; Jorgenson, J. W. J. Chromotogr. 1993, 633, 213-220. (6) Lemmo, A. V.; Jorgenson, J. W. A n d . Chem. 1993.65, 1576-1581.
0003-2700/94/0366-0955$04.50/0 @ 1994 Amerlcan Chemical Soclehr
pmm is the electroosmotic mobility, and Dav is the average
diffusion To double the resolution between two analytes, a 4-fold increase in the applied potential is required. The use of large applied voltages, therefore, is necessary when one desires to try to resolve analytes with exceedingly similar electrophoretic mobilities. These large applied potentials create a practical problem because of the thermal effects (Joule heating) they induce and because of realistic limits to highvoltage use in the laboratory (e. g., a t 60 kV a 15-cm electrical arc can be produced between the electrode and ground). The other approach for enhancing resolution consists of arranging for electroosmotic flow to at least partially counterbalance electrophoretic mobility, thereby delaying the migration of an analyte through the capillary.’ If the electroosmotic flow coefficient is just equal but opposite in sign to the average electrophoretic mobility of the pair of analytes one wishes to resolve, then the resolution will approach infinity. This enhanced resolution, however, is obtained at the expense of long analysis times, as the analytes are brought toa virtual standstill by theelectroosmotic flow. The practical problem with this approach is that it has not been realistically feasible to arrange for electroosmotic flow to exactly counterbalance an analyte’s electrophoretic migration, because electroosmotic flow is difficult to control in a completely predictable manner and tends not to be constant over long periods of time (many hours). A possible solution to this problem is offered by the recent discovery that electroosmotic flow can be controlled using (radial) electric fields external to the ~ a p i l l a r y . ~However, -~~ thelarge range of control needed to reverse or halt an analyte’s migration at pH’s above 5 is not achievable at present.12 In addition, both theory’l and experiment1&13indicate that this large range of control is also dependent upon small capillary inner and outer diameters, low buffer concentrations, and large radial fields. As we desire an approach to increasing resolution in CZE that is able to work independently of buffer pH, radial fields, and buffer concentrations (up to cases where Joule heating becomes a problem), we are currently investigating the use of a pressure-induced flow to counterbalance analyte migration. The use of pressure to increase the efficiency and the resolving power of CZE has been explored previously, to various extents, by several groups. This work has concentrated (7) Jorgenson, J. W.; Lukaar, K.D. A w l . Chem. 1981,53, 1298-1302. ( 8 ) Giddings, J. C. Sep. Sci. 1%9, 4, 181-189. (9) Lee, C. S.;Blanchard, W. C.: Wu, C. T. AMI. Chem. 1990.62, 155&2. (10) Lee,C. S.;McManigill,D.; Wu, C.T.;Patel,B. AMI. Chem. 1991,63,151923. (11) Hayes, M. A.; Ewing, A. W. AMI. Chem. 1992, 64, 512-516. (12) Hayes, M. A.; Kheterpal, I.; Ewing, A. W. AMI. Chem 1993,65, 27-31. (13) Culbertson, C. T.; Jorgenson, J. W. Personal communication, 1993.
Analytical Chemistty, Vol. 66, No. 7, Aprll 1, 1994 955
on using small pressures (less than 1 bar) applied either in conjunction with or in opposition to the electrokinetic migration direction to compensate for resistance to mass-transfer bandbroadening sources in excess of longitudinal diffusion. The excess band broadening sources studied have included Joule heating,14 gravity-induced hydrodynamic flows,I5 and backpressure-induced hydrodynamic flows.I6 Our approach to increase the efficiency of CZE using pressure control differs from these previous approaches because we are not using small pressures to reduce sources of excess band broadening; rather, we are using much larger pressures to halt and/or reverse the electromigration of an analyte pair keeping them in the capillary under an applied potential until an adequate separation is obtained. This technique has been designated flow counterbalanced capillary electrophoresis (FCCE). Below we describe some of the initial theoretical considerations involved in using FCCE and the general instrumental setup, along with the results of amino acid and small peptide separations which show that the resolving power of CE can be increased by at least an order of magnitude. THEORY In FCCE, pressure is applied to the buffer reservoir at one end of the capillary, and it induces a bulk flow of buffer which runs counter to the electrokinetic migration direction. This induced flow is used to actively retard, halt, or even reverse an analyte's electrokinetic migration through a capillary. By retarding, halting, or moving the analytes back and forth across the detection window, one effectively keeps the analytes of interest in the separation field much longer than under normal separation conditions, thereby increasing both the efficiency and the resolving power of the separation. The introduction of a pressure-induced counterflow, however, also introduces an additional band-broadening mechanism, beyond that of longitudinal diffusion, which lowers the rate at which theoretical plates are generated; thereby, slowing the separation process. Normally in CZE, because of the essentially flat flow profile, longitudinal diffusion alone is responsible for the irreducible minimum amount of variance that can be expected. The magnitude of this diffusion, expressed as a variance, ug2,is given by the Einstein equation cB2= 2Dt
dc2vP:t/96D
(3)
where dc is the capillary diameter and upa is the pressureinduced average flow v e l ~ c i t y . l ~The - ~ ~total minimum band broadening expected in FCCE, therefore, is due to a com~~
~
~~
~
~
~~
~~
(14) Gobie, W. A.; Ivory, C. F. J . Chromatogr. 1990, 516, 191-210. (15) Datta, R.; Kotamarthi, V. R. AIChE J . 1990, 36, 916-26. (16) Kok, W. T. Anal. Chem. 1993, 65, 1853-1860. (17) Golay, M. J . E. Gas Chromatography; Buttersworth: London, 1958. (18) Taylor, G. Proc. R. SOC.London A 1953, 219, 186-203. (19) Ark, R. Proc. R. Sor. London A 1956, 235, 67-77.
956
UC2
= u*2:
2Dt = d,2vP,2t/96D
Analytical Chemistry, Vol. 66,No. 7,April 1, 1994
(4)
and solve for d, to get d , = 13.9D/vP, = 5.5 pm
(5)
The pressure (P)required to create a parabolic flow velocity that offsets the forward electrokinetic migration can be found using the Poiseuille equation as follows: Given dc = 5.5 pm, a capillary length (1) of 0.605 m, a buffer solution viscosity ( h ) 0.0010 Pa s, and upa = 7.1 X lo4 m/s, then
(2)
where t is the analyte migration time and D is the analyte diffusion coefficient. In FCCE, the additional band broadening comes in the form of a resistance to mass-transfer term which is due to the pressure-induced parabolic flow profile. This excess band broadening is best described using the equation uc2=
bination of both longitudinal diffusion( uB2) and the induced parabolic flow profile( uc2). Several papers have examined the simultaneous existence of plug and parabolic flow profiles, such as exist in FCCE15,17,2s22 and have shown that the two flow profiles are independent of one another. Thus, the overall flow profile can be expressed as the sum the two profiles and the total variance generated can be given simply as the sum of the variances (U'total = U B ~+ uc2). Reduction of Excess Band Broadening Due to Induced Parabolic Flow Profile. To keep the excess band broadening reasonable, our approach has been to calculate the approximate diameter of capillary needed to reduce the amount of variance generated by the pressure-induced parabolic flow profile to the amount generated by longitudinal diffusion alone (e.g., set uc2 = Q ~ ) .The overall variance generated, therefore, can be limited to twice that of longitudinal diffusion (20.B2). The following example using tetramethylrhodamine isothiocyanate (TRITC) derivatized leucine (L) run at pH 8 illustrates how the approach works. For this initial case the parabolic flow velocity ( u p ) is assumed to beequal but opposite in direction to the electrokinetic flow velocity ( ~ , k ) , so that the analyte pair of interest remains balanced in the detection window. Given an electrokinetic mobility (posm - pep)of 1.43 X l P cm2/V s), an analyte diffusion coefficient (D)of 2.80 X 10" cm2/s, an electric field (Eappl) of 500 V/cm, and an electrokinetic velocity ( U e k ) of Eapp1 (posm - pep)= 0.07 1 cm/s, we set
P = 321huPa/dc2
(6)
P = 4.4 bar or 66 lbslin.' gauge (psig)
The above derivation assumed that the analytes were perfectly balanced in the detection window. For the initial demonstration of FCCE, however, it was easier to move an analyte pair back and forth across thedetection window rather than hold them still. To do this, the sample was injected at the anode and allowed to migrate through the capillary past the detection window (Figure 1A). After the injection plug had traveled three-fourths the length of the capillary, the pressure was then turned on to push the sample back through thedetection window (Figure 1B). The approximate pressure needed to create the desired reverse analyte velocity was (20) Martin, M.; Guiochon, G. Anal. Chem. 1984, 56, 614-20. (21) Martin, M.; Guiochon, G.; Walbroehl, Y.; Jorgenson, J. W. Anal. Chem. 1985, 57, 559-61.
(22) Grushka, E. J Chromatogr. 1991, 559, 81-93.
M-
A
electrokinetic flow only
+
-
detection window
B parabolic + electrokinetic
+ B
~
I
3/4
flow
&
electrokinetic flow only
Derivation of the Efficiency Equation in FCCE. Although moving the analytes back-and-forth across the detection window makes the demonstration of FCCE easier, it makes the derivation of the efficiency equation more difficult because the rate of generation of variance is different in the absence and presence of the pressure induced parabolic flow profile. In the absence of pressure (e.g., during the forward passes across the detection window) the variance generated (df) is solely due to longitudinal diffusion (eq 2). During the time, however, that pressure is applied (e&, the reverse passes) the variance generated(a2,) is due to both longitudinal diffusion (eq 2) and to the excess band broadening brought about by the parabolic flow profile (eq 3). These variances are additive, however, so that the total variance (uztotal)generated is simply the sum of the variances generated from the forward passes (a$) and the reverse passes (aZ,) and is given by
D
J
uuL ... * time
Flgurr 1. Schematlc explanation of FCCE. (A) After the analyte palr is InJected at the anode, a potentlal Is applled across the caplllary causlng the analytes to electromlgrate past the detection window. (B) When the analytes have traveled three-fourths of the way through the capillary, pressure Is applied to push the analytes back through the detectionwlndow. (C) As the analyte palr reaches the point where they are one-fouth of the capllary length fromthe lnjectlon pdnt, the pressure is turned off and the analytes begln to migrate in the forward dlrectlon again. (D) Thls back-and-forth movement across the detectionwindow resutts In the electropherogram as shown.
calculated using the Poiseuille equation. The amount of pressure required to move the analytes in the reverse direction at the same linear velocity as in the forward direction was twice that calculated above to balance the parabolic and electrokinetic flows. After passing through the detection window in the reverse direction, the analytes were pushed back to a position where they were one-fourth the length of thecapillary from theinjection point. At this point the pressure was turned off and the peaks began to migrate in the forward (electrokinetic) direction again (Figure 1C). This back-andforth movement of the analytes across the detection window was continued until adequate separation was obtained giving rise to an electropherogram similar to that in Figure 1D. Moving the analytes back-and-forth across the detection window instead of holding them still in the detection window requires that the pressure be actively turned on and off. Turning the pressure on and off like this requires that a modification be made to eq 3 because the excess band broadening it describes occurs only when the reverse pressure is on. The time ( t ) variable in equation 3 which stood for the total separation time must be changed to the total time spent making the reverse passes (2,). When the analytes are moved back-andforth across the detection window at equal velocities (uek = = '/ZUpa), the time spent making the reverse passes (tr) is equal to one-half of the total run time ( t ) . Because eq 3 is linearly dependent upon time but has a squared dependence on the parabolic flow velocity, moving the analytes backand-forth at equal velocities across the detection window instead of balancing the analytes in the detection window doubles the amount of band broadening due to UC and increases the total amount of band broadening by 50% (to a total of 3UBz instead of 2UB2 ).
where tf is time spent making forward (electrokinetic) passes across the detection window, up is the parabolic flow velocity, and t, is the time spent making reverse passes with the pressure on. Using this revised variance term, the number of theoretical plates expected for FCCE can now be calculated as follows:
N=
2Dtf + 2Dt, + d:vP2tJ96D
where t is the total time over which the separation occurs and uekis the forward or electrokineticvelocity. The electrokinetic velocity is used in the determination of overall plates because the analytes move at a constant velocity relative to one another and to the bulk electroosmotic flow independent of whether the pressure-induced counterflow is operating or not. Using eq 8 above, therefore, and the well-known resolution equation
where tl and t2 are the migration times of the two analytes involved and u1and a2 are their respective baseline widths, the performance characteristics of FCCE can be evaluated and compared with that for conventional CZE.
EXPER IMENTAL SECTION Apparatus. The equipment needed to perform FCCE is readily obtainable or easily made and is described below. The generalized instrumental diagram can be found in Figure 2. (1) Capillary Electrophoresis (CE). Fused silica columns with 5 f 2 pm and 10 f 3 pm nominal inner diameters, and 363-pm outer diameters were obtained from Polymicro Technologies (Phoenix, AZ). The 5-pm capillary was 34.5 f 0.1 cm long (17.3 f 0.2 cm to detection window) and used with an field strength of 870 V/cm. The 10-pm columns were 60.5 f 0.1 cm long (30.3 f 0.2 cm to the detection window) and used with a field strength 500 V/cm. Exact capillary inner diameter measurements were made using a mercury resistance method described below. A pH 8.1, 100 AnaiyticaiChemistry, Vol. 66,No. 7, April 1, 1994
967
1
- mm
I
0
Figure 2. Generalized instrumental diagram for FCCE. The grounded end of the capillary is placed into a buffer vial inside of a pressure reservolr. The other end of the capillary is placed into another buffer reservoir with the hlgh-voltage electrode . Helium gas pressure to the reservolr is controlled using an electrlcally actuated pneumatic valve. The high-voltage power supply and the valve were remotely operated by a personal computer which was also used for data acquisition process.
mM sodium phosphate, 5% methanol, 0.62% triethylamine (TEA) buffer was used in all separations. The capillary was generally preconditioned before runs by flushing at 300 psig (20 bar) for 1 h with 75% 1 M sodium hydroxide/25% methanol followed by a 1-h flush with running buffer. Analytes were injected at 5 kV for 8 s at a concentration of 6X M. (2) Pressure Control. To create a pressure-induced counterflow, the grounded end of a capillary is placed through the Swage-lok fitting of a pressure reservoir and into a buffer vial. The other end of the capillary is placed in another buffer vial with the high-voltage electrode. Helium gas from a tank with a two-stage regulator, at 25 psig (1.7 bar) for the 10-pm column and at 65-75 psig (4.3-5 bar) for the 5-pm columns, was pumped into the pressure reservoir through a Clippard Minimatic ET-3 Electronic/Pneumatic Valve (Clippard Instrument Laboratory, Cincinnati, OH) under computer control. (3) Laser-Induced FluorescenceDetection. A laser-induced fluorescence detection system was used because the extremely short path length of the 5- and 10-pm capillaries would not allow adequate UV-vis sensitivity. It has been found previously that TRITC is an excellent derivatization agent for peptides, amino acids, and other primary amines and that an inexpensive 1.5-mW green (543.5 nm) HeNe laser (MellesGriot obtained through Edmund Scientific Co. Barrington, NJ) matches the excitation wavelength of TRITC (Zex 544 nm) ell.^,^^ The optical system was setup on a breadboard in a light tight box. It consisted of excitation and emission filters (Omega Optical Inc., Battleboro, VT) along with a plano-convex lens to focus the laser beam onto the capillary and a 40X microscope objective (Edmund Scientific Co.) to collect the induced fluorescence off-axis and focus it onto a Hamamatsu R1527 photomultiplier tube (PMT, Bridgewater, NJ.). (4) Computer Control of Run and Data Acquisition. Using a Hewlett-Packard 386 personal computer outfitted with a multifunction data acquisition board (Scientific Solutions, Cochran, OH), in-house software was written with QuickBASIC4.5 (Microsoft Corp., Redmond, WA) to allow one to isolate a peak containing unresolved analytes and to auto-
-
(23) Chen, D. Y.; Swerdlow, H. P.; Harke, H. R.; Zhang, J. 2.;Dovichi, N. J. J . Chromafogr. 1991, 559, 237-246.
950
Analytical Chemlstty, Vol. 66,No. 7, April 1, 1994
matically move this/these peak(s) back and forth across the detection window for up to 8 h or until adequate separation is obtained. (5) Peak Fitting and Plotting Routines. Incompletely resolved peaks were decomposed and fitted to Gaussian equations using the IGOR graphing and data analysis software package by Wavemetrics (Lake Oswego, OR). Baselineresolved peaks were also fitted to Gaussian equations using the same program. Three-dimensional images were generated using the Spyglass Transform Rastor Imaging Program (Spyglass Inc., Champagne, Ill). Reagents. All buffer components and analytes were obtained from Sigma Chemical Co. (St. Louis, MO) and used without further purification. Solutions were made up using distilled deionized water obtained from a Barnstead Nanopure System (Dubuque, IA) and then filtered through 0.2-pm Acrodiscs (from Gelman Sciences obtained through Fisher Scientific, Pittsburgh, PA) just prior to use. Procedure. (1) TRITC Derivatization of Amino Acids (AA)/Peptides. TRITC (Isomer G) obtained from Molecular Probes Inc. (Eugene, OR) or Research Organics (Cleveland, OH) was dissolved in DMSO at a concentration of 10 mg/ mL. Aliquots (25 mL) of this solution were added to 75 mL of the AA/peptide solution consisting of pH 9.0 0.15 M sodium borate b ~ f f e r . ~The , ~ AA/peptide ~ concentrations were in 5-fold molar excess to the TRITC. The derivatization reactions were allowed to proceed for 3 h at room temperature. Afterward, the derivatized peptide solutions were frozen at -20 "C until diluted for use. The tetramethylrhodamine thiocarbamyl amino acid derivatives produced will be referred to below as TRITC-X, where X is the one-letter amino acid abbreviation. (2) Capillary Inner Diameter Measurements. Capillary inner diameter measurements were carried out based on a variation of the mercury resistance method developed by Guthrie et al.25 In this method one end of a capillary and an electrode are placed into a mercury filled vial inside of a pressure reservoir. The capillary is filled with mercury by applying pressure to the mercury-filled reservoir forcing mercury through the capillary. The resistance across the mercury filled capillary is then measured. The measured resistance ( R ) is given by
R = pl/ar2
(10)
where p is the specific resisitivity of mercury (p20 = 95.783 pQ cm), 1 is the capillary length (centimeters), and r is the capillary radius (micrometers). Equation lOcan be rearranged to give the capillary radius (micrometers):
Capillary inner-diameter measurements made using the mercury resistance method described above showed the nominal 10-pm capillaries to be 10.6 f 0.2 pm in diameter and the nominal 5-pm capillaries to be 5.2 f0.2pm in diameter. (24) Haugland, R. Molecular Probes Handbook 1992-4; Molecular Probes Inc.: Eugene, OR, 1992. (25) Guthrie, E. J.; Jorgenson, J. W.; Knecht, L. A.; Bush, S. G. J. HRC CC,J . High Res. Chromatogr. Chromatogr.fCommun. 1983,6, 566-567.
Table 1. Dmurlon Coefflcleni Measurements for TRTICDurfvatlrd I,. L,. QI,. GL,. 061, and GQL
anal*
I L GI GL GGI GGL
T
diffusion coeff (std dev) (cm2/s)
24 24 22.5 22.5 22.5 22.5
2.81 (kO.08) X 10.8 2.79 (A0.04)X 10-8 2.50 (10.05)X 10.8 2.63 (10.14) X 10.8 2.50 (10.06) X 10.8 2.53 (10.05)X 10.8
n ("C) 5 5 5 6 6 8
diffusion coeff
(std dev) corrected % to 24 "C (cm2/e) rsd 2.81 (10.08) X 10.8 2.79 (h0.04)X 10.8 2.59 (10.05)X 10.8 2.72 (A0.14)X 10.8 2.59 (10.06)X 10.8 2.62 (A0.05)X 10.8
1.4 2.8 2.1 5.3 2.3 1.8
(3) Diffusion Coefficients. Diffusion coefficients were measured using the stopped-flow method.26 An analyte was electromigrated onto the column and migrated past the detector. A second electromigration injection of the analyte was made under the same conditions and allowed to migrate approximately one-half the distance to the detector. The highvoltage power supply was turned off for a known amount of time and then restarted allowing the analyte to migrate past the detection window. From the difference in the variances of the peaks in the two runs and from the time the sample was allowed to diffuse with the high-voltage power supply off, a diffusion coefficient was determined. The diffusion coefficients for TRITC-L and TRITC-I were measured in a 5.2pm4.d. 34.5-cm-long capillary run at 588 V/cm with an injection potential of 1 kV for 20 s for isoleucine and 1 kV for 15 s for leucine. The diffusion coefficients for TRITCGL, TRITC-GI, TRITC-GGL, and TRITC-GGI were measured in a nominal IO-pm i.d. 34.5-cm-long capillary run at 588 V/cm with injection potentials of 2 kV for 12 s. The running buffer used and the capillary conditioning procedure for all diffusion coefficient measurements were the same as that described above in the CE section. RESULTS AND DISCUSSION DiffusionCoefficientMeasurements. Diffusion coefficients were measured for TRITC-derivatized L (TRITC-L), I (TRITC-I), G-L (TRITC-GL), G-I (TRITC-GI), G-G-L (TRITC-GGL), and G-G-I (TRITC-GGI) using the stoppedflow method described above. The results of these measurements are shown in Table 1. The diffusion coefficient measurements for the di- and tripeptides were made at 22.5 OC, 1.5 OC lower than the temperature at which the FCCE runs were made. This temperature difference is significant because diffusion coefficients increase a t a rate of approximately 2% per degree Celsius.27 The diffusion coefficients used in calculating the predicted number of theoretical plates below, therefore, have been corrected to 24 OC. Amino Acid/Peptide Separations. Figure 3 shows the FCCE separation of TRITC-L and TRITC-I. The numbered pairsof peaks represent the forward passes across the detection window while the peak pairs labeled with Rs represent the reverse passes. This separation was performed in an 10.6pm-i.d., 60.5-cm-long column with a forward (electrokinetic) velocity of 0.071 cm/s and a reverse velocity ( U p - Uek) of O.O53cm/sec. Figure 4 shows enlarged views of the first, second, fifth, tenth, twentieth, and twenty-ninth forward passes (26) Walbroehl, Y.; Jorgcnson, J. W . J . Microcolumn Sep. 1989, 1 , 41-5. (27) Weast, R. C. CRC Handbook of Chemistry andPhysics, 66th ed.;CRC Press: Boca Raton, FL, 1985.
16
16
20 22 time(seconds x10 ' )
24
26
28
Figure 3. FCCE separation of TRITC-I and TRITC-L in a 10.Bpm-i.d. 60.5-cm-long capillary. The forward passes are represented by the numberedpeaks and the reverse passesare representedby the peaks labeled wlth Rs.The total run tkne was 28 800 seconds (or8 h). The y-axls values for the second half of the flgure have been enlarged by a factor of 10.
I1:i ,*,).Ii 350
400
450
500
,
550
1350
time(seconds)
1
4300
4350
4400
4450
time(sec)
4500
1400 1 4 5 0 time(sec)
1500
I.
1550
4550
time(6ec)
28.30 28.35 28 4 0 28 45 2 8 50 t i m e ( s e c x i ~ ')~
Figure 4. Enlarged views of the first (A), second (e), flfth (C), tenth (D), twentieth (E), and twenty-ninth (F) forward passes of the TRITCI/L Separation. The tlme windows are all 250 s wide.
to better illustrate the increase in resolution between the analytes with time. These windows are displayed at a constant time width of 250 s. The first peak shown in each pair is the TRITC-I analyte. In addition to the separation of the TRITC-derivatized leucine and isoleucine analyte pair, several minor "contaminant" peaks are also resolved. These peaks generally have less than 1% of the area of the major peaks and can be seen Analytcal Chemlstry, Vol. 66, No. 7, Aprll 1, 1994
BSB
i
A.
0.0
100.0 Relative time (seconds)
n
zuu.u
Figure 5. Three-dimensional view of the TRITC-I/L separation. This image was created by stacking ail of the forward runs together and provides a compact way to represent all of the data.
slowly resolving from the leucine and isoleucine analyte pairs near the baseline. Three of the most conspicuous pairs are represented by X , Y, and Z in Figure 4. Analytes X and Y are quickly resolved and have been pushed out of the detection window by the fifth forward pass (Figure 4C). Analyte 2, however, does not even show up until the tenth forward pass (2.5 h into the separation; Figure 4D) and then only as a small shoulder on the TRITC-I peak. After 29 passes (8 h) it has been baseline resolved and is just outside the window shown in Figure 4F. These smaller peaks were probably due to contaminants in the primary amines, or from leucine and isoleucine derivatized with TRITC-(R isomer), a minor contaminant of the TRITC-(G isomer)-derivatizing agent. No attempt to identify them, however, was made. If all of the forward slices are sequentially stacked together, a three-dimensional image is created which shows the increase in resolution over the entire run (Figure 5). This view provides the most concise way in which to represent all of the data. In the theory section above, a capillary inner diameter of 5.5 pm was calculated as necessary to limit the band broadening due to the induced parabolic flow profile to that of longitudinal diffusion (ac2/agZ= 1). In the TRITC-L/I run shown here a larger 10.6-pm-inner diameter capillary was used; and as such, the amount of flow induced band broadening may be expected to increase. In addition, the analytes in the above example are moved back-and-forth across the capillary detection window instead of being held still in the window, which leads to yet a further increase in the amount of flow induced band broadening expected. The amount of total band broadening (atota] 2, expected to occur, however, may be easily calculated using eq 7 and for the run shown above is 7.4 times greater than that from diffusion alone. It is more convenient to express the amount of band broadening as a ratio of ac2/ ug2 to explicitly show how much of the total band broadening is due to the induced flow profile. For the separation above the ac2/ag2 ratio was 6.4. The TRITC-derivatized peptide pairs Gly-Leu/Gly-Ile (GL/GI) and Gly-Gly-Leu/Gly-Gly-Ile (GGL/GGI) were also separated on the same column. The results of these separations are shown in parts A and B of Figure 6, respectively. For conciseness, only the first and last forward passes through the detector are shown. As in the TRITC-L/I separation 960
n
Analytical Chemistry, Vol. 66, No. 7, April 1, 1994
I
B.
300
500
400
11.30
time(sec)
1 1 40
1 1 50
1 1 60
time(secx10' )
Figure 6, (A)First and last (18th) forward passes of the TRITC-GI/OL separation. The total separation time was 17 516 s (or 4.87 h). (B). The first and last (13th) forward passes of the TRITC-GGI/GGL separation. The total separation time was 12 557 s (or 3.49 h).
0
20
40
60
80 time (sec'
'00
125
140
160
'I
Figure 7. Increase in resolution with the square root of time is shown for the TRITC-I/L, TRITC-GVGL, and TRITC-GGI/GGL analyte pairs. The lines passing through the points are the best least squares fRs.
discussed above, several minor contaminant peaks were also resolved over the course of the TRITC-GL/GI and TRITCGGL/GGI separations. The increase in resolution with time for the three pairs of analytes separated above can be seen in Figure 7. The data points shown for the TRITC-I/TRITC-L separation begin at the resolution value calculated for the second forward pass because the first forward pass peak was just off scale which
Table 2. Unoar R~grrrdonEquatlonr for tho Tkorotkal Plate Incroaso wtth Tkm for the Soparattons of TRITGDerivatlzed IIL, QIIQL, and WIIQQL
anal*
I L GI GL GGI GGL
fitted equation for experimental data
r
-8.5 X 10' + 130t 5.8 X 10' 122t 1.OX l0'+ 128t -1.4X l0'+ 128t 5.2X le+115t 4.7 X l0'+ 118t
0.999+ 0.999 0.999+ 0.998 0.997 0.999+
+
% elope slope difference predicted ((experimentalby eq 8 predicted) X 100) 119 119 119 123 116 115
+9.2 +2.5 +7.0 +3.9 +0.1 +2.6
prevented a proper analysis of the peak. The first data points for the dipeptide and tripeptide separations clearly do not fall on the line with subsequent data points. This is due to the fact that the first set of data points in each separation has not been exposed to any of the excess band broadening induced by the parabolic flow profile. The rate of separation at this first point, therefore, is greater than at any subsequent data point where the excess band broadening has occurred. A more thorough examination of this fact will be made in a future paper. With the removal of these two points, however, all of the regression equations yield correlation coefficients of 0.999+. The error in the resolution measurements for the experimentally obtained data was under 5%. The rather large differences in the rates of resolution increase for the three analyte pairs can be attributed to the fact that the structural differences between each analyte in a pair become more apparent as the leucine and isoleucine residues are removed spatially from the larger TRITC tag using small, freely rotating glycine residues. These structural differenceschange the hydrodynamic radii resulting in greater electrophoretic mobility differences between the analytes of each pair as the peptide length increases. Efficiency calculations for all six of the analytes were made (Table 2) and compared to that predicted using eq 8. Because of the many experimentally measured parameters used to construct eq 8, an overall uncertainty of 5-15% in the values obtained for the slope is clearly possible. The diffusion coefficient ( D ) used for calculating the predicted efficiencies in eq 8 can be found in Table 1. The error in the efficiency calculations for the experimentally obtained data was less than 5%. The rate of plate increase is similar for all of the analytes involved and is within experimental error of the predicted line in all cases (Table 2). The first points for the TRITC-GI/GL and TRITC-GGL/GGI separations were not used in the regression analysis for the same reason as they were not used for the resolution data above. The regression analysis yielded fits to the experimental data of at least 0.997 in all six cases. Although the slopes for the equations fitted to the experimentally obtained data were all within the experimental error of the predicted slopes, the slopes for the experimentally obtained data were in all cases greater than the predicted values(Tab1e 2). The results obtained, therefore, were consistently slightly better than theory predicts. A possible explanation for this consistent positiveerror is that thecapillary temperature increases slightly over the extended FCCE run time. This temperature rise could be caused by slight Joule heating effects. A rise in buffer temperature of only 1-5 OC would account for the error seen.
5
I8
20
22 lim.,*.SXIO~
,
24
28
21
Flgure 8. FCCE separatlon of TRITCUI in a 5.2-pm1.d. 34.54111long capillaryfor 28 800 8 (or 8 h). The photometergain was Increased by a factor of 10 at the polnt marked by the 1OX on the flgure.
The more difficult separation of TRITC-derivatized FLEE1 and FLEEL was then attempted using the above-described system but without success. To increase the efficiency and resolving power of the separation, the capillary diameter was decreased from 10.6 to 5.2 rm, the field strength was increased from 500 to 870 V/cm, and the reverse velocity was slowed. The increase in field strength was brought about by a decrease in the capillary length from 60.5 to 34.5 cm. The results for the TRITC-I/TRITC-L pair can be seen in Figure 8. The narrower pairs of peaks indicate forward passes across the detection window. The wider pairs indicate the reverse passes. The reverse pass peak pairs are wider in time because the velocity of the reverse passes is approximately one-sixth that of the forward passes. (The forward (electrokinetic) velocity was 0.0943 cm/s and the reverse velocity was 0.001 5 cm/s). Under these conditions the U C ~ / U Bratio ~ is equal to 1.8 ( ~ 2 = 2 . 8 ~ ~a ~decrease ); of almost 4-fold over the U C ~ / U Bratio ~ found for the 10.6-rm4.d. capillary. It might be noted that one of the peak pairs disappears approximately halfway through the separation (Figure 8b). The computer program used to isolate the analyte pair to be resolved determines whether to turn the pressure on or off depending upon the position of the larger of the two peaks. As such, once the smaller analyte peak becomes 25% of the length of the capillary away from the larger analyte peak, it will migrate or get pushed off the end of the capillary depending upon which side of the main peak it is on. During the thirteenth pass the smaller analyte peak (TRITC-I) was pushed off of the capillary into the high-voltage buffer reservoir. The efficiency of the TRITC-L/I separation in the 5.2pm4.d. column increased 4.5 times over the efficiency of the separation in the 10.6-pm4.d. 60-cm-long capillary from an average of 126 plates/s to an average of 573 plates/s (Figure 9). The efficiency data for the TRITC-I/L separation in the 10.6-rm4.d. capillaries has been included in Figure 10 for comparison purposes. The percent difference between the slopes of the experimentally derived and predicted equations are 10.1%for the TRITC-I analyte and 2.8% for the TRITC-L analyte in the 5.2-pm capillary. Again both of these values are within experimental error. After an 8-h period of separation, the number of theoretical plates calculated for the Ana&kal Chemistry, Vol. 68, No. 7, April 1, l9Q4
861
~
~
~
1
3 0 0
0
TRITC-I for 5.2 Ktn 1.d. capillary TAITC-L for 5 2 Prn i.d. capillary Predicted for 5.2~17' 1.d. capillary TRITC-I for 1 0 . 6 ~ " 1.d. capillary TRITC-L for 10.6pm i.d. caplllary Predicted for 10.6 pm i.d. capillaly
5
10
15
time (sec
20
X O I
25
'
Flguro 9. Rate of theoretical plate increase with time for the TRITCI/L analyte pair In 5.2-pm4.d. 34.5-cm-long and 10.6-pm4.d. 60.5-cm long capillaries. The 10.6-pm4.d. column data has been added to this plot for comparison purposes.
200
240
time(sec)
280
28 65
28 70 28 7 5 time(secxl0 )
Figure 10. First and last (79th) forward passes of TRITCderivatired FLEEI and FLEEL separation. The resolution between the two analytes on the last forward pass Is 0.93. The total separation time was 28 800 s (or 8 h).
peakleft in thedetection window (TRITC-I) was 17.3 million. The resolving power of the separation increased 2.5-fold, from 0.012 to 0.032 units/s'/*. If after an 8-h separation period 17.3 million plates can be produced as above, it should be possible to separate with a resolution of 1.OO two compounds with electrophoretic mobility differences of only 1 X lo-' cm2/(V s), an order of magnitude lower than with normal CZE separation conditions. The separation of FLEEL and FLEEI was again attempted in the 5.2-pm-i.d. 34.5-cm-long capillary. Figure 10 shows the first and last pass. A resolution between the pair of 0.93 was obtained after 8 h. Further decreases in capillary diameter to 2 pm and the use of higher field strengths should improve the resolving power of FCCE so that a complete separation of FLEEL and FLEEI can be accomplished. A practical barrier to the wider application of this technique involves its use of 5-10-pm-i.d. capillaries and fluorescently 962
Analytical Chemistry, Vol. 66,No. 7.April 1, 1994
tagged samples. Unfortunately, when using untreated, bare fused silica capillaries the use of such small inner diameters is necessary because the excess band broadening due to the pressure-induced parabolic flow profile is directly proportional to the square of the capillary's inner diameter as shown in eq 3. If the TRITC-I/L separations performed above were carried out in 25-pm-i.d. columns with the forward and net reverse velocities equal to one another, the excess band broadening expected (e.g., the uc2/ag2 ratio) would be approximately 40 times that due to longitudinal diffusion alone; if these separations were carried out in 50-pm-i.d. columns, under the same conditions, the excess band broadening would be 175 times that of longitudinal diffusion. This is clearly unacceptable. If, however, one can reduce the electroosmotic flow and, therefore, the counter flow required to reverse the analytes, then the use of larger inner diameter capillaries may become feasible. Equation 3 shows that larger capillary diameters may be used as analyte net migration velocity is decreased without increasing the band broadening expected from the pressure-induced parabolic flow profile. Decreasing the net migrationvelocity by a factor of 10 should allow theseparation performed above in the 5.2-pm capillary to be done in a 52pm capillary. With capillaries of this inner diameter the use of the more universal UV absorption detector would be practical. Such careful balancing of the electroosmotic flow and analyte electrophoretic mobilities, as discussed in the introduction, is very difficult to achieve. Perhaps with time, however, the present limitations of external radial field control of electroosmotic flow will be overcome allowing for an extension of FCCE to UV absorption detection systems. Peak Area Decrease with Time. The rate of peak area decrease with time for the three pairs of analytes separated in the 10.6-pm4.d. 60.5-cm-long column differed significantly from one another. The TRITC-L/TRITC-I peak areas decreased at an average rate of 9.0 and 7.l%/h, respectively. The TRITC-GL/TRITC-GI peak areas decreased at an average rate of 5.9 and 5.7%/h respectively. The TRITCGGL/TRITC-GGI peak areas decreased at an average rate of 3.6 and 3.6%/hour respectively. Several reasons for these decreases in peak area with time can be proposed, including photodegredation, adsorption onto the column walls, and the constant separation of very small contaminant peaks from the two main components. A better understanding of the contribution for each of these is needed if longer separations are to be performed or if the analytes are to be balanced precisely in the detection window. ACKNOWLEDGMENT
This work has been supported by NSF Grant CHE9215320. C.T.C. has been supported by a Department of Education fellowship Contract P200A10047-92. We would also like to thank Doug McManigill of Hewlet-Packard for helpful discussions. Received for review November 4, 1993. Accepted January 11, 1994." Abstract published
in Advance ACS Absrracts, February
15, 1994.