Anal. Chem. 2000, 72, 4608-4613
Electrophoretic Separation of Proteins on a Microchip with Noncovalent, Postcolumn Labeling Yingjie Liu,† Robert S. Foote,* Stephen C. Jacobson, Roswitha S. Ramsey, and J. Michael Ramsey
Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, Tennessee 37831-6142
Proteins were separated by microchip capillary electrophoresis and labeled on-chip by postcolumn addition of a fluorogenic dye, NanoOrange, for detection by laserinduced fluorescence. NanoOrange binds noncovalently with hydrophobic protein regions to form highly fluorescent complexes. Kinetic measurements of complex formation on the microchips suggest that the reaction rate is near the diffusion limit under the conditions used for protein separation. Little or no band broadening is caused by the postcolumn labeling step. Lower limits of detection for model proteins, r-lactalbumin, β-lactoglobulin A, and β-lactoglobulin B, were 1100 for the protein-bound dye.) The relative intensities of covalently and noncovalently labeled proteins depend in part on the relative numbers of amine groups and hydrophobic binding sites in the molecule and will thus vary with the protein. Because of the unknown molarities of commercial SYPRO Red and NanoOrange solutions, no attempt was made to compare the reaction rates at identical dye and protein concentrations. The high molar dye/protein ratios used for labeling with OPA (150:1) and NDA (100:1) gave pseudo-first-order reaction kinetics (data not shown), and the plateau fluorescence levels are assumed to represent maximally labeled protein. Among all of the dyes tested, labeling with NDA gave the highest fluorescence intensity at optimum excitation/emission wavelengths (Table 1), but its reaction with the protein was slow (half-time ∼3 min) under the reaction conditions used. Labeling with OPA gave slightly lower fluorescence with much faster reaction kinetics (half-time ∼12 s). However, the OPA reaction is still slow relative to the 3-s postcolumn labeling time on microchips, and the reaction would be only ∼16% complete under these conditions. SYPRO Red and NanoOrange dyes also had very rapid labeling kinetics with β-lactoglobulin A, with NanoOrange giving ∼50-fold greater fluorescence intensity than SYPRO Red and 29% of the intensity of the NDA-labeled protein (Table 1). The kinetics for these dyes were studied by making a series of identical sample injections on the microchip and monitoring the fluorescence intensity of the labeled protein peak at different distances downstream from the mixing tee. The labeling reaction time was calculated as described in the Experimental Section. Data plots for the two dyes are shown in Figure 2. Reaction half-times of 36 and 110 ms were measured for SYPRO Red and NanoOrange, respectively, under the conditions of the experiments. Electroosmotic flow velocities in both the separation and reagent channels were ∼1 mm/s, so that each solution stream occupied half of the mixing channel after merging at the tee to give a combined velocity of ∼2 mm/s. The time required for dye molecules to diffuse halfway into the sample stream (a distance of ∼11 µm) is estimated to be in the range of 100-200 ms, assuming diffusion
coefficients similar to that of other small dye molecules in aqueous solution, e.g., D ) 4.3 × 10-6 cm2 s-1 for rhodamine B.39 The labeling rate for NanoOrange is therefore within the range predicted for a diffusion-limited reaction. Differences in the electrophoretic mobilities of the dye and protein can also effect the mixing and labeling rates in these experiments; i.e., as the dye diffuses laterally into the sample stream it also migrates axially at a velocity which may differ from that of the protein. The resulting electrophoretic mixing could increase the labeling rate above that predicted for passive diffusion alone and may account for the higher rate observed for SYPRO Red labeling. The detection point was set at 6 mm downstream from the mixing tee in the following experiments, unless otherwise noted, so that the total reaction time for postcolumn labeling was ∼3 s. The effect of NanoOrange dye concentration on the detection of β-lactoglobulin A by postcolumn labeling is shown in Figure 3. The amount of protein injected was kept constant for all dye concentrations by maintaining constant gated injection times and field strengths. The fluorescence intensity (area) of the protein peak was determined from the average of four replicate injections at each dye concentration and increased linearly up to 50% v/v of NanoOrange-DMSO in the labeling buffer solution (Figure 3A). The relative standard deviation (RSD) was 8.9% at the lowest dye concentration (1% v/v) and varied from 2.1 to 4.0% for all other concentrations. The signal-to-noise ratio (S/N) for protein detection increased ∼7.5-fold in going from 1 to 32% labeling solution (Figure 3B). The decrease in S/N at 50% dye concentration may be due to precipitation of the dye in the mixing channel causing increased light scattering and background noise. Solid precipitate
(37) Steinberg, T.; Jones, L.; Haugland, R.; Singer, V. Anal. Biochem. 1996, 239, 223-237. (38) Jones, L.; Haugland, R.; Singer, V. FASEB J. 1996, 10, A1512, Abstr. 2954.
(39) Culbertson, C. T., Oak Ridge National Laboratory, personal communication, 2000.
Figure 3. Effect of NanoOrange concentration on the peak area (panel A) and signal-to-noise ratio (panel B) of β-lactoglobulin A (5.7 µM) detected by postcolumn labeling. NanoOrange concentration is given as the percent (v/v) of commercial dye solution in the labeling solution. Buffer, 100 mM sodium borate, pH 9.3; injection time, 0.2 s; field strength, 500 V/cm; detection point, 6 mm downstream from mixing tee. Peak areas are the average of four meaurements; (error bars were within marker dimensions). S/N values were calculated from three determinations.
Analytical Chemistry, Vol. 72, No. 19, October 1, 2000
4611
Figure 4. Variation of plate height for Rhodamine B detected 1 mm above (H-1) and 1 mm below (H+1) the mixing tee as functions of separation field strength. Each data point was the average of three measurements; (error bars not shown were within marker dimensions).
Figure 5. Separation of R-lactalbumin (peak 1), β-lactoglobulin B (peak 2), and β-lactoglobulin A (peak 3) with postcolumn labeling and detection on microchip. Protein concentration, 100 µg/mL each; injection time, 0.4 s; labeling solution, 16% (v/v) of NanoOrange solution in buffer; other conditions as in Figure 3.
was observed at the mixing tee when dye concentrations of 32% or more were used. It should be noted that these concentrations of NanoOrange are much greater than those normally used for protein quantitation in buffer (0.5% v/v).33 The higher concentrations used here for maximum sensitivity may reflect weaker binding of the dye to native β-lactoglobulin A relative to its binding to the detergent coating of proteins in quantitation buffer. As in the previous study of postcolumn labeling of amino acids on microchips,25 the effect of reagent mixing on column efficiency was tested by injecting rhodamine B into the separation column and comparing the plate height of the fluorescent peak at points 1 mm above (H-1) and 1 mm below (H+1) the mixing tee. Figure 4 shows the relative plate heights as a function of separation field strength. The injection plug length was kept relatively constant by varying the injection time in inverse proportion to the field strength. To mimic the protein postcolumn labeling conditions, 16% NanoOrange labeling solution was pumped into the mixing channel at a 1:1 volume ratio to the flow from the separation channel by applying equal field strengths in both channels. Under these conditions, both the length and velocity of an analyte zone 4612 Analytical Chemistry, Vol. 72, No. 19, October 1, 2000
Figure 6. Variation of plate height with the distance downstream from the mixing tee for R-lactalbumin, β-lactoglobulin B, and β-lactoglobulin A. Injection time, 0.2 s; other conditions as in Figure 5. Each data point was the average of three individual measurements (error bars not shown were within marker dimensions).
are increased by 2-fold below the mixing tee, and peak width in the time domain is unchanged unless other band-broadening effects occur between the two detection points. The values of H-1 and H+1 were similar from 100 to 500 V/cm, demonstrating that no significant band broadening occurs as a result of mixing. Both H values declined with increasing field strength up to 300 V/cm and remained relatively constant ((0.2 µm) up to 500 V/cm. A field strength of 500 V/cm was therefore used in all other experiments. Protein Separations. Figure 5 shows the microchip separation of R-lactalbumin (pI 5.4), β-lactoglobulin A (pI 5.13), and β-lactoglobulin B (pI 5.23) in 100 mM sodium borate buffer (pH 9.3), using 16% (v/v) NanoOrange solution in the run buffer as a postcolumn labeling reagent. The separation was completed within 36 s. All three proteins are negatively charged at this buffer pH and elute in order of decreasing pI. β-Lactoglobulin A and β-lactoglobulin B are separated with a resolution of 2.65 at Lsep ) 31 mm, despite having the same molecular weight and a difference of only 0.1 pI unit in their isoelectric points. Figure 6 shows the plate heights of the three proteins at different detection points from 1 to 6 mm below the mixing tee. The plate heights for β-lactoglobulin A and β-lactoglobulin B remain fairly constant with increasing distance from the tee. Interestingly, however, the plate height of R-lactalbumin decreases from 52 µm at 1 mm to a minimum of 20 µm at 4 mm, with no further decrease to 6 mm. This increase in efficiency for R-lactalbumin in the mixing channel may be due to a stacking effect associated with buffer changes at the mixing tee or to protein/ dye complex formation, but it is not clear why only the R-lactalbumin peak is affected. The limits of detection (S/N ) 3) for the three model proteins were determined for microchip CE separations using NanoOrange postcolumn labeling. A 25% solution of the dye was used to avoid the precipitation observed at higher concentrations. Fluorescence was measured 6 mm from the mixing tee using an argon ion laser (488 nm, 20 mW) for excitation. A 488-nm notch filter was used in addition to the band-pass filter to further reduce background signal due to transmitted excitation light. Because the fluorescence response may not be linear at lower protein concentrations,33
minimum concentrations giving a S/N of ∼3 were experimentally determined by analysis of serially diluted protein mixtures, rather than by extrapolation from analyses at higher concentrations. Injection volumes were calculated from the gated injection time, EOF velocity, and channel dimensions. The molar limits of detection for the model proteins were as follows: R-lactalbumin, 85 nM; β-lactoglobulin A, 70 nM; and β-lactoglobulin B, 70 nM. The three proteins had nearly identical mass limits of detection of ∼0.45 pg (24 amol of β-lactoglobulin or 32 amol of R-lactalbumin). These LODs are approximately 10-20-fold higher than those reported for the conventional CE of proteins precomplexed with near-IR dyes.31 The LOD for β-lactoglobulin A, which was used in both studies, was ∼18-fold higher for NanoOrange versus IR125 labeling, when the subunit molecular weight (18 400) was used to calculate molar sensitivity. The higher LOD for NanoOrange is presumably due in part to the higher background and noise caused by the continuous presence of free dye in the postcolumn labeling stream. However, the limit of detection for NanoOrange/ β-lactoglobulin A with postcolumn labeling on microchips was 1020-fold lower than that reported (0.4 fmol-7 pg) for on-column labeling of the same protein with TNS in conventional CZE analysis.30 (In comparison, the sensitivity for detection of a number of proteins in SDS gels is 1-10 ng/band by staining with fluorescent SYPRO dyes or by silver staining and 30-100 ng/ band for staining with Coomassie blue.)37 On-column labeling with NanoOrange was also studied (data not shown). In this method, 16% NanoOrange dye was added to the separation buffer. Under the same operating conditions as used in postcolumn labeling, the separation time was approximately doubled, presumably due to increased negative charge in the protein/dye complexes. The presence of NanoOrange in the separation buffer also decreases the separation efficiency dramati-
cally. The number of theoretical plates for β-lactoglobulin B was 1700 for on-column labeling versus 11 000 for postcolumn labeling. Also, the resolution between β-lactoglobulin A and β-lactoglobulin B was 1.34 for on-column and 1.8 for postcolumn labeling. The lower separation efficiency of on-column labeling may be due in part to its mixing geometry. Because the injected protein plug occupies the entire cross section of the separation channel, mixing between the sample and dye can occur only from the ends of the injection plug. Differences in the mobilities of labeled and unlabeled proteins can cause broadening of the analyte zones before the labeling reaction is completed. Diffusive mixing occurs rapidly in the postcolumn labeling format, and band distortion caused by the difference in mobilities of labeled and unlabeled protein is minimized. The fast reaction kinetics and high fluorescence yields for NanoOrange labeling make this dye suitable for postcolumn labeling of serum proteins separated on microchips. ACKNOWLEDGMENT This research was sponsored by the National Cancer Institute under Grant CA 83238-02. Oak Ridge National Laboratory is managed and operated by UT-Battelle, LLC, under Contract DEAC05-00OR22725 with the U.S. Department of Energy. Y.L. was supported by the Oak Ridge National Laboratory (ORNL) Postdoctoral Research Associates Program, administered jointly by ORNL and the Oak Ridge Institute for Science and Education. The authors thank Judith Eggers and John Cockfield for assistance in microchip fabrication.
Received for review May 31, 2000. Accepted July 20, 2000. AC000625F
Analytical Chemistry, Vol. 72, No. 19, October 1, 2000
4613
