Anal. Chem. 1999, 71, 4014-4022
Factors Affecting Quantitative Electrokinetic Injections from Submicroliter Conductive Vials in Capillary Electrophoresis Robert R. Fuller and Jonathan V. Sweedler*
Department of Chemistry and the Beckman Institute, University of Illinois, Urbana, Illinois 61801
The factors influencing quantitative electrokinetic injections in capillary electrophoresis for custom 340-nL, 10 hµL, and 110-µL stainless steel sample vials have been investigated using a six-analyte mixture containing catecholamines and indolamines. Deleterious sample degradation is increased with smaller sampling vials, decreased capillary-electrode distances, and increased current passed during the injection. Zero-voltage injections from the smallest vials also demonstrate additional injection discrepancies when compared to larger-volume bulk solution injections. These effects are in addition to the electrokinetic bias and complicate the selection of appropriate internal standards. For nanoliter-volume conductive vials, the injection process creates new species and eliminates other electroactive species to such an extent that quantitation becomes problematic. One of the fundamental advantages of capillary electrophoresis (CE) is being able to use and inject from small, limited-volume samples. This feature of CE has resulted in a continual reduction in sample-size requirements. An example of impressive samplesize reduction is the manipulation of single particles, such as cells,1-4 fog droplets,5 and even cellular organelles6,7 into capillaries during injections. Unfortunately, not all samples are amenable to discrete sampling approaches. This has required the further development of sampling techniques adapted to handling extremely small volumes ( OA > Tyr ≈ Trp. As summarized in Table 1, the concept of electrochemical depletion of analytes is consistent with all of our observations, particularly in that the extent of peak discrepancies are analytedependent and that the discrepancies are not observed in larger vials, perhaps due to both limited diffusion from the electrode (container walls) and dilution effects. Furthermore, the fact that OA, which positively migrates at a velocity between that of 5-HT and NAS, is not statistically “depleted”, while the latter two analytes appear to be, does not support the depletion mechanism. If optimal depletion conditions have not been met, then the degree of depletion would be dependent on µi, and OA depletion would be intermediate between 5-HT and NAS. Also, with depletion, integrated analyte signals should reach a maximum when all the (27) Atkins, P. W. Physical Chemistry; W. H. Freeman and Co.: San Francisco, 1978; p 842. (28) Polson, A. Biochem. J. 1937, 31, 1903-1912.
analyte has been sampled even with prolonged injections, but this is not observed, either. Last, in larger-volume vials (10h and 110 µL) with extended sampling periods of several minutes, the same analyte disappearance patterns noted with nanovial sampling are observed (data not shown). These specific observations are explained most simply by electrochemical activity and diffusion on the time scales and sample volumes used in this work and are not explained by depletion mechanisms alone (See Table 1). This degradation occurs independent of vial size, but injections from larger vials (10h and 110 µL) do not show any sampling discrepancy because of the comparably large values of dc-e and large dilution volumes. Changes in analyte composition in nanovials are more readily measured than in larger vials, however, because of the proximity of the capillary to the electrode and the lack of bulk solution to replace analyte-depleted volumes. By using the Einstein-Smoluchowski equation, an approximation can be made to determine the minimum sample volume required to avoid measuring an electrochemically induced sampling discrepancy of 5-10% compared to bulk solution, based on hemiellipsoid and spherical smallvolume vial geometries (Table 2). The calculations in Table 2 do not reflect how hydrolysis affects injection performance due to pH changes and/or formal charge changes on analyte molecules. For smaller inner diameter columns, the volume of material injected and the total current passed during the injection process can be significantly reduced from the 50-µm values assumed here, reducing the minimum required sample volumes. Another possibility for the small-vial measurement discrepancies is that the injection solution matrix is becoming deleteriously modified during the injection. Changes in the sample solution, such as pH changes from water hydrolysis, may affect overall and analyte-specific sampling rates, but are unlikely at these volumes, buffer strengths, and injection conditions. The possibility of pH modification affecting the injection from the nanovials is addressed by varying the electrolyte composition. The effects of ionic strength and buffer capacity on injections from 340-nL, 10h-µL, and 110-µL vials were readily determined for each analyte and injection condition. Representative graphs are presented in Figure 6. While a given data set included at least 27 injections, most of the data did not pass standard normality test requirements (KolmogorovSmirnov test, passing at P g 0.05: 5-HT, P e 0.001; OA, P e 0.001; Analytical Chemistry, Vol. 71, No. 18, September 15, 1999
4019
Table 2. Minimum Sample Volume (nL) Required To Prevent Sampling Errors for Various Electrokinetic Injection Parameters and Sample Vial Geometries
capillary inner radius (R, µM) 0 12.5 25.0 50.0
1.0 kV for 10.0 s (1.6-nL injection) (σ ) (2Dt)1/2 ) 100 µm)a oblate hemihemiellipsoidb spherec 17 19 21 26
17 20 24 33
3.3 kV for 3.0 s (1.6-nL injection) (σ ) (2Dt)1/2 ) 55 µm)a oblate hemihemiellipsoidb spherec 2.8 3.5 4.2 5.9
2.8 3.9 5.2 8.6
a Diffusion coefficient for most amino acids is ∼5 × 10-4 mm2. 2σ was chosen as the distance where diffusion effects become measurable since that is the distance where 5% of the diffused/depleted material will have theoretically diffused at time t, based on a one-sided, onedimensional Gaussian distribution of diffused materials away from the vial walls. Actual percentage of diffused/depleted material will be >5% because of the geometrical constraints of the vial walls. b r1 ) r2 ) R + 2σ; r3 ) 2σ. c r1 ) r2 ) r3 ) R + 2σ.
Figure 6. Effect of vial size, buffer concentration, and injection voltage on 10-s injections of NAS. The amount of NAS injected was decreased when the electrically conductive nanovial was used (5HT showed similar results). dc-e ) 50 ( 20 µm.
NAS, P e 0.001; Trp, P ) 0.0031; Tyr, P ) 0.098) for a three-way ANOVA using buffer strength, injection amount, and vial size as testing levels. Therefore, a nonparametric Friedman repeated measures one-way ANOVA based on ranks was used to determine the effect of treatments on the quantitative measurements (all df ) 2). As expected, each of the different injection voltage conditions (1.0-, 2.0-, and 5.0-kV injections) produced results that were significantly different from each other for every analyte (χ2 ) 18.00, p e 0.001; SNK pairwise comparison, all p < 0.05). Small vial size (340 nL) produced statistically different results for 5-HT and NAS only (χ2 ) 14.89 and 13.556, both p e 0.001; SNK pairwise comparison, p < 0.05); there was no statistical difference in results when comparing 10h- and 110-µL vials for 5-HT and NAS, and there was no dependence on vial size at all for OA, Trp, or Tyr (χ2 ) 2.67, 0.22, 4.67, respectively, p ) 0.328, 0.971, 0.107) for the stated injection parameters. These results exactly corroborate those 4020 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999
Table 3. Signal Intensities Normalized by Injection Voltage and Borate Buffer Concentrationa
analyte 5-HT OA NAS Trp Tyr
borate buffer concn (mM) 26c 140 230 1.00 1.00 1.00 1.00 1.00
1.25 ( 0.11d 1.02 ( 0.03 0.97 ( 0.02 0.93 ( 0.01 0.86 ( 0.04
3.11 ( 0.12d 2.39 ( 0.07 1.00 ( 0.02 0.77 ( 0.02 0.51 ( 0.03
electrophoretic mobility µi (mm2‚V-1‚s-1)b 0.023 ( 0.001 0.012 ( 0.001 0.0055 ( 0.0023 -0.0027 ( 0.0007 -0.0059 ( 0.0006
a Injections from the 10 h- and 110-µL vials were compared at various voltages (1.0, 2.0, and 5.0 kV for 10.0 s) and buffer concentrations. Peak areas for each analyte and voltage condition were divided by those obtained from the 26 mM buffer with the same injection condition and vial size. In this manner, the effect of injection voltage and vial size was eliminated; the tabulated values should only demonstrate the effect of buffer concentration. b Electrophoretic mobilities (µ) are computed relative to anthracene used as an EOF marker in 140 mM borate buffer, µanthracene ) 0.000 mm2‚V-1‚s-1 by definition. Uncertainties are standard errors of the mean (SEM), n ) 3. EOF mobility (µeo) was 0.054 ( 0.002 mm2‚V-1‚s-1. c Signals normalized to 26 mM values. Variances in 26 mM values are reflected in 140 and 230 mM ratios. d Uncertainties shown are SEM, n ) 6, except for 5-HT uncertainties where n ) 3.
observed when electrokinetic injection parameters and vial-size effects were investigated. Borate buffer strength (26 mM) was associated with statistically different results for all analytes except for NAS (χ2 ) 2.89, p ) 0.278), which showed no dependence on buffer strength (5-HT χ2 ) 14.89, p e 0.001; OA χ2 ) 13.56, p e 0.001; Trp χ2 ) 18.00, p e 0.001; Tyr χ2 ) 18.00, p e 0.001). SNK analyses (p < 0.05) on Trp and Tyr data showed that all three buffer strengths produced statistically different injection results, while analyses on 5-HT and OA showed significantly different results only for the 26 mM buffer. Variation in sample ionic strength is expected to affect electrokinetic injections because of changes in the strength of the electric field passing through the sample solution; lower ionic strength solutions have increased electrical resistance and more intense field lines in the sample.29 Electromigration species are therefore subjected to increased electrokinetic sampling bias when sampled from solutions with low ionic strength. This effect is observed in these experiments, as indicated above, and is most easily seen in samples injected from the 10h- and 110-µL vials where no small-vial analyte depletion can occur. Table 3 clearly shows that the expected electrokinetic sampling biases are observed. By studying the effect of sample buffer capacity, it was determined that pH changes from electrolysis during the sample injections were not responsible for the observed small-vial analyte discrepancies. It was suspected that local, rapid changes in pH between the capillary tip and the electrode (vial) surface may have accounted for the varying degree of analyte sampling from the confined sample geometries used here, an effect seen previously.17,30 By altering local pH, an altered analyte µi may have resulted. There are several reasons to believe that pH changes in the sample solution during the injection were not significant enough to cause the observed small-vial discrepancies. First, there would have been a deviation of small-volume samples, as a function of buffer concentration, that would not parallel the ionic strength (29) Kuhn, R.; Hoffstetter-Kuhn, S. Capillary Electrophoresis Principles and Practice; Springer-Verlag: Berlin, 1993. (30) Guttman, A.; Schwartz, H. E. Anal. Chem. 1995, 67, 2279-2283.
trends observed for the larger vials; nanovials containing buffers with low buffering capacities would be most affected by pH changes. Results indicate that all of the nanovial injections mirror the relative patterns of ionic strength effects demonstrated by larger-volume vials. In addition, if pH changes resulted in increased protonation of some of the analytes, these analytes would have been more positively charged and would have migrated more quickly toward the capillary and been injected in larger amounts; measurements indicate that no such signal increase occurs for any analyte and that signal decreases actually occur for 5-HT and NAS. On the basis of the buffering capacity concept, it is possible to estimate the minimum sample volume required to buffer electrogenerated H+ or OH- creation during an injection. For example, suppose the buffering capacity of a buffer is exceeded when the moles of electrogenerated protons approach 50% of the initial moles of buffer present. Also, suppose that a 10-s, 1-kV injection with i ) 1 µA could generate 1 × 10-10 mol of e-. This would suggest that at least 2 × 10-10 mol of buffer must be present during the injection to titrate the generated H+; at 0.1 M, a minimum buffer volume of 2 nL would be required to neutralize the electrogenerated H+. Decreases in buffer capacity and analytedependent pH tolerance levels would increase the required buffer volume or buffer strength to stabilize the pH during the injection. Furthermore, if a 2-nL solution were sampled, consideration must be given to the capillary-electrode distance. However, the sample volume required to reduce small-volume electrokinetic injection discrepancies based on generalized analyte diffusion is sufficiently large to minimize pH-based discrepancies for a 1.0-kV, 10.0-s injection (Table 2). Null (spontaneous or ubiquitous) injections, caused simply by dipping the capillary into the vial and removing it, have long been known in CE.31-33 Fishman et al.32 developed a model of this effect based on the interfacial pressure of the drop remaining on the capillary tip after the capillary is withdrawn from solution. However, their model of null injections does not depend on injection vial volume but indicates that null injections depend on capillary size, shape, and even the hydrophobicity/coating on the capillary tip. Our null injection comparisons revealed that the signal recorded for each analyte after injection from the nanovials was higher than that obtained for the 110-µL wells both before and after adjustment to correct for gravity flow (∼25 pL) of analyte solution. One advantage of our experimental system is the extreme precision of capillary withdrawal rate and position afforded by the micromanipulator system employed for capillary injection. The result that greater amounts are injected from smaller vials is in contrast to reports of reduced quantities of analyte being hydrodynamically injected from nanoliter volume vials, attributed to surface tension behavior.31 Here, when considering every analyte as an item for comparison between the two experimental treatments (injection from the 340-nL and 110-µL vial sizes), a paired, two-tailed t-test showed that the higher values obtained in the 340nL nanovial injections were significant (df ) 5, P < 0.05 for all injection pairs). It is logical, however, that small vials with higher surface tension may deliver more solution into a CE capillary to (31) Rose, D. J.; Jorgenson, J. W. Anal. Chem. 1988, 60, 642-648. (32) Fishman, H. A.; Amudi, N. M.; Lee, T. T.; Scheller, R. H.; Zare, R. N. Anal. Chem. 1994, 66, 2318-2329. (33) Grushka, E. M.; McCormick, R. M. J. Chromatogr. 1989, 471, 421-428.
relieve the high surface tension caused by the greater curvature on the vial surface. This effect is in addition to the electrochemical and electromigration depletion effects described previously. While 1-500-nL sample volume inlet vials are low-volume, multiple injections are possible as one is only injecting 0.1-10% of this volume for each separation. Performing multiple assays from a single vial is advantageous in order to obtain statistics on a sample, to allow standard additions techniques for improved quantitation, and to allow spiking methods to confirm analyte identity. However, multiple injections from the same reconstituted nanovial sample resulted in a continual decline and eventual disappearance (below the limits of detection) of all the analytes studied. Thus, multiple injections from a single nanoliter sample vial are not possible with our arrangement. EOF sampling was not responsible for this behavior since this form of sampling only accounts for 30-µm-i.d. capillaries is problematic and prone to artifacts. Quantitation from larger sample volumes, from several hundred nanoliters or even microliters, can also be affected, but probably only in cases of close proximity of the electrode to the sampling capillary or for large values of ∫i dt during the injection. This points out the need for precisely controlled capillary and electrode positioning between injections, especially for smaller volume vials. It is possible that electrochemical processes can generate an easily detected analyte, in which case, small-volume sampling and extended injections may be advantageous. It was observed that null-electrokinetic injections (0.0 kV) from nanovials tended to produce sampling discrepancies for all analytes when compared to bulk solution null-electrokinetic injections, perhaps due to Analytical Chemistry, Vol. 71, No. 18, September 15, 1999
4021
surface tension in the nanovials. This behavior may have an impact on quantitation when standards and samples are injected from vials with dissimilar volumes and are injected without correcting for these differences. In summary, an electrochemically mediated electrokinetic injection sampling bias is present with analytes with oxidation potentials lower than the solvent or electrolyte for small volumes of sample solution (10 µL). Electrochemical analyte degradation can occur to such an extent that quantitation may no longer be possible for analytes contained in submicroliter sample volumes. The degradation of sample species may also lead to incorrect qualitative assessment of the original sample by introducing or eliminating species of interest. As the irreproducibility dramatically depends on the capillary tip position within the nanovial, automated systems capable of (10 µm capillary tip positioning within the vial are important. Last, this work demonstrates several of the difficulties in choosing internal standards to aid quantitation using CE with nanoliter-volume vials. Internal standards have been employed to (34) Floyd, P. D.; Moroz, L. L.; Gillette, R.; Sweedler, J. V. Anal. Chem. 1998, 70, 2243-2247.
4022 Analytical Chemistry, Vol. 71, No. 18, September 15, 1999
aid quantitation after derivatization or other manipulations prior to the CE injection.9,34 For larger vials, Dose and Guiochon described the need for two internal standards to account for the effective volume injected and mobility of each ion.14 We have shown that one must match these parameters and the electrochemical properties of the analyte, making selection of appropriate standards much more difficult. In addition, due to these effects and because the null injection magnitude depends on vial size, one must perform all calibration separations using the same vials as those used for the analyte separations. Knowledge of the cognizant factors affecting quantitation in such small-volume vials allows appropriate experimental controls to be developed to monitor such effects. ACKNOWLEDGMENT The authors gratefully acknowledge both NIH (Grant NS31609) and NSF (Grant CHE-9622663) for financial support. R.R.F. acknowledges a 1998 University of Illinois Graduate Fellowship. Received for review February 3, 1999. Accepted July 13, 1999. AC990116H
