Anal. Chem. 2000, 72, 2080-2086
Nonaqueous Capillary Electrophoresis of Fatty Acids Derivatized with a Near-Infrared Fluorophore David L. Gallaher, Jr. and Mitchell E. Johnson*
Department of Chemistry and Biochemistry, 308 Mellon Hall, Duquesne University, Pittsburgh, Pennsylvania 15282-1530
Saturated linear fatty acids, derivatized with a nearinfrared absorbing fluorescent dye, were separated in 100% methanol with 12.5 mM tetraethylammonium chloride added as a charge carrier. Separation at 380 V/cm was acceptable for acids that differed in length by a single carbon. The labeled linear fatty acids behaved as random coils in the nonaqueous separation medium, as shown in a fit to a simple theoretical expression. However, even in 100% methanol with a trimethylsilylated capillary, significant adsorption to the capillary wall occurred, which reduced resolution and slowed the separation. Addition of water to the methanol medium caused significant differences in separation behavior of high molecular weight acids (>C16). Addition of a cetyltrimethylammonium bromide surfactant to the separation medium dynamically coated the capillary and greatly improved the separation. The surfactant also interacted with the acyl tail, apparently causing it to collapse. Resolution in an optimal separation medium (20 mM surfactant) ranged from 1.6 to 1.1, depending on chain length, and theoretical plate heights were under 4 µm (N > 105). Resolution was more than adequate to separate stearic (C18:0) from oleic (C18:1) acid, as well as other unsaturated C18 homologues. Nonaqueous separation media are gaining in popularity as a means of expanding the range of mixtures separable by capillary electrophoresis (CE).1,2 Acid-base dissociation constants, capillary wall ζ potentials, and compound solubilities, among others, can be altered dramatically in nonaqueous solutions. Separation of the so-called neutral lipid compounds (which include fatty acids)3 requires nonaqueous solutions to solubilize compounds with chain lengths beyond approximately C12, depending on degree of unsaturation, oxidation, and ionization of the headgroup. Capillary electrophoresis was recently explored as an alternative to gas chromatography (GC) and liquid chromatography (LC) for the separation of fatty acids.4-14 GC is an excellent technique for saturated fatty acids because of its high efficiency, but it is very * Corresponding author. E-mail:
[email protected]. Voice/fax: (412) 3965278. (1) Miller, J. L.; Khaledi, M. G. In High Performance Capillary Electrophoresis; Khaledi, M. G., Ed.; John Wiley & Sons: New York, 1998; pp 525-555. (2) Sahota, R. S.; Khaledi, M. G. Anal. Chem. 1994, 66, 1141-1146. (3) Sweeley, C. C. In Lipids; Lowenstein, J. M., Ed.; Academic Press: New York, 1969; pp 254-267. (4) Collet, J.; Gareil, P. J. Chromatogr. A 1997, 792, 165-177. (5) Drange, E.; Lundanes, E. J. Chromatogr., A 1997, 771, 301-309.
2080 Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
sensitive to compound polarity, the need for multiple derivatizations with oxidized species can make some analyses problematic, and run times are often 1 h or more for long-chain fatty compounds. LC suffers from lower efficiency but is generally more versatile with respect to oxidized compounds.15 CE has the potential to combine the best features of both, but the solubility issue forces the use of a high amount of nonaqueous solvent in the separation medium. There have been a number of reports of separation of fatty acids by capillary electrophoresis in a variety of separation media, including those with organic modifiers,5,7,9,14 often with additional solubilizing agents such as micelles4,6,12,13 or cyclodextrins.10 Very long chain fatty acids require nonaqueous systems.7 Nearly all of these systems employ indirect UV absorbance detection, which suffers mainly from a moderate dynamic range and detection limits in the micromolar range. Analysis of fatty acids from biological samples often requires much lower detection limits. We have described a means of derivatizing carboxylates with a near-infrared absorbing fluorophore.16 The dye itself has a permanent positive charge even in nonaqueous solvents of moderate dielectric constants, which makes it possible to use the usual cathodic detection polarity in a free-zone mode. The chemistry was developed specifically for nonaqueous derivatization, and the quantum efficiency of the dye is enhanced in nonaqueous solvents.17 Fluorescence at near-infrared wavelengths has a number of other advantages, such as low background and simple, robust instrumentation.18 The use of nonaqueous media in CE with these dye molecules has been described by Flanagan et al.19 We describe here the parameters affecting free-zone separation of fatty (6) Gutnikov, G.; Beck, W.; Engelhardt, H. J. Microcolumn Sep. 1994, 6, 565570. (7) Haddadian, F.; Shamsi, S. A.; Warner, I. M. J. Chromatogr. Sci. 1999, 37, 103-107. (8) Koval, M.; Kaniansky, D.; Hutta, M.; Lacko, R. J. Chromatogr. 1985, 325, 151-160. (9) Neubert, R.; Raith, K.; Schiewe, J. Pharmazie 1997, 52, 212-215. (10) Roldan-Assad, R.; Gareil, P. J. Chromatogr., A 1995, 708, 339-350. (11) Salimi-Moosavi, H.; Cassidy, R. M. Anal. Chem. 1996, 68, 293-299. (12) Schmitz, O.; Gaeb, S. J. Chromatogr., A 1997, 767, 249-253. (13) Wan, H.; Blomberg, L. G.; Hamberg, M. Electrophoresis 1999, 20, 132137. (14) Wang, T.; Wei, H.; Li, S. F. Y. Electrophoresis 1998, 19, 2187-2192. (15) Gutnikov, G. J. Chromatogr., B 1995, 671, 71-89. (16) Gallaher, D. L., Jr.; Johnson, M. E. Analyst 1999, 124, 1541-1546. (17) Soper, S. A.; Mattingly, Q. L. J. Am. Chem. Soc. 1994, 116, 3744-3752. (18) Mank, A. J. G.; Lingeman, H.; Gooijer, C. Trends Anal. Chem. 1992, 11, 210-217. (19) Flanagan, J. H.; Legendre, B. L., Jr.; Hammer, R. P.; Soper, S. A. Anal. Chem. 1995, 67, 341-347. 10.1021/ac9909251 CCC: $19.00
© 2000 American Chemical Society Published on Web 03/24/2000
acids, saturated and unsaturated, in 100% methanol using laserinduced fluorescence detection. EXPERIMENTAL SECTION Materials and Reagents. Linear saturated fatty acids C2 (acetic), C3 (propanoic), C4 (butyric), C5 (valeric), C6 (caproic), C7 (enanthic), C13 (tridecylic), C16 (palmitic), C17 (margaric), C18 (stearic), C20 (arachidic), and C22 (behenic) were obtained from Aldrich (Milwaukee, WI). Linear saturated fatty acids C8 (caprylic), C9 (pelargonic), C10 (capric), C11 (undecanoic), and C12 (lauric), as well as the unsaturated γ-linolenic acid (C18:3), N,N ′-dicyclohexylcarbodiimide (DCC), cetyltrimethylammonium bromide (CTAB), and myristyltrimethylammonium bromide (MTAB), were obtained from Acros Organics (Pittsburgh, PA). Brij 35 and anhydrous N,N-dimethylformamide (DMF) were obtained from Aldrich. Urea, originally from ICN (Costa Mesa, CA), tetraethylammonium chloride (TEAC), originally from Eastman Kodak (Rochester, NY), and stearic (C18:0) and oleic (C18: 1) acids, originally from Fisher Scientific (Pittsburgh, PA), were obtained from house stores. Linolenic (C18:2) and petroselenic (C18:1) acids were obtained from Sigma (St. Louis, MO). Reagent grade organic solvents were obtained from house stores and were purified by fractional distillation in all-glass stills and stored over molecular sieves prior to use. Water (18 MΩ cm) used in preparing CE separation media was purified using a Nanopure system (Barnstead, Dubuque, IA). Equipment. The laser-induced fluorescence (LIF) detection system was constructed as follows: Excitation at 780 nm was provided by an InGaAlAs single-mode laser diode (Sharp LTO27MD), collimated using an aspheric collimating lens (Thorlabs, Inc., Newton, NJ) and circularized using an external anamorphic prism pair (Melles Griot, Irvine, CA). The excitation beam was passed through a three-cavity interference band-pass filter centered at 780 nm (Coherent, Auburn, CA) and focused into the capillary using a 40.0 mm achromat doublet lens (Melles Griot). The average power delivered into the capillary was 700 µW. Fluorescence was collected at 90° with a 40× 0.85 NA microscope objective (Fluor 40, Nikon, Tokyo). The objective focused the fluorescence onto a 600 µm pinhole (Melles Griot) placed in the primary image plane of the objective. The fluorescence was filtered immediately behind the pinhole by an interference band-pass filter at 850 ( 20 nm (Omega Optical, Brattleboro, VT). The remaining fluorescence was collected and focused by a pair of achromat lenses (12.5 mm diameter, 24 mm EFL, Rolyn Optics, Covina, CA) onto the active area of an actively quenched photon-counting avalanche photodiode (SPCM-AQ-131, EG&G Optoelectronics Canada, Vaudreuil, Quebec, Canada). The SPCM photodiode also amplified and discriminated the photoelectron pulses, and the TTL output was sent to a multichannel scaler card (MCS II, Oxford Tennelec/Nucleus, Oak Ridge, TN) residing in a Dell Optiplex 466/MXe. Bin width was 250 ms (4 Hz sampling rate). Data collection was controlled by software supplied with the MCS hardware; data were further analyzed using Igor Pro (Wavemetrics, Lake Oswego, OR). Capillary Electrophoresis. Separations were performed in a 50 µm internal diameter, 79 cm (45 cm injection to detection) bare fused-silica capillary (Polymicro Technologies, Phoenix, AZ). Detection windows were prepared by removing a section of the
polyimide coating with a low-temperature flame. New capillaries were flushed sequentially with distilled, deionized water for 30 min, 1.0 M NaOH for 30 min, and distilled, deionized water for 30 min with a syringe pump (Harvard Apparatus, Natick, MA). Capillaries were subjected to a wall deactivation procedure as follows: The capillaries were flushed sequentially with methanol for 10 min, anhydrous acetone for 10 min, and anhydrous chloroform for 10 min. They were then flushed overnight with neat chlorotrimethylsilane. Excess chlorotrimethylsilane was removed by vacuum, and the capillaries were washed with anhydrous chloroform (10 min), followed by acetone (10 min) and methanol (10 min). The capillaries were then filled with the separation medium and allowed to stabilize under high voltage. They were flushed with fresh methanol for at least 30 min each day before injections. Capillary washings between runs were performed using methanol. High voltage power was supplied by a CZE 1000R instrument (Spellman High Voltage Electronics Corp., Plainview, NY). Injections were performed electrokinetically (3 s, 10 kV). For all separations, the detection end of the capillary was held at ground potential and positive high voltage was applied to the injection end. All separations were carried out at ambient temperature. Derivatization Protocol. Stock solutions of the NIR label were prepared in DMF, and those of various fatty acids and DCC were prepared in chloroform. Fatty acids C2-C13 were prepared at a concentration of 1.0 M, and acids C16-C20, at a concentration of 0.1 M. Acid C22 was prepared at a concentration of only 0.025 M because of the reduced solubility of this longer chain acid. The DCC stock solution was prepared at a concentration of 1.0 M, and the solution of the fluorescent label was prepared at a concentration of 0.1 M. For the derivatization of a mixture, 100 µL samples of the acid solutions were combined in a 4 mL conical glass reaction vial with a magnetic stir bar and thoroughly mixed. A 1.5 mL portion of the DCC stock solution was added, and the mixture was stirred for 30 s at room temperature. A 1.0 mL portion of the fluorescent label was then added, and the resulting mixture was allowed to incubate, with continuous stirring, overnight. At the completion of the reaction period, the mixture was filtered through a 0.2 µm PTFE membrane filter to remove precipitated dicyclohexylurea (DCU), and an aliquot of the filtrate was diluted (2.1 × 105)-fold in the CE separation medium for CE analysis. Final injected concentrations were therefore 0.11 µM (C2-C13), 10 nM (C16-C20, 94% reaction efficiency), and 2. nM (C22, ∼80% reaction efficiency). For the derivatization of an individual acid, 100 µL of the acid stock solution was combined with 100 µL of DCC in a conical glass vial with a magnetic stir bar and the mixture was stirred for 30 s. A 100 µL portion of the fluorescent label was then added, and the resultant mixture was allowed to incubate overnight. The individual reaction mixtures were subjected to the same workup as the full acid mixture. All derivatizations were carried out at ambient temperature. RESULTS AND DISCUSSION Free-Zone Electrophoresis. The series of molecules that was investigated was a mixture of derivatized fatty acids with a planar, charged dye and a fully saturated acyl group. The structure of the conjugate is shown in Figure 1. This constituted a fairly welldefined system. The charge on the dye molecule was not Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
2081
Figure 1. Reaction pathway for the derivatization of fatty acids with a near-infrared-absorbing dye. R can be any side chain, but in this case, it is a linear alkane.
Figure 2. (A-C) Separation of derivatized fatty acids in methanol/12.5 mM TEAC at the indicated applied voltages. Acids were derivatized as a mixture. (D) Retention time as a function of acyl chain length for the 25 and 30 kV separations, with linear least-squares fits.
influenced to any significant extent by the length of the acyl chain, according to molecular model calculations and simple resonance arguments. The chemistry of this coupling system was described previously.16 The fatty acids were separated in a nonaqueous electrophoresis medium consisting of methanol and 12.5 mM tetraethylammonium chloride (TEAC). The capillary was deactivated with chlorotrimethylsilane, but to an undetermined extent; deactivation was not complete, as evidenced by significant electroosmotic flow, even in methanol (see below). Figure 2 shows a typical separation of the fatty acid mixture as a function of increasing applied voltage. Note that the overall mobility varied from run to run, because of changes in electroosmotic flow rate due to dynamic modification of the capillary walls, probably by the lipids themselves (see 2082
Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
below). While there was an overall decrease in migration time with voltage, it was not as regular as would be expected. Net mobility (i.e., the vector sum of electrophoretic and electroosmotic mobilities) can written using a Hu¨ckel type equation for the electrophoretic portion:20
µobs ) µeo + Ze/kπηR
(1)
where µ is the mobility, Z is the charge, e is the unit electron charge, η is the viscosity, and R is the solvated radius. The factor (20) Kenndler, E. In High Performance Capillary Electrophoresis. Theory, Techniques, and Applications; Khaledi, M. G., Ed.; John Wiley & Sons: New York, 1998; pp 25-76.
k in the denominator, which is normally written as 6, was left as a variable, as 6 is only strictly applicable for a sphere that does not interact at all with the solvent.21 The solvated radius can be replaced by an exponential function of molecular weight:
µobs ) µeo + Ze/kπη(MW)R
(2)
where R, the molecular weight exponent, is related to shape (0.33 for a sphere, 0.5-0.6 for a random coil, 0.8 for a long rod, 0.667 for a wide, thin disk, and 1 for a free-draining coil).21,22 To derive useful parameters from this equation, migration times for the fatty acids had to be converted to mobilities. The electropherograms were fit to a sum of Gaussians using nonlinear least-squares curve fitting. The Gaussian centroids, i.e., the migration times, were converted to mobilities by dividing them into the injector-to-detector distance and then dividing the result by the electric field strength. The derived mobility data, which contained both the electrophoretic and electroosmotic terms, were then fit with the following equation, weighted by the inverse of the standard deviation of each point, as obtained from the Gaussian fits:
f(MW) ) a0 + a1/(MW)a2
(3)
The mobility data were all well fit, with χ2 values of 1.2-1.4 for the low-percentage water cases. Following the fit, a0 (electroosmotic mobility) was subtracted to obtain only the electrophoretic mobility. Figure 3 shows a plot of the mobility as a function of molecular weight for mobile phases with varying mole fractions of water (Xwater), and Figure 4 shows the corresponding electropherograms. The overall shape of the mobility vs molecular weight (or carbon number) curve was consistent with that reported by Salimi-Moosavi and Cassidy for alkyl sulfate and sulfonate surfactants.11 Electroosmotic mobility values (a0) ranged from -3 × 10-5 (Xwater ) 0) to +5 × 10-5 cm2/(V s) (Xwater ) 0.36). For a coated capillary in 100% organic solvent, the values were reasonable, being about an order of magnitude below those reported previously for similar systems with bare capillaries.19,23,24 The electroosmotic mobility was well correlated with the ratio of the dielectric constant to viscosity.20 The negative value for the 100% methanol case indicated flow toward the anode, which may have been caused by adsorption of TEAC to the capillary wall, providing a slight positive charge. The addition of water changed the direction of the electroosmotic flow back toward the cathode, perhaps by solubilizing the TEAC. The second parameter (a1) decreased with mole fraction of water and also corresponded reasonably well (correlation coefficient 0.8) with the inverse of the solution viscosity (eqs 2 and 3).23,24 The value for the molecular weight exponent (R or a2) was perhaps the most useful result, as it was dependent on the shape (21) Fu, S.; Lucy, C. A. Anal. Chem. 1998, 70, 173-181. (22) Cantor, C. R.; Schimmel, P. R. Biophysical Chemistry. Part III. The Behavior of Biological Macromolecules; W. H. Freeman and Co.: San Francisco, CA, 1980; pp 1019-1039. (23) Wright, P. B.; Lister, A. S.; Dorsey, J. G. Anal. Chem. 1997, 69, 32513259. (24) Schwer, C.; Kenndler, E. Anal. Chem. 1991, 63, 1801-1807.
Figure 3. Electrophoretic mobility obtained from nonlinear leastsquares fitting of total mobility according to eq 3, at the following mole fractions of water in methanol/TEAC: 0 (circles), 0.11 (squares), 0.20 (triangles), 0.28 (inverted triangles), and 0.36 (diamonds). Data were included in the fits up to a large (>50%) increase in χ2. Error bars were smaller than the symbols and were omitted for clarity. All plots are the averages of duplicate data.
of the molecules (see above). Figure 5 shows the value for the molecular weight exponent as a function of water mole fraction. The magnitudes were quite consistent with data obtained by Fu and Lucy for primary ammonium salts in aqueous buffers,21 but were significantly different from those found by Chiesa and Horva´th for maltooligosaccharides in phosphate buffers,25 which was not surprising, given the different natures of the compounds. However, the increase in overall mobility with mole fraction of water (Figure 3) was the opposite of the trends found elsewhere.11,19,23 It was possible that increased water in the separation medium modified the effective charge on the dye molecule by association. Molecular modeling experiments are underway to understand the nature of the charges and ion pairing in these solutions. In terms of polymer dynamics, a coefficient of 0.5-0.6 indicates a random coil, rather than a free-draining coil.22 This model appeared to be completely justified, given the nature of a saturated carbon chain in a nonaqueous solvent. Comparison with behavior in aprotic or hydrocarbon solvents was not attempted. The decrease in the molecular weight exponent with increasing mole fraction of water could be explained as a balling-up of the long hydrocarbon tail as hydrophobic forces become great enough to overcome steric exclusion forces. Note that a sphere would have a coefficient of 0.33. However, this did not explain the observed minimum, though the magnitude of the exponent for the Xwater ) 0.28 case was somewhat suspect. The value was lower than a smooth minimum would show, and the derived value for the electroosmotic flow rate was inconsistent with other data. This is obvious in Figure 3. The reason for this was not clear. (25) Chiesa, C.; Horva´th, C. J. Chromatogr. 1993, 645, 337-352.
Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
2083
Figure 4. Electropherograms corresponding to the mobility data shown in Figure 3 at the indicated mole fractions of water (Xwater) in methanol/ 12.5 mM TEAC separation media. Acids were derivatized as a mixture. Separation was performed at 25 kV.
Figure 5. Molecular weight coefficient (R in eq 2, a2 in eq 3) as a function of mole fraction of water in water/methanol separation media (open circles, top axis) and as a function of CTAB concentration in CTAB/methanol separation media (open squares, bottom axis). Error bars represent (1 standard deviation, as derived from duplicate measurements (i.e., two separate electropherograms), except for the cases of mole fractions 0 and 0.36 and of 5 mM CTAB, where only single measurements were made. These cases were given standard deviations of 5% for comparison purposes.
Note that, in the higher mole fraction of water cases, the longer chain length molecules had mobilities that deviated significantly from the behavior of the shorter chain molecules (Figure 3). For a water mole fraction of 0.28, the deviation began to occur at approximately C20, whereas for a mole fraction of 0.39, the mobility appeared to deviate beginning at C16. While there were not enough data points after the rolloff for a reliable fit, it was clear that the molecular weight exponent was significantly higher, approaching almost 1 for the high-water case. Similar behavior was observed by Gutnikov for underivatized fatty acids above C7 at water percentages greater than 70% with acetone as the modifier.6 Note also that the peak widths were significantly 2084 Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
increased for the very long chains (Figure 4). This behavior suggested an increasing amount of adsorption for the very long chain lipids. Solubility is a significant issue in the analysis of lipids, as discussed above, even when, as in this case, there is a permanent charge on the molecule. As discussed above, shape changes expected at high water content were not consistent with an increase in the molecular weight exponent; rather, the reverse was true, suggesting that the exponent increase at high molecular weight was a consequence not of a shape change but of adsorption and insolubility. Buffer Additives and Adsorption. In addition to the above arguments, there was evidence that significant adsorption occurred at all chain lengths during the separations, even in 100% methanol. Separations could not be performed on a bare capillary. Silylation of the capillary walls was necessary about once a month, depending on the additives used. Significant broadening and loss of resolution could always be minimized by rederivatizing the wall. Given the opposite charges of the fatty acid-dye conjugates and the capillary wall, this was perhaps not surprising. Flanagan et al. and Legendre et al. demonstrated significant adsorption of these types of dyes.19,26 However, even 100% organic separation medium and capillary derivatization failed to prevent adsorption of longer chain compounds. The addition of surfactant is known to improve capillary wall characteristics. Schmitz and Gaeb were successful in eliminating adsorption of fatty acid hydroperoxides by applying Brij 35 surfactant to a commercial C18-coated capillary.12 Cetyltrimethylammonium bromide (CTAB) and myristyltrimethylammonium bromide (MTAB) were investigated as dynamic coating agents, because they also had positively charged headgroups and were expected to provide electrostatic repulsion at the wall. Results are shown in Figure 6 for CTAB. MTAB results were nearly identical. (26) Legendre, B. L., Jr.; Moberg, D. L.; Williams, D. C.; Soper, S. A. J. Chromatogr., A 1997, 779, 185-194.
Figure 6. (A) Separation of derivatized fatty acids in methanol/TEAC. (B-D) Separation of derivatized fatty acids in methanol/TEAC with the indicated amounts of CTAB. Acids were derivatized individually and pooled. The applied voltage was 25 kV for all runs.
Brij 35 had no effect, nor did the addition of urea. Clearly, the addition of CTAB was successful in eliminating most of the adsorption. The electropherograms were much cleaner with at least 10 mM CTAB. Note that these were not micellar solutions, as the retention time remained linearly correlated with the number of carbons and CTAB was not expected to form micelles in 100% organic phases at such low concentrations. There was a slight increase in overall retention time with increasing CTAB concentration, which was probably due to increasing viscosity. Note that there was also a decrease in overall retention relative to the methanol-only case, which supported the electrostatic repulsion argument. Fits of mobility data derived from the CTAB separations (eq 3) yielded electroosmotic mobilities ranging from -8.2 × 10-5 (5 mM) to -1.9 × 10-4 cm2/(V s) (20 mM). The negative values were consistent with adsorption of the positively charged surfactant to the wall. A linear increase in the magnitude of electroosmotic mobility (r2 ) 0.9993) was observed for an increasing amount of surfactant, which was consistent with increased positive charge density on the capillary wall. Note that the value of electoosmotic mobility with 100% methanol was linearly related to the values from the CTAB experiments. This finding also lends additional credence to the values obtained from the fits. The molecular weight exponents (a2; see Figure 5) were significantly lower than the values for 100% methanol. The regular decrease with CTAB concentration suggested significant interaction between the derivatized fatty acids and the surfactant, despite like charges. For 20 mM CTAB, the exponent was 0.34, the same, within experimental error, as the exponent for a sphere. This suggested that the hydrocarbon tail of the surfactant mediated the collapse of the fatty acyl portion of the derivative, although other measurements would be required to confirm this as physical fact. Resolution and Dispersion. Plots of current as a function of applied voltage were strictly linear through 30 kV (380 V/cm),
which indicated a lack of Joule heating. As shown in Figure 2, resolution increased with applied voltage, as expected for cases where longitudinal diffusion is the primary source of dispersion. However, as noted above, significant adsorption certainly occurred when no surfactant was used. Resolution increased by at least 0.1 unit upon the addition of CTAB at 5 mM, and an increasing amount of CTAB increased resolution by ∼0.05 unit for every 5 mM addition beyond that. Resolution and efficiency were difficult to estimate for the data in Figure 2 because peak fits tended to be somewhat unreliable. A “sacrificial” Gaussian (i.e., an extra Gaussian added to the peak-fit equation) was often required to handle the background, which probably gave peak standard deviations that were an underestimate of the true peak widths. Only the 10 mM or greater CTAB separations were reliably fit. (N.B. This problem did not affect the mobility fits, as peak positions were quite accurately determined, but χ2 values may have been underestimated because the inverse of standard deviation was used to compute χ2.) Resolution was a function of peak position, as expected. Migration time was linearly correlated with carbon number, as was peak variance. Therefore, resolution decreased as the square of the migration time (i.e., carbon number), with a peak resolution, for 20 mM CTAB, being greater than 1.6 between C3 and C4 and approaching a limiting resolution of 1.1 at high carbon numbers. With the present equipment, 500 V/cm is easily obtained by cutting back the postdetection portion of the capillary, which would boost resolution to over 1.5 (i.e., baseline) out to approximately C21, all else being equal. With some care, higher fields can be used, which is an additional advantage of nonaqueous CE.2 Plate heights were acceptable, on the order of 3-5 µm, depending on conditions (i.e., (1-2) × 105 theoretical plates). For the CTAB buffers, plate heights were close to 4 µm (1.8 × 105 theoretical plates). Unsaturated Fatty Acids. Figure 7 shows the separation of a series of cis-unsaturated C18 fatty acids in the methanol/TEAC Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
2085
Figure 7. Separation of C18 saturated (E) and unsaturated (A-D) fatty acid homologues: A, γ-linolenic (C18:39,12,15); B, linoleic (C18: 29,12); C, petroselenic (C18:16); D, oleic (C18:19); E, stearic (C18:0).
separation medium at 25 kV. Excellent separation was obtained for this series. Oleic, linoleic, and γ-linolenic acids had an increasing number of unsaturated sites, but in a regular fashion: all had a double bond at C9, linoleic and γ-linolenic acids had a second double bond at C12, and γ-linolenic acid had a third double bond at C15. The addition of rigid bonds in an otherwise freely rotatable system clearly had an effect on the shape of the molecule, as they all had the same effective charge and were efficiently separated. Efficiencies were similar to those for saturated compounds, ranging from 84 000 theoretical plates for oleic acid to 170 000 plates for γ-linolenic acid. That for stearic acid was 130 000 plates. Adsorption might be expected to depend to some extent on the number of double bonds, but the difference in elution time was the opposite of that observed on silica chromatography
2086 Analytical Chemistry, Vol. 72, No. 9, May 1, 2000
columns, so the separation mechanism could be assigned solely to free-zone electrophoresis. Furthermore, peak efficiencies were greater for a higher number of double bonds. Changing the position of the double bond, for example, between C6 and C9 in petroselenic and oleic acids, respectively, also had a profound effect on the effective molecular radius, as oleic and petroselenic acids were well separated (see peaks D and C, respectively, in Figure 7). In general, it was clear that unsaturation had a much greater effect than the chain length, as the overall elution window for the C18 series was roughly the same as the elution window for the entire C2-C22 saturated series under the same conditions (i.e., Figure 2B). Given the data shown above for the saturated acids, monounsaturates cannot be separated from saturates in a single free-zone run under the conditions used here, although this is not necessarily the case for the polyunsaturates. Much longer runs or the addition of buffer additives such as silver ion may be required for the resolution of all acids in a single run; however, this may not be an achievable goal. CONCLUSION We have shown that fatty acids derivatized with cationic fluorophores are very well behaved in simple free-zone media, allowing a certain degree of modeling of their behavior. Cationic surfactants are required to prevent adsorption to the capillary walls, even in 100% organic media, but the surfactant also interacts with the molecule itself. Further work with molecular modeling will aid in the interpretation of these data. Resolution of saturated fatty acids differing by only one carbon has been achieved in
