Anal. Chem. 2001, 73, 1821-1830
Separation of Barbiturates and Phenylthiohydantoin Amino Acids Using the Thermally Tuned Tandem Column Concept Yun Mao and Peter W. Carr*
Department of Chemistry, University of Minnesota, Smith and Kolthoff Hall, 207 Pleasant Street SE, Minneapolis, Minnesota, 55455
There are many more choices of column type than of eluent type for method development in reversed-phase liquid chromatography. It is common to switch between different column types or between the same type from different suppliers to achieve the desired separations. The key difficulty in modulating band spacing by adjusting the column type is that it is a discontinuous, “hit or miss” proposition. The thermally tuned tandem column (T3C) concept effectively solves this problem by connecting two columns in series and independently controlling the two column temperatures. The columns are chosen to have distinctively different chromatographic selectivities (band spacing), so that the unresolved peaks on one column are separated by the other. The optimized separation in the T3C is achieved by simultaneously tuning the two column temperatures. In this study, we used the T3C combination of a carbon and a conventional bonded phase for the separation of barbiturates and phenylthiohydantoin amino acids (PTH-amino acids). Good peak shapes and comparable retention times were observed on the two phases at room temperature. The selectivities on the two phases are quite different. Baseline separations were easily achieved with the T3C set although neither column could individually resolve all the peaks. We further compared the separation of barbiturates optimized by the T3C approach with that optimized by adjusting the mobile phase. We found that T3C gave a better separation. We believe that the T3C combination of a carbon phase and a bonded conventional reversed-phase material provides a powerful and general method to optimize the separation of various mixtures. Reversed-phase liquid chromatography (RPLC) is the most widely used separation technique for a variety of compound types. Optimizing selectivity (band spacing) is the key for method development in RPLC.1-3 By far, the most popular way to adjust band spacing is through the adjustment of the mobile-phase type * To whom the correspond should be addressed. E-mail:
[email protected]. (1) Snyder, L. R.; Glajch, J. L.; Kirkland, J. J. Practical HPLC Method Development, 2nd ed.; Wiley-Interscience: New York, 1996. (2) Schoenmakers, P. J. Optimization of Chromatographic Selectivity: A Guide to Method Development; Elsevier: Amsterdam, 1986. (3) Glajch, J. L.; Snyder, L. R. Computer-Assisted Method Development for HighPerformance Liquid Chromatography; Elsevier: Amsterdam, 1990. 10.1021/ac001155s CCC: $20.00 Published on Web 03/09/2001
© 2001 American Chemical Society
and percentage of organic modifier.1 In practice, it is also quite common to test different column types to achieve the desired separations. As pointed out by DeStefano and co-workers, it is sometimes more desirable to change the separation selectivity by changing the column rather than the components of the mobile phase.4 The vast majority of separations in RPLC utilize a mobile phase composed of acetonitrile (ACN), methanol (MeOH), or tetrahydrofuran (THF). Acetonitrile is usually the preferred organic modifier because it has low viscosity and favorable UV transmittance. THF is not as widely used as the other two because it is easily oxidized and is easily peroxidized. Zhao and Carr have shown that adding water to THF strongly inhibits peroxide formation which might increase the usage of THF as an organic modifier in RPLC.5 However, when the samples need a clean UV detection background or have certain solubility requirements, the choice of mobile-phase types is even more limited. Compared to the narrow selection of eluent type, the choice of stationary phase is nearly overwhelming. During the past twenty years, considerable progress has been made in developing highquality packing materials. Most modern materials are nearly perfectly spherical, monodisperse particles with good masstransfer properties and high reproducibility. Silica-based packing materials are still the most widely used stationary-phase type. More than 100 octyl and octadecyl stationary phases are commercially available in addition to other types of silica-based materials such as cyano, amino, phenyl, and polar-embedded phases. Snyder found that the greatest difference in band spacing occurs with C8, cyano, and phenyl columns.6 To overcome silica’s poor chemical and thermal stability as well as surface chemical heterogeneity, many more new materials have been developed including polymer-, alumina-, and zirconia-based stationary phases. Zhao and Carr quantitatively compared the selectivity difference between various silica-, zirconia-, and polymer-based stationary phases and found quite different selectivities among these phases.7 At Pittcon 2000 alone, fifteen new column series and forty new reversed-phase chromatography columns were introduced.8,9 Considering the huge number of stationary phases available, it is surprising that so little attention is given to the use of column (4) DeStefano, J. J.; Lewis, J. A.; Snyder, L. R. LC-GC 1992, 10, 130-139. (5) Zhao, J. H.; Carr, P. W. LC-GC 1999, 17, 346-352. (6) Snyder, L. R. J. Chromatogr., B 1997, 689, 105-115. (7) Zhao, J. H.; Carr, P. W. Anal. Chem. 1998, 70, 3619-3628. (8) Major, R. E. LC-GC 2000, 18, 262-285. (9) Major, R. E. LC-GC 2000, 18, 356-374.
Analytical Chemistry, Vol. 73, No. 8, April 15, 2001 1821
Figure 1. Block diagram of the T3C system.
selectivity in LC method development. A couple of issues contribute to the underutilization of column selectivity. First, the selectivity of the stationary phase cannot be changed continuously when the column is changed from one type to another. When varying the mobile-phase composition, the band spacing can be varied continuously, which allows fine-tuning of the separation. However, a change of the column type has a discontinuous effect on selectivity. Although the selectivity for some pairs of compounds might improve, other pairs might worsen. Therefore, by using a new column type, most of the effort in method development expended on the previous column will be wasted. Second, many types of phase have quite similar selectivities, especially octyl and octadecyl bonded phases from the same manufacturer. Ill-considered changes in the stationary-phase type usually result in only minor change in selectivity. To have a significant and beneficial effect, a column with rather different selectivity must be chosen. In a series of recent publications,10,11 we reported a new way to optimize column selectivity in liquid chromatography: the thermally tuned tandem column (T3C) concept. Two columns with drastically different selectivities are serially coupled and independently thermally controlled. Selectivity of the T3C system is “tuned” by adjusting the temperatures of the two columns. T3C solves the selectivity discontinuity problem by coupling the two columns in series (see Figure 1). As we described in detail in the previous paper,10 when the temperatures are varied, the effective result is analogous to adjusting the two column lengths continuously and independently. It is very important to choose two phases with distinctively different selectivities; this maximizes the effect of temperature on the overall T3C system selectivity. Our previous work and many other studies have shown that a carbon-based phase has great differences in selectivity from conventional bonded phases.10-15 Thus, a separation that is difficult on a conventional bonded phase may be much simpler on a carbon phase and vice versa. Despite the selectivity differences, both phases are RPLC materials and are compatible with the entire range of organic solvents used in RPLC. This makes it possible to couple two columns in series and pass the same eluent through both columns. In the previous paper, we used the T3C combination of a carbon-coated zirconia (C-ZrO2) phase and an octadecylsilane (ODS) phase for the separation of triazine herbicides and carbamate pesticides.11 We found that, for both samples, there are huge differences in selectivity between these two columns, which allowed us to optimize the separation by the T3C approach. We also found that those polar environmental samples are much (10) Mao, Y.; Carr, P. W. Anal. Chem. 2000, 72, 110-118. (11) Mao, Y.; Carr, P. W. Anal. Chem. 2000, 72, 2788-2796. (12) Jackson, P. T.; Schure, M. R.; Weber, T. P.; Carr, P. W. Anal. Chem. 1997, 69, 416-425. (13) Knox, J. H.; Ross, P. Adv. Chromatogr. 1997, 37, 73-119. (14) Weber, T. P.; Jackson, P. T.; Carr, P. W. Anal. Chem. 1995, 67, 30423050. (15) Weber, T. P.; Carr, P. W.; Funkenbusch, E. F. J. Chromatogr. 1990, 519, 31-52.
1822 Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
more retained on the C-ZrO2 phase than on the ODS phase. Severely tailed peaks were observed on the C-ZrO2 column at room temperature, and the C-ZrO2 column had to be heated above 90 °C to get reasonable retention and good peak shape. We attributed the difference in selectivity and the long retention time of highly polar solutes to the strong dispersion interactions with the carbon phase and to dipole-dipole, dipole-induced dipole, and π-π interactions between the analyte and the polarizable carbon surface. It is interesting to study whether high retention on the C-ZrO2 column is essential to the difference in selectivity; whether the C-ZrO2 phase can be useful at room temperature and, more importantly, whether the T3C combination of a carbon-type phase and a bonded phase will be applicable to less-polar analytes. In this work, we used the T3C approach to separate 10 barbiturates and 13 phenylthiohydantoin-amino acids (PTH-amino acids) and compared the T3C separations with those on the single phases with mobile-phase manipulation. EXPERIMENTAL SECTION Instruments. All chromatographic experiments were conducted with a Hewlett-Packard 1100 chromatography system, equipped with a quaternary pump, a vacuum degasser, an autosampler, a thermostated-column compartment, a variablewavelength UV detector, and a computer-based Chemstation (Hewlett-Packard S. A., Wilmington, DE). An additional experimental heating apparatus from Systec (Systec, Inc., Minneapolis, MN) was used as the second heating zone. This device consists of a mobile-phase preheater assembly and insulating jacket. It allows the column to be heated to 200 °C. The device also cools the column eluent back to a temperature tolerated by the detector. Analytical Columns. The carbon-coated zirconia particles were provided by ZirChrom (ZirChrom Separation, Inc., Anoka, MN 55303). The C-ZrO2 particle has 1.6% (w/w) carbon content and a particle size of ∼3 µm. A 5 cm × 4.6 mm i.d. C-ZrO2 column was packed in this laboratory using the upward stirred slurry method described in ref 16. The Zorbax SB-C18 octadecylsilane silica (5 µm) particles was purchased from Hewlett-Packard. A 5 cm × 4.6 mm i.d. ODS column was packed using the method described in ref 17. This column was used for the separation of barbiturates. Betasil ODS particles (5 µm) were purchased from Keystone Scientific Inc. (Bellefonte, PA). A 5 cm × 4.6 mm i.d. ODS column was packed using the method described in ref 17. This ODS column was used in the separation of PTH-amino acids. Reagents. All reagents used here are reagent grade or better. ChromAR HPLC grade ACN, MeOH, and THF were purchased from Mallinckrodt Chemical Co. (Paris, KY). HPLC water was obtained from Barnsted Nanopure deionizing system (Dubuque, IA) with an “organic-free” cartridge followed by a 0.2-µm particle filter. The water was boiled to remove carbon dioxide. All solvents were filtered through a 0.45-µm filter (Lida Manufacturing Corp., Kenosha, WI) before use. Barbiturates were purchased from the Theta Corp. (Newtown Square, PA) and PTH-amino acids from Sigma Chemical Co. (St. Louis, MO). Other solutes used in this study were purchased from Aldrich (Aldrich Chemical Co., Milwaukee, WI). (16) Jackson, P. T.; Kim, T.-Y.; Carr, P. W. Anal. Chem. 1997, 69, 5011-5017. (17) Tan, L. C.; Carr, P. W.; Abraham, M. H. J. Chromatogr., A 1996, 752, 1-18.
Figure 2. Structures of barbiturates.
Chromatographic Conditions. All measurements were made at a flow rate of 1 mL/min unless otherwise stated. The injection volume was 1 µL with a concentration of 1-2 mg/mL analyte dissolved in pure acetonitrile. Barbiturates are detected at 220 nm and PTH-amino acids at 254 nm. The retention factors (k′) were calculated from the following equation:
k′ ) (tr - to)/(to - tex)
(1)
where tr is the retention time, to is the dead time, and tex is the time that a solute spends outside the column (in the injector, connection tubing, and detector). The dead time was determined with uracil and acetone for the ODS and C-ZrO2 column, respectively. The value of tex was 0.067 min determined by injecting uracil without any column under otherwise the same chromatographic conditions. The resolution is conveniently estimated based on eq 2, in which Rs is resolution, traverage is the
Rs )
xN tr2 - tr1 4 traverage
(2)
average of the retention times, and N is the theoretical plate number.18 N is assumed to be 5000 for a single 5-cm column and 10 000 for the two-column T3C set. Drylab 2000 software version 3.0.08 (LC Resources, Walnut Creek, CA) was used to predict the optimum separation when mobile phases were varied on the ODS column for the barbiturates samples. RESULTS AND DISCUSSION In the previous paper, we used the combination of an ODS phase together with a C-ZrO2 column to separate polar herbicides and pesticides.11 We found that polar compounds are much more retained on the C-ZrO2 compared to the ODS phase due to the (18) Giddings, J. C. Unified Separation Science; Wiley-Interscience: New York, 1991.
Figure 3. Chromatograms showing the separation of barbiturates on (A) ODS at 30 °C, (B) C-ZrO2 at 30 °C, (C) ODS at 60 °C, and (D) C-ZrO2 at 60 °C. Experimental conditions: mobile phase, 20/80 acetonitrile/water; flow rate, 1 mL/min; detection, 220 nm.
extremely strong dipolar interactions between the polar group and carbon surface. In addition, dramatically different selectivities were observed for those polar compounds on two phases, and T3C successfully separated both mixtures. The goal of the present study is to investigate the chromatographic behavior of less-polar compounds on ODS and C-ZrO2 phases and see whether there is enough difference in selectivity between them to make an effective T3C separation. Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
1823
Figure 4. (A) Plot of log k′ on C-ZrO2 versus log k′ on ODS for barbiturates at 30 °C in 20/80 acetonitrile/water mixture. (B) Comparison of elution order (log k′) of the mixture on C-ZrO2 and ODS.
Separation of Barbiturates on T3C Set. Barbiturate drugs were first introduced in the early 1900s. More than 2500 barbiturates have been synthesized, and 50 have been marketed for human use.19 Barbiturates have been used as sedatives, hypnotics, anesthetics, and anticonvulsants, but they can be addictive and abused. Therefore, it is necessary to develop efficient methods for their separation and identification. In this study, we used the T3C approach to separate 10 barbiturates (see Figure 2 for their structures). They all have the same heterocycle base with slightly different substitutent groups. Figure 3 shows chromatograms of a barbiturate mixture in 20/80 acetonitrile/water on the ODS and C-ZrO2 columns, respectively, at 30 and 60 °C. We chose the 20/80 acetonitrile/ water mixture as the mobile phase because acetonitrile is the most preferred organic modifier and this composition gives a reasonable analysis time. The chromatogram in Figure 3A shows that metharbital and aprobarbital (2/3) coelute on the ODS column at 30 °C. While on the C-ZrO2 column at 30 °C (Figure 3B), butethal and hexobarbital (5/7) coelute. In addition, mephobarbital and secobarbital (9/10) are not baseline resolved. We have shown that, on the C-ZrO2, triazines herbicides and carbamate pesticides have extremely long retention with tailing peaks at room temperature.11 However, Figure 3A,B indicates that all these barbiturates are less retained on the C-ZrO2 column than on the ODS column under the same chromatographic conditions. More importantly, the peak shapes on the C-ZrO2 column are as good as those observed on the ODS column even at room temperature. The three-dimensional structures of all barbiturates (not shown here) indicate that the substituted groups and the heterocycle are not coplanar. Because carbon-based phases interact with solutes predominantly by short-range dispersion interactions, solutes that can easily adopt a planar configuration will be strongly retained. We believe that the nonplanar structure of the barbiturate molecules perturb the dipolar interactions between the polar group and carbon surface, which leads to good peak shapes and similar retention range. Panels C and D of Figure 3 are the separations
of barbiturates at 60 °C on the ODS and C-ZrO2 phases, respectively. We noticed that the separation of compounds 2 and 3 improves and compounds 7-9 become overlapped at 60 °C on the ODS column. However, adjusting the temperature alone cannot achieve baseline separation on the ODS column (data not shown). On the C-ZrO2 column, compounds 5 and 7 did not separate at higher temperatures. The elution order and critical pair are the same on C-ZrO2 at 60 and 30 °C, which indicates that varying the temperature alone on the C-ZrO2 is not an effective way to optimize selectivity for this separation. Despite the similar retention range and good peak shapes, Figure 3A,B did show some changes in elution order when the ODS phase is replaced by a carbon phase. For example, solute 7 eluted after solutes 5 and 6 on the ODS column; while on the C-ZrO2 column, solutes 5 and 6 elute after solute 7. The difference in selectivity between the two phases is quantified by a κ-κ plot (Figure 4A), in which the log k′ of the barbiturates on the C-ZrO2 was plotted versus the log k′ on the ODS column under the same experimental conditions. A good linear correlation in a κ-κ plot means that the same or similar retention mechanism operates on both columns while a poor linear correlation implies different retention mechanisms and selectivities.20 Figure 4A shows that there are fairly good correlations between ODS and C-ZrO2 columns for this set of barbiturates. The correlation coefficient is 0.954, and the standard deviation of the linear fit is 0.130. Comparing the correlation coefficients, we observed for the highly polar compounds on these two types of columns (r ) 0.33 for 10 triazine herbicides and r ) 0.85 for 12 carbamate pesticides) (Figure 4A) that the retention mechanism for barbiturates is very similar on the ODS and C-ZrO2 phases. We need to point out that the C-ZrO2 phase is still considered more hydrophobic than the ODS phase because of the lower carbon content of the C-ZrO2 column (1.6% carbon) as compared to the ODS phase (10% carbon). Although the κ-κ plot shows similar selectivity between the ODS and C-ZrO2 phases, it only represents the general or averaged
(19) Foye, W. O.; Thomas, L. L.; Williams, D. A. Principles of Medicinal Chemistry, 4th ed.; Williams & Wilkins: Baltimore, 1995.
(20) Melander, W.; Stoveken, J.; Horvath, C. J. Chromatogr. 1980, 199, 35-56.
1824 Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
Table 1. Experimental Retention Factors of Barbiturates on ODS and C-ZrO2a k′ (ODS)
k′ (C-ZrO2)
solutes
30 °C
60 °C
30 °C
60 °C
(1) barbital (2) metharbital (3) aprobarbital (4) butabarbital (5) butethal (6) butalbital (7) hexobarbital (8) pentobarbital (9) mephobarbital (10) secobarbital
1.64 5.83 5.83 8.50 10.16 11.64 19.39 21.08 24.36 33.80
1.24 4.35 3.85 5.74 6.77 7.54 13.00 13.13 14.00 19.59
0.74 1.80 3.30 3.83 4.76 6.20 4.76 8.62 15.42 16.62
0.56 1.31 2.15 2.46 3.17 3.96 3.17 5.18 8.71 9.26
a
Measured in 20/80 acetonitrile/water mixture.
Figure 6. Chromatograms showing the separation of barbiturate mixture on T3C with (A) ODS and C-ZrO2 column at 30 °C and 1 mL/min, (B) ODS at 80 °C, C-ZrO2 at 40 °C and 1 mL/min, and (C) ODS at 80 °C and C-ZrO2 at 40 °C and 3 mL/min. Experimental conditions: 20/80 acetonitrile/water mixture; detection, 220 nm. Table 2. Experimental Retention Times of Barbiturates on ODS Columna retention time tr (min) ACN/water Figure 5. Window diagram showing the resolution of the critical pair for barbiturates versus the temperature of ODS and C-ZrO2 column.
selectivity difference. In the T3C approach, what really matters are the selectivity differences for the critical pairs. The critical pair is the pair of compounds that has the poorest separation. As long as the critical pairs on the two phases are different, in other words, if the worst separated pair on one column can be resolved at the other, then it is likely that separation will be better using T3C than either individual column. Figure 4B is a detailed comparison of retention factors for this set of barbiturates on the ODS and C-ZrO2 columns, in which the retention orders on the two phases are plotted against their log k′ values. The lines are the simple connection of the same solutes on the two phases. Each crossover in this plot represents a switch of elution order when one phase is replaced by the other. Figure 4B directly indicates whether the separation of the critical pair on one phase is better on the other phase. For example, the circles that represent compounds 2 and 3 are superimposed on the top row of Figure 4B because they coelute on the ODS column. If we follow the lines of compounds 2 and 3, we observe that these two compounds become further apart when the stationary-phase was switched from ODS to C-ZrO2. Similarly, compounds 5 and 7 coelute on the C-ZrO2 and
solutes (1) barbital (2) metharbital (3) aprobarbital (4) butabarbital (5) butethal (6) butalbital (7) hexobarbital (8) pentobarbital (9) mephobarbital (10) secobarbital a
methanol/water
20/80 30/70
30/70
50/50
1.25 3.13 3.13 4.33 5.07 5.74 9.22 9.98 11.45 15.69
1.82 4.36 5.37 8.02 9.17 10.71 13.38 19.78 13.38 31.31
0.83 1.30 1.37 1.7 1.82 1.97 2.33 2.86 2.24 3.76
0.81 1.47 1.33 1.61 1.79 1.97 2.93 2.78 3.40 3.84
THF/water 20/80 30/70 1.67 2.57 4.75 6.05 7.47 9.26 6.00 13.89 8.88 23.19
1.22 1.36 2.47 2.89 3.25 4.00 2.97 4.85 3.70 6.95
Flow rate, 1 mL/min; detection, 220 nm.
their separation improves dramatically when the ODS phase becomes dominant. This plot shows that, for this barbiturate mixture, when the two columns are connected, the two limiting critical pairs must be better resolved at some intermediate combination of the two columns. However, the overall separation is not necessarily improved because a new critical pair can be formed by the tandem column combination. For example, compounds 6 and 7 switch elution order on the ODS and C-ZrO2 columns; this shows that at some combination of the two column temperatures, these two compounds will have the same retention time and coelute on the T3C set. This is the reason we need to tune the Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
1825
Figure 7. Plots of minimum resolution versus the percentage organic modifier of (A) ACN (B) MeOH, and (C) THF in the mobile phase as predicted from Drylab. Simulated Drylab chromatograms showing the separation of barbiturates on ODS column at optimum mobile-phase conditions: (a) 14%ACN, (b) 52% MeOH, and (c) 32.5% THF.
two temperatures to adjust the differential contributions of the two stationary phases to achieve complete separation of the mixture. Selectivity Tuning in T3C. To locate the best temperatures that give the optimal contribution of each column to the total retention in the T3C set, we need to determine the effect of temperature on the retention on both columns. The retention factors for the four initial runs (Figure 3) are summarized in Table 1. Upon a 30 °C increase in temperature, the retention factors on the ODS and C-ZrO2 decrease on average by a factor of 1.53 and 1.54, respectively. With four initial runs on both columns, we calculated the retention on the T3C set as a function of the two column temperatures. First, we fit the retention factors on each column at two temperatures to the van’t Hoff equation, which assumes a linear relationship between the logarithm of the retention factor and the reciprocal of the absolute temperature:
log k′ ) A + B/T 1826
Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
(3)
Here A and B are solute-dependent constants. Equation 3 allows estimation of the retention times at any other temperatures. Next we calculated the solute retention times on the T3C set at any temperature combination (T1, T2) from the following:
tn,i (T1, T2) ) t1,i (T1) + t2,i(T2) - tex
(4)
where tn.i is the net retention time on the T3C set; t1,i and t2,i are the retention time for solute i on the first and second columns, respectively; and tex is the time that a solute spends outside the column (in the injector, connection tubing, and detector). Finally, the resolution of the critical pair was calculated based on eq 2 at each combination of the two column temperatures. The optimization is performed by locating the temperatures that will give the best resolution. The plot of the minimum resolution versus the two column temperatures yields the three-dimensional window diagram given in Figure 5.
Figure 8. Structures of phenylthiohydantoin-amino acids.
Although the temperatures for the four initial runs are 30 and 60 °C, we extrapolated the temperature range to 80 °C on both columns. A rather flat pattern in the window diagram implies that there is only a slight change in the critical pair on the T3C set for the selected temperature range. It also indicates that no coelution take place in the T3C set although coelutions were observed on both columns individually. A resolution larger than 1.5 is considered baseline resolution; therefore, Figure 5 implies that many combinations of the two temperatures in T3C can provide an adequate separation. The highest point in the window diagram corresponds to 30 °C on both columns (labeled (a) in Figure 5) and the predicted resolution is 4.1. The chromatogram under condition a is shown in Figure 6A. All the solutes are baseline resolved in less than 30 min, and the experimental minimum resolution is 3.3. Figure 5 also shows that increasing the temperature of the C-ZrO2 column above 30 °C has a significantly deleterious effect on resolution. In contrast, increasing the ODS column temperature above 30 °C causes only a small loss in resolution but substantially improves the analysis time. When the C-ZrO2 and ODS column temperatures are 40 and 80 °C, respectively (labeled (b) in Figure 5), the predicted resolution is 3.1. The chromatogram under condition b is shown in Figure 6B. As expected, because of the increase in temperature, the analysis time decreases from 30 to 18 min without any significant loss in resolution. The increase in temperature not only shortens the analysis time but also decreases the mobile-phase viscosity. The lower back pressure allows us to further shorten the analysis time using a high flow rate of 3 mL/min. Figure 6C shows the chromatogram at higher temperatures and higher flow rate. The 10 barbiturates are baseline separated in less than 6 min. This was made possible by the superior selectivity provided by the T3C, which allows us to sacrifice the plate count by using a high flow rate yet maintain the baseline resolution. The predicted retention times agree well with the experimental retention times with an average error of 1.3% at condition a and 0.5% at condition b. In
Figure 9. Chromatograms showing the separation of PTH-amino acids on (A) ODS and (B) C-ZrO2 at 40 °C. Experimental conditions: mobile phase, 30/70 acetonitrile/0.1% trifluoroacetic acid; flow rate, 1 mL/min; detection, 254 nm.
summary, a C-ZrO2 phase can be useful at room temperature. It gives good peak shapes and reasonable analysis times for the barbiturate samples. There are quite different selectivities between the carbon and bonded phases even for less-polar compounds which allows us to completely resolve the 10 barbiturates in less than 6 min using the T3C approach. In addition, only four initial runs are needed for the T3C method development. Comparison of T3C Separation with Mobile-Phase Optimization. We have shown that by varying the temperatures on the T3C set one can achieve baseline separation although varying Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
1827
Figure 10. (A) Plot of log k′ on C-ZrO2 versus log k′ on ODS for PTH-amino acids in 30/70 acetonitrile/0.1% trifluoroacetic acid mixture at 40 °C. (B) Comparison of elution order (log k′) of the mixture on C-ZrO2 and ODS.
the temperature alone on either phase did not give adequate separation. Next, we compare the separation of T3C with varying the eluent composition (φ) and type (ACN, MeOH, and THF) on the ODS stationary phase. To facilitate the prediction of the separation as a function of φ, we used Drylab chromatography optimization software. The software can predict the optimum eluent composition based on two initial experiments at two different φ values. The retention times for the initial separations at two φs and three different mobile-phase types are summarized in Table 2. After inserting the retention times into the Drylab program for each eluent type, we obtained the optimization map (window diagram) as shown in Figure 7. The highest point in the window diagram corresponds to the best separation conditions. Figure 7A predicts that when acetonitrile was used as organic modifier, the best separation occurs at 14% ACN, and the corresponding resolution factor is 1.45. The simulated chromatogram at the optimum condition (labeled a in Figure 7A) is shown in Figure 7a. Although a near-baseline separation is predicted, the analysis time for the separation is ∼40 min. By use of the T3C approach, better resolution can be achieved in an analysis time of less than 6 min (Figure 6C). When MeOH and THF were used as the organic modifiers, the best separation is predicted to take place in pure water (Figure 7B,C). However, the corresponding analysis times are 900 and 326 min, respectively, because water is a very weak eluent. In addition, the predicted resolution in pure water (see Figure 7B,C) is very different (1.3 from MeOH/water mixture and 0.82 from THF/water mixture). The difference is due to the limitation of the linear solvent strength theory (LSST) on which the Drylab calculation is based. LSST assumes a linear correlation between the log k′ and φ over the entire range in eluent composition (0 < φ < 1), but the relationship is only quasi-linear. Big deviations from the model are expected when φ is extrapolated to pure water. Therefore, we chose 52% MeOH and 32.5% THF, labeled b and c in Figure 7B,C as the optimum φ. The simulated chromatograms are shown in Figure 7b,c. The highest achievable resolutions are 0.71 and 0.59, respectively, for MeOH and THF. This indicates that use MeOH or THF as the eluent modifier on the ODS column cannot provide baseline separation for all peaks. This example shows that T3C can be in some instance more 1828 Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
Table 3. Experimental Retention Factors of PTH-Amino Acids on ODS and C-ZrO2 k′ (ODS)a
k′ (C-ZrO2)a
solutes
40 °C
60 °C
40 °C
80 °C
(1) PTH-histidine (PTH-His) (2) PTH-asparagine (PTH-Asn) (3) PTH-glutamine (PTH-Gln) (4) PTH-glysine (PTH-Gly) (5) PTH-R-alanine (PTH-Ala) (6) PTH-R-aminobutyric acid (PTH-Aba) (7) PTH-Valine (PTH-Val) (8) PTH-methionine (PTH-Met) (9) PTH-tyrosine (PTH-Tyr) (10) PTH-leucine (PTH-Leu) (11) PTH-isoleucine (PTH-Ile) (12) PTH-Norleucine (PTH-Nle) (13) PTH-phenylalanine (PTH-Phe)
0.59 0.98 1.24 2.21 3.70 5.84 10.20 9.87 4.70 21.32 19.00 22.75 18.41
0.48 0.80 1.01 1.65 2.77 4.50 7.50 7.05 3.26 15.01 13.57 16.32 12.76
0.14 0.93 1.47 1.56 2.01 3.65 6.35 10.90 17.17 11.74 11.30 14.81 27.96
0.003 0.68 0.91 0.89 1.14 2.04 3.48 5.35 8.82 5.76 5.77 6.97 12.95
a
Measured in 30/70 acetonitrile/0.1% trifluoroacetic acid.
powerful than mobile-phase optimization on an ODS column. With T3C, the separation was easily optimized with only four trial runs. Optimization of the Separation of PTH-Amino Acids. Edman degradation is a common method to determine the amino acid sequence of proteins and peptides. It sequentially releases one amino acid at a time and converts them to cyclic compounds called phenylthiohydantoin amino acids.21 In this work, we tried to separate a set of PTH-amino acids using the T3C approach under isocratic conditions. The structures of the 13 compounds are shown in Figure 8 The separations of the PTH-amino acids on individual columns are shown in Figure 9 The chromatogram in Figure 9A shows the separation on the ODS column in 30/70 acetonitrile/0.1% trifluoroacetic acid (TFA) mixture at 40 °C. We observed the coelution of solute pair PTH-Met/PTH-Val (7/8) and solute pair PTH-Phe/PTH-Ile (13/11). Again, good peak shapes and comparable retention times were observed on C-ZrO2 as shown in Figure 9B. However, compounds PTH-Gln3 and PTH-Gly4 coelute and solutes PTH-Met,8 PTH-Leu,10 and PTH-Ile11 are not well separated. (21) Qureshi, G. A. HPLC of Proteins, Peptides and Polynucleotides: Contemporary Topics and Applications; VCH: New York, 1991.
Figure 11. Window diagrams showing the resolution of the critical pair versus the temperature of ODS and C-ZrO2 column: (A) threedimensional plot; (B) resolution contour plot.
The separation comparisons are shown in Figure 10. A fairly good linear regression with small deviation (r ) 0.903, sd ) 0.290) was observed in the κ-κ plot (Figure 10A). We further compared the elution order on the two phases and observed that nine pairs of solutes switch elution order. In addition, it shows that the critical pairs on the ODS (8/7, 13/11) are well separated on the C-ZrO2 and vice versa. Figure 10B indicates that a good separation on the T3C set should be achievable. Again, we optimized the separation by the T3C approach. We performed two additional runs at higher temperatures (60 °C on ODS and 80 °C on C-ZrO2). Table 3 summarizes the retention factors used for T3C optimization. Unlike the highly polar compounds, which have significantly larger ∆H° values on C-ZrO2 than on ODS phases, for PTH-amino acids, the ∆H° s are very similar on these two phases ranging from 1 to 2 kcal/mol. However, there is still enough difference in selectivity to tune the separation in T3C. Using the strategy that we described earlier, a window diagram was constructed in Figure 11. Figure 11A is the window diagram, and Figure 11B is the contour plot. The highest point in the window diagram corresponds to 30 °C on the ODS and 40 °C on the C-ZrO2 and the predicted optimum resolution is 2.0. The chromatogram at the optimized condition is shown in Figure 12. All compounds are well separated, and the experimental resolution is 1.5 between compounds 10 and 11, which means that we overestimated the plate count for the T3C set. This example shows again that, for less-polar compounds, the T3C combination of the ODS and C-ZrO2 phases is an extremely powerful separation tool. CONCLUSIONS The main conclusions of this work are as follows: 1. The T3C combination of aliphatic hydrocarbon and solid
Figure 12. Chromatograms showing the separation of PTH-amino acids mixture on T3C with ODS column at 30 °C and C-ZrO2 column at 40 °C and 1 mL/min. Experimental conditions: 30/70 acetonitrile/ water mixture.
carbon phases is an extremely powerful tool for the separation of various types of mixtures, including both highly polar compounds and less polar compounds. 2. For less-polar compounds, the carbon phase gives good peak shapes and retention times comparable to these on bonded phases. A carbon phase gives significantly different selectivity from an aliphatic hydrocarbon phase. 3. When coeluted or overlapped peaks on one phase can be resolved on the other phase, T3C will very likely provide better separation than either column alone. 4. The separation of the barbiturates and the PTH-amino acids demonstrates that T3C can dramatically improve selectivity and give baseline separations even though the mixtures could not be separated on either individual phase. 5. We also showed that, for barbiturate mixtures, T3C is more powerful than manipulating the eluent type and composition. In Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
1829
addition, the analysis time can be dramatically shortened by using high flow rates because of the high selectivity provided by T3C. 6. Only four initial runs are needed for T3C method development. We believe that T3C is a general approach for method development in HPLC. The combination of an ODS and a carbon phase is widely applicable to a wide variety of polar and nonpolar analytes.
1830
Analytical Chemistry, Vol. 73, No. 8, April 15, 2001
ACKNOWLEDGMENT The authors acknowledge financial support from the National Institutes of Health and LC Resources for the use of the Drylab software. Y.M. acknowledges the Graduate School Dissertation Fellowship from University of Minnesota. Received for review September 27, 2000. Accepted January 27, 2001. AC001155S
