An Automated Orthogonal Two-Dimensional Liquid Chromatograph

Recent progress in online, comprehensive two-dimensional high-performance liquid chromatography for non-proteomic applications. Dwight R. Stoll. Analy...
1 downloads 0 Views 636KB Size
Download PDF
Anal. Chem. 2003, 75, 3484-3494

An Automated Orthogonal Two-Dimensional Liquid Chromatograph Cadapakam J. Venkatramani* and Yury Zelechonok

Pharmacia Corporation, Global Chemical Process R & D, 4901 Searle Parkway, Skokie, Illinois 60077

A simple approach to two-dimensional liquid chromatography has been developed by coupling columns of different selectivity using a 12-port, dual-position valve and a standard HPLC system. The valve at the junction of the two columns enables continuous, periodic sampling (injection) of the primary column eluent onto the secondary column. The separation in the primary dimension is comparable to conventional HPLC, whereas the secondary column separation is fast, lasting several seconds. The high-speed separation in the secondary dimension enables the primary column eluent to be sampled with fidelity onto the secondary column throughout the chromatographic run. One might expect a coupled column liquid chromatography system operating in reverse-phase mode to be strongly correlated and, hence, inefficient. However, by applying a solvent gradient in the primary dimension and by progressively incrementing the solvent strength in the secondary dimension (tuning), the inefficiency or cross correlation between the two dimensions is minimized. In a tuned two-dimensional system, the influence of primary column retention (usually hydrophobicity) is minimal on secondary column retention. This enables subtle differences in component interaction with the two stationary phases to dominate the secondary column retention. The peaks are randomly dispersed over a retention plane rather than along a diagonal, resulting in an orthogonal separation. The peak capacity is multiplicative, and each component has a unique pair of retention times, enabling positive identification. In addition, the location of the component provides two independent measures of molecular properties. The 2D-LC system was evaluated by analyzing a test mixture made of some aromatic amines and non-amines on different secondary columns (ODS-AQ/ODS monolith, ODS/ amino, ODS/cyano). The relative location of sample components in the two-dimensional plane varied significantly with change in secondary column. Among the secondary columns, the amino and cyano columns offered the most complementary separation, with the retention order of several components reversed in the secondary dimension. The theoretical peak capacity of the 2D-LC system was around 450 for a separation lasting 30 min. A 2D-LC system involving amino and cyano columns * To whom correspondence should be addressed. Phone: 847-982-4746. E-mail: [email protected].

3484 Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

resulted in a high-speed separation of the test mixture, with most of the chemical components resolved within a few minutes. A multidimensional technique involving coupled columns is quite powerful, as compared to its one-dimensional counterpart, provided the retention mechanisms in the two dimensions are orthogonal.1-3 Through orthogonality, the cross-information or synentropy existing between dimensions is minimized, resulting in multiplicative peak capacity and, hence, enhanced resolution.4,5 Minimizing synentropy maximizes the efficiency and the information content, a key to complex sample analysis. The importance of multidimensional separations has been known for a long time and had its origin in flat bed chromatography in biological sciences.6-8 A major breakthrough in coupled column chromatography was the invention of Dean’s valveless stream switching device in 1968 that enabled selective transfer of unresolved or partially resolved components from the primary column to a secondary column for further separation.9 Since then, such “heart cutting” methods have been predominantly used in coupled column chromatography. Heart cutting techniques require prior knowledge of the sample component retention, a high-speed switching valve, and an effective focusing mechanism between columns to generate useful separation in the secondary dimension.10 In the past one and a half decades, two different approaches have emerged in comprehensive two-dimensional coupled column chromatography, resulting in orthogonal separations.11-18 In gas (1) Giddings, J. C. Anal.Chem. 1984, 56, 1258A. (2) Giddings, J. C. In Multidimensional Chromatography: Techniques and Applications; Cortes, H. J., Ed.; Marcel Dekker: New York, 1990; pp 1-27. (3) Giddings, J. C. HRC & CC 1987, 10, 319. (4) Phillips, J. B.; Liu, Z.; Venkatramani, C. J.; Jain, V. In Proceedings of the 13th International Symposium on Capillary Chromatography; Riva del Garda, 1991; Dr. Alfred Huethig, Verlag: Heidelberg; p 260. (5) Venkatramani, C. J.; Jingzhen, X.; Phillips, J. B. Anal. Chem. 1996, 68, 1486. (6) Consen, R.; Gordon, A. H.; Martin, J. P. Biochem. J. 1944, 38, 244. (7) Haugaard, G.; Kroner, T. D. J. Am. Chem. Soc. 1948, 70, 2135. (8) Durrum, E. Biochem. J. 1950, 48, 274. (9) Deans, D. R. Chromatographia 1968, 1, 18. (10) Cortes, H. J. In Multidimensional Chromatography Techniques and Applications; Cortes, H. J., Ed.; Chromatographic Science Series; Dekker: New York, 1990; Vol. 50. (11) Liu, Z.; Phillips, J. B. J. Chromatogr. Sci. 1991, 29, 227. (12) Venkatramani, C. J.; Phillips, J. B. J. Microcolumn Sep. 1993, 5, 511. (13) Venkatramani, C. J.; Phillips, J. B. In Proceedings of the 15th International Symposium on Capillary Chromatography, Riva del Garda, 1993; Dr. Alfred Huethig, Verlag: Heidelberg; p 885. (14) Venkatramani, C. J. Ph,D. Dissertation, Southern Illinois University, Carbondale, Illinois, 1994. 10.1021/ac030075w CCC: $25.00

© 2003 American Chemical Society Published on Web 05/31/2003

chromatography, Phillips et al. developed a comprehensive twodimensional gas chromatograph (2D-GC) by selectively tuning the operation parameter (temperature), generating an orthogonal separation.4,5,11-16 In liquid chromatography, Jorgenson et al. developed a comprehensive two-dimensional system by coupling orthogonal techniques to minimize synentropy.17-20,22-25 The first automated, comprehensive 2D-LC system was developed to analyze protein samples by coupling gradient elution ion exchange chromatography (IE) to size exclusion chromatography (SEC).17 Since then, different types of 2D-LC systems have been developed, such as LC/CZE, SEC/CZE, gel electrophoresis/LC, and IE/ LC.18-22,25 However, coupling dissimilar techniques is challenging and complicated, as sampling criteria and the mode of operation differ significantly with increase in dissimilarities between retention mechanisms. An exception to this rule is the work by Murphy et al., who resolved alcohol ethoxylates using an aminopropyl silica column in the primary dimension and a reverse-phase C18 bonded to silica in the secondary dimension.26 Although the twodimensional separation provided information on ethylene oxide (EO) units in the primary dimension (polarity) and alkyl chain length in the secondary dimension (hydrophobicity), some correlation was observed between the two dimensions. Additionally, addressing the sampling needs of the secondary column resulted in long analysis times. We have designed a simple, automated 2D-LC system by slightly modifying the plumbing of a commercial HPLC system and incorporating an electronically controlled, 12-port, dualposition valve between the dimensions.27,28 The primary column eluent is alternatively sampled into a dual (or single) column through a dual, equivalent sampling loop. Since columns differ in their selectivity, chemical components coeluting in the primary column resolve in the secondary column. The primary column separation is comparable to conventional HPLC (minutes), whereas the secondary column separation is very fast (seconds). The highspeed separation in the secondary column enables partial or complete transfer of the primary column eluent to the secondary column. The resulting chromatograms when plotted in an appropriate form result in a comprehensive two-dimensional liquid chromatogram. The frequency of generating the high-speed secondary chromatogram depends on the retention time range of uncharged chemical components in the secondary column. Although column selectivity is critical to 2D-LC, coupling dissimilar stationary phases is not adequate to ensure an orthogonal separation. Chemical components strongly retained on a stationary phase are likely to be strongly retained on different stationary phases, because sample hydrophobicity is the key factor (15) Phillips, J. B.; Xu, I. L. J. Chromatogr. 1995, 703, 327. (16) Liu, Z.; Phillips, J. B. J. Microcolumn Sep. 1989, 1, 249. (17) Bushey, M. M.; Jorgenson, J. W. Anal. Chem. 1990, 62, 161. (18) Bushey, M. M.; Jorgenson, J. W. Anal. Chem. 1990, 62, 978. (19) Bushey, M. M.; Jorgenson, J. W. J. Microcolumn Sep. 1990, 2, 293. (20) Lemmo, A. V.; Jorgenson, J. W. J. Chromatogr. 1993, 633, 213. (21) Rose, D. J.; Opiteck, G. J. Anal. Chem. 1994, 66, 2529. (22) Holland, L. A.; Jorgenson, J. W. Anal. Chem. 1995, 67, 3275. (23) Moore, A. W., Jr.; Jorgenson, J. W. Anal. Chem. 1995, 67, 3448. (24) Hooker, T. F.; Jorgenson, J. W. Anal. Chem. 1997, 69, 4134. (25) Opiteck, G. J.; Jorgenson, J. W. Anal. Chem. 1997, 69, 2283. (26) Murphy, R. E.; Schure, M. R.; Foley, J. P. Anal. Chem. 1998, 70, 4353. (27) Venkatramani, C. J.; Patel, A.; Yury, Z. Comprehensive Two-Dimensional Liquid Chromatography in Pharmaceutical Analysis. Poster presented at the 26th International Symposium on HPLC, Montreal, Canada, 2002. (28) Venkatramani, C. J.; Patel Anurag. Manuscript in preparation.

influencing reverse-phase separation. By tuning the operating parameters, such as mobile-phase strength, temperature, and buffer strength, in conjunction with column selectivity, one can generate an orthogonal separation. Among the experimental parameters, mobile-phase strength is the most significant parameter influencing the separation and is the easiest one to tune. This can be readily achieved by a solvent gradient in the primary dimension and progressive solvent strength increment in the secondary dimension. Increasing the solvent strength in the secondary dimension during a separation progressively reduces the retention power of the secondary column. This, in turn, compensates (or nullifies) the increasing sample component hydrophobicity, enabling subtle differences in component interaction with the two phases to dominate the secondary column retention. In addition, this limits the retention time range of sample components in the secondary dimension. Thus, tuning operation parameters and column selectivity is critical to orthogonal 2DLC. Theory on tuning and its influence in two-dimensional column chromatography can be found in the literature and is not within the scope of this paper.4,5 Unlike other systems, the characteristic feature of this 2D-LC system is its simplicity and flexibility.17-22 With minimal change in plumbing and by incorporating an electronic valve and a timer, any commercial liquid chromatograph can be readily converted into a 2D-LC system. In addition, the design offers the flexibility of using similar or complementary columns in the secondary dimension and use of different mobile phases or gradients in the two-dimensions. However, these modifications will require an additional pump and a detector. The salient features of the above 2D-LC system are discussed below. Operation of both dimensions under comparable reverse-phase conditions enables the use of a conventional HPLC system instead of micro-LC systems cited in the literature.17-22 Micro-LC was mandatory to address the sampling needs resulting from coupling of dissimilar techniques. Since the mobile phase at the head of the secondary column is always stronger than the eluent sampled into the secondary column, effective focusing of sample components takes place throughout the chromatogram. This eliminates the need for additional dilution of primary column eluent prior to secondary column chromatography for effective focusing.22 In addition, earlier designs involving the reverse-phase in the secondary dimension used repetitive gradients to ensure elution of all sample components between successive sampling cycles.22,25 The brief equilibration time between sampling cycles significantly reduces the peak capacity of the system (about one-third). Our approach eliminates the need for repetitive gradients, because the secondary column is at an appropriate solvent strength to receive the primary column eluent throughout the separation. Thus, the entire two-dimensional plane is accessible, and the peak capacity is multiplicative. Another advantage of this system is the highspeed separation lasting 20-30 s, enabling primary column separation to be transferred with fidelity onto the secondary column. Such high-speed separations have been seldom attained in earlier reversed-phase 2D-LC separations. The 2D-LC system was evaluated using a test mixture made of some aromatic amines and non-amines using different columns in the secondary dimension. The relative location of sample components in the two-dimensional plane was evaluated as a Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3485

3486

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

Figure 1. Schematic of 2D-LC system showing the mobile phase flow through the primary (left) and secondary columns (right). The primary column eluent is alternatively sampled into secondary columns S1 and S2 through sampling loops L1 and L2. The flow through the columns is uninterrupted throughout the chromatographic run.

function of column selectivity. The separation is truly orthogonal with sample components randomly dispersed in the two-dimensional plane. The high-speed 2D-LC separation demonstrates the potential of 2D-LC in resolving coeluting peaks and in highthroughput analysis. EXPERIMENTAL SECTION Chemicals and Reagents. Most of the chemicals used in this study [benzylamine, dimethylbenzylamine, methylphenethylamine, methoxyphenamine, propranolol, tryptophan, 3-methyl-2butanone, benzaldehyde, benzoic acid, benzonitrile, and trifluoroacetic acid (TFA)] were purchased from Sigma-Aldrich. Acetonitrile, water, and methanol were purchased from Burdick and Jackson. Stock solutions of benzylamine, dimethylbenzylamine, methylphenethylamine, benzonitrile, benzaldehyde, and methylbutanone were prepared in acetonitrile. Stock solutions of benzoic acid and tryptophan were made in a 1:1 mixture of acetonitrile/water mixture. Stock solutions of salts of propranolol and methoxyphenamine were made in water. A 10-mL test mixture of these components was made by mixing different proportions of individual components in a 4:1 water/acetonitrile mixture. The final concentration of the individual components in the mixture was benzylamine, 240 µg/mL; dimethylbenzylamine, 600 µg/mL; methoxyphenamine, 600 µg/mL; propranolol, 50 µg/mL; tryptophan, 23 µg/mL; 3-methyl-2-butanone; 600 µg/mL; benzaldehyde, 120 µg/mL; benzoic acid, 23 µg/mL; and benzonitrile, 100 µg/mL. Ten microliters of test mixture was injected into the HPLC. All of the stock solutions and the mixtures were refrigerated and found to be stable throughout the study. Chromatographic Equipment. The instrumentation for orthogonal two-dimensional liquid chromatography is shown in Figure 1. For clarity, the events taking place in the two dimensions are shown separately using dark and light shading. The key component of this instrument is an electronically controlled dualposition valve (V). The 12-port, dual position valve enables continuous, alternative sampling of the primary column (C) eluent onto dual secondary columns (S1 and S2) through equivalent, dual sampling loops (L1 and L2). The mobile phase from the pump is spilt at the Valco union (U1), with part of the flow going through the primary column (C) and rest flowing through the secondary columns (S1 and S2). Since the flow rate through the column is a function of its dimension and the tubing used, manipulating these parameters changes the column flow. The flow to the secondary column is further split at a Valco union (U2) with one-half of the flow going through the secondary column (S1) and through the sampling loop (L1 or L3), and the other half flowing through the secondary column (S2) and through the sampling loop (L2 or L3). Comparable volume of the sampling loops L1, L2, and L3 (40 µL each) results in constant flow through the secondary columns. The eluent from the secondary columns merge at the Valco union (U3) before detection. In position 1 (Figure 1, top left-hand corner), the eluent from the primary column flows through the 40-µL sample loop L1 before exiting through the detector (D2, not mandatory) or to waste, whereas the contents of loop L2 from the previous cycle are sampled onto the secondary column S2 (Figure 1, top right-hand corner). The flow through loop L3 (valve port 3 to port 9) maintains undisrupted flow though the secondary column S1. When the valve position is switched, the eluent from the primary

column fills the other 40-µL sample loop L2 (Figure 1, bottom left-hand corner), whereas the contents of loop L1 from the previous cycle are sampled onto the secondary column S1 (Figure 1, bottom right-hand corner). The flow through loop L3 (valve port 9 to port 3) keeps undisrupted flow through the secondary column S2. Thus, there is a continuous flow of mobile phase through the primary and secondary column throughout the separation. Depending upon the retention time range of the chemical components in the second dimension, either partial or whole sample from the primary column is transferred onto the secondary column. In this work, ∼25-60 % of the primary column eluent was sampled onto the secondary column. Since peaks from the primary column last several seconds, at least one scan is obtained across each peak eluting from the primary column, resulting in a comprehensive separation. All instruments used in this study are commercially available. A HP1100 series from Agilent Technologies equipped with a degasser (G1322A0), quaternary pump (G1311A), auto-sampler (G1313A), and diode array detector (G1315A) were used to generate the two-dimensional chromatogram. All the instruments were controlled by Agilent Chemstation software. An electronically activated 12-port, dual-position valve purchased from Valco Instruments Co. Inc. (Houston, Texas) was controlled by a multimode, programmable timer (ICM 500 purchased from ICM Corp., Cicero, NY). The external contact closure from the pump was used to synchronize the ICM timer between successive runs. This enables effective subtraction of system peaks resulting from differences between the mobile-phase strength at the head of the secondary column and the eluent sampled into the secondary column. Details pertaining to the system peak and its elimination are discussed later. The data from Chemstation (primary column retention time and detector response) were exported into Microsoft Excel and were updated with second dimension retention times prior to reprocessing using PSI-Plot. The PSI-Plot purchased from Poly Software International (Pearl River, NY) enabled display of the two-dimensional data either as a contour plot or a threedimensional plot. Chromatographic Conditions. The chromatographic conditions for different studies are summarized below. Experimental Conditions for ODS-AQ/ODS Monolith Column. For this study, a 15 cm × 4.6 mm × 3.0 µm ODS-AQ column from Waters Corporation was used in the primary dimension along with a 10 cm × 4.6 mm Chromolith Performance RP-18e column from Merck Company in the secondary dimension. The flow rates through the primary and secondary column were 0.5 mL/min and 4.0 mL/min, respectively. A single pump was used to deliver the mobile phase to the two columns. Mobile phase A was 0.1% trifluoroacetic acid (TFA) in water, and mobile phase B was 0.06% TFA in acetonitrile. The solvent gradients in both dimensions were the same, starting with 5% mobile phase B and programmed to 50% mobile phase B over 32 min, then to 80% mobile phase by 40 min, followed be reequilibration. A secondary chromatogram was generated every 20 s by sampling the primary column eluent to the secondary column. Approximately 25% of the primary column eluent was sampled onto the secondary column (sample loop volume × number of secondary chromatograms per minute × 100/primary column flow rate; 40 µL × 3 × 100/500 µL). The UV detection was at 272 nm. Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3487

Figure 2. The process of converting high-speed secondary chromatograms into two-dimensional contour plot is shown above. The chromatogram on the left is a partial plot of a HPLC separation lasting 2 min. The stacked plots in the middle are five high-speed secondary chromatograms resulting from fractions transferred to the secondary column (indicated by error bar). The resulting contour plot is shown to the right.

Experimental Conditions for ODS-AQ/Amino Column. For this study, a 15 cm × 4.6 mm × 3.0 µm ODS-AQ column from Waters Corporation was used in the primary dimension along with dual 5 cm × 4.6 mm × 3.5 µm Exsil amino columns from Keystone Scientific in parallel in the secondary dimension. The flow rates through the primary and secondary column were 0.5 mL/min and 1.5 mL/min/column, respectively. A single pump was used to deliver the mobile phase to the two columns. The mobile phase and gradients used in this study were the same as in the earlier experiment involving the ODS columns. A secondary chromatogram was generated every 20 s by sampling the primary column eluent alternatively to the dual secondary columns. Approximately 25% of the primary column eluent was sampled onto the secondary column for further separation (sample loop volume × number of secondary chromatograms per minute × 100/primary column flow rate; 40 µL × 3 × 100/500 µL). The UV detection was at 272 nm. Experimental Conditions for ODS-AQ/Amino/Cyano Columns. For this study, a 15 cm × 4.6 mm × 3.0 µm ODS-AQ column from Waters Corporation was used in the primary dimension. A 5 cm × 4.6 mm × 3.5 µm Exsil amino column from Keystone Scientific and a 3.3 cm × 7.0 mm × 3.0 µm Platinum cyano column from Alltech Associates were used in parallel in the secondary dimension. The flow rate through the primary column was 0.5 mL/min, and the flow rates through the secondary amino and cyano columns were 1.5 mL/min and 2.2 mL/min, respectively. A dual pump was used to deliver the mobile phase to the primary and secondary columns. The mobile phase and gradient used in this study were the same as in the earlier experiment involving 3488

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

the ODS-AQ/ODS monolith columns. A secondary chromatogram was generated every 20 s by sampling the primary column eluent alternatively between the amino and cyano columns every 10 s. Approximately 25% of the primary column eluent was sampled onto each of the secondary columns (sample loop volume × number of secondary chromatograms per minute in a column × 100/primary column flow rate; 40 µL × 3 × 100/500 µL). Two UV detectors set at 272 nm were used in this study. Signals from the two detectors were collected on different data acquisition channels. Experimental Conditions for 1D-HPLC on Amino Column. For this study, a 5 cm × 4.6 mm × 3.5 µm Exsil amino column from Keystone Scientific was used at 0.2 mL/min. Increasing the flow rate to 0.5 mL/min resulted in reduced analysis time without improving the separation. The mobile phase used in this study was the same as an earlier experiment (ODS-AQ/ODS monolith columns). The slope of the gradient was the same as the earlier study, but was terminated at 6 min to reequilibrate the columns. The solvent gradient starting with 5% mobile phase B was programmed to 13.4% mobile phase B over 6 min and then back to initial conditions. The UV detection was at 272 nm. Experimental Conditions for 1D-HPLC on Cyano Column. For this study, a 3.3 cm × 7.0 mm × 3.0 µm Platinum cyano column from Alltech Associates was used at 0.5 mL/min. The mobile phase used in this study was the same as the earlier experiment (ODS-AQ/ODS monolith columns). The slope of the gradient was the same as the earlier study but was terminated at 9 min to reequilibrate the columns. The solvent gradient starting with 5%

Figure 3. An orthogonal two-dimensional separation of test mixture on an ODS-AQ column in the primary dimension (15 cm × 4.6 mm × 3.0 µM) and an ODS monolith column in the secondary dimension (10 cm × 4.6 mm). The primary column flow rate was 0.5 mL/min, and the secondary column flow rate was 4.0 mL/min. The primary column eluent was sampled every 20 s into the secondary column. Approximately 25% of primary column eluent was sampled into the secondary column.

mobile phase B and was programmed to 17.6% mobile phase B over 9 min, then back to initial conditions. The flow rate was set to 2.0 mL/min for a short time period (6.0-7.0 min) to speed the elution of a late-eluting component. The UV detection was at 272 nm. Experimental Conditions for High-Speed 2D-LC on Cyano/Amino Column. For this study, a 3.3 cm × 7.0 mm × 3.0 µm Platinum cyano column from Alltech Associates was used in the primary dimension, and a 5 cm × 4.6 mm × 3.5 µm Exsil amino column from Keystone Scientific was used in the secondary dimension. The flow rate through the primary column was 0.5 mL/min until 6 min, then programmed to 2.0 mL/min until 7 min, then back to 0.5 mL/min. The flow rate in the primary column was increased momentarily to reduce the retention time of the last eluting component by few minutes. However, this has little bearing on the secondary column retention. The flow rate through the secondary column was 2.5 mL/min. A dual pump was used to deliver the mobile phase to the primary and secondary columns. The mobile phase and gradient used in this study were the same as the one used in the cyano column HPLC analysis discussed above. A secondary chromatogram was generated every 20 s. Approximately 25% of the primary column eluent was sampled to the secondary column for further separation (sample loop volume × number of secondary chromatograms per minute × 100/ primary column flow rate; 40 µL × 3 × 100/500 µL). The UV detection was at 272 nm. Experimental Conditions for High-Speed 2D-LC on Amino/Cyano Column. For this study, a 5 cm × 4.6 mm × 3.5 Exsil amino column from Keystone Scientific was used in the primary dimension, and a 3.3 cm × 7.0 mm × 3.0 µm Platinum cyano column

from Alltech Associates was used in the secondary dimension. The flow rate through the primary amino and secondary cyano columns was 0.2 mL/min and 4.0 mL/min, respectively. A dual pump was used to deliver the mobile phase to the primary and secondary columns. The mobile phase and gradient used in this study were the same as the one used for the amino column HPLC analysis discussed above. A secondary chromatogram was generated every 20 s. Approximately 60% of the primary column eluent was sampled to the secondary column (sample loop volume × number of secondary chromatograms per minute × 100/primary column flow rate; 40 µL × 3 × 100/200 µL). The UV detection was at 272 nm. RESULTS AND DISCUSSION Generating Two-Dimensional Contour Plots From HighSpeed Secondary Chromatograms. In comprehensive 2D-LC, the data acquisition system records the detector response as a function of time. Interpretation of these data requires conversion of these high-speed secondary chromatograms into a twodimensional contour plot or a three-dimensional plot using an appropriate software package. The process of converting the highspeed secondary chromatograms to a two-dimensional contour plot is shown in Figure 2. The chromatogram on the left side corner is a partial plot of a hypothetical chromatogram lasting a couple of minutes in the primary column. The error bars on top of this chromatogram indicate sections of the primary column separation sampled into the secondary column. The stacked plot in the middle shows the five high-speed secondary chromatograms resulting from these fractions. The series of high-speed secondary chromatograms can be converted into a two-dimensional contour Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3489

Figure 4. An orthogonal two-dimensional separation of test mixture on an ODS-AQ column in the primary dimension (15 cm × 4.6 mm × 3.0 µM) and an amino column in the secondary dimension (5 cm × 4.6 mm × 3.5 µM, dual). The primary column flow rate was 0.5 mL/min, and the secondary column flow rate was 1.5 mL/min/column. The primary column eluent was alternatively sampled into dual secondary columns every 20 s. Approximately 25% of primary column eluent was sampled into the secondary column.

plot by exporting the data into a spreadsheet program and updating with secondary column retention times, followed by processing using graphics software to generate the two-dimensional chromatogram. The two-dimensional contour plot shown on the right side of Figure 2 was generated by stacking these five high-speed chromatograms and by linear interpolation of the data between secondary chromatograms. The peak eluting around 27 min from the primary column is actually a two-component mixture that is resolved in the secondary dimension. Thus, the two-dimensional contour plot has three spots corresponding to the three components present in the sample. The primary and the secondary column retention times form the two axes of this chromatogram, with the relative intensity being the third axis. The number of contours for each component is a measure of its intensity. Computing exact secondary column retention times and retention ranges of chemical components in the secondary dimension is discussed below. Retention Time Ranges of Chemical Components in the Second Dimension. Exact retention times of chemical components in the second dimension can be determined by selectively transferring a few components from the primary column to the secondary column (sampling once a minute). Retention time in the secondary dimension is the difference between peak elution time and corresponding valve switching time. Alternatively, one could assign an arbitrary retention time of zero for the least retained component in the secondary dimension, because knowledge of it is not critical. It is the relative location of a component in the two-dimensional plane that enables its identification, and not the absolute retention time. 3490

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

The frequency of generating high-speed secondary chromatograms is limited to retention time ranges of sample components in the second dimension. Generating secondary chromatograms faster than the retention time ranges of the components will result in a wrap-around or aliasing problem; i.e., components with strong secondary column retention will appear at incorrect locations (usually at low retention time) in the two-dimensional plane, resulting in misinterpretation of data. However, this can be readily addressed by increasing the second dimension duration. Orthogonal 2D-LC Separation Using Similar Phase in the Secondary Dimension. An orthogonal 2D-LC separation of a test mixture of some aromatic amines and non-amines on an ODSAQ column (15 cm × 4.6 mm × 3.0 µm) in the primary dimension and a C18 monolith column (10 cm × 4.6 mm) in the secondary dimension is shown in Figure 3. One would expect a 2D-LC system with ODS columns in both dimensions to be strongly correlated, with peaks along the diagonal. However, tuning the operation parameters enables polarity differences between columns to dominate the secondary column retention. Thus, the peaks are randomly distributed in the two-dimensional plane, and the separation is orthogonal. Propranolol and benzonitrile have similar retention characteristics in the primary column, but differ significantly in their secondary column retention behavior as a result of their polarity differences. Under the experimental conditions, the polar propranolol has lower retention than benzonitrile in the nonpolar ODS monolith column. Similarly, benzoic acid, with an acidic functionality, exhibits retention characteristics intermediate to neutral benzaldehyde and polar propranolol. Tryptophan, an amino acid with

Figure 5. An orthogonal two-dimensional separation of test mixture on an ODS-AQ column in the primary dimension (15 cm × 4.6 mm × 3.0 µM) and dual cyano (3.3 cm × 7.0 mm × 3.0 µM) and amino (5.0 cm × 4.6 mm × 3.5 µM) columns in parallel in the secondary dimension. The primary column flow rate was 0.5 mL/min, and the secondary column flow rates were 2.2 mL/min on cyano and 1.5 mL/min on the amino column. The primary column eluent was sampled alternatively into secondary columns every 10 s. Approximately 25% of primary column eluent was sampled into each of the secondary columns.

both acidic and basic functionality, has very low retention in the secondary dimension. Absence of short-chain hydrophobic and or acidic components in the sample results in some empty space in the left upper corner of the chromatogram. In this separation, the primary column offers mixed-mode retention (hydrophobicity and polarity), and the secondary column separation is based on sample hydrophobicity. Approximately 25% of the primary column eluent was sampled into the secondary column. Figure 4 shows the 2D-LC separation of sample mixture on the ODS-AQ column (15 cm × 4.6 mm × 3.0 µm) in the primary and dual amino columns (5 cm × 4.6 mm × 3.5 µm in parallel) in the secondary dimension. Under acidic conditions, the amino columns are positively charged, as are the aromatic amines, resulting in their strong repulsion with the secondary column stationary phase. This is reflected by low retention of amines in the secondary column, irrespective of their primary column retention. Thus, all the amines have comparable retention on the secondary column and lie in a straight line. However, benzoic acid interacts strongly in the amino column and has high secondary column retention. Tryptophan with acidic and basic functionality behaves more like an aromatic amine on this column. However, the acidic group in tryptophan results in marginally higher retention, as compared to the amines. The retention of neutral compounds in both the dimensions is

based on their hydrophobicity. In this separation, both primary and secondary columns offer mixed mode retention, but differ from one another, resulting in an orthogonal separation. In the ODS-AQ column, a combination of hydrophobicity and polarity dominate retention, whereas in the amino column, it is the combination of hydrophobicity, polarity, and ionic interactions contributing to retention. The relative position of some of the peaks in this chromatogram is different from the earlier chromatogram in Figure 3. Approximately 25% of primary column eluent is sampled into the secondary column. Orthogonal 2D-LC Separation Using Complementary Phase in the Secondary Dimension. Figure 5 shows a 2D-LC separation resulting from dual, complementary columns (in parallel) in the secondary dimension. The primary column eluent from the ODS-AQ column (15 cm × 4.6 mm × 3.0 µm) was alternatively sampled into an amino (5 cm × 4.6 mm × 3.5 µm) and a cyano column (3.3 cm × 7.0 mm × 3.0 µm) in the secondary dimension. Two UV detectors were used to monitor the secondary chromatograms at 272 nm. The cyano (also called nitrile)-bonded phase interacts strongly with the polar functional groups in the test mixture compounds and is used in resolving components on the basis of functional group differences. The cyano column is more selective for double Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3491

Figure 6. Comparison of high-speed 2D-LC separations (bottom plots) and one-dimensional separations (top plots) on cyano and amino columns: A, benzylamine; B, methylbutanone; C, dimethylbenzylamine; D, methylphenethylamine; E, tryptophan; F, benzoic acid; G, benzaldehyde; H, benzonitrile; I, methoxyphenamine; and J, propranolol.

bonds than any hydrocarbon phases.29 The 2D-LC separations in the ODS-AQ/cyano column (top) and ODS-AQ/amino column (bottom) are complementary. Methylbutanone, benzoic acid, benzaldehyde, and benzonitrile that are strongly retained on the amino column (bottom) are poorly retained on the cyano column (top). The opposite is true for the aromatic amines. The separations on the ODS-AQ/amino column in Figures 4 and 5 are comparable, suggesting the system is highly reproducible. The relative intensity of sample components in 2D-LC depends on the peak fraction that is transferred onto the secondary dimension. Thus, a transfer from the center of a peak will produce an intense spot in the 2D-LC plane, as compared to the fraction from the front or tail of a peak. This explains the differences in the relative intensity of a few sample components in the two complementary 2D-LC separations shown in Figure 5. Spot intensities of some components, such as benzylamine, dimethylbenzylamine, and methylphenethylamine, are comparable; they elute as broad peaks in the primary dimension. Systems involving complementary columns are quite powerful because they provide twice the information per unit time. Since the primary column eluent is sampled alternatively to different columns in the secondary dimension and the eluents from these columns do not mix, the primary column eluent is sampled every (29) HPLC column selection guide, Alltech Associates, no. 343.

3492

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

10 s instead of 20 s. Thus, in the above chromatogram, ∼50% of the primary column eluent is sampled into the secondary columns, with one-half going through each secondary column. Peak Capacity in Two-Dimensional Liquid Chromatography. A key feature of multidimensional separations is the multiplicative peak capacity resulting from orthogonal separation. In the two-dimensional separation shown in Figures 3-5, 90 highspeed secondary chromatograms of 20 s duration were generated in 30 min. This results in a primary column peak capacity of 45, assuming a primary column component is sampled twice into the secondary column (number of high-speed secondary chromatograms/number of times a component is sampled, 90/2). In the secondary dimension, the peak width at half-height is ∼2 s, resulting in secondary column peak capacity of 10 (duration of each secondary chromatogram/peak width at half-height, 20 s/2 s). Since peak capacity is multiplicative in multidimensional chromatography, the total peak capacity of the system is ∼450 (45 × 10) in a 30-min separation.1-3 Peak capacities of this magnitude are difficult to realize in one-dimensional HPLC; however, in reality, the net peak capacity will be lower, because parts of the chromatogram are inaccessible. High-Speed 2D-LC Using Complementary Columns. The top two chromatograms in Figure 6 are one-dimensional HPLC separations of the test mixture on cyano (3.3 cm × 7.0 mm × 3.0

Figure 7. Stacked plot of blank sample chromatogram with and without blank subtraction. The blank-subtracted sample chromatogram C was generated by subtracting sample chromatogram B from blank chromatogram A.

µm) and amino (5.0 cm × 4.6 mm × 3.0 µm) columns, respectively. The cyano column partially resolves some of the sample components (Figure 6, top left chromatogram) and the amino column resolves them into aromatic amine and non-amine groupings (Figure 6, top right chromatogram). On the basis of this onedimensional separation, it is very unlikely that a combination of these columns will be effective in 2D-LC. However, among the columns evaluated in 2D-LC, cyano and amino provided the most complementary separation and, hence, should be effective in 2DLC separations of the test mixture. The high-speed, two-dimensional contour plots shown in Figure 6 (bottom plots) clearly demonstrate the power of these columns in 2D-LC analysis of the sample mixture. Among the 10 sample components, 6 of them are baseline-resolved in the secondary dimension, with the other 4 coeluting as two spots. The retention order of sample components in these chromatograms is consistent with the earlier separation shown in Figures 4 and 5. In the 2D-LC separation involving the cyano/amino columns (bottom left, Figure 6), the amines exhibit low retention, whereas methylbutanone, benzoic acid, benzaldehyde, and benzonitrile exhibit strong retention in the amino column. In the 2D-LC separation involving the amino/cyano columns (bottom right,

Figure 6), the cyano column resolves most of the coeluting peaks from the primary column. These high-speed 2D-LC separations clearly demonstrate the potential of 2D-LC in resolving coeluting peaks and in high-throughput screening. System Peaks, Origin and Elimination. Partial plots of 2DLC chromatograms of a blank and a sample mixture (detector response as a function of primary column retention time) are shown in Figure 7A and B. These chromatograms show the presence of periodic spikes or disturbances in the baseline (system peaks). The number of system peaks in these chromatograms corresponds to the number of fractions transferred from the primary to the secondary column. These system peaks result from the differences in the strength of the mobile phase at the head of the secondary column and the eluent sampled into the secondary column. Although system peaks are inherent in 2D-LC and are reproducible, they can be effectively subtracted prior to twodimensional data processing, as shown in Figure 7C. CONCLUSIONS In this paper, we have successfully demonstrated the design of a simple, automated 2D-LC system. A commercial HPLC system can be readily modified into a 2D-LC system with minimal change Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

3493

in plumbing and by incorporating an electronic valve and a timer. These systems can operate either with single columns in the secondary dimension or with complementary columns. The 2DLC separation of the test mixture clearly demonstrates that the selectivity of sample components in the two-dimensional plane can be manipulated by changing columns and by tuning the operation parameters. Since the retention in the two dimensions are independent, the separation is truly orthogonal, and the peak capacity is multiplicative. Each chemical component has a unique pair of retention times, enabling positive identification. In addition, the location of the chemical component in the two-dimensional plane is a function of its physical properties. The high-speed twodimensional separation of sample components lasting a few minutes demonstrates the potential of this technique in resolving coeluting peaks and in high-throughput screening.

3494

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

ACKNOWLEDGMENT C. J. Venkatramani, expresses sincere gratitude to the late Professor Dr. John B. Phillips for introducing him to comprehensive two-dimensional gas chromatography, on which the current work is based. The authors thank Miss Cara Weyker, Dr. Rick Rhinebarger, and Pharmacia Corporation for all of their encouragement, support, and review of the manuscript. The authors also thank Mr. Anurag Patel, a colleague at Pharmacia Corporation, for his suggestions and help in drawing parts of the valve schematics.

Received for review March 2, 2003. Accepted April 23, 2003. AC030075W