Anal. Chem. 2003, 75, 1173-1179
Variable-Gradient Generator for Micro- and Nano-HPLC Achille Cappiello,* Giorgio Famiglini, Chiara Fiorucci, Filippo Mangani, Pierangela Palma, and Antonella Siviero
Istituto di Scienze Chimiche “F. Bruner”, Universita` di Urbino, Urbino, Italy
A new, simple device generates accurate nano- and microflow rate gradients from any conventional HPLC system. The core of the new device is represented by an electric-actuated, computer-controlled, multiposition HPLC valve. The valve hosts six reservoirs for as many different mobile-phase compositions of increasing strength. A low flow rate stream pushes the weakest solvent through the column as long as required and at the desired flow rate, until the chromatographic run is started. From this time on, the electric actuation allows one to select which reservoir will be on-line with the column and for how long, thus generating a specific solvent gradient, through a sequence of controlled segments of precise mobile-phase composition. This permits one not only to exactly reproduce the programmed slope but also to achieve different gradient shapes (i.e., linear, convex, concave) for different separation needs. The new device has proven to be reliable and reproducible even at the lowest flow rate tested (250 nL‚min-1) and in different chromatographic conditions. Since its first introduction in 1967 by Horva´th and co-workers,1,2 microcolumn liquid chromatography (LC) has raised a great deal of interest, as indicated by the large number of research and review articles that have appeared in the literature over the years. All aspects of micro-HPLC have been deeply investigated ever since: theory, different packing technologies, instrumentation, detection, injection volumes, etc.3,4 and references therein. Even though conventional LC is still widely used in routine analysis, microcolumn LC plays a complementary role, thanks to several key advantages: the ability to work at very low flow rates, which leads to a significantly lower solvent consumption; the possibility to inject very low sample size; the enhanced detection using concentration-sensitive detectors due to a much lower chromatographic dilution during the separation process (mass sensitivity).5,6 These attributes allowed the rapid development of micro* Corresponding author. Tel: +3907224164. Fax: +3907222754. E-mail:
[email protected]. (1) Horva´th, C. G.; Preiss, B. A.; Lipsky, S. R. Anal. Chem. 1967, 39, 14221428. (2) Horva´th, C. G.; Lipsky, S. R. Anal. Chem. 1969, 41, 1227-1234. (3) Vissers, J. P. C.; Claessens, H. A.; Cramers, C. A. J. Chromatogr. 1997, 779, 1-28. (4) Vissers, J. P. C. J. Chromatogr. 1999, 856, 117-143. (5) Jinno, K.; Fujimoto, C. LC-GC 1989, 7, 328-338. (6) Jinno, K. Chromatographia 1988, 25, 1004-1011. 10.1021/ac026125z CCC: $25.00 Published on Web 01/25/2003
© 2003 American Chemical Society
and capillary-LC, especially in combination with mass spectrometry (MS). Recent development of new LC-MS interfaces, such as highly sensitive matrix-assisted laser desorption/ionization (MALDI) and electrospray (ESI), requires even lower flow rates (0.9 µL in the above-mentioned example). In this case, three subsequent segments can be simultaneously present in the chamber, resulting in a faster and deeper mixing of the three eluents. This phenomenon, extended to all loop changes, leads to a linear-like gradient, as demonstrated experimentally.
Figure 7. Plot profiles relative to linear gradients from 100% H2O to 100% CH3CN using the MP valve with a mixing chamber in: (a) 6, (b) 18, and (c) 30 min at (1) 2, (2) 700, (3) 500, and (4) 300 nL‚min-1. The volume of the mixing device was 5 µL for the highest flow rate and 1 µL for the others. Arrows indicate the gradient start.
Figure 8. Plot profiles relative to linear gradients obtained in split mode, from 100% H2O to 100% CH3CN in (a) 6, (b) 18, and (c) 30 min at (A) 2, (B) 700, (C) 500, and (D) 300 nL‚min-1. Arrows indicate the gradient start.
Preliminary tests carried out using 500-µm-i.d. PEEK tubings of 6-, 10-, and 15-µL volume and 3-, 5-, and 8-cm length were unsatisfactory, giving incomplete smoothing of the plot. As a matter of fact, the shape of the mixing chamber is as important as its volume, since for a given volume, a long and narrow chamber would not allow a sufficient mixing of the mobile phase. A high-pressure gradient microsplitter valve (P-470 Upchurch Scientific), with a known and variable internal volume (1.2-2.8 µL) was used as a mixing chamber. The valve has three openings: by simply closing one of them, it is possible to create a variable internal volume that can be adjusted by a manual knob. The volume offered by this valve is suitable when working at a flow rate of 1 µL‚min-1. The gradient shape, smooth in all cases, presents a considerable delivery delay, probably due to the asymmetric internal geometry of the valve. Moreover, working at flow rates higher than 1 µL‚min-1, the maximum volume of the valve (2.8 µL) was too small, and the consequent plot was stepwise, while at flow rates lower than that, the minimum volume (1.2 µL) was too large causing an unacceptable delay. Better results were obtained using a modified union, shown in Figure 6, in which the internal volume was increased by removing the separation septum and by varying the connection fittings in order to generate three mixing chambers of 1-, 3-, and 5-µL volume. This device was tested at flow rates of 2, 700, 500, and 300 nL‚min-1with gradients completed in 6, 18, and 30 min. The volume of the mixing device was selected according to the
flow rate. In Figure 7, the relative plots are reported. The mixing chamber volume selected for the experiments at 2 µL‚min-1 flow rate was set at 5 µL. The gradient shape obtained was smooth and the mobile-phase delay was limited, except for the slowest run for which a larger mixing volume would be more advantageous. In any event, the time delay was lower than that obtained in split mode, with a high correspondence between the expected and the real gradient. For the 700 and 500 nL‚min-1 experiments, the chamber was reduced to 1 µL. In all cases, the gradients were very smooth, with a limited time delay. For 300 nL‚min-1 experiments, even the smallest chamber of 1 µL was too large, causing an excessive time delay of the gradient. However, at such low flow rate, no additional mixing volume was needed, because the gradient shape was already smoothed, as stated above. As a comparison, in Figure 8, linear gradients were obtained in split mode, starting from 100% H2O to 100% CH3CN in 6, 18, and 30 min at 2, 700, 500, and 300 nL‚min-1. Delays were observed in all cases, with unsatisfactory correspondence between actual and expected gradient profile. These experiments show that, when operating in split mode, the flow rate of the mobile phase should not be below the microliter-per-minute range to ensure excellent delivery with a limited time delay of the gradient. The different mobile-phase compositions delivered by the six loops can be displaced as required by the shape of the gradient Analytical Chemistry, Vol. 75, No. 5, March 1, 2003
1177
Table 1. Retention Time Reproducibility Test peak
retention times (min)
X h
RDS (%)
1
18.34; 18.51; 18.03; 18.15; 17.92; 18.01; 18.54; 18.25; 18.18; 17.98; 18.43; 18.27; 18.04; 18.25; 18.16 19.73, 19.38; 19.31; 19.09; 19.52; 19.45; 19.44; 19.28; 19.01; 19.13; 19.27; 19.71; 19.68; 19.10; 19.17 20.04; 20.36; 20.58; 19.98; 20.52; 20.08; 20.65; 20.43; 19.98; 20.03; 20.08; 20.31; 20.28; 20.17; 20.13 21.50; 21.74; 22.08; 21.68; 21.90; 21.50; 21.65; 21.73; 21.99; 22.04; 21.45; 21.57; 21.77; 21.66; 21.85
18.20
1.02
19.35
1.17
20.24
1.07
21.74
0.88
2 3 4
Figure 9. (a) Convex (A), linear (B), and concave (C) gradient profiles obtained from 100% H2O to 100% CH3CN at 700 nL‚min-1 in 18 min, using different valve programs. An arrow indicates the gradient start; (b) expected theoretical plots.
Figure 10. Separation of lactoalbumin tryptic digest, using a Agilent Technologies nano-HPLC column Zorbax 100-µm i.d. × 15-cm, C18 3.5-µm particle size; flow rate, 250 nL‚min-1; linear gradient from 100% H2O to 100% CH3CN in 18 min. Arrows indicate the peaks of interest.
and the complexity of the separation. Rapid, subsequent switching of the valve produces sharp gradients, while longer intervals produce more slack variations. Each switching time can be adjusted independently from the others, thus generating nonlinear gradients (i.e., convex, concave). The effective and direct control over the mobile-phase composition at a given time offers, even at nanoflow rates, the same capabilities of a conventional one. The effectiveness of complex valve switching programs on the gradient shape at very low flow rates is shown in Figure 9a, where three, 18-min-long gradient profiles, obtained at 700 nL‚min-1, using different valve programs, are reported. The loop switching times were the following: for curve A, which represents the convex 9 min 5 min 3 min 1 min
gradient, 2 f 3 f 4 f 5 f 6; for curve B, which represents
Figure 11. Separation of a mixture of five human hormones, testosterone, 19-nortestosterone, (+)-17-methyltestosterone, diethylstilbestrol, and medroxyprogesterone acetate, as they appear in their elution order, in split mode, and with the MP valve at micro- and nanoflow rates. Gradients from 100% H2O to 100% CH3CN were completed in 6, 18, and 30 min at 2 and 500 nL‚min-1. Injection volume, 60 nL. 250-µm i.d. × 25-cm, C18, 5-µm particle size capillary column was used at 2 µL‚min-1 flow rate, while a 100-µm i.d. × 15cm, C18, 3-µm particle size capillary column was used at 500 nL‚min-1 flow rate. UV detector wavelength, 254 nm.
4.5 min 4.5 min 4.5 min 4.5 min
the linear gradient, 2 f 3 f 4 f 5 f 6; for curve C, 1 min 3 min 5 min 9 min
which represents the concave gradient, 2 f 3 f 4 f 5 f 6. The expected theoretical plots are reported in Figure 9b. The absorbance traces are very well differentiated, as expected from the selected time programs. This result is particularly noteworthy, considering the extremely reduced operating flow rate. The initial 1178 Analytical Chemistry, Vol. 75, No. 5, March 1, 2003
dent of the convex curve C is probably due to an imperfect segment mixing at high percentage of water. In fact, when a reverse gradient is performed, this drawback is no longer observed. The dotted line represents the mirror image of the absorbance trace obtained changing the mobile-phase composition
from 100% CH3CN to 100% H2O, following the above-reported switching time program. Valve reproducibility was evaluated in real conditions injecting a lactoalbumin tryptic digest solution at 250 nL‚min-1, using a Agilent Technologies nano-HPLC column Zorbax 100-µm i.d. × 15-cm C18 3.5-µm particle size (Figure 10 and Table 1). Such a low flow rate was chosen to test the new system in extreme conditions, as often required in proteomics. The gradient was linearly varied from 100% H2O to 100% CH3CN in 18 min. Relative standard deviation of the retention times was calculated on 15 replicates and demonstrates that the system generates highly reproducible gradients. Separation Tests. Separation tests on a mixture of five human hormones, testosterone, 19-nortestosterone, (+)-17-methyltestosterone, diethylstilbestrol, and medroxyprogesterone acetate, listed in their elution order, were carried out with both split and MP valve systems at micro- and nanoflow rates (Figure 11). Gradients from 0% CH3CN to 100% CH3CN in H2O were completed in 6, 18, and 30 min at 2 and 500 nL‚min-1. The injection volume was 60 nL in all cases, and the absolute amount injected was 120 ng of each compound for microscale experiments and 30 ng for nanoscale experiments. A 250-µm i.d. × 25-cm, C18, 5-µm particle size capillary column was used at 2 µL‚min-1 flow rate, while a laboratory-made, 100-µm i.d. × 15-cm, C18, 3-µm particle size capillary column was employed at 500 nL‚min-1 flow rate. The UV detector was set at 254 nm. As one can note, a sixth peak showed up in most chromatograms. This is due to a diethylstilbestrol impurity that absorbs at the selected wavelength. The chromatographic profiles obtained with the MP valve demonstrate, in all cases, a higher resolution and a better distribution of the peaks. In particular, using the slowest gradient at both flow rates, the hormones are better separated with the MP valve (fully at 2 µL‚min-1), with a flatter baseline than that obtained with the split device. Gradient generation obtained in split mode may not exactly reproduce the one set at the pumps, and it explains different elution order compared to those obtained with the MP valve. This phenomenon is particularly stressed at very low flow rates. It also explains the scarce variation of retention times using different gradient slopes. In addition, at the lowest flow rate and for all gradients, the MP valve outperforms the split device in terms of peak distribution and overall resolution.
CONCLUSIONS Most of the recent liquid chromatography-mass spectrometry interfaces are based on reduced flow rate devices for very sensitive detection and low contamination. Dedicated small-scale HPLC systems are slowly appearing in the market; however, their use is limited to micro- and nanoapplications. On the other hand, conventional HPLC can be commonly found in modern laboratories for multipurpose use. As a consequence, it is desirable to be able to inexpensively convert a conventional instrument into unfailing and easy-to-use micro- and nanoequipment. We studied a simple and relatively inexpensive approach that could be afforded and operated in any laboratory. The simplicity was not obtained to the detriment of performance that, on the contrary, is of an excellent level, as demonstrated by the experimental results. The inexpensiveness of this system can be appreciated also in the long run, since the use of organic solvents is actually minimized and confined only to low flow rates. No useless waste of costly, hazardous, or environmentally dangerous solvents is needed to operate the new system, while a limited consumption of water is required for high flow rate operations. The advantages of this device are not represented only in a mere system scaling-down, since the system stands out for its ability to directly and effectively control the shape of the gradients even at the lowest flow rate considered (250 nL‚min-1), producing very reproducible linear and nonlinear gradients. Difficult separations can benefit from the accuracy of the gradient generation, and the results can match those obtained at higher flow rates. ACKNOWLEDGMENT We sincerely appreciate the discussions with Prof. Umberto Giostra, Centro di Modellistica Ambientale, Facolta` di Scienze Ambientali, Universita` di Urbino, about the dynamic of fluids. We thank Dr. Gerard Rozing from Agilent Technologies for providing the nano-HPLC columns. We are also grateful to Upchurch Scientific for providing the P-470 high-pressure graduated microsplitter valve and for their constant support to our work.
Received for review September 11, 2002. Accepted January 2, 2003. AC026125Z
Analytical Chemistry, Vol. 75, No. 5, March 1, 2003
1179
