A Method To Reduce Gradient Delay Time of NanoLC - American

Feb 17, 2007 - Hongji Liu,* Jeffrey W. Finch, and John C. Gebler. Life Sciences R&D, Waters Corporation, 34 Maple Street, Milford, Massachusetts 01757...
0 downloads 0 Views 81KB Size
Anal. Chem. 2007, 79, 2961-2964

A Method To Reduce Gradient Delay Time of NanoLC Hongji Liu,* Jeffrey W. Finch, and John C. Gebler

Life Sciences R&D, Waters Corporation, 34 Maple Street, Milford, Massachusetts 01757

A novel method to dramatically reduce the delay time of a nanoLC gradient is described. The gradient is divided into two parts, and each part is formed at different flow rates. The beginning part is formed and delivered to the inlet of the column at a higher-than-normal flow rate. With the formation of the rest of the gradient at a normal flow rate, the whole gradient is further delivered through the column at the same normal flow rate. To form the gradient with the desired slope, the volumetric gradient slope was kept constant, independent of the flow rate. A gradient delay time reduction of 12.5-16 min was observed with the reported method. The resulting gradient profiles and chromatograms were very similar to those obtained with a conventional method. Comparable retention time reproducibility was observed between the two methods. Liquid chromatography operated with a capillary column at a nanoliter per minute scale (nanoLC) has become an effective method for high sensitivity analysis.1-3 With a dramatic reduction in the amount of sample needed for analysis, nanoLC is especially attractive in the area of proteomic studies,4-9 where the analyte amount is sometimes limited. The reduction in the sample demand at a lower flow rate results from the reduction in the dilution effect of the mobile phase on the loaded sample. Another benefit of using nanoLC for proteomic studies is its excellent flow rate compatibility with electrospray ionization interfaced mass spectrometry (ESI-MS), one of the most powerful methods for protein identification. In spite of these benefits, a drawback of nanoLC has been noticed, which is the extensive time delay in the retention times * Corresponding author. Fax: 508-482-3625. E-mail: [email protected]. (1) Martin, S. E.; Shabanowitz, J.; Hunt, D. F.; Marto, J. A. Anal. Chem. 2000, 72, 4266-74. (2) Quenzer, T. L.; Emmett, M. R.; Hendrickson, C. L.; Kelly, P. H.; Marshall, A. G. Anal. Chem. 2001, 73, 1721-5. (3) Haskins, W. E.; Wang, Z.; Watson, C. J.; Rostand, R. R.; Witowski, S. R.; Powell, D. H.; Kennedy, R. T. Anal. Chem. 2001, 73, 5005-14. (4) Link, A. J.; Eng, J.; Schieltz, D. M.; Carmack, E.; Mize, G. J.; Morris, D. R.; Garvik, B. M.; Yates, J. R., III. Nat. Biotechnol. 1999, 17, 676-82. (5) Shen, Y.; Zhao, R.; Belov, M. E.; Conrads, T. P.; Anderson, G. A.; Tang, K.; Pasa-Tolic, L.; Veenstra, T. D.; Lipton, M. S.; Udseth, H. R.; Smith, R. D. Anal. Chem. 2001, 73, 1766-75. (6) Shen, Y.; Tolic, N.; Masselon, C.; Pasa-Tolic, L.; Camp, D. G., II; Hixson, K. K.; Zhao, R.; Anderson, G. A.; Smith, R. D. Anal. Chem. 2004, 76, 14454. (7) Bergen, H. R. 3rd; Vasmatzis, G.; Cliby, W. A.; Johnson, K. L.; Oberg, A. L.; Muddiman, D. C. Dis. Markers 2003, 19, 239-49. (8) Ottens, A. K.; Kobeissy, F. H.; Wolper, R. A.; Haskins, W. E.; Hayes, R. L.; Denslow, N. D.; Wang, K. K. Anal. Chem. 2005, 77, 4836-45. (9) Tolson, J. P.; Flad, T.; Gnau, V.; Dihazi, H.; Hennenlotter, J.; Beck, A.; Mueller, G. A.; Kuczyk, M.; Mueller, C. A. Proteomics 2006, 6, 697-708. 10.1021/ac061942a CCC: $37.00 Published on Web 02/17/2007

© 2007 American Chemical Society

of the separated species with gradient elution, a major elution method for proteomic LC separations. The delay in the retention time is a result of a significant duration of the gradient delay time, which is a time period between the moment when the gradient is formed and the moment when it reaches the column. The gradient delay time is caused by the existence of system components located between the gradient mixing point and the column inlet, such as the sample loop, the injection valve, and other the connection parts (tubing, unions, tees, etc.). The total volume inside these components is the gradient delay volume. Given a constant gradient delay volume, the lower the flow rate, the longer the gradient delay time. For example, the delay time caused by a 5 µL delay volume is 0.005 min if the flow rate is 1 mL/min, but is 20 min if the flow rate is 250 nL/min. Currently, the gradient delay time can be reduced in several ways, one of which is to reduce the size of the LC components contributing to the gradient delay volume.10,11 But there is a practical limit to reducing the gradient delay volume. For example, to allow a 5 µL sample to be injected for analysis, the sample loop size has to be at least 5 µL. Running nanoLC at a higher-thannormal flow rate is another way to reduce the gradient delay time, but one has to tolerate the loss of sensitivity due to the use of the higher flow rate. Splitting the flow immediately prior to the column allows a sample to be run through the column at a normal flow rate with the reduction in the gradient delay time, but since the splitting point is placed downstream the injector, a large portion of the injected sample is diverted to waste along with the split mobile phase.12 In this work, we report a novel method, which minimizes the gradient delay time without causing any significant drawbacks. To reduce the delay time of a gradient with the new method, part of the gradient is formed and delivered at a relatively higher flow rate to the column inlet. Forming a gradient at a relatively higher flow rate than used for delivery has previously been used by others, mostly to form a more accurate gradient that could not otherwise be possibly formed at the flow rates used for separations.13-17 In any of these previous studies, however, a (10) Chervet, J. P.; Ursem, M.; Salzmann, J. P. Anal. Chem. 1996, 68, 15071512. (11) Ivanov, A. R.; Zang, L.; Karger, B. L. Anal. Chem. 2003, 75, 5306-16. (12) Alexander, J. N.; Poli, J. B.; Markides, K. E. Anal. Chem. 1999, 71, 23982409. (13) Katz, E.; Scott, R. P. J. Chromatogr. 1982, 253, 159-178. (14) Schachterle, S.; Alfredson, T. Anal. Chem. 1986, 58, 1368-1372. (15) Berry, V.; Rohwer, E. J. Liq. Chromatogr. 1990, 13, 1529-1558. (16) Davis, M. T.; Stahl, D. C.; Lee, T. D. J. Am. Soc. Mass Spectrom. 1995, 6, 571-577.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007 2961

dedicated gradient storage device was added to the LC system, along with separate pumping systems for gradient formation and delivery. In addition, the entire gradient had to be formed at the higher flow rate. With the new method, only part of gradient is formed at a relatively higher flow rate, and no dedicated gradient storage device is needed. Instead, the part of gradient formed at the higher flow rate is stored in the existing LC components contributed to the gradient delay volume. With the new method, the same LC system is used for both gradient formation and delivery. Different from other reported methods, the main purpose of the new method is to minimize the gradient delay time. Methods. With the new method, a linear gradient of B0 to Bg (gradient g) is divided into two linear gradients (gradient 1 and gradient 2), where B0 and Bg are the starting and ending percentages of eluent B for gradient g. Gradient 1 starts from B0 and gradient 2 ends at Bg. The ending mobile phase composition of gradient 1 (B1) is the same as the starting composition of gradient 2, which can be calculated as described below. The volume of gradient 1 is approximately the same as the LC system’s gradient delay volume, which can be measured experimentally. The volume of gradient 2 equals the volume of gradient g, deducting that of gradient 1. Gradient 1 is formed and delivered to the column inlet at a relatively higher flow rate, while gradient 2 is formed at a normal flow rate. The mobile phase of gradient 2 along with the mobile phase used following gradient 2 for column reequilibration are employed to deliver the whole gradient through the column at a normal flow rate. Due to the use of a higher flow rate to form and deliver gradient 1, the gradient delay time is substantially reduced. Since gradients 1 and 2 are both delivered through the column at a normal flow rate, the use of a higher flow rate during the formation of gradient should not affect system sensitivity. The time needed to form gradient 1 (T1) can be calculated from the gradient delay volume of the LC system (Vd) and the flow rate used to form gradient 1 (F1) as

T1 ) Vd/F1

(1)

B1 ) [Vd(Bg - B0)/FgTg] + B0

(2)

B1 can be calculated as

where Fg and Tg are the flow rate and the gradient time for gradient g. The gradient time for gradient 2 is calculated as:

T2 - T1 ) Tg - (Vd/Fg)

(3)

where T2 is the time counted from the start of the formation of gradient 1 to the end of the formation of gradient 2. T3, the total time needed from the start of the formation of gradient 1 to the end of delivery of the whole gradient (gradients 1 and 2) through the column, can be calculated as

T3 ) Tg + T1

(4)

EXPERIMENTAL SECTION A nanoACQUITY UPLC nanoLC system from Waters Corp. (Milford, MA) was employed for separations. The system included 2962 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

Table 1. Conventional Gradient Table time (min) 0 30 50

flow rate (nL/min)

B%

curvea

250 250 250

3 60 3

6 1

a Curve numbers indicate changes in flow rate and mobile phase composition between adjacent lines. Curve 6 specified a linear change from the previous line to the line where the curve number was located, while Curve 1 suggests an immediate change of conditions at the time specified on the previous line to the line where the curve number is located.

Table 2. Converted Gradient Table for the New Methoda time (min)

flow rate (nL/min)

B%

curve

0 4 (T1) 4.01 18 (T2) 34 (T3)

1000 1000 250 250 250

3 33.4 33.4 60 3

6 1 6 1

a

Vd ) 4 µL, F1 ) 1000 nL/min.

a binary gradient UPLC pump, a nanoflow controller, and an autosampler with a 2 µL sample loop installed. A Q-Tof micro MS system was connected to the LC system to monitor separations. A PicoTip emitter from New Objective Inc (Woburn, MA) was used for ESI. When monitoring gradient profiles were required, the MS system was replaced with a K-2501 model UV detector (wavelength, 205 nm; flow cell, 3 nL) purchased from Knauer ASI (Berlin, Germany). A Waters Atlantis dC18 column (75 µm × 10 cm, 3 µm particle) was connected to the system with the temperature thermostatically controlled at 35 °C. The gradient delay volume of the system was about 4 µL. Eluents A and B were 0.1% formic acid in water and 0.1% formic acid in acetonitrile, respectively. No formic acid was contained in Eluent A when the gradient profiles were monitored with UV. A MassPREP tryptic enolase digest (Waters) was used as a test sample. The sample (100 fmol/µL) was injected with a full-loop injection mode. A linear gradient of 3-60% B at 250 nL/min for 30 min was applied to the column for separation. Two gradient formation methods were used: the conventional method and the new method. With the conventional method, the gradient was formed and delivered both at 250 nL/min. With the new method, the first part of the gradient was formed and delivered to the column inlet at a 1 µL/min flow rate, with the second part being formed and the whole gradient being delivered through the column at 250 nL/min. The chromatograms were acquired with a centroid MS mode at an acquisition rate of one spectrum per second, with the m/z range being 400-1700. RESULTS AND DISCUSSION Conversion of Conventional Gradient. Table 1 is a gradient table used for the conventional method. Converted from Table 1 with eqs 1-4, a gradient table for the new method is described in Table 2 (Vd ) 4 µL, F1 ) 1 µL/min). Plotted with the data shown (17) Cappiello, A.; Famiglini, G.; Fiorucci, C.; Mangani, F.; Palma, P.; Siviero, A. Anal. Chem. 2003, 75, 1173-9.

Figure 1. Theoretical gradient formation profiles. The curve with a dash line describes the gradient slope for gradient g with the conventional method, and the curve with a solid line illustrates the gradient slopes of gradients 1 and 2 with the new method. Vd ) 4 µL. F1 ) 1 µL/min. Gradient g: 3-60% B in 30 min at 250 nL/min.

Figure 2. Monitored gradient delivery profiles with UV (205 nm). (a) Conventional method. (b) New method. No column was connected.

in Tables 1 and 2, Figure 1 illustrates the necessity to increase the gradient slope when a higher flow rate was employed to form gradient 1. The gradient slope of gradient 1 (0-4 min, in Table

2), which was formed at a flow rate 4 times the normal flow rate, was 4 times of that of gradient 2 (4-18 min in Table 2). However, the volumetric gradient slope is the same for gradients 1, 2, and g (7.6 B%/µL). Gradient Profiles. Figure 2 compares the gradient profiles obtained with (a) the conventional method and (b) the new method. With similar profiles observed, the gradient delay time was reduced by 12.5 min as compared to the conventional method. The similarity of the two profiles can be reflected from the following data. The UV absorbance difference between the highest and the lowest points of the profiles shown in Figure 2 was 0.146 (a) and 0.150 (b), while the time difference between the highest and lowest points was 26.9 (a) and 27.4 min (b). The curve slope in profile a was 0.0584 AU/min from 23 to 28 min, and 0.046 AU/ min from 35 to 40 min, while the curve slope in profile b was 0.0584 AU/min from 10 to 15 min, and was 0.048 AU/min from 22.5 min to 27.5 min. With these similarities observed, there was a major difference noticed, which was located in the middle of the gradients. There was a ∼1 min isocratic hold in profile b at ∼18 min. The hold was likely caused by the flow rate change from 1 µL/min to 250 nL/min during the gradient formation at 4 min. An optimized flow control algorithm might be helpful to reduce the hold time. As indicated from the obtained chromatograms (see below), however, the hold did not generate a significant change to the whole chromatographic profile. Chromatograms. Figure 3 illustrates the similarity of chromatograms acquired with (a) the new method (Vd ) 4 µL) and (b) the conventional method. The similar chromatographic profiles obtained with the each of the methods indicate that, for most peptides, the separation was not affected by the use of the new gradient formation-delivery method. The reduction in the gradient delay time can be exhibited by the approximately 12.5 min retention time reduction in Figure 3a than in Figure 3b for both

Figure 3. Comparison of chromatograms acquired with the conventional method and the new method. Sample: 200 fmol of enolase digest. (a) The new method with Vd ) 4 µL. F1 ) 1 µL/min (see Table 2 for gradient). (b) The conventional method (see Table 1 for gradient). (c) The new method with Vd ) 5 µL. F1 ) 1 µL/min (a converted gradient was formed with the same method as used for Table 1).

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2963

the first and the last peaks. As result of the reduction, the runto-run time was reduced from 50 min to 37 min. Figure 3c is a chromatogram acquired with the new method with Vd equaled 5 µL, about 1 µL more than the delay volume of the LC system. The retention time differences of the first peaks (and the last peaks as well) between Figures 3b and 3c was 16 min, about the same as the gradient delay time as observed with the conventional method in Figure 3b. This indicates that the gradient delay time in Figure 3c was nearly eliminated. The success in the use of a Vd greater than the delay volume of the LC system suggests the acceptability of delivering an initial part of the gradient through the column at a higher-than-normal flow rate (as long as the delivery flow rate returns to normal before the elution of the first interested analyte). Excellent retention time reproducibility was observed with the new method. The standard deviations of the retention times of five enolase peptides were 0.04-0.09 min (n ) 6), comparable to the reproducibility data obtained with the same system using the conventional method. Pressure and Column Stability. Theoretically, gradient 1 can be formed at any high flow rate. Practically, however, the magnitude of the flow rate is restricted by such factors as the upper pressure limit of a LC system. During the current study, the pressure generated on the analytical column (10 cm long, packed with 3 µm particles) was approximately 2400 (160 bar) and 650 psi (40 bar) at 1000 nL/min (for gradient 1) and 250 nL/ min (for gradient 2), respectively. These pressures were low enough for the method to be carried out with any conventional nanoscale HPLC instrument. Even if a longer (e.g., 15 cm) column of the same type is used, the generated pressure should still be tolerable for any HPLC instrument. A 20 cm or longer column

2964

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

should readily be used with the method if packed with 5 µm particles, a more conventional particle size. For applications that require a gradient formation pressure intolerable by conventional LC (e.g., when a longer column is packed with particles smaller than 3 µm, and/or a much higher flow rate for gradient 1 is used), an LC system with a higher pressure limit may become necessary. Alternatively, the pressure can be reduced by heating the column under a higher temperature. Another solution is to add a short trap column to the system. The short column is connected between the injector and the analytical column, so that the analytical column can be bypassed when gradient 1 is formed. Due to a “buffering” effect from the liquid in the pumps and the connection tubing, it took 0.4 min for the pressure to be dropped from 2400 to 800 psi when the flow rate was instantly changed during the gradient, reducing the impact of the pressure change on the column stability. To minimize the impact, a column packed at a much higher pressure (∼9000 psi) was used for the current study. No performance degradation was observed after 30 runs were carried out with the column. Columns packed with the same conditions had previously been tested under a similar pressure fluctuation, and no performance degradation was observed after hundreds of injections on a single column. ACKNOWLEDGMENT We thank G. Gerhardt and S. Ciavarini for their support on system instrumentation. We also thank I. S. Krull and G. Gerhardt for reviewing the manuscript. Received for review October 14, 2006. Accepted January 16, 2007. AC061942A