Anal. Chem. 2010, 82, 7015–7020
Selection of Column Dimensions and Gradient Conditions to Maximize the Peak-Production Rate in Comprehensive Off-Line Two-Dimensional Liquid Chromatography Using Monolithic Columns Sebastiaan Eeltink,*,† Sebastiaan Dolman,‡,⊥ Gabriel Vivo-Truyols,§ Peter Schoenmakers,§ Remco Swart,‡ Mario Ursem,‡ and Gert Desmet† Department of Chemical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium, Dionex Corporation, Abberdaan 114, 1046 AA Amsterdam, The Netherlands, and Van ’t Hoff Institute for Molecular Sciences, University of Amsterdam, Nieuwe Achtergracht 166, 1018 WV Amsterdam, The Netherlands The peak-production rate (peak capacity per unit time) in comprehensive off-line two-dimensional liquid chromatography (LC/×/LC) was optimized for the separation of peptides using poly(styrene-co-divinylbenzene) monolithic columns in the reversed-phase (RP) mode. A firstdimension (1D) separation was performed on a monolithic column operating at a pH of 8, followed by sequential analysis of all the 1D fractions on a monolithic column operating at a pH of 2. To obtain the highest peak-production rate, effects of column length, gradient duration, and sampling time were examined. RP/×/RP was performed at undersampling conditions using a short 10 min 1D gradient. The peak-production rate was highest using a 50 mm long 2D column applying an 8-10 min 2D gradient time and was almost a factor of two higher than when a 250 mm monolithic column was used. The best way to obtain a higher peakproduction rate in off-line LC/×/LC proved to be an increase in the number of 1D fractions collected. Increasing the 2D gradient time was less effective. The potential of the optimized RP/×/RP method is demonstrated by analyzing proteomics samples of various complexities. Finally, the trade-off between peak capacity and analysis time is discussed in quantitative terms for both onedimensional RP gradient-elution chromatography and the off-line two-dimensional (RP/×/RP) approach. At the conditions applied, the RP/×/RP approach provided a higher peak-production rate than the 1D-LC approach when collecting three 1D fractions, which corresponds to a total analysis time of 60 min. Porous polymer monolithic columns have emerged in the 1990s1,2 and have become a viable alternative for packed-column * Corresponding author. Tel.: +32 (0)2 629 3324. Fax: +32 (0)2 629 3248. E-mail:
[email protected]. † Vrije Universiteit Brussel. ‡ Dionex Corporation. § University of Amsterdam. ⊥ Present address: Bruker Biosciences, 1/28A Albert Street, Preston VIC 3072, Australia. (1) Hjerte´n, S; Liao, J.-L; Zhang, R. J. Chromatogr., A 1989, 473, 273–275. 10.1021/ac101514d 2010 American Chemical Society Published on Web 07/28/2010
technology for the gradient-elution separation of peptides and proteins.3-5 Macroporous polymer monolithic materials were initially developed by Svec et al. in large I.D. column formats via a molding process.6,7 Macroporous poly(styrene-co-divinylbenzene) (PS-DVB) monolithic rods were operated at high flow rates (up to 25 mL/min) for the gradient separations of proteins.7 This group also demonstrated that the porous properties could be controlled by optimizing the polymerization mixture, i.e., type and composition of the porogenic solvent, and the percentage of difunctional monomer (“cross-linker”) in the mixture.8,9 Huber and co-workers developed styrene-based monoliths in situ in 200 µm I.D. capillary columns and applied these columns for liquidchromatographic mass-spectrometric (LC-MS) analysis of samples from proteomics and genomics studies, including peptides, proteins, single-stranded oligonucleotides, and double-stranded DNA fragments.10-12 Excellent separation performance was, for example, demonstrated by a baseline separation of phosphorylated oligodeoxyadenylic acids, ranging in size from the 12-mer to the 60-mer, and yielded peak widths of only 5.7 s when applying a 7.5 min gradient,13 resulting in a peak capacity of approximately 80. Recently, we reported the use of a 1 m long poly(styrene-codivinylbenzene) monolithic column yielding a peak capacity in excess of 1000 for a peptide separation when applying a 600 min gradient.14 (2) Peters, E. C.; Petro, M.; Svec, F; Frechet, J. M. J. Anal. Chem. 1997, 69, 3646–3649. (3) Ivanov, A. R.; Zang, L.; Karger, B. L. Anal. Chem. 2003, 75, 5306–5316. (4) Geiser, L.; Eeltink, S.; Svec, F.; Frechet, J. M. J. J. Chromatogr., A 2008, 1188, 88–96. (5) Levkin, P. A.; Eeltink, S.; Stratton, T. R.; Brennen, R.; Robotti, K.; Killeen, K.; Svec, F.; Frechet, M. J. M. J. Chromatogr., A 2008, 1200, 55–61. (6) Svec, F.; Frechet, J. M. J. Anal. Chem. 1992, 54, 820–822. (7) Wang, Q. C.; Svec, F.; Frechet, J. M. J. Anal. Chem. 1993, 65, 2243–2248. (8) Viklund, C.; Svec, F.; Frechet, J. M. J. Chem. Mater. 1996, 8, 744–750. (9) Eeltink, S.; Geiser, L; Svec, F; Frechet, J. M. J. J. Sep. Sci. 2007, 30, 407–413. (10) Toll, H.; Wintringer, R.; Schweiger-Hufnagel, U.; Huber, C. G. J. Sep. Sci. 2005, 28, 1666–1674. (11) Holzl, G.; Oberacher, H.; Pitsch, S.; Stutz; Huber, C. G. Anal. Chem. 2005, 77, 673–680. (12) Walcher, W.; Toll, H.; Ingendoh, A.; Huber, C. G. J. Chromatogr., A 2004, 1053, 107–117. (13) Oberacher, H.; Mayr, B. M.; Huber, C. G. J. Am. Soc. Mass Spec. 2004, 15, 32–42. (14) Eeltink, S.; Dolman, S.; Detobel, F.; Swart, R.; Ursem, M.; Schoenmakers, P. J. J. Chromatogr., A 2010, in press.
Analytical Chemistry, Vol. 82, No. 16, August 15, 2010
7015
For the analysis of complex sample mixtures, as for example encountered in proteomics research, the peak capacities that can be realized with one-dimensional (1D) LC often do not suffice to achieve complete separation of all compounds. Multidimensional separation approaches have the potential to separate thousands of components within a reasonable time. In first approximation, the maximum peak capacity in two-dimensional LC is the product of the peak capacities of the individual dimensions, provided that the two retention mechanisms are orthogonal.15 Comprehensive two-dimensional liquid chromatography, where all essential fractions of the sample are being analyzed in both dimensions, can be subdivided into two main categories, i.e., online comprehensive two-dimensional LC (or LC×LC) and off-line comprehensive two-dimensional LC (or LC/×/LC).16 In LC×LC, column dimensions must be selected such that the eluent composition and transfer volume match the LC conditions used in the second dimension. Schoenmakers showed that the maximum flow rate in the first-dimension and, consequently, the diameter of the 1D column depends on the maximum injection volume in the second dimension and on the second-dimension analysis time.17 The optimization of sampling time in online two-dimensional LC has recently been reviewed by Guiochon et al.18 When the sampling rate is too low, resolution achieved in the first dimension is partially lost. However, more time is available to achieve a high peak capacity in the second dimension. The fraction of potential peak capacity of the twodimensional combination of columns that is lost due to the selection of a too small modulation frequency is described by the “Nobuo factor” (modulation efficiency).19 Typically, for LC×LC, two cuts per first-dimension peak provides the best trade-off between the loss of resolution in the first dimension and the analysis time in the second dimension.19,20 However, from a practical point of view, this is difficult to achieve since modulation-phase effects and the random sampling process must be considered. The off-line approach (LC/×/LC) offers more flexibility for the selection of column dimensions and elution conditions, since the second-dimension analysis time can be optimized independently from the first-dimension sampling time. In addition, the organic modifier can be evaporated, so that fractions can be concentrated or dissolved in a different solvent prior to reinjection.18,21 To enhance the preconcentration of peptides or proteins on a trap column, a strong ion-pairing agent can be added prior to the second-dimension separation.21 As a result, the flow rates, transfer volumes, and the compatibility of (first and second dimension) eluent composition are less critical issues in LC/×/LC. The LC/×/LC setup can be optimized for proteomics applications. A large I.D. first-dimension column can be used, which provides (15) Giddings, J. C. Anal. Chem. 1984, 65, 1258A–1270A. (16) Schoenmakers, P. J.; Marriott, P.; Beens, J. LC-GC Eur. 2003, 16, 335– 339. (17) van der Horst, A.; Schoenmakers, P. J. J. Chromatogr., A 2003, 1000, 693– 709. (18) Guiochon, G.; Marchetti, M.; Mriziq, K.; Shalliker, R. A. J. Chromatogr., A 2008, 1189, 109–168. (19) Horie, K.; Kimura, K.; Ikegami, T.; Iwatsuka, A.; Saad, N.; Fiehn, O.; Tanaka, N. Anal. Chem. 2007, 79, 3764–3770. (20) Davis, J. M.; Stoll, D. R.; Carr, P. W. Anal. Chem. 2008, 80, 461–473. (21) Eeltink, S.; Dolman, S.; Swart, R.; Ursem, M.; Schoenmakers, P. J. J. Chromatogr., A 2009, 1216, 7368–7374.
7016
Analytical Chemistry, Vol. 82, No. 16, August 15, 2010
sufficient loadability for the analysis of samples with a broad dynamic range. A small I.D. column can be applied in the second dimension, minimizing chromatographic dilution and maximizing ionization efficiency when coupling LC with mass spectrometry via an electrospray interface. Huber and co-workers compared the sequence coverage obtained after an LC/×/LC-MS/MS separation of a tryptic digest of C. glutamicum using either a strong cationexchange (SCX) column or a reversed-phase (RP) column operated at high pH in the first dimension and a reversed-phase monolith operated at low pH in the second dimension.22,23 They observed that the SCX/×/RP and RP/×/RP approaches complemented each other. Recently, our group compared the 1D-LC performance of a 50 mm long monolithic column with that of an LC/×/LC approach employing a weak-anion exchange and an RP monolithic column in the first and second dimensions, respectively.24 At the conditions applied, the WAX/×/RP approach provided a better peak-capacity-per-analysis-time ratio for separations requiring a peak capacity of 400 or higher. The present contribution discusses how to maximize the peakproduction rate in off-line comprehensive two-dimensional liquid chromatography for the reversed-phase separations (RP/×/RP) of peptides performed at high and low pH using monolithic column technology. The effects of the first-dimension (1D) column dimensions (length and diameter) and LC conditions (flow rate and gradient time) on peak width and sampling volume are discussed. In addition, the effects of 2D column length and gradient time on peak-production rate in 2D-LC are demonstrated at “undersampling” conditions (i.e., containing fewer fractions than required to essentially maintain the first-dimension separation). Finally, the potential of RP/×/RP is demonstrated with separations of proteomic samples of varying complexity. EXPERIMENTAL SECTION Chemicals and Materials. Acetonitrile (ACN, HPLC supragradient quality), heptafluorobutyric acid (HFBA, ULC/MS quality), and trifluoroacetic acid (TFA, ULC/MS quality) were purchasedfromBiosolve(Valkenswaard,TheNetherlands).Ammonium bicarbonate (min. 99%), dithiothreitol (min. 99%), iodoacetic acid (approximately 99%), guanidine-HCl, sodium chloride (analytical reagent grade), cytochrome c (bovine heart), and apo-transferrin (bovine, g98%) were purchased from Sigma-Aldrich (Steinheim, Germany). Lysozyme (hen egg white), alcohol dehydrogenase (yeast), serum albumin (bovine, assay >96%), β-galactosidase, and sodium phosphate monobasic dihydrate (analytical reagent grade) were obtained from Fluka (Buchs, Switzerland). Escherichia coli (E. coli, strain K12) protein sample (lyophilized) was obtained from Bio-Rad Laboratories (Veenendaal, The Netherlands). Preconcentration and desalting of peptides prior to the analytical separation was performed on a 5 mm × 0.2 mm I.D. monolithic trap column (Pepswift RP, Dionex Benelux, Amsterdam, The Netherlands). HPLC separations were performed with 50 mm × 0.2 mm and 250 mm × 0.2 mm monolithic PepSwift RP columns (22) Delmotte, N.; Lasaosa, M.; Tholey, A.; Heinzle, E; Huber, C. G. J. Proteome Res. 2007, 6, 4363–4373. (23) Toll, H.; Oberacher, H.; Swart, R.; Huber, C. G. J. Chromatogr., A 2005, 1079, 274–286. (24) Eeltink, S.; Dolman, S.; Detobel, F.; Desmet, G.; Swart, R.; Ursem, M. J. Sep. Sci. 2009, 32, 2504–2509.
Table 1. Effect of 1D-LC Conditions on Sampling Time and Number of Fractions Collected When Applying a Modulation Ration (MR) ∼2 per 1D Peaka I.D. (mm)
flow rate (µL/min)
gradient time (min)
4σ peak width (s)
sampling time (s)
sampling volume (µL)
no. fractions
0.2 1
2 60
10 10
5.5 5.8
2.8 2.6
0.09 2.4
214 231
a
Column length is 50 mm; 1 µL injection of the six-protein digest; aqueous acetonitrile gradient from 1-26% (ACN) at pH ) 8; detection at 214
nm.
and with a 50 mm × 1 mm monolithic ProSwift RP-10R column (Dionex). Preparation of Tryptic Digests. The six-protein mixture prepared from transferrin, bovine serum albumin, β-galactosidase, alcohol dehydrogenase, lysozyme, cytochrome, and E. coli proteins were digested according the following procedure. Proteins were reduced for 1 h at 60 °C in the presence of 7 mol/L guanidine and 1 mol/L dithiothreitol, followed by alkylation for 30 min at room temperature by adding 1 mol/L iodoacetic acid. To consume any unreacted iodoacetic acid, 1 mol/L dithiothreitol was added. The reduced and alkylated proteins were then dialyzed against 50 mM ammonium bicarbonate (pH 8) for 24 h in dialysis sacks (Sigma) with a cutoff 0.997). Therefore, the relationship 2wb ) a + b(2tG) can be established, which holds at a constant length. After fitting, the a and b parameters were used to predict the peak capacity as a function Analytical Chemistry, Vol. 82, No. 16, August 15, 2010
7017
2D
ξ)
nc
(3)
ttot
Figure 2 illustrates the effect of the 2tG on the ξ using a 50 mm and 250 mm long monolithic column. The 1D RP separation was performed using a 50 mm × 1 mm monolithic column, applying a 10 min gradient at pH ) 8. The 2D analysis included a preconcentration and desalting step on a monolithic trap column (2tdesalt) and a RP gradient separation at pH ) 2 on the capillary column. Initially, a steep increase in ξ can be observed. In this region, ξ is dominated by the contributions of 2t0 and 2 teq time to the total analysis time in the second dimension. With increasing gradient time, the peak capacity initially strongly increases (Figure 1B) and the peak-production rate reaches a maximum. When applying even longer gradients, the peak-production rate decreases. This is because the peak capacity marginally increases with the gradient time at longer gradient duration (see Figure 1B). The existence of a maximum can also be demonstrated by taking the derivative in eq 3 with respect to 2 tG. Forcing this derivative to be 0 yields
2
tG,max )
Figure 1. Effect of gradient time on peak width (A) and peak capacity (B) using a 250 mm × 0.2 mm (closed symbols) and a 50 mm × 0.2 mm (open symbols) monolithic column. Sample: six-protein digest (transferrin, bovine serum albumin, β-galactosidase, alcohol dehydrogenase, lysozyme, and cytochrome c), 1 µL injection (0.5 pmol/ µL); flow rate: 2 µL/min; aqueous acetonitrile gradient from 1% to 35% with 0.05% TFA ion-pairing agent; column temperature: 60 °C. Detection at 214 nm using a 3 nL flow cell.
(t a b
1
st
td
1
G
)
+ 1 + 2td
(4)
where 2tG,max refers to the gradient time in the second dimension that yields the highest value for ξ. 1td ) 1t0 + 1tdwell, and 2td ) 2 tdesalt + 2t0 + 2tdwell + 2twash + 2teq. The maximum peakproduction rate using the 50 mm long monolithic column is obtained at a 2tG ∼ 9 min, and the maximum shifts to a higher
of tG (Figure 1B). For gradient times below 60 min, the peak capacity of the 50 mm long monolith was higher than that of the 250 mm monolith. This can be explained by the difference in morphology of the two monoliths.14 At undersampling conditions, the total peak capacity (2Dnc) and total analysis time (ttot) in off-line two-dimensional LC are given by 1 2D
nc )
tG 2tG × 2 st W
(1)
1
ttot ) 1t +
tG 2 ·t st
(2)
where 2tG is the second-dimension gradient time, 2W is the average second-dimension peak width (equivalent to 4σ), 1t is the first-dimension analysis time, 2t is the second-dimension analysis time, and st is the sampling time. The first-dimensional analysis time is the sum of the column holdup time (1t0), the dwell time (1tdwell), and the gradient time (1tG). The seconddimension analysis time is the sum of 2tdesalt, 2t0, 2tdwell, a wash step (2twash), and the second-dimension column equilibration time (2teq), which corresponds to the time needed to flush the column with three column volumes to obtain good retention time stability. The peak production rate (ξ, min-1) is defined as 7018
Analytical Chemistry, Vol. 82, No. 16, August 15, 2010
Figure 2. Effect of 2D gradient time on peak-production rate using 250 mm (a) and a 50 mm (b) long second-dimension monolithic column with a sampling time of 30 s (solid line), 50 mm column with sampling time of 60 s (- - -; c), and 50 mm column with sampling time of 120 s (---; d). First-dimension separation on a 50 mm × 1 mm long column, 1D delay time is 4 min, 10 min 1D gradient time. 1D gradient from 1% to 26% acetonitrile at pH ) 8 (10 mM aqueous ammonium carbonate buffer) at a flow rate of 60 µL/min. Seconddimension separation on a 250 mm column operating at a flow rate of 2 µL/min included a 1 min preconcentration, delay time of 0.4 min, t0 time of 3.53 min, 0.5 min wash step, and 10.6 min equilibration time; 2D gradient from 1% to 35% with 0.05% TFA ion-pairing agent. Second-dimension separation on a 50 mm column operating at a flow rate of 2 µL/min included a 1 min preconcentration, delay time of 0.4 min, t0 time of 1.36 min, 0.5 min wash step, and 4.1 min equilibration time; 2D gradient from 1% to 35% with 0.05% TFA ionpairing agent.
Figure 3. Peak-production rate versus total analysis time when either a 50 mm (closed symbols) or a 250 mm (open symbols) long 2D monolithic column is used at optimal 2tG and two or more fractions are collected. LC conditions described in Figure 2.
value (20 min) when the 250 mm long monolith is used. This is because the a/b ratio increases with the column length (see Figure 1A); the slope of 2tG vs 2wb is lower for the 250 mm column, which implies that the b is lower. Also, the a decreases with column length. In addition, the ξ of the 50 mm long monolith is more than a factor of 2 higher than that of the 250 mm monolith. This is due to two effects. First, the region of 2tG considered (from 1 to 60 min) contemplates situations in which the 250 mm column yields broader peaks than the 50 mm column (see Figure 1A), reducing 2nc for the 250 mm column compared to the 50 mm monolith for a short gradient duration (Figure 1B). Second, with increasing column length, t0 and teq increase, affecting 2td and, hence, the ξ (see eq 3). Figure 2 also shows that the optimal 2tG depends on the sampling rate. This is clearly perceived from eq 4, where 2tG,max depends on st. The higher st, the higher the value of 2tG,max. Effect of Sampling Time on Peak-Production Rate. Figure 3 illustrates the ξ and total analysis time when either a 50 or a 250 mm long 2D monolithic column is used at optimal 2tG and two or more fractions are collected. When a 50 mm 2D column is used, first, a strong increase in ξ is observed. At higher sampling rates, the increase levels off and ξ tends to reach a maximum around 10 peaks/min. When a 250 mm 2D column is used, ξ increases slightly from 4.5 peaks/min (2 fractions) to a maximum of 5 peaks/min after sampling only three fractions. Apparently, when a 50 mm long 2 D column is used and a slow sampling rate is applied, ξ is significantly affected by the contribution of the 1D analysis time to the total analysis time. With decreasing sampling time or when longer 2D columns (longer t0 and teq) are used, ξ is dominated by the 2D analysis time. Figure 4 shows the RP(pH)8)/×/RP(pH)2) separation of a sixprotein digest when applying a sampling time of 60 and 30 s, respectively. The 1D separation was performed using a 50 mm × 1 mm monolithic column applying a gradient time of 10 min. The 2D separation was executed on a 50 mm × 0.2 mm long monolithic column, applying a 2tG of 7.5 min. The maximum theoretical peak capacity of the separation shown in Figure 4A is 1200, and the separation was completed in 98 min. The
Figure 4. RP (pH ) 8)/×/RP (pH ) 2) separation of a digest of six proteins showing the effect of sampling time (st ) 60 s (A); st ) 30 s (B)) on peak capacity and total analysis time. 50 mm × 1 mm 1D column and 50 mm × 0.2 mm 2D column. Further conditions as described in Figure 2.
Figure 5. Off-line 2D-LC separation of an E. coli digest using a sampling time of 15 s. 50 mm × 1 mm 1D column and 50 mm × 0.2 mm 2D column. Further conditions as described in Figure 2.
separation in Figure 4B yielded a maximum 2D-LC peak capacity of 2400 in 164 min. Whereas the theoretical peak capacity doubles, the total analysis time increased only by 60%. It should be noted that the separation space was not completely filled. This is because the retention mechanisms of the reversedphase separations performed at pH ) 8 and 2, respectively, are not completely independent. As a consequence, the fraction of the total peak capacity that is actually used (“sample peak capacity”) is lower than the theoretical peak capacity. For the analysis of more complex samples, such as an E. coli digest, the most productive way to obtain a higher peak capacity in LC/×/LC, while still working at undersampling conditions, is to increase the number of 1D fractions collected; see Figure 2. Increasing 2tG is less effective. Figure 5 shows the LC/×/LC separation of E. coli digest. The sampling rate (15 s) was selected such that a maximum theoretical peak capacity of 6700 could be achieved within a total analysis time of 740 min. Trade-Off between 1D-LC and the Optimized LC/×/LC Approach. Figure 6 illustrates the trade-off between the 1D-LC system using a 50 mm and 250 mm long monolithic column and the RP(pH)8)/×/RP(pH)2) using a 50 mm long monolithic column in each dimension. The total 1D-LC analysis time is the sum of the desalting time (0.5 min), the column holdup Analytical Chemistry, Vol. 82, No. 16, August 15, 2010
7019
can already be achieved in 60 min using LC/×/LC when analyzing only three 1D fractions. The peak capacity linearly increases with the number of fractions collected and approximately doubles every 60 min.
Figure 6. Trade-off between 1D-LC performance using 50 mm (solid line) and 250 mm (dotted line) long monolithic columns and the optimized off-line 2D-LC approach collecting g2 fractions (solid symbols). 1D-LC conditions described in Figure 1; 2D-LC conditions described in Figure 2.
time, the dwell time, and the gradient time. The lines depict the 1D-LC performance. The closed circles represent the RP(pH)8)/×/RP(pH)2) performance collecting discrete numbers of fractions (two and more). It is evident that below 45 min the highest ξ is obtained using 50 mm long monolithic columns in the 1D-LC mode. This yields a maximum peak capacity of 320. At a total analysis time of 70 min, 50 mm and 250 mm long monolithic columns show comparable performance. For 1 D-LC analysis longer than 70 min, the 250 mm monolithic column provides the ξ. The maximum 1D-LC peak capacity of 475 can be achieved within 3 h. However, the ξ is much smaller than what can be realized by LC/×/LC. A peak capacity of 480
7020
Analytical Chemistry, Vol. 82, No. 16, August 15, 2010
CONCLUDING REMARKS An approach to select column dimensions and LC conditions in LC/×/LC to maximize ξ have been demonstrated. Although the LC/×/LC optimization strategy has been performed with monolithic columns, this approach can be used when applying other types of LC columns. At undersampling conditions, the 1D separation performance is subsidiary of sampling rate. As a result, a short first-dimension analysis time is optimal. This can be achieved using short 1D column length and applying a short 1tG. The optimal 2tG depends on 2D column length and on sampling rate. The optimal 2tG shift to lower values when shorter 2D columns are used yields higher values for ξ. For separations requiring a maximum peak capacity up to 340, 1 D-LC using a 50 mm long monolithic column was found to be superior to LC/×/LC in terms of analysis time. For more demanding separations, 250 mm long monoliths can be used, applying longer gradient times (nc, max = 475). However, the ξ in LC/×/LC is superior for an analysis taking longer than 60 min (collecting three fractions and more), and the achievable peak capacity increases linearly with analysis time. ACKNOWLEDGMENT Support of this work by a grant of the Research Foundation Flanders (G.0919.09) is gratefully acknowledged. Received for review June 8, 2010. Accepted July 17, 2010. AC101514D
