Active Solvent Modulation – A Valve-Based Approach to Improve

1 – Gustavus Adolphus College. 6. Department of Chemistry. 7. 800 West College Avenue. 8. Saint Peter, MN, USA. 9. 10. 2 – Agilent Technologies. 1...
2 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Active Solvent Modulation: A Valve-Based Approach To Improve Separation Compatibility in Two-Dimensional Liquid Chromatography Dwight R. Stoll,*,† Konstantin Shoykhet,‡ Patrik Petersson,§ and Stephan Buckenmaier‡ †

Department of Chemistry, Gustavus Adolphus College, 800 West College Avenue, Saint Peter, Minnesota 56082, United States R&D and Marketing GmbH & Co KG, Agilent Technologies, Hewlett-Packard-Str. 8, 76337 Waldbronn, Germany § Global Research, Novo Nordisk A/S, Novo Nordisk Park, DK-2760, Måløv, Denmark ‡

S Supporting Information *

ABSTRACT: Two-dimensional liquid chromatography (2DLC) is increasingly being viewed as a viable tool for solving difficult separation problems, ranging from targeted separations of structurally similar molecules to untargeted separations of highly complex mixtures. In spite of this performance potential, though, many users find method development challenging and most frequently cite the “incompatibility” between the solvent systems used in the first and second dimensions as a major obstacle. This solvent strength related incompatibility can lead to severe peak distortion and loss of resolution and sensitivity in the second dimension. In this paper, we describe a novel approach to address the incompatibility problem, which we refer to as Active Solvent Modulation (ASM). This valve-based approach enables dilution of 1D effluent with weak solvent prior to transfer to the 2D column but without the need for additional instrument hardware. ASM is related to the concept we refer to as Fixed Solvent Modulation (FSM), with the important difference being that ASM allows toggling of the diluent stream during each 2D separation cycle. In this work, we show that ASM eliminates the major drawbacks of FSM including complex elution solvent profiles, baseline disturbances, and slow 2D re-equilibration and demonstrate improvements in 2D separation quality using both simple small molecule probes and degradants of heat-treated bovine insulin as case studies. We believe that ASM will significantly ease method development for 2D-LC, providing a path to practical methods that involve both highly complementary 1D and 2D separations and sensitive detection.

T

(APIs),8,9 excipients,10 and drug products.10,11 Readers interested in more background on these developments are referred to any of a number of recent review articles.12−14 At the heart of any 2D-LC system is a device that collects portions of effluent from the first dimension (1D) column and transfers them to the second dimension (2D) column for further separation of analytes that coeluted from the 1D column. This device is most commonly a valve or ensemble of valves fitted with one or more sampling loops, depending on the number of fractions that need to be transferred and the manner in which the collection and transfer is carried out. The device is sometimes referred to more broadly as a modulator. Egeness and co-workers recently reviewed the different devices and approaches that have been used in 2D-LC to couple the two dimensions of the system.15 One of the most commonly cited difficulties in implementing 2D-LC methods, both in the literature16−19 and among practitioners, is the negative

he limitations of modern one-dimensional liquid chromatography (1D-LC) are quite clear. Peak capacities on the order of 300 to 600 (depending on molecule type and conditions) that are achievable in about 1 h are insufficient for separating complex mixtures containing hundreds or thousands of different chemical species.1 Sometimes, it is even difficult to develop a single method for analysis of relatively simple mixtures of closely related compounds using a single column.2 After decades of relatively slow development, recent improvements in commercially available instrumentation and software for two-dimensional liquid chromatography (2D-LC) are enabling an increasingly broad group of users to develop innovative and highly effective methods to address challenging analytical problems in a variety of fields. Historically, 2D-LC has been a mainstay in the analysis of synthetic polymers3 and in proteomics.4 It is used for both targeted and untargeted analyses of natural materials ranging from foods and beverages5,6 to Traditional Chinese Medicines7 and industrial chemicals. Most recently, there has been tremendous growth in the use of 2D-LC for the analysis of biopharmaceutical materials, including active pharmaceutical ingredients © 2017 American Chemical Society

Received: May 29, 2017 Accepted: July 19, 2017 Published: July 19, 2017 9260

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Article

Analytical Chemistry influence that the 1D effluent can have on the performance of the 2D separation. This issue has been discussed in detail elsewhere.12 The problem is especially pronounced when the volume of each fraction transferred is increased to improve detection sensitivity at the outlet of the 2D column; in practice, it is not uncommon for the fraction volume to approach or even exceed the dead volume of the 2D column.20−23 If the 1D effluent has properties that result in low retention of analytes in the 2D column when the fraction is injected, broad and sometimes distorted peaks will be observed. For example, in the case where reversed-phase separations are used in both dimensions, the effluent sampled from the 1D separation would contain a high proportion of organic solvent that could cause low initial retention on the 2D column. As described by Egeness and co-workers,15 a number of strategies have been developed by several groups to overcome this problem. Among them are the use of trap columns or cartridges between dimensions for solvent exchange,24,21,22,25,26 partial evaporation of the solvent,27,28 dilution of the 1D effluent with weak solvent prior to injection into the 2D column,20,29,23,26 and modulation of the temperature of a trapping cartridge.30,25 Of course, when coupling two reversed-phase separations in a 2D-LC separation, it is generally advised to place the phase with less retention in the first dimension so that less organic solvent is required to elute compounds from the 1D column relative to the 2D column.31 Inevitably, however, this approach has limitations just like every other approach. For example, oftentimes the two phases with the most complementary selectivities for a particular set of analytes have similar retention, leading to a solvent mismatch situation. Also, if a weakly retentive phase is used in the first dimension, some analytes of interest may not be sufficiently retained and may be lost to the column dead volume. Therefore, a more general solution to this so-called “solvent mismatch” problem is highly desired. In this paper, we describe a new approach to address this problem, which we refer to as Active Solvent Modulation (ASM). The term ASM describes the approach, which is enabled by a modification to existing valve technology that allows dilution of the 1D effluent stream with weak solvent but without the need for an additional pump: (1) it is active in the sense that the dilution is only activated for a portion of each 2D separation cycle; (2) it operates on the sample solvent, i.e., the 1D effluent injected into the 2D column; (3) it modulates the properties of the solvent by dilution with 2D mobile phase. A detailed description of the characteristics of the valve and its operation in the context of 2D-LC experiments is given below (see Results and Discussion). After a thorough comparison of a related approach, we refer to it as Fixed Solvent Modulation (FSM), and we demonstrate the potential of ASM to improve 2 D separation performance using simple small molecule probes. We then go on to demonstrate the utility of the ASM approach in a representative application of Multiple-Heartcutting (MHC) 2D-LC aimed at discovering the degradants of bovine insulin produced upon exposure to accelerated aging conditions.

them. System B was equipped with the Multiple Heartcutting option (G4242A) and operated using Agilent OpenLAB CDS ChemStation Edition, version C.01.07. Acetophenone, butylparaben, and valerophenone were obtained from Sigma-Aldrich (St. Louis, MO) and used as received. Stock solutions were prepared at 10 mg/mL in acetonitrile (ACN). Analytical mixtures were prepared in 50/50 ACN/water with each of the three analytes at 10 μg/mL. The samples analyzed with system B (see below) were generated by exposing 1 mg/mL bovine insulin (Sigma-Aldrich, St Louis, MO) in 25 mM aqueous sodium phosphate buffered at pH 7 to 50 °C for 3 days. System A. System A was used for the separations of small molecule probes shown in Figure 4. Samples were infused directly into the interface valve (i.e., port 3 in Figure S1); thus, only the second dimension of the system was used. Second Dimension. The second dimension of the system involved the following instrument modules: A binary pump (G4220A) with the 35 μL JetWeaver mixer, column thermostat compartment (G7116B), and a diode array UV absorbance detector (DAD, G7117B) equipped with a low dispersion flow cell (G4212-60038; 10 mm path length; V(σ) = 0.6 μL). Chromatograms were recorded using a 160 Hz acquisition rate. The column was a Zorbax SB-C18 30 × 2.1 mm, 3.5 μm column (Agilent Technologies), and it was thermostated at 40 °C. Solvent A was water, solvent B was ACN, and the total flow rate was 2.5 mL/min. The gradient elution program was 2−2− 30−65−2−2%B from 0 to 0.3−0.31−0.56−0.57−0.75 min for the no dilution and 1:5 dilution cases and from 0 to 0.1−0.11− 0.36−0.37−0.50 min for the 1:1 dilution case. For the 1:1 dilution case, a 200 mm × 120 μm i.d. stainless steel capillary was used as the bypass connection in Figure 3, and a capillary of the same dimensions was connected in series with the sample loops L1 and L2. For the 1:5 dilution case, a 150 mm × 170 μm i.d. stainless steel capillary was used as the bypass, and the 200 mm × 120 μm i.d. capillary was again connected in series with the sample loops. In these situations, a large majority of the flow restriction through each path depends on these capillaries, and thus, their relative restrictions determine the flow split ratio through the sample loop and bypass paths. System B. System B was used for the separations of bovine insulin, and its degradants are shown in Figures 5, 6, and S2. First Dimension. The first dimension of the system involved the following instrument modules: A high speed (binary) pump (G7120A); vial sampler (G7129B); column thermostat compartment (G7116C); a diode array UV absorbance detector (DAD, G4212A) equipped with a standard flow cell (G4212-60008; 10 mm path length; V(σ) = 1.0 μL). Chromatograms were recorded using a 20 Hz acquisition rate. The column was a Poroshell HPH C18 150 × 2.1 mm, 2.7 μm column (Agilent Technologies), and it was thermostated at 40 °C. Solvent A was 0.1% TFA in water, and solvent B was 0.09%TFA in 80/20 ACN/Water (v/v). The gradient elution program was: 31−45−80−80−31%B from 0 to 28−29−30−31 min, and the flow rate was 300 μL/min. The 100 μL JetWeaver (mixer) was used in the pump. Fractions #1−6 (as highlighted in Figure 5) were taken as 120 μL fractions starting at 12.35, 13.6, 14.27, 14.87, 15.47, and 16.93 min, respectively. Second Dimension. The second dimension of the system involved the following instrument modules: A binary pump (G4220A) with the 35 μL JetWeaver mixer, column thermostat compartment (G7116B), and a diode array UV absorbance detector (DAD, G4212A) equipped with a 10 mm flow cell

■

EXPERIMENTAL SECTION Data were acquired using two similar but slightly different Agilent 1290 Infinity 2D-LC systems (Agilent Technologies); we refer to them here as System A and System B. The differences between systems primarily lie in variations between the main modules (e.g., pumping systems); we expect that the principal conclusions of this study apply equally well to both of 9261

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Article

Analytical Chemistry

useful to illustrate the value of this approach. Suppose a fraction of 1D effluent is collected for further separation in the second dimension, which contains 50% ACN. If the compounds of interest elute from the 2D column with much less ACN (e.g., 35% using a RP column), then terrible peak shapes will be observed, especially when large volumes of the solution containing 50% ACN are injected. With the FSM approach, the 2D pump can be set to a low percentage of ACN in the early part of each 2D cycle (e.g., 2% ACN) for the time period during which the 1D effluent fraction is being displaced from the sampling loop L2 in Figure 1. If the flow restrictions through the bypass connector and sample loop path are equal, the flows through these two paths will be equal, and the 1D effluent fraction will be mixed 1:1 with the mobile phase from the 2D pump. In the scenario described here, this would effectively modulate the ACN content of the 1D effluent fraction from 50% down to 26%. This results in much better peak shape in the 2D separation because the compound of interest is retained on the 2D column head under these conditions (i.e., the analyte band is focused) before it starts to migrate when the solvent gradient reaches a composition that promotes elution. This dilution of the 1D effluent fraction is accompanied by an increase in the volume of sample that must be injected into the 2D column (i.e., in this case 2-fold more), but it has been shown that the benefit of the decreased solvent strength, and thus a focusing effect that analytes experience, outweighs the cost of having to inject a larger volume.20,33 One significant limitation of the FSM approach is that the bypass connection is fixed, and flow will go through both the valve and the bypass connector throughout the analysis. Since the total flow from the 2D pump is split into two paths, the actual flow rate through the sample loop is lower than the total flow and dependent on the split ratio. This splitting of the two flow paths leads to a complex solvent profile propagating through the column during elution, which can lead to significant baseline disturbances; these are discussed below in reference to Figure 6. In the example described above, the flow rate through the loop would be 50% less than the total flow. This flow rate decreases as the split ratio is increased (e.g., by lowering the restriction of the bypass connector) to provide a greater degree of dilution of 1D effluent fractions with weak solvent. This can then result in a significant fraction of the 2D cycle time being wasted while waiting for strong solvent to be flushed from large sampling loops at low flow rates, particularly at the end of gradient elution programs. An example of this is illustrated with the experimental data shown in Figure 2. The black trace shows a solvent gradient programmed for use with FSM. The pregradient isocratic hold at 2% B enables dilution of the 1D effluent fraction as it is displaced from the sample loop. This period is then followed by a step up to the initial 30%B used in the gradient and then a normal linear gradient, followed by return to the initial condition at 19 s. The red trace shows the actual solvent composition delivered to the 2D column, as measured by a UV-absorbing tracer in the B solvent (i.e., without a column installed). The onset of the step from 2% to 30% B is delayed as expected due to the gradient delay volume associated with the pumping system and connecting tubing. The most instructive aspect of these data, however, is the return of the solvent composition from 65% to 2% B at the end of the cycle. We see that the composition drops rapidly from about 65% to about 15% and then more slowly from 15% to 2% B. We attribute the rapid drop to the displacement of the B solvent from the bypass connector and the slower drop that

(G4212-60008). Chromatograms were recorded using a 40 Hz acquisition rate. The column was a Poroshell HPH C18 50 × 2.1 mm, 2.7 μm column (Agilent Technologies), and it was thermostated at 40 °C. Solvent A was 0.1% formic acid (FA) in water, and solvent B was 0.09% FA in 80/20 ACN/Water (v/ v). The gradient elution program was: 6.25−6.25−26−35−90− 90−6.25%B from 0 to 1.0−1.1−3.1−3.2−3.3−3.4 min. The total 2D cycle time was 5.0 min, and the flow rate was 400 μL/ min. The MHC 2D-LC interface is shown in Figure S1. A detailed description of the operation of this interface was described previously.32 Briefly, two 14-port/6-position selector valves are connected to a 8-port/2-position interface valve. Each selector typically bears a cluster of six 40 μL sampling loops. This provides two parking decks (A and B) with a total of 12 loop positions. Switching the central interface valve places a deck either in the 1D flow path to capture 1D effluent or in the 2D flow path to inject fractions into the 2D column. Switching the selector valves provides access to the discrete loop positions. For the work described here, the commercially available interface was equipped with 120 μL loops (instead of 40 μL), and the selector valves were connected to a version of the central interface valve that was modified as shown in Figure 3 (see accompanying text for details). In this case (1:1 dilution), 170 mm × 120 μm i.d. stainless steel capillaries were used for the connections between the central ASM valve and the loop selector valves shown in Figure S1 (i.e., port 4 of the ASM valve to IN of Deck B, port 5 of the ASM valve to OUT of Deck B, and the analogous connections for Deck A), and a 340 mm × 120 μm i.d. stainless steel capillary was used as the bypass connection.

■

RESULTS AND DISCUSSION Approaches for Solvent Modulation: FSM and ASM. A valve-based solution for diluting 1D effluent fractions prior to injecting them into a 2D column for further separation was described previously by Gron29 and in a recent paper.23 A representative configuration used for this approach, which we refer to as Fixed Solvent Modulation (FSM), is shown in Figure 1. In this configuration, the implementation of any valve

Figure 1. Example configuration of the valve interface used for 2D-LC with Fixed Solvent Modulation (FSM). The flow restrictions through the sampling loops L1/L2 and the bypass connection determine the extent to which the fractions of 1D effluent are diluted with 2D eluent at point A prior to introduction into the 2D column.

normally used for LC × LC (e.g., 10-port/2-position or 8-port/ 2-position as is shown in Figure 1) is augmented with a “bypass connector” that establishes a fluidic connection between the 2D pump and 2D column in parallel with the path through the valve. These connections can be established using simple tee connectors at points A and B in Figure 1. A concrete example is 9262

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Article

Analytical Chemistry

Figure 2. Comparison of the programmed (black) and actual (red) solvent composition profiles measured at the inlet to the 2D column when using the FSM scheme. Because of the low flow rate through the sample loop itself when large dilution factors are used, a long flush-out period is needed at the end of the 2D cycle to remove strong solvent used in the gradient elution program. (A) This is a pregradient dilution period (9 s) with the 2D pump set at 2%B; (B) gradient time (9 s); (C) flush-out period for the bypass connector; (D) flush-out period for the sampling loop. Conditions: total 2D flow, 2.5 mL/min.; A solvent, 50/50 ACN/water; B solvent, A solvent with 10 ppm uracil added; solvent gradient, 30−65%B in 9 s; FSM dilution factor, 4 parts 2 D eluent: 1 part 1D effluent fraction; sampling loop volume, 40 μL.

Figure 3. Configuration of the interface valve used for Active Solvent Modulation (ASM). This valve design is a modification of the 8-port/2position design that has been discussed elsewhere.1 The ASM design has four positions instead of two. In positions 1 and 3, the bypass path is isolated. In positions 2 and 4, the bypass connects the 2D pump to the 2D column in parallel with the sampling loop. Positions 2 and 4 are used during displacement of 1D effluent fractions from the loops; positions 1 and 3 are used during elution from the 2D column and during re-equilibration of the column prior to the next injection. In positions 2 and 4, the bypass behaves the same as in FSM, and the extent to which 1D effluent fractions are diluted with 2D eluent depends on the ratio of restrictions through the sampling loop and bypass paths.

follows to the displacement of the B solvent from the sampling loop, which occurs at a low flow rate (i.e., split ratio: 1 part through loop and 4 parts through bypass). In this case, the slow flush-out of the loop adds about eight seconds to the 2D cycle time that are not required when FSM is not used. This additional time becomes more of a factor as 2D cycle times are shortened, split ratios are increased, and sampling loop volumes are increased. The limitation of FSM described above motivated us to develop a related concept, which we refer to as Active Solvent Modulation (ASM). For this approach, a variation of the 8-port/ 2-position valve is needed, which is illustrated in Figure 3. The following description of the valve function is applicable to the case of LC × LC where two nominally identical sample loops are used in an alternating fashion to transfer fractions of 1D effluent to the 2D column. The primary differences between this valve and the one shown in Figure 1 are (1) the addition of two ports that are fitted with a bypass connector; (2) the addition of two additional rotational positions (i.e., for a total of four). The role of the bypass connection and each of the four rotational positions of the valve are illustrated in the four panels of Figure 3. In Postion 1, the bypass is isolated such that the entire flow from the 2D pump goes through loop L1 to the 2D column. When the valve is in this position, solutes that were injected into the 2D column previously are eluted from the column, and the next fraction of 1D effluent to be injected into the 2D column is collected in loop L2. When the valve is switched to Position 2, the bypass becomes connected to the 2 D pump and column in parallel with the sampling loop. This is functionally similar to FSM, except that the connections between the bypass and 2D pump and column are made internally inside the valve body. The valve is held in Position 2 until the entire 1D effluent fraction is displaced from loop L2, at which point it is diluted with 2D mobile phase (weak in elution strength; e.g., mostly aqueous in the case of RP separations) prior to entering the 2D column. The reasons for doing this are

the same as those described above for FSM. As soon as the 1D effluent fraction is displaced from loop L2, the valve is then switched to Position 3 so that all of the flow from the 2D pump goes through loop L2, until the end of the 2D cycle (i.e., including elution and re-equilibration). At the same time, the next fraction of 1D effluent is collected in loop L1. Note that the bypass connector will contain solvent used during the ASM step when it is isolated as in Positions 1 and 3. At the end of this cycle, the valve is switched to Position 4, where the contents of loop L1 are displaced from the loop and diluted with 2D mobile phase (weak in ACN). When the contents of loop L1 have been completely displaced from the loop, the valve is switched to Position 1, and the cycle repeats as many times as are needed to complete the 2D-LC analysis. We have chosen to describe ASM here in the context of LC × LC for simplicity. The principles can be extended to multiple heartcutting (mLC-LC) and selective comprehensive (sLC × LC) 2D-LC by replacing the two single loops L1 and L2 shown in Figure 3 with sets of loops fixed to selector valves as shown in Figure S1. Demonstration of ASM Using Small Molecule Probes. Figure 4 shows results obtained from separations of a mixture of three small molecule probes to demonstrate the potential for ASM to improve compatibility between the first and second dimensions under well controlled conditions. The conditions used mimic those encountered in the second dimension of LC × LC separations involving RP separations in both dimensions. Only the second dimension of a 2D-LC system was used, and mixtures of the three probe molecules were infused directly into the valve shown in Figure 3 via the first dimension column path at a flow rate sufficient to overfill the loop during the sampling period. The 2D gradient was from 30% to 65% ACN over 15 s, 9263

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Article

Analytical Chemistry

Figure 4. Improvement in signal-to-noise ratio (S/N) and resolution for a set of small molecule test solutes using ASM. Conditions (second dimension): total flow, 2.5 mL/min.; A solvent, water; B solvent, ACN; gradient elution from 30% to 65−30−30% B from 0 to 15−16− 30 s; sample solvent, 50/50 ACN/water; sampling loop volume, 80 μL; solutes: 1, acetophenone; 2, butylparaben; 3, valerophenone. In the case of no ASM and ASM with 1:5 split, the pregradient dilution period was 18 s, and in the case of ASM with 1:1 split, 9 s. The time scale of the blue chromatogram has been shifted slightly so that the peaks overlay with the others for easier visual comparison.

following a pre-gradient dilution period at 2% ACN like that shown in Figure 2. The injected sample contained 50% ACN and 50% water, to mimic the type of difficult situation we often face in LC × LC separations involving RP columns in both dimensions. Figure 4 shows the results of 80 μL injections (i.e., 80 μL is the volume of the sampling loop itself) of this sample without ASM (black) and with ASM at either 1:1 (blue) or 1:5 (red) dilution of the sample with 2D mobile phase containing 2% ACN. These results show in dramatic fashion how ASM can be used to significantly improve both the detection sensitivity and resolving power of 2D separations in cases where the injected 1D effluent fraction contains more strong solvent than the 2D mobile phase. Whereas the peaks obtained without ASM are split, broad, and thus short, the peaks obtained with ASM are more symmetrical, narrower, and thus much taller. Increasing the split ratio from 1:1 to 1:5 provides increasing benefit as measured by these peak attributes. The peaks in the 1:5 case are slightly taller because they are slightly narrower (i.e., peak area is conserved). Application of ASM to Discovery of Degradants of Heat-Treated Bovine Insulin. Figure 5 shows a representative chromatogram obtained from the 1D separation of heattreated bovine insulin (see Experimental Section for preparation details) and the timing of fractions of 1D effluent that were transferred to the second dimension. Panel A shows the 1D separation at full scale, and Panel B shows an expanded view of the baseline that reveals several small peaks. A total of six fractions were transferred to the second dimension, as indicated by the blue highlighted regions. It is important to note that in this case a multiple heartcutting (MHC) interface was used as shown in Figure S1, equipped with 120 μL sampling loops. This interface allowed for storage of 1D fractions 2−6 in loops 1−5 of parking deck A while fraction #1 was injected into the 2 D column from loop 1 of deck B. The use of 120 μL loops enabled facile collection of 1D peaks in single fractions without concern for losses from the loop due to small shifts in 1D retention time. Such shifts can be especially problematic for peptides and proteins because of the sensitivity of their retention times to small changes in the organic solvent fraction of the mobile phase34 (log k ∝ 0.25(MW)0.5). The ability to

Figure 5. Representative chromatogram from 1D-separations of the degraded bovine insulin and indications of when fractions of 1D effluent were collected for subsequent transfer to the 2D column and further separation. (A) Full scale chromatogram showing the major peaks of interest; (B) zoomed view showing trace-level degradants. A total of six fractions of 1D effluent were collected as indicated by the light blue highlighted rectangles. For detailed chromatographic conditions, see the Experimental Section.

focus analytes at the inlet of the 2D column becomes particularly valuable under these circumstances because the sampling loop volume of 120 μL is larger than the dead volume of the 2D column itself (about 105 μL). Figure 6 shows representative chromatograms that were obtained from 2D separations of fractions #1, 2, 4, 5, and 6 shown in Figure 5, using three different transfer approaches. In several instances, we observe that the 2D separation resolves peaks that had coeluted in the 1D separation. This is most likely due to selectivity differences between the two dimensions arising from different gradient slopes34 and mobile phase modifiers,35 even though the same stationary phase was used in both dimensions. Panel A shows the results when the 8-port/2position interface is used normally, that is, without any dilution of the 1D effluent fractions. Panels B and C show the results obtained when using FSM or ASM, respectively. Restrictions for the two flow paths, through loop and through bypass, were matched such that the contents of each loop were diluted by a factor of 2 (i.e., 1:1), which in FSM occurred permanently and in ASM temporarily (i.e., until the fraction had been fully displaced from the loop). In the case of Panel A, we see that broad and distorted peaks are generally observed in each 2D chromatogram. We attribute this to the combination of the large injection volume of 120 μL for each 1D effluent fraction and the fact that each of these fractions contains more ACN than the mobile phase that peptides elute in from the 2D column.36−38 Taking into consideration gradient delay volumes of the pumps, dead times of the columns, and the delay of the 2D gradient program 9264

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Article

Analytical Chemistry

Figure 6. 2D chromatograms for fractions #1, 2, 4, 5, and 6 obtained from the separations shown in Figure 5 with 120 μL sampling loops. (A) Interface operated as unmodified 8-port/2-position valve, switching between positions 1 and 3 in Figure 3. (B) Interface operated in FSM mode, switching only between positions 2 and 4 in Figure 3. (C) Interface operated in ASM mode, using all four positions shown in Figure 3. Results in Panels B and C were obtained with a split ratio of 1:1 for through-loop and bypass flows. Blue arrows indicate baseline distortions observed only in the FSM case. For detailed chromatographic conditions, see Experimental Section.

are much better resolved in Panel B, and in fact, the first of the three peaks begins to show evidence of a fourth peak. Finally, in the case of fraction #4, which is arguably the most interesting as it was collected from the trailing edge of the main peak from the 1D separation, we see that the 2D separation reveals four additional peaks that had coeluted with the main peak in the first dimension. Panel C of Figure 3 shows the same 2D separations as in panels A and B but with the valve shown in Figure 3 operated in ASM mode. In this case, 55 s after beginning the displacement of a 1D effluent fraction from the sampling loops in the MHC interface, the valve was switched to isolate the bypass connection (i.e., Positions 1 or 3 of Figure 3) such that all of the flow from the 2D pump went through the loop for the duration of the 2D separation. In general, we observe that peaks in the chromatograms in Panels B and C look similar. This is the expected result, as the separations are greatly impacted by the dilution of the 1D effluent fractions with weak solvent, and this step is identical in the implementations of FSM and ASM used here. The important differences, however, are observed in the early and latter parts of these separations. The blue arrows in Panel B indicate the type of distortion in separations that can occur in the FSM mode. This was

during the dilution step, we estimate that all of the fractions from the 1D column contain between 29% and 30% ACN, whereas the peaks observed in Figure 6 elute in the range of 25% to 26.5% ACN. This differential, although it is slight, is critical because of the strong dependence of peptide retention on the fraction of ACN in the mobile phase.34 The broad peaks observed in Panel A are obviously undesirable for resolving peptides that coelute in the first dimension and for detecting low concentration species.20 Panel B of Figure 6 shows the same 2D separations as in Panel A but obtained with the valve shown in Figure 3 operated in FSM mode. In general, the peak shapes are much better with FSM; the narrower widths improve resolution and also translate to larger heights (i.e., improved S/N, since the noise levels are nominally the same). This is expected, since the ACN content of the 1D effluent fraction is reduced by about a factor of 2 (e.g., from 30% to about 18%) after diluting with the 5% ACN-containing 2D mobile phase used during the first 1 min of each 2D separation cycle. Several pronounced improvements are observed for the 2D separations of fractions #1, 2, and 4. In the case of fraction #1, only one 2D peak is observed in Panel A, whereas three peaks are easily observed in Panel B. In the case of fraction #2, the three poorly resolved peaks in Panel A 9265

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Analytical Chemistry

■

discussed conceptually above in reference to Figure 2. These distortions will obviously interfere with detection and quantitation of small peaks that might elute in those areas of the chromatogram. We consider the space in between the first and last blue arrows in the chromatograms of Panel B, to be the “useful elution window” because the baseline is relatively flat and enables reliable peak integration. In the case of ASM, the useful window spans the range from about 2.4 to 3.8 min, whereas for FSM, the useful window is between 2.8 and 3.5 min. In the case of ASM, after 1 min, the entire 2D flow goes through the sample loop, and we expect the baseline profiles to be the same as those observed in Panel A. Readers interested in the origins of the baseline distortions observed in the FSM mode will find a detailed discussion of this topic provided in Section S2 and Figure S2. We find that the additional baseline features observed in FSM mode have a big impact on the useful elution window; in this case, it is roughly 50% smaller in the case of FSM compared to ASM, and we see that the last major peak from fraction #6 elutes outside of this window in the case of FSM.

■

Article

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.7b02046. A detailed diagram of the sampling interface for multiple heartcutting, as well as a detailed analysis of the baseline distortions observed in FSM mode relative to ASM (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 507-933-0699. ORCID

Dwight R. Stoll: 0000-0002-4070-9132 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS All of the instrumentation and columns used for the parts of this work carried out in the Stoll Laboratory were provided by Agilent Technologies. Carston Dammann and Tyler Brau are acknowledged for their efforts to collect the data shown in Figure 4. Parts of this work have been supported by funding from the National Science Foundation (CHE1508159).

CONCLUSIONS

In this work, we have described a new valve-based approach to fraction transfer in online 2D-LC that enables effective focusing of analyte bands at the inlet of the 2D column without the need for any additional instrument hardware; we refer to this approach as Active Solvent Modulation (ASM). To demonstrate the performance potential of the ASM approach, we have first shown improvements in the peak shape and width for simple small molecule probes under conditions commonly encountered in 2D-LC. We then went on to demonstrate improvements in peak shape and width and detection sensitivity in a representative application of Multiple-Heartcutting 2D-LC for the discovery of degradants of heat-treated bovine insulin. On the basis of this work, we have arrived at the following principal conclusions: (1) The ASM approach offers advantages over the related Fixed Solvent Modulation (FSM) approach to analyte focusing at the 2D column inlet. Foremost among these are (A) more rapid flushing of high solvent strength eluent from the sampling loops at the end of each 2D separation cycle; (B) less complex mobile phase profiles delivered to the 2D column during gradient elution, which results in less baseline disturbances and a larger practical elution window in each 2D separation. (2) The ability to overcome the influence of the 1D effluent solvent on the 2D separation is very powerful. It allows us to consider the use of very large sampling volumes relative to the dead volume of the 2D column. This provides more flexibility in developing a method and provides a practical path to improve detection sensitivity at the outlet of the 2D column. (3) A practical limit to the ASM approach, which is shared by related approaches, is simply the time it takes to displace the fractions of effluent collected from the 1D column from the sampling loop into the 2D column. For 2D-LC separations with long 2D cycle times (e.g., in LC + LC39), this is not a problem, but for short 2D cycles, it can become a significant fraction of each 2D cycle. We believe the development of ASM represents a significant step forward in terms of improving the ease-of-use of 2D-LC and the overall detection sensitivity of the technique.

■

REFERENCES

(1) Stoll, D. R.; Carr, P. W. Anal. Chem. 2017, 89, 519−531. (2) Venkatramani, C. J.; Wigman, L.; Mistry, K.; Chetwyn, N. J. Sep. Sci. 2012, 35, 1748−1754. (3) Schoenmakers, P.; Aarnoutse, P. Anal. Chem. 2014, 86, 6172− 6179. (4) Wu, Q.; Yuan, H.; Zhang, L.; Zhang, Y. Anal. Chim. Acta 2012, 731, 1−10. (5) Cacciola, F.; Farnetti, S.; Dugo, P.; Marriott, P. J.; Mondello, L. J. Sep. Sci. 2017, 40, 7. (6) Donato, P.; Cacciola, F.; Tranchida, P. Q.; Dugo, P.; Mondello, L. Mass Spectrom. Rev. 2012, 31, 523−559. (7) Li, Z.; Chen, K.; Guo, M.; Tang, D. J. Sep. Sci. 2016, 39, 21−37. (8) Sorensen, M.; Harmes, D. C.; Stoll, D. R.; Staples, G. O.; Fekete, S.; Guillarme, D.; Beck, A. mAbs 2016, 8, 1224−1234. (9) Ouyang, Y.; Zeng, Y.; Rong, Y.; Song, Y.; Shi, L.; Chen, B.; Yang, X.; Xu, N.; Linhardt, R. J.; Zhang, Z. Anal. Chem. 2015, 87, 8957− 8963. (10) Li, Y.; Hewitt, D.; Lentz, Y.; Ji, J.; Zhang, T.; Zhang, K. Anal. Chem. 2014, 86, 5150−5157. (11) He, Y.; Friese, O. V.; Schlittler, M. R.; Wang, Q.; Yang, X.; Bass, L. A.; Jones, M. T. J. Chromatogr. A 2012, 1262, 122−129. (12) Stoll, D.; Carr, P. Anal. Chem. 2017, 89, 519. (13) Stoll, D.; Danforth, J.; Zhang, K.; Beck, A. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2016, 1032, 51−60. (14) Zhang, K.; Wang, J.; Tsang, M.; Wigman, L.; Chetwyn, N. Am. Pharm. Rev. 2013, 16, 39−44. (15) Egeness, M.; Breadmore, M.; Hilder, E.; Shellie, R. A. LCGC Eur. 2016, 5, 268−276. (16) Kivilompolo, M.; Pol, J.; Hyotylainen, T. LC-GC Eur. 2011, 24, 232 , 234, 236, 238, 240−243. (17) François, I.; Sandra, K.; Sandra, P. Anal. Chim. Acta 2009, 641, 14−31. (18) Wang, Y.; Lu, X.; Xu, G. J. Sep. Sci. 2008, 31, 1564−1572. (19) Mihailova, A.; Malerød, H.; Wilson, S. R.; Karaszewski, B.; Hauser, R.; Lundanes, E.; Greibrokk, T. J. Sep. Sci. 2008, 31, 459−467. (20) Stoll, D. R.; Talus, E. S.; Harmes, D. C.; Zhang, K. Anal. Bioanal. Chem. 2015, 407, 265−277. 9266

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267

Article

Analytical Chemistry (21) Gargano, A. F. G.; Duffin, M.; Navarro, P.; Schoenmakers, P. J. Anal. Chem. 2016, 88, 1785−1793. (22) Vonk, R. J.; Gargano, A. F. G.; Davydova, E.; Dekker, H. L.; Eeltink, S.; de Koning, L. J.; Schoenmakers, P. J. Anal. Chem. 2015, 87, 5387−5394. (23) Petersson, P.; Haselmann, K.; Buckenmaier, S. J. Chromatogr. A 2016, 1468, 95−101. (24) Venkatramani, C. J.; Al-Sayah, M.; Li, G.; Goel, M.; Girotti, J.; Zang, L.; Wigman, L.; Yehl, P.; Chetwyn, N. Talanta 2016, 148, 548− 555. (25) Sweeney, A.; Shalliker, R. J. Chromatogr. A 2002, 968, 41−52. (26) Oda, Y.; Asakawa, N.; Kajima, T.; Yoshida, Y.; Sato, T. J. Chromatogr. A 1991, 541, 411−418. (27) Ding, K.; Xu, Y.; Wang, H.; Duan, C.; Guan, Y. J. Chromatogr. A 2010, 1217, 5477−5483. (28) Tian, H.; Xu, J.; Xu, Y.; Guan, Y. J. Chromatogr. A 2006, 1137, 42−48. (29) Gron, O. The Modulator in Comprehensive Two-Dimensional Liquid Chromatography; CoSMoS, San Diego, CA, USA, 2015. (30) Verstraeten, M.; Pursch, M.; Eckerle, P.; Luong, J.; Desmet, G. Anal. Chem. 2011, 83, 7053−7060. (31) Stoll, D. R.; Li, X.; Wang, X.; Carr, P. W.; Porter, S. E. G.; Rutan, S. C. J. Chromatogr. A 2007, 1168, 3−43. (32) Pursch, M.; Buckenmaier, S. Anal. Chem. 2015, 87, 5310−5317. (33) Groskreutz, S. R.; Swenson, M. M.; Secor, L. B.; Stoll, D. R. J. Chromatogr. A 2012, 1228, 41−50. (34) Snyder, L. R.; Kirkland, J. J.; Dolan, J. W. In Introduction to modern liquid chromatography; Wiley: Hoboken, N.J, 2010; p 591. (35) Shibue, M.; Mant, C. T.; Hodges, R. S. J. Chromatogr. A 2005, 1080, 68−75. (36) Jeong, L. N.; Sajulga, R.; Forte, S. G.; Stoll, D. R.; Rutan, S. C. J. Chromatogr. A 2016, 1457, 41−49. (37) Snyder, L. R. J. Chromatogr. A 1964, 13, 415−434. (38) Groskreutz, S. R.; Weber, S. G. J. Chromatogr. A 2015, 1409, 116−124. (39) Stephan, S.; Jakob, C.; Hippler, J.; Schmitz, O. J. Anal. Bioanal. Chem. 2016, 408, 3751−3759.

9267

DOI: 10.1021/acs.analchem.7b02046 Anal. Chem. 2017, 89, 9260−9267