Journal of Chromatography A, 1339 (2014) 174–184

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Coupling state-of-the-art supercritical fluid chromatography and mass spectrometry: From hyphenation interface optimization to high-sensitivity analysis of pharmaceutical compounds Alexandre Grand-Guillaume Perrenoud, Jean-Luc Veuthey, Davy Guillarme ∗ School of Pharmaceutical Sciences, University of Geneva, University of Lausanne, Boulevard d’Yvoy 20, 1211 Geneva 4, Switzerland

a r t i c l e

i n f o

Article history: Received 25 November 2013 Received in revised form 26 February 2014 Accepted 1 March 2014 Available online 11 March 2014 Keywords: SFC–MS UHPSFC–MS Interfacing approach Detection sensitivity Pharmaceutical application

a b s t r a c t The recent market release of a new generation of supercritical fluid chromatography (SFC) instruments compatible with state-of-the-art columns packed with sub-2 ␮m particles (UHPSFC) has contributed to the reemergence of interest in this technology at the analytical scale. However, to ensure performance competitiveness of this technique with modern analytical standards, a robust hyphenation of UHPSFC to mass spectrometry (MS) is mandatory. UHPSFC–MS hyphenation interface should be able to manage the compressibility of the SFC mobile phase and to preserve as much as possible the chromatographic separation integrity. Although several interfaces can be envisioned, each will have noticeable effects on chromatographic fidelity, flexibility and user-friendliness. In the present study, various interface configurations were evaluated in terms of their impact on chromatographic efficiency and MS detection sensitivity. An interface including a splitter and a make-up solvent inlet was found to be the best compromise and exhibited good detection sensitivity while maintaining more than 75% of the chromatographic efficiency. This interface was also the most versatile in terms of applicable analytical conditions. In addition, an accurate model of the fluidics behavior of this interface was created for a better understanding of the influence of chromatographic settings on its mode of operation. In the second part, the most influential experimental factors affecting MS detection sensitivity were identified and optimized using a design-of-experiment approach. The application of low capillary voltage and high desolvation temperature and drying gas flow rate were required for optimal ESI ionization and nebulization processes. The detection sensitivity achieved using the maximized UHPSFC–ESI-MS/MS conditions for a mixture of basic pharmaceutical compounds showed 4- to 10-fold improvements in peak intensity compared to the best performance achieved by UHPLC–ESI-MS/MS with the same MS detector. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Supercritical fluid chromatography (SFC) is currently experiencing a remarkable rebirth of interest in the field of separation science. The recent reconsideration of this technique has been primarily triggered by the shortage of acetonitrile in 2008, which increased the operating costs of liquid chromatography (LC). In this context, SFC has been considered as a greener alternative [1,2]. There are also many other advantages offered by SFC that have been recently rediscovered by a wider audience. From a theoretical point of view, supercritical conditions allow significant increases in analytes diffusion (DM ), while the low mobile phase

∗ Corresponding author. Tel.: +41 22 379 34 63; fax: +41 22 379 68 08. E-mail address: [email protected] (D. Guillarme). http://dx.doi.org/10.1016/j.chroma.2014.03.006 0021-9673/© 2014 Elsevier B.V. All rights reserved.

viscosity maintains a reasonable column pressure [3]. Thus, both fast analysis at high linear velocity and enhanced chromatographic resolution with long columns can be easily achieved with SFC. Originally more dedicated to the analysis of lipophilic compounds, modern SFC has also been applied as a powerful technique for the analysis of molecules exhibiting a broad range of polarity including pharmaceuticals [4,5], natural products [6], ionic compounds [7], and even more recently peptides [8] thanks in part to the modification of the supercritical CO2 mobile phase with small amount of polar organic solvents [9]. Using such a binary mobile phase, the technique is still referenced as supercritical fluid chromatography for the sake of simplification even if, above a certain amount of organic modifier, the mobile phase is generally no longer rigorously under supercritical state but rather in a subcritical state. Nevertheless, no phase disruption between these two states is observed on the detected signal and the advantageous mobile phase properties

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

are conserved in the subcritical region. Practically, SFC offers the additional advantage that it can be easily used in reversed phase and normal phase without the need for any major change in terms of analytical conditions. Nowadays, a huge effort is consented in the development of polar stationary phase for SFC that can provides alternative selectivity compared to reversed phase LC (RPLC). The current interest for the technique is particularly marked at the analytical scale, since a new generation of SFC instruments adapted to the current analytical standards entered the market in 2011. Modern analytical SFC systems benefit from technological advances in the pumping system and backpressure regulation allowing better control of the compressibility of the mobile phase [10]. In addition, these new systems integrate the technical features of ultra-high performance LC (UHPLC), namely higher upper pressure limits and reduced void volumes [11], allowing better compatibility with the most recent stationary-phase technologies such as a sub-3 ␮m core–shell and fully porous sub-2 ␮m particles [12–15]. With the advent of state-of-the-art columns, the kinetic performance achieved with new SFC setups is comparable to those observed in UHPLC [16,17], suggesting the introduction of the term ultra-high performance SFC (UHPSFC) [6,16]. Due to its high sensitivity and selectivity, mass spectrometry is considered as one of the most powerful and versatile detection methods. LC–MS platforms are currently considered the gold standard for most analytical applications including investigations of complex mixtures and biological matrices [18,19]. The potential of SFC–MS has not been as thoroughly realized but is nonetheless promising and holds many advantages, particularly in terms of alternative selectivity and complementarity to LC–MS [20–22]. SFC–MS hyphenation was first described in the 1980s [23]. At this time, chemical ionization (CI) or electron impact (EI) ionization sources were employed in capillary SFC [24]. With the development of packed-column SFC and use of a non-negligible proportion of organic modifier in the mobile phase, SFC–MS quickly adopted the use of LC-like atmospheric pressure ionization (API) sources. Highly versatile and robust atmospheric pressure chemical ionization (APCI) [25] and electrospray ionization (ESI) [26] sources were both successfully used in the early 1990s and are still considered as first choice when using SFC–MS systems [27–29]. Due to the physical nature and compressibility of SFC mobile phase, moving column effluent toward the ionization source is less straightforward than in LC. Routing interfaces should limit the risk of analyte precipitation caused by SFC mobile phase loss of solvating power and density drop related to its decompression. Such effects can be disastrous for chromatographic fidelity, peak shapes and detector response [30]. Thus, dedicated interfaces including the control of backpressure (passive or active restrictor) must be employed. An extensive review focusing on the advantages and drawbacks of SFC–MS interfaces available in 2005 was published by Pinkston [30]. Most of these interfaces could still be mounted on the last generation of SFC instrumentations and commercial ones have been proposed. However, no study assessing their compatibility with UHPSFC–MS specifications, their flexibility in terms of operating conditions and their ability to maintain the high chromatographic performance have been published. The present study investigates the influence of different interface geometries and configurations with or without the use of a make-up fluid on the chromatographic performance and on the MS detection sensitivity. A complete characterization of the most promising interface was also conducted to evaluate the influence of chromatographic conditions such as flow rate, mobile phase composition and backpressure on the analyte quantity or concentration that reaches the API probe. Finally, the absolute detection sensitivity achieved using representative but maximized UHPSFC–MS/MS conditions were compared to that observed with separately fully optimized UHPLC–MS/MS approach for a mixture of basic drugs.

175

2. Experimental 2.1. Reagents and columns Pressurized liquid CO2 , 3.0 grade, (99.9%) was purchased from PanGas (Dagmerstellen, Switzerland). LC–MS grade solvents (methanol, ethanol, isopropanol, acetonitrile and heptane) were purchased from VWR (Radnor, PA, USA). Formic acid and ammonium hydroxide (ULC–MS grade) were purchased from Biosolve BV (Valkenswaard, The Netherlands). Water was obtained from a Milli-Q Water Purification System from Millipore (Bedford, MA, USA). Alprazolam, clonazepam, prazepam, triazolam and methadone were purchased from Lipomed AG (Arlesheim, Switzerland). Hydroxyzine, indapamide, noscapine, papaverine, theophylline and polyethylene glycol (PEG) were purchased from Sigma–Aldrich (Buchs, Switzerland). Acquity UPC2 BEH (100 mm × 3.0 mm, 1.7 ␮m) and Acquity UPLC BEH C18 columns (50 mm × 2.1 mm, 1.7 ␮m) for UHPSFC and UHPLC experiments, respectively, were purchased from Waters (Milford, MA, USA). Different column dimensions were chosen for both chromatographic techniques in order to limit the instrumental contribution to band broadening in both cases. 2.2. Instrumentation 2.2.1. UHPSFC system The Waters Acquity UPC2 system was equipped with a binary solvent delivery pump, an autosampler that included a 10 ␮L loop for partial loop injection, a column oven, a UV detector fitted with an 8 ␮L flow-cell and a two-steps (passive + active) backpressure regulator. The passive component maintains pressure higher than 104 bar while the active component allows further back pressure increase and fine backpressure adjustments. The connection tube between the injector and column inlet was 600 mm long (preheater included) and had an I.D. of 0.175 mm; the capillary located between the column and detector was 600 mm long and had an I.D. of 0.175 mm. On this instrument, the extra-column volume of the system was measured at 59 ␮L and the extra-column variance was measured at 85 ␮L2 , whereas the gradient delay volume was 440 ␮L. The hyphenation interface and splitter for UHPSFC–MS are detailed in Section 2.2.4. 2.2.2. UHPLC system The Waters Acquity UPLC was equipped with a binary solvent manager, an autosampler with a 2 ␮L loop operating in the fullloop injection mode, and a column oven. The connection tube between the injector and column inlet was 300 mm long (preheater included) and had an I.D. of 0.125 mm. The column outlet capillary was 250 mm long and had an I.D. of 0.125 mm and was directly connected to the ESI probe. On this instrument, the extra-column volume of the system was measured at 13 ␮L and the extra-column variance was measured at 8 ␮L2 , while the gradient delay volume was 100 ␮L. 2.2.3. MS/MS detector Both chromatographic systems were hyphenated with the same Waters TQD triple quadrupole mass spectrometer fitted with a Z-spray electrospray ionization (ESI) source. Ionization and MS detection were carried out in the ESI positive mode and with selected reaction monitoring (SRM), respectively. The source temperature, cone gas flow and source extractor voltage were identical in both UHPSFC and UHPLC modes (120 ◦ C, 20 L/h and +3 V, respectively). The capillary voltage, desolvation gas temperature and flow rate were optimized using a design-of-experiments (DoE) approach.

176

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

Fig. 1. Schematic representations of the three UHPSFC–ESI-MS/MS interfaces tested in this study. (A) Pre-BPR splitter with sheath pump interface and (B) Pre-UV and BPR splitter without sheath pump interface.

Cone voltages, collision energies and SRM transitions were optimized for each compound by direct infusion. The optimal values were not affected by the chromatographic mode and remained equal in UHPSFC and UHPLC. Finally, dwell times and inter-channel delays were set to 15 and 5 ms, respectively, to achieve a sufficient number of data points across the peaks [31]. 2.2.4. UHPSFC–MS splitter interface configurations The UHPSFC system was hyphenated with an MS detector using 2 different interface configurations illustrated in Fig. 1 and described as follows: (A) pre-UV-BPR-split: The UHPSFC column outlet tube was replaced by a 400 mm long, 0.175 mm I.D. capillary connected to a zero-dead-volume T-union allowing mobile phase splitting. The first part of the flow was directed toward the MS detector using a 750 mm long, 0.050 mm I.D. PEEK-sil transfer line, while the other part was directed toward the back-pressure regulator (BPR) through the UV detector using the original 600 mm long, 0.175 mm I.D. tube; and (B) pre-BPR-split + make-up pump: This interface kit purchased from Waters was composed of two serial zero-dead-volume T-unions connected to the UV detector or column outlet with 410 mm long, 0.175 mm I.D. tubing. CO2 miscible make-up liquid delivered by a Waters HPLC 515 make-up pump was added and mixed to the chromatographic effluent in the upstream T-union, while the downstream T-union acted as a flow splitter. A fraction of the total flow was directed from the downstream T-union to the ESI source through a 750 mm long, 0.050 mm I.D. PEEK-sil transfer line, while the remaining mobile phase was directed to the BPR via a 1270 mm long, 0.250 mm I.D. connection. 2.2.5. Software Instrument control, data acquisition and data handling of the UHPSFC–MS and UHPLC–MS systems were performed with Masslynx 4.1 (Waters). The physicochemical properties of the pharmaceutical compounds were predicted using ACD/ADME suite software (version 5, Advanced Chemistry Development, Inc., Toronto, ON, Canada, www.acdlabs.com, 2012). Calculations and modeling of splitter behavior were performed using MS Excel Software. Modde software (version 7.0.0.1, Umetrics, Umeå, Sweden) was used for DoE generation, statistical data processing and response-surface modeling. 2.3. Procedure and methodology 2.3.1. UHPSFC–MS interface evaluations The influence of the splitter interface on chromatographic and detection performance was assessed for the 2 configurations described in Section 2.2.4 using a mixture of 6 pharmaceutical compounds (100 ppb each) dissolved in a ternary injection solvent system consisting of EtOH:IPA:heptane, 1:2:7, v:v:v. The UHPSFC mobile phase was an isocratic mixture of 92/8 (v/v)

CO2 /MeOH + 20 mM NH4 OH delivered at 1 mL/min for the total flow introduction interface and at 2.0 mL/min for both the pre-UVBPR-split and pre-BPR-split + make-up pump interfaces. The 20 mM NH4 OH adjunction was required to achieve reasonable peak shape for basic compounds under SFC conditions [32]. Column temperature was set at 40 ◦ C and BPR at 120 bar for the two latter interface configurations. For the pre-BPR-split + make-up pump interface, pure MeOH was used as the sheath liquid at a fixed flow rate of 0.6 mL/min. The capillary voltage, desolvation gas temperature and flow rate were set at 1.0 kV, 400 ◦ C and 800 L/h, respectively. The most intense SRM transitions were monitored for each compound, e.g., prazepam (325 → 271), theophylline (181 → 124), noscapine (414 → 220), clonazepam (316 → 270), papaverine (340 → 202) and indapamide (366 → 132). Detection intensity was obtained from individual extracted ion current (XIC). The effects of additional splitter tubing on peak broadening, selectivity and generated pressure were further investigated for the pre-BPR-split + make-up pump interface using the same analytical conditions and compared to the chromatographic performance and behavior achieved for the UHPSFC–UV configuration. The same mixture was used when UV detection was performed at 210 nm, but the concentration was increased to 10 ppm. 2.3.2. Splitter behavior modeling The total amount of MeOH entering the ESI source, split ratio and sample dilution factor of the pre-BPR-split + make-up pump interface were modeled based on splitter tubing lengths and internal diameters. For this purpose, the Hagen–Poiseuille relationship was employed: F=

d4 P × 128 L

(1)

where F is the MeOH flow rate entering the MS, d is the capillary internal diameter,  is the fluid viscosity, P is the change in capillary pressure and L is the capillary length. The viscosity of the final fluid (homogenous phase composed of CO2 and MeOH coming from both pumps) was estimated based on experimental correlations proposed by Ouyang [33] and applied recently to SFC by Grand-Guillaume Perrenoud et al. [11]. To check the reliability of predictions, the MeOH flow rate delivered through the transfer capillary toward the ESI probe was collected for 3 min using a homemade MeOH trap. For this purpose, 81 experimental conditions including 3 UHPSFC flow rates (1.0, 2.0 and 3.0 mL/min), 3 UHPSFC mobile phase compositions (5, 10 and 20% of MeOH in CO2 (v/v)), 3 sheath pump flow rates (0.3, 0.6 and 0.9 mL/min) and 3 backpressure values (120, 150 and 180 bar) were tested. The trap employed for flow rate measurement consisted of a 50 mL Falcon tube containing 20 mL of PEG (Supplementary material Fig. 1). Two holes were made in the Falcon tube cap allowing both the admission of the chromatographic mobile phase and the discharge of decompressed CO2 . PEEK tubing that was 100 mm long and had a 0.250 mm I.D. was directly connected to the splitter PEEK-sil transfer line and used as a dip tube for mobile phase admission in the Falcon tube. The mobile phase was bubbled into the PEG, trapping the MeOH, while the exhaust decompressed CO2 came out of the Falcon tube through a stainless steel needle. Every single measurement was performed during 3 min. The whole device was weighed before and after each measurement to determine the exact amount of trapped MeOH. This amount was then converted into volume and then into flow rate using MeOH density at the controlled laboratory temperature of 21 ◦ C, data are shown in Supplementary material Table 1. 2.3.3. Maximizing MS detection sensitivity DoE experiments were carried out using UHPSFC–ESI-MS/MS (gradient mode 5–20% of MeOH + 20 mM NH4 OH in CO2 (v/v) in

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

Fig. 2. TIC chromatograms for the separation of 6 drugs obtained with the two UHPSFC–ESI-MS/MS interfaces. (A) Pre-UV and BPR splitter without sheath pump interface (green trace) and (B) Pre-BPR splitter with sheath pump interface (blue trace). Compounds: (1) Prazepam, (2) theophylline, (3) noscapine, (4) clonazepam, (5) papaverine, and (6) indapamide. (For interpretation of the references to color in text, the reader is referred to the web version of this article.)

3 min at 2 mL/min) in SRM detection mode. For the sake of comparison, the gradient steepness (product of gradient slope and column dead time) was identical for the two other investigated flow rates, i.e., 1.0 and 3.0 mL/min. A mixture of 6 drugs displaying different retention times was monitored using the most intense SRM transitions, e.g., hydroxyzine (375 → 201), alprazolam (309 → 281), noscapine (414 → 220), triazolam (343 → 308), papaverine (340 → 202) and methadone (310 → 265). The mixture was dissolved (100 ppb each) in a ternary injection solvent system consisting of EtOH:IPA:heptane, 1:2:7, v:v:v. DoE was also performed using UHPLC–ESI-MS/MS on the same mixture of 6 compounds dissolved at the same concentrations in pure water. The mobile phase consisted of ACN/H2 O both containing 0.1% formic acid (FA), and a gradient run from 15% to 55% ACN in H2 O (v/v) in 3 min at 0.6 mL/min was employed. The gradient steepness was identical for the two other investigated flow rates, i.e., 0.3 and 0.9 mL/min. For both techniques, the same ESI-MS instrument was used. 3. Results and discussion 3.1. Characterization of UHPSFC–MS interfaces 3.1.1. Evaluation of efficiency, sensitivity and flexibility Because of the specific nature of the supercritical mobile phase and its high compressibility, well-designed SFC–MS interfaces are required to avoid loss in chromatographic performance due to mobile phase decompression. In the first part of the study, 2 different interface configurations were compared based on the following features: preservation of chromatographic integrity and performance, detection sensitivity and user-friendliness. To facilitate the visualization, the total ion current (TIC) chromatograms are shown in Fig. 2. The chromatographic parameters and signal intensities of the 6 compounds were measured using each individual extracted ion current (XIC) and were summarized in Table 1. The first routing interfaces, namely “pre-UV-BPR-split interface” was considered mainly because its configuration limits as much as possible the extra-column band broadening since the splitter, consisting of a zero-dead-volume T-union, is placed prior to the UV detector flowcell. In addition, the splitter is placed under the direct control of the active BPR and should provide a relatively good flexibility as demonstrated in the past in several study [34,35]. The chromatogram obtained for the separation of the tested drug mixture with this first interface is reported in Fig. 2A. The efficiency was

177

measured for each individual compound, and the average N value was 8700 plates. A similar separation was obtained using the second interface, namely “pre-BPR-split + make-up pump interface” (Fig. 2B). Make-up addition of 0.6 mL/min MeOH was not found to be deleterious to the chromatographic integrity because the backpressure was still actively controlled by the BPR and since no perturbation related mixing effect was observed. In addition, no significant variations in system total pressure (less than 1 bar), retention time (less than 1%) or retention factor (less than 0.5%) were observed between the two configurations, meaning that additional tubing required for the BPR-split + make-up pump interface had a negligible effect on separation selectivity. Surprisingly, a slightly improved average N value of 9800 plates was obtained with this configuration. This result was counterintuitive because, in this second configuration, the extra-column volume related to the UVcell (8 ␮L) and to the connecting tubes should broaden the peaks detected in MS since these elements are placed in the same line as the MS detector whereas they are separated in the first interface. This increase in the total system dispersion should logically increase the band spreading and decrease the overall chromatographic efficiency of the second configuration compared to the first one. In the present case, the observed N improvement phenomenon second routing interfaces could be explained by the make-up fluid adjunction. Indeed, in both interfaces, the CO2 –MeOH mobile phase decompresses along the MS transfer line (the same dimensions for the two interfaces). Within this capillary that is no longer under the direct influence of the BPR, the fluid overall density and inherent eluting strength drops, leading to a decrease in the analyte’s speed. Thus, the analyte detection duration and its inherent peak width increase. Nevertheless, using the second interface, the MeOH composition of the mobile phase that enters the MS transfer line is higher due to the presence of additional make-up fluid. This higher proportion of MeOH tends to limit the overall density and eluting strength drops of the mobile phase along the transfer capillary. In this more “incompressible” mobile phase, the analyte speed will decrease to a lesser extent than in absence of make-up fluid, leading to shorter detection duration and a thinner peak width. In the present case, the efficiency of individual compounds is influenced by both higher system dispersion that tends to decrease N and lower density and eluting strength drops that significantly increases N. A significantly better chromatographic efficiency was observed using the pre-BPR-split + make-up pump interface for the 5 more retained compounds (k values between 2.1 and 9.1) which were positively impacted by the lower density and eluting strength drops phenomenon and showed an improvement of efficiency between 8% and 16%. Only the less-retained compound (Prazepam, k = 1.2), which should be the most affected by additional band broadening, displayed a 5% broader peak (0.020 vs. 0.019 min) with the pre-BPR-split + make-up pump interface compared to the pre-UV-BPR-split configuration. In contrast, the five remaining and more retained drugs. Detection intensity of each individual compound acquired in SRM mode, the differences between both splitting interfaces were noticeable. All 6 compounds displayed higher peak intensities in the presence of additional make-up fluid. The detection sensitivity was improved by a factor of 5–30 depending on the compounds using the pre-BPR-split + make-up pump interface compared to the pre-UV-BPR-split configuration, while the background noise remained equivalent. This increase in sensitivity was related to the additional MeOH coming from the sheath pump which acted as a supplementary proton source that enhanced the ionization process. 3.1.2. Chromatographic performance: UHPSFC–UV vs. UHPSFC–MS The pre-BPR-split + make-up pump configuration was found to be the best compromise in terms of chromatographic performance,

178

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

Table 1 Performance achieved in MS detection using 2 different UHPSFC–MS routing interfaces. Compound

Pre-UV-BPR-split (A)

Prazepam Theophylline Noscapine Clonazepam Papaverine Indapamide a

Pre-BPR-split + make-up pump (B)

ka

W50% a

Na

Intensitya

ka

W50% a

Na

Intensitya

1.20 2.16 2.46 3.51 4.43 9.62

0.019 0.024 0.024 0.027 0.027 0.059

4240 6230 7200 9390 13,810 11,100

4450 249 10300 1260 9450 1300

1.18 2.15 2.43 3.47 4.36 9.42

0.020 0.023 0.023 0.024 0.026 0.054

4160 6790 7910 12,200 14,970 12,850

71,600 8150 124,000 12,300 54,800 12,600

Values measured based on the MS/MS extracted ion current (XIC).

Fig. 3. Comparison of peak broadening between the UHPSFC–UV configuration (A., purple trace) and the UHPSFC–ESI-MS/MS configuration using the pre-BPR splitter with sheath pump interface and on-line UV detector (B., TIC, deep blue trace) and without on-line UV detector (C., TIC, light blue trace). Compounds: (1) prazepam, (2) theophylline, (3) noscapine, (4) clonazepam, (5) papaverine, and (6) indapamide. (For interpretation of the references to color in text, the reader is referred to the web version of this article.)

flexibility and above all detection sensitivity. However, since significant volumes are added to the system using this configuration and because of the mobile phase density and eluting strength drops phenomenon, its influence on chromatographic behavior and performance needed be further evaluated. For this purpose, the splitter was first completely removed from the instrument, and the UV detector outlet was connected directly to the BPR. The chromatogram obtained for the mixture of 6 pharmaceutical compounds using only UV detection is shown in Fig. 3A and the chromatographic performance values are summarized in Table 2. Within this configuration, an average N value of 13,200 plates was obtained. The same separation was performed with MS

detection including the pre-BPR-splitter + make-up pump. The splitter inlet was directly connected to the outlet port of the UV detector, increasing the total extra-column volume by 18 ␮L (up to 77 ␮L). Separation selectivity was maintained between the UV trace (Fig. 3A) and MS trace (Fig. 3B). However, the MS peaks displayed in average a 25% lower N value (9800 plates) than the peaks detected by UV. The first peak (prazepam) was logically the most affected one, but the compounds with longer retention times still displayed significant broadening. The 8 ␮L UV cell and 10 ␮L UV detector-BPR connecting tube were bypassed, and the splitter was directly connected to the column outlet. This third configuration displayed an equivalent extra-column volume than the original UV set-up in Fig. 3A (59 ␮L). The peak width was expected to be significantly improved using this configuration, but surprisingly, an average N value of 10,100 plates was measured (Fig. 3C). Only the first peak (prazepam) displayed a significant plate-count improvement in the absence of the UV flow-cell and connecting tubes. The overall plate count achieved for this third separation was comparable to that observes in presence of the UV cell and connecting tubes and presented a 23% lower N value than for the original UV separation. These results corroborate the observations made in Section 3.1.1, which noted the mobile phase decompression phenomenon as the major contributor for band spreading in the MS configuration. 3.2. Modeling of splitter interface behavior 3.2.1. Influence of chromatographic parameters Due to its design, the operation of the pre-BPR-split + make-up pump configuration is directly under the influence of chromatographic parameters. The simplest way to understand the principle of operation of this interface was to measure the amount of mobile phase per time unit that is actually directed toward the ESI probe. Because the CO2 is passing from a homogenous condensed phase to a gaseous state while running across the transfer capillary, the only measurable parameter is the remaining MeOH content that can be collected at the capillary outlet. A homemade MeOH trap preventing the potential evaporation of liquid MeOH was designed and directly connected to the transfer line outlet in place of the ESI probe. Eighty-one different chromatographic conditions were used to assess the influence of backpressure, flow rates (SFC and

Table 2 Loss in chromatographic performance between UV and MS detections. Compound

Prazepam Theophylline Noscapine Clonazepam Papaverine Indapamide a

UV (A)

Pre-BPR-split + make-up pump with UV (B)

Pre-BPR-split + make-up pump without UV (C)

k

W50%

N

ka

W50% a

Na

ka

W50% a

Na

1.14 2.09 2.37 3.41 4.29 9.35

0.015 0.016 0.019 0.022 0.024 0.047

6930 13,090 11,000 14,450 16,700 16,800

1.18 2.15 2.43 3.47 4.36 9.42

0.020 0.023 0.023 0.024 0.026 0.054

4160 6790 7910 12,200 14,970 12,850

1.18 2.15 2.45 3.50 4.40 9.42

0.017 0.023 0.022 0.024 0.026 0.054

5570 6740 8220 12,590 14,980 12,940

Values measured based on the MS/MS extracted ion current (XIC).

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

179

Fig. 4. Model of total MeOH amount entering the ESI probe using the pre-BPR splitter with sheath pump interface as a function of the UHPSFC mobile phase flow rate (x-axis) and sheath pump flow rate (y-axis) for two different UHPSFC mobile phase (CO2 /MeOH) compositions, 95/5 (v/v) for (A) surfaces and 80/20 (v/v) for (B) surfaces, and three different fixed backpressure values, 120 bar (A1 and B1), 150 bar (A2 and B2) and 180 bar (A3 and B3).

make-up pumps) and SFC mobile phase composition variations on the amount of MeOH collected per time unit. First, preliminary observations showed that an increase in the backpressure lead to a higher amount of collected MeOH. This effect is predictable because the higher resistance generated by the BPR obviously redirects a more important part of the mobile phase toward the transfer line. In the same way, an increase in the make-up pump flow rate or in the SFC mobile phase MeOH content evidently extended the measured amount of MeOH in the trap. In contrast, faster SFC mobile phase flow rates with constant MeOH composition led to a significant reduction in MeOH directed toward the transfer capillary. This counterintuitive effect is related to the fact that the active BPR leg of the splitter compensates pressure generated by the mobile phase to maintain a constant outlet backpressure. In the event of higher flows, the BPR is more open and offers a path of least resistance and hence the MS capillary receives less flow. Now, in contrast, when the mobile phase flow rate is low, the BPR must close in order to maintain system pressure, hence generating greater resistance along that leg and send more flow through the MS capillary.

Based on these preliminary observations and considering the tubing dimensions and Eq. (1), the MeOH flow rate entering the ESI probe was modeled over the whole range of chromatographic conditions using Excel. Some examples are shown in Fig. 4, including different mobile phase compositions (CO2 /MeOH: 95/5 (v/v) for (A) surfaces and 80/20 (v/v) for (B) surfaces) and various backpressures (120, 150 and 180 bar for surfaces 1, 2 and 3, respectively). As shown, the shape of these surface responses confirms the previous experimental observations. The validity of the model was then experimentally verified with 12 randomly selected sets of flow rates, backpressures and an extended set of mobile phase compositions (between 2% and 40% MeOH in CO2 (v/v)). Less than 10% difference was measured between the amount of MeOH physically collected in the trap and the value predicted by the computer simulation, which can therefore be considered as valid. In addition, it is worth mentioning that thanks to the dynamic adaptation of the split ratio between transfer capillary and the BPR depending on the chromatographic conditions, the total flow rate of MeOH entering the ESI probe tends to be leveled. The MeOH flow rate is always

Fig. 5. Model of the split ratio (A) and UHPSFC mobile phase dilution factor caused by the sheath flow (B) for the pre-BPR splitter with sheath pump interface as a function of UHPSFC mobile phase flow rate (x-axis) and sheath pump flow rate (y-axis) for a fixed UHPSFC mobile phase (CO2 /MeOH) composition of 90/10 (v/v) and 150 bar backpressure.

180

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

Table 3 Investigated levels of the variables involved in Design of Experiment (DoE) methodology. Level

−1 0 +1

UHPSFC

ESI

SFCF (mL/min)

BPR (bar)

SheaF (mL/min)

CapV (kV)

DesoT (◦ C)

DryF (L/h)

1.0 2.0 3.0

120 150 180

0.3 0.6 0.9

+1.0 +2.5 +4.0

200 325 450

300 650 1000

between 125 and 300 ␮L/min when typical operating conditions are applied in both isocratic and gradient modes (e.g., SFC flow rate between 1.0 and 3.0 mL/min, make-up pump flow rate between 0.3 and 0.6 mL/min, backpressure between 120 and 150 bar and mobile phase compositions between 2% and 40% MeOH in CO2 (v/v)). This MeOH flow rate range is well suited to preventing analyte precipitation inherent to CO2 decompression and to ensure both optimal ESI spray formation and good proton transfer during the ionization process. 3.2.2. Split ratio and band dilution factor The passive changes in the split ratio related to chromatographic conditions within the pre-BPR-split + make-up pump interface were further characterized by our computer simulation. Fig. 5A illustrates the amplitude of the split ratio for different SFC mobile phases and make-up flow rates at a fixed backpressure of 150 bar and a constant mobile phase composition of CO2 /MeOH: 90/10 (v/v). A split ratio between 2 and 8 between the mobile phase directed toward the BPR and the MS is expected depending on the flow rate settings. Lower backpressure value or higher MeOH content tend to further direct the mobile phase in the direction of the BPR (split ratio up to 12) due to lower resistance from the BPR or higher generated pressure within the transfer line, respectively. Regarding MS detection sensitivity, a high split ratio toward the BPR could constitute a significant concern in the case of a mass-flow sensitive ionization technique such as APCI. Because only a limited amount of analyte would be directed to the MS probe, the loss of sensitivity expected would be proportional to the split ratio but would be less than 1 order of magnitude. In contrast, the detection sensitivity achievable with a concentration-dependent ionization technique such as ESI would not be affected by the split ratio because flow division has no effect on the analyte concentration in the mobile phase. Nevertheless, detection sensitivity in ESI could be affected by make-up fluid adjunction into the mobile phase. The additional amount of MeOH from the sheath pump introduces a dilution factor that reduces the analyte concentration in the mobile phase. This dilution factor was plotted for different flow rate conditions in Fig. 5B and was found to be between 1.1 and 3.5. However, the latter high value is only observed in “exotic” conditions where a high sheath pump flow rate (1.0 mL/min) is added to a low mobile phase flow rate (0.4 mL/min). Under more typical conditions, the dilution factor was much more reasonable and varied between 1.1 and 1.5. Finally, it is worth mentioning that using a constant makeup flow rate when performing a separation in gradient mode will cause the later eluting peaks to be more diluted than less retained peaks.

daily use. The impact of chromatographic and ionization parameters on sensitivity was assessed in UHPSFC–ESI-MS/MS, using the pre-BPR-split + make-up pump interface following a 2-step designof-experiment (DoE) methodology. In parallel, the optimization of analytical parameters was also performed on the same ESI-MS/MS platform hyphenated to the UHPLC system. The constraints related to latter instrument have been taken into account in selecting the dimensions of the chromatographic column (2.1 mm × 50 mm, 1.7 ␮m) and representative mobile phase conditions for RPLC analysis for pharmaceutical molecules were preferred. Finally, the absolute detection sensitivity performance obtained on the two individually optimized analytical platforms were compared. 3.3.1. Screening for the most important operating factors in UHPSFC–ESI-MS/MS First, the critical experimental variables were selected for the ESI ionization source (i.e., capillary voltage (CapV), desolvation temperature (DesoT), drying gas flow rate (DryF)) and UHPSFC (i.e., mobile phase flow rate (SFCF), backpressure value (BPR) and sheath pump flow rate (SheaF)). The investigated range for each variable is summarized in Table 3, and lower/higher levels were chosen to cover a sufficiently wide but rational range of operating conditions. The peak intensities of the 6 basic model compounds were selected as the analytical response. A half-fractional factorial design (HFFD) was selected, and the central conditions were repeated 6 times to estimate the experimental error. A total of 22 runs (16 for HFFD and 6 central points) were randomly carried out. The model coefficients for the 6 parameters, including 95% confidence intervals, are presented in Fig. 6. The coefficient size and direction (toward a positive or negative value) represent the effect on detection sensitivity when a single operation parameter varies from 0 to 1, while the 5 remaining factors are maintained at their average values. As expected, the SFCF and BPR parameters did not explain ESI-MS sensitivity (confidence interval crosses zero). These parameters do not affect the concentration of the analytes in the mobile phase or the ESI spray stability because our splitter

3.3. Maximizing MS detection sensitivity This section focuses on the influence of various SFC and ESI source parameters on MS detection sensitivity. To be routinely applicable, these parameters and their ranges of investigation were chosen taking into account additional constraints related to conditions of UHPSFC–ESI-MS/MS for the analysis of pharmaceutical compounds. Thus, the nature of the mobile phase (MeOH) and column dimensions (3.0 mm × 100 mm, 1.7 ␮m) were fixed beforehand in order to be as representative as possible of UHPSFC–MS

Fig. 6. Detection sensitivity optimization. Screening the most important operating parameters in UHPSFC–ESI-MS/MS and their influence on detection sensitivity of basic drugs: noscapine (deep blue), papaverine (green), hydroxyzine (gray), alprazolam (purple), triazolam (light blue) and methadone (red). Parameters: capillary voltage (CapV), desolvation temperature (DesoT), drying gas flow rate (DryF), sheath pump flow rate (SheaF), UHPSFC mobile phase flow rate (SFCF), backpressure (BPR) and UHPLC mobile phase flow rate (LCF). (For interpretation of the references to color in text, the reader is referred to the web version of this article.)

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

181

Fig. 7. Ionization source conditions optimization. Detection sensitivity modeling for Alprazolam using the UHPSFC–ESI-MS/MS (A) and UHPLC–ESI-MS/MS (B) configurations. Drying gas flow rate was fixed at 1000 L/h in these representations.

tends to ensure regular and sufficient mobile phase flow toward the ESI source. Regarding the SheaF, a small but significant decreased intensity effect was highlighted when the make-up pump delivered a higher flow rate. This was not surprising because higher MeOH content tends to dilute the analyte concentration in the mobile phase. Thus, a low SheaF should be selected to limit the loss of signal intensity. Finally, ESI variables were shown to play the most significant roles for detection sensitivity, with a positive influence of high DesoT and DryF and a negative influence of CapV, and were therefore selected for further investigations. 3.3.2. Sensitivity modeling and optimization To optimize the values for the three most influent parameters on detection sensitivity (i.e., CapV, DesoT and DryF), a face-centered central composite design (CCD) was selected. The explored ranges for the three ESI parameters remained identical to those reported in Table 3, while the chromatographic parameters for the mobile phase flow rate and the BPR were held constant at 2.0 mL/min and 120 bar, respectively. The make-up pump flow rate, which also affects the response, was set at 0.3 mL/min to limit solvent waste and attain reasonable sensitivity. Experimental error was again estimated using 6 additional trials under average conditions, and a simple statistical study showed acceptable repeatability. The model validity was judged sufficient because the determination coefficients (R2 ) of the response surface models were between 82% and 91% after removal of one outlier trial. The models displayed different intensity amplitudes depending on the investigated compound but exhibited the same response surface shape regardless of the analyte. As an example, Fig. 7A illustrates the response surface model obtained for alprazolam with DryF fixed at its optimal value, i.e., 1000 L/h. The curved shape clearly reflects an interaction effect between parameters. For all compounds, the best sensitivity response was achieved at the lowest CapV value of 1.0 kV and the highest investigated DryF and DesoT values, i.e., 1000 L/h and 450 ◦ C, respectively. Investigations of more extreme values (toward the lowest CapV (1000 L/h) and DesoT (>450 ◦ C)) could not be tested due to instrumental limitations or instability. For a fair comparison of detection sensitivity, the experimental conditions were also optimized to achieve the highest possible sensitivity in the UHPLC–ESI-MS/MS configuration using the same MS detector. Taking into account the UHPLC column dimensions, separations were performed at 0.6 mL/min to achieve fast analysis close to the optimal linear velocity. The effect of ionization parameters (CapV, DryF and DesoT) on detection sensitivity was determined

using a methodology similar to that employed in UHPSFC–ESIMS/MS. Fig. 7B shows the response surface model for detection sensitivity achieved in UHPLC–ESI-MS/MS for alprazolam (DryF fixed at its optimal value, i.e., 1000 L/h). Interestingly, the response surfaces obtained in UHPLC–ESI-MS/MS conditions were very close to those achieved in UHPSFC–ESI-MS/MS, only slight changes in the shape curvatures were observed. Moreover, the optimal conditions for maximizing sensitivity were rigorously identical between both techniques. Initially, this observation is surprising since the composition of mobile phase is very different between UHPSFC and UHPLC (i.e., CO2 /MeOH vs. H2 O/ACN, respectively). However, the fact that CO2 decompresses in the transfer capillary under the UHPSFC conditions must be kept in mind. Then, its solvating power is severely decreased, and analytes are certainly only solubilized in liquid MeOH when they reach the ESI source before being ionized. These conditions are close to those encountered in liquid chromatography conditions, when molecules are dissolved in a hydro-organic liquid mixture. The requirements for nebulization processes (drying gas and desolvation temperature) must therefore be similar. For the latter parameter, the need for a high temperature to evaporate the aqueous part of the hydro-organic mobile phase in UHPLC is obvious, but it might be more surprising for the UHPSFC mobile phase. Most likely, a high DesoT is required to compensate for the low MeOH temperature inherent to CO2 decompression, which is a strongly endothermic process. 3.3.3. Sensitivity comparison The most notable difference between UHPLC and UHPSFC is obviously the solvent composition of the mobile phases (MeOH/CO2 vs. hydro-organic mixture). Although this composition has no influence on the optimal ESI settings, as shown previously, it is highly probable that it impacts the efficiency of ionization and especially the nebulization processes that further govern detection sensitivity. This phenomenon can be illustrated within the same chromatographic technique when comparing the peak intensity achieved with the best ESI conditions to that obtained when the worst ionization settings determined with our model are applied. Fig. 8 shows the chromatograms obtained with the mixture of 6 basic drugs achieved by UHPSFC (traces A) and UHPLC (traces B). The top traces (red; A1 and B1) were obtained using the best ESI conditions, i.e., 1.0 kV, 1000 L/h and 450 ◦ C for CapV, DryF and DesoT, respectively; middle traces (green; A2 and B2) were observed using average source conditions, i.e., 2.5 kV, 650 L/h and 325 ◦ C for CapV, DryF and DesoT, respectively; and bottom traces (blue; A3 and B3) were achieved using the worst ESI settings i.e.,

182

A. Grand-Guillaume Perrenoud et al. / J. Chromatogr. A 1339 (2014) 174–184

Fig. 8. Detection sensitivity (TIC) changes as a function of ionization parameter variations for UHPSFC–ESI-MS/MS (A: traces) and UHPLC–ESI-MS/MS (B: traces) configurations using a mixture of basic drugs ionized at optimal source conditions (A1 and B1 (red traces)), at average source conditions (A2 and B2 (green traces)) and at the worst source conditions (A3 and B3 (blue traces)). Compounds: (1) noscapine, (2) papaverine, (3) hydroxyzine, (4) alprazolam, (5) triazolam, and (6) methadone. (For interpretation of the references to color in text, the reader is referred to the web version of this article.)

4.0 kV, 300 L/h and 250 ◦ C for CapV, DryF and DesoT, respectively. It is worth mentioning that the SRM baseline noise remained very low (