Online NMR and HPLC as a Reaction Monitoring Platform for

Sep 4, 2013 - David A. Foley,* Jian Wang, Brent Maranzano, Mark T. Zell, Brian L. Marquez, Yanqiao Xiang, and George L. Reid. Analytical Research and ...
0 downloads 0 Views 2MB Size
Letter pubs.acs.org/ac

Online NMR and HPLC as a Reaction Monitoring Platform for Pharmaceutical Process Development David A. Foley,* Jian Wang, Brent Maranzano, Mark T. Zell, Brian L. Marquez, Yanqiao Xiang, and George L. Reid Analytical Research and Development, Pfizer Worldwide Research and Development, Eastern Point Road, Groton, Connecticut 06340, United States ABSTRACT: Detector response is not always equivalent between detectors or instrument types. Factors that impact detector response include molecular structure and detection wavelength. In liquid chromatography (LC), ultraviolet (UV) is often the primary detector; however, without determination of UV response factors for each analyte, chromatographic results are reported on an area percent rather than a weight percent. In extreme cases, response factors can differ by several orders of magnitude for structurally dissimilar compounds, making the uncalibrated data useless for quantitative applications. While impurity reference standards are normally used to calculate UV relative response factors (RRFs), reference standards of reaction mixture components are typically not available during route scouting or in the early stages of process development. Here, we describe an approach to establish RRFs from a single experiment using both online nuclear magnetic resonance (NMR) and LC. NMR is used as a mass detector from which a UV response factor can be determined to correct the high performance liquid chromatography (HPLC) data. Online reaction monitoring using simultaneous NMR and HPLC provides a platform to expedite the development and understanding of pharmaceutical reaction processes. Ultimately, the knowledge provided by a structurally information rich technique such as NMR can be correlated with more prevalent and mobile instrumentation [e.g., LC, mid-infrared spectrometers (MIR)] for additional routine process understanding and optimization.

T

and the reaction mixture is quenched and analyzed, but the use of online HPLC systems is becoming increasingly popular.6 Online systems, including those coupled to LC instruments have significant advantages, including minimizing operator error and exposure during sample preparation and providing close to real time data, which is critical for reaction monitoring. LC analytes are routinely detected and quantitated using UV detection, which exploits the molar absorptivity of a compound to calculate its concentration. The ultraviolet absorbance spectra and associated detector response vary on the basis of structure, chromatographic conditions, and wavelength, even for structurally related substances.7 Detector response for quantitative measurements is an extremely important consideration when dealing with multicomponent reaction mixtures. It is of critical importance to account for this response difference, which is corrected by establishing UV response factors for the species being analyzed by HPLC. A number of methods for establishing the relative response factor are available, including chemiluminescent nitrogen-specific detector (CLND)8 and charged aerosol detectors (CAD).9 Limitations are associated with these detectors: CLND is specific for

he use of chromatography, including gas chromatography(GC)1 and liquid chromatography (LC)2 [high performance LC (HPLC) or ultra high performance LC (UPLC)], shows strong precedent as process analytical chemistry (PAC)3 tools in process development. With the use of well developed and appropriate analysis methods, the chromatographic conditions provide analyte specificity with time (i.e., each component to be measured in the sample is presented to the instrument detector at a different and characteristic retention time), and the components are subsequently detected. Many detectors are available for both gas4 and liquid5 chromatography and can impart additional selectivity or sensitivity for the compounds of interest, e.g., mass selective detector (MSD), nitrogen phosphorus detector (NPD), or ultraviolet (UV). Broadly applicable detectors are thermal conductivity detector (TCD) and refractive index (RI). Chromatography combines specificity and detector dynamic range which is linear over several orders of concentration, allowing both major and minor components to be analyzed in the same run. LC is the dominant analytical tool employed in in-process control (IPC) analysis of reaction mixtures in the pharmaceutical industry, due to the strengths listed above, plus applicability to the analytes generally tested. As a result, LC in IPC benefits from an abundance of equipment and a workforce skilled in the operation of these tools. The majority of LC analyses are conducted in an offline manner, where the reactor is sampled © XXXX American Chemical Society

Received: July 30, 2013 Accepted: September 4, 2013

A

dx.doi.org/10.1021/ac402382d | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Letter

Figure 1. Online NMR and HPLC instrumentation setup.

we present details of a combined online NMR and HPLC platform which enables simultaneous reaction monitoring by both analytical techniques, where the NMR data is used to calibrate the LC-UV instrument for quantitative weight percent results.

nitrogen containing compounds, and the commonly used organic solvent acetonitrile cannot be used in the separation method, while CAD is only suitable for compounds that carry a charge under electrospray type ionization conditions and both detection methods require the use of volatile solvent/buffer systems (e.g., MSD compatible mobile phases). The definitive approach to quantify analytes on the basis of weight percent is to employ a reference standard to determine the relative response factor (RRF) for each analyte in a given detector and set of conditions. To generate and characterize reference materials is resource-intensive if the goal is to account for the actual UV detector response of each analyte. At the early stages of active pharmaceutical ingredient (API) process development when the synthetic route, starting materials, and impurities are not set, it is often not feasible to invest such extensive resources into establishing RRFs for each component in the synthetic route. The use of in situ process analytical technology (PAT) is routinely performed in the pharmaceutical industry.10 A number of PAT tools are commercially available to monitor reactant consumption and product formation in organic reactions. Mid-infrared (MIR), near-infrared (NIR), UV, and Raman spectroscopies are regularly used in pharmaceutical process development and manufacturing to monitor and to potentially control reaction progression. While each of these techniques has the ability to track individual components in the reaction mixture, calibration is required to provide quantitative data. Nuclear magnetic resonance (NMR) spectroscopy is an inherently quantitative technique; the number of nuclei is proportional to the area of the peak that is assigned to a particular group in the molecule, obviating the need for mass percent calibration. LC-UV RRFs determined by NMR have been shown to not significantly differ from those established by traditional reference standard determinations.11 In this work,



EXPERIMENTAL SECTION

4-Fluorobenzaldehyde (0.60 mL, 5.59 mmol) was dissolved in 85 mL of methanol−acetonitrile (1:1 mixture) in a 100 mL reaction vessel. Aniline (0.51 mL, 5.59 mmol) was added in one portion. The reaction was conducted at 25 °C, which was replicated throughout the sample loop and probe region of the NMR. The equipment setup is outlined in Figure 1. A continuous stream of the reaction mixture was circulated via 1/16 in. o.d. PTFE tubing from the reaction vessel to the sample loop at 10 mL/min. The flow was split using a needle valve located close to the spectrometer, which diverted some of the liquid stream at a slower rate (3−4 mL/min) into a removable NMR flow cell, inserted in the NMR probe region of the magnet. This slower flow rate was necessary to ensure sufficient residence time in the cell for accurate quantitation of the NMR signals. The flow cell was designed to be used with any regular NMR magnet and probe.12 Once detected, the reaction solution was returned through the transfer tubing to the reaction vessel. HPLC. Online HPLC analysis was performed using an Agilent 1200 HPLC system (Agilent Technologies, Palo Alto, CA). Sampling for the HPLC was conducted via a Valco switching valve (Valco Instrument Co., Houston, TX) on the return line, where ∼100 μL was removed and transferred to a FIA Lab 3200 sampling system (FIA lab instrument Inc., Bellevue, WA) for dilution and mixing. The diluted sample was then transferred to the injection loop for HPLC analysis. B

dx.doi.org/10.1021/ac402382d | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Letter

Figure 2. Stacked plot of 19F NMR spectra. 4-Fluorobenzaldehyde was monitored using the resonance at δF −105.1 ppm and imine at δF −110.1 ppm. The X-axis is chemical shift in ppm, and the Y-axis is the NMR spectrum number, acquired at 5 min intervals for the first 2 h and at 15 min intervals for the remainder of the reaction.

Scheme 1. Reaction of Aniline and 4-Fluorobenzaldehyde

without the need for isolation of those species. The use of online NMR as a reaction monitoring tool is increasingly common, and its quantitative nature has been established.13 The advantages of the technique described here is 2-fold; UV RFFs can be calculated for multiple species in a mixture without generating reference standards for those individual components; second, this facilitates the use of online HPLC as an accurate reaction monitoring technique, for further optimization of a process. In order to demonstrate of the utility of this technique, the reaction of 4-fluorobenzaldehyde with aniline was monitored simultaneously by online NMR and HPLC. The reaction was conducted in a 1:1 solvent mixture of acetonitrile−methanol at 25 °C, and a constantly flowing stream of the reaction mixture was circulated around the sample loop. 19F NMR spectra were recorded at regular intervals over the course of the reaction, and characteristic resonances for the aldehyde starting material and imine product were identified. The signal at δF −105.1 ppm was used to track the progress of 4-fluorobenzaldehyde, while imine formation was monitored by the appearance of a signal at −110.1 ppm (Figure 2). Samples were taken for HPLC by extracting an aliquot of the flowing reaction mixture into a second sample loop before injection onto the HPLC column for analysis. The reaction did not go to completion due to the reverse reaction of the imine with water generated in the

Chromatograms were recorded at 18 min intervals over the course of the reaction. NMR. Online NMR reaction monitoring was conducted by circulating a flowing stream of the reaction mixture from the reaction to a modified flow cell in the coil region of the NMR. Spectra were recorded using a Varian 400 MHz NMR with Auto Switchable probe. 19F NMR spectra were acquired unlocked, initially at 5 min intervals for the first 2 h and at 15 min intervals for the remainder of the reaction. Each spectrum had a total acquisition time of 123 s, which included 4 scans with a 30° flip angle and a relaxation delay of 30 s between each pulse, to ensure complete relaxation. The two sets of data were synchronized using the time stamps which are associated with the acquisition parameters of the NMR data and the sampling time of the LC sample loop.



RESULTS AND DISCUSSION NMR and HPLC working in tandem to monitor a reaction can provide an opportunity to develop a symbiotic relationship between the structurally information rich, inherently quantitative method of NMR and the sensitivity and selectivity of a chromatographic method. An example is outlined which combines online NMR and HPLC to analyze, develop, and optimize organic chemistry reactions. Ultimately, employing this method, the UV relative response factor for components in the reaction mixture can be calculated from a single experiment, C

dx.doi.org/10.1021/ac402382d | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Letter

RRF (Aldehyde) =

RRF (Imine) =

Aldehyde (mol%NMR) = 1.24 Aldehyde (area%HPLC)

Imine (mol%NMR) = 0.86 Imine (area%HPLC)

The RFF value for the aldehyde starting material was calculated using the integral area of the NMR resonance at −105.1 ppm, which was corrected for the number of nuclei represented by the resonance of interest (in this case, one). The integral area is then converted to mol % by dividing by the total integral of aldehyde and imine resonances in the 19F NMR spectrum. The HPLC area % is obtained in a similar way using the area under the peak in the UV chromatogram. Then, a ratio is calculated at a number of time points over the course of the reaction using the equations above, and the average of these values gives the RRF. The combination of online NMR and HPLC analysis allowed the RRF to be established for the aldehyde and imine and thus corrected the HPLC UV responses without the need for reference standards of each component. Furthermore, when dealing with the dynamic system of a constantly changing organic reaction mixture, reference standards may not be easily obtained, as labile species may degrade upon isolation. This technique could also be applied using other NMR active nuclei, in particular 1H NMR, where solvent suppression techniques could be employed to minimize signals due to solvent resonances. While this will impact the quantitative nature of signals close to the region of solvent suppression, 1H NMR spectra of reaction components typically possess more than one characteristic resonance which can be tracked for reaction monitoring purposes.

Figure 3. HPLC monitoring of the reaction mixture. LC conditions: column, dinitrophenyl (DNP), 4.6 × 250 mm, 5 μm at ambient temperature; mobile phase, 90:10 (heptane−ethanol); flow rate, 1 mL/min; UV, 254 nm.

reaction, resulting in the equilibrium process shown in Scheme 1. The 19F NMR resonances for 4-fluorobenzaldehyde and imine product were integrated to obtain the reaction profile. The corresponding HPLC (UV area percent) data was also processed. A representative chromatogram and chromatographic conditions are shown in Figure 3, and the two sets of data were plotted on the basis of time (NMR data represented by squares, HPLC by lines), as shown in Figure 4. An initial overlay of the two sets of data clearly shows a discrepancy in the results generated from the techniques. Analysis of the HPLC results suggests a UV under response for 4fluorobenzaldehyde and an apparent over response for the imine product, when compared with the corresponding NMR results. Using the results obtained from NMR monitoring as a molar detector, a RRF for both components was established using the following equations:

Figure 4. Application of NMR established relative response factor (RRF) to HPLC data. D

dx.doi.org/10.1021/ac402382d | Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry



Letter

AUTHOR INFORMATION

Corresponding Author

*E-mail: davidanthony.foley@pfizer.com. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors would like to thank Bruker for the collaboration to develop an NMR flow cell and Mestrelab Research for NMR reaction monitoring software.



REFERENCES

(1) Luijten, W. Analusis 1998, 26 (1), M24−M28. (2) Brinkman, U. A. T. Chromatographia 1987, 24, 190−200. (3) Workman, J.; Lavine, B.; Chrisman, R.; Koch, M. Anal. Chem. 2011, 83 (12), 4557−4578. (4) Larson, P. Gas Chromatogr. Tech. Appl. 2001, 122−139. (5) (a) Kaushal, R.; Kaur, N.; Upadhyay, A.; Suri, O. P.; Thakkar, A. Int. Res. J. Pharm. 2011, 2 (5), 1−7. (b) Swartz, M. J. Liq. Chromatogr. Relat. Technol. 2010, 33 (9−12), 1130−1150. (6) (a) Schafer, W. A.; Hobbs, S.; Rehm, J.; Rakestraw, D. A.; Orella, C.; McLaughlin, M.; Ge, Z.; Welch, C. J. Org. Process Res. Dev. 2007, 11 (5), 870−876. (b) Welch, C. J. Chirality 2008, 21 (1), 114−118. (c) Welch, C. J.; Gong, X.; Cuff, J.; Dolman, S.; Nyrop, J.; Lin, F.; Rogers, H. Org. Process Res. Dev. 2009, 13 (5), 1022−1025. (7) (a) Lunn, A. K.; Morton, R. A. Analyst 1952, 77, 718−731. (b) Kushwah, D.; Patel, H. B.; Sinha, P. K.; Jana, P. K. E-J. Chem. 2011, 8 (4), 1504−1511. (8) Nussbaum, M. A.; Baertschi, S. W.; Jansen, P. J. J. Pharm. Biomed. Anal. 2002, 27 (6), 983−993. (9) Sun, P.; Wang, X.; Alquier, L.; Maryanoff, C. A. J. Chromatogr., A 2008, 1177 (1), 87−91. (10) (a) Pomerantsev, A. L.; Rodionova, O. Y. J. Chemom. 2012, 26 (6), 299−310. (b) Munson, J.; Stanfield, C. F.; Gujral, B. Curr. Pharm. Anal. 2006, 2 (4), 405−414. (c) Xiang, Y.; Lucas, J.; Van Alsten, J.; Li, B.; Preston, B.; Lovdahl, M.; Hayward, C. M. Am. Pharm. Rev. 2012, 15 (3), 56−61. (d) Reid, G. L.; Ward, H. W., II; Palm, A. S.; Muteki, K. Am. Pharm. Rev. 2012, 15 (4), 49−55. (11) Webster, G. K.; Marsden, I.; Pommerening, C. A.; Tyrakowski, C. M.; Tobias, B. J. Pharm. Biomed. Anal. 2009, 49 (5), 1261−1265. (12) Marquez, B.; Fey, M.; Colson, K.; Krull, R.; Bez, E.; Piroli, D.; Maas, W. NMR flow cell. US 20120092013 A1, 2012. (13) (a) Maiwald, M.; Fischer, H. H.; Kim, Y.-K.; Hasse, H. Anal. Bioanal. Chem. 2003, 375 (8), 1111−1115. (b) Maiwald, M.; Fischer, H. H.; Kim, Y.-K.; Albert, K.; Hasse, H. J. Magn. Reson. 2004, 166 (2), 135−146. (c) Foley, D. A.; Doecke, C. W.; Buser, J. Y.; Merritt, J. M.; Murphy, L.; Kissane, M.; Collins, S. G.; Maguire, A. R.; Kaerner, A. J. Org. Chem. 2011, 76 (23), 9630−9640. (d) Barrios Sosa, A. C.; Williamson, R. T.; Conway, R.; Shankar, A.; Sumpter, R.; Cleary, T. Org. Process Res. Dev. 2011, 15 (2), 449−454.

E

dx.doi.org/10.1021/ac402382d | Anal. Chem. XXXX, XXX, XXX−XXX