Article pubs.acs.org/ac
Chromatographic Resolution of Closely Related Species in Pharmaceutical Chemistry: Dehalogenation Impurities and Mixtures of Halogen Isomers Erik L. Regalado,* Ping Zhuang, Yadan Chen, Alexey A. Makarov, Wes A. Schafer, Neil McGachy, and Christopher J. Welch* Merck Research Laboratories, Rahway, New Jersey 07065, United States S Supporting Information *
ABSTRACT: In recent years, the use of halogen-containing molecules has proliferated in the pharmaceutical industry, where the incorporation of halogens, especially fluorine, has become vitally important for blocking metabolism and enhancing the biological activity of pharmaceuticals. The chromatographic separation of halogen-containing pharmaceuticals from associated isomers or dehalogenation impurities can sometimes be quite difficult. In an attempt to identify the best current tools available for addressing this important problem, a survey of the suitability of four chromatographic method development platforms (ultra high-performance liquid chromatography (UHPLC), core shell HPLC, achiral supercritical fluid chromatography (SFC) and chiral SFC) for separating closely related mixtures of halogen-containing pharmaceuticals and their dehalogenated isosteres is described. Of the 132 column and mobile phase combinations examined for each mixture, a small subset of conditions were found to afford the best overall performance, with a single UHPLC method (2.1 × 50 mm, 1.9 μm Hypersil Gold PFP, acetonitrile/methanol based aqueous eluents containing either phosphoric or perchloric acid with 150 mM sodium perchlorate) affording excellent separation for all samples. Similarly, a survey of several families of closely related halogen-containing small molecules representing the diversity of impurities that can sometimes be found in purchased starting materials for synthesis revealed chiral SFC (Chiralcel OJ-3 and Chiralpak IB, isopropanol or ethanol with 25 mM isobutylamine/carbon dioxide) as well as the UHPLC (2.1 × 50 mm, 1.8 μm ZORBAX RRHD Eclipse Plus C18 and the Gold PFP, acetonitrile/methanol based aqueous eluents containing phosphoric acid) as preferred methods.
I
many commonly used synthetic transformations involving low valent transition metals (e.g., catalytic hydrogenation using a palladium/charcoal catalyst) can give rise to halogen loss via hydrodehalogenation,16 sometimes generating unanticipated “stealth” impurities that can coelute with the major product of interest. Several illustrations of the problems that can arise with desfluoro-impurities and fluorine-containing regioisomers are presented in Figure 1. Reversed phase HPLC of the drug, aprepitant, and the related desfluoro impurity highlights the very small separations that are often observed (Figure 1a). Compounds containing multiple halogens can give rise to fairly complex mixtures of dehalogenation isomers. For example, in Figure 1b, catalytic hydrogenation of a pyridone intermediate leads to formation of the desired product, along with three different dehalogenated products, two of which can be resolved
n recent years, the use of halogen-containing molecules has proliferated in the pharmaceutical industry, where the incorporation of halogens, especially fluorine, has become vitally important for blocking metabolism and enhancing the biological activity of pharmaceuticals.1−3 As a case in point, four of the five top selling small molecule drugs in 2011 contained halogens: Lipitor (atorvastatin, F), Plavix (clopidogrel bisulfate, Cl), Abilify (aripiprazole, Cl), and Advair (fluticasone/salmeterol, F).4 The increased use of fluorine in pharmaceuticals has been accompanied by an evolution of tools and technologies for enabling synthetic chemistry,5−8 spectroscopy9,10 and chromatographic separation11−13 of fluorine-containing pharmaceutical candidates. The chromatographic separation of dehalogenation impurities and mixtures of halogen positional isomers can be quite challenging. Fluorine-containing pharmaceuticals are typically slightly more hydrophobic than their corresponding proteo analogs, and conventional chromatography techniques can sometimes be effective in resolving desfluoro-impurities from fluorine-containing drugs.14,15 However, in many instances, desfluoro-impurities can prove difficult to resolve. Furthermore, © 2013 American Chemical Society
Received: October 14, 2013 Accepted: December 10, 2013 Published: December 10, 2013 805
dx.doi.org/10.1021/ac403376h | Anal. Chem. 2014, 86, 805−813
Analytical Chemistry
Article
Figure 1. (a) Reversed phase achiral HPLC method for separation of aprepitant and related impurities. Column: YMC ODS-AQ C18 (4.6 mm × 250 mm, 5 μm). Room Temperature. Detection: UV 220 nm. Sample: 20 μL injection of 1 mg/mL sample in methanol. Flow rate: 1.5 mL/min. Mobile phase: 0.1% H3PO4 in H2O/CH3CN (ramp from 65:35 to 20:80 in 20 min, hold for 10 min). (b) Reversed phase achiral HPLC method for separation of a hydrogenation product intermediate and related impurities. Column: Thermo Hypersil Gold Phenyl (4.6 mm × 150 mm, 3 μm). Temperature: 40 °C. Detection: UV 210 nm. Flow rate: 0.75 mL/min. Mobile phase: 0.05% TFA in H2O/0.05% TFA in CH3CN (ramp from 65:35 to 56:44 in 25 min, then ramp to 10:90 in 5 min). (c) HPLC loading study for purification of a development compound. Column: Kromasil KR60-10SIL (4.6 mm × 250 mm, 5 μm). Room Temperature. Detection: UV 254 nm. Sample: 100 μL injection of 100 mg/mL sample in methanol. Flow rate: 1.5 mL/min. Mobile phase: 40% EtOH/heptane.
by reversed phase HPLC using a phenyl column. Interestingly, one of the dehalogenation isomers is not resolved using these conditions but is detected in a separate GC assay (Supporting Information, Figure S1). The use of multiple chromatographic methods to address such complex samples is not uncommon. Once dehalogenation impurities have been created, their removal can be quite difficult, as these impurities can often cocrystallize or coelute with the major product of interest. In addition, commercial halogen-containing starting materials may contain significant amounts of the proteo analog or of other halogen-containing isomers, leading to the formation of impurities that can persist through several synthetic steps. Figure 1c shows small scale preparative chromatographic purification of a development candidate containing a desfluoro impurity using a 4.6 mm i.d. analytical silica column. Conditions obtained in this
preparative loading study were then applied to the purification of 640 g of the candidate using an 11 cm i.d. preparative column. It is important to point out that, while sometimes quite challenging, the general problem of the chromatographic separation of dehalogenation impurities and mixtures of halogen isomers is typically solved on a case by case basis using a variety of different chromatographic conditions. While such an approach is often successful, method development can often be timeconsuming owing to the shortage of specific recommendations for rapidly identifying a suitable chromatographic method. Given the increasing importance of the separation of fluoro/desfluorospecies and mixtures of fluorine-containing regioisomers, we undertook a survey of chromatographic method development platforms with the aim of identifying the best solutions that are well suited for this general class of problem. 806
dx.doi.org/10.1021/ac403376h | Anal. Chem. 2014, 86, 805−813
Analytical Chemistry
■
Article
EXPERIMENTAL SECTION Instrumentation. Reversed phase achiral ultra highperformance liquid chromatography (UHPLC) screening experiments were performed with a Waters Acquity UHPLC H-Class (Waters Corp., Milford, MA, USA) system equipped with a quaternary solvent delivery pump, a sampler manager (FTN autosampler), two auxiliary column managers allowing six installed columns, a 80 Hz photodiode array detector, and Waters MassLynx software for instrument control and data processing. Core shell reversed phase HPLC screening experiments were performed on an Agilent 1100 system (Agilent Technologies, Palo Alto, CA, USA). The Agilent system comprised a G1312A binary pump, a G1379A degasser, a G1367A WPALS autosampler, a G1316A column compartment, a G1315B diode array detector, and a 6120 Quadrupole LC/MS detector with electrospray ionization in the positive and negative mode. The system was controlled by Chemstation software. As dry gas nitrogen with a gas flow of 12 L/min (350 °C) was used, the nebulizer was adjusted to 20 MPa (35 psig) and the capillary voltage to 3000 V. Achiral supercritical fluid chromatography (SFC) screening was carried out on an Agilent 1100 with an Aurora A5 Fusion SFC module. The Agilent system comprised a modified G1322A vacuum degasser, a modified G1312A binary pump, a G1367A WPALS autosampler, two G1316A column compartments, and a G1315B diode array detector. The system was controlled by ChemStation software. Chiral SFC screening and optimization experiments were carried out on Waters Acquity UPC2 (WatersCorp., Milford, MA, USA) systems equipped with a fluid delivery module (a liquid CO2 pump and a modifier pump), a sampler manager (FL autosampler), two auxiliary column managers allowing six installed columns, a photodiode array detector, and MassLynx software. Chemicals, Reagents, and Preparation of Buffer Solutions. All chemicals used were of analytical grade. A complete list of chemicals and reagents is provided in the Supporting Information, as well as the protocols for preparation of the buffer solutions used as mobile phase. Stationary Phases. UHPLC Platform. The 2.1 mm i.d. × 50 mm length (1.8 μm particles) ZORBAX (RRHD Eclipse Plus C18, SB-CN, SB-C8, and SB-phenyl) columns were purchased from Agilent Technologies. The 2.1 mm × 50 mm, 1.9 μm Hypersil (Gold AQ and Gold PFP) columns were obtained from Thermo Scientific (Rockford, IL, USA). Core Shell HPLC Platform. The Poroshell 120 (SB-C18, AQ, and Phenyl hexyl) columns were purchased from Agilent Technologies. The Ascentis Express (C8, ES-CN, and F5) columns were obtained from Supelco Analytical (Bellefonte, PA, USA). Dimensions and particle size for all columns were 3.0 mm × 100 mm, 2.7 μm. Achiral SFC Platform. The 4.6 mm × 150 mm, 3 μm Luna HILIC column was purchased from Phenomenex (Torrance, CA, USA). The 4.6 mm × 250 mm, 5 μm GreenSep (BASIC, Nitro, DEAP, Ethyl Pyridine, and Ethyl Pyridine II) columns were purchased from ES Industries (West Berlin, NJ, USA). The 4.6 mm × 250 mm, 5 μm Kromasil (Silica, Diol, NH2, and CN) columns were obtained from AkzoNobel, Separation Products (Bohus, Sweden). The 4.6 mm × 250 mm, 5 μm ZymorSPHERHAP and ZymorSPHER-HADP columns were purchased from Zymor, Inc. (Wayne, NJ, USA). Chiral SFC Platform. Columns packed with Chiralpak (AD-3, AS-3, IA, IB, IC, IC, IE, IF) and Chiralcel (OD-3, OJ-3, OZ-3) were purchased from Chiral Technologies (West Chester, PA,
USA); Lux (Amylose-2 and Cellulose-4) columns were purchased from Phenomenex. Dimensions and particle size for all columns were 4.6 mm × 150 mm, 3 μm. Separation Systems and Screening Methodologies. All screening methods used in this study were performed by standard protocols on four different systems (UHPLC, core shell HPLC, achiral SFC and chiral SFC). A detailed description of all methods and conditions is shown in Table 1. Table 1. Separation Platforms and Screening Methods Used in This Study conditions column size (mm) particle size (μm) flow rate (mL/min) initial hold (min) gradient (%)c hold (min) post equilibration (min)
UPLC
core shell HPLC
50 × 2.1
100 × 3.0
1.8 and 1.9b
achiral SFC
chiral SFC 150 × 4.6
2.7
150a and 250 × 4.6 3a and 5
3
0.6
0.75
3
3
0
0
4
4
5−95 in 5.5 min 0.7 0
10−95 in 8 min 2 4
4−40 in 6 min 5 2
4−40 in 6 min 5 2
All columns used in the achiral SFC platform were 250 × 4.6 mm and 5 μm, except the Luna HILIC column (150 × 4.6 mm, 3 μm). bThe particle size of ZORBAX (RRHD Eclipse Plus C18, SB-CN, SB-C8, and SB-Phenyl) columns is 1.8 μm, while the particle size of Hypersil Gold (AQ and PFP) columns is 1.9 μm. cWhen CH3OH is used as mobile phase in the UPLC screening, its concentration is kept constant at 25% as follows: CH3CN/CH3OH/aqueous solution (from 3.7:25:71.3 to 70:25:5 in 5.5 min). A temperature of 40 °C was used for all screening methods. The backpressure regulator was set at 200 bar for SFC screens. UPLC mobile phases: (1) 0.1% H3PO4 in H2O/ CH3CN. (2) 150 mM NaClO4 in 0.02% HClO4/CH3CN. (3) 5 mM (NH4)2HPO4 in H2O (pH 8.0)/CH3CN. (4) 35 mM KPF6 in 0.1% H3PO4/CH3CN. (5) 10 mM CH3COONH4 in H2O/CH3CN (pH 6.5). (6) 0.1% H3PO4 in H2O/CH3OH/CH3CN. (7) 150 mM NaClO4 in 0.02% HClO4/CH3OH/CH3CN. (8) 5 mM (NH4)2HPO4 in H2O (pH 8.0)/CH3OH/CH3CN. (9) 35 mM KPF6 in 0.1% H3PO4/CH3OH/CH3CN. Core shell HPLC mobile phases (MS compatible): (1) 0.05% TFA in H2O/0.05% TFA in CH3CN. (2) 0.1% HCOOH in H2O/0.1% HCOOH in CH3CN. (3) 2 mM NH4HCO2 in H2O (pH 3.5)/2 mM NH4HCO2 in CH3CN (pH 3.5). Achiral SFC mobile phases: (1) 25 mM IBA in CH3OH/CO2. (2) 25 mM IBA in IPA/CO2. Chiral SFC mobile phases: (1) 25 mM IBA in CH3OH/CO2. (2) 25 mM IBA in IPA/CO2. (3) 25 mM C2H5OH in IPA/CO2. a
■
RESULTS AND DISCUSSION In an attempt to represent the type of real world separation challenges encountered with halogen-containing pharmaceuticals, we selected twelve different compound mixtures, ranging from halogen-containing pharmaceuticals and their dehalogenated analogs to mixtures of fairly simple halogen-containing compounds resembling starting materials (Supporting Information, Figure S2). The selected drugs are structurally diverse, wellknown pharmaceuticals where both the halogenated and dehalogenated analogs are commercially available. While two of the samples contain chlorine (Mixtures 7 and 10), most of the sample sets are fluorine-containing pharmaceuticals, along with the corresponding desfluoro impurities. Mixtures 10, 11, and 12 807
dx.doi.org/10.1021/ac403376h | Anal. Chem. 2014, 86, 805−813
Analytical Chemistry
Article
Figure 2. 132 expanded chromatograms obtained from the injections of aprepitants (mixture 1) in four different separation platforms using standard screening conditions as shown in Table 1. A total of 12 different reversed phase columns in combination with a variety of eluents from low to high pH were evaluated, as well as 24 packed-columns in SFC with several polar modifiers in CO2. UHPLC (six columns and nine mobile phases), core shell HPLC (six columns and three mobile phases), achiral SFC (12 columns and two mobile phases), and chiral SFC (12 columns and three mobile phases).
analytical methods,17−19 including reversed phase achiral UHPLC (6 columns × 9 eluents), reversed phase core shell HPLC (6 columns × 3 eluents), achiral SFC (12 columns × 2 eluents), and chiral SFC (12 columns × 3 eluents). Limited investigation with several additional chromatography method development screening tools (reversed phase chiral HPLC, polar-organic chiral HPLC) showed only modest performance. A complete listing of the columns, eluents, and methods used in the study is presented in the Experimental Section. The resulting 132 chromatograms for each of the 12 sample mixtures provided a somewhat bewildering array of information that required
each contain a family of closely related halogen-containing small molecules, reflecting the diversity of impurities that can sometimes be found in purchased starting materials for synthesis. While most of the drugs and impurities are present as single stereoisomers, several are racemic (e.g., warfarins in mixture 7, desfluoro voriconazole in mixture 3, and ofloxacins in mixture 9). Consequently, each of these mixtures would be expected to afford two possible peaks by achiral chromatography and as many as four peaks by chiral chromatography. We examined each of these mixtures using several chromatographic method development screening systems that are routinely used within these laboratories for the development of 808
dx.doi.org/10.1021/ac403376h | Anal. Chem. 2014, 86, 805−813
Analytical Chemistry
Article
categorization and scoring to allow straightforward interpretation. Examination of aprepitant and its desfluoro analog (mixture 1) afforded the results shown in Figure 2, with chromatograms expanded to highlight the peaks of interest. Several of the UHPLC platform conditions provided baseline separation for this mixture, with the SB-CN column affording the best separation with an eluent of either acetonitrile or methanol with phosphoric acid. Very good separation was also obtained on the Gold PFP, SB-Phenyl and Eclipse C18 columns. The elution order for the two compounds varies from condition to condition, with the smaller peak, corresponding to the desfluoro compound, generally eluting first. Separation of mixture 1 on the core shell HPLC platform shows very good separation for all columns, with a marked preference for the acetonitrile/water/ TFA eluent. Again, the elution of desfluoro before fluoro compound is noted, consistent with general expectations for reversed phase separations, given the slightly higher log P of the fluoro species (see Supporting Information for a complete listing of calculated log P values, Table S1). Separation of mixture 1 on the achiral SFC platform shows a number of conditions that resolve the two species, with a strong preference for methanol as the polar modifier. Interestingly, the elution order again follows the trend of desfluoro before fluoro, even though this is now technically a normal phase separation. Clearly, factors other than Log P (e.g., molecular weight, polarizability, ability to participate in halogen bonding interactions) are also important for this separation. Separation of mixture 1 on the chiral SFC platform shows a number of excellent separations, with elution order varying depending on column and conditions. The overall best separation obtained for the aprepitant mixture was provided by the Chiralcel OJ-3 using a MeOH/CO2/IBA eluent, where baseline resolution of the components of interest is obtained in slightly more than 1 min. In order to simplify comparison across platforms, a scoring system was developed to reflect the quality of the separations obtained with each condition (Figure 3). We chose to focus on resolution and speed, however, a variety of different scoring systems emphasizing other aspects could be imagined. Separations that did not achieve at least partial resolution of all components received 0 points, while marginal separations with resolution less than 0.8 (valley extending to ∼half peak height) received 1 point, partial resolutions (0.8 < Rs < 1.5) received 2 points, and baseline resolutions (Rs ≥ 1.5) received 5 points. In addition, a bonus point was awarded for fast baseline resolutions (
