HILIC to the Rescue: Pharmaceutical Development Case Examples

Hydrophilic interaction liquid chromatography (HILIC) has rapidly matured. What was once a somewhat niche technique is becoming more commonplace in the liquid chromatographer's toolbox. This broader acceptance is likely due to the rapidly growing body of research literature, leading to a better understanding of the separation mechanism, applications, advantages, and shortcomings. As a result, column vendors have realized HILIC's potential, and now offer method development literature, as well as application notes and practical guides. When compared to alternative analytical techniques for analyzing polar molecules, which are difficult to retain using conventional reversed-phase conditions, HILIC has some distinct benefits, such as the ability to be run easily on LC instruments setup for reversed-phase separations and electrospray ionization compatibility. We have found HILIC to be an extremely useful technique for supporting small molecule pharmaceutical development. Described here are some case examples where HILIC not only enabled generation of key data, but could also be validated to the same rigor as traditional reversed-phase LC methods. The first example involves underivatized amino acid analyses. Generalized, isocratic conditions were identified to accelerate method development. Ultraviolet (UV) detection was suitable for high levels, while mass spectrometry (MS) could perform low level impurity monitoring. In the second example, due to the thermal degradation of the main sample component (a pharmaceutical synthetic intermediate), the typical means of assaying acetamide by gas chromatography could not be applied. HILIC–MS provided the means to analyze acetamide at the required parts-per-million levels. Finally, a simple and direct means of monitoring potential genotoxic impurities aziridine and 2-chloroethylamine by HILIC–MS was developed. Furthermore, eliminating derivatization and extraction yielded a robust, sensitive method.

Analytical methods in the pharmaceutical industry can span a wide range of target performance goals. In the early development space, methods may only be used to support single experiments to gather knowledge and experience to enable key decisions regarding process development. As development evolves, so do the methods, progressing into fit-for-purpose techniques that can still rapidly generate knowledge with increasing robustness and decreasing error, but may not be at the level required for full validation and external transfer. A successful pharmaceutical asset progressing to market necessitates the highest level of analytical quality, with methods potentially distributed to a global QC network, and executed for many years to come. Reversed-phase liquid chromatography (LC) remains the most widely used separation technique for the characterization of starting materials, intermediate species, process streams, active pharmaceutical ingredients (APIs), and drug products, due to its robustness, reliability, and availability across the worldwide analytical community. Hydrophilic interaction chromatography (HILIC), despite using similar column technology and the same instrumentation (without needing to "convert," as in normal-phase LC), sees far less usage than RPLC, although its use is growing (1). Typical analytes amenable to HILIC are polar, hydrophilic molecules with partition or distribution coefficients (logP or logD, respectively) less than zero, that also have appreciable solubility in diluents of not more than 50% water (with the remainder frequently being acetonitrile or methanol). Although many APIs themselves fail these requirements (typically with a logP or logD value greater than zero), many starting materials and intermediates used to synthesize small molecule APIs, as well as their related impurities or degradants, are HILIC compatible. Frequently, efforts to analyze these types of compounds by reversed-phase LC, even using columns run under 100% aqueous conditions, fail to obtain sufficient retention, leading the analyst to seek alternative separation mechanisms, such as normal-phase LC or mixed-mode columns.

Assuming target analytes are am-enable to HILIC, potential reasons analytical scientists may avoid HILIC are lack of familiarity, perception of a lack of method development tools (workflows or software), perception of long re-equilibration times (compared to reversed-phase LC), or a general feeling that HILIC is not robust. At present, these areas have all been addressed to some extent. A growing body of literature, conference presentations, and HILIC column manufacturer messaging have all increased awareness and understanding of the HILIC mode, as well as its advantages and disadvantages. In addition to literature sources (2), all major HILIC column vendors have released white papers, brochures, or books to assist method developers, such as (3,4). Computational models are being developed for method prediction and optimization (5). Additional understanding of the HILIC mechanism and operational studies have shown that re-equilibration can be on par with reversed-phase LC (6). As is discussed below, HILIC can also be utilized both for simple fit-for-purpose methods and for robust, fully validated methods, providing a powerful asset in the separations scientist's toolbox.

Materials and Methods

Chemicals and Sample Preparation

Aziridine (1 mg/mL in methanol) was from Spex CertiPrep. 99% 2-chloroethylamine hydrochloride, HPLC gradient grade acetonitrile, HPLC gradient grade methanol, >99.999% ammonium acetate, LC–MS grade formic acid, >99% acetamide, and >97% amino acids were from Sigma-Aldrich. Water was purified in-house, using an Advantage A10 purification system (MilliporeSigma). Proprietary compounds were synthesized in-house at Bristol-Myers Squibb. Amino acid assays were prepared in methanol. Acetamide assays were prepared in acetonitrile. Aziridine and 2-chloroethylamine hydrochloride assays were prepared in 10 mM ammonium acetate in acetonitrile:water (95:5).

Instrumentation and Chromatographic Conditions

Analyses were performed using a Waters Acquity HClass UHPLC with a PDA (proline and platform amino acid methods) or QDA (alanine and acetamide methods) detector and an Acquity BSM UHPLC with an SQD detector (aziridine and 2-chloroethylamine hydrochloride methods). Single-ion monitoring (SIM) using positive electrospray ionization (ESI) was performed at 44.07 (aziridine), 59.90 (acetamide), 79.90 (2-chloroethylamine hydrochloride), and 90.00 (alanine) Da. QDA tuning parameters were: capillary at 1.5 kV; cone at 10 V; and probe at 600 °C. SQD tuning parameters were: source temperature of 150 °C; desolvation temperature of 450 °C; desolvation gas flow of 900 L/h; capillary at 3 kV; cone at 30 V; and extractor at 3 V. A divert valve was used for MS analyses to direct the main components' peak to waste. Flow splitting (1:1) used a LC–MS post-column splitter (Sigma-Aldrich).

Amino acid and acetamide analyses used a Waters Atlantis HILIC 4.6 mm × 150 mm, 3-µm particle size column maintained at 25 °C with a flow rate of 1 mL/min. For amino acid analyses, mobile phase A was acetonitrile, and mobile phase B was 10 mM ammonium acetate in acetonitrile:water (50:50); isocratic compositions are given in Table I. Injection volumes for proline and alanine were 5 and 50 µL, respectively. Acetamide analyses used 0.2% formic acid in acetonitrile:water (97:3) as mobile phase. Acetamide was injected at 10 µL. Aziridine and 2-chloroethylamine hydrochloride analyses used a Waters BEH HILIC 2.1 mm × 50 mm, 1.7-µm particle size column maintained at 30 °C with a flow rate of 0.5 mL/min; mobile phase was 10 mM ammonium acetate in acetonitrile:water (95:5). Aziridine and 2-chloroethylamine hydrochloride samples were injected at 2 µL. Other specific chromatographic conditions are described in the Results and Discussion section.

Results and Discussion

HILIC Method Development

A comprehensive HILIC method development approach looks at analyte physicochemical properties, a variety of stationary phases (such as bare silica and bonded phases with ion exchanging or zwitterionic characteristics), various mobile phase pHs with varying buffer concentrations, various types and levels of organic in the MP, and temperature. This approach, which can be automated to some degree, can be resource and time intensive. Within the constraint of small molecule pharmaceutical development, the process can be streamlined as the analytes encompass a more limited class of properties. Additionally, the method requirements are typically to quantitate a single or a few analytes without interferences from other sample components, as opposed to a comprehensive full purity or impurity stability-indicating method tasked with resolving all components in a complicated mixture. Frequently, this need manifests as a targeted impurity analysis or measuring starting material versus the end-product of a reaction. Within this limited scope, a narrower set of HILIC conditions can be screened initially, as shown in Figure 1, with a more exhaustive method development approach being taken only should the first-pass fail.

Figure 1: Flow of streamlined approach HILIC method development. See the section HILIC Method Development for a full description of the workflow. Green denotes a positive response; red denotes a negative response; purple denotes no response required.

The simple screen begins with assessing the target analyte(s)' logP or logD to help determine if HILIC is a suitable technique, with a general guidance that negative values could be viable. These values can come from internal sources, literature, or predicted either by commercial (such as ACD/Percepta) or free programs (such as chemicalize.com). For the chromatography, mass spectrometric compatible mobile phases of either 0.2% formic acid or 10 mM ammonium acetate in acetonitrile:water (95:5) are used, with the chosen pH dictated by the desired analyte charge state (typically ionized to increase retention). Only unbonded type stationary phases are evaluated, with a bridged ethylene hybrid (BEH) being tested first, followed by a bare silica column if there is insufficient retention (k <1). Although usually less retentive than bare silica, we have found hybrid particles tend to have longer column lifetimes when subjected to the variety of reaction stream samples generated during pharmaceutical process development; in general, both BEH and silica are more robust with regard to these types of samples as compared to bonded phases. Once retention is established, resolution of the target analyte(s) from interferrants is assessed. If the simple screen affords the proper selectivity, additional method optimization can then proceed if needed. Should the conditions fail to retain or resolve the selected components, more comprehensive method conditions are screened, such as described in references (3,4). The following three case examples highlight instances where this streamlined approach to method development quickly yielded successful HILIC methods.

Amino Acid Analyses

There are numerous techniques available to the analytical chemist for assaying amino acids, such as titration, liquid chromatography, and capillary electrophoresis, with or without derivatization. When considering the enablement of process chemists to perform their own analyses, traditional reversed-phase LC with UV detection is ideal. As an example, a small-molecule API was being developed that exhibited stability issues. To address this, API co-crystals with amino acids were being evaluated, which necessitated a method to assay the amino acid content. The initial form utilized proline. As the API has a logP of about 0, and proline has a logD of <-2 at neutral pH, HILIC seemed suitable. The streamlined development as shown in Figure 1 identified a bare silica column as having enough retention of proline to avoid interference from the API, which was unretained. As both the API and proline were highly soluble in methanol, a concentration of 20 mg/mL of the co-crystal (corresponding to 8 mg/mL proline) could be assayed at 210 nm. The slightly optimized conditions (Figure 2) were easily transferred to the process chemists for use in their own laboratories.

Figure 2: HILIC assay for proline in a co-crystal sample. Column: Waters Atlantis HILIC 4.6 mm × 150 mm, 3-µm particle size; mobile phase A: acetonitrile; mobile phase B: 10 mM ammonium acetate in acetonitrile:water (50:50); isocratic at 55% mobile phase B; flow rate: 1 mL/min; column temperature: 25 °C; detection: ultraviolet (UV) absorbance (AU) at 210 nm; injection volume: 5 µL; diluent: methanol.

As other amino acid forms were being investigated, it became apparent that the proline conditions could be generalized to accommodate many other common amino acids. The amino acids could be grouped by logD at pH 7, and then isocratic conditions at various levels of organic content could be used to maintain a roughly 10 min run time (Table I). An example of the logD group of ca. -2 is shown in Figure 3. Although not optimized for specific sample or amino acid cases, this pro-active workflow approach significantly reduced the need for "on-request" new methods to be developed.

Figure 3: HILIC conditions for amino acids with LogD from -2.0 to -2.8. Column: Waters Atlantis HILIC 4.6 mm × 150 mm, 3-µm particle size; mobile phase A: acetonitrile; mobile phase B: 10 mM ammonium acetate in acetonitrile:water (50:50); isocratic at 50% mobile phase B; flow rate: 1 mL/min; column temperature: 25 °C; detection: ultraviolet (UV) absorbance (AU) at 210 nm (scale is normalized); injection volume: 5 µL; diluent: methanol.

An assay of a relatively pure sample, with the major peak being unretained, essentially reduces the separation to a filter. A more challenging example was a request to identify alanine as an impurity in a synthetic intermediate crude product. In this case, the sample had many more potentially interfering components, and the target analyte was to be quantitated at a much lower level (0.1% w/w ). To achieve a quantitative response by UV, a working concentration of alanine on the order of at least 1 mg/mL would be required. Unfortunately, to achieve the required sensitivity, this would require the synthetic intermediate to be prepared at 1 g/mL, which is outside of the solubility range of the sample.

Although less ideal when considering transferability, switching from UV to MS detection afforded not only the requisite sensitivity, but also reduced the need to separate interfering sample components, as only the main peak needed to be resolved from alanine since no impurities isobaric to alanine were observed in the samples. Due to the high organic content of the mobile phase in HILIC, electrospray ionization MS is enhanced (7), further reducing the sample concentration requirements. The final HILIC method conditions were similar to the platform conditions given in Table I, with a slightly weaker eluent strength to help better separate the slightly retained main peak from alanine. Sufficient sensitivity was obtained at 4 µg/mL alanine, requiring a main component concentration of 4 mg/mL (Figure 4). In this particular case, the process actually was generating alanine in relatively high amounts (~1% w/w ), and the HILIC method allowed for monitoring it as the route was altered to reduce the levels.

Figure 4: HILIC method for monitoring alanine as an impurity. Column: Waters Atlantis HILIC 4.6 mm × 150 mm, 3-µm particle size; mobile phase A: acetonitrile; mobile phase B: 10 mM ammonium acetate in acetonitrile:water (50:50); isocratic at 45% mobile phase B; flow rate: 1 mL/min; column temperature: 25 °C; detection: selected ion monitoring (SIM) at m/z = 90.00; injection volume: 50 µL; diluent: methanol. Ultraviolet (UV) absorbance (AU) trace at 200 nm shows the main component and other sample impurities.

Acetamide Analyses

While most methods to assay organic impurities in the synthesis of an API require sensitivity on the order of 0.05% w/w based on the ICH Q3A(R2) guideline (8), some impurities, such as genotoxic impurities, frequently require much lower level detectability, such as parts-per-million (ppm) for higher dosage APIs (9). While these types of potent impurities may be reagents, intermediates, or side-products of the synthetic route, one frequently encountered known carcinogen is acetamide, which can be generated from the common processing solvent acetonitrile (10).

Similar to amino acids, there are a myriad of documented ways to assay acetamide. However, only a subset have the requisite selectivity and sensitivity to detect ppm levels of acetamide in a synthetic intermediate or API matrix. Based on acetamide's physicochemical properties (such as low molecular weight, highly polar, reasonable vapor pressure, lack of chromophore) gas chromatography with either a flame ionization of mass spectrometer detector, depending on the required sensitivity, is a typical first choice in our field. However, sample matrix issues, especially at the high concentrations needed for ppm analyses, can cause problems either in the separation or in recovery.

In 2001, Olson demonstrated that HILIC using base silica columns was a suitable alternative to GC for acetamide (11). However, the method described utilized a near neutral phosphate-based mobile phase and low wavelength UV detection, which would only be suitable down to 0.05% w/w sensitivity. When converting to HILIC–MS using the strategy shown in Figure 1, the acetonitrile-ammonium adduct ion at 59 Da would interfere when performing acetamide selected ion monitoring at m/z of 60, leaving formic acid as the preferred mobile phase modifier. Alternatively, acetone could be used as the MP organic component. While acetamide retention was increased over acetonitrile, the ESI signal was much reduced in the acetone system (Figure 5), similar to the report by Heaton and associates (12). Based on this, a formic acid modified acetonitrile system was preferred and, to improve retention, only 3% water was used.

Figure 5: Comparison of acetamide sensitivity in HIILC-MS using acetonitrile versus acetone as the organic component. Column: Waters Atlantis HILIC 4.6 mm × 150 mm, 3-µm particle size; mobile phase: 0.2% formic acid in water:acetonitrile or acetone (3:97); flow rate: 1 mL/min; column temperature: 25 °C; detection: selected ion monitoring (SIM) at m/z = 59.90; injection volume: 10 µL; sample: 500 ng/mL acetamide in acetonitrile.

The initial request for an alternative to GC–MS involved a relatively polar prepenultimate to an API that was thermally labile and failing recovery for acetamide. A sample concentration of 25 mg/mL was needed to achieve the method target of 100 ppm (2.5 µg/mL acetamide). Additionally, the sample itself was well retained in the HILIC mode. Once acetamide eluted a step gradient to 50% water was used to remove the main component, followed by a 10 min re-equilibration. At a 1 mL/min flow rate, this is a similar amount of column volumes as would be used in reversed-phase LC. The method was successfully validated as a limit test (Table II) in support of a new drug application, demonstrating the robustness of straightforward HILIC methods. As the method had been rigorously tested, it became a starting point for other acetamide analyses, essentially becoming a platform method. For one API, with excellent solubility where the concentration could be doubled, reduction in the flow to the MS improved the sensitivity whereby the method was able to be qualified at 10 ppm.

Aziridine and 2-Chloroethylamine Analyses

2-chloroethylamine is a common synthetic chemistry reagent that can also undergo self-cyclization to form aziridine (13). Both compounds have been found to be mutagenic, and must be controlled as genotoxic impurities in pharmaceuticals either in the API or further upstream in the synthesis (9,14). The highly reactive nature of these compounds makes analyses challenging, and prior literature reports utilized a derivatization or extraction step, or both, in the sample preparation prior to chromatography to achieve low-level detection from a concentrated matrix (15–19). Each additional unit operation of a method adds complexity, time, and the potential for increased error. Direct analysis is preferred to improve simplicity and robustness, especially when considering method transfer and end-user empowerment.

Douša and colleagues recently described a direct method using HILIC–MS for measuring 2-chloro-N-(2-chloroethyl) ethanamine in vortioxetine (20). The method was rapid (LC analysis of ~10 min), sensitive (1 ng/mL LOQ), and met typical validation requirements. We followed a similar approach to develop a method suitable to detect 0.2 µg/mL of aziridine and 2-chloroethylamine in a process intermediate. At a sample concentration of 10 mg/mL, this correlates to a sensitivity of 20 ppm. Ideally, both target analytes would be assayed in the same method. Using the streamlined workflow (Figure 1), minimal HILIC chromatographic method development was required.

A short (5 cm) UHPLC (sub-2-µm particle size) column was initially selected in order to maximize speed while maintaining highly efficient and resolved analyte peaks. Paired with a neutral mobile phase, a 90 s isocratic run time with k ' of ~4 (aziridine) was obtained, allowing for adequate resolution from the void volume containing the unretained and overloaded main components (Figure 6). The benchtop MS manufacturer's suggested default tuning parameters for high eluent flow rates had sufficient sensitivity at the required reporting limit, with little improvement observed by altering them. The method was qualified as a limit test and was able to monitor aziridine and 2-chloroethylamine to support process development.

Figure 6: (a) Ultraviolet (UV) absorbance (AU) trace at 205 nm shows the main component. Trace level monitoring of (b) aziridine (AZ), and (c) 2-chloroethylamine (CEA). Column: Waters BEH HILIC 2.1 mm × 50 mm, 1.7-µm particle size; mobile phase: 10 mM ammonium acetate in acetonitrile:water (95:5); flow rate: 0.5 mL/min; column temperature: 30 °C; detection: selected ion monitoring (SIM) at m/z = 44.07 (AZ) and 79.90 (CEA); injection volume: 2 µL; diluent: 10 mM ammonium acetate in acetonitrile:water (95:5).

Conclusions

HILIC has rapidly evolved into a mature technique that, once understood, can generate methods with comparable robustness when compared to reversed-phase LC. The now broadly available tools and guides are helping to facilitate analytical scientists investigating HILIC for different method development applications. The applicability of HILIC to highly polar compounds can be far more straightforward than alternative approaches, such as GC or reversed-phase LC with derivatization, and can often lead to a simpler final method. Additionally, the ability to run on typical, widely available reversed-phase LC instruments, without having to deal with conversion to normal phase LC mode, enables easier transfer to external laboratories.

In the case examples described here, HILIC was an enabling tool. Although there may be techniques with superior figures of merit, the HILIC methods were fit-for-purpose and readily transferred. For example, derivatization and chromatography or electrophoresis of amino acids could also be employed. However, in our experience, the variable and highly concentrated sample matrices complicate derivatization and electrophoresis equipment is not commonly available across small-molecule quality control networks. Indeed, the absence of derivatization or extraction typically results in more robust processes and significantly reduces sample processing time. When speed to knowledge is the primary endpoint, HILIC can be a lifesaver.

References

(1) R. Majors, LCGC North Am. 30(1), 20–34 (2012).

(2) R.I. Chirita, C. West, A.L. Finaru, and C. Elfakir, J. Chromatogr. A 1217(18), 3091–3104 (2010).

(3) E.S. Grumbach and K.J. Fountain, Comprehensive Guide to HILIC, Waters Corporation, Milford, MA, 2010.

(4) Thermo Fisher Scientific, HILIC Separations Technical Guide P/N TG21003-EN 0714S , (2014).

(5) M. Taraji, P.R. Haddad, R.I.J. Amos, M. Talebi, R. Szucs, J.W. Dolan, and C.A. Pohl, Anal. Chim. Acta 1000, 20–40 (2018).

(6) D.L. Shollenberger and D.S. Bell, LCGC North Am. 34(12), 894–899 (2016).

(7) H.P. Nguyen and K.A. Schug, J. Sep. Sci. 31(9), 1465–1480 (2008).

(8) Impurities in New Drug Substances , ICH Q3A(R2), International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (2006).

(9) Assessment and Control of DNA Reactive (Mutagenic) Impurities in Pharmaceuticals to Limit Potential Carcinogenic Risk, ICH M7(R1), International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (2017).

(10) G. Szekely, M.C.A. De Sousa, M. Gil, F.C. Ferreira, and W. Heggie, Chem. Rev. 115(16) 8182–8229 (2015).

(11) B.A. Olsen, J. Chromatogr. A , 913 (1–2) (2001) 113–122 (2001).

(12) J. Heaton, M.D. Jones, C. Legido-Quigley, R.S. Plumb, and N.W. Smith, Rapid Commun. Mass Spectrom. 25(24) (2011) 3666–3674.

(13) T. Eicher, S. Hauptmann, and A. Speicher, The Chemistry of Heterocycles (Wiley-VCH Verlag & Co., Weinheim, Germany, 2012) pp. 17–37.

(14) L. Verschaeve and M. Kirsch-Volders, Mutat. Res., Rev. Genet. Toxicol. 238(1), 39–55 (1990).

(15) E.M. May, D.C. Hunt, and J. Sloggem, J. Pharm. Biomed. Anal. 5(1) 65–70 (1987).

(16) D.J. Evans, R.J. Mayfield, and I.M. Russell, J. Chromatogr. A 115(2), (391–395 (1975).

(17) P.E. de Haan, D. de Jong, J.H.M. van den Berg, and C.G. Kruse, J. High. Resolut. Chromatogr. 12(9), 604–607 (1989).

(18) P.G. Demertzis and R. Franz, Eur. Food Res. Technol. 204(3), 227–230 (1997).

(19) J. Zapata, J. Temprado, L. Mateo-Vivaracho, and V. Ferreira, J. Pharm. Biomed. Anal. 55(3), 458–465 (2011).

(20) M. Douša, R. Klvaňa, J. Doubský, J. Srbek, J. Richter, M. Exner, and P. Gibala, J. Chromatogr. Sci. 54(2), 119–124 (2016).

Jonathan Shackman is a principal scientist at Bristol-Myers Squibb, in New Brunswick, New Jersey. Direct correspondence to: jonathan.shackman@bms.com