ARTICLE pubs.acs.org/ac
Toward Understanding Molecular Heterogeneity of Polysorbates by Application of Liquid ChromatographyMass Spectrometry with Computer-Aided Data Analysis Oleg V. Borisov,*,† Junyan A. Ji,‡ Y. John Wang,‡ Felix Vega,† and Victor T. Ling† †
Protein Analytical Chemistry Department, and ‡Late Stage Pharmaceutical and Process Development, Genentech, 1 DNA Way, South San Francisco, California 94080, United States
bS Supporting Information ABSTRACT: Polysorbates (PS) are widely used as oil-in-water emulsifiers, stabilizers, wetting agents, solubilizers, and dispersants in the agricultural, food, personal care, and pharmaceutical industries due to their cost effectiveness, biocompatibility, formulation flexibility, low toxicity, and good stabilizing and protecting properties. The polysorbates are often pictured as polyoxyethylated sorbitan monoesters of saturated and/or unsaturated fatty acids. In reality, polysorbates are complex mixtures of multiple components, as follows from the reactions involved in their production. In this work, we report a novel application of liquid chromatographymass spectrometry (LCMS) for the characterization of polysorbates. This method takes advantage of accurate mass measurements and information on the identity of a fatty acid from “in-source” generated characteristic dioxolanylium ions. The method allowed us to perform quick profiling of fatty acids in PS 20 and 80 which, combined with a computer-aided peak assignment algorithm, facilitated detailed characterization of their constituents. As a major finding, we determined that different samples of PS 20 varied from 0% to 15% in relative amounts of unsaturated oleic acid. Although the consequences of this difference were not fully evaluated in this work, one might expect that PS 20 with larger amounts of oleic acid will be more prone to autoxidation, thus potentially having greater impact on the oxidative degradation of the biotherapeutics it formulates.
B
iotherapeutic proteins require proper formulation to preserve their structural integrity over time. Many recombinant proteins, including monoclonal antibodies (MAbs), necessitate formulation concentrations of 100 mg/mL or higher. These concentrations can compromise the chemical and physical stability of proteins in solutions.1,2 Increased intermolecular interactions and perturbations at surfaceliquid and/or airliquid interfaces can result in protein partial unfolding and aggregation, often associated with reduced bioactivity and, even more importantly, increased immunogenicity of therapeutic proteins.35 Various surfactants are commonly used in formulations of biotherapeutic proteins as stabilizers. In particular, nonionic amphiphilic surfactants, polysorbates (often referred in the literature by their trade name as Tweens), are widely used in food and pharmaceutical industries due to their biocompatibility, low toxicity, and good stabilizing properties against physical damage of proteins.1,2,69 According to the European Pharmacopoeia10 and the U.S. PharmacopeiaNational Formulary,11 polysorbates are mixtures of partial esters of fatty acids with sorbitol and its anhydrides, ethoxylated with approximately 20 moles of ethylene oxide (EO) for each mole of sorbitol and sorbitol anhydrides. The amphiphilic nature of polysorbates is provided by a r 2011 American Chemical Society
hydrophilic poly(oxyethylene) (POE) sorbitan headgroup and a lipophilic fatty acid tail group.9 The differences between types of polysorbates are in the nature of their fatty acid side chain and degree of esterification. Two major polysorbates commonly used in the pharmaceutical industry, PS 20 and 80, have monolaurate and mono-oleate as predominant esters, respectively. The amphiphilic nature of polysorbates provides a stabilizing effect on biotherapeutics in formulations by preventing protein losses through surface adsorption and aggregation. This protective action of polysorbates is attributed to their thermodynamically favored coating of hydrophobic airliquid and surface water interfaces.9,12 In some cases, direct interaction of polysorbates with proteins, reported for membrane13 and some other proteins,1,6 dictates selection of a type and concentration of polysorbate for proper protection. In contrast, the colloidal stabilization of immunoglobulins against stress, favored by the presence of polysorbates, is likely not related to a specific immunoglobulinsurfactant binding as this interaction was demonstrated to be minimal.1 Received: March 5, 2011 Accepted: April 14, 2011 Published: April 14, 2011 3934
dx.doi.org/10.1021/ac2005789 | Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
Figure 1. Chemical structure of major expected POE esters in polysorbates.
Commercial polysorbates are chemically diverse polymeric compounds, the polydispersity of which is attributed to the differences in lengths of EO chains. This, in addition to molecular heterogeneity attributed to the existence of esters of several fatty acids with sorbitol anhydrides, makes characterization of polysorbates an analytical challenge. NMR14,15 and several mass spectrometry methods, including OH negative ion chemical ionization16 and, more recently, matrix-assisted laser desorption/ ionization (MALDI) mass spectrometry1719 have been used for direct characterization of polysorbates and products of their degradation. These reports emphasize the molecular complexity of polysorbates and demonstrate the presence of three major polyoxyethylated components based on ethylene glycol and mono- and dianhydrides of sorbitol (sorbitan and isosorbide, respectively). These species are present as polyols and as monoand diesters of fatty acids. Major esters identified in commercial polysorbates are shown in Figure 1. It was calculated20 and demonstrated experimentally18 that polyoxyethylates of uncyclized sorbitol are present only in trace amounts, indicating almost complete dehydration of sorbitol during the manufacturing process. Another commonly reported characteristic of polysorbates of different types and vendors is that they contain on average more than 20 EO units per molecule. Despite the recent advancements in analysis of polysorbates, only limited information is available on batch-to-batch consistencies of polysorbates of the same type.15 Compared to hyphenated techniques, such as LCMS, direct methods of analysis, such as MALDI, will more likely fail to detect some differences due to the overlapping profiles of different species with different ionization efficiencies.18 For example, MALDI spectra were dominated by nonesterified polyols, whereas fatty acid esters, the major components of polysorbates, were present as minor peaks.18,19 Dang et al.15 demonstrated the effectiveness of LCMS as a method for the compositional profiling of two batches of polysorbate 60, manufactured by different processes. More recently, Hewitt et al.21 used LCMS to profile PS 20 and 80 and their degradation products. In this work we report the development of a novel LCMS method of profiling of polysorbates based on the characteristic signal of dioxolanylium ions of fatty acids generated by “in-source” fragmentation. The method offers detection and identification of even minor differences in distribution of the fatty acid esters between batches of polysorbates. An improved
understanding of the molecular heterogeneity of surfactants and ability to screen the incoming raw materials can help with the production of consistent protein formulations.
’ EXPERIMENTAL SECTION Materials. Polysorbates 20 and 80 were obtained from Croda (Rancho Cucamonga, CA) and NOF (Tokyo, Japan). Two grades of polysorbates (Croda) manufactured by standard and super-refined processes were evaluated. For LCMS, 50 mg of polysorbate was dissolved in 5 mL of water in volumetric flasks to give 10 mg/mL stock solutions, stored at 28 °C and protected from light. All solvents including water and acetonitrile were HPLC grade and were purchased from Burdick and Jackson (Muskegon, MI). Formic acid (FA) was obtained from Pierce (Rockford, IL). HPLC Separation of Polysorbates. Reversed-phase (RP) separation was performed on an Agilent 1100 series HPLC system (Santa Clara, CA). A Vydac 214MS (Grace, Deerfield, IL) C4, 2.1 mm 250 mm column packed with 300 Å pore-sized 5 μm particles was operated at 25 °C and a flow rate of 0.250 mL/ min. At initial condition the column was held at 95% of mobile phase (MP) A (0.1% FA in water) and 5% of MP B (0.1% FA in acetonitrile). The analytical gradient consisted of a quick 0.1 min ramp at 5 min to 20% MP B, followed by a linear ramp to 30% MP B at 15 min, a ramp to 85% MP B at 45 min, and to 100% MP B at 50 min, followed by a column wash for 15 min and return to initial conditions at 65.1 min. Initial conditions were held for 10 min for column re-equilibration. Typically, 0.8 μg of aqueous PS solution was loaded onto the column. Mass Spectrometry. Analyses were performed using a Waters QTOF Premier (Milford, MA) quadrupole time-of-flight mass spectrometer equipped with the LockSpray dual-ionization source. The instrument was operated in a positive ionization mode with electrospray voltage at 3.5 kV, source block and desolvation temperatures at 125 and 200 °C, respectively, and desolvation nitrogen gas flow rate at 900 L/h. Prior to analyses, the instrument was externally calibrated using a sodium/cesium iodide calibration mixture (Waters). Improved mass measurement accuracy was achieved by aspiration of a 1 pM/μL in 50% acetonitrile and 0.1% FA in water solution of (Glu1)-fibrinopeptide B (Bachem, Torrance, CA) via the reference sprayer. 3935
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
This LCMS method included two MS functions with cone voltages of 18 and 45 V, promoting “in-source” fragmentation. Full-scan spectra were collected over the m/z range of 1502000 for both functions. In additional experiments, collision-induced dissociation (CID) fragmentation was performed on selected ions with collision cell argon gas pressure of 6.5 mbar. MassLynx v.4.1 software (Waters) was used to control the instrument and for data processing. Data Analysis. The polydispersity of polysorbates stipulates the presence of homologous ion series in their mass spectra. The assignment of these series to the major structures (Figure 1) was conducted by determining a value, k, from the following equation: Mexp Mcore zMOH k¼
∑i MFAi
MEO
ð1Þ
where Mexp is the experimental molecular weight, assuming that the observed molecular ion could have either proton, sodium ion, or potassium ion as an adduct for singly charged ions, or any combination of the above for doubly charged species, Mcore is the formula weight of the “core” for the structures based on sorbitan (96.0575 Da) or isosorbide (112.0524 Da), z is the number of terminating hydroxyl groups, MOH and MEO are formula weights of the hydroxyl group (17.0027 Da) and the EO moiety (44.0262 Da), respectively, MFA is the formula weight of a fatty acid, and integer i is either 1 or 2 in case of a mono- or a diester, respectively (tri- and tetraesters were not considered in this calculation). Approximation of the number of EO units, k0, by rounding value k to the nearest integer was performed for each of the esters in Figure 1. Since the molecular weight, MW, of the species in a homologous series is a discrete function of k, its residual value, ΔMW, resulting from rounding of k to k0, determines the error of the molecular weight estimation. This error was calculated in ppm from the following equation: errorppm ¼
ΔMW MOH jΔkj 106 ¼ 106 Mexp Mexp
ð2Þ
where |Δk| = |k k0| is the absolute difference between k, calculated by eq 1, and its rounded-to-nearest-integer value, k0. Structures were then ranked by their ppm errors, and a structure with the smallest value (within the typical mass measurement error of the QTOF instrument) was considered to be the most probable. A custom program, written in Visual Basic for Applications (VBA) supplied with Microsoft Excel, was developed to automate this task.
’ RESULTS AND DISCUSSION RP Separation of Polysorbate 20 Components. A general understanding of the composition of commercial polysorbates can be derived from considering the typical manufacturing processes.15,22,23 The first step, anhydrization of sorbitol, is an acid-catalyzed process producing a mixture of sorbitans dominated by 1,4-sorbitan and 1,4:3,6-sorbitan (or isosorbide), with 2,5-sorbitan, 3,6-sorbitan in smaller amounts. Then, the anhydro sorbitols react, in the presence of base, with fatty acids, producing esters distributed among the various available hydroxyl groups. In the last step, the oxyethylation is conducted under conditions promoting ester interchange, resulting in more or less random addition of EO to the hydroxyl groups. Each step of the
manufacturing generates a mixture of products contributing to the overall molecular heterogeneity of polysorbates. Thus, strictly speaking, a commonly used formula of polysorbates as being POE 1,4-sorbitan monoesters of fatty acids is an oversimplification, only showing the tip of the iceberg, as these monoesters account for only about 30% of the total content of various polysorbates.20 A representative total ion current (TIC) chromatogram of PS 20 in Figure 2a shows two distinct regions with a series of overlapping peaks eluting around 16 min and a cluster of partially resolved peaks eluting after 25 min. The hydrophobicity of PS is attributed to the presence of fatty acids. The most hydrophilic components of PS eluting under the peak at 16 min contained three overlapping ion envelopes, identified as sodiated species of the major polyols (POE, POE isosorbide, and POE sorbitan). Although the heterogeneity of this peak is not immediately clear from the TIC, it can be discerned by reconstructing the elution profiles of the individual polyoxyethylates. As shown in the inset to Figure 2a, these polyols elute in the order POE < POE isosorbide < POE sorbitan. Previously, Dang et al.15 identified POE sorbitan under the early eluting peak of LCMS profile of PS 60. Nonesterified polyols were also reported as a dominating species in the MALDI spectra of PS 20.18 These polyols lack amphiphilic properties and thus add no value to formulation. It is likely that the majority of these polyols exists as manufacturing byproducts due to incomplete esterification and/or, even more likely, ester interchange reactions during oxyethylation. Thus, it is reasonable to expect that the above oxyethylated polyols are also present as esters of fatty acids. In addition, the hydrolysis of fatty acid esters, the well-known mechanism of PS degradation,24 can partly account for the presence of the above polyols and growth of the early eluting peak in aged samples (data not shown). Identification of Fatty Acid Esters with “In-Source” CID. The vast majority of published data on the analysis of polysorbates indicates a highly heterogeneous nature with significant diversities in chemical structure and concentration of esterified species, challenging a detailed characterization of these surfactants. A simplification of the peak assignment process would be highly desirable. Examination of mass spectra under the peaks eluting after 25 min revealed a prominent series of low m/z ions with a mass difference of 28 Da, corresponding to two methylene groups and consistent with the mass difference between fatty acids. These low m/z ions likely form readily “in-source” even during soft electrospray ionization (ESI) process and are characteristic of particular fatty acid esters, making them useful for obtaining a quick fatty acid profile polysorbates. Dang et al.15 reported m/z 283 and 311 ions in the spectra of major components of PS 60, respectively assigned to ions of ethyl palmitate and stearate, and explained as final products from spontaneous “in-source” fragmentation of POE esters. Similarly, the formation of diagnostic ions, revealing the nature of the fatty acid ester, was observed in the CID spectra of triethanolaminebased esterquat fabric softener.25 In that case, the loss of the tertiary amine yielded cyclic fatty dioxolanylium ions. In this work, POE sorbitan monolaurate, a major component of PS 20, eluting at 32.1 min (Figure 2a), produced a spectrum with abundant m/z 227.20 ions (i.e., laurate with two methylene groups). This agreed with the notion that the presence of this ion is characteristic to a laurate-containing species. In a more detailed study, the origin of these diagnostic ions was examined by subjecting several laurate-containing species to CID. Tandem 3936
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
Figure 2. Representative total ion current (TIC) profile of PS 20 (a), heat map generated from the reconstructed ion chromatograms (RICs) of characteristic low m/z dioxolanylium ions of fatty acids formed during “in-source” fragmentation (b), and RIC for m/z 227.20 (2-undecyl-1,3dioxolanylium ion) indicating the elution of laurates (c), with major peaks labeled as POE sorbitan monolaurate (1), POE isosorbide monolaurate (2), POE sorbitan mixed diesters (37), POE sorbitan trilaurate (8), and POE sorbitan tetralaurate (9).
mass spectrometry (MS/MS) spectra of ions with m/z values matching doubly sodiated ions of POE (26) sorbitan monolaurate, dilaurate, and mixed laurate/mayristate are shown in Figure 3. Common to these spectra is the presence of chargereduced (M þ Na)þ ions, likely formed by expulsion of Naþ. This feature is not typically observed in spectra of protonated ions. However, the tendency for an alkali ion to be expelled from doubly charged ions of peptides was reported to increase with the size of the adduct ion.26 Low m/z regions of the CID spectra were dominated by m/z 227.20 and 255.23 ions, confirming the presence of lauric and myristic acids in the parent ions in Figure 3. It should be noted that these ions were the major fragments, produced by cleavage of a covalent bond during CID of the doubly sodiated ions of POE sorbitan esters, suggesting the formation of resonance-stabilized cyclic dioxolanylium ions of fatty acids (Figure 3d). The above results allowed us to conclude that occurrence of characteristic low m/z ions indicates the presence of a particular fatty acid ester. LCMS performance was further improved by addition of a second scan function with elevated “in-source” potentials to promote partial fragmentation of polysorbate constituents. Reconstructing the LCMS profile for m/z 227.20 from
the second function gave an elution profile of all the lauratecontaining species, indicating that there are at least nine chromatographically resolved of such species in PS 20 (Figure 2c). Spectra under several peaks in Figure 2c (peaks 1, 2, and 5) were examined as shown in Figure 4. Knowledge of the identity of the fatty acid eluting under these peaks significantly facilitated structural assignments conducted by solving eqs 1 and 2. The spectrum under peak 1 (Figure 4b) contained three charge-state envelopes of sodiated POE sorbitan monolaurate with mass accuracy within 5 ppm for the majority of the ions in homologous series. Several minor ion series, positioned within these envelopes, were identified as being due to proton, sodium, and potassium adducts, and/or a combination of the above depending on charge state. Peak 2 was determined to contain overlapping profiles of two species: POE isosorbide monolaurate and POE monolaurate (Figure 4, parts c and d) with only partial chromatographic resolution (Figure 4a). It is important to note that these two species exhibited very similar polydispersity profiles with average numbers of EO units of about 12 as shown in Figure 4, parts c and d. Interestingly, the presence of POE monoesters in polysorbates has not been reported previously, but their logical existence follows from a consideration of the 3937
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
Figure 3. Collision-induced dissociation (CID) spectra (ac) of doubly charged sodiated ions of POE (26) sorbitan monolaurate (a), POE (26) sorbitan dilaurate (b), and POE (26) sorbitan laurate/myristate (c) with m/z 768.46, 859.54, and 873.56, respectively, and the proposed mechanism of formation of dioxolanylium ions (d) during CID experiments.
manufacturing process. It is likely that the majority of POE esters form during the manufacturing process, rather than by being cleaved from POE anhydro sorbitol esters, a known path for oxidative degradation of polysorbate.9,24,27 Peak 5 (Figure 4e) was identified as the POE sorbitan dilaurate with polydispersity similar to that of POE sorbitan monolaurate (Figure 4b), suggesting that the higher-order esters are forming as a result of ester interchange in the course of oxyethylation. Examination of the elution profiles of POE anhydro sorbitol esters clearly exhibited additional degrees of heterogeneity in these species. For example, POE (11) isosorbide monolaurate eluted under at least three partially resolved peaks (Figure 4a). A similar phenomenon was observed for POE sorbitan and isosorbide polyols (Supporting Information, Figure S-1). In each case, spectra for the species under these peaks (data not shown) exhibited no mass difference and very similar polydispersity profiles, pointing to the occurrence of isomers, although their nature was not determined in this work. Data gleaned from “in-source” fragmentation greatly simplified characterization of all fatty acid esters in a sample. The heat map in Figure 2b presents reconstructed elution profiles for the major detected fatty acids in PS 20. In general, species with a particular fatty acid eluted in the following order: POE sorbitan monoester < POE isosorbide monoester e POE monoester < diesters < tri- and higher-order esters. Furthermore, for each of the fatty acids, multiple di- and higher-order esters were detected. For example, myristic and lauric acids gave a characteristic signal
at 43.8 min (peak 6 in Figure 2b) indicating the presence of POE sorbitan mixed laurate/myristate diester, as confirmed by eqs 1 and 2, with a mass accuracy of 6 ppm. Owing to such a complexity in the diesters, corresponding LC peaks were dominated by POE sorbitan species, but the presence of POE isosorbide diesters was also observed upon careful examination. A detailed list of species in PS 20 detected in this work is given in Table 1. Comments on the Polydispersity of Oxyethylation. The polydispersity of POE sorbitan monolaurate is shown in an inset to Figure 4b. In accord with published data,15,1719,21 polysorbates studied in this work contained on average more than 20 EO units per molecule of POE sorbitan monoester. This often causes controversy in the literature since polysorbates are frequently pictured by a structural formula of POE (20) sorbitan fatty acid ester.11 In fact, this single formula does not adequately represent the real emulsifier and even contradicts with the formal definition of polysorbates as esters of sorbitol and its anhydrides copolymerized with approximately 20 moles of ethylene oxide for each mole of sorbitol and sorbitol anhydrides.10,11 This issue was raised by Brandner20 in his 1998 manuscript, where the author points to the complexity of polysorbates, inherited from their manufacturing processes, during which anhydrization of sorbitol leads to a mixture of sorbitans and isosorbides, i.e., species with four and two reactive hydroxyl groups, respectively. Thus, reaction of 1 mol of the above mixture with 20 mol of EO produces oxyethylates with an average number of EO moieties per reactive hydroxyl site of greater than 5. 3938
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
Figure 4. Portion of total ion current (TIC) of PS 20 (a, top trace) and reconstructed ion chromatograms (RICs) (a, bottom traces) for various monolaurates (a) and the spectra under the 31.5 (b), 32.8 (c), 33.1 (d), and 41.2 min (e) peaks, identified as POE sorbitan monolaurate (b), POE isosorbide monolaurate (c), POE monolaurate (d), and POE sorbitan dilaurate (e).
Table 1. LCMS Characterization of the Major Peaks in PS 20 (Figure 2a) RT, min
identification
16.0
POE (919), POE (1023) isosorbide, POE (1737) sorbitan
27.4
POE sorbitan monocaprylate, POE isosorbide monocaprylate, POE monocaprylate
29.8 30.8
POE sorbitan monocaprate POE isosorbide monocaprate
32.1
POE (1738) sorbitan monolaurate
33.4
POE (1022) isosorbide monolaurate, POE (919) monolaurate
34.3
POE sorbitan monomyristate
36.1
POE isosorbide monomyristate, POE monomyristate
36.5
POE sorbitan monopalmitate, POE sorbitan mono-oleate
38.3
POE sorbitan caprylate/laurate, POE isosorbide monopalmitate, POE sorbitan monostearate, POE isosorbide mono-oleate
40.0 41.7
POE sorbitan caprate/laurate, POE isosorbide monostearate, POE sorbitan caprylate/myristate POE sorbitan dilaurate, POE sorbitan caprate/myristate
43.8
POE sorbitan laurate/myristate
45.9
POE sorbitan laurate/palmitate, POE sorbitan laurate/oleate
49.8
POE sorbitan trilaurate (as dominating species)
55.0
POE sorbitan trimyristate, POE sorbitan tetralaurate, other high-order mixed esters
For simplicity, we assumed that the relative amounts of hydrophilic polyols under the 16 min peak (Figure 2a) are representative of
all the species in polysorbate. From the peak areas under the reconstructed chromatograms (inset to Figure 2a) it was estimated 3939
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
that the molar fraction of species with four hydroxyl groups (sorbitan) and two reactive sites (isosorbide and ethylene glycol) was ca. 0.49 and 0.51, respectively. Then, assuming similar reactivities of all the hydroxyl groups toward oxyethylation, reaction of 20 mol of EO with 1 mol of the mixture generates POE chains with on average ca. 6.7 EO units per arm. Interestingly, based solely on PS 20 analytical constants given in the National Formulary, Brandner determined that the “correct” number of EO units reacted with each of hydroxyls is 6.4, so on the average the monoanhydride portion (sorbitans) contains 26 EO units, and the dianhydrides portion is 13.20 Table 2. Distribution of POE Anhydro Sorbitol Esters in PS 20 POE sorbitan esters method
mono-
RIC m/z 227 (this work)
75.9
Brandner (ref 20)
78.6
POE isosorbide ester mono24.1 21.4
mono-
di-
tri-
RIC m/z 227 (this work)
43.0
36.5
20.5
Brandner (ref 20)
49.6
37.7
12.7
Indeed, POE sorbitan laurate exhibited on average 26 EO units (Figure 2b, inset), whereas POE isosorbide and POE laurates both contained 1213 units per molecule (Figure 2). Moreover, the average number of EO units was found to be largely independent of the ester nature and the degree of esterification as can be seen from Figure 4, parts b and e. The distribution of EO units in different types of species seems to depend solely on the number of free hydroxyl groups in the corresponding polyols. The distribution of EO chain lengths during oxyethylation has been shown to approximate a Poisson distribution with the assumption that the number of propagating molecules remains constant throughout the polymerization and that the rate of oxyethylation is equal regardless of a polymer length.23,28 The distribution predicted for the reaction of 25 mol of EO with 1 mol of sorbitan fits experimental data as shown by the solid line in the inset to Figure 4b. The discrepancies are probably due to the fact that sorbitans with increasing POE chain length are not quite equally reactive toward EO. Similarly, Birkmeier and Brandner28 proposed a slight decrease of EO addition rate constants with increasing polymer length, resulting in a sharper peak of EO distribution in polyglycol stearates compared to a distribution described by a Poisson fit. Distribution of Fatty Acids in Polysorbate 20 and 80. Gas chromatographic (GC) analysis of released and methylated fatty acids is a classical method for the determination of fatty acid
Figure 5. Distribution of major fatty acid esters in PS 20 (a) and 80 (b) based on peak areas of corresponding dioxolanylium ions under the peaks of POE sorbitan monoesters, measured during LCMS analysis under promoting “in-source” CID conditions. The inset shows the details on the distribution of C18 fatty acids in the tested PS 20 lots. 3940
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry composition in polysorbates, recommended by European and U. S. Pharmacopoeias for quantitative analysis of these emulsifiers. However, this method lacks species-specific information, since it requires the fatty acids to be saponified from the esters. In this work LCMS was used to estimate the distribution of laurates in their major chromatographically resolved peaks in PS 20 (Figure 2c), reconstructing the chromatogram for m/z 227 dioxolanylium ion. Due to the potential differences in ionization and “in-source” fragmentation efficiencies of various species, such data should only be treated as a semiquantitative. However, it may become useful for batch-to-batch comparison studies, for example. Table 2 compares these results with the ones recalculated from previously published by Brandner,20 whose calculations were based solely on the bulk analytical constants of PS 20 without actual measurements of the amounts of each individual fraction. Nevertheless, the two sets of data are in a reasonable agreement, indicating that, although the POE sorbitan monolaurate is the major component, PS 20 contains significant amounts of other species. Reconstructed profiles of characteristic dioxolanylium ions were also used to estimate the distribution of fatty acids in PS 20 and 80, as summarized in Figure 5, parts a and b. For simplicity, relative amounts were determined based only on the POE sorbitan monoester portion of the samples. Although these results were not verified by the GC analysis, the data are within the ranges listed in the European Pharmacopoeia.10 As expected, laurates and oleates are the major components of PS 20 and 80, respectively. On the basis of multiple analysis of the same PS 20 sample, reproducibility of the method gave relative standard deviations of below 5% for major fatty acids, increasing to about 20% for fatty acids with relative amounts of below 5%. Polysorbate 20 exhibited similar distributions of fatty acids between samples of the same manufacturing process. However, greater differences were observed between samples produced by different processes and manufacturers. Interestingly, the major difference between the emulsifiers produced by Croda’s standard and super-refined processes was in the distribution of C18 fatty acids with significantly higher amounts of unsaturated oleic acid in the super-refined PS 20. The distribution of fatty acids in PS 20 from NOF 20 was found to resemble that of super-refined material from Croda. The relative amount of oleic acid in the tested lots of PS 20 ranged from 0% to 15% as shown in Figure 5a. We speculate that these differences can potentially have important implications on the properties of PS 20 in biotherapeutic formulations. Autoxidation of polysorbates has recently become a great concern for formulation development of biotherapeutics, and it has been linked to one of the major routes of oxidative degradation of proteins.7 It is expected that oxidative degradation of proteins will be affected by the degree to which formulation excipients are susceptible to autoxidation. The oxidizability constant of PS 80 was determined to be 2.65 times greater, compared to that of PS 20.27 It is the presence of large amounts (about 80%) of unsaturated oleic acid in PS 80 that can explain this difference.9,27 Considering the natural susceptibility of unsaturated fatty acids to autoxidation, we raise a question of whether emulsifiers with higher contents of oleic acid can have a stronger effect on the oxidation of biotherapeutic proteins they formulate, compared to surfactants with lower amounts of unsaturated acids. This might be an interesting subject for a future investigation.
ARTICLE
’ CONCLUSIONS PS 20 and 80 are widely used in formulations of biotherapeutics as nonionic amphiphilic surfactants, protecting proteins from surface adsorption and aggregation. As shown in this and previous reports, polysorbates are complex mixtures of various components of polyoxyethylated sorbitol anhydrides and their mono- and higher-order esters with fatty acids. Such molecular heterogeneity is the expected result of the manufacturing process. In fact, a good understanding of the manufacturing process can easily explain the presence of each of the species and predict their polydispersity as was shown by Brandner.20 Although Brandner published his results over a decade ago, our experimental data on the distribution of major species in PS 20 were in good agreement with his calculations. This points to a good production consistency of polysorbates throughout the years. In this study, we applied LCMS to probe the molecular heterogeneity of polysorbates. Significant simplification was achieved by monitoring the signals of the characteristic dioxolanylium ions of fatty acids, promoted by the use of mild “insource” CID conditions. This method was successfully applied to profiling the distributions of the major fatty acids in polysorbates. Precise knowledge of the identity of a fatty acid, accompanied by accurate mass measurements and the computer-aided structure ranking process, facilitated characterization of polysorbates. It is worth mentioning that upon examination of multiple production lots of PS 20, we observed statistically significant differences in relative amounts of species with unsaturated oleic acid, ranging from 0% to 15%. As a result of this comprehensive analysis, previously unreported species of POE fatty acid esters were identified, the presence of which logically follows from considering the manufacturing process. Although this LCMS method was not verified for being quantitative, it can be useful for fast profiling of polysorbates where semiquantitative information is sought with potential for use in monitoring the consistency of raw materials and for comparability screening. ’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank Dan Hewitt of Genentech, Inc. for providing polysorbate samples and helpful discussions. ’ REFERENCES (1) Garidel, P.; Hoffmann, C.; Blume, A. Biophys. Chem. 2009, 143, 70–78. (2) Patapoff, T. W.; Esue, O. Pharm. Dev. Technol. 2009, 14, 659–664. (3) Cromwell, M. E. M.; Hilario, E.; Jacobson, F. AAPS J. 2006, 8, E572–E579. (4) Wang., W.; Singh, S.; Zeng, D. L.; King., K.; Nema, S. J. Pharm. Sci. 2007, 96, 1–26. (5) Wang, W. Int. J. Pharm. 2005, 289, 1–30. 3941
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
Analytical Chemistry
ARTICLE
(6) Deechongkit, S.; Wen, J.; Narhi, L. O.; Jiang, Y.; Park, S. S.; Kim, J.; Kerwin, B. A. J. Pharm. Sci. 2009, 98, 3200–3217. (7) Wang, W.; Wang, Y. J.; Wang, D. Q. Int. J. Pharm. 2008, 347, 31–38. (8) Chou, D. K.; Krishnamurthy, R.; Randolph, T. W.; Carpenter, J. F.; Manning, M. C. J. Pharm. Sci. 2005, 94, 1368–1381. (9) Kerwin, B. A. J. Pharm. Sci. 2008, 97, 2924–2935. (10) European Pharmacopoeia, 7th ed. [Online]; European Directorate for the Quality of Medicines & HealthCare, Council of Europe, Strasbourg, France, http://online6.edqm.eu/, accessed March 1, 2011. (11) The United States PharmacopeiaNational Formulary [Online]; USP 33-NF 28; The United States Pharmacopeial Convention, Rockville, MD, http://www.uspnf.org/, accessed March 1, 2011. (12) Randolph, T. W.; Jones, L. S. Pharm. Biotechnol. 2002, 13, 159–175. (13) Bowie, J. U. Curr. Opin. Struct. Biol. 2001, 11, 397–402. (14) Khossravi, M.; Kao, Y.-H.; Mrsny, R. J.; Sweeney, T. D. Pharm. Res. 2002, 19, 634–639. (15) Dang, H. V.; Gray, A. I.; Watson, D.; Bates, C. D.; Scholes, P.; Eccleston, G. M. J. Pharm. Biomed. Anal. 2006, 40, 1155–1165. (16) Brumley, W. C.; Warner, C. R.; Daniels, D. H.; Andrzejewski, D.; White, K. D.; Min, Z.; Chen, J. Y. T.; Sphon, J. A. J. Agric. Food Chem. 1985, 33, 368–372. (17) Frison-Norrie, S.; Sporns, P. J. Agric. Food Chem. 2001, 49, 3335–3340. (18) Ayorinde, F. O.; Gelain, S. V.; Johnson, J. H., Jr.; Wan, L. W. Rapid Commun. Mass Spectrom. 2000, 14, 2116–2124. (19) Raith, K.; Schmelzer, C. E. H.; Neubert, R. H. H. Int. J. Pharm. 2006, 319, 1–12. (20) Brandner, J. D. Drug Dev. Ind. Pharm. 1998, 24, 1049–1054. (21) Hewitt, D.; Alvarez, M.; Robinson, K.; Ji, J.; Wang, Y. J.; Kao, Y.-H.; Zhang, T. J. Chromatogr., A 2011, 1218, 2138–2145. (22) Stockburger, G. J. Process for Preparing Sorbitan Esters. U.S. Patent 4,297,290, July 17, 1980. (23) Stockburger, G. J. J. Am. Oil Chem. Soc. 1979, 56, 774A–777A. (24) Donbrow, M.; Azaz, E.; Pillersdorf, A. J. Pharm. Sci. 1978, 67, 1676–1681. (25) Saraiva, S. A.; Abdelnur, P. V.; Catharino, R. R.; Nunes, G.; Eberlin, M. N. Rapid Commun. Mass Spectrom. 2009, 23, 357–362. (26) Tang, X.-J.; Thibault, P.; Boyd, R. K. Org. Mass Spectrom. 1993, 28, 1047–1052. (27) Yao, J.; Dokuru, D. K.; Noestheden, M.; Park, S. S.; Kerwin, B. A.; Jona, J.; Ostovic, D.; Reid, D. L. Pharm. Res. 2009, 26, 2303–2313. (28) Birkmeier, R. L.; Brandner, J. D. J. Agric. Food Chem. 1958, 6, 471–475.
3942
dx.doi.org/10.1021/ac2005789 |Anal. Chem. 2011, 83, 3934–3942
