oxidative degradation of polysorbate surfactants studied...

RESEARCH ARTICLE – Pharmaceutical Biotechnology

Oxidative Degradation of Polysorbate Surfactants Studied by LiquidChromatography–Mass Spectrometry

OLEG V. BORISOV,1 JUNYAN A. JI,2 Y. JOHN WANG2

1Protein Analytical Chemistry, Genentech, South San Francisco, California 940802Late Stage Pharmaceutical and Process Development, Genentech, South San Francisco, California 94080

Received 10 July 2014; revised 3 November 2014; accepted 2 December 2014

Published online 11 January 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24314

ABSTRACT: Polysorbates (PSs), as acquired from manufacturing processes and chemical nature of fatty acids (FAs) used in productionof biotherapeutic formulations, are heterogeneous mixtures of structurally related compounds, covering a wide range of physicochemicalproperties. Such complexity presents a certain challenge for analysis of these important surfactants and demands the use of methods offeringsufficient resolution to monitor individual classes of species and detect changes upon stress. A liquid chromatography mass spectrometrymethod, benefiting from the use of low m/z marker ions, simplifies profiling of PSs by providing detailed information on FA compositioneven of chromatographically overlapping peaks. The ability of the method to monitor individual components and follow their changesbecause of oxidative stress was explored. A water-soluble azo compound was used as a model oxidizer. Major degradation products of PS80, because of reactions involving double bond, were identified as oxo-C9:0, keto-C18:1, hydroxyl-C18:1, epoxy-C18:0, and hydroperoxy-C18:1. Stability of PS 20 components was found to depend on the carbon number of polyethoxylated (POE) sorbitan FA ester and its order.Rates of oxidative degradation increased with the length of the FA ester and, moreover, POE sorbitan diesters degraded significantly fasterin comparison to the corresponding monoesters upon the oxidative stress. C© 2015 Wiley Periodicals, Inc. and the American PharmacistsAssociation J Pharm Sci 104:1005–1018, 2015Keywords: analytical chemistry; polymers; liquid chromatography; mass spectrometry; protein formulation; polysorbate; oxidative degra-dation; excipient; surfactant

INTRODUCTION

Polysorbates (PSs) are nonionic surfactants that have many in-dustrial applications and are widely used in the preparationof cosmetics and food products. because of their biocompatibil-ity, low toxicity, and good stabilizing, emulsifying, and wettingproperties, PSs are important excipients in the pharmaceuticalindustry.1–4 There are a number of commercially available PSs;however, PS 20 and PS 80 are the two most commonly used sur-factants in formulations of biotherapeutic proteins.5–7 PSs aredescribed as mixtures of partial esters of fatty acids (FAs) withsorbitol and its anhydrides, ethoxylated with approximately 20moles of ethylene oxide (EO) for each mole of sorbitol and sor-bitol anhydrides.8,9 Inherently, this formal definition implies ahigh degree of heterogeneity of these surfactants, which is asso-ciated with the processes involved in their production.10,11 Themolecular heterogeneity of a PS comes from variability withinits three functional groups, that is, polydispersity with regardto the length of EO chains in the hydrophilic head, differencesin chain length (carbon number) and degree of unsaturationof a FA in the hydrophobic tail, and existence of species withcores consisting of sorbitol mono- and di-anhydrides (sorbitans

Abbreviations used: AAPH, 2,2′-azobis(2-methylpropionamidine) dihydrochlo-ride; CID, collision-induced dissociation; EO, ethylene oxide; FA, fattyacid; PS, polysorbate; POE, polyethoxylated; RP, reversed-phase; LC–MS,liquid chromatography–mass spectrometry; TIC, total ion current.

Correspondence to: Oleg V. Borisov (Novavax, Inc, 20 Firstfield Dr, Gaithers-burg, MD 20878; Telephone: +240-268-2168; E-mail: [email protected])

This article contains supplementary material available from the authors uponrequest or via the Internet at http://onlinelibrary.wiley.com/.

Journal of Pharmaceutical Sciences, Vol. 104, 1005–1018 (2015)C© 2015 Wiley Periodicals, Inc. and the American Pharmacists Association

and isosorbides, respectively).12 Furthermore, depending on thenumber of FA esters per structure, PSs can contain mono-, di-,tri-, and tetra-esters. For simplicity, however, it is customary toconsider PS 20 and PS 80 as polyethoxylated (POE) sorbitanmonoesters of lauric and oleic acids, respectively.9

Brandner12 previously estimated that POE sorbitan mono-,di-, and tri-esters were the major components of PS surfac-tants. These estimates were calculated based on the bulk an-alytical constants of PSs, such as hydroxyl and saponificationvalues, yields of polyols and FAs upon saponification, and mo-lar weights of FAs, rather than actual measurements of theamounts of each individual fraction.12 For example, in PS 20and PS 80, the weight percentages of POE sorbitan mono-esterswere calculated to be 32% and 28%, respectively. A summary ofthe compositions of various PSs, as reported by Brandner,12 isprovided in Table 1. In fact, the complex composition of the var-ious PSs is presumed to be the reason why they are excellentsurfactants having many applications.12

During storage, aqueous solutions of PSs may undergodegradation via two possible mechanisms: ester hydrolysis,which leads to liberation of free FA; and autoxidation, whichcan be accelerated in the presence of light and/or a transitionmetal catalyst.13 Both pathways can potentially have negativeeffects on the stabilizing properties of a PS used in a biother-apeutic formulation. Free FAs formed during ester hydrolysiswould be largely solubilized in micelles up to a certain satu-ration point,13 beyond which, the parenteral product could berejected because of visible particles detected during inspection.This, in addition to an overall decrease in a number of micelles,can result in a lowering of the cloud point (the temperatureabove which a solution becomes turbid) and solution turbidity

Borisov, Ji, and Wang, JOURNAL OF PHARMACEUTICAL SCIENCES 104:1005–1018, 2015 1005

1006 RESEARCH ARTICLE – Pharmaceutical Biotechnology

Table 1. Calculated Weight Percentage Composition of Polysorbates12*

Polyoxyethylene Sorbitan as Polyoxyethylene Sorbidea as

Polysorbate Polyol Mono-Ester Di-Ester Tri-Ester Tetra-Ester Polyol Mono-Ester Di-Ester

20 15.7 32.0 24.3 8.2 1.0 7.8 8.7 2.440 15.7 35.8 29.5 10.6 1.5 2.6 3.3 1.060 12.5 33.3 32.9 14.1 2.2 1.7 2.5 0.980 8.5 28.1 34.1 17.8 3.4 2.2 4.2 1.8

*Reproduced with permission via Copyright Clearance Center from Brandner JD. 1998. Drug Dev Ind Pharm 24:1049–1054.aThe term isosorbide is used in the manuscript, which is equivalent to sorbide used in the original article by Brandner.12

at room temperature. This was demonstrated experimentallyby observing the differences in the cloud points of PS 20 andcetomacrogol (a nonionic surfactant of the polyethylene gly-col family that lacks an ester linkage) upon their short-termstorage at 25◦C–50◦C.13 The absence of a FA ester linkage inthe structure of cetomacrogol protects it from hydrolysis andensures cloud-point stability during storage at those tempera-tures. In contrast, lowering the cloud point in PS 20 solutionsthat are stored at near room temperatures has been explainedby the release of a free lauric acid (a major FA ester of PS 20).13

The propensity of PSs to produce peroxides in situ as theresult of autoxidation has become increasingly relevant be-cause of the wide use of these surfactants in biopharmaceu-tical formulations.4,5 These peroxides can, in turn, adverselyaffect stability of formulated proteins, causing their oxidativedegradation.7,14,15 Thus, the excellent stabilization propertiesof PSs on proteins in solutions can be overshadowed by theirsusceptibility to autoxidation. This dual effect of PS 80 on thestability of a model protein was reported by Wang et al.16

In general, hydroperoxides are the primary products of au-toxidation of many organic compounds. Autoxidation of lipidscontaining saturated and unsaturated FAs has been exten-sively studied, and its free radical-initiated mechanism, con-sisting of chain initiation, propagation, and termination steps,is well established.17 Briefly, the process begins with an initi-ation step that involves hydrogen abstraction that generatesreactive-free radical intermediates,18 that quickly react withmolecular oxygen to produce peroxy radicals. This, in turn, ab-stracts a hydrogen atom intra- or intermolecularly from an-other site to provide another carbon-centered radical, thus en-abling chain propagation. Both the initiation and propagationH-abstraction steps are rate limiting because the oxygen ad-dition is at or near the diffusion-controlled rate under typi-cal experimental conditions. Hydroperoxides are the primaryproducts of autoxidation, stabilities of which are governed bytheir position within a molecule.19 The formation of hydroper-oxides of oleic acid is the direct result of an attack of peroxidesor peroxy radicals on the double bond,20,21 during which hy-droperoxy group can appear at positions C8, C9, C10, and C11,as reviewed by Porter et al.22 Hyroperoxides are prone to sub-sequent decomposition reactions producing various secondaryoxygenated products, the nature of which depends on the start-ing compound and conditions. Formation of cis- and trans-9,10-epoxystearates, in which an oxirane ring replaces existing dou-ble bond of oleic acid, are well-documented products of methyloleate autoxidation.21,23–26 Degradation products with keto andhydroxy functions were also reported to form as secondary prod-ucts of the autoxidation of FA methyl esters.23 For example,reaction mixture of autoxidized methyl oleate contained about

30%–35% peroxide, 25%–30% hydroxy compounds, 20%–25%epoxy compounds, and 15%–20% other unsaturated carbonylcompounds.20

Hydroperoxide degradation can be either acid or base cat-alyzed, or because of a homolytic cleavage of the O–O bond,which ensures a great diversity of degradation products.27 Itwas demonstrated that a variety of secondary species includingesters, formates, aldehydes, and products with a shorter POEchain as a result of its scission are generated during autoxida-tion of even relatively simple POE surfactants.19,27 In light ofthis discussion, it is not surprising that for complex moleculessuch as PS 20 and PS 80 a comprehensive list of degradationproducts, detected by a battery of appropriate analytical tech-nologies, was recently reported by Kishore et al.28 to containabout 60 different species.

Because ether subunits dominate the structure of a PSmolecule, autoxidation at EO moieties explains the formation ofperoxides, changes in acidity, and production of short-chain or-ganic acids in PS 20.13 In general, autoxidation of polyethers inthe presence of air follows the "-activation mechanism, whereether oxygen atoms have the activating effect on neighboringcarbon atoms. This has been extensively studied for ethoxy-lated surfactants, and their degradation products have beenidentified.19,27–30

Kerwin4 and Yao et al.5 recently reported that the presenceof an unsaturated FA in PS 80 plays an important role in the in-creased “oxidability” of this surfactant as compared with PS 20.In general, autoxidation can occur anywhere within a moleculeas long as the kinetic characteristics of H-abstraction at a givensite are favorable. For example, comparison of the partial rateconstants of H-abstraction by a sulfate radical anion at roomtemperature from the C–H bond of a secondary carbon (1.0 ×108 M−1 s−1) is about an order of magnitude greater than thatof an "-C–H ether bond (2.1 × 107 M−1s−1).31 For the autoxida-tion of PSs, these values support experimental data where anunsaturation site of oleic acid was reported to be a prevalentpathway for PS 80 degradation,4,5 whereas PS 20 undergoesdegradation mostly at the "-carbon of EO moieties.13

Although autoxidation of unsaturated FAs and in partic-ular, oleic acid, has been extensively studied for decades be-cause of their importance in the oxidation of lipids in foods andoils,20,21,23–26,32 less is known about the actual degradation path-ways and stabilities of the individual components of PSs underoxidative stress. Recently, however, Hvattum et al.33 investi-gated the degradation of PS 80 using liquid chromatography–mass spectrometry (LC–MS) and proton nuclear magnetic res-onance (1H NMR) following thermal stress at 40◦C for 8 weeks.Characterization of major species, by resolving and followingthe individual FA esters of PS 80, suggested the presence of

Borisov, Ji, and Wang, JOURNAL OF PHARMACEUTICAL SCIENCES 104:1005–1018, 2015 DOI 10.1002/jps.24314

RESEARCH ARTICLE – Pharmaceutical Biotechnology 1007

short-chain POE oleic acid (C18:1) esters and hydroxy, keto,and epoxy products of oxidation of C18:1 esters.

Previously, we described a LC–MS method to determine theheterogeneity of PS, which takes advantage of knowing theidentity of a FA tail of esters from marker dioxolanylium ions,generated by in-source collision-induced dissociation (CID).34

In the study described herein, we expanded the applicationof LC–MS to the investigation of major degradation productsin PS 20 and PS 80 upon oxidative stress induced by 2,2′-azobis(2-methylpropionamidine) dihydrochloride (AAPH). Themodel oxidizer, AAPH, is a water-soluble azo compound thatis often used as a free radical generator, particularly in stud-ies of lipid peroxidation and characterization of antioxidants.Decomposition of AAPH produces molecular nitrogen and twocarbon-centered radicals, which may combine to produce sta-ble products or react with molecular oxygen to generate peroxyradicals.35 The half-life of AAPH is about 54 h (at 40◦C), makingthe rate of free radical generation essentially constant duringthe first several hours in solution.5

Despite the heterogeneity of PSs, their structural compo-nents are conserved and very predictable based on knowledgeof the manufacturing process and published analytical data.We suggest that (1) the presence of marker ions at the low m/zrange of the mass spectrum is a result of the partial cleavage ofPS molecules during LC–MS analysis and is indicative of thenature of the FA ester, and (2) careful examination of partiallyseparated polydisperse mass profiles following reversed-phase(RP) chromatography of intact PS species at the higher m/zrange of the mass spectrum are extremely useful in addressingthe structural characterization of these surfactants and theirdegradation products. We believe that a better understandingof the structure of PSs and their stability characteristics in for-mulation will help in the selection process of the appropriatesurfactants for use in drug manufacturing.

EXPERIMENTAL

Materials

Polysorbates 20 and 80 were obtained from Croda (RanchoCucamonga, California) and were used as received. Each PS(50 mg) was dissolved in 5 mL of water in volumetric flasksto give 10 mg/mL stock solutions, which were stored at 2◦C–8◦C and protected from light. All solvents including waterand acetonitrile were high-performance liquid chromatography(HPLC) grade and were purchased from Burdick and Jackson(Muskegon, Michigan). Formic acid was obtained from Pierce(Rockford, Illinois), and AAPH was purchased from Sigma–Aldrich (St. Louis, Missouri).

LC–MS Analysis of PSs and Data Analysis

The RP chromatography method and MS conditions were usedas previously reported.34 Briefly, separation was performed ona Vydac 214MS (Grace, Deerfield, Illinois) C4, 2.1 × 250 mm2

column packed with 300 A pore-sized 5 :m particles, oper-ated at 25◦C and a flow rate of 0.25 mL/min. PSs were chro-matographed by a RP gradient using mobile phases consistingof 0.1% formic acid in water and in acetonitrile as solvents Aand B, respectively. Typically, 0.8 :g of aqueous PS solutionswere loaded onto the column.

Mass spectrometric analyses were performed using a Wa-ters QTOF Premier (Milford, Massachusetts) quadrupole

time-of-flight mass spectrometer, operated in a positive elec-trospray ionization mode and controlled by MassLynx v. 4.1software. The MS method consisted of two acquisition func-tions with sampling cone voltages of 18 V during the first scanevent and 45 V for increased in-source fragmentation duringthe second scan.34 In additional experiments, CID fragmenta-tion was performed on selected ions with collision cell argongas pressure of 6.5 mbar.

Structural elucidation was performed by matching experi-mentally measured molecular weights of polyoxyethylates withthe ones calculated from formulas of potential constituents ofPSs.34 Information on a FA tail of the polyoxyethylates was ob-tained from the in-source CID data by detecting the low m/zsignal characteristic of the dioxolanylium ion of a FA, as pre-viously reported.34 Hereafter, these ions will be referred to as“marker ions.” Data analysis was facilitated by using a cus-tom program, written in-house in Visual Basic for Applicationssupplied with Microsoft Excel R©.

Oxidative Degradation of PSs with AAPH

Unless otherwise stated, stock PS solutions were diluted to0.04 mg/mL with water and spiked with a freshly prepared,light-protected stock solution of 15 mM AAPH to give its fi-nal concentration in a sample in the range of 1.5–10 mM, asspecifically indicated. Spiked samples were mixed, transferredinto 1-mL glass HPLC vials, capped, and placed into an HPLCautosampler maintained at 40◦C for the duration of the exper-iments. Samples were periodically injected onto a column forLC–MS analysis.

Enrichment of PS 20 Fractions

To reduce sample complexity and to enable the evaluation ofthe oxidative degradation of PS 20-related substances, namely,POE sorbitan mono- and di-laurates, fractions were collectedduring multiple, 5-:g injections of PS 20 onto the column fol-lowed by RP separation according to the LC method describedabove. Both fractions were collected on ice and stored at 2◦C–8◦C in glass tubes. Fractions were further evaporated undernitrogen to remove excess organic solvents and acids from mo-bile phases and to concentrate the contents. Final concentrationadjustments were performed with water to give a LC–MS sig-nal response for the substances to approximately match theresponse of a corresponding substance during LC–MS of an un-fractionated sample. It should be noted that no verification ofconcentrations in these final fractions with orthogonal methodswere performed.

RESULTS AND DISCUSSION

Characterization of PS 80 with LC–MS

The LC–MS total ion current (TIC) chromatogram for a typ-ical PS 80 sample is shown in Figure 1a. Overall, the profilewas very similar to that from previously published data on theanalysis of PS 80 surfactants,33,36 although some variability islikely caused by the differences in the RP gradients employed.In general, PS 80 exhibited fewer peaks compared with the pre-viously reported data for PS 20.34 This result also agrees withthe reduced complexity observed in the matrix-assisted laserdesorption/ionization spectra of PS 80 reported by Ayorindeet al.37

DOI 10.1002/jps.24314 Borisov, Ji, and Wang, JOURNAL OF PHARMACEUTICAL SCIENCES 104:1005–1018, 2015


Figure 1. Typical LC–MS TIC chromatograms of PS 80 (a), heat map generated from the RICs of FA marker ions, induced by in-sourcefragmentation (b), and RIC for m/z 309.3 indicating the elution of oleates (c), with major peaks labeled as POE sorbitan mono-oleate (1),POE isosorbide mono-oleate (2), POE sorbitan di-oleate (3), POE isosorbide di-oleate (4), and POE sorbitan tri-oleate (5). Distribution of FAs,determined from RICs of corresponding marker ions under elution peaks of POE sorbitan mono-esters, is given in the insert.

The RP gradient separates components in the order oftheir increasing hydrophobicity, that is, total hydrocarbon load.Characteristic of the PS surfactants, polyoxyethylated nones-terified polyols elute in the order of POE < POE isosorbide <

POE sorbitan, followed by a series of esterified species withincreasing lengths of FA chains, eluting in the order of POEsorbitan ester < POE isosorbide ester < POE ester; specieswith shorter FA chains elute prior to the species with longerFA chains; and di- and higher-order esters elute last on the RPgradient.

In particular, the PS 80 profile (Fig. 1c) is dominated bymono-oleates of POE sorbitan (Peak 1) and POE isosorbide(Peak 2), followed in turn by di-oleates of POE sorbitan (Peak3) and POE isosorbide (Peak 4), and POE sorbitan tri-oleateas a major component of Peak 5. Elution profiles (Fig. 1b)for each of the esters were investigated by reconstructing thechromatograms for the marker ions determined experimen-tally and verified against expected FA composition of the sur-factant, presented in the PS 80 monograph in the EuropeanPharmacopoeia.8 As shown in Figure 1b, oleic acid (C18:1) ac-counts for about 77% of all the FA ester content of PS 80,whereas other FA esters are present in minor amounts. Be-cause of the isobaric overlap between the monoisotopic m/z 311peak of stearic (C18:0) acid and the third (M+2) isotopic peakof oleic (C18:1) acid, elution profile of m/z 311 marker ion ex-hibited three peaks in the range of expected elution of mono-stearates between retention times of 35–40 min (labeled withasterisks in Fig. 1b). Even though the relative abundance ofthe M+2 isotope of C18:1 is below 3% of its monoisotopic peakabundance, its contribution complicates profiling for C18:0 es-ters, because of the high abundance of oleates in PS 80. In thiswork, adjustments were made by subtracting the contribution

of the M+2 isotopic signal of C18:1 oleates from the total m/z311 signal, allowing an estimation of the abundance of C18:0stearate species with the characteristic m/z 309 marker ion.

We emphasize the fact that the LC–MS method is not in-tended as a method for absolute quantitation of FA ester loadin PSs. Multiple factors, including dynamic range limitation,ionization, and in-source fragmentation efficiencies of differentcomponents may bias the results. Undoubtedly, any methodol-ogy used for absolute quantitation must meet certain criteria,including rigorous fit-for-purpose evaluation and qualificationfor the intended use, which was outside the scope of this work.In our opinion, however, the LC–MS method enables the detec-tion of FA esters in PSs with the opportunity to resolve chro-matographic overlaps and detect degradation species, and itallows the reconstruction of elution profiles of different com-ponents, suggesting their semiquantitative distribution. Themethod can be useful in monitoring stability profiles of PSsand in lot-to-lot comparisons. The results of FA profiling basedon the reconstructed signal of POE sorbitan mono-esters aregiven in a table insert to Figure 1.

The distribution of major POE sorbitan oleates was esti-mated from the corresponding peak areas of m/z 311 markerion (in Fig. 1c). Table 2 compares the relative amounts of thesespecies with data previously reported by Brandner12 for PS 80(as recalculated from Table 1 for POE sorbitan esters). The dis-tribution of the POE sorbitans roughly follows 8.2:10.0:5.1:1.0ratio for mono-:di-:tri-:tetra-esters, respectively. Despite thatthese data sets were determined by different methods andare separated in time by over a decade, the relative amountsof major POE sorbitan esters were similar; the results fromboth studies suggested that POE sorbitan di-oleates are likelythe major components of PS 80, despite the fact that POE



Table 2. Distribution of Major POE Sorbitan Esters in PS 80Determined by LC–MS and Reported by Brandner12

Relative Amount ofEsters (%)

Peak RT (min) Identification RIC, m/z 309 Brandner12

For All Sorbitans (Total 100%)

1 36.6 POE sorbitanmono-oleate

31.0 33.7

3 48.9 POE sorbitan di-oleate 47.1 40.95 55.5 POE sorbitan tri-oleate 21.9 21.3– – POE sorbitan

tetra-oleateN/D 4.1

LC–MS, liquid chromatography–mass spectrometry; N/D, not detected; POE,polyethoxylated; PS, polysorbate; RIC, reconstructed ion chromatograms; RT, re-tention time.

sorbitan mono-oleates are typically presented in the literatureand compendia8 as the major species of PS 80.

Monitoring Oxidative Degradation of PS 80 with LC–MS

Application of the water-soluble free radical generator, AAPH,to study oxidation in heterogeneous systems has been reportedpreviously. In liposomes, for example, AAPH was shown to haveaccess to all layers of the multilamellar assembly, even whenadded after the liposomes were formed.38 In the same study,efficiencies of lipid- and water-soluble initiators and the corre-sponding rates of oxidation of linoleic acid, incorporated intosodium dodecyl sulfate micelles, were reported to be identicalwithin the experimental error.38 By analogy, in this work, itwas assumed that the free radical generator has access to bothhydrophilic and hydrophobic layers of the core-shell sphericalmicelles39 of PS 80 and, thus, AAPH was used to simulate au-toxidation of PSs.

Forced oxidation was monitored by LC–MS, and the occur-rence of a new peak in TIC was first examined for the pres-ence of marker ions followed by the detailed examination ofthe corresponding mass spectra. Mass spectra under individualTIC peaks were carefully examined for the presence of a massshift relative to the homologous peaks for major species in thespectra of the nondegraded sample. Information on degrada-tion products was supplemented by the examination of markerions under these peaks, which further hinted if modificationoccurred at the FA tail. Because of the limited ability of theRP gradient to separate multiple components of PSs, chro-matograms were reconstructed for each of the marker ionsand the most representative spectra under these reconstructedpeaks were subjected to the examination. Individual compo-nents could be distinguished even in cases of chromatographicoverlaps. Furthermore, tentative assignments of degradationproducts were verified by consulting with the abundant liter-ature on the known degradation pathways of compounds con-taining oleic acid.20,21,23–26,32 We believe that the careful exami-nation of MS and partial tandem MS (in-source CID) data fromLC–MS analysis is sufficient to tentatively characterize majorcomponents and their degradation products in PSs. Orthogo-nal methods, such as 1H NMR, often require purified materialfor proper structural elucidation, and in the case of PS 80, itprovided only limited information to support structural elucida-tion of degradation products in the bulk PS 80 sample because

Table 3. Polysorbate 80 Stressed with 1.5 mM AAPH at 40◦C

PS 80 Major SpeciesRate Constant, k

(10−3 min−1) Linear Fit, R2

POE sorbitan mono-oleate 3.76 0987POE sorbitan di-oleate 7.77 0.994POE sorbitan tri-oleate 10.44 0.990

Pseudo first-order rate constants determined for the decomposition of majorPOE sorbitan oleates.

AAPH,2,2′-azobis(2-methylpropionamidine)dihydrochloride;POE,polyethoxylated; PS, polysorbate.

of the heterogeneity of the material and polydispersity of itscomponents.33

Upon reaction with 1.5 mM AAPH, the degradation of POEsorbitan C18:1 mono-, di-, and tri-esters under Peaks 1, 3, and 5,in Figure 1c, respectively, followed a pseudo first-order reactionkinetics with rate constants summarized in Table 3 (see Fig.S-1 in Supplementary Material for plot details). The rate con-stants for the decomposition increased linearly (slope = 3.3 ×10−3 min−1 and R2 = 0.987) with the number of FA esters (esterorder) per molecule of POE sorbitan. The result suggests thatdegradation of PS 80, under AAPH stress, is dominated by ox-idation at the oleic acid tail. From a practical viewpoint, thisresult indicates that the presence of higher-order esters in PS80 can significantly contribute to its overall degradation andeven to the loss of its formulation-protecting properties.

Representative TIC chromatograms, recorded at 12.5 and40.0 h following the addition of AAPH, are shown in Figure 2a.The early eluting peaks containing nonesterified polyols ataround 15 min, barely change their shape or abundance, sug-gesting the absence of FA ester hydrolysis or oxidative decom-position of nonesterified POE chains, which otherwise couldhave resulted in increased abundance of these polyols. At thesame time, major components under Peaks 1–5 significantlydecreased, and a series of new Peaks 6–10 emerged.

First, the new peaks were examined for the presence of themarker ions. Abundant marker ion at m/z 199.1 was detectedunder Peak 6. Furthermore, Peak 7 contained m/z 323.3 markerion, Peaks 8 and 9 had several marker ions at m/z 307.3, 323.3,and 325.3, and Peak 10 exhibited a pair of m/z 325.3 and 307.3marker ions. Reconstructing the profiles for these marker ionsrevealed that m/z 199.1 ion eluted under a single, broad chro-matographic peak, whereas each of the other three marker ions(m/z 307.3, 323.3, and 325.3) exhibited several peaks in theirelution profiles [the reconstructed ion chromatograms (RICs)for these marker ions are shown in Fig. S-2, of the Supplemen-tary Material]. We also noted that m/z 325.3 and 307.3 markerions had fully overlapping elution profiles suggesting that theseions likely originated from the same degradation product of PS80. We thus assumed that m/z 307.3 ion is the secondary frag-ment ion produced by the loss of water from m/z 325.5 ion,which is less stable toward in-source CID.

Second, mass spectra under chromatographic peaks, corre-sponding to the elution apex of each of the marker ions, wereexamined for the presence of mass shifts, exhibited by homol-ogous ions series of ethoxylated species. Figures 2c–2d exhibitthe spectra corresponding to POE sorbitan esters, eluting un-der Peaks 6–8 in comparison to POE sorbitan mono-oleateunder Peak 1 of the nondegraded sample. An abundant m/z199.1 marker ion and the corresponding homologous ion series



Figure 2. Liquid chromatography–mass spectroscopy TIC chromatograms of a fresh (top trace) and incubated with 1.5 mM AAPH at 40◦C for12.5 h (middle trace) and 40 h (bottom trace) PS 80 (a) and spectra under Peaks 1 of nondegraded PS 80 (b), Peak 8 (c), Peak 7 (d), and Peak 6(e) of PS 80 at 12.5 h after treatment with AAPH.

exhibited mass shift of 110.2 Da [309.3–199.1 and 2 × (809.5–754.4), respectively, in the marker and the ethoxylated speciesion regions]. As can be seen by comparing the spectra inFigures 2b and 2e, suggesting the presence of oxo-nonanoic acid(azelaic semialdehyde) under Peak 6, a known product of oleicacid oxidation, forming by scission of the carbon chain at C9position.24 However, the occurrence of oxo-nonanoic esters wasnot previously reported as the product of oxidative degradationof PS 80.

Peaks 7–10 were only partially resolved, although markerions at m/z 323.3 and 325.3 (�M = 14.0 and 16.0 Da, respec-tively) suggested that these are likely oxidation products ofoleic acid. Moreover, both m/z 323- and 325-containing prod-ucts were detected as POE sorbitan, POE isosorbide, and POEesters, eluting in the same order as the original componentsof the nondegraded PS 80 (i.e., POE sorbitan esters < POEisosorbide esters < POE esters). Hvattum et al.33 recently re-ported the presence of m/z 307.3 and 325.3 as ion pair and them/z 323.3 ion in thermally degraded PS 80. The authors at-tributed the former pair of fragment ions to hydroxy or epoxymodifications of C18:1 and the latter ion to keto-C18:1 acid. Inthe present work, strong m/z 325.3 marker ion was detectedunder Peaks 8, 9, and 10 (Fig. S-2 of the Supplementary Ma-terial). Spectrum for species eluting under Peak 8 is shown inFigure 2c. On the basis of the 16.0-Da mass shift, we concludedthat these species were likely because of POE sorbitan epoxy-stearate, following the replacement of the unsaturation site ofoleic acid with oxirane ring. Similar conclusions were madeby Hvattum et al.,33 who studied thermal degradation of PS80. Spectra under two other chromatographic peaks, contain-

ing m/z 325.3 marker ions, were attributed to POE isosorbideepoxy-stearate (under Peak 9) and POE epoxy-stearate (underPeak 10) (refer to Figs. S-3 and S-4, Supplementary Material,for details).

Peak 7 was abundant in the TIC profile during initial phasesof the forced oxidation of PS 80 (Fig. 2a, middle trace) and ex-hibited characteristic m/z 323.3 marker ion, suggesting the ad-dition of a carbonyl oxygen atom (�M = 14.0 Da). The m/z 323.3was previously attributed to a keto-oleate product of oleatedegradation.33 However, the corresponding ion series of POEhomologues under Peak 7 were shifted by 32 Da [2 × (825.5–809.5)] relative to the corresponding ion series of POE sorbitanoleate under Peak 1 in the nondegraded sample (Figs. 2b and2d). Furthermore, compared with the POE (29) sorbitan oleate(m/z 809.5), the corresponding m/z 825.5 ion of the 32-Da shiftedspecies is significantly less stable toward CID, as shown inFigure 3. Tandem MS spectra of this species, in Figure 3b,exhibited neutral losses of 18 and 34 Da as well as series ofneutrals with mass difference of 14 Da. Simultaneous lossesof water (18 Da) and hydrogen peroxide (34 Da) are charac-teristic of hydroperoxides,24 whereas neutrals with proposedcomposition of CnH2n+2O (n = 5–8), originating from the FA es-ter hydrocarbon tail, indicate its overall reduced stability, com-pared with unmodified oleate tail. This led us to conclude thatspecies exhibiting 32.0 Da mass shift were likely because of thepresence of hydroperoxy-oleate (C18:1), which upon mild in-source CID rapidly lose a water molecule to produce the abun-dant m/z 323.3 marker ion. Furthermore, Peak 7 was the mostabundant degradation peak at early time points of the reactionwith AAPH and exhibited the decrease in its abundance as the



Figure 3. Portions of tandem mass spectra for doubly charged sodi-ated precursor ions with m/z 809.48 (a) and 825.46 (b), correspondingto the major components under Peak 1 (a) and Peak 7 (b) and identifiedas POE (26) sorbitan oleate (a) and its +32-Da analog (b) obtained atcollision cell Argon gas pressure of 6.5 mbar and CID energy of 31 V.

reaction proceeded, exhibiting the behavior characteristic ofintermediates. These facts allowed us to propose that thesespecies were because of hydroperoxides typically observed dur-ing autoxidation reactions.

Two possible pathways involving POE sorbitan oleates wereexamined for the radical-induced degradation of PS 80. Reactiv-ity of the double bond of oleic acid produces POE sorbitan esterswith modifications at oleic acid, whereas initiation at a randomEO moiety of the esterified POE arm, leads to the productionof POE esters. Figure 4a schematically presents these two sce-narios of radical site formation at the double bond of oleic acid(Path A) and along POE chain of the esterified arm (Path B)of PS 80 and subsequent degradation products. At the initialphases of the reaction with AAPH, major degradation prod-ucts under Peaks 6–10 (Fig. 2a) were experimentally identifiedas POE sorbitan esterified with hydroperoxy-oleate (C18:1),epoxy-stearate (C18:0), and oxo-nonanoate (C9:0). These prod-ucts were likely products of Path A reactions. Tentative struc-tures for these compounds structures IIa, Va, and VIa, respec-tively, are presented in Table 4. Degradation of the originalamounts of POE sorbitan oleates, present in the nondegradedPS 80, at some point of the reaction exhausted Path A reac-tions. At later time points, Path A products, in turn, followedPath B reaction pathways, producing corresponding POE estersas secondary products of the forced oxidation.

A qualitative evaluation of time-resolved profiles of selectedspecies is presented in Figure 4b. Monitoring the abundancesof Path A (POE sorbitan esters) and PathB (POE esters) prod-ucts suggested that both pathways likely occur simultaneously,but with significantly different rates. For example, POE oleate(structure IIb in Table 4) can form by Path B reaction of oleate-containing species in the nondegraded sample. POE oleate waspresent at low amounts in the original nondegraded materialand in the course of the reaction with AAPH slightly increasedin abundance (in part, because it could originate from mul-tiple sources), following by the sharp decrease, presumablybecause of the fast reactivity of the double bond of oleicacid. In contrast, Path A species, dominated by POE sorbitan

Figure 4. Proposed degradation pathways for polyoxyethylated C18:1 arm of PS 80 (a), and time-resolved changes in abundance of (◦) POE (26)sorbitan C18:1 at m/z 809.5 (structure Ia); (�) sum of POE (26) sorbitan oxo-C9:0 at m/z 754.5 (structure VIa) and POE (26) sorbitan epoxy-C18:0at m/z 817.5 (structure Va); (�) POE (5) C18:1 at m/z 525.4 (structure IIb); (�) sum of POE (5) keto-C18:1 at m/z 539.4 (structure IVb), POE(5) epoxy-C18:0 at m/z 541.3 (structure Vb), and POE (5) oxo-C9:0 at m/z 415.2 (structure VIb); (×) and POE (26) sorbitan hydroperoxy-C18:1(structure IIa) at m/z 825.5 during reaction of PS 80 with 1.5 mM AAPH at 40◦C (b). Tentative structures are given in Table 4.



Table 4. Tentative Structures of Compounds Formed During Oxidative Degradation of PS80 Because of the Reactivity at the Double Bond ofOleic Acid (POE Sorbitan Ester Structures Ia–VIa) or at the POE Chain and the Double Bond of Oleic Acid (POE Ester Structures Ib–VIb)

Structure Structure

Ia Ib

IIa IIb

IIIa IIIb

IVa IVb

Va Vb

VIa VIb

epoxy-stearate (structure Va) under Peak 8, was the majordegradation product until over 90% of the original POE sor-bitan oleate degraded. At that point, Path A reactions becameexhausted and subsequent Path B reactions produced corre-sponding POE esters, dominated by POE epoxy-stearate (struc-ture Vb) under Peak 10.

Oxidative Degradation of PS 20

Liquid chromatography–mass spectrometry profile of PS 20,shown in the top trace of Figure 5a, visually has increased com-plexity, compared with PS 80 in Figure 1a. However, the majorchromatographic features of the two PSs are similar, exhibit-ing regions with nonesterified polyols, eluting before 20 min,mono-esters between 25 and 40 min, and di- and higher-orderesters at 39–50 min.34 The main difference is because of the dif-ferences in abundances of FAs, with C12:0, C14:0, C16:0, andC18:0 abundant in PS 20 versus a dominating presence of C18:1in PS 80. Profiles in Figure 5b, reconstructed for marker ions,show elution of corresponding esters in nondegraded PS 20. Forexample, laurate-containing species (marker ion m/z 227) is the

major components under the 31.7-, 33.2-, and 41.5-min peaks,corresponding to POE sorbitan mono-laurate, POE isosorbidemono-laurate, and POE sorbitan di-laurate, respectively.

In contrary to PS 80, where the presence of the unsatu-rated site within the alkyl chain of oleic acid governs oxidativedegradation of the surfactant, scission at ether bonds has beendescribed as the main oxidative degradation pathway for PS20.13,28 The variation in stability toward the oxidative stresswas observed for different components of the PS 20 sampleupon the reaction with 5 mM AAPH at 40◦C. In Figure 5a,profiles of degraded PS 20 exhibited almost complete depletionof POE sorbitan di- and higher (tri- and tetra-) order esters19 h after the treatment with AAPH, whereas peaks because ofPOE sorbitan mono-esters showed only slight change even af-ter 40 h of the reaction. Reconstructed profiles for marker ions,shown in Figures 5b and 5c, provided further evidence of thephenomenon. For example, we noted the quantitative depletionof di-laurates, eluting at around 41.5 min in the nondegradedsample (compare m/z 227 RICs in Figs. 5b and 5c). In contrast,the POE sorbitan mono-laurate peak at around 31.7 min did



Figure 5. Liquid chromatography–mass spectroscopy TIC chromatograms of the fresh (top trace) and incubated with 5 mM AAPH at 40◦C for19 h (middle trace) and 40 h (bottom trace) PS 20 (a) and RICs for marker ions with m/z 199, 227, 255, 283, and 311, characteristic to caprate(C10:0), laurate (C12:0), myristate (C14:0), palmitate (C16:0), and stearate (C18:0) containing species, respectively, from a fresh (b) and after40 h of AAPH-treated PS 20 sample (c). Peaks labeled as 1, 2a, 2b, and 3 refer to POE sorbitan mono-ester, POE isosorbide mono-ester, POEmono-ester, and POE sorbitan di-ester, respectively.

not exhibit a significant change after the treatment, indicatingthe overall greater stability of mono-esters toward the oxidativestress compared with the corresponding di-esters.

It was also observed that esters with longer FA chains (largercarbon number) exhibited greater susceptibility toward theoxidative degradation, compared with their shorter analogs,as can be seen by comparing profiles in Figures 5b and 5c.For example, POE sorbitan mono-palmitate (C16:0, m/z 283)and POE sorbitan mono-stearate (C18:0, m/z 311), eluting at36.2 and 38.7 min, respectively, in Figure 5b, are affected toa greater extent compared with POE sorbitan mono-caprate(C10:0, m/z 199) and POE sorbitan mono-laurate (C12:0, m/z227), at 29.5 and 31.8 min, respectively. Furthermore, the oc-currence of peaks (labeled as 2b in Fig. 5c) becomes prominentin the reconstructed profiles of ethoxylated esters after treat-ment with AAPH. These peaks were identified because of POEesters, containing 3–9 EO units. For example, in the RIC traceof myristates (C14:0, m/z 255), peak 2b at 36.7 min was identi-fied as POE mono-myristate. In general, POE esters increase inabundance more rapidly for FAs with larger carbon numbers,compared with shorter esters. The formation of free POE estersis consistent with the previously reported data on autoxidativedegradation of PSs.28 Radical-induced ether bond scission along

the esterified POE arm is the accepted mechanism of degrada-tion of ethoxylated surfactants.27

Degradation rate constants for major POE sorbitan esterswere determined based on the pseudo first-order kinetics as-sumption, as shown in Figure 6. During the reaction of PS20 and PS 80 with 1.5 mM AAPH, the rates of degradationwere found to depend on the FA carbon number, with POEsorbitans esterified with shorter-chain FAs being more stabletoward the oxidative stress with AAPH. Table 5 further com-pares rate constants for reactions of PS 20 and PS 80 with 1.5and 10 mM AAPH. As can be noted, degradation rates of POEsorbitan mono-esters of saturated FAs in PS 20 monotonouslyincrease with their carbon number. For example, POE sorbitanmono-C8:0 are about 2.7-times more stable toward the oxida-tive stress with 10 mM AAPH, compared with POE sorbitanmono-C18:0 esters.

Exact reasons for the effect of the FA carbon number onthe stability of POE esters toward oxidative stress with AAPHare not well understood at this time. However, factors suchas a possibility of the preferential ether bond scission withinesterified POE arm can be argued. For example, in aque-ous solutions, an extensive network of H-bonding stabilizesthe conformation of POE chains. The H-bonding between two



Figure 6. Kinetic plots for the reaction of major POE sorbitan esters in PS 20 and PS 80 with 1.5 mM AAPH at 40◦C, based on peak areasof RICs for marker ions with m/z 199 (mono-C10:0), 227 (mono- and di-C12:0), 255 (mono-C14:0), 283 (mono-C16:0), 311 (mono-C18:0), and 309(mono-C18:1).

Table 5. Polysorbate 20 Stressed by 1.5 and 10 mM AAPH at 40◦C

Rate Constant, k (10−3 min−1)

POE Sorbitan Ester 1.5 mM AAPH 10 mM AAPH

Mono-caprilate (C8:0) N/A 0.87Mono-caprate (C10:0) N/A 1.09Mono-laurate (C12:0) ∼0.03 1.19Mono-myristate (C14:0) 0.21 1.47Mono-palmitate (C16:0) 0.57 2.36Mono-stearate (C18:0) 0.89 2.38Di-laurate (C12:0) 2.84 6.67Mono-oleate (C18:1)a 3.87 11.86

Pseudo first-order rate constants for the decomposition of POE sorbitan esters.aObtained from stressing PS 80 with AAPH.AAPH,2,2′-azobis(2-methylpropionamidine)dihydrochloride;POE,

polyethoxylated; PS, polysorbate.

adjacent POE oxygen atoms and a molecule of water stabilizesthe trans-gauche-trans helical conformation of POE.40 In con-trast, the physically stretched POE chain changes conformationto the planar trans-trans-trans (or “all-trans”) (Fig. 9 in Ref.44).In this conformation, the increased distance between oxygenatoms perturbs POE hydration, allowing only a single H-bondbetween a water molecule and the polymer backbone. This per-turbed hydration prevents the formation of solvation bridgesand destabilizes water on the surface of this conformation, asthe result of only weak interactions of the “all-trans” conforma-tion with water molecules. Such conformation is observed, forexample, in a stressed (or physically stretched) POE.41 As theesterified POE arm can be viewed as a linker between the hy-drophobic core and the hydrophilic shell of PS, its conformationcan differ from a free POE chain in an aqueous solution and canpossibly resemble that of the physically stretched (planar “all-trans” conformation). This conformation can presumably makeesterified POE arm of the surfactant more vulnerable to theattack by AAPH. We speculate that this can arguably explain

observed differences between the degradation rates of esterswith different carbon numbers.

Polyethoxylated sorbitan esters with unsaturated FAs werefound to be the least stable species with rate constants in an or-der of higher magnitude compared with esters with saturatedFAs under the same experimental conditions (Table 5). Thisresult is consistent with the literature, indicating a nearly 10-fold difference in partial rate constants of H-abstraction fromC–H bond of a double bond versus reactivity of "-C–H bond ofethers.31 Furthermore, di-esters were found to be significantlyless stable than the corresponding mono-esters. For example,POE sorbitan di-C12:0 degraded 5.6-times faster than POE sor-bitan mono-C12:0 during the reaction with 10 mM AAPH. Thisresult exceeds the twofold difference of the rate constants antic-ipated considering statistically equivalent probabilities of scis-sion along esterified POE arms. Reasons behind such a signifi-cant difference in degradation kinetics of POE sorbitan mono-and di-esters induced by AAPH are not clear at the moment.We note, however, that scission along one of the esterified POEarms in di-esters can produce corresponding mono-esters, ef-fectively increasing concentration of the later species in thesample, as will be discussed below. Also, differences in physicalproperties of micelles formed by different fractions of PS andtheir relative stabilities toward external stress might need tobe considered, which was beyond the scope of this work.

Unlike the oxidative degradation of PS 80, where reactiv-ity of the double bond of oleic acid toward the oxidative stressleads to the formation of chemically novel species, scission ofan ether bond dominating the degradation of PS 20 can shortennonesterified POE arms and shift the polydispersity profile ofPS 20, but not necessarily lead to the formation of new species.In the attempt to overcome this issue, degradation of purifiedfractions of POE sorbitan mono- and di-C12:0 were examinedseparately, as shown in Figure 7. The fractions were collectedand enriched during repeated injections of the fresh PS 20 sam-ple onto the RP column. To simplify the following discussion,these fractions are referred to as the mono- and the di-esterfractions.



Figure 7. Distribution of major species from purified and enriched POE sorbitan mono- (a) and di-laurate (b) fractions at T0 and after treatmentwith 5 mM AAPH for 40 h, and the proposed scheme for major reactions during oxidative degradation of POE sorbitan mono-esters (c). Arrowsin (a) and (b) indicate shifts in polydispersity profiles.

Polydispersity profiles of the enriched fractions exhibited thecharacteristic bell shape with respect to the distribution of EOmoieties (diamond symbols in Figs. 7a and 7b), with good sym-metry around POE (24 EO units) sorbitan esters. After 40 hof the reaction with 5 mM AAPH, POE sorbitan, POE sorbi-tan mono-C12:0, and POE mono-C12:0 products were detectedin both of these fractions; the only detectable difference wasthe relative abundance of these compounds. At the same time,the presence of POE sorbitan di-C12:0 was not detected in thedi-ester fraction, indicating its complete depletion.

For POE sorbitan mono-esters, two types of complimentaryproducts formed, depending on whether bond scission occurredalong its esterified or nonesterified POE arms, as depicted inFigure 7c. Ether bond scission along the esterified POE armproduced complimentary POE sorbitan and POE mono-C12:0products (Path I). It was noted that these degradation productshad polydispersity profiles that were shifted toward the fewernumber of EO units per molecule. For example, POE sorbitanscontained on average 18–19 EO units, whereas complimentaryPOE esters had 4–5 average number of EO units. Path II re-actions, cleaving along one of the nonesterified POE chains,produced free POE and, in part, shifting polydispersity profileof the starting POE sorbitan mono-esters toward species with

fewer EO units (polydispersity shift is indicated by arrows inFig. 4a). It should be noted that free POE products of the PathII reaction were not reliably detected by the LC–MS method,likely because of their hydrophilic nature and their likely poorretention of the column. In contrast, degradation of the di-esterfraction by the Path I reaction produced complimentary POE es-ters and POE sorbitan mono-esters with polydispersity shiftedtoward species with fewer EO units (Fig. 4b).

Radical-induced scission of ether bonds generates productswith several possible terminating groups. Cleaved productshave been reported to terminate with alcohol, aldehyde, hemi-acetal, and formate groups.21,29 Out of these products, alcoholsare homologous to the original nondegraded compounds, buthaving lower molecular weights because of the cleavage of anumber of EO units from the PEO chain (i.e., with mass dif-ferences multiple of the mass of EO unit). Degradation prod-ucts, exhibiting mass shifts that are not multiple of the massof EO, likely contain alternative terminating groups. Possiblegroups are schematically represented in Figure 7c by differentR groups, which were experimentally measured to have massshifts of 0, −16, and +16 Da, suggesting alcohol-, formate-, andhemiacetal-terminating groups, respectively. Spectra for prod-ucts of the oxidative degradation of POE sorbitan mono-oleate



suggested that products with these terminating groups wereproduced in roughly equal amounts during the reaction withAAPH (Fig. S-6 of the Supplementary Material). Although theexact structural elucidation of terminating groups was beyondthe scope of this work, tentative structures presented in Figure7c are consistent with autoxidative mechanisms of degrada-tion of ethoxylated compounds, which were detailed by Bodinet al.27, Kishore et al.,28 and Currie et al.29

Polydispersity profiles of POE esters produced in the mono-and the di-ester fractions are similar. However, there is nearlya twofold difference in the rates of production of these speciesbetween the two fractions, as is evidenced by the increasedabundance of POE esters in the di-ester fraction. This resultsuggests that the rate constants of ether bond scission alongesterified arms of POE sorbitan esters were likely similar forboth mono- and di-ester species. As was noted above, the dif-ference in degradation rate constants of POE sorbitan mono-and di-C12:0 in unfractionated PS 20 were found to exceed theanticipated twofold difference. Cleavage along one of the ester-ified POE arms of di-ester species can generate correspondingmono-esters, essentially increasing the concentration of thesespecies in the sample with homologs, having fewer number ofEO moieties per molecule. This can complicate a true compar-ison of degradation rates of mono- and di-ester species in anunfractionated sample. In this work, the comparison was per-formed between POE (28 EO units) sorbitan mono- and di-esterspecies from the enriched fractions. Species with the numberof EO units per molecule above the average (EOAvg = 24) valuein their polydispersity profiles were chosen for the compari-son. This choice reduces the potential contribution of specieswith number of EO > EOAvg to the formation of homologousspecies with the number of EO < EOAvg because of the shorten-ing of nonesterified POE chains during their oxidative degrada-tion. In this case, we noted that the rate constants were nearlytwofold (2.2 times) different as would be expected for speciescontaining one and two esterified POE arms (shown in Fig. S-6of the Supplementary Material). This result suggests that theactual degradation rate constant of scission along the esteri-fied POE arm was likely similar for POE sorbitan mono- anddi-esters.

Potential Relationship Between Degradation Mechanisms of PSs

The catalytic activity of transition metals toward the oxida-tion of lipids is well known.42 In aqueous phases, metals arepresent as positively charged cations with strong affinities to-ward negatively charged surfaces, such as charged outer sur-faces of droplets of oil-in-water emulsions. The greater thesurface-charge density on such a droplet, the greater its abilityto attract oppositely charged counter-ion. It was demonstratedthat, when catalyzed by iron, the rate of lipid oxidation in emul-sions containing anionic, nonionic, and cationic surfactants wasthe highest when droplets carried a negative charge.42 The ratewas significantly lower in cases when the droplets were posi-tively charged or uncharged.42

Free FAs were reported to promote the oxidation of chargedcolloidal lipid systems dispersed in water.43,44 Because free FAsare surface-active compounds, they diffuse and concentrate atthe interface of micelles, potentially increasing their net nega-tive surface charge. In the literature, it was attributed to thefact that negatively charged emulsion droplets attract tran-sition metals, resulting in increased metal-lipid interaction,

which promotes oxidation.43,44 For example, addition of 1% ofoleic acid to 1% oil-in-water emulsion of stripped soybean oilpromoted the formation of hydroperoxides at pH ≥6, correlatingwell with an increase of a negative surface charge of droplets.43

In contrast, replacement of oleic acid with methyl oleate didnot have the pro-oxidative effect. Furthermore, lipid oxidationin these emulsions was shown to be inhibited by ethylenedi-aminetetraacetic acid, further indicating that the most likelymechanism of the pro-oxidant effect of free FAs is the attrac-tion of cations of transition metals to droplet surfaces, whereinteraction with lipids promotes their oxidation.

Polysorbates are nonionic, and micelles of their fresh aque-ous solutions carry no surface charge. However, undergoinghydrolytic degradation, PSs can release free FAs,28,45 with con-centrations increasing over time. Extrapolating from the pre-vious discussion, free FAs can contribute to the buildup of anegative charge on the surface of micelles by concentrating atthe micelle–water interface. Thus, the more significant the hy-drolytic degradation, the greater the surface charge density is.Attraction of transition metal cations by the negative chargeat micelles’ surface would increase their local concentration atthe surface of the micelle, which, in turn, could result in accel-eration of autoxidation of the surfactant.

Although this hypothesis was not examined in thiswork, the evidence present in the literature on the pro-oxidative effect of free FAs of oil-in-water emulsions sug-gests that it might also be applied to PS solutions at cer-tain conditions, such as pH, buffer composition, and storagetemperature.

CONCLUSIONS

Polysorbates are known to undergo degradation in solutionsvia autoxidation, which is often catalyzed by metal ions andultraviolet (UV) radiation as a seed for free radicals, and/orby hydrolytic cleavage of ester bond.28 These reactions haverecently sparked attention within the biotech community asthey can directly or indirectly hinder the intended stabilizingproperties of PSs in protein formulations. Peroxides generatedduring autoxidation reactions can, in turn, promote oxidation oflabile amino acids, typically Met and Trp residues, which canbe a major cause of the degradation of therapeutic proteins.Autoxidation of PSs recently has become an important areaof research.4,5 It is interesting to note that the ability of PS80/Fe(III) system to generate high levels of peroxy radicals bya Fenton-like reaction, without the use of additional hydrogenperoxide, allowed researchers to use this system as a stressingmethod for ranking the oxidizability of small molecules.46

This study applies LC–MS to further investigate the degra-dation products of PSs upon oxidative stress with AAPH in wa-ter. The described LC–MS method takes advantage of the useof dioxalanylium low m/z marker ions, which are characteris-tic of the FA ester of sample components, separated by the RPgradient. In good agreement with the published data, oxidationat the double bond of oleic acid was shown to be the primarypath for degradation of PS 80. Using the LC–MS method, ma-jor degradation products of PS 80 were identified and tentativemechanisms of their formation were proposed. These are con-sistent with the abundant reports on oxidation of oleic acid andits lipid derivatives. In contrast, PS 20, in which the majorityof esters are with saturated FAs, ether bond scission is clearly



the major mechanism of the oxidative degradation of this sur-factant.

In addition to the direct effect on proteins’ stability, degra-dation of PSs might affect the physical properties of these sur-factants by shifting their cloud point and critical micelle con-centration, induced, for example, by shortening POE chains.Furthermore, shifts in the chemical composition of these sur-factants by liberation of free FAs and/or POE esters, can belinked to the occurrence of cloudiness of a formulated solution.

It should be noted, however, that this work focused on ap-plications of the LC–MS methodology to study compositionsand degradation products of complex species, such as PSs. Thewater-soluble AAPH was used as a model oxidizer as describedin a previous report to have equal access to all layers of the mul-tilamellar assembly of liposomes.38 We believe that AAPH canbe used to simulate autoxidation, catalyzed by the presence ofa transition metal or exposure to UV radiation, by providing arepresentative access to functional groups within the micelles.Convenience of the AAPH method was illustrated by the factthat by adjusting the concentration of the free radical gener-ator, the degradation rates can be controlled to desired levelsfor a particular experiment. These studies, however, were car-ried out in water and effect of buffer composition and pH of thesolution has not been studied. The proposed LC–MS methodenables structural characterization and speciation analysis ofPSs and can be used to monitor their degradation pathways, in-fluencing stabilizing properties of these surfactants and, thus,important for the formulation development of biotherapeutics.

We believe that a better understanding of the properties ofPSs can be gained from a careful examination of physicochem-ical properties of individual components of these surfactants,their relative amounts, and speciation, which is often neglectedfor these surfactants. With the use of LC–MS, this goal is nowwithin reach.

ACKNOWLEDGMENTS

The authors thank Mary Cromwell and Jamie Moore fromGenentech for their support and Barthelemy Demeule andSandeep Yadav from Genentech for critically reviewing themanuscript.

REFERENCES

1. Joint FAO/WHO Expert Committee on Food Additives. 1974. “Toxi-cological evaluation of some food additives including anticaking agents,antimicrobials, antioxidants, emulsifiers and thickening agents”, WHOFood Additives Series No. 5. World Health Organization.2. Garidel P, Hoffmann C, Blume A. 2009. A thermodynamic analysisof the binding interaction between polysorbate 20 and 80 with humanserum albumins and immunoglobulins: A contribution to understandcolloidal protein stabilization. Biophys Chem 143:70–78.3. Patapoff TW, Esue O. 2009. Polysorbate 20 prevents the precipitationof a monoclonal antibody during shear. Pharm Dev Technol 14:659–664.4. Kerwin BA. 2008. Polysorbates 20 and 80 used in the formulation ofprotein biotherapeutics: Structure and degradation pathways. J PharmSci 97:2924–2935.5. Yao J, Dokuru DK, Noestheden M, Park SS, Kerwin BA, Jona J,Ostovic D, Reid DL. 2009. A quantitative kinetic study of polysorbateautoxidation: The role of unsaturated fatty acid ester substituents.Pharm Res 26:2303–2313.6. Deechongkit S, Wen J, Narhi LO, Jiang Y, Park SS, KimJ, Kerwin BA. 2009. Physical and biophysical effects of polysor-

bate 20 and 80 on darbepoetin alfa. J Pharm Sci 98:3200–3217.7. Chou DK, Krishnamurthy R, Randolph TW, Carpenter JF, ManningMC. 2005. Effects of Tween 20 and Tween 80 on the stability of Al-butropin during agitation. J Pharm Sci 94:1368–1381.8. European P. European Directorate for the Quality of Medicines &HealthCare. supplement 6.6, 7th ed., Strasbourg, France: Council ofEurope 2011: p 5311–5313.9. United States Pharmacopeia and National Formulary (USP 33-NF28). online edition., Rockville, MD: The United States PharmacopeialConvention, 2011.10. Stockburger GJrocess for Preparing Sorbitan Esters. July 17, 1980Patent US4297290.11. Stockburger GJ. 1979. Ethoxylation. J Am Oil Chem Soc 56:774A–777A.12. Brandner JD. 1998. The composition of NF-defined emulsifiers: Sor-bitan monolaurate, monopalmitate, monostearate, monooleate, polysor-bate 20, polysorbate 40, polysorbate 60, and polysorbate 80. Drug DevInd Pharm 24:1049–1054.13. Donbrow M, Azaz E, Pillersdorf A. 1978. Autoxidation of polysor-bates. J Pharm Sci 67:1676–1681.14. Ha E, Wang W, Wang YJ. 2002. Peroxide formation in polysorbate80 and protein stability. J Pharm Sci 91:2252–2264.15. Lam XM, Lai WG, Chan EK, Ling V, Hsu CC. 2011. Site-specifictryptophan oxidation induced by autocatalytic reaction of polysorbate20 in protein formulation. Pharm Res 28:2543–2555.16. Wang W, Wang YJ, Wang DQ. 2008. Dual effects of Tween 80 onprotein stability. Int J Pharm 347:31–38.17. Yin H, Porter NA. 2005. New insights regarding the autoxida-tion of polyunsaturated fatty acids. Antioxid Redox Signal 7:170–184.18. Denisov ET, Afanas’ev IB. 2005. Chain mechanism of liquid-phaseoxidation of hydrocarbons. In Oxidation and antioxidants in organicchemistry and biology. Boca Raton, Florida: Taylor & Francis, pp 27–55.19. Bodin A, Linnerborg M, Nilsson JLG, Karlberg AT. 2002. Novel hy-droperoxides as primary autoxidation products of a model ethoxylatedsurfactant. J Surfactants Deterg 5:107–110.20. Swern D, Coleman JE. 1955. Reactions of fatty materials with oxy-gen. XX. Recent developments in the autoxidation of methyl oleate andother monounsaturated fatty materials. J Am Oil Chem Soc 32:700–703.21. Lercker G, Rodriguez-Estrada MT, Bonoli M. 2003. Analysis of theoxidation products of cis- and trans-octadecenoate methyl esters bycapillary gas chromatography-ion-trap mass spectrometry. I. Epoxideand dimeric compounds. J Chromatogr A 985:333–342.22. Porter NA, Caldwell SE, Mills KA. 1995. Mechanisms of free radicaloxidation of unsaturated lipids. Lipids 30: 277–290.23. Marmesat S, Velasco J, Dobarganes MC. 2008. Quantitative deter-mination of epoxy acids, keto acids and hydroxy acids formed in fatsand oils at frying temperatures. J Chromatogr A 1211:129–134.24. Giuffrida F, Destaillats F, Skibsted LH, Dionisi F. 2004. Struc-tural analysis of hydroperoxy- and epoxy-triacylglycerols by liquid chro-matography mass spectrometry. Chem Phys Lipids 131:41–49.25. Velasco J, Marmesat, Bordeaux O, Marquez-Ruiz G, Dobarganes C.2004. Formation and evolution of monoepoxy fatty acids in thermoxi-dized olive and sunflower oils and quantitation in used frying oils fromrestaurants and fried-food outlets. J Agric Food Chem 52:4438–4443.26. Velasco J, Berdeaux O, Marquez-Ruiz G, Dobarganes MC. 2002.Sensitive and accurate quantitation of monoepoxy fatty acids in ther-moxidized oils by gas-liquid chromatography. J Chromatogr A 982:145–152.27. Bodin A, Linnerborg M, Nilsson JLG, Karlberg AT. 2003. Struc-ture elucidation, synthesis, and contact allergenic activity of a majorhydroperoxide formed at autoxidation of the ethoxylated surfactantC12E5. Chem Res Toxicol 16:575–582.28. Kishore RSK, Kiese S, Fischer S, Pappenberger A, Grauschopf U,Mahler HC. 2011. The degradation of polysorbates 20 and 80 and its



potential impact on the stability of biotherapeutics. Pharm Res 28:1194–1210.29. Currie F, Andersson M, Holmberg K. 2004. Oxidation of self-organized nonionic surfactants. Langmuir 20:3835–3837.30. Backtorp C, Borje A, Nilsson JLG, Karlberg AT, Norrby PO,Nyman G. 2008. Mechanisms of air oxidation of ethoxylatedsurfactants–computational estimations of energies and reaction behav-iors. Chem Eur J 14:9549–9554.31. Khursan SL, Semes’ko DG, Teregulova AN, Safiullin AN. 2008.Analysis of the reactivities of organic compounds in hydrogen atomabstraction from their CH bonds by the sulfate radical anion SO4

•−.Kinet Catal 49:202–211.32. Karnojitsky V. 1972. Biochemical importance of lipid peroxides.Russ Chem Rev 41:630–650.33. Hvattum E, Yip WL, Grace D, Dyrstad K. 2012. Characterizationof polysorbate 80 with liquid chromatography mass spectrometry andnuclear magnetic resonance spectroscopy: Specific determination of ox-idation products of thermally oxidized polysorbate 80. J Pharm BiomedAnal 62:7–16.34. Borisov OV, Ji JA, Wang YJ, Vega F, Ling VT. 2011. Toward un-derstanding molecular heterogeneity of polysorbates by application ofliquid chromatography–mass spectrometry with computer-aided dataanalysis. Anal Chem 83:3934–3942.35. Niki E. 1990. Free radical initiators as source of water- or lipid-soluble peroxyl radicals. Methods Enzymol 186:100–108.36. Zhang R, Wang Y, Tan L, Zhang HY, Yang M. 2012. Analysis ofpolysorbate 80 and its related compounds by RP-HPLC with ELSDand MS detection. J Chromatogr Sci 50:598–607.37. Ayorinde FO, Gelain SV, Johnson JH Jr, Wan LW. 2000. Analysis ofsome commercial polysorbate formulations using matrix-assisted laserdesorption/ionization time-of-flight mass spectrometry. Rapid CommunMass Spectrom 14:2116–2124.

38. Barclay LRC, Locke SJ, MacNeil JM, VanKessel J, Burton GW,Ingold KU. 1984. Autoxidation of micelles and model membranes.Quantitative kinetic measurements can be made by using eitherwater-soluble or lipid-soluble initiators with water-soluble or lipid-soluble chain-breaking antioxidants. J Am Chem Soc 106:2479–2481.39. Aizawa H. 2009. Morphology of polysorbate 80 (Tween 80) micellesin aqueous 1,4-dioxane solutions. J Appl Cryst 42:592–596.40. Oesterhelt F, Rief M, Gaub HE. 1999. Single molecule force spec-troscopy by AFM indicates helical structure of poly(ethylene-glycol) inwater. New J Phys 1:6.1–6.11.41. Wang RLC, Kreuzer HJ, Grunze M. 1997. Molecular conforma-tion and solvation of oligo(ethylene glycol)-terminated self-assembledmonolayers and their resistance to protein adsorption. J Phys Chem B101:9767–9773.42. Mei LY, McClements DJ, Wu JN, Decker EA. 1998. Iron-catalyzedlipid oxidation in emulsions as affected by surfactants, pH and NaCl.Food Chem 61:307–312.43. Waraho T, Cardenia V, Rodriguez-Estrada MT, McClements DJ,Decker EA. 2009. Prooxidant mechanisms of free fatty acids in strippedsoybean oil-in-water emulsions. J Agric Food Chem 57:7112–7117.44. McClements DJ, Decker EA. 2000. Lipid oxidation in oil-in-wateremulsions: impact of molecular environment on chemical reactions inheterogeneous food systems. J Food Sci 65:1270–1282.45. Hewitt D, Alvarez M, Robinson K, Ji J, Wang YJ, Kao YH, ZhangT. 2011. Mixed-mode and reversed-phase liquid chromatography–tandem mass spectrometry methodologies to study composition andbase hydrolysis of polysorbate 20 and 80. J Chromatogr A 1218:2138–2145.46. Harmon PA, Kosuda K, Nelson E, Mowery M, Reed RA. 2006. Anovel peroxy radical based oxidative stressing system for ranking theoxidizability of drug substances. J Pharm Sci 9:2014–2028.


本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页文献云下载图书馆入口外文数据库大全疑难文献辅助工具

http://www.xuebalib.com/cloud/

http://www.xuebalib.com/

http://www.xuebalib.com/cloud/


http://www.xuebalib.com/vip.html

http://www.xuebalib.com/db.php

http://www.xuebalib.com/zixun/2014-08-15/44.html


oxidative degradation of polysorbate surfactants studied...

Documents