Thomas Meyer1

Dietmar Waidelich2

August Wilhelm Frahm1

1Chair of PharmaceuticalChemistry,Albert-Ludwigs-University,Freiburg im Breisgau, Germany

2Applied Biosystems,PE Germany GmbH,Langen, Germany

Separation and first structure elucidation ofCremophor EL-components by hyphenatedcapillary electrophoresis and delayed extraction-matrix assisted laser desorption/ionization-timeof flight-mass spectrometry

The polyethoxylated heterogeneous components of the so far poorly characterizednonionic emulsifier Cremophor EL (polyoxyl 35 castor oil) (CrEL) were fractionatedby cyclodextrin-modified micellar electrokinetic capillary chromatography (CD-MEKC).Due to the low UV absorbance of most of the CrEL-components an indirect UV detec-tion was used with phenobarbital-sodium as background absorber. For a preciseassignment of the resulting peaks to the corresponding components capillary electro-phoresis (CE) had to be combined with delayed extraction-matrix assisted laser desorp-tion/ionization-time of flight-mass spectrometry (DE-MALDI-TOF-MS) as detectionsystem. For this purpose, the fractionating robot Probot was employed which enablesboth the on-line fractionation of the CE eluate on a MALDI target during the electro-phoretic separation and the simultaneous dosage of the MALDI matrix solution. Theapplied CrEL amount was optimized by varying the CE injection parameters time, pres-sure and concentration of the sample in order to obtain homologue peak series of suf-ficient intensity without decreasing the separation efficiency. Evaluation of the massspectra was performed by comparing the residue masses of the homologue peakseries with the calculated residue masses of potential CrEL-components. However,the high number of polyethoxylated components leads to overlapping of homologuepeak series with isobaric residue masses. These isobaric interferences were detectedby a high mass accuracy of the measurements (obtained by internal calibration withpolyethylene glycol (PEG) 1000 and by means of the residue mass plot, the newlydeveloped evaluation method. The combination of these techniques allowed the firstdetailed structure analysis of the CrEL-components showing glycerol polyoxyethylene(POE) monoricinoleate and POE monoricinoleate to be the two main components ofthe emulsifier. Furthermore, the coupling of CE with DE-MALDI-TOF-MS is generallyapplicable to the fractionation and identification of polymers.*

Keywords: Capillary electrophoresis / Cremophor EL / Delayed extraction-matrix assisted laserdesorption/ionization-mass spectrometry / Nonionic polyethoxylated surfactant / Probot EL 4841

1 Introduction

Cremophor EL (CrEL), a heterogeneous polyethoxylatednonionic surfactant, is obtained from the reaction of 1 molcastor oil (referring to the postulated main componentglycerol triricinoleate) and 35 mol ethylene oxide [1]. Thefirst identification approach performed by Müller et al. in

1966 was based on saponification, consecutive extrac-tion and thin layer chromatography of the obtained frag-ments [2]. Based on these studies, Müller postulatedglycerol polyoxyethylene (POE) ricinoleates, POE ricinole-ates, glycerol POE ether and at least unsubsituted poly-ethylene glycol (PEG) to be the main components of CrEL(Fig. 1). The exact composition especially of the ricinoleicacid containing esters was not known.

Since CrEL does not cause hemolysis despite of its sur-face active properties, it is used in pharmaceutical pre-parations for parenteral application, e.g., as solubilizer ofhydrophobic drugs like the immunosuppressant ciclos-

Correspondence: Prof. Dr. A. W. Frahm, Chair of Pharmaceuti-cal Chemistry, Albert-Ludwigs-University, Albertstr. 25, D-79104Freiburg, GermanyE-mail: awfrahm@sun2.ruf.uni-freiburg.deFax: �49-761-2036351

Abbreviations: CCaEx, cation-exchange column chromatogra-phy; CrEL, Cremophor EL; DE, delayed extraction; DHB, 2,5-dihydroxybenzoic acid; POE, polyoxyethylene

* Dedicated to Prof. Dr. G. Rücker on the occasion of the 70.anniversary of his birthday

Electrophoresis 2002, 23, 1053–1062 1053

ª WILEY-VCH Verlag GmbH, 69451 Weinheim, 2002 0173-0835/02/07-08–04–1053 $17.50+.50/0

CE

and

CE

C

Figure 1. Assumed components of CrEL according to [2].Hydrophobic portion: (a) Esters of ricinoleic acid andglycerol polyoxyethylene ethers without differentiationbetween glycerol monoricinoleate POE (R1 = R2 = H);glycerol POE diricinoleate (R1 = ricinoleate; R2 = H); gly-cerol POE triricinoleate (R1 = R2 = ricinoleate). (b) Estersof ricinoleic acid and PEG without differentiation betweenPOE monoricinoleate (R1 = R2 = H); POE diricinoleate (R1 =ricinoleate; R2 = H); POE triricinoleate (R1 = ricinoleate;R2 = ethoxylated ricinoleate). Hydrophilic portion: (c) gly-cerol polyoxyethylene ether; (d) PEG.

porin (Sandimmun ) [3]. However, serious side effects ofparenterally administered CrEL, especially anaphylacticreactions, are observed [4–7] requiring detailed informa-tion about both the chemical structures and the pharma-cokinetic properties of its components. Therefore, weelaborated a potentiometric titration method for thequantitation of CrEL in plasma of patients undergoingSandimmun therapy [8–10].

For the fractionation of CrEL-components we developeda cyclodextrin-modified micellar electrokinetic capillarychromatographic method (CD-MEKC) with a phenobarbi-tal-sodium containing running buffer, the optimization ofwhich is described in [11]. Other capillary electrophoretic(CE) separation techniques for nonionic surfactants de-scribed so far in literature [12–14] are based on buffer sys-tems with a high content of acetonitrile (40% v/v) andsodium dodecyl sulfate (SDS) (70 mM) resulting in asso-ciation complexes between the nonionic analytes andthe anionic surfactant, whereas the micelle formation ofSDS is inhibited. These methods enable good separationefficiencies for well-defined, simple-structured polymers

of the alkylphenol polyethoxylate type. However, themuch more complex mixture CrEL cannot be fractionatedwith this technique. For UV transparent polyethoxylatessuch as �,�-diamino-PEG both the indirect UV detection[15] and derivatization techniques, e.g. with 1,2,4-benze-netricarboxylic anhydride [16], are described. In the caseof CrEL, we employed the indirect UV detection withphenobarbital-sodium as background absorber [11] dueto the heterogeneity of the components which hindersa reproducible derivatization. The main problem of themethod was the impossibility to assign the UV signals tothe corresponding polyethoxylates since exactly definedreference substances are not available. Thus, the CEseparation technique had to be combined with a massspectrometric detection device.

Delayed extraction (DE)-MALDI-TOF-MS, generally apowerful method for the characterization of high-masssynthetics and biopolymers [17–19], is so far the only suit-able tool to elucidate the structures of polyethoxylatedmulticomponent mixtures. The soft ionization conditionsachieved with MALDI are of prime importance since astrong fragmentation would prevent the analysis of suchheterogeneous polymer systems. However, the applica-tion of DE-MALDI-TOF-MS to polymer systems contain-ing varying end groups proved to be rather difficult[20–22], as the high number of polyethoxylated compo-nents leads to overlapping of homologue peak serieswith isobaric residue masses. Therefore, we developedtwo new evaluation techniques, the residue mass plotand the abundance plot, which are essential for thedetection of such overlapping series [23].

The key problem of coupling CE with DE-MALDI-TOF-MSis the fractionation of the CE eluate during the electro-phoretic separation since the outlet vial has to be removedand thus the voltage to be switched off. The analyte zonemust then be eluted by applying low pressure and col-lected into a micropipet containing a minimal volume ofwater (ca. 5 �L). Thereafter, the obtained analyte solutionhas to be mixed with the matrix solution on the MALDI-TOF target. This type of off-line coupling described forthe analysis of proteins [24, 25] becomes especially trou-blesome, if the resolution of consecutive peaks is toosmall and a separate hydrodynamic elution is impossible.In addition, first attempts of an on-line coupling were made[26–28] which are, however, susceptible to interferencesand so far not suitable for commercial use.

For the CE fractionation and consecutive MS of CrEL, thefractionating robot Probot was employed representing anon-line fractionation system of the CE eluate directly on aMALDI target without interrupting the electrophoreticseparation [29], rather than a classical on-line coupling.The main module of the system is a metallic cannula with

1054 T. Meyer et al. Electrophoresis 2002, 23, 1053–1062

Figure 2. Scheme of the fractionating robot Probot(taken from [29]).

the integrated capillary outlet (Fig. 2). This cannula pro-vides simultaneously the MALDI matrix solution. In con-sequence, CE eluate and matrix solution are forming amixed droplet directly onto the MALDI target. This dropletrepresents the conductive connection between capillaryand cannula which additionally functions as cathode ofthe electrophoretic system. Thus, the application of therequired high voltage is enabled. The MALDI target isplaced on a sample table automatically moving to thenext sample position after a definite period of time,whereas the cannula including the capillary outlet is fixedpreventing negative feedback and thus an increasedbaseline noise. In this paper, we will focus on the appli-cability of Probot to the coupling between CE and DE-MALDI-TOF-MS as well as to the mass spectrometricmeasurement, calibration and evaluation technique whichshould enable both a high mass accuracy and a preciseassignment of the peak series and thus the detailedstructure analysis of CrEL.

2 Materials and methods

2.1 Chemicals

CrEL was donated by BASF AG (Ludwigshafen, Ger-many). Polyethylene glycol (PEG) 1000 used as referencesubstance was purchased from Aldrich-Chemie (Stein-heim, Germany). �-Cyclodextrin (�-CD, Cavamax W8Pharma) was of pharma grade (Wacker-Chemie, Burg-hausen, Germany) and SDS of CE-grade (Bio-RadLaboratories, München, Germany). Phenobarbital andphenobarbital-sodium (both pharma-grade) were pur-chased from Synopharm (Barsbüttel, Germany). Water

was bidistillated with a Seralpur Delta system (Seral ErichAlhäuser, Ransbach-Baumbach, Germany). Methanol(MeOH) was of HPLC-grade (Janssen Chimica, Geel, Bel-gium). Sodium acetate (NaOAc) and 2,5-dihydroxyben-zoic acid (DHB) were of analytical grade (Aldrich-Chemie).All reagents and solvents were used without further purifi-cation.

2.2 CE and fractionation of the CE eluate

CE was performed on a Crystal 310 CE instrument(ATI Unicam, Kassel, Germany) with a high-voltage powersupply (0–30 kV). On-column detection was carried outwith a Unicam 4225 UV detector using the indirect UVmode (� = 260 nm). The running buffer (pH 10.0) con-tained 20 mM phenobarbital (corresponding to 0.5071 gphenobarbital-sodium and 0.0012 g phenobarbital per100 mL buffer), 10 mM SDS, and 20 mM �-CD as aqueoussolution. The outlet buffer merely contained 20 mM phe-nobarbital as aqueous solution. Untreated fused-silicacapillaries (50 �m internal diameter, 64.5 cm effectivelength, 80.0 cm total length), obtained from PolymicroTechnologies (Phoenix, AZ, USA) were employed. Newcapillaries were first pretreated by flushing with 1 N

NaOH for 10 min, 2 N HCl for 10 min, water for 2 min, andfinally with the running buffer for 20 min. During all condi-tioning steps a pressure of 1000 mbar was applied.Between runs, the capillary was flushed with 0.1 N NaOH(10 min, 2000 mbar) followed by running buffer (5 min,2000 mbar). The temperature of the capillary was main-tained at 20�C, the voltage during CE separation wasadjusted to 15 kV resulting in a current of 6.4 �A. The runtime was 20 min. Chromatographic data were collectedby means of an Unicam 4880 Version 2.04 data system.On-line fractionation of the CE eluate was performed bythe fractionating robot Probot. Controlling of the systemwas carried out with PROBOT control Version 3.23 (bothfrom BAI, Lautertal, Germany). The matrix solution wasprepared by dissolving 750 mg DHB in 20 mL MeOH,mixing with 5 mL of a 1 mM solution of NaOAc and fillingup with water to 50 mL. The fractionation of the CE eluatestarted with a delay time of 10 min. During this period oftime the cannula with the capillary outlet dips into a vialfilled with outlet buffer. The collecting time per samplepoint was 30 s, thus 20 fractions per run were collected.One �L of the matrix solution was added per sample point(flow rate, 12 �L min–1).

2.3 DE-MALDI-TOF-MS

DE-MALDI-TOF-MS was performed with the VoyagerDE STR Workstation (Applied Biosystems, Framingham,MA, USA). The mass spectra were evaluated with

Electrophoresis 2002, 23, 1053–1062 Separation and structure elucidation of Cremophor EL-components 1055

DataExplorer Version 4.0.0.0 (Applied Biosystems). Thereflectron positive ion mode was used for all experi-ments. The total acceleration voltage was set to 20 kV,the voltage on the first grid to 68.5%. All experimentswere carried out with a delay time of 200 ns betweenion production and extraction. Six-hundred single lasershots were accumulated for each mass spectrum.

3 Results and discussion

3.1 CD-MEKC of CrEL

First, the CE fractionation of CrEL using the Probot sys-tem instead of usual outlet vial filled with running bufferwas investigated with special regard to the separationefficiency, the stability of the baseline, and possible fluc-tuations of the current. The CD-MEKC of approximately100 ng CrEL (see Section 3.2) yields under the givenconditions the five peaks CrEL 1–5 as well as two EOF-system peaks S (Fig. 3a). The assignment of the peaksto CrEL-components and to the EOF, respectively, wasverified by blank samples [11]. The reason for the appear-ance of positive peaks (CrEL 4 and 5) despite of the indi-rect UV detection mode is not yet fully understood. Prob-ably, complexes are formed between CrEL-components,�-CD, and the background absorber phenobarbital result-ing in different UV profiles. Comparison with the electro-pherogram obtained after CE fractionation using the Pro-bot system as outlet does not show significant differ-ences (Fig. 3b). Of course, longer migration times were

observed due to the approximately 20 cm lengthenedcapillary. Nevertheless, neither fluctuations of the currentnor an unstable baseline were observed. The electropher-ogram merely shows a decrease of the baseline directly infront of CrEL 4 probably caused by different conductiv-ities of buffer and sample zone which are more pro-nounced in case of the CE-Probot-coupling. Since themass spectra of the obtained fractions did not supplyany homologue peak series of the CrEL-polyethoxylates,an increase of the applied CrEL amount was necessary.

3.2 Optimization of the applied CrEL amount mi

The applied CrEL amount mi can be calculated by meansof Eq. (1) derived from Hagen-Poiseuille’s law:

mi � ��ptid4w

12800�Lges(1)

where �p is the injection pressure (mbar), ti the injec-tion time, d the inner diameter of the capillary (50 �m),w the CrEL content of the sample solution (%) (m/V),� the dynamic viscosity of the buffer (about 1), and Lges

is the total length of the capillary (80.0 cm). With regardto the injection parameters �p = 20 mbar, ti = 6 s and w =4.0%, mi is calculated as 92 ng. By individual variation ofthese parameters mi was stepwise increased up to 460 ng(Table 1). The separation efficiency was determined bythe resolution R of CrEL 4 and 5 which was known to besensitive to variations of the chromatographic conditions.The resolution was calculated with a special technique

�

Figure 3. (a) CD-MEKC fractionation of CrEL (w = 4.0% (m/V)) with a buffer filled outlet vial. Buffer:20 mM phenobarbital, pH 10.0, 10 mM SDS, 20 mM �-CD; Capillary, 62.0 cm total length, 45.5 cmeffective length; voltage, 15 kV; resulting current, 7.9 �A; separation temperature, 20�C; indirect UVdetection (� = 260 nm). – : CE peaks CrEL 1–5; S1–S2: system peaks. (b) CD-MEKC fractionationof CrEL (w = 4.0% (m/V)) with the Probot system as outlet. Conditions analogous to (a). Capillary,80.0 cm total length; 64.5 cm effective length; voltage, 15 kV; resulting current, 6.4 �A.

1056 T. Meyer et al. Electrophoresis 2002, 23, 1053–1062

Table 1. Applied CrEL amount mi and resolution R ofCrEL 4 and 5 versus the injection parameterscontent of sample solution w, injection pressure�p, and injection time ti

w (CrEL) (%)(m/V)�p = 20 mbar;ti = 0.1 min

�p (mbar)ti = 0.1 min;w = 4.0%

ti (min)�p = 20 mbar;w = 4.0%

mi (CrEL)(ng)

4.0 20 0.10 92R 1.325 1.325 1.325

8.0 40 0.20 184R 1.151 1.097 1.130

12.0 60 0.30 276R 1.184 1.060 1023

16.0 80 0.40 368R 0.905 0.910 0.957

20.0 100 0.50 460R � 0.5 0.880 0.882

The applied CrEL amount mi was calculated according toEq. (1) considering the inner diameter d (50 �m), thedynamic viscosity of the buffer � (about 1), and the totallength of the capillary Lges (80.0 cm).

developed for asymmetric peaks [30]. Table 1 showsthat R continuously worsens with increasing �p and ti,respectively. However, the best resolution (R = 1.18) con-cerning mi values � 92 ng was achieved with a 12.0%CrEL solution and unchanged �p- and ti-parametersresulting in an mi value of 276 ng. Under these conditions,homologue peak series of the CrEL-polyethoxylatesappeared for the first time in the mass spectra of the frac-tions. Since their intensities were quite low and could notbe amplified with this univariate approach, we furthertried to increase the applied CrEL amount up to 368 ngby simultaneous variation of the injection parameters.For this purpose, ti was doubled to 0.2 min and w reducedfrom 12.0 to 8.0% (�p = 20 mbar) resulting in both a sig-nificantly improved resolution of 1.43 and polymer seriesof high intensity in the corresponding mass spectra. Thus,these conditions were chosen for the following work.

3.3 Evaluation of DE-MALDI-TOF mass spectra

The assignment of a homologue peak series to the corre-sponding polyethoxylate [31] and thus the identification ofa CrEL-component is achieved by comparing the residuemass of a potential component mres, theo with the experi-mental determined residue mass mres, aver of the series.Equation (2) describes the calculation of mres, theo:

mres, theo = mend, theo � mcat – n�mmon, theo (2)

where mend,theo is the mass of the end group of a poly-ethoxylate, e.g. ricinoleic acid in the case of POE rici-noleate, mcat the mass of the cation (Na� or K�) com-plexed by the polyethoxylate, mmon, theo the mass of aC2H4O-monomer (44.026215 Da), and n the number ofthis monomer leading to an mres, theo value between 0 and44.026215 Da. This sophisticated way is required in orderto assign a homologue peak series to the correspondingpolyethoxylate since the mend, theo values cannot directlybe determined from the mass spectra.

The experimental residue masses obtained from thehomologue peak series of the mass spectra can be deter-mined by both a graphical technique, the “linear regres-sion method”, and a calculating alternative, the “aver-aging method” [32, 33]. Since the latter has been shownto be more precise [23, 33] it has been used to evaluatethe mass spectra. With this technique, first the residuemass of every single peak mres, calc within a series is calcu-lated from Eq. (3):

mres, calc = msignal – n*�mmon, theo (3)

where msignal is the mass of every single oligomer peakand n* is the number of monomer units which has to besubtracted from the mass of a single peak to obtain a resi-due mass between 0 and 44.026215 Da in analogy tomres, theo. Finally, all mres, calc values are averaged leadingto the mean mres, aver.

There are two main problems with the evaluation of themass spectra independent of the type of determinationof the residue masses: first, both the Na�- and the K�-series can principally be observed with a theoreticalmass difference �m = 15.9739 Da [34, 35] significantlycomplicating the assignment of the series [23]. However,running the CE before the mass spectrometric analysisresults in a lack of K� series due to the separation of K�

salts genuinely contained in the CrEL sample because ofthe polymerization of the emulsifier with KOH. Second,overlapping of residue masses with differences as low as0.05 Da (Table 2) occurs due to the high number of poten-tial CrEL-components and of 13C-isotopic peaks. There-fore, we developed two additional evaluation methods,the residue mass plot and the abundance plot [23] wherethe residue masses mres, calc and the abundances of allsignals, respectively, are plotted as a function of n*. Theresidue mass plot shows a parallel to the abscissa inthe case of an undisturbed series. An overlapping of aseries is indicated by a continuous increase or by adecrease of mres, calc depending on the residue mass ofthe interfering series. This interpretation can be verifiedby the abundance plot which supplies a multimodal dis-tribution in the case of overlapping related to the numberof interferences. The residue mass plot allows an inter-

Electrophoresis 2002, 23, 1053–1062 Separation and structure elucidation of Cremophor EL-components 1057

Table 2. Residue masses mres, theo of potentialCrEL-components

Alkanolic part Acidic part Alkali cation mres, theo (Da)

POE ric a) K� 29.0310POE ric/ol b) K� 29.1190POE ric/ric Na� 29.1400glyc POE c) ric/ric/lin d) K� 29.1707POE ric/ric/ol Na� 29.2280

The residue masses mres, theo were calculated according toEq. (2) with respect to the composition of genuine castoroil which contains about 85% ricinoleic acid, 7% oleicacid, 3% linoleic acid, 2% palmitic acid and 1% stearicacid.a) Ricinoleic acidb) Oleic acidc) Ether of glycerol with PEGd) Linoleic acid

pretation without any reservation, whereas the applica-tion of the abundance plot requires strong intensities ofthe series.

3.4 Calibration of DE-MALDI-TOF mass spectra

The main requirement for the applicability of the residuemass plot is a high mass accuracy achieved by internalcalibration with polymer standards. In previous measure-ments of two CrEL-fractions obtained by cation-ex-change column chromatography (CCaEx) internal calibra-tion was carried out with a mixture of polypropyleneglycol(PPG) 1000, 2000 and PEG 3000 resulting in a mass accu-racy of 0.0013 Da [23]. However, in the case of the CEProbot coupling the intensities of the polymer series aremuch lower compared to the intensities of these two frac-tions despite of the method optimization undertaken toincrease the applied CrEL amount (see Section 3.2). Inconsequence, addition of an internal standard to thematrix solution would completely suppress the homol-ogue peak series of the CrEL-components.

Therefore, a first measurement was taken with internalcalibration by means of the Na� adduct signal of �-CD(m = 1319.4126 Da) (Fig. 4) the accuracy of which exclu-sively allows the determination of the nominal residuemasses mres, nom of the series and not of mres, aver values.The assignment of the polymer series to the polyethoxy-lates was then indirectly achieved by comparing themwith the corresponding series of the two CCaEx fractionswhich have been precisely evaluated by the residuemass plots. This procedure was simplified due to themissing K� series and thus the reduced number of over-lapping series.

Figure 4. DE-MALDI-TOF-mass spectrum of the Probotfraction 12. (�) Na� �-CD-signal (m = 1319.4126 Da)(internal calibration used), (�) series with the nominal resi-due mass 13 Da � Na� POE monoricinoleate, (�) serieswith the nominal residue mass 43 Da � Na� glycerol POEmonoricinoleate.

In a second series of experiments the Probot fractions onthe MALDI target were spiked with a very small amount ofPEG 1000 standard in order to determine the mres, aver

values of the series and thus to verify the results fromthe first measurements. The polyethoxylate series wereassigned to the CE peaks by calculating the times of elu-tion from the migration times considering the effectiveand total length of the capillary. The fractionation wasstarted with a delay time of 10 min since no signalappears within this period of time. Then 20 fractionswere collected.

3.5 DE-MALDI-TOF-MS measurements undercalibration with Na� �-CD

Table 3 shows the polymer series identified in the Probotfractions. Fractions 1–6 are free of any polymer seriessince the first CrEL-components elute approximately13.5 min after starting the CE fractionation. In analogy tothe principle of CD-MEKC, the migration times are grow-ing with increasing lipophilicity of the separated compo-nents. The hydrophilic polyethoxylates PEG and glycerolPOE ether are found in fractions 7–10, whereas fraction19 only contains polymers with completely esterifiedhydroxyl groups. In fraction 20 polymer series are notany longer observed.

POE monoricinoleate (mres, nom = 13 Da) and glycerol POEmonoricinoleate (mres, nom = 43 Da) take special positionssince they appear in nearly all fractions (Fig. 4). On the onehand, these monoesters are quite hydrophilic due to theirunderivatized hydroxyl groups. Thus, they are alreadyobserved in fractions 7–10 together with PEG and gly-cerol POE ether. On the other hand, they are able to inter-

1058 T. Meyer et al. Electrophoresis 2002, 23, 1053–1062

Table 3. Polyethoxylate series identified in the Probot fractions 1–20

Fraction mres, noma)b) Component c) Distributionb) Max.b)d) Int.e)

1� 6 Free of polyethoxylate series

7 27 glyc POE ether 1039–1744 1304 551041 PEG 833–1185 965 248243 glyc POE 1ric 1364–1980 1628 758

8 13 POE 1ric 937–1422 1113 249027 glyc POE ether 1039–1744 1304 332941 PEG 833–1097 965 147543 glyc POE 1ric 1364–2068 1628 3727

9 13 POE 1ric 981–1245 1113 74027 glyc POE ether 1083–1303 1304 84241 PEG 657–1142 789 2734

10 27 glyc POE ether 955–1656 1259 328541 PEG 613–1142 833 4481

11 13 POE 1ric 893–1421 1113 395143 glyc POE 1ric 1363–1980 1628 3015

12 13 POE 1ric 805–1334 1113 330727 glyc POE 1ol 1392–1920 1612 164041 POE 1ol 921–1142 1053 147543 glyc POE 1ric 1100–2068 1584 1963

13 1 POE 3ric 1278–2070 1586 360813 POE 1ric 849–1950 1113 425727 glyc POE 1ol 1040–2052 1524 619129 POE 2ric 1218–2054 1570 218741 POE 1ol 877–1670 1142 198843 glyc POE 1ric 968–2068 1628 4502

14 15 glyc POE 2ric 688–2525 2128 30031 glyc POE 3ric 1924–2056 1968 9641 POE 1ol 833–1318 1054 1479

Fraction mres,noma)b) Component c) Distributionb) Max.b)d) Int.e)

15 13 POE 1ric 938–1466 1069 328915 glyc POE 2ric 1644–2393 1996 41427 glyc POE 1ol 951–1744 1304 399841 POE 1ol 833–1274 965 318643 glyc POE 1ric 1232–2025 1584 1841

16 15 glyc POE 2ric 1600–2436 1952 117629 POE 2ric 1130–1878 1305 104131 glyc POE 3ric 1704–2541 2277 34243 glyc POE 1ric 1584–2245 1892 331

17 15 glyc POE 2ric 1600–2349 1952 149529 POE 2ric 954–1967 1305 98631 glyc POE 3ric 2012–2585 2233 402

18 15 glyc POE 2ric 1644–2481 1908 100429 POE 2ric 734–1878 1262 89331 glyc POE 3ric 1792–2673 2189 344

19 15 glyc POE 2ric 1556–2392 1952 101529 POE 2ric 865–1746 1262 168731 glyc POE 3ric 1837–2541 2233 349

20 Free of polyethoxylate series

The collecting time per sample point was 30 s; the fractio-nation of the CE eluate started with a delay time of 10 min.a) mres, nom: nominal residue mass. The nominal residue

masses are calculated according to Eq. (2) with Na�

as complexing cation.b) All masses are given in Da.c) For abbreviations see Table 2.d) Maximum: mass of the most intensive peak within a

homologue series.e) Intensity: height of the most intensive peak within a

homologue series.

act with the more hydrophobic polyethoxylates due totheir lipophilic constituent ricinoleic acid. The intensitiesof the two Na� series (Table 3) identify POE monoricinole-ate and glycerol POE monoricinoleate to constitute themain components of CrEL, whereas the so far postulatedlead component glycerol POE triricinoleate and the otherdi- and triesters play only a minor role.

Furthermore, Table 3 shows that the series with the mres, nom

values 41 and 27 Da include each two polyethoxylatecomponents: both the Na� PEG (mres, theo = 41.0003 Da)and the Na� POE monooleate (mres, theo = 41.0884 Da)can be assigned to the series with the nominal residuemass of 41 Da and both the Na� glycerol POE ether(mres, theo = 26.9847 Da) and the Na� glycerol POE mono-oleate (mres, theo = 27.0727 Da) to the series with the nom-inal residue mass of 27 Da. The nominal residue massesdo not allow a differentiation, but shifts in both the mole-cular mass distribution and the distribution maximum of

the series to higher masses are observed for the Probotfractions 10 and 12, respectively. These shifts as wellas significant changes in the residue mass plots wereobserved for the corresponding series of the two CCaEx

fractions [23] clearly indicating the presence of the fourcomponent series mentioned above. Therewith both thehydrophilic PEG and the glycerol POE ether are elutedup to and including Probot fraction 10, whereas both thePOE monooleate and the glycerol POE monooleate areeluted beginning with fraction 12. The intensities of theirhomologue peak series show that both monooleates arecontained in CrEL to a significant extent, whereas theywere not considered by Müller [2].

The assignment of the identified polyethoxylates to thecorresponding CE peaks (Table 4) finally shows that noneof the peaks is caused by a single CrEL component but byvarious polymers probably forming mixed micelles. Theassignment of PEG to the peaks CrEL 1 and 2 is sup-

Electrophoresis 2002, 23, 1053–1062 Separation and structure elucidation of Cremophor EL-components 1059

Table 4. CrEL components identified in CE peaksobtained with indirect UV detection

CE peak CrEL component

CrEL 1 Glycerol POE etherPEG

Glycerol POEmonoricinoleate

CrEL 2 POE monoricinoleate Glycerol POE etherPEG

CrEL 2/3 a) POE monoricinoleate Glycerol POE monooleatePOE monooleate Glycerol POE

monoricinoleate

CrEL 3 POE triricinoleate POE monoricinoleateGlycerol POE monooleate POE diricinoleatePOE monooleate Glycerol POE

monoricinoleateCrEL 3/4 a) POE monooleate

CrEL 4 POE monoricinoleate Glycerol POE monooleatePOE monooleate Glycerol POE

monoricinoleate

CrEL 5 Glycerol POE diricinoleate POE diricinoleateGlycerol POE triricinoleate

a) CrEL 2/3, 3/4 are the baseline interval zones betweenthe peaks CrEL 2 and 3 and CrEL 3 and 4, respectively.

ported by the fact that these peaks increase in aqueousCrEL solutions after longer storage leading to hydrolysisof POE ricinoleates. The mass spectra of the fractions inthe range between the peaks CrEL 2 and 4 supply a multi-tude of intensive series, whereas the electropherogramonly shows the weak signalCrEL 3 surrounded by the inter-vals 2/3 and 3/4. This discrepancy could be explained bymeans of overlapping of negative and positive signalsresulting in the baseline intervals 2/3 and 3/4, respectively.

3.6 DE-MALDI-TOF-MS measurements undercalibration with Na� PEG 1000

As mentioned in Section 3.4, an addition of a standardpolymer to the matrix solution completely suppressesthe polymer signals of the CrEL-components. Thus, theProbot fractions were spiked directly on the MALDI targetwith 1.0 �L of a PEG 1000 solution (3:1 v/v methanol/water mixture) each. After gradually reducing the concen-tration of PEG 1000 to 10 �gmL–1 corresponding to anabsolute amount of 10 ng homologue peak series of theCrEL-components were observed for the first time withinthis series of measurements. Compared to the experi-ments without internal standard, the intensities of the sig-nals were weakened but sufficiently strong for the evalua-tion of most series. Table 5 shows the characteristics ofthose series, the intensities of which allowed a precisecalculation of the mres, aver values as well as an evaluationvia the residue mass plots. The mass accuracies of thePEG 1000 series ranging from 0.0005 to 0.0031 Da weresimilar to those obtained for the two CCaEx fractions.Furthermore, the residue mass plot again proved to be aprecise technique for the identification of overlappingpeak series. Figure 5 exemplarily shows the residuemass plot of Na� glycerol POE monoricinoleate (mres, nom

= 43 Da) obtained from fraction 10 clearly indicating thelack of interferences in the range 30 � n* � 42. In con-sequence, a significantly better standard deviation of�/�0.0037 Da is achieved if calculated only for the giveninterval (�/�0.0191 Da for the whole series).

The Na� series of POE triricinoleate (mres, nom = 1 Da) andglycerol POE monooleate (mres, nom = 27 Da) could not beverified by mres, aver values and by the corresponding resi-due mass plots due to the very low intensities of these

Table 5. Na� polyethoxylate series from Probot fractions identified by mres, aver values

Component Fractiona) mend, theob) c) mres, theo

b) d) mres,averb) e) mmono,aver

b) f) Distributionb) Maximumb) g)

Glycerol POE ether 9 92.0474 26.9857 26.9971�/� 0.0236 44.0309�/� 0.0340 952–1700 1260/26POE monoricinoleate 10 298.2508 13.0571 13.0467�/� 0.0161 44.0250�/� 0.0219 938–1554 1202/20Glycerol POE monoricinoleate 10 372.2876 43.0676 43.0693�/� 0.0191 44.0235�/� 0.0240 1364–2068 1672/29Glycerol POE diricinoleate 16 652.5278 15.1243 15.1445�/� 0.0210 44.0263�/� 0.0261 1556–2437 1908/28Glycerol POE triricinoleate 16 932.7680 31.2073 31.1952�/� 0.0294 44.0316�/� 0.0395 1924–2453 2277/30POE diricinoleate 16 578.4911 29.1400 29.1455�/� 0.0240 44.0228�/� 0.0313 1086–1878 1306/16

Measurement under internal calibration with Na� PEG 1000.a) Probot fractions giving rise to the respective series.b) All masses are given in Da.c) mend,theo: mass of the nonethoxylated end group of a component.d) mres,theo: theoretical residue mass of a component calculated with Eq. (2).e) mres,aver: experimentally determined residue mass of a component calculated with Eq. (3).f) mmon,aver: mass of an ethylene oxide unit calculated from the consecutive peaks within a homologue series.g) Maximum: mass and corresponding number of ethylene oxide units of the most intensive peak within a homologue

series.

1060 T. Meyer et al. Electrophoresis 2002, 23, 1053–1062

Figure 5. Residue mass plot of the series with the nom-inal residue mass 43 Da (Na� glycerol POE monoricinole-ate) exhibiting the residue masses mres, calc (calculatedaccording to Eq. 3) of the single peaks of a homologueseries in dependence of the number of the ethylene oxideunits n*.

series. In analogy, the mres, aver value of Na� POE mono-oleate (mres, theo = 41.0884 Da) could not be determinedsince the series is clearly dominated by the Na� series ofthe internal standard PEG 1000 (mres, theo = 41.0003 Da).However, the residue mass plot allows the identificationof POE monooleate as CrEL component, in that its serieswith the nominal residue mass of 41 Da obtained fromfraction 10 (Fig. 6a) shows a good correlation betweenthe mres, calc values and mres, theo of Na� PEG 1000. Merelythe last three signals deviate to lower masses probablydue to their low intensities. However, the correspondingseries obtained from fraction 14 (Fig. 6b) shows thatonly the range 14 � n* � 18 is dominated by the internalstandard PEG 1000, whereas the higher molecular massrange is clearly overlapped by Na� POE monooleate in-dicated by continuously increasing mres, calc values. Thisinterpretation shows the molecular mass distribution ofPEG to be located at lower masses than that of POEmonooleate. This finding correlates with the results ob-tained from the two CCaEx fractions and from the measure-ments of the CE-Probot-coupling under Na� �-CD cali-bration (see Section 3.5).

Finally, the mean degree of ethoxylation of the evaluatedseries was compared with the values postulated by Müller[2], who assumed the hydroxyl groups of glycerol to beethoxylated with 10–11 EO units each (Fig. 1). In conse-quence, the glycerol POE esters and glycerol POE ethershould be ethoxylated 30–33-fold, the POE esters merely10–11-fold. The results of our measurements (Table 5)only prove the postulates concerning the glycerol con-taining polyethoxylates, whereas the POE ricinoleatesare ethoxylated to a major degree with 16–20 EO units. Inanalogy, the molecular mass of the underivatized PEGsignificantly deviates from the postulated value of 600 Da

Figure 6. Residue mass plot of the series with the nom-inal residue mass 41 Da (Na� series of PEG 1000)obtained from (a) Probot fraction 10 and (b) fraction 14,respectively.

[2]. The mass spectra of the Probot fractions showedmaximum masses ranging from about 800 to 950 Da(Table 3). In our ongoing work, we are involved in the elu-cidation of the structures of two further conspicuouspolymer series which could so far not be assigned to anypolyethoxylates.

4 Concluding remarks

In this work, a new coupling technique between CE andDE-MALDI-TOF-MS and its applicability to the separationof the heterogeneous nonionic ethoxylated emulsifierCrEL is described. In contrast to the well-known off-linecoupling techniques working with the hydrodynamic elu-tion of the analytes, the fractionating robot Probot allowsthe on-line fractionation of the CE eluate directly on aMALDI target without interrupting the electrophoreticseparation. Furthermore, the MALDI matrix solution issimultaneously added to the CE eluate resulting in amixed droplet of the two solutions. This droplet repre-sents the conductive connection between a metallic can-nula working as cathode and the capillary. Thus, runningthe CE is enabled without any fluctuations of the current.

Electrophoresis 2002, 23, 1053–1062 Separation and structure elucidation of Cremophor EL-components 1061

After increasing the applied CrEL amount from 92 to368 ng by optimizing the injection parameters the massspectra of the fractions supplied polymer series of suffi-cient intensity. These series were first indirectly assignedto the polyethoxylates by comparing the nominal residuemasses of the series with the corresponding series of twopreviously obtained CrEL fractions which had been pre-cisely evaluated. In a second measurement series theProbot fractions were spiked on the MALDI target with10 ng of PEG 1000 each. The achieved high mass accura-cies of the resulting spectra allowed the verification of thefirst results by calculating the exact residue masses of theseries and by comparing them with the theoretical values.An overlapping of different series was identified by theresidue mass plot, the newly developed evaluation tech-nique. Thus, the structures of the most important CrELcomponents, glycerol POE- and POE-monoricinoleate,glycerol POE- and POE-monooleate, glycerol POE ether,and at least of free PEG, were precisely elucidated for thefirst time. Furthermore, the described coupling techniquegenerally describes a system applicable to the fractiona-tion and the consecutive identification of heterogeneoussynthetic polymers.

We gratefully acknowledge the financial support fromFonds der Chemischen Industrie (Frankfurt am Main, Ger-many). We also would like to thank Applied BiosystemsGmbH (Langen, Germany) for extensive support inMALDI-TOF-MS and BASF AG (Ludwigshafen, Germany)as well as ICI Inc. (Essen, Germany) for generous dona-tions of CrEL and PEGs.

Received August 28, 2001

5 References

[1] BASF AG, Technisches Merkblatt: Cremophor EL, Ludwigs-hafen 1987.

[2] Müller, K., Tenside 1966, 2, 37–45.[3] Nürnberg, E., Surmann, P., Hagers Handbuch der Pharma-

zeutischen Praxis, 5th ed., Springer, Berlin/Heidelberg 1991,p. 767.

[4] Dye, D., Watkins, J., Br. Med. J. 1980, 280, 1353–1357.[5] Moneret-Vautrin, D. A., Laxenaire, C., Viry-Babel, F., Br. J.

Anaesth. 1983, 55, 469–471.[6] Howrie, L., Ptachcinski, R. J., Griffith, B. P., Hardesty, R. J.,

Rosenthal, J. T., Burckart, G. J., Venkataramanan, R., DrugIntell. Clin. Pharm. 1985, 19, 425–427.

[7] Habazettl, H., Vollmar, B., Röhrich, F., Conzen, P., Doenicke,A., Baethmann, A., Anaesthesist 1992, 41, 448–456.

[8] Kunkel, M., Böhler, J., Keller, E., Frahm, A. W., Pharmazie1998, 53, 314–321.

[9] Kunkel, M., Meyer, Th., Böhler, J., Keller, E., Frahm, A. W.,J. Pharmaceut. Biomed. 1999, 21, 911–922.

[10] Meyer, Th., Böhler, J., Frahm, A. W., J. Pharmaceut. Biomed.2001, 24, 495–506.

[11] Kunkel, M., Böhler, J., Keller, E., Frahm, A. W., Pharmazie1997, 52, 109–121.

[12] Heinig, K., Vogt, C., Werner, G., J. Chromatogr. A 1996, 745,281–292.

[13] Heinig, K., Vogt, C., Werner, G., Fresenius’ J. Anal. Chem.1997, 357, 695–700.

[14] Vogt, C., Heinig, K., Tenside Surf. Det. 1998, 35, 470–475.[15] Stefani, M., Negrell, C., Braud, C., Vert, M., J. Chromatogr. A

1999, 864, 335–344.[16] Barry, J. P., Radtke, D. R., Carton, W. J., Anselmo, R. T.,

Evans, J. V., J. Chromatogr. A 1998, 800, 13–19.[17] Karas, M., Hillenkamp, F., Anal. Chem. 1988, 60, 2299–2301.[18] Bahr, U., Deppe, A., Karas, M., Hillenkamp, H., Gießmann,

U., Anal. Chem. 1992, 64, 2866–2869.[19] Karas, M., Bahr, U., Deppe, A., Stahl, B., Hillenkamp, H.,

Gießmann, U., Macromol. Symp. 1992, 61, 397–404.[20] Jackson, A. T., Jennings. K. R., Scrivens, J. H., J. Am. Soc.

Mass Spectrom. 1996, 8, 132–139.[21] Juhasz, P., Martin, S. A., Vestal, M. L., Proceedings of the

44th ASMS Conference on Mass Spectrometry and AlliedTopics, Portland, OR 1996.

[22] Montaudo, G., Montaudo, M. S., Pugliesi, C., Samperi, F.,Rapid Commun. Mass Spectrom. 1994, 8, 981–984.

[23] Meyer, T., Kunkel, M., Frahm, A. W., Waidelich, D., J. Am.Soc. Mass Spectrom. 2001, 12, 911–925.

[24] Chakel, J., Pungor, E., Hancock, W., Svedberg, S. A., J.Chromatogr. B 1997, 689, 215–220.

[25] Choudhary, G., Chakel, J., Hancock, W., Torres-Duarte, A.,McMahon, G., Wainer, I., Anal. Chem. 1999, 71, 855–859.

[26] He, L., Li, L., Lubman, D. M., Anal. Chem. 1995, 67, 4127–4132.

[27] Chang, S. Y., Yeung, E. S., Anal. Chem. 1997, 69, 2251–2257.

[28] Preisler, J., Foret, F., Karger, B. L., Anal. Chem. 1998, 70,5278–5287.

[29] BAI GmbH, Manual: PROBOT – Sample robot for HPLC andCZE, Version 3.2, Lautertal (Germany) 1999.

[30] Schoenemakers, P. J., Strasters, J. K., Bartha, A., J. Chro-matogr. 1988, 458, 355–370.

[31] de Koster, C. G., Duursma, M. C., van Rooij, G. J., Heeren,R. M. A., Boon, J. J., Rapid Commun. Mass Spectrom.1995, 9, 957–962.

[32] van Rooij, G. J., Duursma, M. C., Heeren, R. M. A., Boon,J. J., de Koster, C. G., J. Am. Soc. Mass Spectrom. 1996,7, 449–457.

[33] Koster, S., Duursma, M. C., Boon, J. J., Heeren, R. M. A.,J. Am. Soc. Mass Spectrom. 2000, 11, 536–543.

[34] Hoberg, A.-M., Haddleton, D. M., Derrick, P. J., Jackson,A. T., Scrivens, J. H., Eur. Mass Spectrom. 1998, 4, 435–440.

[35] Wang, Y., Rashidzadeh, H., Baochuan, G., J. Am. Soc. MassSpectrom. 2000, 11, 639–643.

1062 T. Meyer et al. Electrophoresis 2002, 23, 1053–1062