c h roma t o g ra p h i c s c i e nc e - harvard apparatus...r.p.w. scott a.m. siouffi larry t....

A Study of Polyethoxylated Alkylphenols by Packed Column SupercriticalFluid ChromatographyB.J. Hoffman and L.T. Taylor ..........................................................................................................61

An Isocratic Liquid Chromatographic Method with Diode-Array Detection forthe Simultaneous Determination of -Tocopherol, Retinol, andFive Carotenoids in Human SerumS. Gueghuen, B. Herbeth, G. Siest, and P. Leroy ............................................................................69

A Procedure for Sampling and Analysis of Air for Energetics andRelated CompoundsM.A. Hable, J.B. Sutphin, C.G. Oliver, R.M. McKenzie, E.F. Gordon, and R.W. Bishop..................77

Displacement Study on a Vancomycin-Based Stationary Phase UsingN-Acetyl-D-Alanine as a Competing AgentI. Slama, C. Ravelet, A. Villet, A. Ravel, C. Grosset, and E. Peyrin..................................................83

Influence of Hydrolysis, Purification, and Calibration Method onFurosine Determination Using Ion-Pair Reversed-Phase High-PerformanceLiquid ChromatographyM.A. Serrano, G. Castillo, M.M. Muñoz, and A. Hernández ..........................................................87

The Use of Nonendcapped C18 Columns in the Cleanup of Clenbuterol and aNew Adrenergic Agonist from Bovine Liver by Gas Chromatography–TandemMass Spectrometry AnalysisM. Fiori, C. Cartoni, B. Bocca, and G. Brambilla ...........................................................................92

Simultaneous High-Performance Liquid Chromatographic Determination ofParacetamol, Phenylephrine HCl, and Chlorpheniramine Maleate inPharmaceutical Dosage FormsH. Senyuva and T. Özden..............................................................................................................97

Evaluation of Select Variables in the Ion Chromatographic Determination ofF–, Cl–, Br–, NO3

–, SO4–2, and PO4

–3 in Serum SamplesZ. Benzo, A. Escalona, J. Salas, C. Gómez, M. Quintal, E. Marcano, F. Ruiz, A. Garaboto,and F. Bartoli...............................................................................................................................101

Temperature Effect on Peak Width and Column Efficiency in SubcriticalWater ChromatographyY. Yang, L.J. Lamm, P. He, and T. Kondo......................................................................................107

A High-Pressure Liquid Chromatographic–Tandem Mass SpectrometricMethod for the Determination of Ethambutol in Human Plasma,Bronchoalveolar Lavage Fluid, and Alveolar CellsJ.E. Conte, Jr., E. Lin, Y. Zhao, and E. Zurlinden ...........................................................................113

Departments

Cover: Photomicrograph of MoS2 by Michael W. Davidson, National Magnetic Field Laboratory, Florida State University atTallahassee. Microscopes provided by Nikon Instrument Group, Melville, NY.

To Subscribe Call (847) 647-2900 x322 • Fax (847) 647-1155 • Web www. j - c h ro m - s c i . c o m

Chromatography Problem Solving & Troubleshooting .........................119New Products .................................................3A

New Literature................................................5A

New Software .................................................6A

Meetings ..........................................................8A

Ad Index ..........................................................8A

Contents

C hroma t o gra p h i c S c i enc eJOURNAL OF Vol. 40, No. 2 FEBRUARY 2002Preston Publications

Div. of Preston Industries, Inc.6600 W. Touhy Ave., Niles, IL 60714

Phone (847) 647-2900, Fax (847) 647-1155www.j-chrom-sci.com

President: S. Tinsley Preston, III

PublisherS. Tinsley Preston, III

Co-EditorsBert M. Gordon Huba Kalász

Managing EditorJ. Eric Brewer

[email protected]

Editorial AssistantKevin Bailey

Associate EditorsDean Rood Roger K. Gilpin

Brian A. Bidlingmeyer

Advertising SalesS. Tinsley Preston, III

[email protected]

Director of Marketing ServicesJanice Gordon

ProductionRoberta Knight, Manager

Dana Neiman

ArtLynne Surma, Director

Stephanie Graffuis-Cain, WebMasterPamela Kintzel

Editorial Advisory Board

Editorial contributions, manuscripts (6 copies), letters tothe editor, photos, etc. to J C S Edit. Dept., P.O. Box48312, Niles, IL 60714-0312. Materials accepted forpublication subject to revision, at publisher’s discretion,to meet editorial standards/style. The act of mailingmanuscripts, letters, photos, and/or materials shallconstitute an express warranty by the contributor that thematerial is original, has not been published/submittedelsewhere in similar form, and is in no way an infringe-ment upon the rights of others. Publisher makes everyeffort to ensure careful handling of all photos, but is notresponsible for incidental loss/damage. Mention of anymethod/product does not constitute endorsement by J C S.

LIST RENTAL: Janice Gordon, 6600 W. Touhy Av e . ,Niles, IL 60714, Phone: 847/647-2900 x321, e-mail:j g o r d o n @ j - c h r o - s c i . c o m .SUBSCRIPTIONS: JCS Circ. Dept, P.O. Box 48312, Niles,IL 60714-0312; phone (847) 647-2900; fax (847) 647-1155; [email protected]. U.S.—1 yr/ $ 3 3 5 ;2 yr/$560. Outside U.S.—1 yr /$350; 2 yr/$590 (U.S.funds/U.S. bank). First-Class/Airmail service available at$ 90/yr additional. VISA, MasterCard, and American Expressaccepted. Single copy: $35. Back issues, microfiche/microfilm, and reprints available from publisher.

JOURNAL OF CHROMATOGRAPHIC SCIENCE (I S S N0 0 2 1 - 9 6 6 5): Published monthly, by Preston Publications,D i v. Preston Industries, Inc., 6600 W. Touhy Ave., Niles,IL 60714-4588. Periodicals postage paid at Chicago andadditional mailing offices. Copyright 2002; all rightsreserved. Registered with Copyright Clearance Center.(Printed on acid-free paper effective Vol. 23, No. 2.)

POSTMASTER: Send address changes to: JOURNAL OFC H R O M ATOGRAPHIC SCIENCE, P.O. Box 48312,Niles, IL 60714-0312.

Printed in U.S.A. by St. Croix Press, Inc. Canadian Publications Agreement #1559842

Lars BlombergPhyllis R. Brown

Kenneth A. CohenTibor Cserháti

Neil D. DanielsonWilliam A. DarkGerald D. Dupré

R. GilpinG. GuiochonJaroslav Janák

Kiyokatsu JinnoS. Bart Jones

R. KaliszanJ.J. KirklandS.F.Y. LiC.E. LinC.H. LochmüllerFernando M. LançasDavid C. LockeFred RabelM.L. RiekkolaR.P.W. ScottA.M. SiouffiLarry T. Taylor

Donald E. Willis

mailto:[email protected]






http://www.j-chrom-sci.com


http://www.j-chrom-sci.com

61

Alkylphenol polyethoxylates (APEs) are a widely used group ofnonionic surfactants in commercial production. Characterization ofthe composition of APE mixtures can be exploited for thedetermination of their most effective uses. In this study samplemixtures contain nonylphenol polyethoxylates and octylphenolpolyethoxylates. The separation of individual alkylphenols byethoxylate units is performed by supercritical fluid chromatography(SFC)-UV as well as normal-phase high-performance liquidchromatographic (HPLC)-UV employing packed columns. Thestationary phase and column length are varied in the SFC setup toproduce the most favorable separation conditions. Additionally,combinations of packed columns of different stationary phases aretested. The combination of a diol and a cyano column is found toproduce optimal results. An advantage of using packed columnsinstead of capillary columns is the ability to inject large amounts ofsample and thus collect eluted fractions. In this regard, fractionsfrom SFC runs are collected and analyzed by flow injectionanalysis–electrospray ionization–mass spectroscopy in order topositively identify the composition of the fractions. In comparingthe separation of APE mixtures by SFC and HPLC, it is found thatSFC provides shorter retention times with similar resolution. Inaddition, less solvent waste is produced using SFC.

Introduction

Alkylphenol polyethoxylates (APEs) are referred to as nonionicsurfactants. Since the mid 1940s, APEs have been used commer-cially for their surfactant ability. The term surfactant includessurface-active compounds characterized by their ability to con-centrate at surfaces and form micelles in solution (1). They havebeen used in a wide variety of applications including industrialprocess aids, dispensing agents in paper and pulp production,emulsifying agents in latex paints and pesticide formulations,flotation agents, industrial cleaners (metal surfaces, textile pro-cessing, and food industry), and household cleaners (1). Thesecompounds are commercially available as oligomeric mixtureswith varying ethoxylate chain lengths as well as varying alkylsizes. Certain APEs have been determined to be estrogenic in fish,birds, and mammals (2).

APEs contain two main molecular regions: the polyethoxylate(POE) chain (EO) is polar and thus hydrophilic and the alkyl-phenol is the hydrophobic area. The hydrophilic nature of the EOis attributed to the hydration of the ether-linked oxygen atoms(3). A technical synthesis of APEs start with phenol, which is alky-lated by trimethylpentane and thus produces octylphenol (OP), orby nonene isomers, which forms nonylphenol (NP) in an acid-cat-alyzed process. Ethoxylation is performed by using KOH–ethanolas a catalyst with a known ratio of ethylene oxide to thealkylphenol (1). The reaction results in an oligomeric mixture ofthe alkylphenol containing an EO chain of varying lengths.

The separation and identification of the components of an APEmixture can be useful for the determination of their most effec-tive applications. Several different types of chromatography havebeen studied previously in efforts to achieve better separationconditions. Gas chromatography (GC) coupled with flame ioniza-tion detection as well as mass spectrometry (MS) has been used inthe analysis of APEs (4). Isomers of each oligomer tend to be sep-arated into clusters by GC. Usually, it is necessary to derivatizesamples containing APEs for analysis by GC, because the com-pounds are not very volatile. GC poorly separates higher molec-ular-weight oligomers because of their lower volatility.

High-performance liquid chromatography (HPLC) has beenused to separate APEs of higher mass oligomers. Both reversed-phase (3) and normal-phase (5–7) chromatographic separationshave been performed on solutions containing APEs. Eacholigomer is separated by an ethoxylate unit, and isomers of eacholigomer tend to coelute. Recently, Gundersen used a graphiticcarbon column in research to separate isomers of individualethoxylated alkylphenols by HPLC (8). Ferguson et al. usedreversed-phase HPLC–electrospray ionization (ESI)–MS to ana-lyze APEs and their metabolites in aquatic environments (9).Normal-phase HPLC–ESI–MS was used by Shang et al. to quanti-tate NPEOs in marine sediment (10).

In addition to traditional forms of chromatography, supercrit-ical fluid chromatography (SFC) has been employed for APE sep-aration. SFC has advantages over both HPLC and GC. SFC canoperate at lower temperatures than GC, allowing samples that arethermally labile to be analyzed. Supercritical fluids have densitiessimilar to liquids and diffusivities similar to gases. These qualitiesallow large molecular-weight molecules that are not volatile to be

Abstract

A Study of Polyethoxylated Alkylphenols by PackedColumn Supercritical Fluid Chromatography

Brian J. Hoffman and Larry T. TaylorVirginia Tech, Department of Chemistry, Blacksburg, VA 24061-0212

Reproduction (photocopying) of editorial content of this journal is prohibited without publisher’s permission.

Journal of Chromatographic Science, Vol. 40, February 2002


62

separated by SFC similar to HPLC but with shorter retentiontimes because of the physical properties of supercritical fluids.This reduces solvent waste and decreases the total analysis time.Capillary-column SFC using flame ionization detection (11,12)has been used to separate both NPEO and OPEO. Because asample is generally destroyed by this method, it is not possible todirectly determine analyte identity. Peak identity can be surmisedby comparing retention times of samples with other APE mix-tures that contain a large fraction of a known single oligomer. Adisadvantage associated with capillary columns is the inability toinject large sample volumes, which precludes semipreparativefraction collection.

In addition, OPEO mixtures have been separated on packed-column SFC using reversed-phase (13,14) and normal-phase(15,16) packing material. Both Takeuchi and Saito and Giorgettieet al. used C18 packed columns to separate OPEO samples bySFC. Takeuchi and Saito found that a microcolumn (1.0 × 500mm) had the best separation performance, but a semimicro-column (1.7 × 250 mm) produced the best results. A conventionalcolumn (6.0 × 250 mm) was used in their research for preparativepurposes. Packed-column SFC allows larger amounts of sampleto be injected into the system for the semipreparative collectionof analyte fractions. Giorgettie et al. studied mixed mobile phasesusing the addition of a modifier in order to make their mobilephase more polar. They used pressure programming and a modi-fier additon to produce optimum separations. Highly efficientseparations were produced under constant modifier concentra-tion and pressure programming.

The object of this study was to compare the ability of normal-phase packed columns to separate APEs on an SFC system.Individual packed columns as well as stacked packed columns ofdifferent stationary phases were used in the SFC experiments.Additional goals of this study were to identify the componentsthat gave rise to the chromatographic peaks in hopes of pro-ducing individual ethoxylated alkylphenol standards. Fractionsthat contain a single ethoxylate compound could later be used asstandards for quantitating APEs in a variety of applications. Acomparison of the ability of SFC and HPLC to separate APEsusing normal-phase packed columns was also studied.

Experimental

Packed-column SFCA Berger (Newark, DE) SFC system was used in the SFC anal-

ysis. A Berger autosampler with a 10-µL injection loop was usedfor conventional sample analysis, and a 75-µL injection loop wasused for the injection of semipreparative samples. SFC-gradecarbon dioxide (Air Products and Chemicals, Inc., Allentown, PA)was used with methanol (Burdick & Jackson, Muskegon, MI) as amodifier. The mobile phase flow rate was 2.0 mL/min. The oventemperature was set at 60°C, and the outlet pressure was kept at120 atm. Absorbance was read at 225 nm by a diode-arraydetector. The detection wavelength was determined by finding themaximum absorbance of an individual APE sample by obtainingits UV–vis spectrum. Supelcosil LC-Diol, Supelcosil LC-CN(Supelco, Bellefonte, PA), and Spherisorb NH2 (Waters, Millford,

MA) columns were used for the chromatographic separation ofthe APE mixtures. All columns measured 4.6 × 250 mm with a 5-µm particle size. A diol bonded silica guard column was used.

Normal-phase HPLCFor HPLC analysis, a Hewlett-Packard (Little Falls, DE) 1050

Series HPLC system was used with a variable wavelength detector(reading 225 nm) and an inline vacuum degasser. Injections weremade manually with a Rheodyne (Rohnert Park, CA) injectorequipped with a 20-µL injection loop. Data were collected andchromatograms were processed by MassLynx software (FisionsInstruments, Altricham, U.K.). A Supelcosil LC-Diol column (4.6× 250 mm, 5 µm) was used for the chromatographic separation ofthe APE mixtures.

Flow injection analysis–MSA Fisions Instruments VG Platform MS was used for the mass

analysis of collected sample fractions. All samples were analyzedunder positive ESI. A syringe pump (Harvard Apparatus, SouthNatick, MA) supplied an 80:20 methanol–water mobile phase tothe probe. Samples were injected by a Rheodyne injectorequipped with a 20-µL injection loop. Nitrogen was used as boththe drying and sheath gas. Data were collected and analyzed byMassLynx software.

Alkylphenol samplesPOE-(4)-NP (ChemService, West Chester, PA) and Triton N-101

(Sigma-Aldrich, Milwaukee, WI) were used as NPEO mixtures.POE-(5)-tert-OP (ChemService) was used as an OPEO mixture.All of the samples that were analyzed by SFC were dissolved inmethanol, and samples analyzed by normal-phase HPLC weredissolved in hexane. The Triton N-101 sample that was used forHPLC was dissolved in 9:1 hexane–acetone in order to increasesolubility. HPLC samples were prepared at approximately 1.0mg/mL, and SFC samples were prepared at approximately 2.0-mg/mL concentrations.

Semipreparative SFCA tee was placed inline between the column and diode-array

detector of the SFC system, splitting effluent approximately 75%to the collection and 25% to the detector. Eluent was divertedusing a portion of fused-silica capillary tubing. Fractions werecollected in preweighed 16-mL collection vials. Absorbance wasmonitored, and fractions were collected manually between min-imum absorbance values. POE-(4)-NP and POE-(5)-tert-OP wereseparated in this fashion. Fractions were evaporated by nitrogenblow-down on a hot plate. The remaining residue was weighed.The fractions were then diluted to 10.0 mL with methanol.Fractions were analyzed by SFC-UV followed by flow injectionanalysis (FIA)–ESI–MS for purity.

FIA–ESI–MS methodSFC-collected fractions were evaporated by nitrogen blow-

down and weighed. Collected fractions were then dissolved inmethanol. Optimal MS settings were found by injecting each frac-tion and tuning the instrument. Fractions were then reinjected,and mass-spectral data were recorded and analyzed. The sourcetemperature was set at 100°C. ESI nebulizing gas flow was set at

kgurski

A syringe pump (Harvard Apparatus, South

kgurski

Natick, MA)


63

20 L/h, and the drying gas flow was 300 L/h. Samples wererecorded in full-scan mode from m/z 200 to 700. The cone voltageranged from 52 to 75 V, and the high voltage lens and ESI capil-lary voltage were kept at 0.88 and 3.46 kV, respectively.

HPLC methodHexane and isopropanol were used as the mobile phase. A linear

gradient was used starting with 100% hexane and then changingto 70:30 hexane–isopropanol over 30 min. From t = 30 to 35 min,the mobile phase was returned to 100% hexane and held for 5 min in order to equilibrate. POE-(4)-NP and POE-(5)-tert-OPwere separated in this fashion.

Results and Discussion

APEs are complex mixtures that provide moderate challengesfor chromatographic techniques. Our research studied how thetotal column length, stationary phase, and column stacking orderof different stationary phases affect the SFC separation of ethoxy-late units in APE mixtures. Our goal was to find a setup that pro-duced the best separation. In order to accomplish this we kept allsystem parameters constant throughout the study other thancolumn setup and modifier gradient. All of the columns usedwere uniform in size (4.6 × 250 mm, 5 µm) in order to allow us toverify the effect of column length and packing material. POE-(4)-NP was used in all of the diol column studies because of its shortelution time.

POE-(4)-NP was separated on a combination of one-, two-, andthree-packed diol columns connected in series to study the effectof column length (Figure 1). A single diol column poorly sepa-rated the sample. Baseline separation was not achieved with a

single column. SFC separation on two diol columns increasedseparation, but early eluting peaks were not baseline separated.Using two diol columns, SFC separation was comparable withnormal-phase HPLC using one diol column. For comparison,POE-(4)-NP, POE-(5)-tert-OP, and Triton N-101 were separated bySFC on two diol columns and HPLC on one diol column (Figures2–4). The retention time of the chromatographic peaks for SFCseparation using two diol columns was considerably lower thannormal-phase HPLC separation using one diol column (Tables Iand II shows data for the NPEO sample and Table III shows datafor the OPEO sample). The addition of a third diol column to theSFC system generated a better separation, but later-eluting peaksbegan to broaden.

The effect of the stationary phase on separation was sequen-tially tested using a single diol, amino, and cyano column (Figure5). The retention of oligomers with longer ethoxylated unitsvaried with each stationary phase tested. The diol column had theleast retention, the amino column had intermediate retention,and the cyano column had the greatest retention. It was not pos-sible to elute all of the compounds off the cyano column using thecorresponding gradient. In general, a larger methanol modifierconcentration was needed to elute longer ethoxylate-chain com-

Figure 2. Chromatograms of POE-(4)-NP using (A) normal-phase HPLC-UVwith one Supelcosil LC-Diol column and (B) SFC-UV with two Supelcosil LC-Diol columns. The peak annotations represent the number of ethoxylate units.

Figure 1. Packed-column supercritical fluid chromatograms using stacked diolcolumns: (A) one Supelcosil LC-Diol column, (B) two Supelcosil LC-Diolcolumns, and (C) three Supelcosil LC-Diol columns. The sample used in eachchromatogram was POE-(4)-NP (2.0 mg/mL). A linear modifier gradient wasused by the following program: 10.0% methanol was increased to 26.0% at arate of 0.6%/min with a 2.0-min hold and then returned to 10.0% in 4.0 minfollowed by a 2.0-min hold.


64

pounds. Because of this, we can conclude that APEs with a longerethoxylate chain are more polar than those with shorter chains.Following this reasoning, the cyano column must be the mostpolar stationary phase because it retained the more polar compo-nents longer, and the diol column is the least polar.

Columns with different stationary phases were coupled inseries to test how the arrangement would affect the retention ofan APE sample. Two column arrangements were tested. The firstconsisted of one diol column followed by one cyano column. Thesecond setup contained three columns, a diol column, a cyanocolumn, and an amino column in series (Figure 6). A steeper gra-dient was needed than previously used in order to elute all of thecompounds because of the presence of the cyano column (as pre-viously mentioned). The modifier gradient that was used isdescribed in Figure 6.

One of our goals in this study was to achieve separation thatwould allow us to easily collect individual oligomers for use asstandards. The combined diol–cyano setup rendered shorterretention times than the combined diol–cyano–amino setup;therefore, this arrangement was used for preparative fraction col-lection. In the chromatograms of stacked columns using differentstationary phases, peak splitting was observed for later-eluting

peaks. POE-(4)-NP and POE-(5)-tert-OP were separated, and fivefractions of each sample were collected. A large volume (75 µL) ofconcentrated sample was injected six to eight times in the collec-tion process. Isolated fractions were reanalyzed both by SFC forpurity (Figures 7 and 8) and FIA–ESI–MS for identification. Theconcentrations used for the semipreparative work caused thechromatographic peaks to significantly broaden and in somecases combine. Because of this phenomenon we were not able tocollect individual fractions of the two initial oligomers of POE-(4)-NP and fractions of the three initial oligomers of POE-(5)-tert-OP as evidenced by the SFC-UV of the early fractions.

FIA–MS was used to identify the components in each fraction.ESI–MS was chosen because it is amenable to high-molecular-weight analytes and works well with liquid mobile phases.Samples were dissolved in methanol (a compatible solvent forESI–MS), which made ESI–MS a desirable tool for fraction iden-tification. It was possible to produce sodium-adducted molecularions rather easily. In order to create an optimum response, thefractions were first injected and the cone voltage varied in orderto produce the greatest response for each individual analyte. AfterMS tuning conditions were perfected, the fractions were rein-jected into the instrument. A spectrum was created between

Figure 4. Chromatograms of Triton N-101 using (A) normal-phase HPLC-UVwith one Supelcosil LC-Diol column and (B) SFC-UV with two Supelcosil LC-Diol columns. The peak annotations represent the number of ethoxylate units.

Figure 3. Chromatograms of POE-(5)-tert-OP using (A) normal-phase HPLC-UVwith one Supelcosil LC-Diol column and (B) SFC-UV with two Supelcosil LC-Diol columns. The peak annotations represent the number of ethoxylate units.


65

m/z 200 and 700 by averaging scans of the injected sample.Figures 9 and 10 show the average mass spectrum of each fraction.The spectra confirm that each chromatographic peak varied byone ethoxylated unit (a separation of m/z 44 represents an ethoxy-late unit). It was possible to identify NP3EO through NP7EO inbasically pure collected fractions of POE-(4)-NP and OP5EOthrough OP8EO in fractions collected from POE-(5)-tert-OP.

Major ion peaks consisted of Na+ adduct ions, and minor peakswere produced by K+ adduct ions under positive electrospray con-ditions. Trace levels of sodium and potassium must be present inthe mobile phase that was used for FIA–ESI–MS because elec-trolyte was not added to the solutions. According to Okada’sresearch (17), APEs have an affinity for alkali metals and have a

flexible structure that allows them to form complexes with alkalimetals. This explains the ion pairing seen in the mass spectra.Crescenzi et al. performed an experiment to see if the detectorresponse would decrease because of the complexation ofoligomers competing for the limited metal pool available. Whenequivalent amounts of ethoxylated compounds were analyzed byESI–MS, it was found that the detector response increased expo-nentially from 1 to 6 EO units and then leveled off at 8 EO units(the scope of the study) (18). A decrease in signal was most notice-able for lower ethoxylated oligomers. This can be explained bynoting that ethoxylated compounds can form increasingly stablecomplexes with alkali metal ions as the EO unit number increases(17).

Table I. Chromatographic Peak Retention Times of POE-(4)-NP (NPEO) Separated by SFC Using TwoSupelcosil LC-Diol Columns and HPLC Using OneSupelcosil LC-Diol Column

EO unit SFC RT* HPLC RT

2 7.18 9.143 7.86 9.934 8.64 10.665 9.68 11.826 10.61 13.087 11.56 14.468 12.49 15.839 13.43 17.28

10 14.37 18.7411 15.16

* RT, retention time.

Table II. Chromatographic Peak Retention Times of TritonN-101 (NPEOs) Separated by SFC Using Two SupelcosilLC-Diol Columns and HPLC Using One Supelcosil LC-Diol Column


2 7.29 9.883 8.02 10.684 8.83 11.845 9.75 13.086 10.66 14.337 11.57 15.558 12.48 16.809 13.36 18.06

10 14.20 19.3911 15.03 20.6812 15.84 22.2913 16.61 24.1114 17.3715 18.1016 18.8117 19.5818 20.10


Table III. Chromatographic Peak Retention Times of POE-(5)-tert-OP (OPEOs) Separated by SFC Using TwoSupelcosil LC-Diol Columns and HPLC Using OneSupelcosil LC-Diol Column


2 6.79 9.283 7.48 10.064 8.20 10.925 9.05 12.086 10.00 13.397 10.96 14.758 11.91 16.109 12.88 17.51

10 13.86 18.9611 14.8012 15.76


Figure 5. Packed-column supercritical fluid chromatograms using singlecolumns of different polar packing material: (A) Supelcosil LC-Diol column, (B) Spherisorb NH2 column, and (C) Supelcosil LC-PCN column. The sampleused in each chromatogram was POE-(4)-NP (2.0 mg/mL). A linear modifiergradient was used by the following program: 10.0% methanol was increasedto 26.0% at a rate of 0.6%/min with a 2.0-min hold and then returned to 10.0%in 4.0 min followed by a 2.0-min hold.


66

It was important to perform chromatographic separations withabsorbance detection on the fractions as well as MS analysis, thusallowing us to positively identify sample components because MScould not detect all of the compounds present. The first fractionof both POE-(4)-NP and POE-(5)-tert-OP contained more thanone compound (as seen in their SFC-UV chromatograms). Thesodium ion affinity of the smaller ethoxylate chain compounds islower than the larger chain oligomers, and because of this they

were not detectable in the mass spectra.APEs can be categorized by their average ethoxylate unit value.

According to Wang and Fingas (3), all of the oligomers havealmost identical molar absorptivity, which allows integrated chro-matographic peak areas to be used directly to determine the molefraction of each oligomer. POE-(4)-NP contained NP predomi-nantly with short ethoxylate chains. NP2EO through NP11EOwere observed in its SFC-UV separation. An average ethoxylate

Figure 6. Packed-column supercritical fluid chromatograms using stackedcolumns of different polar stationary phases: (A) one Supelcosil LC-Diolcolumn and one Supelcosil LC-PCN column and (B) one Supelcosil LC-Diolcolumn, one Supelcosil LC-PCN column, and one Spherisorb NH2 column.The sample used in each chromatogram was POE-(4)-NP (2.0 mg/mL). Multiplelinear modifier gradients were used by the following program: 10.0% methanolwas increased to 13.2% by 0.5%/min and then continued to 14.4% at0.7%/min, 16.6% at 0.8%/min, 20.0% at 1.0%/min, 40.0% at 8.0%/min (heldfor 5.0 min), and then returned to 10.0% at 15.0%/min.

Figure 7. Supercritical fluid chromatograms of collected POE-(4)-NP fractions.Separation was conducted on one Supercosil LC-Diol column and oneSupelcosil LC-PCN column in series (the system settings were the same asFigure 3).

Figure 8. Supercritical fluid chromatograms of collected POE-(5)-tert-OP frac-tions. Separation was conducted on one Supelcosil LC-Diol column and oneSupelcosil LC-PCN column in series (the system settings were the same asFigure 3).

Figure 9. Positive-ion FIA–ESI–MS of POE-(4)-NP fractions operated in full-scanmode. Ions were in the form of (M+Na)+ and each were separated by m/z 44(the mass of one ethoxyl unit): (A) fraction 1, cone voltage of 59 V; (B) fraction2, cone voltage of 53 V; (C) fraction 3, cone voltage of 62 V; (D) fraction 4, conevoltage of 65 V; and (E) fraction 5, cone voltage of 67 V. Each spectrum wasaveraged over the sample injection peak.


67

unit value of 4.20 was calculated from peak areas. POE-(5)-tert-OP had a similar distribution as POE-(4)-NP. Its average ethoxy-late unit value was calculated as 4.48, and it contained OP2EOthrough OP12EO in its SFC-UV separation. Triton N-101 con-tained a greater range of NPEOs. Its calculated average ethoxylateunit was 9.97. NP2EO through NP18EO were observed in its SFC-UV chromatogram. Higher EO peaks were detected in SFC sepa-rations, which were not detected by HPLC analysis. Wang andFingas produced similar average EO unit values from their capil-lary SFC data. Their analysis of Igepal CO430 (trade name forPOE-(4)-NP), Triton X-45 (trade name for POE-(5)-tert-OP), andTriton N-101 produced average EO values of 4.14, 4.50, and 9.52,respectively (11,12). We used the chromatographic data from theSFC-UV separations on two diol columns to calculate our averageEO values.

Conclusion

Normal-phase packed-column SFC produced a similar separa-tion of APE mixtures compared with normal-phase HPLC.Column length, stationary phase, and column combinations withdifferent stationary phases all affected the separation of the APEmixtures tested. Longer column lengths increased the separationof oligomers. More-polar stationary phases retained oligomerswith larger ethoxylate units for a longer time. A combination ofcolumns with different stationary phases produced separationscombining both the effects of longer columns and the separationability of each stationary phase. Retention times for SFC separa-

tions were notably shorter than normal-phase HPLC. One ofSFC’s advantages is its ability to use longer combined columnlengths without elevated back pressure, which occurs in HPLC.Combining multiple columns with different stationary phasesseemed to provide the best separation.

An advantage of using packed columns over the use of capillarycolumns is the ability to inject larger amounts of sample and col-lect eluted fractions. It is possible to isolate and identify individualAPEs. Additionally, it is possible to identify the remaining chro-matographic peaks because of each peak differing by one ethoxy-late unit. Our study demonstrated the importance of using bothabsorbance detection as well as MS. MS alone did not show all thecomponents of our initial fractions because of the decreaseddetector response.

Less solvent waste was produced using SFC compared withHPLC. Each SFC separation that used cyano packing as part of itscolumn arrangement used 6.7 mL of methanol. The remainingSFC setups (the studies of column length and stationary phase)used 11.8 mL of methanol. All separations performed by normal-phase HPLC used 34.75 mL of hexane and 5.25 mL isopropanolfor a combined volume of 40 mL. The HPLC system used almost600% more solvent than the SFC system using a cyano stationaryphase and over 330% more than the other SFC setups studied(this is not including the volume of solvent needed to initiallyequilibrate the systems). The reduction of solvent waste is animportant step of reducing pollution.

Because of the fact that APEs are used as industrial cleaners andother processing aids, they enter wastewater and end up insewage treatment plants. Some APE waste is transferred into theenvironment and metabolized into lower ethoxylated alkylphe-nols, which are considered endocrine disrupters (2). APEs havebeen found in fish, river sediment, and other environmental sam-ples through analytical techniques (1,4,9,10,18–22). The resultsof our study could lead to the further use of the method developedfor applications in the analysis of environmental samples.Additionally, our method could be altered for use in a futurelarge-scale separation and collection of individual ethoxylatedalkylphenols. Access to standards of individual ethoxylatedalkylphenols is important for their quantitative analysis.

Acknowledgments

We would like to acknowledge Dr. Clifford P. Rice (USDAARS/NRI/EQL, Beltsville, MD) for APE information and AirProducts and Chemicals, Inc. for supplying SFC-grade carbondioxide.

References

1. B. Thiele, K. Gunther, and M.J. Schwunger. Alkylphenol ethoxylates:trace analysis and environmental behavior. Chem. Rev. 97: 3247–72(1997).

2. R. White, S. Jobling, S.A. Hoare, J.P. Sumpter, and M.G. Parker.Environmentally persistent alkylphenolic compounds are estrogenic.Endocrinology 135: 175–82 (1994).

Figure 10. Positive-ion FIA–ESI–MS of POE-(5)-tert-OP fractions operated infull-scan mode. Ions were in the form of (M+Na)+ and each were separated bym/z 44 (the mass of one ethoxyl unit): (A) fraction 1, cone voltage of 63 V; (B)fraction 2, cone voltage of 68 V; (C) fraction 3, cone voltage of 65 V; (D) frac-tion 4, cone voltage of 75 V; and (E) fraction 5, cone voltage of 75 V. Each spec-trum was averaged over the flow injection peak.


68

3. Z. Wang and M. Fingas. Rapid separation of nonionic surfactants ofpolyethoxylated octylphenol and determination of ethylene oxideoligomer distribution by C1 column reversed-phase liquid chro-matography. J. Chromatogr. 673: 145–56 (1993).

4. C. Whalberg, L. Renberg, and U. Wideqvist. Determination ofnonylphenol and nonylphenol ethoxylates as their pentafluoroben-zoates in water, sewage sludge, and biota. Chemosphere 20: 179–95(1990).

5. R.E.A. Escott, S.J. Brinkworth, and T.A. Steedman. The determinationof ethoxylate oligomer distribution of nonionic and anionic surfac-tants by high-performance liquid chromatography. J. Chromatogr.282: 655–61 (1983).

6. I. Zeman. Applications of bonded diol phases for separation ofethoxylated surfactants by high-performance liquid chromatography.J. Chromatogr. 363: 233–30 (1986).

7. A.M. Rothman. High-performance liquid chromatographic methodfor determining ethoxymer distribution of alkylphenoxy poly-oxyethylene surfactants. J. Chromatogr. 253: 283–88 (1982).

8. J.L. Gundersen. Separation of isomers of nonylphenol and selectnonylphenol polyethoxylates by high-performance liquid chro-matography on a graphitic column. J. Chromatogr. A 914: 161–66(2001).

9. P.L. Ferguson, C.R. Iden, and B.J. Brownawell. Analysis ofalkylphenol ethoxylate metabolites in the aquatic environment usingliquid chromatography–electrospray mass spectrometry. Anal.Chem. 72: 4322–30 (2000).

10. D.Y. Shang, M.G. Ikonomou, and R.W. MacDonald. Quantitativedetermination of nonylphenol polyethoxylate surfactants in marinesediment using normal-phase liquid chromatography–electrospraymass spectrometry. J. Chromatogr. A 849: 467–82 (1999).

11. Z. Wang and M. Fingas. Quantitative analysis of polyethoxylatedoctylphenol by capillary supercritical fluid chromatography. J. Chromatogr. 641: 125–36 (1993).

12. Z. Wang and M. Fingas. Analysis of polyethoxylated nonylphenolsby supercritical fluid chromatography and high-performance liquidchromatography. J. Chromatogr. Sci. 31: 509–18 (1993).

13. A. Giorgettie, N. Pericles, H.M. Widmer, K. Anton, and P. Datwyler.

Mixed mobile phases and pressure programming in packed column supercritical fluid chromatography: a unified approach. J. Chromatogr. Sci. 27: 318–24 (1989).

14. M. Takeuchi and T. Saito. Combination of semi micro and micropacked column supercritical fluid chromatography with some otherinstruments for qualitative analysis. J. High Resolut. Chromatogr. 14:347–51 (1991).

15. T.A. Dean and C.F. Poole. Solventless injection for packed columnsupercritical fluid chromatography. J. High Resolut. Chromatogr. 12:773–78 (1989).

16. E.S. Francis, M.L. Lee, and B.E. Richter. Modifier addition in micro-column supercritical fluid chromatography with a high pressurepulsed valve. J. Microcolumn Sep. 6: 449–57 (1994).

17. T. Okada. Efficient evaluation of poly(oxyethylene) complex forma-tion with alkali-metal cations. Macromolecules 23: 4216–19 (1990).

18. C. Crescenzi, A. Di Corcia, R. Sampri, and A. Marcomini.Determination of nonionic polyethoxylate surfactants in environ-mental waters by liquid chromatography/electrospray mass spec-trometry. Anal. Chem. 67: 1797–1804 (1995).

19. T.L. Keith, S.A. Snyder, C.G. Naylor, C.A. Staples, C. Summer, K. Kannan, and J.P. Giesy. Identification and quantification ofnonylphenol ethoxylates and nonylphenol in fish tissues fromMichigan. Environ. Sci. Technol. 35: 10–13 (2001).

20. H.B. Lee, T.E. Peart, D.T. Bennie, and R.J. Maguire. Determination ofnonylphenol and their carboxylic acid metabolites in sewage treat-ment plant sludge by supercritical fluid carbon dioxide extraction. J. Chromatogr. A 785: 385–94 (1997).

21. T.L. Potter, K. Simmons, J. Wu, M. Sanchez-Olvera, P. Kostecki, andE. Calabrese. Static Die-away of a nonylphenol ethoxylate surfactantin estuarine water samples. Environ. Sci. Technol. 33: 113–18 (1999).

22. M. Petrovic and D. Barcelo. Determination of anionic and nonionicsurfactants, their degradation products, and endocrine-disruptingcompounds in sewage by liquid chromatography/mass spectrometry.Anal. Chem. 72: 4560–67 (2000).

Manuscript accepted December 7, 2001.

69

An isocratic high-performance liquid chromatography (HPLC)method for the simultaneous determination of αα-tocopherol,retinol, and five carotenoids (lutein–zeaxanthin, ββ-cryptoxanthin,lycopene, and αα- and ββ-carotene) in human serum is described.Serum samples are deproteinized with ethanol and extracted oncewith n-hexane. Resulting extracts are injected onto a C18 reversed-phase column eluted with methanol–acetonitrile–tetrahydrofuran(75:20:5, v/v/v), and full elution of all the analytes is realizedisocratically within 20 min. The detection is operated using threechannels of a diode-array spectrophotometer at 290, 325, and 450 nm for tocopherol, retinol, and the carotenoids, respectively.An internal standard is used for each channel, which improvesprecision. The choice of internal standards is discussed, as well asthe extraction protocol and the need for adding an antioxidantduring the extraction and chromatographic steps. The analyticalrecoveries for liposoluble vitamins and carotenoids are more than85%. Intra-assay relative standard deviation (RSD) values (n = 20)for measured concentrations in serum range from 3.3% (retinol) to 9.5% (lycopene), and interassay RSDs (n = 5) range from 3.8%(αα-tocopherol) to 13.7% (ββ-cryptoxanthin). The present method isused to quantitate the cited vitamins in healthy subjects (n = 168)from ages 9 to 55 years old.

Introduction

Retinol (vitamin A) and α-tocopherol (vitamin E) are nonen-zymatic antioxidants (1). Vitamin A acts as a direct “scavenger”of reactive oxygen species (ROS) and is also thought to inhibitfree radical synthesis via increasing the activity of detoxifyingsystems (2).

Vitamin E protects unsaturated fatty acids located in both celland organelle membranes against endo- and exogenous free rad-icals and ROS, which are involved in the initiation and extent ofmembrane damages caused by nonenzymatic lipid peroxidation(3,4). Carotenoids act as ROS and free radical scavengers (5),stimulants of immune response (6), and anticarcinogenic agents(7). Because of their wide variety of functions and biologicalroles, clinical interest in the evaluation of retinol, α-tocopherol,and carotenoids has increased in recent years owing to their roleas antioxidants, which may be important in reducing the risk ofnumerous diseases including cancer (8–11), coronary heart dis-ease (12,13), and diabetes mellitus (14–18).

Thus, rapid, simple, sensitive, and selective methods for thesimultaneous determination of these antioxidants in biologicalfluids are needed. As a matter of fact, the measurement of anindividual class of antioxidants such as thiols (19), hydrophilic,or liposoluble vitamins provides more information for the mech-anistic evaluation of a clinical disease linked to oxidative stressthan a total antioxidant status assay (20).

Numerous spectroscopic and separative methods have alreadybeen reported for the assay of retinol, α-tocopherol, andcarotenoids in plasma or serum, and among them high-perfor-mance liquid chromatography (HPLC) is one of the most pow-erful analytical tools for this purpose (21–25).

Both normal-phase (26–28) and reversed-phase (29–35) HPLCconditions have been widely used. However, many of thesemethods include gradient elution (36–39), flow rate (34,36),wavelength time-programmation (36,40), a switching devicebetween coupled columns (41,42), and the use of two differentdetectors in series (43,44). All of these approaches are time-con-suming because of their long-equilibration period between eachrun and troublesome because of the hyphenated systems needed.

Indeed, the main difficulty for the simultaneous determinationof liposoluble vitamins and carotenoids results from their dif-ferent spectral characteristics (absorption maxima vary in the

Abstract

An Isocratic Liquid Chromatographic Method withDiode-Array Detection for the SimultaneousDetermination of αα-Tocopherol, Retinol, and Five Carotenoids in Human Serum

Sonia Gueguen1, Bernard Herbeth1, Gérard Siest1, and Pierre Leroy2

1Inserm U525, Centre de Médecine Préventive, 2 rue du Doyen Jacques Parisot, 54500 Vandoeuvre-lès-Nancy, France and 2Thiols andCellular Functions, Faculté de Pharmacie, Université Henri Poincaré Nancy 1, 30 rue Lionnois, 54000 Nancy, France



* Author to whom correspondence should be addressed: email [email protected].


70

range of 292 to 450 nm). This problem has been solved by usingmultichannel UV–vis spectrophotometric detectors (31,37,40,45–47). More recently, a technique combining both isocratic elu-tion in reversed-phase mode and diode-array detection wasreported, providing selectivity between the three classes ofliposoluble vitamins and thus a convenient way for their simul-taneous measurements (32).

For all these methods, the preanalytical treatments, especiallythe extraction procedure relying upon either liquid–liquid(26–28,30–35,39,43,47,48) or solid–liquid (38,49,50) partition,are critical steps to obtain reliable data.

This study deals with some improvements of a previouslyreported method (32); the full validation of the optimized assay;and its use to quantitate retinol, α-tocopherol, lutein–zeaxan-thin, β-cryptoxanthin, lycopene, and α- and β-carotene inhealthy subjects.

Experimental

Chemicals, reagents, and standardsAll solvents and reagents used were of analytical- or HPLC-

grade. Ultrapure water was prepared using a Milli-Q system

(Millipore Milford, MA). Tert-butylated hydroxytoluene (BHT)was purchased from Sigma-Aldrich (St. Quentin Fallavier,France).

All-trans retinol (henceforth simply referred to as retinol),retinol acetate, α-tocopherol, α-tocopherol acetate, and β-carotene standards were obtained from Fluka (Buchs,Switzerland). Zeaxanthin and β-cryptoxanthin were a generousgift from Hoffman-Laroche (Basle, Switzerland). Lycopene andechinenone were purchased from CaroteNature (Lupsingen,Switzerland). Stock solutions of retinol, α-tocopherol, and theircorresponding internal standards (acetate form) were preparedin ethanol (EtOH) added with 0.01% (w/v) BHT. Carotenoidswere prepared in tetrahydrofuran (THF) added with 0.01% BHT.Stock solutions were protected from light in ambered glass bot-tles, titrated by spectrophotometry using their specificabsorbance (Table I), and stored under nitrogen at –80°C for upto 2 mo. The concentrations of stock solutions were 0.25–0.5mg/mL for retinol and retinol acetate, 3–4 mg/mL for α-toco-pherol and α-tocopherol acetate, and 0.1–0.2 mg/mL forcarotenoids.

Daily working solutions for calibration curves were preparedby diluting stock solutions in EtOH containing 0.01% BHT. Theranges of tested concentrations are indicated in Table II. Aninternal standard mixture containing retinol acetate, α-toco-

pherol acetate, and echinenone was also prepareddaily following a similar procedure (combining100 µL of each stock solution of internal standardand diluting the volume to 20 mL withEtOH–0.01% BHT). All the operations were per-formed by handling solutions in darkness and ice.

The standards of β-carotene and zeaxanthinwere used to quantitate α-carotene and bothlutein and zeaxanthin, respectively.

Blood collection and storage conditions Blood was collected at the antecubital vein of

168 healthy control subjects from ages 9 to 55years old (informed consent was obtained, andthe research protocol was in agreement with theHelsinki Declaration) in a reclined position in drytubes (Vacutainer Tube, Becton Dickinson,Grenoble, France). Blood samples were cen-

Table I. Characteristics of Standards Used

MaximumMolecular weight wavelength

Compounds (g/mol) (nm) A1%1 cm* εε (mol–1/L/cm–1)

Retinol 286.5 325 1835 (32,61) 52573Retinol acetate 328.5 326 1550 (32,61) 50912α-Tocopherol 430.7 292 75.8 (45) 3265α-Tocopherol acetate 472.8 290 40 (32) 1891Echinenone 550.9 458 2244 123622

(Hoffmann-Laroche data source)

Lutein–zeaxanthin 568.9 452 2765/2416 (45) 157301/137446β-Cryptoxanthin 552.9 452 2486 (45) 137451Lycopene 536.9 472 3450 (32,61) 185231β-Carotene 536.9 450 2620 (35) 140667

* In EtOH as the solvent. Data references appear in the parentheses.

Table II. Equations of Calibration Curves and Values of LODs and LOQs*

Equations of calibration curves

Concentration Slope† Intercept Correlation LOD LOQ range (µmol/L) (SD‡, n = 5) (SD, n = 5) coefficient† (µmol/L) (µmol/L)

Retinol 0.45–7.50 0.16 (0.015) 0.021 (0.016) 0.998 0.45 0.66α-Tocopherol 4.80–80.0 0.01 (0.000) 0.027 (0.008) 0.996 2.64 5.36Lutein–zeaxanthin 0.10–1.90 0.35 (0.034) 0.024 (0.006) 0.997 0.06 0.11β-Cryptoxanthin 0.09–1.50 0.34 (0.031) 0.022 (0.019) 0.996 0.03 0.09Lycopene 0.12–1.90 0.24 (0.018) 0.018 (0.020) 0.997 0.03 0.08β-Carotene 0.13–2.00 0.35 (0.019) 0.014 (0.006) 0.997 0.03 0.06

* Each calibration curve included six points, and each point was assayed in five replicates.† Calculated by internal standardization: (standard peak area/internal standard peak area)/standard concentration.‡ SD, standard deviation.


71

trifuged (1500 g for 15 min at 4°C) within 2 h after collection,and resulting serum samples were frozen in liquid nitrogen untilHPLC analysis.

Serum sample treatmentAll the handling operations were carried out in darkness. The

serum samples were rapidly thawed at room temperature,homogenized, and 200 µL was transferred into a borosilicateglass tube kept on ice and 300 µL of the internal standard mix-ture added. After mixing with a vortex for 20 s, proteins were pre-cipitated by adding 200 µL of EtOH–0.01% BHT, and the volumewas diluted to 1 mL with ultrapure water. After mixing with anorbital shaker at 2500 rpm for 1 min, 2 mL of n-hexane–0.01%BHT was added. The samples were shaken for 1 min and cen-trifuged at 2700 g for 20 min at 4°C.

The organic layer was carefully transferred into a glass tubeand evaporated to dryness under a stream of nitrogen at roomtemperature. The dried residue was redissolved in 25 µL ofTHF–0.01% BHT and vortexed for 30 s. A 75-µL amount ofmobile phase was added, and the resulting mixture was vortexedfor another 30 s. Samples were then transferred to 200-µL insertvials and placed into the HPLC autosampler.

HPLC system and operating conditionsThe HPLC system consisted of an isocratic solvent delivery

pump (Model Kontron Instruments 422), an autosamplerequipped with a 20-µL injection loop, a cooling sample tray anda column oven (Model AS-300, ThermoQuest, Les Ulis, France),a UV–vis diode-array detector (Model Gold LC-168, BeckmanCoulter, Fullerton, CA), and data-processing software (Gold New,Beckman).

A guard column (8- × 3-mm i.d.) packed with Nucleosil C18 (5 µm) (Macherey Nagel, Duren, Germany) and an analyticalcolumn (250- × 3-mm i.d.) packed with Nucleosil 100 C18 (5 µm)(Macherey Nagel) were eluted with a mobile phase consisting ofa mixture of methanol–acetonitrile–tetrahydrofuran (75:20:5,v/v/v) containing 0.01% (w/v) BHT. The mobile phase was filteredthrough a 0.45-µm Nylon membrane and was used at a columntemperature of 35°C and a flow rate of 0.6 mL/min. Three chan-nels corresponding with different wavelength values were usedto acquire data for the selective monitoring of α-tocopherol (290nm), retinol (325 nm), and carotenoids (450 nm) and theirrespective internal standard. During analysis, the tray compart-ment containing sample vials was cooled at 5°C. After eachworking period (approximately 50 samples), it was necessary torinse the column with methanol at a flow rate of 0.6 mL/min for20 min to eliminate highly hydrophobic compounds and preventthe loss of column efficiency.

CalculationThe vitamin concentrations were determined from a standard

curve of the peak-area ratio of the analyte–internal standardplotted against the concentration of analyte (expressed in micro-moles per liter). A linear least-square regression analysis wasperformed for each analyte, and the standard curve was repeatedif the correlation coefficient was below 0.990.

The detection limit (LOD) and the quantitation limit (LOQ)were expressed, respectively, as:

LOD = (a0 + 3sa0) / a1 Eq. 1

and

LOQ = (a0 +10sa0) / a1 Eq. 2

where a1 is the slope, a0 the intercept, and sa0 the standard devi-ation of the intercept (51).

Quality controlA human serum pool made with 1 mL of fresh serum from 100

healthy subjects and stored at –80°C was used for the routinequality control. Aliquots were extracted and analyzed accordingto the same procedure that was described previously. Evaluationof the method performance was assessed by comparing theresults of the quality control with the means and relative stan-dard deviations (RSDs) calculated using results from several pre-liminary runs (n = 20 per day for five days).


Optimization of sample treatment and HPLC techniqueThe basic method used in this study has been described by

Talwar et al. (32). Some modifications relating to the internalstandards, the sample preparation procedure, and the use of anantioxidant during both the extraction and chromatography pro-cesses have been made. We chose this method because it allowsin a fast and easy way the simultaneous separation of the twoclasses of lipophilic vitamins (namely retinol, α-tocopherol, andcarotenoids). Our main objective was to measure simultaneouslylipophilic vitamins and carotenoids, which are the most abun-dant in human serum. Thus, the separation of the isomers ofretinol, α-tocopherol, and carotenoids did not appear relevantfor our present epidemiological studies.

In most methods, the use of an antioxidant during sampletreatment was demonstrated to be necessary to prevent a signif-icant loss in carotenoid contents, especially lycopene and β-carotene (32,37,39,40,47). Thus, we initially added 0.01%ascorbic acid to the organic solvents used for the standardspreparation (EtOH and THF) and to the mobile phase, as indi-cated by Talwar et al. (32). After analyzing the same sample several times, we observed a decrease of the carotenoid concen-trations, indicating degradation as a function of time. We testedanother antioxidant (BHT) that is widely used in other methods(37,39,47) and added it to the mobile phase and all the solvents(EtOH, THF, and hexane) used for the standard and samplepreparation. Indeed, hexane containing BHT efficiently pro-tected the carotenoids from degradation during the evaporationof the extractive organic layers, and the addition of BHT to themobile phase also prevented any loss of these analytes and prob-ably helped increase the longevity of the column by neutralizingperoxides present in THF. Moreover, we observed that decreasingthe evaporation temperature from 40°C to room temperaturesignificantly increased carotenoid recoveries, as already noted bydifferent authors (39,43).

Other parameters have to be optimized in order to provide thebest conditions for the extraction of liposoluble vitamins and


72

carotenoids. The addition of ultrapure water to the deproteinizedserum with EtOH has been noted to improve the recoveries ofcarotenoids and liposoluble vitamins (37,52). We tested severalEtOH–water proportions in the 1:4 to 1:1 range (v/v) in order toobtain the highest recoveries, and we selected the 1:1 (v/v) pro-portion. Single and double extraction steps with n-hexane (anincrease of the shaking period) were tested, but no significantimprovement of recoveries was observed.

The method previously described (32) used two internal stan-dards: retinol acetate as an internal standard for retinol, and α-tocopherol acetate as an internal standard for both α-toco-pherol and carotenoid. We used a third internal standard (echi-

nenone) for the quantitation of the carotenoids. Echinenone is asynthetic carotenoid and has a structure and chemical propertiesvery similar to the naturally occurring carotenoids in serum.Thus, the use of echinenone is preferable to the use of retinolacetate and α-tocopherol acetate or tocol currently used in othermethods (34,43,47), because it is detected at the same wave-length as the other carotenoids and is light- and temperature-sensitive as other carotenoids. Thus, the use of three internalstandards allows for a better quality control and helps to correctanalytical variations occurring for each liposoluble vitamin andcarotenoid during the extraction and chromatography pro-cesses.

Because no loss of analytes was observed in serum extractskept in darkness for at least 24 h at 5°C, as already reported (39),the automation of the technique was possible with a highthroughput of samples (approximately 30 per day).

Several methods have been developed to measure the main

Figure 1. Typical chromatograms corresponding with a mixture of retinol, α-tocopherol, and carotenoid standards: (A) channel 1, diode-array detectionat 290 nm for α-tocopherol and α-tocopherol acetate; (B) channel 2, diode-array detection at 325 nm for retinol and retinol acetate; and (C) channel 3,diode-array detection at 450 nm for carotenoids and echinenone. The peaknumbers are as follows: (1) 26 µmol/L α-tocopherol, (2) α-tocopherol acetate(the internal standard), (3) 2.43 µmol/L retinol, (4) retinol acetate (internal stan-dard), (5) 0.62 µmol/L lutein–zeaxanthin, (6) 0.49 µmol/L β-cryptoxanthin, (7)echinenone (internal standard), (8) 0.62 µmol/L lycopene, (9) α-carotene, and(10) 0.65 µmol/L β-carotene.

A

B

C

Figure 2. Typical chromatograms corresponding with an extract of a humanserum sample: (A) channel 1, diode-array detection at 290 nm for α-tocopheroland α-tocopherol acetate; (B) channel 2, diode-array detection at 325 nm forretinol and retinol acetate; and (C) channel 3, diode-array detection at 450 nmfor carotenoids and echinenone. Peak numbers are the same as Figure 1.

A

B

C


73

carotenoids present in serum in one run simultaneously with α-tocopherol and retinol (30–34,37,47). Most carotenoids aredetected at 450 or 473 nm, but α-tocopherol and retinol can onlybe detected at 290 and 325 nm, respectively. Most of the previ-ously mentioned methods therefore require the use of severaldetectors in series (43,44) and a multiwavelength detector eitherwith simultaneous monitoring at different wavelengths(31,36,37,40,53) or a change in the detection wavelength duringthe run (30,32,44,47). The need for simultaneous detection atdifferent wavelengths is illustrated by the retinol andlutein–zeaxanthin that elute within a 0.3-min interval and haveto be detected at 325 and 450 nm, respectively. Typical chro-matograms of a standard mixture and an extracted humanserum are shown in Figures 1 and 2. The chromatogramsrevealed elution and baseline resolution between all the analytesof interest except for lutein and zeaxanthin, which are not sepa-rated by this method. The internal standard echinenone waseluted between β-cryptoxanthin and lycopene and thus did notinterfere with the other carotenoids analyzed. Several additionalcarotenoids not identified as of yet appeared between the peak oflutein–zeaxanthin at 4 min and β-cryptoxanthin at 8 min. Beforevalidation of the HPLC method, we have realizeda selectivity study, and BHT has been analyzedwith other analytes to see any potential chro-matographic interference. BHT elutes with ashort retention time (within 3 min) and is onlydetectable at 290 nm, thus no interference withvitamins was observed.

Assay validation and quality control of theHPLC method

The quantitation was achieved using theinternal standardization mode. Data concerninglinearity (the linearity range for each liposolublevitamin was selected according to its physiolog-ical values), LOD, and LOQ are indicated in TableII (and precision in Table III).

The LOD and LOQ values agree with previous data in the liter-ature (32). In order to calculate recoveries, a pooled serum wasspiked with 20 µL of combined standards to provide the addedconcentrations of 0.7 µmol/L retinol, 8.7 µmol/L α-tocopherol,and 0.15 to 0.2 µmol/L carotenoids. The spiked serum samples (n = 5) were then extracted using a single extraction step with n-hexane. Recoveries found were 99.6% ± 11.1% for retinol,91.2% ± 2.0% for retinol acetate, 109.4% ± 13.4% for α-toco-pherol, 101.2% ± 3.0% for α-tocopherol acetate, 112.6% ±22.2% for lutein–zeaxanthin, 104.3% ± 9.1% for β-crypto-xanthin, 109.4% ± 31.0% for lycopene, 85.1% ± 8.5% for β-carotene, and 95.6% ± 9.5% for echinenone. The differentbehaviors of carotenoids with regard to extraction using n-hexane has already been reported by Barua et al. (48). The cal-culated recoveries in this study are satisfactory and comparablewith previously reported values (30,32,33).

In order to check the precision of the method, a human serumpool was analyzed 20 times during the same day to assess therepeatability. This operation was repeated 5 times over a periodof one month to evaluate the interassay precision. The intra- andinterassay variations calculated for each vitamin are shown in

Table III. Precision of the HPLC Assay of Liposoluble Vitamins andCarotenoids in Serum

Within run Between run

Analyte Concentration* (µmol/L) %RSD Concentration† (µmol/L) %RSD

Retinol 1.90 (0.06) 3.3 2.1 (0.09) 4.4α-Tocopherol 34.9 (1.31) 3.8 29.3 (1.1) 3.8Lutein–zeaxanthin 0.65 (0.02) 3.8 0.51 (0.02) 4.5β-Cryptoxanthin 0.13 (0.01) 7.8 0.10 (0.01) 13.7Lycopene 0.53 (0.05) 9.5 0.28 (0.04) 12.5α-Carotene 0.18 (0.02) 8.8 0.14 (0.02) 12.1β-Carotene 0.57 (0.04) 6.7 0.52 (0.05) 9.1

* Mean (standard deviation), n = 20.† Mean (standard deviation), n = 5.

Table IV. Concentrations of Retinol, αα-Tocopherol, and Carotenoids in Millimoles per Liter Measured in the Serum of 168Healthy Subjects from Ages 9 to 55 Years Old and a Comparison with Other Studies

Present study*Men Women Talwar Steghens Olmedilla Sowell

Compound 9–20 years old 21–55 years old 9–20 years old 21–55 years old et al.*,† (32) et al.*,‡ (37) et al.*,§ (54) et al.**,†† (31)

Retinol 1.37 (0.36) 2.18 (0.43) 1.36 (0.31) 1.86 (0.53) 2.00 (0.60) 1.84 (0.80) 1.71 (0.39) 1.91 (1.05–2.97)α-Tocopherol 20.6 (4.08) 29.7 (8.16) 23.6 (10.9) 26.6 (6.38) 29.6 (7.60) 33.0 (6.67) 32.7 (7.40) 25.7 (13.9–47.0)Lutein 0.42 (0.12)‡‡ 0.43 (0.24)‡‡ 0.49 (0.23)‡‡ 0.52 (0.25)‡‡ 0.26 (0.11)‡‡ 0.71 (0.30)‡‡ 0.24 (0.21)‡‡ 0.36 (0.14–0.74)‡‡

Zeaxanthin –‡‡ –‡‡ –‡‡ –‡‡ –‡‡ 0.09 (0.05) 0.07 (0.04) –‡‡

β-Cryptoxanthin 0.13 (0.08) 0.13 (0.11) 0.19 (0.14) 0.17 (0.12) 0.55 (0.11) 0.35 (0.27) 0.60 (0.47) 0.22 (0.05–0.52)Lycopene 0.33 (0.16) 0.28 (0.16) 0.31 (0.16) 0.32 (0.22) 0.37 (0.18) 0.56 (0.43) 0.42 (0.24) 0.40 (0.11–0.80)α-Carotene 0.08 (0.06) 0.10 (0.13) 0.13 (0.11) 0.14 (0.14) 0.07 (0.04) 0.36 (0.26) 0.07 (0.05) 0.08 (0.02–0.22)β-Carotene 0.49 (0.43) 0.42 (0.29) 0.60 (0.37) 0.64 (0.72) 0.38 (0.20) 0.81 (0.45) 0.37 (0.23) 0.34 (0.07–0.88)

* Means (standard deviation).† Concentrations in serum for men and women ranging from ages 19 to 62 years old, n = 111.‡ Concentrations in serum for women ranging from ages 35 to 50 years old, n = 96.§ Concentrations in serum for women ranging from ages 25 to 59 years old, n = 54.

** Concentrations in serum for men and women ranging from ages 4 to 93 years old, n = 3480.†† Means (concentration range).‡‡ Sum of lutein and zeaxanthin (peaks not separated).


74

Table III. The RSDs ranged from 3.3% (retinol) to 9.5%(lycopene) for intra-assay precision and 3.8% (α-tocopherol) to13.7% (β-cryptoxanthin) for interassay precision. The RSDvalues obtained for some carotenoids were comparable withthose reported for most of the other assays (16,31,32). However,the RSDs obtained for retinol and α-tocopherol were lower thanthose reported in these methods. This serum pool was then usedfor routine quality control.

Assay of liposoluble vitamins and carotenoids in a healthy population

The validated method was applied to the measurement ofretinol, α-tocopherol, and five carotenoids in the serum of 168healthy Caucasian subjects (Table IV). In comparison with previ-ously published studies, including more than 25 subjects(31,32,37,54), the value ranges were comparable for most of theliposoluble vitamins and carotenoids measured except forlutein–zeaxanthin, which was higher than the value found byTalwar et al. (32) but similar to other studies (31,54). Similarresults were demonstrated in a previous study by De Leehneer etal. (55). This fact can probably be explained by differencesbetween the populations involved in the different studies. We canalso notice that lutein–zeaxanthin and lycopene were in theserum in significant quantities, thus β-carotene could not bemeasured alone as a representative marker of the serumcarotenoids. As a matter of fact, the carotenoids exhibited dif-ferent distributions between subjects, tissues (56,57), and food(58). Moreover, their antioxidant capacities and functions maydiffer at the cellular level (59). More recently, an HPLC systemcoupling two different C18 columns has been reported for theseparation of 13 carotenoids in plasma (60), but the overall runtime for one sample reached 50 min, which limits thethroughput, and thus no important additional epidemiologicalinformation was given.

Conclusion

The reported HPLC method devoted to the assay of liposolublevitamins and carotenoids in serum permits the separation of themain carotenoids (lutein–zeaxanthin, β-cryptoxanthin, lyco-pene, α- and β-carotene, retinol, and α-tocopherol) within 20min, which allows a high throughput of samples. The methodwas run for several months in the routine laboratory and hasclearly proven its reliability. Because of its specificity and sensi-tivity for a great number of liposoluble vitamins correspondingwith important serum antioxidant biomarkers, this method hasan evident interest for nutritional and epidemiological studiesand is now applied to various pathological groups such as alco-holic and Type I diabetic patients.

Acknowledgments

This project was supported in part by a grant from theAssociation de la Recherche sur le Cholesterol (ARCOL). The

authors gratefully acknowledge the technical staff of the Centerof Preventive Medicine for their kind participation. The authorsthank the Société Francophone des Biofacteurs et Vitamines forgiving them the opportunity to participate in an interlaboratoryquality control.

References

1. E. Niki. Antioxidants in relation to lipid peroxidation. Chem. Phys.Lipids 44: 227–53 (1987).

2. K. Satoh, Y. Ida, H. Sakagami, T. Tanaka, and S. Fujisawa. Effect ofantioxidants on radical intensity and cytotoxic activity of eugenol.Anticancer Res. 18: 1549–52 (1998).

3. L. Tesoriere, A. Bongiorno, R. Re, and M.A. Livrea. Reciprocal pro-tective effects of all-trans retinol and alpha-tocopherol during lipidperoxidation in retinal membranes. Biochem. Mol. Biol. Int. 37: 1–7(1995).

4. L. Tesoriere, A. Bongiorno, A.M. Pintaudi, R. D’Anna, D. D’Arpa,and M.A. Livrea. Synergistic interactions between vitamin A andvitamin E against lipid peroxidation in phosphatidylcholine lipo-somes. Arch. Biochem. Biophys. 326: 57–63 (1996).

5. G.W. Burton and K.U. Ingold. beta-Carotene: an unusual type oflipid antioxidant. Science 224: 569–73 (1984).

6. A. Bendich and J.A. Olson. Biological actions of carotenoids. FASEBJ. 3: 1927–32 (1989).

7. N.I. Krinsky. Carotenoids and cancer in animal models. J. Nutr. 119:123–26 (1989).

8. R.M. Niles. Vitamin A and cancer. Nutrition 16: 573–76 (2000).9. D.C. McMillan, N. Sattar, D. Talwar, D.S. O’Reilly, and C.S.

McArdle. Changes in micronutrient concentrations following anti-inflammatory treatment in patients with gastrointestinal cancer.Nutrition 16: 425–28 (2000).

10. A.A. Botterweck, P.A. van den Brandt, and R.A. Goldbohm.Vitamins, carotenoids, dietary fiber, and the risk of gastric carci-noma: results from a prospective study after 6.3 years of follow-up.Cancer 88: 737–48 (2000).

11. D.S. Michaud, D. Feskanich, E.B. Rimm, G.A. Colditz, F.E. Speizer,W.C. Willett, and E. Giovannucci. Intake of specific carotenoids andrisk of lung cancer in 2 prospective US cohorts. Am. J. Clin. Nutr.72: 990–97 (2000).

12. E.S. Ford and W.H. Giles. Serum vitamins, carotenoids, and anginapectoris: findings from the National Health and NutritionExamination Survey III. Ann. Epidemiol. 10: 106–16 (2000).

13. R.W. Evans, B.J. Shaten, B.W. Day, and L.H. Kuller. Prospectiveassociation between lipid soluble antioxidants and coronary heartdisease in men. The multiple risk factor intervention trial. Am. J.Epidemiol. 147: 180–86 (1998).

14. F.S. Facchini, M.H. Humphreys, C.A. DoNascimento, F. Abbasi, andG.M. Reaven. Relation between insulin resistance and plasma con-centrations of lipid hydroperoxides, carotenoids, and tocopherols.Am. J. Clin. Nutr. 72: 776–79 (2000).

15. F. Granado, B. Olmedilla, E. Gil-Martinez, I. Blanco, I. Millan, andE. Rojas-Hidalgo. Carotenoids, retinol and tocopherols in patientswith insulin-dependent diabetes mellitus and their immediate rela-tives. Clin. Sci. (Colch.) 94: 189–95 (1998).

16. B. Olmedilla, F. Granado, E. Gil-Martinez, I. Blanco, and E. Rojas-Hidalgo. Reference values for retinol, tocopherol, and maincarotenoids in serum of control and insulin-dependent diabeticSpanish subjects. Clin. Chem. 43: 1066–71 (1997).

17. M.C. Polidori, P. Mecocci, W. Stahl, B. Parente, R. Cecchetti, A. Cherubini, P. Cao, H. Sies, and U. Senin. Plasma levels oflipophilic antioxidants in very old patients with type 2 diabetes.Diabetes Metab. Res. Rev. 16: 15–19 (2000).

18. M.A. Abahusain, J. Wright, J.W. Dickerson, and E.B. de Vol. Retinol,alpha-tocopherol and carotenoids in diabetes. Eur. J. Clin. Nutr. 53:630–35 (1999).


75

19. J.F. Salazar, H. Schorr, W. Herrmann, B. Herbeth, G. Siest, and P. Leroy. Measurement of thiols in human plasma using liquid chro-matography with precolumn derivatization and fluorescence detec-tion. J. Chromatogr. Sci. 37: 469–76 (1999).

20. G. Cao and R.L. Prior. Comparison of different analytical methodsfor assessing total antioxidant capacity of human serum. Clin.Chem. 44: 1309–15 (1998).

21. T. Arnhold and H. Nau. “Vitamin A”. In Modern AnalyticalMethodologies in Fat- and Water-Soluble Vitamins. John Wiley &Sons, Inc., New York, NY, 2000, pp. 3–49.

22. H.J. Nelis, E. D’Haese, and K. Vermis. “Vitamin E”. In ModernChromatographic Analysis of Vitamins, 3rd ed. Marcel Dekker, Inc.New York, NY, 2000, pp. 143–228.

23. R.R. Eitenmiller and W.O. Landen. “Simultaneous Determination ofFat-Soluble Vitamins in Food, Feed, and Serum”. In ModernAnalytical Methodologies in Fat- and Water-Soluble Vitamins. JohnWiley & Sons, Inc, New York, 2000, pp. 171–221.

24. A.B. Barua, H.C. Furr, J.A. Olson, and R.B. van Breemen. “VitaminA and Carotenoids”. In Modern Chromatographic Analysis ofVitamins, 3rd ed. Marcel Dekker, Inc., New York, NY, 2000, 1–74.

25. R.R. Eitenmiller and W. Landen. Vitamin Analysis for the Health andFood Sciences. CRC Press, Boca Raton, FL, 1998, 1–501.

26. R. Ohmacht, G. Toth, and G. Voigt. Separation of serum carotenoidsand vitamin A on chromsil-amino and -cyano phases by a bi-direc-tional gradient elution technique. J. Chromatogr. 395: 609–12(1987).

27. G. Panfili, P. Manzi, and L. Pizzoferrato. High-performance liquidchromatographic method for the simultaneous determination oftocopherols, carotenes, and retinol and its geometric isomers inItalian cheeses. Analyst 119: 1161–65 (1994).

28. R.B. McGeachin and C.A. Bailey. Determination of carotenoid pig-ments, retinol, and alpha-tocopherol in feeds, tissues, and bloodserum by normal phase high performance liquid chromatography.Poult. Sci. 74: 407–11 (1995).

29. P. Grolier. “Dosages des Caroténoïdes Sanguins”. In Le statut vita-minique. G. Le Moël, A. Saverot-Dauvergne, T. Gousson, and J.L.Guéant, Eds. EM Int., Cachan., France, 1998, pp. 42–48.

30. Z. Zaman, P. Fielden, and P.G. Frost. Simultaneous determination ofvitamins A and E and carotenoids in plasma by reversed-phaseHPLC in elderly and younger subjects. Clin. Chem. 39: 2229–34(1993).

31. A.L. Sowell, D.L. Huff, P.R. Yeager, S.P. Caudill, and E.W. Gunter.Retinol, alpha-tocopherol, lutein/zeaxanthin, beta-cryptoxanthin,lycopene, alpha-carotene, trans-beta-carotene, and four retinylesters in serum determined simultaneously by reversed-phase HPLCwith multiwavelength detection. Clin. Chem. 40: 411–16 (1994).

32. D. Talwar, T.K. Ha, J. Cooney, C. Brownlee, and D.S. O’Reilly. A routine method for the simultaneous measurement of retinol,alpha- tocopherol and five carotenoids in human plasma by reversephase HPLC. Clin. Chim. Acta 270: 85–100 (1998).

33. D. Hess, H.E. Keller, B. Oberlin, R. Bonfanti, and W. Schuep.Simultaneous determination of retinol, tocopherols, carotenes andlycopene in plasma by means of high-performance liquid chro-matography on reversed phase. Int. J. Vitam. Nutr. Res. 61: 232–38(1991).

34. M.A. Abahusain, J. Wright, J.W. Dickerson, M.A. el Hazmi, and H.Y. Aboul Enein. Determination of retinol, alpha-tocopherol,alpha- and beta-carotene by direct extraction of human serum usinghigh performance liquid chromatography. Biomed. Chromatogr. 12:89–93 (1998).

35. J.P. Steghens, B. Lyan, G. Le Moël, C. Galabert, V. Fayol, H. Faure,P. Grolier, N. Cheribi, F. Dubois, and F. Nabet. Measurement ofcarotenoids by high pressure liquid chromatography: from difficul-ties to solutions. Ann. Biol. Clin. (Paris) 58: 327–35 (2000).

36. O. Sommerburg, L.Y. Zang, and F.J. van Kuijk. Simultaneous detec-tion of carotenoids and vitamin E in human plasma. J. Chromatogr.B Biomed. Sci. Appl. 695: 209–15 (1997).

37. J.P. Steghens, A.L. van Kappel, E. Riboli, and C. Collombel.Simultaneous measurement of seven carotenoids, retinol and alpha-

tocopherol in serum by high-performance liquid chromatography. J. Chromatogr. B Biomed. Sci. Appl. 694: 71–81 (1997).

38. H.M. Arafa, F.M. Hamada, M.M. Elmazar, and H. Nau. Fully auto-mated determination of selective retinoic acid receptor ligands inmouse plasma and tissue by reversed-phase liquid chromatographycoupled on-line with solid-phase extraction. J. Chromatogr. A 729:125–36 (1996).

39. Q. Su, K.G. Rowley, and K. O’Dea. Stability of individualcarotenoids, retinol and tocopherols in human plasma during expo-sure to light and after extraction. J. Chromatogr. B Biomed. Sci.Appl. 729: 191–98 (1999).

40. D.W. Nierenberg and S.L. Nann. A method for determining con-centrations of retinol, tocopherol, and five carotenoids in humanplasma and tissue samples. Am. J. Clin. Nutr. 56: 417–26 (1992).

41. A. Sobczak, B. Skop, and B. Kula. Simultaneous determination ofserum retinol and alpha- and gamma-tocopherol levels in type IIdiabetic patients using high-performance liquid chromatographywith fluorescence detection. J. Chromatogr. B Biomed. Sci. Appl.730: 265–71 (1999).

42. H. Moriyama, H. Yamasaki, S. Masumoto, K. Adachi, N. Katsura,and T. Onimaru. Rapid determination of vitamins A and E in serumwith surfactant as a diluent by column-switching high-performanceliquid chromatography. J. Chromatogr. A 798: 125–30 (1998).

43. L. Yakushina and A. Taranova. Rapid HPLC simultaneous determi-nation of fat-soluble vitamins, including carotenoids, in humanserum. J. Pharm. Biomed. Anal. 13: 715–18 (1995).

44. K.S. Epler, R.G. Ziegler, and N.E. Craft. Liquid chromatographicmethod for the determination of carotenoids, retinoids and toco-pherols in human serum and in food. J. Chromatogr. 619: 37–48(1993).

45. R.W. Browne and D. Armstrong. Simultaneous determination ofserum retinol, tocopherols, and carotenoids by HPLC. MethodsMol. Biol. 108: 269–75 (1998).

46. J.R. Lane, L.W. Webb, and R.V. Acuff. Concurrent liquid chromato-graphic separation and photodiode array detection of retinol, toco-pherols, all-trans-alpha-carotene, all-trans-beta-carotene and themono-cis isomers of beta-carotene in extracts of human plasma. J. Chromatogr. A 787: 111–18 (1997).

47. M.H. Bui. Simple determination of retinol, alpha-tocopherol andcarotenoids (lutein, all-trans-lycopene, alpha- and beta-carotenes)in human plasma by isocratic liquid chromatography. J. Chromatogr. B Biomed. Appl. 654: 129–33 (1994).

48. S. Barua, S. Tarannum, L. Nahar, and M. Mohiduzzaman. Retinoland alpha-tocopherol content in breast milk of Bangladeshi mothersunder low socio-economic status. Int. J. Food Sci. Nutr. 48: 13–18(1997).

49. K.E. Savolainen, K.M. Pynnonen, S.P. Lapinjoki, and M.T. Vidgren.Determination of fat-soluble vitamins in a pharmaceutical dosageform by solid-phase extraction and reversed-phase liquid chro-matography. J. Pharm. Sci. 77: 802–803 (1988).

50. H. Berg, C. Turner, L. Dahlberg, and L. Mathiasson. Determinationof food constituents based on SFE: applications to vitamins A and Ein meat and milk. J. Biochem. Biophys. Methods 43: 391–401(2000).

51. M. Feinberg. La Validation des Méthodes d’Analyse. Une ApprocheChimiométrique de l’Assurance Qualité au Laboratoire. Masson,Paris, France, 1996.

52. M. Jezequel-Cuer, G. Le Moël, J. Mounie, J. Peynet, C. Le Bizec,M.H. Vernet, Y. Artur, A. Laschi-Loquerie, and S. Troupel.Determination of serum or plasma alpha-tocopherol by high perfor-mance liquid chromatography: optimization of operative models.Ann. Biol. Clin. (Paris) 53: 343–52 (1995).

53. J. Arnaud, I. Fortis, S. Blachier, D. Kia, and A. Favier. Simultaneousdetermination of retinol, alpha-tocopherol and beta-carotene inserum by isocratic high-performance liquid chromatography. J. Chromatogr. 572: 103–16 (1991).

54. B. Olmedilla, F. Granado, I. Blanco, and E. Rojas-Hidalgo. Seasonaland sex-related variations in six serum carotenoids, retinol, andalpha-tocopherol. Am. J. Clin. Nutr. 60: 106–10 (1994).


76

55. A.P. De Leenheer, H.J. Nelis, W.E. Lambert, and R.M. Bauwens.Chromatography of fat-soluble vitamins in clinical chemistry. J. Chromatogr. 429: 3–58 (1988).

56. W. Stahl, W. Schwarz, A.R. Sundquist, and H. Sies. cis-trans Isomersof lycopene and beta-carotene in human serum and tissues. Arch.Biochem. Biophys. 294: 173–77 (1992).

57. W. Stahl and H. Sies. Uptake of lycopene and its geometrical iso-mers is greater from heat-processed than from unprocessed tomatojuice in humans. J. Nutr. 122: 2161–66 (1992).

58. A.R. Mangels, J.M. Holden, G.R. Beecher, M.R. Forman, and E. Lanza. Carotenoid content of fruits and vegetables: an evaluationof analytic data. J. Am. Diet. Assoc. 93: 284–96 (1993).

59. F. Khachik, G.R. Beecher, M.B. Goli, W.R. Lusby, and J.C. Smith.

Separation and identification of carotenoids and their oxidationproducts in the extracts of human plasma. Anal. Chem. 64: 2111–22(1992).

60. B. Lyan, V. Azais-Braesco, N. Cardinault, V. Tyssandier, P. Borel,M.C. Alexandre-Gouabau, and P. Grolier. Simple method for clin-ical determination of 13 carotenoids in human plasma using an iso-cratic high-performance liquid chromatographic method. J. Chromatogr. B Biomed. Sci. Appl. 751: 297–303 (2001).

61. The Merck Index. An Encyclopedia of Chemicals, Drugs andBiologicals. Centennial Edition, 11th ed. Merck & Co., Inc., Rahway,NY, 1989.


77

A procedure for the sampling and analysis of energetics and relatedcompounds in the atmosphere is described. The basic procedureconsists of the collection of air samples using sampling cartridgescontaining XAD-2 resin, extraction of the resin with isoamyl acetate,and an analysis of the extract using gas chromatography withelectron capture detection. Modifications and additions to thisprocedure are discussed, such as the use of a prefilter before theresin sampler to collect particulates and the use of a mass selectivedetector to analyze for some propellant compounds of interest orfor quantitative confirmation purposes. Two differing sizes ofsamplers are evaluated according to the air volumes required forcollection. The procedure is tested through the analysis of spikedresin samples, which had air pulled through them for periods oftime corresponding with the required sampling volumes. Thisprocedure has application toward the measurement of energeticresidues in atmospheres resulting from weapons testing andoperations during training exercises involving munitions.

Introduction

The quantitative measurement of the residual amounts of ener-getics and related compounds in the environment has been rou-tinely performed for over three decades. There are numerousmethods used to analyze soils and waters for nitroaromatics,nitramines, and other compounds related to U.S. munitions(1–6). The U.S. Army has used many of these methods in thecourse of environmental monitoring to protect the health andsafety of soldiers and the general population. It has also relied onthese methods to measure soil and water contamination fromexplosives during environmental cleanup operations. The proce-dures generally involve gas chromatographic (GC) and high-per-formance liquid chromatographic (HPLC) analyses, but there arealso thin-layer chromatography and immunoassay methods thatare useful as field screening tests (7–8).

Additionally, there are methods used to monitor selected com-pounds such as trinitrotoluene (TNT) and dinitrotoluenes inworkplace atmospheres (9–10). This monitoring is used to ensurethat munitions workers are not exposed to harmful levels of these

compounds. However, there has been little done toward environ-mental air monitoring for energetics other than the specific caseof stack emissions produced during weapons destruction byincineration. The primary impetus for stack monitoring has beento determine destruction efficiencies associated with the pro-cesses used to burn the munitions feedstocks. The measurementof energetic and related compounds in the general atmospherefrom a health-risk standpoint has become an issue only in the lastfew years.

The U.S. Army has recognized the need to perform air moni-toring for energetics, partially because of public concern aboutair-quality issues in areas near U.S. military reservations. Thereare operations during weapons testing and training that arepotentially capable of putting measurable quantities of energeticsand related compounds into the atmosphere. As a result, theArmy Center for Health Promotion and Preventive Medicine(USACHPPM) has determined the need to modify current air-sampling methodologies and analytical techniques to providemonitoring efforts for a suite of explosives compounds, includingthose commonly analyzed for by soil and water methods. The listof compounds of concern includes the nitroaromatics (such asTNT, tetryl, and their precursors and breakdown products) andnitramines (such as hexahydro-1,3,5-trinitro-1,3,5-triazine(RDX) and octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine(HMX)). There are also other propellant compounds of occasionalconcern, including nitroglycerin, dibutyl- and dioctyl-phthalates,diphenylamine, and pentaerythritol tetranitrate (PETN).

There are U.S. Environmental Protection Agency (EPA) air-sampling procedures that employ sampling devices containingXAD-2 resin to trap polynuclear aromatic hydrocarbons fromambient air and semivolatile organic hazardous compounds instack emissions (11,12). USACHPPM has successfully used modi-fications of several types pertaining to the XAD-2 sampling trainsfor the collection of stack emissions for energetic residues. Wedecided therefore to investigate the use of glass cartridges packedwith XAD-2 resin for general atmospheric sampling for the ener-getics and propellant compounds. Preliminary tests were con-ducted using PS-1 cartridges manufactured for EPA Air ToxicsMethod TO-13 for polynuclear aromatic hydrocarbons, and a fieldstudy was successfully performed using these cartridges.Recently, newly designed cartridges have been employed. These

Abstract

A Procedure for Sampling and Analysis of Air forEnergetics and Related Compounds

Michael A. Hable, Joseph B. Sutphin, Curtis G. Oliver, Robert M. McKenzie, Eleonor F. Gordon, and Richard W. BishopThe U.S. Army Center for Health Promotion and Preventive Medicine, 5158 Blackhawk Road, Aberdeen Proving Ground, MD 21010-5403




78

cartridges are somewhat more robust during shipping and han-dling than the original types and are compatible with the sam-pling requirements of the U.S. Army during weapons testing.These cartridges are of two types: the first being a modification ofthe original PS-1 design used for high-volume sampling (Figure1A) and the second a smaller two-section cartridge designed forshorter test intervals (Figure 1B).

The analytical approach has generally been to use a modifica-tion of existing USACHPPM GC procedures for the nitro-con-taining compounds. These procedures use electron-capturedetection and have been used for many years in our laboratoriesto reliably quantitate these compounds from a variety of matrices(3,4,10,13). GC using a mass selective detector (MSD) was chosenas the most expedient means of analyzing for the phthalate estersand diphenylamine, because all can be done in a single GC run.The XAD-2 resin was solvent desorbed with isoamyl acetate inorder to place the analytes into solution prior to analysis. Isoamylacetate has proven to be an excellent solvent for the compoundsof interest. It also provides superior response and reproducibilitywith the electron-capture detector compared with other solventstried (such as acetonitrile). Finally, it helps to minimize chro-matographic problems that can arise with moisture-laden sam-ples because it is not water miscible (and thus does not retain thewater).

The sampling cartridges, the chromatographic and analyticalprocedures used to analyze for the compounds of concern, andthe test results from the spike studies conducted with the sam-plers will be described.

Experimental

Air-sampling cartridgesThe initial testing was done using PS-1 sampling cartridges

packed with 55 g of XAD-2 resin. Subsequent tests with these

samplers used 50 g rather than 55 g of the XAD-2, primarily as amatter of convenience. These cartridges are designed such thatthe glass cartridge contains a metal screen at the bottom to retainthe resin during sampling, and the resin is sandwiched above thescreen between two sections of glass wool. Design specificationsfor these cartridges vary, but the basic size of the inside samplingbed is 4 × 2.25 inches. The actual sampler and its calibration anduse has been described elsewhere (11).

The modified sampling cartridges were manufactured by AceGlass Inc. (Vineland, NJ). The larger size being tested stillemployed 50 g of XAD-2 resin, but the resin was retained by ametal screen/metal mesh combination at both ends of the car-tridge, with some glass wool at the outlet end only. The glass car-tridge body and contents were held together using Teflon endfittings. The smaller size cartridge used two 10-g sections of XAD-2 resin separated by glass wool. It also contained metal screensand mesh at both ends and glass wool between the screen/meshand resin at the outlet end of the cartridge. The inner diameter ofthe glass body was smaller, but the cartridge was similar to thelarge one in its use of metal screens and Teflon end fittings (asshown in Figure 1B). The second section of the smaller cartridgewas used as a back-up to measure breakthrough. If more than20% of the total of an analyte was found in this section, then thecartridge was considered to have been oversampled. Both types ofcartridges were compatible with the sampling devices used withthe original PS-1 cartridges and were used in the same fashion.

The XAD-2 resin used for packing the cartridges was astyrene–divinylbenzene porous polymer. It was purchased fromRestek Corporation (Bellefonte, PA) under the name “Ultra CleanXAD-2 Resin”. It was found to be sufficiently clean because it didnot require further purification for application toward energeticsampling. It was noted, however, that its appearance variedbetween different lots of the material. This did not seem to affectthe resin’s adsorbent properties, but it had other effects (as will bedescribed).

Recovery testsThe ability of the XAD-2 cartridges to retain the compounds of

concern while large volumes of air were passed through them wastested. Solutions containing known amounts of the analytes inacetonitrile were spiked into the front part of the resin within acartridge (in the case of a two-section cartridge the front sectionwas spiked). The cartridges were placed in a PS-1 sampling appa-ratus, and clean ambient air was pulled through in the same wayas it is generally done with actual sampling in the field. The car-tridges were then returned to the laboratory for the analysis andevaluation of analyte retention.

Analytical proceduresThe XAD-2 from sampled cartridges was transferred to 250-mL

glass bottles with Teflon-lined caps. The two section cartridgesused one bottle per section. Isoamyl acetate (Aldrich, Milwaukee,WI) was added to the containers to desorb the analytes of interestfrom the resin. The nominal amounts used were 100 mL for thelarge size samples and 25 mL for the small size samples. Usually,these amounts were sufficient to cover all the resin in a jar, butoccasionally they had to be increased to 125 mL and 40 mL,

A B

Figure 1. Cartridge designs for XAD-2 samplers used with energetics samplingin air: (A) 50-g cartridge with a modified PS-1 design used for high-volumesampling and (B) 10-g two-section XAD-2 cartridge used for shorter time sampling.


79

respectively, as a result of the excessive swelling of the resin whenplaced in the solvent. The reason for this was not known butappeared to vary with the lot of resin used to pack the cartridge.The jars or vials were agitated for 2 h on a platform-type shaker inorder to ensure adequate resin–solvent contact, then allowed tosit for at least 18 h in a refrigerator at 5°C. A portion of the solvent

extract was subsequently placed in an autosampler vial prior toanalysis for energetics and propellant compounds.

Energetic stock standards were purchased as 1.0-mg/mL solu-tions. All of them were available from AccuStandard Inc. (NewHaven, CT) except nitroglycerin, which was purchased fromCerilliant (Austin, TX). The diphenylamine and phthalate esterswere available as neat materials from Aldrich. Working standardsin isoamyl acetate were prepared from the stock standards (orneat materials). The energetic and nitroglycerin standards rangedfrom 0.01 to 2.0 µg/mL (0.02 to 4.0 µg/mL for HMX), and thestandards for MSD analysis were from 0.5 to 5.0 µg/mL.

The chromatographic analyses for the energetics and nitroglyc-erin were conducted using Agilent Technologies (Wilmington,DE) Model 6890 GCs equipped with electron-capture detectors.Chromatographic runs (shown in Figure 2) were typically madeusing a DB-1 column (J&W Scientific, Folsom, CA) (0.53-mm i.d.,1.0-µm film thickness) cut to 7 m in length. The GC oven wastemperature programmed from 80°C at 15°C/min to 140°C, thento 170°C at 3°C/min, and finally to 200°C at 5°C/min and held for3.0 min. The helium carrier gas was programmed from 2.0 psig(held for 13.0 min) to 4.0 psig at a rate of 150 psig/min and held.The dilution gas was nitrogen at 30 mL/min. The injection-porttemperature was set at 225°C, and the injection-port liner was aSilcosleeve (Restek) with a Silcosteel seal used in splitless mode.The Ni-63 electron capture detector temperature was 250°C. Dataprocessing was done using Turbochrom (PE Nelson, Shelton,CT). An Agilent Model 7673A autosampler was used to make theinjections (the injection volume was 1.0 µL).

Analyses for the phthalate esters and diphenylamine were doneusing an Agilent Technologies 5792 MSD interfaced with anAgilent 5890 GC. The analytical column was an RTX-5ms column(Restek) (0.25-mm i.d., 0.25-µm film thickness) that was 30 m inlength. The GC oven was temperature programmed from 80°C at30°/min to 260°C and then held for 6.0 min. The helium carriergas was set to a constant pressure of 14 psi. The injection-porttemperature was set at 275°C, and the injection-port liner was aSilcosleeve with a Silcosteel seal used in splitless mode. TheGC–MSD interface temperature was 260°C. An Agilent Model7673 autosampler was used to make the injections (the injectionvolume was 3.0 µL). The detector was scanned from m/z 45 to 300after a 5-min solvent delay. Data processing was done usingAgilent ChemStation software. Figure 3 shows a typical chro-matogram of the three analytes.


SamplingAll recovery tests were done using spiked cartridges. We recog-

nize that the ideal way to evaluate the cartridges would have beento sample atmospheres containing known concentrations of thetarget analytes, but unfortunately this was not an option. It wouldbe difficult (if not impossible) to generate stable atmospheres of aknown vapor concentration for many of the compounds. The sit-uation was further complicated by the requirement to samplevery large air volumes (a small test chamber would be inadequatefor such testing). Fortunately, the ability of XAD-2 to trap ener-

Figure 2. Chromatogram for energetics analysis on a 7-m, 0.53-mm-i.d., 1.0-µm film DB-1 column with electron-capture detection: (1) nitrobenzene, RT =1.57; (2) 2-nitrotoluene, RT = 2.03; (3) 3-nitrotoluene, RT = 2.29; (4) 4-nitro-toluene, RT = 2.41; (5) nitroglycerin, RT = 3.26; (6) 1,3-dinitrobenzene, RT =3.97; (7) 2,6-dinitrotoluene, RT = 4.13; (8) 2,4-dinitrotoluene, RT = 4.84; (9)3,4-dinitrotoluene, RT = 5.37; (10) 1,3,5-trinitrobenzene, RT = 6.32; (11)2,4,6-TNT, RT = 6.88; (12) RDX, RT = 8.62; (13) 4-amino-2,6-dinitrotoluene,RT = 11.04; (14) 2-amino-4,6-dinitrotoluene, RT = 12.14; (15) tetryl, RT =13.80; and (16) HMX, RT = 19.56.


80

getics has been demonstrated by actual field sampling. XAD-2resin has been successfully used by USACHPPM over the lastdozen years to sample for some of the target compounds of thisstudy. It has also been used for gas sampling during the testing ofproprietary methodology used for the destruction of chemicalmunitions. The multisection sampling tubes used in bothinstances were of a different design than the cartridges describedin this study, but the resin was the same. Recent field-samplingevents using the cartridges described in this study have alsoshown that the resin is effective in trapping the energetics withminimal breakthrough. We are confident that these spiking testsprovide an adequate further indication of the utility of these car-tridges for energetic trapping and retention.

The recovery results for seven replicate tests on the smaller car-

tridges are presented in Table I. These cartridges had 2.6 to 2.7 m3

of air pulled through them. Similar tests on the larger cartridges(but with 130 m3 of air) are shown in Table II.

Figure 3. Total ion chromatogram for propellant analysis of 5.0 µg/mLdiphenylamine and phthalate esters on a 30-m, 0.25-mm-i.d., 0.25-µm filmRTX-5MS column with mass selective detection: (1) diphenylamine, RT =5.25; (2) di-n-butylphthalate, RT = 6.35; and (3) dioctylphthalate, RT = 11.55.

Table I. Tests Using a Small Cartridge Containing Two 10-g XAD-2 Resin Sections*

%Relative Compound %Recovery standard deviation

2,6-Dinitrotoluene 95 18.82,4-Dinitrotoluene 93 18.13,4-Dinitrotoluene† 100 21.92,4,6-TNT 101 14.5RDX 125 17.3HMX 118 14.32-Nitrotoluene 77 16.93-Nitrotoluene 94 5.04-Nitrotoluene 95 19.3Nitrobenzene 85 15.61,3-Dinitrobenzene 93 17.71,3,5-Trinitrobenzene 94 16.54-Amino-2,6-dinitrotoluene 102 15.52-Amino-4,6-dinitrotoluene 109 14.3Tetryl 100 18.3Nitroglycerin 125 17.8Diphenylamine 91 8.3Di-n-butylphthalate 113 8.4Dioctylphthalate 109 8.2

* 2.6–2.7 m3 volume sampled. Seven spikes at 15 µg (energetics) or 75 µg (propellants).† Surrogate compound. Four replicates were tested.

Table II. Tests Using a Modified PS-1 CartridgeContaining One 50-g XAD-2 Resin Section*

%Relative Compound %Recovery standard deviation

2,6-Dinitrotoluene 88 8.12,4-Dinitrotoluene 89 7.23,4-Dinitrotoluene† 87 6.22,4,6-TNT 91 6.9RDX 101 5.2HMX 107 17.72-Nitrotoluene 99 13.93-Nitrotoluene 103 18.84-Nitrotoluene 114 17.2Nitrobenzene 89 9.11,3-Dinitrobenzene 87 7.41,3,5-Trinitrobenzene 85 7.94-Amino-2,6-dinitrotoluene 93 6.62-Amino-4,6-dinitrotoluene 103 5.1Tetryl 96 11.2Nitroglycerin 104 16.0Diphenylamine 88 3.9Di-n-butylphthalate 100 3.5Dioctylphthalate 106 3.3

* 130 m3 volume sampled. Seven spikes at 100 µg (energetics) or 500 µg (propellants).† Surrogate compound. Six spikes were done for this compound.


81

The recoveries were acceptable and no breakthrough wasobserved in any of the tests of the small cartridges. The effective-ness of desorbing the resin via shaking was tested by multipleextractions of spiked resins. One hour appeared to be sufficient torecover all the analytes, but two hours (followed by standingovernight) is recommended to ensure for full recovery. Theshakeout procedure was much simpler than the Sohxlet proce-dure used for desorbing XAD-2 with methylene chloride as man-

dated in the EPA organics procedures (11,12). A Sohxlet extrac-tion using isoamyl acetate would be difficult to conduct becauseof the high boiling temperature of the solvent. Extraction usingmethylene chloride is not recommended because it is a poor sol-vent for the nitramine compounds.

AnalysisThe chromatographic procedures used to analyze the resin

extracts were not complex for most of the analytes, but severalpotentially required some adjustment of analytical conditions.The nonpolar (DB-1) primary column we used for energetic anal-ysis was capable of separating all of the analytes of interest exceptPETN in a relatively short time. Most of the compounds were notsubject to interferences when used to analyze XAD-2 extractsfrom ambient air samples. However, there may be occasionalbackground interferences with the peaks for the nitrotoluene iso-mers or nitroglycerin depending on the lot of resin, isoamylacetate used, or both. When necessary, a column containing a dif-ferent liquid phase was used to quantitate compounds that couldnot be determined with the primary column. Secondary columnanalysis was also routinely done in order to verify positive detec-tions on the primary column. A polar DB-210 column (J&WScientific) was useful for this purpose. The temperature programwas from 80°C to 240°C, and the carrier gas (H2) was pro-grammed from 1.5 to 9 psig. Chromatographic conditions can bevaried, but a short column is recommended if HMX verification isrequired (shown in Figure 4). HMX is very reactive; a fast flow rateand temperature program is required to get it through the polarcolumn before peak degradation begins to occur.

One important factor that must be considered when per-forming GC analyses for energetic compounds is the use of aclean, properly silanized injection-port liner. Commercially pre-pared liners such as Silcosleeve are recommended. Peaks for themore reactive compounds (especially HMX and the aminodinitro-toluene isomers) will show distorted peak shapes or disappearentirely if the liner is dirty or not silanized. On-column injectionsare not recommended with this analysis because reproducibilityis not as good as with the splitless injections and column life maybe shortened.

A 30-m RTX-5ms column is recommended for the propellantcompound analysis on the GC–MSD, but a shorter column (10 m)can be used. The only consideration with a shorter column is theseparation of the diphenylamine from the isoamyl acetate solvent(there is not much separation between the two). If any of the later-eluting energetic compounds (trinitrobenzene and subsequent)(Figure 2 shows the elution order for energetics on the DB-1 andRTX-5ms columns) are present in the samples, they may bedetected during the propellant compound scan if they are presentin high enough concentrations. The earlier compounds elutewith the solvent front and HMX is not seen. HMX possibly breaksdown either when it contacts the metal parts of the detector or isso slowly eluted from the column that its peak flattens out com-pletely (or both).

The reporting limit for the energetics and nitroglycerin basedon the lowest injected standard is 0.4 µg for each compound (0.8µg for HMX) per cartridge for the small cartridge if 40 mL of thedesorbing solvent is used. It is 1.0 µg for each compound (2.0 µgfor HMX) for the larger cartridge desorbed with 100 mL. Similarly,

Figure 4. Chromatogram for energetics verification analysis on a 9-m, 0.53-mm i.d., 1.0-µm film DB-210 column with electron capture detection: (1)nitrobenzene, RT = 1.58; (2) 2-nitrotoluene, RT = 1.86; (3) 3-nitrotoluene, RT= 2.12; (4) 4-nitrotoluene, RT = 2.25; (5) 2,6-dinitrotoluene, RT = 4.17; (6)nitroglycerin, RT = 4.31; (7) 1,3-dinitrobenzene, RT = 4.39; (8) 2,4-dinitro-toluene, RT = 4.89; (9) 3,4-dinitrotoluene, RT = 5.66; (10) 2,4,6-TNT, RT =6.53; (11) 1,3,5-trinitrobenzene, RT = 6.75; (12) 4-amino-2,6-dinitrotoluene,RT = 7.21; (13) 2-amino-4,6-dinitrotoluene, RT = 7.65; (14) RDX, RT = 7.86;(15) tetryl, RT = 8.68; and (16) HMX, RT = 12.67.


82

the reporting limits for the propellant compounds are 20 µg and50 µg, respectively, based on their lowest injected standard.

The propellant compound PETN was not included during theconduct of these tests because it was difficult to separate chro-matographically from RDX. The two compounds coelute on boththe primary and secondary GC columns that were used duringthe test analyses. Because RDX is a major component of manyU.S. high explosives, it was considered the more important com-pound to evaluate during these cartridge studies. A subsequentcheck using other chromatographic columns has shown that aDB-1301 column (J&W Scientific) can separate the two com-pounds. This column can be used when both RDX and PETN arepotentially present in a sample. PETN is structurally related tonitroglycerin and thus is probably similarly collected and retainedby the XAD-2 resin. PETN spiked onto XAD-2 and desorbed withisoamyl acetate showed good extraction efficiency, but no testshave been conducted using spike cartridges with air pulledthrough them.

The XAD-2 resin cartridge was designed to collect the analytesof interest in the vapor state, but probably serves as a particulatetrap, also. A prefilter can be placed within the cartridge (or in aseparate housing before the cartridge) if differentiation betweenthe two physical forms is desired. A filter sampled separately fromthe resin should be placed in desorbing solvent soon after collec-tion, because many of the compounds of interest will evaporate orsublimate from a filter. A filter included within a cartridge wouldjust transfer evaporating compounds to the adjoining resin. Thisis the basis of the design of small sampling tubes containingTenax resin used in industrial hygiene applications involving airsampling for TNT and other substances (9,13).

Conclusion

A sampling and analytical procedure has been devised and vali-dated for the measurement of energetics and related compoundsin the atmosphere. The sampling cartridge is a successful modifi-cation of other resin-containing devices used for the collection ofsemivolatile organic compounds in air. GC with electron-capturedetection provides a very sensitive and reliable technique for thequantitation of nitro-compounds collected on the samplingmedia. Other compounds that may be of interest (such as the pro-pellant components described in this study) can easily be quanti-tated using GC–MSD.

References

1. U.S. Environmental Protection Agency. Test Methods for EvaluatingSolid Waste, Physical/Chemical Methods, SW-846 Update III. Officeof Solid Waste, Washington, D.C., 1997.

2. M.E. Walsh and T. Ranney. Determination of nitroaromatic,nitramine, and nitrate ester explosives in water using solid-phaseextraction and gas chromatography-electron capture detection: comparison with high-performance liquid chromatography. J. Chromatogr. Sci. 36: 406–16 (1998).

3. M. Hable, C. Stern, C. Asowata, and K. Williams. The determinationof nitroaromatics and nitramines in ground and drinking water bywide-bore capillary gas chromatography. J. Chromatogr. Sci. 29:131–35 (1991).

4. F. Belkin, R.W. Bishop, and M.V. Sheely. Analysis of explosives inwater by capillary gas chromatography. J. Chromatogr. Sci. 24:532–34 (1985).

5. M.E. Walsh and T. Ranney. Determination of Nitroaromatic,Nitramine, and Nitrate Ester Explosives in Soils Using GC-ECD.CRREL Special Report 99-12. U.S. Army Cold Regions Research andEngineering Laboratory, Hanover, NH, 1999.

6. T.F. Jenkins, M.E. Walsh, P.W. Schumacher, P.H. Miyares, C.F. Bauer,and C.L. Grant. Liquid chromatographic method for the determina-tion of extractable nitroaromatic and nitramine residues in soil. J. AOAC 72: 890–99 (1989).

7. P.G. Thorne and K.F. Myers. Evaluation of Commercial EnzymeImmunoassays for the Field Screening of TNT and RDX in Water.CRREL Special Report 97-32. U.S. Army Cold Regions Research andEngineering Laboratory, Hanover, NH, 1997.

8. S. Nam. On-Site Analysis of Explosives in Soil. Evaluation of Thin-Layer Chromatography for Confirmation of Analyte Identity. CRRELSpecial Report 97-21. U.S. Army Cold Regions Research andEngineering Laboratory, Hanover, NH, 1997.

9. OSHA Sampling and Analytical Methods, ORG 044, 2,4-Dintitrotoluene (DNT) and 2,4,6-Trinitrotoluene (TNT). U.S. Dept. ofLabor. Salt Lake City, UT, 1983, http://www.oshaslc.gov/dts/sltc/methods/organic/org044/org044.html.

10. R.W. Bishop, J.L. Kennedy, G.E. Podolak, and J.L. Ryea, Jr. A fieldevaluation of air sampling methods for TNT and RDX. Am. Ind. Hyg.J. 49(12): 635–38 (1988).

11. U.S. Environmental Protection Agency. Second Supplement toCompendium of Methods for the Determination of Toxic OrganicCompounds in Ambient Air, Method TO13. Revision EPA/600/4-89/018. Office of Solid Waste, Washington, D.C., June 1988.

12. U.S. Environmental Protection Agency. Test Methods for EvaluatingSolid waste, Physical/Chemical Methods. SW-846 Update III, Method0010. http://www.epa.gov/epaoswer/hazwaste/test/0010.pdf.

13. R.W. Bishop, T.A. Ayers, and D.S. Rinehart. The use of a solid sorbentas a collection medium for TNT and RDX vapors. Am. Ind. Hyg. J.42(8): 586–89 (1981).


83

The analysis of the binding data of D,L-dansyl amino acids on avancomycin stationary phase is investigated in relation to theaddition of N-acetyl-D-alanine in the mobile phase. This eluentadditive acts as a specific competing agent for the aglycone pocketof the immobilized chiral selector. A model taking into account bothstereoselective and nonstereoselective interactions between thesolutes and the stationary phase is used to fit the experimental data.From the results, the theoretical approach is considered to beadequate to describe the competing agent dependence on soluteretention. To the best of our knowledge, this report constitutes thefirst example of a displacement study on a macrocyclic antibioticstationary phase. This work shows that dansyl amino acids bind tothe active aglycone pocket of the selector and that this interaction isenantioselective. The results also demonstrate that additionalenantioselective sites at the vancomycin surface are involved in thechiral discrimination of these solutes.

Introduction

Various experimental approaches have been proposed to ana-lyze the mechanisms of enantioseparation on chiral stationaryphases (CSPs). For example, some studies have examined thetemperature effects on retention and enantioselectivity (1–4). Thechanges in enthalpy and entropy associated with the transfer ofthe solute can be extracted from linear van’t Hoff plots and ana-lyzed in order to obtain information about the driving forcesimplied in the association process. Another approach for studyingthe interactions between an enantiomer and a selector involvesthe variation of the mobile phase composition (5–10). Such aninvestigation has been carried out for several solute–CSP associa-tions by varying the proportion of the organic modifier (6), the pH(7,8), or the ionic strength (9,10) of the mobile phase.

The competitive (or displacement) approaches constituteanother powerful tool to examine the retention behavior of enan-tiomers in the chiral selective environment. Classically, it may becarried out by injecting one compound as the solute while a fixed

concentration of a possible competing agent is passed throughthe column in the mobile phase. Several examples of such studieshave been reported with immobilized proteins. Wainer et al.(11–13) have studied thoroughly the competitive displacement ofvarious drugs from a human serum albumin (HSA) stationaryphase by different competing agents. Hage et al. (14–16) have alsostudied the effects of additives (such as digitoxin, acetyldigitoxin,or chiral compounds) on the solute retention for immobilizedHSA. More recently, this approach has been applied with successto the study of enantiomer binding to the cellobiohydrolase sta-tionary phase, using cellulose as a competing compound (17).Also, the investigation of the retention and the enantioselectivityof a new CSP (immobilized fatty-acid-binding protein) has beencarried out for a large number of chiral compounds through dis-placement studies (18).

In this study, the displacement concept was applied specificallyto the investigation of the enantioselective and nonselectivebinding of test solutes (D,L-dansyl amino acids) on a vancomycinstationary phase. In order to obtain information about the role ofthe vancomycin aglycone pocket in the enantioselectivity process,the solute retention factor was plotted against the concentrationof the eluent N-acetyl-D-alanine (Ala). N-acetyl-D-Ala was used asa competing agent because it is able to bind specifically to this site(19). Using a general model describing the competing agent con-centration dependence on the solute binding, the retentionparameters as well as the association constant between N-acetyl-D-Ala and vancomycin were determined. The results will be dis-cussed in relation to the variation of enantioselectivity in order toprovide a precise picture of the enantiomer–selector associationprocess.

For high-performance affinity chromatography in the zonalelution mode, the total retention factor (k) is a direct measure ofthe solute interactions within the column. The parameter k isrelated to the number of binding sites of the analyte to the sta-tionary phase and the equilibrium constants for the solute atthese sites (13):

Eq. 1

where βi is the equilibrium constants at the individual sites (i), ji

Abstract

Displacement Study on a Vancomycin-Based StationaryPhase Using N-acetyl-D-Alanine as a Competing Agent

Ines Slama, Corinne Ravelet, Annick Villet, Anne Ravel, Catherine Grosset, and Eric Peyrin*Laboratoire de Chimie Analytique, UFR de Pharmacie de Grenoble, UJF, Domaine de la Merci, 38700 La Tronche, France





84

the fraction of each type of site, m the total moles of binding sites,Vm the void volume, and ki the respective contributions to thetotal retention factor.

The binding of a chiral compound to a CSP can involve twokinds of sites at the surface of the selector: one class of nonselec-tive binding sites with lower affinity and another class of enan-tioselective sites with higher affinity (17). Thus, k can besimplified as follows:

Eq. 2

where the two terms are the sums of the retention factors corre-sponding with selective (s) and nonselective (ns) interactions,respectively. For a pair of enantiomers, it is expected that the knsterms are identical while the ks contributions differ in relation tothe stereoselectivity.

Williams et al. (18,19) have previously shown by nuclear mag-netic resonance and modeling studies that a ligand such as N-acetyl-D-Ala is specifically bound in a 1:1 stoichiometry to thepocket of the aglycone of vancomycin (the structure of van-comycin is shown in Figure 1) via hydrophobic interactions andhydrogen bonds. This is explained by the fact that this antibioticacts on bacteria by binding to cell wall mucopeptide precursorsterminating in D-Ala. If N-acetyl-D-Ala is used as a mobile phaseadditive, it is expected that it will interfere on the retention ofsolutes interacting with the specific aglycone pocket. Thus,assuming that the pocket of the vancomycin aglycone constitutesone of the enantioselective sites of vancomycin for dansyl aminoacid, equation 2 can be modified as follows:

Eq. 3

where kls is the part of the retention factor implying the soluteenantioselective binding to the aglycone pocket, kns the part of

the retention factor involved in the nonspecific binding, and ksthe part of the retention factor corresponding with enantioselec-tive interaction unaffected by the competing agent. K is the asso-ciation constant between N-acetyl-D-Ala and vancomycin and cthe N-acetyl-D-Ala (competing agent) concentration. This simpli-fied model allows for a simple estimate of the role of the aglyconepocket on the enantioseparation of dansyl amino acids on immo-bilized vancomycin. Also, it constitutes a valuable tool to explorethe exact contributions of enantioselective interactions in theoverall retention process of these solutes.

Experimental

ApparatusThe HPLC system consisted of an LC 10AT Shimadzu pump

(Touzart et Matignon, Courtaboeuf, France), a Rheodyne Model7125 injection valve (Interchim, Montluçon, France) fitted with a20-µL sample loop, and a Shimadzu SPD-10A UV–vis detector. AnAstec 150- × 4.6-mm Chirobiotic V HPLC column (packed with astationary phase produced by the chemical bonding of the macro-cyclic glycopeptide vancomycin to a 5-µm silica gel) was usedwith controlled temperature in an Igloocil oven (Interchim). Themobile phase flow rate was 0.8 mL/min.

Reagents and operating conditionsD,L-amino acids were obtained from Sigma Aldrich (Saint-

Quentin, France). Methanol (HPLC grade), trisodium citrate, andcitric acid were supplied by Prolabo (Paris, France). Water wasobtained from an Elgastat option water purification system (Odil,Talant, France) fitted with a reverse osmosis cartridge. Thecolumn temperature was maintained at 25°C for all the experi-ments. The mobile phase consisted in citrate buffer (pH7.0)–methanol (90:10, v/v). The variation range of the N-acetyl-D-Ala concentration was 0 to 20mM. In order to examine the con-centration dependencies of the solute retention correspondingwith the binding capacity of the immobilized vancomycin, reten-tion measurements were related to varying amounts of injectedsolute. Solute samples were prepared at different concentrationsin the mobile phase from 0.125 to 10 µg/mL. The retention factorversus the sample amount plots exhibited a plateau at a sampleconcentration lower than 0.625 µg/mL, followed by a smalldecrease at higher solute concentrations. Thus, 20 µL of eachsolute at a concentration of 0.250 µg/mL was injected in triplicate(i.e., in linear conditions) (20).


The retention factor values for D,L-dansyl amino acids onimmobilized vancomycin were determined in relation to the con-centration of N-acetyl-D-Ala in the mobile phase (0 to 20mM). Thecoefficients of variation for the k values were < 0.5%, indicating ahigh reproducibility and a good stability for the chromatographicsystem. The k values were plotted against c for all the compounds.Figure 2 shows the k versus c plots of D,L-dansyl amino acids. In

Figure 1. Vancomycin structure with the aglycone binding pocket (indicatedby the arrow) (19).


85

all cases, the retention factors decreased when the competingagent concentration increased. Equation 3 was fitted to the exper-imental data using a nonlinear regression procedure. The valuesof the various parameters of equation 3 are shown in Table I. Thenonlinear regression coefficients (R) were higher than 0.987.Thus, it appears clearly that the behavior of the solutes was well-described by the model taking into account the competitionbetween N-acetyl-D-Ala and dansyl amino acids at the aglyconepocket of vancomycin. For all the dansyl amino acid enantiomers,the association constant between the competing agent and van-comycin varied between 820 and 1300 M–1. This demonstratedthat all the solutes studied (D- and L-enantiomers) interacted withthe aglycone pocket. The K average value was similar to the asso-ciation constant value reported for the N-acetyl-D-Ala binding tothe macrocycle binding pocket (i.e., 1300 M–1 at 23°C) (18). Fromthe magnitude of the retention factor kls, approximately 45–50%of the total binding observed for these compounds was dependenton the association with the aglycone pocket. The L-enantiomersexhibited a value of approximately 45%, and the D-enantiomerswere bound at approximately 50% to this active site. This demon-strates that the aglycone pocket is implied in the chiral discrimi-nation. Moreover, the apparent enantioselectivity (kD/kL)decreased when c increased (as shown in Figure 3). However,when the aglycone site was saturated by N-acetyl-D-Ala at a high

displacer concentration (as shown in Figure 1, in which a roughlyconstant value of the solute retention factor is observed for theconcentration range between 10 and 20mM), a substantial enan-tioselectivity remained for all the enantiomeric pairs. This suggests that other enantioselective sites are involved in the chiral discrimination of dansyl amino acids. The enantiose-lectivity values from the solute binding to the aglycone pocket(αp= klsD/ klsL) as well as the “residual” enantioselectivity values (αr =knsD+ksD/knsL+ksL) were calculated (Table II). Such an obser-vation is unusual in comparison with the results of other dis-placement studies carried out on protein CSPs. Usingimmobilized protein, only one enantioselective site is generallyinvolved in the chiral discrimination process (11–17). This orig-inal behavior, observed for the chiral recognition on a van-comycin stationary phase, can be explained by the fact that thistype of macrocycle contains several accessible chiral interactionsites. Moreover, Berthod et al. (21) have previously shown thatcarbohydrate moieties of the teicoplanin stationary phase offerenantioselective binding sites for various enantiomers. Thus, it isquite possible for dansyl amino acid enantiomers to interact withsome chiral environments of the selector in addition to the activeaglycone site. For the four pairs of dansyl amino acid enan-tiomers, a significant difference was observed for the αp values(from 1.26 to 1.48). This demonstrates that the aglycone pocket is

Table I. Determination of the Model Parameters byFitting Equation 3 to the D,L-Amino Acid EnantiomerRetention Factors on Immobilized Vancomycin

kns + ks kls K (M–1)

D-dansyl valine 1.51 1.58 820L-dansyl valine 1.32 1.07 1118D-dansyl serine 0.99 1.01 882L-dansyl serine 0.91 0.68 1249D-dansyl leucine 1.49 1.62 918L-dansyl leucine 1.45 1.18 1144D-dansyl phenylalanine 2.49 2.58 1080L-dansyl phenylalanine 2.29 2.00 1305

Figure 3. Plot of the apparent enantioselectivity (αapp) versus c for all enan-tiomeric pairs: D,L-dansyl valine (); D,L-dansyl serine (+); D,L-dansyl leucine(); and D,L-dansyl phenylalanine ().

Figure 2. Plot of k versus c for D,L-dansyl amino acids: dansyl valines D (+) andL ( ), dansyl serines D () and L (), dansyl leucines D () and L ( ), anddansyl phenylalanines D () and L ().

Table II. Determination of the Enantioselectivity Valuesfor All the Enantiomer Pairs

αapp* αp† α r

‡

(kD/kL) (klsD/klsL) (knsD + ksD/knsL + ksL)

Dansyl valine 1.26 1.47 1.14Dansyl serine 1.29 1.48 1.09Dansyl leucine 1.18 1.26 1.11Dansyl phenylalanine 1.18 1.28 1.08

* The apparent enantioselectivity.† The enantioselectivity resulting from the compound binding to the aglycone pocket.‡ The residual enantioselectivity.


86

responsible for different enantioselective interactions in relationto the structure of the compounds. Two solute groups can be dis-tinguished: (a) D,L-dansyl leucine and dansyl phenylalanine witha lower αp and (b) D,L-dansyl valine and dansyl serine with ahigher αp (Table II). Such a behavior can be explained by a sterichindrance phenomenon. The bulky dansyl amino acid (the firstsolute group mentioned) could limit access to the aglyconebinding pocket, and thus the chiral recognition would bereduced. This is confirmed by the fact that D,L-dansyl tryptophanenantiomers (the most bulky of the dansyl amino acids) were notseparated on this CSP (data not shown). Also, this result agreeswell with the findings of the enantioseparation of amino acidderivatives on a teicoplanin stationary phase (21,22).

Conclusion

This work demonstrates the interest to use N-acetyl-D-Ala as acompeting agent for the aglycone pocket of the macrocyclicantibiotic stationary phase. From the results, it is shown that thedansyl amino acids interact substantially with this active site(nearly 50% of the k value) of the vancomycin stationary phase.Furthermore, it is demonstrated that the aglycone pocket is aneffective enantioselective site, but another class of sites is alsoinvolved in the chiral discrimination. This type of displacementstudy could be applied similarly to the investigation of the relativecontribution of enantioselective and nonselective interactions forthe solute binding on other types of commercially availablemacrocycles (such as teicoplanin or ristocetin A) that contain asimilar aglycone pocket.

References

1. B. Sellergren and K.J. Shea. Origin of peak asymmetry and the effectof temperature on solute retention in enantiomers separations onimprinted chiral stationary phases. J. Chromatogr. A 690: 29–39(1995).

2. C.B. Castells and P.W. Carr. A study of the thermodynamics and influ-ence of temperature on chiral high performance liquid chromato-graphic separation using cellulose tris(3,5-dimethylphenyl-carbamate) coated zirconia stationary phases. Chromatographia 52:535–42 (2000).

3. V. Tittelbach and R.K. Gilpin. Species dependency of the liquid chro-matographic properties of silica-immobilized serum albumins. Anal.Chem. 67: 44–47 (1995).

4. E. Peyrin, Y.C. Guillaume, and C. Guinchard. Peculiarities of dansylamino acid enantioselectivity using human serum albumin as achiral selector. J. Chromatogr. Sci. 36: 97–103 (1998).

5. E. Peyrin and Y.C. Guillaume. Reanalysis of solute retention onimmobilized human serum albumin using fractal geometry. Anal.Chem. 71: 1496–99 (1999).

6. C. David, M.C. Millot, and B. Sebille. High-performance liquid chro-matographic study of the interactions between immobilized β-cyclodextrin polymers and hydrophobically end-capped polyethyl-

ene glycols. J. Chromatogr. B. 753: 93–99 (2001).7. A.M. Stalcup and K.H. Gahm. A sulfated cyclodextrin chiral sta-

tionary phase for high-performance liquid chromatography. Anal.Chem. 68: 1369–74 (1996).

8. E. Peyrin, C. Ravelet, E. Nicolle, A. Villet, C. Grosset, A. Ravel, and J. Alary. Dansyl amino acid enantiomer separation on a teicoplaninstationary phase: effect of eluent pH. J. Chromatogr. A 923: 37–43(2001).

9. A.M. Stalcup, K.H. Gahm, and M. Baldueza. Chiral separation ofchloroquine using heparin as a chiral selector in high-performanceliquid chromatography. Anal. Chem. 68: 2248–50 (1996).

10. Y.C. Guillaume, E. Peyrin, A. Villet, A. Nicolas, C. Guinchard, J. Millet, and J.F. Robert. Use of the Na+ ion as an RPLC retentionmarker to investigate the association of dansyl amino acids with per-methylated β-CD. Chromatographia 52: 753–57 (2000).

11. E. Domenici, C. Bertucci, P. Salvadori, S. Motellier and I.W. Wainer.Immobilized serum albumin: rapid HPLC probe of stereoselectiveprotein-binding interactions. Chirality 2: 263–68 (1990).

12. V. Andrisano, T.D. Booth, V. Cavrini, and I.W. Wainer.Enantioselective separation of chiral arylcarboxylic acids on animmobilized human serum albumin chiral stationary phase. Chirality9: 178–83 (1997).

13. T.A.G. Noctor, I.W. Wainer, and D.S. Hage. Allosteric and competi-tive displacement of drugs from human serum albumin by octanoicacid, as revealed by high-performance liquid affinity chromatog-raphy, on a human serum albumin-based stationary phase. J. Chromatogr. 577: 305–15 (1992).

14. B. Loun and D.S. Hage. Characterization of thyroxine-albuminbinding using high-performance affinity chromatography.Interactions at the warfarin and indole sites of albumin. J. Chromatogr. 579: 225–35 (1992).

15. A. Sengupta and D.S. Hage. Characterization of minor site probes forhuman serum albumin by high-performance affinity chromatog-raphy. Anal. Chem. 71: 3821–27 (1999).

16. D.S. Hage and A. Sengupta. Characterization of the binding of digi-toxin and acetyldigitoxin to human serum albumin by high-perfor-mance affinity chromatography. J. Chromatogr. B 724: 91–100(1999).

17. H. Henriksson, G. Pettersson, and G. Johansson. Discriminationbetween enantioselective and non-selective binding sites on cel-lobiohydrolase-based stationary phases by site specific competingligands. J. Chromatogr. A 857: 107–15 (1999).

18. G. Massolini, E. De Lorenzi, E. Calleri, C. Bertucci, H.L. Monaco, M. Perduca, G. Caccialanza, and I.W. Wainer. Properties of a sta-tionary phase based on immobilized chicken liver basic fatty-bindingprotein. J. Chromatogr. B 751: 117–30 (2001).

19. D.H. Williams, M.S. Searle, J.P. Mackay, U. Gerhard, and R. Maplestone. Toward an estimation of binding constants inaqueous solution: studies of associations of vancomycin groupantibiotics. Proc. Natl. Acad. Sci. USA 90: 1172–78 (1993).

20. E. Peyrin, Y.C. Guillaume, and C. Guinchard. Characterization ofsolute binding at HSA site II and its geometry using biochromato-graphic approach. Biophys. J. 77: 1206–12 (1999).

21. A. Berthod, X. Chen, J.P. Kullman, D.W. Armstrong, F. Gasparrini, I. D’Acquarica, C. Villani, and A. Carotti. Role of the carbohydratemoieties in chiral recognition on teicoplanin-based LC stationaryphases. Anal. Chem. 72: 1767–80 (2000).

22. E. Peyrin, A. Ravel, C. Grosset, A. Villet, C. Ravelet, E. Nicolle, and J. Alary. Interactions between D,L dansyl amino acids and immobi-lized teicoplanin: study of the dual effect of sodium citrate on chiralrecognition. Chromatographia 53: 645–50 (2001).


The influence of HCl concentration (6M, 8M, and 10M) and theratio of sample protein to acid (1 or 5 mg of protein per mL ofacid) on furosine formation during sample hydrolysis is studied.The conditions that maximize furosine formation are 10M HCl inthe ratio of 1 mg of protein to 1 mL of acid. Purification of thehydrolysate by solid-phase extraction is also considered byexamining the effect of hydrolysate volume and volume of 3M HClused to elute the furosine. Furosine quantitation is carried outusing the standard additions and external standard methods. Theresults indicate that there is no interference by the sample matrixand that external calibration is adequate.

Introduction

Furosine (ε-N-(2-furoylmethyl)-L-lysine) is an amino acidformed during the acid hydrolysis of such Amadori products asfructoselysine, lactuloselysine, and maltuloselysine, which aregenerated in the early stages of the Maillard reaction during theheat processing of foods (1). For that reason, estimates of theextent of protein damage caused by heating in the first stagesof that reaction are often based on determinations of theamount of furosine that forms during the acid hydrolysis offoods.

Furosine determinations may be carried out by ion-exchangechromatography (IEC) (2–4), gas chromatography (5), andion-pair reversed-phase high-performance liquid chromatog-raphy (RP-HPLC) (6–12). Recently, capillary electrophoresis(CE) methods of furosine determination have also been devel-oped (13–15).

The first step in furosine analysis is hydrolysis of the sampleproteins. Although optimization of the hydrolysis step has

been considered in previous research (2,3,8–12,16), Resmini etal. (8) pointed out that more research on the effect of acidhydrolysis on furosine formation was needed. Additional issuesassociated with the assay include whether hydrolysate cleanupby solid-phase extraction improves assay performance andwhat calibration strategy produces optimum performance.

As a consequence of the foregoing, the object of this studywas to examine the influence of HCl concentration and ratio onfurosine formation and to establish the most suitable condi-tions for hydrolysate purification and furosine quantitationbased on the chromatographic conditions developed by Del-gado et al. (9).

Experimental

Chemicals and reagentsThe furosine standard with a purity of approximately 70%

was obtained from Neosystem (Strasbourg, France). HPLC-grade acetonitrile was from Scharlau (Barcelona, Spain). Theother analytical reagent-grade chemicals were from Merck(Darmstadt, Germany). Water was quartz-distilled and deion-ized using the Milli-Q system (Millipore, Bedford, MA).

EquipmentThe HPLC apparatus consisted of a Model 110B pump and a

Model 210A injector from Beckman (Berkeley, CA) equippedwith a 20-µL loop and a KNK-029.757 UV–vis detector (KonikInstruments S.A., Barcelona, Spain). Peak areas were deter-mined with the aid of an SP-4290 recorder–integrator (Spectra-Physics, San Jose, CA).

SamplesThe trials were performed using two samples (A and B) of

a powdered enteral formula and two samples (A and B) of

Abstract

M.A. Serrano, G. Castillo, M.M. Muñoz, and A. Hernández*Departamento de Nutrición y Bromatología, Campus Universitario, Universidad de Alcalá, 28871-Alcalá de Henares (Madrid), Spain

Reproduction (photocopying) of editorial content of this journal is prohibited without publisher’s permission. 87



Influence of Hydrolysis, Purification, and CalibrationMethod on Furosine Determination UsingIon-Pair Reversed-Phase High-PerformanceLiquid Chromatography


88

powdered whole milk. Caseinate was the protein source for theenteral formula, and the protein content was 20%. The proteincontent of the powdered whole milk was 25%.

Sample hydrolysisIn order to determine the influence of HCl concentration

and ratio on furosine formation during hydrolysis, three HClconcentrations (i.e., 6M, 8M, and 10M) and two ratios of sampleprotein to acid volume (i.e., 1 and 5 mg of protein per mL ofacid) were tested. A quantity of sample accurately weighed waspoured into a 250-mL Pyrex screw-cap flask, and an appropriatequantity of acid was added. Hydrolysis was carried out in anitrogen atmosphere at 110ºC for 24 h. After the hydrolysatehad cooled to room temperature, it was filtered through No. 52Whatman paper (Whatman, Maidstone, U.K.), and the screw-cap flask was washed out with Milli-Q water. All the liquids werecollected in a volumetric flask that was then topped off withMilli-Q water.

For the sample protein-to-acid ratio trials using 1 mg ofprotein per mL of acid, approximately 0.20 g of the powderedmilk or 0.25 g of the enteral formula was weighed out andhydrolyzed with 50 mL of acid, and the volume was diluted to100 mL. For the sample protein-to-acid ratio trials using5 mg of protein per mL of acid, approximately 0.40 g of thepowdered milk or 0.50 g of the enteral formula was weighedand hydrolyzed with 20 mL of acid, and the volume was dilutedto 50 mL.

Hydrolysate purificationThe different sample hydrolysates were purified using a

Sep-Pak C18 cartridge (WAT020515, Waters, Milford, MA). Thecartridges were prewetted with 5 mL of methanol and 10 mLof Milli-Q water before use.

Different trials were performed to ascertain the optimal con-ditions for purification. Aliquots of 0.5 and 1 mL of filteredhydrolysate of an enteral formula were gradually loaded ontothe cartridge, and the displaced liquid was collected in an evap-oration flask, being careful not to allow air to enter the car-tridge. Elution of the furosine on the cartridge was then carriedout using 3 or 5 mL of 3M HCl, the eluate being collected inthe same flask. The solution thus obtained was evaporated todryness in a rotary evaporator at 30ºC. The dry residue wasreconstituted with acetonitrile–Milli-Q water–formic acid(20:79.8:0.2). Other aliquots of the same filtered hydrolysatewere injected onto the chromatograph without undergoingpurification.

Other parallel trials were performed using 0.5 mL filteredhydrolysate of the enteral formula and the powdered wholemilk in the same conditions to corroborate the effect of col-lecting or discarding the displaced liquid when running thehydrolysate through the cartridge.

Quantitative analysisQuantitation was performed using the external standard

method. A stock solution with approximately 140 µg/mL ofpure furosine standard was prepared by dissolving the totalamount of a commercial vial in 0.1M HCl. This stock solutionwas stored under refrigeration at 4ºC.

An appropriate aliquot of the stock solution was evaporatedto dryness in a rotary evaporator, and the dry residue wasreconstituted with an appropriate volume of acetonitrile–Milli-Q water–formic acid (20:79.8:0.2). Using that same solvent,eight standard dilutions ranging in concentration from 1 to 8µg/mL were prepared from the solution that was obtained. Acalibration curve was obtained by plotting the peak areas versusthe micrograms per milliliter of furosine injected.

Calibration by the standard additions method was also tested.For this purpose, five solutions of each sample (powderedenteral formula and powdered milk) were prepared by taking auniform quantity of hydrolysate and increasing quantities offurosine standard (ranging from 1 to 5 µg/mL). Thus, twostandard additions curves were obtained.

All standard solutions and samples were injected twice.

Chromatographic conditionsFurosine was determined by ion-pair RP-HPLC according to

the method of Delgado et al. (9). Separations were carriedout on a Spherisorb ODS2 5-µm column (250- × 4.6-mm i.d.)(Phenomenex, Torrance, CA) thermostatted at 30ºC. Themobile phase was 5mM sodium heptane sulfonate with 20%acetonitrile and 0.2% formic acid at a flow rate of 1.2 mL/min.Detection was carried out at 280 nm.


The chromatographic method of Delgado et al. (9) wasselected because it was an isocratic method in which furosinewas eluted after a short retention time, thereby reducingthe analysis time. In view of the discrepancies concerninghydrolysate purification contained in the literature, the firstquestion addressed was whether or not a purification stagewas necessary and what the optimum purification conditionswere.

Figure 1 presents the chromatograms for an unpurified andpurified hydrolysate of an enteral formula. The furosine elutedafter a retention time of 10 min. On the chromatogram for theunpurified hydrolysate, a series of small spikes can be observedalong the entire baseline, with two spikes being located quiteclose to the furosine peak. The chromatogram for the purifiedhydrolysate presented a more stable baseline and a smallernumber of peaks, which translates into better separation andintegration. Furthermore, the lifetime of the chromatographiccolumn was extended by purification of the hydrolysates.

Two trials were run to test the purification conditions. Thefirst trial was carried out on a hydrolyzed enteral formula.Two different volumes of hydrolysate (0.5 and 1 mL) were puri-fied, with the displaced liquid collected in both cases. In addi-tion, two different volumes (3 and 5 mL) of 3M HCl for furosineelution were tested. Other aliquots of hydrolysate were injectedwithout purification of any kind. Four replications were per-formed for each set of conditions.

The results are presented in Table I. There was no significantdifference in the furosine values obtained under either set ofconditions when 0.5 or 1 mL of the hydrolysate was run


89

through the cartridge. Thus, the step ofsolid-phase extraction can be used notonly for purification of the hydrolysatebut also to increase the amount of furo-sine in the dry residue obtained, providedthat the displaced liquid is also collected.This advantage may be quite useful whenanalyzing samples containing smallerquantities of furosine, such as pasteur-ized milk.

In addition, the amount of furosineobtained on elution with 3 mL of HCl was9–12% lower than the amount obtained inthe unpurified hydrolysate injecteddirectly onto the column. However, elu-tion carried out using 5 mL of HCl yieldedthe same amount of furosine as in theunpurified hydrolysates. Thus, collectingthe displaced liquid when running thehydrolysate through the cartridge andeluting with 5 mL of 3M HCl implies afurosine recovery of 100% and makes itunnecessary to use a correction factor.

Another trial was carried out to con-firm the previously mentioned results andto try to elucidate the effect of discardingor collecting the displaced liquid whenrunning the sample through the car-tridge. In this trial samples of both anenteral formula and powdered milk wereused. Two hydrolysates were obtainedfrom each sample. Parallel trials were per-formed using 0.5 mL of each hydrolysate,discarding or collecting the displacedliquid from the cartridge, and eluting thefurosine with 3 or 5 mL of 3M HCl. TableII summarizes the results, which showsthat the furosine values in all cases werehigher when the displaced liquid was notdiscarded than when it was discarded,though the difference was small (approx-imately 1% to 3%) and only statisticallysignificant for the enteral formula. Thevalues were also higher when the furo-sine was eluted using 5 mL of acid insteadof 3 mL, but in this case the differencewas higher (approximately 9%) and sta-tistically significant for both samples. Useof a larger volume of HCl for elution inlaboratory tests confirmed that the sameresults were obtained.

It can therefore be concluded that, inthe conditions of the experiment, theeffect of using a larger volume of HCl toelute the furosine was greater than col-lecting or discarding the displaced liquid.Nevertheless, if a larger volume ofhydrolysate were to undergo purification,

Figure 1. Chromatograms for an (A) unpurified and (B) purified hydrolysate of an enteral formula.

Table I. Effect of Hydrolysate Volume Purified and 3M HCl Volume Used as theElution Solvent on the Furosine* Determination in a Powdered Enteral Formula†

Volume of hydrolysate run through the Sep-Pak cartridge

Enteral formula A 0.5 mL 1 mL

Unpurified hydrolysate 66.21 ± 2.48 66.70 ± 0.71Purified hydrolysate

Elution with 3 mL 3M HCl 58.22 ± 0.78 60.73 ± 0.95Elution with 5 mL 3M HCl 66.29 ± 0.71 66.68 ± 0.01

* Milligrams per 100 g of product. Values are the means of four replications ± standard deviation.† Hydrolysate prepared using 6M HCl at a ratio of 5 mg protein to 1 mL HCl.

Table II. Effect of Collecting or Discarding the Liquid Displaced from theCartridge by Hydrolysate and the 3M HCl Volume Used as the ElutionSolvent on the Furosine* Determination in a Powdered Milk† and a PowderedEnteral Formula†

Displaced hydrolysate

Sample Discarded Collected

Powdered milk BElution with 3 mL 3M HCl 120.83 ± 6.00 124.05 ± 2.34Elution with 5 mL 3M HCl 131.94 ± 2.87 132.86 ± 2.55

Powdered enteral formula BElution with 3 mL 3M HCl 128.25 ± 4.56 133.12 ± 4.41Elution with 5 mL 3M HCl 139.49 ± 4.99 143.53 ± 6.91

* Milligrams per 100 g of product. Values are the means of four replications ± standard deviation.† Hydrolysate prepared using 10M HCl at a ratio of 1 mg of protein to 1 mL HCl.


90

the error produced by discarding the displaced liquid couldbecome considerable. The rest of the trials were performed inthe optimum conditions described previously.

Conflicting information has been reported in terms of themost appropriate calibration method to use with both the useof the external standards (8,17) and standard additions (9,12)mentioned. Because calibration by means of the standard addi-tions method makes analysis both longer and more compli-cated because a separate calibration curve is required for everysample with a different composition, the calibration condi-tions were studied. Two calibration curves (one for the pow-dered milk and another for the enteral formula) were obtainedby the standard additions method by adding increasing quan-tities from 1 to 5 µg/mL of the furosine standard to the corre-sponding previously hydrolyzed samples. Two other externalstandard calibration curves were also obtained.

Table III summarizes the results of the calibration study.The slopes of the two calibration curves, the external stan-dard calibration curve, and the standard additions calibrationcurve were quite similar. A statistical comparison of the slopesof the two curves found no statistically significant differencesbetween them (P < 0.05). Therefore, it can be concluded that

there were no effects attributable to the hydrolysate matrix ofeither of the samples tested, even though they differed con-siderably in composition.

The furosine concentration in the samples was also calcu-lated using the two calibration curves, and the furosinerecovery (%) was calculated by the ratio between furosinedetermined with standard additions and external standardcurve equations (Table III). The recovery was 100%, whichagain confirmed that there were no differences attributable tothe sample matrix.

One of the drawbacks to the standard additions method isthat sample concentration is calculated by extrapolation ratherthan by interpolation while it is in the external standard cali-bration method. For this reason, the accuracy of the externalstandard method was also evaluated by calculating the recoveryof furosine as the known (increasing quantities of the standardwere added to the samples). The recoveries are shown in TableIV. Mean recovery values for the two samples were nearly 100%,with coefficients of variation lower than 0.2%. Thus, calcula-tion error resulting from extrapolation would appear to beminimal, and calibration by the external standard method isadequate.

Table III. Calibration Curve Equations and Furosine Recovery for a Powdered Enteral Formula and a Powdered Milk

External standard calibration Standard additions calibration

Furosine injected Furosine injected RecoverySample Curve equation (µg/mL) Curve equation (µg/mL) (%)

Enteral formula A y = 71526.1x – 2547.1 3.39 ± 0.01 y = 70428.9x + 239409 3.40 100.2s.e.* = 1454 s.e. = 941r2† = 0.9999 r2 = 0.9998

Powdered milk A y = 71481.4x – 2100.8 2.10 ± 0.01 y = 70304.2x + 147529 2.09 99.7s.e. = 2157 s.e. = 1210r2 = 0.9999 r2 = 0.9999

* s.e., standard error of estimation.† Determination coefficient.

Table IV. Percentage Furosine Recovery for a PowderedEnteral Formula and a Powdered Milk

Furosine (µg/mL)Recovery

Sample Initial Added Recovered (%)

Enteral formula A 3.39 0.95 4.30 99.081.42 4.78 99.381.89 5.25 99.432.37 5.72 99.31

Mean value 99.30 ± 0.15

Powdered milk A 2.10 0.95 3.02 99.021.89 3.94 98.752.84 4.89 98.993.79 5.84 99.15

Mean value 98.98 ± 0.17

Table V. Effect of HCl Concentration and Ratio DuringHydrolysis on Furosine* Formation in a PowderedEnteral Formula and a Powdered Milk

Hydrolysis conditions Enteral formula A† Powdered milk A†

1 mg protein to 1 mL acid10M HCl 546.22 ± 4.39 (100) 322.95 ± 0.95 (100)8M HCl 460.45 ± 2.41 (84.3) 289.37 ± 4.55 (89.6)6M HCl 332.66 ± 1.24 (60.9) 195.63 ± 0.43 (60.6)

5 mg protein to 1 mL acid10M HCl 493.59 ± 4.40 (100) 306.39 ± 0.98 (100)8M HCl 428.14 ± 5.56 (86.7) 268.89 ± 5.53 (87.8) 6M HCl 320.26 ± 2.56 (64.9) 161.66 ± 1.47 (52.8)

* Milligrams per 100 g protein. Values are the means of three replications ±standard deviation.

† Difference in percentage furosine formation with respect to the values obtainedusing 10M HCl in brackets.


91

The effect of different hydrolysis conditions on the amountof furosine formed was tested under optimal purification con-ditions using external standard calibration. Three acid con-centrations (6M, 8M, and 10M) and two quantities of proteinper mL of acid (1 mg of dilute hydrolysis and 5 mg of concen-trated hydrolysis) were tested. Samples of the enteral formulaand the powdered milk were used, and the results wereexpressed as milligrams of furosine per 100 g of protein in thesample. The results of these trials appear in Table V.

The amount of furosine formed increased with increasingacid concentration in both the dilute and concentrated hydrol-ysis of both samples. Furosine formation was thus highest forthe 10M HCl. The difference in the level of furosine formationwas greater between the 6M and 8M acid concentrations thanbetween the 8M and 10M acid concentrations for both themilk and the enteral formula.

The protein-to-acid ratio during hydrolysis also affects furo-sine formation. Furosine formation was in all cases higher forthe ratio of 1 mg of protein to 1 mL of acid (in other words, fordilute hydrolysis). Nevertheless, the influence of HCl concen-tration was much greater than that of the sample protein-to-acid ratio during hydrolysis. The dilute ratio has customarilybeen recommended for the analysis of total amino acids (16).

The increase in furosine formation according to HCl acidconcentration is in agreement with the results obtained for dif-ferent foods by other researchers (2,3,11,12,17).

No information concerning the influence of the protein-to-acid ratio on furosine formation was found in the literature.

Based on the experimental results, it can be concluded thatoptimum hydrolysis conditions to maximize furosine forma-tion are 10M HCl in the ratio of 1 mg of protein to 1 mL of acid.Because furosine is formed from the Amadori products duringthe hydrolysis of foodstuffs, its concentration was used to eval-uate thermal damage sustained by foods during processing, andit will therefore in all cases be appropriate to try to maximizefurosine formation to ensure the maximum correctness of theevaluation. This is even more important in samples thatundergo milder heat processing.

In liquid foods the concentration of the HCl added to thesamples should be higher, so that the final acid concentrationduring hydrolysis will be 10M. Although the concentration ofcommercial HCl acid concentrate is approximately 11.9M, con-centrations higher than 10M were not tested in order to ensurecomparability of the results for the solid and liquid samples.

Acknowledgments

This study was supported by a grant awarded by theComisión Interministerial de Ciencia y Tecnología (Project

ALI97-0606-C02-02) of the Spanish Ministry of Education andScience.

References

1. H. Erbersdobler and H. Zucker. Untersuchungen zum gehalt anlysin und verfügbarem lysin in trockenmagermilch. Milchwis-senschaft 21: 564–68 (1966).

2. A. Brandt and H. Erbersdobler. Zur bestimmung von furosin innahrungs- und futtermitteln. Landwirtsch. Forsch, Sonderheft28/II: 115–19 (1973).

3. I. Molnár-Perl, M. Pintér-Szakács, R. Wittmann, M. Reutter, andK. Eichner. Optimum yield of pyridosine and furosine originatingfrom Maillard reactions monitored by ion-exchange chromatog-raphy. J. Chromatogr. 361: 311–20 (1986).

4. J. Hartkopf and H. Erbersdobler. Stability of furosine during ion-exchange chromatography in comparison with reversed-phasehigh-performance liquid chromatography. J. Chromatogr. 635:151–54 (1993).

5. W. Büser and H. Erbersdobler. Determination of furosine by gas-liquid chromatography. J. Chromatogr. 346: 363–68 (1985).

6. E. Schleicher and O.H. Wieland. Specific quantitation by HPLCof protein (lysine) bound glucose in human serum albumin andother glycosylated proteins. J. Clin. Chem. Clin. Biochem. 19:81–87 (1981).

7. G.H. Chiang. A simple and rapid high-performance liquid chro-matographic procedure for determination of furosine, lysine-reducing sugar derivative. J. Agric. Food Chem. 31: 1373–74 (1983).

8. P. Resmini, L. Pellegrino, and G. Battelli. Accurate quantificationof furosine in milk and dairy products by a direct HPLC method.Ital. J. Food Sci. 3: 173–83 (1990).

9. T. Delgado, N. Corzo, G. Santa-Maria, M. L. Jimeno, andA. Olano. Determination of furosine in milk samples by ion-pairreversed phase liquid chromatography. Chromatographia 33:374–76 (1992).

10. T. Maroni and P. Lazzari. Determinazione di furosina in HPLC conmetodo isocratico. Ind. Aliment. 33: 964–66 (1994).

11. T. Henle, G. Zehetner, and H. Klostermeyer. Fast and sensitivedetermination of furosine. Z. Lebensm. Unters. Forsch. 200:235–37 (1995).

12. E. Guerra-Hernández and N. Corzo. Furosine determination inbaby cereals by ion-pair reversed-phase liquid chromatography.Cereal Chem. 73: 729–31 (1996).

13. A. Tirelli and L. Pellegrino. Determination of furosine in dairyproducts by capillary zone electrophoresis. A comparison with theHPLC method. Ital. J. Food Sci. 4: 379–85 (1995).

14. L. Del Giovine and A. Bocca. Elettroforesi capillare applicataall’analisi della furosina nel latte. Riv. Sci. Aliment. 3: 247–52 (1996).

15. A. Tirelli. Improved method for the determination of furosine in foodby capillary electrophoresis. J. Food Prot. 61: 1400–1404 (1998).

16. J.W. Finley. Digestibility and Amino Acid Availability in Cerealsand Oilseeds. J.W. Finley and D. Hopkins, Eds. Am. Assoc. CerealChem., St Paul, MN, 1985, pp. 15–30.

17. F. Evangelisti, C. Calcagno, S. Nardi, and P. Zunin. Deteriorationof protein fraction by Maillard reaction in dietetic milks. J. DairyRes. 66: 237–43 (1999).

Manuscript accepted November 26, 2001.

More specific official methodology is needed to survey the illegaluse of clenbuterol in animal production plus the synthesis of newcompounds that currently elude routine analytical methods. Theidentification of a new adrenergic agonist, N1-(2-(4-amino-3,5-dichlorophenyl)-2-hydroxyethyl)-N1-isopropyl-propanamide(known as compound A) in animal feed has prompted studies toverify if the existing cleanup procedures developed for clenbuterolare really effective. This study considers the ion-exchangemechanism on cyanopropyl (CN), sulfonic cation exchange (SCX),mixed phase (MPH) (C8 + SCX), and nonendcapped C18 (C18NE)solid-phase extraction (SPE) columns. Results indicate thatcompound A (by contrast with clenbuterol) is not efficientlyretained on the CN, SCX, and MPH SPE columns (recovery< 10%). This finding thus leads to the development of a gaschromatography–tandem mass spectrometry procedure based onC18NE SPE that is able to purify both agonists from bovine liversspiked at 0.5, 1.0, and 2.0 ppb with a mean recovery of 93% forclenbuterol and 92% for compound A.

Introduction

The development of a multiresidue analytical strategy tosurvey the abuse of adrenergic agonist drugs used as growthpromoters in animal production is mainly based on the selec-tivity of the cleanup procedures and the specificity of the detec-tion systems (1,2). Such requirements have been fundamentalsteps for the analytical toxicology investigations related tohuman poisonings described after the ingestion of clenbuterol-contaminated bovine liver and meat (3,4). Most of the analyticalprocedures reported as effective include an ion-exchange step in

the multiresidue cleanup of beta agonists. This procedureinvolves the interaction between the secondary amino groupshared among all beta agonists and the cyanopropyl (CN) (5)and sulfonic cation exchange (SCX) (6,7) functional groups ofsolid-phase extraction (SPE) columns. The recent characteri-zation of a new clenbuterol-like drug, N1-(2-(4-amino-3,5-dichlorophenyl)-2-hydroxyethyl)-N1-isopropyl-propanamide (8)(known as compound A), with an amide substitution of thenitrogen atom on the alkylic chain (Figure 1) prompted ourgroup to verify if the cleanup procedures most in use werereally selective for such a molecule and explore any possiblealternatives. The aim was to limit as much as possible falsenegative results that could compromise the reliability of theresults of forensic investigations, thus exposing consumers tothe previously mentioned toxicological risk.

Experimental

EquipmentCleanup procedures were performed on a Supelco (Milan,

Italy) vacuum manifold device. A high-performance liquid chro-matography (HPLC) System Gold, a Model 126 pump, a Model168 diode-array detector (DAD), and a Model 501 autosampler(Beckmann Analytical, S. Ramon, CA) were used to assess theperformance of standards on different SPE columns. Chro-matographic conditions consisted of a reversed-phase (RP) C18Lichrosphere Select B column (250 × 4 mm, 5 µm) (Merck,Darmstadt, Germany), a mobile phase of 0.01M sodium acetate(pH 3.0) (A) and acetonitrile (B), and a linear gradient from 10%to 100% B in 20 min. The flow rate was 1.0 mL/min, the DADwas set at 245 and 305 nm, the bandwidth was 4 nm, and thespectra recorded in the range of 220 to 350 nm.

An analytical performance on spiked livers at residue levels

Abstract

Maurizio Fiori, Claudia Cartoni, Beatrice Bocca, and Gianfranco Brambilla*Istituto Superiore di Sanità—Laboratorio di Medicina Veterinaria, viale Regina Elena 299, I-00161 Rome, Italy

Reproduction (photocopying) of editorial content of this journal is prohibited without publisher’s permission.92


* Author to whom correspondence should be addressed.

The Use of Nonendcapped C18 Columns in the Cleanupof Clenbuterol and a New Adrenergic Agonist fromBovine Liver by Gas Chromatography–TandemMass Spectrometry Analysis

93


of 0.5, 1.0, and 2.0 ng/g were carried out on a GCQ ion trapdetector (ThermoQuest Italia, Milan, Italy) with a CP-SIL 8CB-MS FS 30X.25(.25) capillary column (Chrompack Italia, Milan,Italy). The injector temperature was set at 250°C and was insplitless mode. The constant velocity of the carrier gas (He) was40 cm/s. The oven temperature program raised from 70°C to230°C in 11 min (20°C/min), then raised to 280°C in 10 min(5°C/min), and was held for 5 min at 280°C. The GCQ acquisi-tion was in electron-impact mode (70 eV), the multiplier wasset at 1300 V, and the resolution 0.5 amu. For clenbuterol theprecursor ion was m/z 262; the width was 4; the excitation volts1.1; and the product ions were m/z 188–192, 225–229, and262–264. For compound A the precursor ion was m/z 262; thewidth was 4; the excitation volts 1.2; and the product ionswere m/z 73–77, 188–192, and 262–264.

Liver extraction was performed by an ultraturrax apparatusand a rotary evaporator (Buchii, Zurich, Switzerland).

Evaporation under a nitrogen stream and trimethylsilylderivatization of the extracts were carried out on a Heat Block(Pierce Italia, Milan, Italy).

MaterialsClenbuterol HCl (Sigma, Milan, Italy) and compound A

(courtesy of Prof. G. Boatto, Dept. of Pharmacology Universityof Sassari, Sassari, Italy) were used as pure standards. Meth-anol, acetonitrile, and all other reagents and solvents were ofanalytical grade. Bakerbond CN propyl SPE (100 mg) (CN), aro-matic sulfonic acid (100 mg) (SCX), light-load (500 mg) C18nonendcapped (C18NE) (J.T. Baker Italia, Milan, Italy), BondElut certify columns (200 mg) (mixed phase, MPH) (VarianItalia, Milan, Italy), and Extrelut 20 columns (Merck) werealso used.

A standards phosphate buffer (PBS) (0.1M, pH 6.0) and 100%and 20% methanol stock solutions were prepared at 1 mg/mLand stored at +4°C. Working solutions were freshly obtained bydilutions in the appropriate solvents in the range of 100.0 to0.1 µg/mL.

SamplesSeven different incurred beef livers (previously tested as

negative for beta agonists) were spiked at levels of 0.5, 1.0,and 2.0 ng/g each by the addition of 25, 50, and 100 µL of the0.1-µg/mL PBS working solutions, respectively. Such spikinglevels have been chosen according to the pharmacokineticsdata of clenbuterol in calves (9). Analyses were repeated onthree different sessions, and recoveries and reproducibilitywere calculated according to Gowick et al. (10).

Analytical procedureDifferent protocols were followed depending on the SPE

sorbent tested. From fractions of 1 mL from the applications,washing and elution steps were collected separately, brought todryness, and resuspended in a 200-µL HPLC mobile phase toassess recovery. For each SPE procedure considered, repro-ducibility was assessed on 12 replicates on two different days.

The CN procedure was done according to Musch and Massart(5). Columns were conditioned with 2 mL of MeOH followed by1 mL H2O, not allowing the column to dry. Then, 1 mL of20% of a 100-µg/mL MeOH standard working solution wasapplied at a flow rate of 0.5 mL/min. The column was washedwith 3 mL of MeOH and eluted by another 3 mL of MeOH,which had 1% triethylamine (TEA) as the counter ion.

The SCX procedure was as follows. Columns were condi-tioned with 2 mL MeOH and 1 mL 0.1M acetic acid. Then, 1 mLof the standard working solution (20% MeOH) was applicatedat a flow rate of 0.5 mL/min. After washing with 3 mL MeOH,elution was performed with 3 mL MeOH (1% TEA).

The MPH columns were performed as follows. According tothe procedure described by Montrade et al. (6), columns wereconditioned with 2 mL MeOH and 0.1M PBS (pH 6.0). Appli-cation of the standard working solution (100 µg/mL) in PBS(0.1M, pH 6.0) was performed at a flow rate of 0.5 mL/min.Rinsing with 1 mL 1.0M acetic acid, washing with 3 mL MeOH,

Figure 1. Structures of compound A and clenbuterol.

Table I. Recovery Expressed as the Percentage of Compound A and Clenbuterol* on CN, SCX, MPH, and C18NE SPEColumns†

CN SCX MPH C18NE

Fraction Compound A Clenbuterol Compound A Clenbuterol Compound A Clenbuterol Compound A Clenbuterol

Application 22 ± 3 n.d.‡ 78 ± 4 n.d. n.d. n.d. n.d. n.d.First washing 64 ± 5 2 ± 2 22 ± 3 n.d. 56 ± 81 2 ± 1 n.d. n.d.Second washing 13 ± 3 n.d. n.d. n.d. 27 ± 5 n.d. n.d. n.d.Third washing 3 ± 4 n.d. n.d. n.d. 12 ± 3 n.d. 18 ± 5 n.d.First elution n.d. 71 ± 4 n.d. 82 ± 3 n.d. 78 ± 4 29 ± 6 66 ± 3Second elution n.d. 28 ± 3 n.d. 24 ± 1 n.d. 21 ± 2 49 ± 3 29 ± 2Third elution n.d. n.d. n.d. n.d. n.d. n.d. 4 ± 2 1 ± 2

* Mean ± standard deviation, n = 12.† 100 µg of each compound loaded.‡ n.d., not determined.


94

and eluting by 3 mL MeOH (1% TEA) was then performed.C18NE columns were conditioned with 3 mL MeOH and

1 mL H2O. Application of the working standard solution (20%MeOH) involved washing with 3 mL of 50% MeOH and elutionwith 4 mL MeOH (1% TEA).

Blank and spiked liver samplesFrom each liver 5 g was sampled with the addition of 200 µL

of the appropriate internal standard. A 0.1-µg/mL working

solution in PBS (0.1M, pH 6.0) (clenbuterol for compound Aanalysis and vice-versa) corresponded with 4 ng/g. The liversamples were minced in 50-mL polypropylene tubes by ultra-turrax in 20 mL HCl (0.5 N). After sonication (RT = 15 min),the samples were allowed to hydrolyze overnight at room tem-perature under shaking. Supernatants were recovered by cen-trifugation (3000 g, 30 min, and 4°C), thus removing the uppersolid lipid layer. After the addition of 200 µL NaOH (10 N)under vortexing, the liquid was adsorbed on Extrelut 20

Table II. Recovery Study* by GC–MS–MS from 20 Livers Spiked at Residue Levels of 0.5, 1.0, and 2.0 ng/g

Retention Ions Spiked level MeanDrug time (m/z) (ppb) N recovery CVR CC alpha CC beta

Clenbuterol 14.37 262 0.5 21 93.9 4.2 0.55 0.58225 1.0 21 93.0 1.6188 2.0 21 94.1 1.8

Compound A 20.47 262 0.5 21 93.2 3.1 0.53 0.56188 1.0 21 93.4 2.473 2.0 21 91.4 2.5

* Mean ± standard deviation.

Figure 2. Elution profile of clenbuterol (RT = 10.73) and compound A (RT = 15.69) from RP-HPLC–DAD analysis with a linear gradient of 10% to 100%acetonitrile in 20 min.


95

columns (1), and the elution was performed by a 60-mL mix-ture of n-hexane–dimethyl chloride (8:2, v/v). Organic phasecollected in a 100-mL glass round bottom flask was evapo-rated to dryness. Residue resuspended in 1 mL 20% MeOHwas loaded on C18 SPE columns, according to a 2-mL washing(50% MeOH) and 4-mL MeOH (1% TEA) elution. The eluatewas evaporated under a nitrogen stream on the heater blockand derivatized with 20 µL N,O-bis(trimethylsilyl)trifluoro-acetamide (1% trimethylchlorosilane, 60 min, 60°C) (1). Weinjected 2 µL in the gas chromatography (GC)–tandem massspectrometry (MS–MS) system. Calibration curves for eachanalytical session (in the range of 0.20 to 4.00 ng injected) werebuilt for each analyte.

CalculationsFor HPLC analysis, calibration curves were built by plotting

the peak area of the analyte versus its nominal concentration.For the GC–MS–MS analysis validation study, we consideredthe sum of the area of the precursor ion and two product ionsboth for the analyte and the internal standard (clenbuterol forcompound A analysis and vice-versa).The peak-area ratio (ana-lyte–internal standard) was plotted against the nominal con-centration of the analyte. The calculation of the decision limitsfor CC alpha (the smallest content of the analyte in liver thatmay be confirmed with 95% probability) and CC beta (thesmallest content of the analyte from which sample is trulyviolative with a confidence limit of 99%) on livers spiked atresidue levels was performed according to the statisticalapproach of Gowick et al. (10).


SPEThe elution profile expressed as the recovery rate of clen-

buterol and compound A from different SPE columns is

reported in Table I. The coefficient of regression for the cali-bration curves were r = 0.9976 for clenbuterol and r = 0.9973for compound A.

Method validationThe results of the recovery study by GC–MS–MS on livers

spiked at the residue levels are reported in Table II with methodperformances. Regression curves over three analytical sessionswere r = 0.9985 for compound A and r = 0.9987 for clen-buterol. It is worth noting that the European Commissionsuggested the acquisition of one parent ion and two productions for the unambiguous GC–MS–MS identification of theforbidden substances. For this purpose, the chromatograms ofa blank liver extract for compound A and a liver extract spikedat 1.0 ng/g (clenbuterol as the internal standard added at 4.0ng/g) are shown in Figure 2, in which the traces reported onthe figure from the top to the bottom refer to the total ion cur-rent, the precursor ion, the first product ion, the secondproduct ion, and the sum of the ions (precursor + productions), respectively.

DiscussionA comparison of the results of the recovery study on SCX and

CN columns for clenbuterol and compound A indicates theformer bases’ binding mainly on the ion-exchange interactionof the secondary amino group. The lack of such a function incompound A greatly affects the recovery, thus demonstratingthat the primary amino group shared among both compoundsis weakly active charged (its pKa is lowered) and sterically hin-dered by the two chlorinated atoms (Figure 1). The behavior ofcompound A (more retained on propyl CN columns during theapplication) can be mainly addressed to hydrophobic interac-tions (Table I). Such hydrophobic interactions that are presentin MPH columns as C8 alkylic chains are not sufficient to retaincompound A during the methanolic washing, which is a basicstep to improve the selectivity of such “cleanup” procedures.

These considerations suggest the use of stronger hydro-

Figure 3. GC–MS–MS analysis of a blank and spiked liver for compound A: (A) clenbuterol used as the internal standard spiked at 4.0 ng/g, (B) the blank liverextract for compound A, and (C) a liver extract spiked with compound A at 1.0 ng/g.

A B C


96

phobic interactions, coupled with the presence of an ion-exchange mechanism for clenbuterol (represented by the freesilanols groups in the case of C18NE columns). In order toallow the electrostatic interactions of clenbuterol, the flowrate in the application was reduced to 0.5 mL/min. The infor-mation derived from the RP-HPLC analysis of such drugsshowed that compound A was eluted only in the presence of100% acetonitrile (Figure 3). This behavior has been con-served on C18NE SPE columns that allow up to 2 mL 50%MeOH washing without any appreciable loss of compound A.

On this basis, an appropriate SPE procedure was developedon liver extracts. In order to limit as much as possible thepresence of interfering substances that could act as counterions, a preliminary extraction step on Extrelut columns atalkaline pH was carried out, which proved to be effective toconcentrate clenbuterol and anilino-like compounds in theorganic phase. The limited losses in the recoveries reported inTable II referred to the overall procedure and can be reasonablyaddressed to the sample handling.

Conclusion

The results encourage the use of C18NE SPE columns,which are able to work at the same time with RP and ion-exchange mechanisms and give satisfactory recoveries for twocompounds with quite different chromatographic behavior.The pharmacokinetics evidence that clenbuterol could persistin bovine liver at residue levels above 1 ng/g for at least 400 hfrom the withdrawal of a growth promoting treatment (9) sug-gests that this approach could represent a realistic tool forthe survey of the illegal use of such a drug and its related sub-stances in animal productions, with an irrelevant probability ofhaving false positive (CC alpha) or false negative (CC beta)results.

Acknowledgments

We would like to thank Prof. Gianpiero Boatto (Departmentof Pharmacology, University of Sassari) for the characterization

of compound A and Mr. Giovanni Bartolini (Istituto Superioredi Sanità) for the technical help. This work was supported byMinistero della Sanità, Project No. 98/JG.

References

1. The Use of Immunoaffinity Columns in Multi-Residue and Con-firmation Analysis of Beta Agonists in Biological Samples. L.A. vanGinkel, H.J. van Rossum, and R.W. Stephany, Eds. Commission ofthe European Communities, Bilthoven, The Netherlands, April22–26, 1991.

2. A. Polettini, M.C. Ricossa, A. Groppi, and M. Montagna. Deter-mination of clenbuterol in urine as its cyclic boronated derivativeby gas chromatography-mass spectrometry. J. Chromatogr. B 564:529–35 (1991).

3. G. Brambilla, T. Cenci, F. Franconi, R. Galarini, A. Macrì, F. Ron-doni, M. Strozzi, and A. Loizzo. Clinical and pharmacologicalprofile in a clenbuterol epidemic poisoning of contaminated beefmeat in Italy. Toxicol. Lett. 114: 47–53 (2000).

4. Centro Nacional de Epidemiología. Intoxicación alimentaria rela-cionada con Clenbuterol. España 1993–1994. Bol. Epidem.Microbiol. 1(12): 229–31 (1993).

5. G. Musch and D.L. Massart. Isolation of basic drugs from plasmausing solid-phase extraction with a cyanopropyl-bonded phase.J. Chromatogr. 432: 209–22 (1988).

6. M.P. Montrade, B. Le Bizec, F. Monteau, B. Siliart, and F. Andrè.Multi-residue analysis of beta agonist drugs in urine of meat pro-ducing animals by gas chromatography-mass spectrometry. Anal.Chim. Acta 275: 253–68 (1993).

7. B.F. Spisso, C.C. Lopez, M.A.S. Marques, and F.R.A. Neto. Deter-mination of beta 2 agonists in bovine urine: comparison of twoextraction/clean up procedures for high-resolution gas chro-matography-mass spectrometry analysis. J. Anal. Toxicol. 24:146–52 (2000).

8. B. Neri, R. Cozzani, M. Di Pietrogiacomo, G. Brambilla, M. Fiori,C. Testa, and G. Boatto. HRGC-MS EI and CI for the identificationand characterisation of a new clenbuterol-like substance. Adv.Mass Spectrom. 15: 587–88 (2001).

9. M.J. Sauer, R.J.H. Pickett, S. Limer, and S.N. Dixon. Distributionand elimination of clenbuterol in tissues and fluids of calves fol-lowing prolonged oral administration at growth-promoting dose.J. Vet. Pharmacol. Therap. 18: 81–86 (1995).

10. P. Gowik, B. Julicher, and S. Uhlig. Multi-residue method fornon-steroidal anti-inflammatory drugs in plasma using high-per-formance liquid chromatography-photodiode-array detection.Method description and comprehensive in-house validation.J. Chromatogr. B Biomed. Sci. Appl. 716(1-2): 221–32 (1998).


97

A rapid, precise, and specific high-performance liquidchromatographic method is described for the simultaneousdetermination of paracetamol, phenylephrine HCl, andchlorpheniramine maleate in combined pharmaceutical dosageforms. The method involves the use of a µBondapak CN RPanalytical column (125 Å, 10 µm, 3.9 × 150 mm) at 22°C as thestationary phase with the mixture of acetonitrile and phosphatebuffer (pH 6.22, 78:22) as the mobile phase. Derivatization of thedrugs is not required. The method is applied to commercialpediatric cough–cold syrups, tablets, and capsules marketed inTurkey. The relative standard deviation for 10 replicatemeasurements of each drug in the medicaments is always less than 2%.

Introduction

Combinations of decongestant, antihistaminic, and analgesicpreparations are widely used for cough and cold treatment.

Several methods have been described for the quantitative deter-mination of these drugs. High-performance liquid chromatog-raphy (HPLC) methods have been investigated by many workers.Most of them were based on ion-pair formation, and the detectionmethods were typically based on measuring the UV absorbance ofthe analytes (1–9).

The methods given in the literature were applied to these activeingredients, but the methods could only determine two compo-nents simultaneously. In the given methods, paracetamol andphenylephrine HCl give peaks with the same retention times.Other analytical techniques such as derivative spectrophotometry(2), high-performance thin-layer chromatography (10), liquidchromatography (LC)–mass spectrometry (11), gas–liquid chro-

matography (12), and spectrofluorometric (13) methods havealso been reported, but none of the methods are applicable for thesimultaneous determination of three components with a largeexcess of paracetamol content.

In USP 24, the determination of these components has alsobeen performed with HPLC, but all of them were determined sep-arately and the method does not involve simultaneous determi-nation (14).

In this study, we propose to employ UV detection to determineactive ingredients in cough–cold syrups, tablets, and capsulesafter HPLC separation.

The advantages of the proposed method are that the methodworks well for all cold drugs with UV absorptivity, the detectorresponse for all drugs are similar, they are easily applicable inlarge excess amounts of paracetamol drugs without any fittinginto one another, and they have a very short analysis time ofapproximately 4 min.

Experimental

ApparatusThe system consisted of a Hewlett Packard (Waldborn,

Germany) Series 1100 LC including an HP UV–vis detector,vacuum degasser, gradient pump module, auto injector with avariable injection valve, and column compartment oven. AµBondapak CN RP analytical column from Waters (Milford, MA)(125 Å, 10 µm, 3.9 × 150 mm) was used. Instrumental settingswere a flow rate of 1.5 mL/min, a column temperature at 22°C,and a detector wavelength of 265 nm.

Materials and reagentsAll the drugs were of USP quality. Methanol and acetonitrile

were obtained from J.T. Baker (Griesheim, Germany) in HPLCgradient grade. Orthophosphoric acid and triethylamine wereobtained from Merck Inc. (Darmstadt, Germany). The water used

Abstract

Simultaneous High-Performance LiquidChromatographic Determination of Paracetamol,Phenylephrine HCl, and Chlorpheniramine Maleate inPharmaceutical Dosage Forms

Hamide çenyuva* and Tuncel ÖzdenInstrumental Analysis Centre, Scientific and Technical Research Council of Turkey, 06530, Ankara, Turkey



* Author to whom correspondence should be addressed: [email protected].

S


98

was distilled and deionized by using a Millipore (Vienna, Austria)Milli-Q ultrapure system. Other chemicals were of analytical orHPLC grade.

Mobile phaseThe mobile phase consisted of an aqueous solution of phos-

phate buffer (pH 6.22) and acetonitrile (22:78, v/v). The phosphatebuffer was prepared by dissolving 1.36 mL orthophosphoric acidin 1 L water. Triethylamine was added to the phosphate buffersolution in order to adjust the pH to 6.22. Acetonitrile and waterwere previously filtered under vacuum through 0.45-µm nylonfilters before injection into the HPLC apparatus.

Standard stock solutionsStandard solutions were prepared by dissolving the drugs in

methanol and diluting them to the desired concentrations.

ParacetamolA 320-mg sample of paracetamol was accurately weighed and

dissolved with methanol up to volume in a 10-mL volumetricflask. A 31.2-µL volume of this solution was again diluted withmethanol to volume in a 10-mL volumetric flask.

Phenylephrine HClA 10-mg sample of phenylephrine HCl was accurately weighed

and dissolved with methanol up to volume in a 10-mL volumetricflask. A 1000-µL volume of this solution was again diluted withmethanol to volume in a 10-mL volumetric flask.

Chlorpheniramine maleateA 10-mg sample of chlorpheniramine maleate was accurately

weighed and dissolved with methanol up to volume in a 50-mLvolumetric flask.

Standard mixture solutionA standard mixture solution was prepared from these stock

solutions by mixing 2000 µL of a paracetamol standard solution,152 µL of a phenylephrine HCl standard solution, 14.8 µL of achlorpheniramine maleate standard solution, and 1232 µLmethanol.

Sample preparationsSyrup samples

The syrup solution was homogenized by shaking and dilutedwith methanol to give a final concentration of 40 to 120 µg forparacetamol, 1 to 4 µg for phenylephrine HCl, and 0.3 to 1 µg for

chlorpheniramine maleate in 1 mL.

Tablet and capsulesTwenty tablets or capsule contents were

weighed, their mean weight determined, and theywere finely powdered. An equivalent weight of thetablet or capsule content was transferred into a10-mL volumetric flask containing 6 mLmethanol, ultrasonicated for 20 min, and dilutedto 10 mL with methanol. The solution was filteredthrough a 0.45-µm nylon filter.

This solution was again diluted with methanolto give a final concentration mentioned in thesyrup samples.


Calibration and linearityAn external standard method was used for quan-

titative determinations. Triplicate 1-, 3-, 6-, 8-, 9-,10-, and 15-µL injections were made for the stan-dard mixture solution. The retention times of thestandards were 1.13 min for paracetamol, 2.13min for phenylephrine HCl, and 3.44 min forchlorpheniramine maleate. A typical HPLC chro-matogram of the standard mixture is shown inFigure 1. The calibration graphs were obtained byplotting the peak area against the concentration ofthe drugs. In the simultaneous determination, thecalibration graphs were found to be linear in thementioned concentrations (the correlation coeffi-cients are shown in Table I).

Precision (reproducibility)The precision of the method was studied by

Figure 1. A typical HPLC chromatogram of the standards: paracetamol (1.13 min), phenylephrine HCl(2.13 min), and chlorpheniramine maleate (3.44 min).

Figure 2. A typical HPLC chromatogram of pediatric cough–cold syrup: paracetamol (1.13 min),phenylephrine HCl (2.13 min), and chlorpheniramine maleate (3.44 min).


99

determining the concentrations of each drug in a syrup, capsule,and tablet ten times. The results of the precision study (shown inTable I) indicate that the method is reliable (relative standarddeviation percentage < 2).

Recovery testsRecovery tests were performed by adding a known amount of

each drug to a cough–cold syrup and tablet where it was knownto be absent.

The mean results of five analyses ranged from 97.10 to 99.53(Table II), and these can be considered to be good recoveries.

Determination of the limit of detection and quantitationThe limit of detection (LOD) was defined as the concentration

of phenylephrine HCl and chlorpheniramine maleate (calculatedas 0.0325 µg/mL and 0.0279 µg/mL, respectively) that produceanalytical signals equal to thrice the deviation of the backgroundsignals. The limit of quantitation (LOQ) was the lowest levels ofphenylephrine HCl and chlorpheniramine maleate (determinedto be 0.251 µg/mL and 0.184 µg/mL, respectively) in the simulta-neous quantitative assay. The relative standard deviation per-centage results of the LOQ studies were 1.29 for phenylephrineHCl and 2.51 for chlorpheniramine maleate (n = 10).

SelectivitySelectivity was assessed by a quality control of the chro-

matograms obtained from samples and placebo. Possible interfer-ences resulting from substances present in the medicamentswere not observed.

Determination of active ingredients in pharmaceutical dosage forms

The contents of three drugs in ten different pediatriccough–cold syrups, capsules (Coldeks), and tablets for each brand(Dristan and Deflu) were determined by the proposed method,and the results are presented in Table III.

The chromatogram of a pediatric cough–cold syrup is shown inFigure 2.

Conclusion

The concentration of phenylephrine HCl, chlorpheniraminemaleate, and a large excess of paracetamol in pharmaceuticalsamples can be satisfactorily determined using HPLC with UVdetector. This study has shown that UV detection is a sensitive,reliable, reproducible, and accurate method for the determina-tion of the active ingredients in pediatric cough–cold syrups, cap-sules, and tablets.

The method is straightforward and simpler than the commonlyused HPLC methods involving ion pairing or derivatization. Ascan be seen in the figures, 3.5 min is enough for all of the activeingredients to be released.

This method has been found suitable for the routine analysis ofthe pharmaceutical dosage forms in quality control and R&D lab-oratories for products of similar type and composition.

References

1. O.W. Lau and C.S. Mok. High-performance liquid chromatographicdetermination of active ingredients in cough–cold syrups with indi-rect conductimetric detection. J. Chromatogr. A 693: 45–54 (1995).

Table I. Results of the Precision and Linearity Study*

Precision Coefficient of Linearity Ingredients (%RSD†) correlations (µg/mL)

Paracetamol 0.13 0.9999 25–120Phenylephrine HCl 1.95 0.9999 0.3–10Chlorpheniramine maleate 1.36 0.9999 0.2–3

* n = 10.† %RSD, relative standard deviation percentage.

Table III. Assay Results of Active Ingredients inCommercial Samples*

Label Found %Label Samples Ingredient value (mg) (mg) claim

Syrup (mg/mL) Paracetamol 32 32.100 100.30Phenylephrine HCl 1 1.000 100.00Chlorpheniramine

maleate 0.2 0.200 100.00

Dristan Paracetamol 325 325.010 100.00Phenylephrine HCl 5 5.037 100.74Chlorpheniramine

maleate 2 2.019 100.99

Deflu Paracetamol 300 300.051 100.02Phenylephrine HCl 5 5.019 100.38Chlorpheniramine

maleate 2 2.010 100.50

Coldeks Paracetamol 325 325.015 100.01Phenylephrine HCl 5 5.012 100.24Chlorpheniramine

maleate 1 1.006 100.60

* Average of 10 analyses.

Table II. Results of the Recovery Tests for the DrugsUnder Study by the Proposed Method*

Amount Matrix Ingredient added % Recovery†

Syrup (mg/mL) Paracetamol 30.0 99.73 (0.18)Phenylephrine HCl 1.0 98.40 (1.55)Chlorpheniramine maleate 0.2 98.00 (4.55)

Tablet (mg) Paracetamol 325.0 99.99 (0.04)Phenylephrine HCl 5.0 100.56 (1.56)Chlorpheniramine maleate 2.0 99.60 (0.65)

* n = 5.† Relative standard deviation shown in parentheses.


100

2. N. Erk and M. Kartal. Simultaneous high performance liquid chro-matographic and derivative ratio spectra spectrometry determinationof chlorpheniramine maleate and phenylephrine hydrochloride. Il Farmaco. 53: 617–22 (1998).

3. R.N. Raju, P. Srilakshimi, and U.M. Krishina. Simultaneous determi-nation of chlorpheniramine maleate, ephedrine hydrochloride andcodeine phosphate in cough syrups by reverse-phase high-perfor-mance liquid chromatography. Indian Drugs 29: 408–11 (1992).

4. G. Liang, H. Qian, L. Zhu, Y. Wang, and L. Sun. Simultaneous deter-mination of paracetamol and chlorpheniramine by RP-HPLC.Zhongguo Yaoxue Zazhi. 29: 46–48 (1994).

5. I.L. Honigberg, J.T. Stewart, and A.P. Smith. Liquid chromatographyin pharmaceutical analysis: determination of cough–cold mixtures. J. Pharm. Sci. 63: 766–69 (1974).

6. J.O. De Beer, C.V. Vandenbroucke, and D.L. Massart. Experimentaldesign for the rapid selection of separation conditions for methyl andpropyl parahydroxybenzoate, phenylephrine hydrochloride andchlorpheniramine maleate by ion-pair liquid chromatography. J. Pharm. Biomed. Anal. 12: 1379–96 (1994).

7. B.R. Thomas, X.G. Fang, P. Shen, and S. Ghodbane. Mixed ion pairliquid chromatography method for the simultaneous assay ofascorbic acid, caffeine, chlorpheniramine maleate, dextromethor-phan HBr monohydrate and paracetamol in Frenadol sachets. J. Pharm. Biomed. Anal. 12: 85–90 (1990).

8. A.I. Gasco-Lopez, R. Izquierdo-Hornillos, and A. Jiminez.Development and validation of high performance liquid chromatog-

raphy method for the determination of cold relief ingredients inchewing gum. J. Chromatogr. A 775: 179–85 (1997).

9. G. Indrayanto, A. Sunarto, and Y. Adriani. Simultaneous determina-tion of phenylpropanolamine hydrochloride, caffeine, paracetamol,glycerylguaiacolate and chlorpheniramine maleate in Silabat tabletusing HPLC with diode array detection. J. Pharm. Biomed. Anal. 13:1555–59 (1995).

10. M. El-Sadek, A. El-Shanawany, A. Aboul-Khier, and G. Ruecker.Determination of the components of analgesic mixtures using high-performance thin layer chromatography. Analyst 115: 1181–84(1990).

11. C. Celma, J.A. Allue, J. Prunonosa, C. Peraire, and R. Obach.Simultaneous determination of paracetamol and chlorpheniraminemaleate in human plasma by liquid chromatography–tandem massspectrometry. J. Chromatogr. A 870: 77–86 (2000).

12. O.W. Lau and Y.M. Cheung. Simultaneous determination of someactive ingredients in cough–cold syrups by gas–liquid chromatog-raphy. Analyst 115: 1349–53 (1990).

13. J.A. Arancibia, A.J. Nepote, G.M. Escandar, and A.C. Olivieri.Spectrofluorimetric determination of phenylephrine in the presenceof large excess of paracetamol. Anal. Chim. Acta. 419: 159–68(2000).

14. The United States Pharmacopoeia 24. The National Formulary 19,U.S. Pharmacopeial Converntion Inc., Rockville, MD, 2000.


101

A full experimental design at two levels is applied for the estimationof the significance of select factors that may influence the ionchromatography (IC) determination of F–, Cl–, Br–, NO3

–, SO4–2, and

PO4–3 in serum samples. The factors studied are various sample

deproteinization procedures, eluent composition, and flow rates.Deproteinization using either acetonitrile–NaOH or ultrafiltrationcan be used in order to obtain a significant protein removal beforeIC analysis; however, the former is recommended because it is lesstime-consuming and cheaper. Better resolution is obtained when a sodium hydroxide solution is used as the eluent. There is noinfluence of the sample’s deproteinization procedures on thechromatographic resolution.

Introduction

The ability to determine the concentration of physiologicallyand pathologically related anions with a single sample treatmentand technique is of great importance. Ion chromatography hasbecome one of the most powerful tools for the quantitative anal-ysis of anions in a wide variety of matrices. One of the problemsassociated with serum anion determination by this technique isthe lack of similar sensitivity for all the anions to be determinedsimultaneously. Of all the inorganic anions present in humanserum, chloride, bicarbonate, and phosphate are the most rou-tinely determined. However, the demand for sulfate, bromide,nitrate, and fluoride serum analysis is increasing because of theclinical information derived from the levels in serum. The impor-tance in human physiology of the inorganic anions sulfate (1),bromide (2–4), nitrate (5,6), and chloride (7) have been described.

The biochemical analysis of metabolites in whole blood andserum requires initial deproteinization to remove hemoglobinand other proteins that may interfere with the assays. Different

agents have been used. One of them is perchloric acid (8,9); how-ever, it has the disadvantage that percholorate inhibits someenzyme reactions considerably and it has to be removed by pre-cipitation with potassium hydroxide. A simple deproteinizationprocedure has been reported by Khan et al. (10) involving the useof sulfosalycylic acid (SSA). However, the SSA agent is often con-taminated with sulfate. Tricholoroacetic acid has been reported(11) as a time-consuming procedure and interferes with theanion elution profile. The precipitation method based on acetoni-trile (ACN) (12) has not been successful because proteins are notquantitatively precipitated by this method, which limits the life-time of the separator column to a maximum of two months.Centrifugation or ultracentrifugation has been reported (11) asan efficient method for the deproteinization of serum samples.

A number of studies have reported the determination of someinorganic anions in serum by ion chromatography. These haveincluded the simultaneous determination of inorganic phos-phate, bromide, nitrate, and sulfate in human serum (1) and thedetermination of thiocyanate (13), bromide (14), and sulfate (15).In measuring systems in which the signal is linearly related to thecomponent of interest, matrix components, and instrumentalparameters, factorial analysis is a convenient tool to apply.Factorial approaches to experimental designs contrast with sim-plex approaches in that several experiments can be performedsimultaneously and are used to calculate the main effects and theinteraction effects of several factors. Full factorial designs at twolevels of variation for the input factors are often used in analyticalchemistry (16,17).

This research deals with the optimization of a method for thesimultaneous analysis of chloride, fluoride, nitrate, bromide, sul-fate, and phosphate in human serum by isocratic ion chromatog-raphy. A factorial design at two levels was applied in order toestimate the magnitude of the main effects and various two-factorinteractions under the experimental conditions of interest. It fur-ther allowed for a better interpretation of the results obtained.The evaluation of parameter significance is a very important stepin the optimization procedure. The selection criterion forchoosing the factors involved in this design was dictated by

Abstract

Evaluation of Select Variables in the IonChromatographic Determination of F –, Cl–, Br–, NO3

–, SO4–2, and PO4

–3 in Serum Samples

Z. Benzo1,*, A. Escalona3, J. Salas1, C. Gómez1, M. Quintal1, E. Marcano1, F. Ruiz1, A. Garaboto1, and F. Bartoli21Centro de Química and 2Dpto. de Biologia estructural, Instituto Venezolano de Investigaciones Científicas, IVIC, Apdo. Postal 21827,Caracas 1020-A, Venezuela and 3Centro de Química Analítica, Facultad de Ciencias, Universidad Central de Venezuela, Apdo. Postal 47102,Caracas 1041-A, Venezuela





102

variables that may influence the resolution such as deproteiniza-tion treatment, eluent composition, and flow rate.

Experimental

Apparatus and reagentsThe ion chromatographic equipment used was a Dionex

(Sunnyvale, CA) DX 500 system with a 100-µL injection loop, anIon Pac AS11 analytical column (4 × 250 mm), and an Ion PacAG11 Guard column (4 × 50 mm) (Dionex). The column temper-ature was 25°C. The apparatus was equipped with an ion fibersuppressor ASRS-11-4 mm (Dionex Anion self regenerating) thatwas continuously regenerated by water and a conductimetricdetector (Dionex) PeakNet Chromatography Workstation.

All chemicals were of analytical-reagent grade. Sodiumhydroxide was obtained from Merck (Darmstadt, Germany), andNa2CO3, NaHCO3, NaCl, NaNO3, NaF, Na2SO4, Na3PO4, and KBr (> 99.9% pure) were from Aldrich (St. Louis, MO). Milli-Q deion-ized water was used throughout (Milli-Q water purificationsystem, Millipore, MA) as well as HPLC-grade ACN (EM Science,NJ). Ultrafree-Cl low binding cellulose 10,000 nominal molecularweight limit filters were used (Millipore) as well as 0.45- and 0.47-µm nylon filters (Alltech, Deerfield, IL).

A protein assay was performed with the Coomassie Plus Proteinassay reagent kit (Pierce, Rockford, IL).

Serum samplesA pooled serum sample was obtained from the Medical Center

at our research institute. Each deproteinization method was per-formed on an aliquot of this pooled sample.

Sample treatmentA preliminary study was carried out in order to assess the effi-

ciency of protein removal by different methods. For this reason,three deproteinization treatments that have been reported in theliterature were tested: ACN (18,19), ACN–NaOH (20), and ultrafil-tration (19). Protein quantitation was carried out on an aliquot ofthe same serum sample after each deproteinization treatment,measuring the absorbance at 595 nm.

The ACN treatment consisted of mixing 0.2 mL of serumsample with an equal volume of ACN, then centrifuging at 2000 × g for 5 min.

For the ACN–NaOH treatment, 500 µL of a serum sample, 50 µL of NaOH (2M), and 150 µL of deionized water were addedand shaken for a few seconds. Then, 1 mL of ACN was added andvortex mixed for 10 s. The resulting mixture was centrifuged for5 min at 755 × g. Finally, 1 mL of the supernatant solution wasdiluted with 5 mL of deionized water and injected into the chro-matograph.

For the ultrafiltration, a 1:10 diluted serum sample was filteredfor 30 min and injected.


Protein removalThe sample preparation for ion chromatographic analysis of

serum is much more elaborate than for other physiological fluids.

Proteins that are present in high concentration must be removedbefore the sample is injected, because they negatively affect theseparation efficiency of the ion-exchange columns.

The principal objective of this study was to find an efficient, fast,and reliable deproteinization method that yields a solution suit-able for subsequent ion chromatographic analysis. The threedeproteinization methods described in the Experimental sectionwere applied, and the resulting percentages of protein removalwere 21.5%, 93.3%, and 98.6% for the ACN, ACN–NaOH, andultracentrifugation treatments, respectively.

After the efficiency of these methods was verified, it was decidedto choose the ACN–NaOH mixture and ultrafiltration depro-teinization procedures as variables to be considered in the exper-imental design in order to observe their influence on thechromatographic resolution.

OptimizationIn order to obtain proper information on the significance of the

factors mentioned, a full 24 factorial design at two levels wasapplied. This type of design involved sixteen experiments. Fourfactors were examined in this design: a sodium hydroxide solu-tion (X1), a sodium carbonate–sodium bicarbonate mixture (X2),

Table I. Experimental Variables Considered in theApplication of the 24 Full Factorial

Natural Coded variable variable Level (+1) Level (–1)

NaOH X1 12mM 6mMNa2CO3–NaHCO3 X2 3mM,2.4mM 0Flow rate X3 1.5 mL/min 1.0 mL/minDeproteinization

procedure X4 ultrafiltration NaOH–ACN

Table II. 24 Experimental Design*

Resolution Experiment X1 X2 X3 X4 criterion

1 –1 –1 –1 –1 11.362 1 –1 –1 –1 10.623 –1 1 –1 –1 12.304 1 1 –1 –1 7.415 –1 –1 1 –1 11.036 1 –1 1 –1 10.997 –1 1 1 –1 11.508 1 1 1 –1 7.149 –1 –1 –1 1 10.0410 1 –1 –1 1 10.8411 –1 1 –1 1 12.6912 1 1 –1 1 7.5613 –1 –1 1 1 10.5214 1 –1 1 1 11.0715 –1 1 1 1 11.3216 1 1 1 1 9.26

* Resolution taken as the response.


103

the flow rate (X3), and a deproteinization procedure (X4) (Table I).Two levels were chosen associated with the –1 and +1 levels of thecorresponding coded variables. These were selected on the basisof the literature review on this topic, and the deproteinizationprocedures were selected according to the results obtained anddescribed in the previous section.

The selected experimental response was a criteria of peak reso-lution (Rs) and the peak area. This criterion of resolution con-sisted in the addition of the resolution values of all consecutivepeaks plus the number of peaks that appears in the chro-matogram (21). Table II shows the results.

The significance was evaluated through the application of the

student t-test with a 0.05 significance level. Table III shows theseresults. The variables of X1 at the –1 level (6mM), X2 at the –1level, and the interaction of X1 and X2 resulted as being significantat the 95% level. The influence of X1 revealed that a better reso-lution was obtained when this sodium hydroxide concentrationwas used. The use of the Na2CO3–NaHCO3 mixture did not influ-ence the resolution, and this agreed with the result of the signifi-cance test. The X1–X2 interaction revealed that an improvementon the resolution was obtained when both mixtures were used.

IdentificationFigure 1 shows the chromatograms of an aqueous solution run

at an eluent composition of 6 and 12mM NaOH (Figures 1A and1B) and a serum sample using isocratic conditions (Figures 1Cand 1D). Peak identification was based on retention times.

As it can be seen from these chromatograms, in general, theaqueous standard and the serum samples exhibited a well-definedresolution and symmetrical peaks (not broadened) in less than 16min. Shorter retention times, mainly for the monovalent anddivalent anions, were observed when changing the concentrationof the eluent mixture (12mM). Furthermore, higher signals wereobtained at this condition. Observed in the chromatogram of theserum sample was the appearance of a second peak (probablyacetate), which was coeluted with fluoride.

Figures 1C and 2 show the effect on the resolution from thedeproteinization treatments. It can be observed that there was nota significant change between the results obtained. Therefore,either treatment (deproteinization by using ACN–NaOH or byultrafiltration) could be used. This result was consistent with thatobtained in the experimental design in which the deproteiniza-

Table III. Calculation of the Effects

Estimated Standard Experimental Probability Effect value deviation t-value level

b0 10.42 0.1613 64.58 0.0000b1 –1.0574 0.1613 –6.55 0.0012*b2 –0.4545 0.1538 –2.95 0.0317*b3 –0.0619 0.1613 –0.10 0.7039b4 0.1256 0.1613 0.77 0.4714b12 –1.0655 0.1538 –6.92 0.0012*b13 0.2638 0.1945 1.35 0.2331b14 0.1955 0.1613 1.21 0.2797b23 –0.0945 0.1538 –0.61 0.5659b24 0.2514 0.1538 1.63 0.1630b34 0.0621 0.1613 0.38 0.7158

* Significant effect.

Figure 1. Chromatograms of an aqueous solution run at two different eluent compositions (A and B) and a serum sample using two different isocratic conditions (C andD): (A) an aqueous standard of 6mM NaOH eluent and 1-mL/min flow rate; (B) an aqueous standard of 12mM NaOH eluent and 1-mL/min flow rate; (C) a serumsample of 6mM NaOH eluent, ACN–NaOH deproteinization, and 1-mL/min flow rate; and (D) a serum sample of 12mM NaOH eluent, ACN–NaOH deproteiniza-tion, and 1-mL/min flow rate.


104

tion treatment did not have any statistical significance. However,we recommend the ACN–NaOH mixture because it is less time-consuming (the sample can be deproteinized in approximately

5 min compared with the 30 min that it takes for the ultrafiltra-tion procedure).

Even when the flow rate was not significant, according to theresults from the significance test a flow rate of 1.5mL/min (+1 level, X3) is more convenient if anal-ysis time is considered important (as shown inFigures 1C and 3).

The significance of the b12 interaction was veri-fied by the use of the interaction diagrams (22).The b12 interaction at the –1 level revealed that theresolution was improved when the NaOH (6mM)and Na2CO3–NaHCO3 mixtures were used. Thechromatogram resulting from this experimentalcondition is shown in Figure 4. However, this con-dition is not recommended when fluoride has tobe determined, because its peak appeared withinthe water dip.

In order to further evaluate the efficiency of thedeproteinization procedures, an evaluation of thesignal sensitivity was carried out for each anion bytaking into account the peak area as a measure-ment of the response in the experimental design.The result of this study (not shown) led to the con-clusion that the treatment procedure does nothave any significant effect on the signal sensitivity.

Because of the complexity of the serum matrix,the column efficiency was checked in order to seeif any modification (probably occasioned by thematrix) could have affected it. For this reason, acontrol solution containing all the anions consid-ered in this study was passed through the columnbefore and after running the serum samples fortheir analysis (shown in Figure 5). In order toevaluate column efficiency, the theoretical plateswere calculated for each anion before and after theset of experiments involved in this research, andno significant change was observed. No change inthe signal sensitivity and resolution was obtained.Therefore, it can be said that the serum matrix didnot alter the column’s original characteristicswith it being able to analyze the inorganic anionsin this matrix under the conditions developed inthis work.

Conclusion

It has been shown in this work that the simulta-neous determination of six important physiolog-ical anions in human serum is possible using ionchromatography under isocratic conditionswithout altering the column’s original conditionsand thus its efficiency. This is an important aspect,taking into account the complexity of the serummatrix. Scarce information has been given aboutcolumn integrity after this type of analysis.

The application of the experimental design to

Figure 2. Chromatogram of a serum sample with ultrafiltration: 6mM NaOH eluent and 1-mL/min flowrate.

Figure 3. Chromatogram of a serum sample with ACN–NaOH: 6mM NaOH eluent and 1.5-mL/minflow rate.

Figure 4. Chromatogram of a serum sample with ultrafiltration: 6mM NaOH–Na2CO3–NaHCO3eluent and 1-mL/min flow rate.


105

this particular application for evaluating the significance of inputparameters for analytical determinations allows for the obtainingof important information on the variables that most influence thedetermination of the anions in this matrix and on the interactionsbetween chemical and instrumental parameters.

There was no influence of the sample’s deproteinization proce-dures on the chromatographic resolution.

Acknowledgments

The authors gratefully acknowledge the valuable support of theprogram BID-CONICIT through the project QF-10 and Prof. LuisGómez of the chromatography laboratory at the AnalyticalChemistry Center (UCV) for his valuable assistance.

This research is dedicated to our late and dearest colleagueSantos Melendez who could never fulfill his dream of workingwith us in this project.

References

1. P. De Jon and M. Burggraaf. An ion chromatographic method for thesimultaneous determination of inorganic phosphate, bromide, nitrateand sulphate in human serum. Clin. Chim. Acta. 132: 63–71 (1983).

2. Z. Khalkhali and B. Parsa. “Measurement by Non-DestructiveNeutron Activation Analysis of Bromine Concentration in theSecretions of Nursing Mothers. In Nuclear Activation Techniques inthe Life Sciences. International Atomic Energy Agency, Vienna,Austria, 1972, pp. 461–66.

3. L.-O. Plantin. In Nuclear Activation Techniques in the Life Sciences.International Atomic Energy Agency, Vienna, Austria, 1972, pp.461–66.

4. D.L. Trump and M.C. Hochberg. Bromide intoxication. JohnsHopkins Med. J. 138: 119–23 (1976).

5. M. Ferrant. Methemoglobinemia. J. Pedriatr. 29: 585–92 (1946).6. M. Cornblath, A.F. Hartmann. Methemoglob-inemia in young

infants. J. Pedriatr. 33: 421–25 (1948).7. K. Arai, F. Kusu, N. Noguchi, K. Takamura, and H. Osawa. Selective

determination of chloride and bromide ions in serum by cyclic volat-mmetry. Anal. Biochem. 210: 109–13 (1996).

8. G.A. Reichard, Jr., O.E. Owen, A.C. Haff, P. Paul, and W.M. Bortz.Ketone body production and oxidation in fastingobese humans. J. Clin. Invest. 53: 508–15 (1974).

9. A. Tizianello, G.D. Ferrari, G. Garibotto, and C. Robaudo. Amino acid metabolism and the liverin renal failure. Am. J. Clin. Nutr. 33: 1354–62(1980).

10. K. Khan, E. Blaak, and M. Elia. Quantifying inter-mediary metabolites in whole bood after a simpledeproteinization step with sulfosalicylic acid. Clin.Chem. 37: 728–33 (1991).

11. L. Politi, R. Chiaraluce, V. Consalvi, N. Cerulli, andR. Scandurra. Oxalate, phosphate and sulphatedetermination in serum and urine by ion chro-matography. Clin. Chim. Acta 184: 155–66 (1989).

12. C. Reiter, S. Muller, and T. Muller. Improvedmethod for the determination of sulphate in humanserum using ion chromatography. J. Chromatogr.413: 251–56 (1987).

13. Y. Michigami, T. Takahashi, F. He, Y. Yamamoto,and K. Ueda. Determination of thiocyanate inhuman serum by ion chromatography. Analyst 113:389–92 (1988).

14. M.E. Miller and C.J. Cappon. Anion-exchange-chromatographic determination of bromide inserum. Clin. Chem. 30: 781–83 (1984).

15. D.E.C. Cole and D.A. Landry. Determination ofinorganic sulfate in human saliva and sweat by controlled-flow anion chromatography. J. Chromatogr. 337: 267–78 (1985).

16. R.G. Brereton. Chemometrics in analytical chem-istry, a review. Analyst 112: 1635–57 (1987).

17. E. Marengo, M.C. Gennaro, and C. Abrigo.

Figure 5. Column performance (A) before and (B) after 40 injections of serum sample. Data for A andB are shown in Tables IV and V, respectively.

Table IV. Data for the Anions in Figure 5A

Anion Retention time (min) Area

Fluoride 1.82 71263Chloride 2.30 674127Nitrite 2.52 42713Bromide 3.48 91697Nitrate 3.60 294013Sulfate 4.48 1065847

Table V. Data for the Anions in Figure 5B

Anion Retention time (min) Area

Fluoride 1.82 71282Chloride 2.27 666512Nitrite 2.48 37662Bromide 3.43 107169Nitrate 3.55 284737Sulfate 4.25 1074038


106

Investigation by experimental design and regression models of theeffect of five experimental factors on ion-interaction high-perfor-mance liquid chromatographic retention. Anal. Chim. Acta 321:225–36 (1996).

18. “Ions in Physiological Fluids”. In Dionex Application Note 107.Dionex, Sunnyvale, CA.

19. R. Sakuma, T. Nishina, and M. Kitamura. Deproteinizing methodsevaluated for determination of uric acid in serum by reversed-phaseliquid chromatography with ultraviolet detection. Clin. Chem. 33:1427–30 (1987).

20. H. Itoh. Determination of trichloroacetate in human serum and urine

by ion chromatography. Analyst 114: 1637–40 (1989).21. O.N. Obrezkov, A.V. Pirogov, I.V. Pletnev, and O.A. Shpigun.

Simplex-optimization with a new criterion. Applications to dual-column ion chromatography. Mikrochim. Acta I: 293–302 (1991).

22. G.E.P. Box, W.G. Hunter, and J.S. Hunter. Statistics for Experimenters:An Introduction to Design Data Analysis and Model Building. JohnWiley and Sons, New York, NY, 1978, pp. 415–17.

Manuscript accepted September 5, 2001.

107

Subcritical water has been recently employed as the mobile phaseto eliminate the use of organic solvents in reversed-phase liquidchromatography. Although the influence of temperature onretention in subcritical water chromatography has been reported,the temperature effect on peak width and column efficiency has notyet been quantitatively studied. In this work, several polar andchlorinated compounds are separated using pure subcritical wateron Zorbax RX-C8, PRP-1 (polystyrene–divinylbenzene), HypersilODS, and ZirChrom-polybutadiene columns. Isothermal separationsare performed at temperatures ranging from 60°C to 160°C. Theretention time and peak width of analytes are reduced withincreasing temperature. However, the column efficiency is eitherimproved or almost unchanged with the increasing temperature inthe low-temperature range (lower than the 100°C to 120°C range),but it is decreased when temperature is further raised in the high-temperature range (higher than the 100°C to 120°C range).Therefore, a maximum in column efficiency is obtained attemperatures within the 100°C to 120°C range in most cases.

Introduction

Reversed-phase liquid chromatography (RPLC) is a very popular separation and analysis technique used today.Unfortunately, organic solvents are required to achieve separa-tion in RPLC. An enormous amount of these organic solvents isconsumed every day worldwide. These organic solvents areexpensive in terms of both purchasing and waste disposal. Inaddition, they are also potentially harmful to the laboratory envi-ronment and the operator. Therefore, searching for nontoxic sol-vents as the mobile phase for RPLC is of great interest.

Ambient water is too polar to serve as an eluent for reversed-phase separation. Fortunately, the polarity of water decreaseswith increasing temperature. Therefore, the solubility of organiccompounds is dramatically increased in water at elevated tem-

peratures (1–4). For example, the solubility of some pesticidesand polycyclic aromatic hydrocarbons is increased several ordersof magnitude by raising the water temperature from ambient to200°C (1–3). Thus, liquid chromatographic (LC) separations canbe achieved by using high-temperature (subcritical) water(5–11). With two additional components (an oven and a back-pressure regulator or restrictor), a conventional high-perfor-mance liquid chromatographic (HPLC)–UV system can be easilymodified to a subcritical water separation system. The oven isused to provide the temperature for subcritical water separation,and the backpressure regulator prevents water from boilingwhen working with temperatures higher than 100°C. The UVdetector is placed outside the oven. Depending on the separationtemperature and the flow rate of water used, the temperature ofthe water eluent in the UV flow cell varies, but it is lower than theoven temperature. If a backpressure regulator or a short packedLC column is used to provide the backpressure, they are nor-mally connected to the outlet of the UV flow cell. Thus, the flowcell is under pressure and may be damaged.

Most reports on subcritical water chromatography mainlyfocus on testing the feasibility of using subcritical water as themobile phase for reversed-phase separation (5–11). Even thoughthe effect of water temperature on the retention is mentioned insome of these reports (5–11), a quantitative study of the temper-ature effect on peak width and column efficiency in subcriticalwater separation has not yet been reported. It should be pointedout that the effects of temperature on retention, viscosity, diffu-sivity, and the number of plates have been well-investigated inconventional HPLC (12–17). However, the temperature rangewas generally much narrower and normally went up to 80°C. Inaddition, organic solvents were involved in the mobile phases ofthese studies (12–17).

In this work, pure water at elevated temperatures and pres-sures was used as the eluent to separate several polar analytesand chlorophenols on four commercial columns, whichincluded the Zorbax RX-C8, polymeric PRP-1, Hypersil ODS, andZirChrom-PBD columns. Separations were performed at tem-peratures ranging from 60°C to 160°C in an isothermal manner.The peak width was monitored and the number of theoreticalplates was calculated to evaluate the temperature effect oncolumn efficiency.

Abstract

Temperature Effect on Peak Width and ColumnEfficiency in Subcritical Water Chromatography

Yu Yang1,*, Lori J. Lamm1, Ping He1, and Toru Kondo2

1Department of Chemistry, East Carolina University, Greenville, NC 27858 and 2Fuji Silysia Chemical Ltd., Kasugai-Shi, Aichi-Ken, 487-0013, Japan





108

Experimental

Separation columnsA polystyrene–divinylbenzene column (PRP-1, 250- × 4.1-mm

i.d.) was purchased from Hamilton Company (Reno, NV). AZorbax RX-C8 column (250- × 4.6-mm i.d.) was obtained fromDuPont (Wilmington, DE). A Hypersil ODS column (100- × 4.6-mm i.d.) (Keystone, Bellefonte, PA) was used to separate a phenolmixture. Because the recently developed zirconia-based columnshave shown excellent thermal stability and column efficiency(18–20), a ZirChrom-polybutadiene (PBD) column (100- × 2.1-mm i.d.) (ZirChrom Separation Inc., Anoka, MN) was alsoemployed in this study. The particle size was 3 µm for theZirChrom-PBD column and 5 µm for the other three columns.

ReagentsAll analytes used in this study were obtained from Sigma (St.

Louis, MO). The stock solutions of the solutes were prepared inmethanol (HPLC grade) (Fisher Scientific, Fair Lawn, NJ). Thedeionized water (18 MΩ) was prepared in our laboratory using aSybron/Barnstead (Boston, MA) system. All mobile phases werepurged using helium gas prior to each use.

Subcritical water separationA homemade subcritical water chromatography–UV system

was employed in this work. A Hewlett-Packard (Avondale, PA)gradient pump Series 1050 was used to deliver the mobile phase.The flow rate was 0.2 mL/min for the ZirChrom-PBD column and1 mL/min for the other three columns. The outlet of the pumpwas connected to a Valco injector fitted with a 2-µL sample loop(purchased from Keystone Scientific). The injector was locatedjust outside a Fisher Isotemp oven. A piece of stainless steeltubing (100-cm × 0.005-inch i.d.) (Keystone) was connectedbetween the injector and the separation column as a preheatingcoil. Both the preheating coil and the separation column wereplaced inside the oven. The preheating coil acted like a high-tem-perature water reservoir to ensure that the water eluent reachedthe desired temperature before entering the separation column.Because water will be vaporized at 1 atm and temperatures at100°C or higher, backpressure must be applied to the outlet of thecolumn in order to keep water in the liquid state. There are sev-eral reasons for avoiding steam in subcritical water chromatog-raphy–UV systems. The water mobile phase may stay in liquidnear the column inlet, thus causing steam to form inside the sep-aration column near the outlet end if there is not enough back-pressure applied. Thus, the mobile phase exists as two separatephases (liquid water and steam) in the separation column. Inaddition, the UV signal strongly fluctuates if steam exists in thesystem. This means that the UV detector is not stable when steampasses through the flow cell. In this study, a capillary restrictor (7-cm × 75-µm) (Polymicro Technologies Inc., Phoenix, AZ) wasplaced outside the oven and between the separation column andthe UV flow cell in order to ensure that the water inside the sepa-ration column stayed in the liquid state at higher temperatures.Connection unions (1/16 inch to 1/16 inch) (Supelco, Bellefonte,PA) were used to connect the restrictor. By 1/16-inch stainless steeltubings, the inlet of union 1 and the outlet of union 2 were con-nected to the column and UV detector, respectively. The fused-

silica capillary restrictor was connected with both unions usinggraphite ferrules (Alltech, Deerfield, IL). We evaluated the influ-ence of the restrictor dimension on the retention time usingrestrictors having 7 to 30 cm in length and 51 to 103 µm in innerdiameter. However, there was no significant effect of the restrictordimension on the retention time. A restrictor with a length of 7 cm and a 75-µm inner diameter was chosen for all of the exper-iments reported in this work. An LDC variable wavelengthdetector (spectro Monitor 3200, Riviera Beach, FL) was used inthis separation system. The UV detector was set at a wavelength of254 nm for the entire work.

After purging the deionized water with helium, the water wascontinuously pumped through the separation column at either0.2 or 1 mL/min, depending on the columns used. Then, theoven was turned on and set to a desired temperature. In order toensure that separations were carried out at the set temperature,the first injection was not made until approximately 20 min afterthe oven temperature was reached. This allowed the stationaryphase in the packed column and the mobile phase to equilibrateto the desired temperature. It should be noted that the tempera-ture of the stationary phase and the mobile phase inside thecolumn lagged behind the oven temperature by approximately 5to 20 min, depending on the temperature employed. A HewlettPackard 3396 Series II integrator was used as the data-recordingdevice. The peak width monitored in this work was at half-height, and the number of theoretical plates (N) was computedusing the following equation:

N = 2π(tRH/A)2 Eq. 1

where H and A are the peak height and area, respectively.


Zorbax RX-C8 columnThe Zorbax RX-C8 column was first used to study the temper-

ature effect on the peak width and column efficiency. The testsolutes in this study included pyridine, benzamide, catechol, andguaiacol. The temperature used for the separation of these ana-lytes ranged from 60°C to 100°C because this column wasproven to be thermally stable at temperatures up to 100°C forseveral thousand column volumes (18). In case of coelution, theanalytes were injected individually. It is known that the retentiontime is decreased with increasing temperature. The same trendwas observed in this study with all four solutes tested. Forexample, pyridine was not eluted until approximately 44 min at60°C (as shown in Figure 1A) (t0 = ~2.4 min). However, the sameanalyte was eluted within approximately 16 min at 100°C. Itshould be noted that the decrease in retention with increasingtemperature was in an almost linear fashion (Figure 1A). Figure1B demonstrates the temperature effect on the peak width forthe test analytes. Because the viscosity of water decreased dra-matically when the temperature was raised (as shown in Table I)(21,22), the diffusivity was greatly enhanced. Thus, narrower


109

peaks were obtained at elevated temperatures. Similar to theretention time, the reduction in the peak width with increasingtemperature was not dramatic.

The influence of temperature on column efficiency is demon-strated in Figure 1C. Based on the type of curves in Figure 1C,the solutes can be divided into two groups. The first groupincludes benzamide and pyridine. The peak efficiency of thesetwo solutes was significantly improved with increasing tempera-

ture. The number of theoretical plates obtained at 100°C was53–84% higher than that at 60°C for benzamide and pyridine.This uptrend temperature effect on efficiency was achievedbecause the reduction in retention was slower than the reduc-tion in peak width when the temperature was raised. This phe-nomenon can be seen from Figures 1A and 1B. When thetemperature was increased from 60°C to 100°C, the reduction inretention time and peak width for benzamide was 49% and 62%,respectively. However, the plate number of catechol and guaiacolwas almost unchanged when the temperature was raised from60°C to 100°C. This means that both the retention time and peakwidth of catechol and guaiacol were decreased with similar rateswhen the temperature was increased.

PRP-1 columnBecause the polymeric PRP-1 column is thermally stable at

temperatures up to 160°C based on our previous work (18), thetemperature range was expanded to 160°C to evaluate the tem-perature effect on the column efficiency with a greater tempera-

Table I. Temperature Effect on the Viscocity of Water*

Viscocity (cP)

Temperature (°C) At 50 bar At 100 bar

25 0.8898 0.888950 0.5479 0.5487

100 0.2836 0.2849150 0.1832 0.1844200 0.1345 0.1357250 0.1061 0.1075

* Obtained from references 21 and 22.

Figure 1. Temperature effect on (A) the retention time, (B) peak width, and (C) number of theoretical plates for separation on the Zorbax RX column.

Figure 2. Temperature effect on (A) the retention time, (B) peak width, and(C) number of theoretical plates for separation on the PRP-1 column.


110

ture range. Therefore, the temperature effect on the peak effi-ciency of the same or similar solutes (resorcinol, catechol, ben-zamide, and pyridine) was also investigated by using the PRP-1column. The separation temperatures ranged from 60°C to160°C at an interval of 20°C. The analytes were injected individ-ually in case of coelution at higher temperatures. The retentiontime and peak width were also significantly decreased withincreasing temperature as shown in Figure 2 (t0 = ~2.0 min). Thereduction in retention time and peak width was up to 80% forthese analytes by raising the separation temperature from 60°Cto 160°C. This means that the analysis was 7–9 times faster at160°C than that at 60°C.

The effect of temperature on peak efficiency is depicted inFigure 2C. Similar to the separation on the Zorbax column,uptrend curves of temperature versus column efficiency wereobtained in the temperature range of 60°C to 120°C. However, thenumber of theoretical plates was decreased when the temperaturewas further raised to 160°C. Thus, the peak efficiency reached amaximum at temperatures in the 100°C to 120°C range. Thismaximum of peak efficiency shows that the reduction in reten-tion time was smaller than the reduction in peak width at the low-temperature range, and the decrease in retention time wasgreater than the decrease in peak width at the high-temperaturerange. This phenomenon can be clearly seen from Figures 2A and2B. For example, the retention time of resorcinol was reduced35% while its peak width experienced a 48% reduction when tem-perature was raised from 60°C to 80°C (low-temperature range).However, the opposite phenomenon was observed at the highertemperature range. When the temperature was increased from140°C to 160°C, the decrease in retention time and peak width forcatechol was 20% and 8%, respectively.

Hypersil ODS columnIn order to further explore the effect of temperature on

column efficiency, a mixture of phenol, 2-chlorophenol, and 2,3-dichlorophenol was separated using a Hypersil ODS column.Even though the thermal stability of this column is poorer thanthe PRP-1 column, we still used a temperature range of 60°C to140°C. Because the column was exposed to high temperaturesonly for several hours (the most) in this study, the thermal sta-bility did not get significantly worse within this short period oftime based on our previous study (18). Again, both the retentiontime (t0 = ~1.0 min) and peak width were decreased withincreasing temperature as shown in Figures 3A and 3B. Thenumber of theoretical plates (equivalent to a 25-cm column) wasslightly increased for chlorophenols but stayed almostunchanged for phenol when the temperature was increased from60°C to 100°C. Further raising the temperature from 100°C to140°C caused a significant decrease in column efficiency (asdemonstrated in Figure 3C). This was in agreement with theresults obtained by using the PRP-1 column even though dif-ferent analytes were used.

ZirChrom-PBD columnBased on references 19 and 20, the ZirChrom-PBD column

was stable at temperatures up to the range of 150°C to 200°C.Therefore, the phenol mixture was also separated on theZirChrom-PBD column at temperatures ranging from 60°C to

140°C. Because the inner diameter of the ZirChrom-PBDcolumn was 2.1 mm, a flow rate of 0.2 mL/min was used for thiscolumn. Similar to separations on the Hypersil ODS column, thecolumn efficiency was either increased or unchanged when thetemperature was raised from 60°C to 100°C (as shown in Figure4C), but the plate number (equivalent to a 25-cm column) wasdecreased when the temperature was further increased from100°C to 140°C. However, the decrease in efficiency was less sig-nificant for this zirconia-based column compared with that forthe Hypersil column. As can be seen from Figure 4C, increasingthe temperature from 60°C to 140°C resulted in either nodecrease or a very little decrease (approximately 15%) in effi-ciency with the ZirChrom-PBD column but a typical 40%decrease with the Hypersil ODS column. This means that the

Figure 3. Temperature effect on (A) the retention time, (B) peak width, and (C) number of theoretical plates (equivalent to a 25-cm column) for separationon the Hypersil ODS column.


111

zirconia-based column is more suitable for separations at highertemperatures. Another benefit associated with the ZirChrom-PBD column is that the analysis time required by this columnwas much shorter than that required by the Hypersil column. Asshown in Figures 3 and 4, separation on the ZirChrom-PBDcolumn was approximately four times faster than that on theHypersil column, but the column efficiency of the ZirChrom-PBD under the fast analysis conditions was still competitivecompared with that obtained by the Hypersil column.

Mechanism for the temperature effect on column efficiency insubcritical water chromatography

To the best of our knowledge, this is the first report that quan-titatively describes the effect of water temperature on column

efficiency over a wide temperature range in subcritical waterchromatography. In many cases, a maximum in column effi-ciency in subcritical water chromatography was obtained whenthe water temperature was varied in this work. We believe thatthis maximal efficiency was caused by two factors that are asso-ciated with water temperature. The first one is the mass transferthat improves the efficiency, and the second is the longitudinaldiffusion that worsens the column efficiency.

It is well-known that the diffusivity or diffusion coefficient ofthe mobile phase (Dm) is directly proportional to the absolutetemperature and inversely proportional to the viscosity of themobile phase. Because the viscosity of water is decreased withincreasing temperature as demonstrated in reference 22 (Table I),the diffusivity (mass transfer) of water is dramatically increasedand the mass transfer resistance (the Cm term in the van Deemterequation) is greatly decreased at elevated temperatures. Thus,narrower bands and higher column efficiency should be expectedwith increasing temperature. This is why the number of theoret-ical plates was generally increased when the temperature wasraised from 60°C to the 100°C to 120°C range (as illustrated inFigures 1–4). Therefore, we believe that mass transfer may domi-nate the subcritical water separation process at the lower temper-ature range (lower than the 100°C to 120°C range).

By increasing the temperature from 100°C to 120°C, the diffu-sivity is further increased and even better mass transfer results.However, the better mass transfer also causes a greater axialmolecular diffusion (longitudinal diffusion, the B term in the vanDeemter equation), which makes the column efficiency becomepoorer. Therefore, the higher the temperature, the greater thelongitudinal diffusion (B is directly proportional to Dm in the vanDeemter equation) and the lower the column efficiency. This isthe reason why the number of plates was decreased when thetemperature was raised from the 100°C to 120°C range to the140°C to 160°C range (Figures 2–4). Therefore, longitudinal dif-fusion may be the dominating factor that controls the subcriticalwater separation at the higher temperature range. Carr et al.(20) reported that the column efficiency was decreased for sepa-rations using organic solvent–water mixtures at temperatures of150°C and 200°C, although the authors indicated that this mightbe caused by the interaction of molecules with the column wallsat higher temperatures (20).

Because increasing the separation temperature causes lowermass transfer resistance (the C term in the van Deemter equa-tion decreases) but also greater longitudinal diffusion (the Bterm in the van Deemter equation increases), a maximal columnefficiency may be observed. However, if the decrease in the Cterm and increase in the B term are similar in the lower temper-ature range, then they compensate each other. Thus, the effi-ciency will stay unchanged when the temperature is increasedfrom low temperature to the 100°C to 120°C range. This is evi-denced in Figures 1, 3, and 4. However, at a higher temperaturerange the increase in the B term always exceeds the decrease inthe C term. Therefore, the column efficiency was alwaysdecreased when the temperature was further raised. This mayexplain why the number of plates was always decreasing with allof the columns and solutes tested when the temperature wasincreased from the 100°C to 120°C range to the 140°C to 160°Crange (as shown in Figures 2–4).

Figure 4. Temperature effect on (A) the retention time, (B) peak width, and (C) number of theoretical plates (equivalent to a 25-cm column) for separationon the ZirChrom-PBD column.


112

Conclusion

The elution of several polar analytes has been achieved byusing pure water and four different types of commercially avail-able reversed-phase columns at elevated temperatures undermoderate pressure to keep the water in the liquid state. A fused-silica capillary restrictor was connected between the separationcolumn and the UV flow cell to provide the backpressure neededto avoid water from boiling at higher temperatures. For all of theanalytes studied and columns used, the peak width wasdecreased with increasing water temperature. For example, thepeak width of benzamide obtained on the PRP-1 column wasreduced by as much as 93% by raising the temperature from60°C to 160°C. However, the column efficiency was eitherimproved or remained unchanged initially but then decreasedwith increasing temperature. Thus, a maximum in efficiency wasobserved at temperatures in the 100°C to 120°C range in mostcases.

Acknowledgments

This research was supported by an award from ResearchCorporation (CC4607). The authors would also like to thankZirChrom Separations Inc. for providing the ZirChrom-PBDcolumn. Toru Kondo acknowledges Fuji Silysia Chemical Ltd.and the Department of Chemistry at East Carolina University forproviding the opportunity to conduct this research.

References

1. N.D. Sanders. Observation of the solubility of heavy hydrocarbonsin near-critical water. Ind. Eng. Chem. Fundament. 25: 171–74(1986).

2. D.J. Miller and S.B. Hawthorne. Method for determining the solu-bility of hydrophobic organics in subcritical water. Anal. Chem. 70:1618–21 (1998).

3. D.J. Miller, S.B. Hawthorne, and A.A. Clifford. Solubility of poly-cyclic aromatic hydrocarbons in subcritical water from 298 K to 498 K. J. Chem. Eng. Data 43: 1043–46 (1998).

4. Y. Yang, S.B. Hawthorne, and D.J. Miller. Toluene solubility andorganic partitioning from gasoline and diesel fuel into water at ele-vated temperatures and pressures. J. Chem. Eng. Data 42: 908–13(1997).

5. R.M. Smith and R.J. Burgess. Superheated water—a clean eluent forreversed-phase high-performance liquid chromatography. Anal.Commun. 33: 327–29 (1996).

6. D.J. Miller and S.B. Hawthorne. Subcritical water chromatographywith flame ionization detection. Anal. Chem. 69: 623–27 (1997).

7. R.M. Smith and R.J. Burgess. Superheated water as an eluent for reversed-phase high-performance liquid chromatography. J. Chromatogr. A 785: 49–55 (1997).

8. B.A. Ingelse, H.-G. Janssen, and C.A. Cramers. HPLC–FID withsuperheated water as the eluent: improved methods and instrumen-tation. J. High Resolut. Chromatogr. HRC 21: 613 (1998).

9. Y. Yang, A.D. Jones, and C.D. Eaton. Retention behavior of phenols,anilines, and alkylbenzenes in liquid chromatographic separationsusing subcritical water as the mobile phase. Anal. Chem. 71:3808–13 (1999).

10. O. Chienthavorn and R.M. Smith. Buffered superheated water as aneluent for reversed-phase high-performance liquid chromatography.Chromatographia 50: 485 (1999).

11. T.M. Pwlowski and C.F. Poole. Solvation characteristics of pressur-ized hot water and its use in chromatography. Anal. Commun. 36:71–75 (1999).

12. W.R. Melander, B.-K. Chen, and C. Horváth. Mobile phase effects inreversed-phase chromatography I. Concomitant dependence of retention on column temperature and eluent composition. J. Chromatogr. 185: 99–109 (1979).

13. F.V. Warren and B.A. Bidlingmeyer. Influence of temperature oncolumn efficiency in reversed-phase liquid chromatography. Anal.Chem. 60: 2821–24 (1988).

14. T. Welsch, M. Schmid, J. Kutter, and A. Kalman. Temperature of theeluent: a neglected tool in high-performance liquid chromatog-raphy? J. Chromatogr. 728: 299–306 (1996).

15. L.C. Sander and S.A. Wise. Subambient temperature modification ofselectivity in reversed-phase liquid chromatography. Anal. Chem.61: 1749–54 (1989).

16. A. Tchapla, S. Heron, H. Colin, and G. Guiochon. Role of tempera-ture in the behavior of homologous series in reversed-phase liquidchromatography. Anal. Chem. 60: 1443–48 (1988).

17. L.A. Cole and J.G. Dorsey. Temperature dependence of retention inreversed-phase liquid chromatography. 1. Stationary-phase consid-erations. Anal. Chem. 64: 1317–23 (1992).

18. P. He and Y. Yang. “Thermal Stability of Reversed-Phase ColumnsUnder Subcritical Water Conditions”. Presented at EasternAnalytical Symposium, Atlantic City, NJ, November 2000.

19. J. Li and P.W. Carr. Effect of temperature on the thermodynamicproperties, kinetic performance, and stability of polybutadiene-coated zirconia. Anal. Chem. 69: 837–43 (1997).

20. J. Li, Y. Hu, and P.W. Carr. Fast separation at elevated temperatureson polybutadiene-coated zirconia reversed-phase material. Anal.Chem. 69: 3884–88 (1997).

21. Y. Yang, M. Belghazi, A. Lagadec, S.B. Hawthorne, and D.J. Miller.Elution of organic solutes from different polarity sorbents using sub-critical water. J. Chromatogr. A 810: 149–59 (1998).

22. L. Haar, J.S. Gallagher, and G.S. Kell. National Bureau ofStandards/National Research Council Steam Tables. HemispherePublishing Corporation, New York, NY, 1984.


113

A technique is presented for the specific and sensitive determinationof ethambutol concentrations in plasma, bronchoalveolar lavage(BAL), and alveolar cells (AC) using a high-pressure liquidchromatographic (HPLC)–tandem mass spectrometric (MS–MS)method. The preparation of samples requires a deproteinization stepwith acetonitrile. The retention times for ethambutol, neostigminebromide, and propranolol are 2.0, 1.4, and 1.1 min, respectively, witha total run time of 2.8 min. The detection limits for ethambutol are0.05 µg/mL for plasma and 0.005 µg/mL for the BAL supernatants andAC suspensions. The assay has excellent performance characteristicsand has been used to support a study of the intrapulmonarypharmacokinetics of ethambutol in human subjects.

Introduction

Ethambutol has a primary role in the treatment of tuberculosisand is recommended with isoniazid, rifampin, and pyrazinamideas initial therapy (1). Ethambutol is rapidly absorbed and has abioavailability of 7% after oral administration (2,3). Under fastingconditions, the maximum concentration (mean ± standard devi-ation, SD) of the drug in serum is 4.5 ± 1.0 µg/mL and the timeto maximum concentration is 2.5 ± 0.9 h (2). The minimuminhibitory concentration of ethambutol for M. tuberculosisranges from 0.5 to 2 µg/mL in broth media (4). A microbiologicalassay (detection limit of 0.4 µg/mL) using M. smegmatis as thetest organism for determining ethambutol in serum has beenreported (5). A gas chromatographic (GC)–mass spectrometric(MS) method has been used for the determination of ethambutolin tablets (6). A GC–liquid chromatographic (LC) assay withimproved performance characteristics (detection limit of 0.1µg/mL in plasma) has been used to study the pharmacokinetics of

ethambutol in humans (3,7–9) and rabbits (10), and a high-pres-sure liquid chromatographic (HPLC) method for the determina-tion of ethambutol in plasma (detection limit of 10 ng/mL) andurine (detection limit of 10 µg/mL) has been described (11).

We report the use of a sensitive HPLC–tandem mass spectro-metric (MS–MS) technique to measure ethambutol in humanplasma, bronchoalveolar fluid (BAL) (detection limit of 0.05µg/mL), alveolar cells (AC) (detection limit of 0.005 µg/mL), andplasma (detection limit of 0.05 µg/mL). Compared with othermethods, the technique has the advantages of increased sensi-tivity and a capability to analyze small sample volumes. The speci-ficity of HPLC–MS–MS detection greatly minimizes the risk ofinterference from other substances. This is especially importantwhen analyzing specimens from patients such as those with AIDSwho are taking numerous concomitant medications. It currentlyis being used to support a phase-one study of the intrapulmonarypharmacokinetics of ethionamide in normal subjects and sub-jects with AIDS.

Experimental

ChemicalsAll solvents and chemicals were HPLC grade except ammo-

nium acetate, which was certified. A 1.0-mg/mL solution ofethambutol HCl (Lederle Laboratories, Wayne, NJ) was made in50% methanol and stored refrigerated. This solution was furtherdiluted to produce working stock solutions of 0.1, 1.0, and 10µg/mL of ethambutol. Stock solutions of 1.0 mg/mL neostigminebromide (Aldrich Chemical Co., Milwaukee, WI) and propranolol(USP Reference, Rockville, MD) were prepared in 50% methanol.Neostigmine bromide and propranolol were then diluted to a con-centration of 0.050 µg/mL in acetonitrile and used as the internalstandard for plasma, and propranolol was diluted to 0.300 µg/mLand used as the internal standard for BAL and AC.

Abstract

A High-Pressure Liquid Chromatographic–Tandem Mass Spectrometric Method for the Determination ofEthambutol in Human Plasma, Bronchoalveolar LavageFluid, and Alveolar Cells

John E. Conte, Jr.1,2,3,*, Emil Lin4, Yeping Zhao4, and Elisabeth Zurlinden1

1Department of Epidemiology and Biostatistics, Infectious Diseases Research Laboratory, 2Department of Medicine, 3Department ofMicrobiology and Immunology, and 4Department of Biopharmaceutical Sciences, University of California, San Francisco, 350 ParnassusAvenue, Suite 507, San Francisco, CA 94117



* Author to whom correspondence should be addressed.


114

InstrumentalChromatography

The mobile phase (containing 80% acetonitrile, 4mM ammo-nium acetate, and 0.10% trifluoroacetic acid) was run through ahypersil silica column (50- × 4.6-mm i.d., 5-µm particle size) at a

flow rate of 0.8 mL/min using a Shimadzu (Columbia, MD) LC-10AD pump. Extracts from samples were injected onto the systemwith a Waters (Milford, MA) Intelligent Sample Processor 717Plus. The retention times for ethambutol, neostigmine, and pro-pranolol were 2.0, 1.4, and 1.1 min, respectively, with a total runtime of 2.8 min.

MSWe used two different MS systems during the development and

validation of this assay to explore different types of MS equipment.Neostigmine bromide was the internal standard used for theplasma and BAL that were assayed on the PE Sciex API III(PerkinElmer, Foster City, CA), whereas propranolol was used asthe internal standard for the assays in plasma, BAL, and ACs per-formed on the Micromass (Manchester, U.K.) Quattro LC. Peakdetection and area determinations for some plasma and BAL were

Figure 1. Daughter ion spectra and chemical structures of ethambutol using theSciex APCI mode.

Figure 2. Daughter ion spectra and chemical structures of neostigmine (theinternal standard) using the Sciex APCI mode.

Figure 3. Daughter ion spectra and chemical structures of ethambutol using theMicromass Quattro LC electrospray mode.

Figure 4. Daughter ion spectra and chemical structures of propranolol (theinternal standard) using the Micromass Quattro LC electrospray mode.


115

carried out with a PE Sciex API III.The MS used the following settings and conditions. The mul-

tiple reaction monitor scanning mode was set at m/z 205–116 forethambutol and m/z 209–71 for neostigmine (Figures 1 and 2).Atmospheric pressure chemical ionization (APCI)–positive ion-ization was used. The sample inlet used a heated nebulizer at450°C. The discharge current was +3 µA. The gas curtain flow was1.2 L/min (N2 = 99.999%). The nebulizer pressure was 551.4 kPa.The collision gas consisted of a 9.99% nitrogen–90.01% argonmixture (set at 250 × 1012 molecules/cm2). Peak detection for theACs and some plasma and BAL specimens was carried out on aMicromass Quattro LC. For these specimens the reaction channelwas m/z 205.35–116.10 for ethambutol and m/z 260.18–115.95

for propranolol (Figures 3 and 4). Electrospray–positive ioniza-tion with a flow rate of 0.2 mL (5-to-1 split ratio of 1.0 mL/min)to the Micromass system was used. The sample inlet used a heatednebulizer. The sample cone was set to 25 V for ethambutol and 35 V for propranolol. The energy collision was set to 15.0 eV forboth ethambutol and propranolol. A Macintosh Quadra 800 com-puter (Apple Computers, Cupertino, CA) was used for peak inte-gration and analysis.

Sample preparationStandard curves

Plasma standard curves were prepared by adding appropriatevolumes of ethambutol working stock solutions into 0.2 mL ofblank plasma to yield the concentrations of 0.05, 0.10, 0.20, 0.40,0.80, 1.2, 1.6, and 2.4 µg/mL of ethambutol. The standards forBAL supernatants were spiked to yield concentrations of 0.005,0.010, 0.020, 0.040, 0.080, 0.160, 0.320, and 0.640 µg/mL ofethambutol. The AC suspension standards were spiked to yieldconcentrations of 0.005, 0.010, 0.020, 0.040, 0.100, 0.400, 0.800,1.600, and 2.000 µg/mL ethambutol. Standard curves were con-structed by plotting a 1/y weighted least-squares linear regressionof ethambutol to the internal standard peak-area ratios versus thespiked concentration of ethambutol.

Preparation of plasma standards and samplesIn order to ensure consistency of recovery, 200 µL of acetoni-

trile containing 0.050 µg/mL neostigmine or propranolol as theinternal standard was added to 0.2 mL plasma standards and sam-ples. After vortexing, an additional 0.2 mL of the internal standardsolution was added. After vortexing and then centrifuging for 5 min at 1800 × g, the solvent phase was transferred to a 400-µLmicrofuge tube, and 2.0 µL were injected onto the HPLC system.

Preparation of BAL supernatants and AC pellet standardsand samples

A cell count and differential was performed on the BAL lavagefluid, then a 30-mL aliquot was centrifuged at 400 × g for 5 minand the supernatant immediately separated from the cells. BALsupernatant standards and samples were prepared by adding 0.5mL of the internal standard solution (0.015 µg/mL neostigmineor 0.150 µg/mL propranolol) to 0.25 mL of the sample, vortexing,and then centrifuging for 5 min at 1800 × g. The solvent phasewas transferred to a 400-µL microfuge tube, and 2.0 µL wereinjected onto the HPLC system.

ACs were resuspended volumetrically in deionized water andsonicated for 2 min on a Fisher 550 dismembrator (FisherScientific, Santa Clara, CA) to lyse the cells. A 250-µL volume ofthe internal standard (0.300 µg/mL propranolol) was added to 250µL of an AC cell suspension and vortexed. A 250-µL volume of ace-tonitrile was added and mixed by vortexing. Following centrifu-gation for 5 min at 1800 × g, 2 µL of the solvent phase wasinjected onto the HPLC system.

Preparation of controls for method validationTwo sets of stock solutions were prepared; one was used for

spiking standards and the other for spiking controls. Measuredamounts of plasma were spiked at 0.15, 0.4, 0.8, and 1.4 µg/mL;aliquoted; and frozen at –70°C for stability studies. Aliquots were

Figure 5. Chromatograms of blank plasma: (A) the internal standard and (B) ethambutol.

Figure 6. Chromatograms of a study subject’s plasma obtained 4 h after the fifthdose of 15 mg/kg ethambutol administered once a day: (A) the internal standardand (B) ethambutol. The ethambutol concentration was 0.734 µg/mL.

A

B

A

B


116

analyzed in duplicate weekly over a period of six weeks. In orderto assess interday reproducibility, standard curves with controlsspiked at concentrations of 0.1, 0.3, 1.2, and 2.4 µg/mL were ana-lyzed on five different days. Intraday reproducibility was assessedby analyzing six preparations of each of the four concentrationson the same day. The validation for BAL supernatants was carriedout in the same time frames as for plasma, with controls spiked atconcentrations of 0.015, 0.04, 0.16, and 0.24 µg/mL. The valida-tion for ACs was performed at concentrations of 0.010, 0.40, and1.60 µg/mL.

StatisticsThe statistical analysis was performed using the PROPHET

Computer Resource (12). Linearity (r2), precision (coefficient ofvariation, CV), recovery (relation of test result to the true concen-

tration) (13), and percentage accuracy (14) were calculated. Thedetection limit was defined as the lowest point of the standardcurve. Drug concentrations in epithelial lining fluid (ELF) werecalculated using the urea diffusion method, and AC concentra-tions were calculated using cell counts in alveolar fluid as we havepreviously reported (15–17).


Linearity, assay precision, recovery, and accuracy assessmentsHPLC–MS–MS chromatograms of ethambutol and the internal

standard in plasma, BAL supernatant, and AC suspension areshown in Figures 5–10. The detection limits for ethambutol were0.05 µg/mL for plasma and 0.005 µg/mL for the BAL supernatants

Figure 7. Chromatograms of blank BAL supernatant: (A) the internal standardand (B) ethambutol.

Figure 10. Chromatograms of a study subject’s AC suspension obtained 4 hafter the fifth dose of 15 mg/kg ethambutol administered once a day: (A) theinternal standard and (B) ethambutol. The ethambutol concentration was 0.316µg/mL.

Figure 9. Chromatograms of blank AC suspension: (A) the internal standard and(B) ethambutol.

Figure 8. Chromatograms of a study subject’s BAL supernatant obtained 4 hafter the fifth dose of 15 mg/kg ethambutol administered once a day: (A) theinternal standard and (B) ethambutol. The ethambutol concentration was 0.053µg/mL.

A

B

A

B

A

B

A

B


117

and AC suspensions. The detection limit referred to the lowestpoint of the standard curve and was at least five times the noiselevel. The mean ± SD of the r2 from 24 standard curves (8 inplasma, 8 in BAL, and 8 in ACs) was 0.9941 ± 0.0060. Results forthe assay precision, recovery, and accuracy assessments in theplasma, BAL, and AC suspensions are summarized in Tables I–III.

CVThe mean (± SD) CVs and the ranges of the assay for intraday

and interday determinations together for plasma, BAL super-natants, and ACs were 7.81% ± 2.02% (ranging from 3.9% to10.14%), 6.46% ± 3.69% (ranging from 1.42% to 11.42%), and

12.67% ± 4.59% (ranging from 6.0% to 20.0%), respectively(Tables I–III).

The mean (± SD) recoveries and the ranges of the assays forintraday and interday determinations together in plasma, BALsupernatants, and ACs were 105.91% ± 7.73% (ranging from93.3% to 119.0%), 95.94% ± 10.43% (ranging from 80.0% to106.88%), and 105.48% ± 3.60% (ranging from 100.00% to110.00%), respectively (Tables I–III). The accuracy ranges for allof the determinations in plasma, BAL supernatants, and ACs were–6.67% to 19.0%, –20.0% to 6.88%, and 0.0% to 10.0%, respec-tively (Tables I–III).

StabilityThe results of repeated determinations of ethambutol in spiked

plasma, BAL supernatants, and ACs stored at –70°C revealed nosignificant degradation of the drug. These determinations wereperformed over a period of 4 mo for plasma, 7 weeks for BAL

Table II. Assay Precision, Recovery, and Accuracy forEthambutol Determination in BAL Supernatant

Measured Spiked concentration

concentration (mean ± SD) Recovery* Accuracy†

(µg/mL) (µg/mL) CV (%) (%) (%)

Intraday‡ (n = 6)0.240 0.248 ± 0.004 1.4 103.33 3.330.160 0.171 ± 0.004 2.2 106.88 6.880.040 0.040 ± 0.002 6.2 100.00 0.000.015 0.012 ± 0.001 4.4 80.00 –20.0

Interday§ (n = 12)0.240 0.238 ± 0.023 9.9 99.17 –0.830.160 0.165 ± 0.011 6.4 103.13 3.130.040 0.038 ± 0.004 11.4 95.00 –5.00.015 0.012 ± 0.001 9.8 80.00 –20.0

* Measured/spiked × 100%.† (Measured – spiked)/spiked × 100%.‡ Six separately spiked samples at each of four concentrations.§ Plasma spiked at four concentrations and analyzed in duplicate on six different days.

Table III. Assay Precision, Recovery, and Accuracy forEthambutol Determination in Alveolar Cells



(µg/mL) (µg/mL) CV (%) (%) (%)

Intraday‡ (n = 6)1.600 1.643 ± 0.099 6.0 102.69 2.690.400 0.423 ± 0.053 12.6 105.75 5.750.010 0.010 ± 0.001 14.5 100.00 0.00

Interday§ (n = 10)1.600 1.707 ± 0.207 12.1 106.69 6.690.400 0.431 ± 0.047 10.8 107.75 7.750.010 0.011 ± 0.002 20.0 110.00 10.00

* Measured/spiked × 100%.† (Measure – spiked)/spiked × 100%.‡ Six separately spiked samples at each of three concentrations.§ Plasma spiked at three concentrations and analyzed in duplicate on five different days.

Table I. Assay Precision, Recovery, and Accuracy forEthambutol Determination in Plasma



(µg/mL) (µg/mL) CV (%) (%) (%)

Intraday‡ (n = 6)2.4 2.24 ± 0.227 10.1 93.33 –6.671.2 1.30 ± 0.111 8.6 108.33 8.330.3 0.32 ± 0.012 6.1 106.67 6.670.1 0.12 ± 0.011 9.2 119.00 19.00

Interday§ (n = 10)2.4 2.35 ± 0.168 7.2 97.92 –2.081.2 1.26 ± 0.114 9.1 105.00 5.000.3 0.33 ± 0.013 3.9 110.00 10.000.1 0.107 ± 0.009 8.3 107.00 7.00

* Measured/spiked × 100%.† (Measured – spiked)/spiked × 100%.‡ Six separately spiked samples at each of four concentrations.§ Plasma spiked at four concentrations and analyzed in duplicate on five different days.

Table IV. Ethambutol Concentrations* in Plasma, ELF, andAC in Five Adult Volunteer Subjects

Subject Subject Subject Subject Subject Sample #1 #2 #3 #4 #5

Plasma (2 h after fifth dose†) 3.41 1.79 1.15 1.75 0.86Plasma (4 h after fifth dose) 4.99 1.15 2.11 1.90 2.39ELF‡

(4 h after fifth dose) 3.61 1.14 3.05 1.80 2.51AC§

(4 h after fifth dose) 64.82 18.92 59.98 108.9 35.42

* All concentrations are given in micrograms per milliliter.† A single oral daily dose of 15 mg/kg was given for 5 days.‡ The amount of ELF collected in the BAL fluid was calculated from the urea

concentration in BAL and serum, as previously reported (15–17).§ The concentration of ethambutol in ACs is given as micrograms per milliliter of

cell volume and was calculated as previously reported (15–17).


118

supernatant, and 9 mo for ACs (data not shown). The mean (± SD)CV of the stability studies at four concentrations in plasma andBAL supernatant were 8.38% and 7.36%, respectively. Repeatanalyses of BAL pellets from four study subjects resulted in amean (± SD) CV of 0.10%.

Patient dataThe concentrations of ethambutol in plasma, BAL supernatant,

and ACs in five of forty subjects who participated in an NIH-sup-ported study of the intrapulmonary pharmacokinetics of ethamb-utol are summarized in Table IV. Bronchoscopy and BAL wereperformed, and blood was drawn at 4 h following the last dose of a5-day course of 15 mg/kg ethambutol. Blood samples were alsoobtained 2 h after the last dose. From this preliminary analysis, itcan be seen that ethambutol concentrations in plasma ranged from0.86 to 3.41 µg/mL at 2 h and 1.15 to 4.99 µg/mL at 4 h after the lastdose was administered. The concentrations in ELF ranged from1.14 to 3.61 µg/mL and in AC ranged from 18.92 to 108.9 µg/mL.

Conclusion

We have developed a sensitive HPLC–MS–MS assay that pro-vides specific, rapid, and reliable determinations for ethambutolin small volumes of plasma, BAL, and AC. The preparation ofplasma, BAL supernatant, and AC samples requires a depro-teinization step. The stability data indicated that no significantdrug degradation occurred in plasma, BAL supernatant, or ACsstored at –70°C over a period of 4 mo, 6 weeks, and 9 mo, respec-tively. The linearity of the standard curve in the range describedwas excellent. Assay precision was high for plasma, BAL, and ACs.The performance characteristics of this assay make the methodsuitable for clinical and pharmacological studies, particularlythose that are designed to quantitate the intrapulmonary concen-tration of drugs.

This method is currently being used to support phase-onestudies of the pulmonary pharmacokinetics of ethambutol inpatients with tuberculosis and normal volunteers. In this prelim-inary analysis, ethambutol concentrations in plasma and ELFappear to be similar (i.e., the drug diffuses passively from plasmainto ELF). Ethambutol concentrations are considerably greaterin the AC than in plasma or ELF, indicating that ethambutol isconcentrated in ACs. This finding may be of importance in thetreatment of tuberculosis, which is an intracellular infection. Acomplete analysis of this pharmacokinetic study will be publishedelsewhere.

Acknowledgments

This work was carried out with funds provided by NIH Grant#AI36054 and NIH Grant #5 MO1 RR-00079 (General ClinicalResearch Center) at the University of California, San Francisco.

The authors would like to thank Ganfeng Wong for assay devel-opment, Margareta Andersson for performing the assays, and EveBenton for manuscript preparation.

References

1. J.B. Bass, Jr., L.S. Farer, P.C. Hopewell, R. O’Brien, R.F. Jacobs, F. Ruben, D.E. Snider, Jr., and G. Thornton. Treatment of tuberculosisand tuberculosis infection in adults and children. American ThoracicSociety and The Centers for Disease Control and Prevention. Am. J. Respir. Crit. Care Med. 149: 1359–74 (1994).

2. C.A. Peloquin, A.E. Bulpitt, G.S. Jaresko, R.W. Jeliffe, J.M. Childs, andD.E. Nix. Pharmacokinetics of ethambutol under fasting conditions,with food, and with antacids. Antimicrob. Agents Chemother. 43:568–72 (1999).

3. C.S. Lee, J.G. Gambertoglio, D.C. Brater, and L.Z. Benet. Kinetics oforal ethambutol in the normal subject. Clin. Pharmacol. Ther. 22:615–21 (1977).

4. C.N. Lee and L.B. Heifets. Determination of minimal inhibitory con-centrations of antituberculosis drugs by radiometric and conven-tional methods. Am. Rev. Respir. Dis. 136: 346–52 (1987).

5. P.R. Gangadharam and E.R. Candler. Microbiological assay ofethambutol. J. Antimicrob. Chemother. 3: 57–63 (1977).

6. T.L. Ng. A gas chromatographic/mass spectrometric study of thetrimethylsilylation of ethambutol and a tablets assay method basedon the trimethylsilyl derivative. J. Chromatogr. Sci. 20: 479–82(1982).

7. C.S. Lee and L.Z. Benet. Gas–liquid chromatographic determinationof ethambutol in plasma and urine of man and monkey. J. Chromatogr. 128: 188–92 (1976).

8. C.S. Lee and L.Z. Benet. Micro and macro GLC determination ofethambutol in biological fluids. J. Pharm. Sci. 67: 470–73 (1978).

9. C.S. Lee, D.C. Brater, J.G. Gambertoglio, and L.Z. Benet. Dispositionkinetics of ethambutol in man. J. Pharmacokinet. Biopharm. 8:335–46 (1980).

10. M.M. Chen, C.S. Lee, and J.H. Perrin. Absorption and disposition ofethambutol in rabbits. J. Pharm. Sci. 73: 1053–55 (1984).

11. M. Breda, P. Marrari, E. Pianezzola, and B.M. Strolin. Determinationof ethambutol in human plasma and urine by high-performanceliquid chromatography with fluorescene detection. J. Chromatogr. A729: 301–307 (1996).

12. Prophet. Abtech Corporation, Charlottesville, VA, 1999.13. S.A. Signs, T.M. File, and J.S. Tan. Antimicrob. Agents Chemother.

26: 652–55 (1984).14. R.V. Smith and J.T. Stewart. Textbook of Biopharmaceutic Analysis.

Lea and Febiger, Philadelphia, PA, pp. 79–94, 1981.15. J.E. Conte, Jr., J.A. Golden, S. Duncan, E. McKenna, E. Lin, and

E. Zurlinden. Single-dose intrapulmonary pharmacokinetics ofazithromycin, clarithromycin, ciprofloxacin, and cefuroxime in vol-unteer subjects. Antimicrob. Agents Chemother. 40: 1617–22(1996).

16. J.E. Conte, Jr., J.A. Golden, S. Duncan, E. McKenna, and E. Zurlinden.Intrapulmonary pharmacokinetics of clarithromycin and of ery-thromycin. Antimicrob. Agents Chemother. 39: 334–38 (1995).

17. J.E. Conte, Jr., J.A. Golden, S. Duncan, E. McKenna, and E. Zurlinden.Intrapulmonary concentrations of pyrazinamide. Antimicrob. AgentsChemother. 43: 1329–33 (1999).


c h roma t o g ra p h i c s c i e nc e - harvard apparatus...r.p.w. scott a.m. siouffi larry t....

Documents