Direct measurement of urinary testosterone and epitestosterone conjugates using high-performance liquid chromatography/tandem mass spectrometry

JOURNAL OF MASS SPECTROMETRYJ. Mass Spectrom. 35, 50–61 (2000)

Direct measurement of urinary testosteroneand epitestosterone conjugates usinghigh-performance liquidchromatography/tandem mass spectrometry

David J. Borts† and Larry D. Bowers*Athletic Drug Testing and Toxicology Laboratory, Department of Pathology and Laboratory Medicine, Indiana UniversityMedical Center, Medical Science Building A-128, 635 Barnhill Drive, Indianapolis, Indiana 46202-5120, USA

Measurement of the ratio of testosterone (T) and epitestosterone (E) in urine has been used as an indicationof ‘natural’ steroid supplementation for a decade. The direct measurement of the glucuronide and sulfateconjugates of testosterone and epitestosterone by high-performance liquid chromatography/tandem massspectrometry (HPLC/MS/MS) should resolve a number of issues regarding unusual metabolism due to eithergenetic disposition or attempts to avoid detection of abuse. Determination of nanomoles per liter (0.1 ppb)concentrations of analytes in a complex biological matrix by HPLC/MS/MS is complicated by sample matrix-specific ion suppression during ESI. Deuterated internal standards of all compounds were used to overcome theeffects of suppression. Comparison of the HPLC/MS/MS method with a two-part gas chromatographic/massspectrometric method showed statistical equivalence in urine samples. Analysis of urine samples with elevatedT-glucuronide to E-glucuronide ratios did not show that a significant number could be explained by an elevatedexcretion of epitestosterone sulfate. The HPLC/MS/MS method was also used further to characterize geneticand metabolic factors that give rise to unusual T/E ratios. Copyright 2000 John Wiley & Sons, Ltd.

KEYWORDS: steroids; glucuronide; sulfate; electrospray; ion suppression

INTRODUCTION

The use of anabolic steroids to improve athletic perfor-mance in major sports has been rumored for more than twodecades, and the recent documentation of a government-sponsored doping program1 has substantiated these fears.The International Olympic Committee (IOC) and othernational and international sport governing bodies havedeveloped programs to combat drug use. A critical com-ponent of any anti-drug program is effective testing, fre-quently of urine samples. Decreased use, prompted byimprovements in analytical testing methods and increasedgovernmental control of synthetic anabolic agents, is oftencredited for falling rates of steroid detection. More skepti-cal observers suggest that decreasing positive rates are theresult of more sophisticated users and ‘designer drugs.’

* Correspondence to: L. D. Bowers, Athletic Drug Testing andToxicology Laboratory, Department of Pathology and LaboratoryMedicine, Indiana University Medical Center, Medical Science Build-ing A-128, 635 Barnhill Drive, Indianapolis, Indiana 46202-5120,USA.E-mail: [email protected]† Present address: Research Triangle Institute, Analytical and Chemi-

cal Sciences, P.O. Box 12194, Research Triangle Park, North Carolina27709-2194, USA.

Contract/grant sponsor: National Collegiate Athletic Association.Contract/grant sponsor: National Football League.Contract/grant sponsor: United States Olympic Committee.

Anabolic steroids which are normally present in thebody, such as testosterone (T) and its precursors dehy-droepiandrosterone, androstenedione and androstenediol,present a particularly challenging analytical problem inathletic drug testing. It is now clear that supraphysiolog-ical doses of T can enhance muscle mass and strength.2

The range of T concentrations observed in random urinecollections is large, such that the definitive detection ofpharmacological testosterone use would be impossiblebased on this measurement alone. Donikeet al.3 demon-strated that by computing the ratio of the urinary T con-centration to that of a metabolically unrelated steroid,epitestosterone (E), many physiological factors that affectthe urine concentration could be eliminated. Populationstudies performed in a number of laboratories4–6 havedocumented that the mode of the logrithmic distributionof urinary T/E ratios in men and women is in the range0.9–1.6 with a standard deviation of¾1.0. Based onDonike et al.’s work, the IOC established a T/E ratio of>6 as an indication of exogenous use of testosterone-enhancing compounds. Catlinet al.7 reported five sampleswith urine T/E ratios>6 among the 1510 samples ana-lyzed at the 1984 Olympic Games, although this rateof incidence could be artificially increased by individu-als using exogenous testosterone-enhancing compounds.Dehennin and Scholler8 have estimated the incidenceof T/E ratio >6 in a teenage athletic population at 15in 10 000.

The analytical problem in T/E ratio determination isfurther complicated by the fact that steroids undergo

CCC 1076–5174/2000/010050–12 $17.50 Received 29 April 1999Copyright 2000 John Wiley & Sons, Ltd. Accepted 11 October 1999

HPLC/MS/MS OF URINE TESTOSTERONE AND EPITESTOSTERONE 51

conjugation (phase II metabolism) with glucuronide (G),sulfate (S) and potentially other polar species prior toexcretion in urine. Gas chromatography/mass spectrom-etry (GC/MS) is the preferred technique for the detectionand/or quantification of urinary steroids for athletic drugtesting. However, the steroids need to be deconjugated andderivatized before the sample can be analyzed. Deconjuga-tion is achieved by enzymatic or acidic hydrolysis. Enzy-matic hydrolysis is accomplished with-glucuronidaseextracted fromE. coli, beef liver,P. vulgataor H. poma-tia. The preparations fromP. vulgata and H. pomatiacontain bothˇ-glucuronidase and aryl sulfatase activity.The 17-sulfate androgens are reported to resist hydroly-sis by the aryl sulfatase fromH. pomatiaandP. vulgatawhich have a weak sulfatase activity that is highly spe-cific for 3ˇ-hydroxy-5 - and -3 -hydroxy-5-steroids.9,10

Consequently, the aryl sulfatase treatment does not cleavethe TS or ES conjugates. Thus Donikeet al.’s originalwork actually measured the ratio of the T and E glu-curonides (TG/EG).

It has been demonstrated that administration oftestosterone11–14 or steroid precursors of testosterone15

can increase the TG/EG ratio to>6. Concerns about‘naturally elevated’ TG/EG ratios, however, have beenpublished.16–18 Dehenninet al. reported that increasedexcretion of ES was responsible for several casesof ‘naturally elevated’ TG/EG ratio and proposedmeasurement of the TG/(EGC ES) ratio.18,19 The GC/MSmethods for measuring this ratio are involved, andhave a number of potential problems. We report hereon the development of an a high-performance liquidchromatographic (HPLC)/electrospray ionization tandemmass spectrometric (ESI-MS/MS) method for the directmeasurement of the urinary glucuronide and sulfateconjugates of T and E.

EXPERIMENTAL

Reagents and solvents

Sulfate and glucuronide conjugates of T were purchasedfrom Steraloids (Wilton, NH, USA). Unlabeled (d0) Econjugates and all four 16,16,17-2H3-labeled (d3) con-jugates, used as internal standards (IS), were synthe-sized in our laboratory.20 Ammonium acetate (99.999%),acetic acid (doubly distilled), sodium hydrogencarbon-ate (99.7C%), potassium carbonate (99C%), propanethiol(99C%) and ammonium iodide (99.999%) were obtainedfrom Aldrich (Milwaukee, WI, USA). Highly pure(18 M�) water was obtained by passing house-distilled,deionized water through a Barnstead Nanopure II sys-tem (Syborn/Barnstead, Boston, MA, USA). Methanol(HPLC grade) and ethyl acetate (HPLC grade) wereobtained from Curtin Matheson (Houston, TX, USA).Sodium phosphate (dibasic, 99.7%) and hexanes (HPLCgrade) were from Fisher Scientific (Fair Lawn, NJ, USA).E. coli ˇ-glucuronidase (½200 U ml�1) was supplied byBoehringer Mannheim (Indianapolis, IN, USA). Concen-trated sulfuric acid (instrumental grade) was purchasedfrom EM Science (Gibbstown, NJ, USA), sodium chlo-ride (ACS reagent grade) from Sigma (St Louis, MO,

USA) and N-trimethylsilyltrifluoroacetamide (MSTFA)from Alltech (Deerfield, IL, USA).

Urine sample collection

Untimed urine samples were collected from healthy volun-teers (aged 17–50 years), including 45 males, five femalesand 10 Chinese males. Urine samples from an additional27 individuals that were found to have T/E ratio>6 byroutine GC/MS screening were drawn from the servicebase of the Athletic Drug Testing and Toxicology Labo-ratory at the Indiana University Medical Center. No addi-tives were used and the samples were frozen at�20°Cuntil analysis.

Sample preparation

Solid-phase extraction (SPE) was performed with a 12-port Visiprep SPE vacuum manifold (Supelco, Bellefonte,PA, USA). The extraction conditions were first optimizedwith standard mixtures of the four steroid conjugates.An aliquot of urine (3 ml) was spiked with 100 nmoll�1 each of d3-TG and d3-EG and 50 nmol l�1 eachof d3-TS and d3-ES. The spiked urine samples weremixed with 3 ml of ammonium acetate–acetic acid buffer(buffer I, 0.1 mol l�1, pH 4.5). Disposable C-18 BondElut LRC extraction cartridges with 200 mg of sorbent(Varian, Harbor City, CA, USA) were conditioned sequen-tially with methanol (1 ml), water (1 ml) and ammoniumacetate–acetic acid buffer (buffer II, 15 mmol l�1, pH 4.5)(2 ml). The spiked urine sample was passed through theprimed cartridge. The cartridge was then washed sequen-tially with buffer II (2 ml), methanol–water (2 ml, 2 : 3,v/v), which was 15 mmol l�1 in ammonium acetate andhad the same concentration of acetic acid as that of bufferII, and water (2 ml). The conjugates were eluted withmethanol–water (5.0 ml, 1 : 1, v/v). A slow, dropwiseflow-rate was maintained during washing and elution. Theeluate was dried on a SpeedVac concentrator (Savant,Farmingdale, NY, USA) and stored at�20°C until prepa-ration for the appropriate analysis.

HPLC/MS/MS assay

A Beckman Model 126 programmable solvent modulewas used to deliver eluent through a Hypersil BDSC18 column (150ð 1 mm i.d., 3µm, 120A) (KeystoneScientific, Bellefonte, PA, USA), at a flow-rate of50 µl min�1. Isocratic elution with a 43% (v/v) solution ofmethanol–7.5 mmol l�1 NH4OAc, 0.11% glacial HOAcbuffer was used throughout. Samples were reconstitutedin 100 µl of mobile phase and 10µl were injected. ARheodyne (Cotati, CA, USA) Model 8125 injection valveequipped with a 20µl loop was housed in a DuPont(Wilmington, DE, USA) forced air oven thermostatedat 37°C. After every 15 injections the column wasflushed for 30 min with a 90% (v/v) solution ofmethanol–7.5 mmol l�1 NH4OAc, 0.11% glacial HOAcat 50µl min�1 followed by a 30 min wash with theelution solvent before the subsequent injection. The HPLCcolumn was connected to the IonSpray probe by a fused-silica capillary (100µm i.d. ð 170 µm o.d.) (Polymicro

Copyright 2000 John Wiley & Sons, Ltd. J. Mass Spectrom. 35, 50–61 (2000)

52 D. J. BORTS AND L. D. BOWERS

Technologies, Phoenix, AZ, USA) and all effluent wasdirected to the mass spectrometer.

A Perkin-Elmer SCIEX (Norwalk, CT, USA) APIIII PLUS triple-quadrupole mass spectrometer operated inthe positive ion electrospray mode was used for detec-tion. The electrospray voltage was 4.0 kV and the inter-face temperature was set at 60°C. Zero grade air (GasTech, Indianapolis, IN, USA) was used as the nebulizinggas at an operating pressure of 40 psi. The curtain gaswas ultra-high purity nitrogen (Gas Tech) flowing at 1.2 lmin�1. Both quadrupoles Q1 and Q3 were mass calibratedwith poly(propylene glycol) and operated at unit resolu-tion. Ultra-high purity argon (Linde, Danbury, CT, USA)at a thickness of 3.0ð 1011 molecules cm�2 was used asthe collision gas.

The data acquisition scheme was divided into threeperiods in the Routine Acquisition and Display soft-ware (Perkin-Elmer SCIEX) to apply an optimized set ofparameters to each steroid conjugate. The orifice voltagewas set at 75 V for TG (period 1), at 75 V for TS andES (period 2) and at 65 V for EG (period 3). The colli-sion energy in the laboratory frame of reference was set at20 eV for TG, at 18 eV for TS and ES and at 15 eV forEG. The mass spectrometer was operated in the selectedreaction monitoring (SRM) mode. Q1 was used to selectprecursor ions corresponding to the protonated pseudo-molecular ions (m/z 465 for d0-TG andd0-EG; m/z 369for d0-TS andd0-ES; m/z 468 for d3-TG and d3-EG;andm/z 372 for d3-TS andd3-ES). Q3 was stepped tomonitor the intact precursor ions and product ions cor-responding to the deconjugated steroid nucleus (m/z 289(d0), m/z 292 (d3)), the singly (m/z 271 (d0), m/z 274(d3)) and doubly (m/z 253 (d0), m/z 256 (d3)) dehydrateddeconjugated steroid nucleus. The dwell time for eachtransition was 200 ms, which gave¾0.5 scans s�1, orabout 30 scans across a chromatographic peak. The peakareas corresponding to the transitionsm/z 465! 289 ford0-TG,m/z 369! 289 ford0-TS,m/z 369! 271 ford0-ES, andm/z 465! 289 ford0-EG were measured usingMACQUAN (Perkin-Elmer SCIEX) software. The peakareas of the corresponding transitions for the deuteratedinternal standard steroid conjugates were used to calcu-late the peak area ratio used for calibration. The measuredconcentrations of TG, EG, TS and ES were used to calcu-late the various ratios. Relative ion abundances for eachcompound were computed from the peak areas of the indi-vidual transitions.

GC/MS assay

Urine residues from SPE were reconstituted with 400µlof 1 : 1 (v/v) methanol–water, split in half and dried. Thefirst portion was reconstituted in 1.0 ml of 0.2 mol l�1

sodium phosphate buffer (pH 7.0) to which 50µl of E. coliˇ-glucuronidase solution (containing½10 U of enzyme)were added, and the solution was vortex mixed and incu-bated for 1 h at 60°C. The solution was cooled and100 mg of solid sodium hydrogencarbonate–potassiumcarbonate (3 : 2, w/w) and 5 ml of hexanes were added.The solution was shaken for 10 min and centrifuged for5 min at 2000g. The hexane layer was removed, evapo-rated to dryness and the residue derivatized as describedbelow. The second portion was reconstituted with 1 mlof ethyl acetate and 2µl of aqueous 4.0 mol l�1 sulfuric

acid solution were added. The solution was vortex mixed,incubated for 1 h at 40°C and cooled, after which 1 mlof 5% (v/v) aqueous sodium hydrogencarbonate and 4 mlof ethyl acetate were added. After shaking for 10 minand centrifugation for 5 min at 2000g, the ethyl acetatelayer was removed and washed with 2.5 ml of saturatedNaCl solution. The organic solvent was removed, evap-orated to dryness and the residue derivatized with 50µlof MSTFA–1-propanethiol–NH4I (1000 : 3 : 2, v/v/w) for45 min at 60°C to form TMS-enol–TMS-ether deriva-tives. A 1 µl portion of the derivatization mixture wasinjected directly without further treatment.

We used a Varian (Walnut Creek, CA, USA) Sat-urn III GC/MS system equipped with a Model 3400gas chromatograph for all GC/MS analyses. Injec-tions were made with a Model 8200 autosampler intoa Model 1078 split/splitless injector operated in thesplitless mode with a 2 mm i.d. glass wool-packedliner. The split vent was closed for the first 0.7 minafter injection and the injection rate was 0.2µl s�1. Apolysilarylene–polydimethylsiloxane bonded-phase capil-lary column (DB-5ms; 30 mð 0.25 mm i.d., 0.25µm filmthickness) (J&W Scientific, Folsom, CA, USA) was usedthroughout. The He carrier gas flow-rate was adjusted to1 ml min�1 at 250°C. The initial GC column temperaturewas 170°C, ramped to 260°C at 20°C min�1, to 305°Cat 2.7°C min�1 and finally to 320°C at 20°C min�1.

Ion current chromatograms were obtained by scanningover them/z 400–450 range. The data acquisition softwarewas set to sum three microscans per analytical scan, whichgave five analytical scans s�1, or about 30 scans over achromatographic peak. Peak areas from them/z 432 (MCžfor di-TMS d0-T and -E) andm/z 435 (MCž for di-TMSd3-T and -E) extracted ion profiles were used for quanti-tation. Calibration standards and urine samples analyzedby GC/MS were corrected for the [MC 3]Cž (m/z 435)isotopic contribution of thed0 standards to the MCž (m/z435) d3 internal standard extracted ion profile peak area.The theoretical contribution was calculated to be 3.1%and was observed to be in the range 2.9–3.1%. Correc-tions were made by subtracting 3.0% of them/z 432 peakarea from the internal standard peak area for all standardsand samples prior to preparing the calibration curve orinterpolating the measured concentration. Measured con-centrations were used to compute the ratios of all species.

Calibration curves

Calibration standards containing each of the fourd0steroid conjugates were prepared in HPLC mobile phase atconcentrations corresponding to urinary concentrations of0, 10, 25, 75, 150, 300, 750 and 1500 nmol l�1. Each stan-dard was spiked with the four deuterated internal standardsat the same level as the urine samples. For HPLC/MS/MS,the standards were reconstituted in mobile phase andinjected directly. For GC/MS, the standards were dried,derivatized and analyzed in the same way as the urinesamples.

RESULTS AND DISCUSSION

Our development of direct analytical methods formeasurement of steroid conjugates arises from our interest



in steroid metabolism. The importance of HPLC/MS/MSin the study of anabolic steroid metabolism, especiallyphase II metabolism, was demonstrated earlier, when thefacile hydrolysis of 17-methyl-17 -sulfate conjugateswas shown to be the origin of the previously unexplainedpresence of unconjugated steroid metabolites in urine.21

We have continued to develop HPLC/MS/MS methods forbroad spectrum urinary screening for steroid conjugates.22

It became apparent in the early stages of this work thatbecause of the large number of structurally similar steroidconjugates present in urine and the limited resolvingpower of HPLC, the synthesis of stable isotope-labelledinternal standards would be critical in development of themethod.20

Although GC/MS is the accepted methodology forsteroid analysis, there are a number of drawbacks tothis approach. The first is hydrolysis of the conjugates.In addition to glucuronidase and sulfatase activity,H.pomatia juice is reported to contain 3-hydroxysteroiddehydrogenase/5–4 isomerase activity which can con-vert 5-androstene-3,17 -diol and 5-androstene-3,17 -diol into T and E, respectively.23 As mentioned above,the sulfate conjugates cannot be cleaved enzymatically,so acid hydrolysis must be used. Under acidic hydroly-sis conditions, ES is reported to undergo acid-catalyzedWagner–Meerwein rearrangement.18 Such a rearrange-ment involves a shift of an adjacent group, such as the18-methyl group, to the carbocation center of the inter-mediate produced under acidic conditions. Methanolysis,which has been reported to cleave both the glucuronideand sulfate conjugates in a single step,18,19,24 has resultedin variable cleavage and rearrangements in our hands.When trimethylsilyl derivatives are used, it is importantto note that the peak area for thed3 internal standards hadto be corrected for the [MC 3]Cž isotopic contribution ofthe d0 signal prior to calculation of the peak area ratios.This contribution arises primarily from the two siliconatoms in the TMS groups used to derivatize the steroidsfor GC/MS. Ignoring this contribution results in about a23% error at the high concentration end of the calibrationrange described above. Finally, Linnet25 has described thepresence of urine matrix effects in the GC/MS analysisof T and E. While these limitations can be overcome,HPLC/MS/MS would appear to have a number of advan-tages over GC/MS approaches.

Microbore HPLC was chosen to maximize sensitivitysince the concentrations of the sulfate conjugates of T

and E have been reported to be in the 0.35–350 nmol l�1

range.18 HPLC conditions were chosen to give baselineresolution of the steroid conjugates (retention times: TG,11.8 min; TS, 14.7 min; ES, 18.5 min; EG, 22.3 min),to minimize interferences from any isobaric endogenoussteroid conjugates present in the urine samples, and toallow optimized mass spectrometric parameters to beapplied to each analyte individually (Fig. 1). Isocraticelution was employed to eliminate the need for time-consuming solvent cycling necessary after gradient elu-tion, although in practice this benefit was probably notachieved owing to the need to wash the analytical columnafter each 15 injections. Small, sample-specific retentiontime fluctuations, presumably due to column overload,were noted throughout these analyses. Retention times forrepeated injections of a particular urine sample were veryreproducible. These fluctations presented no problem inquantification since each conjugate had a correspondingdeuterated internal standard. Column overload conditionsare unusual in HPLC analyses. In this case the use of amicrobore column and the difficulty in isolating the polarsteroid conjugates from other urine matrix componentsduring the sample preparation were the likely explanationfor this problem.

Because of the low concentrations of the steroid conju-gates, the mass spectrometer was operated in the SRMmode. Optimization of mass spectrometric conditionsfor each conjugate was necessary because each of thefour steroid conjugates behaves differently with respectto adduct formation and collision-induced dissociation(CID). The mass spectrometric conditions were optimizedby infusing standard solutions of each conjugate and firstadjusting the orifice voltage to give the maximum sig-nal for any adducted form of the conjugate to serve asthe precursor ion, and then adjusting the collision energyto give the maximum signal for the most abundant ionof three characteristic product ions. The three charac-teristic product ions for each conjugate corresponded tothe deconjugated steroid nucleus, the singly dehydrateddeconjugated steroid nucleus and the doubly dehydrateddeconjugated steroid nucleus. Bowers and Sanaullah havedescribed the tendencies of TG and EG to form differ-ent adducts as a function of orifice voltage.22 They alsonoted the relative lability of the glycosidic bond of EGat higher orifice potential.20 TG and EG give protonated,singly ammoniated, singly sodiated and doubly sodiated

Figure 1. Selected LC/MS/MS ion current profiles for a typical male urine sample.



adducts. The singly ammoniated adduct of EG was themost abundant ion at low orifice voltages, but the proto-nated adduct was most abundant at higher orifice voltagesand was found to give better signal-to-noise ratios withreal urine samples. TS and ES exhibit a larger variety ofadducts (Fig. 2), forming doubly ammoniated and sodi-ated, ammoniated adducts in addition to those describedfor the glucuronides. The singly protonated adduct waschosen as the precursor for each conjugate as it was themost abundant adduct for all but EG. TS and ES requiredcollision energies similar to TG to cause similar amountsof precursor fragmentation (Fig. 3). Both of the sulfate

conjugates, however, fragmented more completely thanthe glucuronides to give more abundant low-m/z productions. TS gave lower characteristic product ion abundancesthan the other conjugates.

Detection limits for the HPLC/MS/MS system usingstandard solutions of each steroid conjugate and a cri-teria of a signal-to-noise (S/N) level½5 were deter-mined to be 2 pmol on-column. This corresponds to aconcentration of about 6.7 nmol l�1 in urine for thesample preparation described above. Detection limits forurine samples were approximately the same but werehighly sample dependent. The main source of the sample

Figure 2. Dependence of (A) testosterone sulfate and (B) epitestosterone sulfate pseudomolecular ion abundances on orifice voltage.Pseudomolecular ions: �, m/z 369, protonated adduct; �, m/z 386, ammoniated adduct; °, m/z 391, sodiated adduct; ž, m/z 403,doubly ammoniated adduct; M, m/z 408, singly ammoniated plus singly sodiated adduct; N, m/z 413, doubly sodiated adduct.

Copyright 2000JohnWiley & Sons,Ltd. J. MassSpectrom. 35, 50–61 (2000)


Figure 3. Dependence of precursor and product ion abundances on collision energy for (A) testosterone sulfate and (B) epitestosteronesulfate. Precursor ions: �, m/z 369, [MC H]C. Product ions: �, m/z 289, [M� SO3]C; °, m/z 271, [M� SO3 � H2O]C; M, m/z 253,[M� SO3 � 2H2O]C.

dependencewas extensivesignal suppressiondue to thepresenceof co-eluting matrix components.Signal sup-pression with ESI has been discussedby Ikonomouet al.,26 Buhrmanet al.27 and Matuszewskiet al.28 Thephenomenonoccurswhenoneor moreco-elutingcompo-nentscausesa decreasein the measuredion abundanceof an analyte.Gas-phaseproton transfer reactionshavebeensuggestedas a mechanismfor signal suppression,but other mechanismsare also possible.26 In the presentstudy, we found that for equal amountsof d3-internalstandardinjectedon-column,the integratedpeakareawasmuchlessfor urine extractsthanstandardsolutions.This

signal suppressionvaried greatly betweensamplesandalso betweenthe four conjugatesof interest in individ-ual urine samples(Fig. 4). The extentof suppressionforsteroid conjugateswithin a particular samplewas cor-related for the four conjugates,and may be a functionof the overall concentrationof the original urine sampleas reflectedby its specific gravity. The extent of signalsuppression,given as the differencebetweenthe peakareacountsfor the internal standardin standardsolutionand the urine extractsas a percentageof the peak areacountsfor the internalstandardin standardsolutions,wasestimatedto be 81š 13% (range42–98%) for d3-TG,



Figure 4. Extent of ionization suppression for 10 representative male urine samples. Sample D shows the most suppression whereassample B shows relatively little suppression.

79š 14% (range46–96%) for d3-TS, 69š 17% (range20–98%) for d3-ES and 57š 19% (range24–99%) ford3-EG in a 25 sample subsetof the 45 sample maleurine referencegroup.The trendtoward lesssuppressionat longer retentiontimes may be relatedto the amountand/ortypeof interferingmatrix componentsco-elutingateachtime. Sincetheanalyseswereperformedin theSRMmode,interferingmatrix componentswerenot identified.The signal suppressioneffect demonstratesthe necessityof having a deuteuratedinternal standardfor eachana-lyte in orderto achieveaccuratequantitation.Becauseofthe great variation in the extent of suppression,the useof anything less than a comprehensiveset of deuteratedinternal standardsis inadequate.Further work is underway to characterizethe interferingcomponents’ion abun-dancesuppression.Improvementsin samplepreparationmethodsshoulddecreasethe signalsuppressioneffects.

This finding appearsto contradict the prevailing atti-tuderegardingthe ability to minimize samplepreparationprior to HPLC/ESI-MSanalysisandto degradechromato-graphicresolutionto increasethespeedof analysis.Thereare two featuresof this analysisthat enhancethe poten-tial for ion suppression.First, the relative polar steroidconjugatesaredifficult to separatefrom the urine matrixcomponentsandotherglucuronideandsulfateconjugates.Sincethe samplepreparationandHPLC separationwereboth basedon reversed-phasechromatography,it is notsurprising that some componentsmight co-elute. Sec-ond, given the expectedconcentrationsof steroid con-jugate to be quantified and the concentrationof otherconjugatedspeciesin urine, ion suppressionmight beexpected.Nevertheless,one must be careful in assess-ing systemperformance.Bean and Henion29 reportedarapid HPLC/MS/MSsystemfor T andE conjugatedeter-mination with little chromatographicresolution, relying

on MS/MS specificity to achieveaccurateresults.Giventhe potentialfor ion suppressionandthe numberof natu-ral isobaricsteroids,we feel that this approachis fraughtwith difficulties andshouldbe avoided.

For regulatorypurposes,measurementof the presenceof a prohibited substancerequires identification of thecompoundeither by full scanor by consistencyof ionabundanceratios betweena referencematerial and theunknown.In thecaseof GC/electronionizationMS, eitherthreeor four ion abundanceratiosof structurallycharac-teristic ions are requiredto be within š10–20% of thatobservedin a referencematerial analyzedat the sametime.30 Forchemicalionization,whichwould includeESI,control of conditionsis certainly more complexand thishasbeenrecognizedby requiring that the ion abundanceratiosfall within š20–25% of thoseobservedin a refer-encecompound.Our useof positive ESI for conjugatesthatwould benegativelychargedin solutionwasdueto alackof characteristicionsin thenegativeionizationmode.Even with positive ESI, the most commonly observedfragmentsin MS/MS fragmentationof steroidconjugatesarethe lossof theconjugate(glucuronideor sulfate)moi-ety and sequentialloss of H2O from the steroidnucleus.Although there is structural information in the relativeabundanceof the product ions of steroid conjugates,31

thepresenceof isobaricsteroidisomerprecursorionsandoverlapsin the production abundanceratioscould makeunambiguousidentification of steroid conjugatesin theurine matrix difficult.

Several groups have discussedion abundanceratioidentificationcriteria for HPLC/MS/MS with recommen-ded ranges of š10% or š20%.32–34 Based onobservationswith standardsolutions, the latter figureseemsto reflect better the instrumentalcapabilitiesforthe steroid conjugates.Within run, the relative standard



deviation of ion abundance ratios in standards was<8%for all four compounds. In samples, however, 19% ofthe TG, 7% of the ES and 0% of the EG peaks didnot meet aš20% identification criterion. Interestingly,5% of thed3-TG, 17% of thed3-ES and 0% of thed3-EG peaks also did not agree with their respective ionratios observed in standard solutions. TS is not includedin this discussion owing to the small number of sampleswith measurable concentrations. It is interesting to notethat the failure to meet identification criteria parallels thedegree of ion suppression observed for the three peaks. Ionsuppression might be expected to impact significantly onthe determination of ion abundance ratios due to changesin signal-to-noise ratios for lower abundance ions.

Comparison with a GC/MS method

In order to compare the LC/MS/MS method for determin-ing T/E ratios with the established technique of GC/MS,a reference group of 45 male urine samples that gave T/E

ratios<6 on routine screening were quantified by bothtechniques. Because of the potential problems involvedwith the deconjugation and derivatization steps requiredfor GC/MS, deuterated internal standards of TG, TS, ES,and EG were also used for the GC/MS analysis. Sev-eral statistical approaches were used to determine thecomparability of the two techniques (see Table 1). Lin-ear regression analysis of the TG/EG ratios measuredby LC/MS/MS and GC/MS for the 45 reference groupsamples gave a slope of 0.848 (š0.045), an intercept of0.16 (š0.047) and a correlation coefficient of 0.890. Stan-dard linear regression and Deming regression gave sim-ilar results (Fig. 5). Inspection of the residuals revealeda random distribution. A pairedt-test supported the nullhypothesis that there was no difference between the meth-ods at the 95% confidence level. A plot of the differencebetween the methods as a function of their mean for eachsample35 showed a distribution about the zero differenceline. If a 2% bias between methods and a 5% imprecisionis allowed, all concentrations for the comparison of TG,

Table 1. Linear fit parameters for LC/MS/MS versus GC/MSa

Parameter TG TS ES EG TG/EG ratio

Slopeš SD 1.045š 0.040 0.586š 0.109 0.935š 0.053 0.969š 0.044 0.957š 0.056(1.080š 0.027) (0.881š 0.088) (0.998š 0.038) (1.010š 0.031) (1.025š 0.040)

Interceptš SD �4.195š 8.350 2.312š 1.496 1.101š 2.069 5.908š 10.784 0.090š 0.065(�9.714š 5.723) (0.142š 1.215) (�1.015š 1.488) (�1.752š 7.666) (0.023š 0.046)

R2 0.941 0.403 0.878 0.920 0.875(0.970) (0.635) (0.937) (0.956) (0.935)

SyÐx 36.3 8.48 6.91 48.5 0.224(36.7) (9.18) (7.02) (49.0) (0.228)

a Values in parentheses are for the Deming regression assuming uncertainties in both axes.

Figure 5. Correlation between T/E ratios determined by GC/MS and LC/MS/MS. The dashed line represents classical linear regressionand the solid line represents the Deming regression method that assumes errors in the values for both axes.



EG and ES fall within the acceptance limits. In the caseof TS, nearly half of the samples had no measurable con-centration, making comparison impossible. Although thiscomparison shows that the two methods give equivalentresults when multipled3-internal standards are used, theLC/MS/MS method still has several advantages.

Comparison with literature population studies

Comparison of data generated with the HPLC/MS/MSmethod with those reported by Dehenninet al. can alsoassist in determining the validity of the new method. Themean concentrations of the glucuronide conjugates forour reference group are higher (Table 2) than values of120 nmol l�1 for TG and 128 nmol l�1 for EG deter-mined by Dehenninet al. using GC/MS and a 90-samplereference group.19 The mean concentrations of the sul-fate conjugates for our reference group are much lower

than the values of 25 nmol l�1 for TS and 70 nmol l�1

for ES determined by Dehenninet al. It is difficult toexplain the differences in the sulfate conjugate concentra-tions, but they may simply be the result of using smallreference groups in both studies. The differences do notappear to be simply the result of different instrumentaltechniques, as our LC/MS/MS results were confirmed byGC/MS with a comprehensive set of deuterated internalstandards. It should be noted, however, that Dehenninet al.19 used methanolysis to cleave both conjugate types,did not have a conjugated deuterated internal standard tocorrect for any side reactions during hydrolysis and cal-culated the sulfate concentrations by difference betweenthe total [(TGC TS), (EGC ES)] and glucuronide frac-tion (TG, EG) measurements. In spite of the differencesin average conjugate concentrations, the values of thecritical conjugate concentration ratios (Table 3) were sim-ilar. We found an average TG/EG ratio of 1.08 for our

Table 2. Concentrations of urinary testosterone and epitestosterone conjugates (nmol l−1)a

Parameter TG TS ES EG T E

Reference group (n D 45):Mean 162.5 6.6 32.7 183.9 169.2 216.6Standard deviation 148.2 10.9 19.6 169.3 147.1 179.8Range 19.6 750.9 0.0 51.5 0.0 92.2 21.1 810.7 19.6 750.9 21.7 892.4

TG/EG > 6 group (n D 27):Mean 427.2 73.1 64.3 49.4 500.3 113.7Standard deviation 383.3 100.6 81.5 44.5 435.9 118.9Range 43.2 1592.0 0.0 341.8 1.8 399.3 5.0 178.0 44.2 1933.8 6.8 577.3

Chinese male group (n D 10):Mean 40.7 13.7 44.5 92.5 54.4 137.0Standard deviation 65.4 13.4 28.4 59.3 70.1 80.9Range 4.3 209.8 0.9 41.7 11.6 103.9 27.7 217.1 7.2 236.9 50.7 321.6

Female group (n D 5):Mean 20.8 6.5 28.5 20.8 35.0Standard deviation 10.4 4.6 31.5 10.4 35.5Range 12.7 35.8 2.8 13.5 5.9 83.6 12.7 35.8 9.9 97.1

a 1 nmol is equivalent to 288 ng of free steroid.

Table 3. Ratios of urinary testosterone and epitestosterone conjugates

Parameter TG/EG TG/E T/E TS/ES EG/ES TG/TS

Reference group (n D 45):Mean 1.08 0.86 0.90 0.18 6.94 27.65Standard deviation 0.69 0.50 0.50 0.27 6.85 49.20Range 0.12 3.08 0.10 2.26 0.12 2.43 0.00 0.95 1.89 33.41 1.20 212.38

TG/EG > 6 group (n D 27):Mean 9.45 4.90 5.44 0.89 1.28 15.74Standard deviation 3.80 3.60 3.37 0.86 1.13 22.70Range 6.17 23.02 1.92 19.51 2.85 19.51 0.00 2.69 0.32 5.57 1.20 102.00

Chinese male group (n D 10):Mean 0.52 0.29 0.43 0.35 2.46 5.49Standard deviation 0.85 0.42 0.46 0.33 1.34 6.87Range 0.08 2.71 0.06 1.32 0.07 1.49 0.01 1.14 0.76 4.76 0.13 22.10

Female group (n D 5):Mean 1.22 0.85 0.85 4.28Standard deviation 0.79 0.42 0.42 2.60Range 0.43 2.31 0.37 1.37 0.37 1.37 1.48 7.38



reference group, while Dehenninet al.’s study gave a ratioof 1.3. These ratios also agree well with previous popula-tion studies using GC/MS.4–6 The average value for theTG/(EGC ES) ratio by the HPLC/MS/MS method is 0.86,whereas Dehenninet al. reported 0.7. Since our averageES concentration was much lower than Dehenninet al.’s,the difference between the TG/EG and TG/(EGC ES)ratios is not as large in our study. In eight of the 45 indi-viduals in our reference group, the TG/(EGC ES) ratiowas 30% or more lower than the TG/EG ratio. If theTS concentrations are considered, the average value forthe (TGC TS)/(EGC ES) ratio is 0.90, whereas Dehen-nin et al. found a ratio of 0.8. Since the TS concentrationsare much lower than for the other three conjugates, theireffect on the overall T/E ratio is small for most sam-ples in our reference group. The (TGC TS)/(EGC ES)ratio for four of the 45 individuals in our reference groupdid, however, change by 30% or more relative to boththe TG/EG and the TG/(EGC ES) ratios. Interestingly,the (TGC TS)/(EGC ES) ratio was increased relative tothe TG/(EGC ES) ratio in most of these individuals.We would propose, therefore, rationing the total concen-trations of the two compounds, rather than calculatingthe TG/(EGC ES) ratio as initially suggested by Dehen-nin et al.

Dehennin et al. reported that two individuals withTG/EG ratios in the 5–8 range had TG/(EGC ES) ratioswithin their reference range, while a single positive dop-ing case with a TG/EG of 75 had a TG/(EGC ES) ratioof 28.19 Although this finding is interesting, the trueevaluation of an improved test would be for individualswhose TG/EG ratios were in the 6–15 range. We analyzed27 samples that gave TG/EG ratios by routine GC/MSmethodology in the range 6.2–23.0, specifically 19 sam-ples in the 6–10 range, 7 samples in the 10–15 range

and one sample with a TG/EG ratio of 23. The mean TG,TS and ES concentrations are higher for this group thanfor the reference group (Table 2). The mean EG concen-tration for the T/E> 6 group is much lower than thatfor the reference group. The increase in T concentration,combined with a decrease in E concentration, could beconsistent with administration of exogenous testosterone.To examine the possibility that some of the samples inthe TG/EG group give elevated ratios due to physiolog-ical causes, the TG/(EGC ES) ratio was calculated andcompared with a proposed cutoff ratio. Dehennin hassuggested that a threshold for the TG/(EGC ES) ratiobe 2.85.18 Dehennin has shown that for 12 individualswith physiologically high TG/EG in the range 4–12, whowere known to be free from exogenous steroid administra-tion, the TG/(EGC ES) ratios always fell below the 2.85threshold.18 The cutoff value of 2.85 was calculated bytaking the average of the TG/(EGC ES) ratio value for areference group and adding four standard deviations. Thesame calculation with our reference group gives a valueof 2.86, nearly identical with Dehennin’s value. For thegroup of elevated TG/EG ratio samples, the mean ratio fellfrom 9.45 for TG/EG to 4.90 for the TG/(EGC ES) ratioor to 5.44 for the (TGC TS)/(EGC ES) ratio (Table 3).Of perhaps more importance, only three of the 27 sampleswould have been judged to have fallen within the refer-ence range for the TG/(EGC ES) ratio and only one ofthe 27 fell below the cutoff for the (TGC TS)/(EGC ES)ratio. Not surprisingly, the single determination whichchanged classification was the slightly elevated result(TG/EGD 6.2). Hence although the TG/(EGC ES) ratiomay be of value in some cases of alleged doping withtestosterone-enhancing substances, the fraction of indi-viduals with increased ES excretion is not as large assuggested by Dehennin’s studies.

Figure 6. T/E ratios determined by HPLC/MS/MS for 10 Chinese males.



There have been some suggestions that the TG/EGratio test is unfairly biased with respect some racialgroups. We were provided with the race of the individ-uals in the ‘normal’ and ‘elevated T/E’ groups abovewithout disclosure of their identity. The only two racialgroups that had sufficient numbers of tests to warrantanalysis were caucasian (16 normal; elevated) and Afro-American (24 normal; 9 elevated). There were no statis-tically significant differences between the mean TG, EGor ES concentrations or the TG/EG, TG/(EGC ES) or(TGC TS)/(EGC ES) ratios for the two groups with nor-mal or elevated findings. Owing to the relatively smallnumber of samples in this comparison, it is importantnot to over-interpret this result. Nevertheless, it is clearthat there are no major differences in the populationvalues.

During the course of the study, it was noted that severalvolunteers of Chinese descent had relatively low con-centrations of all of the steroid conjugates. It has beenreported that the urinary excretion of TG by Japanesemales is low compared with Caucasian and Ainu males,although no significant difference was noticed in theplasma levels of T.36,37 Given the ability of HPLC/MS/MSto measure steroid conjugates directly, we tested thehypothesis that the difference in excretion is the result ofphase II metabolism. The majority of samples collectedfrom 10 healthy Chinese males had substantially lowerTGC TS levels than the non-Chinese males. In contrast,almost all of these samples had EGC ES levels compa-rable to those of the Caucasian population. Consequently,the majority of these samples have lower TG/EG ratios(Fig. 6). These results strongly suggest that the geneticdifferences in T metabolism are not due to the enzymesof the conjugation system.

CONCLUSIONS

HPLC/MS/MS offers faster sample preparation thanGC/MS, eliminates the quantitative recovery problemsassociated with the hydrolysis and derivatization stepsof GC/MS and allows greater experimental flexibility insteroid metabolism studies. Significant ion suppressionwas observed for many of the urine samples; the useof four deuterated internal standards assured accuratedetermination of concentrations but did not assist inmeeting ion abundance ratio criteria. The developmentof a universally applicable validated method will requireimproved sample preparation procedures.

We used the HPLC/MS/MS method to quantify testos-terone and epitestosterone conjugates in urine samplesfrom a reference popoulation and random samples pre-viously found to have an elevated TG/EG ratio. Theresults obtained on the reference population were simi-lar to those of other reported studies. Based on the resultsreported here, the finding of elevated ES concentrationsas an explanation for a ‘naturally elevated’ T/E ratio isan uncommon occurrence, although we cannot rule it outin selected cases. We were also able to rule out phase IImetabolism as an explanation for racial differences in uri-nary T/E ratios.

Acknowledgements

This research was funded by grants from the National CollegiateAthletic Association, the National Football League, and the UnitedStates Olympic Committee. We acknowledge the contributions ofDr Sanaullah to synthesis of standard materials and preliminary effortsin this work.

REFERENCES

1. Franke WW, Berendonk B. Clin. Chem. 1997; 43: 1262.2. Bhasin S, Storer TW, Berman N, Callegari C, Clevenger B,

Phillips J, Bunnell TJ, Tricker R, Shirazi A, Casaburi R.N. Engl. J. Med. 1996; 3: 351.

3. Donike M, Barwald KR, Klostermann K, Schanzer W, Zim-mermann J. In Sport: Leistung und Gesundheit, Heck H,Hollmann W, Rost R (eds). Deutscher Artze-Verlag: Cologne,1983; 293 298.

4. Ayotte C, Goudreault D, Charlebois A. J. Chromatogr. 1996;687: 3.

5. Catlin DH, Hatton CK, Starcevic SH. Clin. Chem. 1997; 43:1280.

6. Baenziger J, Bowers LD, Unpublished data.7. Catlin DH, Kammerer RC, Hatton CK, Sekera MH, Merdink JL.

Clin. Chem. 1987; 33: 319.8. Dehennin L, Scholler R. Path. Biol. 1990; 38: 920.9. Roy AB. Biochem. J. 1956; 62: 41.

10. Houghton E, Grainger L, Dumasia MC, Teale P.Org. MassSpectrom. 1992; 27: 1061.

11. Kicman AT, Brooks RV, Collyer SC, Cowan DA, Nanjee MN,Southan GJ, Wheeler MJ. Br. J. Sports Med. 1990; 24: 253.

12. Dehennin L, Matsumoto AM. J. Steroid Biochem. Mol. Biol.1993; 44: 179.

13. Palonek E, Gottlieb C, Garle M, Bjorkhem I, Carlstrom K. J.Steroid Biochem. Mol. Biol. 1995; 55: 121.

14. Perry PJ, MacIndoue JH, Yates WR, Scott SD, Holman TL.Clin. Chem. 1997; 43: 731.

15. Bowers LD. Clin. Chem. 1999; 45: 295.16. Oftebro H. Lancet 1992; 339: 941.

17. Raynaud E, Audran M, Pages JC, Fedou C, Brun JF,Chanal JL, Orsetti A. Clin. Endocrinol. 1993; 38: 351.

18. Dehennin L. J. Endocrinol. 1994; 142: 353.19. Dehennin L, Lafarge P. Dailly Ph, Bailloux D. Lafarge J-P. J.

Chromatogr. B 1996; 687: 85.20. Sanaullah, Bowers LD. J. Steroid Biochem. Mol. Biol. 1996;

58: 225.21. Edlund PO, Bowers LD, Henion JD. J. Chromatogr. 1989;

487: 341.22. Bowers LD, Sanaullah. J. Chromatogr. B 1996; 687: 61.23. Venturelli E, Cavalleri A, Secreto G. J. Chromatogr. B 1995;

671: 363.24. Tang PW, Crone DL. Anal. Biochem. 1989; 182: 289.25. Linnet K. Biol. Mass Spectrom. 1993; 22: 412.26. Ikonomou MG, Blades AT, Kebarle P. Anal. Chem. 1990; 62:

957.27. Buhrman DL, Price PI, Rudewicz PJ. J. Am. Soc. Mass Spec-

trom. 1996; 7: 1099.28. Matuszewski BK, Constanzer ML, Chavez-Eng CM. Anal.

Chem. 1998; 70: 882.29. Bean KA, Henion JD. J. Chromatogr. B 1997; 690: 65.30. Baldwin R, Bethem RA, Boyd RK, Budde WL, Cairns T, Gib-

bons RD, Henion JD, Kaiser MA, Lewis DL, Matuzak JE,Sphon JA, Stephany RW, Trubey RK. J. Am. Soc. MassSpectrom. 1997; 8: 1180.

31. Schaff J, Bowers LD. Proceedings of the 46th Asms Confer-ence on Mass Spectrometry and Allied Topics, 1997; 83.

32. McLaughlin LG, Henion JD. Kijak PJ. Biol. Mass Spectrom.1994; 23: 417.



33. Schilling JB, Cepa SP, Menacherry SD, Bavda LT, Heard BM,Stockwell BL. Anal. Chem. 1996; 68: 1905.

34. Li YTL, Campbell DA, Bennett PK, Henion JD. Anal. Chem.1996; 68: 3397.

35. Peterson PH, Stockl D, Blaabjerg O, Pederson B, Birke-mose E, Thienpont L, Lassen JF, Kjeldsen J. Clin. Chem.1997; 43: 2039.

36. Okamoto M, Setaishi C, Horiuchi Y, Mashimo K, Moriya K,Itoh S. J. Clin. Endocrinol. Metab. 1971; 32: 673.

37. Kobayashi T, Lobotsky J, Lloyd CW. J. Clin. Endocrinol.Metab. 1966; 26: 610.


Documents

Direct measurement of urinary testosterone and epitestosterone conjugates using high-performance liquid chromatography/tandem mass spectrometry