Steroids 69 (2004) 461–471

Development of a highly sensitive and specific new testosteronetime-resolved fluoroimmunoassay in human serum

Jean Fieta,b,∗, Frank Gitona, Ibrahim Fidaac, Alain Valleixd,Hervé Galonse, Jean-Pierre Raynaudf

a Emi Inserm 03-37, Centre de Recherche Chirurgicale CHU Henri Mondor, Faculté de Médecine,8 rue du Général Sarrail 94010 Créteil France

b Laboratoire de Biochimie, Faculté de Pharmacie, 75006 Paris, Francec Laboratoire de Biochimie Hormonale, Hˆopital Saint-Louis, 1 Avenue Claude Vellefaux, 75475 Paris France

d Service des Molécules Marquées CEA Saclay, 91191 Gif/Yvette, Francee Laboratoire de Chimie Organique, Faculté de Pharmacie, 75006 Paris, France

f Université Pierre et Marie Curie, 4 place Jussieu, 75252 Paris, France

Received 26 February 2004; received in revised form 1 April 2004; accepted 14 April 2004

Available online 24 June 2004

Abstract

A new time-resolved fluoroimmunoassay (TR-FIA) of testosterone in serum is described, using a biotinylated testosterone tracer, witha long spacer arm between biotin and testosterone, coupled to the C3 of the testosterone: a biotinylaminodecane carboxymethyloximetestosterone.

This tracer affords a great sensitivity of the standard curve, because a amount of 0.3 pg of testosterone can be significantly mea-sured on the testosterone standard curve. The “functional” sensitivity is at least equal to 21 pg/ml of serum. The specificity of the assayis insured by a celite chromatographic step on new minicolumns before immunoassay. The variation coefficient of inter-series repro-ducibility measured on low and normal testosterone levels in untreated and testosterone treated hypogonadal men were between 2.17and 5.07%. The accuracy test, (overload and dilution tests) gave satisfying results. Moreover, in a comparison with GCMS, it appearedthat the correlation coefficient was 0.992 and no significant difference could be exhibited between the two methods. Consequently, thisspecific, sensitive reproducible and easy to use method is well suited to the measurement of testosterone in clinical and pharmacologicalconditions.© 2004 Elsevier Inc. All rights reserved.

Keywords:Time-resolved fluoroimmunoassay; Celite chromatography; Testosterone; Steroid

1. Introduction

Synthesis of different biotinylated steroid tracers al-lowed us to develop several steroid immunoassays us-ing time-resolved fluorometric detection of europium(Delfia Technology)[1–9]. These steroid immunoassaysexhibited usually higher sensitivity than those obtainedwith tritiated tracers, enzyme conjugated tracers andchallenged the immunoassays using125iodine labelledsteroids with the advantages of stable, non-radioactivetracer.

∗ Corresponding author. Tel.:+33 1 4981 3558; fax:+33 1 4981 3552.E-mail address:jean.fiet@libertysurf.fr (J. Fiet).

Although numerous publications concerning testosteroneimmunoassay measurements have been reported, to ourknowledge one publication using testosterone TR-FIA havebeen published[10]. We report a new testosterone TR-FIAwith a new biotinylated tracer with the ambition to obtaina maximal sensitivity and specificity, thanks to biotinylatedtracer with a long arm between biotin and the steroid andto new convenient minicolumns for celite chromatographybefore time-resolved fluoroimmunoassay (TR-FIA). More-over, an extensively controlled excellent reproducibility wasequally obtained.

This new testosterone TR-FIA has been developed forthe measurement of testosterone in untreated and treatedhypogonadal men, and in women consulting for hirsutism,acne or menstrual cycle irregularities.

0039-128X/$ – see front matter © 2004 Elsevier Inc. All rights reserved.doi:10.1016/j.steroids.2004.04.008

462 J. Fiet et al. / Steroids 69 (2004) 461–471

2. Materials, apparatus, methods and patients

2.1. Steroids and reagents

2.1.1. Commercial steroids, radioactive testosterone andchemical reagents

All steroids used as calibrators and for the synthesisof biotinylated conjugates were purchased from Steraloids(RI, USA). Tritiated (1,2,6,7-3H) testosterone came fromAmersham Biosciences (91898 Saclay, France). Chemi-cal reagents for conjugate synthesis were obtained fromSigma/Aldrich (BP 701, 38297 Saint-Quentin Fallavier,France).

2.1.2. Synthesis of biotinylated testosterone tracers(Scheme 1)

2.1.2.1. 10-Boc-amino-decylamine (1). To a cold (5◦C)solution of 1,10-diaminodecane (17.2 g, 100 mmol) in90 ml dioxane di-tert-tbutyldicarbonate [(Boc)2O] (5.04 g,25 mmol) in 30 ml dioxane was added. Stirring was pur-

NH NH

S

O

COOH

N

OH

H H

NH NH

S

O

O

NH

(CH2)10

NH3

H H

NH NH

S

O

O

NH

(CH2)n

NH3

H HO

O OH

N

OH

NH NH

S

O

O

NH

NH

O

O(CH2)n

H H

+

CF3COO_

3b

4

2a : n = 32b : n =10

5a : n = 35b : n = 10

1. iso-BuOCOCl

2. H2N(CH2)10NHBoc, 1

1. iso-BuOCOCl

2.

CF3COO

_

+

NH NH

S

O

O

NH

(CH2)10

NH

OO

C(CH3)3

H H

2b

CF3COOH

Scheme 1. Preparation of two biotinylated tracers.

sued overnight. The mixture was concentrated under re-duced pressure and extracted with CH2Cl2 75 ml, washedtwice with water and chromatographed on a silica gelcolumn using CH2Cl2/EtOH/Et3N 90:8:2 as eluent. Rf(CH2Cl2/EtOH/Et3N 75:20:5 v/v/v)= 0.42; yield =46%;1H NMR (CDCl3) δ: 1.3 (m, 16H, 8CH2); 1.5 (s, 9H, Boc);2.6 (t, 2H, CH2–NH2); 3.0(q, 2H, CH2–NH–Boc); 4.5 (bs,1H, NHBoc); 7.2(bs, 2H, NH2).

2.1.2.2. 10-Boc-amino-decylbiotinylamide (2b). To acold (5◦C) solution of biotin (3.14 g, 11.5 mmol) inDMSO (12 ml), was added tributylamine (2 ml) andiso-butylchloroformate (1.50 ml, 11 mmol) in 3 ml diox-ane. After 10 m stirring, 10-Boc-amino-decylamine (2.72 g,10 mmol) and triethylamine (2 ml) in dioxane was added.After stirring overnight the mixture is poured in 100 ml ofice water. The precipitate is filtered and washed twice with10 ml ice water. The solid is dried overnight in a desicca-tor containing P2O5. Rf (CH3CN/EtOH/Et3N 5:3:1 v/v/v)= 0.7; yield= 90%; m.p.= 120–125◦C. 1H NMR (CDCl3)δ: 1.2 (m, 16H, 8CH2); 1.5 (s, 9H, Boc); 1.6 (m, 6H, 3CH2);

J. Fiet et al. / Steroids 69 (2004) 461–471 463

2.8 and 3d and dd (CH2–S); 4.2 and 4.4(2m, 2H, 2CH–N);4.5 (t, 1H, NHCO); 5.5 and 5.9 (2bs, 2H, 2NH–biot).

2.1.2.3. Biotinylaminodecyl-ammonium trifluoroacetate(3b). To a solution of2b in CH2Cl2 30 ml was added an-hydrous trifluoroacetic acid 10 ml. The mixture was stirredat rt for 2 h. After evaporation of the solvent, the remain-ing oil crystallized upon trituration twice with 20 ml ofcyclohexane and 20 ml of Et2O. The solid was filtered andwashed with Et2O. Yield = 95%; m.p.= 145–152◦C; 1HNMR (CDCl3) δ: 1.4–1.6 (m, 16H, 8CH2); 2.05 and 2.15(2t, 4H, 2CH2); 2.8 (m, 6H, 4CH2); 3(m, 3H, CH2S andCHS); 4.10 and 4.30 (2m, 2H, 2CH–N); 7.5 and 8.5 (2bs,7H, 4NH and NH3).

2.1.2.4. Final synthesis of biotinylated tracers (5a–b). Thepreparation of biotinylaminopropylammonium trifluoroac-etate3a has been previously reported[1]. Iso-butylchloro-formate (35�l, 0.22 mmol) was added to a cold (5◦C)solution of testosterone-3-carboxymethyloxime4 (0.1 g,0.25 mmol) and triethylamine (40�l, 0.3 mmol) in diox-ane. After 5 min of stirring, the solution was added to asolution of biotinylaminoalkyl–ammonium trifluoroacetate3a–b (0.25 mmol) and triethylamine (50�l, 0.45 mmol) in2 ml DMSO. After 4 h of stirring, the mixture was dilutedin 10 ml cold water. The tracers, which precipitated, werefiltered and washed with 2 ml of cold water and purifiedby column chromatography using EtOAC/MeOH (99.5:0.5v/v) as eluent.

5a: yield 46%; m.p. 147–152◦C; 1H NMR (CDCl3, TMS)δ: 0.72 (s, 3H, 18-CH3); 0.97 and 1.02 (2s, 3H, 19-CH3);2.62 and 2.82 (d and dd, 2H, CH2–S); 3.09 (m, 1H, CH–S);3.25 and 3.35 (2q, 2× 2H, 2CH2NCO); 4.25 and 4.45 (2m,2H, CH–N); 4.41 and 4.48 (2s, 2H, O–CH2CON); 5.18 and5.95 (2bs, 2H, NH); 5.69 and 6.32 (2s, 1H, 4-H E and 4-HZ); 6.60 and 6.75 (2t, 2 CH2–NH–CO).

5b: yield 63%; m.p. 112–116◦C; 1H NMR (CDCl3, TMS)δ: 0.72 (s, 3H, 18-CH3); 0.97 and 1.02 (2s, 3H, 19-CH3);2.62 and 2.82 (d and dd, 2H, CH2–S); 3.09 (m, 1H, CH–S);3.25 and 3.35 (2q, 2× 2H, 2CH2NCO); 4.25 and 4.45 (2m,2H, CH–N); 4.41 and 4.43 (2s, 2H, O–CH2CON); 4.70 and5.32 (2bs, 2H, NH); 5.65 and 6.20 (2t, 2CH2–NH–CO), 5.75and 6.35 (2s, 1H, 4-H E and 4-H Z).

These two biotinylated tracers5a and5b were a mixtureof two E/Z stereoisomers. We separated the5b stereoisomersusing HPLC procedure: HPLC Shimadzu SCL 10AS Appa-ratus, Hypurity C18; 250× 4.6 column; acetonitrile/H2O:45/55; flow, 1.5 ml/min, at 35◦C; detection, UV240 nm.

2.1.3. Preparation of the biotinylated testosteronesolutions (tracers)

The stock solutions of biotinylaminopropane car-boxymethyloxime testosterone (5a reagent, see above) andbiotinylaminodecane carboxymethyloxime testosterone (5breagent, see above) tracers were prepared by dissolving,respectively, 4mg of5a reagent and 5 mg of5b reagent in

100 ml of ethanol. The biotinylated tracer intermediary so-lutions were prepared by diluting the stock solutions 1/100in ethanol. Stock and intermediary ethanolic solutions werekept at+4◦C.

2.1.4. Preparation of the anti-testosterone antibodyThe anti-3-carboxymethyloxime-testosterone antibody

was prepared in rabbits according to Vaitukaitis’ methodand described previously[11]. Specificity was studied byassessing cross-reactivities for a displacement of 50% ofthe biotinylated tracers by various steroids.

2.1.5. Special reagents, devices and apparatus for TR-FIA(described previously)[5]

Microtitration plates: 12 × 8 NUNC (Ref. 1244-550Perkin-Elmer 91945 Courtaboeuf France).

Goat anti-rabbit antibody: Valbiotech (75010 France;4 mg/2 ml) for coating.

Washing solution: physiological saline solution withTween 20 (0.1%).

Eu-labelled streptavidin: 0.1 mg/ml, 2.5 ml (Ref. 1244-360, Wallac), diluted 1/1000 in the following buffer: BSA5 g/l + Tween 40, in physiologic saline solution+ Na azide(0.5%), diethylene triamine penta-acetic acid 15 mg/l, ad-justed to pH 7.8 with Tris–HCl.

Delfia Enhancement solution (Ref 1244-105, Perkin-Elmer).

Coating buffer: 0.05 M, pH 9.6. Anhydrous Na carbonate(Na2CO3) 1.55 g Na hydrogen carbonate (NaHCO3) 2.97 gqsp 1000 distilled water, adjusted to pH 9.6 with dilutedHCl.

Saturation solution: pH 7.4 0.05 M. Disodium hydrogenphosphate (Na2HPO4·2H2O) 7.2 g, sodium di-hydrogenphosphate (NaH2PO4·2H2O) 1.3 g, distilled water qsp1000 ml. Dissolve bovine serum albumin 12 g (SigmaA-9647) and Na azide 2 g (Merck, Ref. 106-688), kept at+4◦C.

Assay buffer: pH 7.4 0.05 M. Sodium chloride (NaCl)9 g, disodium hydrogen phosphate (Na2HPO4·2H2O) 7.2 g,sodium di-hydrogen phosphate (NaH2PO4·2H2O) 1.3 g,distilled water qsp 1000 ml. Dissolve bovine serum albu-min 3 g (Sigma A-9647) Na azide 2 g (Merck, Ref. 106-688).

Fluorometer: time-resolved fluorescence was measuredwith a 1234 Delfia Fluorimeter (Perkin-Elmer, 91945Courtaboeuf, France). A model 1296-024 platewash device(Perkin-Elmer, Wallac) was also used.

GC/MS: GC/MS analysis were performed with HP6890for the gas chromatography and with HP5973 for the massspectrometry.

2.1.6. Coating and saturation of microtiter platesMicrotiter plate wells were coated by adding 250�l

of goat anti-rabbit antibody, diluted 1000-fold in coatingbuffer. After washing three times with a washing solu-tion, free well sites were saturated by adding 300�l of the

464 J. Fiet et al. / Steroids 69 (2004) 461–471

saturation solution. The microtitration plates were coveredwith sealing tape (Corning 430454) and kept at+4◦C.

2.1.7. Testosterone calibrator solutions for establishingtwo standard curves—a classical standard curve(CSC)—and a sensitized standard curve (SSC)

Mother solutions containing 10 mg of testosterone per100 ml of pure ethanol, were used to prepare alcoholicdaughter solutions 100 times less concentrated. Theseethanolic standard solutions were kept at+4◦C and usedto prepare extemporaneously the calibrator solutions in theassay buffer. The amounts of testosterone added into thewells of the microtitrate plates to establish the CSC werein pg/well (pmol): 600 (2.08), 300 (1.04), 150 (0.52), 75(0.26), 37.5 (0.13), 18.75 (0.06), 9.38 (0.03), 4.69 (0.016).For establishing the SSC, the quantities of testosteroneput into the wells were in pg/well (pmol): 75 (0.26), 37.5(0.13), 18.75 (0.06), 9.38 (0.03), 4.69 (0.016), 2.34 (0.008),1.13 (0.004), 0.58 (0.002).

2.1.8. Reagents solvents and devices for extraction andchromatography

Iso-octane, cyclohexane, ethylacetate, dichloromethaneand ethylene glycol were supplied by Merck (Nogent-sur-Marne, France), Carlo-Erba (Rueil Malmaison, France) andProlabo. Celite, i.e. “Celite analytical filter aid” with verysmall pore size (median pore size 2.5�m) was from Lom-poc, California 93438-0518, USA. This celite was condi-tioned by treating with cyclohexane extensive washing, driedthen kept at 100◦C, before packing into chromatographicminicolumns. Minicolomns (Supelco, St. Quentin-FallavierCedex, France), Visiprep apparatus (St. Quentin-FallavierCedex, Ref. 57250-U).

Fig. 1. Celite-ethylene glycol separation chromatography of testosterone from other steroids, using solvents of increasing polarity. Percentageof recoveriesof the following 3H tritiated steroids: ( ) DELTA4 = ∆4-androstenedione; () ANDRO = androsterone; () DHEA = dehydroepiandrosterone; () DHT= 5α-dihydrotestosterone; () TESTO = testosterone; () 17OHP = 17-OH-progesterone; () 11BETA = 11βOH-∆4-androstenedione; () ADIOL= 5α-androstan-3α,17β-diol; ( ) DELTA5 = 5-androsten-3β,17β-diol.

2.2. Procedures for testosterone assays in serum

2.2.1. TR-FIA procedure

2.2.1.1. Steroid extraction step from serum.In order tomonitor losses occurring during the extraction, chromatog-raphy and redissolution steps before immunoassay, 0.1 mlof a aqueous tritiated testosterone solution (containing4000 dpm) was added to 0.1–1 ml of serum in 16 ml propy-lene tubes (Sarstedt Ref. 55515). After 30 min of incubationwith intermittent shaking, extraction was carried out with10 ml of a mixture of ethylacetate/cyclohexane 50/50 (v/v),by mixing for 2 min with a multivortex. Centrifugate at2500 t/min× 5 min. Freeze the aqueous lower phase andtransfer the upper organic layer to 14 ml glass tubes. Evap-orate at 37◦C under a stream of filtered air. We obtained adry extract containing the steroids.

2.2.1.2. Chromatographic separation steps (Fig. 1). Thedry extracts were redissolved in isooctane (1.5 ml), vor-texed 30 s, ultrasonicated 5 min and vortexed again for 30 s.This organic solution was layered onto celite chromato-graphic minicolumns as already reported[9] and passedthrough with the aid of a negative pressure (−5 kg Pa), us-ing Visiprep apparatus. This negative pressure was constantduring passing the elution chromatographic solvents. Eachaddition of solvents into the chromatographic columnswas performed while the vacuum pump was stopped. Twomilliliter of pure isooctane was added into the columnand this volume was not collected. Five milliliter of pureisooctane were added and collected if necessary (this frac-tion contains androstenedione.). Five milliliter of a mixtureof isooctane+ dichloromethane 88/12 (v/v) were added

J. Fiet et al. / Steroids 69 (2004) 461–471 465

and collected or not (this fraction contains DHT). At lastthe addition of 5 ml of a more polar mixture of isooc-tane+ dichloromethane 80/20 (v/v) allowed the elution oftestosterone. This collected organic chromatographic elu-tion fraction was evaporated to dryness. The dry residue isredissolved in assay buffer (0.5–1 ml) with the aid of vor-texing and sonication. In this aqueous solution, testosteronewas immunoassayed by TR-FIA.

2.2.1.3. TR-FIA of testosterone.The assays were donein duplicate. The assay was carried out by adding to eachcoated well, 0.1 ml of the testosterone aqueous chromato-graphic solution or 0.1 ml of standards, then 0.05 ml ofthe appropriate dilution of the labelled biotin-testosteronetracer and 0.05 ml of adequate anti-3-carboxymethyloxime-testosterone/BSA. Microtitration plates were incubatedwhile being agitated 350 rev/min, at room temperature,for 2 h. The immunoreaction was stopped by washing thewells three times with washing solution. Then, 0.2 ml of aEu-labelled steptavidin was added and the plates were agi-tated for 20 min at 350 rev/min before washing three times.The Eu was dissociated by adding 0.2 ml of enhancementsolution to each well, then agitated 250 rev/min for 20 minbefore measuring time-resolved fluorescence. (the appropri-ated dilution of the biotin-testosterone tracer on one handand of the anti-3-carboxymethyloxime-testosterone/BSA onthe other hand were reported in results, § 3.2.1).

2.2.2. RIA procedure

2.2.2.1. Competition RIA of testosterone with tritiatedtestosterone as tracer[11] . After extraction followedby celite chromatographic on 5 ml Kimble glass pipettesas columns, the aqueous solution of testosterone wasassayed by RIA using the same anti-3-carboxymethyloxime-testosterone, tritiated testosterone as tracer and a scintilla-tion proximity reagent (RPN 140) (Amersham/Biosciences,91898, Orsay, France) for separation and counting as pre-viously described[11].

2.2.2.2. Competition RIA of testosterone with125

iodinetestosterone as tracer (IM 1087).Commercial Kit ofBeckman/Immunotech, Marseille Cedex 09, France: briefly,a chromatographic step was carried out before testosteroneimmunoassay, only in women.

2.2.3. GC/MS procedure.[12]Briefly, 1 ml of serum was extracted, using deuterated

testosterone as internal standard. The derivatisation reagentwas pentafluorobenzoyl chloride (Sigma–Aldrich).

2.3. Patients

Testosterone TR-FIA was developed in the aim, firstly toinvestigate the androgenic hormonal status of hypogonadal

men before and after testosterone treatment, and secondlyto study women suffering from hirsutism, acne or menstrualcycle irregularities. The clinical and pharmacological resultswill not be reported in the present paper.

3. Results

3.1. Synthesis of the tracers

Acylation of biotinylaminopropylammonium and bi-otinylaminodecylammonium with testosteronene-3-CMOafforded two biotinylated tracers5a and 5b. The E and Zstereoisomers of the5b tracer were separated in two distinctpeaks with a good resolution. All the analytical validationand clinical assays were carried out using the mixture ofthe5b tracer (E and Z).

3.2. Characterization of the anti-testosterone antiserum

3.2.1. Titers of the antiserum anti-testosterone-3CMO-BSAused

Dilution of the anti-T-3-CMO/BSA used in the RIAwas 2 × 10−4 [11]. In the standard TR-FIA methodol-ogy, we obtained a CSC, using a final anti-testosteroneantiserum dilution of 0.16× 10−4 and a quantity ofthe biotinylaminodecane-3-CMO-testosterone tracer (5bScheme 1, E and Z stereoisomeres mixture) of 192 pg/well.In the sensitized TR-FIA methodology, we chose a greaterdilution of antiserum, 0.66× 10−5 and 55 pg/well ofthe same tracer. With the biotinylaminopropane-3-CMO-testosterone tracer (5a Scheme 1), the dilution of antiserumwas 0.5× 10−5, and the quantities of tracer was 50 pg/well.

3.2.2. Specificity of the anti-testosterone-3-CMO/BSAThe specificity of the anti-testosterone-3-CMO/BSA was

previously reported using tritiated testosterone as tracer[11]. The use of biotinylated tracer, biotinylaminodecanecarboxymethyloxime testosterone (5b tracer) or biotiny-laminopropane carboxymethyloxime testosterone (5a tracer)did not modify significantly the main cross-reactivitiesstudied (delta-4-androstenedione 1.5%, dihydrotestos-terone (DHT) 29%, dehydroepiandrosterone (DHEA)0.5%, 5�-androstane, 3�,17�-diol 0.1%, 5�-androstane,3�,17�-diol 0.1%, 17-OH progesterone<0.001%, an-drosterone<0.1%. Equally the use of one of the twostereoisomeres E and Z of5b tracer, did not modify the crossreactivities obtained with the mixture of the E and Z isomers.

3.3. Establishement of testosterone standard curves

The amounts of T (expressed in pg) put into the tubesand into the wells of the microtitration plates necessary forestablishing, respectively, the immunocompetition standardcurves for RIA and TR-FIA method were: 600, 300, 150,

466 J. Fiet et al. / Steroids 69 (2004) 461–471

75, 37.5, 18.75, 9.38 and 4.69, which correspond to 2.08,1.04, 0.52, 0.26, 0.13, 0.065, 0.032 and 0.016 pmol of T.The quantities of testosterone in the wells for establishingthe TR-FIA sensitized standard curve were 75, 37.5, 18.75,9.38, 4.69, 2.34, 1.17 and 0.59 pg.

3.3.1. Testosterone standard curves

3.3.1.1. Percentages of mean bindings.The RIA andTR-FIA standard curves are presented inFig. 2. Theywere established from 12 intra-assay determinations. Themeans (±SD) of the binding for the following calibrators(pg/well) 4.69, 9.38, 18.75, 37.50, 75, 150, 300 and 600were, respectively, 91.2 (6.9), 85.6 (3.5), 78.1 (3.4), 66.5(2.2), 54.1 (2.6), 44.7 (2.7), 30.1 (1.5) and 18.8 (1.1) forthe RIA standard curve (with a tritiated tracer) and 80.5(1.1), 67.7 (1.6), 48.2 (1.3), 27.6 (0.6), 14.2 (0.4), 8.2 (0.2),4.5 (0.1) and 3.05 (0.1) for the classical TR-FIA standardcurve (CSC; established using the5b tracer) and 85.5 (1.5),78 (1.5), 62.2 (1.1), 42.1 (0.9), 27.2 (0.9), 19.1 (0.5), 16.4(0.2) and 15.6 (0.3) (established using the5a tracer). Themeans (±SD) of the binding for the following calibrators(pg/well) 0.56, 1.17, 2.34, 4.69, 9.38, 18.75, 37.50 and 75of the sensitized TR-FIA standard curve (SSC) were 92.2(1.2), 87.5 (0.9), 76.1 (1.0), 59.3 (0.8), 36.6 (0.5), 20.6 (0.3),11.4 (0.2) and 6.7 (0.1). The means (±SD) of the bindingfor the Beckman-Immunotech Commercial Kit calibrators(pg/well) 4.70, 17.5, 75, 250 and 1000 were 97.1 (2.31),81.1 (1.9), 50.9 (1), 31.1 (0.8) and 15.1 (0.5).

Fig. 2. TR-FIA sensitized standard curve (); TR-FIA classical standard curve (�); RIA (125I testosterone as tracer) standard curve ofBeckman-Immunotech Kit (); RIA (3H testosterone as tracer) standard curve (�); IC50 of the standard curves.

3.3.1.2. Radioactivity and fluorescence unit ranges.Themean range of the radioactivity of the RIA standard curve(with tritiated tracer) was between 5440 dpm for the zerostandard calibrator and 1012 dpm for the 600 pg calibrator(with a mean background of 180 dpm)[11]. The mean rangeof the arbitrary fluorescence units of the TR-FIA standardcurve was between 220,200 for the zero calibrator (absenceof analyte), and 6650 for the last calibrator (600 pg; with abackground of less than 600 fluorescence units). The rangeof the sensitized TR-FIA standard curve was between 37.200units of fluorescence for the zero calibrator and 2500 for thelast calibrator (75 pg; with a background of less than 250).

3.3.1.3. Sensitivity of the testosterone standard curves.They were established as the quantities of testosterone pertube or per well which displaced 20, 50 and 80% of thetracers. These displacements were obtained for the RIAstandard curve by 21.5, 110 and 478 pg of testosterone;for the classical TR-FIA standard curve by 4.70, 16.8 and65.0 pg of testosterone; and for the TR-FIA sensitizedstandard curve by 1.5, 6.9 and 17 pg of testosterone.

In the standard curve of the Beckman-ImmunotechTestosterone Kit, the 20, 50 and 80 displacements of the125iodine testosterone tracer were obtained with 18, 75 and650 pg, respectively.

3.3.1.4. Lowest detection limits of testosterone from thestandard curves. The least quantities of T that signifi-cantly displaced the tracers, i.e. equivalent to B0 or mean

J. Fiet et al. / Steroids 69 (2004) 461–471 467

binding of tracer in the absence of analyte-3SD were foundto be 4.10 pg/well on the RIA standard curve (with a tritiatedtracer), 0.7 pg/well on the classical TR-FIA standard curveand 0.35 pg/well on the sensitized TR-FIA standard curve.

3.4. Validation of the testosterone TR-FIA method

3.4.1. Specificity of the testosterone TR-FIADHT which cross-reacts with anti-testosterone antiserum

was completely eluted in an iso-octane/DCM 12% chro-matographic fraction and no DHT is present in the more po-lar chromatographic elution fraction, iso-octane/DCM 20%,in which the testosterone was eluted. Consequently, DHTdoes not interfere at all in the testosterone assay. Moreover,the �4-androstenedione is eluted before DHT, the DHEAand androsterone are eluted in the DHT elution fraction,the more polar steroids 11-keto, 11-hydroxy-testosterone,11�-hydroxy-�4-androstenedione and androstanediolswere eluted in more polar elution fractions after the testos-terone elution fraction. 17-OH progesterone, a pregnenesteroid present in the testosterone elution fraction do notcross react with the anti-testosterone antibody.

3.4.2. Recovery of the minute doses of tritiated testosteroneadded to the serum samples for monitoring the extraction+ celite chromatography steps

The mean recovery of minute doses of tritiated testos-terone added to each serum samples to monitor the recoveryafter the two-step extraction+ chromatography was equalto 82.8%± 6.7 (n = 350). The individual result of recoveryof each assay was taken into account to calculate the serumtestosterone concentration.

3.4.3. Detection limits

3.4.3.1. Detection limits in serum calculated from the low-est detection limits of testosterone on the standard curves.The theoretical detection limit of testosterone in the serumdepends on the methodology used. With 1 ml of extractedserum, 1 ml of assay buffer used to redissolve steroid afterchromatography, 0.1 ml of aqueous chromatographic solu-tion assayed by TR-FIA (§ 2.2.1), a mean recovery of testos-terone of 82.8% (§ 3.4.2), and the lowest detection limits oftestosterone from the TR-FIA standard curves (§ 3.3.1.4),the theoretical calculated detection limits of testosterone inserum was 8.5 pg/ml using the methodological conditions toobtain the classical testosterone standard curve and 4.3 pg/mlusing the sensitized methodological conditions and the sen-sitized testosterone standard curve. Comparatively, the the-oretical detection limit using the RIA was 49 pg/ml.

3.4.3.2. The quantification limit or “lower limit ofquantification” (LLOQ) or “functional” sensitivity. LLOQwas determined in assaying 20 times, four sera whose Tconcentrations were, respectively, 60, 40, 20 and 15 pg/ml.

We obtained a mean level of 44 pg/ml (CV= 11.2%) forserum containing theoretically 40 pg/ml using the classi-cal methodologic conditions (consequently the LLOQ was<44 pg/ml) and a mean value of 21 pg/ml for a serum con-taining 20 pg/ml (CV 7.6%) using the sensitized method-ologic conditions (consequently the LLOQ was<21 pg/ml).

3.4.4. Reproducibility

3.4.4.1. Reproducibility of duplicate.The CV of duplicateTR-FIA measurements were always below 2%.

3.4.4.2. Intra-series precision.Four pools of sera in whichthe testosterone concentrations (ng/ml) were 0.15, 0.30, 0.60and 1.20 were aliquoted in 10 samples for each level (con-sequently 40 samples in all). Each sample (in which theconcentration was unknown to the technician) was assayedin the same run according to the sensitized methodology (n= 40 samples). The same experience, but using the standardmethodology, was performed on sera pools in which testos-terone levels (ng/ml) were 2, 4, 8 and 16. The results werereported inTable 1.

3.4.4.3. Inter-series precision.The testosterone inter-assay precision was determined in 80 serum assay seriesof hypogonadal men (three control sera (ng/ml): 0.75, 1.50and 3.00 were assayed in each series) and in 17 assay seriesof testosterone treated hypogonadal men sera (control sera(ng/ml): 2.50, 4.00 and 6.00). The means± SD (ng/ml),(CV, %) were, respectively, 0.755± 0.037 (4.883); 1.53±0.078 (5.070); 3.024± 0.126 (4.179) and 2.456± 0.112(4.580); 4.084± 0.145 (3.555) and 6.004± 0.131 (2.174).In each run, a steroid-free charcoal-stripped serum gaveundetectable testosterone concentration.

3.4.5. Accuracy

3.4.5.1. Recovery experiments.Four sera (two male andtwo female sera) were overloaded with four concentrationsof testosterone. The testosterone concentrations of the over-loaded sera were assayed by TR-FIA and the recoveries

Table 1Intra-assay precision

Theoretical testosteronelevels (ng/ml)

Measured testosterone levels(mean± S.D.; ng/ml)

CV (%)

0.15 0.16± 0.01 6.840.30 0.313± 0.02 5.010.60 0.613± 0.02 4.001.20 1.145± 0.04 3.222.00 2.016± 0.06 3.164.00 4.00± 0.10 2.468.00 8.056± 0.33 4.15

16.00 15.523± 1.00 6.47

TR-FIA measurement of 80 samples containing 8 levels of testosterone.Each level was assayed in 10 samples in the same run.

468 J. Fiet et al. / Steroids 69 (2004) 461–471

Table 2TR-FIA measurements of testosterone (ng/ml) in undiluted and diluted sera (three female serum-levels 1.2.3 and three male serum-levels 4.5.6)

Serum dilutions T level 1 T level 2 T level 3 T level 4 T level 5 T level 6

Undiluted 0.66 0.76 1.06 2.83 4.88 5.36Dilution 1/2 0.68 0.72 1.00 2.72 4.82 5.32Dilution 1/4 0.64 0.64 1.04 2.92 4.72 5.52Dilution 1/8 0.64 0.80 0.96 2.96 4.80 5.20

calculated. The analytical recoveries of T added to serawere between 90 and 110%.

3.4.5.2. Dilution test. Testosterone concentrations weremeasured in the serum of six patients, three hyperandro-genic women, and three testosterone treated hypogonadalmen. Testosterone was assayed at various dilutions rangingfrom undiluted to eight-fold diluted. The woman sera wereassayed using sensitized methodology and the men serawith the classical methodology. The results are reported inTable 2.

3.4.6. Comparison of results with RIA method and withGC/MS

3.4.6.1. Comparison of woman testosterone levels usingTR-FIA sensitized methodology and RIA[11] . The testos-terone levels were assayed simultaneously in 67 samplesof consecutive women consulting for hirsutism or acne.

0

1

2

3

4

5

6

7

8

TR

-FIA

(ng

/ml)

0 1 2 3 4 5 6 7 8

GC-MS (ng/ml)

Y = 1.064 . X - 0.036

R = 0.983 2

Fig. 3. Comparison of testosterone serum levels measured by GCMS (X-axis) and by TR-FIA (Y-axis) in 13 untreated and 13 testosterone-treatedhypogonadal men, and in 10 women suffering from hirsutism, acne or period disturbances. Regression curve, square of the correlation coefficientR2.

The two compared methods included a chromatographicstep. The equation of the regression curve was:Y = 0.961X+ 0.032 (Y: TR-FIA; X: RIA), with R2 = 0.971 (Fig. 3).

3.4.6.2. Comparison with GC/MS.Sample sera of un-treated hypogonadal men (n = 13) sera samples of testos-terone treated hypogonadal men (n = 13) and sera samplesof hyperandrogenic women (n = 10) were simultaneouslyassayed by TR-FIA and GC/MS (GCMS; LLOQ= 50 pgof testosterone per ml of serum). The equation of the re-gression curve wasY = 1.0644X − 0.0358 (Y: TR-FIA; X:GCMS), with R2 = 0.9832. There is no significant differ-ence between the two paired series of GCMS and TR-FIAresults.

3.4.7. Stability of the tracerThe stock solution and intermediary alcoholic tracer,

biotin-aminodecane 3-CMO-testosterone, have been storedfor 24 months so far at+4◦C with no loss of activity.

J. Fiet et al. / Steroids 69 (2004) 461–471 469

4. Discussion

The aim of this work was to establish a sensitive, spe-cific and accurate non-isotopic method to assay serumtestosterone in testosterone treated and untreated hypog-onadal men and in hyperandrogenic women, to comparewith GCMS and with a previously published RIA methodapplied in hyperandrogenic women[11].

According to some steroid non-isotopic assay methodsusing biotinylated tracer[1–9], we synthesized two biotiny-lated testosterone tracers with different length of the spacerarm between biotine moiety and steroid (tracera with 3Cand tracerb with 10C). The two tracers were studied andusing HPLC, we separated the two stereo isomers E and Zof the tracerb.

The use of E or Z stereoisomer of biotinylaminodecanecarboxymethyloxime testosterone (tracerb) instead of themixture of the two, did not increase significantly the sensi-tivity of the testosterone standard curves and did not signifi-cantly decrease the cross-reactivities of the anti-testosteroneimmunserum. This was at variance with previously reportsfor 21-deoxycortisol[13] radioimmunoassays, in whichthe sensitivity of the standard curve was depending onthe geometric isomerism of the125iodine 21-deoxycortisoltracers.

The aim of the synthesis of the two tracersa andb wasto reach the maximal sensitivity of the testosterone standardcurves. It was reported that the chemical structure of thespacer arm between, biotin and steroid has a relevant influ-ence on the process of steroid antibody recognition[14], andthat generally the sensitivity increases as the length of thespacer arm increases[15]. Indeed it was clearly shown inthe chemiluminescence immunoassay of testosterone usinga biotinylated tracer[16], that the sensitivity of the testos-terone standard curve greatly increased with the length ofthe spacer arm between biotin and the steroid, and that onlya C11 alkyl spacer arm allowed to obtain a satisfying stan-dard curve. Spacer arms of three or six carbon atoms at-tached to carbon 7 of the testosterone were not suitable forestablishing testosterone standard curves.

In our work, we report a slightly greater sensitivity of thestandard curve using the biotinylated tracer with the longerspacer arm (10 carbon atoms in tracerb) compared to thebiotinylated tracer with three carbon atoms (tracera). How-ever, this difference of sensitivity was not as striking as thosedescribed previously in testosterone chemiluminescence as-say. This discrepancy may be attributed to the different sitesof attachment of the spacer arm on testosterone backbone-onthe seven carbon in the testosterone chemiluminescence im-munoassay, and on the three carbon in our report. Indeed,another testosterone TR-FIA using also biotine-testosteronetracer with a spacer arm of six carbon atoms attached onC3[10] exhibited also fair sensitivity performances in salivatestosterone immunoassay.

In this work, all the validation and clinical assays werecarried out using the tracerb.

The TR-FIA testosterone assay sensitivity was adjustedby varying the tracerb concentration and anti-testosteroneantibody dilution. Thus, we obtained on one hand, a CSCwith a significant least detectable dose of 0.70 pg/well, and a50% tracer displacement (IC50) of 18 pg/well (Fig. 2), and onthe other hand, a SSC with a significant least detectable doseof 0.35 pg/well and an IC50 of 6 pg/well. This sensitivityappeared superior to those given by testosterone RIA usingtritiated testosterone tracer[11], and even125iodine labelledtestosterone tracer (Fig. 2).

In taking into account these least detectable doses onthe testosterone standard curves, we can calculate detectionlimits in serum. They were 9.3 and 4.7 pg/ml using method-ologies with the CSC and the SSC, respectively. Thesedetection limits were either lower or in the same range ofthose of other non-isotopic testosterone assays previouslypublished[16,10].

In practice, the functional sensitivity was strictly lower(respectively, 44 pg/ml with CSC (with a CV= 11.2%),and 21 pg/ml with the SSC (CV= 7.6%), than those of theonly reported in chemiluminescence testosterone assay[16](equal to 230 pg/ml with a CV= 20%).

The CVs of intra-series precision measured 10 times in 4low level control sera using the more sensitive methodology(with SSC) and in 4 high level control sera using the classi-cal methodology (with CSC) were between 3.16 and 6.84%consequently lower than those usually reported, intra-seriesCV between 7.7 and 9.1%[16], and between 8.9 and 14.6%[10].

The inter-series precision measured in 80 series of hy-pogonadal sera with low level of testosterone and in 17 se-ries of testosterone treated hypogonadal men sera with highor normal level of testosterone was expressed by CVs be-tween 2.17 and 5.07%. The CV reported in previous pub-lications of non-isotopic methods, were rather higher, withCV between 7.9 and 8.7%[10] and between 12.8 and 15.9%[16].

The specificity of our testosterone immunoassay wasbased on a celite-ethylene glycol chromatographic step car-ried out before the immunoassay. The use of celite-ethyleneglycol as a chromatographic phase has been developedsince long time ago[17]. Testosterone was eluted in asolvent volume whose polarity was chosen so as to com-pletely separate testosterone from DHT which was elutedin a less polar elution fraction. As reported in § 3.4.1.,the less polar steroids androsterone,�4-androstenedione,DHEA were eluted before testosterone elution fractionwhile more polar such as androstanediols, 11-OH and11-ketotestosterone and 11�-HO-�4-androstenedione wereeluted after testosterone elution fraction. According to ourexperience, the only steroid which was eluted simulta-neously with testosterone was the 17-OH progesterone.However, this steroid did not cross-react at all with theanti-testosterone-3-CMO-BSA, and did not interfere in ourtestosterone immunoassay. The use of a celite with verysmall median pore size (2.5�m) allowed to reach such a

470 J. Fiet et al. / Steroids 69 (2004) 461–471

good resolution. The chromatographic step before steroidimmunoassays, today, is no more largely widespread, be-cause it is considered to be cumbersome and time consum-ing. Thus, numerous steroids, including testosterone, aremeasured by immunoassays, RIA particularly, directly onserum or only after an extraction step. Consequently, it re-sults in false high serum testosterone levels, particularly inwomen[18]. The use of polypropylene minicolumns filledwith celite+ ethylene glycol, and the migration of differentsolvents through celite columns using a slight depressionin Visiprep has increased considerably the practicabilityof this chromatographic step compared to previous celitechromatographic procedures[11]. So 48 samples using2 Visipreps can be easily processed simultaneously, andthe testosterone containing fractions will be collected inless than 1 h and half. Moreover, the recovery of minutedoses of tritiated testosterone added to each serum sam-ples to calculate the testosterone level was high (82%) andhomogenous.

As a consequence of the good precision specificity andsensitivity, we could expect a good exactitude. Indeed, thestandard tests exploring the exactitude quality, the dilutionand recovery tests were satisfying.

Moreover, we compared the testosterone levels in con-secutive women consulting for hirsutism using this reportedTR-FIA and RIA previously published[11]. We found agood correlation.

At last, 36 sera were comparatively measured using ourmethod and GC/MS. No significant difference and no sys-tematic bias were found between the two assay methodsand the correlation was correct (R2 = 0.98). This resultis an additional argument for the good accuracy of ourtestosterone TR-FIA method. Considering several steroidhormones (including testosterone) measurements, similaranalytical performances were also recently reported[19]between radioimmunoassays and GC/MS.

In conclusion, this new non-isotopic testosterone TR-FIAhas the advantage on RIA to use only very little quantities ofradioactivity, the new minicolumns developed for the celitechromatographic step increase the practicability and this newTR-FIA present the analytical qualities required for a spe-cific, sensitive reproducible and accurate measurements oftestosterone in serum samples.

Acknowledgements

We thank Dr Alain Belanger (Molecular EndocrinologyLaboratory, CHUL Sainte Foy, Quebec, Canada) for com-parison with GC/MS.

References

[1] Boudi A, Fiet J, Galons H. Preparation and acylation of a newbiotinylation reagent. Synth Commun 1999;29:3137–41.

[2] Boudi A, Giton F, Galons H, Villette JM, Soliman H, Fiet J.Development of a plasma 17�-hydroxyprogesterone time resolved-fluorescence immunoassay involving a new biotinylated tracer.Steroids 2000;65:103–8.

[3] Fiet J, Boudi A, Giton F, Villette JM, Soliman H, MorineauG, et al. Plasma 21-deoxycortisol: comparison of a time-resolved fluoroimmunoassay using a biotinylated tracer with aradioimmunoassay using125iodine. J Steroid Biochem Mol Biol2000;72:55–60.

[4] Fiet J, Giton F, Boudi A, Soliman H, Villette JM, GalonsH. Plasma 17-OH pregnenolone: comparison of a time-resolvedfluoroimmunoassay using a new tracer 17-OH pregnenolone-3-oxyacetyl-biotine with a radioimmunoassay using125I 17-OHpregnenolone-3-hemisuccinate-histamine. Steroids 2001;66:81–6.

[5] Fiet J, Giton F, Boudi A, Vexiau P, Galons H, Raynaud JP. Highlysensitive and specific time-resolved fluoroimmunoassay (TR-FIA) ofcyproterone acetate and free cyproterone. J Steroid Biochem MolBiol 2001;75:315–22.

[6] Fiet J, Giton F, Boudou Ph JM, Villette JM, Soliman H, Morineau G,et al. A new specific and sensitive time-resolved fluoroimmunoassayof 11-deoxycortisol in serum. J Steroid Biochem Mol Biol2001;7:143–50.

[7] Fiet J, Giton F, Boudi A, Boudou Ph, Soliman H, Villette JM,Galons H. Development of a sensitive and specific new plasma 4-androstene-3,17-dione time-resolved fluoroimmunoassay (TR-FIA).Steroids 2001;66:609–14.

[8] Fiet J, Giton F, Auzerie J, Galons H. Development of a newsensitive and specific time-resolved fluoroimmunoassay (TR-FIA) ofchlormadinone acetate in the serum of treated menopausal women.Steroids 2002;67:1045–55.

[9] Ibrahim F, Giton F, Boudou P, Villette JM, Julien R, Galons H, etal. Plasma 11�-hydroxy-4-androstene-3,17-dione: comparison of atime-resolved fluoroimmunoassay using a biotinylated tracer with aradioimmunoassay using a tritiated tracer. J Steroid Biochem MolBiol 2003;1874:1–6.

[10] Tschöp M, Behre HM, Nieschlag E, Dressendörfer A, Strasburger CJ.A time-resolved fluorescence immunoassay for the measurement oftestosterone in saliva: monitoring of testosterone replacement therapywith testosterone busiclate. Clin Chem Lab Med 1998;36:223–30.

[11] Fiet J, Gosling JP, Galons H, Boudou P, Aubin PH, Bélanger A, et al.Coordinated radioimmunoassays for eight plasma steroids relevantof the investigation of hirsutism and acne in women. Clin Chem1994;40:2296–305.

[12] Bérubé R, Gauvin D, Blais M, Noreau L, Bélanger A. Déterminationof testosterone, 5-androstenediol-3,17-diol, dehydroepiandrosteroneand androstenedione in human serum using high-resolution gaschromatography and electron impact mass spectrometry. In:Proceeding of the 47th ASMS Conference on Mass Spectrometryand Allied Topics. Dallas, Texas; 1999 [Abstract].

[13] Fiet J, Villette JM, Galons H, Boudou PH, Burthier JM,Hardy N, et al. The application of a new, highly-sensitiveradioimmunoassay for plasma 21-deoxycortisol to the detectionof 21-hydroxylase deficiency. Ann Clin Biochem 1994;31:56–64.

[14] Hauptmann H, Paulus B, Kaiser T, Herdtweck E, Huber E, LuppaPB. Concepts for the syntheses of biotinylated steroids. Part I.Testosterone derivatives as immunochemical probes. Bioconjug Chem2000;11:239–52.

[15] Mikola H, Sundell AC, Hänninen E. Labeling of estradiol andtestosterone alkyloxime derivatives with a europium chelate for time-resolved fluoroimmunoassays. Steroids 1993;58:330–4.

[16] Luppa P, Brückner C, Schwab I, Hauck S, Schmidmayr S, BirkmayerC, et al. 7�-Biotinylated testosterone derivatives as tracers fora competitive chemiluminescence immunoassay of testosterone inserum. Clin Chem 1997;43:2345–52.

J. Fiet et al. / Steroids 69 (2004) 461–471 471

[17] Abraham GE, Buster JE, Lucas LA, Corrales PC, Teller RC.Chromatographic separation of steroid hormones for use inradioimmunoassay. Anal Lett 1972;5:509–17.

[18] Fitzgerald RL, Herold DA. Serum total testosterone: immunoassaycompared with negative chemical ionization gas chromatography–mass spectrometry. Clin Chem 1996;42:749–55.

[19] Dorgan JF, Fears TR, Mc Mahon RP, Friedman LA, Patterson BH,Greenhut SF. Measurement of steroid sex hormones in serum: acomparison of radioimmunoassay and mass spectrometry. Steroids2002;67:151–8.