Qualitative Determination of Synthetic Analogues of Insulin in Human

However, the misuse of insulin in sports has been reported often, and the international doping control system requires a reliable and robust assay to ...
16 downloads 10 Views 387KB Size
Download PDF
Anal. Chem. 2005, 77, 3579-3585

Qualitative Determination of Synthetic Analogues of Insulin in Human Plasma by Immunoaffinity Purification and Liquid Chromatography-Tandem Mass Spectrometry for Doping Control Purposes Mario Thevis,*,† Andreas Thomas,† Philippe Delahaut,‡ Alain Bosseloir,§ and Wilhelm Scha 1 nzer†

Institute of Biochemistry, German Sport University Cologne, Carl-Diem Weg 6, 50933 Cologne, Germany, Laboratory of Hormonology, Centre d’Economie Rurale, Rue du Point du Jour 8, 6900 Marloie, Belgium, and ZenTech, Parc Scientifique du Sart-Tilman, Avenue du Pre´ -Aily 10, 4031 Angleur, Belgium

Synthetic insulins such as Humalog Lispro, Novolog Aspart, or Lantus Glargine, are commonly employed for the treatment of insulin-dependent diabetes mellitus owing to convenient handling and fast or prolonged bioavailability. However, the misuse of insulin in sports has been reported often, and the international doping control system requires a reliable and robust assay to determine the presence or absence of related drugs prohibited by the World Anti-Doping Agency. Qualitative evidence of administered substances, which is preferably obtained by mass spectrometry, is of utmost importance. Plasma specimens of 2 mL were fortified with three synthetic insulin analogues and purified by immunoaffinity chromatography, and extracts were analyzed by microbore liquid chromatography and tandem mass spectrometry. Product ion scan experiments of intact proteins enabled the differentiation between endogenously produced insulin and its synthetic analogues by collisionally activated dissociation of multiply charged precursor ions. This top-down sequencing-based assay allows the assignment of individual fragment ions, in particular, of those comprising modifications that are originating from Ctermini of B-chains. Recoveries of synthetic insulins from plasma aliquots ranged from 91 to 98%, and detection limits were accomplished at 0.5 ng/mL for all target analytes. The determination of human insulin and analogues, such as porcine insulin or synthetic rapidly or slowly acting variants, for example Humalog (Lispro) and Novolog (Aspart) or Lantus (Glargine), is of high interest for different fields of analytical chemistry such as doping control and clinical and forensic analysis. The rapidly or slowly acting insulins are only slightly different from human insulin. For example, with Humalog, the proline and lysine residues at B28 and B29 (B-chain) are switched in position (Figure 1b), and in the case of Novolog, the proline residue at * Corresponding author. Phone: 0221-4982-7070. Fax: 0221-497-3236. Email: [email protected]. † German Sport University Cologne. ‡ Centre d’Economie Rurale. § ZenTech. 10.1021/ac050066i CCC: $30.25 Published on Web 04/30/2005

© 2005 American Chemical Society

position B28 is substituted by an aspartic acid residue (Figure 1c).1 These modifications reduce the protein’s affinity to self-associate to hexamers,2 because insulin’s biological activity is present only in monomers. Thus, the bioavailability of insulin after subcutaneous injection is accelerated, as compared to preparations of recombinant human insulin.3 In contrast, Lantus is a long-acting insulin analogue with an exchange of the A21 asparagine by a glycine (A-chain), and it contains two additional arginines at the C-terminus of the B-chain (Figure 1d). Compared to human insulin, an elevated isoelectric point is obtained, which causes a decreased solubility of the drug at physiological pH values and, thus, a microprecipitation of glargine after subcutaneous injection. The absorption from the injection site is delayed, and the duration of action is extended.4-6 In terms of clinical investigations, the detailed knowledge about presence, concentration, and metabolism of insulin and its related compounds is fundamental for a medical treatment of patients suffering from different types of diabetes under individual conditions. Additionally, physicians need to identify surreptitious applications by individuals or a third party, pretending an insulindependent diabetes mellitus.7 Forensic scientists identified the power of insulin as a lethal weapon in 1958,8 and an astonishing number of attempted or successful homicides and suicides by administration of the pancreatic hormone was found.9,10 Elite athletes are tempted to abuse insulin preparations to artificially improve their performance. In particular, rapidly acting insulins are in demand owing to their improved controllability, as compared to conventional human insulin formulations. The misuse of insulin in sports has often been reported, in particular (1) Rosak, C. Internist 2001, 42, 1523-1535. (2) Fabris, D.; Fenselau, C. Anal. Chem. 1999, 71, 384-387. (3) Barnett, A. H.; Owens, D. R. Lancet 1997, 349, 47-51. (4) Campbell, R. K.; White, J. R.; Levien, T.; Baker, D. Clin. Ther. 2001, 23, 1938-1957. (5) Levien, T. L.; Baker, D. E.; White, J. R.; Campbell, R. K. Ann. Pharmacother. 2002, 36, 1019-1027. (6) Reinhart, L.; Panning, C. A. Am. J. Health-Syst. Pharm. 2002, 59, 643-649. (7) Polonsky, K. S.; O’Meara, N. M. In Endocrinology, 4th ed.; Jameson, J. L., Ed.; Saunders: Philadelphia, 2001; pp 697-711. (8) Birkinshaw, V. J.; Gurd, M. R.; Randall, S. S.; Curry, A. S.; Price, D. E.; Wright, P. H. Br. Med. J. 1958, 463-468. (9) Bauman, W. A.; Yalow, R. S. J. Forensic Sci. 1981, 26, 594-598. (10) Haibach, H.; Dix, J. D.; Shah, J. H. J. Forensic Sci. 1987, 32, 208-216.

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005 3579

Figure 1. Primary structures of human insulin (a, MW ) 5807 Da), Humalog (b, MW ) 5807), Novolog (c, MW ) 5824), and Lantus (d, MW ) 6064).

in body building,11-14 but the question of whether insulin is a performance-enhancing drug is not entirely answered yet. Several facts support the hypothesis that insulin administration positively influences performance, for example, by the increase of muscle glycogen utilizing hyperinsulinaemic clamps prior to sports events or during recovery phases. In addition, recent studies have indicated that the use of insulin increases muscle size by its chalonic action in inhibiting protein breakdown.12,15-18 Since 1999, insulin has been included in the list of prohibited substances of the International Olympic Committee (IOC) and the World AntiDoping Agency (WADA), with the exemption for those athletes (11) Reverter, J. L.; Tural, C.; Rosell, A.; Dominguez, M.; Sanmarti, A. Arch. Intern. Med. 1994, 154, 225-226. (12) Evans, P. J.; Lynch, R. M. Br. J. Sports Med. 2003, 37, 356-357. (13) Rich, J. D.; Dickinson, B. P.; Merriman, N. A.; Thule, P. M. J. Am. Med. Assoc. 1998, 279, 1613. (14) Dawson, R.; Harrison, M. Br. J. Sports Med. 1998, 31, 259. (15) Sonksen, P. H. J. Endocrinol. 2001, 170, 13-25. (16) Wolfe, R. R. Curr. Opin. Clin. Nutr. Metab. Care 2000, 3, 67-71. (17) Wolfe, R. R. Curr. Opin. Clin. Nutr. Metab. Care 2005, 8, 61-65. (18) Tipton, K. D.; Wolfe, R. R. Int. J. Sport Nutr. Exercise Metab. 2001, 11, 109-132.

3580

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

demonstrably suffering from diabetes,19,20 and in several recent hearings in terms of doping offenses, insulin misuse has been frequently mentioned. Different types of immunoassays, such as radioimmunoassay (RIA), enzyme-linked immunosorbent assay (ELISA), or microparticle enzyme immunoassay (MEIA), have been commonly employed for identification and quantitation of human and animal insulin21 since the first assay was introduced by Yalow and Berson in 1959.22 Nevertheless, several problems have been observed in the past concerning the accuracy of immunoassays, on the basis of cross reactivities to either precursors (such as proinsulin or split proinsulin) or degradation products of insulin or analogues to human insulin, for example, originating from synthetic analogues or animal sources. Although these disadvantages have (19) List of Prohibited Classes of Substances and Prohibited Methods of Doping; International Olympic Committee: Lausanne, 2003. (20) The 2004 Prohibited List; http://www.wada-ama.org/docs/web/standards_harmonization/code/list_standard_2004.pdf (accessed: 12-22-2004). (21) Chevenne, D.; Trivin, F.; Porquet, D. Diabetes Metab. 1999, 25, 459-476. (22) Yalow, R. S.; Berson, S. A. Nature 1959, 184, 1648-1649.

mainly been overcome with improved immunoassays,23-25 several applications utilizing mass spectrometry have been developed for a quantitative analysis of human insulin, its precursors, and animal analogues.26-28 In the present study, we demonstrate the possibility to identify synthetic analogues of human insulin, Humalog (Lispro), Novolog (Aspart), and Lantus (Glargine) in addition to endogenous insulin in human plasma by means of immunoaffinity purification followed by microbore liquid chromatography and tandem mass spectrometry. This top-down sequencing-based assay29,30 enables the facile and fast determination of diabetesrelated drugs in doping control samples, in which detailed information about administered substances is of paramount importance in terms of clarification of origin and purpose of (ab)use. EXPERIMENTAL SECTION Materials and Chemicals. Oasis HLB solid-phase extraction cartridges (60 mg) were obtained from Waters (Milford, MA). Acetonitrile (HPLC grade), trisodium phosphate dodecahydrate (p.a.), sodium chloride (p.a.), and acetic acid (glacial) were purchased from Merck (Darmstadt, Germany), and trifluoroacetic acid (99+%), human plasma (lyophilized), bovine insulin, and recombinant human insulin were from Sigma (St. Louis, MO). Insulin analogues Humalog, Novolog, and Lantus were supplied by Eli Lilly (Indianapolis, IN), Novo Nordisk (Princeton, NJ), and Aventis (Kansas City, MO), respectively, and the anti-insulin immunoaffinity gel (0.5 mL/column, 10 mg IgG/mL) was obtained from CER (Marloie, Belgium). Tris(carboxyethyl)phosphine (TCEP-HCl) was puchased from Pierce (Rockford, IL). All buffers and solutions were prepared using deionized water (Milli-Q grade). Plasma Samples. All experiments and validation steps were performed utilizing commercially available plasma, but to probe for specificity of the presented assay, 10 different plasma samples of healthy volunteers were obtained from the Department of Cardiology and Sports Medicine of the German Sport University, Cologne, Germany. Stock and Working Solutions of Insulins. Bovine insulin was used as the internal standard in all samples. Its stock solution contained 10 pmol/µL dissolved in 2% acetic acid, and a working solution (0.1 pmol/uL) was prepared freshly with each batch of samples. The stock solutions of Humalog, Novolog, Lantus, and human insulin were prepared in 2% acetic acid at 100 pmol/µL and diluted to working solutions at 0.01 pmol/µL shortly before sample preparation. All working solutions contained 0.1 pmol of bovine insulin/µL to avoid loss of target analytes during the dilution process. Stock solutions were stored at 4 °C. (23) Lindstro¨m, T.; Hedman, C. A.; Arnqvist, H. J. Diabetes Care 2002, 25, 10491054. (24) Sapin, R.; Galudec, V. L.; Gasser, F.; Pinget, M.; Grucker, D. Clin. Chem. 2001, 47, 602-605. (25) Butter, N. L.; Hattersley, A. T.; Clark, P. M. Clin. Chim. Acta 2001, 310, 141-150. (26) Sto ¨cklin, R.; Vu, L.; Vadas, L.; Cerini, F.; Kippen, A. D.; Offord, R. E.; Rose, K. Diabetes 1997, 46, 44-50. (27) Kippen, A. D.; Cerini, F.; Vadas, L.; Sto ¨cklin, R.; Vu, L.; Offord, R. E.; Rose, K. J. Biol. Chem. 1997, 272, 12513-12522. (28) Darby, S. M.; Miller, M. L.; Allen, R. O.; LeBeau, M. J. Anal. Toxicol. 2001, 25, 8-14. (29) Kelleher, N. L.; Lin, H. Y.; Valaskovic, G. A.; Aaserud, D. J.; Fridriksson, E. K.; McLafferty, F. W. J. Am. Chem. Soc. 1999, 121, 806-812. (30) Reid, G. E.; McLuckey, S. A. J. Mass Spectrom. 2002, 37, 663-675.

Preparation of Phosphate-Buffered Saline. The phosphatebuffered saline (PBS) consisted of 120 mM Na3PO4 and 0.5 M NaCl in deionized water. The pH was adjusted to 8 by the addition of aqueous hydrochloric acid (3 M). Disulfide Reduction of Human Insulin and Humalog. To characterize the fragmentation behavior of Humalog in comparison to human insulin, both compounds were subjected to disulfide reduction to provide isolated A- and B-chains for mass spectrometric studies. Aliqouts of 500 pmol of human insulin or Humalog were dissolved in 500 µL of 10 mM TCEP-HCl in deionized water and incubated for 10 min at 60 °C. After cooling to ambient temperature, the solution was diluted with 500 µL of a mixture of deionized water and acetonitrile (1:1, v:v) and analyzed by ESIMS/MS. Mass Spectrometry of Intact Insulins and Isolated A- and B-Chains of Human Insulin and Humalog. Product ion spectra of bovine insulin, human insulin, and all synthetic insulins were recorded from the 5-fold protonated molecules, that is, m/z 1147.6 obtained from bovine insulin, m/z 1162.4 from human insulin, m/z 1162.4 from Humalog, m/z 1166.0 from Novolog, and m/z 1213.8 from Lantus, providing characteristic fragment ions of respective compounds. Owing to identical precursor ions of human insulin and Humalog, the determination of diagnostic fragment ions enabling the differentiation of these analytes by mass spectrometry is required. Hence, A- and B-chains of both drugs were prepared by disulfide reduction of the intact proteins, and product ion mass spectra of B-chains (3430 Da) comprising the discriminative sequence of the amino acids B28 and B29 were recorded from the 3-fold protonated molecule at m/z 1144.3. All spectra were measured on an Applied Biosystems QTrap mass spectrometer (Foster City, CA). Solutions containing 0.5 pmol/µL of intact proteins or cleaved A- and B-chains (see section titled Disulfide Reduction of Human Insulin and Humalog) were introduced into the mass spectrometer by means of a syringe pump at a flow rate of 5 µL/min and an ion spray voltage of 5000 V. The precursor ions were isolated at low resolution ((0.7 u) and fragmented by CAD at a collision gas pressure of 4.5 × 10-5 Torr and a collision energy of 80 eV. Plasma Sample Preparation. A 2-mL portion of human plasma was fortified with 2 pmol of bovine insulin (ISTD, 20 µL of working solution) and divided into 1-mL fractions placed in two 1.5-mL Eppendorf tubes. The samples were vortexed for 5 s and centrifuged for 5 min at 6000g, and the combined supernatants were transferred onto an immunoaffinity chromatography (IAC) column that was preconditioned with 2 mL of PBS. Another 2 mL of PBS was added, the IAC column was vortexed for 5 s, and the mixture was incubated at room temperature. After 30 min, the diluted plasma sample was eluted and discarded, the gel was washed three times with PBS, and two volumes of 1 mL of 2% AcOH were used to elute the target analytes directly onto an Oasis solid-phase extraction (SPE) cartridge. The SPE cartridge was preconditioned with 2 mL of acetonitrile and 2 mL of acetic acid (2%), washed after loading of the IAC extract with 2 mL of a mixture of acetonitrile and acetic acid (2%) (1:9, v:v), and eluted with a volume of 1.2 mL of of a mixture of acetonitrile and acetic acid (2%) (4:6, v:v) into a 1.5-mL Eppendorff tube. The sample was evaporated to dryness utilizing a vacuum centrifuge at 40 °C for 90 min, and the dry residue was resuspended in 40 µL of a Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3581

mixture of H2O containing 0.5% acetic acid and 0.01% TFA and acetonitrile (72:28, v:v). LC-MS/MS Analysis. LC-MS/MS analyses of intact insulins were performed on an Agilent 1100 Series high performance liquid chromatograph (Palo Alto, CA) interfaced to an Applied Biosystems QTrap mass spectrometer (Foster City, CA). The LC was equipped with a Zorbax 300SB-C8 guard column (2.1 × 12.5 mm, particle size 5 µm) and a Zorbax 300SB-C18 analytical column (1.0 × 50 mm, particle size 3.5 µm, pore size 300Å) operated at 30 °C. The mobile phase consisted of (A) H2O containing 0.5% acetic acid and 0.01% TFA and (B) 80% CH3CN and 20% H2O, also containing 0.5% acetic acid and 0.01% TFA. A sample volume of 20 µL was injected into the LC-MS/MS system, and a gradient of 28% B to 55% B in 15 min at a flow rate of 60 µL/min was used to separate human insulin and its analogues. Reequilibration of the column was accomplished within 23 min. The mass spectrometer was operated in the positive mode at an ion spray voltage of 5500 V, and the declustering potential was optimized for efficient isolation of the 5-fold protonated molecules of human insulin (m/z 1162.4), Humalog (m/z 1162.4), Novolog (m/z 1166.0), and Lantus (m/z 1213.8). Nitrogen was used as collision gas at 4.5 × 10-5 Torr, and product ion scan spectra were recorded at collision energies of 80 eV. To probe for carryover effects and to prolong column durability, each plasma sample analysis was followed by an injection of deionized water. Here, the gradient was modified to 28% B, increasing to 100% B in 30 min, maintained at 100% B for 10 min, followed by reequilibration for 20 min. The column oven temperature was increased to 55 °C to allow efficient purging. Assay ValidationsRecovery. The recoveries of Humalog, Novolog, and Lantus were determined by analyses of six blank plasma specimens fortified with 0.8 ng/mL of each compound and compared to analyses of another six blank plasma samples, the final extracts of which were spiked with 1.6 ng of each drug. To both sets of samples, 2 pmol (12 ng) of bovine insulin was added to the eluate from the SPE cartridge before evaporation. Recovery was calculated by comparison of mean peak area ratios of the analytes and the ISTD of samples fortified prior to and after extraction. Assay Validation: Lower Limit of Detection. The lower limit of detection (LLOD) was defined as the “lowest content that can be measured with reasonable statistical certainty”31 at a signalto-noise ratio g3. Six blank plasma samples spiked with internal standard (ISTD) only and another six blank plasma specimens fortified with 0.5 ng/mL of Humalog, Novolog, and Lantus in addition to the ISTD were prepared and analyzed according to the established protocol. The noise in blank plasma samples was measured, and a 3-fold standard deviation was added. Average signal intensities obtained for Humalog, Novolog, and Lantus from fortified plasma specimens were used to calculate respective signal-to-noise ratios. Assay Validation: Specificity. Ten different blank plasma specimens were prepared as described above to probe for interfering peaks at the expected retention times of Humalog, Novolog, and Lantus in product ion scan experiments. In the case of Humalog, endogenous insulin generates a closely related mass spectrum but differs by a characteristic product ion at m/z 217 (31) Kromidas, A. Validierung in der Analytik; Wiley-VCH: Weinheim, 1999.

3582

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

(vide infra), which was utilized to distinguish between these two compounds. Assay Validation: Linearity. Six plasma samples of a volume of 2 mL were fortified with the reference compounds Humalog, Novolog, or Lantus at concentrations of 0.5, 0.75, 1.0, 1.25, 1.5, and 2.0 ng/mL.32 The specimens were prepared and analyzed as described above, and linearity was proven by the Mandel test.33,34 RESULTS AND DISCUSSION Mass Spectrometry of Intact Insulins and Isolated A- and B-Chains of Human Insulin and Humalog. All synthetic, bovine, and human insulins generated informative product ion spectra upon collisionally activated dissociation of the 5-fold protonated molecules. Owing to their composition of two peptide chains connected by two disulfide bonds in addition to one intrachain disulfide bridge, efficient fragmentation required high collision energies of 80 eV, resulting in small product ions primarily ranging from m/z 100 to 250. In Figure 2, the product ion spectra of human insulin, Humalog, Novolog, and Lantus are presented. Bovine insulin , Novolog, and Lantus differ from human insulin in their amino acid composition and, thus, their molecular weight, allowing a separation of these analytes by mass selection. The rapid-acting insulin Humalog differs from endogenous human insulin only by two amino acid residues (B28 and B29) with switched positions. Hence, chromatography demonstrated similar behaviors, and the molecular masses are identical, which required a differentiation of both analytes by tandem mass spectrometry to identify the abuse of the synthetic drug. The product ion spectra of (M + 5H)5+ at m/z 1162.4 of intact human insulin (a) as well as Humalog (b) containing abundant common fragment ions at m/z 120, 129, 136, and 143 are shown in Figure 2, whereas human insulin contains an individual ion at m/z 226 and Humalog a corresponding fragment ion at m/z 217. To substantiate the assumption that these product ions originate from the C-termini of the B-chains, both compounds, human insulin and Humalog, were treated with TCEP-HCl to cleave disulfide bonds and, thus, chemically separate A- and B-chains. The 4-fold protonated molecules of both isolated B-chains (m/z 858.3) were dissociated, yielding abundant fragment ions at m/z 226 and 217, which were assigned as y3-y1 and y2 for human insulin and Humalog, respectively (Figure 3). These product ions result from prolinedirected fragmentation, a phenomenon that has been described in the literature for various peptides and proteins.35 These characteristic product ions allow a distinction between the endogenously produced insulin and the synthetically derived insulin Humalog in plasma samples, as shown below. LC-MS/MS Analysis of Plasma Samples. The scope of the study was the development of an assay allowing the mass spectrometric determination of synthetic insulin analogues in human plasma doping control samples. Normal endogenous serum insulin levels for fasting or nonfasting nonobese subjects are expected between 12 and 76 fmol/mL27 or 35-525 fmol/mL28 (32) International Conference on Harmonization. Validation of Analytical Procedures: Methodology; http://www.ich.org/MediaServer.jser?@_ID)418&@ _MODE)GLB. (accessed 22-12-04). (33) Mandel, J. The Statistical Analyses of Experimental Data; John Wiley & Sons: New York, 1964. (34) Gottwald, W. Statistik fu ¨ r Anwender; Wiley-VCH: Weinheim, Berlin, New York, 2000. (35) Loo, J. A.; Edmonds, C. G.; Smith, R. D. Anal. Chem. 1993, 65, 425-438.

Figure 2. ESI-product ion spectra of (a) (M + 5H)5+ of human insulin (m/z 1162.4), (b) (M + 5H)5+ of Humalog (m/z 1162.4), (c) (M + 5H)5+ of Novolog (m/z 1166.0), and (d) (M + 5H)5+ of Lantus (m/z 1213.8).

Figure 3. ESI-product ion spectra of B-chains of (M + 4H)4+ at m/z 858.3 of (a) human insulin, and (b) Humalog. The abundant fragment ions at m/z 226 or 217 resulting from respective y3-y1 or y2 ions allow a mass spectrometric distinction between these closely related compounds.

(corresponding to 0.1-0.5 ng/mL or 0.2-3.2 ng/mL), respectively. According to patient information provided by the manufacturer of synthetic insulins, the approximate insulin requirement is 0.5-1.0 U/kg/day, and approximately 50-70% can be supplied by rapid-acting analogues.36 For a person of 70 kg, this demand accounts for 0.6-1.8 mg of insulin per day. The half-life of synthetic insulin was published for Humalog37 at 26 min after (36) Novo Nordisk. NovoLog Insulin AspartsInformation for the Patient; 01-062002.

intravenous administration and 60 min after subcutaneous application of 0.1 U/kg. Because no information regarding amounts and cycles of insulin abuse have been published, plasma concentrations in cheating athletes’ samples as well as potential periods of detection are difficult to estimate. Thus, the present assay was targeted at the detection of normal endogenous insulin levels, because a hyperinsulinaemic administration of synthetic analogues is expected on the basis of recent studies describing the anticata(37) Eli Lilly. Humalog Insulin Lispro InjectionsDescription; 01-06-2002.

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3583

Figure 4. LC-MS/MS analysis of a plasma sample fortified with 1.0 ng/mL of (a) Novolog and (b) Lantus. Both chromatograms represent the total ion current generated byproduct ion scan experiments on respective 5-fold charged precursor ions at m/z 1166.0 and 1213.8.

Figure 5. LC-MS/MS analyses of (a) a plasma sample fortified with 0.8 ng/mL of Humalog, and (b) a blank plasma specimen. The product ion mass spectrum of part a contains a fragment ion at m/z 217, proving the presence of Humalog, whereas part b is lacking this diagnostic criterion.

bolic effect of locally increased insulin levels.17,38,39 Plasma specimens were fortified with the insulin therapeutics Humalog, Novolog, and Lantus, which are prohibited in sports for all healthy athletes, purified by immunoaffinity chromatography, and analyzed (38) Tessari, P.; Inchiostro, S.; Biolo, G.; Vincenti, E.; Sabadin, L. J. Clin. Invest. 1991, 88, 27-33. (39) Biolo, G.; Declan Fleming, R. Y.; Wolfe, R. R. J. Clin. Invest. 1995, 95, 811819.

3584 Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

by LC-ESI-MS/MS. A typical chromatogram obtained from a plasma sample spiked with 1.0 ng/mL of Novolog and Lantus is shown in Figure 4. Distinct signals are obtained for Lantus at 22.74 min and Novolog at 23.10 min. A plasma sample fortified with 0.8 ng/mL of Humalog is shown in Figure 5a, and the product ion mass spectrum corresponding to the signal at 23.44 min contains the expected fragment ion at m/z 217, which is not present in the product ion mass spectrum of a blank plasma sample, as

Table 1. Drug-Specific Analytical Parameters and Assay Validation Results compd

MW (Da)

precursor ion (M + 5H)5+ (m/z)

collision energy (eV)

recovery (%)

LLOD (ng/mL)

human insulin Humalog Novolog Lantus

5807 5807 5824 6064

1163.4 1163.4 1166.0 1213.8

80 80 80 80

n/a 94 98 91

n/a 0.5 0.5 0.5

depicted in Figure 5b. Here, only a fragment ion at m/z 226 is observed, and both facts combined prove the presence of human insulin and the absence of Humalog. Assay Validation. To utilize the established procedure for screening or confirmation purposes in doping control analyses, an assay validation was conducted covering the aspects recovery, lower limit of detection, specificity, and linearity. The tested parameters are summarized in Table 1. The determined recoveries for Humalog, Novolog, and Lantus ranged from 91 to 98% at precisions of 14-19%. The efficient extraction and purification of trace amounts of insulins allowed an LLOD of 0.5 ng/mL for all synthetic analogues. The noise in blank plasma samples was measured, and a 3-fold standard deviation was added. The signals obtained for Humalog, Novolog, and Lantus provided a signal-to-noise ratio g3 at 0.5 ng/mL. According to normal nonfasting insulin plasma levels, calibration curves were prepared for synthetic insulins from 0.5 to 2.0 ng/ mL, all of which demonstrated linearity. The obtained equations were y ) 0.1144x + 0.0357 for Humalog at r2 ) 0.993, y ) 0.1058x + 0.0104 for Novolog at r2 ) 0.992, and y ) 0.0566x + 0.0049 for Lantus at r2 ) 0.990. Owing to product ion scan experiments utilized for all analyses in the present study, peak areas include the total ion current at respective retention times of target analytes. As described above, endogenous insulin and Humalog generate identical precursor ions at m/z 1163.4 and elute at nearly identical retention times. Thus, the calibration curve of Humalog is composed by synthetic and endogenous insulin, which is indicated also by an elevated axis intercept, as compared to Novolog and Lantus. But becuase all calibration points were prepared from the same plasma pool, the amount of endogenous insulin remained constant and allowed a linear calibration curve for Humalog. The assay was tested for specificity with 10 different plasma samples prepared and analyzed as described above. No sample generated interfering signals at expected retention times of target analytes, and signals corresponding to human insulin were distinguished from Humalog by the absence of the diagnostic fragment ion at m/z 217. CONCLUSION The developed procedure allows a fast and reliable determination of synthetic insulins in human plasma utilizing a top-down

calibration curve 0.5-2.0 ng/mL n/a linear linear linear

sequencing-based mass spectrometry approach. The selective nature of mass spectral data eliminates common shortcomings of procedures employing immuno assays26 that are amenable for cross reactivities to precursors or degradation products of endogenous insulin. Hence, mass spectrometry allows an unambiguous identification of target analytes, a crucial and essential factor in doping control analysis, considering the consequences of a positive test result that entails suspension or termination of an athlete’s career. The materials used are all commercially available, an important point for ensuring the possibility of transferring this method to other accredited doping control laboratories. The selective nature of IAC purification yielded clean extracts of plasma samples, and the employed tandem mass spectrometer provided abundant fragment ions in product ion scan experiments in low m/z ranges, enabling the identification of Novolog and Lantus as well as the differentiation between endogenous human insulin and its synthetic analogue Humalog. The low quantity of required volumes (2 mL) is advantageous, because blood sampling is still seldom used in doping control analysis and, if any, only small amounts are withdrawn from athletes. Finally, the described assay requires less than 4 h for sample preparation, also an important item with regard to high throughput analysis in limited time frames during great sporting events, such as the Olympic Games or world championships. ACKNOWLEDGMENT This project has been carried out with support from the World Anti-Doping Agency (WADA). In addition, the authors thank the Manfred Donike Gesellschaft e.V., Cologne for financial support and PD Dr. Petra Platen (German Sport University, Department of Cardiology and Sports Medicine) for providing blank plasma specimens.

Received for review January 12, 2005. Accepted April 7, 2005. AC050066I

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3585