Study of 4-Quinolone Antibiotics in Biological Samples by Short

Extraction of pipemedic acid (1), enoxacin (2), norfloxacin (3), ciprofloxacin (5), ... determines the laboratory frame collision energy, was varied b...
0 downloads 0 Views 339KB Size
Download PDF
Anal. Chem. 1997, 69, 4143-4155

Study of 4-Quinolone Antibiotics in Biological Samples by Short-Column Liquid Chromatography Coupled with Electrospray Ionization Tandem Mass Spectrometry Dietrich A. Volmer,* Bashir Mansoori, and Steven J. Locke

Institute for Marine Biosciences, National Research Council, 1411 Oxford Street, Halifax, Nova Scotia B3H 3Z1, Canada

Simultaneous detection and confirmation of 15 quinolone antibiotics was accomplished by fast short-column liquid chromatography coupled with electrospray ionization tandem mass spectrometry (LC/MS/MS). Several physicochemical parameters such as hydrophobicity and aqueous dissociation constants were calculated from the structural formulas of the quinolone drugs, and their impact on both chromatographic and mass spectrometric behavior was studied. Additionally, a possible influence of bulk solution pH on electrospray detection sensitivity of 4-quinolones was investigated and compared to predictions based on solution-phase equilibria. A signal intensity comparison of the MH+ ions at different pH values for all 15 compounds did not reveal any pH effect, despite variations by several orders of magnitude in equilibrium concentrations in bulk solution. To demonstrate the potential of the LC/MS/MS method, its application to trace analysis in several biological matrices such as milk, salmon, and human urine was investigated. The method was shown to be sensitive with detection limits down to 1 ppb in both milk and salmon tissue. The versatility of the method was also exhibited by utilizing it for rapid identification of urinary metabolites of ciprofloxacin. Finally, a new complementary approach is described for confirmatory analyses of 4-quinolones by means of a quasi-MS/ MS/MS technique involving in-source collision-induced dissociation. It is shown that LC/quasi-MS/MS/MS can significantly enhance structural information and, thus, the specificity of analysis for the investigated 4-quinolones. Most of the antimicrobial agents introduced for clinical use in recent years are derivatives of nalidixic acid. Their common skeleton is termed 4-oxo-1,4-dihydroquinoline. These drugs (Table 1) are, however, better known under their generic name, 4-quinolones.1 The progenitor of 4-quinolones, nalidixic acid (13), was introduced in the early 1960s, followed by oxolinic acid (12) and cinoxacin (11) in the 1970s. The antibacterial activity of 4-quinolones was shown to result from selective inhibition of bacterial DNA synthesis.1 Of particular importance today are the 6-fluorinated piperazinyl derivatives 2-10, with antibacterial activities approaching 1000 times that of nalidixic acid.2,3 These newer 6-fluoroquinolones are not only highly active against a wide * To whom correspondence should be addressed. E-mail: Dietrich. [email protected]. (1) The Quinolones; Andriole, V. T., Ed.; Academic Press: London, 1988. (2) Gootz, T. D.; Brighty, K. E. Med. Res. Rev. 1996, 16, 433. S0003-2700(97)00425-3 CCC: $14.00

© 1997 American Chemical Society

spectrum of Gram-negative bacteria but also moderately active against Gram-positive bacteria.4 The use of 4-quinolones, however, is not limited to clinical applications. These agents are also widely applied in the treatment and prevention of veterinary diseases in food-producing animals and in commercially farmed fish such as salmon and catfish.5,6 Several agents were specifically developed for veterinary medicine. For instance, danofloxacin (8), enrofloxacin (9), and sarafloxacin (10) are used to treat respiratory and enteric bacterial infections in cattle, swine, chicken, and turkey7,8 and diseases in aquacultured fish.9 These compounds are very similar to those given to humans. Enrofloxacin, for example, is widely used for animal treatment in Europe. Its major metabolite is ciprofloxacin (5), which is currently the most used clinical antibiotic in the world. Other 4-quinolones, such as ofloxacin (4), have been or are still used in both human and veterinary medicine. The wide application range and the extensive use and misuse of 4-quinolones in veterinary medicine represents a potential hazard because residues of these drugs may persist in edible tissues or milk. Together with the misuse of 4-quinolones in human medicine, this has led to the emergence and spread of drug-resistant bacterial strains.10 Moreover, nalidixic acid and oxolinic acid have been linked to cancer in rats,11 and 4-quinoloneinduced acute arthropathy has been observed in several animal species.12 Although severe cases of 4-quinolone-induced arthropathy have been observed only rarely in humans, incidents of transient arthralgia have also been reported.12 These observations have precluded the use of 4-quinolones in children and pregnant women. The need to identify 4-quinolones in various biological tissues and fluids is obvious. Because of their outstanding effectiveness, inadmissible use of nonapproved 4-quinolones in certain applications is very likely (enrofloxacin, for example, is not approved for (3) Mokrushina, G. A.; Charushin, V. N.; Chupakhin, O. N. Pharm. Chem. J. 1995, 9, 590. (4) Fernandes, P. B.; Chu, D. T. W. Annu. Rep. Med. Chem. 1988, 23, 133. (5) Scheer, M. Vet. Med. Rev. 1987, 2, 104. (6) Bauditz, R. Vet. Med. Rev. 1987, 2, 122. (7) Kemp, I.; Gesbert, M.; Guiltel, M.; Bennejean, G. Res. Vet. Sci. 1992, 53, 257. (8) Mann, D. D.; Frame, G. M. Am. J. Vet. Res. 1992, 53, 1022. (9) Food Chem. News 1993, 35, 51; 1996, 38, 4. (10) Neu, H. C. Science 1992, 257, 1064. Acar, J. F.; Goldstein, F. W. Clin. Infect. Dis. 1997, 24, S67. Acar, J. F. Clin. Drug Invest. 1995, 9, 45. (11) Yamada, T.; Maita, K.; Nakamura, J.; Okuno, Y.; Hosokawa, S.; Matsuto, M.; Yamada, H. Food Chem. Toxicol. 1994, 32, 397. (12) Linseman, D. A.; Hampton, L. A.; Branstetter, D. G. Fund. Appl. Toxicol. 1995, 28, 59.

Analytical Chemistry, Vol. 69, No. 20, October 15, 1997 4143

Table 1. Structural Formulas and Nominal Molecular Masses of the 4-Quinolones Studied

treatment of cattle in several European Union [EU] countries). Proposed restrictive measures and establishment of regulatory levels for 4-quinolones have not yet been widely implemented in veterinary medicine, despite calls by regulatory agencies for restriction.13 The EU set a maximum residue limit (MRL) of 30 parts-per-billion (ppb) for the sum of enrofloxacin and its metabolite, ciprofloxacin, in muscle, kidney, and liver.14 The MRL for milk is expected to be set at or below 10 ppb.15 Switzerland limits the concentration of oxolinic acid to 10 ppb in fish and the maximum level for enrofloxacin to 30 ppb in milk, eggs, and meat.16 The great variety of 4-quinolones, and the possibility of trace residues in edible tissue and milk, make it necessary to develop sensitive multiresidue screening methods. (13) Anim. Pharm. 1995, 327, 13; 328, 20. (14) EC Regulations 2377/90 and 2901/93. (15) Hammer, P.; Heeschen, W. Milchwissenschaft 1995, 50, 513. (16) Charriere, R.; Leiser, W.; Dousse, R. Mitt. Gebiete Lebensmittelunters. Hyg. 1996, 87, 223.

4144

Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

Several methods have been reported for determination of 4-quinolones in various biological matrices. Vega et al.17 described a high-performance thin-layer chromatography method for oxolinic acid in fish tissue, with detection limits around 10 ppb. The majority of analytical methods for 4-quinolones, however, are based on liquid chromatography (LC). Methods for flumequine (14), nalidixic acid, oxolinic acid, piromidic acid, and/or sarafloxacin in catfish, salmon, and shrimp have been reported with ultraviolet (UV) and fluorescence detection (FLD).16,18-21 Detec(17) Vega, M.; Rios, G.; Saelzer, R.; Herlitz, E. J. Planar Chromatogr. 1995, 8, 378. (18) Munns, R. K.; Turnipseed, S. B.; Pfenning, A. P.; Roybal, J. E.; Holland, D. C.; Long, A. R.; Plakas, S. M. J. AOAC Int. 1995, 78, 343. (19) Pfenning, A. P.; Munns, R. K.; Turnipseed, S. B.; Roybal, J. E.; Holland, D. C.; Long, A. R.; Plakas, S. M. J. AOAC Int. 1996, 79, 1227. (20) Meinertz, J. R.; Dawson, V. K.; Gingerich, W. H.; Cheng, B.; Tubergen, M. M. J. AOAC Int. 1994, 77, 871. (21) Degroodt, J. M.; Wyhowski de Bukanski, B.; Srebrnik, S. J. Liquid Chromatogr. 1994, 17, 1785.

tion limits down to the low-ppb level were sometimes obtained, particularly when fluorescence detection was used. Hormaza´bal and Yndestad22 described an LC/FLD method for enrofloxacin with limits of quantification of 3 ppb in milk and 5 ppb in meat. In addition, numerous studies deal with HPLC assays for determination of pharmacokinetic parameters for clinically important 4-quinolones and their metabolites. For example, Scholl et al.23 described a postcolumn derivatization LC/FLD technique for determining picogram amounts of ciprofloxacin and its four known metabolites in urine, serum, feces, bile, and tissue. Other methods have been described24-26 for similar purposes. Most of the described methods allow the simultaneous analysis of only a few 4-quinolones. More importantly, using these methods does not allow concurrent confirmation of a tentatively identified 4-quinolone residue on the basis of the retention time alone, or elucidation of the structures of unknown residues or metabolites. Hammer and Heeschen15 developed a specific enzyme-linked immunosorbent assay (ELISA) for enrofloxacin and ciprofloxacin with detection limits of 1.56 ppb in milk. Physicalchemical confirmatory techniques, however, are usually needed to ensure that such analytical methods do not give rise to falsepositive responses. Generally, it is accepted that mass spectral (MS) analysis, particularly when combined with LC or capillary electrophoresis (CE), provides unequivocal proof of the presence of drug residues in biological samples.27,28 Only a few examples showing LC/MS analysis of 4-quinolones have been reported so far. Horie et al.29 utilized thermospray LC/MS to detect oxolinic acid, nalidixic acid, and piromidic acid in fish by monitoring their MH+ ions. D’Agostino and co-workers30 reported in-source collision-induced dissociation (CID) electrospray (ESI) spectra for several 6-fluoroquinolones and were able to produce structurally significant product ions using this technique. Doerge and Bajic31 also utilized in-source CID with an atmospheric pressure chemical ionization (APCI) interface to generate product ions for oxolinic acid, nalidixic acid, flumequine, and piromidic acid. The major product ions described in this work,31 however, correspond to loss of water from the MH+ ions, which is a very common neutral loss and, therefore, not particularly structure-specific. Detection limits of 1 ng in fortified fish extracts were reported.31 A specific assay for the confirmatory identification of danofloxacin residues in chicken and cattle tissues was described by Schneider and coworkers,32 who utilized electrospray tandem mass spectrometry (MS/MS) to monitor the MH+ ion and two significant product ions. Optimum electrospray and MS/MS operating conditions permitted32 the detection of danofloxacin residues in liver extracts down to 50 ppb. The most extensive study on LC/MS of (22) Hormaza´bal, V.; Yndestadt, M. J. Liquid Chromatogr. 1994, 17, 2911. (23) Scholl, H.; Schmidt, K.; Weber, B. J. Chromatogr. 1987, 416, 321, (24) Beermann, D.; Wingender, W.; Foerster, D.; Zeiler, H. J.; Graefe, H. K.; Schacht, P. J. Clin. Pharmacol. 1984, 24, 403. (25) Zeiler, H. J.; Beermann, D.; Wingender, W.; Foerster, D.; Schacht, P. Infection 1988, 16, S19. (26) Aoki, H.; Osshima, Y.; Tanaka, M.; Okazaki, O.; Hakusui, H. J. Chromatogr. B 1994, 660, 365. (27) Volmer, D. A. Rapid Commun. Mass Spectrom. 1996, 10, 1615. (28) Bateman, K. P.; Locke, S. J.; Volmer, D. A. J. Mass Spectrom. 1997, 32, 297. (29) Horie, M.; Saito, K.; Nose, N.; Tera, M.; Nakazawa, H. J. Liquid Chromatogr. 1993, 16, 1463. (30) D’Agostino, P. A.; Hancock, J. R.; Provost, L. R. Rapid Commun. Mass Spectrom. 1995, 9, 1038. (31) Doerge, D. R.; Bajic, S. Rapid Commun. Mass Spectrom. 1995, 9, 1012. (32) Schneider, R. P.; Ericson, J. F.; Lynch, M. J.; Fouda, H. G. Biol. Mass Spectrom. 1993, 22, 595.

4-quinolones, so far, was recently presented by Schilling and coworkers.33 Despite the fact that the study deals with just one compound (sarafloxacin), this excellent paper describes in detail a multidimensional approach which combines the specificity of sample preparation, liquid chromatography, and electrospray tandem mass spectrometry. Schilling et al.33 used ion intensity ratios from the MS/MS experiment as an additional aid in confirmation and were able to achieve MS/MS product ion ratios with 0.998) in the investigated concentration range (40 pg/mL to 40 ng/mL). MS/MS. In this study, LC/MS/MS was used as a confirmatory method for 4-quinolones. It was, therefore, of interest to study the dissociation behavior of the MH+ ions of these compounds upon collisional activation. Table 4 lists the ESI MS/MS data obtained at a collision offset voltage, ∆Vc, of 30 V and a collision gas pressure of p(N2) ) 2.75 mTorr. Two major fragmentation pathways were observed in the CID analysis of MH+ of 7-(N4′-alkylpiperazinyl)-6-fluoroquinolones (1-10): (a) loss of H2O from the carboxyl function at C-3, followed by loss of R1 (as C2H4 or C3H4) from the ethyl or cyclopropyl groups at N-1. (Sarafloxacin, 10, did not exhibit such a loss of R1.) (b) Initial loss of CO2 from the carboxyl group, followed by loss of C2H5N (1-3, 5, 6, 10), C3H7N (4), or C4H9N (9). As an illustrative example, the proposed interpretation of product ions of ciprofloxacin (5) is shown in Scheme 1. The analytical importance of certain product ions in this scheme will be discussed in more detail in the following section. Danofloxacin (8) and N-desmethyldanofloxacin (7) exhibited losses of 57 (C3H7N) and 43 amu (C2H5N), respectively, from the [MH H2O]+ ions as well. Because of their bicyclic structures at R7, however, the resulting product ions at m/z 283 have an azetidinetype structure at C-7, as compared to the aziridine-type structures of the related product ions of the other 7-(N4′-alkylpiperazinyl)6-fluoroquinolones. The product ions at [MH - 87]+ (corresponding to two consecutive neutral losses [CO2 + C2H5N] from the MH+ ions, see above), observed with several 4-quinolones, can be used for sensitive class-specific screening of 7-piperazinyl-6-fluoroquinolones in sample extracts. Moreover, the optimum collision offset voltages, ∆Vc, for these transitions were almost identical (about 30 V) for all investigated compounds (Figure 2). That is, high sensitivity in neutral-loss scan mode could be obtained with a single set of experimental parameters for all compounds investigated (see Figure 1b). An application of this observation to a classspecific screening procedure in milk samples will be shown in

the following section. N′-4-Alkylated 4-quinolones (e.g., 9) gave rise to similar ions with, of course, higher neutral loss masses (101 amu for N′-4-methyl and 115 amu for N′-4-ethylpiperazine derivatives). Careful examination of the CID spectra of 7-(N4′-alkylpiperazinyl)-6-fluoroquinolones also revealed several significant structurediagnostic ions of lower abundance. For example, ions corresponding to HF loss from MH+ or from other secondary precursor ions were observed with most of the fluorine-containing 4-quinolones (Table 4). The intensities of these ions were low, as was also observed by D’Agostino and co-workers.30 These intensities, however, could be greatly increased by quasi-MS/MS/MS (see below). In addition, loss of CO from several dehydrated ion species was observed in most spectra (for example, at m/z 286, 203, and 189 in Scheme 1) among other minor fragmentations. Scheme 1 summarizes the principal dissociation reactions observed for 7-(N4′-alkylpiperazinyl)-6-fluoroquinolones, exemplified by the proposed fragmentation pathway of 5. The CID spectra obtained for the acidic 4-quinolones (1115) contained fewer product ions due to the lack of reactive C-7 substituents. The principal ion fragmentation pathway involved loss of H2O from the MH+ ion, followed by sequential neutral losses of R1 and one or two CO molecules (Table 4). Flumequine (14) exhibited a neutral loss of 42 amu from [MH - H2O]+ to give the ion at m/z 202, probably corresponding to loss of CH3CHdCH2 from the ring structure between N-1 and C-8. Interestingly, the acidic 4-quinolones did not show any [MH - CO2]+ ions, in contrast to behavior observed for 7-(N′-alkylpiperazinyl)-6fluoroquinolones (see above). Analytical Applications. To evaluate the short-column LC/ MS method, the linearities, limits of detection (LODs), and limits of confirmation (LOCs) for several 4-quinolones were measured from fortified control milk, fish, and urine samples. 4-Quinolone antibiotics usually expected in these matrices were chosen. The procedures used for cleanup and extraction of the 4-quinolones from the various matrices investigated were modifications from previously reported methods (see Experimental Section). AlAnalytical Chemistry, Vol. 69, No. 20, October 15, 1997

4149

Scheme 1. Proposed Interpretation of Main Fragment Ions of MH+ (m/z 332) of Ciprofloxacin (5)

Figure 2. MS/MS absolute ion intensities (arbitrary units) of the [MH - 87]+ transitions of 7-piperazinyl-6-fluoroquinolones as a function of the collision offset voltage, ∆Vc (individual data points not shown). For curve assignment, refer to Table 1.

though improving these well-established methods was not a goal of this study, their suitability for combination with short-column LC/MS/MS was evaluated here. External standard calibration curves were determined from the individual peaks of the 4-quinolones in time-scheduled SIM experiments (see Table 5). Quantifications were performed in the SIM mode because of its better precision than that obtainable using MS/MS.27 4150 Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

7-(N′-Alkylpiperazinyl)-6-fluoroquinolones in Milk. Seven 4-quinolone drugs were targeted in milk: 1, 2, 3, 5, 6, 9, and 10 (Table 1). As described in the previous section, the piperazinyl group of 7-piperazinyl-6-fluoroquinolones gave rise to generic [MH - 87]+ ions which could be used to monitor the presence of residues of most 7-piperazinyl-6-fluoroquinolones. Enrofloxacin (9), with its N′-4-ethyl substituent, could not be monitored by this procedure, and determination of enrofloxacin is discussed in the next section. The strategy used here to detect, confirm, and quantify 7-piperazinyl-6-fluoroquinolones in milk samples was as follows: (a) prescreening of target and nontarget 7-piperazinyl6-fluoroquinolones by constant neutral-loss (CNL) scan experiments using the [MH]+ f [MH - 87]+ transition; (b) quantification of identified target compounds by monitoring their protonated molecules in a time-scheduled SIM experiment; and (c) further confirmation, if necessary, by selected reaction monitoring using three characteristic transitions. The application of both MS/MS prescreening and timescheduled SIM quantification is demonstrated in Figure 3 for a milk sample spiked with four 7-piperazinyl-6-fluoroquinolones at the 10 ppb level. Residues were readily identified, confirmed, and quantified at this concentration level. The LODs in milk, defined as the minimum detectable amount of analyte with a signal-tonoise ratio of 3:1 in time-scheduled SIM of MH+ ions, were 0.2-2 ppb for the six 7-piperazinyl-6-fluoroquinolones and 0.5 ppb for enrofloxacin (see Table 5). LOCs by MS/MS were in a similar range, between 1 and 2 ppb (the LOC was defined as the MRM

Table 5. Linear Regression Analysis of Peak Areas with Time-Scheduled SIM Limits of Detection (LOD) and Confirmation (LOC) for the Investigated 4-Quinolones in Milk, Fish, and Urine Samplesa matrix milk

salmon

urine

4-quinolone

slope/104

y-intercept/103

r2

LODb (ppb)

LOCc (ppb)

pipemedic acid (1) enoxacin (2) norfloxacin (3) ciprofloxacin (5) lomefloxacin (6) enrofloxacin (9) sarafloxacin (10) oxolinic acid (12) nalidixic acid (13) flumequine (14) piromidic acid (15) enoxacin (2) norfloxacin (3) ofloxacin (4) ciprofloxacin (5) enrofloxacin (9) cinoxacin (11)

0.36 2.01 0.27 0.98 0.87 1.36 2.10 0.41 0.14 0.24 0.13 24.63 28.12 17.37 15.46 34.67 7.62

0.33 -12.24 -5.42 -0.89 -1.23 0.78 7.66 -7.60 -2.88 -7.03 -1.80 1658 1851 872 1253 2610 1219

0.999 0.999 0.998 0.998 0.999 0.998 0.999 0.993 0.995 0.996 0.995 0.993 0.994 0.996 0.996 0.992 0.994

2 1 2 1 0.5 0.5 0.2 1 1 1 1 ndd nd nd nd nd nd

2 2 2 2 1 2 1 2 5 2 5 nd nd nd nd nd nd

a External calibration curves were determined by spiking the sample matrices in the following concentration range (5-7 concentration levels): milk, 1-100 ppb; salmon, 1-50 ppb; human control urine, 1-250 µg/mL. The compounds are listed in order of increasing retention times within a given sample matrix. b Estimated by signal-to-noise of 3:1 from the low end of the calibration curves. c Estimated limits of confirmation by MRM (three transitions). d Not determined.

Figure 3. LC/MS/MS analysis of a milk extract. The original milk sample was spiked with 4-quinolones at the 10 ppb (nanograms per milliliter) level. (a) Prescreening of a control milk sample, CNL of 87 amu; (b) prescreening of spiked milk sample, CNL of 87 amu; (c) quantification of spiked milk sample, time-scheduled SIM traces of the MH+ ions. For peak assignment, refer to Table 5.

limit of detection [signal-to-noise ratio ) 3:1] for [MH]+ f [MH - 87]+ transitions; the LOC value for 9 is that for the quasi-MS/ MS/MS procedure described in the following section). Acidic 4-Quinolones in Salmon Muscle Tissue. Four acidic 4-quinolones (12-15), which are typical in aquaculture applications, were investigated in salmon muscle tissues. LC/MS

Figure 4. LC/MS/MS analysis of a salmon muscle tissue extract. The original salmon sample was spiked with 4-quinolones at the 50 ppb (nanograms per milliliter) level. (a) Time-scheduled SIM traces (MH+) of control salmon; (b) time-scheduled SIM (MH+) of spiked salmon; (c-e) MS/MS confirmation of nalidixic acid (13) by MRM (∆Vc ) 40 V). For peak assignment, refer to Table 5 (11 was investigated as a possible internal standard).

chromatograms for a salmon control sample and for a muscle tissue sample spiked at the 50 ppb level are shown in Figure 4a,b. The time-scheduled SIM LODs in muscle tissue, as estimated by a signal-to-noise ratio of 3:1 from the low end of the calibration curve, were ∼1 ppb for all acidic quinolones (see Table 5). Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

4151

Table 6. Summary of Retention Data, Physicochemical Properties, and Major Product Ions of Ciprofloxacin (5) and Its Urinary Metabolites (M1-M4)

c

peak no.

tr (min)

Mn

pKa1 (-COOH)

pKa2 (-Rx,y,zNH+)

log Da (pH ) 2.75)

MH+ m/z

M1 5

3:50 3:85

305 331

6.24 6.27

10.18 10.17

-2.72 -2.42

306 332

M2 M3

4:71 5:25

411 345

6.27 6.24

0.31 0.43c

412 346

M4

6:28

359

6.27

0.42c

360

major product ionsb m/z (relative abundance) 306 (10), 288 (50), 268 (85), 245 (25), 227 (60), 217 (100) 332 (60), 314 (85), 288 (100), 294 (5), 274 (20), 268 (10), 245 (95), 231 (85) 332 (30), 314 (100), 294 (10), 245 (5), 231 (25) 346 (10), 328 (100), 287 (50), 285 (30), 258 (20), 245 (5), 231 (10), 217 (100) 360 (2), 342 (60), 217 (100)

a Apparent partition coefficient at I ) 0.02 M. b Collision-induced dissociation at ∆V ) 30 V and p(N ) ) 2.75 mTorr (see Experimental Section). c 2 log P.

Because the MH+ ions of these compounds did not exhibit generic, chemical class-specific fragment ions under collisional activation in MS/MS (except relatively unspecific [MH - H2O]+ ions), characteristic product ions were monitored for confirmation. Commonly, guidelines for confirmatory analyses suggest at least three structure-diagnostic ions per compound, including the MH+ ion33,46 (criteria for selection of these ions are addressed in more detail in the next section). In this study, MH+ and the following ions were used for confirmation: the [MH - H2O - 28]+, [MH - H2O - 56]+, and [MH - H2O - 84]+ ions were chosen for 12 (at m/z 216, 188, and 160) and 13 (at m/z 187, 159, and 131), corresponding to neutral losses of the R1 moiety, (R1 + CO), and (R1 + 2 CO), respectively. The diagnostic ions used for 14 were [MH - H2O - C3H6]+ (m/z 202), [MH - H2O - C3H6 - CO]+ (m/z 174), [MH - H2O - C3H6 - 2CO]+ (m/z 146); for 15, [MH - H2O - CO]+ (m/z 243), [MH - H2O - CO - C3H6]+ (m/z 201), and [MH - H2O - 2CO-C3H6]+ (m/z 173) were selected. An example of such a confirmation procedure is illustrated in Figure 4c-e for 13. The LOCs were defined as the MRM limits of detection for spiked fish samples for the weakest of the three transitions chosen, as estimated by a signal-to-noise ratio of 3:1 for these ions, and ranged between 1 and 5 ppb (Table 5). Rapid Identification of Metabolites in Human Urine Samples. Several clinically important 4-quinolones were investigated in this study. The calibration curves obtained for all studied compounds in control urine samples are summarized in Table 5. Of particular interest was ciprofloxacin (5). The identification and quantification of ciprofloxacin metabolites excreted via urine after oral administration of the drug was studied here by means of the shortcolumn LC/MS/MS methodology reported above. The main biotransformation route of 7-(N′-alkylpiperazinyl)-6fluoroquinolones occurs at the C-7 piperazine ring1 (refer to Table 1). Desethylenyl- (M1), N-sulfonyl- (M2), oxo- (M3), and Nformylquinolones (M4) have been reported to be the major urinary metabolites of ciprofloxacin.1,34 The structural formulas are summarized in Table 6. In order to identify these metabolites (and possibly others as well) by LC/MS/MS, precursor ion (46) Sphon, J. J. Assoc. Off. Anal. Chem. 1978, 6, 1247.

4152

Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

scanning was conducted. Precursor ions were selected in order to detect modifications at C-7. Two ions were chosen: m/z 217 and 231. The product ion at m/z 231 was selected in order to cover metabolites with intact piperazine rings (M2-M4), while m/z 217 was selected for both the intact ring species and openring metabolites (M1). The results of these experiments are summarized in Figure 5b,c and clearly illustrate that rapid identification of metabolites was readily possible by using only two generic precursor ions. Subsequently, confirmation of structures of tentatively identified metabolites on the basis of their precursor ion response was achieved via product ion scans of the corresponding MH+ ions. The product ion spectra are summarized in Table 6, along with calculated physicochemical properties of M1-M4. The observed product ions are consistent with the tentative chemical structures assigned to these metabolites (Table 6). Because pure standards of M1-M4 were not available, quantification of metabolite concentrations was based on the assumption that their ionization efficiencies were similar to that of the original precursor compound, 5. This assumption seemed reasonable because the ESI molar responses of all 4-quinolones investigated in this study did not depend substantially on their structures (compare Table 3). Two samples were taken within 11.5 h after a single dose of 500 mg. A third urine sample was taken 5.5 h after a second dose of 500 mg (16 h after the first dose). The summed concentrations (5 + M1-M4) measured in all urine samples ranged between 110 and 360 mg/mL (see, for example, Figure 5d). From our results on the quantification of urinary metabolites, the metabolic compositions after a single dose of ciprofloxacin (500 mg, oral) were as follows: in 0-11.5 h urine (two samples), unchanged 5 (83.7%), M1 (2%), M2 (8.2%), M3 (6.0%), and M4 (0.07%). These numbers correspond well with metabolite levels in urine reported previously for 5, e.g., unchanged 5, 89% + 11% combined metabolites (serum half-lives reported1 for 5 were between 3 and 4.5 h with ∼30-33% excretion into urine). Quasi-MS/MS/MS Analysis of 7-(N′-Alkylpiperazinyl)-6fluoroquinolones. As mentioned above, guidelines for use of mass spectrometry as a confirmatory method usually recommend

Figure 5. LC/MS/MS analysis of 0.5 mL of a human urine sample of a volunteer (collection period 1.5-11.5 h) after oral administration of 500 mg of ciprofloxacin (5). (a) Full-scan analysis (m/z 200-600); (b) precursor ion scan chromatogram of m/z 217; (c) precursor ion scan chromatogram of m/z 231; (d) quantification by SIM of the MH+ ions. Experimental conditions: MS/MS, ∆Vc ) 40 V. For peak assignment, refer to Table 6.

monitoring at least three structurally diagnostic ions per analyte in SIM experiments.46 Confirmation criteria for MS/MS experiments are less clearly defined. As pointed out by Schilling and co-workers,33 choice of diverse product ions and is necessary for a high degree of specificity in the MS/MS analysis. Often, the choice of diverse product ions in MS/MS is not possible because low-energy CID does not lead to the formation of the desired ions. For example, Schilling et al.33 used three transitions in the MRM analysis of sarafloxacin (10) in catfish tissues, one corresponding to reaction in the piperazine group ([MH - CO2 - C2H5N]+, m/z 299) and two involving the carboxyl function ([MH - H2O]+ and [MH - CO2]+, m/z 368 and 342, respectively). No other intense ions were formed upon CID,33 in agreement with the present work (Table 4). While the 7-(N4′alkylpiperazinyl)group and the corresponding ion formed under CID certainly are highly diagnostic and specific for certain 4-quinolones, losses of water and CO2 from the protonated molecules are very common and may not be regarded as sufficiently specific for confirmation. Ideally, product ions reflecting the substituting groups at N-1, C-6, C-7, and C-8 (Table 1) are needed.30 An additional degree of specificity was introduced to analysis of 4-quinolones by using a quasi-MS/MS/MS approach. This quasi-MS/MS/MS technique was recently reported as a means of analyzing isomeric sulfonamide antibiotics in milk.28 In quasiMS/MS/MS, in-source CID is used as a first quasi-MS/MS stage to generate first-generation fragment ions from the protonated

molecules. In a second MS/MS step, the ions of interest are isolated and made to undergo CID in the collision quadrupole to yield second generation product ions, which can be used as additional structure-specific ions. In this study, the dehydrated species [MH - H2O]+ were chosen as intermediate ions. These [MH - H2O]+ ions are highly reactive, as shown by Colorado and Brodbelt47 in collisional activation experiments of several protonated 4-quinolones in a quadrupole ion trap. Additionally, the use of [MH - CO2]+ ions as potential precursors for further collisional activation was investigated here. The second generation product ions from CID of [MH - CO2]+, however, were virtually identical to the product ions obtained from single-step CID using MS/MS of the MH+ ions, for all investigated 4-quinolones, and thus did not provide an additional degree of specificity. The results of the quasi-MS/MS/MS experiments are summarized in Table 7 (ions used for confirmations are shown in italics). The MS/MS and quasi-MS/MS/MS spectra for 10 are compared in Figure 6. Upon collisional activation, protonated sarafloxacin (m/z 386) dissociates to produce three main peaks (Figure 7a), corresponding to dissociation in the carboxyl and the piperazinyl functions (m/z 368, 342, and 299, see discussion above). Likewise, all other 7-(N4′-alkylpiperazinyl)-6-fluoroquinolones followed this fragmentation behavior (Table 4). In contrast, the quasi-MS/MS/MS analysis of 10 via [MH - H2O]+ intermediates yielded additional fragment ions (Figure 7b) including the structure diagnostic [(MH - H2O) - HF]+ (m/z 348), [(MH H2O) - HF - CO]+ (m/z 320) and [(MH - H2O) - HF - Pip]+ ions (m/z 270, corresponding to cleavage of the entire piperazinyl function R7 with simultaneous hydrogen migrations). That is, the combination of MS/MS and quasi-MS/MS/MS could greatly enhance structural information. There are, however, differences between MS/MS and quasi-MS/MS/MS with respect to the total signal intensities. To produce the necessary intermediate [MH - H2O]+ ions, the API potential difference (orifice-skimmer voltage difference) was increased to 95 V, which caused a loss of ∼30% in total signal but gave maximum yield of [MH-H2O]+ ions. Absolute signals were approximately 10 to 100 times lower for the fragments obtained in the quasi-MS/MS/MS process than for fragments obtained by regular MS/MS of the MH+ ions. These high losses are mainly a result of the relatively inefficient in-source CID process with the particular instrument used. The [MHH2O]+/[MH]+ ratios in the quasi-MS/MS stage were 0.1 to 0.4 for the investigated quinolones at ∆Vi)95V in our experiments. These ratios can easily be increased by increasing ∆Vi, but the absolute signals decreased rapidly for ∆Vi>100V. That is, ∆Vi)95V was found to be optimal with respect to maximum ion currents of [MH-H2O]+ ions. Nevertheless, absolute limits of confirmation in the picogram range were observed. Consequently, highly sensitive and selective analyses, in conjunction with the LC method described in this paper, were possible. An illustrative example for quasi-MS/MS/MS confirmation of enrofloxacin (9) in milk at the 10 ppb level is shown in Figure 7. A mixed-mode (MS/MS + quasi-MS/MS/MS) experiment was used for analysis. Four MRM transitions were monitored, three pairs originating from a regular MS/MS experiment ([MH]+ f [MH - H2O]+, [MH]+ f [MH - CO2]+, and [MH]+ f [MH CO2 - C4H9N]+) at low API potential difference (∆Vi ) 35 V) plus a fourth transition in quasi-MS/MS/MS mode at ∆Vi) 95 V ([MH]+ f [MH - H2O]+ f [(MH - H2O) - CO - C2H4]+. The (47) Colorado, A.; Brodbelt, J. Anal. Chem. 1994, 66, 2330.

Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

4153

Table 7. Second-Generation Product Ions of [MH - H2O]+ Ions in the Quasi-MS/MS/MS Analysis of 7-(1-Piperazinyl)-6-fluoroquinolones product ions, m/z (relative abundance)a compound pipemedic acid (1) enoxacin (2) norfloxacin (3) ciprofloxacin (5) lomefloxacin (6) enrofloxacin(9) sarafloxacin (10)

precursor MH+ - H2O 286 (60) 303 (60) 302 (100) 314 (70) 334 (100) 342 (40) 368 (100)

-HF 283 (10) 282 (50) 294 (10) 314 (50) 322 (5) 348 (90)

-28 258 (20) 275 (10) 274 (10) 306 (10) 314 (10)

-HF - 28

other major ions (100)b

255 (10) 254 (20) 286 (10) 294 (5) 320 (90)

233 (10), 215 258 (10), 232 (100), 204 (20) 231 (50), 226 (10), 204 (10) 231 (100) 258 (15), 249 (20), 243 (20) 300 (10), 286 (100), 271 (10), 258 (20), 243 (20), 217 (20) 292 (50), 270 (80), 189 (30)

a API interface potential difference (orifice-skimmer voltage difference) for [MH - H O]+ ion generation at ∆V ) 95 V; collision-induced 2 i dissociation in q2 at ∆Vc ) 40 V and p(N2) ) 2.75 mTorr (see Experimental Section). b Complementary structure diagnostic ions produced from the quasi-MS/MS/MS procedure with sufficient intensities for analytical use are shown in italics.

Figure 6. MS/MS and quasi-MS/MS/MS spectrum of sarafloxacin (10). (a) Product ions of MH+ (m/z 386) at ∆Vc ) 30 V and ∆Vi ) 35 V; (b) product ions of [MH - H2O]+ (m/z 368) at ∆Vc ) 40 V and ∆Vi ) 95 V. For mass assignments, refer to Table 7.

SIM trace (MH+ at m/z 360, Figure 7a) shows several interferences, especially at shorter retention times. In the MS/MS (Figure 7b-d)and quasi-MS/MS/MS traces (Figure 7e), however, no responses were obtained for ions characteristic of 9, other than those at the correct retention time. CONCLUSIONS The described LC/MS/MS method for 4-quinolones is sensitive and specific. It eliminates the need for separate quantification and confirmation procedures as required by most literature methods for 4-quinolones. The present method is suitable for a variety of different matrices (milk, fish, and urine) and has also proven valuable for rapid identification and quantification of 4154 Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

Figure 7. MS/MS and quasi-MS/MS/MS analysis of a milk extract. The original milk sample was spiked with enrofloxacin (9) at the 10 ppb (nanograms per milliliter) level. (a) SIM trace of the MH+ ion (m/z 360); (b-e) simultaneous mixed-mode MRM traces (two experiments, 250 ms dwell time each). (b-d) Experiment 1 (MS/MS): MRM of three transitions at ∆Vc ) 30 V and ∆Vi ) 35 V. (e) Experiment 2 (quasi-MS/MS/MS): MRM of one transition at ∆Vc ) 40 V and ∆Vi ) 95 V.

metabolites. The major advantages of this method are the rapidity of separation and the specificity of the MS detection technique used. In this study, conventional, well-established sample preparation methods based on liquid/liquid extractions were used. While effective and reliable, these tedious and time-consuming procedures lack the sophistication of modern separation and detection techniques such as those described here. It will be important in the future to develop efficient and effective sample preparation techniques that match the rapid turnaround time of the analytical procedures. Possible examples of such preparation techniques

are automated on-line solid-phase extraction, solid-phase microextraction, and immunoaffinity extractions with drug-specific monoclonal antibodies. ACKNOWLEDGMENT The authors thank Dr. Robert K. Boyd (IMB/NRC, Halifax, NS, Canada) for his valuable discussions and suggestions in the preparation of this manuscript, Mrs. Carolyn McCarther for

providing urine samples, and Dr. Santosh Lall (IMB/NRC) for providing salmon samples. Received for review April 23, 1997. Accepted July 28, 1997.X AC970425C X

Abstract published in Advance ACS Abstracts, September 1, 1997.

Analytical Chemistry, Vol. 69, No. 20, October 15, 1997

4155