In phase I metabolism, methods to distinguish formation of N-oxides from sulfoxides and C-hydroxylations could have a significant impact in the rapid metabolite identification/profiling of drug molecules by mass spectrometry. Analysis by LC/tandem mass spectometry can easily distinguish oxidative possibilities when they occur in positions characterized by distinct product ions. However, in cases where various oxidations can occur, for example, on the same ring system, the metabolite of interest must be produced in large quantities either by in vitro or in vivo experiments, followed by isolation in sufficient quantities for structure determination by NMR. This process is time-consuming and tedious. Hence, it is advantageous to develop simple chemical methods for the selective conversion of N-oxides to their corresponding amines. The method should be practical, and the reagent employed should enable us to perform the transformation in the presence of various biological matrices such as plasma and urine to allow subsequent identification by LC/MS.

It is of course well established that, due to the NAD(P)H reductase system, liver preparations can reduce compounds as well as oxidize them, and this process has been harnessed to convert N-oxides to amines in drug metabolism studies (Powis et al., 1982; Kitamura et al., 1999). Yet this is not a selective means of reducing N-oxides. Although several promising chemical reagents have been reported in the literature for the selective reduction of N-oxides to amines in the presence of other sensitive functional groups (Seaton et al., 1984; Malinowski and Kaczmarek, 1987), this process has apparently not been widely employed in the identification/profiling of drug metabolites. We became interested in this subject following the report of Prakash et al. (1997), in which the authors employed TiCl₃ to prove the absence of N-oxide functionality during the characterization of novel benzisothiazole ring-cleaved products of the antipsychotic drug ziprasidone.

Mather and Thomas (1972) also used TiCl₃ to characterize a purported N-hydroxyamide metabolite formed from lidocaine in humans. This case illustrates the problem of using a chemical method such as this as the sole means of characterization. Subsequent studies of lidocaine metabolism by Nelson et al. (1974 and 1978) confirmed that TiCl₃ successfully reduces the N-hydroxyamide back to lidocaine in the urine matrix, but in those later studies no evidence for the formation of such a hydroxyamide metabolite in vivo was found. The discrepancy is probably due to the presence of another lidocaine metabolite they found in urine, N-hydroxy-2,6-dimethylaniline, which can also be reduced by TiCl₃. Reduction by TiCl₃ was also employed by Beckett et al. (1971) in their analysis of nicotine-1′-N-oxide in urine in the presence of nicotine and cotinine. Gorrod and Gooderham (1981) used TiCl₃ to aid in the characterization of N,N-dimethylaniline-N-oxide and to distinguish it from phenol oxidation products in the in vitro metabolism of N,N-dimethylaniline, and this method was later generalized by Seto and Guengerich (1993) in their study of cytochrome P450 2B1 oxidation of a number of N,N-dialkyl anilines. In the case of these simple anilines, however, TiCl₃ was used more as a classic chemical reagent to quantitate the amount of aniline-N-oxide formed in the incubation after the metabolite was extracted and isolated from the matrix.

Although TiCl₃ was employed in the examples discussed above, a variety of other reagents, including complex hydrides, trivalent phosphorous compounds, various sulfur- and selenium-containing compounds, metals in acids, and catalytic hydrogenations have been reported in the literature for the reduction of amine N-oxides. A comprehensive list of reagents for this transformation has recently been reviewed (Chandrasekhar et al., 2002). We chose three reagents, TiCl₃ (McCall and TenBrink, 1975; Seaton et al., 1984), TiCl₄/NaI (Balicki, 1990), and PMHS (Chandrasekhar et al., 2002), to evaluate their potential for applications in drug metabolism. PMHS was chosen because it has been demonstrated to perform reduction of N-oxides under milder conditions compared with titanium chloride-based reductions.

The advent of electrospray and ionspray ionization techniques has led to the routine use of LC/MS methods for profiling metabolites present in a wide variety of matrices (typically urine, plasma, and in vitro incubations) in drug discovery today. In this study, we wished to explore whether TiCl₃ might be the reagent of choice for rapid confirmation of a very commonly observed metabolite, the N-oxide, which is often impossible to distinguish by mass spectrometry alone from metabolites formed by C-hydroxylation in the same area of the molecule. We wished to explore the conditions under which such a reagent could be used in the matrices themselves or after isolation of a peak by preparative LC. Our aim is also to explore how selective the reagent might be, not only to other oxidation products such as sulfoxides, epoxides, and alcohols, but also to other sensitive and labile functionalities in drugs that we have studied. Herein, we describe these efforts to selectively reduce a diverse set of N-oxides with the above reagents and provide real-world drug metabolism examples to demonstrate the utility of this method in the context of rapid metabolite profiling by LC/MS.

Materials and Methods

Materials. Isoquinoline N-oxide (1), sodium iodide, titanium(III) chloride (ca. 10 wt. % solution in 20-30 wt. % hydrochloric acid), titanium(IV) chloride (1 M solution in dichloromethane), poly(methylhydrosiloxane), and tetrakis(triphenylphosphine)palladium(0) were purchased from Sigma-Aldrich (St. Louis, MO). N-Oxides 2 through 14 (for structures see Fig. 1) were obtained from the internal Global Sample Management System at Eli Lilly and Co. (Greenfield, IN). All solvents used for LC/UV/MS were of chromatographic grade.

Reduction of Isoquinoline N-Oxide (1) to Isoquinoline. 1) TiCl₃: TiCl₃ (300 μl) was added to an ice-cold solution of 14.1 mg of 1 in 1 ml of methanol. The mixture was kept on ice for 1 h and neutralized with 500 μl of 5 N KOH. 2) TiCl₄/NaI: TiCl₄ (500 μl) and 30 mg of NaI were added to a stirred solution of 10.7 mg of 1 in 1 ml of acetonitrile at room temperature. After 1 h, the reaction mixture was neutralized with 500 μl of 5N KOH. 3) PMHS: PMHS (36 μl) and 2.5 mg of Pd(PPh₃)₄ were added to a solution of 29.04 mg of 1 in 2 ml of THF at room temperature, and the mixture was stirred for 2 h. LC/MS analysis was performed on a Shimadzu VP Series (Shimadzu Scientific Instruments, Inc., Columbia, MD) consisting of a SIL-10AXL autosampler and Surveyor PDA detector. Separations were carried out on an Inertsil ODS3 column (2.1 × 150 mm, 5-μm particle size; MetaChem Technologies Inc., Torrance, CA) with a flow rate of 0.25 ml/min. The mobile phase consisted of 0.2% formic acid (mobile phase A) and methanol (mobile phase B). Analytes were eluted using a gradient method (min/% B): 0/5, 3/5, 30/50, 30.1/90, and 35/90. Prior to analysis, an aliquot of each reaction mixture was diluted with methanol to a concentration of 1 mg/ml, diluted 10 times further with water, and centrifuged, and the clear supernatant was injected onto the column. MS was performed on a Finnigan LCQ spectrometer (Thermo Electron Corporation, San Jose, CA) under positive ESI conditions.

Reduction of Leurosine N-Oxide (2) to Leurosine. 1) TiCl₃: TiCl₃ (30 μl) was added to an ice-cold solution of 1 mg of 2 in 1 ml of methanol at 5°C. After 90 min at 5°C, an aliquot (20 μl) of the reaction mixture was diluted with 980 μl of 0.2% aqueous formic acid, centrifuged, and analyzed by LC/MS. 2) PMHS: PMHS (3 μl) and 1 mg of Pd(PPh₃)₄ were added to a stirred solution of 1 mg of 2 in 1 ml of THF at room temperature. After 2 h, the reaction mixture was treated as before for LC/MS analysis.

Reduction of Leurosine N-Oxide (2) to Leurosine in the Presence of Rat Plasma. 1) TiCl₃: TiCl₃ (30 μl) was added to an ice-cold solution of 1 mg of 13 in 500 μl of methanol and 500 μl of rat plasma at 5°C. After 90 min at 5°C, the reaction mixture was treated as before for LC/MS analysis. 2) PMHS: PMHS (3 μl) and 1 mg of Pd(PPh₃)₄ were added to a stirred solution of 1 mg of 2 in 500 μl of THF and 500 μl of rat plasma at room temperature. After 2 h, the reaction mixture was treated as before for LC/MS analysis.

Reduction of Leurosine N-Oxide (2) to Leurosine in the Presence of Rat Urine. The reductions were performed as described before in the presence of plasma, and LC/MS system and conditions were as described before, except that a modified gradient method [(min/% B): 0/20, 3/20, 28/45, and 30/80] was used.

Reduction of N-Oxides 3 through 13. TiCl₃ (7.5 μl) was added to each 250-μg sample of substrates 3 through 13 in 500 μl of methanol at 5°C. After 2 h at 5°C, an aliquot from each reaction mixture was diluted 25 times with a 3:1 mixture of 0.2% aqueous formic acid/acetonitrile and centrifuged, and the clear supernatant was analyzed by LC/MS. Analysis revealed <50% conversion in most instances; hence, an additional 10 μl of TiCl₃ was added, and after 90 min at 5°C, an aliquot from each reaction mixture was treated as before and analyzed by LC/UV/MS.

Reduction of N-Oxides 3 through 13 in the Presence of Rat Urine. Rat urine (250 μl) and 7.5 μl of TiCl₃ were added to each 250-μg sample of substrates 3 through 13 in 250 μl of methanol at 5°C. After 2 h at 5°C, an aliquot from each reaction mixture was diluted 25 times with a 3:1 mixture of 0.2% aqueous formic acid/acetonitrile and centrifuged, and the clear supernatant was analyzed by LC/UV/MS. Analysis revealed <50% conversion in most instances; hence, an additional 10 μl of TiCl₃ was added, and after 90 min at 5°C, an aliquot from each reaction mixture was treated as before and analyzed by LC/UV/MS.

Reduction of N-Oxides 3 through 13 in the Presence of Rat Plasma. TiCl₃ (15 μl) was added to each 250-μg sample of substrates 3 through 13 in 250 μl of methanol and 250 μl of rat plasma at 5°C. After 90 min at 5°C, an aliquot from each reaction mixture was treated as before and analyzed by LC/UV/MS.

LC/UV/MS analysis was performed on an Agilent 1100 series liquid chromatographic system (Agilent Technologies, Palo Alto, CA). Chromatographic separations were performed on an XTerra MS C₁₈ column (2.1 × 50 mm, 3.5-μm particle size; Waters Corporation, Milford, MA) with a flow rate of 1 ml/min at 60°C. The mobile phase consisted of 0.2% aqueous ammonium formate (mobile phase A) and 0.2% ammonium formate in a 1:1 mixture of acetonitrile/methanol (mobile phase B), and the analytes were eluted using a gradient profile (min/% B): 0/5, 7/100, 8/100, and 8.05/5. MS studies were conducted in positive and negative ESI modes using a Micromass ZQ spectrometer (Waters Corporation); for the analysis, a portion (300 μl) of the effluent from the diode array detector was directed to a mass spectrometer. Compounds were quantified by integrating UV peaks using MassLynx integration software version 3.5.

Reduction of N-Oxide Metabolite of a Pyridazine Analog by TiCl₃. TiCl₃ (1 μl) was added to an ice-cold solution of approximately 1 μg of pyridazine N-oxide 14 in 100 μl of methanol. After 1 h at 0°C, an aliquot of the reaction mixture was diluted with 0.2% aqueous formic acid and analyzed by LC/MS as described under “Reduction of Leurosine N-Oxide (2) to Leurosine.”

Reduction of N-Oxide Metabolites 15, 16, and 17 Directly from Dog Urine by TiCl₃. A single dose of 6 mg/kg of a proprietary compound was administered intravenously to a beagle. Urine samples collected from 0 to 48 h were pooled. TiCl₃ (5 μl) was added to an ice-cold 250-μl urine sample containing three oxidative metabolites (M + 16). After 90 min at 0°C, an aliquot of the urine was diluted two times with 0.2% aqueous formic acid and analyzed by LC/MS. HPLC separations were performed on a Shimadzu system using an Inertsil ODS3 column (3.0 × 150 mm, 5-μm particle size) with a flow rate of 0.4 ml/min. The mobile phase consisted of 10 mM ammonium acetate (mobile phase A) and acetonitrile (mobile phase B), and analytes were eluted using a gradient method (min/% B): 0/10, 3/10, 13/30, 33/50, and 49/90.

Selective Reduction of Sulfoxide, N-Oxide Analogs 18 and 19 to Sulfoxide Analogs 20 and 21 by TiCl₃. TiCl₃ (3 μl) was added to an ice-cold solution of approximately 10 μg of 18 in 100 μl of methanol. After 1 h at 0°C, an aliquot of the reaction mixture was diluted 10 times with 0.2% aqueous formic acid and analyzed by LC/MS. The HPLC system consisted of a Shimadzu system attached to a Micromass Qtof2 mass spectrometer (Waters Corporation). Separations were performed on a Supelco Discovery C₁₈ column (2.1 × 50 mm, 5-μm particle size; Supelco, Bellefonte, PA) with a flow rate of 0.2 ml/min. The mobile phase consisted of 10 mM ammonium acetate (mobile phase A) and a 9:1 mixture of acetonitrile/isopropanol (mobile phase B), and the analytes were eluted using a gradient profile (min/% B): 2/10, 25/60, 25.1/90, and 30/90.

TiCl₃ (3 μl) was added to an ice-cold solution of approximately 5 μg of 19 in 100 μl of methanol. After 90 min, an aliquot of the reaction mixture was diluted five times with 10 mM aqueous ammonium acetate and analyzed by LC/MS. The HPLC consisted of a Shimadzu system attached to a Thermo Finnigan LCQ DECA mass spectrometer (Thermo Electron). Chromatography was performed as described before for 18.

NMR Analysis of 4 and 5. NMR spectra were acquired on an Inova 500 MHz NMR system equipped with a Varian 5-mm cold triple-resonance probe (Varian Inc., Palo Alto, CA). Compounds were dissolved in CD₃OD, and the spectra were referenced with respect to residual solvent signal at 3.30 ppm.

Results and Discussion

Figure 1 shows the structures of N-oxide substrates used in this study. As a test substrate, we chose isoquinoline N-oxide (1) to evaluate the efficiency of the reducing agents. The LC/MS profile of the reduction of isoquinoline N-oxide to isoquinoline by these reagents is shown in Fig. 2. As evident from the chromatogram, complete conversion was observed with TiCl₃ and PMHS, and >90% conversion was observed with TiCl₄/NaI.

With these results in place, we first investigated the reduction of leurosine N-oxide (2), a complex and multifunctional vinca alkaloid, with TiCl₃ and PHMS. Under the reaction conditions described under Materials and Methods, reduction of N-oxide 2 to leurosine was accomplished to the extent of 40% by TiCl₃ (Fig. 3) and quantitatively by PMHS (Table 1). To test the ability of these reagents to reduce N-oxides in the presence of biological matrices, we then investigated the transformation of 2 in the presence of rat urine and plasma. As illustrated in Fig. 3, these matrices seem to enhance the reduction efficiency of TiCl₃ under the same reaction conditions. In contrast, PMHS effected poorer conversion in the presence of urine or plasma (Table 1), and extended reaction time and/or the addition of more reagents did not influence the course of the transformation. First, we considered the possibility that water in the reaction mixture might itself account for the inability of the palladium catalyst to function. However, quantitative conversion to leurosine was observed when the reaction was carried out in a 1:1 mixture of THF/water. Next, we considered that proteins present in the plasma and urine might inactivate PMHS and/or Pd(PPh₃)₄. We chose albumin as a model protein to study the effect of protein concentration on the reduction efficiency of PMHS. The results revealed almost no conversion at 20 mg/ml albumin, 38% at 2 mg/ml, and nearly quantitative conversion at 0.2 mg/ml, clearly suggesting the role of proteins in the inactivation of PMHS and/or Pd(PPh₃)₄.

We expanded the scope to a diverse set of N-oxides 3 through 13 with TiCl₃ as the single reducing agent because of the reaction's simplicity (it involves the addition of a single reagent) and its ability to reduce N-oxides in the presence of matrices. The reactions were performed in the absence and presence of matrices (urine and plasma) as before. As apparent from Table 1, the reduction efficiency of TiCl₃ is very high for most of the substrates, and the presence of matrices had little or no effect on the conversion rates. Only two structural types, represented by N-oxide 9 and N-oxides 10 through 12, seem to be poor substrates for reduction by TiCl₃. However, gratifyingly, the conversion rates for these substrates are respectable in the presence of both biological matrices when compared with the rates in the absence of the matrix (Table 1). When working with small amounts of substrates, as in these cases, we have found that neutralization of acidic TiCl₃ with base is not necessary prior to HPLC analysis. However, we have observed extensive degradation with some compounds when the temperature of the reaction mixture was kept at room temperature for prolonged periods of time.

The LC/UV/MS profile of the reaction mixture of 5 displayed two peaks with nearly equal intensity. Examination of their mass spectra revealed identical MH⁺ ions at m/z 238 that correspond to the reduced amine. Thus, it is logical to assume that N-oxide 5 might have existed in E and Z isomeric forms. To verify this possibility, an NMR investigation was undertaken. Contrary to the assumption, the initial NMR spectrum of 5 measured in CD₃OD showed the E isomer exclusively. However, after standing in the NMR tube over a period of time (8 days), the E isomer slowly converted to the Z isomer until equilibrium had been reached (Fig. 4). Unlike the case of the NMR experiment, rapid E to Z isomerization was observed in the LC/MS study, probably due to the acidic mobile phase conditions employed during sample preparation and HPLC. As anticipated, similar results were observed with the furan analog 4.

With a satisfactory method in hand for the transformation of N-oxides to amines, we then turned our attention to applying this methodology to real-world drug metabolism cases. Because of its simplicity and effectiveness in reducing N-oxides in the presence of biological matrices, TiCl₃ is clearly the preferred reagent. The first example tested was a proprietary compound containing a pyridazine moiety, which is easily oxidized in vivo. Thus, when it was administered to rats, a metabolite having a mass 16 Da higher than the parent was found to be the major circulating metabolite in plasma. Tandem mass spectometry revealed the site of oxidation to be on the pyridazine ring. No typical product ions, the loss of 16 Da (an oxygen atom), or 17 Da (a hydroxyl radical) were observed from the protonated molecular ion that would suggest oxidation to the N-oxide; hence, the initial structure was proposed with C-hydroxylation. To verify this proposal, a sufficient amount of the metabolite was purified for NMR spectroscopy, and the structure was eventually determined to be the N-oxide 14. Treatment of the purified metabolite with TiCl₃ subsequently produced the parent amine in quantitative yields (Table 1).

In our next attempt, we performed the N-oxide reduction with TiCl₃ directly on a urine sample. LC/MS analysis of a dog urine sample collected after the administration of a proprietary compound containing both pyridine and 4-substituted piperidine moieties showed three oxidative metabolites with masses 16 Da higher than the parent, the structures of which were suggested by the fragmentation analysis to be the pyridine N-oxide 15 and the two diastereomeric piperidine N-oxides 16 and 17 (Fig. 5A). After treatment with TiCl₃, the LC/MS analysis showed only the parent molecule (Fig. 5B), confirming the MS observations that all three metabolites are indeed N-oxides. Thus, TiCl₃ can be successfully employed to perform N-oxide transformations to amines directly on biological matrices without the need for prior isolation of metabolites.

Finally, we were able to demonstrate the selectivity of TiCl₃ for N-oxides in the presence of sulfoxides. Rat hepatic microsomes metabolized a proprietary drug molecule possessing thioether and piperidine functionalities, among other metabolites, to a metabolite in which the sulfur and the piperidine nitrogen were oxidized to sulfoxide and N-oxide, respectively. The structure of the bis-oxidized metabolite 18 (Fig. 6A) was unambiguously established by MS and NMR methods. The purified metabolite upon treatment with TiCl₃ under the reaction conditions described under Materials and Methods showed partial conversion of the bis-oxidized metabolite 18 to sulfoxide 20 (Fig. 6B), demonstrating the reagent's selectivity toward N-oxides. For an analogous metabolite, 19, in which a dimethylamino group replaces the piperidine ring, treatment of the purified metabolite with TiCl₃ again resulted in essentially complete and selective reduction of the N-oxide to give the sulfoxide 21. It has been reported, however, that under refluxing conditions, simple alkyl and aryl sulfoxides could be reduced to sulfides by TiCl₃ (Ho and Wang, 1973).

As described in the Introduction, there have not been many reports describing the use of a chemical reagent such as TiCl₃ for drug metabolism applications, even though its use as a chemical reagent for N-O reduction is well established. Furthermore, among the cases we have seen, only a couple described the use of TiCl₃ directly in a typical matrix for drug metabolism studies, lidocaine metabolites in urine from a human metabolism study (Mather and Thomas, 1972), and nicotine metabolism in humans (Beckett et al., 1971). Yet we have found TiCl₃ to be a very useful reagent for reducing N-oxides of a wide variety of structure types; furthermore, our results show that TiCl₃ can be employed directly in urine and plasma in addition to selectively reducing small amounts of N-oxide metabolites isolated after preparative LC. Although we were interested in the characterization of N-oxides, it should be noted that TiCl₃ would reduce most N-O bonds. Two such examples are found in the case of lidocaine metabolism, the characterization of its putative hydroxyamide metabolite (Mather and Thomas, 1972), and the subsequent demonstration of a confounding presence in urine of another lidocaine metabolite reduced by TiCl₃, N-hydroxy-2,6-dimethylaniline (Nelson et al., 1974 and 1978). However, our results are consistent with the known selectivity of TiCl₃ for reduction of N-oxides in the presence of alcohols and phenols and also demonstrate its selectivity against sulfoxides, epoxides, and a number of other chemically labile groups present in structurally complex drugs such as leurosines (2) and β-lactams (13).

In conclusion, we have described an optimized, simple qualitative method for the conversion of N-oxides to amines in the presence of other sensitive functional groups, including alcohols, epoxides, and sulfoxides, suitable for drug metabolism applications. The best results were obtained when N-oxides were exposed to TiCl₃ at 0 to 5°C for 60 to 90 min. Notably, our results also indicate that TiCl₃ can be successfully used for N-oxide reductions in the presence of biological matrices, including plasma and urine. Finally, the results presented here also demonstrate selectivity of TiCl₃ toward N-oxides, with the notable exception of other N-oxidation products.