Abstract
The accelerated pace of contemporary drug discovery and development in the pharmaceutical industry has generated increasing demands for early information on the metabolic fate of candidate drugs to guide the selection of new compounds for clinical evaluation. In response to these demands, we have developed a procedure for the rapid analysis of complex biological mixtures for the presence of drug-related materials and have embarked on the development of novel computer-based approaches whereby such procedures can be automated. The goal of this work was to rapidly identify drug metabolites (derived either from a single substrate or from a mixture of substrates) formed in vivo or in vitro. The approach that we have developed relies on the use of generic chromatographic and mass spectrometric methods for analysis of mixtures of drugs and metabolites and on correlation analysis of tandem mass spectrometry spectra to distinguish drug-related components from endogenous materials. Cross-correlation of the spectra also is used to identify the relationship between each metabolite and its respective parent drug in the mixture. In this manner, metabolites of a mixture of several drugs may be analyzed in the time it normally would take to analyze the products from a single substrate. We show that this rapid analytical approach can, with only minor sacrifices in the completeness of the data, significantly increase the number of compounds whose metabolic fate can be elucidated in a given time.
Recently, metabolite characterization has shifted from the development stage of the drug discovery process and has become an integral component of early discovery-stage research. Concomitant with this shift has been a change in the type of information sought from metabolic studies. Although detailed qualitative and quantitative analysis of the metabolic pathway of a lead candidate continues to be an essential element of the drug development process, biotransformation information gleaned at the discovery stage can be used to guide synthetic chemistry efforts to either block or enhance metabolism with a view to optimizing the pharmacokinetic and safety profiles of newly synthesized compounds. Additionally, data on the metabolic fate of a large number of compounds may facilitate the development of structure-activity relationships for metabolism.
The availability of tandem mass spectrometers (McLafferty, 1980; Busch et al., 1988) has given the drug metabolism scientist a very powerful tool for rapidly analyzing biological samples for drugs and metabolites (Fenselau, 1992; Gelpi, 1995). Researchers outlined a liquid chromatography with tandem mass spectrometry (LC-MS/MS1) procedure for detecting drugs and metabolites in biological matrices by using a combination of constant neutral loss scans, parent scans, and product ion spectra (Perchalski et al., 1982; Lee and Yost, 1988). This strategy continues to be used routinely to analyze biological fluids for drugs and their metabolites (Straub, 1987; Vrbanac et al., 1992;Jackson et al., 1995).
A limitation of the use of constant neutral loss scans and parent ion scans is that any change in the structure which results in a change in the mass of the neutral loss or the m/z of the product ion monitored may result in major metabolites not being detected (Naylor et al., 1993). To overcome this limitation, drug metabolism scientists predict likely metabolic transformations and use full-scan liquid chromatography with mass spectrometry (LC-MS) data to search for predicted products of metabolism (Gibson and Skett, 1986). This approach also has limitations. Metabolites with masses that are difficult to predict, such as oxidative cleavage and rearrangement products, may escape detection.
In addition to these more traditional approaches which identify metabolites from a single substrate, product ion spectral libraries are being generated to help identify structurally related components (Lee et al., 1995). Unknown samples are analyzed by obtaining product ion spectra from each component in the sample. The product ion spectrum of each unknown is compared to product ion spectra in the library in order to identify unknown mixture components that are structurally related to known compounds in the library. Although the library search algorithms are well established for peptide sequences and small molecule electron impact mass spectra, the application of these techniques to small molecule product ion spectra is not well developed. This is due primarily to the difficulty associated with establishing experimental conditions that provide the same product ions and intensities among instruments and operators (Lemire and Busch, 1996).
With the demand for metabolism data on larger numbers of compounds has come the need to increase the pace of analysis for biotransformation products. This challenge has been addressed in part by automating much of the data acquisition and analysis (Penman et al., 1995; Cole et al., 1996). Although this approach accelerates the process and reduces demand on the analyst, its use is restricted largely to the analysis of biotransformation products of single substrates. In this article, we describe a procedure for the characterization of biotransformation products from a mixture of substrates. This procedure has the potential of significantly increasing the number of compounds that can be analyzed at one time.
The key to our strategy is the combined use of LC-MS/MS and correlation analysis. The ability of correlation analysis to detect similarities and differences between two product ion spectra is used to distinguish endogenous background materials in biological isolates from drug-related components and to identify the relationship between each metabolite in the mixture and its respective parent drug. In this report, we describe the LC-MS/MS and data analysis procedures, together with our findings from a series of preliminary in vitro experiments where several substrates for metabolism were incubated as a mixture. When the results of this approach are compared to those obtained by more traditional LC-MS methods for investigating drug metabolism in vitro, it is evident that the detection and identification of metabolites in the mixture is accurate and reliable, and that the time savings far outweighs the potential drawback that minor metabolites may not be observed during screening. By using mixtures, metabolites of several drugs may be analyzed in the time it normally would take to analyze the products from a single substrate.
Materials and Methods
The drug candidates used for this study, namely, L-770,970, L-771,644, L-773,417, L-773,558, and L-775,215, are Merck investigational compounds whose structures are shown in Fig.1. Human liver microsomal preparations, pooled from 10 donors, were prepared in-house. Glucose 6-phosphate, NADP, glucose-6-phosphate dehydrogenase, and p-nitrophenol were obtained from Sigma (St. Louis, MO). Trifluoroacetic acid was obtained from Pierce (Rockford, IL). Other chemicals and solvents were obtained from Fisher (Fair Lawn, NJ).
Incubations.
The five drug candidates (Fig. 1) were incubated, both individually and as a mixture (2 μM each), at 37°C for 1 h with pooled human liver microsomes (2 mg protein/ml) in the presence of an NADPH-generating system containing 10 mM glucose 6-phosphate, 1 mM NADP, and 2 I.U./ml glucose-6-phosphate dehydrogenase, all in 100 mM K2PO4 buffer, pH 7.4, and 3 mM MgCl2. All concentrations are expressed as the final concentration in the incubation mixture. The final volume of the incubation was 500 μl. Because differences between the control incubation and the experimental incubations are important for identifying possible metabolites, several control incubations were performed. These included incubation with microsomes inactivated by heating at 100°C for 10 min, incubation in the absence of NADP, and a “combination control” incubation made by combining drug incubated in the absence of microsomes with microsomes incubated in the absence of drug, so that the same amount of reagent was present in the control and the experiment incubations. The products of the individual and mixed incubations were reconstituted in 20% acetonitrile (ACN)/water after protein precipitation with 2 volumes of ACN containing the analytical internal standard, and then analyzed by LC-MS. Additionally, equal volumes of each of the five individual incubations were pooled after protein precipitation to generate a combined incubation (combined single incubations) and analyzed by LC-MS.
LC-MS/MS Analysis.
Products of incubations with liver microsomal preparations were analyzed by high-pressure liquid chromatography (HPLC) using an HP 1050 pump and autosampler (Hewlett-Packard, Wilmington, DE) equipped with a narrow bore reversed-phase column (Zorbax Rx-C8, 150 mm x 2 mm; MAC-MOD Anal., Inc., Chadds Ford, PA) and a linear ACN/water gradient (15–75% organic in 20 min) containing 0.1% formic acid at a flow rate of 200 μl/min. The effluent from the HPLC column was introduced into a Sciex API IIIplus triple quadrupole mass spectrometer (Thornhill, Ontario, Canada) via the Turbo IonSpray interface. Mass spectrometric analysis was performed in the positive ionization mode at unit resolution over the mass range 200 to 700 Da. MS/MS analyses were performed at unit resolution (0.7 Da peak width at half-height) in both quadrupoles over the mass range 50 to 600 Da. Scan time for both MS and MS/MS analyses was adjusted to 1.5 s/scan. Instrument parameters were optimized for detection and fragmentation of L-773,417, the compound used as the positive control for the program. The orifice potential was set at 45 V and collision energy was −35 eV with an argon gas thickness of 225 × 1012atoms/cm2.
Manual Data Reduction.
A summary of the procedures for the identification of biotransformation products in mixtures is shown in Table1. Manual data analysis was used to develop and validate the automated data reduction routines. The manual procedure used the data analysis software MultiView from Sciex to manipulate and review chromatograms and spectra. The first step in the analysis procedure (see Table 1, step 3) is to subtract the total ion current chromatogram (TIC) for the control chromatogram from the TIC of the corresponding experiment chromatogram. Chromatographic peaks found in this difference chromatogram are examined to find the ion or ions responsible for the chromatographic peak. These ions are then used to generate the extracted ion current chromatogram (XIC) for both the control and experiment. The list of m/zand retention time of peaks unique to the experiment is used in the subsequent step to program the acquisition of product ion spectra (Table 1, step 4). This procedure works well for incubations with relatively high concentrations of drugs (>2 μM), but at lower substrate concentrations, the reduced signal-to-noise ratio in the full scan TIC precludes detection of metabolites that provide sufficient response to yield useful product ion spectra. To compensate for this, a window consisting of five to eight successive scans is stepped across the chromatogram. Any m/z value whose intensity rises and falls in the form of a chromatographic peak over the steps is extracted from both the experiment and control. The m/z values and retention times of discrete peaks are used later to program the acquisition of product ion spectra. To be thorough, this process may take 1 to 3 h/sample depending on the complexity of the sample.
The next step in the manual data processing is to sort the lists ofm/z and retention time to permit efficient generation of product ion spectra (Table 1, step 4). The grouping allows as many peaks as possible to be analyzed in a single chromatographic run. For a complex sample, multiple injections may be required to obtain product ion spectra for all of the identified components. After the sorting is complete, manual interpretation of the product ion spectrum of each component is used to distinguish metabolites from endogenous background material (Table 1, step 6), to relate metabolites to their precursor substrates (Table 1, step 7), and to assign a likely site of metabolism in the substrate that gives rise to each metabolite.
Automated Data Reduction.
Because the above manual procedure proved to be time-consuming and often somewhat subjective, a set of computer programs was developed using AppleScript and the data analysis software MultiView from Sciex. The first script requires LC-MS data files both from the experimental incubation and from the control incubation (Table 1, step 3). MultiView calculates the “average mass spectrum” for the entire chromatographic run and identifies every mass in the mass spectrum. Each mass identified in the average mass spectrum for the experimental incubation is extracted from the TIC. The program then evaluates the resulting XIC to determine whether the signal strength is adequate to obtain a product ion spectrum. If the selected signal threshold is exceeded (in this case 10,000 counts), the program extracts the same mass from the control TIC and compares the XICs of the experiment and the control. The program stores those peaks found only in the experiment. When all the criteria are met, the program writes them/z value of the metabolite [M+H]+ ion, the retention time at which it was detected, and the intensity of the chromatographic peak to a text file. This text file then is used to program the acquisition of the product ion spectra (Table 1, step 4). A copy of the XIC from both the experiment and the control is printed out for subsequent manual inspection. Thus, the program is able to rapidly identify species that likely arose from metabolism of the parent drug(s) used as substrate(s) in the incubations.
The second script accesses the MS/MS chromatogram obtained, extracts the product ion spectrum of each peak in the chromatogram, writes a text file of the m/z values and intensities for the product ion spectrum of each analyte in a format that can be read directly into the correlation analysis program, and prints out the spectra (Table 1, step 5). Automation of the above procedure (Table 1, steps 3–5) saves 4 to 6 h of analyst time per experiment.
The next step in the automated data analysis is to decide which components in the mixture are related to the substrates and which components are endogenous (Table 1, step 6). Once again, we have written a computer program to automate this step. We have developed a pattern recognition program that uses correlation analysis of product ion spectra (Owens, 1992) to determine the degree of similarity between each unknown mixture component and all of the substrates. Full details of this data processing procedure will be reported elsewhere. This step is based on the premise that the more similar two product ion spectra are to one another, the more likely the compounds in question are structurally related. The output from this program determines whether there is a sufficiently high degree of similarity between the product ion spectra of each unknown component and each substrate for the unknowns to be considered likely metabolites. Once this decision has been made, assignments of substrate–metabolite relationships is carried out (Table 1, step 7). This step is particularly difficult when mixtures of substrates are used. The pattern recognition program is used once again to provide a more detailed analysis of the differences and similarities between the product ion spectra of each unknown and each substrate. Again, the more similar the spectra are to one another, the more closely related the structures are likely to be. Metabolites are assigned to their corresponding substrates based on how closely the product ion spectrum of each metabolite matches that of each substrate. The best match between a metabolite product ion spectrum and a substrate product ion spectrum is taken to indicate the most likely substrate–metabolite relationship.
Results
Identification of Biotransformation Products.
As expected, microsomal incubation of a mixture of five drug candidates significantly increased the complexity of the TIC generated by LC-MS analysis relative to that obtained from a single substrate (Fig.2). Nevertheless, the TIC chromatogram exhibited near-baseline resolution of all five substrates (denoted A - E, Fig. 2). To segregate peaks corresponding to biotransformation products from those associated with endogenous materials, parallel control incubations were performed. Peaks in the active incubation (for example, the peak at 11.66 min, denoted with an arrow in Fig. 2) that were not found in the corresponding control sample indicated possible metabolites. Table 2 summarizes the [M+H]+ ions of possible metabolites identified by the AppleScript program developed here (Table 1, step 3). Also included in Table 2 are the lists of [M+H]+ions of possible metabolites obtained by the more conventional methods of manual examination of the TIC of the experiment and control incubations, and a directed mass search of possible products resulting from phase I metabolism (+14, +16, +18, +30, −2, −14, −18 Da). The script detected all the possible metabolites identified by manual examination. It also detected eight endogenous components, as opposed to only five endogenous components suggested by the manual examination routine. Residual substrates and synthetic impurities were rejected as potential metabolites by the program and by manual examination. In contrast, the directed mass search failed to identify 60% of the metabolites, since these compounds proved to be fragments of their respective substrates generated by internal metabolic cleavage and therefore were difficult to predict. The components identified as possible metabolites then were examined further by MS/MS analysis (see below). Data were reduced with the in-house AppleScript program (seeMaterials and Methods and Table 1, step 3) which afforded an output of the form shown in Fig.3. The script produced a listing of all peaks, sorted by retention time, with their respective [M+H]+ ions, chromatographic peak retention times, and peak intensities. In total, more than 20 chromatographic peaks were identified as deriving from possible metabolites.
It is important to note here that the proper choice of control incubation will determine the success of identifying possible metabolites. The best control incubation is that whose background TIC is the same as the experiment incubation. We investigated the use of control incubations prepared with boiled microsomes, incubations omitting NADP from the NADPH-generating system, and the combination control (see Materials and Methods). In all three cases, the control TIC background was different, such that when comparing to the experiment TIC to identify possible metabolites, varying endogenous species are identified as possible metabolites. The “combination control”, where all the components found in the experiment incubation are found but where metabolism has been prevented, proved to be the best choice of control incubation.
Assignment of Metabolites to Parent Drug.
To identify from which parent drug a metabolite was derived, product ion spectra were obtained of the [M+H]+ ion of the potential metabolite (Table 1, step 4). The product ion spectra of as many as 17 different components were obtained in one chromatographic run. From each product ion spectrum, a list of m/z values and intensities was generated by a second AppleScript program (seeMaterials and Methods and Table 1, step 5).
Data derived from the product ion spectra were analyzed by correlation analysis in a two-step process. The first step determined whether a given compound was likely drug-related (Table 1, step 6). The ability of correlation analysis of product ion spectra to detect structurally related compounds in mixtures relies on the fundamental assumption that the more similar two chemical structures are to one another, the more similar their product ion spectra will be. This similarity is reflected in the calculated correlation value. By comparing the product ion spectra of potential metabolites to those of each parent drug, it is possible to deduce drug-relatedness. As an example, Fig.4 shows the product ion spectrum of L-751,644 ([M+H]+ at m/z 535), one of its metabolites (metabolite 2, [M+H]+ at m/z 507), and endogenous background ([M+H]+ at m/z 375) found in both the experiment and the control incubations. Visual comparison of the product ion spectrum of L-751,644 (A) and that of its metabolite (B) identifies similarity between the two spectra. Additionally, comparison of the spectrum of the endogenous material (C) with that of L-761,644 clearly indicates lack of similarity between the two spectra. Correlation values of 0.752 and 0.084 were calculated from the comparison of the spectrum of L-751,644 with those of its metabolite and the endogenous material, respectively.
A correlation value of 0.150 was defined as the cutoff for distinguishing drug-related from endogenous material. Differentiation of endogenous substances from drug-related materials (synthetic impurities, degradates, and metabolites) was confirmed by the data obtained in the single incubations. The correlation analysis identified endogenous materials very effectively. Only three endogenous compounds out of 85 possible correlations were assigned incorrectly as possible metabolites, whereas all of the metabolites found in the mixture were identified correctly as drug-related materials.
Table 3 shows the results obtained when the product ion spectrum of each compound in the mixed incubation sample was correlated with that of L-771,644, listed in order of decreasing correlation value. Metabolites of L-771,644 were expected to have high correlation values, and therefore to appear near the top of the list of drug-related components. This is exactly what was found. Metabolites shown subsequently to be derived from L-771,644 (see below) are identified in Table 3 by correlation values in bold type.
The second step in the analysis determined the relationship between all other parent drugs and their metabolites (Table 1, step 7). To accomplish this, the product ion spectrum of each potential metabolite was cross-correlated with that of each parent drug (Table4). The higher values in the table indicate a greater degree of similarity between the product ion spectra being compared. Assignment of metabolites to their respective parents was based on the largest correlation value (highlighted in bold type) within a row in Table 4. These assignments are summarized in Table5 under the heading “Mixed”.
Using correlation analysis, all metabolites found in the mixture incubation were assigned correctly to their respective parent drugs (Table 5). Twenty-one metabolites were found in the individual incubations (five from A and E, four from B and C, and three from D). Of these, 16 (76%) were found in the mixed incubation and 15 (71%) were found in the combined single incubations (see Materials and Methods and Table 5). Several metabolites were not readily observed. The majority of metabolites not identified in the mixtures were from L-773,558 (A, Table 5), which exhibited a poor response under the LC-MS conditions employed in this study. Possibly the increased background noise in the mixed incubations impeded detection of L-773,558 metabolites. To determine whether this indeed was the case, a targeted search was conducted in the XICs of the mixture and combined individual incubations for components identified in the chromatogram of the individual incubations. This search verified the presence of the “missing” metabolites in both the mixed and combined incubations at levels below the threshold for obtaining reliable product ion spectra.
Discussion
LC-MS/MS.
Experimentally, the LC-MS/MS procedure used in this study proved to be robust from both chromatographic and mass spectrometric considerations. It has been shown previously that a single chromatographic system (with only slight modifications from one class of substrates to the next) provides the necessary information for most of the compounds tested (Lee et al., 1995). The HPLC column and mobile phases used here for the gradient are rarely changed. The mobile phase composition of the gradient may be changed, but the shape of the gradient and the timing remain virtually constant. This permits reliable analysis with very little method development time. Similarly, little time is required to determine appropriate mass spectrometric conditions. For investigation of phase I oxidative pathways, we find that tuning the instrument to optimize analysis of the substrate provides good ionization and fragmentation conditions for the metabolites. One area of variability is the collision energy used to generate the product ion spectra. The collision energy can have dramatic effects on the appearance of collision-induced dissociation spectra (Lemire and Busch, 1996). Since this pattern is the basis for the computer-based identification of each metabolite, it is important to choose an appropriate collision energy for ions over the mass range investigated. The collision energy was set at a value that affords informative product ion spectra for the substrate. Typically, this occurs at a collision energy that attenuates the precursor ion signal by approximately 90 to 95%. The only change, which is used when necessary, is to vary the collision energy with m/z to maintain similar center of mass collision energies. This prevents excessive fragmentation of lower m/z ions, while assuring that some fragmentation will occur for higher m/z ions. This is most important when the expected m/z values cover a wide mass range, since the use of an appropriate collision energy allows useful spectra to be generated for more components in each chromatographic run.
Experimental Controls.
We have found that a control incubation is essential for the effective elimination of background components. Merely preventing metabolism by either using boiled microsomes or omitting NADP from the NADPH-generating system added to the incubation mixture does not suffice. Each of these two controls differs substantially from the experiment in that neither allows for the metabolism of endogenous material. Likewise, incubations that lack only substrate do not effectively control for any synthetic impurities, products of chemical instability, instrument artifacts, or NADPH-independent reactions, which may be mistaken as products of substrate metabolism.
The combination control contains all of the mixture components, including substrate, cofactors, enzyme and matrix, but no products of metabolism. As described in Materials and Methods, drug incubated in buffer at 37°C is combined with a blank microsomal incubation lacking drug, thereby generating a much closer representation of the experiment. This control reduces the number of possible metabolites identified from the full scan TIC by increasing the discrimination between endogenous background and drug-related material. Importantly, the number of mixture components that require product ion spectra to determine whether they are endogenous or drug-related is reduced substantially, resulting in appreciable savings in both time and sample.
We also included in our protocol a biological internal standard to monitor the presence of drug–drug interactions. Typically, the biological internal standard is the lead candidate in the therapeutic program and is incubated both individually and in the mixture under study. Any discrepancy between the two incubations indicates possible drug–drug interactions and a need for further investigation. For routine metabolic screening, the use of a biological internal standard serves an additional purpose, i.e., it serves as a positive control in the liver microsomal incubations, allowing data obtained on different occasions to be compared.
Mixture Analysis.
Two mixture types were used in this investigation: 1) mixed, where multiple substrates were coincubated and 2) combined, where individual incubations with single substrates were pooled before analysis. Although the use of a biological internal standard in a mixed incubation may give an indication of when drug–drug interactions are occurring, it is possible that the metabolite profiles of some substrates may change when compared to the single substrate incubations. By incubating compounds individually and then pooling to a mixture for analysis, potential drug–drug interactions can be eliminated while maintaining the increased speed of analysis. Pooling incubations has the disadvantage of diluting the samples and decreasing the response of low-level components, as well as increasing the possibility of matrix interference. The choice of procedure to be used will depend on the demands of the analysis.
In comparison to standard LC-MS analysis of individual incubations, the most significant benefit afforded by our approach using mixtures is the marked savings in both instrument and analyst time. For a series of five drug candidates, 10 full–scan and 10 MS/MS chromatograms must be acquired for the individual incubations. In contrast, only two full–scan and three MS/MS chromatographic runs need be acquired for the mixed or the combined individual incubations. This reduces the analysis time 4-fold. Furthermore, the task of identifying possible metabolite peaks and setting up the MS/MS experiments manually takes from 4 to 6 h of analyst time, depending on the complexity of the chromatogram. The in-house AppleScript programs developed as part of this work reduce the analyst time to 1 to 2 h. For the studies presented here, the mixtures were restricted to five components. This is by no means the experimental limit. We have applied our approach to mixtures of up to 13 compounds with promising results. The actual upper limit will vary depending on the chromatographic and mass spectrometric properties of the compounds being assayed. As with any analytical measurement, the practical limits are determined by resolution and sensitivity. Thus, any advance that increases either of these two parameters should enable the number of drugs included in the mixture to be increased.
Correlation Analysis.
The use of correlation analysis makes metabolism studies with mixtures of substrates to support drug discovery feasible on a routine basis. Our correlation analysis approach has the advantage of requiring only product ion spectra from the starting substrate and each suspected product of metabolism, without the need to use libraries, training sets or metabolic mass shifts. Additionally, the product ion spectra do not need to be interpreted to identify substrate-related material from endogenous material or to establish substrate-metabolite relationships.
Correlation analysis of product ion spectra in our metabolism screening is based on two related assumptions, namely, that the pattern of ions found in the MS/MS spectrum reflects the structure of the molecule, and that the more similar two chemical structures are, the more similar their product ion spectra will be. Although the former assumption is generally true in all cases, the latter may not be. It is possible that the MS/MS characteristics of the metabolite differ appreciably from those of the parent drug, in which case the metabolite may not be detected. Within mixtures of related compounds, some metabolic cleavage products may be common to more than one of the parent drugs. The ranking of the values from the correlation analysis indicates which parent MS/MS spectrum is most similar to the metabolite spectrum. It does not, however, rule out the possibility that the metabolite may be formed from more than one parent drug. The analyst has to decide when further investigation is needed. Fortunately, correlation analysis indicates when such investigation might be necessary. As an example, consider the metabolite with an [M+H]+ at ion at m/z 290 (Table 4). Although correlation analysis correctly identified L-775,215 as the parent drug, two other substrates, L-773,417 and L-771,644, also had relatively high correlation values, suggesting the possibility that the metabolite may be derived from more than one parent drug. Comparison with individual incubations indicated that this metabolite, in fact, also was generated enzymatically from L-773,417, but not from L-771,644. To avoid having common metabolites in the mixture, the analyst could limit the substrates used in the mixtures to very different structural classes.
Finally, the spectral similarities deduced by correlation analysis must be considered in terms of the structure of the parent and metabolite before any final conclusion may be drawn on a mixture of unknowns. This requires a detailed interpretation of the parent MS/MS spectrum. Work is in progress to automate the time-consuming processes involved with interpreting low energy collision-induced dissociation MS/MS spectra.
Keeping pace with the increasing demand for metabolic studies requires new approaches to monitoring drug metabolism. We have shown that work with mixtures of drug candidates allows several compounds to be studied in the same amount of time that traditional approaches might take to analyze a single compound. The analysis of a mixture of substrates requires a means of establishing the relationship between substrate and metabolite in the presence of many other related substrates and metabolites. Correlation analysis of the product ion spectra is a highly effective means of distinguishing potential metabolites from endogenous material and assigning these metabolites to their respective parent drugs. Further development of this technology is ongoing, and is likely to have a major impact on the role of drug metabolism in the drug discovery process.
Acknowledgments
We thank the late Dr. John Gilbert for his support; Dr. Donald Slaughter for providing human liver microsomal preparations; Drs. Michael Patane and Jacob Hoffman for providing the Merck investigational compounds; and Maureen Hetzel for her help in the preparation of this manuscript.
Footnotes
-
Send reprint requests to: Dr. Carmen Fernández-Metzler, Merck Research Laboratories, WP75–200, West Point, PA 19486.
- Abbreviations used are::
- HPLC
- high–performance liquid chromatography
- LC-MS
- liquid chromatography with mass spectrometry
- LC-MS/MS
- liquid chromatography with tandem mass spectrometry
- ACN
- acetonitrile
- TIC
- total ion current chromatogram
- XIC
- extracted ion current chromatogram
- Received May 15, 1998.
- Accepted July 15, 1998.
- The American Society for Pharmacology and Experimental Therapeutics