REVIEWS
`
`·
`
`HPLC-@/M_S/1\!S: a powerful
`tool for integrating drug
`metabolism into the drug
`discovery process
`
`Walter A Korfmacher, Kathleen A Cox, Matthew S. Btyant,
`John Veals, Kwokei Ng, Robert Watkins and Chin-Chung Lin
`
`HPLC combined with atmospheric pressure ionization
`
`{API) mass spectrometry {MS) has become a very useful
`
`tool in the pharmaceutical industry. The technique of
`
`HPLC-API/MS/MS is becoming very important for both
`
`drug discovery and drug development programs. In the
`
`drug discovery area, it has three major uses: {1) rapid,
`
`quantitative method development,
`
`(2) metabolite
`
`identification, and (3) multi-drug analysis. The sensitivity
`
`of the API source and the selectivity provided by tandem
`
`mass spectrometry {MS/MS) enable rapid, quantitative
`
`method development for drugs in plasma. Early
`
`information on the metabolism of candidate drugs can
`
`guide structural modifications, thereby improving the
`
`activity and/or bioavailability.
`
`employed in selected cases for quantitative analysis. More
`recently, with the introduction of commercially available
`HPLC combined with atmospheric pressure ionization/
`tandem mass spectrometry
`(HPLC-API/MS/MS), both
`metabolite identification and quantitative analyses are being
`performed routinely using this new technique.
`Until recently, most analytical methods for the determi(cid:173)
`nation of a candidate drug were based on either GC or HPLC
`methods. The introduction of HPLC-APJ/MS/MS systems has
`provided new opportunities for rapid method development
`and metabolite identification, and this has helped to inte(cid:173)
`grate dmg metabolism into the drug discovery process.
`Recent articles have described HPLC/ API/MS/MS technol(cid:173)
`ogy and provided some examples of its utility for specific
`assays 1- 11 In this report, the utility of HPLC-APVMS/MS sys(cid:173)
`tems for integrflting drug metabolism into the dmg discov(cid:173)
`ery process will be discussed. A comparison wHl be made
`witl1 GC and HPLC methodologies.
`
`The use of mass spectrometry (MS) to suppo11
`
`metabolism studies of drugs in development is
`well documented1---6. In the past, various ·off-line'
`MS methods were employed for metabolite identi(cid:173)
`ficationi these procedures were typically reserved for candi(cid:173)
`date drugs in the development stage. In addition. GC-MS "''1S
`
`Rapid method development
`Currently) \'arious techniques are available to generate large
`numbers of compounds for biological screening, such as
`combinatorial synthesis and isolation from natural sources.
`The active compounds identified in these screens often
`require pharmacokinetic and in vitro metabolic evaluation.
`Over 100 compounds may have to be screened for their
`
`Walter A. Korfmacher•, Kathleen A. Cox, MatthewS. Bryant, John Veals, Kwokei Ng and Chin·Chung lin, Department of Drug
`Metabolism and Pharmacokinetics, and Robert Watkins. Department of Pharmacology, Schering-Piough Research Institute. 2015
`Galloping Hill Road, Kenilworth, NJ 07033·0539, USA. 'tel: + 1 908 2983183, fax: + 1 908 2983966, e-mail: walter.korfmacher@spcorp.com
`
`532
`
`Copyright© Elsevier Science Ltd. All rights reserved. 1359--6446/97/$17.00. Pll: $1359-6446(97)01121-5
`
`DDT Vol. 2, No. 12 December 1997
`
`MYLAN - EXHIBIT 1028
`
`

`
`
`
`
`
`
`
`in vitro stability or in vivo pharmacokinetic parameters before
`a lead compound can be chosen. These in vivo and in vitro
`studies require large numbers of samples in various biologi(cid:173)
`cal matrices that need to be subjected to quantitatiYe analy(cid:173)
`sis; rapid pharmacokioetic evaluation is therefore required.
`In the past, standard methods used various sampJe clean(cid:173)
`up methods (for example liquid/liquid phase extraction and
`solid-phase extraction) and GC or HPLC techniques for sepa(cid:173)
`ration and detection of compounds 12. As shown in Figure 1,
`these sample preparation methods typically required a sam(cid:173)
`ple volume of 200-1000 fd of plasma or serum. The sepa(cid:173)
`ration method depended on the nature of the compound,
`and a significant amount of time was required to obtain the
`proper analytical conditions necessary to resolve the various
`components in the matrix. Often, a derivatization l:itep wai:i
`necessary to obtain enhanced sensitivity. This multi-step
`procedure could take 1-3 weeks to develop.
`However, rapid method development (Figure 1) is poss(cid:173)
`ible when samples are analyzed by HPLC-APJ/MS/MS.
`
`Standard sample
`preparation
`
`Rapid sample
`preparation
`
`40 ~I plasma
`j
`Add 100 111 CH3CN
`+ Internal standard
`j
`Vortex
`j
`Centrifuge
`j
`Transfer to microvials
`I
`Analyze by HPLC-API/MS/MS
`
`200-1 , 000 r•l plasma
`I
`Add internal standard
`
`Liquid/liquid or solid-phase
`extraction
`j
`Evaporate to dryness
`j
`Add solvent
`j
`Vortex
`j
`Centrifuge
`j
`Transfer to microvials
`j
`Analyze by HPLC-UV or GC
`
`Method development
`time: 1-3 weeks
`
`Method development
`time: 1-2 days
`
`Figure 1. Comparison ofstandard sample
`preparation versus rapid sample preparation.
`
`L_-----------~--"
`
`REVIEWS
`
`Sample clean-up involves protein precipitation. This only
`requires 20--40 ~tl of plasma or semm. Current API source
`designs are well suited to accept hundreds of injections of
`this type of sample, with minimal source cleaning needed
`between sample sets. In addition, HPLC method develop(cid:173)
`ment is less demanding because of the added selectivity
`provided hy the mass spectrometer. Compounds with dif(cid:173)
`ferent molecular weights do not need to be resolved.
`Typically, 1 or 2 days is sufficient to develop a method for a
`compound that is part of an ongoing series of compounds
`heing screened for their pharmacokinetic parameters.
`By using one HPLC-API/MS/MS system, one opemtor can
`typically develop methods for two compounds and the
`methods used to an~lyze one set of samples for each com(cid:173)
`pound per week. This can result in a rapid turnaround of
`:)amples so that the required pharmacokinetic information
`can be provided to the drug discovery team in a timely man(cid:173)
`ner. Evolving strategies (see below) may help to improve
`this throughput in the near future.
`The two main reasons why HPLC-APJ/MS/MS allows
`rapid method development are the inherent selectivity and
`sensitivity of the instrumentation. The selectivity of HPLC(cid:173)
`API/MS/MS is on account of the specificity provided by tan(cid:173)
`dem mass spectrometry (MS/MS)". \V'hen used for quantita(cid:173)
`tive analysis, tandem mass spectrometty with a triple stage
`quadrupole mass spectrometer is used as follows. A specific
`ion is selected in the first quadmpole, typically the proton(cid:173)
`ated molecule (MH+) for the compound of interest. The
`:-;elected ion then enterS the collision cell (the second
`quadrupole) of the instmment and fragments into one or
`more smaller (product) ions. These fragment ions are ana(cid:173)
`lyzed in the third quadmpole and detected. Because of the
`mass selectivity provided by <1 tandem mass spectrometer,
`the high chromatographic resolution typically required for
`UV or fluorescence detection is not necessary. Usually,
`extensive scunple clean-up procedures are not required. The
`sensitivity of the HPLC-API/MS/AlS system is provided in
`part by the API source. For typical pharmaceutical mol(cid:173)
`ecules, assays can be readily developed for concentrations
`in the mnge of 10-25 ng/ml using only 40 r<l of plasma.
`Thus, preconcentration or derivatization techniques are
`generally not required.
`
`Serial bleeding in rats
`Previously, one rat per time point was the standard pro(cid:173)
`cedure for phannacokinetic studies in the rat. Currently1
`using HPLC-API/MS/MS, methods can be developed based
`
`DDT Vol. 2. No. 12 December 1997
`
`533
`
`

`
`
`
`
`
`
`
`REVIEWS
`
`on the analysis of 40 ~tl of plasma or serum (Figure ll". or
`even on 10--20 ~ll of plasma or serum if lower sample volumes
`are necessmy13. Because of these small sample volume re(cid:173)
`quirements, it is possible to analyze samples from rats that
`are subjected to serial bleeding. In serial hleeding, a single
`rat is dosed with the test compound and sampled at mul(cid:173)
`tiple time points (typically five to eight). At each time point
`a small volume of blood is removed (typically 200 ftl), For
`mean data, three rats can be dosed. The advantages of
`serial bleeding in rats are:
`
`• Fewer rats are needed to obtain phannacokinetic data
`on a compound.
`• Smaller amounts of the test compound are needed for
`dosing.
`• There is less variability in the data.
`• The entire set of pharmacokinetic parameters ( cm.1X' <lrea
`under cutve, tv2) are available for each animaL
`
`Figure 2 shows data obtained from (a) a serial bleeding
`study and (b) a multi-rat study for the same compound
`dosed orally at 10 mg/kg. It can be seen that there is a
`better correlation between the individual data and the mean
`data for the serial bleeding study than for the multi-rat study,
`
`Metabolite identification
`Previously, metabolite identification was reserved for com(cid:173)
`-pounds in the development phase. One reason for this wns
`that the standard method for metabolite identification relied
`on radiolabeled chug. As shown in Flgure 3~ the process of
`synthesizing the drug and collecting, purifying. and analyz(cid:173)
`ing the metabolites typically takes 2-4 months. This time
`frame is not acceptable by current dmg discovery standards.
`In the drug discovety phase, HPLC-API.'MS/MS can be used
`for metabolite identification. As shown in Figure 3, exten(cid:173)
`sive metabolite identification can be completed in 1 week
`In many cases, 1 day is sufficient to obtain useful infor(cid:173)
`mation on the plasma metabolites of a drug dosed in an
`experimental animal. A more extensive look at other fluid.s
`(urine, bile) or other tissues (e.g. brain, heart) could take 1
`or 2 weeks, Jn either case, significant amounts of infor(cid:173)
`mation on the metabolism of a compound can he obtained
`in a relatively short period of time.
`Metabolite identification in dn.1g discovety provides early
`information that can lead to stmcnu·al changes in the current
`lead compound, improving such pharmacokinetic param(cid:173)
`eters as oral bioavailability, half-life (11;2 ), or c;11~x· Often
`
`(a)
`
`10000
`
`l Ol
`-S
`c
`0
`
`~ -c "' L)
`
`c
`0
`0
`
`1000
`
`100
`
`0
`
`2
`4
`3
`Time after dosing (h)
`
`5
`
`6
`
`(b)
`
`10000
`
`1000
`
`'
`
`' •
`
`'
`+
`
`' •
`
`+
`+
`
`~
`-S
`c
`0
`~
`E
`1l c
`
`0
`0
`
`100
`
`'
`0
`
`2
`4
`3
`Time after dosing (h)
`
`5
`
`6
`
`Flgure2. Comparison of data obtained from (a) a
`serial bleeding study vem1s (b) a multi-mt study for
`the same compound dosed orally at 10 mg!kg.
`
`these parameters can be ch<mged by improving the metabolic
`stability of the compound, In order to improve metabolic
`stability, it is \·ery important to know how a compound is
`metabolized. The go:1l of drug discovery is to progress a
`lead compound into a final candidate drug that can he
`placed in the development stage. The traditional role of
`clmg metabolism in drug discovery was often limited and, jn
`rhe past, consisted mainly of producing a pharmacokinetic
`profile of lead compounds. Early metabolite identification
`can provide informmion on how to improve the metabolic
`stability of the lead stmcture. In this way, future lead com(cid:173)
`pounds might be a metabolite identified from the previous
`lead drug or an analog of the previous drug designed to
`block the major route of metabolism. In either case, metab(cid:173)
`olite information in early drug discovety may lead to a much
`faster progression from the early lead drug to the final
`candidate drug.
`
`534
`
`DDT Vol. 2, No. 12 December 1997
`
`

`
`
`
`
`
`
`
`REVIEWS
`
`Standard metabolite
`Identification
`
`Metabolite identification
`using HPLC·API/MS/MS
`
`Full scans
`(find MH+)
`
`Suspected metabolites
`
`Neutral loss scans
`
`//
`
`~
`
`Product ion scans
`
`I l
`
`Parent ron scans
`
`Make radiolabeled drug
`)
`Collect HPLC peaks
`)
`Purify
`)
`Analyze via MS
`
`I Time: 2-4 months I
`
`40 ~I plasma
`)
`Add 100 f'l CH3CN
`)
`Vortex
`
`Centrifuge
`)
`Analyze by HPLC-API/MS/MS
`I Time: 2-5 days
`
`Ftgut·e 3. Compattson of standard metabolite
`identification procedures versus metabolite detection
`and identification using HPLC-API!MS/MS.
`
`The methodology for using tandem mass spectrometry
`(MS/MS) to identity metabolites is well documentedl-5.'='.1·1;
`however, the application of this methodology to drug dis(cid:173)
`covel)' is a relatively recenl development. Figure 4 shows a
`general strategy for the usc of MS/IVIS for the detection and
`identification of metabolites without the need for a radio(cid:173)
`i<lbeled drug. A full-scan mass spectrum displays all masses
`detected O\'cr the desired mass range. This provides no
`structural information on the compounds detected, so there
`could be ~L1bstantial contributions from unrelated com(cid:173)
`pounds, solvent ion!-l or cbemk-al noise. Stmctural infor(cid:173)
`mation can be obtained on masses of interest by subjecting
`them to MStiVIS. The most common type of MS/MS experi(cid:173)
`ment is a product ion scan. Jn this case, ions of interest are
`isolated and fragmented into several product ions. These
`product ions provide characteristic structural information
`ahout the original ion. For example, a suspected metabolite
`ion would fragment into product ions that either resemhle
`the original parent dmg or differ from the product ion
`masses by the mass of the metabolite modification. A second
`MS/IvlS mode is a precursor ion scan. Here, a characteristic
`product ion is specified and all ions that fragment to form
`this product are monitored. In this manner, a characteristic
`fragment of the parent dmg can be selected to screen for
`possible metabolites that might produce the same fragment.
`The third iv1S/MS mode is a neutral loss scan. This is similar
`
`Flgut·e 4. General strategy for the use of tandem mass
`spectromeily (1119111$) for the detection and
`identification of metabolites.
`
`to a precursor ion scan except that the characteristic frag(cid:173)
`ment selected is a neutral species rather than an ionic
`species. Conjugated species such as glutathione or sulfate
`give characteristic neutral losses when fragmented in a mass
`spectrometer; therefore, these metabolites can often be
`detected with a neutral loss scan.
`In addition to the identification of metabolites of lead
`. compounds, it is ve1y important to give some estimate of the
`levels of these metabolites compared with the dosed drug, If
`a metabolite is minor (1 o/o or less of the dosed drug), it may
`not be important to block its route of metabolism. If the
`metabolite is major (>20% of the dosed drug), it may
`become the new lead drug, or modification of the site of
`metabolism may lead to a better compound. The best way to
`quantitate a metabolite is to compare its response with an
`auth~ntk standard of the metabolite (as is done with the
`dosed dmg). However, a metabolite standard is not usually
`available, and the only- estimate that can be made is to use
`the calibration cu1ve for the dosed drug to estimate the level
`of the metabolite. It is known that MS responses for various
`compounds can be quite different; but, in many cases, the
`response of the metabolite wm be within a factor of two to
`three of the dosed drug. Thus, while this method of estimat(cid:173)
`ing the le,·el of one or more metabolites is not ideal, it can
`provide useful information.
`Figure 5 shows a mass chromatogram of drug B dosed at
`10 mg/kg orally in the rat. The top trace shows the internal
`standard, which is a structurally similar analog that is used
`for quantitation. The second trace shows the peak for drug
`B. The third trace shows three IV1+16 metabolites that were
`monitored for this compound. The bottom trace is the
`reconstructed ion chromatogram which is the sum of the top
`
`DDT Vol. 2, No. 12 December 1997
`
`535
`
`

`
`
`
`
`
`
`
`REVIEWS
`
`m/z ~ 183.2; DA of 456.50> 457.50 ~
`1 :~
`1
`1 ~~f/z~411.2;0Aof437.50>438.50 ~
`
`100
`
`5~]
`
`m/z = 427; DA of 453.50> 454.50
`
`.
`
`AM
`
`c (\
`
`(a)
`
`(b)
`
`(c)
`
`100 RIC
`
`(d) 5~1
`
`A
`
`6:40
`
`8:20
`
`1:40
`
`3:20
`
`5:00
`
`Time (min)
`
`5
`r3.823 X E
`
`5
`r 1.226 X E
`
`3
`r6.427 X E
`
`5
`r1.334 X E
`
`Figure 5. Mass cbromatogran~>from the HPLC-APIIMS/MS cma(ysis of an
`extract of a rat plasma sample jinm a 10 mglkg om/ dosing stut{)''
`(a) internal standard, (b) dosed druR. (c) peaks (designated A,B and C)
`from three M+16 metabolites, (d) reconstntcted ion chromatogram (RIC);
`DA ~ daughter ion.
`
`1000
`
`three traces. Each mass chromatographic trace is normalized
`to the highest peak. This figure shows the excellent selec(cid:173)
`tivity of the HPLC-API/MS/MS technique: this is typical of
`the type of data that can be obtained by this procedure.
`Figure 6 shows an example of a discovety drug that was
`extensively metabolized. The orally dosed drug (parent)
`reached a maximum level of Jess than 100 ng/ml. Three of
`its metabolites reached levels that were estimated to be sig(cid:173)
`nificantly higher than the parent drug. A fourth metabolite
`showed very low levels. This information provided impor(cid:173)
`tant clues concerning what part of the molecule needed he
`modified in order to improve the metabolic stability of the
`compound,
`
`Multiple analysis
`Another advantage of the HPLC-APJ/MS/MS technology is
`that it can be used for multiple analyses, \vhich can be
`defjned as the determination of more than one compound in
`one chromatographic assay. Use of this capability is a recent
`development for HPLC-API/MS/MS. This feature can be
`utilized in at least two different ways:
`
`• analysis of multiple studies and
`• multi-drug study analysis.
`
`In either case, the analytical chal(cid:173)
`lenge is the same.
`In multiple analysis, spiked plasma
`(or serum, etc.) calibration samples
`are prepared with more than one
`analyte. The simplest example is two
`analytes, although the analysis of
`samples containing as many as 10 or
`20 analytes bas been reported!'. The
`basic principle is the same regardless
`of the number. Figure 7 shows the
`results of the analysis of four drugs
`plus one internal standard in a single
`HPLC-API/MS/MS assay, in which
`each analyte has a unique MS/MS
`'analytical window'. The four drugs
`are compounds with similar struc(cid:173)
`tures that are spiked at 50 ng/ml each
`into rat plasma, resulting in four cali(cid:173)
`bration cu1ves (one for each of the
`four drugs) for the single analytical
`mn. Thus, in the same analytical run,
`samples from one or more single
`
`--------- Parent
`--"'-- M + 16 (B)
`·-·e-·- M-82
`- .. +- .. M + 16 (A)
`---Q-- M-14
`
`~
`5
`·~
`~ c:
`
`0
`0
`
`10
`
`<j~'{:>--~-
`
`-~-
`
`1
`0
`
`4
`3
`2
`Time after dosing (h)
`
`5
`
`6
`
`Figut·e 6. Results from the ana(J~is of samples
`obtaimd from a discovery drug that was dosed oral()'
`at 10 mglkg; the dosed compound levels are compared
`with estimated metabolite levels. Jhe metabolites are
`identified i<l' their mass diJ!'ermce from the drug. Jhe
`two AI+ 16 metabolites are distinguished as A and B.
`
`[
`
`536
`
`DDT Vol. 2, No. 12 December 1997
`
`

`
`
`
`
`
`
`
`0
`
`0
`
`(b)
`
`(c)
`
`(d)
`
`(e)
`
`(a) 1 OO (IC; DA of 590.29
`1 OO (IC; DA of 560.29
`1 O~ (IC; DA of 544.29
`1 O~ (IC; DA of 540.29
`1 o: riC; DA of 61 0.29
`(f) 10: ('c
`
`L
`
`h "
`
`A
`~
`
`A
`
`[1.131 X E4
`
`[1.579xE4
`
`r8.477 X E'
`
`r2.884 X E'
`
`[7.112xE3
`
`[3.011 X E5
`
`1:40
`
`3:20
`
`5:00
`
`&
`
`6:40
`
`8:20
`
`Time (min)
`
`Figure 7. Mass chromatogmmsfrom the HPLC-APIIMS/;JJS ana!J•sis of an
`extmct of a rat plasma sample spiked u•ith four drugs (tmces a. b. d, e) at a
`level of 50 nginzl each plus tbe internal standard (c); trace( ts the
`reconstructed ion chromatogram (RIC); DA = daughter ion.
`
`REVIEWS
`
`utility of multi-drug studies is still
`unclear.
`
`Conclusion
`HPLC-API/MS/MS has been shown to
`have an important role in integrating
`dmg metabolism into the drug dis(cid:173)
`covery process. The technique has
`three primary uses:
`
`• rapid method development,
`• metabolite identitlcation,
`• multi-component analysis.
`
`In addition, because of its sensitivity,
`the technique can be used for the
`analysis of the low volume samples
`obtained from serial bleeding studies
`vdth rats. These capabilities guaran(cid:173)
`tee that HPLC-API!MS/MS will con(cid:173)
`tinue to be the technique of choice
`for drug discovery method develop(cid:173)
`ment and study sample analysis for
`the foreseeable future.
`
`dose studies can be analyzed by monitoring the four ana(cid:173)
`Iytes and one internal standard. Alternatively, samples from
`a multi-drug study (i.e. the four dn1gs dosed in one animaD
`can also be analyzed using this technique.
`In a multi-drug .study, more than one compound is dosed
`~imultaneously in a JaboratOJY animaJI5-17. These studies
`are also referred to a.s combinatorial pharmacokinetic
`studies, ·cassette dosing' or 'N-in-one' dosing studies 10. The
`advantage of a multi-drug study is that more compounds
`can be dosed in a shorter time with fewer anim<1ls used.
`Another advantage is that the sample analysis is more effi(cid:173)
`cient. On the other h<md, a disadvantage is that dmg-drug
`interactions may lead
`to misleading pharmacokinetic
`results 16. One way to minimize potential drug--dmg inter(cid:173)
`actions is to reduce the dosing leveL A recent approach is
`to dose at 10/n mg/kg, where n is the number of com(cid:173)
`pounds being dosed. Another disadvantage is that method
`development and sample calculation time is increased over
`the more common single dosing/single assay studies. A third
`disadvantage is that metabolite identification is not practical
`with multi-drug studies. A metabolite may have the same
`mass as another drug in the ttssay. For these reasons, the
`
`REFERENCES
`Kd:wk P. ·,md T~mg. L 0993) .-Jna/. Cbem. 2.2, 9!2.A-986A
`2 Gelpi. E. (1995) J. Chroma/ogr. 703. S9-80
`3 Ower. TIL Let>. E.D.;-~nd Henion. J.D. f J9H6) A no!. Chem. 'lR,
`2453-2..j61J
`-1 Korfnr.tcher, \Y/.A. e/ a{. 0995) in Encyclopedia of Aua{l•rtcal Science,
`pp . .'lll:F-303-J. At":J()emk !lre,.;.~
`') Korfmacher. \V.A. era!. ( 1990) Blomed. !im•iron. ,If ass SjXcfrom. 19,
`
`191-:~0I
`6 Bailie. TA. ( 1992) lni.J. Jlass Spec! rom. /Oil Process. llR-'119,
`2..'-\9---31-1
`..., Allen, G.D. e/ a/. (19%) LCGCH. 510-5I·i
`H \'olme1~ DA and \'ollmer, D.l. (11)96) LG-G'C H. 236-2-t2
`9 Dulik. D.J\l. C'l a/. 09%) in Mass SfJ¥ctromeiiJ' in tbe Biological Scieuces
`fBurlinganw. A. :md Carr, S . .A. .. eds). pp. --12H29. Human~t
`10 Olah. T.V. el af. 099·1)]. Pbe1rm. Biomed. Anal. 11. 705--712
`I 1 Constanzer. M. cf ul. 0995)]. Cbromata,t:p: 666. 117-126
`12
`B1)~1Dt. lil.S. e/ a/. 0997)}. Chromatogr. 7]7, 6H)6
`13 Chilton. A. c/ u/. (1997) Proceedings of the 45th ASMS Conference on Mass
`Spectmmclry and Allied Topic_,~ 1-5 june (in press)
`l'l de Hoffm;mn, E. (1996) .f. Ilfasf Spectrom. 31, 129-137
`!-5 Halm. K. (1996J Proceedings of the 44/b ASJIS Conj'en?JICE 011 Mass
`j]>ec/romei!J'tmd .41/led Topics, 12-16 May. p. 146S
`flennan, J. el a/. 099?) .f. Med. Chem. 40, 829----~331
`16
`17 Olah. T.\'. er nl. \1997) Rapid Comnum. Mass Spectrom. 11. 17-23
`
`DDT Vol. 2, No. 12 December 1997
`
`537