Progress in Medicinal Cheniistry - Val. 40.
`Series Editors: F.D. King and A.W. Oxford
`Guest Editors: A.B. Reitz and S.L. Dax
`( ' 2002 Elsevier Science B.V. All rights reserved
`
`1 The Discovery and Development
`of Atorvastatin, a Potent Novel
`Hypolipidemic Agent
`
`BRUCE D. ROTH*
`
`Department qf Chemistry, Pfizer- Global Research and Development,
`Ann Arbor Laboratories, 2800 Plymouth Road, Ann Arbor,
`MI 481 70, U.S.A.
`
`ABSTRACT
`
`The search for potent and efficacious inhibitors of the enzyme HMC-CoA re-
`ductase (HMGRI) was the focus of considerable research in the 1980s. Building
`on the discovery of the fungal metabolite-derived inhibitors, mevastatin, lo-
`vastatin, pravastatin and simvastatin, a number of totally synthetic inhibitors
`were discovered and developed. This manuscript describes the discovery and
`development of one of those synthetic inhibitors, atovastatin calcium, currently
`marketed in the United States as LIPITOR". This inhibitor was designed based
`in part on molecular modeling comparisons of the structures of the fungal
`metabolites and other synthetically derived inhibitors. In addition to develop-
`ment of the structure-activity relationships which led to atorvastatin calcium,
`another critical aspect of the development of this area was the parallel im-
`provement in the chemistry required to prepare compounds of the increased
`synthetic complexity needed to potently inhibit this enzyme. Ultimately, the
`development of several chiral syntheses of enantiomerically pure atorvastatin
`calcium was accomplished through a collaborative effort between discovery
`and development. The impact of the progress of the required chemistry as well
`as external factors on internal decision-making with regards to the development
`of atorvastatin calcium will be discussed.
`
`*Tel (734) 622-7737; Fax (734) 622-3107
`
`1
`
`NCI Exhibit 2023
`Page 1 of 22
`
`

`
`2
`
`DISCOVERY AND DEVELOPMENT
`
`The biosynthesis of cholesterol from acetyl-CoA involves a process of more
`than 20 biosynthetic steps (Figure 1.1) [I]. This tightly controlled pathway is
`regulated by the levels of low-density lipoprotein (LDL)-receptors on liver cells
`as the means of ensuring whole-body cholesterol homeostasis [2]. It has been
`known since the late 1950’s and early 1960’s that inhibition of cholesterol
`biosynthesis was an effective means of lowering plasma cholesterol in both
`animals [3] and man [4]. What was unclear was whether it could be done
`safely. In fact there were many doubts based on the experience with the tri-
`paranol (MER-23, I, Figure 1.2) which caused cataracts in humans [5]. Despite
`this setback, the criteria for a safe and effective inhibitor of cholesterol bio-
`
`Acetyl CoA
`
`t
`
`HMO-COA
`
`Mevalonate
`
`lsopentenyl A Dimethallyl
`
`, I
`
`lsopentenyl
`Adenine
`(tRNA)
`
`t
`Farnesyl - proteins
`Pyrophosphate - Pyrophosphate
`
`ma pmtsln
`famesyrtmnsfererr
`
`famesylatd
`
`cis-prenyi
`trans fomae
`
`t
`t
`
`Ubiquinone
`
`Squalene
`
`+
`
`Cholesterol
`
`t
`
`Dolichol
`
`Figure
`
`I 1. The Cholesterol hio.synihelic pathway,
`
`NCI Exhibit 2023
`Page 2 of 22
`
`

`
`BRUCE D. ROTH
`
`3
`
`Figure 1.2. Tripurunol (MER-29).
`
`synthesis were clearly articulated by Curran and Azamoff in 1957 [6]. Their
`postulate was that safe inhibition could be achieved by blocking the pathway
`after the formation of acetoacetate, but prior to the formation of squalene.
`When in-depth mechanistic studies were performed with triparanol, not sur-
`prisingly, it was discovered that it broke this rule by inhibiting the pathway at
`the pentultimate step in the biosynthetic pathway, as evidenced by the accu-
`mulation of desmosterol in the plasma and tissues of patients treated with this
`drug [7]. Further studies demonstrated that it was also desmosterol which ac-
`cumulated in the lens of patients [8], emphasizing the potential dangers of
`inhibiting steps late in the biosynthetic pathway and causing medical concerns
`that would follow this area of research for decades. In fact, as recently as 1992,
`there were calls for a moratorium on the use of cholesterol-lowering drugs in
`primary prevention of myocardial infarction (MI) due to the lack of data from
`long-term clinical trials demonstrating a reduction not just in cardiovascular
`mortality, but in total mortality as well [9]. This concern was not completely
`alleviated until the results of the Scandinavian Simvastatin Survival Study were
`published in 1994 demonstrating reductions in total mortality with long-term
`statin treatment [lo].
`Despite the findings with triparanol, the search for cholesterol biosynthesis
`inhibitors continued unabated heled by the hope that inhibition pre-squalene
`would avoid the formation of non-metabolizable sterol intermediates, such as
`desmosterol, and result in a safe and effect treatment for hypercholester-
`olemia [I 1).
`The enzyme which became the focus of attention in the search for cholesterol
`biosynthesis inhibitors was 3-hydroxy-3-methylglutaryl-coenzyme A reductase
`(HMGR, EC 1.1.1.34),
`the rate-limiting and first committed step in the bio-
`synthetic pathway. This membrane-bound, endoplasmic reticulum localized
`enzyme catalyzes the two-step conversion of (S)-3-hydroxy-3-methylglutaryl-
`coenzyme A to 3-(R)-mevalonic acid through a putative hemi-thioacetal
`(Figure 1.3) [12]. Given that this hemi-thioacetal most likely represents a
`
`NCI Exhibit 2023
`Page 3 of 22
`
`

`
`4
`
`DISCOVERY A N D DEVELOPMENT
`
`transition-state intermediate, it should not have been surprising when screening
`of fermentation beers resulted in the isolation of compounds that closely mi-
`micked this structure.
`The first fermentation product identified was isolated almost simultaneously
`by two separate laboratories and given the name compactin (later changed to
`mevastatin) by the Beecham group, who isolated it as an antifungal from strains
`of the microorganism Penicillium brevicompactum and determined its mole-
`cular structure by x-ray crystallography (2, Figure 1.4) [ 131. The second group,
`from the Fermentation Research Laboratories at Sankyo, isolated the identical
`compound from cultures of Penicillium citrinum, but discovered it was a potent
`and competitive inhibitor of rat liver HMGR in vitro and sterol synthesis in
`vivo and gave it the code number ML-236B [14]. Further studies with ML-
`2368 demonstrated that it decreased serum total and LDL-cholesterol in dogs
`[ 151, monkeys [ 161, and human patients with heterozygous familial hyper-
`cholesterolemia [ 171. Shortly after the discovery of compactin, a second fungal
`metabolite (3), differing from compactin by a single methyl group was isolated
`from cultures of Aspergillus terreuS by workers at Merck [18] (and named
`mevinolin) and from Monascus ruber by the Sankyo group (and named
`Monacolin K) [19]. This compound was found to inhibit rat liver HMGR twice
`as potently as compactin (Ki of 0.6 nM vs 1.4 nM) [ 181 and was later renamed
`lovastatin. Ultimately, 3 would be the first HMGR inhibitor (HMGRI) approved
`by the U.S. Food and Drug Administration for the treatment of hypercholes-
`terolemia and would be marketed by Merck Sharpe and Dohme under the trade
`name Mevacot" . Two other potent fungal metabolite HMGRIs would ulti-
`mately become marketed drugs, pravastatin (4), produced by microbial hy-
`droxylation of compactin [20] and simvastatin (5) produced by synthetic
`modification of lovastatin [21]. Although all of these compounds were potent
`HMGR inhibitors and effective cholesterol-lowering agents, in the early 1980's
`concern over the viability of these compounds was created by the termination
`of the development of compactin in 1980 due to safety concerns created by
`results from preclinical toxicology experiments [22]. This apparently also led to
`a temporary suspension of the development of lovastatin. Thus, even though the
`fungal metabolites as a class would ultimately prove extremely safe and ef-
`fective in clinical trials, in the early 1980's there was at least a perceived need
`
`NCI Exhibit 2023
`Page 4 of 22
`
`

`
`BRUCE D. ROTH
`
`5
`
`HorYo
`
`2, compactin (mevastatin)
`
`3, lovastatin (MEVACORB)
`
`C02Na
`
`4, simvastatin (ZOCORS)
`
`5, pravastatin (PRAVACOLB)
`
`Figure I . 4. Fimgui mrtubolite inhibitors of'HMGR.
`
`for structurally novel HMGR inhibitors such that any non-mechanism related
`toxicity would be avoided.
`The first indication that the complex hexahydronaphthalene portion of the
`fungal metabolites could be replaced with a simpler ring system without loss of
`biological activity appeared in a patent application [23], then in publication
`form, from the Merck, Sharpe and Dohme Research Labs [24]. In this dis-
`closure, it was revealed that ortho-biphenyl containing 3,5-dihydroxy-6-hep-
`tenoic acids and their lactones, such as 6 (Figure 1.5), were equipotent to the
`fungal metabolites at inhibiting HMGR in vitro. This disclosure led us to de-
`velop the hypothesis that the key requirements for potent inhibition of HMGR
`were a mevalonolactone/3,5-dihydroxy-heptanoic or -6-heptenoic acid moiety
`and a large lipophilic group held in the correct spatial relationship by a spacer
`or template group [25]. If this were true, then virtually any ring system which
`fulfilled this requirement would lead to a series of potent inhibitors. This hy-
`pothesis was apparently shared by other laboratories and a large number of
`
`NCI Exhibit 2023
`Page 5 of 22
`
`

`
`6
`
`DISCOVERY AND DEVELOPMENT
`
`HoY-Jo
`
`Hoyro
`
`6
`
`Pyrrole Template
`
`Figure 1.5. HMGR inhibitor tetnplates.
`
`diverse series of inhibitors were discovered and developed based on this
`model [26].
`We selected the 1H-pyrrole ring system as our starting template to test this
`hypothesis, primarily because these could readily be prepared from 1,4-dike-
`tones through the classical Paal-Knorr condensation [27] (see retrosynthesis in
`Scheme I. I) and these 1,4-diketones, in turn, were potentially available pos-
`sessing a wide variety of I - and 4-substituents employing the thiazolium salt
`chemistry developed by Stetter [28].
`In practice, this scheme proved highly effective and a large number of I ,2,5-
`trisubstituted pyrroles were prepared using several omega-aminopropionitriles
`as the amine component to introduce a latent aldehyde. Unveiling of these
`latent aldehydes by DIBAL reduction followed by condensation with the dia-
`nion of methyl or ethyl acetoacetate employing the procedure of Weiler [29]
`introduced the remaining carbons needed in the targeted compounds. Un-
`fortunately, though expedient, this chemistry introduced the 5-hydroxyl as a
`racemic mixture, a problem that would need to be corrected later. Despite this
`less than optimal solution to the stereochemical requirements at C-5, we were
`able to control the relative configurations of the 3- and 5-hydroxyls by appli-
`cation of the predominantly syn-selective reduction of P-hydroxy ketones de-
`veloped by Narasaka and Pai [30]. In general, this protocol afforded
`approximately a 10: 1 ratio of syn/anti diasteromers. Lactonization by reflux in
`toluene produced the corresponding lactones from which the cis-diastereomer
`could be removed by recrystallization. The pure trans-diasteromers were then
`ring-opened by base hydrolysis to provide the biologically active dihydroxy-
`acids. This general synthesis is illustrated by the synthesis of the 2-(4-fluoro-
`phenyl)-5isopropyl analog (12) shown in Scheme 1.2.
`The initial question addressed was determination of the optimal spacing
`between the mevalonolactone and the pyrrole ring. This was rapidly narrowed
`
`NCI Exhibit 2023
`Page 6 of 22
`
`

`
`BRUCE D. ROTH
`
`H O T c o z R
`
`Pyrrole Template
`
`CHO
`
`R'i
`
`'Rp
`
`Scheme 1.1.
`
`Pvrrole inhibitor re frosynthesis.
`
`to a two-atom linker through the synthesis of a small group of analogs
`(Table 1.1).
`With this established, we next prepared a series of approximately thirty 2,5-
`disubstituted analogs possessing a range of substituted aromatic, cyclic, bran-
`ched and straight-chain aliphatic groups to define the optimal substituents at the
`2- and 5-positions. The conclusion from this exercise was that the distance
`across the pyrrole ring from the tip of the 2-substituent to the tip of the 5-
`substituent could be no longer than 10 angstroms with the size of the 2-sub-
`stituent being no more than 5.9 angstroms and the 5-substituent being no more
`than 3.3 angstroms. Further refinement of this analysis revealed that best
`potency was contained in compound 12 possessing a 4-fluorophenyl in the
`2-position and an isopropyl in the 5-position of the pyrrole ring [25].Un-
`fortunately, this compound still possessed only one-tenth of the inhibitory
`potency of mevastatin (Table 1.2). Taking into account the likely scenario that
`all of the biological activity was contained in one stereoisomer, we were still
`considerably short of the target potency and had come to the limit of what could
`be accomplished using the current synthetic route. In these circumstances, the
`options are to find alternate series or to attempt to ascertain the source of the
`deficiency. To this end, a simple molecular modeling exercise was undertaken
`to compare the differences between our best compound and those reported by
`
`NCI Exhibit 2023
`Page 7 of 22
`
`

`
`8
`
`DISCOVERY AND DEVELOPMENT
`
`“o\ CHO ++
`
`Ei3N
`
`Fq&
`
`HO
`
`58%
`
`1) H$I-CN
`HOAc, reflux. 73%
`2) DiBAL, 92%
`
`1
`NaH, n-BuLi, THF, -78’ -
`
`CH3COCH&O,CH,
`
`64%
`
`1) Bu3B, NaBH4,
`THF, -78’
`2) NaOH, H202
`
`PhCH3, reflux
`
`52% as a 97:3
`misture of
`transcis
`diastereorners
`
`Scheme 1.2. Synthesis of 1,2.5-irisubstiiuied pyrrole inhibitors.
`
`Merck. The simple overlay of these molecules (see Figure 1.6) revealed the
`presence of a methyl group in the Merck compound in a region of space not
`occupied by our inhibitors.
`To determine the importance of occupying this space, bromine and chlorines
`were introduced into the 3- and 4-positions of our most potent analog (12)
`employing the synthetic route described in Scheme 1.3 [31]. After testing the
`ability of these compounds to inhibit rat-liver HMGR, we were gratified to find
`that both compounds possessed inhibitory potencies comparable to the fungal
`metabolites (Tabfe 1.3).
`Although initially we were excited by this finding, the 3,4-dibromo
`analog 19 was taken into early preclinical development and rapidly found to
`display considerable toxicity [32]. As it turned out, much of the toxicology
`had been observed by others and was found to be specific to rodents or was
`derived from exaggerated pharmacology at high dosage levels and was most
`
`NCI Exhibit 2023
`Page 8 of 22
`
`

`
`BRUCE D. ROTH
`
`9
`
`Table I , I . OPTIMIZATION OF THE LINKER GROUP
`
`" O D 0
`
`X
`
`20
`
`24
`
`>I00
`
`53
`0.5
`
`I
`
`8
`
`9
`
`10
`I f
`
`'' Inhibition otI"C]-acetate conversion to cholesterol employing crude rat liver homogenate (ref. 24).
`
`Table 1.2. VARIATION AT THE PYRROLE 5-POSITION
`
`I I
`12
`13
`14
`15
`16
`17
`
`0.57
`0.40
`1.6
`20
`2.2
`17
`>I00
`
`"Inhibition of [I4C]-acetate conversion to cholesterol using a crude rat liver homoyenate. Mevastatin
`lCsr,= 0.026 pM (ref. 24).
`
`NCI Exhibit 2023
`Page 9 of 22
`
`

`
`DISCOVERY AND DEVELOPMENT
`
`CH3
`
`Figirre 1.6 Overluy of HMGR inhihitor teniplutrs
`
`severe with very bioavailable inhibitors which achieved high plasma and
`tissue concentrations [33,34]. Once again, we were faced with a decision
`point in the pyrrole series. Since we did not know whether the toxicity
`observed was related to the mechanism of action, the pyrrole series or the
`presence of the bromines in the 3- and 4-positions, rather than abandoning
`
`TBDMSO
`
`t-BuMepSiC1
`imidazole, DMF
`
`100%
`
`H
`
`H
`
`
`
`12
`
`
`
`NBS or NCS
`DMF. 0'
`100%
`
`H
`
`H
`
`I
`
`TBDMSO
`
`HOAc, THF
`
`35%
`
`x
`
`x
`
`
`
`x
`
`x
`
`
`
`Scheme 1 3. Sjw/hesis oj the 3,4-dihu/osuh.slituted unulogs
`
`NCI Exhibit 2023
`Page 10 of 22
`
`

`
`BRUCE D. ROTH
`
`Table 1.3. EFFECT OF HALOGEN SUBSTITUTION, AT THE PYRROLE
`3- AND 4-POSITIONS
`
`12
`18
`19
`2 (mevastatin)
`
`H
`CI
`Br
`
`0.23
`0 . 0 2 ~
`0.028
`0.030
`
`"Inhibition of the conversion of D, L-[ "C]-HMG-CoA to inevalonic acid using partially purified rat
`liver HMGR (ref. 30).
`
`the pyrrole series, a two-pronged approach was taken of both looking for
`alternative series [35,36] and synthesizing 3,4-non-halogen-substituted
`pyrroles in the hope that these compounds would retain activity, but lack
`toxicity. Unfortunately, the requirement for a penta-substituted pyrrole also
`required the development of an entirely new synthetic route to effectively
`develop the SAR at the 3- and 4-positions, since the existing route was
`limited only to those substituents that could be introduced by electrophilic
`substitution. A possible solution was presented through the 3 + 2 cycload-
`dition of azlactones and acetylenes pioneered by Huisgen [37]. This
`chemistry proved to be a very versatile means of preparing pentasubstituted
`pyrroles from a-amidoacids and acetylenes containing at least one electron
`withdrawing group (esters, nitriles, carboxamamides) [3 11. Although yields
`were best with acetylenes containing two electron withdrawing groups (e.g.,
`dimethylacetylene dicarboxylate), acceptable yields could be obtained with
`those possessing only one electron withdrawing group (4MO%). As
`significant, in the case of the unsymmetrical phenylacetylenes, considerable
`regiocontrol over the orientation of the 3- and 4-substituents could be
`achieved by adjustment of the substituents derived from the amide and
`amino acid precursors (Scheme 1.4). Using this methodology, followed by
`application of the Weiler dianion chemistry and stereoselective reduction
`used previously, a series of compounds were made with the already opti-
`mized 2-(4-fluorophenyl) and 5-isopropyl substitution and a variety of
`phenyl, substituted phenyl, ester, amide and nitriles at the 3- and 4-positions
`(Table 1.4).
`
`NCI Exhibit 2023
`Page 11 of 22
`
`

`
`12
`
`DISCOVERY AND DEVELOPMENT
`
`*
`
`HOOC 0
`
`Ac,O, reflux
`
`n
`O Y O
`
`n
`O Y O
`
`*
`
`Ac,O, reflux
`
`Schenic. 1.4 Rrgiocnntrnl in 3 + 2 cvclouddilion tnetlrulrd /yrrole synthesis.
`
`Table 1.4. SUMMARY OF 2-(4-FLUOROPHENYL)-5-ISOPROPYL-3,4-D1SUBSTlTUTED
`PY RROLES
`
`20
`21
`22
`23
`24
`24
`( + ) - 24
`(-)- 24
`2( mevastatin)
`
`C02Me
`C02Et
`Ph
`C02Et
`Ph
`Ph
`Ph
`Ph
`
`C02Me
`COzEt
`C02Et
`Ph
`C02CHZPh
`CONHPh
`CONHPh
`CONHPh
`
`0.18
`0.35
`0.17
`0.050
`0.040
`0.025
`0.007h
`0.44'
`0.030
`
`Inhibition of conversion of D,L -[I4C]-HMG-CoA to mevalonic acid using partially purified rat
`crude liver HMGR (ref. 30).
`Contaminated with 3% of (-)-24.
`"Contaminated with 3Yu of ( + )-24.
`
`NCI Exhibit 2023
`Page 12 of 22
`
`

`
`BRUCE D. ROTH
`
`13
`
`1,2,3,5-Tetrasubstituted analogs were available by application of the Stetter
`chemistry
`to substituted cinnamoylesters
`followed by decarboxylation
`(Scheme 1.5).
`Due to the difficulty in synthesis, a total of only 20 analogs were prepared,
`with best activity found in the 3-pheny1, 4-carboxamidophenyl analog (24).
`Separation of the two enantiomers of 24 by synthesis and separation of the
`diastereomeric R-a-methylbenzylamides followed by hydrolysis demonstrated
`that, as expected, all of the biological activity resided in one stereoisomer,
`(+)-24. This isomer was later confirmed to be the R,R-stereoisomer by total
`synthesis [3 I] and x-ray crystallography and found to possess inhibitory po-
`tency approaching that of simvastatin in vitro. Scale-up of this analog and
`preliminary testing in vivo in both casein-fed rabbit and cholestyramine-primed
`dog models of hypercholesterolemia demonstrated that ( + )-24 possessed po-
`tency and efficacy in vivo comparable to that found with lovastatin (un-
`published data). In subsequent studies done under more carehlly controlled
`conditions with larger groups of animals, it was determined that atorvastatin
`was actually more potent and efficacious than lovastatin at lowering LDL-
`cholesterol in rabbits [38] and guinea pigs [39] and triglycerides in rats [39].
`Having identified a potent and efficacious HMGR inhibitor, we were now
`faced with a critical decision, that of whether to develop our compound as the
`racemate or the pure stereoisomer. In fact, Sandoz when faced with this de-
`
`F
`
`n
`
`+
`
`
`
`Et3N
`
`/
`
`CHO
`
`0
`
`f i
`
`HO
`
`C02Me
`
`CH30H
`
`1 NaoH
`EiO - p-TSA. toluene
`
`EtoYoEt
`
`d
`
`25
`
`reflux. 71%
`
`30%
`
`Schrnie 1.5. Svntlie.vis of l,2.33 5-tetru substituted pyrro1e.s.
`
`NCI Exhibit 2023
`Page 13 of 22
`
`

`
`14
`
`DISCOVERY AND DEVELOPMENT
`
`cision in the development of fluvastatin chose to develop it as the racemate
`[26]. We chose to develop atorvastatin as the pure stereoisomer, for several
`reasons: 1) to avoid the unnecessary burden to the patient of having to meta-
`bolize 50% of possibly inert material (the wrong enantiomer) and 2) the desire
`to avoid having an obvious disadvantage (potency) in a compound entering the
`marketplace potentially 10 years after the fungal metabolite-derived inhibitors.
`Having made this decision, we formed two teams of chemists working in
`parallel towards a chiral synthesis, one in Discovery Chemistry in Ann Arbor
`and a second in Chemical Development in Holland, Michigan. The first chal-
`lenge was actually not the chiral synthesis, but scaling the achiral parts of the
`existing process that would be needed for the ultimate chiral synthesis. One of
`the initial problems was scaling the 3 + 2 cycloaddition reaction used pre-
`viously, in that, excess phenylamidocarbonyl phenylacetylene was required to
`achieve good yields, but this proved very difficult to separate from the product
`on large scale. Conceptually, the solution could be derived from the Paal-Knorr
`cyclization, if an appropriate amine would cyclodehydrate with the properly
`substituted 1,Cdiketone. This route would also open up the possibility of a
`convergent synthesis employing a fully elaborated side-chain ( Scheme I . 6).
`We therefore set about the preparation of the requisite I ,4-diketone using the
`Stetter methodology. However, we were disappointed to find that we were
`unable to achieve the desired cyclodehydration under a variety of conditions
`(Scheme 1.0. Fortunately, the Holland group had better success with this
`transformation (vide infra).
`Because of our inability to affect the Paal-Knorr condensation with the fully
`substituted diketone, as an alternative, we examined the synthesis of the tetra-
`substituted pyrrole 25 (Scheme 1.7) based on the assumption that the carbox-
`amide could be introduced later in the sequence. In the event, Paal-Knorr
`cyclization of the less highly substituted diketone proceeded smoothly to
`produce 25 in modest yield (Scheme I . I I ) . Subsequent introduction of the
`N-phenyl carboxamide proceeded smoothly by bromination with N-bromo-
`succinimide, followed by lithium halogen exchange and reaction of the re-
`sultant heteroaryl lithium with phenyl isocyanate. Hydrolysis then afforded the
`aldehyde 26 prepared previously employing the 3 + 2 cycloaddition protocol
`[3 I]. All of these transformations were scalable and proceeded in acceptable
`yield. Our strategy in Ann Arbor for introducing the 5-R-hydroxyl involved
`application of the diastereoselective aldol condensation of Braun [40] to al-
`dehyde 26 (Scheme I. 13). Thus, condensation of 26 with the magnesium dia-
`nion of S-( + )-2-acetoxy- l, l ,2-triphenylethanol afforded a 96:4 ratio of the S,R
`and S,S-diastereomers in 60% yield. This ratio could be improved to 98:2 with
`one recrystallization [3 11. Ester exchange with sodium methoxide followed by
`reaction with excess lithio-t-butylacetate afforded the R-6-hydroxy-P-ketoester
`made previously as the racemate. Reduction with Bu3B-NaBH4 as before the
`
`NCI Exhibit 2023
`Page 14 of 22
`
`

`
`BRUCE D. ROTH
`
`15
`
`a) Retrosynthesis -
`
`F%
`
`0 CONHPh
`
`+
`
`(+)-24
`
`a) Model system
`
`"0- +
`
`CHO
`
`CONHPh
`
`Et3N
`
`*
`
`0
`
`HO
`
`H~N-'~
`
`H2N
`
`CONHPh
`58%
`
`NO
`REACTION!
`
`f
`
`Scheme 1.6. Optinid retros-vnthesis vf (+ 1-24 and,failed model .mdl,.
`
`R S N , CH(0Et)Z
`
`afforded the syn-P,&dihydroxyester which after hydrolysis, acidification and
`lactonization afforded crude lactone ( + )-24. Fortuitously, the d,I-pair crys-
`tallized out of ethyl acetate-hexanes and 100% enatiomerically pure (+)-24
`could be isolated from the mother liquors.
`
`NCI Exhibit 2023
`Page 15 of 22
`
`

`
`16
`
`DISCOVERY AND DEVELOPMENT
`
`1) NBS. DMF, 0’. 100%
`2) n-BuLi. THF, PhNCO. 69%
`3) H30*, 86%
`
`*
`
`CJ
`
`25
`
`a ‘CoNHPh
`
`26
`
`1) J o J P h
`
`Ph
`
`Ph
`PLDA, MgBr2, -78’
`60%, 9 7 % ~
`
`2) NaOCH3
`
`Ph
`
`OLi
`
`2) Et3B, NaBH4
`
`(+)-24, >99%ee
`
`Although this route was successful in producing gram quantities of en-
`atiomerically pure ( + )-24, because of the linear nature of this route, the
`number of low-temperature reactions involved and the relatively low yields in
`some of the final steps such as the final purification, its potential for scale-up to
`provide the kilogram quantities needed for further development was low. Thus,
`for the synthesis to be economically viable, the Holland group was forced to
`develop an entirely different approach [41,42]. A critical component of this
`effort was an extensive investigation of the Paal-Knorr conducted by Alan
`Millar in Chemical Development which finally resulted in a successful cyclo-
`dehydration in the model system when a full equivalent of pivalic acid was
`used as catalyst (Scheme 1.8). This afforded pentasubstitued pyrrole 27 in 43%
`yield and demonstrated that a totally convergent synthesis was possible. This
`now became the ultimate goal.
`
`NCI Exhibit 2023
`Page 16 of 22
`
`

`
`BRUCE D. ROTH
`
`17
`
`______)
`1 equiv. pivalic acid
`THF. reflux, 43%
`
`CONHPh
`
`To this end, several routes passing through the known (S)-methyl-4-bromo-
`3-hydroxybutyrate 28, an intermediate used in prior syntheses of HMGRIs [43],
`were developed [41]. This key intermediate was derived most efficiently from
`isoascorbic acid as has been reported previously 143451, such that it was
`produced as a single stereoisomer (Scheme 1.9). Protection of 28 as the t-butyl-
`dimethylsilylether [43], followed by conversion to the nitrile provided an ad-
`vanced intermediate (29) that could be taken in several directions.
`Thus, 29 could be hydrolyzed to the acid and chain extended by activation
`with N,N-carbonyldiimidazole followed by reaction with the magnesium salt of
`potassium t-butyl malonate [46]. Acidification followed by deprotection with
`buffered fluoride afforded the 6-hydroxy-j3-ketoester 30 which was converted to
`the syn-l,3-diol employing NaBH4 and Et,BOMe, a slight modification of
`the original procedure [47]. Protection of the diol as the acetonide produced the
`nicely crystalline nitrile 31 in 65% yield and with diastereoselectivity in the
`range of 100: 1. One recrystallization improved this ratio to >350: 1. Reduction
`of the nitrile with molybdenum-doped Raney-Nickel catalyst then afforded the
`desired side-chain (32) with outstanding enantiomeric excess
`(>99.5)
`(Scheme I.IO) [41].
`An alternate, shorter route involved reaction of the alcohol derived from 29
`with 3 4 equivalents of lithium tert-butyl acetate to afford an excellent 7540%
`yield of hydroxyketone 30 without the need for prior protection of the alcohol
`and with no detectable reaction with the nitrile (Scheme I. I I ) . Although these
`routes still involved a low-temperature reduction, both could still be scaled to
`kilogram quantities [41].
`Cyclization of the fully functionalized, stereochemically pure side-chain 32
`with the filly substituted diketone under carefully defined conditions ( 1 eq.
`pivalic acid, I :4: 1 toluene-heptane-THF, Scheme 1.12) then afforded a 75%
`yield of pyrrole 33. Deprotection and formation of the hemi-calcium salt
`produced stereochemically pure atorvastatin calcium in a convergent, com-
`mercially viable manner which accomplished the original vision for the
`
`NCI Exhibit 2023
`Page 17 of 22
`
`

`
`18
`
`DISCOVERY AND DEVELOPMENT
`
`”%
`
`0 -
`OH
`
`HO
`
`lsoascorbic acid
`
`H202, CaC03
`
`KZc03
`
`OH
`HO&CO2K
`R :
`OH
`
`I
`J
`
`HBr, HOAc
`CHsOH
`
`1) t-BuMezSiCl
`imidazole, 4-DMAP
`
`2) NaCN, DMSO
`
`OTBDMS
`NC,),,CO~CH,
`
`R
`
`29
`
`Scheme I . 9. Svnthesis of (S)-methyl-4-bron~o-3-hv~rorvb~~~are
`28.
`
`synthesis developed in discovery chemistry, but was reduced to practice in
`chemical development.
`Although one might have expected that the decision to take atorvastatin
`calcium into clinical development would be straight-forward, it was not. By the
`time we completed the preclinical studies needed to file an Investigational New
`Drug Application (IND) with the Food and Drug Administration (FDA) in late
`1989, Mevacot‘”’, ZOCOT”, and PravacoI“ had all been approved for marketing
`by the FDA. Thus, we were faced with the expectation of coming into the
`marketplace nearly a decade after at least three HMGRIs and possibly more
`(LescoP was approved several years later by the FDA). Fortunately, by this
`time, evidence from preclinical efficacy studies was beginning to emerge
`suggesting that atorvastatin calcium may be more potent and efficacious than
`the fungal metabolite derived inhibitors at lowering total and LDL-cholesterol,
`at least in some animal models [38,39]. Encouraged by this positive data and
`
`NCI Exhibit 2023
`Page 18 of 22
`
`

`
`BRUCE D. ROTH
`
`19
`
`OTBDMS
`
`1) NaOH
`
`NC+2CH3
`
`29
`
`2) CDI, Mg(O&CHzCO@U),
`3) Bu~NF, HOAC, THF
`
`* NC&COzCH3
`
`OH 0
`
`R
`
`30
`
`1) NaBH,. Et,BOMe
`CHsOH. -90'
`2) (CH3)2C(OCH3)2
`CH~SOJH
`
`HZN
`
`0x0
`
`32
`
`H P , Ra-Ni,
`CH30H, 50 PSI
`95%
`
`O X 0
`
`NC*2CH3
`
`31, 65%
`
`Sclierne 1.10. C'kirtrl side-chrrin .synthe.~is.
`
`now having a scaleable process for synthesis of enantiomerically pure drug
`substance, the decision was taken by Dr. Ronnie Cresswell, then President of
`Parke-Davis Research, to move atorvastatin calcium into clinical trials in the
`hope that an improved efficacy profile would be observed in man over the then
`marketed drugs. To the delight of all those involved in the discovery and de-
`velopment of atorvastatin calcium, the merits of the drug were rapidly de-
`monstrated in the phase 1 clinical trials in healthy volunteers where reductions
`in LDL-C approaching 60% were observed at the high dose of 80 mg/day
`(Tuhk 1.5) [48]. This data provided the impetus for further development, since
`this level of efficacy was not achievable with other HMGRIs at approved doses
`or, in fact, with any other cholesterol-lowering drug. Since that original study in
`healthy volunteers, the outstanding potency and efficacy at lowering total
`cholesterol, LDL-cholesterol [49] and triglycerides [50] of atorvastatin cal-
`cium, now marketed in the United States as Lipitor", has been reproduced and
`confined in numerous clinical studies and in many thousands of patients
`[51,52]. Today it has brought benefit to inillions of patients and is one of the
`most widely prescribed pharmaceuticals in the world.
`
`OH
`
`NC&CO,CH~ R
`
`1) 3 equiv. LiCHpCOpt-Bu
`
`> NcJ,,J.,,co~cH~
`
`75%
`
`OH 0
`R
`
`30
`
`Sc,lienie 1.11. .4lte,nute rliisul de-chain sJwhesis.
`
`NCI Exhibit 2023
`Page 19 of 22
`
`

`
`20
`
`DISCOVERY AND DEVELOPMENT
`
`F%
`
`0 ‘ONHPh
`
`1 equiv. pivalic acid
`
`+
`X
`0 0
`
`*
`
`1:4:1 toluene-heptane-THF
`reflux, 75%
`
`H2N
`
`32
`
`33
`
`Scheme I . 12.
`
`Convergent. chirut synthe.fi.7 of atorvustution culcium
`
`Atorvastatin calcium
`
`Table I .5. MULTIPLE-DOSE TOLERANCE AND PHARMACOLOGIC EFFECT OF
`ATORVASTATIN CALCIUM IN HEALTHY VOLUNTEERS (REF. 48)
`
`Dose
`( m d 4
`
`Placebo
`10
`20
`40
`80
`
`% change (mg/dL),
`Total Cholesterol
`
`%t change (mgJdL)
`LDL-Cholesterol
`
`% change (mgJdL)
`Triglvcericles
`
`-3
`-22
`-30
`-36
`-45
`
`-3
`-3 1
`-39
`-47
`-58
`
`-3
`-12
`-30
`0
`-22
`
`NCI Exhibit 2023
`Page 20 of 22
`
`

`
`BRUCE D. ROTH
`
`REFERENCES
`
`21
`
`Bloch. K. Science 1965, l50(692), IP-28.
`Goldstein, J. L., Brown, M. S . J. Lipid Re.$. 1984, 25. 145M1
`Blohm, T. R., MacKenzie. R. D. Arch. Biochem. BiophjJs. 1959, XS. 245.
`Avigan, J.. Steinberg, D., Vroman, H. E., Thompson, M. J., Mosettig, E. J. B i d . Chein. 1960,
`235, 3123-3126.
`Laughlin, R. C., Carey, T. F. J. Anrer. Mecl. Assoc. 1962, 1 8 / , 339340.
`Curran, G. L., AzamoK D. L. Arch. Internu1 Med. 1958, 101. 68S689.
`Steinberg, D., Avigan. J. J. Biol. Chem. 1960, 23s. 3127-3129.
`Avigan. J., Steinberg, D.. Thompson, M. J., Mossettig, E. Biochem. Biophj~.~. Rrs. Conimun.
`1960, 2, 63-65,
`Davey-Smith, G.. Pekkanen, J. BMJ 1992, 304, 431434.
`Scandinavian Simvastatin Survival Study Group. Lancet 1994, 344, 1383-1 389.
`Boots, M. R., Boots, S. G., Noble, C. M., Guyer, K. E. J. Pharm. Sci, 1973, 62. 952.
`Rogers, D. kl., Panini, S. R., Rudney. H. . ~ - H ? . ~ r / ~ . ~ ~ - 3 - r n e t h ~ ~ l g l i f ~ u r ~ ~ /
`CfJW<LJM A Relrcruse:
`Sabin