SAXA-DEF-00324
`
`AstraZeneca Exhibit 2003
`Mylan v. AstraZeneca
`IPR2015-01340
`
`Page 1 of 26
`
`

`
`SAXA-DEF-00325
`
`Page 2 of 26
`
`Page 2 of 26
`
`

`
`CHAPTER 1
`
`The Discovery of the Dipeptidyl
`Peptidase-4 (DPP4) Inhibitor
`Onglyzat: From Concept to
`Market
`
`JEFFREY A. ROBL AND LAWRENCE G. HAMANN
`
`Bristol-Myers Squibb Research & Development, Department of Discovery
`Chemistry – Metabolic Diseases, P.O. Box 5400, Princeton, NJ 08543, USA
`
`1.1 Introduction
`
`The prevalence of diabetes in developed and now emerging countries represents
`a significant health burden to a large portion of the world’s population. Type-2
`diabetic patients, characterized in part by elevated fasting plasma glucose of
`1 (7.0 mmol L
`1) and glycosylated hemoglobin (HbA1c) Z 6%,
`4125 mg dL
`are at increased risk for the development of both microvascular (retinopathy,
`neuropathy, nephropathy) as well as macrovascular complications (myocardial
`infarction, stroke). As such, diabetes is the leading cause of blindness, kidney
`failure, and limb amputation worldwide.1 Diabetes is a progressive disease,
`with morbidity and mortality risk increasing with both duration and severity of
`hyperglycemia. Additionally, diabetes is also now impacting different popula-
`tion sectors (adolescents, developing countries) not typically associated with
`the disease 30 years ago. Consequently, the continually increasing diabetes
`prevalence is placing greater strain on both health care systems and economies
`on a global scale. In 2007 alone, studies have shown that diabetes cost the US
`
`RSC Drug Discovery Series No. 4
`Accounts in Drug Discovery: Case Studies in Medicinal Chemistry
`Edited by Joel C. Barrish, Percy H. Carter, Peter T. W. Cheng and Robert Zahler
`r Royal Society of Chemistry 2011
`Published by the Royal Society of Chemistry, www.rsc.org
`
`1
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`SAXA-DEF-00326
`
`Page 3 of 26
`
`Page 3 of 26
`
`

`
`Cl
`
`OEt
`
`Chapter 1
`
`HO
`
`H2N
`
`3
`
`N
`
`O
`
`CN
`
`H
`
`H O
`
`2
`
`O
`
`OH
`
`HO
`
`HO
`
`N
`
`CO2H
`
`O
`
`O
`
`O
`
`N
`
`e
`
`O M
`
`1
`
`OMe
`
`2
`
`Figure 1.1 Late-stage development candidates from the Bristol-Myers Squibb dia-
`betes research portfolio.
`
`economy $174 billion in medical expenses and lost productivity.2 While death
`rates related to heart disease, stroke, and cancer have all decreased since 1987,
`the death rate due to diabetes has increased by 45% during this same period.3
`Thus, the discovery and development of new therapies for treating and pre-
`venting diabetes continue to be a major emphasis of health care companies.
`In response to this landscape, the Discovery organization at Bristol-Myers
`Squibb (BMS) made the strategic decision to refocus efforts in the late 1990’s
`towards identifying and progressing novel targets for the treatment of diabetes.
`This was in part aimed at building upon BMS’s already established presence in
`the anti-diabetes market through the Glucophaget franchise and in recognition
`of the significant unmet medical need for novel, more efficacious, and well
`tolerated treatments for the disease. It was from these efforts that advanced
`clinical candidates such as muraglitazar (1, Pargluvat, dual PPAR agonist),4
`dapagliflozin (2, SGLT2 inhibitor),5 and saxagliptin (3, Onglyzat, DPP4 inhi-
`bitor)6 were discovered within the BMS Discovery organization (Figure 1.1).
`
`1.2 Modulation of GLP-1 in the Treatment of
`Diabetes
`
`At the start of this effort, several oral anti-diabetic agents (OADs) were
`available to patients suffering from type-2 diabetes. These included hepatic
`glucose suppressors (e.g. metformin), insulin secretagogues (e.g. sulfonylureas),
`glucose absorption inhibitors (e.g. acarbose), and insulin sensitizers (e.g. thia-
`zolidinediones or TZDs such as rosiglitazone and pioglitazone). While all have
`shown utility in lowering HbA1c levels in diabetic patients, current OADs come
`with a variety of safety and/or tolerability issues. The biguanindes such as
`metformin, currently the most widely prescribed therapy for diabetes, have
`issues related to gastrointestinal (GI) tolerability and lactic acidosis.7 Sulfo-
`nylurea treatment is often accompanied by higher incidences of hypoglycemia
`and weight gain,8 while glucose absorption inhibitors exhibit modest efficacy
`and GI disturbance.9 Finally, TZDs have been associated with edema, wor-
`sening of congestive heart failure, negative effects on bone fracture rate, and, in
`recent studies, mixed results regarding cardiovascular (CV) safety profiles.10
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00327
`
`Page 4 of 26
`
`Page 4 of 26
`
`

`
`The Discovery of the Dipeptidyl Peptidase-4 (DPP4) Inhibitor Onglyzat
`
`3
`
`With this background in mind, we sought to identify new targets which would
`not only provide an efficacious alternative mechanism for lowering blood
`glucose and HbA1c levels, but would also present an opportunity for achieving
`a superior safety and tolerability profile when compared to current standards of
`care. Ideally, such a drug would be suitable for combination with existing
`agents, as poly-pharmacology with multiple OADs is emerging as the standard
`treatment paradigm for type-2 diabetes therapy.
`Glucagon like peptide-1 (GLP-1) is a 30-amino acid peptide incretin hor-
`mone derived from processing of pro-glucagon and is secreted by the L-cells of
`the intestinal mucosa in response to glucose stimulation. Since the early 1990’s,
`GLP-1 had been known to be a potent insulin secretagogue and glucagon
`suppressor, with robust anti-diabetic and pro-satiety effects in diabetic
`humans,11,12 but efforts to advance GLP-1 itself as a pharmaceutical agent
`were hampered by its extremely short pharmacokinetic half-life in vivo (plasma
`t1/2E2 min). As a result, considerable effort in the drug discovery community
`was expended toward the identification of small-molecule GLP-1 receptor
`agonists that would capture the beneficial effects of GLP-1 while exhibiting oral
`bioavailability and a superior pharmacokinetic duration of action. Unfortu-
`nately, efforts to identify such small-molecule agonists have to date been
`unsuccessful, due in part to a dearth of viable bona fide screening hits.13 In light
`of this shortcoming, a number of pharmaceutical and biotech companies have
`advanced subcutaneously administered, peptide GLP-1 receptor agonists with
`superior duration of action in vivo. Among the most advanced agents are
`exenatide (Byettat)14 and liraglutide (Victozat),15 both of which have been
`approved by regulatory agencies for the treatment of type-2 diabetes. While
`these drugs are effective in lowering HbA1c and demonstrate a net beneficial
`effect on weight gain and other CV risk factors, they require parenteral
`administration (once or twice daily dosing), and patient uptake of these agents
`has been limited despite their robust efficacy and promising safety profile.
`
`1.3 Dipeptidyl Peptidase-4 as a Target for Diabetes
`Treatment
`
`While the advancement of orally active, small-molecule GLP-1 receptor ago-
`nists remains elusive, another opportunity to modulate GLP-1 receptor activity
`in vivo focused on preventing the degradation of endogenous GLP-1 with small-
`molecule inhibitors of the primary peptidase responsible for the in vivo
`degradation of GLP-1, dipeptidyl peptidase-4 (DPP4), a non-classical serine
`protease.16 Our initial interest in DPP4 inhibitors was piqued by a report from
`Holst and Deacon, wherein the authors outlined a compelling argument for the
`utility of DPP4 inhibition in the treatment of diabetes, primarily via the
`potentiation of endogenous GLP-1.17 DPP4 belongs to a family of aminodi-
`peptidases and is both a cell surface and circulating enzyme. Historically, it had
`also been identified as the lymphocyte cell surface marker CD26, and as such
`DPP4/CD26 exhibits pharmacology related to cell membrane-associated
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00328
`
`Page 5 of 26
`
`Page 5 of 26
`
`

`
`4
`
`Chapter 1
`
`activation of intracellular signal transduction pathways and cell–cell interac-
`tion in addition to its peptidase enzymatic activity.18 It is widely believed that
`the signaling function of DPP4/CD26 is distinct from its enzymatic function.
`The concept of generating targeted protease/peptidase inhibitors as thera-
`peutic agents is well documented in the literature.19 In the majority of cases,
`selective enzyme inhibitors have been used to prevent the conversion of an
`endogenous ‘‘non-functional’’ peptide/protein precursor (e.g. angiotensin I) to
`a physiologically active peptide/protein (e.g. angiotensin II), thereby attenu-
`ating formation of the protagonist bio-active enzyme product to effect ameli-
`oration of the disease state. Use of such approaches has led to marketed drugs
`for numerous indications, including ACE and renin inhibitors for hyperten-
`sion,20 HIV protease inhibitors for AIDS,21 and thrombin inhibitors for
`DVT.22 Less prevalent are approaches targeting proteases/peptidases that
`degrade endogenous substrates which are known to exert a beneficial effect. For
`example, neutral endopeptidase (NEP) proteolytically degrades the endogen-
`ous vasodilator atrial natriuretic peptide (ANP) to inactive fragments. By
`retarding this degradation, NEP inhibitors have found use in the treatment of
`hypertension.23 In common with NEP, where inhibition of protease mediated
`protein degradation was the pharmacological objective, BMS and several other
`research groups engaged in the search for DPP4 inhibitors to maximize the
`beneficial effects of endogenously released GLP-1.
`
`1.4 Early Inhibitors of DPP4
`
`To jump-start the BMS DPP4 chemistry program, the group was able to
`capitalize on groundwork laid in the mid-late 1990’s when several potent
`inhibitors of DPP4/CD26 were reported that could be classified as either
`‘‘irreversible’’ or ‘‘reversible’’, depending on the mechanism of inhibition
`(Figures 1.2 and 1.3). Hydroxamates such as 4 were proposed to be both
`substrates and inhibitors of DPP4, presumably via direct covalent modification
`of the enzyme through the active-site serine residue (Ser630).24 Phosphate-
`based inhibitors such as 5 were also reported to undergo covalent addition to
`DPP4 but exhibited weak potency.25 The interesting boronate-based inhibitors
`(e.g. 6), originally advanced by the Tufts University and Boehringer Ingelheim
`groups, exhibited exceptionally high inhibitory activity in vitro, presenting
`
`N N
`
`7
`
`B
`
`OH
`OH
`
`H H
`
`Me
`
`N
`
`O
`
`P
`
`O
`O
`
`O
`
`H2N
`NO2
`
`5
`
`Cl
`
`O
`
`H2N
`
`Cl
`
`N
`
`HO
`
`O
`
`6
`
`B
`
`OH
`
`O
`
`O
`
`NH
`
`4
`
`H2N
`
`Me
`
`N
`
`O
`
`O
`
`Figure 1.2 Early examples of ‘‘irreversible’’ DPP4 inhibitors.
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00329
`
`Page 6 of 26
`
`Page 6 of 26
`
`

`
`The Discovery of the Dipeptidyl Peptidase-4 (DPP4) Inhibitor Onglyzat
`
`5
`
`N
`
`O
`
`CN
`
`NH
`
`HN
`
`N
`
`11 (X = CF3)
`12 (X = CN)
`
`N
`
`H2N
`
`H2N
`
`O
`
`8
`
`S
`
`N
`
`O
`
`9
`
`H2N
`
`N
`
`O
`
`10
`
`CN
`
`X
`
`O
`
`N
`
`O
`
`N
`
`O
`
`NH
`
`14
`
`NH
`
`NH
`
`13
`
`Figure 1.3 Early examples of ‘‘reversible’’ DPP4 inhibitors
`
`themselves as ‘‘transition-state’’ inhibitors which presumably form tetrahedral
`boronate esters involving the Ser630 hydroxyl group.26–28 As such, many of
`these compounds exhibited slow tight-binding kinetics with Koff rates of several
`days (versus seconds/minutes with non-covalently bound inhibitors). However,
`these compounds also suffered from poor solution stability arising from
`intramolecular cyclization of the terminal primary amine with the boronate,
`affording an inactive product (e.g. 7). The propensity of compounds of this
`chemotype to undergo internal cyclization limited their viability as drug
`candidates.
`Due to the uncertain risks associated with advancing irreversible inhibitors as
`drug candidates, the team viewed the reversible inhibitors exploited by the
`Mount Sinai, Probiodrug, Ferring Research, and Novartis scientific teams as
`more attractive starting points for a lead finding effort (Figure 1.3). Probiodrug
`had described simple Ile-pyrrolidides (8) and Ile-thiazolidides (9, later advanced
`to the clinic by Merck and Probiodrug as P32/98) which exhibited in vitro
`potency in the nanomolar range, were chemically stable, and, in the case of 9,
`demonstrated glucose area under the curve (AUC) reductions in Zuckerfa/fa rats
`in an oral glucose tolerance test (oGTT).29–31 Reports from Li et al. at Mount
`Sinai highlighted early examples of nitrilo-pyrrolidines specifically designed as
`inhibitors of DPP4.32 Equally intriguing was the work described by Ferring in
`which nitrilo-pyrrolidines such as 10 were identified as exceptionally potent
`inhibitors of the enzyme.33,34 Initially targeted as agents for immunomodula-
`tion (via CD26 inhibition), these compounds represented ‘‘drug-like’’ scaffolds
`and exhibited exceptional
`inhibitory potency. Additionally, a contempo-
`raneous patent application from Novartis35 described the structure of com-
`pound 11 (a related analogue 12 would later be disclosed as Novartis’ first
`clinical compound, DPP-728)36 and its ability to increase plasma insulin in
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00330
`
`Page 7 of 26
`
`Page 7 of 26
`
`

`
`6
`
`Chapter 1
`
`fasted, high-fat fed rats in an oGTT. The inhibitors represented by compounds
`8–12 provided useful insights for the design of DPP4 inhibitors at BMS.
`At the initiation of the program, there were still many unknown factors
`related to the pharmacology and safety of DPP4 inhibition. It was clear from
`earlier work that Fischer 344 rats possessing a naturally occurring loss-of-
`function mutation of the DPP4 gene were healthy, viable, and free of serious
`immunological complications.37 It was later shown that these rats also exhib-
`ited a favorable metabolic phenotype on a high-fat diet and demonstrated
`improved glucose tolerance and GLP-1 secretion.38 Thus, complete ablation of
`DPP4 did not appear to represent a serious safety concern in rats, but rather the
`Fischer rat provided support for the concept that inhibition of DPP4 could be
`both safe and efficacious. Others questions still remained. Was high selectivity
`for DPP4 versus other related peptidases (e.g. DPP8, DPP9, FAP, etc.) an
`absolute requirement for this target? Would inhibition of DPP4 potentiate
`other endogenous peptides, leading to unintended deleterious (or beneficial)
`consequences? Would potentiating endogenous GLP-1 (versus exogenous
`administration) be sufficient to affect a robust anti-diabetic response in
`humans? Finally, what potential mechanism-based toxicological effects, if any,
`would be seen upon chronic administration of a DPP4 inhibitor?
`In light of limited literature in the field, and with no reports of a compound
`having advanced to clinical trials, these questions would ultimately need to be
`addressed during the execution of our discovery and subsequent clinical
`development programs. Despite the unknowns, the positive aspects of this
`target were numerous. Potent small-molecule inhibitors with systemic exposure
`upon oral dosing were known. Although limited, DPP4 inhibitors had
`demonstrated pharmacodynamic efficacy in genetic animal models of type 2
`diabetes in preliminary pharmacological studies. Preclinical proof-of-concept
`for the anti-diabetic actions of GLP-1 was already established and suggested a
`low potential for hypoglycemia. In vitro assays and several in vivo models were
`already described in the literature, enabling rapid program initiation. It was
`against
`this backdrop that significant medicinal chemistry and biology
`resources at BMS were deployed on this newly emerging target in the early
`months of 1999.
`
`1.5 Design of BMS’s DPP4 Medicinal Chemistry
`Program
`
`Given the attractiveness of DPP4 as a therapeutic target, it was anticipated that
`this field would soon become highly competitive, and we therefore sought to
`accelerate the program. High-throughput screening (HTS), routinely a key
`component of drug discovery programs, was deemed as too time consuming to
`rapidly afford a chemical starting point. Thus, we decided to initially adopt a
`design optimization approach,
`improving upon the leads reported by the
`Probiodrug, Ferring, and Novartis research groups; HTS would later be used
`to provide leads for a second-generation effort. From a potency perspective, the
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00331
`
`Page 8 of 26
`
`Page 8 of 26
`
`

`
`The Discovery of the Dipeptidyl Peptidase-4 (DPP4) Inhibitor Onglyzat
`
`7
`
`Figure 1.4 Rationale for generation of conformationally restricted DPP4 inhibitors.
`
`nitrilo-pyrrolidines were deemed to be highly attractive, but were reported to
`exhibit modest pharmacokinetic duration of action and suffered from chemical
`instability.33,34 In solution, the proximal amino group attacks the nitrile
`functionality (see Figure 1.3), eventually leading to the intermediate cyclic
`imidate 13 and ultimately the diketopiperzine 14, both of which are inactive
`versus DPP4. Addressing this issue was viewed as a critical component of the
`medicinal chemistry effort due to considerations regarding both the half-life of
`the compound in vivo, and for high purity processing of the active pharma-
`ceutical ingredient (API) on large scale in a drug manufacturing setting.
`From earlier work by Lin et al.,39 it was demonstrated that replacement of
`the prolyl amide bond with a fluoroalkene isostere resulted in the generation of
`potent DPP4 inhibitor 15 (Figure 1.4). This finding was significant in that it
`suggested that the critical prolyl and amino pharmacophores in 16 may be
`conformationally locked in an extended arrangement which is favorable for
`enzyme inhibition. In addition, because incorporation of the alkene prohibits
`intramolecular attack of the amine onto the acyl hydroxamate, the finding
`suggested novel paths for inhibitor design to retard intramolecular cyclization.
`Taking a cue from earlier work performed at BMS in the design of dual
`ACE/NEP inhibitors,40 we applied the concept of conformationally restricted
`dipetide mimetics in our search for novel inhibitor chemotypes. Many of these
`cores seemed to possess the critical elements required for DPP4 inhibition,
`including a prolyl amide group and a charged amino functionality at the P2
`position. It was hoped that locking the inhibitor conformation by this approach
`would not only enhance binding affinity, but also prevent inactivating cycli-
`zation in compounds possessing an electrophilic pharmacophore (e.g. nitrile,
`phosphate, etc.) on the proline ring. Unguided by the availability of a DPP4
`X-ray crystal structure at that time, our design efforts led to a variety of
`different bi- and monocyclic dipetide mimetics, generically represented in
`Figure 1.4. Unfortunately, all of the compounds generated in this series were
`inactive against DPP4, which we attributed to either incorrect conformational
`geometry required for inhibition, or to steric intolerance for substitution on the
`proline ring. The latter hypothesis was supported by the poor activity exhibited
`by the simple methyl-substituted prolyl derivative 17 as compared to its
`unsubstituted counterpart 8. Interestingly, a recent report from Phenomix
`disclosing 5,5-fused bicyclic lactams such as 18 as potent DPP4 inhibitors
`provides validation for this initial approach.41
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00332
`
`Page 9 of 26
`
`Page 9 of 26
`
`

`
`Chapter 1
`8
`1.6 Design of Cyclopropyl-fused Nitrilo-pyrrolidines
`
`In concert with the effort described above, the discovery team examined
`alternative approaches to the generation of novel inhibitors that might mini-
`mize or obviate the undesired cyclization reaction. Cognizant of the impact of
`pyrrolidine substitution on DPP4 activity (e.g. 17), we proposed that cyclo-
`propyl fusion to the prolyl ring, represented by 19,42 might represent a new and
`viable approach (Figure 1.5). Incorporation of a fused cyclopropane ring
`constituted a minimal steric burden at P1, would impact the planarity of the
`prolyl ring, and importantly have the potential to conformationally retard
`intramolecular attack of the amine to the nitrile. Hence, a series of cyclopro-
`pane-fused nitrilo-pyrrolidines 20–23 were synthesized and assessed for both
`DPP4 inhibition and solution stability (pH 7.2 phosphate buffer, 39.5 1C).43 As
`compared to the unsubstituted prototype 10, the trans-4,5- and trans-2,3-iso-
`mers 22 and 23, respectively, were not well tolerated by the enzyme, negating
`further evaluation of these as viable program leads. In contrast, both the cis-
`4,5- (20) and the cis-3,4- (21) analogues exhibited acceptable, though slightly
`diminished, activity as compared to 10. More interestingly, 20 exhibited a
`significant enhancement in solution stability as compared to either 10 or 21,
`confirming that cis-4,5-cyclopropyl fusion did indeed retard the intramolecular
`cyclization process. Expanding upon this finding, compounds which incorpo-
`rated a corresponding tert-leucine at P2 as in 24–26 minimized potency dif-
`ferences between the respective cores, though solution stability was still
`
`Figure 1.5 Genesis of cyclopropyl-fused nitrilo-pyrrolidine chemotype.
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00333
`
`Page 10 of 26
`
`Page 10 of 26
`
`

`
`The Discovery of the Dipeptidyl Peptidase-4 (DPP4) Inhibitor Onglyzat
`
`9
`
`O
`
`O
`
`N
`
`NH
`
`30
`
`R
`
`NH
`
`N
`
`NH
`
`29
`
`O
`
`R
`
`O
`
`R
`
`N
`
`CN
`
`NH2
`
`28 (anti)
`
`R
`
`N
`
`H2N
`
`O
`
`CN
`
`27 (syn)
`
`Syn oreintation favored with:
`
`- increased steric bulk at R
`- cis-4,5-fused cyclopropyl substitution
`
`Figure 1.6 Factors determining syn versus anti orientation of the methano-nitrilo-
`pyrrolidine chemotype.
`
`favorably preserved in the cis-4,5-methano core (e.g. 24). These and other data
`suggested that increased steric and lipophilic bulk at the P2 position of the
`inhibitor (e.g. ethyl vs. isobutyl vs. tert-butyl, etc.) enhanced both in vitro
`potency as well as solution stability. A similar trend was also observed by the
`Ferring Research group in the simple nitrilo-pyrrolidine series.34
`Clearly both the steric bulk of the P2 side-chain and the cyclopropyl fusion
`on the pyrrolidine ring cooperatively enhanced comformational stability.
`Computational analyses later demonstrated that, primarily because of van der
`Waals interactions,
`increasing the size of the side-chain R (Figure 1.6)
`disfavored the anti orientation 28 required for irreversible intramolecular
`cyclization to the inactive products 29 and 30. For example, where R is Me
`versus tert-butyl, the DDH between the anti (28) and syn (27) orientations
`1. Furthermore, addition of the cis-4,5-
`were calculated to be 0.8 kcal mol
`cyclopropyl moiety also disfavors adoption of the anti orientation by an
`1 as compared to the unsubstituted nitrilo-pyrrolidine
`additional 0.6 kcal mol
`ring. Thus, by combining these conformational elements, we were able to
`identify novel inhibitors of DPP4 with high in vitro potency and enhanced
`solution stability.
`Compound 24 represented a major breakthrough in the chemistry program.
`Profiling of this lead demonstrated a low potential for off-target liabilities
`[hERG inhibition, cytochrome P450 (CYP450) inhibition, broad receptor
`screening, etc.], as well as favorable pharmacokinetic properties in the rat
`(F¼ 77%, t1/2¼ 2.8 h). Additionally, compound 24 was efficacious in Zuckerfa/fa
`rats, reducing glucose AUC in an acute oGTT (ED50¼ 3.3 mg kg
`1, glucose
`challenge 30 min post-dose) when compared to vehicle control. While the
`duration of action of 24 was not particularly long (ED50¼ 92 mg kg
`1, glucose
`challenge 5 h post-dose), the in vivo potency was still significantly greater
`than that of the literature thiazolidine lead 9 (ED50¼ 38 mg kg
`1, glucose chal-
`lenge 30 min post-dose). On the basis of these findings, additional staff were
`assigned to the program to permit a detailed and robust exploration of the
`structure–activity relationship (SAR), primarily focusing on further modifica-
`tions at P2.
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00334
`
`Page 11 of 26
`
`Page 11 of 26
`
`

`
`Chapter 1
`10
`1.7 SAR Optimization Leading to the Discovery of
`Saxagliptin
`
`Building upon the cis-4,5-methano-2-nitrilopyrrolidine core, a wide variety of
`substituents, represented generically in 31, were explored (Figure 1.7). As
`highlighted earlier, greater potency and stability were realized via introduction
`of highly b-branched P2 side-chains. Due to the very limited commercial
`availability of b-quaternary a-amino acids, the team devised an efficient
`strategy to generate novel P2 units by several complementary paths, each
`deriving from cyclic and/or symmetrical ketones to avoid unnecessary intro-
`duction of additional stereocenters.6,43 In one arm of this approach, ketones
`underwent Knoevenagel condensation with a malonate diester, and the
`resulting Michael acceptor was subjected to conjugate addition to introduce
`the alkyl-substituted quaternary center. Mono-hydrolysis and subsequent
`Curtius rearrangement yielded the desired b-branched amino acids. A parallel
`approach began with Horner–Emmons condensation of the ketone to give an
`a,b-unsaturated ester. Reduction of the ester to the primary allylic alcohol
`and esterification with Boc-Gly set up a zinc-mediated ester enolate Claisen
`rearrangement, which provided the desired amino acids with a vinyl functional
`handle at the b-position. Alternatively, a standard Strecker synthesis could
`be used on available aldehydes to ultimately afford the desired amino acid
`building blocks. In all cases, the racemic amino acids were then coupled to the
`
`R1
`
`R3
`
`R1
`
`R3
`
`R1
`
`R3
`
`OH
`
`CO2R
`
`O
`
`O
`
`NH
`
`PG
`
`R1
`
`R3
`
`CO2H
`
`NH
`
`PG
`
`R1
`
`R3
`
`O
`
`R2
`
`R3
`
`R1
`
`R2
`
`R3
`
`R1
`
`R1
`
`R3
`
`CO2H
`
`HO2C
`
`CO2H
`
`RO2C
`
`CO2R
`
`NH
`
`PG
`
`R2
`
`R3
`
`R1
`
`CHO
`
`R2 R3
`
`R1
`
`N
`
`CN
`
`O
`
`31
`
`NH
`
`R
`
`HN
`
`CONH2
`
`Figure 1.7 Various synthetic routes for the generation of highly b-branched amino
`acids.
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00335
`
`Page 12 of 26
`
`Page 12 of 26
`
`

`
`The Discovery of the Dipeptidyl Peptidase-4 (DPP4) Inhibitor Onglyzat
`
`11
`
`N
`
`H2N
`
`O
`
`CN
`
`32
`DPP4 Ki = 3.9 nM
`
`in vivo
`conversion
`
`N
`
`H2N
`
`O
`
`CN
`
`34
`DPP4 Ki = 0.9 nM
`
`HO
`
`H2N
`
`N
`
`O
`
`CN
`
`33
`DPP4 Ki = 7.4 nM
`
`HO
`
`H2N
`
`N
`
`O
`
`CN
`
`3
`DPP4 Ki = 1.3 nM
`
`R2
`
`R3
`
`R1
`
`N
`
`O
`
`CN
`
`NH
`
`R
`
`31
`
`- High in vitro potency
`- Excellent in vivo efficacy & duration
`of action
`- Poor oral exposure and short half life
`
`- High in vitro potency
`- Excellentin vivo efficacy & duration
`of action
`- Good oral exposue and half life
`
`Figure 1.8 Classical PK/PD disconnect in rats leads to hypothesis of active meta-
`bolite generation in vivo, resulting in the discovery of saxagliptin.
`
`enantiomerically pure P1 nitrilo-methanopyrrolidine fragment and the result-
`ing diastereomers were chromatographically resolved.
`From this effort emerged two compounds of particular interest, inhibitors 32
`and 34 (Figure 1.8). In the case of 32, this compound demonstrated superior in
`vivo efficacy and a sustained duration of response for inhibition of plasma
`1 (71% @ 30 min post-
`DPP4 in normal fasted SD rats at a dose of 4 mmol kg
`dose and 64% @ 4 h post-dose). Duration of the pharmacodynamic (PD)
`response was particularly long when compared to compounds generated earlier
`in the program. Intriguingly, this extended PD response was most often
`observed in analogues containing a vinyl substituent (e.g. 32) at P2. Despite its
`robust in vivo activity, compound 32 demonstrated low oral bioavailability
`(5%) and a short half-life (t1/2¼ 1.2 h) in rats. In contrast, the related saturated
`analogue of 32 (vinyl replaced with ethyl) exhibited significantly higher bio-
`availability (31%) and greater in vitro stability in rat and human liver micro-
`somes, but a weaker response in the rat plasma DPP4 inhibition assay (despite
`equivalent in vitro potency). The disconnect between the pharmacokinetic and
`pharmacodynamic profile of 32 immediately suggested in vivo conversion to an
`active metabolite possessing superior target potency and/or reduced clearance
`profile. Based on this hypothesis, a variety of putative or surrogate hydr-
`oxylated metabolites related to 32 were prepared. While unequivocal char-
`acterization of 33 as an active metabolite of 32 was never established, the
`activity of 33 mirrored that of 32 in the aforementioned PD model. While the
`rat half-life of 33 (t1/2¼ 1.3 h) was not significantly longer when compared to
`that of 32, the absolute bioavailability increased 10-fold to 53%. A more
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00336
`
`Page 13 of 26
`
`Page 13 of 26
`
`

`
`12
`
`Chapter 1
`
`striking finding was observed with compound 34. In concordance with estab-
`lished SAR, incorporation of the bulkier adamantyl hydrophobic group in the
`P2 side-chain enhanced in vitro potency (Ki¼ 0.9 nM) as compared to com-
`pounds with smaller fragments in this position. In the normal rat, this com-
`1, po) afforded robust plasma DPP4 inhibition (84% @ 0.5 h
`pound (4 mmol kg
`post-dose and 83% @ 4 h post-dose) despite exceptionally low bioavilability
`(2%) and modest half-life (t1/2¼ 1.4 h). Capitalizing on the learnings from
`compound 33, and considering the propensity of adamantyl groups to undergo
`CYP-mediated hydroxylation, the corresponding metabolite 3 (saxaglipitin)
`was synthesized. The compound proved to be highly potent in vitro, exhibited
`good pharmacokinetic properties in the rat (F¼ 75%, t1/2¼ 2.1 h) and effected
`near maximal (B87%) plasma DPP4 inhibition in the rat at both 0.5 and 4 h
`1.
`post-dose when administered at 4 mmol kg
`In addition to saxagliptin, a number of other inhibitors incorporating
`P2 side-chains with various substituted adamantanes were prepared and
`evaluated, including other positional isomers of hydroxyadamantane, dihy-
`droxyadamantane, and fluoroadamantane. While these other adamantane-
`derived compounds also exhibited potent DPP4 inhibition in vitro, saxagliptin
`provided a comparatively superior PK and PD profile, and was ultimately
`chosen for development.
`
`1.8 Binding of Saxagliptin to Human DPP4: A Slow
`Tight-binding Inhibitor
`
`Through more extensive characterization of the binding kinetics of saxagliptin
`(and select analogs) to DPP4, it was noted that several compounds showed
`evidence of slow tight-binding. Further analysis of SAR patterns relating to this
`slow on/off-rate feature revealed correlations with both steric bulk in the P2
`side-chain (specifically b-quaternary substitution) and the presence of a nitrile
`functionality. Prior to our undertaking of detailed kinetic studies, it had been
`speculated by several groups that a transient covalent bond was formed
`between the hydroxyl group of the active-site Ser630 of the DPP4 catalytic triad
`and the nitrile carbon of nitrile-based inhibitors, yet early X-ray co-crystal
`structures with such compounds lacked adequate resolution to confirm
`appropriate electron density where such a bond would exist.44 Our findings
`were consistent with a hypothesis whereby the bulky P2 side-chain’s displace-
`ment of water in the S2 pocket drove entropic aspects governing on-rate (for 3,
`Kon¼ 4.6105 M
`1 s
`1), and, once anchoring the inhibitor in the active site,
`covalent interaction with the nitrile drove enthalpic aspects governing off-rate
`(for 3, Koff¼ 2310
`5 s
`1), resulting in overall Ki enhancement for certain
`inhibitors possessing these features.45 As a result, the t1/2 for dissociation of
`saxagliptin from DPP4 was determined to be B50 min at 37 1C (B250 min at
`room temperature). In comparison to the rapid off rates of non-nitrilo-pyrro-
`lidines such as sitagliptin, the slow off-rates exhibited by saxagliptin and, to a
`lesser extent, other nitrilo-pyrrolidines such as 12,45b,46 proved to be a unique
`
`Downloaded on 05/02/2015 20:03:18.
`
`Published on 30 September 2010 on http://pubs.rsc.org | doi:10.1039/9781849731980-00001
`
`View Online
`
`SAXA-DEF-00337
`
`Page 14 of 26
`
`Page 14 of 26
`
`

`
`The Discovery of the Dipeptidyl Peptidase-4 (DPP4) Inhibitor Onglyzat
`
`13
`
`attribute of saxagliptin among the clinically advanced inhibitors. The slow off-
`rate kinetics exhibited by saxagliptin likely