Molecular Recognition of Ligands in Dipeptidyl Peptidase IV
`
`Current Topics in Medicinal Chemistry, 2007, 7, 609-619
`
`609
`
`Bernd Kuhn*, Michael Hennig, and Patrizio Mattei
`
`F. Hoffmann-La Roche Ltd, Discovery Research Basel, CH-4070 Basel, Switzerland
`
`Abstract: The serine protease dipeptidyl peptidase IV (DPP-IV) is a clinically validated target for the treatment of
`type II diabetes and has received considerable interest from the pharmaceutical industry over the last years.
`Concomitant with a large variety of published small molecule DPP-IV inhibitors almost twenty co-crystal
`structures have been released to the public as of May 2006. In this review, we discuss the structural characteristics
`of the DPP-IV binding site and use the available X-ray information together with published structure-activity
`relationship data to identify the molecular interactions that are most important for tight enzyme-inhibitor binding.
`Optimized interactions with the two key recognition motifs, i.e. the lipophilic S1 pocket and the negatively charged
`Glu 205/206 pair, result in large gains in binding free energy, which can be further improved by additional
`favorable contacts to side chains that flank the active site. First examples show that the lessons learned from the X-
`ray structures can be successfully incorporated into the design of novel DPP-IV inhibitors.
`
`INTRODUCTION
`Prolyl peptidases are a relatively small family of enzymes
`that are able to cleave peptide bonds after a proline residue
`[1]. One of the best characterized members of this class is the
`serine protease dipeptidyl peptidase IV (DPP-IV, EC
`3.4.14.5), which specifically removes N-terminal dipeptides
`from substrates containing proline, and to some extent
`alanine, at the penultimate position [2]. An important in vivo
`active substrate of DPP-IV is the incretin glucagon-like
`peptide 1 (GLP-1) which has a stimulating effect on insulin
`secretion in a meal-dependent manner [3]. As DPP-IV is
`responsible for rapid degradation of GLP-1 levels in plasma
`the concept of DPP-IV inhibition to increase the half-life of
`this hormone and prolong its beneficial effects has been
`pursued as a new potential therapeutic approach to the
`treatment of type 2 diabetes [4-7].
`As discussed in this issue and in several excellent
`reviews [6,7], there has been an explosion of patents and
`publications
`in
`recent years
`in particular from
`the
`pharmaceutical industry. The high interest in DPP-IV is on
`the one hand because it is a clinically validated target. On
`the other hand, the DPP-IV inhibitor binding site seems to
`be highly drugable in the sense that tight, specific binding to
`the enzyme can be achieved with small molecules without
`compromising favorable physico-chemical properties that are
`important for good pharmacokinetics. While no DPP-IV
`inhibitor is on the market yet, several molecules have
`progressed to phase II clinical trials and beyond: vildagliptin
`1 (Novartis, preregistered) [8], sitagliptin 2 (Merck, preregis-
`tered) [9], saxagliptin 3 (Bristol-Myers Squibb, phase III)
`[10], SYR322 4 (Takeda, phase III) [11], denagliptin 5
`(Glaxo SmithKline, phase II) [12], 815541 (GlaxoSmith
`Kline/Tanabe, phase II), PSN9301 (Osi-Prosidion, phase II),
`NVP-DPP728 6 (Novartis, phase II discontinued) [13], and
`P3298 7 (Merck/Probiodrug, phase II discontinued) [14].
`
`*Address correspondence to this author at the F. Hoffmann-La Roche Ltd,
`Discovery Research Basel, CH-4070 Basel, Switzerland; Tel: +41 61
`6889773; Fax: +41 61 6886459; E-mail: bernd.kuhn@roche.com
`
`The overview of the advanced inhibitors 1-7 in Fig. (1)
`shows that structurally diverse molecules are able to interact
`strongly with DPP-IV. For the design of novel DPP-IV
`inhibitors, it is of interest to identify those protein-ligand
`interactions that are crucial for achieving tight binding. More
`generally, a structural understanding why DPP-IV binds so
`many drug-like molecules might be useful for a better
`assessment of protein drugability in the future [15]. These
`questions can be addressed to some extent by X-ray
`crystallography. The first crystal structure of DPP-IV was
`published in 2003 by Rasmussen et al. [16] revealing its
`complex with the inhibitor valine-pyrrolidide 8. As can be
`seen from Fig. (2), the number of structures deposited to the
`Protein Data Bank (PDB) [17] has increased steadily since
`then, most of them have been solved as protein-inhibitor
`complexes [9,18-24]. The abundance of publicly available
`DPP-IV X-ray information provides a good basis to analyze
`the molecular recognition processes in this enzyme.
`This review focuses on three aspects of DPP-IV inhibitor
`interactions. Firstly, important structural information of the
`DPP-IV binding site gleaned from the existing X-ray studies
`is summarized. Secondly, we highlight the central molecular
`recognition interactions as they have emerged from analyses
`of complex crystal structures and chemical probing. Lastly,
`first attempts to translate the lessons learned from the X-ray
`structures into the design of novel chemotypes are reviewed.
`
`GENERAL STRUCTURAL ASPECTS
`Human DPP-IV is a 766 amino acid transmembrane
`glycoprotein consisting of a cytoplasmic tail (residues 1-6), a
`transmembrane region (residues 7-28), and an extracellular
`part (29-766) [16]. The extracellular region can be further
`subdivided into two domains: a) the catalytic domain
`(residues 508-766) which shows an a/b
` hydrolase fold and
`contains the catalytic triad Ser630 – Asp708 – His740 and b)
`an eight-bladed b propeller domain (residues 56-497) which
`also contributes to the inhibitor binding site [19]. DPP-IV is
`enzymatically active as a homodimer and this is also the
`assembly predominantly
`found
`in
`the
`asymmetric
`
` 1568-0266/07 $50.00+.00
`
`© 2007 Bentham Science Publishers Ltd.
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 1
`
`

`
`610 Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6
`
`Kuhn et al.
`
`O
`
`N
`
`H2N
`
`N
`
`N
`
`F3C
`
`N
`
`F
`
`F
`
`F
`
`N
`
`O
`
`CN
`
`NH
`
`1, Vi ldagli ptin
`IC 50 = 3.5 nM
`Novartis
`
`HO
`
`O
`
`N
`
`O
`
`F
`
`N
`
`N
`
`N
`
`4, S YR -322
`IC50 = 4 nM
`Takeda
`
`N
`
`S
`
`H2N
`
`N
`
`O
`
`7, P3298
`Ki = 123 nM
`M erck/P robiodrug
`
`H2N
`
`N
`
`O
`
`O
`
`NH
`
`HOOC
`
`10, Diprotin A
`apparent IC 50 = 1.1 m M
`
`HO
`
`H2N
`
`N
`
`O
`
`C N
`
`3, Saxagliptin
`IC50 = 0.5 nM
`B ri stol-Myers S quibb
`
`N
`
`O
`
`C N
`
`NH
`
`HN
`
`N
`
`NC
`
`6, NVP-DPP 728
`IC50 = 22 nM
`Novarti s
`
`O
`
`N
`
`N
`
`O
`
`N
`
`N
`
`N
`
`NH
`
`9, B DPX
`Ki = 5.4 m M
`Novo Nordisk
`
`2, S itagliti pin
`IC50 = 18 nM
`Merck
`
`F
`
`F
`
`N
`
`H2 N
`
`O
`
`CN
`
`5, Denaglipt in
`IC50 = 22 nM
`GlaxoS mithKli ne
`
`H2 N
`
`N
`
`O
`
`8, Val-pyrrol idide
`IC5 0 = 2 m M
`
`N
`
`S
`
`N
`
`O
`
`O
`
`NH
`
`12
`IC50 = 39 nM
`Guilford P harma
`
`H2N
`
`11
`IC50 = 40 m M
`S anthera
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 2
`
`

`
`Molecular Recognition of Ligands in Dipeptidyl Peptidase IV
`
`Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6 611
`
`Fig. (1) Contd….
`
`I
`
`N
`
`H2N
`
`O
`
`CN
`
`13
`Ki = 25 nM
`
`F
`
`F
`
`H2N
`
`N
`
`O
`
`16
`Ki = 61 nM
`P fi zer
`
`N
`
`N
`
`S
`
`N
`
`O
`
`O
`
`O
`
`N
`
`21
`IC5 0 = 0.35 nM
`Eis ai
`
`N
`
`S
`
`F
`
`F
`
`N
`
`H2N
`
`O
`
`C N
`
`14
`IC5 0 = 0.6 nM
`Taisho
`
`H2 N
`
`N
`
`O
`
`15
`IC 50 = 43 nM
`M erck
`
`OMe
`
`C l
`
`MeO
`
`N
`
`N
`
`C l
`
`NH
`
`NH2
`
`17
`IC50 = 0.1 nM
`Roche
`
`R
`
`N
`
`H2 N
`
`O
`
`CN
`
`22a, R=H
`Ki = 13 nM
`22b, R=F
`Ki = 3 nM
`J ohnson & Johnson
`
`N
`
`H
`
`NH2
`
`M eO
`
`OMe
`
`19
`IC5 0 = 500 nM
`R oche
`
`O
`
`N
`
`N
`
`N
`
`N
`
`H2N
`
`23
`IC 50 = 10 nM
`Takeda
`
`Fig. (1). Selected inhibitors of DPP-IV. Ligands for which crystal structure information is available have underlined numbers. For
`these ligands, the location of the S1 pocket is indicated by a curved line, the atoms interacting with the Glu 205/206 dyad are in a
`rectangular box, and the atoms forming a covalent bond to Ser 630 are encircled.
`
`unit of the known X-ray structures. The catalytic site lies in
`a large cavity between the two extracellular domains and can
`be accessed through two active site openings. The crystal
`structure of DPP-IV with a decapeptide suggests
`that
`substrates access the catalytic site through a cleft between the
`two domains [20]. The second opening which is located in
`the b propeller domain might be used for the dipeptidic
`product to leave the enzyme [19].
`The shape and hydrophobic/hydrogen bonding properties
`of the inhibitor binding site of DPP-IV are illustrated in Fig.
`
`(3a,b left) for the bound cyanopyrrolidine 6 and xan-thine
`derivative 9 [22]. We use color coding to distinguish
`hydrogen bond acceptor/donor functionalities and aromatic/
`non-aromatic hydrophobic surface patches. The binding site
`is open in the front for solvent access and a considerable
`number of ligand atoms are in contact with surrounding
`water. Hydrophobic and hydrophilic patches in the environ-
`ment of the ligands are roughly equal in size with hydrogen-
`bond acceptor groups dominating the hydrophilic part. As
`can be seen from the protein-ligand interactions in Fig. (3a,b
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 3
`
`

`
`612 Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6
`
`Kuhn et al.
`
`Fig. (2). Annual number of structures deposited to the Protein Data Bank. The grey bar indicates the structures deposited in the first
`quarter of 2006.
`
`Fig. (3). Illustration of solvent-accessible surface of the DPP-IV binding site (left) as well as of important protein-ligand interactions
`(right) for two inhibitors. (a) X-ray complex structure of cyanopyrrolidine NVP-DPP728, 6, with human DPP-IV, (b) crystal structure
`of xanthine derivative BDPX, 9, with porcine kidney DPP-IV (PDB-id: 2aj8). Surfaces are colored by hydrophobic and hydrogen
`bonding (HB) properties: HB acceptor (red), HB donor (blue), HB acceptor/donor (magenta), hydrophobic (grey), aromatic
`hydrophobic (green). Dashed red lines indicate protein-ligand hydrogen bonds and the dashed blue line shows the covalent linkage
`between NVP-DPP728 and Ser 630. Residues Tyr 631, Val 656, Trp 659, Tyr 666, and Val 711 lining the S1 pocket in the back are
`removed for the sake of clarity.
`
`right), this is due to backbone carbonyl groups pointing into
`the binding site at the bottom of the pocket and the
`negatively charged side chains of the Glu 205/206 dyad
`which strongly bind to basic groups such as the secondary
`amines of the two ligands. Crystal structures with the low-
`turnover substrate diprotin A (Ile-Pro-Ile, 10) confirmed that
`
`the two carboxylates of the Glu dyad make short hydrogen
`bonds to the N-terminus of the peptide (d = 2.6Å) providing
`a strong ligand recognition motif [19,25]. For
`ligands
`containing primary amines, the third hydrogen bond is
`typically made to the hydroxyl group of Tyr 662.
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 4
`
`

`
`Molecular Recognition of Ligands in Dipeptidyl Peptidase IV
`
`Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6 613
`
`The two binding modes displayed in Fig. (3) provide a
`good illustration of the interactions made by inhibitors to
`achieve tight binding to DPP-IV. Apart from the ionic
`interactions with the Glu dyad both ligands fill the proline-
`specific S1 pocket in the back with hydrophobic fragments.
`The low micromolar affinity of the small Val-pyrrolidide 8
`and of the b -phenethylamine fragment 11 from Santhera
`Pharmaceuticals [24], shows that substantial binding affinity
`can be gained by optimized interactions with the S1 and Glu
`dyad anchor sites. The
`importance of
`these
`two
`pharmacophores is also underpinned by the fact that almost
`all inhibitor classes share the presence of a lipophilic moiety
`in close proximity to a primary or secondary amine.
`Additional ligand-protein interactions that are seen in
`Fig. (3) are a covalent bond to the catalytic Ser 630 by a
`cyano electrophile mimicking the transition state of peptide
`cleavage, as well as hydrogen bonds to the backbone NH of
`Tyr 631, which forms the oxyanion hole, and to
`the
`backbone carbonyl of Glu 205. The aromatic rings of Phe
`357 and Tyr 547 are exposed to the binding site and provide
`opportunities for additional lipophilic interactions. Each of
`the two ligands in Fig. (3) uses one of these two aromatic
`residues for p -p
` stacking. Finally, the polar Arg 125 and
`Asn 710 side chains are located close to the carbonyl amide
`linking the P1 fragment with the N-terminus of substrates
`(termed P2 amide recognition region throughout this text).
`As observed for the classical serine proteases, few DPP-IV
`inhibitors explore the C-terminal direction of the binding
`
`the
`site (S1’-S2’-…). One noteworthy exception are
`ketopyrrolidines 12, or related ketoazetidines, in which the
`benzothiazole ring is an optimized substituent of the S1’ site
`[26]. Replacing the benzothiazole moiety with a methyl or
`phenyl group renders the molecules inactive, indicating that
`a significant amount of additional binding free energy can be
`gained on the C-terminal end of the scissile bond. In the
`next paragraph, we will use the structural information
`together with examples of published structure-activity
`relationship (SAR) data to investigate the importance of the
`different interaction regions in more detail.
`The large number of X-ray structures allows an assess-
`ment of the flexibility of the DPP-IV binding site, which is
`of relevance for computational drug design. A superposition
`of the available complex crystal structures reveals very little
`movement in the active site (Fig. (4)). Notable exceptions
`are Ser 630 and Tyr 547, which both can adopt two
`conformations, and especially the flexible side chain of Arg
`358 at the end of the binding pocket. The crystallographic
`temperature factors of atoms of Arg 358 are typically high
`and the average position of its side chain is strongly
`influenced by the bound ligand. In contrast, the other
`arginine
`residue, Arg 125,
`shows
`little movement,
`presumably due to the stabilizing effect of its salt bridge
`with Glu 205. The two binding modes in Fig. (3) reveal the
`induced fit effect of the xanthine core which triggers a
`rotation of the aromatic ring of Tyr 547 from its usual
`orientation for better p -p
` interaction.
`
`Fig. (4). Illustration of flexibility in the active site of DPP-IV. Shown are the residues for which different conformations are seen in
`the X-ray complex structures. The cyanopyrrolidine inhibitor 13 is displayed to indicate the positions of the flexible residues
`relative to a bound ligand (PDB-id: 1orw).
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 5
`
`

`
`614 Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6
`
`Kuhn et al.
`
`MOLECULAR RECOGNITION
`This section is divided into several subparagraphs that
`separately discuss the different interaction motifs used by
`DPP-IV ligands.
`
`Catalytic Ser 630 and Oxyanion Hole
`The catalytic machinery of DPP-IV involves a serine
`nucleophile within the catalytic triad Ser-Asp-His, whose
`sequential order, however, is inverse to
`that found in
`classical serine proteases (His-Asp-Ser) [27]. Several early
`inhibitors have been developed that use an electrophilic
`group, mainly a nitrile, to interact covalently with Ser 630
`[6,7]. The concept of covalent binding to nitriles is well
`known from cysteine protease inhibitors [28]. X-ray studies
`confirmed that the nitrile carbon atom changes its hybridi-
`zation state and is in covalent bond distance from the oxygen
`atom of the Ser 630 side chain (see Fig. (3a)) [23]. The
`increase in binding affinity with the additional -CN group is
`substantial, leading to an up to 1000-fold tighter binding to
`the enzyme [5]. For example, cyanopyrrolidine 6 has an IC50
`of 22 nM while the des-cyano analogue binds only with
`16m M to DPP-IV [29]. Enzymatic and biophy-sical studies
`revealed that the covalent interaction of comp-ound 6 is
`reversible and that the activity of the enzyme is regenerated
`upon release of
`the
`inhibitor [29,30]. Compared
`to
`equipotent
`non-covalent DPP-IV
`inhibitors,
`cyano-
`pyrrolidines show both slower association and dissociation
`rates (compound 6: k on = 2.2 x 106 M - 1s- 1, koff = 7.3 x 10- 3
`s- 1, K D = 3.4 x 10- 9 M- 1) [30].
`Covalent coordination from the catalytic serine residue to
`nitriles or other electrophiles has several potential pitfalls.
`Phosphonates are known as irreversible inhibitors of DPP-IV
`[4]. Inhibitors with nitrile or boronic acid electrophiles have
`a limited chemical stability due to intramolecular cyclization
`with the free amino group [6,7]. To prevent the confor-
`mational change required for intramolecular reaction bulky
`substituents were introduced in proximity to the amino
`
`and/or cyano group (compounds 1, 3, 5). These second-
`generation cyanopyrrolidines show significantly higher
`chemical stability.
`Apart from the transition state mimetics that covalently
`bind to Ser 630, few inhibitors use the oxyanion hole, which
`is composed of the backbone NH of Tyr 631 and the side
`chain OH of Tyr 547, for binding. The only ligands for
`which this interaction is confirmed by crystal structures are
`the xanthines 9 (Fig. (3b)) and the related pyrimidine-2,4-
`dione 4 [31], in which a carbonyl group accepts a hydrogen
`bond from the amide NH of Tyr 631. As hydrogen bonding
`is very sensitive to a correct geometry few chemotypes are
`apparently able to interact both with the hydrogen bond
`donor arrangement of the oxyanion hole and the S1 and Glu
`dyad anchors that are mandatory for tight binding.
`
`S1 Pocket
`The specificity pocket S1 is composed of the side chains
`of Tyr 631, Val 656, Trp 659, Tyr 662, Tyr 666, and Val
`711. It is highly hydrophobic as illustrated in Fig. (5a).
`Overlays of the existing X-ray structures reveal very little
`changes in size and shape of the pocket demonstrating its
`high specificity for proline residues. The rigidity of this
`pocket was probed by several groups through modification of
`the ring size of P1 fragments. In isoleucine analogues of the
`non-covalent inhibitor 8 an order of magnitude lower affinity
`was observed when the pyrrolidine moiety was replaced by a
`piperidine or azetidine [32]. The low tolerance for larger
`rings was confirmed for the covalent 2-ketopyrrolidine 12
`while the 2-ketoazetidine analogues were equipotent [26].
`The close-up view of the S1 pocket in Fig. (5a and b)
`reveals a small hydrophobic niche in the back and suggests
`to introduce some asymmetry into the P1 fragment to mimic
`the shape of this pocket. This higher asymmetry can be
`achieved by introducing a sulfur atom into a 5-membered
`ring, as illustrated by the thiazolidine 7, which is approxi-
`mately 2-fold more active than the corresponding pyrrolidine
`[32]. Other favorable modifications of the pyrrolidine ring
`
`Fig. (5). Close-up view of the S1 pocket illustrating the hydrophobic niche that can be optimally filled with small substituents. (a)
`front view of S1 pocket with bound Val-pyrrolidide 8 (PDB-id: 1n1m). See Fig. (3) for the color coding of the surface. (b) top view of
`S1 pocket showing the 2,4-dichloro phenyl ring of aminopyrimidine 17 (cyan, PDB-id: 1rwq) and the phenyl ring of
`aminobenzoquinolizine 20a (blue).
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 6
`
`

`
`Molecular Recognition of Ligands in Dipeptidyl Peptidase IV
`
`Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6 615
`
`are cyclopropanation as in 3 or asymmetric substitution with
`fluorine atoms. Highly potent fluoro-substituted pyrrolidines
`have been published by GlaxoSmithKline 5 [12], Taisho 14
`[33], Merck 15 [34], and Pfizer 16.[35] Hulin et al.
`performed a fluorine scan around the pyrrolidine ring in the
`cyclohexylglycine amide series 16 and found that the activity
`highly depends on the position and stereo-configuration of
`the fluorine substitution [35]. A maximal gain in Ki of ~4-
`fold compared to the unsubstituted pyrrolidine could be
`achieved. As fluorine occupies little more space than
`hydrogen this high sensitivity underlines the stringent shape
`constraints of the S1 pocket.
`A considerably larger gain in binding affinity compared
`to the pyrrolidines could be achieved by small substituents
`on aromatic rings in the S1 pocket [21,36,37]. This is
`illustrated by the aminopyrimidines 17, for which the
`location of the 2,4-dichloro-phenyl ring in the S1 pocket is
`depicted in Fig. (5b) [21]. Table (1) shows the SAR for
`small substitutions around this phenyl ring. Slightly bigger
`substitutents than fluorine, such as chlorine or methyl, are
`best -
` in the optimal para position -
` and improve the IC50
`compared to the unsubstituted phenyl by a factor 30-40.
`Substitutions in
`the meta position
`lead
`to
`repulsive
`interactions with the enzyme and are less favorable.
`
`Table 1. Structure-Activity Relationship of 6-Phenyl Subs-
`tituents of Aminopyrimidines 18 [21]. The Numbers
`are IC50 Data for DPP-IV Inhibition [m M]
`
`N
`
`N
`
`6
`
`R
`
`NH
`
`NH2
`
`18 (R = H: 42)
`
`Cl
`
`2.5
`
`31
`
`1.4
`
`OMe
`
`1.5
`
`80
`
`47
`
`F
`
`14
`
`40
`
`18
`
`R =
`
`ortho
`
`meta
`
`para
`
`Me
`
`1.5
`
`20
`
`1.0
`
`The affinity enhancements observed in the aminopyri-
`midines are even surpassed by the substituent effects in the
`aminobenzoquinolizine (amino-BZQ) series (19 and 20a-j).
`The binding mode of the HTS hit 19 in Fig. (6) reveals the
`potential to substitute the flexible n-butyl side chain with
`better fitting P1 fragments [37]. As can be seen in Table (2),
`amino-BZQ’s with different unsubstituted aromatic and
`lactam moieties (20a-d) inhibit DPP-IV in the high nM to
`low m M range. The X-ray structure with compound 20a
`(Fig. (5b)) reveals the position of the phenyl substituent in
`the S1 pocket. A 40- to 50-fold increase is observed through
`methyl substitution at the correct position, as illustrated by
`20e, 20g, and 20h. The most dramatic improvement
`(>1000-fold) is achieved when the only weakly active pyrrole
`20b is converted into 3-methylpyrrole 20f. The importance
`
`of optimally filling small voids in buried cavities has been
`noted previously [38,39]. This optimi-zation concept is
`nicely exemplified on DPP-IV, with a steep SAR for S1
`pocket substituents.
`It is noteworthy that also some polar groups such as
`pyridine 20c and lactam 20d are tolerated in the S1 pocket
`leading to compounds with an overall more balanced polarity
`pattern. The data in Table (2) underscore the importance of
`the correct orientation of the substituents. Inhibitor 20i has a
`methyl group capable of filling the niche in the S1 pocket in
`the same manner as 20g, however, the polar pyridine
`nitrogen is oriented towards the lipophilic environment in
`the back of this pocket. As a consequence, 20i is only
`slightly more active than the unsubstituted 20c. The penalty
`for unmatched polarities is even more drastic in the lactam
`subseries (20j vs. 20h).
`
`P2 Amide Recognition: Arg 125, Asn 710
`The carbonyl of the amide bond connecting the N-
`terminus with the P1 residue in DPP-IV substrates is located
`in a polar, “electrophilic” environment consisting of the side
`chains of Arg 125 and Asn 710. As shown in Fig. (7, green)
`for the peptide mimetic 13, a hydrogen bond with the amido
`NH2 group of Asn 710 is formed. Conversion of the amide
`to a thioamide leads to a reduction of affinity by 2 and
`replacing the amide by a methylene unit makes the molecule
`inactive, confirming the favorable electrostatic interaction of
`the carbonyl dipole with the protein environment [32].
`Attempts in
`the cyanopyrrolidine series to mimic
`the
`geometry and dipole effect of an amide linker by a trans-
`fluoroolefin lead to a reduction of potency [5]. Interestingly,
`the position of the electronegative oxygen atom can be
`reached by ortho-substitution from aromatic P1 fragments,
`as illustrated in Fig. (7) for the compounds 2 and 17. Merck
`reported a 3-4 fold tighter binding to DPP-IV substituting
`the phenyl in various b -phenethylamine series with a fluorine
`at the ortho-position [36,40]. While the terminal amide
`group of Asn 710 is slightly rotated and not involved in a
`hydrogen bond with the ligand, a favorable electrostatic
`
`Fig. (6). Binding mode and important interactions of the
`aminobenzoquinolizine inhibitor 19 with human DPP-IV [37].
`Dashed red lines indicate protein-ligand hydrogen bonds.
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 7
`
`

`
`616 Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6
`
`Kuhn et al.
`
`Table 2. Structure-Activity Relationship of P1 Substituents of
`Aminobenzoquinolizines [37]. The Numbers are IC50
`Data for DPP-IV Inhibition [nM]
`
`MeO
`
`OMe
`
`N
`
`20b
`9300
`
`N
`
`20f
`5.4
`
`20a
`200
`
`20e
`4.6
`
`R
`
`NH2
`
`N
`
`H
`
`N
`
`20c
`880
`
`20g
`19
`
`N
`
`N
`
`20i
`320
`
`N
`
`O
`20d
`510
`
`N
`
`O
`
`20h
`9.3
`
`N
`
`O
`
`20j
`10000
`
`interaction between the positively charged Arg 125 and the
`Cd+ -Fd-
` dipole moment remains. Using another ortho-
`substituent with high dipole moment, Takeda reported a
`favorable interaction of their 2-cyano group with Arg 125 in
`the pyrimidine-2,4-dione series 4 [31]. For the aminopyri-
`midines (18, Table (1)), small ortho-substitutions (-Me, -Cl,
`-OMe) on the 6-phenyl moiety lead to 17-28 fold lower IC50
`values compared to the unsubstituted ring. The consistent
`gain in affinity with these three substituents indicates that
`the change of torsional angle between the pyrimidine and 6-
`phenyl rings due
`to ortho-substitution might be a
`contributing factor for better protein-ligand fit in this series.
`From the overlays in Fig. (7) and the assembled SAR it can
`be concluded that placing hydrophobic and/or electronegative
`ligand atoms at a very precise location in the vicinity of Arg
`125 and Asn 710 is rewarded with substantial affinity gains.
`
`N-Terminal Recognition: Glu 205, Glu 206, Tyr 662
`Hydrogen bonding interactions with the side chains of
`the two glutamate residues 205 and 206 are, besides the
`filling of the S1 pocket, the second most important anchor
`point for inhibitor binding. Secondary and primary amines
`are recognized by DPP-IV in this region, and for the latter
`ones a third hydrogen bond is formed, typically to Tyr 662.
`This interaction substitutes the binding of the N-terminus of
`
`Fig. (7). Interactions between inhibitors 2 (orange, PDB-id:
`1x70), 13 (green, PDB-id: 1orw), and 17 (blue, PDB-id: 1rwq)
`with the DPP-IV residues Arg 125 and Asn 710 at the P2 amide
`recognition site. The dashed red line indicates the hydrogen
`bond between the carbonyl group of inhibitor 13 and Asn 710.
`
`peptide substrates. The positions of the basic nitrogen atoms
`in the different crystal structures overlap within a sphere of
`only 1.2 Å, making this a very tight pharmacophore
`constraint. Destruction of the hydrogen bond network
`through alkylation of the amino group abolishes activity
`[41]. Furthermore, variation of the basicity of the amine in a
`Roche cyanopyrrolidine series revealed a sharp drop in
`binding affinity by 2 oders of magnitude when the pKa was
`lowered from 7.3 to 6.0 [42]. An interesting class of
`compounds without basic nitrogen are the carbamoyltriazoles
`from Eisai, 21 [43].
`
`Additional Interactions: Phe 357, Tyr 547 and Arg 358
`High nanomolar affinity can be achieved by interactions
`with the S1 pocket, the Glu dyad, and the P2 amide
`recognition site (for ex. 7: Ki = 123nM). However, further
`affinity gains require either covalent binding to the Ser 630
`residue or the exploitation of additional protein-ligand
`interactions. Prime candidates which are used in almost all
`low nanomolar DPP-IV inhibitors are the phenyl rings of the
`two residues Phe 357 and Tyr 547. These are 6-10 Å away
`from S1 pocket and Glu dyad and exposed to the ligand
`binding site. Figures (3a and b) illustrate p -p
` stacking
`interactions between each of the two residues with different
`aromatic ligand fragments. The other energetically favorable
`arrangement, namely aromatic edge-to-face interactions, are
`also seen in crystal structures, one example being the
`interaction between the biphenyl of 22a and Phe 357 [23].
`There are several possibilities to optimize the aromatic
`interactions in both arrangements, as reviewed previously
`[44]. The inhibitor denagliptin 5 is
`interesting as
`it
`presumably makes interactions to both aromatic side chains
`simultaneously. Alternative to aromatic-aromatic interac-
`tions, hydrophobic contacts between Phe 357, Tyr 547 and
`
`Boehringer Ex. 2009
`Mylan v. Boehringer Ingelheim
`IPR2016-01563
`Page 8
`
`

`
`Molecular Recognition of Ligands in Dipeptidyl Peptidase IV
`
`Current Topics in Medicinal Chemistry, 2007, Vol. 7, No. 6 617
`
`large aliphatic groups, such as adamantyl in 1 or 3, can be
`used to achieve low nanomolar IC50 values.
`Lastly, the presence of Arg 358 in close proximity to Phe
`357 makes the positively charged side chain an additional
`interaction partner for substituents on ligand aromatic rings.
`While considerable flexibility of Arg 358 is seen in Fig. (4)
`the observed SAR for several series interacting with Phe 357
`indicates that additional binding free energy can be gained by
`optimizing the electrostatics in this region. For example,
`placing electronegative groups such as trifluoromethyl or
`fluorine next to the positive charge of Arg 358 led to a 4-
`fold increase in binding
`in sitagliptin 2 and in
`the
`cyanopyrrolidine 22b each [9,23].
`
`STRUCTURE-BASED DESIGN OF DPP-IV INHIBI-
`TORS
`The vast majority of known DPP-IV inhibitors have been
`identified by a) derivatization of Xaa-Pro dipeptides, which
`are the products of enzyme cleavage, and b) high-throughput
`screens of compound collections followed by optimization of
`the initial hits [6]. Given
`the
`increasing amount of
`information about the interactions between DPP-IV and
`bound inhibitors from X-ray crystallography during the last
`four years it is timely to review first structure-based design
`attempts for lead finding and optimization.
`A successful example of scaffold hopping from the
`xanthine core of 9 to the quinazolinones 23 through the help
`of molecular modeling and high
`throughput structural
`biology has been reported by Takeda [31]. As shown in Fig.
`(3b), the xanthine moiety of BDPX 9 is involved in a strong
`hydrogen bond with the backbone NH of Tyr 631 and p -
`stacking interactions with Tyr 547, further providing good
`exit vectors to reach the S1 pocket and the Glu 205/206
`dyad. As revealed by a crystal structure, the designed
`quinazolinone scaffold retains these same interactions while
`incorporating carbonyl acceptor function and substituent
`vectors in a single ring. From this series, the inhibitor SYR-
`322 4 with an IC50 of 4 nM emerged, which is currently in
`phase III clinical trials.
`Researchers at Johnson & Johnson identified a series of
`novel biaryl-based DPP-IV inhibitors using a structure-based
`approach [23]. Initial analysis of the X-ray structure of Val-
`pyrrolidide 8 revealed an unoccupied hydrophobic pocket at
`the S2 subsite formed by the side chains of Phe 357 and Arg
`358 (see Fig. (3a)). This cavity could be filled by biaryl
`substituents yielding the cyanopyrrolidine inhibitor 22a with
`a Ki of 13 nM. Further 4-fold improvement was achieved by
`a fluorine atom in the 4-position of the terminal phenyl ring,
`22b. From their X-ray studies, they attributed this increase
`in binding to a better fit to the targeted pocket and stronger
`edge-to-face p -p
` interaction between the biphenyl group and
`Phe 357.
`We have used a structure-based approach to optimize the
`aminobenzoquinolizine class starting from the HTS hit 19
`whose IC 50 is 500 nM [37]. Analysis of its X-ray complex
`structure with DPP-IV (Fig. (6)) revealed the potential for
`tighter binding to the S1 pocket by replacement of the
`flexible n-butyl side chain with more constrained cyclic
`fragments. Virtual screening of monocyclic rings using
`
`docking and similarity search techniques yielded new P1
`substituents for the amino-BZQ scaffold with different
`polarities (Table (2)). As discussed
`in
`the previous
`paragraph, small substitutions at specific positions in these
`rings drastically increased the binding affinity towards DPP-
`IV, the most interesting ones being in the low nanomolar
`potency range and showing both excellent drug-like
`properties as well as promising in vivo activity in rats.
`Finally, virtual screening of larger molecule selections to
`identify novel inhibitor chemotypes for DPP-IV has been
`attempted. After an unsuccessful HTS run yielding no
`suitable starting points for lead generation, a structure-based
`focused screen with emphasis on low molecular weight
`compounds was performed by Ward et al. [45]. Using a
`hierarchical approach of in silico filtering, 3D pharmaco-
`phore matching, docking, and visual inspection a library of
`4000 molecules was assembled and tested at a three times
`higher compound concentration (30 m M) than in the HTS.
`Several novel, but quite weak DPP-IV inhibitors, the most
`potent showing an inhibition of 82% at 30 m M, could be
`identified. Santhera Pharmaceuticals reported results from a
`biased fragment screen with the aim to identify novel S1
`pocket binders [46]. As known DPP-IV inhibitors typically
`have li