Article
`
`pubs.acs.org/jmc
`
`Extended Structure−Activity Relationship and Pharmacokinetic
`Investigation of (4-Quinolinoyl)glycyl-2-cyanopyrrolidine Inhibitors
`of Fibroblast Activation Protein (FAP)
`Koen Jansen,† Leen Heirbaut,† Robert Verkerk,∥ Jonathan D. Cheng,‡ Jurgen Joossens,† Paul Cos,§
`Louis Maes,§ Anne-Marie Lambeir,∥ Ingrid De Meester,∥ Koen Augustyns,† and Pieter Van der Veken*,†
`†Medicinal Chemistry, Department of Pharmaceutical Sciences, University of Antwerp, Universiteitsplein 1, B-2610 Antwerp, Belgium
`‡Fox Chase Cancer Center, 333 Cottman Avenue, Philadelphia, Pennsylvania 19111-2497, United States
`§Laboratory of Microbiology, Parasitology, and Hygiene, Departments of Pharmaceutical Sciences and Biomedical Sciences,
`University of Antwerp, Universiteitsplein 1, B-2610 Antwerp, Belgium
`∥Medical Biochemistry, Department of Pharmaceutical Sciences, University of Antwerp, Universiteitsplein 1, B-2610 Antwerp,
`Belgium
`
`*S Supporting Information
`
`ABSTRACT: Fibroblast activation protein (FAP) is a serine
`protease related to dipeptidyl peptidase IV (DPPIV). It has been
`convincingly linked to multiple disease states involving remodeling
`of
`the extracellular matrix. FAP inhibition is investigated as a
`therapeutic option for several of these diseases, with most attention
`so far devoted to oncology applications. We previously discovered the N-4-quinolinoyl-Gly-(2S)-cyanoPro scaffold as a possible
`entry to highly potent and selective FAP inhibitors. In the present study, we explore in detail the structure−activity relationship
`around this core scaffold. We report extensively optimized compounds that display low nanomolar inhibitory potency and high
`selectivity against the related dipeptidyl peptidases (DPPs) DPPIV, DPP9, DPPII, and prolyl oligopeptidase (PREP). The log D
`values, plasma stabilities, and microsomal stabilities of selected compounds were found to be highly satisfactory. Pharmacokinetic
`evaluation in mice of selected inhibitors demonstrated high oral bioavailability, plasma half-life, and the potential to selectively
`and completely inhibit FAP in vivo.
`
`■ INTRODUCTION
`Fibroblast activation protein (FAP, FAP-α, seprase) belongs to
`the prolyl oligopeptidase family S9, which consists of serine
`proteases that cleave peptide substrates preferentially after
`proline residues. Other members of this family include the
`dipeptidyl peptidases (DPPs: DPPIV, DPP8, DPP9) and prolyl
`oligopeptidase (PREP, POP).1 FAP has been linked to multiple
`disease states involving remodeling of the extracellular matrix
`such as hepatic and pulmonary fibrosis, keloid formation,
`rheumatoid arthritis, and osteoarthritis.2−7 FAP is also highly
`expressed on activated fibroblasts in over 90% of common
`tumors.8,9 It has been demonstrated in
`human epithelial
`syngeneic mouse models that FAP activity promotes tumori-
`genesis and that FAP inhibition attenuates tumor growth.10−12
`The enzyme is furthermore expressed only transiently during
`wound healing and is essentially absent in normal adult tissues
`and in nonmalignant tumors.13 These appealing characteristics
`of FAP account for its ongoing evaluation as a drug target. Both
`immunotherapy and small-molecule based approaches have so
`far been reported, most of them focusing on applications in the
`oncology domain (vide infra).
`FAP possesses both dipeptidyl peptidase and endopeptidase
`activity, catalyzed by the same active center. This is in contrast
`with the DPPs, possessing only the former activity type, and
`
`PREP, which is an enzyme of strict endopeptidase capa-
`bility.14,15 While designing out DPP affinity in FAP inhibitors is
`relatively straightforward, obtaining inhibitors possessing
`selectivity for FAP over PREP is considered to be far more
`challenging. This is, among others, illustrated by the significant
`overlap between in vitro processable substrate sequences for
`FAP and PREP and the fact that numerous reported FAP
`inhibitors have limited or no selectivity with respect to PREP.16
`Given that recent findings indicate that PREP deficient mice
`have impaired spatial learning, memory, and neuronal develop-
`ment, FAP over PREP selectivity could nonetheless be an
`important feature for inhibitors.17,18 The availability of such
`compounds can already have significant impact when applied as
`tool compounds to study the role of FAP in pathophysiology.
`In some cancer
`types for example, FAP and PREP are
`simultaneously overexpressed by cell types that are part of the
`metastatic tumor microenvironments, and there are currently
`no reports available that quantitatively discrimate between the
`potential contributions of both proteolytic activities to disease
`progression.17
`
`January 8, 2014
`Received:
`Published: March 11, 2014
`
`© 2014 American Chemical Society
`
`3053
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3053
`
`

`

`Journal of Medicinal Chemistry
`
`Article
`
`the relevant FAP inhibitors reported in the
`Several of
`literature were used as reference compounds in this study. Val-
`boroPro 1 (talabostat, PT-100) is an nonselective boronic acid
`inhibitor that reached phase II clinical trials for several cancer
`types before it was withdrawn, apparently because of both
`safety and efficacy reasons (Figure 1 and Table 1).19−21
`
`Figure 1. Relevant FAP inhibitors used as references in this study.
`
`Linagliptin 2 has received clinical approval as a DPPIV inhibitor
`but also displays substantial FAP affinity.22 Compound 3 is a
`representative of a series of pyroglutamyl(2-cyanopyrrolidine)
`derivatives reported by Jiaang et al.23 Compounds 4 and 5 are
`part of the quinolinoylglycyl(2-cyanopyrrolidine) class of FAP
`inhibitors reported recently by our group.24 Although no
`PREP-assay results were published by Jiaang, we report here
`that compound 3 does have highly satisfactory FAP over PREP
`selectivity. Therefore, Jiaang’s and our molecules represent the
`only two reported inhibitor chemotypes with potential for full
`FAP selectivity across the panel of related enzymes evaluated.
`Jiaang’s compounds nonetheless were reported to have poor
`pharmacokinetic (PK) behavior in mice. The corresponding PK
`data for selected representatives of our own molecules are
`reported in this manuscript. In addition, Bachovchin et al. have
`also recently reported the D-Ala-boroPro based FAP inhibitor
`6.24 This compound’s approximate 40-fold selectivity toward
`PREP (determined under our assay conditions) is remarkable
`for a boronic acid, certainly since a recently published pseudo
`peptide boronic acid inhibitor with the same D-Ala-boroPro
`scaffold but with an acetyl-Arg-2-(2-(2-aminoethoxy)ethoxy)-
`
`Table 1. IC50 of Reference FAP Inhibitors
`
`acetamide group as the N-substituent instead of a 4-pyridinoyl
`to be selective toward PREP.25
`group was
`found not
`Nonetheless, no in vivo PK data have so far been published
`for 6.26
`In our earlier report on FAP inhibitors related to 4, we have
`investigated the quinolinoyl moiety’s substitution pattern and
`the influence of the position of the heterocyclic N-atom on
`FAP affinity and selectivity. No heterocyclic scaffolds other than
`quinoline and isoquinoline were so far built in at this position.
`Evaluating the impact of other azaheteroaromatic rings was
`therefore considered a first goal of this study with which we
`aimed to significantly expand our current SAR knowledge for
`this part of the molecule. Second, SAR for the P2 glycine
`residue also remained underexplored, as no fragments other
`than glycine have been introduced at this position. Several P2-
`modified analogues were therefore made. Furthermore, these
`molecules were deemed of additional interest to anticipate on
`potential sensitivity of the P2 glycine amide bonds to unspecific
`proteases in vivo. Third, we sought to further optimize the P1
`cyanopyrrolidine residue of our inhibitors. This was done by
`applying predictive SAR data that we derived earlier for
`inhibitors containing a P1-(2-cyanopyrrolidine) moiety.24,27 In
`the P1-modified series, we also investigated the importance of
`the warhead on our molecules and the influence on inhibitory
`potency of changing the carbonitrile function for other
`electrophilic warhead types (boronic acid, chloromethyl
`ketone). Fourth,
`in vitro PK and cellular toxicity data are
`reported for five optimized inhibitors. On the basis of these
`data, we selected four compounds for in vivo PK studies and in
`vivo inhibition experiments in mice.
`
`■ CHEMISTRY
`A total of around 60 novel inhibitors were synthesized for this
`study. All compounds were prepared following the general
`strategies in Schemes 1 and 2, in which target compounds are
`clustered according to the modification type they contain,
`relative to reference compound 4. The preparation of inhibitors
`containing an adapted P2 group (9−11 and 16−27) is
`summarized in Scheme 1. Three types of modifications were
`included in this series: (1) attachment of a side chain to the α-
`methylene group of the P2 moiety, (2) introduction of N-
`substituents, and (3) replacement of the P2−P3 amide bond by
`a reduced analogue or a peptidomimetic triazole ring. For the
`preparation of the α-methylene substituted series (Scheme 1,
`entry 1),
`the corresponding, commercially available Boc-
`protected P2-amino acids were coupled to 2-cyanopyrrolidine.
`The Boc group of the coupled products was then removed
`using tosylic acid in acetonitrile under conditions we identified
`as optimal to avoid Ritter-type addition of the intermediate tert-
`
`compd
`
`FAP
`
`1
`2
`3
`4
`5
`6
`
`0.066 ± 0.011
`0.37 ± 0.002
`0.017 ± 0.001
`0.0103 ± 0.0004
`0.011 ± 0.0004
`0.025 ± 0.001
`
`PREP
`
`0.98 ± 0.06
`>100
`>100
`0.86 ± 0.07
`>50
`0.99 ± 0.04
`
`IC50 (μM)a
`DPPIV
`
`0.022 ± 0.001
`0.0020 ± 0.0002
`>100
`>100
`>100
`>100
`
`DPP9
`NDc
`>100
`>100
`>100
`>100
`>50
`
`DPP2
`
`0.086 ± 0.007
`>100
`>100
`>100
`>100
`>100
`
`SI (FAP/PREP)b
`14.8
`>250
`5882.4
`83.5
`>4500
`39.6
`
`aDetermined under our own assay conditions. bSI stands for “selectivity Index” (calculated as [IC50(PREP)/IC50(FAP)]). cND stands for “not
`determined”.
`
`3054
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3054
`
`

`

`Journal of Medicinal Chemistry
`
`Article
`
`Scheme 1. Synthesis of P2-Modified FAP Inhibitors 11−14 and 16−27a
`
`aReagents and conditions: (a) HATU, N-Boc-AA-OH, DIPEA, 75−80%; (b) TsOH 1.4 equiv, MeCN, 24 h, 84−86%; (c) 4-quinolinoyl chloride
`hydrochloride, DIPEA, 25−86%; (d) (2S)-(2-cyanopyrrolidine) tosylate, HATU, DIPEA, 64%; (e) K2CO3, RNH2, 13−52%; (f) K2CO3, quinolin-4-
`ylmethanamine, 25−49%; (g) NaN3, DMF; (h) CuI, THF, 17−50%.
`
`butyl cation to the nitrile group of products 7−9. These
`products were coupled with 4-quinolinoyl chloride hydro-
`chloride to yield final compounds 11−13. The 2-cyanopyrro-
`lidine starting material was obtained as described earlier.23 For
`molecules involving the introduction of N-substituents or P2−
`P3 amide group modifications, (S)-1-(2-chloroacetyl)-
`pyrrolidine-2-carbonitrile 15 was used as a central intermedi-
`ate28 (Scheme 1, entry 2). Reaction of 15 with primary amines
`and subsequent coupling with 4-quinolinoyl chloride hydro-
`chloride yielded compounds 16−20. To obtain target
`compounds 21 and 22 with a reduced P2−P3 amide group,
`the chloroacetyl group of 15 was substituted with N-substituted
`(4-quinolinyl)methylamines. Finally, the peptidomimetic tri-
`azole containing compounds were made by nucleophilic
`substitution of 15’s chloro group with sodium azide and
`subsequent Cu(I)-catalyzed 1,3-dipolar addition with the
`desired alkynes to give products 24−27.
`Scheme 2 describes the preparation of P3- and P1-modified
`analogues. Two strategies were followed for all
`reported
`compounds bearing a 2-cyanopyrrolidine derivative in P1
`(29−63) (Scheme 2, entry 1). For molecules designed to
`mainly generate SAR data for
`the P3 position, diverse
`heteroarylcarboxylates were coupled to a specific glycyl-(2-
`cyanopyrrolidine) derivative using standard peptide coupling
`reagents. Alternatively, for a number of compounds designed
`mainly to render P1 SAR information, N-(4-quinolinoyl)glycine
`28 was coupled to a desired 2-cyanopyrrolidine derivative. The
`its 4S-fluoro, 4R-fluoro, and 4,4-difluoro
`2-cyanopyrrolidine,
`congeners and all glycine adducts of these molecules were
`
`prepared as reported earlier or as described in the Supporting
`Information.29 Heteroarylcarboxylates were acquired commer-
`cially or, in the case of substituted 4-quinolinoyl derivatives,
`synthesized using the Sandmeyer isatin synthesis followed by
`reaction and decarboxylation.30,31 Detailed
`the Pfitzinger
`procedures for all intermediates and compounds in this report
`are given in the Supporting Information. The boronate, the
`chloromethyl ketone, and unsubstituted pyrrolidine derivatives
`were synthesized in an analogous manner (Scheme 2, entry 2).
`Again, N-(4-quinolinoyl)glycine 28 was coupled to the desired
`pyrrolidine derivative using standard peptide coupling
`techniques. Both the required 2-boronyl- and chloromethyla-
`cylpyrrolidine derivatives were obtained as described in the
`literature.32,33 To obtain boronic acid final compound 66, its
`pinanediol ester precursor 65 was deprotected using phenyl-
`boronic acid.
`
`■ RESULTS AND DISCUSSION
`All synthesized compounds were evaluated as inhibitors of FAP,
`DPPIV, DPP9, DPPII, and PREP.27 No separate screening
`experiments were carried out on DPP8. On the basis of the
`enzyme’s close homology with DPP9 and the outcome of
`earlier directed studies, potencies toward both enzymes can be
`expected to be comparable with a high degree of confidence.
`Table 2 summarizes the assay results of
`the P2-modified
`analogues. In a first set of compounds (11−14), the P2-glycine
`residue of 4 was replaced by D-Ser and related D-Ala, L-Ala, and
`1-amino-1-carboxycyclopropane moieties. The choice for the
`
`3055
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3055
`
`

`

`Journal of Medicinal Chemistry
`
`Article
`
`Scheme 2. General Synthetic Strategies for the Synthesis of P1 and/or P3 Modified Target Compoundsa
`
`aReagents and conditions: (a) YCOOH, HATU, DIPEA, 3 h, 43−67%, or YCOOH, HOBT, EDC, DIPEA, 16 h, 84−86%, or YCOOH, (CH3)2C
`C(Cl)N(CH3)2, DIPEA, 25−70%, or quinoline-4-carbonyl chloride hydrochloride, DIPEA, 55−59%; (b) pyrrolidine derivative, HOBT, EDC,
`DIPEA, 16 h, 65−86%; (c) phenylboronic acid, MTBE−H2O, 79%.
`
`first two residues was based on an extensive screening of
`fluorogenic peptide substrates by Edosada et al., demonstrating
`that FAP’s endopeptidase capability is strictly limited to
`peptides with a P2-Gly, D-Ser, or D-Ala residue.34 As an
`additional
`illustration, a P2-D-Ala residue present in Bach-
`ovchin’s boronic acid FAP-inhibitor 6 also demonstrated low
`nanomolar FAP affinity and good selectivity toward PREP.26
`When introduced as a replacement for compound 4’s glycine
`residue however (11 and 12), the FAP affinities observed
`dropped 600- and 300-fold, respectively. FAP inhibition was
`completely abolished by introducing the closely related amino
`acids L-Ala and 1-amino-1-carboxycyclopropane at P2. While
`these data underscore the importance of a P2-glycine in our
`inhibitors, the lack of affinity of 13 and 14 also serves as
`confirmation that Edosada’s original findings on substrates
`translate similarly for
`inhibitors. Finally, L-Ala containing
`analogue 13 has a PREP potency that stands out among the
`four congeneric inhibitors discussed.
`Moving further with our SAR exploration of the P2 position,
`we decided to synthesize analogues with an N-alkylated and/or
`a reduced P2−P3 amide bond. Taking into account that N-
`alkylation of amide bonds in peptides is a well-known way of
`reducing susceptibility to (non-FAP-related) proteolytic
`activity, both modification types could increase the metabolic
`stability of this potentially labile part of the inhibitors’ basic
`structure. In the N-alkylated amide series (compounds 16−20),
`substituents of varying size and electronic properties were
`tested. All these interventions were found to lead to a nearly
`complete loss of FAP potency, even in the case of the smallest
`
`methyl substituent. Remarkably, PREP affinity was affected to a
`significantly smaller extent, bringing on additional evidence that
`the 4-quinolinoyl substituent of 4 is involved in a specific
`interaction with FAP’s active center that significantly adds to
`target affinity but is disrupted by small structural changes.
`Similar conclusions were drawn from the assay results of the
`reduced analogues: tertiary amines 21−23 also had strongly
`reduced FAP inhibition. It is, however, impossible to state to
`what extent
`the increased conformational
`freedom of
`the
`quinolinylmethyl substituent or
`the lack of a conjugated,
`electron withdrawing carbonyl function contributes to these
`findings. Furthermore, a basic amine functionality is present in
`these molecules. Apparently this feature does not allow picking
`up additional potency from salt-bridge formation with FAP’s
`Glu203 and Glu204 residues, the interaction responsible for
`FAP’s recognition of dipeptide substrates. Neither does it seem
`to be involved in a significant interaction with the other DPPs’
`homologous GluGlu motifs, as reflected by the assay results for
`these enzymes. On the basis of the obtained results with the set
`of
`inhibitors 16−23, we did not prepare any additional
`analogues in this series. Fitting in the same framework of
`metabolic P2−P3 amide stabilization, molecules (24−27) were
`then prepared in which this amide group is replaced by an
`isosteric, 4-substituted 1,2,3-triazole ring. The 4-alkynyl
`derivatized quinolines required to obtain triazole analogues of
`4 were, however, found not to be readily accessible. This led us
`to prepare analogues of older FAP inhibitors that we reported:
`N-acylated glycyl(2-cyanopyrrolidines) with a P3-benzoyl or
`naphthoyl substituent. The compounds obtained were again
`
`3056
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3056
`
`

`

`Journal of Medicinal Chemistry
`
`Article
`
`Table 2. IC50 Data of Inhibitors with a Modified P2 Residue
`
`3057
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3057
`
`

`

`Journal of Medicinal Chemistry
`
`Table 2. continued
`
`Article
`
`aSI stands for “selectivity index” (calculated as [IC50(PREP)/IC50(FAP)]).
`
`found to have reduced inhibitory affinity, with phenyl analogue
`24 possessing around 8-fold less potency compared to the
`benzoyl amide. Naphthyl-containing 27 is characterized by an
`even higher affinity loss compared to its 1-naphthoyl
`substituted parent compound that possesses FAP submicro-
`molar potency. Not satisfied by these results, we concluded that
`the 4-substituted-1,2,3-triazole is not a good amide isostere at
`the glycine P2 position.
`Investigations of the P3 region of inhibitors related to 4 was
`aimed at
`replacing its quinoline ring with various other
`azaheterocycles. Our earlier studies, during which we had
`already investigated quinoline and isoquinoline isomers, clearly
`demonstrated the position of the azaheteroaromatic nitrogen to
`be crucial, with appreciable FAP affinity only present in case of
`the 4-quinolinoyl isomer of compound 4. In all compounds
`reported here, we therefore limited the selection of heterocycles
`to systems that have a azaheteroatom topology for at least one
`nitrogen that is comparable to that of the 4-quinolinoyl ring
`nitrogen. Table 3 summarizes the data obtained for a number of
`(substituted) five- and six-membered heteroaromatic contain-
`ing analogues of 4. The closely related pyridine 29 was found to
`have a 6-fold reduced FAP potency compared to 4. In spite of
`its lower absolute affinity, 29 has the highest ligand efficiency
`(0.38) of all compounds we hitherto prepared. Introduction of
`substituents in 2-position of the pyridine ring as in compounds
`30−35 resulted in a further decrease of inhibitory potency, but
`no additional steric or electronic determinants of FAP potency
`could be discerned here: all compounds from this subset seem
`to have inhibitory potential within the same order of
`magnitude. Additionally, an analogue containing an extra
`methylene linker (36) and a reduced P3-heteroaromatic ring
`system (piperidine 37) were prepared. Similarly,
`these
`molecules did not demonstrate better FAP affinities than 29
`and also convey the hypothesis
`that high FAP-affinity-
`conferring P3 substituents are mainly present in limited parts
`of heteroaromatic chemical space.
`
`Identification of these other potentially interesting parts of
`chemical space was then attempted with compounds 38−44.
`These contain heteroaromatic P3 substituents that are more
`distant from the pyridine- or quinoline-based systems that we
`so far discovered to be optimal. Pyridazine (in inhibitor 38) was
`the only azine with more than one nitrogen heteroatom that we
`evaluated. Already, this compound had a 3-fold decreased FAP
`potency relative to 29. This finding withheld us from investing
`effort in preparing target molecules with a triazine isomer at P3.
`Conversely, we shifted attention to more electron-rich, five-
`membered azaheterocycles. Pyrrole was not selected for this
`subset because of stability reasons, but a suitable (vide supra)
`imidazole derivative and two triazole isomers were included
`(compounds 39−41). These molecules,
`together with the
`homologated 42 and the thia- and oxazole derivatives 43 and
`44, all performed significantly less than 29 in the FAP
`inhibition assay. Finally, mutation of the pyridine ring into an
`aniline was found not to be beneficial for FAP affinity either:
`the aniline 45 and its trifluoroacetylated precursor 46 have IC50
`values in the low micromolar range, more than 100-fold less
`than pyridine 29. Taken altogether, the data in Table 3 were
`not compelling enough to immediately continue further
`structural exploration of monocyclic heteroaromatic P3 rings:
`FAP potencies observed in this series were not able to compete
`with reference 4. In the case of the highly efficiently binding
`inhibitor 29, further optimization did not seem obvious. In
`addition, the FAP/PREP selectivity indices observed, so far not
`mentioned in the framework of this series, were generally very
`satisfactory for the pyridine subset but again not significantly
`better than the optimal quinolines described earlier. It deserves
`highlighting however that as reported by Poplawski et al.,
`introduction of a P3−P4-pyridinoyl substituent has been used
`to obtain a series of highly potent, boronate-based FAP
`inhibitors represented by reference compound 6.26
`For the reasons cited above, we then turned attention to
`bicyclic P3 heteroaromatics with at least one nitrogen atom that
`
`3058
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3058
`
`

`

`Journal of Medicinal Chemistry
`
`Article
`
`Table 3. IC50 Data of Five- and Six-Membered P3 Heterocycles
`
`3059
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3059
`
`

`

`Journal of Medicinal Chemistry
`
`Table 3. continued
`
`Article
`
`aSI stands for “selectivity index” (calculated as [IC50(PREP)/IC50(FAP)]).
`
`is equivalent to the azaheteroatom in quinoline 4 (Table 4).
`With the set of inhibitors 47−51, 2,3-annelated pyridines were
`evaluated. Pyrrolopyridine 47 was found to be a nanomolar
`FAP inhibitor but still almost 5-fold less potent
`than 4.
`Selectivity indices with respect
`to PREP were largely
`comparable for both (SI[FAP/Prep] values were calculated to be
`∼50 and 80 for 47 and 4,
`respectively). Substituted
`imidazopyridines 48 and 49 and certainly triazolopyridine 50
`were significantly less potent as FAP inhibitors than 4,
`suggesting that introduction of additional nitrogen atoms in
`the annelated five-membered heterocycle leads to lower target
`affinities. A similar effect seems to be in place when a six-
`membered heterocycle is annelated with pyridine (1,7-
`naphthyridine 51). The IC50 value of
`this azaquinoline
`derivative is roughly 3 times higher than that of quinoline 4;
`nonetheless, selectivity toward PREP is slightly better than that
`of the reference.
`Since a P3-quinoline residue out of all the heterocycles
`investigated still proved to be optimal, we decided to return to
`this scaffold and investigate the effect of hitherto unexplored
`substitution patterns at its 5 and 7 positions. Our preliminary
`data had already indicated that introduction of 5- or 7-chloro
`and bromo substituents resulted in low nanomolar FAP
`inhibitors with optimal selectivity toward PREP (SI[FAP/Prep]
`of up to 103). Similar FAP potencies were obtained here with
`the methyl-substituted analogues 52 and 53, of which the 5-
`substituted congener clearly had the best FAP/PREP selectivity
`seen so far (SI[FAP/Prep] = 2 × 103). Introduction of bulkier
`
`phenyl-based groups (compounds 54−56) seems less desirable
`at both the 5 and 7 positions: FAP potencies drop close to 1
`order of magnitude. Because of the relatively small size of the
`methoxy substituent, one would furthermore expect the 5-
`methoxylated inhibitor 57 to have potencies in the same range
`as 5 or 53. Surprisingly however, potency was reduced more
`than 1000-fold, again indicating that the supposed specific
`interaction of the quinolinoyl system with FAP’s active center
`can be critically disturbed by small
`structural changes.
`Additional investigation of this finding was undertaken during
`the optimization of the P1 position (vide infra, Table 5).
`As part of an earlier study, we had shown that in the glycyl(2-
`cyanopyrrolidine) series, FAP’s S1 pocket preferentially
`accommodates 2-cyanopyrrolidine and a limited number of
`derivatives with 4-substituents of minimal steric bulk, e.g., a 4S-
`fluoride. Similarly, (2S,4S)-2-cyano-4-fluoropyrrolidine and its
`4,4-difluorinated analogue were also used earlier as P1 residues
`by Jiaang in FAP inhibitors related to 6. Depending on the
`target,
`these residues have also been investigated with
`considerable success in compounds targeting DPPIV and
`DPP8/9. The same strategy was applied to the inhibitor series
`studied here.27 Complementarily, we also prepared the
`(2S,4R)-2-cyano-4-fluoropyridine diastereomer and introduced
`it as a P1 residue. Supposedly relating to its more elaborate
`synthesis involving a chirality inversion step on the 4-position
`of starting material trans-4-hydroxyproline, this isomer so far
`had not been investigated in FAP or related DPP inhibitors. To
`allow efficient assessment of the effect of introducing each of
`
`3060
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3060
`
`

`

`Journal of Medicinal Chemistry
`
`Table 4. IC50 Data of Bicyclic P3 Heterocycles
`
`Article
`
`aSI stands for “selectivity index” (calculated as [IC50(PREP)/IC50(FAP)]). bSolubility problems upon dilution of compounds in aqueous buffer,
`resulting in poor reducibility of inhibition curves.
`
`the three mono- and difluorinated P1 isomers, direct analogues
`of reference compound 4 were prepared (58−60). Grossly in
`line with expectations, the 4S-fluoropyrrolidine 58 displayed
`around 3-fold higher FAP affinity than the parent compound.
`PREP-affinity however had also increased slightly more than
`
`proportionally with FAP-potency. This is consistent with our
`included 4S-fluorinated (2-
`earlier mentioned report
`that
`cyano)pyrrolidine inhibitors.27 Most
`the 4R-
`remarkably,
`fluoropyrrolidine isomer 59 has lost FAP affinity compared to
`4 and 58, amounting to up to 3 orders of magnitude. Making
`
`3061
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3061
`
`

`

`Journal of Medicinal Chemistry
`
`Article
`
`Table 5. IC50 Data of Inhibitors with a Modified P1 Residue
`
`aSI stands for “selectivity index” (calculated as [IC50(PREP)/IC50(FAP)]).
`
`Table 6. Influence of the Warhead on IC50
`
`aSI stands for “selectivity index” (calculated as [IC50(PREP)/IC50(FAP)]).
`
`steric factors less likely to be accountable for this effect, the
`corresponding 4,4-difluorinated analogue of 4 (UAMC-1110,
`vide infra) was one of the most potent FAP inhibitors identified
`to date. Seminal studies by Raines et al. have shown that 4S-
`fluoroproline derivatives possess a hyperconjugatively stabilized
`endo-puckered ring conformation, while in 4R-fluoroprolines
`an exo-puckered ring conformer
`is more favored. This
`difference, which should also be present with 2-cyanopyrroli-
`
`dine derivatives, could be contributing to the large potency
`differences observed between 58 and 59.35 Analogously, 4,4-
`difluoroproline derivatives have been demonstrated to have
`endo- and exo-puckered conformations of comparable energy,
`similar to unsubstituted prolines.36 Although this would imply
`that the slightly higher FAP affinity of 60 compared to 4 cannot
`be rationalized using conformational arguments,
`increased
`hydrophobicity resulting from difluorination might be in play
`
`3062
`
`dx.doi.org/10.1021/jm500031w | J. Med. Chem. 2014, 57, 3053−3074
`
`Petitioner GE Healthcare – Ex. 1006, p. 3062
`
`

`

`Journal of Medicinal Chemistry
`
`Table 7. In Vitro Pharmacokinetic Properties of Selected FAP Inhibitorsa
`
`parameter
`
`kinetic solubility (μM)
`log D
`plasma stability (% unchanged at 6 h) (mouse/rat/human)
`microsomal stability (% unchanged at 6 h) (mouse/rat)
`cytotoxicity (MRC-5 cells) (μM)
`and stands for “not determined”.
`
`4
`
`>200
`0.5
`100/90/nd
`nd
`>64
`
`5
`
`>200
`0.7
`90/nd/100
`75/nd
`>64
`
`68
`
`>200
`0.8
`nd/100/80
`90/nd
`>64
`
`60
`
`>200
`1
`85/95/nd
`90/94/nd
`>64
`
`Table 8. In Vivo PK Properties of Selected FAP Inhibitors in Rats
`
`Article
`
`61
`
`>200
`nd
`nd
`nd
`>64
`
`AUC (μg·h/mL)
`
`Cl (mL/min)
`
`relative bioavailability (%)
`
`Vd (L)
`T1/2 (h)
`Cmax (μg/mL)
`5 (iv)a
`1.75
`11.7
`6.1
`1.73
`6.5
`5 (po)b
`3.5
`23.0
`13.4
`1.7
`5.6
`60 (iv)a
`0.43
`2.83
`23.4
`1.74
`11.8
`60 (po)b
`0.34
`1.55
`76.7
`3.4
`14.6
`61 (iv)a
`0.77
`6.5
`11.1
`1.40
`8.5
`61 (po)b
`0.86
`8.2
`39.39
`1.22
`14.7
`0.75
`aCompound was formulated in PEG200 and administered via single intravenous injection at 5 mg/kg. bCompound was formulated in PEG200 and
`administered per os (gavage) at 20 mg/kg.
`
`Tmax (h)
`
`0.33
`
`0.33
`
`52
`
`74
`
`79
`
`here. Noteworthy, 60 was also found to have better FAP/PREP
`selectivity and a very proficient ligand efficiency of 0.34, which
`is significantly higher than the corresponding value calculated
`for 4 (0.27).
`To further expand the P1-4,4-difluorinated subset, com-
`pounds 61−63 were synthesized. These differ by their P3-
`quinoline residues’ substitution patterns. The 6-methoxyquino-
`line 61 is the difluorinated analogue of an inhibitor that we
`reported in our foregoing study.23 Both the low nanomolar
`FAP potency and thousandfold FAP/PREP selectivity of these
`two molecules are highly comparable. Compounds 62 and 63
`were specifically included to extract additional information on
`the SAR of
`the P3-quinoline’s 5-position (vide supra).
`Surprisingly, 62 and 63 were found to be micromolar FAP
`inhibitors. Although the 5-methoxylated compound 57 (Table
`5) had a similar profile, drawing the general conclusion that
`introduction of substituents at the 5-position reduces FAP
`affinity, would oversee reference molecule 5. The latter all but
`compares to its difluorinated counterpart 62, but together with
`its 5-chlorinated analogue that we also published earlier,
`it
`belongs to the set of most potent and selective molecules
`discovered.29 More subtle hypotheses related to the com-
`pound’s binding kinetics might
`therefore be required to
`tentatively explain the observed behavior.
`It
`is clear that
`invoking only the steric or electronic parameters of
`the
`quinolone substituents is not sufficient to rationalize the assay
`data of all 5-substituted molecules we prepared.
`Finally, we investigated how either omitting or changing the
`warhead function to a boronic acid would impact FAP potency
`and selectivity of our molecules. We also selected the well-
`known chloromethyl ketone warhead to verify whether
`modification to irreversible inhibitors could be possible. All
`molecules prepared were analogues of reference 4 (Table 6).
`First, building in an unsubstituted pyrrolidine residue at P1 in
`64 was found to give a 3 log reduction in potency compared to
`reference compound 4. Qualitatively, this finding is in line with
`earlier reports by our own group and by Jiaang et al. and
`underscores the very considerable contribution to inhibitor
`affinity from the warhead function. Compared to the
`mentioned examples though, the residual micromolar affinity
`of 64 still is significant and one of the best potencies reported
`
`for FAP inhibitors without a warhead group. Next, the boronic
`acid warhead, present in many of the older series of (mostly
`nonselective) FAP inhibitors, was evaluated in 66. This
`molecule was found to be highly potent but also considerably
`less selective toward PREP when compared to its nitrile-based
`counterpart 4 ((SI[FAP/Prep] of 3 vs ∼85,
`respectively).
`Evaluation data of Bacho