PDEStrIAn: PDE Structure and

Ligand Interaction Annotated database

Chapter 2

Chimed Jansen, Albert J. Kooistra, Iwan J. P. de Esch, Rob Leurs and Chris de Graaf

Manuscript in preparation

44

45

CH

AP

TE

R 2

Abstract

A systematic analysis is presented of the 168 phosphodiesterase (PDE) catalytic domain

crystal structures present in the Protein Data Bank (PDB) with a focus on PDE-ligand

interactions. The PDE Structure and ligand Interaction Annotated (PDEStrIAn) database

contains a consistent alignment of 57 PDE ligand binding site residues, that enables the

identification of subtype-specific PDE-ligand interaction features and classification of

ligands according to their binding modes. We illustrate how systematic mining of PDE-

ligand interaction space gives new insights into how conserved and selective PDE

interaction hot spots can accommodate the large diversity of chemical scaffolds in PDE

ligands. A substructure analysis of the co-crystalized PDE ligands in combination with

those in the ChEMBL database provides a toolbox for scaffold hopping and ligand design.

These analyses lead to an improved understanding of the structural requirements of PDE

binding that will be useful in future drug discovery studies.

46

2.1 Phosphodiesterases as Drug Targets

Cyclic nucleotide Phosphodiesterases (PDEs) play a key role in regulating the levels of the

ubiquitous second messengers, cyclic adenosine monophosphate (cAMP) and cyclic

guanosine monophosphate (cGMP). PDEs therefore provide a handle for control of an array

of biochemical pathways throughout the body,1, 2 and have proven to be effective drug

targets.3 There are eleven human PDE families; PDE1, PDE2, PDE3, PDE10 and PDE11

hydrolyze both cAMP and cGMP; PDE4, PDE7 and PDE8 selectively hydrolyze cAMP;

and PDE5, PDE6 and PDE9 selectively hydrolyze cGMP.4 Kinetoplastid parasite PDEs are

also of interest as potential drug targets, the PDEA and PDEB families selectively hydrolyze

cAMP, while the PDEC family hydrolyzes both cAMP and cGMP.5, 6 The hydrolysis of the

cyclic nucleotides occurs in the substrate binding pocket of the PDE catalytic domain and

is catalyzed by two metal ions that occupy the adjacent metal binding region (Figure 1A-

B). The identity of one of the metal ions is required to be Zn2+ in order to maintain activity,

the identity of the second may vary, though in most cases it is Mg2+.7-9 X-ray co-crystal

structures of both the substrate and product give insight into the mechanism of the

hydrolysis, in which the substrate is held in place by interaction with a key conserved

glutamine residue, QQ.50, a “hydrophobic clamp” formed by IHC.32 and FHC.52, and ionic

bonds between the phosphate group and metal ions (Figure 1A-B, Figure 2). Hydrolysis of

the cyclic phosphate ester bond occurs through attack by a water molecule activated by the

metal ions.8, 10

The pervasive and tissue specific expression of PDEs, allows PDE inhibitors to be applied

in a wide range of therapeutic areas.11 To date twelve selective PDE inhibitors have been

approved for use as pharmaceuticals, of which five have been crystalized in complex with

a PDE (Figure 1C). The first of these to reach blockbuster status was sildenafil (Viagra®),

approved for the treatment of erectile dysfunction and pulmonary hypertension, thereby

establishing PDEs as highly successful drug targets. Three PDE5 inhibitors have followed,

tadalafil (Cialis®), vardenafil (Levitra®) and most recently avanafil (Stendra®), bringing

improvements in selectivity.12 Following considerable effort including the late stage failure

47

CH

AP

TE

R 2

of rolipram, two PDE4 drugs have entered the market, drotaverine (No-Spa®) as an

antispasmodic and recently roflumilast (Daxas®) as a treatment for COPD.13, 14 Five PDE3

drugs have reached the market; cilostazol (Pletal®) for the treatment of intermittent

claudication, anagrelide (Agrylin®) for the treatement of thrombocythemia, and the

cardiotonic vasodilators amrinone (Inocor®), enoximone (Perfan®) and milrinone

(Primacor®). Like the PDE4 drugs they display problematic contraindications, notably

nausea and cardiovascular side effects.15, 16 There are several marketed non-selective PDE

inhibitors targeting both PDE3 and PDE4 such as theophylline and several which inhibit a

broad range of PDE families such as dipyridamole. One of the oldest prescribed PDE drugs

is papaverine, which has long been used as a vasodilator. Recently papaverine was found

to be a potent and selective PDE10 inhibitor and it has played a role in generating interest

in PDE10 as a drug target.17 The development of selective inhibitors of PDE10 continues

to receive considerable interest, in spite of the recent failure of PF-02545920 to reach the

market.18 With only 4 of the 11 human PDE families targeted by selective drugs so far, and

drugs targeting PDEs present in parasitic organisms for the treatment of neglected diseases

yet to be developed, the potential for new drugs targeting PDEs is evident.

2.2 The Structure of Phosphodiesterases

All PDE inhibitors crystalized to date have been found to bind to the substrate binding

pocket of the PDE catalytic domain. The catalytic domain consists of up to 16 helices (H1-

H16) and 16 loops (A-N) that fold to form a substrate binding pocket (Figure 1D). When

all published PDE crystal structures are overlaid, it is clear that PDEs share a highly

conserved fold, with an overall RMSD of C-alpha atoms of 1.2Å (Figure 1E). The

conformation of the H-loop, which includes H8 and H9, is conserved across most PDE

crystal structures and borders the metal binding region. However significant differences

occur in certain crystal structures, for example in PDE5 the H-loop folds over the substrate

binding pocket in crystal structures containing a bound ligand. Similarly the M-loop which

also borders the substrate binding pocket shows some flexibility, including cases of induced

fit for bulky ligands. The most highly conserved fold is made up of 10 helices (H5-H7 and

H10-H16), with an RMSD across all PDE structures of 1.0Å (Figure 1F) and fold forms the

48

catalytic site which includes a metal binding region at the apex of H6 – H13 and a substrate

binding region formed between H13 – H16.

The PDE catalytic domain is adjoined by a lengthy variable N-terminus containing one or

more highly ordered structural elements or domains, and a usually short unstructured C-

terminus. The amino-terminal domains include the CaM-binding domain (CaM), GAF

domains, transmembrane domain (TM domain), upstream conserved regions (UCRs) and

PAS domain and may regulate the activity of the catalytic domain, play a role in the

localization of PDEs, or the interaction with protein partners.19-21 Crystal structures are

available of N-terminal GAF domains of PDE2 (complete dimer), PDE5, (GAF-A & GAF-

B), PDE6 (GAF-A) and PDE10 (GAF-B). Ligand binding has been observed for GAF

domains and this may result in ordering the catalytic domain in an open conformation.22

There are 168 PDE crystal structures in the Protein Databank at the time of writing (Figure

1G). The first crystal structure published of the PDE catalytic domain was an unliganded

structure of PDE4B2B containing metal ions, the structure established the 16 helix

nomenclature (H1-H16) for PDE structures.7 The first ligand bound PDE crystal structure,

containing zardaverine bound to PDE4D, provided insights into the role of the catechol

scaffold during binding and into PDE dimerization.23 A subsequent study in PDE4 showed

the structural basis for selectivity between the rolipram enantiomers.24 The first (catalyzed)

substrate bound structure soon followed with AMP bound to PDE4D providing insights into

the catalytic process.10 This was followed by an in depth study that involved PDE1B,

PDE4B, PDE4D and PDE5. As a result, the“glutamine switch” was identified as the

probable mechanism that controls substrate selectivity. Also the term the term “hydrophobic

clamp” was introduced to describe a hallmark ligand-PDE interaction .25 The number of

PDE crystal structures has continued to climb at a rate of about 15 per year and there are

now structures of the catalytic domains of 10 of the 11 human PDE families (a structure of

PDE11 is still lacking). At the subtype level only 13 of the 21 subtypes have been

crystalized (Figure 1G, structures of 1A, 1B, 3A, 6A, 6B, 7B, 8B and 11A are still lacking).

Beyond human PDEs, efforts to control parasite proliferation by means of PDE inhibitors

have shown great promise, yet only a small fraction of these PDE targets have been

crystalized.5, 26, 27

49

CH

AP

TE

R 2

The growth in novel PDE inhibitors included in the ChEMBL database since the year 2000

provides a different view on which PDE targets are of greatest interest (Figure 1H). As

might be expected PDE4 and PDE5 inhibitors make up the largest contribution, however

few PDE9 inhibitors have been published despite the publication of 18 PDE9 crystal

structures. In the case of PDE10, the discovery of novel inhibitors follows the release of a

significant number of crystal structures and in 2012 over half of the novel PDE inhibitors

published in ChEMBL targeted PDE10. Parasite PDEs have also gained significant interest

as targets to treat neglected diseases with growing numbers of both novel inhibitors and

crystal structures published. In the cases of PDE1 and PDE3 there are a significant number

of novel inhibitors published, despite the fact that just one PDE1 and two PDE3 crystal

structures having has been published.28, 29 In the case of PDE3, inhibitors of both PDE3 and

PDE4 may be an added factor in the number of active compounds registered in ChEMBL.30

Along with PDE7, PDE8 and PDE11, these are targets with limited crystallographic data to

support drug discovery, although it can be anticipated that pharmaceutical interest will lead

to structural biology activities for these new targets.

50

Figure 1: The growing structural understanding of PDEs and PDE inhibition

(A-B) The binding modes of cAMP (A, PDB: 2PW3) and AMP (B, PDB: 1PTW) bound to

PDE4D. The key interacting residues are shown colored by the pocket region, HMB.02

51

CH

AP

TE

R 2

(H160), NHC1.25 (N321), IHC.32 (I336), FS.35 (F340), QQ.50 (Q369) and FHC.52 (F372), and

named according to the nomenclature presented in Figure 2.

(C) Molecular structures and indications of the five marketed PDE inhibitors crystalized

with PDEs (1-5), two PDE inhibitors which failed to reach the market (6-7) and the PDE

substrates cAMP and cGMP with arrows indicating the bond broken during hydrolysis by

PDEs to form AMP and GMP (8-11).

(D-F) Structural overviews of PDEs. (C) A schematic diagram of the structure of PDEs.

Loops are lettered A-N (green) and helices are numbered H1-H16 (blue), the region of the

substrate binding pocket is highlighted (yellow). Two loops have been emphasized, the H-

loop (purple) which borders the substrate binding region and the M-loop (red) which

borders the metal binding region. The faded region, including H1 – H7, has been moved

from behind the protein to the side for clarity. (D) An overlay of the backbone ribbons of

all PDE crystal structures. (E) An overview of the conserved helices in PDE structures

showing the position of the substrate binding site as a surface. The colors of the surface

denote regions of the binding pocket as described in detail in Figure 2A.

(G) An overview of PDE crystal structure publications by year. Details of the number of

crystal structures published for each of the 21 subtypes spread over 11 PDE families are

shown to the right of the graph. The two PDE6C (*) crystal structures published are binding

pocket chimeras of PDE6C in PDE5A constructs.

(H) An overview of novel active PDE inhibitors published in the ChEMBL database by year

starting with the year 2000 and excluding earlier PDE inhibitors. This table provides an

indication of the influence of PDE crystal structures on the discovery of novel PDE

inhibitors.

2.3 A Novel PDE Binding Site Nomenclature

In order to describe the interactions between the ligands and the substrate binding pocket in

a consistent manner, we suggest a systematic nomenclature for the residues of the binding

pocket. To identify regions of the pocket involved in ligand binding and to identify

differences in binding across the PDE super family, the pocket was divided into 10 regions

(Figure 2A). The 10 regions consist of; the invariant glutamine (Q) regions Q, Q1 and Q2;

52

the hydrophobic clamp (HC) regions HC, HC1 and HC2; the metal binding (MB) regions

MB, MB1 and MB2; and the solvent filled (S) region S. A total of 57 residues were selected

as pocket residues to allow for variation of the pocket conformation and the range of binding

modes adopted by ligands (Figure 2B). Of the 57 pocket residues, 13 are conserved across

all PDEs, 11 of which are involved in metal binding (Figure 2C). The two other conserved

amino acids play a key role in substrate binding, these are the glutamine residue, QQ.50, and

the phenylalanine residue, FHC.52. The identities of the PDE substrate binding pocket amino

acids at each position can be compared across the PDE subtypes using the alignment

provided in Figure 2D. The only gaps in the alignment are found at Q2.44, as a result of the

variable length of the M-loop, and at Q2.31 where H14 tightens for one turn in TcPDEC.

Multiple splice variants of the PDE subtypes are expressed which affects the numbering of

amino acids, the amino acids in the alignment are numbered according to the canonical

sequence of each PDE subtype. In order to identify PDE pocket residues in a consistent

manner across the PDE families, novel nomenclature combines the amino acid sequence

reference with the pocket region name and the position of amino acid in the pocket sequence

(Figure 2E). Recently, a similar nomenclature was applied to enable the construction of an

automated database of kinase structures for public access in which protein-ligand

interactions are stored as IFPs.31

53

CH

AP

TE

R 2

54

Figure 2: The PDE ligand binding pocket of the catalytic domain

(A) The PDE ligand binding site shown as a surface over a representation of the protein

backbone (PDE4D, PDB: 1OYN). The surface is labeled and colored to show the 10 defined

regions of the binding site. The substrate binding site includes the Q, Q1 and Q2 regions

that surround the important invariant glutamine residue QQ.50, the HC, HC1 and HC2

regions that surround the hydrophobic clamp (I/V/LHC.32 and FHC.52) and the S region

solvent filled sub-pockets. The metal binding site is divided into the MB, MB1 and MB2

regions.

(B) The positions of the C-alpha atoms of pocket residues are shown as spheres in the color

of the regions to which they belong. The pocket residues are labeled according to their

position in the pocket and their position in the PDE sequence.

(C) A WebLogo representation of the conservation of the 57 amino acids of the PDE binding

pocket across the PDE subtypes listed in panel D (color coding as defined in panel A).

(D) An alignment of the pocket residues in each of the PDE subtypes of which crystal

structures have been published. Residue numbers are taken from the canonical sequence of

each PDE subtype. A color bar above the residues indicates the pocket region in which the

residues are found.

(E) A nomenclature is presented that combines the standard amino acid reference

containing the single letter amino acid code (red) and isoform specific residue number

(purple) with the PDE pocket residue region name (blue) and the PDE pocket residue

number (green). When referencing PDE pocket residues of a subtype the isoform number

may be omitted (YHC1.01) and when referencing PDE pocket residues across the families the

amino acid code and isoform number may be omitted (HC1.01).

Four previous analyses of PDE crystal structures have been published. The first described

the binding of 10 inhibitors to PDE4B, PDE4D and PDE5A in 15 crystal structures.32 An

analysis of ligand interactions is provided with key interactions across multiple PDEs

identified. Methods of improving selectivity and potency are discussed in detail with several

examples provided that show the effect of addressing regions of the pocket. An overview

of the therapeutic importance of PDE targets, which included the structural analysis of 20

crystal structures from PDE families 1, 3, 4, 5 and 9, found similar binding motifs to

55

CH

AP

TE

R 2

determine substrate and inhibitor binding.11 In another analysis of the same PDE structures,

in depth descriptions of the binding modes taken by several representative ligands, cAMP,

cGMP, theophylline, IBMX, zaprinast, sildenafil and rolipram are provided.33 The most

comprehensive previous analyses of PDE structures reviewed the first 60 PDE structures

published in the PDB, covering PDE families 1, 2, 3, 4, 5, 7 and 9.34 In that analysis the

core interactions between inhibitors and the invariant glutamine and hydrophobic clamp

were described as essential and multiple recognition elements were used to explain

selectivity in detail for PDE4 and PDE7. With over 100 PDE crystal structures added to the

PDB since the previous analysis by Ke and Wang and with new cheminformatics tools at

hand, the question of the drivers of PDE selectivity can now be revisited.

Through processing the available PDE crystallographic data in a systematic manner,

studying the protein fold, binding site interactions, ligand substructures and decorations,

and combing this data is an accessible format, we constructed the PDEStrIAn database as a

toolbox for drug designers. The utility of this toolbox is enhanced by the difficulty in

achieving selectivity between the PDE families and the potential PDE inhibitors have

shown as drugs indicate its relevance. To facilitate this and future studies of the PDE super

family a novel standardized nomenclature for PDE binding site residues is introduced here.

The PDEStrIAn database contains the most comprehensive structural analysis of PDEs to

date and the key findings of this analysis are presented here.

2.4 Building PDEStrIAn

The PDEStrIAn database was constructed to aid the systematic analysis of PDE crystal

structures published in the PDB. An overview of the method used to construct the

PDEStrIAn database is provided in Figure 3A and a detailed description follows below. The

canonical sequences of human and parasite PDE subtypes were collected from the UniProt

database and aligned using ClustalW (1). The PDE crystal structures containing a catalytic

domain were gathered from the Protein DataBank and a chain was selected for further

analysis according to their B-factor, ligand placement, the presence of gaps, solvent

molecules and Ramachandran plots (2). The structures were aligned together with the

56

canonical sequences using a combined sequence and structural alignment (3). The pocket

residues were defined and the sequence alignment was manually optimized to improve the

alignment of pocket residues (4). The definitive set of pocket residues received a new

consistent nomenclature including the name of the pocket region (5). The ligands, water

molecules, metal ions and pocket residues were isolated and processed using an interaction

fingerprint (IFP) generation protocol (6) forming the basis of the PDEStrIAn database (7).

The interactions were systematically analyzed by pocket region and according to the

substructures of ligands involved in interaction with specific regions (8). Interactions

between the ligand and water molecules and metal ions found in the crystal structures were

processed to assess their role in ligand binding (9). The scaffolds present in crystalized PDE

inhibitors were identified and the occurrence of the scaffolds in crystalized ligands and PDE

inhibitors found in the ChEMBL database was analyzed. The data can be arranged by PDE

pocket region and several example structures for the regions are provided. The PDEStrIAn

database can be applied to aid the design of novel PDE inhibitors by presenting knowledge

from known PDE inhibitors in an accessible format.

2.5 IFP Generation

A key step in the construction of the PDEStrIAn database was the preparation of IFPs from

the crystal structure complexes. The binding mode and IFP of sildenafil (Figure 1C) to

PDE5A (PDB: 1UDT) illustrates the use of IFPs to encode protein-ligand interactions

(Figure 3B-D). The binding mode of sildenafil is presented from the pocket opening

showing all residues interacting with sildenafil (Figure 3B). Figure 3C shows a top view

projection in which sildenafil, the residues forming specific interactions with sildenafil and

the metal ions are shown projected over the surface of the pocket, this is the visualization

method used throughout the remainder of the article. The IFP of the binding mode of

sildenafil is presented in Figure 3D. The IFP is a bit string in which each bit encodes the

presence (1) or absence (0) of a particular interaction type between a protein residue and a

ligand 35. The bit string is made up of five bits per binding site residue; a hydrophobic

interaction bit; a face-to-face π-π interaction bit; an edge-to-face π-π interaction bit; a

hydrogen bond acceptor bit; and a hydrogen bond donor bit. Ionic interactions and cation-

57

CH

AP

TE

R 2

π interactions were excluded from the bit string because these interactions were not found.

In the example of sildenafil bound to PDE5A, 13 residues show the presence of hydrophobic

interactions, QQ.50 additionally shows hydrogen bond acceptor and donor interactions and

FHC.52 additionally shows face-to-face and edge-to-face π-π interactions. The IFP of a

protein-ligand complex can be rapidly compared to those of other ligands bound to the same

or similar proteins, to identify key residues involved in ligand binding, or to find similarities

in the binding modes of multiple ligands.

Figure 3: Building the PDEStrIAn database

(A) An overview of the steps undertaken to build the PDEStrIAn database. For those steps

represented in figures, the relevant figure names are provided. Examples of structures

which show specific interactions with pocket regions have been retrieved from the database

as indicated at the bottom of the flowchart.

58

(B-C) Sildenafil (2D structure presented in Figure 1C) bound to PDE5A (PDB: 1UDT)

shown from the opening of the ligand binding pocket and from above. Residues and areas

of the pocket are color coded according to pocket region.

(D) The IFP bit string for residues interacting with sildenafil. Each residue displays five bit

positions that can either be on (1), indicating the presence of an interaction, or off (0),

indicating the absence of an interaction. The bits that are on are color coded according to

the type of interaction made with the ligand. For clarity only the 13 of the 57 binding site

residues that have at least one interaction with sildenafil are shown in the bit string.

2.6 Ligand-PDE Interaction Analysis

The presence of interactions between ligands and residues in the PDE crystal structures can

be aggregated for each PDE family and plotted into a heat map as shown in Figure 4.

Comparing the frequency of interactions made by each pocket residue across all PDE-ligand

complexes four residues stand out, I/V/LHC.32, F/YS.35, QQ.50 and FHC.52. Interactions with

these residues are seen across all PDE families and in almost all crystal structures.

Hydrophobic interactions with I/V/LHC.32 occur in all but one structure, and with F/YS.35 in

all but two structures. Aromatic interactions with F/YS.35 occur in 57% of structures and

with FHC.52 in all PDE crystal structures. Hydrogen bond donor or acceptor interactions with

QQ.50 are found in over 90% PDE crystal structures. These key interactions are driven by

the core scaffolds of the bound ligands, which consistently include a flat aromatic or fully

conjugated ring system and one or more hydrogen bond donors or acceptors.

The heat map in Figure 4 could potentially also be used to identify interactions that are

specific to particular PDE subtypes. These PDE subtype specific interaction hotspots may

be important drivers for PDE ligand selectivity, although one should be aware that for

several subtypes relatively few structures have been solved (PDE3, PDE3, PDE6, PDE7,

PDE8, parasite PDEs). Still, this data can be applied to quickly identify interactions of

interest when designing selectivity into PDE inhibitors. Examples of selective polar

interactions will be discussed in the paragraphs addressing the pocket regions, Q2 (Figure

8C; PFK interacts with YQ2.33 in PDE10), HC1 (Figure 9A; HBT interacts with NHC1.25 in

59

CH

AP

TE

R 2

PDE4B), S (Figure 9D; 0WL (CHEMBL2180070) interacts with YS.35 in PDE9) and MB1

(Figure 10D; IHM interacts with MMB1.17 in PDE4D).

60

Figure 4: Frequency of interactions between pocket residues and ligands in crystal

structures of PDEs

A heat map of the IFP results for PDE crystal structures showing the frequency of

interaction types between the bound ligands and pocket residues for each PDE family. The

colors indicate the interaction type and the intensity of colors indicates the percentage of

crystal structures of a PDE family in which a specific interaction takes place, with white

representing 0% and solid color indicating 100%.

2.7 Ligand-Water Interactions

Water molecules play an important role in protein ligand binding and an analysis of

interactions between crystallographic water molecules and ligands in PDE crystal structures

was performed to identify key water molecules. The superposed PDE structures allow the

identification of areas in the binding sites in which a significant number of water molecules

have been found within a proximity of 1.5A (Figure 5). From this study, one particular

cluster (D) was shown to contain 49 water molecules from crystal structures of PDE2-PDE7

and PDE9-PDE10. In one crystal structure the water molecule at this particular position has

been displaced by a ligand, one of two conformations observed for zardaverine.32 The

structural water molecules in cluster D are bound by the residues DMB.22 and F/YHC1.1

stabilizing their position. Water at this position does not appear to form hydrogen bonds

with the cyclic nucleotides, but seems to form hydrogen bonds with AMP and GMP, thereby

stabilizing the catalytic product of PDEs. This consistent placement of a water molecule

indicates that water at this position should be considered during molecular modeling and

virtual screening studies.

The other clusters of water molecules are less general for PDEs. For example cluster B

containing 9 water molecules is only found in PDE9 structures, and occurs in a unique sub-

pocket that has not yet been addressed by PDE9 inhibitors. Cluster G contains 10 PDE4

waters and 3 PDE10 waters. However, given the proximity of this cluster to the metal ions,

this may have more to do with the types of inhibitors being developed for these targets than

a particular specificity of water molecules to those PDEs. Two clusters are found at the

61

CH

AP

TE

R 2

solvent exposed side of the pocket and they are dependent on the particular ligand present

in the crystal structure. Cluster I contains water molecules bound to nucleotide-like

moieties, while in structures containing cluster H water molecules, catechols are the most

common ligand scaffolds.

Figure 5: Water molecules involved in PDE ligand binding

(A) Crystalographic water molecules forming interactions with PDE ligands. The oxygen

atom of each water is shown, colored by the PDE family of the crystal structure from which

it was extracted. Nine clusters were identified, these are circled and labeled A-I, with A-F

forming interactions with ligands and residues (magenta), G forming interactions with

ligands, metals and the protein (red) and H-I forming interactions with just the ligand

(cyan). The water molecules are shown over a 1OYN (PDB:) pocket surface.

(B) The water molecules that interact with PDE inhibitors in crystal structures shown with

an overlay of all PDE ligands and metal ions. Water molecules are colored according to

the interactions they form, cyan form interactions with the ligand, magenta form

interactions with ligands and protein residues, purple form interactions with the ligand and

metals and red form interactions with ligands, metals and the protein residues.

2.8 Ligand Scaffold Analysis

62

The structural motifs of PDE ligands can be assessed according to the sub-pockets with

which they interact. This provides a way to quickly identify those functional groups that

have successfully been used to address a given sub-pocket. To aid such efforts a scaffold

analysis was performed using the core scaffold of each ligand crystalized with a PDE

(Figure 6). The information about the binding mode of the scaffold and vectors of the side-

chains was retained by superposing the crystal structures and systematically extracting data

(Figure 7).

63

CH

AP

TE

R 2

64

Figure 6: Scaffold analysis of the ligands in PDE crystal structures

The core scaffold from each ligand crystalised with a PDE was identified. Scaffolds were

defined such that only one scaffold would be identified in each ligand. The scaffolds were

screened against active PDE inhibitors published in the ChEMBL database and the number

of hits are shown in green. The number of unique hits among ligands crystalized with PDEs

are shown, as well as the total number of hits and the PDB code of each structure. The R-

groups were collected for each hit molecule and the number of unique R-groups is provided

at each attachment point, ChEMBL R-groups are in green and crystal structure R-groups

are in black. The R-groups indicate points of attachment and have been numbered

according to the vector of the attachment in each crystal structure. Numbers run

sequentially as the angle of the vector changes from 16 (0º back of the pocket) to 4 (90º

towards metal ions) to 8 (180º towards solvent) to 12 (270º towards QQ.50) as described in

Figure 7. In Figure 6 alternate binding modes of scaffolds have been left out and the R-

groups from alternate binding modes are included with the R-groups of the most common

binding mode of the scaffold.

65

CH

AP

TE

R 2

A B

C D

Figure 7: The steps involved in defining the scaffolds of crystalized ligands. A In this step

the crystal structure is oriented in a consistent manner following alignment. B) A schematic

structure of the ligand is drawn using the orientation of the crystal structure as a guide. C)

The points of attachment around the scaffold of the crystalized PDE ligands were named

according to the vector of the bond broken using the chart shown. D) The placement of the

scaffolds was consistent with the orientation of the ligand in the pocket allowing the vector

to provide information about the placement of R-groups in the PDE binding pocket.

The most common scaffold type among crystalized ligands bound to PDEs are purines, with

IBMX, cAMP/AMP and cGMP/GMP accounting for most cases. The purines are

remarkable in that they often act as both hydrogen bond acceptors and donors to the

conserved glutamine, QQ.50, whereas most scaffolds only act as hydrogen bond acceptors.

66

Another common scaffold type found particularly in PDE4 inhibitors is the catechol-like

moiety. While ethers are not the strongest hydrogen bond acceptors, the combination of two

acceptors targeting QQ.50 has resulted in highly potent PDE4 inhibitors. The quinolines are

another scaffold type commonly deployed in PDE inhibitor design. The nitrogen atoms in

quinolines usually act as acceptors where hydrogen bonds are formed with QQ.50. A common

theme across PDE ligands is the presence of a fused ring system in the scaffold. These fused

ring systems can be optimally accommodated in the narrow hydrophobic clamp between

the aliphatic residues at position I/V/LHC.32 and the conserved aromatic residues at position

FHC.52.

The scaffold was defined as the smallest fragment of each ligand involved in interactions

with the invariant glutamine and hydrophobic clamp, that retains the character of the

moieties involved in those interactions and prevents redundancy across ligands crystalized

with PDEs. A count of the scaffolds in ligands crystalized with PDEs and active PDE

inhibitors found in the ChEMBL database was also made. The bonds broken to isolate the

scaffolds were assigned R numbers according to a cyclic scheme based on the vector of the

bond broken using the orientation of the ligand in the crystal structure as a reference. In this

way R-groups with particular vectors in the pocket could be grouped together to allow the

crossing of R-groups from multiple ligands in a for example a Markush enumeration. The

R-groups found in ligands crystalized with PDEs and in PDE inhibitors found in the

ChEMBL database were gathered into a database for analysis. A Markush library containing

a scaffold and all R-group variations for a vector can be used in combination with docking

to efficiently probe a particular sub-pocket with relevant chemical diversity.

An analysis was also made of the moieties of ligands that interact with regions of the

substrate binding pocket in PDE crystal structures. In this case the interactions identified by

IFP analysis were the starting point for identification of fragments of the ligands involved

in interactions. For example 84 ligands across most PDE families only address the Q1

pocket with fragments that form hydrophobic interactions. Six ligands bound to PDE4 form

aromatic interactions with Q1 and those fragments are found in 13 PDE4 and PDE5 ligands

forming interactions with the Q1 pocket. Hydrogen bonding with the Q1 pocket is seen in

67

CH

AP

TE

R 2

5 crystal structures, 2 in PDE2, 1 in PDE3, PDE5 and PDE10. In each of these cases these

fragments were only found in compounds forming hydrogen bonds with the Q1 pocket

suggesting these fragments could be utilized in more ligands to improve binding.

2.9 Interactions with the Q1 and Q2 Pockets

The Q1 pocket flanks the invariant glutamine QQ.50 forming a sub-pocket deep in the

substrate binding site. Interactions with the Q1 pocket are nearly exclusively hydrophobic,

however two exceptions are shown in Figure 8. The Q1 pocket shows significant water

molecule occupancy in PDE crystal structures, indicating the potential to accommodate

polar functional groups. However, few ligands exploit this potential. Ligands that form

polar interactions with Q1 are among the most potent for each family, showing that the

potential of addressing the Q1 pocket is underutilized in PDE drug discovery efforts. The

high selectivity of the dihydropyridazinones for PDE3 over for example PDE4 and PDE5,36

can be attributed to the hydrogen bond interaction that the pyridazinone carbonyl oxygen

can form with HQ1.27, as illustrated for the PDE3 inhibitor MERCK1 (IC50 = 0.11 nM) in

Figure 8A (PDB: 1SO237). The corresponding YQ1.27 residue in PDE4 forms an hydrogen

bond with the conserved QQ.50 residue in most crystal structures. In PDE5 the QQ1.27 residue

also forms hydrogen bonds with QQ.50 in most crystal structures. However of the ligands

only 3P4 (IC50 = 5.5 nM) shown in Figure 8B (PDB: 2CHM), forms a hydrogen bond with

QQ1.27 in a PDE5 crystal structure.38

The Q2 pocket lies adjacent to QQ.50 towards the opening of the PDE substrate pocket. The

size of the Q2 pocket is family dependent, with significant Q2 pockets seen in structures of

PDE1, PDE10 and the parasite PDEs (LmjPDEB1, TcrPBEC and TbrPDB1). In the parasite

PDEs a sub-pocket of the Q2 pocket, dubbed the P-pocket, has been targeted in attempts to

achieve selectivity over human PDEs.39 The key residues involved in ligand interactions are

located at positions Q2.33, Q2.46 and Q2.49, that form hydrophobic interactions in 55%,

20% and 67% of structures respectively. The Q2 pocket plays a particularly important role

in inhibitor design for PDE10, where 30% of ligands address the Q2 pocket, forming both

π-π interactions and hydrogen bonds with YQ2.33. The selective PDE10 inhibitor PFK (IC50

68

= 17 nM) in Figure 8C (PDB: 3QPN) was discovered using molecular docking of a virtual

combinatorial library with the specific aim of improving selectivity by addressing the Q2

pocket.40 Interestingly, while 90% of ligands in PDE crystal structures form a hydrogen

bond with QQ.50, 24% of PDE10 inhibitors do not form this hydrogen bond, including PF-

02545920 shown in Figures 8D (PDB: 3HR141). A traditional hydrogen bond may be

compensated by the presence of an N-H···π hydrogen bond between QQ.50 and a phenyl ring

in the ligand. The contribution of the N-H···π hydrogen bonds may be significant as

indicated by the potency of PF-02545920 (IC50 = 0.37nM). Additionally, hydrogen bonds

to the Q2 pocket are unique to PDE10, where ligands form hydrogen bonds with YQ2.33

further stabilizing occupation of the Q2 pocket.

69

CH

AP

TE

R 2

Figure 8: Examples of ligands interacting with the Q1 and Q2 pockets.

(A-D) Examples of compounds that interact with the Q1 and Q2 pockets. The binding

modes are shown of; (A) MERCK1 to PDE3A (PDB: 1SO2) (B) 3P4 to PDE5A (PDB:

2CHM), superposed over a 1UDT pocket surface, (C) PFK to PDE10A (PDB: 3QPN),

superposed over a 3HR1 (PDB:) pocket surface, and (D) PF-02545920 to PDE10A (PDB:

3HR1), superposed over a 3HR1 pocket surface).

70

(E) The IFP bit strings of the proteins and ligands shown in A-D. Only those residues that

form at least one interaction with one of the four ligands are shown.

2.10 Interactions with the HC1, HC2 and S Pockets

Almost all ligands in PDE co-crystal structures interact with residues located in the HC1

region around the hydrophobic clamp (98%). Although most ligands only form hydrophobic

interactions with HC1, several also form hydrogen bonds with the residue at position

HC1.25, for example HBT (CID 1021494, IC50 = 50 nM) shown in Figure 9A (PDB:

3HMV) forms a hydrogen bond to NHC1.25 of PDE4B via an amide group.42 The binding of

HBT also shows a unique displacement of QQ.50 away from the pocket, despite its potent

inhibition of PDE4B. A series of halogenated pyrimidinone PDE5A inhibitor analogues

were designed, including fluorinated 5FO (PDB: 3SHY), chlorinated 5CO (PDB: 3SHZ),

and the brominated 5BO (IC50 = 13 nM) shown in Figure 9B (PDB: 3SIE), that target YH1.01

in PDE5A via a putative halogen bond.43 Interestingly, experimentally determined IC50

values showed a good correlation with the calculated halogen bond energies, demonstrating

that the halogen bond is an applicable tool in the design of PDE inhibitors. By contrast, the

HC2 region interacts with just 21% of ligands, and exclusively through hydrophobic

interactions. In Figure 9C (PDB: 2OVY) the quinoxaline scaffold of the potent PDE10A

inhibitor PFJ (CHEMBL219445, IC50 = 6 nM) targets HC2 44, forming hydrophobic

interactions with GHC2.51, AHC2.54 and VHC2.55 in PDE10A. In SAR studies it was found that

this bulky group does not fit the smaller HC2 sub-pocket of PDE3A/B, explaining its high

selectivity for PDE10A over PDE3. The S pocket is primarily addressed by interactions

between the core scaffolds of ligands and the residue at position F/YS.35, a phenylalanine in

all PDEs except PDE9 where a tyrosine is found at this position. The presence of tyrosine

at position YS.35 in S pocket has been used in the structure-based design of PDE9A

inhibitors. In Figure 9D (PDB: 4G2L), the protonated nitrogen atom in 0WL

(CHEMBL2180070, IC50 = 32 nM) forms a hydrogen bond with YS.35.45 The compound

forms part of a structure-based effort to improve brain penetration of the closely related

Alzheimer drug candidate PF-04447943, which completed a phase II trial in 2010. The

water mediated hydrogen bond between PF-04447943 and YS.35 was replaced by a direct

71

CH

AP

TE

R 2

hydrogen bond in the binding of 0WL (CHEMBL2180070) reducing potency and

selectivity, but improving brain penetration.

Figure 9: Examples of ligands interacting with the HC1, HC2 and S pockets.

(A-D) Examples of compounds that interact with the HC1, HC2 and S pockets. The binding

modes of are shown of; (A) HBT bound to PDE4B (PDB: 3HMV), superposed over a 1OYN

72

(PDB:) pocket surface, (B) 5BO bound to PDE5A (PDB: 3SIE) superposed over a 1UDT

(PDB:) pocket surface, (C) PFJ to PDE10A (PDB: 2OVY) superposed over a 3HR1 (PDB:)

pocket surface, and (D) 0WL to PDE9A (PDB: 4G2L) superposed over a 3K3E (PDB:)

pocket surface.

(E) The IFP bit strings of the proteins and ligands shown in A-D. Only those residues that

form at least one interaction with one of the four residues are shown.

2.11 Interactions with the MB, MB1 and MB2 Pockets and the Metal

Ions

The MB pocket contains the metal ions responsible for the hydrolysis of the cyclic

nucleotides cAMP and cGMP to AMP and GMP respectively. The binding modes of the

products are shown in Figure 10A (PDB: 1TB7) showing AMP bound to PDE4D and Figure

10B (PDB: 1T9S) showing GMP bound to PDE5A 25. The cyclic nucleotides adopt very

similar binding modes with each making ionic bonds to both Zn2+ and Mg2+. The pattern of

hydrogen bond donors and acceptors surrounding the adenine ring of AMP and guanine

ring of GMP differ. The PDEs match these differing hydrogen bond patterns by flipping

QQ.50. Although this “glutamine switch” was proposed to drive cyclic nucleotide selectivity,

multiple differences are now thought to control substrate specificity with no individual

residue playing a dominant role.46

Just two of the crystalized PDE inhibitors form tight interactions with the catalytic metal

ions in PDEs, zardaverine (IC50 = 0.39 µM, PDE4D, PDB: 1XOR32) and AN2898 (IC50 =

0.24 µM, PDE4B, PDB: 3O0J47), shown in Figure 10C. These compounds do not display

unusual potency as PDE inhibitors, suggesting that addressing the metal ions directly

provides little additional interaction energy if any, most likely due to the displacement of

structural water molecules during binding. The oxaborole of AN2898 that binds to the metal

ions and HMB.02 is unique amongst crystalized PDE inhibitors, as is the catechol like

dicyanophenoxy ring that interacts with QQ.50. The MB1 and MB2 pockets form few

specific interactions with ligands, only IHM (CID 78225170, IC50 = 0.5 nM) bound to

PDE10A (PDB: 3V9B) shown in Figure 10D forms a hydrogen bond with MMB1.17.48 In

73

CH

AP

TE

R 2

Figure 9C PFJ (CHEMBL219445) is shown to form π-π interactions with FMB1.20, an

interaction shared by two analogues and the recent PDE2 crystal structure containing

BAY60-7550 (PDB: 4HTX). All other inhibitor interactions with MB1 and MB2 are

hydrophobic.

74

Figure 10: Functional groups interacting with the MB, MB1 and MB2 pockets illustrated

with examples.

(A-D) Examples of compounds which interact with the MB, MB1 and MB2 pockets. The

binding modes of are shown of; (A) AMP to PDE4D (PDB: 1TB7), superposed over a 1OYN

(PDB:) pocket surface, (B) GMP to PDE5A (PDB: 1T9S) superposed over a 1UDT (PDB:)

pocket surface, (C) AN2898 to PDE4B (PDB: 3O0J) superposed over a 1OYN (PDB:)

pocket surface, and (D) IHM to PDE4D (PDB: 3V9B) superposed over a 1OYN (PDB:)

pocket surface.

(E) The IFP bit strings of the proteins and ligands shown in A-D. Only those residues which

form at least one interaction with one of the four residues are shown.

2.12 Structure-Based PDE Ligand Discovery and Design

The superposition of all PDE crystal structures combined with the annotation of the binding

pockets, water clusters, and the generated PDE-ligand interaction data, make the

PDEStrIAn database especially suited for ligand design and structure-based optimization.

The consistent manner of the database creation allows for an easy comparison of multiple

structures, co-crystallized inhibitors and their interactions. The fragmentation of crystalized

PDE ligands into scaffolds and R-groups taking the binding conformation into account and

applying this to generate libraries of orientation specific R-groups and scaffolds provides a

toolbox for addressing different pocket regions and scaffold hopping. Together these tools

can support future structure-based PDE ligand discovery and design, including structure-

based virtual screening,26 fragment growing,49, 50 and modulation of PDE activity and

selectivity.19

75

CH

AP

TE

R 2

Figure 11: Drug design case studies: fragment growing and allosteric modulations

(A-D) The growing of the PDE4D fragment screening hit, 3,5-dimethyl-1H-pyrazole-4-

carboxylic acid ethyl ester (A, PDB: 1Y2B), with an N-phenyl (B, PDB: 1Y2C), an

additional para-methoxy group (C, PDB: 1Y2D) and replacing this with a meta-nitro group

(D, PDB: 1Y2K). Each structure is shown superposed over a 1OYN (PDB:) pocket surface.

(E/J) The IFP bit strings of the proteins and ligands shown in A-D. Only those residues

which form at least one interaction with one of the four residues are shown.

(F-I) Interactions between ligands bound to the PDE catalytic site and regions that are able

to fold over the pocket. In the structure of roflumilast bound to the catalytic domain of

PDE4D (F, PDB: 1XOQ, superposed over a PDB: 1OYN pocket surface) the catalytic

pocket is open. In the structure of Icarisid II bound to PDE5A (G, PDB: 2H44, superposed

over a PDB: 1UDT pocket surface) the H-loop folds over the pocket entrance enclosing the

ligand. The structure of atizoram bound to the catalytic domain of PDE4B shows closing of

the pocket by the C-terminal residues (H, PDB: 3KKT, superposed over a surface of PDB:

76

1OYN). In the case of D159153 bound to a PDE4D construct which includes the UCR2

regulatory domain, the pocket is closed off by a helix of UCR2 (I, PDB: 3IAD, superposed

over a PDB: 1OYN pocket surface).

An example of fragment growing in PDE4D is shown in Figures 11A-D. An initial fragment

screening yielded 3,5-dimethyl-1H-pyrazole-4-carboxylic acid ethyl ester as an initial hit

for further optimization (PDE4D IC50 = 82 µM49). Structure-guided fragment growing

starting from the pyrozole hit (Figure 11A, PDB: 1Y2B) and guided with six further PDE4B

and PDE4D co-crystal structures enabled the design of a series potent of PDE4 inhibitors.

The addition of a phenyl group in a first round of synthesis that forms a π-π interaction with

HMB.02 brought a 400-fold increase in potency (PDE4D IC50 = 0.27 µM, Figure 11B, PDB:

1Y2C). The addition of a para-methoxy group (Figure 11C, PDB: 1Y2D) resulted in a

binding mode switch out of the pocket and a 10-fold reduction in potency (PDE4D IC50 =

2.0 µM). Placing a nitro group in the meta position instead resulted in the inhibitor

coordinating with the magnesium ion (Figure 11D, PDB: 1Y2K) and a 4000-fold increase

in potency over the initial screening hit in just two rounds of synthesis (PDE4D IC50 = 0.021

µM49).

Ligands can be designed to not only target the PDE catalytic site (PDE4D, PDB: 1XOQ,

Figure 11F), but also protein regions that fold over the opening of the catalytic site. The

catalytic site can be closed off by the H-loop extending over the pocket, examples of this

fold are seen in PDE2A and PDE5 (Figure 11G, PDB: 2H44) crystal structures. Moreover

in PDE4 structures the C-Terminus can fold over to close the pocket as shown in Figure

11H (PDE4D, PDB: 3KKT). The extent to which a ligand is encapsulated by a PDE pocket

can be influenced by the design of the ligand and this will impact the selectivity and kinetics

of binding. An example of the structure-based design of interactions with residues that close

the PDE pocket is that of UCR2 interactions with PDE4 inhibitors. This interaction has been

utilized to modulate inhibition and achieve selectivity for PDE4D over PDE4B.19, 51 The

proposed mechanism of regulation, following kinetic and structural studies of PDE4

inhibitor binding, is a two-site model with negative cooperativity. In this model PDE4 forms

a dimer and the binding of UCR2 to one monomer reduces the affinity of UCR2 to the other

77

CH

AP

TE

R 2

monomer. Inhibitors which form strong interactions with UCR2 while bound to PDE4 are

therefore likely to stabilize a partial inhibition of PDE4, where one monomer is full and

closed, while the other is empty and open. Through modulation of the interactions between

the ligand and UCR2 it is possible to design partial inhibitors of PDE4 that are less emetic

than full inhibitors. Additionally by forming interactions specific to UCR2 in one PDE4

subtype, PDE4 subtype selectivity becomes possible. Through the application of this

knowledge the authors were able to design a series of PDE4 inhibitors with improved side-

effect profiles based on mouse model studies. One of these inhibitors, D159153 (PDE4D7

IC50 < 1 nM, 89% Imax), is shown in Figure 11I (PDE4D, PDB: 3IAD).

2.13 Conclusion

Crystal structures of PDEs were extensively analyzed across the PDE superfamily resulting

in database of PDE-ligand interactions complemented with an analysis of ligand-water and

ligand-metal interactions. Through the aggregation of this data into a thorough overview,

key interactions and PDE subtype specific interactions are easily identifiable. A novel

nomenclature for PDE pocket residues is proposed that enables cross family comparisons

of PDE crystal structures to be performed systematically. A scaffold analysis of crystalized

PDE ligands provides a toolbox for computational chemists to perform scaffold hopping or

Markush R-group substitution studies when targeting PDEs. By relating R-groups to vectors

in the pocket and by additional analysis of ligand substructures interacting with specific

regions of the PDE catalytic pocket, support will be provided for the design of novel PDE

inhibitors with improved selectivity and potency profiles.

References

1. Conti, M.; Beavo, J., Biochemistry and Physiology of Cyclic

Nucleotide Phosphodiesterases: Essential Components in Cyclic Nucleotide

Signaling. Annu. Rev. Biochem. 2007, 76 (1), 481-511.

2. Houslay, M. D., Underpinning compartmentalised cAMP signalling

through targeted cAMP breakdown. Trends Biochem. Sci. 2010, 35 (2), 91-

100.

78

3. Keravis, T.; Lugnier, C., Cyclic nucleotide phosphodiesterase (PDE)

isozymes as targets of the intracellular signalling network: benefits of PDE

inhibitors in various diseases and perspectives for future therapeutic

developments. Br. J. Pharmacol. 2012, 165 (5), 1288-1305.

4. Bender, A. T.; Beavo, J. A., Cyclic Nucleotide Phosphodiesterases:

Molecular Regulation to Clinical Use. Pharmacol. Rev. 2006, 58 (3), 488-

520.

5. Wang, H.; Kunz, S.; Chen, G.; Seebeck, T.; Wan, Y.; Robinson, H.;

Martinelli, S.; Ke, H., Biological and Structural Characterization of

Trypanosoma cruzi Phosphodiesterase C and Implications for Design of

Parasite Selective Inhibitors. J. Biol. Chem. 2012, 287 (15), 11788-11797.

6. Oberholzer, M.; Marti, G.; Baresic, M.; Kunz, S.; Hemphill, A.;

Seebeck, T., The Trypanosoma brucei cAMP phosphodiesterases

TbrPDEB1 and TbrPDEB2: flagellar enzymes that are essential for parasite

virulence. FASEB J. 2007, 21 (3), 720-731.

7. Xu, R. X.; Hassell, A. M.; Vanderwall, D.; Lambert, M. H.; Holmes,

W. D.; Luther, M. A.; Rocque, W. J.; Milburn, M. V.; Zhao, Y.; Ke, H.;

Nolte, R. T., Atomic Structure of PDE4: Insights into Phosphodiesterase

Mechanism and Specificity. Science 2000, 288 (5472), 1822-1825.

8. Liu, S.; Mansour, M. N.; Dillman, K. S.; Perez, J. R.; Danley, D. E.;

Aeed, P. A.; Simons, S. P.; LeMotte, P. K.; Menniti, F. S., Structural basis

for the catalytic mechanism of human phosphodiesterase 9. Proc. Natl.

Acad. Sci. U.S.A. 2008, 105 (36), 13309-13314.

9. Francis, S. H.; Blount, M. A.; Corbin, J. D., Mammalian Cyclic

Nucleotide Phosphodiesterases: Molecular Mechanisms and Physiological

Functions. Physiol. Rev. 2011, 91 (2), 651-690.

10. Huai, Q.; Colicelli, J.; Ke, H., The Crystal Structure of AMP-Bound

PDE4 Suggests a Mechanism for Phosphodiesterase Catalysis†,‡.

Biochemistry (Mosc). 2003, 42 (45), 13220-13226.

11. Jeon, Y. H.; Heo, Y. S.; Kim, C. M.; Hyun, Y. L.; Lee, T. G.; Ro, S.;

Cho, J. M., Phosphodiesterase: overview of protein structures, potential

therapeutic applications and recent progress in drug development. CMLS,

Cell. Mol. Life Sci. 2005, 62 (11), 1198-1220.

12. Wang, R.; Burnett, A. L.; Heller, W. H.; Omori, K.; Kotera, J.;

Kikkawa, K.; Yee, S.; Day, W. W.; DiDonato, K.; Peterson, C. A.,

Selectivity of Avanafil, a PDE5 Inhibitor for the Treatment of Erectile

Dysfunction: Implications for Clinical Safety and Improved Tolerability.

The Journal of Sexual Medicine 2012, 9 (8), 2122-2129.

79

CH

AP

TE

R 2

13. Houslay, M. D.; Schafer, P.; Zhang, K. Y. J., Keynote review:

Phosphodiesterase-4 as a therapeutic target. Drug Discovery Today 2005, 10

(22), 1503-1519.

14. Hatzelmann, A.; Morcillo, E. J.; Lungarella, G.; Adnot, S.; Sanjar,

S.; Beume, R.; Schudt, C.; Tenor, H., The preclinical pharmacology of

roflumilast – A selective, oral phosphodiesterase 4 inhibitor in development

for chronic obstructive pulmonary disease. Pulm. Pharmacol. Ther. 2010,

23 (4), 235-256.

15. Thompson, P. E.; Manganiello, V.; Degerman, E., Re-Discovering

PDE3 Inhibitors - New Opportunities for a Long Neglected Target. Curr.

Top. Med. Chem. 2007, 7 (4), 421-436.

16. Giembycz, M. A., Life after PDE4: overcoming adverse events with

dual-specificity phosphodiesterase inhibitors. Curr. Opin. Pharmacol. 2005,

5 (3), 238-244.

17. Siuciak, J. A.; Chapin, D. S.; Harms, J. F.; Lebel, L. A.; McCarthy,

S. A.; Chambers, L.; Shrikhande, A.; Wong, S.; Menniti, F. S.; Schmidt, C.

J., Inhibition of the striatum-enriched phosphodiesterase PDE10A: A novel

approach to the treatment of psychosis. Neuropharmacology 2006, 51 (2),

386-396.

18. Banerjee, A.; Narayana, L.; Raje, F. A.; Pisal, D. V.; Kadam, P. A.;

Gullapalli, S.; Kumar, H.; More, S. V.; Bajpai, M.; Sangana, R. R.; Jadhav,

S.; Gudi, G. S.; Khairatkar-Joshi, N.; Merugu, R. R. T.; Voleti, S. R.; Gharat,

L. A., Discovery of benzo[d]imidazo[5,1-b]thiazole as a new class of

phosphodiesterase 10A inhibitors. Bioorg. Med. Chem. Lett. 2013, 23 (24),

6747-6754.

19. Burgin, A. B.; Magnusson, O. T.; Singh, J.; Witte, P.; Staker, B. L.;

Bjornsson, J. M.; Thorsteinsdottir, M.; Hrafnsdottir, S.; Hagen, T.; Kiselyov,

A. S.; Stewart, L. J.; Gurney, M. E., Design of phosphodiesterase 4D

(PDE4D) allosteric modulators for enhancing cognition with improved

safety. Nat Biotech 2010, 28 (1), 63-70.

20. Heikaus, C. C.; Pandit, J.; Klevit, R. E., Cyclic Nucleotide Binding

GAF Domains from Phosphodiesterases: Structural and Mechanistic

Insights. Structure 2009, 17 (12), 1551-1557.

21. Francis, S.; Houslay, M.; Conti, M., Phosphodiesterase Inhibitors:

Factors That Influence Potency, Selectivity, and Action. In

Phosphodiesterases as Drug Targets, Francis, S. H.; Conti, M.; Houslay, M.

D., Eds. Springer Berlin Heidelberg: 2011; Vol. 204, pp 47-84.

22. Pandit, J.; Forman, M. D.; Fennell, K. F.; Dillman, K. S.; Menniti, F.

S., Mechanism for the allosteric regulation of phosphodiesterase 2A deduced

80

from the X-ray structure of a near full-length construct. Proc. Natl. Acad.

Sci. U.S.A. 2009, 106 (43), 18225-18230.

23. Lee, M. E.; Markowitz, J.; Lee, J.-O.; Lee, H., Crystal structure of

phosphodiesterase 4D and inhibitor complex. FEBS Lett. 2002, 530 (1–3),

53-58.

24. Huai, Q.; Wang, H.; Sun, Y.; Kim, H.-Y.; Liu, Y.; Ke, H., Three-

Dimensional Structures of PDE4D in Complex with Roliprams and

Implication on Inhibitor Selectivity. Structure 2003, 11 (7), 865-873.

25. Zhang, K. Y. J.; Card, G. L.; Suzuki, Y.; Artis, D. R.; Fong, D.;

Gillette, S.; Hsieh, D.; Neiman, J.; West, B. L.; Zhang, C.; Milburn, M. V.;

Kim, S.-H.; Schlessinger, J.; Bollag, G., A Glutamine Switch Mechanism

for Nucleotide Selectivity by Phosphodiesterases. Mol. Cell 2004, 15 (2),

279-286.

26. Jansen, C.; Wang, H.; Kooistra, A. J.; de Graaf, C.; Orrling, K. M.;

Tenor, H.; Seebeck, T.; Bailey, D.; de Esch, I. J. P.; Ke, H.; Leurs, R.,

Discovery of Novel Trypanosoma brucei Phosphodiesterase B1 Inhibitors

by Virtual Screening against the Unliganded TbrPDEB1 Crystal Structure.

J. Med. Chem. 2013, 56 (5), 2087-2096.

27. Wang, H.; Yan, Z.; Geng, J.; Kunz, S.; Seebeck, T.; Ke, H., Crystal

structure of the Leishmania major phosphodiesterase LmjPDEB1 and

insight into the design of the parasite-selective inhibitors. Mol. Microbiol.

2007, 66 (4), 1029-1038.

28. Humphrey, J. M.; Yang, E.; Ende, C. W. a.; Arnold, E. P.; Head, J.

L.; Jenkinson, S.; Lebel, L. A.; Liras, S.; Pandit, J.; Samas, B.; Vajdos, F.;

Simons, S. P.; Evdokimov, A.; Mansour, M.; Menniti, F. S., Small-molecule

phosphodiesterase probes: discovery of potent and selective CNS-penetrable

quinazoline inhibitors of PDE1. MedChemComm 2014, 5 (9), 1290-1296.

29. Nikpour, M.; Sadeghian, H.; Saberi, M. R.; Nick, R. S.; Seyedi, S.

M.; Hosseini, A.; Parsaee, H.; Bozorg, A. T. D., Design, synthesis and

biological evaluation of 6-(benzyloxy)-4-methylquinolin-2(1H)-one

derivatives as PDE3 inhibitors. Biorg. Med. Chem. 2010, 18 (2), 855-862.

30. Ochiai, K.; Takita, S.; Kojima, A.; Eiraku, T.; Iwase, K.; Kishi, T.;

Ohinata, A.; Yageta, Y.; Yasue, T.; Adams, D. R.; Kohno, Y.,

Phosphodiesterase inhibitors. Part 5: Hybrid PDE3/4 inhibitors as dual

bronchorelaxant/anti-inflammatory agents for inhaled administration.

Bioorg. Med. Chem. Lett. 2013, 23 (1), 375-381.

31. van Linden, O. P. J.; Kooistra, A. J.; Leurs, R.; de Esch, I. J. P.; de

Graaf, C., KLIFS: A Knowledge-Based Structural Database To Navigate

Kinase–Ligand Interaction Space. J. Med. Chem. 2014.

81

CH

AP

TE

R 2

32. Card, G. L.; England, B. P.; Suzuki, Y.; Fong, D.; Powell, B.; Lee,

B.; Luu, C.; Tabrizizad, M.; Gillette, S.; Ibrahim, P. N.; Artis, D. R.; Bollag,

G.; Milburn, M. V.; Kim, S.-H.; Schlessinger, J.; Zhang, K. Y. J., Structural

Basis for the Activity of Drugs that Inhibit Phosphodiesterases. Structure

2004, 12 (12), 2233-2247.

33. Manallack, D. T.; Hughes, R. A.; Thompson, P. E., The Next

Generation of Phosphodiesterase Inhibitors:  Structural Clues to Ligand and

Substrate Selectivity of Phosphodiesterases. J. Med. Chem. 2005, 48 (10),

3449-3462.

34. Ke, H.; Wang, H., Crystal structures of phosphodiesterases and

implications on substrate specificity and inhibitor selectivity. Curr Top Med

Chem. 2007, 7 (4), 391-403.

35. Marcou, G.; Rognan, D., Optimizing Fragment and Scaffold

Docking by Use of Molecular Interaction Fingerprints. J. Chem. Inf. Model.

2006, 47 (1), 195-207.

36. Edmondson, S. D.; Mastracchio, A.; He, J.; Chung, C. C.; Forrest,

M. J.; Hofsess, S.; MacIntyre, E.; Metzger, J.; O'Connor, N.; Patel, K.; Tong,

X.; Tota, M. R.; Van der Ploeg, L. H. T.; Varnerin, J. P.; Fisher, M. H.;

Wyvratt, M. J.; Weber, A. E.; Parmee, E. R., Benzyl vinylogous amide

substituted aryldihydropyridazinones and aryldimethylpyrazolones as

potent and selective PDE3B inhibitors. Bioorg. Med. Chem. Lett. 2003, 13

(22), 3983-3987.

37. Scapin, G.; Patel, S. B.; Chung, C.; Varnerin, J. P.; Edmondson, S.

D.; Mastracchio, A.; Parmee, E. R.; Singh, S. B.; Becker, J. W.; Van der

Ploeg, L. H. T.; Tota, M. R., Crystal Structure of Human Phosphodiesterase

3B:  Atomic Basis for Substrate and Inhibitor Specificity†. Biochemistry

(Mosc). 2004, 43 (20), 6091-6100.

38. Allerton, C. M. N.; Barber, C. G.; Beaumont, K. C.; Brown, D. G.;

Cole, S. M.; Ellis, D.; Lane, C. A. L.; Maw, G. N.; Mount, N. M.; Rawson,

D. J.; Robinson, C. M.; Street, S. D. A.; Summerhill, N. W., A Novel Series

of Potent and Selective PDE5 Inhibitors with Potential for High and Dose-

Independent Oral Bioavailability†. J. Med. Chem. 2006, 49 (12), 3581-3594.

39. Orrling, K. M.; Jansen, C.; Vu, X. L.; Balmer, V.; Bregy, P.;

Shanmugham, A.; England, P.; Cos, P.; Maes, L.; Adams, E.; van de

Bogaart, E.; Chatelain, E.; Ioset, J.-R.; Stolpe, A.; Zorg, S.; Veerman, J.;

Seebeck, T.; Sterk, G. J.; de Esch, I. J. P.; Leurs, R., Catechol Pyrazolinones

as Trypanocidals: Fragment-Based Design, Synthesis and Pharmacological

Evaluation of Nanomolar Inhibitors of Trypanosomal Phosphodiesterase

B1. J. Med. Chem. 2012.

82

40. Helal, C. J.; Kang, Z.; Hou, X.; Pandit, J.; Chappie, T. A.; Humphrey,

J. M.; Marr, E. S.; Fennell, K. F.; Chenard, L. K.; Fox, C.; Schmidt, C. J.;

Williams, R. D.; Chapin, D. S.; Siuciak, J.; Lebel, L.; Menniti, F.;

Cianfrogna, J.; Fonseca, K. R.; Nelson, F. R.; O’Connor, R.; MacDougall,

M.; McDowell, L.; Liras, S., Use of Structure-Based Design to Discover a

Potent, Selective, In Vivo Active Phosphodiesterase 10A Inhibitor Lead

Series for the Treatment of Schizophrenia. J. Med. Chem. 2011, 54 (13),

4536-4547.

41. Verhoest, P. R.; Chapin, D. S.; Corman, M.; Fonseca, K.; Harms, J.

F.; Hou, X.; Marr, E. S.; Menniti, F. S.; Nelson, F.; O’Connor, R.; Pandit,

J.; Proulx-LaFrance, C.; Schmidt, A. W.; Schmidt, C. J.; Suiciak, J. A.;

Liras, S., Discovery of a Novel Class of Phosphodiesterase 10A Inhibitors

and Identification of Clinical Candidate 2-[4-(1-Methyl-4-pyridin-4-yl-1H-

pyrazol-3-yl)-phenoxymethyl]-quinoline (PF-2545920) for the Treatment of

Schizophrenia†† Coordinates of the PDE10A crystal structures have been

deposited in the Protein Data Bank for compound 1 (3HQW), 2 (3HQY), 3

(3HQW) and 9 (3HR1). J. Med. Chem. 2009, 52 (16), 5188-5196.

42. Kranz, M.; Wall, M.; Evans, B.; Miah, A.; Ballantine, S.; Delves, C.;

Dombroski, B.; Gross, J.; Schneck, J.; Villa, J. P.; Neu, M.; Somers, D. O.,

Identification of PDE4B Over 4D subtype-selective inhibitors revealing an

unprecedented binding mode. Biorg. Med. Chem. 2009, 17 (14), 5336-5341.

43. Xu, Z.; Liu, Z.; Chen, T.; Chen, T.; Wang, Z.; Tian, G.; Shi, J.; Wang,

X.; Lu, Y.; Yan, X.; Wang, G.; Jiang, H.; Chen, K.; Wang, S.; Xu, Y.; Shen,

J.; Zhu, W., Utilization of Halogen Bond in Lead Optimization: A Case

Study of Rational Design of Potent Phosphodiesterase Type 5 (PDE5)

Inhibitors. J. Med. Chem. 2011, 54 (15), 5607-5611.

44. Chappie, T. A.; Humphrey, J. M.; Allen, M. P.; Estep, K. G.; Fox, C.

B.; Lebel, L. A.; Liras, S.; Marr, E. S.; Menniti, F. S.; Pandit, J.; Schmidt,

C. J.; Tu, M.; Williams, R. D.; Yang, F. V., Discovery of a Series of 6,7-

Dimethoxy-4-pyrrolidylquinazoline PDE10A Inhibitors†. J. Med. Chem.

2006, 50 (2), 182-185.

45. Claffey, M. M.; Helal, C. J.; Verhoest, P. R.; Kang, Z.; Fors, K. S.;

Jung, S.; Zhong, J.; Bundesmann, M. W.; Hou, X.; Lui, S.; Kleiman, R. J.;

Vanase-Frawley, M.; Schmidt, A. W.; Menniti, F.; Schmidt, C. J.; Hoffman,

W. E.; Hajos, M.; McDowell, L.; O’Connor, R. E.; MacDougall-Murphy,

M.; Fonseca, K. R.; Becker, S. L.; Nelson, F. R.; Liras, S., Application of

Structure-Based Drug Design and Parallel Chemistry to Identify Selective,

Brain Penetrant, In Vivo Active Phosphodiesterase 9A Inhibitors. J. Med.

Chem. 2012, 55 (21), 9055-9068.

83

CH

AP

TE

R 2

46. Ke, H.; Wang, H.; Ye, M., Structural Insight into the Substrate

Specificity of Phosphodiesterases. In Phosphodiesterases as Drug Targets,

Francis, S. H.; Conti, M.; Houslay, M. D., Eds. Springer Berlin Heidelberg:

2011; Vol. 204, pp 121-134.

47. Freund, Y. R.; Akama, T.; Alley, M. R. K.; Antunes, J.; Dong, C.;

Jarnagin, K.; Kimura, R.; Nieman, J. A.; Maples, K. R.; Plattner, J. J.; Rock,

F.; Sharma, R.; Singh, R.; Sanders, V.; Zhou, Y., Boron-based

phosphodiesterase inhibitors show novel binding of boron to PDE4 bimetal

center. FEBS Lett. 2012, 586 (19), 3410-3414.

48. Park, C. M.; Jeon, D. J., Stereoselective synthesis of novel pyrazole

derivatives using tert-butansulfonamide as a chiral auxiliary. Organic &

Biomolecular Chemistry 2012, 10 (13), 2613-2620.

49. Card, G. L.; Blasdel, L.; England, B. P.; Zhang, C.; Suzuki, Y.;

Gillette, S.; Fong, D.; Ibrahim, P. N.; Artis, D. R.; Bollag, G.; Milburn, M.

V.; Kim, S.-H.; Schlessinger, J.; Zhang, K. Y. J., A family of

phosphodiesterase inhibitors discovered by cocrystallography and scaffold-

based drug design. Nat Biotech 2005, 23 (2), 201-207.

50. Recht, M. I.; Sridhar, V.; Badger, J.; Bounaud, P.-Y.; Logan, C.;

Chie-Leon, B.; Nienaber, V.; Torres, F. E., Identification and Optimization

of PDE10A Inhibitors Using Fragment-Based Screening by

Nanocalorimetry and X-ray Crystallography. J. Biomol. Screen. 2013.

51. Gurney, M.; Burgin, A.; Magnusson, O.; Stewart, L., Small Molecule

Allosteric Modulators of Phosphodiesterase 4. In Phosphodiesterases as

Drug Targets, Francis, S. H.; Conti, M.; Houslay, M. D., Eds. Springer

Berlin Heidelberg: 2011; Vol. 204, pp 167-192.