The Influence of Bioisosteres in Drug Design: Tactical Applications to Address Developability Problems

Abstract

The application of bioisosteres in drug discovery is a well-established design concept that has demonstrated utility as an approach to solving a range of problems that affect candidate optimization, progression, and durability. In this chapter, the application of isosteric substitution is explored in a fashion that focuses on the development of practical solutions to problems that are encountered in typical optimization campaigns. The role of bioisosteres to affect intrinsic potency and selectivity, influence conformation, solve problems associated with drug developability, including P-glycoprotein recognition, modulating basicity, solubility, and lipophilicity, and to address issues associated with metabolism and toxicity is used as the underlying theme to capture a spectrum of creative applications of structural emulation in the design of drug candidates.

An erratum to this chapter can be found at http://dx.doi.org/10.1007/7355_2014_74

1 Introduction

Bioisosterism is a powerful concept that has found widespread application in drug design and continues to be an important tactical element in contemporary medicinal chemistry practices [1–10]. Bioisosterism, which evolved from the concept of shape isosterism exhibited by simple structural elements, was originally recognized by Langmuir almost a century ago, and the first experiments that demonstrated the potential of bioisosterism in a practical setting were reported in 1932 and 1933 [1, 11–13]. In a series of experiments, Erlenmeyer and his colleagues showed that antibodies recognizing ortho-substituted tyrosine moieties in synthetic antigens, derived by reaction with diazonium ions, were not able to distinguish between a phenyl and thiophene ring or O, NH, and CH₂ in a linker element [12, 13]. The description of this phenomenon as bioisosterism is attributed to Harris Friedman who introduced the term in 1951 to define compounds demonstrating similar biological activities [14]. However, Friedman recognized that bioisosterism and isosterism were distinct concepts, an observation that anticipated contemporary drug design principles in which the utility of bioisosteres is frequently dependent on context and relies upon a less than exact structural or physicochemical mimicry for the manifestation of biological effect. The thoughtful deployment of a bioisostere offers potential value in drug design campaigns by providing an opportunity to probe the effect of steric size and shape, the modulation of dipole and electronic properties, lipophilicity and polarity, or pK _a on a biological response, which may be functional mimicry or antagonism of a biological regulator [1, 2, 10]. In addi tion to affecting potency and function, isosteres have demonstrated utility in addressing problems associated with pharmacokinetic and pharmaceutical properties, specificity, toxicity, and metabolic activation pathways in vivo in addition to being a source of novel intellectual property [1–10]. This chapter will summarize some of the more interesting tactical applications of bioisosteres in drug design where problems of the type commonly encountered by medicinal chemists have inspired innovative solutions that offer useful instruction. The vignettes are organized based on a specific property or problem under examination rather than the perhaps more traditional approach of cataloguing based on functional group mimicry.

2 Bioisosteres Designed to Enhance Drug Potency and Selectivity

2.1 Fluorine as an Isostere of Hydrogen

Fluorine has emerged as a remarkable element in drug design, and its applications continue to provide interesting effects on biological activity that can often be somewhat surprising, leading to its broad deployment by the medicinal chemistry community which, in turn, has promoted a deeper understanding of its properties [15–24]. The strategic introduction of fluorine to replace a hydrogen atom can markedly influence potency [25–30]. For example, the l-nucleoside analogue emtricitabine (2, FTC) is the 5-fluoro analogue of lamivudine (1, 3TC) and is consistently the more potent HIV-1 inhibitor in cell culture by four- to tenfold, which is reflected in the inhibition of HIV-1 reverse transcriptase by the respective triphosphate derivatives [25–27]. The F atom ortho to the piperazine and morpholine in the antibacterial agents, the quinolone-based DNA gyrase inhibitor 3 and the protein synthesis inhibitor linezolid 4, respectively, is of importance to both potency and efficacy in vivo while in indole-based HIV-1 attachment inhibitors, the 4-F substituent installed in 6 enhanced the potency of prototype 5 by more than 50-fold [15, 16, 28–30].

In a series of γ-secretase inhibitors that reduced Aβ protein production in a cell-based assay, the modest potency associated with the parent molecule 7 (evaluated in racemic form) was markedly improved by complete fluorination at C-5 although in this particular example gem-dimethyl substitution at C-5 also improved biological activity [31]. The introduction of a F atom at the para-position of the benzylated imide 10 enhanced thrombin inhibitory activity by over fivefold along with selectivity over the related serine protease trypsin by fourfold with 11 exhibiting a 67-fold preference for the coagulation cascade enzyme [17, 32]. An X-ray crystallographic structure of 11 bound to thrombin provided an understanding of the structure–activity relationship (SAR) observation with the F substituent proximal (2.4 Å) to the backbone C_α-H atom of Asn₉₈ and aligned to establish an interaction with the C atom of the backbone amide C=O of Asn₉₈ at a distance of 3.5 Å and an angle of 96° to the plane of the C=O (Fig. 1) [17, 18, 32]. The F to C=O carbon interaction is the one with broader occurrence, and although this interaction is considered to be energetically modest, it appears to be sufficient to differentiate the potency between 10 and 11 by playing a supporting role to the primary interactions between the enzyme and the inhibitor [33]. In an analogous matched pair study, the thrombin inhibitory activity of the amidine 12 is embellished sixfold by F substitution to afford 13, with an X-ray crystallographic co-crystal structure suggestive of a H-bonding or electrostatic interaction between the F atom and backbone NH of Gly₂₁₆ which are 3.47 Å apart, close to the upper distance limit for a H-bond [16]. The importance of this interaction was underscored from a comparison with the co-crystal of 12 and thrombin in which the aromatic ring adopted a significantly different conformation to that of 13 [16].

Fig. 1

Key interactions of the 4-F substituent of 11 with Asn₉₈ of thrombin

2.2 Carboxylic Acid Isosteres to Improve Potency

Carboxylic acid isosteres are a familiar part of the medicinal chemistry landscape and have been deployed to address a number of challenges in drug design, including positively influencing potency. A particularly compelling example that illustrates the importance of carefully selecting the optimal isostere is provided by the SARs that subtended the discovery of the potent angiotensin II receptor antagonist losartan (15) and its analogues 14 and 16–20 [34–38]. A key insight was that the tetrazole moiety in 15 enhances potency over the carboxylic acid analogue EXP-7711 (14) by tenfold, attributed to the topological geometry in which the 2H isomer depicted projects the acidic NH or negative charge 1.5 Å further from the aryl ring than does the carboxylic acid moiety [34, 39, 40]. Interestingly, as summarized by the comparison presented in Table 1, the position of the carboxylic acid moiety on the phenyl ring exerts minimal effect on potency, with 14 and 18 similarly active, while the ortho-tetrazole isomer 19 is over 50-fold more potent than the meta-substituted analogue 20 [38]. The carboxylic acid and tetrazole moieties are considered to bind in the same pocket in the angiotensin II receptor and interact with Lys₁₉₉ [41]. However, this interaction does not appear to depend on the formation of a classic salt bridge since mutation of Lys₁₉₉ to Gln exerted only a modest twofold effect on the binding of losartan (15) and a fivefold effect on the binding of EXP-7711 (14), data that are more consistent with a H-bonding interaction between drug and protein [41]. The poor activity of 20 may be a consequence of increased steric bulk compared to the carboxylic acid analogue 18.

Table 1 Structure–activity relationships associated with the topology of the acidic moiety in angiotensin II receptor antagonists

The importance of topology and the associated geometric implications are also apparent in acylsulfonamide-based angiotensin II antagonists where this functionality functions suitably as an acid surrogate. The CONHSO₂Ph moiety of 16 is a 20-fold less potent ligand than the topologically reversed SO₂NHCOPh isomer 17, an observation that is consistent with the longer C–S bond in 17 more effectively mimicking the topology of the tetrazole 15 by projecting the charge further from the biphenyl core than 16, which more closely resembles the carboxylic acid 14 [34, 38, 42].

The discovery of the macrocyclic hepatitis C virus (HCV) NS3 protease inhibitor BILN-2061 (21) represented a milestone in HCV drug design by providing proof-of-concept for this mechanistic approach to the control of viremia in infected subjects. The disclosure of this series of tripeptide-based inhibitor stimulated considerable interest in further structural optimization, an endeavor that became of particular importance in the context of overcoming the cardiotoxicity that resulted in termination of this pioneering molecule [43–45]. The conversion of the carboxyl terminus to an acylsulfonamide, particularly the cyclopropyl acylsulfonamide moiety, afforded compounds with increased potency as a consequence of the cyclopropyl ring complementing the P1′ pocket and the sulfone oxygen atoms engaging the protein via H-bonding interactions with the catalytic elements of the enzyme. The significant advantage conferred by this structural modification can be readily assessed by comparing the potency of the matched pairs of acyclic inhibitors 22 and 23 and the P4-P2* macrocycles 24 and 25 [46–49].

The optimization of inhibitors of the anti-apoptotic B-cell lymphoma 2 proteins Bcl-2, Bcl-x_L, and Bcl-w, for which the fragment leads biphenyl carboxylic acid 26 and phenol 27 were identified using a nuclear magnetic resonance (NMR) screen, involved tethering of the two molecules which bound weakly to proximal pockets in the Bcl-x_L protein [50–52]. Initial attempts to link acid 26 to phenol 27 focused on substituting at the C atom ortho to the carboxylic acid moiety of 26, a topology that was probed based on an analysis of NMR data [50, 51]. However, this approach was not initially successful in identifying hybrid molecules with increased affinity, attributed to poor geometric vectors that gave suboptimal binding. An acylsulfonamide was conceived to offer an improved vector while preserving the acidic moiety, of importance as a complement to Arg₁₃₉ of the Bcl protein, and a library of 120 molecules were prepared after establishing that the simple methyl acylsulfonamide exhibited comparable affinity for Bcl as 26 [51]. The acylsulfonamide 28 emerged from that exercise with further optimization to navitoclax (ABT-263, 30) proceeding via the intermediacy of compound 29 [50–52].

2.3 Applications of Heterocycles as Isosteres and Isosterism Between Heterocycles

Heterocycles find widespread application in drug design as important scaffolds for deploying functionality or as critical pharmacophoric elements that interact intimately with target proteins, a cue taken from Mother Nature who takes significant advantage of heterocycles. The immense utility of heterocycles resides in their typically facile synthetic accessibility combined with the versatility to project a range of vectors while the electronic properties of the ring can readily be modulated by the introduction of additional heteroatoms and substituent selection. Aspects of heterocycles that are of importance to the medicinal chemist revolve around their ability to act as H-bond donors or acceptors, to influence the properties of substituents by withdrawing or donating charge and to engage in π–π and dipole–dipole interactions. Thus, the careful selection of a heterocycle for a specific application is frequently of considerable importance and there are many examples where the effects of substitution by rings offering a similar silhouette can be quite subtle in nature and not always well understood.

The H-bonding potential of a wide range of functionality, designated pK _BHX, has been determined based on the association of 4-fluorophenol with an acceptor in CCl₄ monitored at room temperature by the shift in the infrared stretching frequency of the OH bond [53–55]. On this scale, a stronger H-bond acceptor is associated with a higher pK _BHX value that shows dependence on several physical properties, including the position of the acceptor atom in the periodic table, polarizability, field and inductive/resonance effects, electronic resonance, and steric effects associated with substituents and their interaction with the acceptor atom. A particularly important aspect of H-bonding potential is that across a range of functionality it is not quantitatively related to basicity (pK _BH ⁺ or pKa) and there are several examples where a H-bonding interaction occurs preferentially at an atom of low basicity in the presence of a more basic acceptor [56–62]. For example, nicotine (31) protonates initially on the pyrrolidine nitrogen atom in H₂O, but H-bonding with 4-fluorophenol in CCl₄ occurs primarily at the pyridine nitrogen atom. For cotinine (32), the amide carbonyl is the primary site for H-bonding, which contrasts with basicity since the amide exhibits a pK _BH ⁺ value of −0.71 compared to 5.20 for the pyridine nitrogen atom where protonation occurs [56–62].

Table 2 presents a comparison of pK _HBX and pK _BH ⁺ (pKa) values for some common 5- and 6-membered ring heterocycles where within a homologous series there is some correlation between H-bonding potential (pK _BH ⁺ ₎ and pK _a values although isoxazole and pyridazine are notable exceptions. In the case of pyridazine, a heterocycle that has been proposed to be a privileged scaffold in medicinal chemistry, the high H-bonding potential for this poorly basic molecule is attributed to the α-effect in which unfavorable lone pair-lone pair interactions are relieved by one of the N atoms engaging in a H-bonding interaction, as depicted in Fig. 2 [63, 64].

Table 2 Comparison of pK _HBX and pK _BH ⁺ (pK _a) values for common 5- and 6-membered ring heterocycles

Fig. 2

H-bonding associated with the pyridazine heterocycle

The topological deployment of an oxazole heterocycle influenced the potency of blood platelet aggregation inhibition associated with the non-prostanoid prostacyclin (PGI₂) mimetics 33 and 34 [65–68]. The oxazole 33 is fivefold more potent than the isomer 34 which, in turn, is comparable to the cis-olefin 35 [65–68]. These data were interpreted in the context of the nitrogen atom of the central oxazole of 33 being ideally positioned to accept a H-bond from a donor within the PGI₂ receptor protein, consistent with earlier SAR observations. This interaction is not available to the isomer 34 since the O atom of an oxazole ring is a poorer H-bond acceptor or the olefin 35, suggesting that the central oxazole ring of 34 is acting merely as a scaffold that is structurally analogous to 35 [65–68]. In the solid-state structures of 33 and 34, although the molecules adopt an almost identical shape, the central oxazole rings accept H-bonds from the CO₂H moiety of an adjacent molecule in the unit cell, leading to a markedly different pattern of crystal packing between the two isomers [68]. Ab initio calculations indicate that a H-bond to the N atom of an oxazole is 10.4 kJ/mol or 2.48 kcal/mol more stable than that to the O atom (Fig. 3), reflected in the essentially exclusive occurrence of H-bonds only to the N atom of oxazoles observed in the Cambridge Structural Database (CSD) [68–70].

Fig. 3

Energetics of H-bonding in oxazoles based on ab initio calculations

In a series of quinolone-based inhibitors of HCV NS5B polymerase, the 1,2.4-oxadiazole 36 exhibited potent enzyme inhibition, IC₅₀ = 41 nM, in part by establishing a H-bond with the backbone NH of Ser₄₇₆, as determined by X-ray crystallographic analysis [71]. The topological isomer 37 is an order of magnitude weaker, consistent with the weaker potential for the O atom to engage in H-bonding despite projecting the benzyl moiety in a similar vector to 36. In order for 37 to present the sterically somewhat hindered N-4 to the Ser₄₇₆ NH, the benzyl moiety would be required to adopt an alternate and unfavorable vector. Interestingly, the several other azole heterocycles 38–40 examined were relatively poor inhibitors of the enzyme, with 38 and 39 perhaps the most surprising given that they combine a strong H-bond acceptor with the preferred topology of the benzyl moiety that is found in 36 [71, 72].

The metal binding properties of azole rings were explored in the context of inhibitors of HIV-1 integrase and conducted with a chemotype in which the binding topology of the fluorobenzyl substituent controlled the presentation of the heteroatom to one of the catalytic magnesium atoms [73–77]. As depicted in Fig. 4, the positioning of the fluorobenzyl moieties in 41 and 42 dominates the binding conformation thereby dictating which azole heteroatom is presented to the active site metal atom, providing a sensitive assessment of metal binding potential, with SARs summarized in Table 3 [73–77]. The importance of Mg²⁺ coordination potential and its impact on potency is most evident in the comparison between the two 1,2,4-oxadiazoles 47 and 48 in which the O atom of 48 is an ineffective metal coordinator, an observation that is concordant with heteroatom H-bonding potential and was recapitulated in a series of 1,6-naphthyridine derivatives [73–75]. Interestingly, in the pyrido[1,2- a]pyrimidine series, the thiazole 44 was selected as the basis for further optimization while in the 1,6-naphthyridine chemotype, the 1,3,4-oxadiazole ring that performs only modestly in this series (refer to compound 46) was the most effective amide surrogate studied, presumably reflecting subtle differences in binding between the two structural classes [75–77].

Fig. 4

Interactions between representative azole-substituted pyrido[1,2- a]pyrimidines and −1,6-naphthyridines and the catalytic Mg²⁺ atoms of HIV-1 integrase

Table 3 Structure–activity relationships for a series of azole-substituted pyrido[1,2- a]pyrimidine-based inhibitors of HIV-1 integrase

The electron-withdrawing properties of heterocyclic rings have been exploited as an important element in drug design with two aspects prominent: the acidifying effect of a heterocycle on a NH substituent, which enhances H-bonding donor properties and confers carboxylic acid mimicry in the case of sulfonamide antibacterial agents, and, secondly, the electron-withdrawing properties of a heterocycle ring that has been used to advantage in the design of mechanism-based inhibitors of serine-type proteases and hydrolases [78–81]. The site of attachment of substituents on a heterocyclic ring has been shown to be of importance to potency in all of these settings, interpretable based on the quantification of electron withdrawal. A scale of electron withdrawal has been formulated as the concept of charge demand, defined as the fraction of π-charge transferred from a negatively charged trigonal carbon atom to the adjacent X group, typically measured by ¹³C-NMR chemical shifts of trigonal benzylic carbanions, as depicted in Fig. 5 [82–85]. The resonance electron-withdrawing capacity of common functional groups and a selection of heterocycles is summarized in Table 4, data that reveals interesting trends and emphasizes the importance of carefully selecting specific heterocycles for a particular application.

Fig. 5

Charge demand is defined as the fraction of π-charge transferred from a negatively charged trigonal carbon atom to the adjacent X group

Table 4 Charge demand associated with functional groups and heterocycles

An important aspect associated with the design of serine protease inhibitors has focused on substrate mimetics that present the enzyme with an electrophilic carbonyl element. This acts as a decoy for the scissile amide bond that reacts with the catalytic serine hydroxyl to form a stable but unproductive tetrahedral intermediate [79]. The equilibrium in favor of the adduct is a function of the electrophilicity of the C=O moiety, with peptidic aldehydes an early vehicle that established the viability of this concept but which was subsequently refined by deploying trifluoromethyl ketones and α-ketoamides. Trifluoromethyl ketones provide an additional example of the beneficial effect on potency of replacing H with F since the corresponding methyl ketones are typically inactive [86]. α-Ketoamides are particularly interesting in inhibitor optimization because they provide an opportunity for the incorporation of structural elements designed to interact more extensively with the S′ pockets of an enzyme [87]. This design concept was further explored in the context of human neutrophil elastase (HNE) inhibitors that focused on tripeptidic, mechanism-based inhibitors that incorporated heterocycles as the carbonyl-activating element [88–94]. α-Ketoamides were included in this study for purpose of comparison and inhibitory potency was found to correlate nicely with the electrophilicity of the carbonyl moiety, as captured in the SARs summarized in Table 5 [89]. In addition to providing the potential for unique interactions with enzyme and the ability to probe interactions with S′ sites, the heterocycle moiety was also considered to offer the potential to sterically interfere with reductive metabolism of the carbonyl moiety [88–90, 94].

Table 5 SAR associated with a series of tripeptidic inhibitors of HNE incorporating activated carbonyl moieties to interact with the catalytic serine

Studies with a matched pair of oxadiazole derivatives found that HNE inhibitory activity was sensitive to the topology of the heterocycle, a structure–activity relationship clearly rooted in the electron-withdrawing properties of the heterocycle [91, 92]. The 1,3,4-oxadiazole 51 is a potent inhibitor of HNE with a K _i = 0.025 nM, but the isomeric 1,2,4-oxadiazole 52 is 20-fold weaker, with a K _i = 0.49 nM [90]. In these inhibitors, the electron-withdrawing effect of the 1,3,4-oxadiazole moiety is similar to an oxazole, whereas the 1,2,4-oxadiazole ring performs more like an imidazole (Table 4). Based on this result, the 1,3,4-oxadiazole moiety was adopted as the carbonyl-activating moiety in more advanced SAR studies that ultimately led to the identification of ONO-6818 (53) as an orally bioavailable, non-peptidic inhibitor of HNE that was advanced into clinical trials [92].

The basic design principles were confirmed by an X-ray crystallographic analysis of a benzoxazole-based HNE inhibitor bound to the enzyme which revealed that the catalytic serine hydroxyl had reacted with the activated carbonyl moiety as anticipated to afford a tetrahedral intermediate [88]. In addition, the benzoxazole nitrogen atom engaged the NH of the imidazole of His₅₇, the residue that is part of the catalytic triad, in a H-bonding interaction, as summarized in Fig. 6 [88].

Fig. 6

Key interactions between a tripeptidic ketobenzoxazole derivative and HNE

2-Keto-oxazole derivatives have featured prominently in the design of inhibitors of fatty acid amide hydrolase (FAAH), a serine hydrolase responsible for degrading endogenous lipid amides, including anandamide and related fatty acid amides that have been identified as neuronal modulators [80, 81, 95–103]. The electronic properties of the oxazole heterocycle were influenced by electron-donating and electron-withdrawing substituents introduced at C-5 of the ring that indirectly influenced the electrophilicity of the carbonyl moiety designed to react with the enzyme serine hydroxyl; a survey that provided the structure–potency relationships is compiled in Table 6. FAAH inhibitory activity was found to correlate with the Hammett σ constant associated with the substituent, with the plot of the data allowing the conclusion that the 5-CO₂H derivative bound as the anion while the 5-CHO and 5-CO.CF₃ analogues bound as their hydrated forms [82, 94–96, 104].

Table 6 Structure–activity relationships associated with a series of ketooxazole-based inhibitors of FAAH

A particularly striking example of the importance of correctly deploying a heterocycle based on its electronic and H-bonding properties that relates closely to biological potency is illustrated by the series of 3 non-peptidic heteroaryl nitriles compiled in Table 7 that have been explored as inhibitors of the cysteine proteases cathepsins K and S [105]. These compounds depend on the electron-withdrawing properties of the heterocycle to activate the nitrile moiety toward addition of the catalytic cysteine thiol of the enzyme to form a stable, covalent but reversible imino thioether-based complex. The 2-cyanopyrimidine 54 is a potent and effective inhibitor of both cathepsins, but the high charge demand at this site (0.43, Table 4) is such that the nitrile moiety is indiscriminately reactive and forms adducts with thiols of microsomal proteins. The isomeric 2-cyanopyridines 55 and 56 are of lower intrinsic electrophilicity based on the reduced charge demand (0.41, Table 4) which would be expected to translate into weaker enzyme inhibition. However, 55 and 56 exhibited markedly different protease inhibition profiles that provided unique insights into drug–target interactions. Pyridine 55 is 44- and 127-fold less potent toward cathepsins S and K, respectively, while the isomer 56 is inactive toward both cysteine proteases [105]. This observation was attributed to the concept that the pyrimidine N-3 is involved in cysteine thiol activation by acting as a general base to engage the thiol H atom, thereby catalyzing addition to the nitrile moiety, as depicted in Fig. 7. Replacing the pyrimidine N-3 atom with a C–H, as in pyridine 56, not only removes this function but also introduces a negative steric interaction between the ring CH and the cysteine SH [105].

Table 7 Inhibitors of cathepsins S and K

Fig. 7

Role of the pyrimidine N3 or pyridine N1 atom as a general base catalyst to facilitate addition of the catalytic cysteine thiol to the activated nitrile in cathepsin S and K inhibitors

In a study inspired by observations of the marked potency differences between the two oxazole-based factor Xa inhibitors 57 and 58, which represent a matched pair differing only by the topology of the oxazole rings, dipole–dipole interactions between the heterocycle ring and an amide moiety of the protein were analyzed [106, 107]. The oxazoles 57 and 58 differed in potency by over tenfold while the isoxazole 59 was the most potent with a K _i = 9 nM. X-ray co-crystal data indicated that these inhibitors were not engaging the enzyme via a H-bond donor–acceptor interaction, but the heterocycle rings were noted to be close to the amide bond between Cys₁₉₁ and Gln₁₉₂ with the planes of each almost parallel, leading to the suggestion of a dipole–dipole interaction, as illustrated in Fig. 8. Indeed, potency was shown to correlate with a favorable dipole–dipole interaction in the more active compound 57 that is mismatched in the less active isomer 58 (Fig. 9). The experimental dipole moment for oxazole is 1.7 D and for isoxazole is 3.0 D and gas phase calculations indicate that under optimal conditions the interaction energies between these heterocycles and N-methylacetamide, which has a dipole moment of 3.7 D, are −2.74 kcal/mol for oxazole and −2.95 kcal/mol for isoxazole. Although the energies in an aqueous environment are expected to be lower based on solvation issues, these differences are sufficient to explain the differences in potency observed with the azoles 57–59. A plot of the calculated energies versus dipole moment for a series of heterocycles interacting with N-methylacetamide exhibited a good correlation, R² = 0.84, while for benzene, pyrazine, and triazine, heterocyclic rings without a dipole moment, the interacting energies increase as the ring becomes more electron deficient. These studies suggested that potency can be optimized by the careful deployment of heterocycle rings in a fashion that aligns dipole moments in an antiparallel orientation compared to proximal amide moieties in a target protein. This effect can be enhanced by decreasing the π-electron density of the heteroarene in order to improve heteroarene-amide π interactions.

Fig. 8

Binding interactions of factor Xa inhibitors with structural elements in the active site

Fig. 9

Dipole–dipole interactions between oxazole-based factor Xa inhibitors and the enzyme that are proposed to subtend the observed potency differences

2.4 Isosteres of Drug and Water Molecules

Displacing a water molecule that mediates protein–ligand binding by introducing an appropriate structural element into the ligand to produce an isostere of the ligand–water complex can provide significant enhancements in potency by establishing new H-bonding interactions directly between the protein and ligand [108]. The increased potency arises from both increasing the enthalpic contribution to the thermodynamic signature of the association and taking advantage of the entropic energy gain that arises from releasing the bound water molecule into bulk solvent [108]. Although this can be a challenging enterprise since success depends on several factors, including the specific topology of the ligand-water-protein complex and the ability to effectively mimic the key interactions, several examples have been described where this approach has been used to considerable advantage [109–119]. The design of a series of cyclic urea-based inhibitors of human immunodeficiency virus-1 (HIV-1) protease provides a particularly compelling pioneering example of the concept that in this case is based on displacing the water molecule that mediates the interaction between the NHs of the 2 flap residues Ile₅₀/Ile_50’ and the P2/P2′ carbonyl O atoms of linear peptidomimetic inhibitors represented by 60, as depicted in Fig. 10 [109–111]. A series of cyclic ureas exemplified by 61 were conceived based on the premise that the urea carbonyl moiety would displace the water molecule and establish H-bonds directly with the flap residues. As an additional benefit, the conformational constraint provided by the cyclic template would preorganize the inhibitor for recognition by the protease, providing an additional entropic advantage [109–111]. The urea 62 was identified from the initial phase of these studies as a potent HIV-1 protease inhibitor, K _i = 0.27 nM, that demonstrated antiviral activity in cell culture, EC₉₀ = 57 nM [109].

Fig. 10

Cyclic HIV-1 protease inhibitors designed based on the premise of replacing a bound water molecule

Poly(ADP-ribose) polymerase-1 (PARP-1) is a DNA repair enzyme and inhibitors are potentially useful when combined with specific anticancer therapeutic agents by prolonging their antitumor activity [113–115]. Co-crystallization of the potent human PARP-1 inhibitor 5-phenyl-2,3,4,6-tetrahydro-1H-azepino[5,4,3-cd]indol-1-one, K _i = 6 nM, with chicken PARP revealed the presence of a water molecule, designated H₂O₅₂, that interfaced between the indole NH and Glu₉₈₈, as depicted in Fig. 11a [113]. A family of 3,4-dihydro-[1,4]diazepino[6,7,1-hi]indol-1(2H)-ones 63–67 (Table 8) was designed that relied upon the concept of topological inversion of the indole heterocycle to an isostere that allowed functionalization at C-3 with substituents capable of displacing H₂O₅₂ and directly engaging Glu₉₈₈ [114]. The (E)-carboxaldehyde oxime 65 depicted in Fig. 11b satisfied the design criteria and is characterized as a potent human PARP-1 inhibitor, K _i = 9.4 nM, that is over tenfold more potent than the prototype 63. Indole 65 is severalfold more potent than the hydroxymethyl derivative 64 and the key structure-activity observations with the O-methylated derivatives 66 and 67 are consistent with the fundamental design hypothesis that the oxime moiety displaced H₂O₅₂. This was definitively confirmed by solving the co-crystal structure of oxime 65 with the enzyme which revealed the oxime OH engaging Glu₉₈₈ as proposed, as captured in Fig. 11b. Interestingly, the potency-enhancing effect of introducing the oxime moiety was muted in a series of compounds that incorporated a C-6 aryl substituent [113–115].

Fig. 11

Depiction of the key intermolecular interactions between chicken poly(ADP-ribose) polymerase-1 and indole-based inhibitors

Table 8 Inhibitor of human poly(ADP-ribose) polymerase-1 by 3,4-dihydro-[1,4]diazepino[6,7,1-hi]indol-1(2H)-ones

The enzyme scytalone dehydratase catalyzes two steps in melanin biosynthesis in Magnaporthe grisea, a plant fungal pathogen [116]. The benzotriazine 68 is a potent scytalone dehydratase inhibitor and modeling of the analogue 72 in the active site of the enzyme (Fig. 12a) recognized its isosteric relationship with the salicylamide 71 which had been co-crystallized with the enzyme. A water molecule that played a critical role in mediating the interaction between the inhibitor and two enzyme residues, Tyr₃₀ and Tyr₅₀, by H-bonding to the benzotriazine moiety of 68 and the amide carbonyl of 72 was viewed as an opportunity for optimization of the scaffold. The removal of the key benzotriazine nitrogen atom that accepted a H-bond from the water molecule afforded the much weaker inhibitor cinnoline 69, but the introduction of a nitrile moiety at the 3-position of the heterocycle provided a molecule, 70, that is over 18,000-fold more potent than the progenitor [116]. The remarkable increase in potency for the addition of just two atoms represents a highly ligand-efficient modification that was also observed with an analogous quinazoline/quinoline matched pair [116, 121]. An X-ray co-crystal structure of 70 bound to the enzyme confirmed the modeling hypothesis and validated the design principle, as summarized in Fig. 12b [116].

Fig. 12

Binding interactions between scytalone dehydratase and inhibitors

A similar strategy was adopted to enhance the potency of the triazine-based p38α MAP kinase inhibitor 73, K _i = 3.7 nM, in which the key N atom of the triazine heterocycle was replaced by C-CN in order to displace a bound water molecule observed in an X-ray co-crystal structure [117]. This rationalization led to the design of the cyanopyrimidine 74 which, with a K _i = 0.057 nM, conferred a 60-fold enhancement of potency. X-ray crystallographic data obtained with a closely analogous compound demonstrated that the cyano moiety engaged the NH of Met₁₀₉ in a H-bonding interaction, displacing the water molecule interfacing Met₁₀₉ with the triazine N of 73, as hypothesized during the design exercise.

The displacement of a water molecule mediating the interaction between a series of heat shock protein 90 (Hsp90) inhibitors and the protein by the introduction of a strategically deployed cyano substituent contributed significantly to potency [119]. The pyrrolopyrimidine 75 was designed from the fragment screening lead 77 but was devoid of significant potency based on the results of a fluorescence polarization assay. This was attributed to the loss of a key H-bond between the imidazole nitrogen atom of 77 and a water molecule that interfaced with Asn₅₁ of Hsp90 and another water molecule in the active site. The introduction of the 3-cyano moiety to 75 gave the significantly more potent Hsp90 inhibitor 76 with the nitrile nitrogen atom engaging Asn₅₁ directly, displacing the water molecule. Further optimization afforded the more potent 78 that engaged the Hsp90 protein principally via the interactions summarized in Fig. 13 [119].

Fig. 13

Principal drug–target interactions between pyrrolopyrimidine 78 and Hsp90

One of the difficulties associated with displacing bound water molecules is illustrated by observations with a series of human prostaglandin D2 synthase (hH-PGDS) inhibitors based on the prototype 79, IC₅₀ = 2.34 nM [120]. In the X-ray co-crystallographic structure of 79 with the enzyme, the isoquinoline nitrogen was observed to engage the hydrogen atom of a water molecule that interacted with Thr₁₅₉ and Leu₁₉₉ of hH-PGDS, leading to a study of the effect of incorporation of structural elements designed to displace the bridging water. The hydroxymethylated naphthalene derivative 80 and its amine analogue 81 were synthesized and confirmed by X-ray studies to perform as anticipated, displacing the water molecule and directly engaging Thr₁₅₉ and Leu₁₉₉ with the full complement of interactions. However, the potency of these compounds was significantly diminished compared to 79 with IC₅₀s of 1,480 nM for 80 and 845 nM for 81, representing 360- and 630-fold differences, respectively [120]. A potential explanation was suggested based on a closer analysis of the crystal structures which revealed that the topographical disposition of the hydroxyl and amino moieties of 80 and 81 was close to the plane of the naphthalene ring (21° and 27°, respectively), an energetically unfavorable arrangement that incurs allylic 1,3 strain [10, 122]. In addition, the dihedral angle between the phenyl and naphthalene rings was ~20° less than the 117° observed for 79, data that taken together suggested that the increased energy associated with the imposed conformational constraints outweighed the entropic gains associated with displacing the water molecule [120].

3 Bioisosteres to Modulate Conformation

3.1 Fluorine as a Hydrogen Isostere

The properties of fluorine make it an intriguing and useful isostere of a hydrogen atom, particularly in the context of alkanes where its unique properties have only recently begun to be examined in detail and more fully appreciated [16–21, 123–125]. Fluorine is the most electronegative of the atoms that forms covalent bonds with carbon and the polarization associated with the C–F bond affords a strong dipole moment, 1.85 D for CH₃F, that interacts with proximal functionality to influence conformational preferences in a fashion that can be exploited in drug design. These effects, the strength of which are dependent on the identity of the interacting substituent, are based on several underlying principles that include dipole–dipole interactions, attractive electrostatic effects, repulsion between fluorine and electronegative atoms, p orbital repulsion, and a hyperconjugative effect between an adjacent C–H bond and the low-lying C–F σ* orbital [21, 124–126]. Density functional theory (DFT) calculations have provided a relative ranking of the interaction energies between a fluorine atom and several elements and functional groups, with interactions with amine, alcohol, amide, and fluorine substituents of sufficient strength to be of practical utility in drug design exercises, data that are summarized in Table 9 [124–126].

Table 9 Energy differences between the gauche and anti conformers of α-substituted fluoroethanes based on DFT calculations

1,2-Difluoroalkanes prefer a gauche conformation between the two fluorine atoms that is favored by 0.8 kcal/mol based upon a hyperconjugative interaction between an adjacent C–H bond donating into the low-lying C–F σ* orbital (Fig. 14a) [126, 127]. For 1,3-difluoroalkanes, the conformational preference depicted in Fig. 14b is favored by dipole–dipole interactions between the C–F bonds that is estimated to amount to a difference of 3.3 kcal/mol [128]. However, while these interactions have been shown to influence the conformational preferences of structurally simple fluoroalkane derivatives, they have yet to be fully exploited in drug design [126–130].

Fig. 14

Conformational preferences of 1,2- and 1,3-difluoroalkanes and the underlying interactions

The conformation of an α-fluoro amide is influenced by the preference for the dipoles of the C–F and C=O moieties to align in an antiparallel fashion that in secondary amides can be reinforced by an electrostatic interaction between the electronegative F atom and amide NH [21, 131–136]. An interesting illustration of the utility of this effect is provided by the biological evaluation of the enantiomers of 83, the α-fluorinated derivative of the vanilloid receptor agonist capsaicin (82) [136]. The trans conformer in which the C–F and C=O dipoles are aligned in the same plane in an antiparallel fashion is calculated to be 6 kcal/mol more stable than the gauche conformer and 8 kcal/mol more stable than the cis conformer based on ab initio calculations, attributed to a favorable dipole alignment and an electrostatic interaction between the F and NH atoms, as depicted in Fig. 15 [131–136]. This conformational preference provided a potentially useful probe of the shape of the receptor based on the stereo-differentiation of the (R)- and (S)-enantiomers of 83 should the alkyl side chains project orthogonally to the plane of the amide moiety in the bound state, as depicted by the hashed bonds. However, the TRPVI receptor failed to discriminate between the two enantiomers of 83, leading to the conclusion that the bound conformation of 82 is that in which the alkyl side chain vector projects in a plane similar to that of the amide moiety which for (R)- and (S)- 83 would be disposed in the parallel fashion depicted [136].

Fig. 15

Conformational analysis of the enantiomers of α-fluoro capsaicin (83)

A fluorine atom deployed β- to the nitrogen atom of an amide also influences conformational preferences by favoring a gauche arrangement between these functionalities, a phenomenon established by X-ray crystallographic and NMR studies in the context of a series of β-peptide derivatives [126, 134, 135, 137–139]. For the difluoro succinamide esters 85 (threo) and 87 (erythreo) and the corresponding acids 84 (threo) and 86 (erythreo), solid-state structures revealed that all 4 molecules adopted a conformation that satisfied a vicinal F–F gauche interaction, a phenomenon that extended to solution based on an analysis ³ J _HF and ³ J _HH coupling constants in the NMR spectra [134, 135]. In the case of the threo isomers 84 and 85, this conformational preference places the amide carbonyl moieties in an anti-periplanar arrangement while for the erythreo isomers, the amide carbonyls are disposed gauche [134, 135].

For the GPR119 agonist 88, EC₅₀ = 868 nM with intrinsic activity of 111% in a cAMP assay, isosteric replacement of the pro-(S) H atom β- to the amine moiety resulted in 89, a compound with tenfold enhanced potency, EC₅₀ = 80 nM, and similar intrinsic agonistic properties, measured as 107% [140]. The F atom was introduced based on an appreciation of its potential to influence conformational preferences, a hypothesis confirmed by NMR analysis which revealed a population of the conformation in which the F and NH atoms are gauche amounting to 75% in CDCl₃ solution [140].

The gauche interaction between a fluorine atom and an amide influences the conformational preference of proline with the 4-(R)-F and 4-(S)-F derivatives 90 and 91, respectively, favoring complementary conformers [124]. The incorporation of (R)- and (S)-4-F-Pro into collagen fibrils markedly affects the properties of the polymer in a fashion that has provided insight into the stabilizing effects of the 4-(R)-hydroxyproline that occurs naturally in collagen [124, 141–146]. These studies have led to the suggestion that the stabilizing effect of 4-(R)-hydroxyproline on collagen, which is manifested as favoring a tightly wound helix structure, is the result of an inductive effect rather than a H-bonding interaction of the 4-OH substituent. The C^γ-endo conformation of proline is preferred by a modest 0.41 kcal/mol, leading to a 2:1 population of this conformer at room temperature while 4-(R)-hydroxyproline favors the C^γ-exo conformer by a calculated 0.48 kcal/mol (Fig. 16) [147]. For 4-(S)-F proline (91), the C^γ-endo conformation is favored by 0.61 kcal/mol, attributed to stabilization by the gauche effect between fluorine and the amide moiety which leads to the F atom adopting an axial disposition, an arrangement that is observed in the single-crystal X-ray structure (Fig. 16) [147, 148]. In contrast, 4-(R)-fluoroproline (92) prefers the C^γ-exo conformation by 0.85 kcal/mol, with the gauche effect between the F and amide moieties favoring an equatorial disposition of the F atom [147]. The C^γ-exo conformation mimics that preferred by 4-(R)-hydroxyproline and provides an explanation for the structural similarity of collagen fragments that incorporate either 4-(R)-hydroxy- or 4-(R)-fluoroproline [149].

Fig. 16

Conformational preferences for proline, 4-(S)-fluoroproline, and 4-(R)-fluoroproline

3-Fluoroproline also exhibits conformational bias as illustrated by studies with the diastereomeric esters 92 and 93 [150]. Both of these compounds crystallize with a trans configuration between the amide and ester moieties, as depicted in the structures of 92 and 93, but the proline rings adopt quite different conformations. In each case, the F atom is axially disposed, favored by a stabilizing gauche interaction with the amide moiety such that the pyrrolidine ring of the (R)-isomer 92 adopts the C^γ-exo conformation while, in contrast, the (S)-isomer 93 favors the C^γ-endo arrangement [150]. In solution, the C^γ-endo conformation of 93 dominated based on the ≤1 Hz coupling constant between the α- and β-protons in the ¹H-NMR spectrum, an observation supported by calculations that indicate that this conformation should be preferred by a ratio of 97:3. For the 3-(R)-isomer 92, unfavorable steric and electronic repulsive effects between the F atom and ester moiety led to a preference for the C^γ-exo conformation, populated to the extent of 69%. Notably, the conformational preferences of each of these 2 proline derivatives were expressed when they were incorporated into short peptide sequences [145, 150, 151].

β-Amino fluoroalkanes exhibit conformational bias based on gauche interactions between the amine and F moieties, the energetics of which depend on the protonation state [21, 124, 152]. In the unprotonated form, a gauche relationship between the NH₂ and F of 2-fluoroethylamine is weakly favored by an energy estimated to be ~1 kcal/mol, an interaction considered to be a bridging H-bond. However, protonation results in a much stronger preference for a gauche relationship between the F atom and the charged amine that is estimated to be 5.8 kcal/mol and attributed to a favorable electrostatic (charge–dipole) interaction between the charged amine and the electronegative F atom [21, 124, 152]. These energetic preferences are such that cyclic amines experience considerable conformational bias, exemplified most effectively by the analysis of the protonated form of 3-fluoro-N-methyl-piperidine (94) as summarized in Fig. 17 [153–155]. Despite experiencing steric compression, the ring F atom of 94 overwhelmingly prefers an axial disposition, with conformer A in Fig. 17 representing the global minimum and favored to the extent of 95–96%, stabilized by an electrostatic interaction between the F and NH⁺ moieties [155]. Conformer D contributes only 4–5% of the population at equilibrium, with a productive electrostatic effect compensating for unfavorable diaxial interactions between the F and CH₃ moieties.

Fig. 17

Conformational preference of N-methyl-3-fluoropiperidine (94) in its protonated form

The synthesis and evaluation of the two fluorinated enantiomers 96 and 97 of γ-aminobutyric acid (GABA, 95) provided some insight into the conformation of the neurotransmitter when bound to cognate receptors or that recognized by the aminotransferase (Fig. 18) [156, 157]. Fluorination of GABA (95) increases the acidity of the carboxylic acid moiety and decreases the basicity of the amine but preserves the zwitterionic nature of the molecule at neutral pH. NMR analysis indicated that all molecules adopted an extended conformation in solution but with 96 and 97 further preferring an arrangement in which a favorable gauche interaction occurs between the fluorine and the NH₃ ⁺ moiety [156]. The available conformations for each of the enantiomers 96 and 97 are captured in Fig. 18 with those that are disfavored based on the absence of a gauche effect noted [156]. Both 96 and 97 activated the cloned human GABA_A receptor with comparable potency, although both were significantly less potent than the natural neurotransmitter 95, results that are consistent with the bound form of the neurotransmitter being the extended conformation represented by B in Fig. 18 that is accessible to all three compounds [156, 157]. However, GABA aminotransferase was able to differentiate 96 and 97, catalyzing elimination of HF from the latter with 20-fold higher efficiency than for 96, an observation that suggested that the enzyme recognizes a conformation in which the NH₃ ⁺ and CH₂CO₂- moieties are disposed in a gauche arrangement, as represented by conformation C in Fig. 18 [156, 157]. Both fluorinated derivatives were agonists at ρ1 and ρ2 GABA_C receptors, with the (R)-isomer 97 tenfold weaker and the (S)-isomer 96 20-fold weaker than the natural ligand 95 [158]. This was attributed to weaker electrostatic interactions due to the reduced basicity of the amine, observations that taken together suggested that GABA (95) binds to the GABA_C receptor in the folded orientation represented by conformation C in Fig. 18 [158].

Fig. 18

Structures of GABA (95), the enantiomeric fluorinated derivatives 96 and 97, and their preferred conformations

In an analogous fashion, the bound conformation of N-methyl-d-aspartate (99), an agonist mimetic of glutamic acid (Glu, 98), at recombinant GluN2A and GluN2B receptors expressed in Xenopus laevis oocytes was probed through the synthesis and evaluation of (2S,3S)-3F-NMDA (100) and (2S,3R)-3F-NMDA (101), the available conformations of which are depicted in Fig. 19 [159]. ¹H- and ¹⁹F-NMR spectral data were consistent with 100 adopting conformation A in solution while DFT calculations suggested that 101 would prefer the conformation represented by B in Fig. 19, which was observed in the single-crystal X-ray structure. (2S,3R)-3F-NMDA (101) exhibited no significant effect in the biological assays while (2S,3S)-3F-NMDA (98) induced currents in oocytes expressing either GluN2A or GluN2B receptors, although the maximal responses were less than that observed with NMDA (99) [159]. These observations are consistent with only (2S,3S)-3F-NMDA (100) being able to adopt a conformation that mimics the receptor-bound form of NMDA (99), designated as A in Fig. 19, results that concur with the X-ray crystallographic structure of NMDA (99) bound to the GluN2D receptor that is highly homologous to GluN2A and GluN2B.

Fig. 19

Structures of glutamate (98) NMDA (99), the enantiomeric fluorinated derivatives 100 and 101, and their preferred conformations

β-Fluoroethanol derivatives do not exhibit a strong conformational preference in the absence of an intramolecular interaction between the F and OH moieties which stabilizes a gauche arrangement by approximately 2 kcal/mol [152, 160, 161]. However, this effect is significantly reinforced by protonation of the alcohol which stabilizes the gauche conformer by an energy estimated at ~7 kcal/mol and which has been attributed to the combination of a stereoelectronic effect and an intramolecular H-bonding-type interaction. The HIV-1 protease inhibitor indinavir (102) and its epimer 103 presented an interesting opportunity to probe the consequences of introducing F-OH interactions on conformational disposition, probed with the congeners 104–107, that could readily be equated with enzyme inhibitory activity, measured as the K _i values that are summarized in Fig. 20 [162]. In addition to influencing conformation, it was anticipated that the introduction of a F atom to indinavir (102) would affect the acidity of the OH, potentially introduce steric interactions between drug and target, and also affect solvation. In the indinavir series, the syn,syn analogue 104 exhibits potency comparable to the prototype 102 while the anti,anti isomer 106 is tenfold less active. The syn,anti epi-indinavir derivative 107 is eightfold more potent than progenitor 103, but the anti,syn diastereomer 105 with a K _i of 5,900 nM is substantially (36-fold) less active [162]. The 2 most potent fluorinated compounds syn,syn (104) and syn,anti (107) were both shown to adopt a fully extended conformation in solution based on ¹H-¹H and ¹H-¹⁹F coupling constants, stabilized by a gauche relationship between the F and OH moieties, which is similar to that of indinavir (102) when bound to HIV-1 protease. However, the solution conformations of the less potent fluorinated diastereomers 105 and 106 were considerably more complex, sampling several additional populations and providing a potential explanation for the weaker protease inhibitory activity.

Fig. 20

The structures and HIV-1 protease inhibitory activity associated with indinavir (102), epi-indinavir (103), and the 4 fluorinated analogues 104–107

3.2 The Conformation of Substituted Phenyl and Heteroaryl Derivatives and Isosterism

The topographical preferences of substituted aryl and heteroaryl rings are dependent upon both ring structure and the identity of the substituent atom [10, 163]. An illustration of the effect of substituent identity on biological activity is illustrated by the structure–activity relationships associated with the three non-prostanoid prostacyclin agonists 108–110 compiled in Table 10 which reveal a significant limitation on one of the simplest of the classic isosteric relationships, that between O and CH₂ [10, 65, 66]. The results have been interpreted based on the conformational preferences captured in Fig. 21 with the ether 108 and cinnamate 110 considered to be more potent based on the facility with which they are able to adopt an overall coplanar configuration with the phenyl ring, which contrasts with 109 in which the alkyl side chain projects orthogonally to the plane of the aromatic ring as a consequence of allylic 1,3-strain [10, 65, 66, 122, 163].

Table 10 Platelet aggregation inhibition associated with a series of non-prostanoid prostacyclin agonists

Fig. 21

Preferred conformations of phenoxyacetic acid, cinnamic acid, and phenylpropanoic acid

For the HIV-1 non-nucleoside reverse transcriptase inhibitor (NNRTI) 111, replacing the OCH₂ moiety with a CH₂ linker afforded 112 which largely preserved antiviral activity, suggesting that in this context there is an isosteric relationship between the OCH₂ and CH₂ moieties [164]. This thesis was supported by the good alignment of the key elements of 112 when docked over the X-ray crystallographic structure of 111 bound to the enzyme and reflects the specific orientation of 111 in the enzyme in which the plane of the chlorophenyl and azaindazole rings approach orthogonality.

The conformational preference for arylmethyl ethers that are devoid of ortho substitution is one in which the MeO moiety is close to coplanarity with the phenyl ring and is rationalized on the basis of a rehybridization of the substituent to maximize electronic overlap between the oxygen lone pair and the π system [163]. However, in trifluoromethoxy- and difluoromethoxy-substituted benzenes, the substituent behaves more like an alkyl moiety, projecting orthogonally to the plane of the aromatic ring [165–169]. This conformation is calculated to be energetically favored by ~0.5 kcal/mol over the coplanar topography, which contrasts with the 3 kcal/mol preference for planarity calculated for anisole. An example where this effect appears to be manifested, at least in part, is provided by the cholesteryl ester transfer protein inhibitor 113, IC₅₀ = 1.6 μM, which is eightfold less potent than the fluorinated analogue 114, IC₅₀ = 0.2 μM, a structure–activity relationship attributed to both the effect of steric presentation of the substituent and a favorable drug–protein interaction [16, 166]. In this example, ab initio calculations indicate that while the ethoxy moiety in 113 is more stable when coplanar with the phenyl ring, the substituent in 114 prefers a vector that is closer to perpendicular.

Heteroaryl ethers, in which the substituent is attached to a carbon atom adjacent to a heteroatom, exhibit a topological preference based on lone pair-lone pair repulsive interactions, the energetics of which can be considerable and of both significance and utility in drug design [170–174]. The conformational preferences for a series of methoxy-substituted 5- and 6-membered heterocycles based on DFT calculations are summarized in Fig. 22 and reveal that, with the exception of furan and thiophene, the equilibrium is significantly biased for all cases examined [170]. This preference was used as a design principle to good effect in the identification of the factor Xa inhibitor ZK-807834 (117), a compound that has its origin in the 2,7-dibenzylidenecycloheptanone 115 [171–173]. The progenitor 115 gave a cause for concern based on the potential for photochemically induced olefin isomerization, prompting conception of the pyridine ether chemotype represented by 116 which was designed based on the conformational analysis summarized in Fig. 23. ZK-807834 (117) is a potent factor Xa inhibitor, K _i = 0.11 nM, and an X-ray co-crystal structure confirmed the predicted topology for the acyclic amidine-substituted phenyl ether although the ring with the cyclic amidine moiety adopted the less stable conformation [173].

Fig. 22

Conformational preferences or heteroaryl ethers based on density functional calculations

Fig. 23

Analysis of the topological equilibrium associated with the 2-phenoxy pyridines

Two more recent examples where this phenomenon may be operative are provided by the corticotropin-releasing factor (CRF) antagonist 119, which has potential for the treatment of depression and anxiety, and the GPR119 agonist 121 that promotes postprandial insulin secretion and may be useful for the treatment of diabetes [174, 175]. In both cases, a heteroaryl phenyl ether is employed to mimic the topology inherent to a fused heterocyclic ring system, the pyrrolo[2,3- d]pyrimidine 118 in the case of 119 and the pyrazolo[3,4- d]pyrimidine 120 in the case of 121. In both examples, the favorable effect of the nonbonded lone pair-lone pair interactions may be augmented by intramolecular steric effects afforded by the heteroaryl CH₃ substituent that is manifested as allylic 1,3-strain [174, 175].

The conformational bias provided by an alkyl substituent bound to a phenyl ring has been used to advantage in the design of the factor Xa inhibitor 123 (Table 11) which is based on the biphenyl prototype 122, in which improved physical properties (lower molecular weight and cLogP) were sought [176]. In this example, a cyclopropyl substituent was designed to replace the phenyl ring distal from the methoxyphenyl moiety that engages the enzyme S1 sub-pocket. A careful analysis of phenylcyclopropane conformation indicated that for compounds with a benzylic H atom, the cyclopropane ring adopts a conformation designated as bisected (conformation A in Fig. 24) that is stabilized by electronic effects associated with overlap of the cyclopropyl carbon-carbon bond orbitals with the π system. However, the introduction of a benzylic substituent alters the preference to favor a perpendicular conformation (conformation B in Fig. 24), 0.7 kcal/mol more stable for CH₃, that would more effectively mimic that of the biphenyl moiety of 122 and project the dimethylamine into the enzyme S4 pocket. Reduction to practice revealed that the phenylcyclopropane 123 is close to an order of magnitude more potent than 122, attributed to optimized hydrophobic interactions with the S4 pocket and slightly reduced strain in the bound geometry, as summarized in Table 11 [176]. In addition, the phenylcyclopropane 123 exhibits a reduced cLogP (1 log₁₀) and a lower molecular weight that, when combined with the improved potency, affords significant improvements in ligand efficiency (LE), ligand-lipophilicity efficiency (LLE), lipophilicity-corrected ligand efficiency (LELP), and Fsp³, the ratio of sp³ C atoms to the total number of C atoms, all factors that are believed to be associated with increased drug durability in development [121, 177–182].

Table 11 Potency and physical parameters of drug–target interactions associated with the factor Xa inhibitors 122 and 123

Fig. 24

Perpendicular and bisected conformations of phenylcyclopropane

The bicyclo[1.1.1]pentane moiety is another isostere that mimics the conformation of a phenyl ring while simultaneously reducing lipophilicity and increasing Fsp³. This moiety originally demonstrated value in the context of the glutamate antagonists 124–126 but more recently has found utility in inhibitors of the amyloid precursor protein processing enzyme γ-secretase 127 and 128 [183–186]. The presentation of the two key vectors by the bicyclo[1.1.1]pentane ring faithfully reproduces that of the para-substituents of a phenyl ring, but the dimensions are such that the distance between the substituents is ~1 Å shorter, as captured in Fig. 25 [186]. In the case of the glutamate antagonist 125, a mimic of 124, this could readily be compensated by deploying the larger tetrazole as a carboxylic acid isostere, although the result was less than impressive [183–185]. However, the propellane-based γ-secretase inhibitor 128 is an effective mimic of the phenyl analogue 127 without resort to additional structural adjustment and this compound offers improved solubility, membrane permeability, and metabolic stability in HLM than the progenitor, data summarized in Table 12 [186]. Moreover, 128 is less lipophilic, ElogD = 3.8 compared to 4.7 for 127, LLE increases from 4.76 to 6.55, and the introduction of five additional sp³ carbon atoms more than doubles Fsp³ from 0.25 to 0.53 [186].

Fig. 25

Comparison of vectors and dimensions of a phenyl and bicyclo[1.1.1]pentane moiety

Table 12 Comparison of physical properties for the γ-secretase inhibitors 127 and 128

4 Bioisosteres to Modulate Drug Developability Properties

4.1 Isosteres to Modulate Permeability and P-Glycoprotein Recognition

The substitution of a H atom ortho to the anilide NH by a fluorine in the two closely related series of factor Xa inhibitors 129–132 and 133–135 resulted in improved Caco-2 cell permeability, as summarized in Table 13 [187, 188]. The observed increase in permeability for 130, 132, and 134 compared to the corresponding unsubstituted analogues 129, 131, and 133, respectively, may be due to an effective masking of the H-bond donor properties of the anilide NH by association with the electronegative F atom in an electrostatic interaction [133, 189, 190]. In contrast, an ortho nitrile substituent in the aminobenzisoxazole chemotype afforded a compound 135 with significantly reduced permeability, presumably a function of increased acidity and H-bond donor capacity of the NH while geometrical constraints prevent the linear cyanide moiety from establishing an intramolecular H-bond.

Table 13 The effect of substituents ortho to an anilide on Caco-2 permeability in two series of factor Xa inhibitors

The presence of a fluorine atom capable of engaging an amide or sulfonamide N–H in an electrostatic interaction has emerged as a common structural motif, particularly in kinase inhibitors, and may contribute to improved oral exposure, as illustrated by the matched pair comparisons between 136 and 137 and 138 and 139 that are compiled in Table 14 [191, 192].

Table 14 Potency and oral bioavailability of Rho kinase (ROCK 1) inhibitors 136 and 137 and the kinase insert domain receptor (KDR) inhibitors 138 and 139

Intramolecular H-bond capture by a pendent F atom was exploited to reduce P-glycoprotein (P-gp)-mediated brain efflux in a series of β-site amyloid precursor protein cleaving enzyme (BACE1) inhibitors [193–195]. The prototype BACE1 inhibitor 140 presented in Table 15 exhibited a large efflux ratio in cell lines expressing human or rat P-gp while analogues 141 and 142 show improved permeability properties, attributed to an intramolecular interaction between the F atom in the amide cap moiety and the NH, consistent with P-gp substrate recognition that relies upon protein–drug H-bonding [193–197].

Table 15 Caco-2 permeability and efflux ratios for the series of BACE1 inhibitors 140–142

A similar tactical approach provided a solution to improved membrane in the series of CNS-penetrant bradykinin B1 antagonists 143–146 compiled in Table 16 which were explored as potential agents for the relief of pain [198]. The introduction of F atoms into the amide moiety resulted in improved passive permeability and reduced P-gp efflux ratios, attributed to a reduction in the strength of the carbonyl marked with an asterisk (*) to act as a H-bond acceptor [198]. However, an intramolecular interaction between the F atoms and the NH may also be contributory.

Table 16 Caco-2 permeability, efflux ratios, and H-bond strength for the series of bradykinin B1 antagonists 143–146

The alternative topology depicted in Fig. 26 offers the potential for a similar electrostatic interaction between the ortho-F atom and the benzamide NH and this structural element is also prevalent in drug design [133, 189]. The potent tachykinin hNK₂ receptor antagonist 147 (Table 17) exhibited poor membrane permeability across a confluent Caco-2 cell layer, attributed to the polar amide NH moiety which was critical for potency and could not be modified by methylation or replaced by an isostere [199]. A halogen atom was introduced ortho to the amide carbonyl moiety to establish an intramolecular interaction with the NH, described as a H-bond with the fluoro derivative 149, that resulted in improved permeability across a parallel artificial membrane (PAMPA) or a Caco-2 cell layer while a Cl substituent (148) was less effective but still superior to H. In this context, the ortho-F substituent in 149 performed somewhat similarly to the nitrogen atom of pyridine 151 which offers a more conventional H-bonding opportunity and markedly improved permeability in both assays compared to the analogous phenyl derivative 150 [199]. The ortho-F atom in the antiandrogen enzalutamide (152), which was approved for the treatment of castration-resistant prostate cancer by the FDA in August 2012, may contribute to its excellent pharmacokinetic profile [200, 201]. A similar motif is presented by ZD-9331 (153), a fluorinated methotrexate analogue which exhibits activity toward ovarian cancer cells resistant to classical thymidylate synthase inhibitors [202, 203].

Fig. 26

The ortho-fluorinated benzamide motif

Table 17 Structure, human NK2 receptor binding affinity, PAMPA, and Caco-2 permeability for a series of phenyl alanine-based antagonists

Another productive F–NH relationship that exerts a beneficial effect on Caco-2 permeability and efflux ratio is provided by the series of cyclic benzamidine-based BACE1 inhibitors 154–157 (compiled in Table 18) [204]. The prototype compound 154 exhibited significant basicity, poor membrane permeability, and a high efflux ratio predictive of the molecule being a P-glycoprotein substrate which was anticipated to further contribute to reduced brain exposure. A key step toward solving the problem was the introduction of a fluorine atom ortho to the amidine moiety to afford 155, a compound with a pK _a that is reduced by 1.3 units that exhibits improved Caco-2 permeability and a reduced efflux ratio while preserving BACE1 inhibitory activity. A similar effect was observed between the matched pairs 156 and 157 in which the F atom is believed to form a weak H-bond with the NH while calculations indicated that the solvation energy of the fluorinated derivative is less negative than for the H analogue, electronic and steric effects that shield the polar nitrogen atom which thus presents less than 2 H-bond donors to the environment [204].

Table 18 Caco-2 and pKa data associated with BACE1 inhibitors 154–157

4.2 Isosteres of Guanidines and Amidines

The high basicity associated with guanidine and amidine moieties limits the fraction of the more permeable unprotonated form that exists at physiological pH, providing an understanding for the generally poor permeability associated with molecules incorporating these structural elements [205, 206]. The prevalence of arginine as the P1 moiety of substrates of the serine protease enzymes that constitute the coagulation cascade has catalyzed the identification of guanidine and amidine surrogates as part of the effort to identify potent, selective, and orally bioavailable inhibitors that offer potential as antithrombotic agents [207–210]. The pK _a of a substituted guanidine moiety can be reduced significantly from the typical 13–14 range by modification to an acylguanidine, pK _a ~8, or an oxyguanidine, pK _a ~7–7.5, both of which represent isosteric replacements of a methylene moiety. However, these modifications have the potential to significantly affect molecular recognition that may be manifested as a reduction in potency, depending on the circumstance under study. Nevertheless, acylguanidines have been successfully deployed in a series of histamine H₂ agonists and NPY Y2 antagonists while an oxyguanidine moiety has proven to be an effective surrogate of arginine in thrombin inhibitors, with RWJ-671818 (158) a representative compound that was advanced into phase 1 clinical trials [211–218].

An extensive survey of benzamidine mimetics was conducted using the potent factor Xa inhibitor SN429 (159), K _i = 13 pM, as the basis for assessing the effect of this kind of structural variation on potency and oral bioavailability (Fig. 27) [219]. The study identified several neutral substituents that functioned as useful and effective amidine surrogates in this context, including the 3-chlorophenyl analogue which demonstrated a K _i of 37 nM, potency similar to that of the 3-aminophenyl derivative, which displays a K _i of 63 nM. However, both of these compounds are tenfold weaker than the more basic 3-aminomethyl compound, which exhibits a K _i of 2.7 nM, and are significantly less potent than the prototype 159. However, the 4-methoxy analogue represented an acceptable compromise between potency (K _i = 11 nM) and pharmacokinetic properties and it is this P1 moiety that is found in the factor Xa inhibitor apixaban (160) which received marketing approval from the FDA in December 2012 as an agent for reducing the risk of blood clots and stroke in subjects experiencing atrial fibrillation that is not related to a heart valve problem [207, 220].

Fig. 27

The structure of the factor Xa inhibitor SN429 (159) and some of the benzamidine surrogates evaluated in an attempt to identify potent enzyme inhibitors with improved oral bioavailability

4.3 Isosteres of Phosphates and Phosphonates

Phosphates and phosphonates are highly acidic elements, and the low pK _a associated with these moieties provides a significant limit to membrane permeability and, hence, oral bioavailability [221–223]. Monofluoro- and difluoro-methylenebisphosphonic acids have been developed as useful chemically and enzymatically stable phosphate isosteres, but these rely upon preserving the inherently high acidity of the prototype and are not useful in the design of orally bioavailable drug candidates without resort to prodrug technology, which has been the most widely applied and successful tactic to deliver this kind of polar structural element [224–229]. Phosphate and phosphonate isosteres have been of particular interest in the design of nucleoside and nucleotide antiviral and anticancer agents and phosphatase inhibitors for which the design principles are based on substrate mimicry [230, 231]. However, the most notable success in the design of phosphate mimics has been accomplished in the arena of inhibitors of the strand transfer reaction catalyzed by HIV-1 integrase where the seminal identification of α,γ-diketo acids as phosphate transfer transition state mimics inspired the design of a wide range isosteres, summarized in Fig. 28, that are compatible with oral bioavailability [232–237].

Fig. 28

A selection of HIV-1 integrase inhibiting motifs that mimic the transition state for phosphate transfer during integration of viral DNA into host chromosomal DNA

The HIV-1 integrase inhibitors raltegravir (161), elvitegravir (162), and dolutegravir (163) have been licensed for marketing by the FDA for the treatment of HIV-1 infection [232–237].

4.4 Isosteres to Modulate Basicity/Acidity and Solubility

The presence of a basic amine in a molecule can be associated with several problems including compound promiscuity, inhibition of the human ether a go- go-related gene type 1 (hERG) cardiac potassium channel, phospholipidosis, and recognition by P-glycoprotein [121, 238–246]. The basicity of an amine can readily be modulated by the introduction of proximal electron-withdrawing substituents or functionality, with the highly electronegative fluorine the most notable and one of the most widely utilized based on its metabolic stability and modest steric volume [247]. The pK _as of fluorinated ethylamines are summarized in Table 19 where the data indicate that the effect of introducing a F atom is additive in nature which allows the change in pK _a to be estimated with reasonable accuracy for aliphatic amines. Each fluorine atom introduced to a C atom β- to the amine reduces the pK _a by 1.7 units, an effect that is dependent on σ-transmission and therefore declines with increasing distance such that the effect at the γ-C is reduced to a shift of −0.7 units while a F at the δ-C reduces pK _a by 0.3 units, data that is reflected in Table 19.

Table 19 The effect of proximal F atoms on the basicity of ethylamine

An insightful example of the application of the effect of F to reduce amine basicity is provided by the design of inhibitors of the motor protein kinesin spindle protein (KSP) explored as a potential therapy for the treatment of taxane-refractory solid tumors [248]. The efficacy of this series of compounds was restricted by P-gp efflux which was determined experimentally to be minimized by adjusting the pK _a of the amine to a range between 6.5 and 8.0. The N-cyclopropyl derivative 164 represented an initial solution to this problem but was associated with time-dependent cytochrome P450 inhibition, a known liability of this structural element [249]. The β-fluoroethyl derivative 165 also satisfied the pK _a requirement but was found to be N-dealkylated in vitro and in vivo to release fluoroacetaldehyde which was oxidized to fluoroacetic acid, a toxin that is metabolized in vivo to a potent inhibitor of aconitase, an enzyme in the tricarboxylic acid cycle [139, 250]. This problem was solved by deploying the fluorine atom in the piperidine ring β- to the amine, an arrangement that produced the cis derivative 166, in which the F atom is axially disposed as confirmed by single-crystal X-ray analysis, and the trans analogue 167, where the F atom adopts an equatorial disposition. The effect of this structural modification on the basicity of the piperidine was dependent on the stereochemical disposition of the F atom with the pK _a of 167 determined to be 6.6 while 168 was more basic with a pK _a of 7.6 and it was this compound, designated as MK-0371, that was selected for clinical evaluation [248].

The electron-withdrawing properties of the CF₃ and CHF₂ deployed β- to an amine reduce basicity to an extent that these functionalities have found application as amide mimics, an isosteric relationship furthered by the similarity of dipoles and the geometry of the N-C-CF₃ and N-C-CF₂, which approximates the 120° associated with an amide moiety, as illustrated in Fig. 29 [251–254]. Although this bioisostere was originally conceived to replace the amide bonds of peptide derivatives, it has begun to find a similar application in drug design, with the most prominent example being the cathepsin K inhibitor odanacatib (170) which has completed phase 3 clinical trials for the treatment of osteoporosis. Cathepsin K is a lysosomal cysteine protease in osteoclasts that is responsible for bone degradation during remodeling and inhibitors prevent bone resorption. L-006235 (168), in which the nitrile moiety is presented to the enzyme as an electrophile to react reversibly with the catalytic cysteine thiol, emerged as a refined cathepsin K inhibitor that exhibited good pharmacokinetic properties. However, this compound has poor selectivity for cathepsin K versus the analogues enzymes cathepsin B, L, and S due to its strongly basic nature, which promoted accumulation in the acidic environment of lysosomes [255]. L-873724 (169) offered good enzyme inhibitory selectivity but poor PK, with further optimization leading to odanacatib (170) which solved both problems by installing a F atom in the Ile residue to block hydroxylation while the cyclopropyl moiety reduced the propensity for amide hydrolysis [255, 256]. While odanacatib (170) has been quite successful clinically, with the phase 3 trial halted early due to the observation of good efficacy and safety, the pharmaceutical properties of this molecule are less than ideal [257–260]. Odanacatib (170) is highly crystalline and exhibits low aqueous solubility, properties that contribute to the ≤10% bioavailability across preclinical species after dosing the drug as a suspension in methocel. In an effort to overcome the dissolution-limited bioavailability, modification of the CF₃ moiety to a CHF₂ was explored as a means of improving the physical properties by increasing the pK _a of the amine [260]. This modification afforded 171 which preserved cathepsin K inhibitory potency and selectivity while decreasing log D by 3 units, attributed to the increased basicity, with the result that bioavailability from a 1% methocel suspension was improved fourfold from 6% with odanacatib (170) to 23% with 171 in the dog and 8–22% in the rat. The increase in basicity with 171 facilitated salt formation with strong acids like HCl although, interestingly, salts offered no advantage over the neutral form in rat PK studies [260].

Fig. 29

Isosterism between an amide and trifluoroethylamine and difluoroethylamine moieties

In a series of tetrahydroisoquinoline-based inhibitors of phenylethanolamine N-methyltransferase (PNMT), the enzyme that catalyzes the final step in epinephrine biosynthesis by methylating norepinephrine using S-adenosyl l-methionine as the cofactor, binding of these compounds to the α₂ adrenoreceptor was a significant issue. The prototypical 3-methyl derivative 172 is only modestly selective (Table 20) [261]. In an effort to address this problem, the effect of modulating the basicity of the amine by the successive introduction of F atoms into the 3-methyl group was explored, an approach that proved to be both productive and informative. The monofluoro analogue 173 exhibited similar affinity for the enzyme as 172 while α₂ binding declined only modestly and this molecule was calculated to be less basic than the prototype by 1.5 pK _a units. The trifluoromethyl homologue 175 exhibited poor potency in both assays and is poorly basic; however, the difluoromethylated compound 174 provided the optimal balance of properties, retaining good affinity for PNMT while reducing α₂ adrenoreceptor binding by ~200-fold and providing the first example of fluorination of an amine modulating target selectivity [261].

Table 20 Calculated basicity and inhibitory potency of a 3-substituted tetrahydroisoquinolines toward human PNMT and the α₂ adrenoreceptor

A detailed analysis of the properties of 3-substituted oxetane rings and applications in the context of broader functionality have established this heterocycle as an advantageous structural element when utilized in a fashion that takes advantage of its electron-withdrawing properties and topographical isosterism with the carbonyl functionality [262–265]. These properties allow the oxetane ring to be deployed as a ketone or amide mimetic depending on the structural context with the electron-withdrawing effects reducing the basicity of an amine in a fashion that is strictly dependent on proximity [262]. As is evident from the homologous series presented in Table 21, the amino oxetane 177 is the weakest base while the aminomethyl and aminoethyl homologues 178 and 179, respectively, exhibit progressively increased basicity which shows a correlation with solubility, with the exception of the parent molecule 176 which is poorly soluble despite being the most basic [262]. It is the weak basicity associated with 177 coupled with the structural mimicry of the carbonyl moiety by the oxetane ring which adopts a planar topography that has led to this motif being considered as a useful isostere of an amide functionality (Fig. 30).

Table 21 Basicity and solubility of a series of oxetane-substituted phenylbutyl amines

Fig. 30

Topographical similarity between the carbonyl and oxetane moieties

4.5 Isosteres to Modulate Lipophilicity and sp² Atom Count

The oxetane ring has also been proposed as a useful replacement for the gem-dimethyl moiety that provides a similar vectorial presentation of the 2 substituents (Thorpe-Ingold effect) and a similar size while reducing the lipophilic burden [121, 262–266]. This concept is particularly useful in an era where there is a considerable focus on the role of physical properties in drug design that has fostered the belief that contemporary practices relies far too heavily on lipophilicity to derive potency, a strategy that is based on taking advantage of entropic rather than enthalpic contributions to binding affinity [121, 177, 267–270]. The rising appreciation of the potential problems associated with this approach to drug design, referred to as molecular obesity, is beginning to be manifested in the description of strategies and tactics to identify scaffolds that replace sp² carbon centers by sp³-based motifs [121, 181, 182, 238, 271–275]. One example of this is provided by studies of oxytocin antagonists of which 180 was the prototype, a compound potent receptor affinity, K _i = 6 nM, but poor aqueous solubility of 6 μg/mL and a low Fsp³ of 0.18 [276]. The effect of replacing the pyrazine ring with piperidine, pyrrolidine, and azetidine was examined from which the azetidine 181 emerged as a preferred compound that bound to the oxytocin receptor with a K _i of 9.5 nM but which exhibited tenfold improved aqueous solubility of 59 μg/mL and a much increased Fsp³ of 0.46 [276].

The establishment of an intramolecular H-bonding interaction can mimic the topology of an aromatic or heterocyclic ring in a fashion that reduces the dependence on sp² carbon centers, and this isosterism-based concept has found application in kinase inhibitor design. The Cdk4 inhibitor 182 (Fig. 31) was identified as a weak lead inhibitor, IC₅₀ = 44 μM, derived by structure-based scaffold generation using a model of Cdk4 and a proprietary modeling program [277]. The SAR associated with the homologous series of pyridine derivatives 183–185 is summarized in Table 22 and clearly reveals the importance of the topological deployment of the N atom, attributed to the formation of an intramolecular H-bond between the 2-pyridyl N atom of 185 and the distal urea NH that favors a cis amide topology. This hypothesis was subsequently confirmed by a co-crystal of the inhibitor with Cdk2 (Fig. 31) and the effect could be recapitulated with the thiazole 186 although this compound is threefold less potent than 185 [277].

Fig. 31

Structure of the Cdk4 kinase inhibitor lead 182 and drug-target interactions for the analogue 185

Table 22 SAR associated with urea-based Cdk4 kinase inhibitors

PD-166285 (187) is a broad-acting kinase inhibitor that was used as a vehicle to explore the concept of replacing the pyridone ring with a mimic based on an intramolecular H-bond to favor the preferred topology [278, 279]. Ab initio calculations predicted that the cis urea is favored by 0.5 kcal/mol in H₂O and 3.2 kcal/mol in the gas phase, a contention supported by a search of the pyrimidinyl urea substructure in the CSD where seven molecules were found, all of which exhibited an intramolecular H-bond (Fig. 32 ). The urea 188 demonstrated broad-based kinase inhibitory activity with potency comparable to 187 and the derivative NVP-BGJ398 (189) was advanced into phase 1 clinical trials [278, 279].

Fig. 32

Preferred conformation of 1-methyl-3-(pyrimidin-4-yl)urea

Another successful example of the application of this concept is provided by the anthranilamide 190 which mimics the phthalazine 191, discovered as an inhibitor of the fibroblast growth factor receptor family of receptor tyrosine kinases by high-throughput screening [280]. The anti topology of the para-chlorophenyl ring depicted is calculated by ab initio methods to be 3.1 kcal/mol more stable than the cis conformer. This facilitated the design concept of replacing the pyridazine ring by an intramolecularly H-bonded moiety which led to compounds with potent kinase inhibitory properties [280].

Other examples from the kinase inhibitor arena where this concept is well represented include the dual PI3Kα/mTOR inhibitor 193 which is derived from 192, Lck inhibitors 194, and the p38α MAP kinase inhibitor 195 [281–283]. The presence of intramolecular H-bonds in molecules of this type in an aqueous environment has recently been verified by NMR [284, 285].

An interesting application of the isosteric interplay between rings and H-bonded surrogates is provided by the design of the B-Raf^V600E kinase inhibitor 200 which has its origins in the benzamide derivative 196 [286]. The lead amide 196 exhibits poor solubility, 3–9 μg/mL, and a high melting point of 226°C and offers only a very weakly basic aminopyridine moiety, pK _a = 0.7, as a vehicle for salt formation, leading to a dependence on an amorphous dispersion-based formulation to deliver adequate oral exposure. The potency of 196 was improved markedly by modifying the aminopyridine moiety to the pyrazolopyridine found in 198 and both compounds were used as vehicles to explore the design of urea-based inhibitors that relied upon intramolecular H-bonding to correctly orient the pharmacophoric elements. This was successful in the context of the ureas 198 and 199, but chemical instability led to further structural refinement with optimization ultimately focusing on the reverse amide series represented by 200, a series with improved physical properties and robust antitumor activity in a B-Raf^V600E mouse xenograft model [286].

The sulfonamide moiety was developed as a useful phenol isostere in the context of β-adrenergic antagonists with advantage since the high polar surface area and H-bonding properties of this moiety frequently restrict blood–brain barrier penetration as a consequence of reduced permeability and recognition by P-gp [1–10, 196, 197, 287–292]. These physical properties have been exploited in restricting drug molecules to peripheral tissues, but the increased polarity may also reduce oral bioavailability due to poor membrane permeability and there are occasions where more lipophilic moieties are desirable [293, 294]. Although the 2,2-difluoro-2-(pyridin-2-yl)ethanamine moiety in the thrombin inhibitor 202 was introduced to interfere with benzylic oxidation by cyctochrome P450 (CYP 450), this moiety may, in essence, be functioning as a sulfonamide isostere based on analogy with the prototype 201 [216–218, 295–297]. In this context, not recognized in the design process, the two electronegative F atoms may be viewed as nonpolar mimics of the sulfone oxygen atoms that also reduce the basicity of the nitrogen atom and enhance the H-bonding donating properties of the NH to more effectively mimic the sulfonamide. The extra atom introduced with the phenethyl moiety may compensate to some extent for the longer bonds associated with the C–S bonds in a sulfonamide. However, the difluoroalkyl moiety has also been invoked as a heteroaryl ether isostere with the caveat that the geometry and topology of this moiety differs from that of an oxygen-based substituent; indeed, the difluoroalkyl substituent may more effectively mimic a fluoroalkyl ether which adopts an orthogonal disposition to the phenyl ring [298]. Interestingly, 202 also incorporates a fluorobenzene moiety to mimic the pyridone with a close interaction between the F atom and the thrombin peptide backbone NH observed in the co-crystal, indicative of isosterism between C–F and C=O bonds.

Another example where a difluoromethyl mimics a sulfone can be found in the matched pair of long-acting dual D2-receptor/β2-adrenoceptor agonists 203 and 204 captured in Table 23 [299]. The affinity of 203 and 204 for the β2-adrenergic receptor and the intrinsic agonistic properties are very similar, but the log D_7.4 and pK _a of the nitrogen atom for the 2 molecules differ, with the difluoromethyl analogue 204 more lipophilic but more basic. The design of 204 was based on the development of a model that identified secondary amines with a pK _a > 8.0 and a log D_7.4 > 2 as molecules possessing an ultra-long duration of action [299].

Table 23 Structure and properties associated with the dual D2-receptor/β2-adrenoceptor agonists 203 and 204

5 Isosteres to Address Metabolism and Toxicity

5.1 Substitution of Hydrogen by Deuterium

The substitution of H by D is the most conservative form of bioisosteric replacement and this tactic has attracted considerable attention recently as a practical approach to drug design. The underlying premise relies upon the effect of deuterium substitution to influence pharmacokinetic properties when deployed in a strategic fashion that takes advantage of the kinetic isotope effect (KIE) associated with the heavier atom [300–304]. The differences between the isotopes are small but measurable, with deuterium 0.140 cm³/mol per atom smaller than hydrogen, the C–D bond is 0.005 Å shorter than a C–H bond, and the log P_oct of D is 0.006 units less lipophilic, an effect that can be measured in per-deuterated alkanes and which has facilitated the chromatographic separation of enantiomers that are based solely on H/D isotopic substitution [305–307]. Deuteration slightly increases the basicity of proximal amines and reduces the acidity of phenols and carboxylic acid derivatives [308–311]. Following the discovery of the isotope, deuteration of drug molecules was quickly adopted as a useful structural modification for the study of metabolic pathways and the origins of toxicity, with the KIE playing the key role in this application [300, 312–314]. However, the study of the deuteration of drug candidates as an approach to improving pharmacokinetic properties by reducing the rate of metabolic modification at sites where H atom abstraction determines the rate of reaction has been adopted only recently, although the first demonstration of the practicality of the approach was described in 1961 [315]. The KIE for a D for H substitution is dependent on the circumstances but usually ranges from one- to sevenfold with calculations suggesting a seven- to tenfold difference; however, in specific cases, much higher isotope effects have been measured [316, 317]. D for H substitution has been shown to translate into meaningful changes in the clinical pharmacokinetic profiles of drugs, exemplified most effectively by CTP-347 (205), a dideuterated analogue of the antidepressant paroxetine, and SD-254 (206), a per-deuterated derivative of the dual serotonin/norepinephrine reuptake inhibitor venlafaxine that also finds utility for its antidepressant properties [303, 318, 319]. In CPT-347 (205), deuterium is introduced at the methylenedioxy moiety, the site of metabolic modification by CYP 2D6, which leads to mechanism-based inhibition of the enzyme through the intermediacy of a carbene-based metabolite, resulting in drug accumulation on repeat dosing due to autoinhibition and precipitating drug–drug interactions [303, 320–322]. CPT-347 (205) does not exhibit significant mechanism-based inhibition of CYP 2D6 in vitro and a phase 1 clinical study demonstrated that subjects dosed with the deuterated drug were able to metabolize the CYP 2D6 substrate dextromethorphan more effectively when compared to historical data for paroxetine [303]. In a phase 1 clinical trial in normal healthy volunteers, SD-254 (206) was reported to be metabolized more slowly than venlafaxine which is O-demethylated by CYP 2D6 [303].

However, selective deuteration at a site of metabolism does not always lead to enhanced pharmacokinetic properties, illustrated by a study of the tyrosine kinase inhibitor imatinib (207) for which N-demethylation to the less active piperazine 208 is a major metabolic pathway [323–325]. The tri-deuterio analogue 209 exhibited increased stability toward N-demethylation in rat and human liver microsomes, as anticipated, but intravenous administration of the compound to rats was not associated with increased exposure and the rate of demethylation of 209 was similar to 207, attributed to a relatively low rate of demethylation of the parent in rat liver microsomes and the low rate clearance of both compounds in rats [323].

An alternative application of deuterium incorporation that led to improved pharmacokinetic properties is illustrated by the HCV NS3 protease inhibitor telaprevir (210) which suffers facile racemization at the P1 residue at higher pH and in human plasma [326]. The (R)-diastereomer is the major metabolite of telaprevir in vivo but is 30-fold weaker than the (S)-isomer which contributes to the need for this drug to be dosed on a TID schedule. Installing a deuterium atom at the configurationally labile carbon atom gave a compound that inhibited HCV NS3 protease with a K _i = 20 nM, approximately twofold more potent than the telaprevir (210), and reduced the propensity for racemization in human plasma: the deuterated compound produced only 10% of the epimer over 1 h of incubation compared to 35% with the protio homologue. Although the stability of the deuterio derivative in rat plasma was lower, this compound exhibited a 13% increased AUC following oral administration to rats when compared to telaprevir (210) [326].

Deuteration at metabolically labile sites can redirect metabolism away from a toxic species, as in the case of the HIV-1 non-nucleoside reverse transcriptase inhibitor efavirenz (211), or enhance bioactivation of prodrugs, as illustrated by the platelet aggregation inhibitor clopidogrel (214) [327, 328]. In rats, the metabolism of efavirenz (211) is complex, with the initial step hydroxylation of the aromatic ring to afford a phenol that is a substrate for sulfation. Subsequent hydroxylation at the propargylic position affords a cyclopropylcarbinol that facilitates the addition of glutathione to the alkyne to afford an adduct from which the glutamate is subsequently cleaved, a process blocked by acivicin, to afford 213, which was determined to be the source of kidney toxicity seen uniquely in rats. As part of this investigation, deuteration at the propargylic position was designed to slow the hydroxylation step and reduce the amount of 213 produced, a successful enterprise since 212 exhibits less severe nephrotoxicity of lower frequency [327].

Clopidogrel (214) requires metabolic activation in order to express its antithrombotic effects, a process that is initiated by oxidation of the thiophene ring by CYP 450 enzymes to produce the thiolactone 215 [328, 329]. Thiolactone 215 is further oxidized to the acylated sulfoxide 216, a species that is readily hydrolyzed to the highly reactive sulfenic acid 217, considered to be the ultimate active principle of the drug that reacts covalently with and blocks the P2Y₁₂ receptor that is activated by adenosine diphosphate (ADP) [328, 329]. However, the metabolism of clopidogrel (214) into the active species is limited, complicated by alternative pathways that involve cleavage of the ester moiety to afford an inactive acid and CYP-mediated oxidation of the piperidine ring [328]. In an effort to direct metabolic activation toward the thiophene ring and enhance the antiplatelet activity of clopidogrel (214) in vivo, the effect of deuteration of the piperidine was examined. The d ₆ derivative 218 exhibited similar metabolic stability to clopidogrel (214) in rat and human liver microsomes but improved conversion to the thiolactone 215, an effect that manifested in vivo in rats as enhanced ex vivo inhibition of ADP-induced platelet aggregation following oral administration of 218 [328].

5.2 Substitution of Carbon by Silicon

The replacement of carbon atoms by silicon has attracted attention as a drug design principle that may offer advantage in specific circumstances where the unique properties of silicon can be exploited [330, 331]. A particularly compelling example of the application of deploying silicon in a tactical sense is provided by studies with the antipsychotic agent haloperidol (219), a dopamine D₂ antagonist, in which the replacement of the carbinol carbon by silicon was examined as a means of eliminating a problematic metabolic pathway [332, 333]. Haloperidol (219) is metabolized, in part, by dehydration of the piperidinol which affords the tetrahydropiperidine 221, a precursor to the pyridinium 222 which is neurotoxic, as summarized in Fig. 33. Sila-haloperidol (220) exhibits biological properties that are quite similar to the progenitor 219, but this compound cannot be metabolized by dehydration to 223 and further oxidized to 224 due to the inherent instability of carbon-silicon double bonds [332]. Indeed, the metabolism of sila-haloperidol (220) follows more conventional pathways associated with the piperidine moiety in which oxidative metabolism occurs adjacent to the N atom leading to dealkylation of the fluorophenyl-containing side chain to afford 225 and ring hydroxylation that produces 228 and 230 (Fig. 34) [334].

Fig. 33

Metabolism of haloperidol (219) to a pyridinium

Fig. 34

Metabolism of sila-haloperidol (220)

5.3 Isosteres of Alkyl Groups and Alkylene Moieties

tert-Butyl moieties are susceptible to metabolic oxidation of the methyl groups and this pathway can be a source of poor pharmacokinetic properties. Replacing a single methyl group with CF₃ in the context of the NK1 receptor antagonist 231 and the vanilloid receptor antagonist 233 afforded 232 and 234, respectively, as compounds that demonstrate enhanced metabolic stability in human liver microsomes compared to the prototypes [335]. A further refinement of this approach identified the trifluoromethylcyclopropyl as a metabolically more stable tert-butyl isostere, exemplified in the type II 5α-reductase inhibitor finasteride (235) where this substitution provided 236 which exhibits a half-life in human liver microsomes of 114 min, a substantial improvement on the 63-min half-life measured for the progenitor [336]. The potent respiratory syncytial virus fusion inhibitor 237 exhibits poor stability in human liver microsomes with a t _1/2 of just 4 min, similar to the isopropyl- and cyclobutyl-substituted homologues while the cyclopropyl derivative 238 uniquely addressed this deficiency exhibiting a tenfold improved t _1/2 of 39 min [337].

Alkanoic acid derivatives can be metabolized by β-oxidation that can lead to poor pharmacokinetic properties in vivo, a problem encountered with iloprost (240), an analogue of prostacyclin (239) in which the chemically labile bicyclic enol ether oxygen atom is replaced with a CH₂ isostere [338]. β-Oxidation of fatty acids is a degradative pathway that proceeds mechanistically as depicted in Fig. 35 and involves the initial formation of a CoA ester as a substrate for oxidation to an α,β-unsaturated CoA ester that is further metabolized to a CoA ester with two fewer carbon atoms [339, 340]. The essential formation of the α,β-unsaturated CoA ester provides opportunity for rational structural modification of xenobiotic carboxylic acids to block this pathway. The introduction of germinal substitution at the α- or β-positions and replacing the β-carbon with a heteroatom are effective tactics that prevent olefin formation and it was the latter that was successfully applied to iloprost (240) leading to the design of cicaprost (241), a compound with longer-lasting hypotensive effects in rats following oral administration [338].

Fig. 35

The β-oxidation pathway for alkanoic carboxylic acids

5.4 Isosteres of Carboxylic Acids to Reduce Reactive Metabolite Formation

Acyl-CoA esters have been shown to be inherently electrophilic acylating agents, providing further impetus to replace the carboxylic acid moiety with an isostere that cannot engage in this metabolic pathway [10, 341–345]. Carboxylic acids are also readily conjugated with glucuronic acid in vivo to afford acyl glucuronides that can undergo sequential rearrangement of the acyl moiety to the adjacent hydroxyl substituents on the pyranose ring, a process summarized in Fig. 36. This becomes problematic when the pyran ring opens to reveal an unnatural aldehyde which can react with amine moieties of proteins, lysine, for example, to generate an imine [346, 347]. The imine intermediate can be trapped with NaB³[H]₄, but this functionality can also undergo an Amadori rearrangement to afford the corresponding aminomethyl ketone, resulting in an irreversible protein modification that may create a hapten. Insights have been gleaned into the properties of a carboxylic acid that promote this rearrangement which has been shown to be more facile for electron-deficient acids but slowed by bulky substituents at the α-carbon atom [348–350]. Guidelines have been developed that attempt to equate the propensity of rearrangement with the potential for idiosyncratic toxicity, with the half-life of the acyl glucuronide in buffer that separated safe from problematic acids estimated to be 3.6 h [351]. The level of mechanistic understanding underlying the potential toxicity of carboxylic acid provides a rationale for structural modification to avoid this metabolic pathway by deploying an acid isostere incapable of undergoing rearrangement after glucuronidation. Figure 37 captures the range of acidic functionality that has been examined to address problems across a wide variety of medicinal chemistry campaigns [10, 344, 345, 352, 353].

Fig. 36

Rearrangement of acyl glucuronides and exposure of an electrophilic aldehyde moiety

Fig. 37

A sampling of carboxylic acid isosteres

5.5 Isosteres of Thiols and Alcohols

The thiol moiety is relatively uncommon in drugs and drug candidates due to potential problems arising from metabolic lability or chemical reactivity, with the angiotensin-converting enzyme inhibitor captopril (243) and penicillamine (244), a chelator used to promote heavy metal excretion, the most prominent exceptions [354, 355].

A useful although somewhat underutilized isostere of a thiol is the CHF₂ moiety which has been applied in an elegant fashion to the design of HCV NS3 protease inhibitors where the P1 cysteine moiety of hexapeptide, substrate-based inhibitors, was considered to be a liability [356]. Capitalizing on earlier observations of the intramolecular H-bonding properties of the CHF₂ moiety in 248, this element was examined as a thiol surrogate in the context of the potent NS3 inhibitor 245 which exhibited a K _i = 40 nM [356, 357]. Replacing the P1 SH with CH₃ led to an almost 20-fold erosion of potency and 246 inhibited the enzyme with a K _i = 700 nM. However, a CHF₂ terminus at P1 proved to be very effective, fully restoring the potency of 247 to that of the cysteine progenitor, with a K _i = 30 nM [356]. This amino acid, referred to as difluoro-Abu, was conceived to be suitable for this context after an insightful analysis of its properties which suggested considerable potential to mimic the cysteine. Mimicry between a SH and a CHF₂ was based on several observations that included the similarity of the van der Waals surfaces of the two structural elements, 46.7 Å for HCF₂CH₃ compared to 47.1 Å for HSCH₃ and electrostatic potential maps, which revealed that the negative potential around the sulfur lone pairs of electrons was similar to that of the two fluorine atoms while there was positive potential around both the SH and CF₂H hydrogen atoms, suggestive of H-bonding capability. Indeed, an X-ray co-crystal structure of a related inhibitor revealed that the CF₂H moiety donated a H-bond to the C=O of Lys₁₃₆ while one of the fluorine atoms was close enough to the C-4-hydrogen atom of Phe₁₅₄ to suggest the presence of a weak C–H to F H-bond [356].

Lysophosphatidic acid (249) is a mitogen that has been shown to interact with four G protein-coupled receptors designated LPA 1–4 in addition to being an agonist at the peroxisome proliferator-activated receptor γ (PPARγ), a nuclear hormone receptor, and antagonists offer potential in a range of clinical conditions including the treatment of liver and lung fibrosis, several cancers, and neuropathic pain [358, 359]. The acyl moiety in lysophosphatidic acid (249) can migrate to the primary alcohol, stimulating the design of analogues that are functionally incapable of entering this rearrangement pathway [360, 361]. In this context the CHF₂ moieties of 250 and 251 were conceived as potential mimics of the OH moiety of 249 that would be chemically stable. Neither compound was recognized by LPA receptors 1, 2, or 3, with both agonistic and antagonistic properties evaluated, but 250 was found to stimulate luciferase production in CV-1 cells transfected with luciferase under the control of a PPARγ-responsive element [360, 361].

5.6 Substitution of C–H by N in Phenyl Rings and C by N in Dihydropyridines

The substitution of the C–H moiety of a phenyl ring by nitrogen is a useful and well-established tactic for mitigating metabolic activation to chemically reactive and potentially toxic species that has found broad-based application and is captured in synoptic form by several illustrative examples [362–367]. The 4,7-dimethoxy indole derivative 252 is a potent HIV-1 attachment inhibitor, but the observation of O-demethylation as a metabolic pathway in HLM gave rise to the concern that the chemically electrophilic quinone 253 could be formed in vivo [362]. The 4,7-dimethoxy-6-azaindole 254 analogue largely preserved the antiviral activity of 252 but, in contrast, would afford the far less electrophilic 255 upon dealkylative metabolism, and it was this compound, designated BMS-488043, that was advanced into proof-of-concept clinical studies [362]. In addition, the mild basicity associated with the azaindole ring of 254 led to improved aqueous solubility.

The phenol 256 is a short-acting calcium-sensing receptor antagonist that was examined for its potential as a treatment for osteoporosis [363, 364]. However, CYP 3A4-mediated oxidation of the phenol ring of 256 afforded the catechol 257 which underwent further oxidation to the corresponding ortho quinone, a metabolite identified as the source of GSH adducts in both human and rat liver microsomes. The strategic introduction of a nitrogen atom into the phenol ring of 256 to give 258 was anticipated to reduce the rate of metabolic activation of the catechol based on quantum chemistry calculations which indicated that oxidation of the aza-catechol derived from 258 to the corresponding quinone was energetically less favorable than for 257. This was confirmed experimentally with a 50-fold reduction in the formation of GSH adducts compared to 256 when 258 was incubated in liver microsomes [363, 364].

The CRF antagonist 259 was found to undergo metabolic activation following dosing to rats followed by trapping of the reactive species by GSH which afforded adducts amounting to 25% of the dose, with 260 identified as the major species produced in vivo [365–367]. This metabolic process was postulated to proceed via iminoquinone formation, providing a rationale for studying the effect of replacing the electron-rich phenyl ring with a pyridine, an approach that, after additional optimization, led to the identification of 261 as a compound with targeted biological and pharmacokinetic properties. As part of the structural refinement of 259, the pyrazinone chlorine substituent was replaced with an electron-withdrawing nitrile moiety in order to reduce metabolic activation of the olefin while the methoxy was introduced into the side chain to redirect metabolism to an alternative site that would provide benign products [365–367].

Another example of the beneficial effect of replacing a C atom with N to interfere with metabolism is provided by the bioisosteric replacement of the dihydropyridine (DHP) ring of the Ca²⁺ channel blocker class of smooth muscle relaxant exemplified by nifedipine (262, R=ortho-NO₂) with a pyrimidinone heterocycle [368–371]. The dihydropyridine ring is subject to rapid first-pass oxidative metabolism in vivo to give the corresponding pyridine 263 which is an inactive metabolite. This led to the design of the pyrimidinone 264 as a probe of the idea that this ring system would be an effective substitute of the DHP ring. The design concept was based on the premise that the amide NH of 264 would mimic the H-bond donor properties of the dihydropyridine NH of 262 while the second nitrogen atom of the ring would provide a site for the introduction of the important carbonyl moiety, initially examined in the context of the methoxycarbonyl derivative 264, as well as providing resistance toward facile oxidation of the heterocycle. However, as an acylated urea derivative, 264 suffered from chemical instability, necessitating replacement by the more robust ureido moiety found in 265. In this analogue, the orientation of the exocyclic carbonyl group is as depicted in 265, favored by both dipole–dipole interactions and an intramolecular H-bond that projects the R substituent in a vector compatible with vasodilatory activity [368–371].

5.7 Isosteres of Heterocycles to Reduce Metabolic Activation

The potent bradykinin B₁ antagonist 266, K _i = 11.8 nM, developed as a potential treatment for pain, was found to produce glutathione adducts when incubated in rat and human liver microsomal preparations, a metabolic pathway also observed in vivo following oral administration of the drug to rats with GSH adducts derived from the drug identified in bile [372]. The site of GSH adduction was determined to be the pyridine ring, characterized as 268 and hypothesized to arise either from the diiminoquinone 267 or the epoxide 270, while the pyridine N-oxide 269 was determined to be a minor contributory pathway.

The design of a suitable isostere of the diaminopyridine ring was based upon the assumption that both NHs were of importance while the pyridine N atom was considered to function as a H-bond acceptor, an analysis that anticipated the simple glycine derivative 271 as the most rudimentary mimic. In order to influence the conformation of 271 in a fashion that allowed mimicry of the topology of the ortho-disposed substituents of 266, the gem-dimethyl derivative 272 was prepared, but this compound expressed only modest affinity for the bradykinin B₁ receptor, K _i = 3.5 μM [372]. Further refinement to the cyclopropyl analogue 273 led to a more than 50-fold increase in potency, attributed to conformational constraint due to π–π hyperconjugation between the cyclopropyl C–C bond and the amide C=O that favors the two conformations depicted in Fig. 38, with that represented by A mimicking the topology associated with 266. In addition, it was noted that the 116° bond angle associate with the cyclopropane ring substituents more closely matches the 120° vectors projected by the pyridine ring.

Fig. 38

Conformational topology and bond angles of spirocyclopropyl glycinamides

Amide and ester isosteres are susceptible to protease and/or esterase-mediated degradation in vivo while alkyl esters can be degraded by oxidative dealkylation by CYP 450 enzymes to afford the corresponding carboxylic acids, providing an impetus to identify surrogates with resistance to metabolic modification [373–377]. Pioneering studies in this area were focused on the design of heterocycles as ester replacements in the context of benzodiazepine derivatives of general structure 274 and 275 and muscarinic agonists based on the naturally occurring arecoline (276) [378–380]. Azole heterocycles are the most common amide and ester replacements that have been explored, captured in synoptic fashion in Fig. 39. This tactic remains an approach of contemporary interest although there can be marked differences between heterocycles in the ability to emulate carbonyl-based functionality dependent on context that may be attributed to subtle effects associated with the underlying electronic properties of a heterocycle that are not always understood [2, 3, 9, 10, 71, 381–387].

Fig. 39

Azole isosteres of esters and amides

5.8 Isosteres of the NO₂ Moiety

The unique properties of the nitro group have often made this a difficult structural motif to emulate by isosteric replacement [388, 389]. For example, the nitrophenyl moiety of 277, an inhibitor of the protein–protein interaction between murine double minute 2 (MDM2) and p53, offers the optimal potency from an extensive survey of potential substitutes [389]. These observations were suggested to be a function of a lipophilic and strongly directional interaction between the nitro group and the MDM2 protein that is also dependent on the electron-withdrawing properties. Where the latter effect is of importance for drug–target interactions, several electron-withdrawing functional groups can substitute for the nitro, including pyridine, pyridine N-oxide, sulfonyl, trifluoromethyl, amide, and a carboxylic acid moiety that is considered to be an exact isostere [104, 390–395]. Intermolecular interactions for the nitro moiety as collected from the Cambridge Structural Database of small molecules are summarized in Fig. 40 [387].

Fig. 40

Intermolecular interactions of nitro groups from the Cambridge Structural Database

6 Epilogue

The design of bioisosteres is a powerful and effective concept in drug design that has been applied to solve a wide range of problems encountered in drug discovery campaigns. This chapter has focused on the practical utility of applying bioisosteric substitution to solve contemporary developability challenges that are encountered almost routinely by the practicing medicinal chemistry. The creativity and ingenuity of medicinal chemists is quite clear from this synopsis which captures many insightful and elegant examples of drug design. The most successful and creative designs are frequently based on a careful analysis of not only the chemistry underlying a particular problem but also a deep and detailed understanding of the physicochemical properties of the atoms and structural elements that are selected to address an issue. While medicinal chemists have developed a broad toolbox of useful structural elements to draw upon to emulate a range of commonly encountered functionalities, it is anticipated that the challenges in drug design that lie ahead will continue to provide a stimulus for creativity that will expand the range of bioisosteric replacements.