Optimization of Inhibitors of Mycobacterium tuberculosis Pantothenate Synthetase Based on Group Efficiency Analysis

Abstract Ligand efficiency has proven to be a valuable concept for optimization of leads in the early stages of drug design. Taking this one step further, group efficiency (GE) evaluates the binding efficiency of each appendage of a molecule, further fine‐tuning the drug design process. Here, GE analysis is used to systematically improve the potency of inhibitors of Mycobacterium tuberculosis pantothenate synthetase, an important target in tuberculosis therapy. Binding efficiencies were found to be distributed unevenly within a lead molecule derived using a fragment‐based approach. Substitution of the less efficient parts of the molecule allowed systematic development of more potent compounds. This method of dissecting and analyzing different groups within a molecule offers a rational and general way of carrying out lead optimization, with potential broad application within drug discovery.

Ligand efficiency has proven to be av aluablec oncept for optimizationofl eads in the early stages of drugd esign.T aking this one step further, group efficiency (GE) evaluates the binding efficiency of each appendage of am olecule, furtherf ine-tuning the drug design process. Here, GE analysisi sused to systematically improve the potency of inhibitors of Mycobacterium tuberculosis pantothenate synthetase, an important target in tuberculosis therapy. Binding efficiencies were found to be distributed unevenly within al ead molecule derived using af ragmentbased approach. Substitution of the less efficient parts of the molecule allowed systematic development of more potent compounds. This method of dissecting and analyzing different groups within am olecule offers ar ational and generalw ay of carrying out lead optimization,w ith potential broad application within drug discovery.
The concept of ligand efficiency (LE) [1][2][3] has been used as an important guidingp rinciple for lead optimization in early stage drug design. Subsequently, the idea of group efficiency( GE) [4] was proposed, andt his has been applied to identify hotpots on proteins and to analyze parts of al igand that makei mportant contributions to binding. [5][6][7][8][9] Herein, we describe the use of GE to optimize af ragment-derived inhibitor of Mycobacterium tuberculosis pantothenate synthetase, an attractive target for developing new drugs against tuberculosis. [10][11][12] Pantothenate synthetase catalyzes the ATP-dependent formationo fa na mide bond between pantoate and b-alanine. [11][12] We have previously reported the identification of fragments 1 and 2 (see Schemes 1a nd 2) from biophysical screens using thermal shift and NMR methods. [13][14] The stepwise growingo fi ndole fragment 1 led to the generation of lead compound 5 (Scheme1;s ee also, Figure S1 in the Supporting Information). [13] In ap arallel study,l inking of fragments 1 and 2 afforded compounds 6-9 (Scheme 2; see also, Figure S2 in the SupportingI nformation). [13,15] Both fragment growinga nd linking approaches rapidly led to relativelyp otent inhibitors against pantothenate synthetase (5: K D = 1.5 mm and 9: K D = 0.9 mm). Based solely on the K D values and LE data obtained for compounds 5 and 8,i tw as not obvious how to identify areas and vectorsf or furthero ptimization of the compounds. Therefore, in an effort to improve binding potencyw hilem aintaining LE, [16][17] aG Ea pproachw as used to analyze compounds 5 and 8.F ollowing as imilar Free-Wilson analysis [18] as proposed by Saxty and co-workers, [6] compounds 5 and 8 were dissected into component parts, and the binding contributions (DDG)f rom these individual building blocks calculated. These data are summarized in Figure1 (for detailed calculations of GE, see the section entitled "Calculations for GE analysis"i nt he SupportingI nformation). This GE analysis around 5 and 8 quickly revealed inefficient bindingc omponents within the molecules, suggesting straightforward approaches to fragment modificationsw ithout indiscriminately increasing the inhibitor size.
As shown in Figure 1, the majority of the binding energy resides in the originali ndole fragment (GE = 0.75). As imilar observation has been observed in other fragment elaboration Scheme1.Afragment-growinga pproach applied against Mycobacterium tuberculosis pantothenate synthetase, generating lead compound 5.  strategies [6] and is mainly due to the high intrinsic binding energies required for fragments to be detected (DG = 4.2 kcal mol À1 [19] is added to the binding of indole group 1 to compensate for the free energy associated with loss of fragment rigidbody entropyd uring binding). Similarly,t he charged acetate side chain in both 5 and 8 also contributes significantly to the overall binding,p ossessing high GE valueso f0 .43 and 0.35, respectively.C onversely,G E analysish ighlights the limited contribution to binding of the acyl sulfonamide, methyl pyridine andt he benzofuran groups, all of which have calculated GE values of 0.16-0.17 (the GE of the sulfamoylg roup is based on the N-(methylsulfonyl)acetamide group).
Previousw orko nf ragment growing and linking [13] have shownt hatt he acyl sulfonamideg roups not onlyc ontributed to additional bindinge nergy by forming additional hydrogen bonds between the sulfoneo xygen and botht he backbone amide of Met 40 and the side chain of His 47,b ut morei mportantly alsos erved as an effective functionalg roup for directing the correct vectors towards both the P1 and P2 pockets without clashingw itht he side of the active siteformed by Met 40 (Figure S4 in the SupportingI nformation). Based on theseo bservations, the initialo ptimization strategyw as focused on replacemento ft he methyl pyridine and the benzofuran groups rather thant he acyls ulfonamidel inker in bothcompounds 5 and 8.
The binding of compounds 10-19 against M. tuberculosis pantothenate synthetase was determined by isothermal titration calorimetry (ITC), and the structure-affinityr elationship (SAR) results are summarized in Ta ble 1; ITC binding data for all compounds are presented in the Supporting Information). Replacing the methylp yridine/benzofuran groups in 5 and 8 generated as eries of sub-micromolar inhibitors (10)(11)(12)(13)(14).The substitution of the methyl pyridine ring (5)b yamore electronrich tolueneg roup (10: K D = 340 nm,L E= 0.32) resulted in approximately af ourfold improvementi na ffinity towards the enzyme. Compound 10 was the most ligand efficient compound tested.T he GE value for the tolueneg roup in 10 is Scheme2.Afragment-linking approach applied against Mycobacterium tuberculosis pantothenate synthetase generating lead compounds 6-9.(X-Y-Z represents the approximate three-atom length of the linker.) Figure 1. A) Groupefficiency (GE) analysisofcompound 5 estimatesthe contributionsofthe binding efficiencies from differentfunctional groups andquickly reveals inefficient binding groups in themoleculefor further optimization of potency.B)Asimilar GE analysisapplied towardscompound 8.*DDG = DGÀDG rigid , DG rigid = 4.2kcalmol À1 . [19] Thecross-sectional view of theX-ray crystalstructure of theactivepocketofpantothenatesynthetaseisshown in green with inhibitors 5 and 8 bound. (Thecross-sectional view wasg enerated by removing residues fromone half of theactivepocketofpantothenatesynthetaseusing theDSvisualizer software. Thes urface on theother halfofthe protein wasg enerated using PyMol v.0.99 [20] ). DG values of thecompounds were determined from titration experimentsusing ITC. TheG Evalue is subsequently calculatedbydividingthe DG contributionfromeachgroup by thenumberofheavy atoms in thegroup. Gratifyingly,t he addition of ab ulkier and more electronegative trifluromethyl group to the indole sulfonamidec ore gave rise to 11,t he most potent compound of this series (K D = 200 nm,L E = 0.30, GE = 0.27). Likewise, the installation of hydrophobic andl arger groups like tert-butylbenzene (12)a nd naphthalene (13)g ave rise to compounds that bind to pantothenates ynthetase with improved affinity (K D = 460 nm and 610 nm,r espectively)c omparedw ith parent compounds 5 and 8.I nterestingly,t he more hydrophilic indole acyl sulfonamides 15-19 (cLog P = 1.0-3.2) bind with lower affinities (K D = 2-17 mm)c omparedw ith more lipophilicc ompounds 10-14 (cLog P = 3.6-5.0). This step of the optimization process was achieved in large part based on GE and SAR analysisw ithoutm uch consideration of structure. However, it dida ssume that the new compoundsb ind at the active site of pantothenate synthetase in as imilar way to the originall ead compounds, 5 and 8.I no rder to establish this, the structures of the four most potent inhibi-tors (10-13)b ound to the enzymew ere solved using protein X-ray crystallography ( Figure 2). The X-ray crystal structures of 10-13 bound to pantothenate synthetase show binding at the active site, with ac onserved binding mode for the indole sulfonamide fragment core. Less obviously,t he substituted groups on all four compounds were seen to bind in the P1 pocket of the enzyme (see Figure 1B). The P1 pocket binds the alkyl groups of the pantoate substrate and is primarily lipophilic, surrounded by the hydrophobic residues Pro 38, Met 40, Val143, Leu 146 and Phe 157 ( Figure S5 in the Supporting Information). In contrast, the P2 site binds the phosphates of ATPa nd is relatively hydrophilic. As can be seen in Figure 2, the bindingo rientationso ft he added groupsa re all similar,a nd no new hydrogen bondsa re formed. The detailed binding interactions of the most potent compound (11) with the P1 pocket residues are shown in Figure S5 in the Supporting Information. In addition to binding assays and X-ray crystallography studies, an inhibition study was carriedo ut that demonstrated that compound 11 inhibits pantothenate synthetase with an IC 50 value of 5.7 mm (see the Supporting Information). The structural data on compounds 10-13 provided the impetus for further elaboration of the series, with av iew to making ac ompound that probes more deeplyi nto the P1 site. It was rationalized that the introduction of am ethylene group between the aromatic and sulfonyl groups shoulda llow the aromatic group to slide below Met 40 and push a para substituent to the back of the P1 pocket ( Figure 3; for detailed binding interactions of 11 with the P1 pocket, see also Figure S5 in the Supporting Information). To test this hypothesis, compound 20 was synthesized, using at rifluoromethyl-substitutedbenzylsulfonamide as an ew coupling substrate.
The X-ray crystal structure of 20 bound to M. tuberculosis pantothenates ynthetase (Figure 3) showedt he hoped for binding with the trifluoromethyl group picking up favorable hydrophobic interactions with Val139, Val142 and Val143. Compound 20 was shown to inhibit the enzymaticr eaction withasignificantly improved IC 50 value of 250 nm as compared with 11. Furthermore, ac ell-based assay against M. tuberculosis showed on-target inhibitory activity leadingtoc ell death. [21] As the use of fragment-based methods expands, the need for subsequentl ead optimization of fragment-derived compounds becomes increasingly important. The work presented here demonstrates the use of GE analysis to critically and thoroughly examinet he binding distribution of al ead compound and illustrates the practicality of applying GE analysist om odify parts of am olecule that are not making efficient contributions to binding.I nt his case, it led to the generation of ar elatively

Experimental Section
Syntheses and characterization of organic molecules, biochemical, X-ray crystallography and isothermal titration calorimetry methods are described in the Supporting Information. Additionally,N MR spectra related to this publication are also available at the University of Cambridge data repository (www.repository.cam.ac.uk/ handle/1810/249084).