Fragment-Based Lead Discovery Strategies in Antimicrobial Drug Discovery

Abstract

Fragment-based lead discovery (FBLD) is a powerful application for developing ligands as modulators of disease targets. This approach strategy involves identification of interactions between low-molecular weight compounds (100-300 Da) and their putative targets, often with low affinity (K_D ~0.1-1 mM) interactions. The focus of this screening methodology is to optimize and streamline identification of fragments with higher ligand efficiency (LE) than typical high-throughput screening. The focus of this review is on the last half decade of fragment-based drug discovery strategies that have been used for antimicrobial drug discovery.

Keywords: FBLD; STD NMR; antimicrobial; inhibitors of pathogenic bacterial enzymes; multidrug resistance; phenotypic screens.

Figure 1

Representatives of the estimated 500 natural and synthetic inhibitors of metallo-β-lactamases.

Figure 2

Fragments guide lead derivatization of NDM-1 based on the natural product aspergillomarasmine A (AMA). Iminodiacetic acid (IDA), a simplified analogue of AMA, was successfully derivatized to compound 2 as NDM-1 inhibitor with IC₅₀ value similar to that of AMA.

Figure 3

Fragments guiding lead derivatization of NDM-1 inhibitors derived from the natural product captopril making use of the de novo molecular design program SPROUT (a computer program for constrained structural generation). The synthesis of fragment 3 [27] is based on the analysis of a crystal structure of NDM-1 complexed with hydrolyzed ampicillin (PDB ID: 3Q6X) [30]. SPROUT [28] was used to identify active sites for in silico-generated fragments. The sites identified were adjacent to the following: the Lys224 side chain; the Zn-2 metal ion; the nucleophilic hydroxide/water that “bridges” the two zinc ions; and a conserved tryptophan (Trp87) that hydrophobically interacts with the aromatic ampicillin C6 side chain [31], crucial for binding of β-lactams to metallo β-lactamases. SAR also identified several other compounds with activity against NDM-1 in the submicromolar range (4–8) [27].

Figure 4

Fragments guide lead derivatization of fragment 9 as inhibitor of NDM-1 predicted by virtual screen of a library of 122,500 fragments. A combination of HTVS, SPR and NMR screening validated the synthesized derivatives of 9 ability to interact with NDM-1 [29].

Figure 5

Fragment 26 synthetic analogs. Fragment 26 inhibition of NDM-1 was predicted by virtual library screening (770,000 fragments). To identify and validate fragment interaction with NDM-1, a prepared library of 10 compounds, HTVS, NMR screening and Density Functional Theory (DFT) calculation, and biochemical assays were used [32]. The compounds with estimated optimal putative activity, i.e., Ki and per cent inhibition at 50 mg/mL were determined to be inhibitors of NDM-1 (29–33) and are shown here. Hit molecule 34 was identified through VS as NDM-1 inhibitor [32].

Figure 6

Representatives of hydroxamate-based advanced LpxC inhibitors.

Figure 7

Fragments guide lead 41 derivatization where at low nanomolar concentration (enzyme functional assay) the glycine moiety complexes with zinc. Unfortunately, it has poor antimicrobial activity. Fragments guide lead 40 derivatization can through an imidazole moiety chelate zinc. Imidazole derivative 46, obtained through structure-guided design, resulted in a 2-(1S-hydroxyethyl)- which exhibits inhibition of LpxC at low nanomolar concentrations. In addition, 46 exhibits antimicrobial activity (minimal albumin effects) against Pseudomonas aeruginosa at a minimum inhibitory concentration (MIC) of 4 μg/mL [42].

Figure 8

Maximal affinity binding with BoNT/F-LC as shown via SPR was for 51 (NSC1014) (K_D: 5.58 × 10^-6). The IC₅₀ of 50 (NSC1011), 51 (NSC1014), and 52 (NSC84094) were 30.47, 14.91 and 17.39 μM, respectively (endopeptidase assay). Survival times (mouse model) were extended by 50 (NSC1011) and 51 (NSC1014).

Figure 9

Fragments identified as N-acetyglucosaminidase inhibitors. Fragment 53 being the top virtual hit, validated by SPR, with K_D of 228 μM for the AtlE [53].

Figure 10

Illustrative examples of small heterocyclic inhibitors of MurA_EC. They inhibit enzyme activity by forming a covalent bond with cysteine in the enzyme active site.

Figure 11

Sulfhydryl active fragments with S_NAr have been identified as inhibitors of the active site cysteine of MurA enzymes from S. aureus (Mur_SA) and E. coli (Mur_EC). The fragments, with fragment 61 being the top hit, have been investigated by glutathione stability HPLC and NMR assays, followed by validated by MurA inhibitory assay.

Figure 12

Illustrative examples of chloroacetamide-based covalent inhibitors for MurA from E. coli (Mur_EC) identified by LC-MS/MS and biochemical assays. Fragment 64 was identified as the most effective inhibitor that binds to MurA Cys115. Inhibition of MurA_EC by these fragments appears not to be associated with their thiol reactivity, as per evaluation of fragments’ electrophilicity using the Ellman reagent (reduced) as a proxy for cysteine thiol. This suggests the possibility for their further development as inhibitors of the MurA enzymes.

Figure 13

AstraZeneca PPAT inhibitors (71, 72), and from Novartis lead compounds (72, 74), and fragment hit 75. When fragments co-crystalized with E. coli PPAT were examined, 75, a methoxy tryptamine derivative, partly overlapped with 73, although its interaction is different from fragment 73. Fragments 73 and 74 served the basis for development of more than 50 analogs based on with improved on-target potency, including 76, piperidine carbamate. Analysis of these analogs verified that compounds with an AZ benzimidazole core exhibited an enhanced ability to permeate Gram-negative bacteria. These findings eventually resulted in discovery of compounds, e.g., 77, with anti-E. coli WT activity, which was confirmed by multiple methodologies (biochemical, SPR, and MICs). Regrettably, further progression of this series was halted due to bacterial efflux actions [67].

Figure 14

Structures of the best hits after ¹⁹F NMR screening and validated by SPR with above average of 9% in ¹H-¹⁵N TROSY NMR from set of 39 fragments. Fragments 78, 79, and 80 demonstrated binding to the secondary binding sites of BambL. The SAR of 78 demonstrated binding pocket identity predicted by computational study of fragment 78 scaffold that is responsible for the binding [70].

Figure 15

Fragment 86 was identified by NMR and X-ray crystallography as an initial analog for scaffold building. The synthesis started from C-2 to access a more polar region of the binding groove. Confirmation of the appropriateness of C-2 as a starting area for binding pocket access was validated by crystal structural analysis of 87, 88, and 89. Subsequent studies using HSQC showed that relative to parent compounds, C-2 analogues exhibited enhanced K_D values. In addition, C-2 analogues inhibited DsbA. This also indicates that there is the potential for development of benzofuran analogues, which could target virulence [73].

Figure 16

Structures of fragment 90 analogues screened by ¹H-¹⁵N HSQC NMR. To elucidate the differences in ligand binding of fragment 91 to VcDsbA vs. EcDsbA, crystal structures were obtained of each, complexed with sodium taurocholate, a known ligand. Crystallography showed that sodium taurocholate exhibits binding orientations in complex with VcDsbA that differ from that which occurs with EcDsbA. In addition, the protein–ligand interactions resulting in stabilization of binding orientations are revealed [76].

Figure 17

Optimal starting point identification using thermodynamic profiling of fragment sized PqsR ligands. A re-evaluation of the 98, 99, and 100 fragment scaffolds showed that 100 has remarkably improved enthalpic efficiency (EE) and ligand efficiency (LE) values. Hit 100 led to 104 optimization resulting in 106, which enabled further optimization through flexible fragment growing. The N-propyl amine linker in 107 shows an increase in potency (20-fold). Crystallography of 107 in complex with PqsR91-319 reveals the extended linker pointed further into the pocket containing the alkyl chain of the natural ligand.

Figure 18

Fragment 108 demonstrated the highest ligand efficiency (LE). Fragment 108 with MabPurC revealed an interaction at the 6-postion of the pyrimidine in the “ribose binding pocket”. This binding position appeared to be tolerated since it also was functional 110 binding. Fragment expansion here permits the addition of the pyridine moiety of fragment 109 into fragment 108. However, the introduction of a flexible linker was necessary for linkage of the fragments since they were almost perpendicular to each other. Synthesis of a library of a dozen compounds, where different linkers were explored, led to identification of compound 112 having the best binding affinity and LE [89].

Figure 19

Inhibitor 113 of TrmD of Gram-positive, Gram-negative bacteria, and Mtb developed using P. aeruginosa and Mtb TrmD crystal structures with a series of thienopyrimidinone derivatives identified through high-throughput screening with nanomolar potency against TrmD [97]. Compound 114 was developed when FBLD approach to develop selective inhibitors was applied against H. influenzae TrmD [99].

Figure 20

Fragment hits for MabTrmD. Elaboration on fragment 117 resulted in 118 (K_D 27 nM, LE 0.34), a low-nanomolar affinity compound with anti-MabTrmD activity. Screening of merged fragments 115 and 116 (e.g., 118) against Mab and Mtb showed promising MIC values. Compound 119 had the best MIC values against Mtb and Mab of 1.6 μM and 2.3 μM in supplemented 7H9 medium [91]. This series of compounds had activity against mycobacteria in vitro and in vivo [97].

Figure 21

Fragment-linking strategy led to synthesis of 18 compounds based on fragments 120 and 121. Compound 122 demonstrated markedly improved MthIMPDH ΔCBS inhibition (LE 0.29; 0.52 μM IC₅₀), which compared to 120 and 121 is significantly more potent (1300-fold); compound 121 X-ray structure with shown MthIMPDH at lower right. The racemate 122 has ~50% the Mth IMPDH ΔCBS inhibition as compared to its (S)-isomer, a pattern similar to that previously reported for other IMPDH inhibitors [109,110,111]. However, anti-Mtb H37Rv activity of the most potent analogues (0–100 μM tested) lacked clinically relevant activity (MIC₉₀ ≥ 50 μM). This lack of activity may be the result of poor cell permeability, metabolic instability, and/or efflux [109]. Additional information about FBLD developments in the search for anti-Mtb compounds can be found in these recent reviews [112,113].

Figure 22

The intermediate product Thr-AMP of ThrRS. Representative inhibitors of ThrRS in the nanomolar range.

Figure 23

Halofuginone (HF), an inhibitor of the plant analog of ThrRS, the prolyl-tRNA synthetase (ProRS). Fragments 123 and 124 are the best representatives of a series of synthetic analogs of HF. Compound 125 demonstrated antibacterial activity (MIC 16 mg/mL) against E. coli and Salmonella enterica [115].

Figure 24

Hits chosen for further optimization based on the STD-NMR and their binding affinities (K_D values; ¹⁵N^–1H HSQC NMR titration).

Figure 25

Best fragment hits against P. falciparum, 132 from a library of 1604 fragments and 133 and 134 from Astra Zeneca HTS screened 500,000 fragment corporate library. Tetrahydro-β-carboline compounds 135 and 136 from a 1604 fragment library against N. meningitidis.