Finding Fragments Thymidylate Synthase Proof of Principle

We first applied Tethering to thymidylate synthase (TS). This enzyme converts de-oxyuridine monophosphate (dUMP) to thymidine monophosphate (dTMP), an activity essential for DNA synthesis. The cancer drug 5-fluorouracil irreversibly inhibits TS, and a selective inhibitor of a non-human form of the enzyme could yield a new antibiotic or antifungal drug [23].

In addition to its biological interest, TS was ideally suited for developing Tethering [11]. It is well characterized both structurally and mechanistically, and the many inhibitors developed for the enzyme demonstrate that it is a "druggable" target. Moreover, the active site contains a nucleophilic cysteine residue. Although the Escherichia coli version of the enzyme we used contains four other cys-teine residues, crystallography revealed these to be largely non-surface exposed, and they did not interfere with our experiments.

Initial experiments screened pools of ten compounds, each present in roughly ten-fold excess over TS, with a total disulfide concentration of about 2 mM and a reducing agent (2-mercaptoethanol) concentration of 1 mM. After screening about 1200 compounds, we saw a strong selection for N-phenyl-sulfonamide-substituted proline fragments, as represented by N-tosyl-D-proline (Fig. 9.4). In a separate experiment, this fragment could even be selected from a pool of 100 compounds, each present at roughly the same concentration as TS. However, larger pools have more compounds with similar molecular weights, making data more challenging to interpret. In practice, pools of five to ten compounds strike a balance between throughput and unambiguous interpretation.

9.3 Finding Fragments: Thymidylate Synthase Proof of Principle 311

9.3 Finding Fragments: Thymidylate Synthase Proof of Principle 311

Fig. 9.4 Improvements in potency of N-tosyl-D-proline. Structural analyses revealed that the glutamate moiety from the mTHF cofactor could be appended to the hit from Tethering, and further elaboration led to a submicromolar inhibitor.

A critical feature of Tethering is that thermodynamics govern disulfide bond formation. To ensure that fragment selection was thermodynamic rather than kinetic, we added a reducing agent (2-mercaptoethanol). Without reducing agent, the active-site cysteine reacts with whichever disulfide it encounters first, usually the solubilizing element common to all library members. Although even a small amount of reducing agent allows disulfide exchange, the N-tosyl-D-proline fragment could tolerate strongly reducing conditions. In fact, even in the presence of

Fig. 9.5 Structures of TS with the N-tosyl-D-proline fragment bound through two different cysteine residues (red, blue) or non-covalently bound (green). Reprinted from [12] with permission.

20 mM of 2-mercaptoethanol, where the ratio of reductant to disulfide was 10:1, a mass corresponding to N-tosyl-D-proline conjugation was still prominent.

Screens with chemically similar fragments showed that although substitutions around the aromatic moiety and in the stereochemistry of the proline residue did not disrupt the fragment's affinity, the proline residue itself was essential. Crystallography of N-tosyl-D-proline covalently linked to TS explained these structure-activity relationships (SAR): the proline residue sits snugly within a hydrophobic pocket, and one of the sulfonamide oxygen atoms makes a hydrogen bond to Asn 177 on the enzyme, but the phenyl ring is in a relatively open area (Fig. 9.5).

To learn whether the disulfide bond itself changed how the fragment binds, we determined the crystal structure of N-tosyl-D-proline bound non-covalently to TS. As shown in Fig. 9.5, the ''free'' fragment binds in a nearly identical manner to the disulfide-linked fragment, demonstrating that the covalent linkage does not affect how the fragment binds.

To test whether nearby cysteines would be suitable for Tethering, we mutated the active-site cysteine to a serine and introduced a new cysteine nearby (C146S, L143C). When we performed Tethering on this mutant enzyme, we also strongly selected N-tosyl-D-proline, and when we solved the X-ray crystal structure we found that this fragment binds in a manner very similar to the other structures, despite the very different trajectories that the disulfide linkage takes (Fig. 9.5). The lack of influence of the disulfide attachment on the fragment's binding mode, along with the fact that the fragment could be strongly selected from more than one cysteine residue, suggested the inherent fragment affinity was more important energetically than the specifics of how it was linked to the protein.

Enzymatic assays determined the inhibitory potential of N-tosyl-D-proline: the fragment has a Ki of 1.1 mM, so weak that it likely would be missed in any conventional screen. However, the crystal structure shows that the phenyl group binds in a similar position to the para-amino-benzoic acid moiety of the natural co-factor, methylenetetrahydrofolate (mTHF); by simply grafting the glutamate moiety from this co-factor onto N-tosyl-D-proline, we boosted the affinity 40-fold to 24 mM. A small library of compounds with substitutions off the proline yielded a compound with a Ki of 330 nM, three orders of magnitude more potent than the original fragment (Fig. 9.4). Overall, applying Tethering to TS demonstrated the capability of mass spectrometry to selectively discover weak, disulfide-containing fragments that were optimized into lead-like compounds.

Was this article helpful?

0 0

Post a comment