Structure-Guided Discovery of Potent Antifungals that Prevent Ras Signaling by Inhibiting Protein Farnesyltransferase

Abstract

Infections by fungal pathogens are difficult to treat due to a paucity of antifungals and emerging resistances. Next-generation antifungals therefore are needed urgently. We have developed compounds that prevent farnesylation of Cryptoccoccus neoformans Ras protein by inhibiting protein farnesyltransferase with 3-4 nanomolar affinities. Farnesylation directs Ras to the cell membrane and is required for infectivity of this lethal pathogenic fungus. Our high-affinity compounds inhibit fungal growth with 3-6 micromolar minimum inhibitory concentrations (MICs), 4- to 8-fold better than Fluconazole, an antifungal commonly used in the clinic. Compounds bound with distinct inhibition mechanisms at two alternative, partially overlapping binding sites, accessed via different inhibitor conformations. We showed that antifungal potency depends critically on the selected inhibition mechanism because this determines the efficacy of an inhibitor at low in vivo levels of enzyme and farnesyl substrate. We elucidated how chemical modifications of the antifungals encode desired inhibitor conformation and concomitant inhibitory mechanism.

Figure 1.

Structures. (A) The farnesyl diphosphate substrate. (B) The L-778,123 inhibitor (1) comprises four flexibly connected rings (A–D), with five freely rotatable bonds. Rings A–C fold to form a compact core. The D ring is located on the surface of that core and free to rotate around χ₂₃. Within the core, the A ring can fold with two orientations, by rotation around χ₁₂. The locations of the R¹-R⁴ sites that were modified in this study are indicated. (C) The CnFTase active site with bound FPP and L-778,123 (PDB 3Q7A). The B ring binds to the catalytic Zn²⁺ (magenta). The locations of the four rings and their modification sites relative to the peptide substrate binding-site and product exit groove are shown.

Figure 2.. The effects of the C325A mutation in the CnFTase β subunit on C. neoformans.

The ram1Δ deletion strain was transformed with either wild-type RAM1 or mutant ram1-C325A. (A) Thermotolerance. Each strain was incubated in five-fold serial dilutions on solid YPD medium at 25°C, 30°C, and 37°C. Growth was assessed by inspection after 48 hrs. (B) Mating. The two MATα strains (ram1Δ+RAM1 or ram1Δ+ram1-C325A) were incubated in mating co-cultures with an opposite mating type strain (KN99a – “MATa”) on MS mating medium. Mating competence was determined after 7 days by assessing the presence (wild-type, right) or absence (C325A mutant, left) of mating hyphae and sexual spore formation (inset).

Figure 3.

CnFTase is the in vivo target of 2f. (A) Ras cellular mislocalization, assessed by epifluorescent microscopy (1000X). Bar = 10 μm. Top. In an untreated C. neoformans strain GFP-Ras1 localizes predominantly at the plasma membrane (left); in the presence of 0.78 μM 2f (sub-inhibitory), the fusion protein remains in the cytoplasm (right). Bottom. Control patterns of Ras localization with wild-type (left) and ram1Δ deletion (right) strains expressing mCherry-Ras1. (B) Overexpression of Ram1 negates 2f antifungal activity. A C. neoformans strain transformed with the galactose-regulated pGAL7-RAM1 over-expression construct was incubated with glucose (uninduced, bottom), and galactose (induced, top). Antifungal activity was assessed by the relative size of the zone of fungal growth inhibition surrounding sterile discs impregnated with 10 μL of 20 mM 2f dissolved in DMSO (bottom disk) or DMSO alone (top disc).

Figure 4.

The effects of phosphate on the steady-state kinetics of CnFTase. (A–C) FPP steady-binding. (D–F) 2f inhibition. (G–I) 2q inhibition. (A) Michaelis-Menten curves fit (2, 3) to the initial rate dependence on FPP concentration (1) determined at six phosphate concentrations (0–50 mM; solid lines, fit; circles, observed initial rates). (D, G) Initial rates (1) measured at 1 uM FPP at six different phosphate concentrations (circles), fit (lines) to extract ^1μMIC₅₀ values (2, 3). (B, E, H) Extracted K_M and ^1μMIC₅₀ values were fit a two-state, single-site, isothermal phosphate-binding curves (2, 3) with phosphate dissociation constants varying 0.1–0.8 μM (lines, fit; circles, extracted values). For FFP (B) and 2f (E), the two states were modeled as constant α and β values in 2. For 2q (H) a linear post-transition baseline (4) was used for β in 2. (C, F, I) The extracted V_max values (C, circles) and inhibitor-free rates (F and I, circles) were fit (lines) to a two-state, single-site, isothermal phosphate-binding curve (2, 3) using linear post-transition baselines (4). The average mid-point of the effect of phosphate on affinities (B, E, H) is ~0.5 mM. (J) A model for the effects of phosphate binding on enzyme activity and inhibition. The enzyme comprises two conformations, L and H, that are in equilibrium with each other. In the absence of phosphate, L (blue). The enzyme has a binding site for phosphate (P) at a location outside the active site (i.e. allosteric), with higher affinity for H than L. Consequently, phosphate binding shifts the equilibrium to H (green). The L and H conformations differ in catalytic efficiency (V_max), FPP affinity (K_M), and the ^1μMIC₅₀ values of 2f and 2q. (K) The occupancies of binding sites for FPP , (black) inhibitor sites I (green) and II (blue) and two phosphate sites (orange: the pseudo-product inhibitor in the active site, ψP, and the allosteric site, aP). Circles indicate occupancy: open, empty; filled, bound ligand. Only one occupancy state corresponds to active enzyme. Inhibitor binding at I and II are mutually exclusive. Binding at I requires FPP occupancy (uncompetitive inhibition). Binding at II is mutually exclusive with FPP binding (competitive inhibition). Pseudo-product inhibition occurs only in the H state with occupied aP, because the affinity of aP is ~500-fold higher than ψP.

Figure 5.

Inhibition by 2f and 2q. Initial velocities, v_o, (circles) were extracted from progress curves (1) measured as a function of FPP (log scale) and different inhibitor concentrations (labeled, nM) at 0 mM (left column) or 10 mM (right column) phosphate, and fit (lines) to inhibition models (5) with binding polynomials for the different inhibition modes: 2f, mixed competitive and uncompetitive (8); 2q, uncompetitive (7). The values of the fit parameters and their error estimates were determined by bootstrapping (insets: distribution histograms of 10,000 trials); affinities are shown as their dissociation constants (nM): K_F, FPP affinity; K_C, competitive inhibitor affinity; K_U, uncompetitive inhibitor affinity). The 2f data set measured in the absence of phosphate was fit with ideal conditions (10); the other three required non-ideal treatment (11–13), because the total enzyme concentration, p_T, was ~10 nM (see insets), close to value of at least one dissociation constant so that the ideality condition (9) does not hold.

Figure 6.

Electron density maps of 2q and 2f bound to CnFTase. Simulated annealing F_o-F_c electron density difference maps (gray, contoured at 4.0 σ; only positive densities are shown) revealed bound compounds, FPP (if present), Zn²⁺ (magenta), and sulfate. (A) 2q binds in the presence of FPP (7T08). (B) 2f displaces FPP (7T0A).

Figure 7.

Structures of inhibitor complexes. (A) 2q (yellow) bound at site I (7T08) in an orientation similar to L-778,123 (light grey). R² butyl modification (dotted circle), present in most compounds, extends the contacts between the inhibitor and exit groove (red surface). The R¹-trifluoromethoxy group (dotted circle) forms halogen bonds (dotted lines) with terminal isoprene 3 of the FPP substrate, with the indole ring of W90β, and within the inhibitor to the R³ chloride. These interactions strengthen the interactions of the inhibitor both with the bound substrate and with the protein. The internal halogen bond reduces flexibility. (B) 2f (grey) bound at site II (7T0A). The R¹-trifluoromethoxy swaps halogen bonds with the bound FPP terminal isoprene 3 at site I for bonds with C201β and C272β. The R⁴ bromide intercalates between W329β and Y269β, and forms a third halogen bond with C272β. R² butyl group forms contacts in the exit groove. The sulfate occupies the bisphosphate-binding site, and corresponds to the putative location of the phosphate pseudo-product inhibitor. (C) and (D) Superposition of 2f (grey) and 2q (yellow) emphasize the differences in the locations of sites I and II, and the pivoting motion of the B ring around the catalytic Zn²⁺. If site II is occupied, the FPP substrate (outline in D) cannot bind. The R² butyl group binds to the exit groove in both locations, but is less deeply inserted at site II. (E) Superposition of 2f and 2q emphasizes the rotation of the D ring around χ₂₃.

Figure 8.

The effects of inhibition mode and concentrations of substrate and enzyme on inhibitor efficacy. IC₅₀ values were calculated (3) for competitive (6) and uncompetitive (7) inhibition of a single-substrate enzyme. Substrate (K_M) and inhibitor affinities for both inhibition modes (K_I) were set to 1 (concentration units are the same for parameters and variables, and therefore omitted). Enzyme concentrations are presented as copy number per cell. Given a cell diameter of 10 μm (approximate size of C. neoformans), the volume is $V_c=\frac{4}{3}\pi r^3=523\mu {\text{m}}^3=5\times {10}^{-13}\text{ L}$. The concentration of a single protein in that cell therefore is $\frac{1}{N_A}\times \frac{1}{V_c}=\frac{1}{6\times {10}^{23}}\times 5\times {10}^{-1}\approx {10}^{-1}=0.1\text{nM}$. (A) The effect of substrate concentration ([S]) on IC₅₀ values for competitive (blue) and uncompetitive (green) inhibition at vanishingly small enzyme concentrations where the ideal numerical conditions apply (9). With these simulation parameters, the critical substrate concentration, S*, where the superior efficacy of these inhibition modes interchange corresponds to the K_M value of the enzyme. In general, S* depends on the three-way relationship between K_M and the two inhibition mode affinities. (B) The effect of enzyme concentration ([E]) on competitive inhibition IC₅₀ values (red: low, favorable values; blue: high, unfavorable), simulated using solvers for non-ideal numerical conditions (11–13). The phase space has been partitioned into four quadrants, using substrate and enzyme concentrations corresponding to K_M as the dividing lines: −/− (low substrate, low enzyme), −/+ (low substrate, high enzyme), +/− (high substrate, low enzyme), and +/+ (high substrate, high enzyme). The −/+ and +/− quadrants behave as the ideal case shown in panel (A). In the −/+ quadrant, the increase in available binding sites enables the competitive and active occupancies (Figure 4K) to co-exist and rather than compete. Consequently, the inhibitor loses efficacy. In the +/+ quadrant the effects of increased protein and substrate concentration reinforce each other. Accordingly, competitive inhibition is most effective only at low substrate and protein concentrations. (C) The effect of enzyme concentration on uncompetitive inhibition IC₅₀ values. The −/+ and +/− quadrants also behave as the ideal case shown in panel (A). In the −/+ quadrant inhibitor efficacy improves because increased enzyme concentrations capture all free substrate. Given that the inhibitor binds only to the enzyme-substrate complex (Figure 4K), IC₅₀ values approach K_I, as is the case in the +/− quadrant. In the +/+ quadrant, high substrate concentrations enable competition between the active and uncompetitive occupancies (Figure 4K), thereby reducing inhibitor efficacy. Accordingly, uncompetitive inhibition is effective either at high substrate, or high enzyme concentrations. (D) Comparison of inhibitor effectiveness under plausible in vivo conditions. Contour plot of 2f IC₅₀ values (contour labels are μM) as a function of FPP concentration and enzyme copy number. The plot was generated by applying 5 with the mixed inhibition polynomial, Q_M (8), with affinities for the high-phosphate, H conformation (Figure 5), and using solvers for non-ideal conditions 10–13. The efficacy quadrants are indicated. (E) Contour plot of 2q IC₅₀ values under plausible in vivo conditions, using the uncompetitive polynomial, QU (7). (F) Ratio of IC₅₀(2q)/ IC₅₀(2f). Grey marks the region where 2f outperforms 2q at least two-fold in the joint −/− efficacy quadrant (it is asymmetrical because 2f has a mixed inhibition mode).

Scheme 1.

Synthesis of potent antifungal agent 2f.^a a Reagents and conditions: (a) m-CF₃O-C₆H₄-NH₂, NaBH(OAc)₃, 4Å molecular sieves, DCE, 20 °C, 56%; (b) ClCH₂COCl, aq NaHCO₃, CH₃CO₂C₂H₅, 0 °C; (c) Cs₂CO₃, DMF, rt, 76%; (d) i. Ph₃CCl, (C₂H₅)₃N, DMF, 20 °C, 75%, ii. (CH₃CO)₂O, pyridine, 20 °C, 80%; (e) i. ArCH₂Br, CH₃CO₂C₂H₅, 55 °C, ii. CH₃OH, 55 °C; 85%. (f) i. LiOH, THF/H₂O, 0 °C, 60%, ii. MnO₂, CH₂Cl₂/1,4-dioxane, 50 °C, 90%; (g) TrtCl, Et₃N, DMF, rt, 94%; (h) p-Br-C₆H₄-CH₂Br, CH₃CN, 60 °C, 1 d, 91%; (i) con HCl, CH₃OH, then aq NaHCO₃; (j) NaBH(OAc)₃, 4Å molecular sieves, DCE, 20 °C, 45%. Note: 9a: p-Br-C₆H₄-CH₂Br