Information about the minimum energy docked scores for the known fungicides is given in Table 4
and more in Supplementary Table S7
. Here, the evaluation of the docked structures for each fungicide was made according to (a) the minimum Gibbs free energy (ΔG
) conformation of all 100 conformations considered (in the latter text ΔG
(100)), (b) the minimum energy conformation for the subset of conformations where the distance Fe-N < 3 Å (in the latter text ΔG
(<3Å)), and (c) the maximum angle (Fe-N-(R)C angle, Figure 1
) conformation for the subset of conformations with a distance Fe-N < 3Å (with binding energy given in Table S7
(max α)). Regarding these three energies, the following relation is valid: ΔG
(100) ≤ ΔG
(<3Å) ≤ ΔG
Besides taking into account the nitrogen vicinity and angle, we took into account which nitrogen was coordinated, with the accent to specific nitrogen, which for imidazoles was N3 imidazole nitrogen, for triazole was N4 triazole nitrogen, and for tetrazole was N4 position nitrogen, which is already established to form a coordinated bond with protoporphyrin iron [11
]. The selection of the most favorable Fe-N-(R)C angle is a priority for the QM/MM minimization for a structure to converge with an Fe-N bond, since it is necessary to establish an as good as possible geometry input overlap between the in-plane azole nitrogen orbital of the donating free electron pair from nitrogen and the electron-accepting free d orbital from iron. The simple use of minimum energy conformation among all 100 runs does not make too much sense for the best QM/MM geometry input, because the scores of the Autodock4 program do not take into account the quantum mechanical and polarization effects of the iron atom.
In comparison with the literature, the ΔG
(100) for the most frequent fungicides (posaconazole, ketoconazole, miconazole, fluconazole, voriconazole, clotrimazole, itraconazole, and VT1161) is correlated with the experimental inhibitory effect of these drugs on C. albicans
Cyp51 (values taken from Reference [11
]) with Pearson′s R
of 0.742. In other publications [29
], between clotrimazole, fluconazole, itraconazole, ketoconazole, and voriconazole, fluconazole had the weakest experimental binding for both C. albicans
and human Cyp51, as was determined also with Autodock4 in this study. In the same study [29
], itraconazole and ketoconazole had the lowest dissociation constant, Kd
, for human Cyp51, while only voriconazole had a lower Kd
for C. albicans
Cyp51 in this study, while Autodock4 ketoconazole and itraconazole had the strongest bindings to C. albicans
Cyp51 among these five considered antifungals. The Kd
for ketoconazole for C. albicans
Cyp51 was determined to be ≈12 nM, while for propiconazole it was ≈38 nM [29
]. Here, for propiconazole and 10 ga runs, the Kd
was 46 nM. Finally, attempts to dock ligands to the Cyp51 protein have already been made [30
] where the proper orientation of the azole nitrogen has been criticized. This means that geometry imperfections due to, e.g., the angle between triazole/imidazole nitrogen and Fe (which we have addressed), can be expected. In Reference [30
] for zebrafish Cyp51, Autodock yielded binding energies of −10.7, −9.7, and −7.4 kcal/mol µM for lanosterol, ketoconazole, and propiconazole, respectively. We calculated a binding energy of lanosterol to the C. albicans
Cyp51 protein of −11.71 kcal/mol. Among 100 conformations, it did not contain any conformation which included any bond with iron atoms, but that was as expected [31
] since the substrate binds differently than inhibitors (antifungals); inhibitors should bind to iron, contrary to the substrate, which may not be the case due to the oxygen mechanism already described [31
]. In conclusion, our results obtained with Autodock4 for 5tz1 seem to be comparable with the literature and might be used for a rough assessment of the binding affinity of new hit compounds.
In addition, in this study we established a high correlation for 64 docked antifungals between 5tz1 and 3ld6 (R2 = 0.917), 5tz1 and 5eqb (R2 = 0.905), and 5tz1 and 5eab (R2 = 0.921).
For the ligands established to contain a bond with Fe other than with nitrogen (i.e., other than Fe-N), we give their corresponding binding affinities: aliconazole Fe-Cl with −9.68, azaconazole Fe-Cl with −7.89, quinconazole Fe-Cl with −9.30, pyrifenox Fe-O with −9.16, pyrisoxazole Fe-Cl with −9.12, terconazole Fe-Cl with −10.84, triadimefon Fe-O with −8.03, triflumizole Fe-F with −6.95, triforine Fe-O with −5.92, uniconazole Fe-O with −7.82, uniconazole-P Fe-O with −7.63 (all values are in kcal/mol, respectively).
Now, all these results have been taken into account and evaluated to determine the filtering criteria for the hit DB compounds. The most important result is for the 5tz1 crystallographic ligand oteseconazole (VT1161), with 10 ga runs and ΔG
(10) = −10.10 kcal/mol. This will be the prefiltering energy criterion in Scheme 1
for hit compounds and is a little bit more strict than average in Table 4
(−9.43) for N-Fe coordinating antifungals, but is slightly lower than the ΔG(100) for oteseconazole. The results of such prefiltering among the 127 G1 hit compounds (111 + 16, i.e., 16 for approach II) are shown in Supplementary Table S8
(regarding only the ΔG(10) column).
Now, in order to pass to the QM/MM simulations, the hit compounds have to satisfy condition 5a with 100 ga runs in Scheme 1
(at least one of the mentioned criteria). Thus, a hit compound has to contain either N close to Fe (<3Å) with a Fe-N-(R)C angle >140 degrees or N close to Fe (<3Å) with at least two conformations of the Fe-N-(R)C angle >120 degrees. The reasons for these criteria are the following: The average angle value in Table 4
is 150.8° and setting that angle cut-off would be too strict, since even the most potent fungicides like posaconazole or itraconazole would not meet that criterion, so it should be set lower than that. The first round number (in 10°) would be 140° to meet posaconazole and itraconazole. If a Fe-N-®C angle >140 is set then exactly 75% of fungicides (all except the lower quartile) in Table 4
obey this criterion. Still, some important fungicides might not meet this, but if more conformations (i.e., at least two conformations) with the N atom among the 100 conformations are close to Fe with a minimum angle of 120°, we flexibly consider such hit compounds while taking the conformation with the highest possible angle. In addition, we allow other conformations than specific azole nitrogen bound to Fe, but only for oxygen (Cl, F, I, or even S are simply too weak ligands to consider for QM/MM simulations). So, if the hit compound does not have any conformation of N with Fe (<3Å), it at least must have oxygen within 3 Å, but such a conformation must be strong. Thus, we set an alternative additional rule for non-nitrogen, but with dist(Fe-O < 3Å). As in the prefiltering, we have already established −10.1 kcal/mol; we raise the bar in absolute terms by at least −1 kcal/mol, obtaining a cut-off value of −11.1 kcal/mol (Scheme 1
) for Fe-O < 3Å.
From these results (Table 5
), the compounds that have finally passed to the last QM/MM stage are those for which the column “For QM/MM” is “yes”; all others failed to pass. More detailed results of this docking stage 5a are presented in Supplementary Table S8
The root-mean-square deviation (RMSD) for the PD144418 ligand between the crystallographic and minimum energy docked structure using Autodock4 is 1.32 Å.
The results of all the Erg2 ligands and literature-important ligands for the 5hk1 protein in Autodock4 (minimum docked energy conformation in kcal/mol) are given for 10 genetic algorithm runs (runs accompanied subsequently with additional analyses with 100 ga are given in brackets): (Erg2 ligands) aldimorph −9.00 (−9.01), dodemorph −10.99 (−10.92), fenpropidin −9.9 (−10.07), fenpropimorph −10.57 (−10.71), tridemorph −8.56 (−9.20), piperalin −9.76 (−9.89), and spiroxamine −9.68 (−9.77). Those literature-important for the σ1 receptor are 4-IBP −12.03 (−12.06), amiodarone −10.83 (−10.37), clomiphene −11.78 (−11.25), emopamil −10.7 (−11.15), fecosterol −11.32 (−11.56), haloperidol −11.77 (−12.01), ifenprodil −11.03 (−11.09), NE-100 −9.33 (−9.89), opipramol −12.87 (−12.69), PD144418 −10.6 (−10.61), pentazocine −8.15 (−8.12), sufentanil −9.5 (−9.80), tamoxifene −11.9 (−11.11), and triparanol −11.74 (−11.83). Among all the considered ligands, PD144418 is a crystallographic ligand of 5hk1 and its score was used as a prefiltering value for the evaluation of hit compounds. The cut-off was used with 100 ga runs but with the top-rated non-hit ligand, which, among all the considered literature ligands, is opipramol (−12.69 kcal/mol).
The full results for all the considered G2 hits docked to 5hk1 are in Autodock4 in Supplementary Table S9
. Here, we present only the important results, i.e., only those hit compounds with a minimum docked energy conformation in kcal/mol using 100 ga runs of score with an absolute value higher than the cut-off for opipramol (−12.69), which are DB06555 (−14.22), DB08746 (−13.25), DB08622 (−13.18), DB00637 (−13.64), DB02491 (−14.19), DB07075 (−15.55), and DB12345 (−13.08) (Table S9
). Only these compounds passed the Autodock4 5b condition (see also Table S9
For ten literature important ligands—opipramol, tamoxifen, haloperidol, triparanol, ifenprodil, amiodarone, emopamil, fenpropimorph, pentazocine, and fecosterol—we established a positive Pearson’s correlation, R
, of +0.864 between the Autodock4 Kd
values determined in this study (when the binding energy in kcal/mol is expressed in Kd
values) and the estimated pharmacophore Kd
values of the same ligands with the Erg2 protein in the work of C. Laggner et al. [20
]. This is a promising result. However, regarding fenpropimorph and tridemorph, the literature [32
] in contrast to our results for Adt4 predicts a much stronger binding to both the Erg2 and σ1 receptor proteins. For tridemorph with 100 ga runs, Autodock4 predicted a Kd
of 144 nM, while for fenpropimorph it predicted 17.5 nM. On the other hand, for tridemorph the experiment determined a Kd
of 0.09 nM for Erg2 and 0.04 for σ1. Meanwhile, for fenpropimorph, it determined a Kd
of 0.05 nM for Erg2 and only 0.011 nM for the σ1 receptor, which is a considerable difference. The difference between the literature and our results can also be seen with tamoxifen; here, we obtained a Kd
of 1.85 nM, compared to the Erg2 value of 1470 nM and the σ1 receptor value of 34 nM [33