Dynamic Basis for Auranofin Drug Recognition by Thiol-Reductases of Human Pathogens and Intermediate Coordinated Adduct Formation with Catalytic Cysteine Residues.

In all the living systems, reactive oxygen species (ROS) metabolism provides resistance against internal and external oxidative stresses. Auranofin (AF), an FDA-approved gold [Au(I)]-conjugated drug, is known to selectively target thiol-reductases, key enzymes involved in ROS metabolism. AF has been successfully tested for its inhibitory activity through biochemical studies, both in vitro and in vivo, against a diverse range of pathogens including protozoa, nematodes, bacteria, and so forth. Cocrystal structures of thiol-reductases complexed with AF revealed that Au(I) was coordinately linked to catalytic cysteines, but the mechanism of transfer of Au(I) from AF to catalytic cysteines still remains unknown. In this study, we have employed computational approaches to understand the interaction of AF with thiol-reductases of selected human pathogens. A similar network of interactions of AF was observed in all the studied enzymes. Also, we have shown that tailor-made analogues of AF can be designed against selective thiol-reductases for targeted inhibition. Molecular dynamics studies show that the AF-intermediates, tetraacetylthioglucose (TAG)-gold, and triethylphosphine (TP)-gold, coordinately linked to one of catalytic cysteines, remain stable in the binding pocket of thiol-reductases for Leishmania infantum and Plasmodium falciparum (PfTrxR). This suggests that the TP and TAG moieties of AF may be sequentially eliminated during the transfer of Au(I) to catalytic cysteines of the receptor.

Introduction

Auranofin (AF), an organogold drug initially approved by FDA to treat rheumatoid arthritis, contains an Au(I) atom forming linear coordinate bonds with triethylphosphine (TP) and tetraacetylthioglucose (TAG) groups, through phosphorous and sulphur, respectively (Figure ).

Figure 1

Schematic representation of AF.

Throughout this manuscript, TP-gold and TAG-gold, together, are referred to as AF-intermediates. Over the years, antibacterial activity of AF against several pathogens such as Helicobacter pylori,[1]Mycobacterium tuberculosis,[2]Staphylococcus aureus,[3−5]Enterococcus faecalis, Enterococcus faecium,[6]Treponema denticola,[7]Clostridium difficile, and so forth has been reported.[1,2,7] Previous studies have shown that AF manifests an antiprotozoal activity against major protozoal pathogens such as Plasmodium falciparum,[8,9]Entamoeba histolytica, Giardia lamblia,[10,11]Trypanosoma brucei,[12] and Leishmania infantum.[13,14] Reported in vitro studies show that AF is also effective against filarial nematode parasites that cause river blindness and lymphatic filariasis.[15] Thus, AF can be a highly repurposed drug as it shows lethal/inhibitory activity against diverse pathogens.

Parasitic organisms are subjected to an endogenous oxidative stress and oxidative challenges imposed by the host’s immune system. The thiol-based redox metabolism involved in scavenging ROS is pivotal to maintain redox homeostasis in parasites. Thiol-reductases (TRs), such as thioredoxin reductases (TrxR), glutathione reductase (GR), thioredoxin-glutathione reductase (TGR), and trypanothione reductases (TryR), are the key enzymes involved in the thiol-based redox metabolism in different pathogens. Apart from protecting the cells against oxidative stress, these enzymes are also essential for proper protein folding and DNA synthesis.[16−18] Previous biochemical and structural studies have shown that AF and related gold drugs are potential inhibitors of TrxR, TryR, and TGR.[19]

TRs are functional homodimers that exist in two different forms, high molecular weight TR (HMW-TR) and low molecular weight TR (LMW-TR). HMW-TRs, predominantly present in eukaryotes, contain a CXYXXC redox active motif, where X is hydrophobic residue and Y is negatively charged residue, at the flavin adenine dinucleotide (FAD)-binding domain, whereas LMW-TrxR, mainly present in prokaryotes and selective eukaryotes, contains a CXXC motif at its nicotinamide adenine dinucleotide phosphate (NADPH)-binding domain.[20] Two catalytic cysteine residues, located on Cys-Val-Asp-Val-Gly-Cys motif, are conserved across the HMW-TRs, TrxR, TryR, and TGR enzymes.

Various inhibitory concentration and toxicity studies have been carried out for AF and its analogues, against bacterial[2] and protozoal pathogens.[21] AF has shown an IC50 of 9.68 μM against LiTryR and cent percent mortality of the pathogen at 50 μM concentration.[14] AF showed Ki of 10 μM concentration against both TrxR and GR activities of TGR of Schistosoma mansoni (SmTGR) after 1 h incubation but had no effect on mammalian cells in vitro.[22] TrxR activities of G. lamblia, E. histolytica, and Toxoplasma gondii were inhibited by AF at an IC50 of 152, 0.4, and 0.28 μM, respectively.[11,23,24] AF (150–300 nM) completely inhibited growth of S. aureus, whereas 1.2 μM of AF achieved the same for H. pylori.[1] IC50 and a minimum inhibitory concentration (MIC) of AF against H. pylori were 88 nM and 2 μM, respectively. A recent study has shown that N-heterocyclic carbenes (NHCs), AF analogues, exhibit an inhibitory concentration (MIC) comparable to AF against H. pylori. Interestingly, toxicity of these analogues for human embryonic kidney cells (HEK-293T cells) is 13–20-fold lower than that of AF.[1] Recently, the AF analogue (referred as analogue 7) has been shown to inhibit the TrxR from E. histolytica (EhTrxR) at nanomolar concentration (IC50 = 55 nM) when compared to sub-micromolar concentration of AF (IC50 = 400 nM).[11]

Angelucci et al. reported the cocrystal structure of SmTGR with gold atoms.[25] Although the authors used the SmTGR–AF complex for crystallization, the crystal structure obtained revealed the presence of only gold [Au(I)] atoms, rather than AF, forming a linear coordinate bridge between the pair of cysteine residues (Cys–Au–Cys) at two different sites: (i) the first redox active di-thiol couple, Cys154 and Cys159, near FAD-binding site and (ii) the second di-thiol couple, Cys520 and Cys574, at the C-terminal tail. The cocrystal structure of LiTryR with AF showed the presence of TAG of the AF and an Au(I) atom linked to the first di-thiol couple.[13] Although LiTryR and SmTGR have different domain architectures and different substrates preferences, in both cases, a gold atom forms a coordinated adduct with the catalytic cysteine residues present near the FAD-binding site. The formation of an irreversible, stable, linear coordinate bond between Au(I) and the catalytic cysteines permanently disables the transfer of the reductive potential from reduced FAD to the cysteines and thus restricts further redox mechanism of the protein.[25] However, no structural information is available either with the AF or its intermediate adducts with target enzymes. Despite having several biochemical and structural studies, the mechanism of binding of AF and transfer of Au(I) from AF to the catalytic cysteines proved to be surprisingly elusive and remains unknown.

Here, we used in silico molecular modeling, docking, and molecular dynamics (MD) simulation approaches to (A) analyze the effect of variation in the binding pocket residues on TRs and AF interactions, (B) understand the dynamics of the intermediate coordinated adduct in the catalytic pocket, and (C) perform a comparative analysis of AF and analogues binding to TrxR, TryR, and TGR from diverse pathogens. This study would provide important leads to understand the molecular mechanisms of differing efficacies of AF and its analogues against TRs from different pathogens and also may shed some light upon the possible mode of transfer of Au(I) from AF to the target enzymes.

Results

Docking and MD Simulation Analysis of AF with HMW-TRs

LiTryR

SiteMap analysis revealed the dimeric interface as the most potent binding site in LiTryR. Docking of AF on LiTryR at the dimeric interface showed that residues from both protomers contribute for AF binding through a network of hydrogen bonds, hydrophobic interactions, and van der Waals contacts (Figure A,B). Glu467, Ser470, and His461 were involved in hydrogen bond or ionic interactions with the TAG moiety of the AF. Phe396, Lys61, and Pro462 residues formed hydrophobic contacts with the TAG moiety. Hydrophobic regions of the TP moiety has established van der Waals or hydrophobic interactions with the side chains of Thr335 and His461 residues (Figure B).

Figure 2

AF docking into LiTryR dimeric interface. (A) AF (red stick) interacts with the dimeric interface residues of TryR. The two protomers of LiTryR are colored in orange (protomer A) and royal blue (protomer B). AF is present near the catalytic Cys residues and FAD (shown in green stick) binding site. (B) Interaction of AF (green ball and stick) with the binding pocket residues of LiTryR is shown. Dashed red lines and straight blue-red lines represent intermolecular H-bond and ionic interactions, respectively. Golden dashed lines represent intramolecular coordinate bonds in AF. Same representation for interaction has been used in Figures , 5, and 8.

Figure 4

Interaction of AF with SmTGR and modeled BmTrxR. (A) AF interacts with the dimeric interface residues of SmTGR near the catalytic Cys residues and FAD-binding site. (B) Modeled structure of BmTrxR with docked FAD and NADPH indicated in green sticks. (C) Interaction of AF with BmTrxR. Blue dotted lines represent the predicted nucleophilic attack on AF.

Figure 5

Interaction of AF with LMW TRs from bacteria and protozoon. Residual interaction of AF with bacterial TrxRs (A) HpTxrR and (B) MtbTrxR. (C) Modeled structure of WolbTrxR and (D) interaction of AF with WolbTrxR. (E) Intermolecular contacts of EhTrxR–AF complex.

Figure 8

Interaction of AF-intermediate coordinated adducts with the dimeric interface residues of LiTryR and PfTrxR. (A) Schematic representation of possible mechanisms of transfer of Au(I) to the catalytic Cys di-thiols, through the formation of coordinated intermediate adducts, to form final Cys–Au(I)–Cys adduct. Intermolecular interactions between Cys coordinated TAG-gold or TP-gold with LiTryR (B,C) and PfTrxR (D,E), respectively.

The thermodynamic stability of the docked complex was analyzed by computing ΔGbind. Molecular mechanics/generalized born surface area (MM/GBSA) calculated ΔGbind of AF–LiTryR complex was found to be −36.72 kcal/mol, indicating a thermodynamically favorable interaction (Table ). The AF–LiTryR docked complex was subjected to 50 ns MD simulation (Figure S1). Network of interactions observed in the docked complex were retained throughout the simulation (Figure S1C), also, low root mean square deviation (RMSD) and root mean square fluctuation (RMSF) of the ligand indicate stability of AF in the binding site (Figure S1A,B).

Table 1

MM/GBSA Calculated ΔGbind for AF Docked on Different TRs

organism	protein	ΔGbind (kcal/mol)
L. infantum	TryRa	–36.72
P. falciparum	TrxRa	–50.05
S. mansoni	TGRa	–51.24
B. malayi	TrxRa	–35.19
H. pylori	TrxRb	–43.05
M. tuberculosis	TrxRb	–54.62
E. histolytica	TrxRb	–66.47
Wolbachia	TrxRb	–46.99

HMW-TR.

LMW-TR.

Comparative sequence analyses of HMW-TRs reveal that the AF binding residues are highly conserved in TryR of trypanosomatids, including Trypanosoma cruzi and T. brucei, and TrxR of other protozoal parasites, including P. falciparum and G. lamblia, which indicate that both the trypanosomatids and other protozoal parasites can be targeted by the AF (Figure ). Though TryR and TrxR uses different substrates, cofactors NADPH- and FAD-binding pocket residues are highly conserved, and two catalytic Cys residues, just before α2 helix, are invariant (Figure ). Therefore, above analyses may explain the susceptibility of these parasites to AF treatment.

Figure 3

AF interacting residues: sequence alignment of HMW-TRs. AF binds to the dimeric interface, formed by residues of α1, the loops connecting β2−α2, α3−α8 of 1st protomer and α11 along with loop connecting β19−α10 and α11, α12−η4 of the 2nd protomer. Residues of these regions interacting with the drug are highlighted in red. Residues highlighted in blue correspond to various substitutions, among the residues interacting with AF. The FAD-binding region is formed by α1, α2, α4, β15, and loops connecting β2−α2, β6−α6, and η4−α8; interacting residues are highlighted in yellow. The NADPH-binding region is formed by α5, β17, and loops connecting β8−α5, β9−α6, β10−β11, and β14−β15, and interacting residues are highlighted in green.

PfTrxR

PfTrxR shares a significant sequence identity with LiTryR (Figure , Table S1). The docking of AF on PfTrxR and MD simulation analysis showed that residues from both monomers of the dimeric interface contributed for the AF binding through hydrogen bond, ionic and hydrophobic interactions (Figure S2), similar to the network of interaction as found in AF docked on LiTryR. Further, low ΔGbind (−50.05 kcal/mol) of AF to PfTrxR (Table ) and insignificant change in RMSD and RMSF of complexes, as function of time with respect to structure of starting complex, suggested that AF and PfTrxR forms a thermodynamically stable complex (Figure S3A,B).

SmTGR and BmTrxR

AF docked on SmTGR also showed the same site of binding and pattern of AF recognition as it binds to LiTryR and PfTrxR. H-bond interactions through His571, Arg450, and Glu576 and hydrophobic interactions with Lys124, Pro572, and Leu508 stabilized the AF at the potent binding site (Figure A). The protein–ligand RMSD and ligand RMSF of SmTGR–AF complex analyzed during the 50 ns MD simulation (Figure S4A–C) showed that AF forms a stable complex with SmTGR. The negative ΔGbind (−51.24 kcal/mol) indicates the thermal stability of AF in the binding pocket of SmTGR (Table ).

The BmTrxR structure was modeled using human TrxR as the template and validated by the Ramachandran plot analysis (Figure S5A). Also, FAD and NADPH, being readily docked in the respective binding pockets, support that stereochemistry of the model was accurate (Figure B). Similar to above discussed TRs, in BmTrxR also AF got docked at the dimeric interface near catalytic cysteine residues with a similar network of interaction. The major AF interacting residues like His611, Glu616, Thr613, and other hydrophobic residues are also present in BmTrxR, which indicates a similar mode of AF binding (Figures , 4C). The negative ΔGbind (−35.19 kcal/mol) affirms the thermal stability of AF in the binding pocket of BmTrxR (Table ). The RMSD of the complex and RMSF of AF were found to be within limits during MD simulation, which indicates high stability of the complex (Figure S4A,D,E).

Recognition of AF by LMW-TrxRs of H. pylori, M. tuberculosis, Wolbachia, and E. histolytica

In the presence of FAD, docking of AF at the dimeric interface yielded a very low score. Upon removing FAD, the SiteMap analysis identified FAD-binding site as a potential binding site. Docking studies of AF were carried out on TrxRs of both H. pylori (HpTrxR) and M. tuberculosis (MtbTrxR) in the absence of FAD. In both cases, a similar network of hydrogen bonds and hydrophobic and ionic interactions facilitated the recognition of AF at the FAD-binding site (Figure A,B). The computed ΔGbind supported the thermodynamically favorable binding of AF to HpTrxR and MtbTrxR at the FAD-binding site (Table ).

TrxR of Wolbachia (WolbTrxR), an endosymbiont bacteria of Brugia malayi, is a LMW-TR and shares significant sequence identity with TrxR of Francisella tularensis (F. tularensis TrxR) (Table S1). Therefore, we used the F. tularensis TrxR structure as the template to model WolbTrxR (Figure C). The modeled structure of WolbTrxR was validated by Ramachandran plot analysis (Figure S5B). The SiteMap analysis and docking studies on WolbTrxR showed the FAD-binding site as a preferable binding site for AF, as found in previously mentioned LMW-TrxRs. The computed MM/GBSA binding energy of AF to WolbTrxR was comparable to ΔGbind of the HpTrxR–AF complex, which indicated that AF may form a stable complex with WolbTrxR (Table ). Intermolecular interactions between AF and WolbTrxR also support that WolbTrxR–AF may form a thermodynamically stable complex, and AF may bind at the FAD-binding site (Figure D).

Though EhTrxR is a protozoal TR, it exhibits a higher sequence and structural similarity with LMW-TR, compared to HMW-TRs of protozoa (Figure , Table S1). Docking, MD simulation, and computed ΔGbind studies indicated that AF binding to EhTrxR is similar to AF binding to MtbTrxR (Figure E, Table ). Comparative sequence analysis of LMW-TRs revealed that the cofactors binding pockets, catalytic cysteine residues, and residues interacting with AF are highly conserved, which support that AF may be recognized similarly by all LMW-TRs (Figure ).

Figure 6

AF interacting residues among LMW-TrxRs. AF binding region is formed by helices (α1, η1, α4), sheets (β8, β16), and the loops connecting β7−α3 and β8−α4. Residues highlighted in red correspond to those interacting with NADPH. Residues highlighted in yellow correspond to those interacting with FAD. Residues highlighted in blue correspond to those interacting with AF (blind docking) and residues highlighted in green correspond to those interacting with both FAD and AF (dockings in bacterial TrxRs were carried out in the absence of FAD). PfTrxR is included in the alignment to reflect upon the difference in AF recognition, between LMW-TRs and HMW-TRs.

AF Analogues Bind to TRs

Preliminary glide XP-docking indicated that among virtually screened analogues of AF (Figure ), analogues belonging to group “A” was more negative and hence showed better docking scores than group “B”. Among analogues in group “A”, A1 exhibited marginally better ΔGbind than LiTryR, PfTrxR, and HpTrxR compared with other analogues (Table ). In contrast, calculated ΔGbind of A3 to EhTrxR is better than other analogues. It is interesting to note that though A5 has exhibited better docking score, ΔGbind of this analogue to TRs is weaker than other analogues (Table ). Considering docking scores and ΔGbind together, these results indicate that AF analogues, A1, A3, and A5, may be optimized to develop more potent drug against specific TRs.[1]

Figure 7

Schematic representation of AF and its analogues.

Table 2

Docking Scores and Corresponding MM/GBSA Calculated ΔGbind Scores (in Parenthesis) of AF Analogues Docked on LiTryR, PfTrxR, EhTrxR, HpTrxR and MtbTrxR

		AF analogues
protein	organism	A1	A2	A3	A4	A5	B1	B2	B3	B4	B5	C
TryRa	L. infantum	–3.37 (−39.0)	–3.31 (−36.2)	–3.58 (−35.4)	–2.74 (−37.8)	–6.63 (−36.6)	–2.62 (−26.0)	–2.56 (−24.0)	–2.43 (−20.8)	–2.47 (−23.4)	–2.90 (−22.9)	–3.35 (−26.9)
TrxRa	P. falciparum	–6.84 (−39.2)	–4.03 (−32.6)	–3.86 (−30.6)	–3.78 (−27.6)	–4.68 (−32.3)	–3.85 (−20.4)	–3.02 (−18.8)	–1.38 (−12.3)	–1.96 (−14.2)	–2.59 (−14.7)	–4.38 (−32.9)
TrxRb	E. histolytica	–3.54 (−29.0)	–5.89 (−44.6)	–5.40 (−52.4)	–3.28 (−44.4)	–7.45 (−45.9)	–3.19 (−27.8)	–2.60 (−30.5)	–3.14 (−31.3)	–2.91 (−30.1)	–4.31 (−33.4)	–6.12 (−35.2)
	H. pylori	–2.40 (−40.5)	–3.78 (−31.0)	–3.58 (−38.2)	–3.63 (−39.9)	–4.94 (−38.1)	–3.12 (−21.1)	–2.23 (−24.6)	–1.68 (−21.6)	–1.69 (−20.0)	–2.46 (−21.7)	–3.18 (−27.7)
	M. tuberculosis		–1.89 (−36.8)	–3.39 (−33.4)	–2.99 (−37.8)	–3.90 (−33.2)		–3.20 (−27.0)	–1.84 (−18.4)	–2.14 (−23.4)	–2.82 (−26.0)	–2.99 (−24.1)

HMWTR.

LMWTR.

Docking and Stability Analysis of AF-Intermediates with LiTryR and PfTrxR

MD simulations of the generated complexes, in which the reaction intermediates of AF, that is, TAG-gold or TP-gold, were coordinately linked to one of the catalytic cysteines, showed the physical stability of the coordinated complex. The TAG-gold–Cys adduct was stabilized through hydrogen bonding and hydrophobic interactions, whereas TP-gold–Cys adduct was stabilized mainly by hydrophobic interactions (Figure B,C). Low RMSD and RMSF of ligand in the binding pocket with respect to the docked structure indicated the stability of the intermediate complexes (Figure S6). For further investigation on the stability of the complex, we examined the network of interactions at the interface between coordinated AF-intermediates and the LiTryR complex at 10 ns MD trajectory. Most of the interactions were preserved during MD simulation, suggesting that these might be the intermediate adducts formed during Cys–Au(I)–Cys linear coordinated adduct formation (Figure S6). Calculated ΔGbind of −23.03 and −17.46 kcal/mol for TP-gold and TAG-gold, respectively, also support the thermodynamic stability in the complex (Table ). The crystal structure[13] and our in silico thermodynamic (ΔGbind) analysis of LiTryR–TAG complex support that the TAG-gold intermediate adduct may be more stable compared to the TP-gold intermediate (Table ).

Table 3

MM/GBSA Calculated ΔGbind Scores of Coordinately Linked AF-Intermediates with LiTryR and PfTrxR

organism	protein	AF intermediate	ΔGbind (kcal/mol)a
L. infantum	TryR	AF**	–26.62
		AG-Au*	–23.03
		TP-Au*	–17.46
P. falciparum	TrxR	AF**	–38.46
		AG-Au*	–30.78
		TP-Au*	–16.66

ΔGbind represents binding energies at 10 ns* and 50 ns** of MD simulation.

We carried out a similar study with TP-gold and TAG-gold on PfTrxR. Intermolecular interactions between AF-intermediates and PfTrxR and computed ΔGbind of TP-gold and TAG-gold moieties with PfTrxR indicated that these intermediates may form a stable coordinated complex with PfTrxR (Figure D,E, Table ) like LiTryR–AF-intermediate complexes. Less than 3 Å RMSD and RMSF during the MD simulation of PfTrxR with AF-intermediates also suggested that, like LiTryR, AF-intermediates can form stable coordinated complex with PfTrxR (Figure S7).

Discussion

AF, a gold(I)-conjugated organometallic drug initially identified against rheumatoid arthritis and later tested to repurpose against several pathogens, is known to selectively target TRs through a highly debated but still unknown mechanism. Cocrystal structures of TRs, SmTGR,[25] LiTryR,[13] and EhTrxR,[31] have shown Au(I) forming a linear coordinate bound with the catalytic cysteine residues present near the FAD-binding site. This explains that “somehow” Au(I) gets transferred from AF to catalytic cysteines. There are many theories behind the metabolism of AF, but its molecular mechanism is largely unknown. Tepperman et al. have discussed about the removal of acetyl groups from AF during its metabolism.[35] In this study, we used docking and MD simulation approaches to explore the possibility of recognition of entire AF by TRs of different pathogens. Based on the fact that Au(I) from AF gets transferred to the catalytic cysteine residues, in TryR, TrxRs, and TGR, we hypothesize that the entire AF should be recognized in the pocket close to the catalytic cysteines so that they can be selectively targeted and also the Cys residues of the off-targets can be avoided. Also, the transfer of Au(I) to the catalytic cysteines should happen in a sequential elimination mechanism, where either TAG or TP moiety gets eliminated forming an intermediate coordinated adduct with one of the catalytic Cys, followed by the elimination of other moiety.

AF Binds to the Dimeric Interface of HMW TRs

First of all, we used LiTryR as the target receptor for AF docking. The binding site mapping analysis showed the dimeric interface as a potential target site. Docking of AF at this identified site revealed that residues from both monomers contribute for AF binding in such a way that the Au(I) gets positioned near the catalytic cysteines (Cys52 and Cys57). Further, MD simulation and calculated ΔGbind strongly support the stability of the LiTryR–AF complex.

Comparative sequence analysis shows that the AF-binding pocket residues are highly conserved across a large group of human pathogens (Figure ). Although, TryR and TrxR use different substrates, trypanothione and thioredoxin, respectively, they have significant sequence and structural similarity. Thus, we considered PfTrxR, SmTGR, and BmTrxR for subsequent docking studies. Like LiTryR, in all the three TRs, interaction profiles, and free energy of the binding support that AF can also bind to dimeric interfaces of PfTrxR, SmTGR, and BmTrxR TRs similar to LiTryR (Table ).

AF Competes with FAD To Bind to LMW-TRs

Previous studies have shown that AF not only inhibits the TrxRs of H. pylori,[1]M. tuberculosis,[2] and E. histolytica(11) but also showed potent growth inhibition against these pathogens. However, bacterial TrxR exhibits an insignificant sequence identity with HMW-TrxRs such as PfTrxR (Figure , Table S1). The main objective was to investigate whether entire AF binds to LMW-TrxRs, HpTrxR, MtbTrxR, WolbTrxR, and EhTrxR. We were also interested to know how differently AF may be recognized by LMW-TrxRs compared to the HMW-TrxRs. Our docking studies of AF into the dimeric interface of HpTrxR and MtbTrxR did not produce a significant docking score. Also, no docking site was found near the catalytic Cys residues. Apart from the dimeric interface, the FAD-binding site is also the one which is close to the catalytic cysteine residues. Therefore, we hypothesized that AF may compete with FAD to reach the catalytic cysteine residues. Our docking, MD simulation, and MM/GBSA calculations not only strongly support our hypothesis but also indicate the stable binding of AF at the FAD-binding site (Figure , Table ). This shows that AF is recognized differently by HMW-TRs and LMW-TRs.

AF Analogues Can Be a Better Alternative To Target Specific TRs

A recent study has shown that few NHCs, AF analogues, exhibited inhibitory concentration comparable to AF against H. pylori with less toxicity,[1] which indicates that analogues can be better drugs than AF. In addition, an AF analogue has been shown to inhibit EhTrxR at nanomolar concentration while the sub-micromolar concentration of AF was known to be inhibitory.[11] We carried out in silico binding analysis, using the analogues outlined in Figure , to examine their binding potential to LiTryR, PfTrxR, HpTrxR, MtbTrxR, and EhTrxR vis-a-vis AF binding to above TRs. We have identified three leads, A1, A3, and A5, which can be further optimized for specific TRs. Because all the A-grouped analogues have TAG moiety, it indicates that TAG may play an important role in drug recognition. The other NHCs having bulkier aromatic rings, as shown in Figure , may determine selectivity for different TRs. This analysis also indicates that rather than “one size fits all”, it is important to design the tailor-made AF analogues, specific to given TRs.

TR Inhibition by AF May Be a Three-Step Catalysis Process

The cocrystal structure of LiTryR with AF has also shown the presence of the TAG fragment of AF at the dimeric interface region of the protein.[13] In this structure, the TAG has been shown as a separate entity which is not linked to Au(I). The presence of the TAG fragment and not the TP fragment in the structure explains that AF gets cleaved into TAG, Au(I), and TP moieties mediated by some unknown mechanism at the dimeric interface. Our studies also show that entire AF is recognized at the dimeric interface of HMW-TrxRs. Based on these facts, we hypothesized that the transfer of Au(I) to the catalytic cysteines may be catalyzed through a three-step reaction mechanism: (i) noncoordinative binding of entire X–Au–Y at binding site; (ii) reaction with one Cys with concomitant elimination of X/Y and formation of Cys–Au–Y/X coordinative adduct (Figure A); and (iii) reaction of the second Cys with concomitant elimination of Y/X and formation of the final Cys–Au–Cys adduct. To investigate the above mechanism, we created the coordinated adduct of TP-gold or TAG-gold moieties with one of the catalytic cysteine residues of LiTryR and PfTrxR. Our MD simulation and MM/GBSA studies showed the stability of the coordinated adduct in the binding pocket which supports our hypothesis (Figures , S6, and S7, Table ).

Conclusion

Based on the above analysis, we conclude that AF may bind at the dimeric interface of HMW-TrxRs, TGR, and TryR TRs. Catalytic cysteines located near the dimeric interface may form a coordinated adduct with Au(I) atom. In LMW-TrxRs, AF may compete with FAD to bind to the FAD-binding site of each monomer. Then, it may form a transient coordinated adduct with one of the cysteine and then form a stable irreversible coordinated Cys–Au(I)–Cys complex that permanently inhibits the catalytic function of the protein. We also suggest that some of the AF analogues can be optimized as potential drugs against the studied TR. This study would also help to unravel the mechanistic insights into metal coordinated adduct formation with target molecules in a group of compounds where gold- (AF, aurothioglucose, and aurotioprol), silver- (Ag(sulfadiazene)), arsenic- (As2O3), ruthenium- (trans-[RuCl4(Me2SO)(Im)]−), or platinum (cis-[Pt(amine)2X2])-based complexes are used as drugs or inhibitors.

Materials and Methods

Comparative Sequence Analysis of TrxR, TryR, and TGR from Different Microorganisms

The sequences of TryRs, TGR, and TrxRs of different pathogens were retrieved from UniProt and NCBI database (Table S2). ClustalW[26] was used for multiple sequence alignment to compare the motifs/residues involved in interaction with AF, and the cofactors (FAD and NADPH). Manual editing of aligned sequences was performed using the BioEdit tool. Multiple sequence alignment was performed using structurally characterized protein/s as a template.

Homology Modeling of TrxRs of B. malayi and Wolbachia

Schrodinger Prime module [Schrödinger release 2017-3: Prime, version 3.8, Schrödinger, LLC, New York, NY, 2014][27] was used for homology modeling of TrxR of B. malayi (BmTrxR) and Wolbachia endosymbiont of B. malayi (WolbTrxR). The sequences of the proteins were retrieved from UniProt database (accession ID: A0A0J9XPT5 for B. malayi and A0A225X7E2 for Wolbachia). Using the module, PSI-BLAST was performed using the PDB database, and the structure with the highest score (including percentage homology and query coverage) was selected as the template. Two chains of the same PDB were aligned against the query sequence, and a homo-dimeric model with the cofactors (present in the template structure) was modeled. The structures of human TrxR (PDB: 2ZZ0)[28] and F. tularensis TrxR (PDB: 6BWT) were used to build the structures of BmTrxR and WolbTrxR, respectively. Human TrxR and BmTrxR share 73% sequence similarity, whereas F. tularensis TrxR and WolbTrxR share 71% sequence similarity. Structural modeling was followed by loop refinement. The modeled structures were validated using Ramachandran plot analysis (Figure S5).

Generation and Quantum Mechanics Optimization of AF and AF Analogues

The 2D structure of AF was downloaded in .sdf format from PubChem (PubChem CID: 70788951) and imported to the Maestro GUI where sulphur of TAG moiety and phosphorus of TP moiety were allowed to form a linear zero-order bond (coordinate bond) with the Au(I) metal ion. Similarly, a set of five NHC were used to generate AF analogues. The AF analogues were categorized into three groups, where (A) TP group was replaced with selected NHCs; (B) TAG group was replaced with selected NHCs and the TP group was replaced with chloride; and (C) TP group was replaced with chloride. AF analogues were generated by modifying the TAG and TP moieties of AF with respective NHCs and chloride, using Maestro 3D builder (Figure ).

Each compound was, then, subjected to quantum mechanics (QM) for geometry optimization, using Jaguar (Schrodinger). Density function theory, MO6-2X hybrid functional, was applied with a LAVCP basis set, with charge 1 and spin multiplicity 1, in gas phase. The threshold of maximum convergence criteria, energy change, and root mean square density matrix change were kept as 100 iterations, 5 × 10–5 hartree, and 5 × 10–6, respectively. For self-consistent field convergence, ligand field theory was considered for initial guess. The vibrational frequency was calculated from this optimization. The optimized geometry was further refined with single-point energy using Jaguar with LACVP**++ basis set, where ** represents polarization and ++ represents diffusion.

Molecular Docking

In this study, we used LiTryR (PDB: 2YAU),[13] SmTGR (PDB: 3H4K),[25] PfTrxR (PDB: 4J56)[29] and B. malayi (BmTrxR; modeled), and LMW-TrxRs of H. pylori (HpTrxR; PDB: 3ISH), M. tuberculosis (MtbTrxR; PDB: 2A87),[30]E. histolytica (EhTrxR; PDB: 4CBQ),[31] and Wolbachia (WolbTrxR; modeled) structures for docking. SiteMap tool of Schrodinger software was used to identify the target sites in the protein structures. The centroid of the identified sites was used for grid generation. The grid was used for performing rigid docking, where the receptor was kept rigid, and the ligand was allowed constrained flexibility. These QM-polarized ligands were docked with the XP precision docking mode of Glide module of Schrodinger. Negative scores indicate tighter binding.

Generation of Coordinated Adducts of AF-Intermediate

The AF-intermediates, that is, TP-gold and TAG-gold were generated, using a 3D builder tool in Maestro Suit, by deleting the TAG and TP, respectively, from the AF. The ligand structures were further energy minimized using the OPLS3 force field. The AF-intermediates were then docked on LiTryR and PfTrxR at the same position where AF was docked. The ligand was manually positioned near the catalytic Cys residues, the disulphide bonds present in the crystal structures were broken, and a formal charge of −1 was assigned to sulphur atoms of both Cys residues. The complexes were further energy minimized, and the sulphur atoms of catalytic Cys residues, present in the vicinity of the Au(I), were allowed to form a zero-order bond (coordinate bond) with the Au(I) of the AF-intermediate. The stability of the AF-intermediate in the positioned region was determined by performing a 10 ns MD simulation.

MD Simulation

MD simulation was performed to investigate the stability of the ligands at the docked site of the proteins. In the current study, we carried out MD simulations for LiTryR, PfTrxR, SmTGR, and BmTrxR docked with AF and LiTryR and PfTrxR with AF-intermediate coordinated adducts using Desmond MD simulations program [Desmond Molecular Dynamics System, version 2.2, D. E. Shaw Research, New York, NY, 2009]. All of the complexes were solvated with single-point charge water model and neutralized with 0.15 M NaCl in an orthorhombic box (a = b = c = 10 Å and α = β = γ = 90°). The systems were minimized and equilibrated with default protocols of the Desmond. The dynamics of the system was calculated with the OPLS3 force field. The long-range electrostatic interactions were calculated using particle-mesh Ewald method.[32] A cutoff radius of 9.0 Å was applied for short-range van der Waals and Coulomb interactions. The systems were simulated under an isothermal–isobaric (NPT) ensemble at 300 K temperature and 1 atm pressure. The temperature and pressure of the system were maintained using Nose–Hoover thermostat[33] and Martyna–Tobias–Klein methods,[34] respectively. An integral time-step of 2 fs was used for the overall simulation. Finally, a 50 ns nonconstrained MD simulation was performed for systems with AF docked on LiTryR, PfTrxR, SmTGR, and BmTrxR. A 10 ns MD simulation was performed for the systems in which coordinated adducts of AF were formed with LiTryR and PfTrxR.

MM/GBSA Calculations

MM/GBSA is a method to calculate free energy of the system. The prime module of Schrodinger calculates free energy considering energies from different components, such as electrostatic, covalent, van der waal, and lipophilic interaction energies.[27] Free energy is calculated as per the following equationwhere E represents the minimized component energies.

ΔGbind of the complexes was determined after docking AF or AF analogues on the TRs and after 10 ns simulation, by performing MM/GBSA calculation. The calculation was performed considering the input partial charges on the ligands. Negative scores indicate tighter binding.