Glutathione redox cycle enzymes as potential targets for heme-mediated oxidation under hemolysis: in silico analysis

Редокс-гомеостаз глутатіону (-глутамілцистеїнілгліцину) в еритроцитах людини залежіть від активності глутатіонпероксидази (GPX1, КФ 1.11.1.9), глутатіонредуктази (GSR, КФ 1.8.1.7), глутаредоксину 1 (GRX1) і NADPH-генеруючих ферментів пентозофосфатного шляху, глюкозо-6-фосфатдегідрогенази (G6PD, КФ 1.1.1.49) і 6-фосфоглюконатдегідрогенази (PGD, КФ 1.1.1.44). Накопичення вільного гему при гемолізі може вплинути на активність протеїнів, у зв'язку з чим був проведений in silico аналіз структури ферментів редокс-циклу глутатіону з метою виявлення можливих гем-зв'язуючих сайтів та залишків цистеїну, здатних до окислення. Анотації протеїнів були взяті з UniProt. Докінг гему проведений у PatchDock з RMSD кластерування 1,5 Å і з використанням PDB структур протеїнів та гему. Потенціал окислення цистеїнів оцінювався за допомогою Cy-Preds. Для мономерів GSR (1DNC, 3DJJ, 3DK9, 2GRT) та димерів (3SQP, 2GH5) передбачено зв’язування гему через His81 біля дисульфідного зв’язку між ланцюгами та через Cys59 біля сайтів зв’язування FAD і GSSG. Гем-зв'язуючі сайти у GPX1 (2F8A) і GPX3 (2R37) також виявлені у ділянці між ланцюгами та у активному центрі (His80). Зв’язування гему з GLRX1 (4RQR) передбачено майже виключно поблизу N-кінця, незважаючи на доступність всіх цистеїнів разом з CPYC мотивом у активному центрі. В мономері G6PD (2BH9, 5UKW) гем-зв'язуючі ділянки виявлені у сайті зв’язування NADP+ та в -спиралі 437–447, розташованій у димері 2BHL на поверхні між ланцюгами. Гем стикувався до PGD (4GWG, 4GWK) у ділянці зв’язування субстрату біля His187. Таким чином, активні центри ферментів та ділянки взаємодії ланцюгів були виявлені в більшості варіантів докінгу гему. У кожному протеїні виявлено від одного (у PGD) до трьох (у GSR) схильних до окислення цистеїнів, в тому числі серед потенційних сайтів зв’язування гему. Опосередкований гемом окислювальний ефект на ферменти редокс-циклу глутатіону у еритроцитах і плазмі крові може явитись важливим механізмом посилення гемолізу при стресі та патології.


Introduction
Glutathione is tripeptide (-glutamylcysteinylglycine) involved in the oxidative stress response and adaptation as water-soluble antioxidant and co-substrate of glutathione-dependent enzymes. Imbalances of glutathione redox homeostasis followed by the accumulation of glutathione disulfide (GSSG) are observed under cardiovascular and neurological diseases, diabetes and other pathologies (Luschak, 2012).
Defense function of glutathione in blood is mostly linked to reduced glutathione (GSH) oxidation under direct interaction with oxidants or in glutathione peroxidase reaction with consequent reduction by NADPH-dependent glutathione reductase (Andersen et al., 1997). The only source of NADPH in erythrocytes is pentose-phosphate pathway (PPP) thereby the deficiency of PPP key enzyme glucose-6phosphate dehydrogenase causes hemolytic anemia (OMIM #300908). GSH also may be attached to proteins cysteine residues and be removed by glutaredoxin having a glutathione-disulfide oxidoreductase activity in the presence of NADPH and glutathione reductase. Glutathionylation contributes to thiol groups defense from oxidation and provides the important reserve of red blood cell glutathione under the oxidative stress (Hanschmann et al., 2013). Glutathione redox cycling in human erythrocytes is, therefore, dependent on the activities of glutathione peroxidase (GPX, EC 1.11.1.9), glutathione reductase (GSR, EC 1.8.1.7), glutaredoxin 1 (GRX1) and NADPH-generating enzymes of PPP: glucose-6-phosphate dehydrogenase (G6PD, EC 1.1.1.49) and 6-phosphogluconate dehydrogenase (PGD, EC 1.1.1.44).
Hemolysis occurring under various stresses and pathological conditions in mammals leads to hemoglobin release with its further degradation and free heme accumulation in blood plasma (Chiabrando et al., 2014;Immenschuh et al., 2017). Heme can directly damage cell structures through its prooxidant and detergent action (Chiabrando et al., 2014), furthermore, its attachment to heme regulatory motifs (HRM), such as Cys-Pro, described in various proteins, is considered to be a signaling event (Mense, Zhang, 2006). Short-term heme binding to proteins is performed mostly through lipophilic amino acids while stable attachment of heme is provided by its covalent binding predominantly to histidine or cysteine residues (Smith et al., 2010). Under significant free heme accumulation the glutathione redox cycle enzymes could become the targets for heme-mediated modification but their affinity to heme molecule as well as susceptibility of their cysteine residues to oxidation have not been investigated yet.
Taking into account direct and indirect prooxidant effects of heme on protein conformation and activity, in silico study of putative heme-binding sites and oxidizable cysteine residues in human proteins involved in glutathione redox cycling was performed as a part of analysis of the mechanisms of heme action on the redox homeostasis under hemolysis.

Materials and methods
The amino acid (AA) sequences and annotations of the proteins selected for study (Table 1) were downloaded from UniProt knowledgebase (http://www.uniprot.org/). Free online tool HemeBIND (http://mleg.cse.sc.edu/hemeBIND/) was used to predict amino acids heme-binding propensity by analysis of protein sequences in *.fasta format. HemeBind performs predictions through comparison with sequences of the known heme-binding complexes (Liu, Hu, 2011).
The experimental data on protein structures ( Among structures known for this moment for human GSR six ones with highest resolution having coenzymes and substrate (GSSG) or product (GSH) as ligands were selected for analysis. PDB-files with dimeric structures of GPX1 and GPX3 were edited in text redactor for analysis of single protein chains. Structure of tetrameric forms of glutathione peroxidase hasn't been yet experimentally investigated.
Free online tool TM-Align (http://zhanglab.ccmb.med.umich.edu/TM-align/) was used for structural alignment which was estimated in Å as RMSD (root mean squire deviation) of the distances between carbons of the aligned protein chains (Zang, Scholnik, 2005).
Docking of heme as a ligand to PDB-structures was carried out by free on-line tool PatchDock, Beta 1.3 Version (http://bioinfo3d.cs.tau.ac.il/PatchDock/) with clustering RMSD 1.5 Å as it was recommended for protein-ligand docking. First 20 docking solutions with the highest scores for each target PDB-structure were analyzed. Scoring was based on both geometric fit and atomic desolvation energy calculation mostly oriented on molecular shape complementarity (Schneidman-Duhovny et al., 2005). PDB-file for heme molecule was downloaded from PubеChem (http://www.ebi.ac.uk/pdbesrv/pdbechem/chemicalCompound/show/HEM). Visualization of PDB-structures in cartoon view with amino acid side chains was performed by the help of PyMOL (The PyMOL Molecular Graphics System, Version 1.3, Schrödinger, LLC, downloaded from https://sourceforge.net/projects/pymol/); analysis of amino acid residues arranged in close proximity to heme iron as well as distances to heme iron was carried out by SwissProtViewer 4.1.0 (SPDBV; http://spdbv.vital-it.ch).
Cysteines oxidation prediction was performed by free online tool Cy-Preds (Soylu, Marino, 2016) that provided energy and similarity evaluations and functional characterization of the cysteine reactivity based on the experimental data collected in RedoxDB. Results of both HAL_C and COPA schemes of Cy-preds algorithm were used for calculations according to the tool documentation.

Results and discussion
Protein sequence analysis by HemeBind predicted predominantly hydrophobic amino acid residues as potential heme-binding sites: 10 hydrophobic amino acids (AA) of totally 14 AA predicted for GSR sequence; 18 AA of 21 AA for GPX1; 21 of 26 AA for GPX3; all 6 AA for GLRX1; 13 AA of 24 AA for G6PD and 44 AA of 57 AA predicted for PGD. Cys235 of GSR and Cys289 of PGD predicted to bind heme are not located in CP/PC or CXXC motives while Cys170 belongs to PC motif and is neighbor to Cys171 (169PCC) in PGD. Moreover only in this sequence histidines (H453, H466) were scored as putative targets for heme binding. Abovementioned histidines and Cys289 are more accessible than Cys170 or Cys171 according to PDB-files. Proline but not cysteine of PC motif (CP77) was predicted to bind heme in GPX1. Percentage of predicted heme-binding sites (as a ratio to the total number of amino acids in the chain) was the biggest in PGD -12% against 3% in GSR, 5% in G6PD; 6% in GLRX1 and 10% in GPX1. Summarizing the results of HemeBind analysis it can be concluded that specific hemebinding sites able to covalent heme attachment were predicted only for GSR and PGD sequences.
Structural alignment of PDB structures known for one protein revealed very close conformations. Thus for GSR variants used in the study (including 2GRT with mutation) RMSD range varied from 0.20 Å (3DK9 aligned to 3DJJ) to 0.45 Å (3SQP aligned to 2GH5); for glucose-6-P dehydrogenase RMSD was close to 1.00 Å; for glutathione peroxidase isoforms RMSD=1.10 Å (alignment of 2F8A and 2R37). Two structures known for human 6-P-gluconate dehydrogenase were found almost identical (RMSD=0.1 Å).
Molecular docking of heme molecule to monomeric GSR (1DNC, 3DJJ, 3DK9 and 2GRT) revealed putative heme-binding sites located predominantly in four areas (Fig. 1A): the loop region 80-96 with Cys90 and His80 neighbor to -helix (a), the cavity in the centre of the molecule near FAD and glutathione disulfide binding sites close to Cys58 and Cys63 (b), near the hydrophobic regions 45-50 (Arg-Ala-Ala-Val-Val) and 123-127 (His-Ile-Glu-Ile-Ile) organized into two adjacent beta-strands (c) and the region with the loop 354-360 and part of -helix with His351 (d).
Chains in the dimers (3SQP and 2GH5, analyzed in SPDBV) revealed rather big contacting areas formed by more than 140 amino acids with about 70 ones from each chain. Some of them were predicted (Table 3) as heme-binding sites in monomeric models, including redox-active Cys58 and Cys63 as well as regions near His80, Thr339, Phe372 and Glu442. Thus heme might interfere dimerization and thereby disturb the formation of the active form of glutathione reductase. In the enzyme dimers (3SQP and 2GH5) the cavity formed by two chains was predicted to become more preferable site for heme binding than other areas (Fig. 1, boxes a-c). Significant portion of these amino acids was also involved in binding of coenzymes (FAD and NADP) and substrate (GSSG). Among amino acids found in these sites only His351 arranged near the surface could provide long-term specific heme binding.  Molecular docking of heme to glutathione peroxidase isozymes revealed different heme-binding areas for dimeric and monomeric forms (Fig. 2, Table 4). In both dimers heme was predicted to bind predominantly in the interchain region (boxes a and c) thereby interacting with amino acids of both chains. Heme ring was attached to the GPX1 dimer surface just in the region of active centre near Gly47.  Near redoxactive disulfide GPX1 monomer (2F8A:A) was predicted to bind heme predominantly near Phe124 (box b) that didn't participate in dimerization. Thus erythrocyte-specific GPX1 in dimer form could be directly inhibited by heme. Lack of information about GPX tetramer structure didn't allow to predict heme action on the assembly of glutathione peroxidase isomers.
GPX3 dimer (2R37) is not symmetrical thereby protein chains interact by different residues: Lys62 and region 87-93 from the side of chain A and Ile36 together with regions 41-44 and 53-57 from chain B. Heme could be fitted into the cleft between chains in dimer while GPX3 monomer (2R37:A) more probably could bind heme in the region of active site (Fig. 2, box d).
Glutaredoxin 1 is the smallest protein among ones investigated in this study. Heme was predicted to bind almost exclusively (19 of 20 best solutions) to short -helix at the N-end (Table 4). Only one solution predicted heme interaction with Tyr24 just near redox-active disulfide Cys22-Cys25 that is close to the surface. No cavities were found in GLRX1 PDB-structure thereby all cysteines are rather accessible to heme binding including CPYC motif in active site.
Molecular docking of heme molecule to glucose-6-phosphate dehydrogenase revealed two main heme-binding areas (Fig. 3A) for monomer (2BH9): NADP+ binding region in the active site (box a) and -helix 437-447 (box b) found in the dimer 2BHL at the interchain surface. G6PD chains in the dimer are bound symmetrically by the interaction of more than 130 amino acids (equally about 65 from each chain) mostly (more than 50%) hydrophobic. Thus heme binding could more probably directly inhibit activity of G6PD at the active centre than disturb tetramer assembly. Heme binding to 6-phosphogluconate dehydrogenase monomer (4GWK) was predicted at only one area (Fig. 3B, box c) known as substrate binding region. This region contains two groups of amino acid residues: S129-G130-G131 and His187-Asn188 (Table 5). No data is available about residues involved in dimerization so only direct inhibition of PGD activity might be suggested as the mechanism of heme action. Almost total identity of two PDB structures available for human PDG (RMSD=0.1 Å) explains coinciding docking results for 4GWK and 4GWG.
Summarizing the results of docking studies performed for human enzymes of glutathione redox cycle it should be concluded that direct heme binding in the active sites areas is highly probable. Regions involved in substrates or/and coenzymes attachment were revealed in the majority of heme docking variants. Heme binding probability was affected by dimerization. Analysis of the first 20 solutions for all variants of PDB structures used in the study revealed approximately 10% higher scores (p<0,001) in the case of dimers compared to monomer forms. GSR dimers 2GH5 and 3SQP had the highest scores (6381229 and 6330137) and the largest contact areas (840+33 and 847+26, correspondently) as well as G6PD dimer 2BHL (6226171 score with contact area 819+38). The lowest scores with small contact areas were found for monomeric variants of GPX1 and GPX3 (528189 and 5338108, areas 686+33 and 686+53) as well as for GLRX1 (5276124, areas 712+34).  It should be mentioned that PatchDock algorithm didn't take into account the types of interactions but scored mostly geometrical fitting of ligand therefore only the presence of certain amino acid in docking area could be used for type of bond predictions.
Cysteines oxidation potential was estimated by the Cy-Preds online tool (Table 6).  The covalent heme binding to proteins is known to be realized predominantly through cysteine, histidine or tyrosine (Smith et al., 2010). These three types of amino acids were predicted at the distances close enough for iron coordination in several solutions for GSR (His81, Cys59, Table 3), G6PD (Tyr249 ,  Table 5), GPX1 (His80) and GLRX1 (Cys83 , Table 4). Moreover, Cy-Preds found at least one cysteine susceptible to oxidation in each protein studied (Table 6). Three oxidizable cysteines were predicted in GSR (Cys59, Cys64, Cys91).
So analysis carried out in this study testifies that glutathione reduction is more likely to be inhibited under hemolysis by heme binding and oxidation of cysteines in glutathione reductase and glucose-6phosphate dehydrogenase. These results are in agreement with the experimental data on the increase of GSSG/GSH ratio in the cells and blood plasma under action of agents causing oxidative stress in mammals (Pandey, Rizvi, 2011). Inhibition of glutathione peroxidases by heme might temporally redirect GSH to non-enzymatic reactions including Fe 3+ reduction that might have pro-oxidant effect through acceleration of Fenton reaction. On the other hand, inhibition of glucose-6-phosphate dehydrogenase in