In Silico Analysis of the Enzymes Involved in Haloarchaeal Denitrification

During the last century, anthropogenic activities such as fertilization have led to an increase in pollution in many ecosystems by nitrogen compounds. Consequently, researchers aim to reduce nitrogen pollutants following different strategies. Some haloarchaea, owing to their denitrifier metabolism, have been proposed as good model organisms for the removal of not only nitrate, nitrite, and ammonium, but also (per)chlorates and bromate in brines and saline wastewater. Bacterial denitrification has been extensively described at the physiological, biochemical, and genetic levels. However, their haloarchaea counterparts remain poorly described. In previous work the model structure of nitric oxide reductase was analysed. In this study, a bioinformatic analysis of the sequences and the structural models of the nitrate, nitrite and nitrous oxide reductases has been described for the first time in the haloarchaeon model Haloferax mediterranei. The main residues involved in the catalytic mechanism and in the coordination of the metal centres have been explored to shed light on their structural characterization and classification. These results set the basis for understanding the molecular mechanism for haloarchaeal denitrification, necessary for the use and optimization of these microorganisms in bioremediation of saline environments among other potential applications including bioremediation of industrial waters.


Introduction
The nitrogen cycle (N-cycle) is one of the most important biogeochemical cycles which is mainly driven by prokaryotes. This cycle contains different pathways with several redox reactions that comprise assimilatory and respiratory processes for energy conservation [1]. Some of the reactions involved make denitrification the most energetically profitable out of all types of anaerobic dissimilatory pathways carried out by bacteria and archaea [2]. It consists of a microbial respiratory process in which nitrate is used as an alternative electron acceptor when soluble oxygen is not available for aerobic respiration. The whole pathway involves the sequential reduction of nitrate (NO 3 − ) to dinitrogen (N 2 ) via nitrite (NO 2 − ), nitric oxide (NO), and nitrous oxide (N 2 O) [3] (Figure 1).
Briefly, two nitrate ions are sequentially reduced to N 2 in a series of four oxidationreduction reactions catalysed by four metalloenzymes known as respiratory nitrate reductase (Nar), respiratory nitrite reductase (Nir), nitric oxide reductase (Nor) and nitrous oxide reductase (Nos) [4]. In this process, a total of 10 electrons are transferred to nitrogen oxides to complete their reduction to dinitrogen [4]. The presence of a total or partial set of denitrification enzymes determines whether a microorganism is classified as complete or incomplete denitrifier, respectively [5][6][7][8].
Haloferax mediterranei revealed that it is very similar to its bacteria counterparts, being the structure of the quinol binding site and the water channel highly conserved [25].
This article presents for the first time the complete set of in silico structural models of the remaining three main enzymes of denitrification pathway in Haloferax mediterranei. These results offer new opportunities to better understand the catalytic mechanisms of the enzymes from haloarchaea and may even support the design of possible mutants that improve their catalytic efficiency for biotechnological applications.

Selection of Species for the Analysis of Sequences Alignments
Two approaches were used to select relevant species to conduct the alignment of the sequences: i) Sequences coding for the enzymes in at least one organism from which the enzymes have been structurally characterized were included in all the alignments analysed; ii) Sequences coding for the enzymes of denitrification in species described as complete denitrifiers were included (two haloarchaea species from each family (this classification is according to Taxonomy NCBI Database). After species selection, the alignments of the sequences were performed using Clustal Omega software (https://www.ebi.ac.uk/ Tools/msa/clustalo/, accessed on 18 June 2020.). This bioinformatic tool is based on the HHalgorithm described by Söding [26,27], and for these alignments default parameters were used.

Model Building
Homology modelling was carried out using the SWISS-Model software (https:// swissmodel.expasy.org/, accessed on 8 November 2020) using default modelling parameters [30,31]. The quality of the model was measured using GMQE and QMEAN parameters. GMQE (Global Model Quality Estimation) provides quality estimation which combines properties from the target-template alignment and the template structure. This estimation reflects the expected accuracy of a model built with that alignment and template, normalized by the coverage of the target sequence. Values of QMGE around zero indicates low quality whereas higher numbers (around one) indicate higher reliability. QMEAN provides an estimation of the degree of nativeness of the structural features of the model compared to what would be expected from the experimental structures [32,33]. Values of QMEAN around zero indicate good quality models meanwhile scores of −4.0 or below indicate low quality. Models was carried out using mature protein sequences (excluding TAT signal peptide). Cofactors have been included in the model structure at the same site than in the templates. Orientations and distances are not optimized with respect to the residues in the model. The calculated distances of the interaction between the atoms in the model and in the cofactors are compared with those in the templates in Table S1.

Analysis of Electrostatic Surface Potential and Illustration
Models structure analysis and illustrations were carried out using PyMol Molecular Graphics System, Version 2.3.2 Schrödinger, LLC. Cofactors were incorporated into the finished homology model at the same site of templates. The residue numbering of all alignments includes the N-terminal signal peptide region. Mature enzymes were used (excluding N-terminal signal peptide region) to carry out the model building. However, the model numbering contains N-terminal signal peptide region to be in concordance with alignments. The electrostatic surface analysis was performed using Adaptative Poisson-Boltmann Solver plug-in in PyMol software (Version 2.3.2 Schrödinger, LLC.)
In Haloferax mediterranei genome the respiratory nitrate reductase gen cluster is in the plasmid pHM300. It contains genes coding for the catalytic and the electron transfer subunit, NarG and NarH, respectively, but there is no a gene that encode a product similar to bacterial NarI [8,20].

N-terminal Analysis Exposes Conflict of Annotation in Several Haloarchaeal Species
In the N-terminal signal peptide region of Haloferax mediterranei NarG, a distinctive twin-arginine translocase (TAT) exportation motif ([ST]RRxFLK) has been identified ( Figure 2A), which is absent in the same region in NarH. This difference indicates that folded NarGH dimer might be transported across the plasma membrane by the TAT system. The TAT signal peptide is a short N-terminal motif characterized by the presence of a positive N-terminal region, a hydrophobic region and a consensus short motif AxA which is the specific cleavage site for a signal peptidase [36,37]. The TAT system transports folded proteins across the plasma membrane in Bacteria and Archaea and it plays a critical role in protein transport in haloarchaea [21,[38][39][40].
All the NarG enzymes from haloarchaea present the TAT signal sequence in the N-terminal region (Figure 2A). In Haloferax mediterranei NarG, the predicted protease cleavage site is between Ser 67 and Asp 68 (AAS-DD), therefore, the mature NarG would begin with an aspartate residue. The size of the Haloferax mediterranei NarG signal peptide is 67 residues, which is larger than those identified in the rest of analysed species except for Halogeometricum pallidum and Halosimplex carlsbadense. This feature could be due to ambiguous annotation. It is important to note that this conflict of annotation has been previously described in other haloarchaeal studies [25,41]. It is possible to assume that narG start codon was GUG instead of AUG, being the first amino acid Val 30 or Val 38 , (Figure 2A). GUG start codon is less frequent than AUG in prokaryotes, nevertheless, its presence is relevant in genomes with high GC content such as the Haloferax mediterranei genome [42].

The Structural Model Shows Spatial Conservation of All Residues Involved in Cofactors Binding
To obtain an insight into the structure of the respiratory nitrate reductase from Haloferax mediterranei, a homology modelling study was conducted. The best template was the structure of the perchlorate reductase (PcrAB) from Azospira oryzae (PDB: 4ydd.1.AB). For NarG modelling, the sequence without the TAT signal peptide was used as query, showing a sequence identity with PcrA of 45%, and a coverage of 95%. The quality of the model was supported by a QMEAN of -2.21 and a QMQE of 0.77. NarH reported an identity of 51%, with a coverage of 93% with PcrB and the QMEAN and QMQE were of −0.46 and of 0.75, respectively. NarG structure was modelled from Pro 16 to Asp 916 (numbering according to mature protein which corresponds to residues 83 and 983 in WP_004056332.1). NarH structure was modelled from Asp 17 to Asp 350 (which maintains the same amino acid numbering in WP_004056335.1).
The structural model of NarGH from Haloferax mediterranei is shown in Figure 3. The largest NarG catalytic subunit conserves the characteristic structure of the Moco active site and the [4Fe-4S] centre (FS0). The NarH is the electron transfer subunit and conserve the residues that ligate the three [4Fe-4S] centres (FS1, FS2, FS3) and the [3Fe-4S] centre (FS4).
In DMSO reductase family members, the Mo atom of the MoCo is ligated, in a trigonal prismatic manner, by four thiolates of the bis-MGD cofactor, an inorganic ion like oxygen, sulphur, or selenium atoms and an amino acid ligand, such as serine, cysteine, selenocysteine or aspartate residues. The nature of this ligand is the basis for the classification of the three types of DMSO reductases. Respiratory nitrate reductases have an aspartate residue as the Mo ligand, like the other members of type II group of the DMSO reductase family [10,43]. The architecture of the MoCo cluster in Haloferax mediterraeni could be similar to the one observed in Azospira oryzae PcrA being Asp 249 the Mo ligand. This Asp residue, involved in catalysis, is homologous to Asp 198 in PcrA and Asp 222 in Escherichia coli NarG. In Azospira oryzae PcrA, the sixth ligand is an oxygen atom from an oxo, hydroxyl, or associated water molecule as in NarG from E.coli [10,35].
Regarding the iron-sulphur clusters in Haloferax mediterranei NarGH model, all [Fe-S] clusters are coordinated by four cysteine residues with exception of the [4Fe-4S] center (FS0) of NarG subunit. The specific pattern for FS0 coordination CxxCxxC(x)nC has been reported in formate dehydrogenases (Fdh), periplasm nitrate reductases (Nap) and citoplasmic assimilatory nitrate reductases (Nas). However, in NarG the FS0 cluster is ligated by three Cys side residues and by one His sidechain. This coordination pattern HxxxCxxxC(x)nC seems to be conserved in all type II DMSO reductases ( Figure S1) [10,35] and other iron-sulphur clusters of 'Rieske' proteins [44].
This coordination pattern is also observed in Haloferax mediterranei NarG (residues His 110 , Cys 114 , Cys 118 , Cys 153 ) ( Figure 2B and Figure S1) and in other haloarchaea. Alignments presented here suggest that all analysed haloarchaea NarG sequences contain this coordination pattern, except for Halobiforma haloterrestris and Halopiger aswanensis ( Figure 2B). The cluster coordination by the His chain has been suggested to increase the reduction potential [44]. The location of FS0, between Mo-bis-MGD and FS1 cluster in NarH, has been hypothesized to play a direct role in the electron transfer mechanism. In fact, a mutagenesis study in which this His is mutated to Cys resulted in the loss of NarGHI activity in Escherichia coli [45]. The close location between the FS0 cluster and the cofactor in both Haloferax mediterranei NarG and PcrA suggests the presence of a similar pathway of electron transfer to the catalytic site despite 46.04% sequence similarity. In E.coli NarGH the loop in which the FS0 ligands are found also include an Asn that could be implicated in substrate binding, and a Trp that is bonded to the S4 in FS0 cluster [10]. Both residues are also conserved in a similar orientation in NarG model from Haloferax mediterranei (Asn 113 and Trp 120 ) but in Azospira oryzae PcrA, the Trp residue is substituted by His. It has been proposed that the hydrogen bond between Trp and FS0 may increase the reduction potential of the FS0 cluster [10]. As well, a conserved Arg, which is near the FS0 cluster, could act increasing the reduction potential in Escherichia coli NarG [10]. Interestingly, the corresponding amino acid is a Lys in Azospira oryzae PcrA and haloarchaea NarG (Lys 155 in Haloferax mediterranei NarG) ( Figure 4 and Figure S1) as well as in periplasmatic Naps (data not shown). With regard to the catalytic mechanism, it has been described that the oxidation state of the Mo ion changes between Mo (VI) and Mo (V) during the oxo-transfer reaction, and an oxygen ligand (OH/H 2 O) is released upon cofactor reduction to Mo (IV). EPR Mo (V) signals revealed that NarG displays low-pH and high-pH forms. In the Escherichia coli NarG structure, Asp 222 is hydrogen bonded to the N of a conserved His 546 , which would be deprotonated in the transition to the high-pH form [10]. One of the two protons involved in the nitrate reduction might be this one associated with the acid-base transition between low-pH and high-pH forms. The His residues, highly conserved in bacterial nitrate reductases is an Asp in halophilic nitrate reductases (Asp 493 in Haloferax mediterranei NarG). The absence of this exchangeable proton with the medium indicates that the protontransfer mechanism associated with the nitrate reduction could be different, as occur with periplasmic nitrate reductase [10], in which the residue equivalent to His is Gln. Orange sticks indicate non-homologous amino acid substitution of tunnel gates (Trp 489 for PcrA and Glu 528 for NarG). Black sticks represent amino acids with possible role in substrate binding (N 63 and N 113 in PcrA and NarG, respectively), reduction potential regulation of FS0 (H 70 and W 120 in PcrA and NarG, respectively), electron transport bridge between FS0 and MOCO cofactor (K 104 and K 155 in PcrA and NarG, respectively), and proton donor during catalysis (H 454 and D 493 in PcrA and NarG, respectively).
In Escherichia coli NapA no solvent-exchangeable protons are detected by EPR in the Mo (V) state, but the catalytic mechanism is similar than in NarG. It seems that Mo (VI) is ligated to a water molecule, which is lost on reduction to Mo (V) allowing the nitrate binding. The Mo (V)-nitrate is then reduced by an electron arising from the FS0 centre to Mo (IV) and subsequent the nitrate is reduced to nitrite. The bounded oxygen of nitrate is then transferred to the molybdenum as a terminal oxo group, which protonate to give the stable Mo (VI)-OH 2 [46,47]. In Desulfovibrio desulfuricans NapA a slightly different mechanism occurs in which nitrate binds to the Mo (IV) form [48]. The catalytic mechanisms in these enzymes are yet not clear and alternative pathways for the reduction of nitrate have been postulated depending on pH and substrate concentration in which nitrate can alternatively binds to Mo (IV) and Mo (V) forms [49]. The catalytic mechanism in Haloferax mediterranei should be assess by protein-film voltammetry and EPR experiments.
Although the electron transfer pathway between NarH and NarG in halophilic archaea could be similar to that described in bacteria, the orientation in the membrane is inverted and a membrane anchoring subunit similar to the bacterial NarI is not conserved [21]. As the enzyme from Haloferax mediterranei has been isolated as a dimer NarGH, it is not clear the identity of the electron donor to NarH [20]. In Figure 3C,D, the surface electrostatic potential of active site tunnel and NarGH are shown. As expected, the surface shows a high negative charge density, which is a general characteristic of halophilic enzymes with a high percentage of acidic residues in surface, likewise the contact surface with the electron donor subunit is easily located as it has neutral and positive charge density in surface. The product of gene Orf7 of the nitrate reductase cluster could be the electron donor to NarH [20]. This protein is similar to soluble b-type cytochromes encoded in the operons of other type II molybdoenzyme, for example SerA in the Thauera selenatis selenate reductase and EbdC in the Aromatoleum aromaticum ethylbenzene dehydrogenase subunit [21]. A trimer structure of NarGH with the protein encoded by Orf7 has been modelling using as template the EdbC form Aromatoleum aromaticum (36.65% of sequence identity, GMQE and QMEAN of 0.63 and −3.17 respectively), showing that this trimer could be established (data not shown). Nevertheless, in a recent proteomic study of denitrification in Haloferax mediterranei, this protein was not identified, and it was postulated that NarH could interact with NarC, which is homologous to the cytochrome b subunit of cytochrome bc1 complex. Further experiments should be done to confirm the identity of the electron donor to NarH.
Miralles-Robledillo and coworkers reported that haloarchaeal NarG is closer to (per)chlorate reductases (PcrA) group than to bacterial NarG [43]. Despite this fact, NarG from haloarchaea and PcrA from Bacteria are classified as two different monophyletic groups. On the one hand, PcrAB catalyses the reduction of oxo chlorates, such as perchlorate and chlorate to form hypochlorite under anaerobic conditions [50]. On the other hand, nitrate reductases (NarGH) carry out the reduction of nitrate to nitrite. However, it has been reported that Azospira oryzae PcrAB can recognize nitrate as substrate. Conversely, Haloferax mediterranei and Haloarcula marismortui nitrate reductases can reduce perchlorate and chlorate [18,51]. The biochemical characterization of NarGH reported that this enzyme can use oxo chlorates as substrates with an efficiency comparable to its natural substrate [21,51]. In addition, bromate and iodate have also been identified as substrates of these enzymes. When the structural model of Haloferax mediterranei NarG and the crystal structure of Azospira oryzae PcrA were compared, some differences in the substrate accessing gate tunnel and in the active site were detected ( Figure 4A,B). Figure 2C shows the alignment of the gate and the active site from the haloarchaeal NarG subunit with Azospira oryzae PcrA and Escherichia coli NarG. In this alignment, we can observe the conserved catalytic aspartate (Asp 249 ) and three residues, Phe 243 , Tyr 244 , and Glu 528, which are part of the substrate gate tunnel. These gating residues overlay accurately with the well-studied structure of Azospira oryzae PcrA. Likewise, Phe 243 and Tyr 244 in NarG correspond to Phe 192 and Tyr 193 in PcrA ( Figure 2C). However, a non-homologous Glu residue (Glu 528 ) has been detected in NarG instead of the Trp 489 observed in PcrA. This change was also reported in Escherichia coli NarG [35] and it is conserved in NarGs from bacteria ( Figure S1). In PcrA catalytic subunit, the substrate access tunnel has positive charges creating a high affinity for oxyanions. Conversely, in Haloferax mediterranei NarG monomer the gate tunnel has predominantly negative charges ( Figure 3C,D).
Mutational studies have been performed in the gate residues and in the catalytic Asp in Azospira oryzae PcrA to determine the impact of residues replacement on cell growth and catalytic efficiency [35]. F192A, Y193A and W489A mutations prevented growth using perchlorate as electron acceptor. Interestingly, PcrA W489E mutant retained its ability to grow with chlorate and still displayed activity to reduce perchlorate although its K m was notably increased (Table 1). Nevertheless, what is more surprising and counterintuitive is that this Trp to Glu mutation greatly decreased the PcrA nitrate affinity. The authors suggested that this feature could be due to both, the flexibility of Glu residue in the modified active site because of the lack of hydrogen bonds with Ser 620; and the stacking interaction with Tyr 193 [35]. However, the enzyme from Haloferax mediterranei, despite presenting the Glu residue at the gate tunnel, has higher affinity for nitrate compared with Azospira oryzae PcrAB and with Escherichia coli NarGHI and therefore, the Glu residue would not hinder the nitrate selection.
The role of this Glu residue in the substrate gate tunnel of nitrate reductases is not clear. It has been proposed in PcrAB the reduction of Mo (VI) to Mo (IV) shifts catalytic Asp to a bidentate coordination of the Mo (IV) leading to a conformational shift of Phe residue in the gate tunnel opening it. It is possible to assume that Glu residue acts in a similar way, preventing the premature entry of the substrate to the catalytic site and changing its conformation during catalytic mechanism. Besides, the negative charge could generate repulsion for larger anionic molecules and along with a narrow tunnel act as selective substrate filter, avoiding the binding of bulky substrates such as dimethyl sulfoxide and trimethylamine N-oxide.

Respiratory Nitrite Reductase
Nitrite reductases catalyse the second denitrification reaction, and they can be divided into two types of enzymes: copper-containing nitrite reductases (NirK) or cytochrome-cd 1dependent nitrite reductases (NirS) [53]. NirKs are homotrimeric enzymes (~35 kDa) which contain two copper-binding sites (type 1 and type 2 copper centre) per monomer whereas cytochrome-cd 1 -dependent nitrite reductases (NirS) are homodimeric and contain haems c and d 1 as cofactors [54,55]. NirK can be classified as Class 1 and Class 2. Both types are homotrimer with 2 cupredoxin domain per monomer (two-domains NirK). Class I and Class II differ in their optical absorption spectrum, with class I being blue and showing maximum absorbance at 590 nm and class II being green and with maximum absorbance at 450-460 nm [56]. New classes of NirKs have recently been identified. Class III NirKs are hexameric enzymes that contain two different type 1 copper per subunit, one of class I and one of class II [57]. Additional three-domain NirK have been identified. They contain an extra domain where c-type haem or cupredoxin is tethered to the N or C-terminal site [58][59][60][61]. Recently, a novel four-domain NirK has been discovered. It has been identified in species from Rhizobiales order, such as Bradyrhizobium sp. ORS 375 [62]. They contain the core domain of "classical" two-domains NirK (cupredoxin domain) to which both a cytochrome c and cupredoxin domain are tethered at the N terminus [62]. Recent studies reported a division of NirK sequences into two different phylogenetic clades. Clade I comprises mainly Alphaproteobacteria whereas clade II harbours more diverse taxonomic groups, including Archaea [56]. Each clade presents a different sequence pattern around the catalytic His: TRPHL and SSFHV/I/P. The analysis of haloarchaeal nitrite reductases revealed that all of them are coppercontaining nitrite reductases and they can be classified as two-domains NirKs containing only cupredoxin domains ( Figure 5B).

N-Terminal Inspection Reveals a TAT Signal Peptide
All haloarchaeal NirK sequences analysed show the TAT motif except from Halogranum amylolyticum. In addition, all of them contain the lipobox, [LVI][ASTVI][GAS]C characteristic of lipoproteins ( Figure 5).
In bacteria, lipoprotein precursors are processed by signal peptidase II (SPase II), which specifically recognizes the conserved lipobox motif. A glyceride-fatty acid lipid is attached by a prolipoprotein diacylglyceryl transferase (Lgt) to the conserved cysteine of lipobox, and SPase II cleaves the precursor immediately upstream of this lipid-modified cysteine. In Gram-negative and some Gram-positive bacteria, the conserved lipobox cysteine residue is also acylated by apolipoprotein N-acyltransferase. Several bioinformatic analysis, supported by experimental data, indicate that most haloarchaeal lipoproteins are TAT substrates, in a greater proportion than in other archaea. Homologs of Lgt, SPase II and acyltransferase have not been detected in archaeal genomes, indicating that the molecular mechanisms underlying archaeal lipoprotein biosynthesis may are distinct from their bacterial counterparts [63,64]. The presence of TAT/lipobox combination indicates that halophilic NirKs could be lipoproteins exported by the TAT system through a mechanism not yet described. NirK from Haloferax mediterranei was overexpressed in Haloferax volcanii and polypeptides of different size were detected in cytoplasm and extracellular medium reporting trimeric structure [65]. The shorter isoform was predominant in cytoplasm fractions and was postulated to be the product of an alternative transcriptional processing. The longer one was the predominant in the extracellular fraction and its amino terminal sequence is not consistent with the bacterial lipoprotein processing (unfortunately the membrane fraction was not analysed) [65]. Therefore, molecular mechanisms beyond this issue remain unclear: it could be due to different processing of the polypeptide or maybe due to overexpression. In fact, both mechanisms could also work simultaneously.
The TAT motif is absent in NirK from the two bacterial species, Ralstonia pickettii and Neisseria gonorrhoeae ( Figure 5A). NirK from Ralstonia pickettii contains a Sec-dependent signal peptide (residues 1 to 31) whereas NirK from Neisseria gonorrhoeae, which is an outer membrane lipoprotein, has the conserved lipobox ( Figure 5A) and [66].

The Overall Structure Is Consistent with an Homotrimer with Conserved Residues Involved in the Catalytic Mechanism
A structural model of respiratory nitrite reductase (NirK) from haloarchaeon Haloferax mediterranei was obtained using the structure of the soluble domain of lipoprotein AniA from the pathogenic bacteria Neisseria gonorrhoeae (PDB: 5tb7.1A) as template. The sequence identity with the template was 44%, with a 93% of coverage and QMEAN and GMQE values of 0.64 and 0.76, respectively. AniA is a trimeric class II NirK with each monomer consisting of two β-sandwich cupredoxin domains [67,68]. As the template, the structural model of NirK from Haloferax mediterranei ( Figure 6A,B) would be organized in three monomers constituting a homotrimer structure. Each monomer is composed by 327 amino acids with two cupredoxin domains to support copper atoms.
Clustal Omega alignment reveals that the most important amino acid residues involved in catalysis and electron transfer as well as copper-centre coordination are conserved in all analysed haloarchaeal species ( Figure 5B and Figure S2). Furthermore, a SSFHV/I/P motif conserved around the active site has been found, being proline the last amino acid in most haloarchaeal genomes. This motif, involving the catalytic His residue, classifies haloarchaea NirKs into Clade II. This clade includes NirKs sequences belonging to Class II or unclassified, while Clade I comprises Class I enzymes [56,69]. Helen et al. (2016) observed that the conserved TRPHL and SSFHV/I/P motifs are not a reliable method of classification into Clade I or Clade II since this region is much more diverse than previously described. Nevertheless, enzymes belonging to Clade II have two deletion regions in two surface loops as common feature. The linker loop between the two cupredoxin domains is, in Haloferax mediterranei and in other NirK enzymes from Clade II, seven residues shorter than in Clade I enzymes ( Figure 6C and Figure S3). In AniA, this shortened linker loop may promote a more intimate interaction between AniA and the surface of the outer membrane [67]. The second loop forms a large pedestal shaped tower structure. In Clade II enzymes the eight residue deletion results in a shortened α-helix and a refolded coil structure. This difference in the surface near the type I copper site could be important for binding of different electron donors [67].
The Haloferax mediterranei NirK model reveals the presence of the two copper sites with different coordination patterns. Firstly, type 1 copper site coordinated by two His residues (His 129 and His 179 ), one Cys (Cys 170 ) and one Met (Met 183 ). Secondly, type 2 copper centre is placed in the interface of two adjacent subunits, and it is coordinated by three His residues involving both subunits. His 134 and His 169 are provided by the N-terminal domain of one monomer and His 329 by the C-terminal domain of the adjacent monomer ( Figure 6A). Both copper ions are connected by His 169 , of the type 2 copper centre, and Cys 170 , of the type 1 ( Figure S2). In well-studied NirK atomic resolution structures, type 2 copper centre has been reported to be coordinated by three His residues and one water molecule which are involved in the catalytic mechanism [70,71]. These coordination patterns (His 2 -Cys-Met and His 3 -H 2 O) have been detected and conserved in others twodomain NirK, three-domain NirK and four-domain NirK enzymes belonging to Bacteria and Archaea domains such as Nitrosomonas europaea [72], Alcaligenes faecalis [73], Alcaligenes xylosoxidans [74], Ralstonia pickettii [75], Haloarcula marismortui [19] or Bradyrhizobium sp. ORS 375 [62]. The catalytic mechanism of NirK involves Asp and His residues that donate two protons necessary for nitrite to nitric oxide conversion. These residues are conserved in Haloferax mediterranei NirK, being His 278 and Asp 132 (Figure 6A), and the proposed mechanism [70,[76][77][78] could also take place in NirK from Haloferax mediterranei. The physiological electron donors to NirK in Haloferax mediterranei is still unknown. It could be a halocyanin that it is encoded in the proximity of NirK [6]. In Figure 6D,E is represented the electrostatic potential of the trimer and possible areas of interaction with the electron donor can be identified.

Nitrous Oxide Reductase
The nitrous oxide reductase (Nos) catalyses the last step of denitrification. This enzyme has been extensively studied during the last twenty years in bacteria, particularly from soils, due to its relationship with climate change [79,80]. Nevertheless, there is scarce information about archaeal NosZ enzymes. To date, about ten Nos from different microorganisms have been isolated and none of them belong to the Archaea domain [81]. Phylogenetically, Nos are divided in two main groups: typical (Clade I) and atypical (Clade II) Nos [79][80][81]. In Clade I, Nos proteins are characterized by the presence of TAT signal peptides related to the TAT export system, whereas Clade II proteins use the Sec system for transport across membranes [81]. Another important difference is that some enzymes of Clade II also carry an additional cytochrome c domain at the C-terminal [81,82]. The Nos "core" structure is remarkably similar between clades, since both clades possess Cu A and Cu Z multicopper centres [79][80][81]. Cu A is a binuclear centre, which is the electron transfer cluster, while Cu Z is a tetranuclear cluster which is part of catalytic centre [81,83].
Haloferax mediterranei nitrous oxide reductase lacks the characteristic c-haem binding motif and the Sec-dependent signal peptide, thus it belongs to the Clade I or "typical" NosZ ( Figure 7A).

N-Terminal Analysis Shows Same Annotation Conflict as in Respiratory Nitrate Reductase
Haloferax mediterranei NosZ harbours, as NirK, the TAT/lipobox signal peptide, pointing out the enzyme could be anchored to the membrane through Cys 74 . The signal peptide is larger than typical ones and the consensus motif is far from the N-terminus ( Figure 7A). This feature could indicate, as in NarG, that the NosZ ORF may not be well annotated, and its start could be present closer to the consensus motif. Three probable ORF starts are Val 30 , Val 34 or Val 44 , being GUG the start codon ( Figure 7A and Figure S4).

Cu A and Cu Z Centres Are Conserved in Haloarchaea
The Haloferax mediterranei NosZ structural model ( Figure 8) has been conducted using as template the crystal structures of NosZ from Marinobacter hydrocarbonoclasticus (Pseudomonas nautica) (PDB: 1qni.1) [84]. The sequence identity was 53%, with a coverage of 96% and a QMEAN of -2.74 and GQME of 0.69. Although GQME was higher using as template Pseudomonas stutzeri (PDB: 6rkz.1, 49% sequence identity, 96% of coverage and QMEAN and GMQE values of -2.09 and 0.81, respectively) we decided to use Marinobacter hydrocarbonoclasticus enzyme for modelling Haloferax mediterranei NosZ, since there was low score differences between models, and Marinobacter hydrocarbonoclasticus is an extreme halotolerant microorganism [85]. This bacterium was considered a better fit for this modelling study given that proteins adapted to salty environments share features among them, such as the presence of acidic residues [86]. In comparing amino acid composition between template and query we found that the Marinobacter hydrocarbonoclasticus enzyme sequence is composed of approx. 8.6% Asp and 5.3% Glu, whereas percentages for Haloferax mediterranei NosZ are 9.8% and 7.3%, respectively. Furthermore, theoretical pI of both enzymes has been determined bioinformatically at 4.7. These values point toward the fact that these enzymes are acidic and reinforce the choice of Marinobacter hydrocarbonoclasticus NosZ as the most suitable template for this study. Template and query show good similarity and superimposing, even considering that NosZ sequence from Haloferax mediterranei is larger than Marinobacter hydrocarbonoclasticus. Structural modelling suggests that Haloferax mediterranei NosZ is, like the template, a homodimer [84]. Each monomer presents two domains that bind the copper centres Cu A and Cu Z . Cu A is a binuclear copper centre located at the C-terminal domain of the enzyme whose main function is the electron transport from small electron carrier proteins to the tetranuclear copper sulphide catalytic centre Cu Z situated at the N-terminal [81,84]. Cu A centre is coordinated by a cupredoxin fold organized in nine β-strands. Structure and coordination of this binuclear centre are well studied in Marinobacter hydrocarbonoclasticus NosZ crystal structure [84]. This coordination is carried out mainly by six residues: Cys 611 and Cys 615 which bind the copper atoms; His 576 and Met 622 that coordinate one of the Cu A centre (Cu A1 ) while Trp 613 and His 619 bind Cu A2 centre [84]. These residues seem to be conserved (or partially conserved) among species, including Haloferax mediterranei and other haloarchaea B and Figure S4) [87]. The Trp that coordinates Cu A2 centre is replaced by another aromatic residue (Tyr or Phe) in almost all the haloarchaea, being it a conservative replacement. The only exception is Halorubrum aidingense, in which Trp is replaced by His.
Cu Z , is located at the N-terminal and has a seven-bladed-β-propeller fold. This center has been only identified in NosZ enzymes and contains four sulphide bridged copper atoms coordinated by seven His residues [81,84]. In the case of Marinobacter hydrocarbonoclasticus NosZ (and other NosZ), each copper is coordinated by two His except from CuIV atom which is coordinated by only one. His residues involved in the coordination of copper atoms in Marinobacter hydrocarbonoclasticus are: His 129 , His 130 , His 178 , His 320 , His 375 , His 426 and His 487 ( Figure 9A and Figure S4). Cu Z4 can attach an oxygen atom from a water molecule or a hydroxide, which is the proposed substrate binding position [84,88,89]. This arrangement receives the name of 4Cu1S or simply Cu Z * centre [81]. Apart from those His residues, there are two important amino acids related to the stabilization of the Cu Z * centre, Lys 447 and Glu 485 . Lys 447 seems to establish a hydrogen bond with a hydroxide ligand which is bound to Cu Z4 and another hydrogen bond with Glu 485 [90][91][92]. On the one hand, sequence alignments display that key coordination residues are conserved among haloarchaea species (Figure 9). The seven His are present together with the Lys and Glu residues. On the other hand, the model maintains its characteristic distorted tetrahedron geometry ( Figure 8A). These data provide evidence to support that the Cu Z centre is well conserved in Haloferax mediterranei and in other haloarchaea and suggest that the catalytic reaction of nitrous oxide reduction to dinitrogen happens in the same way as in Marinobacter hydrocarbonoclasticus.

Electron Transfer Pathways Show Partial Conservation and Similarity to Clade II NosZ
Electron transfer in this enzyme encompasses two steps, electron transfer from a small electron donor to the Cu A centre and electronic transfer from Cu A to Cu Z centre. Using molecular docking approaches, Dell'Acqua and coworkers have proposed different possible conserved electron routes in Marinobacter hydrocarbonoclasticus among other mi-croorganisms [83]. Electrons from small molecules are transferred to the Cu A centre. In the case of Marinobacter hydrocarbonoclasticus, the amino acids which are involved in this process are: H 616 (which is supposed to be the entry point of electrons), Leu 618 , Asp 569 , Ala 545 , Val 574 and Gln 547 ( Figure S4) [83,87]. Then, electrons from Cu A must be transferred to Cu Z centre to achieve a fully Cu Z centre reduction (4Cu + ), which is the most favourable state for N 2 O binding [83,93]. The most probable hypothesis indicates that electrons are transferred firstly to the Trp 613 , subsequently to Phe 614 and finally to the oxygen of the water molecule or hydroxyl group bound to Cu Z4 ( Figure S4). In the other alternative pathway, the electron transference starts at the His 619 , follows at Met 620 , and then it finishes at His 178 which is a ligand of the Cu Z2 ( Figure S4) [83].
Combining these data with sequence alignment, potential presence of this electron pathway has been analysed, not only in Haloferax mediterranei NosZ but also in other haloarchaea. Thus, considering the residues implicated in the first step of the electron transport pathway in Marinobacter hydrocarbonoclasticus, it is concluded that this pathway is not conserved in haloarchaea. Surprisingly, some haloarchaeal residues in these positions are more similar to those in Wolinella succinogenes NosZ, which belong to the Clade II of NosZ. These residues, Ala 545 , Asp 569 and His 616 , are replaced by Arg, Glu, and Ser, respectively (Gln 529 and Val 524 are residues with a certain variability between species) [83]. Furthermore, Arg 557 and Cys 627 have been proposed to carry out electron transport pathway between c-haem-containing domain and Cu A in Wolinella succinogenes [83]. In fact, these residues are well conserved in haloarchaea. Nevertheless, haloarchaeal NosZ lack c-haem-containing domains. Figure 8B,C shows the possible electron binding site on Haloferax mediterranei NosZ. Transport from Cu A to Cu Z is more conserved. The two proposed pathways are possible because all mentioned residues are conservatively replaced. The most probable route only changes Trp 613 for another aromatic residue (except from Halorubrum aidingense), whereas the alternative route replaces Met 620 by Leu.

Conclusions
Sequence analysis of the main denitrification enzymes indicates that they are associated with the plasma membrane on the extracellular side. The conserved lipobox in NirK and NosZ point out that both could be membrane-bound lipoproteins that are exported through a TAT system. Although it is not known with certainty which is the electron donor protein to NarH and how it is associated with the plasma membrane. Even though the orientation in the membrane is reverse, the residues involved in cofactors coordination and in the electron transfer, are highly conserved in haloarchaeal nitrate reductases. Catalytic mechanism and electron transfer in NirK and NosZ are likely remarkably similar to bacteria, as well. The structural models of the enzymes make possible the identification of key residues involved in the discrimination of substrates, electron transfer or catalytic activities allowing the design of directed mutagenesis experiments to confirm the role of these residues. As nitrate reductase from Haloferax mediterranei binds its substrate from the extracellular media, where is expected to be more diluted than in cytoplasm, the substrate affinity must be higher when comparing with bacterial nitrate reductase, as occurs in the comparation with Escherichia coli nitrate reductase. Directed mutagenesis experiments will be needed to confirm the role of the residues assigned to substrate binding and responsible for the binding affinity. Nevertheless, further structural, and spectroscopic analyses are required to confirm the mechanisms responsible for Haloferax mediterranei denitrification to identify the intermediate electron donors and acceptors and to describe the bioenergetics associated with this respiratory pathway configuration. Clearly, the description of haloarchaeal denitrification remains in their infancy compared to denitrification pathways in bacterial model. Importantly, there is a notable lack of crystallographic structural data for haloarchaeal denitrification enzymes due to overexpression problems and low protein stabilization. For this reason, the optimization of the overexpression systems for haloarchaeal protein is necessary. These methodological improvements could increase the number of solved haloarchaeal crystallographic structures; therefore, shedding some light on the electron pathways and catalytic mechanism involved. In conclusion, our results set the basis for the understanding of the molecular mechanism involved in denitrification, which will help in the optimization of the use of these microorganisms in bioremediation of wastewater and brines in the near future.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/biom11071043/s1, Figure S1: Complete alignment of NarG sequences of bacteria and haloarchaeal species included in this study, Figure S2: Complete alignment of NirK sequences of bacteria and haloarchaeal species included in this study, Figure S3: Complete alignment of NirK sequences of Clade I and Clade II species included in this study, Figure S4: Complete alignment of NosZ sequencesof bacteria and haloarchaeal species included in this study, Table S1: Interaction distance (Angstroms) between the enzyme residues and the metal centres in NarG, NirK and NosZ templates and models.