**The Requirement of Inorganic Fe-S Clusters for the Biosynthesis of the Organometallic Molybdenum Cofactor**

#### **Ralf R. Mendel 1, Thomas W. Hercher 1, Arkadiusz Zupok 2, Muhammad A. Hasnat <sup>2</sup> and Silke Leimkühler 2,\***


Received: 18 June 2020; Accepted: 14 July 2020; Published: 16 July 2020

**Abstract:** Iron-sulfur (Fe-S) clusters are essential protein cofactors. In enzymes, they are present either in the rhombic [2Fe-2S] or the cubic [4Fe-4S] form, where they are involved in catalysis and electron transfer and in the biosynthesis of metal-containing prosthetic groups like the molybdenum cofactor (Moco). Here, we give an overview of the assembly of Fe-S clusters in bacteria and humans and present their connection to the Moco biosynthesis pathway. In all organisms, Fe-S cluster assembly starts with the abstraction of sulfur from l-cysteine and its transfer to a scaffold protein. After formation, Fe-S clusters are transferred to carrier proteins that insert them into recipient apo-proteins. In eukaryotes like humans and plants, Fe-S cluster assembly takes place both in mitochondria and in the cytosol. Both Moco biosynthesis and Fe-S cluster assembly are highly conserved among all kingdoms of life. Moco is a tricyclic pterin compound with molybdenum coordinated through its unique dithiolene group. Moco biosynthesis begins in the mitochondria in a Fe-S cluster dependent step involving radical/S-adenosylmethionine (SAM) chemistry. An intermediate is transferred to the cytosol where the dithiolene group is formed, to which molybdenum is finally added. Further connections between Fe-S cluster assembly and Moco biosynthesis are discussed in detail.

**Keywords:** Moco biosynthesis; Fe-S cluster assembly; l-cysteine desulfurase; ISC; SUF; NIF; iron; molybdenum; sulfur

#### **1. Introduction**

As one of the most abundant metals on earth, iron naturally is one of the prevalent metal ions in biological systems [1]. Iron is a major constituent of iron-sulfur (Fe-S) clusters and plays an important role in life on earth. Fe-S centers are essential protein cofactors in all forms of life [2]. They are involved in many key biological pathways including the metabolism of carbon, nitrogen and sulfur, photosynthesis, respiration, biosynthesis of antibiotics, protein translation, replication and DNA repair, gene regulation, protection from oxidizing agents, and neurotransmission. In particular, Fe-S centers are not only involved as enzyme cofactors in catalysis and electron transfer, but they have been revealed to be essential for the assembly of other metal-containing cofactors.

The most common clusters are [2Fe-2S], [3Fe-4S], and [4Fe-4S], and they are the most versatile and presumably oldest cofactors of proteins in the cell (Figure 1). Their synthesis and insertion into apo-proteins require the function of complex cellular machinery [2–4]. In addition to the roles named above, Fe-S cluster-containing proteins play critical roles in the assembly of other metal-containing enzymes or metal-containing cofactors such as the molybdenum cofactor (Moco). In this case, Fe-S cluster assembly has to precede the biosynthesis of this metal-dependent molecule.

**Figure 1.** Three common types of Fe-S clusters. Shown are the structures of a rhombic [2Fe-2S] and a cubane [4Fe-4S] cluster. The [3Fe-4S] cluster can be generated by the loss of iron from a [4Fe-4S] cluster. Sulfur is symbolized in yellow and iron is represented in red.

Fe-S clusters were first discovered in the early 1960s by H. Beinert, R.H. Sands, and others, in photosynthetic organisms [5], nitrogen-fixing bacteria [6], and sub-mitochondrial fractions of mammalian origin [7]. To date, numerous different types of proteins or enzymes containing diverse Fe-S clusters have been identified [2]. Fe-S containing proteins are ubiquitous and unarguably constitute the oldest but the structurally-heterogeneous class of proteins in biology. Fe-S enzymes are quite diverse in function and many of them catalyze key redox reactions in central metabolism under both aerobic and anaerobic conditions. Fe-S clusters can form spontaneously in solution from Fe2+/Fe3+, R-S−, S2−; or from Fe2+, R-SH, and sulfur [8], and it has been suggested that the first biologically meaningful reactions were catalyzed by Fe-S clusters [9]. Presumably, the first Fe-S proteins arose from the incorporation of preformed inorganic Fe-S clusters into polypeptides. In the world as we currently know it, the process of Fe-S cluster biosynthesis in living organisms turns out to be highly regulated and is catalyzed by numerous biogenesis factors that are remarkably conserved among prokaryotes and eukaryotes [4,10].

The complexity of Fe-S cluster biosynthesis became evident in 1998, when a complex gene cluster was discovered in bacteria that codes for proteins that are involved in their controlled assembly [11]. In eukaryotes, the mitochondria were identified as the primary compartment for Fe-S cluster assembly and were found to contain a very similar system as found in prokaryotes [12]. Since the late 1990s, the proteins involved in Fe-S cluster biosynthesis have been studied and characterized extensively. While these studies provided a general outline of in vitro and in vivo Fe-S cluster assembly, a number of major questions remain to be answered. Remaining gaps in our knowledge are: how Fe-S clusters are transferred to their target proteins, how specificity in this process is achieved and, in particular, how the iron for cluster assembly is provided in the cell.

Among the most recent additions to the field of Fe-S dependent enzymes was the discovery of the superfamily of radical/S-adenosylmethionine (radical/SAM) enzymes in 2001 [13]. These enzymes utilize a [4Fe-4S] cluster and SAM to initiate a diverse set of radical reactions, in most cases via the generation of a 5 -deoxyadenosyl radical intermediate. While this superfamily has been already identified in 2001 by studies that were mainly based on bioinformatics, the discovery of new Fe-S containing enzymes that employ SAM to initiate radical reactions still continues. Especially in recent years, it has become obvious that most reaction pathways for the synthesis of complex metal-containing cofactors have recruited radical/SAM chemistry [14]. One example has been the identification of the mechanism of the radical/SAM enzyme MoaA, a GTP 3 ,8-cyclase in the biosynthesis of Moco of the diverse class of molybdoenzymes [15,16]. The number of known radical/SAM-dependent enzymes grew exponentially during the last years, with an initial identification of 600 members of the superfamily by Sofia et al. in 2001 that until today has increased to more than 113,000 members [14]. These enzymes are found across species and catalyze a diverse set of reactions, the vast majority of which have yet to be characterized. Due to their functional diversity, most cellular processes depend on this superfamily of [4Fe-4S]-containing enzymes.

#### **2. The Assembly of Fe-S Clusters in Bacteria**

Three main Fe-S cluster assembly pathways have been identified to date, namely the NIF (nitrogen fixation) system, the SUF (sulfur formation) system, and the ISC (iron sulfur cluster) system (reviewed in Reference [4]) (Figure 2). The three systems have different phylogenetic distributions throughout the three kingdoms of life. For example, in cyanobacteria, the SUF pathway is the major system for Fe-S cluster assembly, while in *Escherichia coli* the ISC has the predominant role, while the SUF pathway is more important under stress conditions [17]. Furthermore, in gram-positive pathogens such as *Mycobacteria* or *Clostridia*, as well as some Archaea, the SUF pathway is essential. In other bacteria like the plant-pathogenic bacterium *Dickeya dandatii*, all three ISC, SUF, and NIF systems are present [18]. In eukaryotes, the Fe-S cluster assembly pathway is further complicated by a different localization to specific organelles [19]. Homologues of the ISC pathway are present predominantly in mitochondria, while SUF homologues are restricted to the chloroplasts of some photosynthetic organisms (Figure 2). In addition, a cytosolic iron sulfur cluster (CIA) machinery is present that appears to be distinct from the SUF and ISC pathway.

**Figure 2.** Possible links of Fe-S cluster assembly systems in prokaryotes and eukaryotes. Bacterial organisms harbor different complements of the NIF (nitrogen fixation), ISC (iron-sulfur cluster), and SUF (sulfur mobilization) systems. The NIF system is specialized for the assembly of nitrogenase in azototrophic bacteria. In bacteria like *E. coli*, both the ISC and SUF systems are present, while Pseudomonads contain only the ISC system and Cyanobacteria contain only the SUF system, and in *Dickeya dandatii* all three systems are present. The ISC assembly machinery of mitochondria is likely inherited from an ancestor of α-proteobacteria, the evolutionary origin of these organelles. The SUF machinery of plastids in higher plants has been likely inherited by endosymbiosis of a photosynthetic bacterium. The cytosolic iron-sulfur protein assembly (CIA) machinery for the maturation of cytosolic and nuclear Fe-S proteins depends on the mitochondrial ISC assembly machinery. These three systems are highly conserved in eukaryotes from humans to yeast and plants.

Fe-S clusters are mainly bound to proteins by cysteine or histidine residues either in the rhombic [2Fe-2S] or the cubic [4Fe-4S] forms (Figure 1) [20]. In model organisms, like *E. coli*, the Fe-S cluster assembly pathways have been well studied. *E. coli* possesses two systems for Fe-S cluster assembly, which share the same basic principles in cluster assembly. While the ISC machinery is transcribed

by the *iscRSUA-hscBA-fdx-iscX* operon, the SUF machinery is organized in the *sufABCDSE* operon (Figure 3) [21–23]. As a starting point, the l-cysteine desulfurases IscS or SufS convert l-cysteine to l-alanine and provide the sulfur in the form of a protein-bound persulfide [23]. IscS was shown to act as housekeeping l-cysteine desulfurase [11], while SufS acts under conditions of iron-limitation and oxidative stress [24]. For house-keeping Fe-S cluster assembly, IscS interacts with IscU, thereby making IscU accessible to receive the persulfide sulfur from IscS. In this step, IscU serves as a scaffold protein for the initial assembly of [2Fe-2S] clusters and [4Fe-4S] clusters (Figure 3) [25,26]. The iron source for nascent Fe-S cluster formation has not been identified yet; however, several proteins have been discussed as candidates [27]. IscS and IscU proteins form a heterotetrameric complex together with CyaY, the bacterial homologue to frataxin (see below) [26,28]. During this step of the assembly of Fe-S clusters electrons are required for persulfide reduction. These electrons are most likely provided by ferredoxin (Fdx) [29–31]. Initially, one [2Fe-2S] cluster is formed per IscU monomer inducing a conformational change within the IscU protein that decreases the stability of the IscS–IscU interaction [23]. This step is expected to further enable the reductive coupling of two [2Fe-2S] clusters to form one single [4Fe-4S] cluster on IscU.

**Figure 3.** A model for the assembly of Fe-S clusters in *E. coli* by the ISC and SUF machinery. For each system, the proteins involved in each step are indicated in addition to their operon organization. The building of the Fe-S clusters in both systems is facilitated by the scaffold proteins IscU or SufB. After the formation of [2Fe-2S] or [4Fe-4S] clusters on IscU, their release is catalyzed with the help of the HscBA co/chaperones in the ISC system. This step in particular still needs to be clarified in the SUF machinery. After formation, the clusters are transferred by the A-type carrier proteins IscA and SufA to target proteins. Both systems are tightly regulated by IscR at the level of Fe-S cluster availability. Under normal conditions, the IscR regulator exists in an [2Fe-2S] cluster bound form and represses its own expression to control Fe-S cluster formation and the Fe-S state of the cell. Under iron limitation, the IscR regulator is accumulated and converted to its apo-form that activates the *suf* operon. The model shows a simplified version of the regulation of both systems, not depicting the regulation by small RNAs or oxygen.

For the release of Fe-S clusters from IscU, HscA, and HscB are involved in an ATP-dependent manner, two members of the DnaK/DnaJ chaperones/co-chaperone family [32]. HscA recognizes a specific motif on IscU, and their interaction is additionally regulated by the co-chaperone, HscB. The mechanism by which the chaperone facilitates cluster release from IscU has been proposed to involve two conformational states of IscU with different affinities to the bound Fe-S clusters [33]. The chaperones thereby favor the low-affinity IscU state and facilitate the release of the Fe-S cluster from IscU.

In comparison, the SUF system has also been well characterized from studies in bacteria [4,24,34–37]. Here, SufS forms a complex with SufE, which together mobilize the sulfur for cluster assembly. In the SUF machinery, SufB is the Fe-S scaffold protein that acts in conjunction with SufC (and in some cases additionally with SufD) [38]. After Fe-S cluster formation, SufA then transfers the Fe-S clusters to target apo-proteins [39] (Figure 3).

Numerous Fe-S carriers have been identified in both prokaryotes and eukaryotes [4]. These include as main carriers the so-called A-type carriers (ATC) IscA, SufA, and ErpA (Figure 4) [40,41]. Other carriers include the highly-conserved NFU-type proteins [42], the monothiol glutaredoxins (Grx 5 in yeast and GrxD in *E. coli*) [43,44], or the P-loop NTPases, (Ind1 in mitochondria, ApbC in Salmonella) [45,46].

Previous phylogenetic studies had classified ErpA and IscA into two different families; while ErpA belongs to family ATC-I, IscA was grouped into family ATC-II [40]. While ATC-I family members interact with the apo-target proteins, the ATC-II family members are predicted to interact with the scaffold proteins instead. However, the ATC proteins were also shown to replace each other in their roles.

The expression of the SUF and ISC system has been revealed to be tightly regulated in *E. coli*. While the ISC system is the house-keeping Fe-S cluster assembly system, the SUF system instead is mainly synthesized under iron-limiting conditions [24]. One of the main regulators that regulate the expression of either the ISC system or the SUF system is the IscR protein [4]. IscR is a transcriptional regulator that exists in the apo-form and in a [2Fe-2S] cluster bound form in the cell [47]. In its [2Fe-2S] cluster bound form, IscR represses its own expression in addition to that of *iscRSUA-hscBA-fdx-iscX.* In contrast, in its apo-form, IscR activates the expression of the SUF system (Figure 3) [4]. This mechanism allows IscR to fine-tune Fe-S cluster synthesis in response to the presence of synthesized Fe-S clusters and iron availability in the cell.

**Figure 4.** Moco biosynthesis and the link to Fe-S cluster assembly in *E. coli*. In Moco biosynthesis, Mo-MPT is formed from 5 GTP with cPMP and MPT as stable intermediates. For enzymes of the DMSO reductase family like TorCA or NarGHI, bis-MGD is formed by the addition of GMP to each MPT unit in the bis-Mo-MPT intermediate. For enzymes like PaoABC, Mo-MPT is further modified by the addition of CMP to form the MCD form of the cofactor. Molybdoenzymes, in general, are complex enzymes, containing additional cofactors like FAD, Fe-S clusters, or hemes. The sulfur for the synthesis of the dithiolene group in Moco is mobilized by IscS with the additional involvement of the TusA protein. CyaY interacts with the IscS-IscU complex and forms a central heterotetramer. Fe-S clusters assemble on the scaffold protein IscU, which receives the sulfur from the l-cysteine desulfurase IscS. Assembly and release of the clusters is catalyzed by the chaperones HscAB. Ferredoxin is delivering the electrons. The carrier proteins IscA and ErpA deliver the matured Fe-S clusters to recipient protein among which is MoaA in Moco biosynthesis.

#### **3. A General Scheme for the Biosynthesis of the Molybdenum Cofactor**

The biosynthesis of the molybdenum cofactor (Moco) is highly conserved among all kingdoms of life. The chemical nature of Moco was first determined by Rajagopalan and coworkers in 1982 [48]. The pathway for the biosynthesis can generally be divided into three steps, with the characteristic of each step being the formation of a stable intermediate (Figure 4) [49–52]. The first step represents the synthesis of cyclic pyranopterin monophosphate (cPMP) from 5 -GTP [53], the second step the introduction of two sulfur atoms into cPMP forming molybdopterin (MPT) by [54], and the third step is the insertion of molybdate into MPT, resulting in the formation of the so-called Mo-MPT cofactor [55]. In prokaryotes, Mo-MPT is further modified by the attachment of GMP or CMP to the phosphate group of MPT, forming the two dinucleotide variants of Moco, MPT-guanine dinucleotide (MGD), [56] and MPT-cytosine dinucleotide (MCD) [57,58] (Figure 4). The characteristics of different forms of Moco are represented further by different ligands at the molybdenum atom. This resulted in the categorization

of molybdoenzymes into three families (Figure 5): the xanthine oxidase (XO) family, the sulfite oxidase (SO) family, and the dimethyl sulfoxide (DMSO) reductase family. Enzymes of the DMSO reductase family generally contain two MPT moieties ligated to the molybdenum atom and are present only in prokaryotes [50]. Here, we will briefly summarize the biosynthesis of Moco.

**Figure 5.** The three families of molybdoenzymes in pro- and eukaryotes. Moco exists in different variants and is divided into three enzyme families. The basic form of Moco is Mo-MPT that coordinates the molybdenum atom in a tri-oxo form. The SO family is present both in pro-and eukaryotes and is characterized by a molybdenum ligation with one oxo-, one hydroxide- and one cysteine ligand from the protein backbone. The XO family is also present in both pro- and eukaryotes and is characterized by an equatorial sulfido-ligand, an apical oxo ligand, and one hydroxide or O- ligand. The XO family contains the sulfurated form of Mo-MPT. In *E. coli*, an additional nucleotide is present in this family, forming the molybdopterin cytosine dinucleotide cofactor (MCD). The DMSO reductase family contains two MPTs (bis-Mo-MPT) or two MGDs (bis-MGD) ligated to one molybdenum atom with additional ligands being an O/S and a sixth ligand which can be a serine, a cysteine, a selenocysteine, an aspartate or a hydroxide ligand.

In the first step of Moco biosynthesis, 5 -GTP is converted to cPMP (Figure 4) [53]. cPMP is a 6-alkyl pterin with a cyclic phosphate group at the C2 and C4 atoms [59]. This reaction is catalyzed by two enzymes, MoaA and MoaC [53,60] (Figure 4). While the individual catalytic functions of MoaA and MoaC have long been unknown, recent studies showed that MoaA catalyzes the conversion of GTP to (8S)-3 ,8-cyclo-7,8-dihydroguanosine 5 triphosphate (3 ,8-cH2GTP), and MoaC catalyzes the conversion of 3 ,8-cH2GTP to cPMP [15]. Details of the recent updates and the details of MoaA mechanism, where [4Fe-4S] clusters play central roles, will be discussed in the next chapter.

In the second step of Moco biosynthesis, two sulfur atoms are inserted into cPMP, and MPT is formed as a stable intermediate [54,61–65]. This reaction is catalyzed by MPT synthase, composed of two MoaD and two MoaE subunits (Figure 4) [66]. The sulfur atoms required for this reaction are present at the C-terminus of MoaD in form of a thiocarboxylate group [67,68]. Studies on the reaction mechanism showed that the first sulfur is added at the C2 position of cPMP in a reaction that is coupled to the hydrolysis of the cPMP cyclic phosphate group [65]. The second sulfur is then transferred to the C1 of the hemisulfurated intermediate and MPT is formed as the product.

In the third step of Moco biosynthesis, molybdate is inserted to the dithiolene sulfurs of MPT and Mo-MPT is formed, one of the forms of Moco. The specific insertion of molybdenum into MPT is catalyzed by the joined action of MoeA and MogA (Figure 4) [69,70]. MogA thereby hydrolyzes ATP and forms the activated MPT-AMP intermediate [71]. This intermediate is then transferred to MoeA, which mediates molybdenum ligation at low concentrations of MoO4 <sup>2</sup><sup>−</sup> [70]. Mo-MPT can be further modified by nucleotide addition in the next step of Moco biosynthesis, which is present only in prokaryotes (Figure 5) [72]. Alternatively, the Mo-MPT cofactor can be directly inserted into enzymes of the SO family. In this family, the molybdenum atom of Mo-MPT is coordinated by a cysteine from the polypeptide backbone of the protein, representing the cofactor in its MPT-MoVIO2 form [50].

The proteins of the DMSO reductase family in bacteria contain the bis-MGD cofactor (Figure 5) [50]. The synthesis of the bis-MGD was shown to occur in a two-step reaction catalysed by MobA using Mo-MPT and Mg-GTP as substrates [72]. In the first reaction, the bis-Mo-MPT intermediate is formed on MobA from two Mo-MPT molecules [73]. In the second reaction, two GMP moieties from GTP are added to the C4 phosphate of bis-Mo-MPT, forming the bis-MGD cofactor [74,75]. After the formation of bis-MGD, the cofactor can be released from MobA and inserted into target enzymes. However, it is expected that bis-MGD does not exist in a free form in the cell. Formed bis-MGD is rather expected to be immediately recruited by Moco-binding chaperones that protect the cofactor from oxidation and specifically interact with their target apo-molybdoenzymes for bis-MGD insertion. After insertion, bis-MGD is generally ligated by a serine, a cysteine, a selenocysteine, or an aspartate from the protein backbone. The other ligand in enzymes of the DMSO reductase family that bind the bis-MGD cofactor is an oxo- or sulfido atom.

The addition of CMP results in the formation of the MPT-cytosine dinucleotide (MCD) cofactor [58], a cofactor that is present in the xanthine oxidase family of molybdoenzymes in bacteria (Figure 5). The formation of MCD is catalyzed by MocA (Moco cytidylyltransferase). After the formation of MCD and the bis-MGD cofactor, a further modification at the molybdenum atom can occur by the addition of a terminal sulfido-ligand [76]. The sulfido-ligand at the equatorial position of the molybdenum atom is a characteristic feature of enzymes in the xanthine oxidase family. In eukaryotes, enzymes of this family do not harbor the additional CMP modification of the cofactor (Figure 5).

#### **4. Linking Moco Biosynthesis and Fe-S Cluster Assembly in Bacteria**

More than 60 different Moco-containing molybdoenzymes have been identified to date [77]. In recent years it has become evident that the biosynthesis of Moco and the assembly of Fe-S clusters are directly connected to each other. Moco biosynthesis directly depends on the presence of Fe-S clusters or components of the Fe-S cluster assembly machinery on several levels (Figure 4). Many molybdoenzymes bind Fe-S clusters as additional cofactors that are involved in intramolecular electron transfer reactions. In Moco biosynthesis, the MoaA protein harbors two [4Fe-4S] clusters and thus directly depends on Fe-S cluster assembly. In addition, the l-cysteine desulfurase IscS is shared between Fe-S cluster assembly and Moco biosynthesis since it also mobilizes the sulfur for the synthesis of the dithiolene group present in Moco. Further, the expression of most molybdoenzymes and proteins involved in Moco biosynthesis in bacteria is regulated by the transcriptional regulator for **f**umarate and **n**itrate **r**eduction, FNR [78,79]. The activity of FNR itself is directly dependent on the availability of Fe-S clusters under anaerobic conditions; consequently, Moco is not synthesized and molybdoenzymes are not expressed when Fe-S clusters are not assembled (Figure 4).

#### *4.1. The Involvement of Radical SAM Chemistry for Moco Biosynthesis*

The first step of Moco biosynthesis, the conversion of GTP into cPMP, directly depends on the assembly of [4Fe-4S] clusters, which proceeds through a complex rearrangement reaction, where C8 of guanine is being inserted between C2 and C3 of ribose [60]. During this reaction, MoaA plays a key role. MoaA is the only protein that binds Fe-S clusters in the pathway of Moco biosynthesis and was grouped into the superfamily of radical/SAM enzymes. In general, the mechanism of radical SAM

chemistry involves the [4Fe-4S]2<sup>+</sup> cluster bound to the C-X3-C-X2-C motif located at the N-terminus of radical SAM superfamily enzymes (Figure 6).

**Figure 6.** A general scheme for the mechanism of radical-SAM enzymes. The [4Fe-4S] that is bound to the conserved C-X3-C-X2-C motif provides the electron required for the reductive cleavage of SAM. In this reaction, methionine and the adenosyl radical (dAdo·) is generated. The formed dAdo· then abstracts a hydrogen atom from the substrate (SH), and a substrate radical (S·) and dAdoH are formed.

The cluster is reduced by one electron to the catalytically-active state and coordinates the substrate 5 S-adenosyl-methionine (SAM). In the reduced [4Fe-4S]<sup>+</sup> state, the cluster transfers an electron to the sulfonium sulfur of SAM, thereby promoting homolytic S–C bond cleavage to generate the 5 -deoxyadenosyl radical intermediate (dAdo·) and methionine. The dAdo· intermediate then abstracts a hydrogen atom from the substrate. The 5 -deoxyadenosine (dAdoH) is formed, and the resulting product radical intermediate may either be the end product or can undergo further transformations (Figure 6).

In the first step of Moco biosynthesis, MoaA and MoaC catalyze the conversion of 5 GTP into cyclic pyranopterin monophosphate (cPMP). MoaA is a member of the radical SAM superfamily and binds the characteristic [4Fe-4S] cluster at the N-terminus and an additional [4Fe-4S] at the C-terminus. For Fe-S cluster insertion into MoaA, recent studies showed that both ErpA and IscA can provide Fe-S clusters to MoaA. Since Δ*erpA*/Δ*iscA* double mutant strains were revealed to be completely devoid of Moco, it was concluded that SufA is unable to substitute the roles of both A-type carrier proteins for Moco biosynthesis (unpublished results).

The mechanism of cPMP formation was first investigated using isotope labeling experiments, which indicated that the C-8 atom of the guanine base of GTP is inserted between the C2 and C3 atoms of the ribose moiety.

Recent studies showed that MoaA catalyzes the conversion of GTP into a 3 ,8-cyclo-7,8 dihydroguanosine 5 -triphosphate (3 ,8-cH2GTP) intermediate (Figure 7). The conversion of GTP into 3 ,8-cH2GTP by MoaA proceeds through a radical formation at C3 by the abstraction of the H-3 atom of GTP by 5 -dAdo· [15,80]. The free radical generated at the C3 of the ribose has been revealed by isotope labeling studies [15,80], where a deuterium atom at the 3 -position was shown to be transferred to dAdoH. The resulting C3 centered radical attacks C8 to form the aminyl radical intermediate, which is then reduced by transfer of an electron and a proton to form 3 ,8-cH2GTP. Therefore, MoaA catalyzes the C–C bond formation between the GTP C3 and the C8 of the guanine, resulting in the

3 ,8-cH2GTP intermediate as the end product of the MoaA catalyzed reaction. The conversion of 3 ,8-cH2GTP to cPMP is catalyzed by MoaC in the next step of the reaction [15]. It has been proposed that two loops in MoaC provide conformational flexibility of the enzyme that facilitates a general acid/base-catalyzed mechanism for the formation of the pyranopterin structure that is coupled to the cyclic phosphate ring formation (Figure 7). This step catalyzed by MoaC involves the irreversible cleavage of the pyrophosphate group that has been proposed to provide the thermodynamic driving force for the overall reaction.

**Figure 7.** The conversion of 5 GTP to cPMP involving the functions of MoaA and MoaC. The colors on 5 GTP and cPMP indicate the source of the carbon and nitrogen atoms in cPMP as determined by isotope labeling studies. The reaction involves the formation of 3 ,8-cH2GTP as the product of MoaA and the substrate for MoaC.

#### *4.2. Sulfur Mobilization Involves Sharing of Protein Functions in Prokaryotes*

In the next step of Moco biosynthesis, cPMP is converted to MPT in a reaction catalyzed by MPT synthase [54,61–64]. In this reaction, two sulfur atoms are inserted into the C2 position of cPMP and the C1 position of the formed hemisulfurated intermediate [65]. In bacteria, MPT synthase is composed as a (αβ)2 heterotetramer of two central MoaE subunits, each binding one MoaD subunit (Figure 4) [66]. The sulfur atoms required for this reaction are bound as thiocarboxylate groups at the C-terminus of each MoaD subunit [67,68]. After the sulfur transfer reaction, the thiocarboxylate group is regenerated on each MoaD subunit. This reaction is catalyzed by MoeB under ATP consumption [81,82]. In the course of the reaction, MoaD dissociates from MoaE and reassociates with MoeB in a (αβ)2 heterotetramer. In the first step, an acyl-adenylate group is formed at the C-terminus of MoaD [81,83,84]. In a second step, sulfur is directly transferred from a sulfur transferase to the activated MoaD-AMP C-terminus, and the MoaD thiocarboxylate (MoaD-COSH) is rebuilt. MoaD-COSH then dissociates from the (MoeB-MoaD)2 complex and MoaD reassociates with MoaE [84,85]. The proteins IscS and TusA are proposed to be involved in the sulfur transfer reaction for the formation of the MoaD thiocarboxylate group in *E. coli* (Figure 4) [86,87]. It has been proposed that IscS first forms a persulfide group on TusA that is then reductively cleaved and transferred to MoaD by attacking the MoaD–AMP bond [88,89].

After MPT formation, the dithiolene group of MPT serves as the backbone for molybdenum ligation. This step is catalyzed by MogA and MoeA under ATP consumption by the formation of an MPT-AMP intermediate catalyzed by MogA. After molybdate insertion into MPT-AMP in a MoeA-dependent reaction, the resulting tri-oxo Mo-MPT is either inserted into enzymes of the SO family or is further modified by nucleotide (GMP or CMP) addition, forming the MCD or bis-MGD cofactors, respectively.

#### *4.3. The Insertion of Di*ff*erent Cofactors into Molybdoenzymes in Bacteria*

Molybdoenzymes are generally composed of different subunits harboring additional prosthetic groups, such as cytochromes, the Fe-S cluster, or FAD/FMN that are involved in intramolecular electron transfer reactions (Figure 4) [50]. The molybdenum atom thereby exists in the oxidation states VI, V, or IV under physiological conditions and acts as a transducer between 2 electron transfer and 1 electron transfer processes often coupled to proton transfer.

In general, Moco is deeply buried within the enzyme that is accessible via a substrate-binding funnel leading to the molybdenum atom [90]. It has been suggested that Moco insertion is accomplished by molecular chaperones that induce the final folding of the enzymes after Moco insertion [91]. The insertion of the different forms of Moco has been best studied in bacteria. In *E. coli*, many molybdoenzymes are located in the periplasm and require the Tat system for their translocation. The transport of enzymes occurs in the folded state after the insertion of the Moco and Fe-S clusters or other cofactors. This translocation presents a "quality control" step accomplished by the chaperones as Moco insertases and ensures that only the matured enzymes that contain Moco are translocated. After subunit assembly, Moco is inserted, final folding of the enzyme is accomplished, the enzyme is directed to the Tat-translocon and finally exported to the periplasm.

For Fe-S cluster insertion into enzymes, several A-type carriers have been identified that facilitate this process [40]. These proteins, which are SufA, IscA, or ErpA in *E. coli,* carry and insert the Fe-S clusters to designated target proteins. Previous reports suggested that ErpA is essential for the formation of an active formate-nitrate reductase complex in *E. coli* [92]. It was shown that *E. coli erpA* mutant strains were devoid of formate dehydrogenase and nitrate reductase activities. In these studies, IscA was able to partially complement the *erpA* mutant, showing that these proteins might have overlapping roles, but ErpA seems to be the more specific enzyme for nitrate reductase and formate reductase maturation [92]. However, the overlapping effects of Fe-S cluster assembly and biosynthesis and insertion of Moco were not accounted for in this study. Since MoaA contains Fe-S clusters, the effect of the *erpA* mutant might have already been a decrease in MoaA activity, leading to a lack of Moco in the cell. Recent studies revealed that NarG is not expressed in *erpA* mutant strains and that inactive FNR precludes the expression of the *narGHJI* operon (unpublished results). Additionally, it was shown that ErpA and IscA are involved in Fe-S cluster insertion into MoaA (unpublished results), so that the inactivity of nitrate reductase and formate dehydrogenase in *erpA* mutant strains is likely rather based on the lack of Moco. It still remains possible that, in addition, ErpA and IscA are involved in Fe-S cluster insertion into these molybdoenzymes. However, based on the lack of expression of the FNR-regulated operons, the involvement of these proteins cannot be analyzed. Thus, it is difficult to dissect the combined effects of Moco biosynthesis and Fe-S cluster insertion for molybdoenzyme maturation.

#### **5. Compartmentalization of Fe-S Cluster and Moco Biosynthesis in Eukaryotes: The Role of Mitochondria**

#### *5.1. Mitochondrial Fe-S Cluster Biosynthesis in Eukaryotes*

In eukaryotes, the main Fe-S cluster assembly is localized in the mitochondria, which were derived from bacteria by endosymbiosis. The key component for Fe-S cluster biosynthesis in mitochondria is the l-cysteine desulfurase NFS1, which forms the central dimer (Figure 8).

The enzyme serves as the general sulfur donor for cellular Fe-S cluster synthesis, also for cytosolic Fe-S clusters. At the NFS1 dimer interface, a dimer of ISD11 is bound that stabilizes NFS1 [93–95]. While ISD11 is conserved in eukaryotes [96], a prokaryotic homolog of ISD11 is not known [93]. ISD11 belongs to the large family of LYRM proteins that fold into a triple-helical bundle [97]. ISD11 contributes to the interaction with the acyl carrier protein ACP1 [93]. While ACP1 is not needed for efficient synthesis of the [2Fe-2S] cluster on ISCU2, the ISD11-ACP1 sub-complex was proposed to regulate Fe-S cluster assembly by linking the energy load of the cells to the Fe-S cluster assembly complex [98]. Each monomer of the NFS1 dimer binds frataxin (FXN), ISCU2, and FDX2 at the two opposite ends. For the conversion of the persulfide sulfur (S0) to sulfide (S2<sup>−</sup>), electrons are required, which are provided by FDX2 [99–102]. First, a [2Fe-2S] cluster is formed on ISCU2, by combining Fe2<sup>+</sup> entry from a still unresolved iron donor and S2<sup>−</sup> provided by NFS1. The role of FXN in this process is still under debate.

The next step in mitochondrial Fe-S protein biogenesis involves the dissociation of the [2Fe-2S] cluster from ISCU2, its trafficking to the monothiol glutaredoxin GLRX5 as an intermediate Fe-S cluster binding partner, and the specific insertion into target proteins (Figure 8) [103–105]. This reaction requires the roles of the dedicated chaperone system of the Hsp40-Hsp70 (DnaJ-DnaK) class [106,107]. For cluster transfer, the DnaJ co-chaperone HSC20 binds to a hydrophobic motif in ISCU2 [97,108,109]. The HSC20-ISCU2 complex then recruits the Hsp70 chaperone HSPA9. In the next step, the cluster is transferred to GLRX5. For [2Fe-2S] cluster transfer to target proteins, no further proteins are required [110]. However, since GLRX5 is not essential, the protein function can be bypassed by direct transfer of [2Fe-2S] from ISCU2.

**Figure 8.** The different compartments for the biosynthesis of Moco in humans and the link to Fe-S cluster assembly in mitochondria. The first step of Moco biosynthesis, the conversion of 5'GTP to cPMP catalyzed by MOCS1A and MOCS1B, is localized in mitochondria. This is also the main compartment for Fe-S cluster biosynthesis in eukaryotes. Fe-S clusters assemble on the scaffold protein ISCU2, which receives the sulfur from the l-cysteine desulfurase complex NFS1/ISD11. Frataxin (FDN) interacts with the NFS1/Isd11-ISCU2 complex to form the quaternary complex. Ferredoxin (FDX2) delivers the electrons for the process. Assembly and release of the clusters from ISCU2 are facilitated by the chaperones HSC20/HSPA9. The carrier proteins GLRX5, ISCA2, and/or IBA57 deliver the Fe-S clusters to target proteins, like MOCS1A in Moco biosynthesis. Synthesized cPMP by MOCS1A/B needs to be transferred to the cytosol, where all further modification steps of Moco are catalyzed. These steps involve the conversion of cPMP to MPT by MOCS2A/MOCS2B (which are activated by MOCS3), the insertion of molybdate by GEPHYRIN (GEPH), and the insertion of Mo-MPT either to the mARC protein (localized at the outer mitochondrial membrane) or to sulfite oxidase (before its translocation to the mitochondrial intermembrane space). For aldehyde oxidase and xanthine dehydrogenase, the formation of the equatorial sulfido-ligand is transferred by HMCS. Dual localization of NFS1 both in mitochondria and the cytosol is predicted. Cytosolic NFS1 acts as a sulfur donor for MOCS3.

The assembly and insertion of [4Fe-4S] clusters additionally require the A-type ISC proteins ISCA1-ISCA2 and IBA57 (Figure 8) [111–114]. How these proteins mechanistically assist the fusion of the [2Fe-2S] into a [4Fe-4S] cluster is a complex process involving the transfer of [2Fe-2S] clusters from GLRX5 to ISCA1-ISCA2 and the assembly of a [4Fe-4S] cluster on the ISCA1-ISCA2 complex [115–117]. IBA57 is not required in this step. IBA57 was shown to form a [2Fe-2S] cluster-mediated complex specifically with ISCA2, involving GLRX5 [111,118,119]. Mitochondrial [4Fe-4S] cluster biosynthesis is a dynamic system, also involving Nfu1, Ind1 and BolA3 in some cases (recently reviewed in Reference [120]). The Fe-S cluster insertion for the human MoaA homologue MOCS1A (see below) has not been investigated so far.

#### *5.2. Cytosolic Fe-S Cluster Assembly: The CIA Machinery*

For cytosolic Fe-S cluster assembly, the mitochondrial ISC system has been proposed to be essential [121,122]. Studies by several groups have shown that the mitochondrial ISC machinery generates a sulfur-containing factor "X-S" that is exported to the cytosol via the mitochondrial ABC transporter ABCB7 and is used for cytosolic Fe-S cluster assembly by the CIA machinery (Figure 9) [12,123–125]. However, models also exist that propose that cytosolic versions of NFS1, ISCU, and FDN are involved in cytosolic Fe-S cluster formation [120]. The CIA machinery is composed of up to 13 known proteins that assemble both cytosolic and nuclear Fe-S proteins. Initially, a [4Fe-4S] cluster is assembled on the CIA scaffold complex formed between CFD1-NBP35 (Figure 9) [125,126]. The initial cluster synthesis on CFD1-NBP35 further requires the electron transfer chain composed of the flavin-dependent oxidoreductase NDOR1 and CIAPIN1, however, the precise role of the electron-transfer for cluster synthesis is not completely understood yet [127,128]. The insertion of the two Fe-S clusters of CIAPIN1 additionally requires the cytosolic monothiol glutaredoxin reductase GLRX3 [129–131]. GLRX3 binds a bridging [2Fe-2S] cluster with BOLA2 which is further transferred to CIAPIN1 (Figure 9) [130,132]. CIAPIN1 can coordinate a pair of [2Fe-2S] clusters or a [2Fe-2S] cluster and a [4Fe-4S] cluster [115,128,133,134]. While the BOLA2-GLRX3 complex is able to transfer [2Fe-2S] clusters to CIAPIN1, the formation of the [4Fe-4S] cluster in CIAPIN1 is still fully undefined.

**Figure 9.** Model of the cytosolic Fe-S cluster assembly (CIA) in humans. Assembly of cytosolic Fe-S clusters starts in mitochondria with components of the early ISC machinery synthesizing a sulfur-containing precursor (X-S) that is subsequently exported to the cytosol by the ABC transporter ABCB7. The CIA machinery involves [4Fe-4S] cluster assembly by the CFD1-NBP35 complex serving as a cluster scaffold. This reaction requires electron input from the flavin-oxidoreductase NDOR1 and the Fe-S protein CIAPIN1. The complex of GLRX3-BOLA2 facilitates [2Fe-2S] cluster insertion into CIAPIN1. In the next step, the CFD1-NBP35-bound [4Fe-4S] cluster is released and transferred to the majority of target Fe-S apoproteins with the help of CIAO3. Fe-S cluster targeting to individual apoproteins is achieved by CIAO1, CIAO2B, and MMS19.

The next step of cytosolic Fe-S assembly involves the trafficking of the [4Fe-4S] cluster from the CFD1-NBP35 complex to CIAO3 and then to the CIA targeting complex (CTC) which is composed of CIAO1, CIAO2B, and MMS19 [135,136]. From this complex, which can be formed with different protein components, the cluster is transferred directly to target proteins (Figure 9). The specific interaction with the target proteins is mediated by the CTC proteins.

#### *5.3. The Formation of cPMP Is Localized in Mitochondria*

The conversion of 5 GTP to cPMP is localized in mitochondria in humans in a reaction that is catalyzed by MOCS1A and MOCS1B [137]. As explained above, this compartment is also the main compartment for the synthesis of Fe-S clusters (Figure 8). Since MOCS1A requires two [4Fe-4S] clusters for activity, it remains speculative that linking that first step of Moco biosynthesis to mitochondrial Fe-S cluster assembly provided a mechanistic advantage [138]. Since MOCS1A and MOCS1B are highly homologous to their bacterial counterparts, MoaA and MoaC, respectively, the conversion of 5 GTP to cPMP is catalyzed by the same mechanism and is therefore not described in detail again (Figure 7). After cPMP formation, the molecule has to be transported to the cytosol, where all further steps of Moco biosynthesis are catalyzed. For plants, it has been suggested that the export of cPMP involves the transporter protein Atm3 [139]. The human counterpart to Atm3 is ABCB7 (Figure 9). Surprisingly, Atm3, like ABCB7 has initially been suggested to transport the "X-S" species essential for cytosolic Fe-S cluster assembly. The precise role of Atm3 is consequently still unknown and the defects in molybdoenzyme activities in Atm3 mutants might also be explained by an overlapping defect in Fe-S cluster assembly. Future studies are necessary to further explore the export of cPMP from mitochondria to the cytosol.

#### *5.4. Moco Is Formed in the Cytosol*

After its transport to the cytosol, cPMP is converted to MPT in the reaction catalyzed by MPT synthase. In humans, MPT synthase is formed by MOCS2A and MOCS2B, the homologues of MoaD and MoeB, respectively (Figure 10). Both the human and bacterial MPT synthases catalyze a similar reaction that is explained in detail for the bacterial counterparts above (Figure 4) [68].

**Figure 10.** Regeneration of the thiocarboxylate group on MOCS2A. Conversion of cPMP to MPT requires the transfer of two sulfur, a reaction catalyzed by MPT synthase. The formation of the MOCS2A thiocarboxylate group is catalyzed after the formation of the (MOCS2A/MOCS3)2 complex in humans. First, MOCS2A-AMP is formed under ATP consumption. MOCS3 receives the sulfur from NFS1, and a persulfide group is formed at the C-terminal domain of MOCS3. This sulfur is then transferred to MOCS2A with the formation of a perthiocarboxylate group as the intermediate. After reductive cleavage, the thiocarboxylate group on MOCS2A is formed and MOCS2A-SH reassociates with MOCS2B.

After the MPT synthase reaction, the C-terminal carboxylate group on MOCS2A is regenerated by MOCS3 (Figure 10) [140]. MOCS3 contains an N-terminal domain that is homologous to the *E. coli* MoeB protein and a C-terminal domain that shares homologies to sulfur transferases (rhodaneses). First, the N-terminal domain of MOCS3 activates the C-terminus of MOCS2A under ATP consumption (Figure 10). In the second part of the reaction, sulfur is transferred from the C-terminal rhodanese-like domain of MOSC3 to the formed MOCS2A-AMP [141,142]. This reaction differs from the reaction described for the *E. coli* proteins, since the MOCS3 protein itself catalyzes both the adenylation and the sulfur transfer reaction. The persulfide group on MOCS3 has been suggested to be formed by the l-cysteine desulfurase NFS1 in the cytosol [143,144]. It has been proposed that small amounts of NFS1 in the cytosol are sufficient for supplying the sulfur for MPT formation [143]. Since for the CIA machinery cytosolic NFS1 might not be required, it remains possible that the role of NFS1 in the cytosol is restricted to Moco biosynthesis. Further studies are necessary to confirm this.

After the MPT formation, molybdate ion is ligated to the dithiolene group of MPT. GEPHYRIN is the homolog of the bacterial MogA and MoeA proteins (Figure 8) [145]. GEPHYRIN is a two-domain protein with one domain being homologous to MogA and the other domain being homologous to MoeA [146].

The chemistry of molybdenum insertion has been studied in detail for the plant counterpart of GEPHYRIN, named CNX1. The AMP-part of MPT-AMP functions as an anchor on the E-domain while the dithiolene moiety of MPT-AMP points to a separate pocket on the E-domain where molybdate is bound and waits to be inserted. Finally, the pyrophosphate bond between AMP and MPT is hydrolyzed and the newly-formed Moco is released [147].

After the completion of Moco, the cofactor can be directly inserted into the molybdoenzymes sulfite oxidase or mARC which belong to the SO family of molybdoenzymes (Figure 8) [148]. For the two enzymes of the XO family, xanthine dehydrogenase and aldehyde oxidase, Moco is further modified by the formation of an equatorial sulfido-group [149,150]. This reaction is catalyzed by a Moco sulfurase, named HMCS (human molybdenum cofactor sulfurase) before the insertion of the cofactor into molybdoenzymes (Figure 8). HMCS is a homodimeric two-domain protein with an N-terminal domain sharing homologies to the bacterial l-cysteine desulfurases IscS [151] and a C-terminal domain that binds Moco. The Moco-sulfido group is directly formed on Mo-MPT bound to the C-terminal domain by transfer of a persulfide group from the N-terminal domain [152,153]. After sulfuration, Moco is then inserted into the target enzymes xanthine dehydrogenase and aldehyde oxidase (Figure 8) [152,153]. So far, it is not known which enzymes of the CIA machinery insert the two [2Fe-2S] clusters into xanthine dehydrogenase or aldehyde oxidase.

#### **6. Conclusions**

In this review, we highlighted the link between the biosynthesis and maturation of molybdoenzymes and the biosynthesis and distribution of Fe-S clusters. Several levels of this link were identified: (a) the synthesis of the first intermediate in Moco biosynthesis requires a radical/SAM-dependent protein; (b) the sulfurtransferase for the dithiolene group in Moco is shared with the synthesis of Fe-S clusters; (c) the modification of the active site with a sulfur atom additionally involves an l-cysteine desulfurase, and (d) most molybdoenzymes require Fe-S clusters as additional redox-active cofactors. While the general pathways of the biosynthesis/assembly of Moco and Fe-S clusters have been studied in detail, numerous open questions still remain. One of the most intriguing questions in Fe-S cluster assembly is the source of the iron atom in the cluster. In particular, iron has to be provided in the correct oxidation state for biological processes. Further, after the biosynthesis of the complex cofactors like Fe-S clusters and Moco, intricate mechanisms have to control distribution, trafficking, and insertion of these cofactors into their specific target proteins. The transfer mechanisms involve cofactor-binding chaperones, as most of these prosthetic groups are extremely fragile and oxygen-sensitive. The specificity of the cluster transfer process into target proteins, their recognition, and the hierarchy in which metal-center insertion for Fe-S clusters and Moco occurs, is still largely unresolved. Also, the compound "X-S" that is delivered by the mitochondrial Fe-S cluster machinery to the CIA machinery is still unknown.

For humans, the molybdoenzyme sulfite oxidase is most crucial for survival and it is located in the mitochondrial intermembrane space. A lack of sulfite oxidase activity usually results in death in early childhood. The dependency on Fe-S clusters for the biosynthesis of Moco directly links Fe-S cluster related diseases to the severe outcome of sulfite oxidase deficiency. The link of Fe-S cluster related diseases to Moco deficiency needs to be investigated in future studies.

**Author Contributions:** Conceptualization, R.R.M. and S.L.; writing—original draft preparation, R.R.M. and S.L.; writing—review and editing, R.R.M., T.W.H., A.Z., M.A.H. and S.L.; visualization, S.L.; supervision, R.R.M. and S.L.; project administration, R.R.M. and S.L.; funding acquisition, R.R.M. and S.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research was supported by continuous individual grants from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) (LE1171/15-2, LE1171/11-2 to S.L. and ME1266/31-1 to R.R.M.) including funding by the DFG Priority Programme SPP1927 'FeS for Life' and by the DFG Research Training Group GRK 2223/1 'PROCOMPAS'.

**Acknowledgments:** The authors thank all current and former members of their research groups in addition to collaboration partners who were involved in the work over the past years.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Tungstoenzymes: Occurrence, Catalytic Diversity and Cofactor Synthesis**

#### **Carola S. Seelmann 1, Max Willistein 1, Johann Heider <sup>2</sup> and Matthias Boll 1,\***


#### Received: 30 June 2020; Accepted: 28 July 2020; Published: 31 July 2020

**Abstract:** Tungsten is the heaviest element used in biological systems. It occurs in the active sites of several bacterial or archaeal enzymes and is ligated to an organic cofactor (metallopterin or metal binding pterin; MPT) which is referred to as tungsten cofactor (Wco). Wco-containing enzymes are found in the dimethyl sulfoxide reductase (DMSOR) and the aldehyde:ferredoxin oxidoreductase (AOR) families of MPT-containing enzymes. Some depend on Wco, such as aldehyde oxidoreductases (AORs), class II benzoyl-CoA reductases (BCRs) and acetylene hydratases (AHs), whereas others may incorporate either Wco or molybdenum cofactor (Moco), such as formate dehydrogenases, formylmethanofuran dehydrogenases or nitrate reductases. The obligately tungsten-dependent enzymes catalyze rather unusual reactions such as ones with extremely low-potential electron transfers (AOR, BCR) or an unusual hydration reaction (AH). In recent years, insights into the structure and function of many tungstoenzymes have been obtained. Though specific and unspecific ABC transporter uptake systems have been described for tungstate and molybdate, only little is known about further discriminative steps in Moco and Wco biosynthesis. In bacteria producing Mocoand Wco-containing enzymes simultaneously, paralogous isoforms of the metal insertase MoeA may be specifically involved in the molybdenum- and tungsten-insertion into MPT, and in targeting Moco or Wco to their respective apo-enzymes. Wco-containing enzymes are of emerging biotechnological interest for a number of applications such as the biocatalytic reduction of CO2, carboxylic acids and aromatic compounds, or the conversion of acetylene to acetaldehyde.

**Keywords:** tungsten enzymes; tungsten cofactor; aldehyde:ferredoxin oxidoreductase; benzoyl-CoA reductase; acetylene hydratase; formate dehydrogenase

#### **1. Introduction**

Tungsten and molybdenum are transition metals of the sixth group and occur in nature predominantly in form of their oxyanions—tungstate (WO4 <sup>2</sup><sup>−</sup>) and molybdate (MoO4 <sup>2</sup>−). While their average abundance in the earth's crust is highly similar, their bioavailability may differ in various aqueous environments. Both metals are present in biological systems in ligation to the so-called metallopterin (or metal binding pterin, MPT, initially introduced as molybdopterin), which occurs as a three-ring pyranopterin compound in most known molybdo- or tungstoenzymes. The entire metal cofactors are referred to as Moco or Wco, respectively [1–6]. There are many variants of MPT-derived cofactors with regard to the metal inserted, the presence of additional ligands of the metal, the number of pyranopterin molecules bound per metal (one or two), and the attachment of additional nucleotides to the MPT core. According to this structural diversity and the underlying amino acid sequence similarities, MPT-containing enzymes can be divided into four unrelated families: (i) xanthine oxidase (XO), (ii) sulfite oxidase (SO), (iii) dimethyl sulfoxide reductase (DMSOR), and (iv) aldehyde oxidoreductase (AOR) families. Wco-containing enzymes belong to the DMSOR and AOR

families, whereas Moco is predominantly found in members of the XO, SO and DMSOR families. In the AOR family, tungsten is always bound by two MPTs, referred to as W-bis-MPT; in DMSOR family members, tungsten or molybdenum is coordinated by two MPT guanine dinucleotide (MGD) moieties, referred to as W-/Mo-bis-MGD cofactor (Figure 1). All tungstoenzymes contain at least one Fe-S cluster next to Wco, and especially the multi-subunit enzymes harbor a number of additional redox-active cofactors such as Fe-S clusters, flavins or hemes.

**Figure 1.** Tungsten cofactor (Wco) found in members of the dimethyl sulfoxide reductase (DMSOR) and aldehyde oxidoreductase (AOR) enzyme families. The former contains either Mo-bis-MPT guanine dinucleotide (MGD) or W-bis-MGD, the latter in most cases use W-bis-metallopterin (MPT) as the active site cofactor.

The majority of Moco- or Wco-containing enzymes catalyze hydroxy- or oxo-transfer reactions such as water-dependent hydroxylations or hydrations via water-, hydroxyl-, or oxo-intermediates bound to the metal in the catalytic course. There are also examples of MPT-dependent enzymes catalyzing hydride or hydrogen atom transfers (e.g., formate dehydrogenase or class II benzoyl-CoA reductase) or sulfur atom transfer reactions (polysulfide reductase). With the exception of acetylene hydratase, all Wco- and Moco-containing enzymes catalyze redox reactions (Figure 2). It is evident that most naturally occurring tungstoenzymes are involved in low-potential (E◦' < −400 mV) electron transfer reactions. This finding can be rationalized by the generally lower redox potential of the biologically relevant redox transitions of W(IV/V/VI) vs. Mo(IV/V/VI). Though acetylene hydratase does not catalyze a redox reaction, the enzyme depends on a low-potential redox activation (see below).

Based on the highly similar physicochemical properties of molybdate and tungstate, the latter was initially considered to act as a general inhibitor of Moco-dependent enzymes, which has been observed in many cases if the replacement of molybdenum by tungsten is enforced by high tungstate concentrations (for an excellent review on this aspect, see [7]). As an example, a recent study used tungstate to inhibit Moco-dependent enzymes from facultatively anaerobic bacteria involved in gut disorders [8]. In facultatively tungsten-dependent enzymes that occur with either of the two metals, the bioavailability of tungstate or molybdate may govern which of the two metals is incorporated into bis-MPT-containing enzymes.

Today, it has become evident that Wco is an essential cofactor of many archaeal and bacterial enzymes that have been discussed in previous reviews with various focuses [7,9–12]. Here, we first aim to provide a state-of-the-art overview of the occurrence and function of obligately and facultatively tungsten-containing enzymes. We then discuss processes that may be involved in discriminating between molybdenum and tungsten during uptake, Mo/W-MPT synthesis and the incorporation into their respective apo-enzymes. This aspect is of particular importance for the emerging number of organisms that simultaneously produce Wco- and Moco-dependent enzymes.

**Figure 2.** Reactions catalyzed by obligately and facultatively tungsten-dependent enzymes. Abbreviations: AOR: aldehyde oxidoreductase; BCR: class II benzoyl-CoA reductase; AH: acetylene hydratase; FDH: formate dehydrogenase; FWD: tungsten-containing formylmethanofuran dehydrogenase; TSR: thiosulfate reductase; DMSOR: dimethyl sulfoxide reductase; NAR: membrane-bound nitrate reductase; NAP: periplasmic nitrate reductase.

#### **2. A**ffi**liation of Tungsten Enzymes in the AOR and DMSOR Enzyme Families**

The presence of Wco is restricted to members of a few clades of the DMSOR and AOR enzyme families. The biochemically characterized tungstoenzymes of the AOR family are affiliated to five clades of AORs (AOR, FOR, GAPOR, GOR, and WOR5) and one clade of the active site subunit BamB of class II benzoyl-CoA reductases (Figure 3A). In addition, the enzyme family contains one clade of archaeal enzymes and two of bacterial enzymes, whose biological functions are yet unknown (WOR4, YdhV and AOR1). Enzymes of the YdhV clade are unique in containing Moco instead of Wco, whereas WOR4 has been identified as a Wco-containing protein. The metal content of the enzymes of the AOR1 clade is unknown, since these have been only identified from genome sequences and not on the protein level. AORs of the XO family [13,14] are solely Moco-dependent and therefore not further discussed in this review.

The DMSOR family is subdivided into three subfamilies and several additional clades, and the tungstoenzymes of this family are affiliated to at least eight clades, which mostly contain either Mocoor Wco-dependent members (in case of formate or formylmethanofuran dehydrogenases, and nitrate-, DMSO-, or thiosulfate reductases [15]). Only the acetylene hydratase clade seems to be predominantly W-dependent, but this is based on only a single biochemically characterized enzyme, and it is unknown whether all these enzymes or closely related clades share the same metal preference (Figure 3B).

**Figure 3.** Phylogenetic trees of the large subunits of selected enzymes of the AOR family (**A**) and the DMSOR family (**B**). Clades or subclades containing tungsten enzymes are labeled in light-orange (bacteria), dark-orange (archaea) or violet (enzymes working with either tungsten or molybdenum, for better visualization additionally marked with a violet dot), and those containing molybdenum enzymes are labeled in light- or dark-blue (bacterial and archaeal enzymes, respectively). Enzyme clades of unknown metal content are not colored. Members of the DMSOR subfamilies 1–3 are highlighted by the outer circle as indicated. Species epithets for unaffiliated AORs: *Pa: Pyrobaculum aerophilum*; *Dg: Desulfovibrio gigas*; *Cf: Clostridium formicoaceticum*. Abbreviations: AOR, FOR, GAPOR, WOR5: archaeal aldehyde oxidoreductases of various specificities (for details see text); BamB: active site subunit of class II benzoyl-CoA reductases; WOR4, YdhV, AOR1: archaeal or bacterial Mo- or W-containing enzymes of unknown function; GOR: glyceraldehyde-3-phosphate oxidoreductases from thermophilic bacteria; FDH: formate dehydrogenases; NAP/NAS: periplasmic and assimilatory nitrate reductases; FMD/FWD: molybdenum- and tungsten-dependent formylmethanofuran dehydrogenases; AH-like: acetylene hydratases and similar hypothetical enzymes; TSR/PSR: thiosulfate/polysulfide reductases; PAD: phenylacetyl-CoA dehydrogenases; ARR, ARS: arsenate reductases/arsenite oxidases; DMSOR: dimethyl sulfoxide reductases; TMAOR: trimethylaminoxide reductases; NAR/anammox: chemolithotrophic nitrite oxidoreductases and putative nitrate reductases from anammox bacteria; NAR-6: nitrate reductases and similar clades; CLR/SER: chlorate/selenate reductases and dimethyl sulfide dehydrogenases; EBDH: ethylbenzene dehydrogenases and related enzymes.

#### **3. Obligately W-Containing Enzymes**

#### *3.1. Aldehyde Oxidoreductases*

Most known clades of obligately tungsten-dependent enzymes show activities as aldehyde oxidoreductases (Figure 2). Four of the five enzymes of the AOR family encoded in the genome of the archaeon *Pyrococcus furiosus* have been characterized as hyperthermophilic and extremely O2-sensitive aldehyde oxidoreductases of various specificities. Together with the tungsten-dependent oxidoreductase WOR4, which was shown to contain Wco and a [3Fe-4S] cluster, but did not show any detectable activity [16], these enzymes define five separate phylogenetic clades of the AOR enzyme family (Figure 3A), namely the subfamilies of archaeal AOR, formaldehyde oxidoreductase (FOR), glyceraldehyde-3-phosphate oxidoreductases (GAPOR), and tungsten-dependent oxidoreductases (WOR4 and WOR5). Attempts to replace tungsten in some of these proteins with molybdenum or vanadium led either to proteins lacking any metal or to molybdenum-containing, inactive variants, proving the necessity of the Wco for their activities [17,18]. Additional members of the enzyme

family emerged recently from obligately or facultatively anaerobic bacteria, namely bacterial AOR and GAPOR (GOR), which represent sister clades of the archaeal AOR and GAPOR, respectively (Figure 3A).

Archaeal and bacterial AORs (*sensu stricto*) represent the most abundant and best-studied clades exhibiting aldehyde oxidoreductase activity. Both contain a subunit of a similar size harboring the W-bis-MPT cofactor with an additional Mg2<sup>+</sup> ion bridging the phosphate groups of the two cofactors, as well as an additional [4Fe-4S] cluster [19–21] (Figure 4). The archaeal AORs, as well as some AORs from strictly anaerobic bacteria, occur as homodimers with a bridging iron atom [20], and utilize ferredoxin as a physiological electron acceptor. In contrast, AORs from *Aromatoleum aromaticum* and other facultatively anaerobic denitrifying bacteria, as well as from the strictly anaerobic *Moorella thermoacetica*, consist of three different subunits in an α2β2γ<sup>2</sup> composition. The additional electron-transferring small subunit with four [4Fe-4S] clusters and the medium-size FAD-containing subunit allows for the use of a broader electron acceptor range, including NAD<sup>+</sup> [21]. All characterized enzymes of the archaeal or bacterial AOR clades oxidize a variety of aldehydes to their respective carboxylic acids. Their main physiological function was proposed to be the detoxification of aldehydes that accumulate in different metabolic pathways. Though known for decades, there is still little knowledge of the reaction mechanisms of AORs and the internal electron transfer events involved, largely because of the difficult genetic accessibility of the respective host organisms. Remarkably, AORs also catalyze the reverse reaction, namely the reduction of non-activated acids to the corresponding aldehydes at E◦' ≈ −560 mV, albeit at a rate lower than 1% compared to those of aldehyde oxidation. This property is reflected in the alternative name "carbonic acid reductase" (CAR) for the enzyme from *M. thermoacetica*. By utilizing this property, the metabolism of fermentative archaea has been artificially shifted from the production of acetic acid to ethanol. In spite of the very low observed reverse in vitro activities of AORs, the reaction seems to work well in whole-cell systems [22]. A similar potential of AORs in the biotechnologically interesting utilization of syngas as a substrate for acetogenic bacteria has been suggested [22,23].

**Figure 4.** Structure of formaldehyde:ferredoxin oxidoreductase from *P. furiosus*, in complex with glutarate (pdb 1B4N) at 2.4 Å solution. Left panel: overall structure; right panel W-bis-MPT and [4Fe-4S] cofactor. The tungsten atom is depicted in blue, Mg2<sup>+</sup> is shown in light green, and Ca2<sup>+</sup> in dark green.

FORs are known from various hyperthermophilic archaea and account together with AOR and GAPOR for the most abundant tungstoenzymes in these organisms. Unlike other AORs, FORs show only low specific activities with short chain aldehydes. They turn over formaldehyde with reasonable rates but show very high *K*m-values for almost all substrates. A possible exception is glutardialdehyde with the lowest *K*m-value of the tested substrates, but the data available do not allow to propose a rational physiological function of FOR [24]. The X-ray structure of the FOR for *P. furiosus* confirms its composition as a homotetramer without a bridging iron atom, as in AORs (Figure 4). Instead, FOR contains an additional Ca2<sup>+</sup> ion ligated to one of the MPT cofactors, which is attributed to a structural function. The binding of W-bis-MPT and the [4Fe-4S] cluster is highly similar in AOR and FOR, and the FOR structure with a bound glutarate as a substrate mimic led to the speculation that diacid semialdehydes may serve as potential physiological substrates [25].

Two further clades comprise GAPOR from archaea and GOR from bacteria, which are involved in glucose metabolism via a modified glycolytic pathway. They specifically couple the oxidation of glyceraldehyde-3-phosphate to 3-phosphoglycerate without 1,3-bisphosphoglycerate as an intermediate to the reduction of ferredoxin. GAPORs are monomeric enzymes containing Zn2<sup>+</sup> as an additional metal [26,27].

Finally, two further clades of tungsten-containing AOR family members of unknown physiological relevance are represented by WOR4 and WOR5 from *P. furiosus*. The former showed no aldehyde-oxidizing activity with any substrate. Furthermore, it is the only enzyme of the family containing a [3Fe-4S] cluster, consistent with the loss of one of the cysteine ligands [16], the latter was found to oxidize a similar aldehyde as AOR [28]. Orthologous genes putatively encoding AOR-like enzymes are found in many hyperthermophilic archaea and anaerobic bacteria, and some of them have been biochemically characterized, such as AORs and FORs from *Thermococcus litoralis*, AORs from *T. paralvinellae*, *Clostridium formicoaceticum*, *Eubacterium acidaminophilum*, *Desulfovibrio gigas*, *Methanobacterium thermoautotrophicum*, and GAPOR from *Pyrobaculum aerophilum* [7,29–31].

#### *3.2. Class II Benzoyl-CoA Reductases*

Benzoyl-CoA-reductases (BCRs) are key enzymes for aromatic compound degradation in anaerobic prokaryotes [32–35]. They catalyze the reduction of the central intermediate benzoyl-CoA to cyclohexa-1,5-diene-1-carboxyl-CoA (dienoyl-CoA) at a redox potential of E'◦ = −622 mV [36]. Two phylogenetically unrelated classes of BCRs have been identified that both catalyze the same reaction. Class I BCRs are present in facultative anaerobes such as the denitrifying bacterium *Thauera aromatica* [37]. They employ [4Fe-4S] clusters as the only cofactors and use reduced ferredoxin as an electron donor. Class I BCRs couple the thermodynamically unfavorable reduction of benzoyl-CoA to the stoichiometric hydrolysis of ATP. Obligately anaerobic bacteria employ an ATP-independent, Wco-containing one megadalton class II BCR complex [38]. Class II BCRs have been purified from the Fe(III)-respiring *Geobacter metallireducens* [39] and the sulfate-respiring *Desulfosarcina cetonica* [40], both are composed of eight subunits (Bam[(BC)2DEFGHI]2, Bam = benzoic acid metabolism). With four W-bis-MPTs, four Zn, >50 Fe-S clusters, four selenocysteines, and six FADs, class II BCRs represent one of the most complex metalloenzyme machineries known (Figure 5). BamB harbors the Wco-binding active site subunit and can be purified along with the electron transferring BamC subunit [41]. The remaining BamDEFGHI subunits are proposed to be involved in the endergonic electron transfer to Wco, driven by a flavin-based electron bifurcation [39]. In this process, the endergonic electron transfer from a donor to a low-potential acceptor, here benzoyl-CoA, is driven by the exergonic reduction of a second high-potential acceptor using the same donor [42]. The anticipated high-potential acceptor(s) of class II BCRs are still unclear, but recent studies suggest that class II BCRs transfer electrons to menaquinone, as evident from the membrane-association of the complex from *G. metallireducens* [39].

**Figure 5.** Scheme of the class II BCR-complex from *G. metallireducens*. The endergonic electron transfer from reduced ferredoxin to benzoyl-CoA may be driven by an exergonic reduction of NAD<sup>+</sup> and/or menaquinone (MQ).

The BamB subunit belongs to the AOR enzyme family and contains Wco, Zn2<sup>+</sup> and a [4Fe-4S] cluster as cofactors. Bioinformatic analyses suggest that all strictly anaerobic bacteria employ a tungsten-containing BamB for benzoyl-CoA reductions during growth with aromatic substrates [34]. Initial studies with *Desulfococcus multivorans* [43] and *G. metallireducens* [38] indicated that growth with aromatics depends on molybdenum and selenium, however, the isolated Bam(BC)2 contained stoichiometric amounts of tungsten, whereas molybdenum was virtually absent [41]. This finding suggests that traces of tungstate in the molybdate stock solution were sufficient to produce an active, Wco-containing class II BCR. A TupABC transporter is induced during growth with aromatics that guarantees selective tungstate uptake from the medium [44].

The 1.9 Å resolution X-ray structure of Bam(BC)2 crystals revealed that Wco is accommodated in a highly hydrophobic pocket where it is coordinated by four dithiolene sulfurs of W-bis-MPT, a thiolate of a conserved cysteine (Cys322) and a sixth non-proteinogenic ligand [45] (Figure 6). Its identity could so far not be resolved unambiguously by spectroscopic methods, whereas computational studies favor a water ligand [46,47]. Neither the substrate nor the product are directly bound to tungsten. The spatial separation of the potential proton donor His260 and the W-atom with the aromatic ring positioned in-between provides the molecular basis for the anticipated biological Birch reduction of aromatic rings [48]. Continuum electrostatic and quantum-mechanical/molecular mechanics calculations suggest that benzoyl-CoA reductions are initiated by a hydrogen atom transfer from a W(IV) species via bound water yielding a W(V)-(OH−) species and a substrate radical intermediate [46,47]. These studies also suggested that the second electron derives from the pyranopterin cofactor, rather than from W(V). Proton transfers from an invariant histidine (His260) likely assist this step. A BamB catalysis is one of the rare examples of an MPT-dependent enzyme that does not involve an oxo- or hydroxyl-transfer. The unusual properties of class II BCRs among MPT-enzymes may be explained by the extremely low redox potential of the substrate/product couple that affords a radical biochemistry.

**Figure 6.** Structure of the BamBC components of class II benzoyl-CoA reductase from *G. metallireducens* (pdb 4Z3W-Z). Left panel: overall structure; the catalytic BamB subunit depicted in red contains the W-bis-MPT cofactor, BamC (light green) contains 3 [4Fe-4S] clusters. Right panel: W-bis-MPT, [4Fe-4S] cluster and the substrate benzoyl-CoA. Tungsten (blue) is coordinated by five thiols (four from the dithiolenes and one from Cys322), and by an inorganic ligand (magenta). A Mg2<sup>+</sup> bridging the phosphate groups of the two MPTs is shown in green.

The Birch reduction of aromatic rings is widely used in synthetic chemistry for many applications [49]. Due to its negative attributes, such as its dependence on alkali metals, and ammonia and cryogenic reaction conditions, alternative procedures are being studied such as the use of photo- [50] and electrocatalytic [51] methodologies. In this light, biocatalytic BCRs may be attractive for future Birch reduction applications under mild and environmentally friendly conditions.

#### *3.3. Acetylene Hydratases*

Anaerobic acetylene degradation by *Pelobacter acetylenicus* [52] is initiated by the hydration and tautomerization to acetaldehyde, catalyzed by acetylene hydratase (AH). So far, only the enzyme from *P. acetylenicus* has been isolated and studied, and a high resolution crystal structure is available (Figure 7) [53–56]. The oxygen-sensitive enzyme belongs to the DMSOR enzyme family and contains a W-bis-MGD and a [4Fe-4S] cluster [53]. When AH was isolated from *P. acetylenicus* grown under tungsten-depletion (2 nm) and a 1000-fold excess of molybdate, a molybdenum-containing variant of AH was identified (45–50% metal occupation) that converted acetylene at 10% of the rate of the Wco-containing enzyme [57]. Because of these highly unphysiological expression conditions, we consider AH here as an obligately tungsten-containing enzyme, but further research on the viability of Mo-containing AH variants and their occurrence is required. Almost all studies were carried out with Wco-containing AH.

AH is unique among all Moco/Wco enzymes in terms of catalyzing a non-redox reaction, and it is the only known enzyme that acts on acetylene as a physiological substrate. Notably, acetylene inhibits many key enzymes of anaerobic metabolism, with the exception of nitrogenase, which reduces acetylene to ethylene [58,59]. AH itself is inhibited by acetylene analogues such as cyanide, carbon monoxide, nitrous oxide or substituted acetylenes [53,54]. Though the reaction catalyzed by AH is redox neutral, AH requires a strong reductant like titanium(III)-citrate or sodium dithionite to be catalytically active *in vitro* [53,56]. The mandatory reductive activation of AH is assigned to the reduction of Wco to the W(IV) state, though the exact redox potential of the Wco is unknown [53,55]. An intact [4Fe-4S] cluster appears to be dispensable for catalysis [56] as supported by the observation that degradation to a [3Fe-4S] cluster did not affect AH activity [53].

**Figure 7.** Structure of acetylene hydratase from *P. acetylenicus*(pdb 2E7Z) at 1.26 Å resolution. Left panel: overall structure. Right panel: W-bis-MGD and [4Fe-4S] cofactors. Tungsten is coordinated by five thiols (four from the dithiolenes and one from Cys141), and by a water molecule. The tungsten atom is depicted in blue.

The mechanism of AH has been subject to many discussions, and two types of mechanistic scenarios were suggested with different roles for the catalytically active W(IV): (i) a first-shell mechanism, where tungsten coordinates acetylene and primes the substrate for a nucleophilic attack by water (or hydroxide), and (ii) a second-shell mechanism that relies on the polarization of tungsten-bound water, followed by an electrophilic attack on the C≡C bond of acetylene [55]. The second-shell mechanism was favored by density-functional theory calculations based on a high resolution crystal structure, but no spectroscopic evidence was obtained to corroborate this mechanism [55]. Based on calculations for the active site mimics of AH, another study concluded that the direct coordination of acetylene by tungsten was favored over the binding of water [60]. It was reported that an inorganic model complex (NEt4)2[WIVO(mnt)2], where mnt = malonitrile, was able to catalyze the same reaction as AH, whereas the W(VI) analogue failed [61]. While these results could not be reproduced, recent developments and thorough characterizations of other biomimetic tungsten complexes strongly support a mechanism of acetylene binding directly to tungsten (first-shell mechanism) [62,63]. Such a metal-acetylene coordination is known from other organometallic molybdenum-/tungsten-complexes [64].

Acetaldehyde is an important building block for many chemicals and pharmaceuticals. Its industrial production from acetylene is typically achieved in the presence of a mercury catalyst-containing mercuric chloride that exhibits adverse effects on human health and the environment [65]. Thus, enzymatic solutions for this industrial process are of potential interest.

#### **4. Enzymes Containing Either Tungsten or Molybdenum**

A number of isofunctional enzymes of the DMSOR family exist in nature that may contain either Moco or Wco. This apparent promiscuity may be explained by an adaptation to molybdate or tungstate enriched/depleted environments rather than to substantial differences in enzyme function. In some organisms, sets of genes exist that specifically code either for Moco- and/or for Wco-dependent variants of MPT-dependent enzymes. The induction of the individual gene clusters in response to varying molybdate/tungstate concentrations has been described in some cases. The molybdenumor tungsten-containing variants may differ in their activity or affinity to the individual substrates. Owing to the lower redox potential of the biologically relevant W(IV/V/VI) vs. Mo(IV/V/VI) states, Wco-containing variants appear to be catalytically more efficient for reactions occurring at low redox potentials and Moco for those at more positive redox potentials.

#### *4.1. Formate Dehydrogenases*

The widely abundant tungsten- or molybdenum-dependent FDHs catalyze the reversible conversion of formate to CO2 at E◦' = −430 mV with various electron donors/acceptors [66–70]. Metal-dependent FDHs are involved in formate-dependent respiration, syntrophy and methanogenesis, acetogenesis, methylotrophy or in fermentations as components of formate hydrogen lyase complexes. Unlike most other Moco/Wco enzymes, an FDH catalysis does not involve an oxygen atom transfer reaction. Though the reaction mechanism of FDHs is still under debate, a hydride transfer from formate to a sulfido-ligand at Mo(VI)/W(VI) appears to be the most plausible as it circumvents the difficult deprotonation of the Cα proton of formate [68,71–73]. Here, we briefly summarize the occurrence and function of Wco-dependent FDHs and refer to recent in-depth reviews of the diversity, structure and function of metal-containing FDHs [66–68].

Many FDHs depend on molybdenum, and elevated tungstate concentrations in the medium yield inactive Mo-dependent variants as reported for FDHs from *E. coli* [74] or *Methanobacterium formicium* [75]. On the other side, a number of FDHs have been isolated with a clear preference for tungsten over molybdenum, whereas only a few FDHs appear to be active with either of the two metals. Wco-containing FDHs are affiliated to two enzyme clades, the membrane-bound FDH-N-like enzymes and the soluble or membrane-associated FDH-H-like enzymes, whose active sites are oriented towards the periplasm or the cytoplasm, respectively. Cytoplasmic Wco-containing FDHs have initially been described in a number of acetogenic clostridia such as *Moorella thermoaceticum* [76], *Clostridium formicoaceticum* [77] or *C. carboxidivorans* [78], where they catalyze the NAD(P)Hdependent reduction of CO2. A tungsten-containing FDH of *Peptoclostridium* (formerly *Eubacterium*) *acidaminophilum* serves either as an electron donor system for amino acid fermentations or as component of a formate hydrogen lyase complex [79]. A tungsten-dependent FDH of this clade is also involved in the oxidation of reduced C1 compounds to CO2 with dioxygen as an electron acceptor in the aerobic methylotrophic *Methylobacterium extorquens* [80]. This finding indicates that the use of Wco is not restricted to anaerobic organisms. Two tungsten-specific FDHs, an FDH-H and a FDH-N type enzyme, were reported from *Syntrophobacter fumaroxidans* during syntrophic propionate fermentations, where they play a role in electron transfers to the methanogenic partner [81]. Further Wco-containing FDHs of the FDH-N clade are known from *Desulfovibrio* species and other sulfate reducing bacteria [82–85]. One of these species, *D. alaskensis*, is known to incorporate either Moco or Wco into the same FDH, depending on the tungstate and molybdate concentrations in the medium [83,85]. In contrast, *D. vulgaris* Hildenborough produces a specific Moco-dependent FDH during growth with molybdate and an FDH active with either metal during growth with tungstate [84]. A similar case of an active FDH-N type enzyme with either Moco or Wco in differently grown cells was reported for *Campylobacter jejuni* [86], which contains only one set of FDH encoding genes.

Biocatalysts that sequester CO2 are of increasing general interest. In this light, FDHs that reduce CO2 to formate as a viable energy source are of a potential biotechnological use. A tungstencontaining FDH has been adsorbed to an electrode and successfully used for electrocatalytical CO2 reduction [87]. Furthermore, an enzyme complex of FDH and hydrogenase (formate hydrogen lyase) reversibly interconverts H2 + CO2 to formate demonstrating that formate may be used as an energy storage/transport compound [88,89]. Finally, the solar-driven CO2 reduction by FDHs in combination with artificial and/or biological photosystems has been demonstrated [90,91].

#### *4.2. Formylmethanofuran Dehydrogenases*

Formylmethanofuran dehydrogenase catalyzes the first step of methanogenesis that is related to the reaction catalyzed by FDHs: the reduction of CO2 yielding formate is bound in an activated form at the methanofuran cofactor [69]. Methanogens have been found to contain isoenzymes being either specific for tungsten or molybdenum [92,93]. The genome of *Methanososarcina acetivorans* encodes four putative formylmethanofuran dehydrogenases. Two of these are proposed to be specific for tungsten or molybdenum, respectively, with one of the tungsten-dependent isoenzymes being specifically required for growth with carbon monoxide [94]. The crystal structure of the tungsten-containing formylmethanofuran dehydrogenase from *Methanothermobacter wolfei* revealed that the enzyme harbors two active sites being separated by a 43 Å tunnel which allows the directed diffusion of the formate from the first active site (the Wco center) to the second one, the formylmethanofuran forming center [95].

#### *4.3. Respiratory Nitrate Reductases*

Two versions of respiratory nitrate reductases belong to the DMSOR enzyme family: the membrane bound enzymes (NarGHI) and the periplasmic ones (NAP). The active site subunit of the former, NarG, is largely conserved among all nitrate-respiring bacteria and archaea and usually contains Moco in the active site subunit. Tungstate usually acts as an inhibitor of the assembly of Moco into NarG, NAP or other respiratory Moco-containing enzymes [8,74]. However, *Paracoccus pantotrophus* is able to grow under denitrifying conditions in the presence of 100 μm tungstate, and indirect evidence was obtained that tungsten was incorporated into its periplasmic nitrate reductase NAP [96]. Furthermore, the nitrate-respiring, hyperthermophilic archaea *Pyrobaculum aerophilum* and *Aeropyrum pernix* grow at temperatures and environments where tungstate is enriched and molybdate is depleted. Under these special growth conditions, a tungsten-containing NarGHI-type enzyme was isolated from of *P. aerophilum* with a two-fold lower turnover number than the Mo-containing enzyme [97].

#### *4.4. Dimethyl Sulfoxide and Trimethylamine N-Oxide Reductases*

DMSOR and trimethylaminoxide reductase (TMAOR) are both required to use DMSO or TMAO as terminal electron acceptors in oxygen-independent respiratory chains. In the DMSOR from *Rhodobacter capsulatus* and *R. sphaeroides*, the Mo bound to the bis-Mo-MGD cofactor was substituted with tungsten, and the resulting Wco-containing enzyme reduced DMSO even at a higher rate, but was inactive in terms of catalyzing the reverse dimethyl sulfide oxidation [15,98]. This finding is rationalized by the generally lower redox potential of Wco in comparison to Moco [99]. Likewise, TMAOR from *E. coli* was also active when molybdenum was substituted by tungsten even with a slightly increased catalytic efficiency [100].

#### *4.5. Thiosulfate Reductases*

A recent report shows that a thiosulfate reductase (TSR) of the DMSOR family is involved in thiosulfate respiration in the archaeon *P. aerophilum*. The enzyme was recombinantly produced in *P. furiosus* and contained either Mo or W, dependent on the growth conditions [101]. The molybdenum-containing enzyme showed a ten-fold higher conversion rate and a twice as high affinity to thiosulfate than the tungsten-containing one. This study represents the first case of producing an active Moco- or Wco-containing enzyme in *P. furiosus*.

#### **5. Tungsten Uptake and Assembly of the Wco**

#### *5.1. Tungsten Uptake*

Tungsten is taken up by prokaryotes in form of the tungstate oxyanion (WO4 <sup>2</sup>−) by three high-affinity ATP-binding cassette (ABC)-type systems: the highly tungstate-specific TupABC system, and the ModABC/WtpABC systems that transport either tungstate or molybdate [102]. They all consist of a periplasmic binding protein (component A), a transmembrane channel (component B), and an ATPase subunit (component C). The ModABC system with similar affinities to molybdate and tungstate in the low to high nM range is highly abundant in bacteria and archaea and allows for the specific transport of molybdate and tungstate even in the presence of high concentrations of similar other oxyanions (e.g., 28 mm sulfate in marine environments). The WtpABC transporter is present in archaea that lack TupABC or ModABC systems, such as *P. furiosus*, and is rather distantly related to these uptake systems. The binding protein WtpA shows by far the highest affinity for tungstate with Kd = 19 pm [103]. The TupABC system has originally been studied in *P. acidaminophilum* [104,105], and later in *C. jejuni* [106,107] and *D. alaskensis* [108,109]. The encoding TupABC genes are widely abundant in bacterial and archaeal genomes [110]. The structural basis by which TupA discriminates between tungstate and molybdate, considering their highly similar atomic radii, is still unknown.

#### *5.2. Metallopterin Cofactor Synthesis*

Most knowledge of MPT biosynthesis derives from research on Moco (for initial and recent reviews on Moco synthesis see [4,5,111–113]). As almost all genes required for Moco synthesis are present in the genomes of prokaryotes that synthesize Wco-dependent enzymes, it is generally accepted that MPT synthesis is catalyzed by similar enzymes and identical intermediates [114]. In contrast, very little is known about the steps that discriminate between molybdenum and tungsten during the insertion of the metals into MPT. Here, we first briefly summarize knowledge of Moco synthesis in the prokaryotic model organism *E. coli*, followed by the presentation of initial insights and hypotheses about Wco synthesis. There, we will address the special challenges of simultaneously synthesizing Moco- and Wco-dependent enzymes.

Moco biosynthesis proceeds via a four-step process: (i) conversion of guanosine-5 -triphosphate (GTP) to the four-ringed cyclic intermediate pyranopterin monophosphate (cPMP), (ii) introduction of the two thiols of the dithiolene group, (iii) adenylation to MPT-adenosine monophosphate (AMP), and (iv) metal insertion and release of AMP; the latter two steps are often merged as a single one (Figure 8) [4,5,111,113]. The formation of cPMP (or "precursor Z") from GTP occurs in two steps involving complex molecule rearrangements. The radical S-adenosylmethionine (SAM) enzyme MoaA catalyzes an initial remodeling of the carbon backbone of GTP and afterwards, MoaC completes the synthesis of cPMP with the concomitant release of the pyrophosphate moiety of the original GTP [115,116].

In the second step, two sulfur atoms are inserted into cPMP at the C1 and C2 position, resulting in MPT formation. This step is catalyzed by MoaD and MoaE, which form a heterotetrameric complex, also referred to as MPT synthase. Two MoaEs bind one cPMP each, while MoaD transfers one sulfur atom at a time from a thiocarboxylate group on its conserved C-terminal glycine residue, producing a hemisulfurated intermediate in the process [117–120]. MoaD itself receives its thiogylcyl modification by the sulfurylase MoeB. MoaD forms a complex with MoeB, which catalyses the Mg2+ and ATP-dependent adenylation of the C-terminal glycine residue of MoaD. This activated MoaD is then sulfurated by the cysteine desulfurase IscS [121,122].

In the third step, MPT is adenylated in an ATP-dependent reaction by MogA (in bacteria) or MoaB (in archaea), yielding MPT-AMP. This activated intermediate is transferred to MoeA, where molybdate is bound to the dithiolene sulfurs of MPT, accompanied by hydrolysis and the release of the AMP moiety [114,123,124]. The product is referred to as Mo-MPT, which represents the Moco found in SO family members, and which serves as a building block for the cofactors of the three other families of MPT-containing enzymes. To generate the cofactor for the XO-family enzymes, MocA catalyzes the cytidylation of MPT with CTP while releasing pyrophosphate. The generated MPT cytosine dinucleotide cofactor is then further modified by the addition of an equatorial sulfido-ligand in the active site of the enzymes [13,125]. Ligation of a molybdenum and a second MPT yields Mo-bis-MPT, which is found in the AOR family member YdhV [126]. The Mo-bis-MGD cofactor of the DMSOR family enzymes is synthesized by the attachment of two GMP moieties from GTP [127,128].

Wco synthesis is considered to involve similar or even the same enzymes for the steps up to metal insertion. So far, the adenylation of MPT by MoaB in the course of Wco synthesis was shown in the hyperthermophilic *P. furiosus* [114]. No other enzyme specifically involved in Wco synthesis has been studied so far.

**Figure 8.** Moco biosynthesis in *E. coli*. In prokaryotes, the steps from GTP to MPT-adenosine monophosphate (AMP) are catalyzed by similar enzymes for Moco and Wco synthesis. YdhV from *E. coli* is the only characterized molybdenum-containing AOR family member [126].

#### *5.3. Strategies for the Selective Insertion of Molybdenum*/*Tungsten into Target Proteins*

The highly similar physicochemical properties of molybdate and tungstate immediately lead to the question regarding which mechanism guarantees the correct transfer of both oxyanions into their target proteins. In particular, organisms producing enzymes that are only active with either of the two metals depend on a highly selective metal insertion system. We distinguish here between three different scenarios: (i) organisms that exclusively produce either Wco- or Moco-dependent enzymes, (ii) organisms that contain genes for both type of enzymes, but only produce either Moco- or Wco-enzymes under certain conditions, and (iii) organisms that produce Moco- and Wco-dependent enzymes simultaneously. In principle, selectivity may be achieved during molybdate/tungstate uptake, Mo-MPT/W-MPT formation, or the insertion of the mature W-bis-MPT/bis-Mo-MPT into the respective target proteins.

In the first scenario, the selective uptake of either molybdate or tungstate would already be sufficient to guarantee selectivity. Specific tungstate uptake may be achieved by the periplasmic binding protein TupA from the TupABC transporter [105], or WtpA from *P. furiosus*, which show much higher affinities for tungstate vs. molybdate [103]. However, no transporter system exhibiting a comparable selectivity for molybdate over tungstate has been described so far, which explains the frequently observed antagonistic effect of tungstate for Moco-dependent enzymatic processes.

In the second scenario, the selective uptake may also play a role but would require at least two selective uptake systems for either of the two metal oxyanions, which may be induced under certain conditions. In addition, MoeAs have been proposed earlier as selectivity generating enzymes that may be specific for either molybdate or tungstate during metal insertion into MPT-AMP [12,85,114]. This proposal was based on the finding that at least two versions of MoeA encoding genes were identified in genomes of archaea [12] or in *D. alaskensis* [85]. The latter produces either Moco or Wco containing versions of the same FDH, depending on the molybdate/tungsten concentrations in the medium. In *D. vulgaris* Hildenborough, the production of Wco- or Moco-containing FDH isoenzymes is regulated by the molybdate/tungstate availability at the transcriptional level [84]. These results are in line with a metal-dependent induction of either tungsten- or molybdenum-specific pairs of MoeA/apo-FDH paralogues.

Such a mechanism is not feasible when Moco- and Wco-depending enzymes have to be produced in parallel (scenario iii). In recent years, a number of bacteria were identified that simultaneously produce obligately Wco- and Moco-dependent enzymes, and they all contain at least two *moeA* copies (Figure 9). *C. jejuni* produces a tungsten-containing FDH under certain growth conditions that may be coupled with molybdenum-dependent nitrate reductase in a menaquinone-dependent respiratory chain [86]. In *A. aromaticum*, a tungsten-dependent AOR and two molybdenum-dependent enzymes, phenylacetyl-CoA dehydrogenase and nitrate reductase, are produced during growth with phenylalanine under denitrifying conditions [21]. When *G. metallireducens* grows with *p*-cresol or 4-hydroxybenzoate under ammonifying conditions, it produces one strictly Wco-dependent class II benzoyl-CoA reductase [38,41], and two Moco-containing enzymes from different families: the 4-hydroxybenzoyl-CoA reductase of the XO family [129,130] and nitrate reductase of the DMSOR family [44]. While *C. jejuni* contains two paralogous *moeA* genes [106], even three paralogues are found in the genomes of *A. aromatoleum* and *G. metallireducens*. In the latter bacterium, the genetic context of the three *moeA* genes indicates the specificity for molybdenum or tungsten (Figure 9). In particular, *moeA-1* is located adjacent to the genes encoding the tungstate-dependent TupABC transporter, and *moeA-3* is adjacent to paralogues of the *bamBC* genes, which putatively code for an isoenzyme of tungsten-dependent class II benzoyl-CoA reductase. Thus, it is tempting to speculate that *moeA-1* and *moeA-3* encode a tungstate-specific metal insertase. In contrast, *moeA-2* is located in a gene cluster together with genes encoding molybdenum-dependent NarGHI-type nitrate reductase. Thus, *moeA-2* likely encodes a molybdenum-specific enzyme. These findings support the idea that the simultaneous production of Moco- and Wco-dependent enzymes is accomplished by MoeA

isoenzymes that are selective for either molybdate or tungstate. Though this assumption appears plausible, any experimental evidence for it is lacking.

**Figure 9.** Genomic location of *moeA* paralogues in bacteria that produce Wco- and Moco-containing enzymes simultaneously.

Finally, the question rises as to how the Mo-MPT/W-MPT cofactors formed by specific MoeAs are specifically processed and targeted into the individual apo-enzymes. Specific targeting excludes that Mo-MPT/W-MPT formed are released from individual MoeAs because the modifying enzymes that bind the Mo-MPT/W-MPT in the next step will hardly distinguish between the molybdenum- or tungsten-containing versions. In tungstoenzymes, the metal is always bound to two pyranopterins (bis-MPT), which affords the conversion of MoeA-bound W-MPT to W-bis-MPT in a similar manner as the MobA-catalyzed process known for molybdenum enzymes of the DMSOR family [128]. Thus, the reported complex formation between MoeA and MobA [131] appears to be essential for maintaining the selectivity of the individual bis-MPT intermediates formed. It is conceivable that individual MoeA/MobA complexes specifically interact with the respective apo-enzymes either before or after the GTP-dependent W-bis-MGD formation, including further modifying enzymes such as FdhD. The latter is involved in sulfuration of the Mo-/W-bis-MPT of FDHs and in its insertion into apo-FDHs [132–134]. MoeAs, specific for either tungsten or molybdenum, may enable or disable such additional interactions. Finally, the formation of supercomplexes between all the components involved in the conversion of the last common MPT-AMP intermediate to the mature cofactors inserted in individual target proteins could substantially facilitate selective Moco/Wco targeting.

#### **6. Conclusions**

Recent research has revealed a higher abundance and catalytic diversity of tungstoenzymes than previously anticipated. While tungsten was originally mostly associated with AORs from hyperthermophilic archaea and some clostridial FDHs, a number of tungstoenzymes with novel functions have been discovered in a large variety of microorganisms comprising fermenting, sulfate-reducing, metal oxide-reducing, denitrifying, aerobic and pathogenic bacteria. Tungstoenzymes are of emerging biotechnological interest due to their involvement in the low-potential reduction processes of CO2, carboxylic acids or aromatic rings, or in the industrially important conversion of acetylene to acetaldehyde. Thus, tungsten appears to be the bio-metal of choice for a number of challenging enzymatic reactions. While knowledge of the structure and function of tungstoenzymes has continuously been increasing over the past two decades, there is still a high research demand concerning the biosynthesis and incorporation of Wco. In particular, the mechanisms that discriminate between tungstate- and molybdate-insertion, and the selective targeting of Mo- or W-bis-MPTs to the individual apo-proteins, are still unknown. Considering the high similarities of molybdate and tungstate, the observed high selectivity for either of the two metals affords sophisticated molecular solutions. Insights into the molecular basis of this selectivity will help to explain how enzymes that strictly rely on tungsten are specifically loaded with this cofactor even in the presence of high molybdate concentrations in the medium. They will also provide a rationale as to why whole-cells may be inhibited by tungstate due to the antagonistic effects of tungstate and molybdate. Bacteria that simultaneously produce tungsto- and molybdoenzymes will serve as valuable model organisms to address the question of selectivity for tungstate vs. molybdate in the future.

**Author Contributions:** Phylogenetic analyses and visualization: J.H., C.S.S.; structural analysis: M.W.; visualization of reaction schemes and genetic organization: C.S.S., M.B.; project supervision: M.B.; the manuscript was written and revised through contributions of all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the German Research Foundation (DFG), grants BO 1565 15-1 and HE 2190 11-1, the collaborative research center SFB 1381, project ID 403222702 (M.B.) and the Synmikro center Marburg (J.H.).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Metal–Dithiolene Bonding Contributions to Pyranopterin Molybdenum Enzyme Reactivity**

#### **Jing Yang 1, John H. Enemark <sup>2</sup> and Martin L. Kirk 1,\***


Received: 2 February 2020; Accepted: 2 March 2020; Published: 5 March 2020

**Abstract:** Here we highlight past work on metal–dithiolene interactions and how the unique electronic structure of the metal–dithiolene unit contributes to both the oxidative and reductive half reactions in pyranopterin molybdenum and tungsten enzymes. The metallodithiolene electronic structures detailed here were interrogated using multiple ground and excited state spectroscopic probes on the enzymes and their small molecule analogs. The spectroscopic results have been interpreted in the context of bonding and spectroscopic calculations, and the pseudo-Jahn–Teller effect. The dithiolene is a unique ligand with respect to its redox active nature, electronic synergy with the pyranopterin component of the molybdenum cofactor, and the ability to undergo chelate ring distortions that control covalency, reduction potential, and reactivity in pyranopterin molybdenum and tungsten enzymes.

**Keywords:** metal–dithiolene; pyranopterin molybdenum enzymes; fold-angle; tungsten enzymes; electronic structure; pseudo-Jahn–Teller effect; thione; molybdenum cofactor; Moco

#### **1. Introduction**

It is now well-established that all known molybdenum-containing enzymes [1–3], with the sole exception of nitrogenase, contain a common pyranopterin dithiolene (PDT) (Figure 1) organic cofactor (originally called molybdopterin (MPT)), which coordinates to the Mo center of the enzymes through the sulfur atoms of the dithiolene fragment. To date, the PDT component [4] of the molybdenum cofactor (Moco) is the only known occurrence of dithiolene ligation in biological systems. This cofactor is also found in anaerobic tungsten enzymes, and it may be one of the most ancient cofactors in biology [5]. The study of metal–dithiolene compounds (metallodithiolenes) has undergone a recent renaissance, with their synthesis, geometric structure, spectroscopy, bonding, and electronic structure having been recently highlighted [4,6–20]. Here, we briefly review the discovery of metallodithiolene compounds [13,21]. This history is followed by a more extensive discussion of key investigations into the myriad roles of the dithiolene ligands in the structure, bonding and reactivity of metal compounds, using multiple spectroscopic techniques, as well as theoretical calculations. Throughout this review, the key implications of these results for Mo and W enzymes are discussed.

**Figure 1.** The reduced tetrahydro form of the pyranopterin dithiolene (PDT) coordinated to Mo in the molybdenum cofactor (Moco). In the enzymes, the Mo ion can redox cycle between the MoIV, MoV, and MoVI oxidation states.

In the early 1960s, several research groups reported intensely colored square planar metal complexes with chelating sulfur-donor ligands that could stabilize metal compounds in a range of formal oxidation states related by one-electron oxidation-reduction (i.e., redox) reactions (Figure 2) [22–24]. McCleverty gave these novel ligands the general name "dithiolene" in order to emphasize their delocalized electronic structures [25]. These ligands are also described as being "non-innocent" due to the participation of the dithiolene ligands in the multiple one-electron reactions of their metal complexes and the inability to assign a specific oxidation state to the metal ion or the dithiolene ligands [11].

**Figure 2.** Square planer metallodithiolene complexes. R = CN, CH3, Ph, CF3.

Importantly, these non-innocent dithiolene ligands can modulate the nature of the covalent bonding with transition metal ions via the various redox states accessible to the dithiolene (Figure 3) [13]. The ene-1,2-dithiolate is the reduced form of the ligand and possesses six π-electrons. This ligand form is both a σ-donor and π-donor that usually forms strong covalent bonds with an oxidized transition metal ion, as is observed in the active sites of most pyranopterin Mo and W enzymes (e.g., Mo(V)/Mo(VI)-dithiolene bonds). The radical anion form with five π-electrons is usually found in molecules chelated by multiple dithiolene ligands, where extended delocalization of the π-electrons and mixed-valency assists in the stabilization of the metal–ligand bonds. The fully oxidized 1,2-dithione form of the ligand possesses only four π-electrons and can be described by two resonance structures (e.g., the 1,2-dithione and 1,2-dithiete). The low-lying empty π\* orbitals of the S=C bonds in the dithione can accept π-electron density from electron-rich low-valent transition metals [16,17], thereby stabilizing such compounds. However, dithione-containing low-valent metal complexes are encountered much less frequently than high-valent transition metal ions coordinated by reduced forms of dithiolene ligands.

**Figure 3.** Dithiolene redox states and resonance structures for the oxidized dithione/dithiete forms. (Adapted with permission from *Inorganic Chemistry*, 2016, 55, 785–793. Copyright (2016) American Chemical Society).

In 1982, Johnson and Rajagopalan proposed that Moco consisted of the Mo ion coordinated by the dithiolene fragment of the PDT (Figure 1), from the results of an elegant series of degradative, analytical and spectroscopic studies of sulfite oxidase [26]. This proposed structure was subsequently confirmed by X-ray crystallography [27,28], and numerous examples are now known [29]. Molybdenum and tungsten enzymes are the only known examples of dithiolene coordination in biology, and given the "non-innocent" behavior of dithiolene ligands in simple metal compounds, one may ask what role does dithiolene coordination play in molybdenum enzymes? Through a series of examples involving small molecules and enzymes, we will address this important question and how it relates to control of metal–ligand covalency, reduction potentials, and reactivity in pyranopterin Mo and W enzymes.

#### **2. Mo–Dithiolene Bonding**

#### *2.1. Early Descriptions of Mo–Dithiolene Bonding*

Some insight into the role of dithiolene coordination in enzymes is provided by the organometallic compounds of the general formula Cp2M(bdt), where Cp is C5H5 −, and M is either Mo, V or Ti. The fold angle of the dithiolene ligand depends on the formal d-electron count of the metal, and this angle ranges from nearly planar (9◦) for Mo (d2), to 35◦ for V (d1), and 46◦ for Ti (d0) (Figure 4). Lauher and Hoffman [30] related this increase in the fold angle with decreased d-electron count to donation from the filled out-of-plane Sπ<sup>+</sup> orbital to the in-plane metal d-orbital (Figure 5). For the molybdenum enzymes, these model compound results imply that the Mo–dithiolene fold angle in Moco could be related to the formal oxidation state of the Mo atom, with Mo(VI) (d0) sites possessing a relatively large fold angle and Mo(V) (d1) and Mo(IV) (d0) sites possessing smaller fold angles. Accurate fold angles are difficult to determine for large protein molecules, but values ranging from 6–33◦ have been calculated for various molybdenum enzymes [31]. The binding of substrate or inhibitors, and/or dynamic conformational changes in the protein, are expected to modulate the active site chelate fold angle and thereby affect enzyme reactivity [4,32].

**Figure 4.** Fold angle distortions as a function of redox orbital electron occupancy in a series of Cp2MIV(bdt) complexes. (Adapted with permission from *J. Am. Chem. Soc.* 2018, 140, 14777–14788. Copyright (2018) American Chemical Society).

**Figure 5.** Pictorial description of how the ligand fold angle modulates the degree of mixing between the dithiolene out-of-plane S orbitals (Sπ+) and the in-plane Mo(xy) redox orbital. The chelate ring fold is along the dithiolene S–S vector. (Adapted with permission from *Proc. Natl. Acad. Sci. USA.* 2003, 100, 3719–3724. Copyright (2003) National Academy of Sciences.

#### *2.2. Spectroscopic Investigations of Mo–Dithiolene Bonding*

#### 2.2.1. Electron Paramagnetic Resonance (EPR) Spectroscopy

An important spectroscopic signature of molybdenum enzymes, such as xanthine oxidase and sulfite oxidase, is a unique Mo(V) electron paramagnetic resonance (EPR) spectrum. The EPR spectra of the enzymes display a relatively large average g-value (gave = 1.97) and relatively small 95,97Mo hyperfine interactions (*hfi*) compared to the EPR spin-Hamiltonian parameters from typical inorganic Mo(V) complexes that possess hard N, O, and Cl donor ligands. The unique EPR parameters for molybdenum enzymes have been ascribed to covalent delocalization of electron density between the Mo(V) center and the sulfur atoms of the coordinated pyranopterin dithiolene unit [33]. The oxo-Mo(V) model compound Tp\*MoO(bdt) (Figure 6, where Tp\* is hydrotris-(3,5-dimethyl-1-pyrazolyl)borate and bdt is 1,2-benzenedithiolate)) displays Mo(V) EPR spin-Hamiltonian parameters that are very similar to those observed in the enzymes. This supports the proposal of dithiolene coordination in Mo enzymes [34], which has been confirmed by X-ray crystal structures [2]. Recent multidimensional variable frequency pulsed EPR studies of sulfite oxidase, where the sulfur atoms of Moco have been isotopically labeled with 33S (I = 3/2), have provided *direct* experimental evidence for delocalization of Mo(V) spin density onto the S atoms of the dithiolene fragment of Moco [35,36]. Density functional theory (DFT) computations show spin polarization effects and strong covalent intermixing between the in-plane metal dxy orbital and out-of-plane pz orbitals of the PDT dithiolene S atoms, which provide a mechanism for the observation of a significant 33S hyperfine interaction [12,36].

**Figure 6.** The Tp\*MoVO(bdt) model. Note that the apical oxo ligand can be changed to a terminal sulfido or nitrosyl to probe the electronic structure of the Mo–dithiolene unit. The bdt ligand can also be conveniently interchanged with a large variety of other dithiolenes.

#### 2.2.2. Electronic Absorption and Resonance Raman Spectroscopies

Experimental investigation of the electronic structures of the Mo centers of enzymes is difficult because of the intense absorptions from other chromophores (e.g., the *b*-type heme in sulfite oxidase and iron sulfur centers and FAD in xanthine oxidase) [37–41]. However, the effects of dithiolene coordination on electronic structure have been investigated for model oxo-Mo(V) compounds (Figure 6) by electronic absorption, XAS, magnetic circular dichroism (MCD), and resonance Raman (rR) spectroscopies [12,14–17,32,33,42–51]. For Tp\*MoO(bdt), the electronic absorptions at 19,400 cm−<sup>1</sup> (Band 4) and 22,100 cm−<sup>1</sup> (Band 5) are assigned to S <sup>→</sup> Mo charge transfer bands (Figure 7A) [12]. These assignments have been confirmed by rR spectroscopy (Figure 7A,B), which shows three resonantly enhanced vibrations at 362.0, 393.0, and 931.0 cm<sup>−</sup>1. The lower frequency vibrations (ν<sup>1</sup> and ν6) can be assigned to symmetric S–Mo–S stretching and bending vibrations, and the 931.0 cm−<sup>1</sup> frequency (ν3) is primarily the Mo≡O stretch. Figure 7C shows a molecular orbital diagram that is consistent with the spectroscopic data of Figure 7A,B. Band 5 of Figure 7A is assigned as <sup>ψ</sup>opa" <sup>→</sup> <sup>ψ</sup>xza", <sup>ψ</sup>yza' (blue arrow, Figure 7C), a transition which formally results in the promotion of an electron from an out-of-plane dithiolene molecular orbital to the nearly degenerate Mo dxz,yz-based orbitals, which are strongly antibonding with respect to the apical Mo≡O bond. This band assignment is supported by the rR enhancement of ν<sup>3</sup> (squares) with excitation into Band 5 (Figure 7A,B). The preferential enhancement of vibrations ν<sup>1</sup> (diamonds) and ν<sup>6</sup> (circles) upon excitation at 514.5 nm (Figure 7A,B) supports assignment of Band 4 as the electronic transition <sup>ψ</sup>ipa" <sup>→</sup> <sup>ψ</sup>xya' (red arrow, Figure 7C), which promotes an electron from the antisymmetric in-plane dithiolene orbital (ψipa") to the half-filled in-plane Mo dxy (ψxya') orbital. The intensity of this electronic transition illustrates the covalency of in-plane metal–dithiolene bonding and suggests that such a pseudo-σ-mediated process could play a role in one-electron transfer steps of enzyme catalysis.

**Figure 7.** (**A**) Solid state resonance Raman profiles and 5K mull electronic absorption spectrum for Tp\*MoVO(bdt). (**B**) Resonance Raman spectrum for Tp\*MoVO(bdt) (293K) using 514.5 nm excitation (75 mW). (**C**) General molecular orbital diagram for Tp\*MoVO(dithiolene) complexes. The z-axis is oriented along the Mo≡O bond and the energies of the molecular orbitals are not drawn to scale. Transitions are described in the text. (Adapted with permission from *Inorganic Chemistry*, 1999, 38, 1401. Copyright (1999) American Chemical Society).

#### **3. Synergistic Interactions between the Dithiolene and Pterin Components of the PDT**

Electronic coupling between the dithiolene and the pterin components of the PDT is most prevalent in the dihydropyranopterin form of the PDT [4,15,20,29,52]. This coupling is dramatically reduced in a tetrahydropyranopterin due to the loss of extended π-conjugation in these systems. Two-electron oxidation of the tetrahydro pyranopterin component of the PDT can result in

an unusual asymmetric dithiolene known as the "thiol–thione" form that leads to bond and electronic asymmetry in the metal–dithiolene core [4,15,52]. As depicted in Figure 8, the two-electron oxidized 10,10a-dihydropyranopterin can undergo an induced internal redox reaction upon protonation at the N-5 position that involves a subsequent charge transfer between the dithiolene chelate and the piperazine ring of the pterin. This protonation results in a dominant monoanionic thiol–thione chelate form of the ligand when bound to Mo or W. This thiol–thione character can also occur in the absence of protonation by the concept of resonance, which may also be described as configurational mixing between the thiol–thione and dithiol states. This type of thiol–thione chelate has been observed and studied in a small molecule Mo(IV) systems [4,15,20,52]. In these systems, excited state thiol–thione character was shown to be admixed into the ground state configuration using a variety of spectroscopic and computational probes of the electronic structure. The analysis of the data indicates that a two-electron oxidized pterin is inherently electron withdrawing, allowing for a low-lying dithiolene → pterin intraligand charge transfer (ILCT) state to mix with the ground state to provide a variable degree of thiol–thione character in the electronic ground state.

**Figure 8.** Oxidized PDT ligands: dihydropyranopterin (**a**) and protonated dihydropyranopterin (**b**) yielding the thiol/thione.

Definitive spectroscopic signatures are associated with the presence of a dihydropterin form of the PDT ligand. It is observed that the dithiolene → pterin intraligand charge transfer (ILCT) band is intense (E = 20,000–27,500 cm<sup>−</sup>1; ε ~ 10,000–16,000 M−<sup>1</sup> cm<sup>−</sup>1) [52], and there is considerable resonance enhancement of numerous Raman vibrations that can be assigned as originating from pterin and dithiolene C = C and C = N vibrations. Key resonance enhanced vibrational modes that can be used to characterize the presence of dihydropterin thiol–thione character in the enzymes include the 1508 cm−<sup>1</sup> and 1549 cm−<sup>1</sup> pyranopterin–dithiolene stretching frequencies that were observed in this Mo(IV) cyclized pyranopterin dithiolene model compound. This oxidized pyran ring closed form of the PDT has yet to be definitively observed in any pyranopterin Mo enzyme, but its presence would have profound implications on the electronic structure of the Mo site. Namely, the change in ligand charge from −2 to −1 leads to an asymmetric reduction in the charge donated by the monoanionic ligand compared to the dianionic dithiolene. Charge effects on oxygen atom transfer catalysis have recently been explored in model compounds showing dramatic rate enhancements in the oxidative half reaction that leads to substrate reduction [53]. This reactivity correlates with a large shift in the Mo(VI/V) reduction potential between cationic [Tpm\*MoO2Cl]<sup>+</sup> (−660 mV vs. Fc+/Fc) and charge neutral Tp\*MoO2Cl (−1010 mV vs. Fc+/Fc) [53]. The same effect on redox potential and reactivity would be expected in enzymes that could adopt an oxidized PDT with a thiol–thione configuration. The presence of a thiol–thione form of the PDT in an enzyme would also have a considerable impact on the active site electronic structure, and enable the pyranopterin to play a more significant role in catalysis by fine-tuning the Mo redox potential and providing a π-pathway for electron transfer regeneration of the active site [52]. Additionally, the asymmetry in the dithiolene (thiol/thione) charge donation would be expected to result in a significant *trans* effect or *trans* influence on oxo or sulfido ligands that are coordinated to the Mo or W ion and oriented *trans* to the thione sulfur.

#### **4. The Electronic Bu**ff**er E**ff**ect and Fold Angle Distortions**

#### *4.1. Photoelectron Spectroscopy (PES) Studies*

A common structural feature of the large group of pyranopterin Mo enzymes that catalyze a wide range of oxidation/reduction reactions in carbon, sulfur, and nitrogen metabolism is coordination by the sulfur atoms of one (or two) unique dithiolene groups derived from the side chain of a novel substituted pterin (PDT, Figure 1). Given the electronic lability of the dithiolene, a possible role of dithiolene coordination in molybdoenzymes is to buffer the influence of other ligands and changes in the formal oxidation state of the metal. Gas-phase photoelectron spectroscopy (PES) is a powerful tool for probing metal–ligand covalency in isolated molecules. Gas-phase ultraviolet PES of the molybdenum model complexes with the general formula Tp\*MoE(tdt) (Figure 6, where E = O, S, or NO, and tdt = 3,4-toluenedithiolate), exhibit nearly identical first ionization energies (6.88–6.95 eV) even though there is a dramatic difference in the electronic structure properties of the axial ligand. Collectively, these results have provided direct experimental evidence for the "electronic buffer" effect of dithiolene ligands [54].

Additional evidence for the electronic buffer effect of dithiolene ligands has been provided by gas-phase core and valence electron ionization energy measurements of the series of molecules Cp2M(bdt) (Figure 4, Cp = η5-cyclopentadienyl, M = Ti, V, Mo, and bdt = benzene-1,2-dithiolate). Comparison of the gas-phase core and valence ionization energy shifts provides a unique quantitative energy measure of valence orbital overlap interactions between the metal and the sulfur orbitals that is separated from the effects of charge redistribution. The results explain the large amount of sulfur character in the redox-active orbitals and the electronic buffering of oxidation state changes in metal–dithiolene systems. The experimentally determined orbital interaction energies also reveal a previously unidentified overlap interaction of the predominantly sulfur HOMO of the bdt ligand with the filled π orbitals of the Cp ligands, suggesting that direct dithiolene interactions with other ligands bound to the metal could be significant for other metallodithiolene systems in chemistry and biology [55].

#### *4.2. A Large Fold Angle Distortion in a Mo(IV)–Dithione Complex*

Mo(IV)–dithione complexes are much rarer than Mo(V)/Mo(VI)-dithiolene complexes. Recently, a detailed spectroscopic and computational study was performed on a novel Mo(IV)–dithione complex, MoO(SPh)2( i Pr2Dt0) (where <sup>i</sup> Pr2Dt0 = *N*,*N* -isopropylpiperazine-2,3-dithione) [17]. The structure of this unusual molecule was determined by x-ray crystallography and displays a remarkably large dithiolene fold angle (η = 70◦). This large fold angle was compared to that observed in more than 75 other metallodithiolene complexes found in the Cambridge crystallographic database, where fold angles were found to range from 0.3◦ to 37.3◦ with an average value for η of 12.5◦ [17]. The large fold angle distortion in the metallodithiolene ring of MoO(SPh)2( i Pr2Dt0) is reflected in its unusual electron absorption spectrum. The combination of an electron rich Mo(IV) center and electron donating thiolate (SPh) ligands results in the presence of low-energy Mo(IV) d(x2-y2) <sup>→</sup> dithione MLCT and thiolate → dithione LL'CT transitions as a result of the strong π-acceptor character of the dithione ligand. These spectral assignments are supported by resonance Raman profiles constructed for the 378 cm−<sup>1</sup> S–Mo–S symmetric stretch and the 945 cm−<sup>1</sup> Mo≡O stretch in addition to the results of TDDFT computations. The donor–acceptor nature of the complex was revealed in a molecular orbital fragments analysis using a donor fragment, [(PhS)2Mo(IV)]2<sup>+</sup> (F1) and an acceptor fragment, [<sup>i</sup> Pr2Dt0] (F2). The analysis showed that 21% of the F1 HOMO was mixed into the F2 fragment LUMO at a 70o fold angle. In contrast, only 5% of the F1 HOMO was mixed into F2 fragment LUMO in a planer configuration (η = 0◦), correlating the effective π-acceptor ability of the dithione with the ligand fold angle. The effects of this HOMO-LUMO mixing also affects the HOMO-LUMO gap, with the HOMO-LUMO gap increasing at larger fold angles (Figure 9). The increased covalency that results from the fold angle distortion represents an example of a strong pseudo-Jahn–Teller effect, vide

infra, involving vibronic coupling between the ground state and a low-energy excited state in the non-distorted (η = 0◦) geometry of this molecule. A scan of the potential energy surface as a function of this fold angle distortion coordinate results in an asymmetric double well potential (Figure 10), with the global minimum representing a ground state geometry with the dithione ligand fold distorted toward the apical oxo ligand. Thus, an oxidized dithione form of the PDT present in an enzyme active site would be expected to possess a very large ligand fold angle, unless the polypeptide enforces a more planer fold angle geometry.

**Figure 9.** Frontier orbital energies as a function of fold angle in MoIVO(SPh)2( i Pr2Dt0), which possesses a dithione π-acceptor ligand. (Adapted with permission from *Inorganic Chemistry*, 2016, 55, 785–793. Copyright (2016) American Chemical Society).

**Figure 10.** A double well in the ground state potential energy surface of MoIVO(SPh)2( i Pr2Dt0) as a function of the ligand fold angle. (Adapted with permission from *Inorganic Chemistry*, 2016, 55, 785–793. Copyright (2016) American Chemical Society).

#### *4.3. Low-Frequency Pyranopterin Dithiolene Vibrational Modes in Xanthine Oxidase*/*Dehydrogenase*

Low-frequency dithiolene distortions that are coupled to large electron density changes at the Mo ion represent an example of the electronic buffer effect [54], and have been probed in bovine xanthine oxidase (XO) and *R. capsulatus* xanthine dehydrogenase (XDH) using resonance Raman spectroscopy [40]. Computations have shown that exciting into a low-energy Mo(IV) → product metal-to-ligand charge transfer (MLCT) band results in a large degree of change transfer from the Mo(IV) HOMO to the product LUMO, resulting in an excited state with significant Mo(V) hole character (e.g., Mo(IV)–P<sup>0</sup> <sup>→</sup> Mo(V)–P·). Thus, the optical charge transfer process mimics the instantaneous one-electron oxidation of the Mo ion, which is encountered in the electron transfer reactions of the enzymes.

The Mo(IV) → 2,4-TV and Mo(IV) → 4-TV (2,4-TV = 2,4-thioviolapterin; 4-TV = 4-thioviolapterin) MLCT bands are red-shifted relative to the Mo(IV) → violapterin MLCT band [39,40,56–59]. The red-shift of the MoIV–2,4-TV and MoIV–4-TV MLCT bands eliminates spectral overlap with the absorption envelope of the 2Fe–2S spinach ferredoxin clusters and FAD. The elimination of the

FAD fluorescence background and spurious signals deriving from 2Fe–2S vibrations contributing to the Raman spectrum allow for the acquisition of very high-quality resonance Raman data. Multiple low-frequency (200–400 cm−1) Raman vibrations are observed to be enhanced when using laser excitation on resonance with the Mo(IV) → product MLCT band [40], and these have been assigned as a vibrational mode involving dithiolene folding, Mo≡O rocking, and pyranopterin motions (Band A: MoIV–4-TV = 234 cm<sup>−</sup>1; MoIV–2,4-TV = 236 cm<sup>−</sup>1), a ring distortion vibration that possesses both Mo–SH and pyranopterin motions (Band B: MoIV–4-TV = 290 cm<sup>−</sup>1; MoIV–2,4-TV = 286 cm<sup>−</sup>1), the symmetric S–Mo–S dithiolene core stretching vibration (Band C: MoIV–4-TV = 326 cm−1; MoIV–2,4-TV = 326 cm<sup>−</sup>1), and the corresponding asymmetric S–Mo–S dithiolene stretch (Band D: MoIV–4-TV = 351 cm<sup>−</sup>1; MoIV–2,4-TV = 351 cm<sup>−</sup>1) (Figure 11). Thus, the instantaneous generation of a hole on the Mo center (Mo(IV)–P0 <sup>→</sup> Mo(V)–P·) by photoexcitation is felt by the dithiolene chelate and extends all the way to the amino terminus of the PDT. The most resonantly enhanced mode in this spectral region is Band C, the symmetric S–Mo–S dithiolene core stretching, and the frequency of this mode and Band D are similar to those observed in Tp\*MoO(bdt) [12,32], which were assigned as the chelate ring symmetric S–Mo–S stretching and bending vibrations, respectively. Band A is significant, since it possesses dithiolene ring folding character indicating that electron density changes at Mo are buffered by a distortion along this low-frequency coordinate, as has been observed in the various model systems described in this review. These observations strongly support an electron transfer role for the PDT in catalysis, with the dithiolene contributing to the Mo–S covalency necessary for increasing the electronic coupling matrix element for electron transfer (*HDA*) and to affect the Mo reduction potential via the covalency in the Mo–Sdithiolene bonds.

**Figure 11.** Low-frequency rR spectra for wt, Q102G, and Q197A XDH, MoIV−4-TV (**a**) and MoIV−2,4-TV (**b**). Raman spectra were collected on resonance with the Mo(IV) → P MLCT band using 780 nm laser excitation (Adapted with permission from *Inorganic Chemistry*, 2017, 56, 6830–6837. Copyright (2017) American Chemical Society).

#### **5. Vibrational Control of Covalency**

A combination of MCD, electronic absorption, electron paramagnetic resonance, resonance Raman, and photoelectron spectroscopies has been used in conjunction with theory to reveal vibrational control of metal–ligand covalency in a series of Cp2M(bdt) complexes (M = Ti, V, Mo; Cp = η5-C5H5) [60] (Figure 4). The work is important because it has allowed for a detailed understanding of how redox orbital electron occupancy (Ti(IV) = d0, V(IV) = d1, Mo(IV) = d2,) affects the nature of the M–dithiolene bonding scheme at parity of the ligand set and at parity of charge. In this series of complexes, large changes in the metallodithiolene fold angle and electronic structure are observed as electrons are successively removed from the redox orbital (Figure 4). These electron occupancy effects on the fold angle distortion are now understood in terms of the pseudo-Jahn–Teller effect (PJT). PJT-derived molecular distortions originate from the mixing of the electronic ground state (Ψ0) with specific excited states (Ψi) [61,62]. The ground state–excited state energy gap (2Δ), the matrix elements (*F*0i) of the

vibronic contribution to the force constant (*F*), and the primary non-vibronic force constant (*K*0) all govern the degree of the ligand fold distortion according to:

$$F\_{0i} = \left\langle \Psi\_0 \middle| \frac{\partial H}{\partial \mathbf{Q}} \middle| \Psi\_i \right\rangle \tag{1}$$

$$F^2 > \Delta \cdot K\_0 \tag{2}$$

At the critical threshold defined by Equation (2), the metallodithiolene centers of Cp2M(bdt) can distort along the dithiolene fold angle coordinate to yield a double well potential energy surface (Figure 12), and the magnitude of the PJT distortion is maximized by a large *F*, a small Δ, and a small *K*0. Thus, the PJT distortion in these Cp2M(bdt) complexes effectively couples soft fold angle bending modes in the M-dithiolene chelate ring to the inherent electronic structure of the system via the d-electron count. Importantly, the mixing of low-energy charge transfer states into the ground state by the PJT effect controls the covalency of the M–S bonds.

**Figure 12.** The excited state (black) and ground state (red) potential energy surfaces associated with varying values of F<sup>2</sup> (Dotted: F<sup>2</sup> = 0, solid: F<sup>2</sup> = <sup>Δ</sup>·K0, dashed: F<sup>2</sup> = 2Δ·K0) When the condition F<sup>2</sup> > Δ·K0 is met one observes that the single-well ground state potential energy surface distorts into a double-well potential. The F2 <sup>&</sup>gt; <sup>Δ</sup>·K0 criteria describe a strong PJT effect. (Adapted with permission from *J. Am. Chem. Soc.* 2018, 140, 14777–14788. Copyright (2018) American Chemical Society).

One of the unique aspects of Mo–S and W–S bonding is the small energy gap between filled dithiolene-based orbitals and the lowest energy metal-based orbital, which naturally leads to low-energy charge transfer states that can mix with the ground state. Mode softening along the dithiolene fold coordinate is important in pyranopterin Mo and W enzymes since this leads to a potential energy surface where a large range of dithiolene fold angles may be sampled without paying a prohibitive energy penalty. This effect is maximized when *F*<sup>2</sup> - Δ·*K*0. Thus, a low-energy pathway is operative that can minimize energetically unfavorable reorganizational energy contributions along the reaction coordinate, which accompany redox changes at the metal ion. As mentioned previously, these fold angle distortions have been shown to be kinematically coupled to low frequency pyranopterin modes in XO and contribute to low-energy barriers for electron transfer regeneration of the active site. However, in the enzymes there may be either a competing or additive relationship between active site distortions that are driven via the d-electron count of the metal ion and distortions that are imposed by the protein. Vibronic coupling effects that derive from different occupancy numbers for the redox-active orbital will function to modulate the enzyme reduction potential in the oxidative and reductive half reactions of pyranopterin Mo and W enzymes, and this occurs by modulating the degree of metal–ligand covalency via low-frequency distortions at the active site.

#### **6. Conclusions**

This review focuses on the electronic structures, molecular structures, and spectroscopic properties of well-characterized metallodithiolene compounds in order to provide deep insight into the role(s) of metal–dithiolene bonding in pyranopterin dithiolene containing enzymes (Figure 1). The discovery, in the early 1960s, that transition metal dithiolene compounds undergo a series of one-electron oxidation-reduction reactions (Figure 2), provided the first evidence for the "non-innocence" of dithiolene ligands and the highly covalent nature of metal–dithiolene bonding. Additional links between metal–dithiolene covalency and electronic and molecular structure were posited from theoretical studies of bent metallocene−dithiolene compounds (Figure 4) by Lauher and Hoffman in 1976 [30], who related metal−dithiolene chelate ring "folding" with the metal ion d-electron configuration. Investigations of Mo dithiolene compounds by electronic absorption, resonance Raman, and EPR spectroscopies showed that S → Mo charge transfer bands dominate the visible spectrum and that there is substantial delocalization of spin density onto the S atoms of the dithiolene. Recent comprehensive studies of bent metallocene−dithiolene compounds have shown that low-energy ligand fold distortions arise from a pseudo-Jahn−Teller (PJT) effect, which involves vibronic coupling of the electronic ground state with electronic excited states to control metal−ligand covalency [60] (Section 5). This vibronic coupling process may play critical roles in the catalytic cycles of pyranopterin Mo and W enzymes by dynamic and/or static modulation of redox potentials and providing a superexchange pathway for electron transfer through the PDT framework. However, greater understanding of how geometric and electronic structure control reactivity, and define function in Mo and W enzymes, will require linking the concepts that have been developed for metallodithiolenes to the emerging results from studies of well-characterized compounds that mimic the pterin component of PDT (Section 3). Exploring the synergistic interactions between the dithiolene and pterin components of the PDT and the metal ion will be challenging, but such research promises to provide important insights into these critically important enzymes.

**Author Contributions:** J.H.E., J.Y., and M.L.K. collectively conceived and drafted this article. All authors have read and agreed to the published version of the manuscript.

**Funding:** M.L.K.'s research contributions to this article were funded by the National Institutes of Health (R01-GM-057378).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Structure: Function Studies of the Cytosolic, Mo- and NAD**+**-Dependent Formate Dehydrogenase from** *Cupriavidus necator*

**Russ Hille 1,\*, Tynan Young 2, Dimitri Niks 1, Sheron Hakopian 3, Timothy K. Tam 2, Xuejun Yu 4, Ashok Mulchandani <sup>5</sup> and Gregor M. Blaha 1,\***


Received: 19 May 2020; Accepted: 28 June 2020; Published: 6 July 2020

**Abstract:** Here, we report recent progress our laboratories have made in understanding the maturation and reaction mechanism of the cytosolic and NAD+-dependent formate dehydrogenase from *Cupriavidus necator.* Our recent work has established that the enzyme is fully capable of catalyzing the reverse of the physiological reaction, namely, the reduction of CO2 to formate using NADH as a source of reducing equivalents. The steady-state kinetic parameters in the forward and reverse directions are consistent with the expected Haldane relationship. The addition of an NADH-regenerating system consisting of glucose and glucose dehydrogenase increases the yield of formate approximately 10-fold. This work points to possible ways of optimizing the reverse of the enzyme's physiological reaction with commercial potential as an effective means of CO2 remediation. New insight into the maturation of the enzyme comes from the recently reported structure of the FdhD sulfurase. In *E. coli*, FdhD transfers a catalytically essential sulfur to the maturing molybdenum cofactor prior to insertion into the apoenzyme of formate dehydrogenase FdhF, which has high sequence similarity to the molybdenum-containing domain of the *C. necator* FdsA. The FdhD structure suggests that the molybdenum cofactor may first be transferred from the sulfurase to the C-terminal cap domain of apo formate dehydrogenase, rather than being transferred directly to the body of the apoenzyme. Closing of the cap domain over the body of the enzymes delivers the Mo-cofactor into the active site, completing the maturation of formate dehydrogenase. The structural and kinetic characterization of the NADH reduction of the FdsBG subcomplex of the enzyme provides further insights in reversing of the formate dehydrogenase reaction. Most notably, we observe the transient formation of a neutral semiquinone FMNH·, a species that has not been observed previously with holoenzyme. After initial reduction of the FMN of FdsB by NADH to the hydroquinone (with a kred of 680 s−<sup>1</sup> and Kd of 190 μM), one electron is rapidly transferred to the Fe2S2 cluster of FdsG, leaving FMNH·. The Fe4S4 cluster of FdsB does not become reduced in the process. These results provide insight into the function not only of the *C. necator* formate dehydrogenase but also of other members of the NADH dehydrogenase superfamily of enzymes to which it belongs.

**Keywords:** nicotinamide adenine dinucleotide (NADH); electron transfer; enzyme kinetics; enzyme structure; formate dehydrogenase; carbon assimilation

#### **1. Introduction**

The molybdenum- and tungsten-dependent formate dehydrogenases have drawn increased attention over the past 5–10 years due to the demonstration that under the appropriate conditions, most, if not all, are able to catalyze the reverse reaction, reduction of CO2 to formate, under the appropriate conditions. Indeed, some enzymes of this family, which include the closely related formylmethanofuran dehydrogenases of the Wood–Ljungdahl pathway, function physiologically to reduce CO2 to formate in what is probably the most evolutionarily ancient mechanism of carbon fixation.

*Cupriavidus necator* H16 (previously known as *Ralstonia eutropha*) has four formate dehydrogenases, of which, two are cytosolic enzymes that utilize NAD<sup>+</sup> as oxidizing substrate [1,2]. One of these contains molybdenum and is encoded by the *fdsGBACD* operon, the other possesses tungsten and is encoded by the *fdwAB* operon (presumably enlisting additional subunits from the *fds* operon) [3]. The molybdenum-containing enzyme was originally isolated and characterized by Bowien and coworkers, who showed that the mature holoenzyme belongs to the NADH dehydrogenase superfamily of enzymes [4–6]. The FdsA, FdsB, and FdsG subunits are homologous to corresponding subunits in the cytosolic arm of NADH dehydrogenase [7–10]. FdsA is homologous to the Nqo3 subunit of the NADH dehydrogenase from *Thermus thermophilus* [8,9], although the latter lacks a molybdenum center as found in the former [7]. The homology between FdsA and Nqo3 includes the presence of a histidine ligand to one of the Fe4S4 clusters. The C-terminal domain of FdsA, containing the molybdenum center, is also homologous to the crystallographically characterized FdhF formate dehydrogenase of *E. coli*, with the molybdenum-coordinating Cys 378 of FdsA equivalent to Sec 140 in FdhF [11]. The Fe4S4 and FMN-containing FdsB subunit is homologous to the *T. thermophilus* Nqo1 subunit and like Nqo1 also possesses a binding site for NADH/NAD+. The FdsG subunit is homologous to the *T. thermophilus* Nqo2 subunit and has a single Fe2S2 cluster. We report here recent work from our laboratories on both mechanistic and structural aspects of the *C. necator* enzyme.

#### **2. Catalysis of CO2 Reduction by the** *C. necator* **Formate Dehydrogenase**

Under physiological conditions, reducing equivalents enter the *C. necator* formate dehydrogenase holoenzyme at its molybdenum center and leave at the FMN, reducing NAD<sup>+</sup> to NADH; electron transfer between the molybdenum and flavin, which are separated by some 55 Å [10], is mediated by the intervening iron–sulfur clusters. Although originally reported to be unable to catalyze the CO2 by NADH, we have recently shown that the enzyme is indeed capable of doing so when CO2 (not bicarbonate) is used as substrate [12]. This is consistent with a mechanism for formate oxidation involving direct hydride transfer of the Cα-H to the MoVI = S group of the L2MoVIS(S-Cys) molybdenum center (where L is the bidentate enedithiolate-coordinated pyranopterin cofactor found in molybdenum- and tungsten-containing enzymes, present in this enzyme as the guanine dinucleotide) to L2MoIV(SH)(S-Cys), with CO2 rather than bicarbonate as the immediate product of the reaction [13]. The steady-state kinetic parameters have been determined in both the forward and reverse directions and are shown in Table 1.


**Table 1.** Steady-state kinetic parameters for the *C. necator* formate dehydrogenase.

The Haldane relationship for these parameters for an enzyme, such as formate dehydrogenase operating via a ping-pong mechanism with separate sites for the reductive and oxidative half-reactions, is as follows:

$$\mathbf{K\_{eq}} = \frac{\frac{\left(\mathbf{k\_{cat}}^{\text{forward}}\right)}{\left(\mathbf{k\_{m}^{\text{form}}}\right)} \cdot \frac{\left(\mathbf{k\_{cat}^{\text{forward}}}\right)}{\left(\mathbf{K\_{m}^{\text{AND}}}\right)}}{\frac{\left(\mathbf{k\_{cat}^{\text{reverse}}}\right)}{\left(\mathbf{k\_{cat}^{\text{current}}}\right)} \cdot \frac{\left(\mathbf{k\_{cat}^{\text{reverse}}}\right)}{\left(\mathbf{k\_{cat}^{\text{current}}}\right)}} = \frac{\frac{\left(201 \text{ s}^{-1}\right)}{\left(130 \text{ }\mu\text{M}\right)} \cdot \frac{\left(201 \text{ s}^{-1}\right)}{\left(310 \text{ }\mu\text{M}\right)}}{\frac{\left(10 \text{ s}^{-1}\right)}{\left(2700 \text{ }\mu\text{M}\right)} \cdot \frac{\left(10 \text{ s}^{-1}\right)}{\left(46 \text{ }\mu\text{M}\right)}} = 1250\tag{1}$$

The value 1250 compares favorably with the Keq calculated from the 100 mV difference between ΔE0 for the NAD+/NADH and CO2/formate couples of 2100, the disparity reflecting a ~10% uncertainty in kcatforward and kcatreverse, which are squared terms in numerator and denominator, respectively, of Equation (1).

The reaction can be pushed significantly in the direction of CO2 reduction by the addition of an NADH regeneration system [14]. As shown in Figure 1 left, addition of a catalytic amount of formate dehydrogenase to a solution that is 29.5 mM in CO2 (at 30 ◦C) and 300 μM in NADH results in the formation of 120 μM formate (as quantified by ion chromatography) and 130 μM NAD+—in other words, the reaction is tightly coupled. When the experiment is repeated with the addition of 50 mM glucose and a catalytic amount of glucose dehydrogenase for NADH regeneration, the amount of formate generated increases approximately 10-fold to 1.0 mM (Figure 1 right). This illustrates the potential commercial use of the enzyme for generation of formate from CO2 using a suitable source of reducing equivalents.

**Figure 1.** Catalysis of CO2 reduction to formate by *C. necator* formate dehydrogenase. (**Left**) the reaction of 0.2 μM holoenzyme with 29.5 mM CO2(aq) and 300 μM NADH, 20 mM Bis-Tris propane, pH 6.3, 30 ◦C. The absorbance change reflects the consumption of 130 μM NADH. The amount of formate accumulated as determined by ion chromatography was 120 μM [12]. (**Right**) the accumulation of formate under the same conditions upon addition of 50 mM glucose and catalytic glucose dehydrogenase to the reaction mix. Adapted with permission from *Biochemistry* **2019**, *58*, 1861–1868 [14]. Copyright (2019) American Chemical Society.

#### **3. Insertion of the Molybdenum Cofactor into Formate Dehydrogenase and Other Members of the DMSO Reductase Family**

As is seen with all members of the DMSO reductase family of molybdenum enzymes, the active site molybdenum center is deeply buried in the C-terminal domain of the FdsA subunit, and the means by which it is introduced into the apoenzyme is not presently understood. Given the extensive structural homology of this domain to the *E. coli* FdhF, one can consider the overall topology of the latter which consists of three interlaced domains that constitute the body of the protein, and a contiguous C-terminal domain that constitutes a "cap" over the cofactor in the holoenzyme (Figure 2A [11]). It is now well-accepted that all the Mo- and W-containing formate dehydrogenases possess a M=S group that is inserted into the metal coordination sphere as the final step of cofactor maturation [15]. Sulfur forms a more covalent bond with molybdenum and tungsten than does the more electronegative oxygen. As a result, there is less formal negative charge on the sulfur, e.g., making it better able to accept a hydride in

the course of the reaction. A similar argument has been made in the case of xanthine oxidase and related enzymes, which also require a Mo=S [10]. The sulfur transferase catalyzing this reaction (the product of the *fdhD* gene in *E. coli*, *fdsC* in the *C. necator* operon) is also thought to play a role in the insertion process, and the X-ray crystal structure of FdhD has recently been reported with two equivalents of GDP bound at the presumed position of the maturing molybdenum center [16]. Modeling the cofactor into the structure yields a complex with the apex of the molybdenum coordination sphere pointing into the channel through which the catalytically essential sulfur is delivered (via a cysteine desulfurase), with the *bis*(MGD) portion of the cofactor presenting a concave basket to the surface of the complex; the pyranopterin portion of both cofactors is solvent-exposed with the principal interactions with FdhD principally involving the guanine dinucleotide extensions of the cofactor (Figure 2B). In this orientation, the cofactor cannot be transferred directly to the body of the apo FdhF, where it is also oriented concave outwards with the two guanine dinucleotide arms extended into the protein, not out toward solvent. On the other hand, the molybdenum center in FdhF interacts with the cap domain of the protein principally via its pyranopterin rings, interacting minimally with the guanine dinucleotide arms. This being the case, it is possible to dock the cap (with bound cofactor) to the dimeric FdhD2·(GDP)2 complex in such a way that the GDP arms overlap; both proteins are in a position to interact optimally with the cofactor sandwiched between them. If this interaction is physiologically significant, the implication is that FdhD passes the now sulfurated and mature cofactor not to the body of apo FdhF but to its cap, which then closes over the body swinging the cofactor into position in the active site. In this way the highly unstable cofactor is never released into free solution. There have been a number of efforts in the past to identify regions on the surface of DMSO reductase family enzymes that interact with the cofactor insertion machinery. If the above analysis is correct, then this surface, on the face of the C-terminal cap domain that interfaces with the body of the enzyme, is buried in the structure of the holoenzyme. In support of this model, the C-terminal cap domains of the apo forms of both *E. coli* trimethylamine-*N*-oxide reductase TorA [17] and *E. coli* periplasmic nitrate reductase NapA [18] in complex with their respective chaperones TorD and NapD (that recognize the proteins' N-terminal twin-Arg signal sequence that targets them to the periplasm) have been reported to assume an open position in readiness to accept the mature molybdenum cofactor. This indicates that the C-terminal cap domain is indeed able to adopt the type of open configuration that is proposed here.

**Figure 2.** A comparison of the structures of FdhD and FdhF. Panel (**A**) top, the structure of FdhF (PDB 1FDO), with the body of the enzyme in *gray* and the C-terminal cap domain in *blue*. Middle, the body

of FdhF with the C-terminal cap domain removed, revealing the buried molybdenum center. Bottom, an enlargement of the molybdenum center, illustrating its disposition in the body of FdhF. Panel (**B**) top, views of the putative interfaces of the dimeric FdhD in complex with two equivalents of GDP (*gray*; PDB 4PDE) and the C-terminal cap domain of FdhF with the molybdenum center shown (*blue*). Bottom, the structures shown rotated toward one another illustrating the overlap between the GDP's of the FdhD structure and the molybdenum center of FdhF.

#### **4. The Crystal Structure of FdsBG**

The X-ray crystal structure of the FdsBG subcomplex of the *C. necator* formate dehydrogenase has also been determined at a resolution of 2.3 Å [19]. This fragment of the holoenzyme contains FMN and a Fe4S4 cluster in the FdsB subunit and a Fe2S2 cluster in FdsG; it lacks the FdsA subunit that contains the molybdenum center and additional iron–sulfur clusters. As expected, the structure of each subunit is closely related to its cognate subunits in the NADH dehydrogenases from *T. thermophilus* and *Aquifex aeolicus*, FdsB being homologous to the *T. thermophilus* Nqo2 and *A. aeolicus* NuoF subunits and contains FMN and a Fe4S4 cluster, and FdsG to the *T. thermophilus* Nqo1 and *A. aeolicus* NuoE subunits and contains a Fe2S2 cluster [20]. Compared to the *A. aeolicus* NuoF, FdsB has an RMSD of 1.48 Å for 394 Cα atoms and contains a Rossmann-like fold [21] encompassing the FMN binding site, a ubiquitin-like and a four-helix domain containing the Fe4S4 cluster [22] (Figure 3); all these structural elements are shared with NuoF. The Fe4S4 cluster is in a mostly hydrophobic environment close to the protein surface and is bound by C443B, C446B, C449B, and C489B (superscript B indicating the residue is in FdsB). The principal structural differences between FdsB and NuoF are in surface loops of the protein. However, the N-terminal portion of FdsB consists of a thioredoxin-like fold [20] not seen in NuoF or Nqo1. This domain lacks the Fe2S2 cluster typical of thioredoxins owing to mutation of the iron-coordinating cysteines (P10B, A15B, S45B, and F49B). This resembles the C-terminal portion of FdsG (RMSD 1.97 Å for 71 Cα atoms of the core of the domain) that contains the Fe2S2 cluster (vide infra).

The FdsG subunit is homologous to the NuoE subunit of *A. aeolicus* NADH dehydrogenase, consisting of an N-terminal four-helical bundle (residues 29G–74G) and a C-terminal thioredoxin-like domain (residues 79G–159G) that possesses the subunit's Fe2S2 cluster. These two domains are connected by a four-amino acid linker and are rotated by ~26◦ relative to the orientation seen in NuoE.

It is to be noted that while the N-terminal domains of FdsG and NuoE both have four helices, the first helix of FdsG runs parallel, but in NuoE, it runs across the second and third helices. The C-terminal thioredoxin-like domain of FdsG resembles the N-terminal thioredoxin-like domain of FdsB, but unlike the cofactorless FdsB domain, the FdsG domain contains a spinach ferredoxin-like Fe2S2 cluster [19]. This iron–sulfur cluster is again in a hydrophobic environment near the surface, coordinated by C86G, C91G, C127G, and C131G. Like *A. aeolicus* NuoE (but unlike *T. thermophilus* Nqo1), the C-terminus of FdsG is truncated at its C-terminus and lacks a disulfide bond.

At the FdsB–FdsG interface, two helices of the N-terminal four-helical bundle domain of FdsG contact the Rossmann-like domain of FdsB, as seen in the subunit interface between NuoE and NuoF. FdsG's C-terminal thioredoxin-like domain interacts with the same surfaces of the ubiquitin-like and Rossmann-like domains of FdsB as do the corresponding structural elements of NuoE and NuoF. The different placement of the ubiquitin and of the 183B–190B loop in FdsB as compared to NuoE results in an increase in the distance between the Fe2S2 cluster of FdsG and the Fe4S4 cluster of FdsB, being of ~1.0 Å as compared to the separation seen in NuoEF.

At a distance of nearly 21 Å (edge-to-edge), direct electron transfer between the two Fe/S clusters is expected to be extremely slow and unlikely to be physiologically relevant. On the other hand, each cluster is within 12 Å of the FMN isoalloxazine ring (Figure 4). Given the model that has been constructed for the holoenzyme based on the structures of FdhF and Nqo123, the Fe4S4 cluster of FdsB clearly lies on-path for electron transfer between the molybdenum center and FMN of the formate dehydrogenase, while the Fe2S2 cluster of FdsG is off-path. In the NADH dehydrogenases, it has been

suggested that this cluster may serve to temporarily store electrons during electron transfer out of the FMN once it has been reduced by NADH (vide infra). The FMN in FdsB is surrounded by a network of interactions that is also conserved in other NADH dehydrogenase-like enzymes [20,23] and is part of the solvent-accessible cavity that constitutes the NAD+/NADH binding site. The C8-methyl of the FMN points towards the Fe4S4 cluster, with the C4=O facing the Fe2S2 cluster (Figure 4).

**Figure 3.** The structure of FdsBG. (**Top**) the overall structure of FdsBG, with FdsB in *gray* and FdsG in *blue* (PDB 6VW8). The center structure is rotated about the horizontal or vertical axis as indicated to give the structures to the left and right, respectively. The two iron–sulfur clusters are rendered as CPK spheres and the FMN as CPK sticks. (**Bottom Left**) the domain structure of FdsB, with the thioredoxin-like domain in *red*, the Rossmann fold in *gray*, the ubiquitin-like domain in *green*, and the four-helix domain in *blue*; (**Center**) the domain structure of FdsG, with the N-terminal domain in gray and the thioredoxin-like domain in *red*; (**Right**) a comparison of the thioredoxin-like domains of FdsB (*upper*) and FdsG (*lower*).

**Figure 4.** Disposition of the redox-active centers of FdsBG. From left to right the Fe4S4 and FMN of FdsB and the Fe2S2 of FdsG (PDB 6VW8).

#### **5. EPR Characterization of the Iron–Sulfur Clusters of FdsBG**

Extended incubation of FdsBG complex with sodium dithionite (pH 7.0) yields two signals (as seen in Figure 5 top). The first is seen at liquid nitrogen temperatures as high as 200 K, with g1,2,3 = 2.000, 1.948, and 1.920 and linewidths of 1.4, 1.7, and 1.6 mT, respectively. These values are in good agreement

with previously reported parameters for the cluster designated as Fe/S1 for holoenzyme (Figure 5 middle, dotted spectrum) [13]. The second signal is seen only below 20 K and has g1,2,3 = 2.039, 1.955, and 1.891, and linewidths of 4.5, 1.4, and 5.3 mT (Figure 5 middle, dashed spectrum). This signal has not been previously seen with the holoenzyme [13], owing to its low intensity, broad linewidths, and overlap with the much stronger Fe/S3 signal, and is designated Fe/S5. In the *A. aeolicus* NuoEF and *T. thermophilus* Nqo12 NADH dehydrogenases, only the Fe2S2 clusters exhibit EPR signals at 77 K, while the Fe4S4 signals appear only below 50 K [24]. Further, the g-values g1,2,3 = 2.004, 1.945, and 1.917 for the N1a Fe2S2 cluster of the 24 kDa subunit of bovine complex I [25] (the homolog to FdsG) are in very good agreement with the Fe/S1 signal seen here in FdsBG, and accordingly, the Fe/S1 signal has been assigned to the Fe2S2 cluster of FdsG. The signal assigned to the N3 Fe4S4 cluster in the NADH dehydrogenase systems, with g-values of g1,2,3 = 2.037, 1.945, and 1.852, is also in good agreement with the Fe/S5 signal of the FdsBG complex, and we accordingly assign the Fe/S5 signal to the Fe4S4 cluster of FdsB. The Fe/S1 signal seen with holoenzyme was initially assigned to the His-coordinated Fe4S4 cluster of FdsA (Figure 5 bottom, dotted spectrum) [13], but that is clearly incorrect as FdsBG exhibits the signal. The Fe/S5 signal has similar g-values to those reported for the previously observed Fe/S3, but the signal-giving cluster couples to the molybdenum center and must be the proximal Fe4S4 cluster to the molybdenum center in FdsA; the Fe/S5 signal must, therefore, be a new signal not seen previously with holoenzyme. The assignment of EPR signal Fe/S5 to the Fe4S4 cluster of FdsB is consistent with our previously published structural model of holoformate dehydrogenase placing the Fe4S4 cluster near FdsA in the intact holoenzyme and on-path between the molybdenum and FMN of the enzyme [10].

**Figure 5.** EPR spectra of the iron–sulfur clusters of the FdsBG complex. (**Top**) the observed iron–sulfur EPR spectrum (solid) and simulated composite spectrum (dashed) of dithionite-reduced FdsBG, collected at 9 K with modulation amplitude of 8 Gauss and microwave power of 2 μW. The sample was prepared by incubation of 125 μM of FdsBG complex in 100 mM potassium phosphate, pH 7.0 with 2 mM buffered sodium dithionite under anaerobic conditions for 1 h at room temperature prior to freezing. (**Middle**) the individual signals resulting from the simulation of the composite spectrum above. The spectrum corresponding to the previously assigned Fe/S1 (dotted) and an additional Fe/S5 signal (dashed) are resolved only in the FdsBG complex. (**Bottom**) a comparison of the spectra seen with FdsBG (dashed) and holoenzyme, at 20 K (dotted). The dashed lines mark the location of g1 and g3 features of the Fe/S3 component seen in the spectrum of the reduced holoenzyme (see text) [19].

#### **6. Rapid-Reaction Kinetics of FdsBG Reduction by NADH**

The rapid-reaction kinetics of FdsBG reduction by NADH have also been examined. This is the reverse of the physiological reaction for the formate dehydrogenase, but is the physiological direction seen in the NADH dehydrogenases. A typical kinetic transient is shown in the inset of Figure 6 left. The reaction is biphasic, with the fast phase of the reaction completing within 120 ms. A plot of the observed kfast as a function of NADH is hyperbolic (Figure 6 left), yielding a limiting kred of 680 s−<sup>1</sup> and Kd NADH of 190 μM at 5 ◦C. We note that this rate is more than sixfold faster than the limiting rate of reduction of holoenzyme at high formate, some 20-fold faster than kcat for formate oxidation (13), and 350-fold faster than kcat for CO2 reduction [12] under comparable conditions.

**Figure 6.** The kinetics of FdsBG reduction by NADH. (**Left**) a plot of kfast vs. NADH, yielding a limiting kred of 680 s−<sup>1</sup> and a Kd NADH of 190 μM. The *inset* shows a typical biphasic transient, as described in the text. The reaction conditions were 100 mM potassium phosphate, pH 7.0, 5 ◦C. (**Center**) spectra seen at the times indicated in the reaction of 10 μM FdsBG with 5 μM NADH, as obtained with a photodiode array detector on the stopped-flow. (**Right**) difference spectra obtained using the spectra obtained at the times indicated. The formation of FMNH· in the fast phase of the reaction is reflected in the positive feature in the 500–650 nm region in the *dotted* difference spectrum and its loss during the slow disproportionation phase of the reaction by the negative feature in the same region in the *dashed* difference spectrum [19].

When the reaction was repeated using substoichiometric NADH (so that the FdsBG1e- formed on disproportionation will not be further reduced by reaction with a second equivalent of NADH), the absorbance increase seen at the end of the fast phase of the reaction implies formation of the neutral semiquinone FMNH· (Figure 6 center, dashed). This species has not seen in previous work with holoenzyme and must be the result of transfer of a single electron from the fully reduced flavin hydroquinone (formed on its initial two-electron reduction by NADH) to the Fe2S2 cluster of FdsG iron, leaving the flavin as FMNH· (and giving rise to the Fe/S1 EPR signal; see further below and Scheme 1). On a second, slower timescale, two equivalents of the FdsBG2e- formed upon reaction with NADH disproportionate to give one equivalent each of FdsBG1e- and FdsBG3e-. To the extent this occurs, FdsBG1e- will have its Fe2S2 cluster (Fe/S1) reduced and its FMN oxidized, whereas FdsBG3ewill have both Fe/S1 and the FMN reduced.

$$
\begin{bmatrix}
\mathsf{F}\mathsf{e}\_{2}\mathsf{S}\_{2}^{\otimes\mathsf{x}} \\
\end{bmatrix}
\begin{bmatrix}
\mathsf{F}\mathsf{e}\_{2}\mathsf{S}\_{2}^{\otimes\mathsf{x}} \\
\end{bmatrix}
\begin{bmatrix}
\mathsf{F}\mathsf{e}\_{2}\mathsf{S}\_{2}^{\otimes\mathsf{x}} \\
\end{bmatrix} \begin{bmatrix}
\mathsf{F}\mathsf{e}\_{2}\mathsf{S}\_{2}^{\otimes\mathsf{x}} \\
\end{bmatrix} = \\
\mathsf{F}\mathsf{e}\_{4}\mathsf{S}\_{4}^{\otimes\mathsf{x}} & \mathsf{G}\mathsf{e}\_{4}\mathsf{S}\_{4}^{\otimes\mathsf{x}} \\
\end{bmatrix}
$$

**Scheme 1.** Proposed electron transfer in FdsBG upon NADH reduction.

The UV/Visible signature of FMNH· [26] is seen as the positive feature in the 500–650 nm region in the difference spectrum between fully oxidized enzyme and that collected at 0.3 s after the initial reaction of FdsBG with NADH (Figure 6 right, black spectrum). The extended negative feature in the 300–500 nm region is due to reduction of the Fe/S clusters. The subsequent disappearance of the

FMNH· formed in the fast phase of the reaction is reflected in the pronounced negative feature in the 500–650 nm region seen in the 0.3–10 s difference spectrum (Figure 6 right, dashed difference spectrum). Here, the two broad, negative peaks centered at 570 and 605 nm reflect the presence of FMNH· at 0.3 s but its loss over the next 10 s. Formation of FMNH· has independently been confirmed in a freeze-quench EPR experiment. Figure 7 shows the spectrum seen at 150 K when FdsBG is reacted with NADH and frozen after ~40 ms (Top), which exhibits the EPR signatures of both FMNH· and Fe/S1. In addition to demonstrating the formation of FMNH·, this experiment establishes that it is the Fe2S2 cluster of FdsG that becomes reduced in forming FMNH·. Subtracting out the Fe/S1 signal from the measured EPR spectrum yields an isotropic EPR signal with a giso = 2.003 and a linewidth of ~1.9 mT, reflecting formation of the neutral rather than anionic semiquinone (Figure 7 middle) [27,28]. No additional iron–sulfur signal attributable to the Fe/S4 signal is seen at 9 K, even when the experiment is repeated in the presence of excess NADH, indicating that it does not get reduced in the course of reduction of FdsBG by NADH (Figure 7 bottom).

**Figure 7.** EPR of the neutral flavin semiquinone of FdsBG. (**Top**) The EPR spectrum seen by rapid-freeze quench (quenching time ~40 ms) on reaction of 40 μM FdsBG with 0.8 mM NADH at 0 ◦C. The spectrum was obtained at 150 K with modulation amplitude of 8 Gauss and microwave power of 0.4 mW. (**Middle**) the spectrum of FADH·, obtained by subtracting out the Fe/S1 contribution (solid line) along with a simulation (dashed line). (**Bottom**) The EPR spectrum seen on mixing 160 μM of FdsBG with 60 μM NADH at room temperature for 10 s prior to freezing. The EPR spectrum was collected at 9 K with modulation amplitude of 8 Gauss and microwave power of 2 μW, and it exhibits only the Fe/S1 signal with no contribution from Fe/S4. All samples were prepared in 100 mM potassium phosphate, pH 7.0, under anaerobic conditions [19].

#### **7. The Thioredoxin-Like Domain of FdsB**

The N-terminal thioredoxin-like domain of FdsB is extremely similar to the Fe2S2 ferredoxin domain from *A. aeolicus* [21], which like the *C. necator* formate dehydrogenase is known to dimerize. The largest contact interface between the two FdsBG protomers within the asymmetric unit of the crystal is between the two FdsB thioredoxin-like domains, suggesting that these contribute significantly to the dimerization of the holoenzyme.

The buried area is small (~560 Å2), however, with only a few specific interactions across the interface, which are not highly conserved among the NADH dehydrogenase family. In addition, FdsBG remains monomeric throughout purification. It thus appears that the dimer seen in the asymmetric unit is simply the result of crystal packing [29,30].

#### **8. Electron Transfer in FdsBG**

During normal turnover with holoformate dehydrogenase, reducing equivalents from the oxidation of formate pass from the active site molybdenum center through a chain of iron–sulfur clusters to the FMN, which ultimately reduces NAD+. FdsBG is able to reduce NAD+, as are the corresponding subcomplexes from NADH dehydrogenases [31,32] and the NAD+-reducing hydrogenase [33]. Consistent with the Fe4S4 cluster of FdsB being on-path between the molybdenum center and FMN, it lies close to the C8-methyl of FMN and to highly conserved surface residues that are implicated in the interaction of FdsB with FdsA (i.e., E441B–S487B, I451B–G452B, and K292B–L298B). By contrast, the Fe2S2 cluster of FdsG lies further from the region implicated in FdsA binding. Nevertheless, as summarized in Scheme 1, the present evidence indicates that the initial electron transfer event out of the FMN in FdsB upon reduction by NADH is to the off-path Fe2S2 cluster, which has the higher reduction potential relative to the on-path Fe4S4 cluster. In the case of the NADH dehydrogenases, a similar electron transfer to the Fe2S2 cluster is thought to minimize formation of neutral flavin semiquinone, FMNH·, thereby reducing the accumulation of reactive oxygen species [20,22].

#### **9. Concluding Remarks**

The cytosolic and NAD+-dependent formate dehydrogenase is fully capable of catalyzing the reduction of CO2 to formate using NADH as a source of reducing equivalents. The values for the forward and reverse steady-state kinetic parameters are consistent with the expected Haldane relationship. The addition of an NADH-regenerating system consisting of glucose and glucose dehydrogenase increases the yield of formate approximately 10-fold, suggesting the commercial potential of the enzyme as an effective means of CO2 remediation.

A consideration of the recently reported structure of the FdhD sulfurase that transfers a catalytically essential sulfur to the maturing molybdenum cofactor prior to insertion into apoenzyme suggests that the cofactor may first be transferred to the C-terminal cap domain of apo formate dehydrogenase, which then closes over the body of the enzyme to assemble the active site, rather than transferring the cofactor directly to the body of the protein.

The X-ray crystal structure of the FdsBG fragment of the *C. necator* formate dehydrogenase is reviewed, as are the kinetics of its reduction by NADH. Notably, the neutral semiquinone FMNH· is observed transiently in the course of the reaction of FdsBG with NADH. In a reaction analogous to the physiological reduction of NADH dehydrogenases by NADH, after initial reduction of the FMN of FdsB by NADH to the hydroquinone (with a kred of 680 s−<sup>1</sup> and Kd of 190 μM), one electron is rapidly transferred to the Fe2S2 cluster of FdsG, leaving FMNH·, as characterized by both UV/visible spectroscopy and EPR. The Fe4S4 cluster of FdsB does not become reduced in the process.

Finally, we note that the cryo-EM structure of the *R. capsulatus* formate dehydrogenase, which is very similar to the *C. necator* enzyme that has been the focus here, has very recently been reported at a resolution of 3.3 Å [34]. The overall structure of the FdsBG fragment of the *R. capsulatus* protein is very much in agreement with that described here for the *C. necator* protein, and the overall structure is also in good agreement with that predicted previously on the basis of the structures of the *E. coli* FdhF formate dehydrogenase and the *T. thermophilus* NADH dehydrogenase [10]. Not previously anticipated is the proximity (9.5 Å edge-to-edge) of the Fe4S4 clusters designated A4 in the two FdsA subunits of the dimeric Fds(ABGD)2 dimer, which suggests reducing equivalents are able to pass easily between protomers (a situation also seen in the even more complex formylmethanofuran dehydrogenase from *Methanothermobacter wolfeii* [35]). Interestingly, the A4 Fe4S4 cluster is the one possessing histidine as a ligand and presumably having a higher rather than lower reduction potential. Importantly, the cryo-EM work clearly demonstrates that the small FdsD subunit is an integral component of the mature holoenzyme, interacting intimately with the C-terminal domain of FdsA and apparently stabilizing the inserted molybdenum cofactor. The position of FdsD is not inconsistent with the model for cofactor incorporation considered here, although its role in cofactor insertion remains to be elucidated. Very interestingly, evidence is presented that cryo-EM is able to distinguish between oxidized and reduced iron–sulfur clusters in complex systems. If found to be generally applicable, this would constitute a very significant advantage of the cryo-EM method over other structural methods and provide information presently obtainable by spatially resolved anomalous dispersion refinement [36].

**Author Contributions:** R.H. supervised the EPR and kinetic work described here, and prepared the manuscript. T.Y., S.H., and T.K.T. analyzed the X-ray crystallographic work with FdsBG under the supervision of G.M.B., who also contributed to preparation of the manuscript. D.N. performed the EPR and kinetic work described. X.Y. performed the work involving CO2 reduction to formate under the supervision of A.M., who also contributed to preparation of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by a grant from the U.S. Department of Energy (DE-SC0010666 to R.H.).

**Acknowledgments:** We thank the staff of the following beamlines for their assistance in data collection: 5.0.1 and 5.0.2 of the Berkeley Center of Structural Biology at Advanced Light Source (ALS), 24-ID-C of the Northeastern Collaborative Access Team at Advanced Photon Source (APS), and BL7-1 of the Stanford Synchrotron Radiation Lightsource (SSRL). The Berkeley Center for Structural Biology is supported in part by the Howard Hughes Medical Institute. The Advanced Light Source is a Department of Energy Office of Science User Facility under Contract No. DE-AC02-05CH11231. The Pilatus detector on 5.0.1 was funded under NIH grant S10OD021832. The ALS-ENABLE beamlines are supported in part by the National Institutes of Health, National Institute of General Medical Sciences, grant P30 GM124169. The Northeastern Collaborative Access Team beamlines are funded by the National Institute of General Medical Sciences from the National Institutes of Health (P30 GM124165). The Advanced Photon Source is a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. The Stanford Synchrotron Radiation Lightsource, SLAC National Accelerator Laboratory, is supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Contract No. DE-AC02-76SF00515. The SSRL Structural Molecular Biology Program is supported by the DOE Office of Biological and Environmental Research and by the National Institutes of Health, National Institute of General Medical Sciences (including P41GM103393). The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of NIGMS or NIH.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
