*Article* **Crystal Structures of [Fe]-Hydrogenase from** *Methanolacinia paynteri* **Suggest a Path of the FeGP-Cofactor Incorporation Process**

#### **Gangfeng Huang 1, Francisco Javier Arriaza-Gallardo 1, Tristan Wagner 1,2 and Seigo Shima 1,\***


Received: 18 August 2020; Accepted: 15 September 2020; Published: 17 September 2020

**Abstract:** [Fe]-hydrogenase (Hmd) catalyzes the reversible heterolytic cleavage of H2, and hydride transfer to methenyl-tetrahydromethanopterin (methenyl-H4MPT<sup>+</sup>). The iron-guanylylpyridinol (FeGP) cofactor, the prosthetic group of Hmd, can be extracted from the holoenzyme and inserted back into the protein. Here, we report the crystal structure of an asymmetric homodimer of Hmd from *Methanolacinia paynteri* (pHmd), which was composed of one monomer in the open conformation with the FeGP cofactor (holo-form) and a second monomer in the closed conformation without the cofactor (apo-form). In addition, we report the symmetric pHmd-homodimer structure in complex with guanosine monophosphate (GMP) or guanylylpyridinol (GP), in which each ligand was bound to the protein, where the GMP moiety of the FeGP-cofactor is bound in the holo-form. Binding of GMP and GP modified the local protein structure but did not induce the open conformation. The amino-group of the Lys150 appears to interact with the 2-hydroxy group of pyridinol ring in the pHmd–GP complex, which is not the case in the structure of the pHmd–FeGP complex. Lys150Ala mutation decreased the reconstitution rate of the active enzyme with the FeGP cofactor at the physiological pH. These results suggest that Lys150 might be involved in the FeGP-cofactor incorporation into the Hmd protein in vivo.

**Keywords:** [Fe]-hydrogenase; FeGP cofactor; guanylylpyridinol; conformational changes; X-ray crystallography

#### **1. Introduction**

[Fe]-hydrogenase (Hmd) catalyzes the reversible hydride transfer to methenyltetrahydromethanopterin (methenyl-H4MPT<sup>+</sup>) from H2 (Figure 1a) [1,2]. This reaction is involved in the hydrogenotrophic methanogenic pathway [2,3]. [Fe]-hydrogenase forms homodimer and contains an active site cleft at the two dimeric interfaces (Figure 1b) [4,5]. The active-site cleft binds the iron-guanylylpyridinol (FeGP) cofactor as the prosthetic group (Figure 1c) [6]. The FeGP cofactor contains a low spin Fe(II), which is ligated with two CO, one acyl-C and pyridinol nitrogen [5,7–12]. This cofactor is covalently bound to the protein via cysteine-S ligand at the iron site [5,10,13] (Figure 1c). The pyridinol ring is substituted with two methyl- and one guanosine monophosphate (GMP) groups [8]. The GMP part is bound to the mononucleotide-binding site of the Rossmann-fold-like structure of the Hmd protein [4].

**Figure 1.** Structure and function of [Fe]-hydrogenase. (**a**) Reaction catalyzed by [Fe]-hydrogenase. The side chain of H4MPT was omitted. (**b**) Crystal structure of [Fe]-hydrogenase (holoenzyme) from *Methanococcus aeolicus* (PDB: 6HAC). Two monomers are shown as a cartoon model (grey and blue). The FeGP cofactor is depicted as a ball and stick model. A water molecule is bound to the iron in the open resting state. (**c**) Chemical structure of the FeGP cofactor. Cys176 thiolate is covalently bound to the iron of the cofactor.

The FeGP cofactor is extractable from [Fe]-hydrogenase in the presence of 60% methanol, 1-mM 2-mercaptoethanol and 1% NH3 [14]. The isolated cofactor is stabilized by 2-mercaptoethanol, which makes a complex at the iron site [10,12,15]. The FeGP cofactor can also be isolated in the presence of 50% acetic acid. An acetate ligand binds to the iron site to stabilize the iron complex of the cofactor [12]. The FeGP cofactor is decomposed by UV-A/blue light [16] and hydrogen peroxide [17,18]. A decomposition product of the organic part of the FeGP cofactor is guanylylpyridinol (GP), in which the acyl-methyl substituent was hydrolyzed to a carboxymethyl group [12].

Hmd proteins can be heterologously produced in *Escherichia coli* as an apo-form that does not contain the FeGP cofactor [19]. Crystal structure of the apo-Hmd from *Methanocaldococcus jannaschii* indicated that the active-site cleft of the apoenzyme is in a closed conformation [4]. When the extracted FeGP cofactor is mixed with the apoenzyme, the cofactor binds to the protein in the active-site cleft and generates the active holoenzyme [19]. Crystal structures of holoenzymes without substrate bound have always been observed in an open conformation, in which the active-site cleft is exposed to bulk solvent [5]. The structure of the Hmd from *Methanococcus aeolicus* holoenzyme in complex with methenyl-H4MPT<sup>+</sup> showed that upon binding of methenyl-H4MPT<sup>+</sup>, the active-site cleft closes [20]. In the closed tertiary complex, the iron site of the FeGP cofactor is activated by expulsion of the water molecule bound on the iron site. The empty iron coordination site is proposed to be the H2-binding site [20].

Many mimic complexes of the FeGP cofactor have been synthesized [2,21–26]. Some of the mimics are composed of similar complex structures to the FeGP cofactor containing Fe(II) [27] or Mn(I) [25] as the metal center with two CO and pyridinol but lack the GMP moiety. Several mimic complexes exhibit catalytic activities of H2 activation and hydride transfer to chemical compounds [22–25]. Recently, reconstituted active semi-synthetic [Fe]-hydrogenases were produced by incorporation of the mimic complexes to the apoenzyme [25,28]. The reconstituted enzyme exhibited only a few percent of the activity of the native enzyme [28]. The reconstitution requires a longer period than that with the FeGP

cofactor to achieve full activity [28]. Addition of GMP during the reconstitution of the mimic-complexes increased the enzymatic activity to a certain extent [28].

In this work, we heterologously produced the Hmd apo-form from a mesophilic methanogenic archaeon, *Methanolacinia paynteri* (pHmd, National Center for Biotechnology Information Reference Sequence: WP\_048153035.1), which belongs to the *Methanomicrobiales* order. The structures of Hmd from the *Methanomicrobiales* order have not been reported so far. We crystallized pHmd after in vitro reconstitution with the isolated FeGP cofactor from native Hmd purified from *Methanothermobacter marburgensis* [14]. These crystallization trials yielded two unexpected structures. The first one was an asymmetric pHmd homodimer containing one monomer bound with the FeGP cofactor and the other monomer in the apo-form. The second crystal structure was a symmetric homodimer in complex with the broken product of the FeGP cofactor (GP). In addition, co-crystallization experiments of pHmd with GMP and an iron mimic complex (see Section 3) yielded a symmetric homodimer in complex with only GMP. The crystal structures of asymmetric Hmd, and Hmd bound with GMP or GP have not been reported before. Based on the structures, we propose a possible trajectory of the binding process of the FeGP cofactor to the protein to produce the active holoenzyme.

#### **2. Results and Discussion**

#### *2.1. Crystal Structure of the Asymmetric Homodimer of pHmd*

Reconstitution of the pHmd holoenzyme was performed in the presence of a slight excess of FeGP cofactor (0.18 mM) relative to the pHmd protein (0.13 mM). The specific activity of the reconstituted enzyme was variable in the holo-form obtained in each reconstitution experiment (60–250 U/mg at 40 ◦C) under standard assay conditions (oxidation of methylene-H4MPT), which was substantially lower than that of the native Hmd from *M. marburgensis* (~400 U/mg at 40 ◦C) [14], the reconstituted Hmd from *Methanocaldococcus jannaschii* (jHmd) (~400 U/mg at 40 ◦C) [14], and the reconstituted Hmd from *Methanococcus aeolicus* (3000 U/mg at 40 ◦C) [20]. The 2.1 Å crystal structure obtained from this reconstituted preparation contained one homodimer in the asymmetric unit, in which one monomer is bound to the FeGP cofactor and the other is in an apo-form (Figure 2, Table 1). The active-site cleft bound with the FeGP cofactor was in the open conformation and the second monomer without the cofactor was in the closed conformation. The formation of the asymmetric structure indicated that the open/closed conformational change of a monomer of the Hmd homodimer occurs independently from another monomer.

**Figure 2.** Crystal structure of the asymmetric homodimer of pHmd. (**a**) Overview of the asymmetric homodimer of pHmd. One monomer (left) is the apo-form and the second monomer (right) is bound to the FeGP cofactor. (**b**) Comparisons between the apo-form of the asymmetric pHmd homodimer (orange) and apo-form of jHmd (pink, PDB: 2B0J). The chain from the next monomer of pHmd in the central domain is distinguished by the green color. (**c**) Comparison between the holo-form of the asymmetric pHmd homodimer (green) and holo-form of jHmd (blue, PDB: 3F47). The chain from the next monomer of pHmd in the central domain is distinguished by the orange color. The FeGP cofactor is shown by ball and stick models. For both pictures in (**b**,**c**), the N-terminal domain is superposed.




**Table 1.** *Cont.*

<sup>a</sup> Values relative to the highest resolution shell are within parentheses. <sup>b</sup> Rfree was calculated as the Rwork for 5% of the reflections that were not included in the refinement. <sup>c</sup> rmsd, root mean square deviation.

The N- (residues 1–241) and C- (residues 253–342) terminal domains of the two monomers of the asymmetric homodimer are very similar to each other: 217 Cα superposed with a root mean square deviation (rmsd) of 0.192 Å for the N-terminal domain and 74 Cα superposed with a rmsd of 0.131 Å for the C-terminal domain. The N-terminal domains of the apo- and holo-form monomers of the asymmetric pHmd overlapped with those of apo- and holo-forms of Hmd from *M. jannaschii* (jHmd), respectively (Figure 2b,c). However, when the central domain composed of the dimeric C-terminal domains are superposed, a deviation of the N-terminal domain is observed between pHmd apo- and holo-forms compared to those of apo- and holo-forms of jHmd, respectively (Figure S1). Such variation is caused by the asymmetric structure; the holo conformation impacts the apo conformation at the central domain and vice versa.

According to the similar N-terminal domain structures, the holo-form of pHmd shows an identical binding mode of the FeGP cofactor as observed in the jHmd holoenzyme (Figure S2). In contrast, the apo-forms of pHmd and jHmd showed a slight difference at the loop involved in the FeGP cofactor coordination. One of the differences between the pHmd and jHmd apo-forms is the location of Lys150, which moved outside from the active site in jHmd (Figure S3). This movement might also be attributed to the crystal packing. Lys150 is conserved in the Hmd-encoding genes in the genomes reported with exceptions found in fifteen genomes of *Methanobrevibacter* species (e.g., *M. smithii*), where the lysine position varies to glutamate (Figure S4). Notably, the Hmd activity of the cell extract from the *Methanobrevibacter* species was very low [29,30].

#### *2.2. Crystal Structure of pHmd in Complex with GMP*

We obtained crystals from the solutions containing the reconstituted pHmd with an iron complex in the presence of GMP. This mimic complex has been used for reconstitution of semisynthetic jHmd (see Section 3) [28]. The 1.55 Å crystal structure revealed a symmetric pHmd homodimer in the closed conformation, which was fully occupied with GMP but without mimic complex (Figure 3, Table 1). The N-terminal domain of the GMP-binding structure and the apo-form monomer of pHmd in the asymmetric homodimer superposed well (295 Cα superposed with rmsd of 0.416 Å, Figure 3a). The GMP binding site is identical to that of the FeGP cofactor observed in the pHmd–FeGP complex of the asymmetric pHmd. The residues in direct contact with GMP and the GMP moiety of the FeGP cofactor have the same orientations (Figure 3b). The Lys150 side chain adopts the same orientation as in the apo form; however, binding of the GMP slightly rearranged the loop 110–116 and rigidified its surroundings as observed by a lower B-factor profile.

**Figure 3.** Structural comparison of the apo-form of the pHmd asymmetric-homodimer (grey) and the GMP-bound form (orange) of pHmd. (**a**) Monomers of each structure is shown by cartoon model. The N-terminal domains were superposed. (**b**) Zoom-up view of the GMP binding site. GMP is shown as ball and stick model. Hydrogen bonds are indicated by dashed lines.

#### *2.3. Crystal Structure of pHmd in Complex with GP*

From the reconstituted pHmd holoenzyme with the FeGP cofactor, in addition to the asymmetric homodimer crystal, another crystalline form was obtained. This form diffracted to 1.7 Å resolution (Table 1). The asymmetric unit of the crystal contained two homodimers in the closed conformation. Contrary to the other reconstituted holo-Hmd structures, the pyridinol part was only partially visible in the electron density and could be modelled for only one monomer in the four monomers in the asymmetric unit. The FeGP cofactor appeared to be decomposed to GP during crystallization process. The structures of the N-terminal domain and the residues binding GMP and GP superposed well (Figure 4a). Binding of GP induces the local conformational rearrangement of the two loops containing Lys150 and Cys175, respectively, compared to the apo form. In the pHmd–GP structure, the van der Waals interaction between the hydroxyl group of the pyridinol ring and amino group of Lys150 (2.7 Å) stabilizes the pyridinol ring of GP in this structure (Figure 4). The Cys175 side chain was modelled in two conformations of 80% and 20% occupancy in the broad electron density. The interaction between the amino group of Lys150 and the pyridinol does not appear to be optimal for hydrogen-bonding. However, in the incorporation process of the FeGP cofactor, a slight tilt of the pyridinol ring would improve the interaction.

**Figure 4.** Structural comparison of the GMP-bound form (orange) and the GP-bound form (cyan) of pHmd. (**a**) Monomers of each structure are shown by a cartoon model. The N-terminal domains are superposed. (**b**) Zoomed-in view of the active site. GMP and GP are shown by ball and stick models. Hydrogen bonds are indicated by dashed lines.

The carboxymethyl part of the pyridinol group was not visible, which indicates that the elongated carboxymethyl substituent of the GP does not interact with the protein. This observation contrasts with the structure of hexameric Hmd from *M. marburgensis* in the open conformation obtained in an oxidized broken state. The carboxymethyl group was visible and bound with an Fe atom coordinated by His203 and Cys172, and Asp189 from loop of another dimer (Figure S5a) [18]. The loop containing Asp189 is not conserved in the pHmd sequence. The absence of the loop structure explains why the GP binding mode is different and the iron was removed from the cleft after decomposition (Figure S5b).

The closed conformation of the active-site cleft of the pHmd in complex with GP and also GMP indicated that binding of the GMP and pyridinol parts of the FeGP cofactor does not induce the opening of the active-site cleft. The formation of the Cys175-S–Fe bonding and the resulting conformational change of the loop at Cys175 might trigger the open state. The crystal structure of the reconstituted Cys176Ala-mutated jHmd holoenzyme (Cys176 of jHmd is equivalent to Cys 175 of pHmd) has been reported [10]. This enzymatically inactive jHmd holoenzyme, lacking the Cys176-S–Fe bonding, was crystallized in the open conformation even in the presence of methylene–H4MPT bound, which indicated that the Cys176Ala mutation hindered the open/closed conformational change essential for the catalytic activity.

In the pHmd holo-form in the asymmetric homodimer, Lys150 is dissociated from the pyridinol ring and the loop containing Lys150 moved slightly away (Figure 5). Lys150 should move further away from the active site in the closed conformation induced by methenyl-H4MPT<sup>+</sup> binding, because the Lys150 side chain clashes with the phenyl ring part of methenyl-H4MPT<sup>+</sup> observed in the holoenzyme structure Hmd from *Methanococcus aeolicus* in complex with methenyl-H4MPT<sup>+</sup> (Figure S6) [20]. In the ternary complex, the Lys150 comes into contact with the side chain of methenyl-H4MPT<sup>+</sup>, mainly via a water network, which might therefore have an additional role in the substrate binding.

**Figure 5.** Structural comparison of the GP-bound form (cyan) of pHmd and the FeGP-bound form (yellow) of the pHmd asymmetric homodimer. (**a**) Monomers of each structure are shown by a cartoon model. The N-terminal domains were superposed. (**b**) Zoomed-in view of the active site. GP and FeGP are shown by ball and stick models. In the GP-bound form, the carboxymethyl group of GP is not visible. Hydrogen bonds are indicated by dashed lines.

#### *2.4. E*ff*ect of Lys150Ala Substitution on the Reconstitution Rate of pHmd*

The structural analysis of pHmd in complex with GP indicated that Lys150 could interact with the FeGP cofactor in the incorporation process. This hypothesis was tested using the pHmd Lys150Ala variant. Prior to the reconstitution assay, we measured the kinetic parameters of the wild-type enzyme and Lys150Ala variant reconstituted under the standard reconstitution condition. Unexpectedly, the Lys150Ala variant exhibited much higher *V*max and *K*<sup>m</sup> values (640 U/mg and 160 μM) than those of the wild-type enzyme (66 U/mg and 6 μM) in the oxidation reaction of methylene-H4MPT at pH 6.0 (Figures S7 and S8). The *V*max and *K*<sup>m</sup> of the reduction reaction of methenyl-H4MPT<sup>+</sup> with H2 at pH 7.5 are 1300 U/mg and 62 μM for the wild-type enzyme and 820 U/mg and 110 μM for the K150A variant, respectively. Increase of the *K*<sup>m</sup> values of the Lys150Ala variant is consistent with the observation of the contact of Lys150 with the side chain of methenyl-H4MPT described in the last section.

The formation of the holoenzyme by binding of the FeGP cofactor was followed by monitoring the increase of enzymatic activity after addition of the FeGP cofactor (4–700 nM) to the assay solution, which contained the 4-nM apoenzyme, and 20-μM methenyl-H4MPT<sup>+</sup> under H2 at pH 7.5 or 20-μM methylene–H4MPT under N2 at pH 6.0. The time course of the change of the absorbance at 336 nm was recorded (Figure 6a,b,e,f). We calculated the specific activity (U/mg) at each time point (per 2 s) from the slope of the absorbance change (Figure 6c,d,g,h). In these experiments, we assume that the increase of the Hmd activity indicates the increase of the active holoenzyme in the assay by incorporation of the FeGP cofactor into the apoenzymes. Hence, the specific activity is a function of the concentration of the reconstituted enzyme and the residual substrate concentration in the assay. In the assay condition at pH 7.5, which is the physiological pH [31], the wild-type enzyme was quickly reconstituted and reached the maximum activity within 25 s in the presence of 700-nM FeGP cofactor (Figure 6c). The Lys150Ala variant also exhibits activity, but the increase of the specific activity at the same condition was much slower than the case of the wild-type enzyme; to reach the maximal activity, 130 s was required (Figure 6d). These results support the hypothesis that Lys150 contributes to the binding kinetics of the FeGP cofactor to the protein. In the case of the reverse reaction, oxidation of methylene–H4MPT at pH 6.0, the Lys150Ala mutation did not affect the reconstitution of the holoenzyme (Figure 6e–h). To obtain the reconstitution rate from the data, we simulated the reconstitution curves (Figures S9 and S10). The simulated curves of reconstitution fitted to the observed curve when the reconstitution rate (*k*<sup>2</sup> in Figure S9) of the FeGP cofactor to the wild and Lys150Ala apoenzymes are assumed as 0.01 <sup>±</sup> 0.005 <sup>μ</sup>M−1·s−<sup>1</sup> and 0.0024 <sup>±</sup> 0.0003 <sup>μ</sup>M−1·s−<sup>1</sup> at pH 7.5 and 0.10 <sup>±</sup> 0.01 <sup>μ</sup>M−1·s−<sup>1</sup> and 0.14 <sup>±</sup> 0.03 <sup>μ</sup>M−1·s−<sup>1</sup> at pH 6.0, respectively. These data support the function of Lys150 in the binding kinetics of incorporation of the FeGP cofactor in the physiological pH. Protonation of the 2-hydroxy group of the FeGP cofactor might affect the incorporation rate of the FeGP cofactor into the protein.

**Figure 6.** Reconstitution of the [Fe]-hydrogenase holoenzyme from the apoenzyme of the wild-type (**a**,**c**,**e**,**g**) and Lys150Ala variant (**b**,**d**,**f**,**h**) with the FeGP cofactor. Changes of the absorbance at 336 nm according to the reduction of methenyl-H4MPT<sup>+</sup> (**a**,**b**) and oxidation of methylene-H4MPT (**e**,**f**) were recorded. The color area indicates the standard deviation of three tests. The enzymatic activity (U/mg) (**c**,**d**,**g**,**h**) of each 2 s was calculated from the data of (**a**,**b**,**e**,**f**), respectively. Concentrations of the FeGP cofactor were 700 (•), 350 (+), 100 (Δ), 50 (-), 10 (-) and 4 nM (♦). The plots of the data of the wild type and the Lys150Ala variant are shown in red and blue, respectively.

#### **3. Materials and Methods**

#### *3.1. Chemicals and Reagents*

Tetrahydromethanopterin (H4MPT) and methenyl-H4MPT<sup>+</sup> were isolated from *M. marburgensis* cells [32]. Methylene-H4MPT was produced by the reaction of H4MPT with formaldehyde [14]. The FeGP cofactor was isolated from [Fe]-hydrogenase from *M. marburgensis* as described previously [14]. All other chemical compounds used in this work were purchased from Sigma-Aldrich (Darmstadt, Germany).

#### *3.2. Gene Synthesis of [Fe]-Hydrogenase from Methanolacinia Paynteri*

The [Fe]-hydrogenase gene from *M. paynteri* (NCBI Reference Sequence: WP\_048153035.1) was modified for the codon usage optimization as shown below and synthesized by GenScript. The DNA synthesized was inserted into the expression vector pET-24b(+) at the *Nde*I and *Sal*I restriction-enzyme digestion sites. Genes of the Lys150Ala variant were synthesized using the template of the wild-type gene.

5 -CATATGACAATAAAGAAGGTAGCTATACTAGGAGCAGGGTGTTATAGGACTCACTCA GCGACCGGCATTACCAACTTTGCGCGTGCGTGCGAGGTGGCGGAAATGGTTGGTAAACCGG AGATCGCGATGACCCACAGCACCATTGCGATGGCGGCGGAACTGAAGTACCTGGCGGGCAT CGACAACATCGTGATTAGCGATCCGAGCTTCGCGGGCGAGTTTACCGTGGTTAAGGACTTCG ATTACAACGAAGTTATCAAGGCGCACAAAGAGAACCCGGAAACCATCATGCCGAAGATTCGT GAGAAAGTGAACGAACTGGCGAAAACCGTTCCGAAGCCGCCGAAAGGCGCGATCCACTTTG TGCACCCGGAGGACCTGGGTCTGAAGGTGACCACCGACGATCGTGAAGCGGTTCGTGACG CGGATCTGATCATTACCTGGCTGCCGAAGGGTGACATGCAGAAAGGCATCATTGAGAAGTTC GCGGGTGATATCAAGCAAGGCGCGATCATTACCCACGCGTGCACCATTCCGACCACCCTGTT CTACAAAATCTTTGAGGAACTGGGCATTGCGGATAAGGTGGAAGTTACCAGCTATCACCCGG GTGCGGTGCCGGAGATGAAAGGCCAGGTTTACATCGCGGAAGGTTATGCGAGCGAGGAAGC GATCAACACCATTTACGAGCTGGGTAAGAAAGCGCGTGGTCATGCGTTTAAGCTGCCGGCGG AACTGATTGGTCCGGTTTGCGACATGTGCGCGGCGCTGACCGCGATTACCTACGCGGGTCTG CTGGTGTATCGTGATGCGGTTATGAACATTCTGGGTGCGCCGGCGGGTTTCAGCCAGATGAT GGCGACCGAGAGCCTGGAACAAATCACCGCGTATATGAAGAAAGTGGGTATTAAAAACCTGG AGGAAAACCTGGACCCGGGTGTTTTCCTGGGCACCGCGGATAGCATGAACTTTGGCCCGATT GCGGAGATTCTGCCGACCGTTCTGAAGAGCCTGGAAAAGCGTGCGAAATAAGTCGAC-3 .

#### *3.3. Enzyme Production, Purification and Reconstitution*

The apoenzymes of [Fe]-hydrogenase from *M. paynteri* were heterologously overproduced in *E. coli* BL21(DE3). The recombinant *E. coli* was cultivated in the tryptone–phosphate (TP) medium containing 50 μg/mL kanamycin at 37 ◦C [33]. When the optical density of *E. coli* at 600 nm became 0.6–0.8, 1-mM isopropyl β-d-thiogalactopyranoside (IPTG) was added to induce expression of the targeted gene. Cells were harvested by centrifugation using Avanti JXN-26 centrifuge with JLA-10.500 rotor (Beckman-Coulter, Krefeld, Germany) at 8000 rpm for 30 min at 4 ◦C. The wet cells (5–10 g) were suspended in 50-mM 3-(*N*-morpholino)propanesulfonic acid (MOPS)/KOH pH 7.0 containing 1-mM dithiothreitol (DTT). Cells were disrupted on ice by sonication for 10 min using SONOPULS GM200 (Bandelin, Berlin, Germany) with KE76 tip with 50 cycles. The cell debris and the unbroken cells were removed by centrifugation using an Avanti JXN-26 centrifuge with a JA-25.50 rotor (Beckman-Coulter) at 15,000 rpm for 30 min at 4 ◦C. Ammonium sulfate (2-M final concentration) was added to the supernatant. Precipitates were removed by centrifugation using an Avanti JXN-26 centrifuge with JA-25.50 rotor at 15,000 rpm for 30 min at 4 ◦C. The supernatant was loaded on a Phenyl Sepharose High Performance column (75 mL, GE Healthcare Life Sciences, Solingen, Germany) and eluted with a linear gradient of ammonium sulfate from 2 M to 0 M in 50-mM MOPS/KOH buffer pH 7.0 containing 1-mM DTT. Fractions containing the apoenzyme of pHmd were collected and concentrated by using Amicon Ultra-4 Centrifugation filters (30-kDa cut-off). To further purify the apoenzyme of pHmd, the concentrated apoenzyme sample was loaded to a HiPrep 16/60 Sephacryl S-200 HR gel filtration column (120 mL, GE Healthcare Life Sciences) using 25-mM Tris(hydroxymethyl)aminomethane (Tris)/HCl buffer pH 7.5 containing 150-mM NaCl, 5% glycerol and 2-mM DTT. To increase the purity of protein, the gel filtration repeated two times using the same conditions [20]. Finally, the purified apoenzyme was concentrated to 50–100 mg/mL.

Protein concentration was measured by Bradford method using bovine serum albumin as the standard. Reconstitution was performed under dark conditions in an anoxic tent (Coy) with a gas phase of 95%N2/5%H2 at 8 ◦C by mixing the 0.13-mM apoenzyme and the 0.18-mM FeGP cofactor (at a molecular ratio of 0.75:1), respectively, as previously described [14].

#### *3.4. Enzyme Activity Assay*

The enzyme activity was anaerobically measured as previously described using a 1-mL quartz cuvette containing 0.7 mL assay mixture [14]. For the reduction of methenyl-H4MPT<sup>+</sup> with H2, 120-mM potassium phosphate pH 7.5 containing 1-mM ethylenediaminetetraacetic acid (EDTA) was used as the assay buffer under 100% H2. For the oxidation of methylene-H4MPT under N2, 120-mM potassium phosphate pH 6.0 containing 1-mM EDTA was used as assay buffer under 100% N2. The assay was started by adding 10 μL of 0.01 mg/mL enzyme solution (final concentration in the assay was 0.14 μg/mL), and the decrease (reduction)/increase (oxidation) of absorbance at 336 nm was recorded. Its specific activity was calculated using the extinction coefficient of methenyl-H4MPT<sup>+</sup> (ε336nm = 21.6 mM−1·cm<sup>−</sup>1) [14]. One unit (U) activity is the amount of the enzyme catalyzing the formation or consumption/formation of 1 μmol methenyl-H4MPT per min.

The reconstitution rate of the holoenzyme was kinetically analyzed in a quartz cuvette. The assay solution for the reduction reaction contained 120-mM potassium phosphate buffer pH 7.5, 1-mM EDTA, 4-nM (0.15 μg/mL) pHmd apoenzyme and 20-μM methenyl-H4MPT<sup>+</sup> under 100% H2. In the oxidation reaction, the assay solution contained 120-mM potassium phosphate buffer pH 6.0, 1-mM EDTA, 4-nM (0.15 μg/mL) pHmd apoenzyme and 20-μM methylene-H4MPT under 100% N2 The reaction was started by the addition of 4–700-nM cofactor (final concentration). A change of the absorbance at 336 nm was recorded at 40 ◦C.

#### *3.5. Simulation of the Enzyme Kinetics Data*

Michalis–Menten kinetics-parameters were obtained by simulation of the substrate consumption and product formation by numerical integration of the equations derived from mass action kinetics (Figure S7). The reconstitution rate with the FeGP cofactor and simulated enzymatic activities were calculated by numerical integration using the equations shown in Figure S9. All simulations were coded in Python 3.7 using Spyder 4.1 development environment and the following libraries: SciPy [34], NumPy [35], Matplotlib [36] and pandas [37].

#### *3.6. Crystallization*

[Fe]-hydrogenase holoenzyme from *M. paynteri* (pHmd) was crystallized under 95%N2/5%H2 at 8 ◦C using 96-well two-drop MRC crystallization plates (sitting drop vapor diffusion method). For the crystallization of pHmd–GP and pHmd–FeGP complexes, 0.7 μL of 25 mg/mL reconstituted holoenzyme was mixed with 0.7 μL reservoir solution (from crystallization kits) under yellow light and incubated under dark conditions. The best diffracting crystal of pHmd–GP came out within two weeks in 25% *w*/*v* polyethylene glycol 1500 and 100-mM succinic acid/sodium dihydrogen phosphate/Glycine (SPG) buffer pH 8.5 (JBScreen Wizard 3&4 HTS, Jena Bioscience, Jena, Germany). For pHmd asymmetric dimer form, the crystals grew within two weeks in 30% *w*/*v* polyethylene glycol 4000, 200-mM lithium sulfate and 100-mM Tris pH 8.5 (JBScreen Wizard 3&4 HTS, Jena Bioscience).

The reconstituted holoenzyme with the Fe complex was prepared as previously described [28]. An Fe(II) complex [(2-CH2CO-6-HOC5H3N)Fe(CO)3I] (complex 5 in [26]) was dissolved in methanol containing 1% acetic acid [28]. For reconstitution, 0.4 mL of 10-mM Fe complex solution was mixed anaerobically with a 7.6 mL solution of 100-mM sodium acetate pH 5.6, 0.02 mM apoenzyme and 2-mM GMP (final concentrations) and incubated on ice for one hour. The buffer of the reconstituted enzyme was exchanged by three-time concentration/dilution cycles using a 30 kDa ultrafilter with 10 mM MOPS/KOH pH 7.0 and finally concentrated to 25 mg/mL. For crystallization, 0.7 μL of the reconstituted pHmd solution was mixed with 0.7 μL reservoir solution (from crystallization kits) under

yellow light and incubated under dark conditions at 10 ◦C. The best crystal appeared within one month in 20% *w*/*v* polyethylene glycol 3350 and 200-mM magnesium formate reservoir solution (JBScreen Wizard 3&4 HTS, Jena Bioscience).

#### *3.7. Data Collection and Refinement*

The crystals of the asymmetric homodimer and the one containing GP were flash-frozen (3–5 s) in their crystallization reservoir solution supplemented with 10% *v*/*v* glycerol under 95%N2/5%H2. The crystal of pHmd bound with GMP was flash-frozen (3–5 s) in its crystallization reservoir solution containing 30% *v*/*v* glycerol under 95%N2/5%H2. All diffraction experiments were performed at 100 K on beamline BM30A (French Beamline for Investigation of Proteins) at the European Synchrotron Radiation Facility (ESRF) equipped with an ADSC Q315r charge-coupled device detector. The data were processed with XDS [38] and scaled with SCALA from the CCP4 suite [39]. The structure of pHmd–FeGP was determined by molecular replacement with PHASER [40] by decoupling the N- and C-terminal domain of the native Hmd from *M. marburgensis*in complex with 2-naphthylisocyanide (PDB: 4JJF) as templates. The structures of pHmd–GP and pHmd–GMP were solved with PHASER by using the monomer in the closed conformation of the asymmetric homodimer. The models were manually built with COOT [41] and refined with Phenix [42] and BUSTER (Bricogne G., Blanc E., Brandl M., Flensburg C., Keller P., Paciorek W., Roversi P, Sharff A., Smart O.S., Vonrhein C., Womack T.O. (2017). BUSTER version 2.10.3. Cambridge, United Kingdom: Global Phasing Ltd.). The final models were validated using the MolProbity server (http://molprobity.biochem.duke.edu) [40]. Data collection, refinement statistics and PDB code for the deposited model are listed in Table 1. The hydrogens were omitted in the final deposited model. The figures were generated and rendered with PyMOL (version 1.7, Schrödinger, Cambridge, UK). Alignments were performed by Clustal Omega [43]. The figures were made using ESPript 3.0 [44].

#### **4. Conclusions**

Based on the crystal structures of the pHmd apo- and holo-form—the latter bound with GMP, GP or the FeGP cofactor—we propose a trajectory of the isolated FeGP cofactor incorporation into the apoenzyme. First, binding of the GMP part guides correct positioning of the FeGP cofactor, which slightly opens the active-site cleft to engage the binding of the pyridinol part at the flexible Cys175 loop and induces small local conformational change at the loop containing Lys150. The iron site of the FeGP cofactor forms a covalent Cys175-S–Fe bonding upon exchange with the acetate/2-mercaptoethanol ligand bound in the free cofactor. Sequential binding of the GMP and pyridinol moieties might allow a correct covalent bonding between Cys175-thiolate and the Fe site. Lys150 might guide the binding of the pyridinol part to the specific position. The Lys150Ala mutation analysis supported this hypothesis. These results are of general interest for studying how nucleotide-containing cofactors and coenzymes are incorporated into the protein, and for developing semi-synthetic [Fe]-hydrogenase using mimic complexes. A plausible strategy of incorporation of the mimic complexes into the Hmd protein is to synthesize mimic compounds, which contains the GMP moiety at the position 4 of the pyridinol ring. Another strategy might be modifying the Hmd apoenzyme to enhance a smooth Fe–S bond formation of the mimic cofactor in the absence of the GMP moiety, in which the loop containing Lys150 might be the target of modification.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-6740/8/9/50/s1, Figure S1. Observation of subtle rearrangements of the asymmetric pHmd when the C-terminal domain of holo-jHmd and apo-jHmd are superposed, Figure S2. Superposition of the FeGP cofactor binding sites in pHmd (PDB: 6YKA), Hmd from *Methanococcus aeolicus* (aHmd) (PDB: 6HAC) and jHmd (PDB: 3F47), Figure S3. Comparison of the loop involved in the FeGP cofactor coordination in the jHmd and pHmd apo-forms. Figure S4. Alignments of Hmd amino acid sequences from different organisms, Figure S5. Comparison of the GP-binding states in the structure of Hmd from *M. marburgensis* obtained in an oxidized broken state with that in the structure of GP bound form of pHmd, Figure S6. Superposition of the active sites of apo- and holo-forms of Hmd. Figure S7. Equations used for the simulation of the modelled reaction and calculation of the kinetic parameters. Figure S8. Simulation of the progressive curves of the reactions. Figure S9. Equations used for the simulation of the binding

constant of the FeGP cofactor to the pHmd apoenzyme, Figure S10. Simulation of the change of the activity in the reconstitution assay of pHmd.

**Author Contributions:** Conceptualization, S.S.; methodology, S.S.; software, T.W.; validation, T.W.; formal analysis, G.H., T.W., F.J.A.-G. and S.S.; investigation, G.H., T.W. and F.J.A.-G.; resources, S.S..; data curation, G.H., F.J.A.-G., T.W. and S.S.; writing—original draft preparation, S.S.; writing—review and editing, G.H., F.J.A.-G., T.W. and S.S.; visualization, G.H and T.W.; supervision, S.S.; project administration, S.S.; funding acquisition, S.S. and T.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Max Planck Society (to T.W. and S.S.) and the Deutsche Forschungsgemeinschaft priority program (SPP 1927) (SH 87/1-1, to S.S.).

**Acknowledgments:** We thank Xile Hu (École polytechnique fédérale de Lausanne), who provided us the iron complex. The authors thank the staff from the BM30A (FIP) beamline at the European Synchrotron Radiation Facility (ESRF) for their availability and advice during data collection.

**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* **Theoretical Insights into the Aerobic Hydrogenase Activity of Molybdenum–Copper CO Dehydrogenase**

**Anna Rovaletti 1, Maurizio Bruschi 1, Giorgio Moro 2, Ugo Cosentino <sup>1</sup> and Claudio Greco 1,\* and Ulf Ryde 3,\***


Received: 14 September 2019; Accepted: 4 November 2019; Published: 9 November 2019

**Abstract:** The Mo/Cu-dependent CO dehydrogenase from *O. carboxidovorans* is an enzyme that is able to catalyse CO oxidation to CO2; moreover, it also expresses hydrogenase activity, as it is able to oxidize H2. Here, we have studied the dihydrogen oxidation catalysis by this enzyme using QM/MM calculations. Our results indicate that the equatorial oxo ligand of Mo is the best suited base for catalysis. Moreover, extraction of the first proton from H2 by means of this basic centre leads to the formation of a Mo–OH–Cu<sup>I</sup> H hydride that allows for the stabilization of the copper hydride, otherwise known to be very unstable. In light of our results, two mechanisms for the hydrogenase activity of the enzyme are proposed. The first reactive channel depends on protonation of the sulphur atom of a Cu-bound cysteine residues, which appears to favour the binding and activation of the substrate. The second reactive channel involves a frustrated Lewis pair, formed by the equatorial oxo group bound to Mo and by the copper centre. In this case, no binding of the hydrogen molecule to the Cu center is observed but once H2 enters into the active site, it can be split following a low-energy path.

**Keywords:** CO dehydrogenase; dihydrogen; hydrogenase; quantum/classical modeling; density functional theory

#### **1. Introduction**

Carbon monoxide dehydrogenases (CODHs) have proved to be an essential component for the biogeochemical carbon monoxide (CO) consumption. They contribute to maintenance of sub-toxic concentration of CO in the lower atmosphere by processing approximately 2 × 108 tons of it annually [1]. To date, only two enzymes have been found to be able to use CO as carbon and energy source [2]. The first type is represented by the oxygen-sensitive Ni,Fe-dependent CO dehydrogenase enzyme (NiFe-CODH), found in anaerobic bacteria and archea, whereas in aerobic carboxido-bacteria, such as *Oligotropha carboxidovorans*, the same function is performed by the oxygen-tolerant MoCu-dependent CO dehydrogenase (MoCu-CODH) enzyme [3]. These bacterial enzymes catalyze the oxidation of CO to CO2 following the reaction:

$$\text{CO} + \text{H}\_2\text{O} \longrightarrow \text{CO}\_2 + 2\text{H}^+ + 2\text{e}^-$$

In addition to this reaction, both enzymes show the ability to catalyze other reactions, although at lower activity. Reduction of CO2, following the reverse mechanism, is explicated by NiFe-CODH, while the ability of oxidizing H2 to protons has been reported for MoCu-CODH [3,4].

The latter belongs to the xanthine oxidase family of enzymes, whose members are usually characterised by a mononuclear redox-active site, LMoVIOS(OH), with a square-pyramidal coordination geometry [5]. The oxo (O2−) group is an apical ligand and the equatorial plane has a dithiolene ligand from a molybdopterin–cytosine dinucleotide (MCD) cofactor, a catalytically labile hydroxyl group (OH<sup>−</sup>) and a sulphide ion (S2<sup>−</sup>) [6]. MoCu-CODH presents a noncanonical binuclear active site, in which the sulphido ligand bridges to a second metal, –S–Cu–S-Cys and the equatorial hydroxyl group is deprotonated to form another oxo group. The copper ion is found in a +1 oxidation state and it is also coordinated by a weakly-bound water molecule [7].

The protein is a (αβγ)2 heterodimer. The active site is located in the large subunit (CoxL; 89 kDa), the medium subunit (CoxM; 30 kDa) contains a FAD cofactor, whereas the small one (CoxS; 18 kDa) possesses two [2Fe–2S] iron–sulfur clusters [7].

The protein resting state is characterised by a MoVI(=O)SCuI core. Binding of either CO or H2 is believed to occur at the CuI centre, allowing the proper placement and activation of the substrate [8,9]. The subsequent steps lead to the transfer of two electrons to the Mo ion. This process is favoured by the presence of a highly delocalised redox-active orbital over the Mo–μS–Cu moiety [10]. The catalytic activity of MoCu-CODH is also promoted by the presence of highly conserved amino acids in the metal second coordination sphere [11]. A glutamate residue (Glu763) positioned in proximity of the equatorial oxygen ligand of molybdenum is believed to promote deprotonation events [9,12]. In addition, a phenylalanine (Phe390), situated in a flexible loop in front of the copper centre, has been proposed to contribute to the proper orientation of the Cu-bound ligand [13].

Notwithstanding the fact that the general features of the catalytic reaction cycle of the enzyme has been clarified thanks to previous experimental and theoretical efforts, a detailed understanding of the reaction mechanisms for the oxidation of both CO and H2 has not yet been reached. As for H2, a mechanism for the dihydrogen molecule splitting was proposed by Wilcoxen and Hille (see Figure 1) [9]. According to electron paramagnetic resonance (EPR) spectroscopy studies, H2 is believed to coordinate Cu<sup>I</sup> in a side-on fashion. Once the hydrogen is bound, it would be activated by the metal and could easily be split by extraction of a proton thanks to a nearby base. Subsequently, a second proton would leave the active site and the two electrons would be transferred to the Mo centre. To date, no theoretical data have been reported in support of the experimental proposal. Only one previous theoretical work investigating the hydrogenase activity has been published so far [14]. In this study, it was unexpectedly found that the copper ion could not bind the H2 substrate. Therefore, the authors suggested that changes in the active site are needed, for example, by protonation events, before the catalytic cycle can start. In such context, it is interesting to notice that, to the best of our knowledge, side-on binding of H2 to molecular species containing a Cu(I) ion has been previously reported only in a couple of gas-phase case studies, that is, H2–CuCl [15] and H2–CuF [16]. Somehow, the interaction between H2 and the enzyme appears to find a closer parallel with the case of porous materials containing electron-rich copper ions, such as certain zeolites proposed for H2 storage applications [17–19] and Prussian-blue analogues [20].

**Figure 1.** Reaction mechanism for H2 activation proposed by Wilcoxen and Hille on the basis of experimental results [9].

In the present work we present a plausible mechanistic picture for the MoCu-CODH hydrogenase activity, taking into account the effects of the protein matrix by means of hybrid quantum mechanical/molecular mechanical (QM/MM) enzyme models.

#### **2. Results and Discussion**

#### *2.1. Study of H*<sup>2</sup> *Binding Modes to the Copper Centre*

We started with an investigation of the most plausible binding sites for the H2 substrate. Considering the resting state of the protein, two types of coordination to the copper centre were considered (**1HR** and **2HR** in Figure 2). We find that H2 can bind to the Cu ion both when the latter is bi-coordinated (μS–Cu–S-Cys388) and also when the Cu coordination sphere includes a weakly bound water molecule, as proposed in previous investigations of the enzyme resting state [7,13]. The hydrogen molecule is found to be slightly activated, with a H–H distance of 0.77 Å in both cases (the equilibrium distance of H2 at BP86/def2-TZVP level is 0.75 Å).

For both binding modes, we estimated the binding energy of H2 to Cu<sup>I</sup> by comparing the energy of the coordinated structure to the corresponding one in which H2 is not linked to copper but it is already present in the active site (**1HNoB** and **2HNoB** in Figure 3). The resulting QM/MM total energy differences were found to be +46.9 and +46.4 kJ/mol, in the absence (**1H**) or presence of the coordinated water molecule (**2H**), respectively (see Table 1).

**Table 1.** Relative energies (kJ/mol) calculated using the quantum mechanical/molecular mechanical (QM/MM) approach at B3LYP-D3(BJ)/def2-TZVPD level.


**Figure 2.** Binding modes of H2 to CuI characterised by means of QM/MM calculations considering different copper coordination modes (**1HR**, bi-coordinated; **2HR**, tri-coordinated), active site protonation states (**3HR**, protonated Cys388) and the stable intermediate **1H-FLP\_R**. All distances are in Å.

**Figure 3.** QM/MM optimised structures of intermediates where the H2 molecule is present in the active site but not bound to the Cu ion, considering different copper coordination modes (**1HNoB**, bi-coordinated; **2HNoB**, tri-coordinated), active-site protonation states (**3HNoB**, protonated Cys388). All distances are in Å.

Thus, formation of these reactants was found to be strongly unfavourable from an energetic point of view. Therefore, we contemplated the possibility that a sterically hindered Lewis acid–Lewis base pair (a so-called frustrated Lewis pair, FLP), may form in the active site, which would eventually lead to heterolytic H2 cleavage. Indeed, combination of Lewis acid and bases have been demonstrated to activate a wide range of small molecules [21]. Moreover, the plausible involvement of an FLPs in enzymatic catalysis has been recently proposed for the H2 splitting by [NiFe]-hydrogenase [22] and an FLP might actually be active also in the case of [FeFe]-hydrogenases [23–25].

We found a variant of **1HR** (**1H-FLP\_R** in Figure 2), in which H2 assumes a different position in the active site, such that the Mo–Oeq atom and the copper centre appear to be able to act as a FLP. The **1H-FLP\_R** reactant is characterised by a H2 molecule positioned in front of the copper atom, with long Cu–HH2 distances (2.80 and 2.93 Å) and a Oeq–HH2 distance of 2.33 Å. **1H-FLP\_R** was found to be 0.8 kJ/mol lower in energy than the **1HNoB** species in which the hydrogen molecule is separated from the bimetallic centre.

Based on previous literature, which proposed that the active site could be protonated prior to substrate binding [14,26], we also evaluated the influence of protonation on the coordination of H2 to the metal. Protonation of the bridging sulphide ion turned out not to favour dihydrogen binding, as no H2-bound geometry was found on the QM/MM potential energy surface. On the other hand, protonation of the sulphur atom of Cys388 favours coordination of H2 to Cu. The latter process would lead to the formation of a H2-coordinated geometry (**3HR** in Figure 2), in which dihydrogen is activated to a H–H bond of 0.78 Å and the computed binding energy is +10.9 kJ/mol.

It is noticeable that energy minimizations of the one-electron reduced counterparts of **1HR**, **2HR** and **3HR** lead to the detachment of the hydrogen molecule from the copper centre. This result is not compatible with the experimental outcomes that identified a H2-bound MoVCuI species by means of EPR experiments [9]. However, since one-electron reduced counterparts do not represent plausible intermediates in the catalytic cycle—there are no evidences that H2 binding is followed by Mo reduction, prior to substrate oxidation—we have not investigated this aspect any further. The aim of the work is focused on the mechanistic characteristics of the H2 oxidation reaction, the details of which are described in the subsequent sections.

#### *2.2. Exploring Basic Residues in the Active Site*

Next, we investigated which functional group may represent the base for the abstraction of the first proton from the activated hydrogen molecule. We considered the Glu763 residue, which has been suggested to take part into deprotonation events (*vide infra*) and the equatorial oxo ligand of molybdenum. As for the reactants, we considered not only **1H-FLP\_R** and **3HR** but also **1HR** and **2HR**. In fact, even though H2 binding in the latter two cases was computed to be disfavoured, one cannot completely rule out the hypothesis that such adducts have some minor role in dihydrogen oxidation catalysis, a possibility that would be enabled in case the subsequent reaction steps are associated with an overall smooth energy profile.

The two bases would lead to the formation of two copper hydrides (see Figure 4), namely either GluOOH–CuI H (**P1**) or Mo–OH–Cu<sup>I</sup> H (**P**). For the **1HR** and **2HR** states (see Figure 2), the oxygen of the glutamate residue is closest (the OGlu–HH2 distance is 1.87 Å and 1.89 Å for **1HR** and **2HR**), while the Mo–oxo ligand is 3.09 and 3.28 Å from the closest H atom of H2, respectively. However, for **3HR**, Oeq is the closest base (the O–H distance is 1.91 Å) while the oxygen of Glu763 is 3.83 Å away.

The search for the GluOOH–CuI H products was found to be challenging, as the glutamic acid sidechain in the guess geometries turned out to release a proton to reform a Cu-bound H2 along optimizations. Therefore, to optimize the intermediates **1HP1** and **2HP1**, (see Figure 5) we had to fix the OGlu–H distance; this was done in the hypothesis of the existence of a barrier for proton transfer, although so small that convergence on the desired minimum could be hampered by the choices made on the starting input structures to be optimized. The total energy differences of these structures, with respect to the corresponding reagents, were found to be 57.7 and 83.3 kJ/mol for **1H** and **2H**, respectively. In the case of **3H**, no GluOOH–Cu<sup>I</sup> H (**3HP1**) product was found.

**Figure 4.** Copper–hydride complexes formed by proton extraction by means of Glu763 (**P1**) or Mo–O*eq* (**P**).

**Figure 5.** QM/MM optimised structures of the copper hydride complexes formed by proton abstraction by means of Glu763 (**P1**). All distances are in Å.

The product formed by extraction of the proton by the equatorial ligand of Mo (**P**) was not found in the case of **2H**, whereas it corresponds to structures with a total energy difference of −2.9 and 28.0 kJ/mol with respect to the reagents for **1H** and **3H**, respectively (**1H-FLP\_P** in Figure 6 and **3HP** in Figure 7). **1H-FLP\_P** can also represent the product for the H2 splitting reaction starting from the **1H-FLP\_R** structure. In this case, the resulting Cu–hydride was found to be 44.4 kJ/mol less stable than the reagent.

**Figure 6.** QM/MM structures of intermediates and transition states (in square brackets) of species involved in the catalytic cycle **B** (see Figure 8). All distances are in Å.

**Figure 7.** QM/MM structures of intermediates and transition states (in square brackets) of species involved in the catalytic cycle **A** (see Figure 8). All distances are in Å.

#### *2.3. Plausible Activation Mechanisms for H*<sup>2</sup> *Splitting*

Based on the above results on the relative energies of the intermediates potentially involved in the catalysis, we proceeded with the computation of the activation barriers in order to estimate which reaction mechanism would represent the most energetically favourable path for the hydrogenase activity of MoCu-CODH.

Starting from the H2-bound conformation **1HR**, we have just shown that the equatorial Mo–oxo ligand represents the most favourable basic group for the first proton abstraction, leading to a rather stable product (**1H-FLP\_P**). However, despite extensive efforts, we were not able to locate any transition state for the reaction **1HR** → **1H-FLP\_P**. This issue is probably due to the large distance between the labile proton and the base. We also considered the **1HR** → **1HP1** reaction because, even if the GluOOH–Cu<sup>I</sup> H product is not as stable as the Mo–OH–CuI H product, the whole catalysis may be driven by subsequent exoenergetic reactions. However, again, no transition state was found.

On the other hand, the **1H-FLP\_R** species can be linked to the **1H-FLP\_P** product through a low-lying transition state (**1H-FLP\_TS**), reported in Figure 6. The reaction **1H-FLP\_R** → **1H-FLP\_P** was found to be endergonic (ΔE = 44.4 kJ/mol) with an associated activation barrier of 45.2 kJ/mol. The latter results support the hypothesis that Cu and the totally oxidised equatorial ligand of molybdenum could play the role of a FLP.

For the complex with a water molecule at the copper centre (**2HR**), we could not locate any transition state geometry that would allow the deprotonation of the hydrogen molecule by the anionic Glu763 residue.

Finally, we considered protonation of Cys388 and found that such protonation can be functional for the hydrogenase activity. In fact, in this case the activation energy for H2 cleavage involving the equatorial oxo-ligand is only 29.3 kJ/mol (**3HTS** in Figure 7), that is, 15.9 kJ/mol lower than for **1H-FLP**.

Independently of the protonation state of Cys388, our results support a mechanism in which the equatorial oxo group of Mo extracts the first proton. Such process is characterised by a rather low activation barrier that may allow a relatively rapid proton exchange, in accordance with experimental evidence [9]. However, differently from what was hypothesised in the literature [9], we propose that this rapid proton exchange is promoted by the oxo ligand on Mo, rather than by Glu763. Moreover, the equilibrium of this step favours the H2 reactant complex over copper hydride in both cases, as discussed by Wilcoxen and Hille in light of the pH independence of the reaction. Finally, the intermediates **3HP** or **1H-FLP\_P** should be converted to the fully reduced form of the enzyme by the transfer of the Cu-bound hydride to the (now protonated) Oeq, giving MoIVO(OH2)CuI [8,13,27] (**P2**; see Figure 8). In both cases, the resulting reduced intermediates were found to be very stable, −90.8 and −82.8 kJ/mol for **3HP2** and **1H-FLP\_P2**, respectively (see Table 1 and Figures 6 and 7). Moreover, the reaction barriers for the above two reactions are relatively low (28.4 and 34.7 kJ/mol, respectively). Finally, the reverse reactions are energetically impeded since the reverse activation barriers exceed 140 kJ/mol in both cases (TSs energies are and 147.2 and 161.9 kJ/mol respectively). Again, this is in good agreement with experiments, showing no production of H2 by the enzyme [9].

**Figure 8.** Proposed catalytic cycles for the oxidation of H2 by means of MoCu-CODH considering either the presence of protonated (**A**) and deprotonated (**B**) Cys388.

#### **3. Methods**

#### *3.1. The Protein*

All calculations were based on the crystal structure of MoCu-CODH hydrogenase in its oxidised form (PDB ID: 1N5W) [7]. Only the large subunit (CoxL) of one monomer, containing the active site, was considered in this study. The enzyme was set up in the same way as in our previous study [28]. The protonation state of all the residues was determined based on calculations with PROPKA [29] and on studies of the hydrogen-bond pattern, of the solvent accessibility and of the possible formation of ionic pairs. All Arg, Lys, Asp and Glu residues were assumed to be charged, with exception of Glu29 and Glu488 that were protonated on OE2, whereas Asp684 was protonated on OD1. Cysteine ligands coordinating to metals were deprotonated. Among the His residues, His61, 339, 766 and 793 were

protonated on the ND1 atom, His177, 178, 210, 213, 243, 700, 753, 754 and 788 were assumed to be protonated on NE2 atom, whereas the other His residues were assumed to be doubly protonated. The protein was solvated with water molecules, forming a sphere with a radius of 60 Å around the geometric centre of the protein. The added protons and water molecules were then optimised by a 1-ns simulated-annealing molecular-dynamics simulation, followed by energy minimization [28] (for details on the adopted force field, see the next subsection).

#### *3.2. QM/MM Calculations*

The QM/MM calculations were performed with the ComQum software [30,31]. According to this approach, the protein and the solvent are split into three subsystems—System 1 corresponds to the QM region and it was relaxed by QM methods. It consisted of the molybdenum ion and its first coordination sphere (two O2−, the bridging sulphide and the MCD cofactor, truncated to exclude the phosphate and cytosine moieties), the copper ion, the ligand molecules H2, H2O, as well as the sidechains of residues Cys388 and Glu763 (see Figure 9).

**Figure 9.** Composition of the QM system within the hybrid QM/MM model of the enzyme, in the form free from H2/H2O ligands at the Cu centre.

System 2 consists of all residues with any atom within 6 Å of any atom in System 1, whereas System 3 contains the remaining part of the protein and the water molecules. The latter two systems were kept fixed at the crystallographic coordinates in the QM/MM calculations.

In the QM calculations, System 1 was represented by a wavefunction whereas all the other atoms were represented by an array of partial point charges.

When there is a covalent bond between the QM and MM systems, the QM system was truncated using the hydrogen link-atom approach [32]. The QM system was capped with hydrogen atoms (hydrogen link atoms, HL), the position of which are linearly related to the corresponding carbon atoms (carbon link atoms, CL) in the full system [30]. All atoms were included in the point-charge model, except the CL atoms. ComQum employs a subtractive scheme with electrostatic embedding and van der Waals link-atom corrections [33]. The total QM/MM energy is calculated as

$$E\_{\rm QM/MM} = E\_{\rm QM1+ptch23}^{\rm HL} + E\_{\rm MM123,q\_1=0}^{\rm CL} - E\_{\rm MM1,q\_1=0}^{\rm HL} \tag{1}$$

where EHL QM1+ptch23 is the QM energy of System 1 truncated by HL atoms and embedded in the set of point charges modeling Systems 2 and 3. EHL MM1,*q*1=<sup>0</sup> is the MM energy of System 1, truncated by HL atoms, without any electrostatic interactions. ECL MM123,*q*1=<sup>0</sup> is the classical energy of the whole system, with CL atoms and with the charges in System 1 set to zero, to avoid double counting of the electrostatic interactions. The QM calculations were carried out using TURBOMOLE 7.1 software [34]. Geometry optimizations and TS searches (the latter based on minimizations with geometric constraints imposed step-wise to selected atoms, along the putative reactive path) were performed using the BP86 functional [35,36] in combination with def2-TZVP basis set [37]. All calculations included Grimme's

dispersion correction with Becke–Johnson damping (D3(BJ)) [38]. The resolution-of-identity technique was employed to accelerate the calculations [39].

The MM calculations were performed with the Amber software [40], using the Amber ff14SB force field for the protein [41] and the general Amber force field [42] with restrained electrostatic potential (RESP) charges [43] for H2 and MCD. The two Fe2S2 clusters were described with RESP charges and a non-bonded model (they are kept fixed in the calculations).

#### **4. Conclusions**

We have studied dihydrogen oxidation by MoCu-CODH using QM/MM calculations. Our results indicate that the equatorial oxo ligand of Mo is a better base than Glu763 during catalysis. Moreover, extraction of the first proton from H2 by means of this basic centre leads to the formation of a Mo-OH–CuI H hydride that allows for the stabilization of the copper hydride, otherwise known to be very unstable [44]. It is intriguing to notice that our proposal of a direct involvement of the Mo-bound oxo group in H2 splitting finds a conceptually similar case in a recently published mechanistic study on the [Fe]-hydrogenase. In the latter enzyme, a deprotonared OH group that belongs to the iron–guanylylpyridinol cofactor and that is in γ-position with respect to the metal-activated H2, was suggested to be involved in substrate splitting [45]. Such mechanistic picture on the [Fe]-hydrogenase was the result of crystallographic studies that took advantage of the fact that the latter enzyme exists in two forms—an open substrate-accessible form and a closed catalytically active form. This is not the case for MoCu-CODH; however, key details on the proton transfer events following H2 binding might still be gained at experimental level, most likely by means of application of high-sensitivity infrared spectroscopy techniques (see for example, Reference [46]).

In light of our results, two plausible mechanisms for the hydrogenase activity of the enzyme can be proposed, as reported in Figure 8. The first reactive channel (cycle **A** in Figure 8) depends on protonation of the sulphur atom of Cys388, which appears to favour the binding and activation of the hydrogen molecule. The second reactive channel involves a frustrated Lewis pair, formed by the Mo–O*eq* oxo group and the copper centre (cycle **B** in Figure 8). In this case, no binding of the hydrogen molecule to the Cu center is observed but once H2 enters into the active site, it can be split following a low-energy path. All in all, we cannot exclude that both pathways may be active in MoCu-CODH, thus enhancing the overall catalytic turnover. The two mechanisms for H2 oxidation are characterized by energy profiles that were found to be quite similar (see Figure 10) and although the FLP-based mechanism features a ~20 kJ/mol higher rate-determining barrier, it does not require any preliminary active-site protonation. However, the H<sup>+</sup> source for protonation of Cys388 might be the H2 molecule itself, as a result of a possible proton transfer following initial H2 splitting at the active site by means of the FLP mechanism. Alternatively, protonation of Cys388 may occur as a parallel, concerted process at the onset of H2 oxidation catalysis. Notably, the existence of a pair of distinct proton pathways has been recently suggested for [FeFe]-hydrogenases. There, a "regulatory" protonation of the [4Fe–4S] portion of the hexanuclear active site was proposed to occur in concomitance with a "catalytic" protonation step occurring at the di-iron portion of the same active site [47]. The results presented in the latter study and—hopefully—those discussed in the present paper are likely to provide useful information for future efforts to deepen insights into proton transfer mechanisms from and to the MoCu-CODH active site.

**Figure 10.** QM/MM energy profiles (in kJ/mol) for the oxidation of H2 in MoCu-CODH, considering either the presence of protonated (profile traced in green) or deprotonated (profile traced in blue) Cys388.

**Author Contributions:** Conceptualization, C.G. and U.R.; investigation and formal analysis, A.R., C.G. and U.R.; data curation, A.R.; writing—original draft preparation, A.R.; writing—review and editing, U.C., M.B., G.M., C.G. and U.R.

**Funding:** This investigation has been supported by grants from the Swedish research council (project 2018-05003). The computations were performed on computer resources provided by the Swedish National Infrastructure for Computing (SNIC) at Lunarc at Lund University.

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

#### **References**


© 2019 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/).
