The oxidative inactivation of FeFe hydrogenase reveals the flexibility of the H-cluster

The oxidative inactivation of FeFe hydrogenasereveals the flexibility of the H-clusterVincent Fourmond1*, Claudio Greco2, Kateryna Sybirna3, Carole Baffert1, Po-Hung Wang5,

Pierre Ezanno1, Marco Montefiori5, Maurizio Bruschi2, Isabelle Meynial-Salles4, Philippe Soucaille4,

Jochen Blumberger5, Herve Bottin3, Luca De Gioia6 and Christophe Leger1

Nature is a valuable source of inspiration in the design of catalysts, and various approaches are used to elucidate themechanism of hydrogenases, the enzymes that oxidize or produce H2. In FeFe hydrogenases, H2 oxidation occurs atthe H-cluster, and catalysis involves H2 binding on the vacant coordination site of an iron centre. Here, we show that thereversible oxidative inactivation of this enzyme results from the binding of H2 to coordination positions that are normallyblocked by intrinsic CO ligands. This flexibility of the coordination sphere around the reactive iron centre confers on theenzyme the ability to avoid harmful reactions under oxidizing conditions, including exposure to O2. The versatile chemistryof the diiron cluster in the natural system might inspire the design of novel synthetic catalysts for H2 oxidation.

1

Hydrogenases are enzymes that catalyse the reversible oxidation

2 of dihydrogen into protons and electrons. Their catalytic3 capacities have attracted considerable interest. It is believed4 that they could be used as H2 oxidation catalysts in fuel cells, or5 for H2 production by photosynthetic microorganisms. They are6 also a source of inspiration for the design of artificial catalysts1–3.7 Hydrogenases are divided into two classes based on the metal8 content of their active site. FeFe hydrogenases exhibit the highest9 turnover rates4. Their active site, the so-called H-cluster, is com-

10 posed of a standard [4Fe4S] cluster covalently attached by a cysteine11 residue to a [Fe2(CO)3(CN)2(dtma)] subsite (dtma¼ dithiomethyl-12 amine) (Fig. 1). The irons of this FeFe subsite are named proximal13 (Fep) or distal (Fed) according to the distance to the [4Fe4S] cluster.14 The hydrogenase from the green alga Chlamydomonas reinhardtii15 (Cr) has no other cofactor than the H-cluster. Others host a variable16 number of additional redox cofactors: two [4Fe4S] clusters in the17 case of the enzyme from Desulfovibrio desulfuricans (Dd), and18 three [4Fe4S] clusters and one [2Fe2S] cluster in the enzyme from19 Clostridium acetobutylicum (Ca).20 During catalysis, the active site binds H2 in the so-called Hox state21 (in which the FeFe subsite is in the electronic state Fe(II) Fe(I)), then22 loses one electron and one proton to give the Hred state. Hox is recov-23 ered upon transfer of one electron and one proton.24 The coordination around the distal iron has a ‘rotated’, square-25 pyramidal conformation, with a vacant coordination position,26 marked by an asterisk in Fig. 1, where H2 or the exogenous inhibitor27 CO bind2,5. According to density functional theory (DFT)Q2 calcu-28 lations, this conformation is locked by a hydrogen bond between29 the CN2 ligand on Fed and a nearby lysine residue6 that is essential30 for activity7. However, Roseboom et al. showed that all CO ligands31 coordinated to the distal iron in the CO-inhibited state could be32 exchanged for exogenous 13CO under illumination8. This obser-33 vation suggests that the intrinsic CO ligands on Fed are not static,34 but it is unclear whether this mobility plays a role during catalysis.

35It has been proposed that the bridging CO moves to a terminal pos-36ition in certain catalytic intermediates8,9.37By studying the oxidative inhibition of the enzyme10 using a38combination of experimental and theoretical techniques, we found39that the coordination sphere of the distal Fe is more flexible than40had been anticipated. This flexibility is functionally important, as41it prevents the oxidative destruction of the H-cluster.

42Results and discussion43FeFe hydrogenases inactivate both reversibly and irreversibly at44high potential. When an enzyme is adsorbed or attached11 to a45rotating disc electrode, the current response is proportional to46turnover frequency12. Changes in activity can therefore be47followed as changes in current, with a high temporal resolution.48Poising the electrode potential in a well-designed sequence of49potential steps gives a current response that can be interpreted to50learn about redox-driven inactivation/reactivation processes13. In51the experiment shown in Fig. 2, the hydrogenase from Cr was52attached to a rotating disc graphite electrode spun at a high rate53in a solution saturated with H2; stepping the electrode potential E54from 65 mV to 165 mV at t¼ 140 s results in an instant increase55in H2-oxidation current (red upward arrow in Fig. 1b) due to the56instant increase in driving force, followed by a slow decrease (red57downward arrow) that reveals the inactivation of the enzyme.58Stepping the potential back to 65 mV at t¼ 280 s results in an59instant decrease in current (green downward arrow), followed by60a slow recovery of activity (green upward arrow), which shows61that the inactivation is reversible.62However, a comparison of the initial and final current values63(purple arrows in Fig. 2) shows an overall decrease in current64during the course of the experiment. Because the covalent attach-65ment of the enzyme molecules on the electrode prevents film66loss11, this decrease reveals a slow irreversible inactivation. Such a67combination of reversible and irreversible inactivation was already

1Laboratoire de Bioenergetique et Ingenierie des Proteines, CNRS, Aix-Marseille University UMR 7281, 31 ch Joseph Aiguier, 13009 Marseille, France Q1,2Department of Environmental Science, Universita degli Studi di Milano-Bicocca, Milan 20126, Italy, 3iBiTec-S, SB2SM, LMB (UMR CNRS 8221), DSV, CEA,91191 Gif-sur-Yvette, France, 4Universite de Toulouse, INSA, UPS, INP, LISBP, INRA:UMR792,135 CNRS:UMR 5504, avenue de Rangueil, 31077 Toulouse,France, 5Department of Physics and Astronomy, Gower Street, London WC1E 6BT, UK, 6Department of Biotechnology and Biosciences, Universita degli Studidi Milano-Bicocca, Milan 20126, Italy. *e-mail: [email protected]

ARTICLESPUBLISHED ONLINE: XX XX 2014 | DOI: 10.1038/NCHEM.1892

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry 1

1 apparent in cyclic voltammetry experiments carried out with the2 enzyme from Dd10. Cyclic voltammograms for Cr and Ca hydroge-3 nases are shown in Supplementary Fig. 1: all FeFe hydrogenases4 behave in a similar way but the reversible inactivation of the clostri-5 dial enzyme is less pronounced, which makes quantitative analysis6 difficult (Supplementary Figs 1 and 2). This is the reason we7 focus here on chronoamperometric data obtained with Cr

8hydrogenase, but we show in Supplementary Fig. 1 that the voltam-9mograms obtained with Cr and Ca hydrogenases can be simulated

10using the mechanistic information gained from our study.

11Two inactive states are reversibly formed under oxidizing12conditions. Fitting a mono-exponential decay to transients such13as those in Fig. 2b did not yield satisfactory results14(Supplementary Fig. 3). Only bi-exponential decays could15convincingly reproduce the data. This shows that two inactive16species are reversibly formed. We call them I′ and I′′, and we call17A the active form of the enzyme:

I′ Nk′a

k′iAN

k′′i

k′′aI′′

18The first-order rate constants have a subscript ‘a’ or ‘i’ to designate19activation or inactivation reactions, respectively. Fitting kinetic20equations derived from the above model (Supplementary21Section 1.1) to transients such as those in Fig. 2b allows the22determination of the four rate constants of interconversion23between species A, I′ and I′′. In the fitting procedure, we24simultaneously adjusted 13 parameters: two sets of five rate25constants (one set of ka

′, ki′, ka

′′, ki′′ and one rate constant of

26irreversible inactivation at each electrode potential where27(in)activation occurs) and three values of the steady-state current28response for the fully active enzyme (one at each potential).29Figure 2b (fits) and c (residuals) shows that the fits are excellent.30Repeating such experiments for different values of the potential31steps and at different pH values and H2 concentrations gave the32dependence of the rate constants on these parameters. In the33following, we interpret these data to learn about the mechanism34of the inactivation process.

35Inactivation proceeds by a chemical reaction followed by36oxidation. We observed that the rate constants of formation of37both I′ and I′′ (ki

′ and ki′′) are independent of potential (blue and

38red Q3circles in Fig. 3a,b), irrespective of pH (Supplementary Fig. 8).39This indicates that the first step of the formation of I′ and I′′ is a40chemical process, not a redox step. These chemical steps are41indicated by red and blue horizontal arrows in Fig. 3c.42By contrast, the rate constants of reactivation of both I′ and I′′

43strongly depend on potential. As shown using brown and light44blue triangles in Fig. 3a and b and Supplementary Fig. 8, the45values of ka

′ and ka′′ are approximately proportional to

46exp(–FE/RT) at low potentials, that is, a factor of 10 over 60 mV, as47expected for a fast, one-electron redox reaction (F is the Faraday48constant, R the gas constant and T the absolute temperature). The49values plateau at high potentials (the slope becoming less than a50factor of two over 150 mV). These observations are very robust51when the pH is changed (Supplementary Fig. 8). This indicates52that each inactive species is present under two redox states53(Supplementary Section 1.3). We shall use the subscript ‘overox’54to denote the states that are more oxidized than Hox (Fig. 3c). The55Hoverox states reactivate in two ways: (1) a slow, direct reaction,56depicted by orange and green arrows in Fig. 3c Q4, for which the poten-57tial-independent rate constant ka

′direct or ka′′direct is given by the high

58potential plateaux in Fig. 3a,b; (2) a reduction followed by a fast59chemical reaction, indicated by the brown and blue arrows in60Fig. 3c, for which the potential-dependent rate constants are61termed ka

′redox and ka′′redox. The dependence on potential of the

62four rate constants can therefore be described by six parameters—63ki

′, ka′direct, ki

′′, ka′′direct, ka

′redox at E¼ 0 V and ka′′redox at E¼ 0 V

64(referred to as ka′redox@0 V and ka

′′redox@0 V)—the values of65which were obtained by fitting Supplementary equation (6) to the66data sets shown in Fig. 3a,b and Supplementary Fig. 8.

FepFed*

a

Fed

*

b

Figure 1 |Q12 a, The active site H-cluster of FeFe hydrogenases, and its

surroundings (adapted from pdb 3C8Y; ref. 50). The vacant coordination

position on the distal iron is marked by an asterisk. The phenylalanine

residue is discussed in the text. b, The same structure rotated by 908,showing the phenylalanine and the CO ligand of Fed using a space-

filling model.

b

2

4

6

8

10

i (μA

)

Data

Fit

c

–0.5

0.0

0.5

1.0

Δi (μA

)

0 100 200 300 400 500

t (s)

0 100 200 300 400 500

t (s)

0 100 200 300 400 500

t (s)

a

–0.1

0.0

0.1

E (

V v

ersu

s S

HE

)

Figure 2 | Chronoamperometric data obtained with a film of Cr

hydrogenase covalently attached to a rotating graphite electrode.

a, Sequence of potential steps applied to the electrode. b, Resulting current

(black), together with a fit to the kinetic model (dashed blue line, see text).

c, Residuals of the fit. Conditions: pH 7, 1 bar H2, 1 8C, electrode rotation rate

5,000 r.p.m. The capacitive current is much smaller than the faradaic

current, and need not be subtracted. Red and green arrows in b indicate the

instant changes in activity due to changes in driving force (vertical arrows)

and slow changes (bent arrows) due to inactivation. Purple arrows indicate

the amount of activity definitely lost during the experiment.

ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.1892

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry2

1 The formation of H′ox and H′′

ox from Hox is endergonic: ki′, ka

′

2 and ki′′ , ka

′′. Nevertheless, the inactivation proceeds quantitatively3 at high potential because overoxidation pulls the endergonic4 chemical step.5 We also examined the dependence on potential of the rate of irre-6 versible inactivation, which is responsible for the fact that not all7 activity is recovered after transient exposure to high potentials8 (purple arrows in Fig. 2b). The analysis of the data in9 Supplementary Fig. 4 shows that only Hox irreversibly inactivates;

10 this occurs by a slow one-electron oxidation. By contrast, reversibly11 inactivated species I′ and I′′ are protected from irreversible inacti-12 vation. Experiments in which the reversibly inactivated enzyme is

13exposed to O2 (Supplementary Section 1.4 and Supplementary14Figs 9 and 10) demonstrate that I′ and I′′ are also completely pro-15tected from O2-induced oxidative damage.

16Oxidative inactivation is triggered by H2 binding and coupled to17deprotonation. To determine the protonation states of the inactive18species, we studied the dependence on pH of the six rate constants19deduced from the analysis of data such as those in Fig. 3a,b. Of these20six rate constants, only ka

′redox@0 V and ka′′redox@0 V depend on

21pH, and it is as expected for a one-proton, one-electron reaction22(dashed line in Fig. 3e and Supplementary Fig. 6). This indicates23that the two Hoverox states have one fewer proton than Hox.24To identify the protons that are abstracted upon overoxidation,25we studied the influence of H2 concentration on the electrochemical26response, and found that the inactivation rate constants (ki

′, ki′′) are

27proportional to the partial pressure of H2 (Fig. 3d), whereas all other28rate constants are almost independent of H2 concentration29(Supplementary Fig. 5). This indicates that inactivation is initiated30by the binding of H2 to the Hox state. It follows that the deprotona-31tions coupled to the oxidation to H′

overox and H′′overox reveal the

32formation of hydrides from bound H2.33With the electrochemical data showing unambiguously that inac-34tivation is triggered by H2 binding, the question arises as to where35H2 binds and how this inactivates the enzyme. We answer this ques-36tion in the following using a combination of molecular dynamics37(MD), density DFT calculations and site-directed mutagenesis.

3

4

5

6

d(F

e d –

δ2 C

) (Å

)0 2 4 6 8 10

Time (ns)

MD

Crystal

a

b

Fep

Fed

CO

Phe296

δ2C

Figure 4 | Thermal fluctuations of Dd FeFe hydrogenase Phe296 probed

by MD. a, Snapshot of a MD run of solvated hydrogenase (in red) is

superimposed on the corresponding crystal structure (pdb code 1HFE, in

blue). The distance between the distal Fe atom of the H-cluster (Fed) and

the d2C atom of Phe296 is indicated by a dashed line Q14. b, Thermal

fluctuations of this distance as a function of simulation time (red line). The

dashed blue horizontal line corresponds to the distance in the crystal

structure. The snapshot in a was taken at t¼ 7.8 ns. At this conformation

the isomerization of the CO ligand is sterically feasible. The backbone atom

root-mean-square deviation with relation to the X-ray structure is 1.8 Å.

A

10–2

10–1

10–1

1

10

Rat

e co

nsta

nt (

s–1 )

0.0 0.1 0.2

E (V versus SHE)

0.0 0.1 0.2

E (V versus SHE)

A

a

10–2

10–1

1

10

Rat

e co

nsta

nt (

s–1 )

b

1

10

6 7 8

pH

karedox @ 0 V

karedox @ 0 V

0.00

0.05

Rat

e co

nsta

nt (

s–1 )

Rat

e co

nsta

nt (

s–1 )

0.0 0.5 1.0

pH2 (bar)

d e

c

I´ I˝

I´ I˝

A

HoxHox Hox

Hoverox HoveroxHoverox

+H2, k i +H2, k i

–H+

E 0 E0

karedox ka

redox

–e––e–

–H+

–e–ka

direct kadirect

ki

ki

ka

ki

ki

ka

Figure 3 | Rate constants for the inactivation/reactivation processes

determined from electrochemical experiments. a,b, Dependence on

potential of the rate constants of interconversion between the active A

species and the two inactive species (I′ and I′ ′): A↔ I′ (a) and A ↔ I′ ′ (b)

(1 bar H2, pH 7, T¼ 1 8C). Dashed lines are fits to Supplementary equation

(6). c, Kinetic model that explains the dependence on potential, pH and H2

concentration of the rate constants. H′overox and H′ ′

overox species reactivateQ13 in

two ways: (1) through a direct chemical reactivationQ13 (indicated by green and

yellow arrows, rate constants ka′direct and ka

′ ′direct), or (2) first through a

reduction to the H′ox and H′ ′

ox species and then a chemical reactivation

(indicated by brown and light blue arrows, rate constants ka′redox and

ka′ ′redox). d, Dependence of inactivation rate constants (ki

′ and ki′ ′) on H2

partial pressure at pH 7. e, pH dependence of ka′redox and ka

′ ′redox at E¼0 V,

1 bar H2. The dashed line has a slope of one decade per pH unit.

NATURE CHEMISTRY DOI: 10.1038/NCHEM.1892 ARTICLES


1 Mobility of the CO ligands around the distal iron. DFT2 computations show that no stable H2 adduct is formed upon H23 approach to any iron of the cubane subcluster (Supplementary4 Section 3.1) and only one coordination site is vacant in the5 natural H-cluster (on Fed), so the non-productive binding of H26 should be dependent upon prior displacement of one intrinsic CO7 ligand to the normal site of H2 binding (the CN2 ligand on Fed is8 locked by hydrogen-bond interactions6). From the crystal9 structure of the Dd enzyme, any movement of the CO ligands on

10 Fed seems unlikely as the side chain of phenylalanine 296 (F23411 in Cr) obstructs the path between equatorial and axial positions12 (Fig. 1). We therefore probed the flexibility of F296 and of the H-13 cluster using MD simulations starting from the structure of the14 Dd enzyme. The distance fluctuations between Fed and the closest15 C-atom of F296 (d2C) at T¼ 300 K are shown in Fig. 4b.16 Whereas the mean distance is close to the value found in the X-17 ray structure, the distance fluctuations are rather large even on the18 nanosecond timescale of the present simulations. The largest19 distance we have observed within 10 ns is 6.1 Å. A snapshot of20 this configuration is shown in Fig. 4a. In this configuration, F29621 is sufficiently far from Fed that CO-ligand isomerization becomes22 sterically feasible. Figure 4a also highlights that the distance23 fluctuations are mainly due to fluctuations of the protein24 backbone. The present MD simulation unambiguously shows that

25the conformational flexibility of the active site pocket is26sufficiently large to allow the rearrangement of the CO ligand.27We carried out DFT calculations on large models of the active28site (Supplementary Section 3) to compute the energy difference29between the H-cluster as observed in the X-ray structure of the30Hox state and corresponding isomers differing in the orientation31of the CO ligands coordinated to Fed (Fig. 5). Isomers featuring a32vacant coordination position either between the two iron atoms33(1) or in a pseudo-equatorial position (2) both correspond to34energy-minimum structures. They are respectively 2.4 and358.5 kcal mol21 higher in energy than Hox, indicating that H-36cluster forms featuring CO ligands in unusual positions correspond37to minor species (1% and 1025%, respectively), consistent with the38fact that these species had not been suspected from previous39Fourier-transform infrared (FTIR) Q5or X-ray investigations.

40Binding of H2 to minor vacant sites. Having established that 1 and412 correspond to minor H-cluster isomers, we examined the42energetics of H2 binding (Fig. 6). The computed reaction energies43for H2 binding to 1, 2 and Hox models are similar (binding44energies are within 4 kcal mol21 of each other), indicating that all45isomers can bind H2; the calculated structures of H2-bound states46are shown in Fig. 6. Dihydrogen is h2-coordinated to the47proximal atom in 1H2, and to the distal Fe in 2H2. The H–H

Fe Fe

S S

CO

CN

NCOC

CysS

NH

CO

H

Fe Fe

S S

CO

CN

NCOC

CysS

NH

CO

H2

Fe Fe

S S

CO

CN

NCOC

CysS

NH

CO

*

Fe Fe

S S

H

CN

NCOC

CysS

NH

CO

C

Fe Fe

S S

H2

CN

NCOC

CysS

NH

CO

C

Fe Fe

S S

*

CN

NCOC

CysS

NH

CO

C

Fe Fe

S S

CO

CN

NCOC

CysS

NH

*C

I' I''

A

Very slow

–H+

–e–

–H+

–e––H

+

–e–

1H

1H2

1

2H

2H2

2

+H2 +H2

FastFast

+H+

+e–+H2

+H+ , +e–

O

Fe Fe

S S

CO

CN

NCOC

CysS

NH

H2

CO

Fe Fe

S S

CO

CN

NCOC

CysS

NH

H

CO

O

O

O

Catalytic cycle

Figure 5 | States of the active sites involved in catalysis (within the green triangle) and in high-potential inactivation. The vacant coordination position is

indicated by a red asterisk. Species 1H and 2H correspond to H′overox and H′ ′

overox and 1H2 and 2H2 to H′ox and H′ ′

ox. Grey frames delimit the I′ and I′ ′ inactive

species. Assignment of 1, 1H2 and 1H to I′ (rather than I′ ′) is arbitrary, as there is no way to tell which is which. Calculated structures of 1H2 and 2H2 are

shown in Fig. 6.



1 bond is slightly elongated, suggesting activation of the H2 molecule2 (1.036 Å in 1H2 and 0.882 Å in 2H2, compared to 0.740 Å in3 molecular H2, Supplementary Table 2). Note that 1, 2 and Hox4 are five coordinated species that can interconvert, but the binding5 of H2 prevents isomerization between 1H2, 2H2 and the normal6 H2-bound cluster (Supplementary Section 3.2).7 It was recently shown14 that H2 binding to the Hox form of the8 H-cluster triggers, under catalytic conditions, the one-electron9 oxidation of the cluster, which is expected to be facile if it is

10 coupled to concomitant heterolytic cleavage of H2 by the N atom11 of dtma15–17. By contrast, the oxidation of 1H2 and 2H2 to 1H12 and 2H should be difficult because no suitable base residue is13 located in the proximity of the H2 ligands. As a consequence, we14 expect that the one-electron oxidation of 1H2 and 2H2 can only15 take place under conditions that are more oxidizing than during16 normal turnover. For the same reason, Hþ abstractions from17 (1H2)þ and (2H2)þ (and also from the corresponding hydride18 species 1H and 2H formed after abstraction of one proton from19 H2) are also expected to be slow, resulting in a turnover that is so20 slow that the fully oxidized enzyme essentially appears inactive.21 Although 1 and 2 are never formed quantitatively, the inactive22 states accumulate significantly when hydrogen binding and oxi-23 dation pull the reaction towards the formation of 1H and 2H.

24 Phenylalanine 234 to tyrosine mutation abolishes reversible25 inactivation. In an attempt to hinder the movement of CO26 around the distal iron, we examined the effect of replacing27 phenylalanine 234 in Cr hydrogenase (F296 in the enzyme from28 Dd; Fig. 4) with a tyrosine. The mutant retains 20% of the wild-29 type (WT)Q6 activity and gives a very similar electrochemical30 response in the low potential region, but cyclic voltammograms of31 F234Y display a much higher proportion of irreversible32 inactivation at high potential and almost no reactivation on the33 return sweep (hence, no reversible inactivation; Supplementary34 Fig. 7). Analysis of chronoamperometric experiments similar to35 those in Fig. 2 returned inactivation rate constants (ki

′ and ki′′)

36 �20 times smaller than those of the WT enzyme (Supplementary37 Fig. 7). This supports the above conclusion that H2-dependent38 reversible inactivation protects the WT enzyme against irreversible39 oxidative damage.

40 Discussion41 All FeFe hydrogenases studied so far by protein film voltammetry42 (Cr, Ca and Dd) exhibit the same behaviour under oxidative con-43 ditions: a biphasic inactivation at high potential followed by44 partial reactivation upon reduction (this work and ref. 10). The inac-45 tive species are probably not the O2-protected Hox inactQ7 state that is46 present in aerobically purified samples of Dd FeFe hydrogenase18–20,47 as this state has never been observed with Ca and Cr hydrogenases.48 We also consider it unlikely that oxidative inactivation would stem49 from the binding of the N atom of the dtma bridge to Fed (as

50suggested in ref. 21). Indeed, the formation of the two inactive51states is coupled to the abstraction of two different protons52(Fig. 3e), but the H-cluster in the Hox state has only one proton53(on the dtma bridge), which is not easily released, as discussed in54Supplementary Section 3.3. Moreover, theoretical calculations55suggest that the protein matrix prevents the binding of the bridging56N (ref. 22).57From the observed linear dependence of inactivation rates on H258pressure (Fig. 3d), and calculations, we concluded that reversible59inactivation results from the non-productive binding of H2 to60minor isomers of the Hox state, in which one or both CO ligands61bound to Fed have moved, leaving to a vacant coordination position62either in an axial (1 in Fig. 5) or equatorial (2) position. In the63crystal structure of the enzyme from Dd, phenylalanine 296 seems64too close to Fed for CO to hop from an equatorial position to an65axial position, but MD simulations show that the Fed phenylalanine66distance fluctuates enough to permit movement (Fig. 4). Replacing67this amino acid (F234 in the enzyme from Cr) with a tyrosine68abolishes reversible inactivation (Supplementary Fig. 7), suggesting69that it prevents isomerization. Although this could simply result70from steric factors, the tyrosine OH group is probably sufficiently71close to the N of the peptide bond of alanine 78 or isoleucine 8272(Cr numbering; Supplementary Fig. 11) to be locked by hydrogen73bonding. CO mobility is supported by the earlier observation that74all CO ligands on the Fed of CO-inhibited Dd hydrogenase can be75replaced with 13CO under illumination8. The very small proportion76of Hox isomers (lower than 1%) is the reason these species evade77detection by spectroscopic techniques and crystallography.78Binding of H2 to one of the minor isomers results in a H2 adduct79(Fig. 6) that can be oxidized to form a stable hydride. The latter reac-80tivates either by reversing the inactivation steps (reprotonation,81reduction and H2 release) or by slowly closing an alternative cataly-82tic cycle (further proton and electron abstraction). The absence of a83dedicated base should not totally prevent deprotonation of the inac-84tive states: the inactivation processes are �105 times slower than85turnover, so any weak base—such as the thiols from dtma—could86transfer protons on such a slow timescale23–26. In fact, slow proton87transfer is probably one of the reasons these species are essentially88inactive, unable to dispose of bound H2 and the resulting hydride89at a rate-matching normal turnover.90A fraction of the enzyme also irreversibly inactivates at high91potential. The dependence on potential of the rate of this process92(Supplementary Fig. 4) shows that the irreversible inactivation93begins with a very slow oxidation of Hox. This should lead to a difer-94rous state, which we refer to as Hoverox. It may be that Hoverox, which95is very electrophilic and has a vacant coordination position, is so96reactive that it binds any nucleophilic group present, resulting in97disruption of the H-cluster. This is probably the reason Silakov98and co-workers observed oxidative damage in their attempts to99oxidize the Hox state of Cr hydrogenase in the absence of H2

100(ref. 27). By contrast, 1H and 2H are protected from oxidative101damage (anaerobic or aerobic, Supplementary Sections 1.2 and1021.4), probably because Fed has a saturated coordination sphere in1031H and 2H and is thus much less prone to nucleophilic attack104than the 5-coordinated Fed of the active species. That equilibrium105redox titrations cannot be carried out in the presence of H2 is106likely to preclude the spectroelectrochemical detection of any of107the inactive species shown in Fig. 5.108Isomerization of the distal coordination sphere of the H-cluster109and the resulting reversible inactivation probably play an important110functional role, because reversibly inactivated forms are protected111from oxidative damage (aerobic or anaerobic). It is remarkable112that the phenylalanine residue shown in Fig. 4, the substitution of113which makes the enzyme prone to oxidative damage, is highly con-114served: it is present in more than 90% of the 125 sequences of the115genes categorized as FeFe hydrogenases in ref. 28 (it is replaced

1H2 2H2

Figure 6 | Structures of 1H2 and 2H2, as predicted by DFT calculations.

Colour code: orange, Fe; white, C or H; yellow, S; red, O; blue, N. The

computed structure of Hox–H2 is shown in Supplementary Fig. 15.



1 with a tyrosine in the remaining 10%). It is possible that this residue2 helps maintain the delicate balance between excessive flexibility of3 the active site, which would be detrimental for stability, and exces-4 sive rigidity, which would prevent the reversible inactivation that5 protects against oxidative damage. This relation between flexibility6 and protection is reminiscent of the recent proposal that a change7 in coordination of an FeS cluster is linked to resistance to O2 in8 certain NiFe hydrogenases29–31. It may be that changes in coordi-9 nation also induce resistance against chemical stress in other metal-

10 loenzymes, but only using a unique combination of techniques was11 it possible here to demonstrate the functional relevance of12 such flexibility.13 The catalytic cycle of FeFe hydrogenases was once proposed to14 involve a bridging hydride32,33, as observed in synthetic complexes2,15 before this hypothesis was ruled out27. We found that H2 binding to16 a site other than the axial position on Fed does occur in the natural17 system, where it is essentially non-productive because of the lack of18 an efficient base. Therefore, binding in a bridging position may be19 catalytically relevant for synthetic complexes, provided a proton20 relay is engineered accordingly3.

21 Methods22 Samples of WT C. acetobutylicum and C. reinhardtii FeFe hydrogenases were23 prepared as described in refs 34–37.

24 Mutagenesis and purification of CrHydA1. Site-directed mutagenesis was25 performed on Chlamydomonas reinhardtii hydA1 gene cloned into the pBBR-26 hydA1N vector34 using the QuikChange Site-Directed Mutagenesis Kit (Agilent27 Technologies). The mutagenesis primers used for creating mutant F234Y were28 F234Yup 5′-ccggcgtgctgtacggcaccaccgg-3′ and F234Ydown 5′-29 ccggtggtgccgtacagcacgccgg-3′ (with nucleotide triplets coding a mutated amino acid30 shown in bold). The recombinant plasmid was introduced and expressed in31 Shewanella oneidensis AS52 strain as described in ref. 34. An average amount of32 0.4 mg of purified active protein per litre of culture was obtained for both native and33 mutant enzymes. Purification and activity assays were performed in a glovebox filled34 with N2.

35 Hydrogen evolution assay of Cr FeFe hydrogenases. Hydrogen evolution was36 assayed by amperometry using a modified Clark-type electrode (Hansatech)37 reversely polarized (þ0.7 V; Pt versus Ag/AgCl), and standardized using an38 aliquot of H2-saturated solution. Turnover rates of 810 and 160 mmol H239 min21 mg21 protein were measured at pH 6.7, T¼ 20 8C, for the WT enzyme and40 the F234Y mutant, respectively, using reduced methyl viologen (5 mM) as41 electron donor.

42 Protein film electrochemistry experiments. Experiments were carried out in a43 glovebox filled with N2, using the electrochemical set-up and equipment as44 described in ref. 38. Reference 11 describes the procedure for attaching the FeFe45 hydrogenases we study to rotating disc pyrolytic graphite edge electrodes (RDPGEs).46 The two-compartment electrochemical cell was kept at the desired temperature47 using a water circulation system. All potentials are quoted versus the standard48 hydrogen electrode (SHE). The concentration of H2 was varied by mixing pure H249 with argon in an Aalborg gas proportioner. We analysed and fitted the data using in-50 house programs SOAS39 and QSoas. The former is available free of charge at51 http://bip.cnrs-mrs.fr/bip06/software.html. It is being replaced by an entirely new,52 powerful, open-source program called QSoas, which will become available soon.53 Both programs embed the ODRPACK software for nonlinear54 least-squares regressions40.

55 Molecular dynamics simulations. Simulations were carried out starting from the56 crystal structure of Dd FeFe hydrogenase, pdb code 1HFE (L and S subunits only),57 because among the available structures of holoenzymes, it is the closest in size to Cr58 hydrogenase. We added the bridging CO that is missing in this crystal structure. The59 protein was described with the GROMOS96 43a1 united atom force field41. Point60 charges for the H-cluster atoms were parametrized for the Hox state using DFT61 calculations (Supplementary Table 1). The protein was solvated with SPC/E water42

62 and protonation states were chosen according to pH 7. After energy minimization,63 harmonic restraints were applied to all protein and metal cluster atoms, and the64 solvent equilibrated for 1 ns in the NPT ensemble at 300 K and 1 bar. The harmonic65 restraints were then released and the protein equilibrated for �1 ns. The following66 9 ns of dynamics were used to monitor the dynamics of the H-cluster and67 surrounding protein residues. All simulations were carried out with the GROMACS68 package, version 4.5.5 (ref. 43). The full simulation protocol is given in the69 Supplementary Information.

70 Density functional theory. Optimizations were carried out with the TURBOMOLE71 program suite44, using the BP86/def2-SVP level of theory45–47. Reported energies are

72electronic energies. The antiferromagnetic coupling that characterizes the [4Fe4S]73cubane included in the H-cluster was treated using the broken symmetry (BS)74approach48 using a recently developed technique for convergence control in BS75self-consistent field calculations49. Details on the atomic composition of the models76and on the geometric constraints used to take into account the effects of the77protein matrix surrounding the H-cluster are discussed in detail in the78Supplementary Information.

79Received 15 April 2013; accepted 11 February 2014;80published online XX XX 2014

81References821. Carroll, M. E., Barton, B. E., Rauchfuss, T. B. & Carroll, P. J. Synthetic models for83the active site of the [FeFe]-hydrogenase: catalytic proton reduction and the84structure of the doubly protonated intermediate. J. Am. Chem. Soc. 134,8518843–18852 (2012).862. Tard, C. & Pickett, C. J. Structural and functional analogues of the active sites of87the [Fe]-, [NiFe]-, and [FeFe]-hydrogenases. Chem. Rev. 109, 2245–2274 (2009).883. Liu, T., Dubois, D. L. & Bullock, R. M. An iron complex with pendent amines89as a molecular electrocatalyst for oxidation of hydrogen. Nature Chem. 5,90228–233 (2013).914. Fontecilla-Camps, J. C., Volbeda, A., Cavazza, C. & Nicolet, Y.92Structure/function relationships of [NiFe]- and [FeFe]-hydrogenases. Chem.93Rev. 107, 4273–4303 (2007).945. Hsieh, C-H. et al. Structural and spectroscopic features of mixed valent FeIIFeI

95complexes and factors related to the rotated configuration of diiron hydrogenase.96J. Am. Chem. Soc. 134, 13089–13102 (2012).976. Greco, C., Bruschi, M., Fantucci, P., Ryde, U. & De Gioia, L. Isocyanide in98biochemistry? A theoretical investigation of the electronic effects and energetics99of cyanide ligand protonation in [FeFe]-hydrogenases. Chem. Eur. J. 17,

1001954–1965 (2011).1017. Knorzer, P. et al. Importance of the protein framework for catalytic activity of102[FeFe]-hydrogenases. J. Biol. Chem. 287, 1489–1499 (2012).1038. Roseboom, W., De Lacey, A. L., Fernandez, V. M., Hatchikian, E. C. &104Albracht, S. P. J. The active site of the [FeFe]-hydrogenase from Desulfovibrio105desulfuricans. II. Redox properties, light sensitivity and CO-ligand exchange as106observed by infrared spectroscopy. J. Biol. Inorg. Chem. 11, 102–118 (2006).1079. De Lacey, A. L., Fernandez, V. M., Rousset, M. & Cammack, R. Activation and108inactivation of hydrogenase function and the catalytic cycle:109spectroelectrochemical studies. Chem. Rev. 107, 4304–4330 (2007).11010. Parkin, A., Cavazza, C., Fontecilla-Camps, J. C. & Armstrong, F. A.111Electrochemical investigations of the interconversions between catalytic and112inhibited states of the [FeFe]-hydrogenase from Desulfovibrio desulfuricans.113J. Am. Chem. Soc. 128, 16808–16815 (2006).11411. Baffert, C. et al. Covalent attachment of FeFe hydrogenases to carbon electrodes115for direct electron transfer. Anal. Chem. 84, 7999–8005 (2012).11612. Leger, C. & Bertrand, P. Direct electrochemistry of redox enzymes as a tool for117mechanistic studies. Chem. Rev. 108, 2379–2438 (2008).11813. Fourmond, V., Infossi, P., Giudici-Orticoni, M-T., Bertrand, P. & Leger, C. ‘Two-119step’ chronoamperometric method for studying the anaerobic inactivation of an120oxygen tolerant NiFe hydrogenase. J. Am. Chem. Soc. 132, 4848–4857 (2010).12114. Greco, C., Bruschi, M., Fantucci, P., Ryde, U. & De Gioia, L. Mechanistic and122physiological implications of the interplay among iron–sulfur clusters in123[FeFe]-hydrogenases. A QM/MM perspective. J. Am. Chem. Soc. 133,12418742–18749 (2011).12515. Barton, B. E., Olsen, M. T. & Rauchfuss, T. B. Aza- and oxadithiolates are126probable proton relays in functional models for the [FeFe]-hydrogenases. J. Am.127Chem. Soc. 130, 16834–16835 (2008).12816. Camara, J. M. & Rauchfuss, T. B. Combining acid–base, redox and substrate129binding functionalities to give a complete model for the [FeFe]-hydrogenase.130Nature Chem. 4, 26–30 (2011).13117. Camara, J. M. & Rauchfuss, T. B. Mild redox complementation enables132H2 activation by [FeFe]-hydrogenase models. J. Am. Chem. Soc. 133,1338098–8101 (2011).13418. Fernandez, V. M., Aguirre, R. & Hatchikian, E. C. Reductive activation and redox135properties of hydrogenase from Desulfovibrio gigas. Biochim. Biophys. Acta 790,1361–7 (1984).13719. Dijk, C. v., Berkel-Arts, A. v. & Veeger, C. The effect of re-oxidation on the138reduced hydrogenase of Desulfovibrio vulgaris strain Hildenborough and its139oxygen stability. FEBS Lett. 156, 340–344 (1983).14020. Fernandez, V. M., Hatchikian, E. C., Patil, D. S. & Cammack, R. ESR-detectable141nickel and iron-sulphur centres in relation to the reversible activation of142Desulfovibrio gigas hydrogenase. Biochim. Biophys. Acta 883, 145–154 (1986).14321. Olsen, M. T., Rauchfuss, T. B. & Wilson, S. R. Role of the azadithiolate cofactor144in models for [FeFe]-hydrogenase: novel structures and catalytic implications.145J. Am. Chem. Soc. 132, 17733–17740 (2010).14622. Miyake, T. et al. Does the environment around the H-cluster allow coordination147of the pendant amine to the catalytic iron center in [FeFe] hydrogenases?148Answers from theory. J. Biol. Inorg. Chem. 18, 693–700 (2013).



1 23. Zaffaroni, R., Rauchfuss, T. B., Fuller, A., De Gioia, L. & Zampella, G.2 Contrasting protonation behavior of diphosphido vs dithiolato diiron(I)3 carbonyl complexes. Organometallics 32, 232–238 (2012).4 24. Apfel, U.-P. et al. Models for the active site in [FeFe] hydrogenase with iron-5 bound ligands derived from bis-, tris-, and tetrakis(mercaptomethyl)silanes.6 Inorg. Chem. 49, 10117–10132 (2010).7 25. Dong, W. et al. An insight into the protonation property of a diiron azadithiolate8 complex pertinent to the active site of Fe-only hydrogenases. Chem. Commun.9 305–307 (2006).

10 26. Ezzaher, S., Gogoll, A., Bruhn, C. & Ott, S. Directing protonation in [FeFe]11 hydrogenase active site models by modifications in their second coordination12 sphere. Chem. Commun. 46, 5775–5777 (2010).13 27. Silakov, A., Kamp, C., Reijerse, E., Happe, T. & Lubitz, W.14 Spectroelectrochemical characterization of the active site of the [FeFe]15 hydrogenase HydA1 from Chlamydomonas reinhardtii. Biochemistry 48,16 7780–7786 (2009).17 28. Vignais, P. M. & Billoud, B. Occurrence, classification, and biological function of18 hydrogenases: an overview. Chem. Rev. 107, 4206–4272 (2007).19 29. Fritsch, J. et al. The crystal structure of an oxygen-tolerant hydrogenase uncovers20 a novel iron–sulphur centre. Nature 479, 249–252 (2011).21 30. Shomura, Y., Yoon, K-S., Nishihara, H. & Higuchi, Y. Structural basis for a [4Fe–22 3S] cluster in the oxygen-tolerant membrane-bound [NiFe]-hydrogenase.23 Nature 479, 253–256 (2011).24 31. Mouesca, J-M., Fontecilla-Camps, J. C. & Amara, P. The structural plasticity of25 the proximal [4Fe3S] cluster is responsible for the O2 tolerance of membrane-26 bound [NiFe] hydrogenases. Angew. Chem. Int. Ed. 52, 2002–2006 (2013).27 32. Bruschi, M., Fantucci, P. & De Gioia, L. DFT investigation of structural,28 electronic, and catalytic properties of diiron complexes related to the [2Fe]H29 subcluster of Fe-only hydrogenases. Inorg. Chem. 41, 1421–1429 (2002).30 33. Bruschi, M., Fantucci, P. & De Gioia, L. Density functional theory investigation31 of the active site of [Fe]-hydrogenases: effects of redox state and ligand32 characteristics on structural, electronic, and reactivity properties of complexes33 related to the [2Fe]H subcluster. Inorg. Chem. 42, 4773–4781 (2003).34 34. Sybirna, K. et al. Shewanella oneidensis: a new and efficient system for expression35 and maturation of heterologous [Fe–Fe] hydrogenase from Chlamydomonas36 reinhardtii. BMC Biotechnol. 8, 1–8 (2008).37 35. Baffert, C. et al. CO disrupts the reduced H-cluster of FeFe hydrogenase. A38 combined DFT and protein film voltammetry study. J. Am. Chem. Soc. 133,39 2096–2099 (2011).40 36. Girbal, L. et al. Homologous and heterologous overexpression in Clostridium41 acetobutylicum and characterization of purified clostridial and algal Fe-only42 hydrogenases with high specific activities. Appl. Environ. Microbiol. 71,43 2777–2781 (2005).44 37. Lautier, T. et al. The quest for a functional substrate access tunnel in FeFe45 hydrogenase. Faraday Discuss. 148, 385–407 (2011).46 38. Liebgott, P-P. et al. Relating diffusion along the substrate tunnel and oxygen47 sensitivity in hydrogenase. Nature Chem. Biol. 6, 63–70 (2010).48 39. Fourmond, V. et al. SOAS: a free program to analyze electrochemical data and49 other one-dimensional signals. Bioelectrochemistry 76, 141–147 (2009).50 40. Boggs, P. T., Donaldson, J. R., h. Byrd, R. & Schnabel, R. B. Algorithm 676:51 ODRPACK: software for weighted orthogonal distance regression. ACM Trans.52 Math. Softw. 15, 348–364 (1989).Q8

5341. Scott, W. R. P. et al. The GROMOS biomolecular simulation program package.54J. Phys. Chem. A 103, 3596–3607 (1999).5542. Berendsen, H. J. C., Grigera, J. R. & Straatsma, T. P. The missing term in effective56pair potentials. J. Phys. Chem. 91, 6269–6271 (1987).5743. Hess, B., Kutzner, C., van der Spoel, D. & Lindahl, E. GROMACS 4: algorithms58for highly efficient, load-balanced, and scalable molecular simulation. J. Chem.59Theory Comput. 4, 435–447 (2008).6044. Ahlrichs, R., Bar, M., Haser, M., Horn, H. & Kolmel, C. Electronic structure61calculations on workstation computers: the program system TURBOMOLE.62Chem. Phys. Lett. 162, 165 (1989). Q96345. Becke, A. D. Density-functional exchange-energy approximation with correct64asymptotic behavior. Phys. Rev. A 38, 3098 (1988). Q96546. Perdew, J. Density-functional approximation for the correlation energy of the66inhomogeneous electron gas. Phys. Rev. B 33, 8822 (1986). Q96747. Weigend, F. & Ahlrichs, R., Balanced basis sets of split valence, triple zeta valence68and quadruple zeta valence quality for H to Rn: design and assessment of69accuracy. Phys. Chem. Chem. Phys. 7, 3297–3305 (2005).7048. Noodleman, L. & Norman, J. G. Electronic-structure, optical-spectra, and71photochemistry of Ag-2 and Cu-2. Abstr. Pap. Am. Chem. S. Q107249. Greco, C., Fantucci, P., Ryde, U. & Gioia, L. D. Fast generation of broken-73symmetry states in a large system including multiple iron-sulfur assemblies:74investigation of QM/MM energies, clusters charges, and spin populations. Int. J.75Quant. Chem. 111, 3949–3960 (2010).7650. Pandey, A. S., Harris, T. V., Giles, L. J., Peters, J. W. & Szilagyi, R. K.77Dithiomethylether as a ligand in the hydrogenase H-cluster. J. Am. Chem. Soc.78130, 4533–4540 (2008).

79Acknowledgements80The authors thank S. Dementin for helpful discussions and critical reading of the81manuscript. The authors acknowledge funding from CNRS, AMU and ANR Q11(ANR-12-82BS08-0014). L.D.G. acknowledges support from MIUR (Prin 2010M2JARJ).83P.W. acknowledges the Ministry of Education, Republic of China (Taiwan) for a84PhD scholarship, and J.B. thanks the Royal Society for a University Research Fellowship.85The UK’s High Performance Computing Materials Chemistry Consortium (funded by the86Engineering and Physical Sciences Research Council, EP/ F067496) is acknowledged for87access to the high-performance computing facility HECToR.

88Author contributions89V.F., C.G., C.B., J.B., L.D.G. and C.L. designed the experiments. V.F., C.G., P-H.W., P.E.,90M.M. and M.B. performed the experiments. V.F., C.G., J.B., L.D.G. and C.L. analysed the91data. K.S. and H.B. provided the enzyme from Cr and constructed the mutants. I.M-S. and92P.S. provided the enzyme from Ca. V.F., C.G., C.B., J.B., L.D.G. and C.L. co-wrote the paper.

93Additional information94Supplementary information is available in the online version of the paper. Reprints and95permissions information is available online at www.nature.com/reprints. Correspondence and96requests for materials should be addressed to V.F.

97Competing financial interests98The authors declare no competing financial interests.99



Publisher: Nature

Journal: Nature Chemistry

Article number: nchem.1892

Author (s): Fourmond et al.

Title of paper: The oxidative inactivation of FeFe hydrogenase reveals the flexibility of the H-cluster

Query no.

Query Response

1 Please check that all addresses are correct as presented and ensure that all addresses are provided in full (including city)

2 DFT expanded OK?

3 I have changed this to blue and red (not red and blue) to match the figure – please check

4 Changed to Fig 3c here – OK?

5 FTIR expanded OK?

6 WT expanded OK?

7 Hox inact correct? Please check

8 Please check author names in ref. 40.

9 Ref 44, 45, 46 – please confirm these are single-page refs or provided as article numbers; otherwise please provide end pages

10 Ref 48 – please provide full publication details, and check title

11 Please expand acronyms in the acknowledgements section

12 Please provide general bold title for Figure 1

13 “reactivate” correct? Or react?

14 Fig 4a – there appear to be two dashed lines – should we give a colour here?

The oxidative inactivation of FeFe hydrogenase reveals the flexibility of the H-cluster

Documents