Crystal structure and biochemical characterization of Chlamydomonas FDX2 reveal two residues that, when mutated, partially confer FDX2 the redox potential and catalytic properties of FDX1

The green alga Chlamydomonas reinhardtii contains six plastidic [2Fe2S]-cluster ferredoxins (FDXs), with FDX1 as the predominant isoform under photoautotrophic growth. FDX2 is highly similar to FDX1 and has been shown to interact with specific enzymes (such as nitrite reductase), as well as to share interactors with FDX1, such as the hydrogenases (HYDA), ferredoxin:NAD(P) reductase I (FNR1), and pyruvate:ferredoxin oxidoreductase (PFR1), albeit performing at low catalytic rates. Here we report the FDX2 crystal structure solved at 1.18 Å resolution. Based on differences between the Chlorella fusca FDX1 and C. reinhardtii FDX2 structures, we generated and purified point-mutated versions of the FDX2 protein and assayed them in vitro for their ability to catalyze hydrogen and NADPH photo-production. The data show that structural differences at two amino acid positions contribute to functional differences between FDX1 and FDX2, suggesting that FDX2 might have evolved from FDX1 toward a different physiological role in the cell. Moreover, we demonstrate that the mutations affect both the midpoint potentials of the FDX and kinetics of the FNR reaction, possibly due to altered binding between FDX and FNR. An effect on H2 photo-production rates was also observed, although the kinetics of the reaction were not further characterized. Electronic supplementary material The online version of this article (doi:10.1007/s11120-015-0198-6) contains supplementary material, which is available to authorized users.


Introduction
Ferredoxins (FDXs) are small, ubiquitous proteins that typically contain iron-sulfur clusters and mediate electron shuttling among multiple metabolic pathways. The green Marko Boehm and Markus Alahuhta authors have contributed equally to the work.
Electronic supplementary material The online version of this article (doi:10.1007/s11120-015-0198-6) contains supplementary material, which is available to authorized users.
alga Chlamydomonas reinhardtii (Chlamydomonas throughout) contains six genes that encode for chloroplastlocalized ferredoxins. These proteins have been categorized into three separate groups, according to the plant nomenclature: FDX1 and FDX5 belong to the photosynthetic category (or leaf type), FDX2 belong to the non-photosynthetic group (or root type), and FDX3, FDX4, and FDX6 are divergent, being more closely related to bacterial FDXs (Terauchi et al. 2009). In their oxidized state, FDXs display absorption maxima in the visible region at about 330, 420, and 464 nm, and their redox potentials are very negative; FDX1, for example, is centered at around -398 mV (Wada et al. 1974;Orme-Johnson 1973;Matsubara and Sasaki 1968;Hutson et al. 1978;Hase et al. 1976). FDXs harbor a CX 4 CX 2 CX n C motif required for [2Fe2S]-cluster ligation and form specific electrostatic complexes with a variety of interacting enzymes, using carboxyl side-chain groups of conserved amino acids that generally interact with lysine or arginine counterparts on the binding partner.
As electron shuttles, Chlamydomonas FDXs are particularly important for hydrogen production, because they are the natural electron donor to the hydrogenases in vivo. Chlamydomonas contains two [FeFe]-hydrogenase enzymes, CrHYDA1 and CrHYDA2, which catalyze the formation of hydrogen from two electrons and two protons, either under light-anoxic or dark-anoxic conditions (Ghirardi et al. 2000;Harris 2009). FDXs represent a branch point for three hydrogen production pathways (Meuser et al. 2012;Winkler et al. 2010): the PSII dependent, PSII independent, and fermentative.
To further characterize the role of the algal FDXs, Peden et al. (2013) identified specific targets for each of the six Chlamydomonas FDXs and assessed CrFDX1 and CrFDX2 specificity toward selected metabolic pathways. Using yeast two-hybrid and pull-down assays, they detected binding partners for the two CrFDXs and confirmed that CrFDX2 can also interact with common CrFDX1 interaction partners. CrFDX2 amino acid sequence is highly homologous to that of CrFDX1 (67 % identity for the mature proteins), and it contains conserved residues that are known to be important for interactions with CrFDX1 enzyme targets (Winkler et al. 2009b). This sequence conservation might explain why CrFDX2 participates in electron-transfer reactions with similar redox partners as CrFDX1, although at lower rates. Both proteins contain three highly conserved, negatively charged, solvent-exposed regions that were proposed to be responsible for mediating protein-protein interactions in FDXs (Kameda et al. 2011). Typically, these regions form a highly conserved structure that facilitates cluster insertion to the apo-protein and electron transfer to/from the mature protein (Bertini et al. 2002;Kameda et al. 2011). Four conserved amino acid residues present in the two CrFDXs specifically contribute to electron donor/acceptor selectivity in vivo (Terauchi et al. 2009). Indeed, CrFDX2 has recently been shown to mediate electron transfer (although not as efficiently as CrFDX1) to known CrFDX1 interaction partners, i.e., CrFNR1, CrHYDA1, and CrPFR1, (Noth et al. 2012;Peden et al. 2013;Terauchi et al. 2009;van Lis et al. 2013). However, it is unknown whether or under which conditions this happens in vivo.
Structural models have predicted differences in surface charge distribution on the two CrFDXs, which may explain CrFDX1 0 s more negative redox potential (-398 vs -321 mV for CrFDX2) (Galván and Márquez 1985;Terauchi et al. 2009). These distinct physical characteristics might determine their interaction specificity and influence the binding of various electron acceptors/donors to each protein (Terauchi et al. 2009). Notably, CrFDX2 lacks a phenylalanine (F62) residue that may be required for the proper interaction of CrFDX1 with CrHYDA (Winkler et al. 2009a) and possibly with CrFNR1 (Hurley et al. 1997;Mayoral et al. 2005); it also lacks a C-terminus tyrosine residue, Y95, present in CrFDX1. The lower capability of CrHYDA1 to generate hydrogen using CrFDX2 as the electron donor might be indicative not only of their different redox potentials but also of different interaction mechanisms between CrHYDA1 and either CrFDX1 or CrFDX2.
In an effort to better define these interactions and the function of CrFDX1 and CrFDX2, we over-expressed the two proteins in E. coli. The mature versions of these proteins were purified, used for crystallography studies, and characterized by different types of spectroscopic techniques (see supplemental data 1, 2, 3, and 4). Here we report the first 3D structure of a Chlamydomonas ferredoxin, CrFDX2, at atomic resolution of 1.18 Å . The CrFDX2 folding motif is similar to that of previously published plant-type FDX structures from other organism (Bes et al. 1999;Fish et al. 2005), one of which represents a FDX1 type from Chlorella fusca (Cf), with high sequence homology to the noncrystallized CrFDX1. Based on the high degree of similarity between Cf and Cr FDX1, comparison with the Cr FDX2 structure, and published data, we selected two amino acid residues present on the interaction surface of CrFDX1 with FNR and hydrogenase (but absent or present as a different residue on CrFDX2). We mutated CrFDX2 F62 to M62 (the equivalent residue in CrFDX1) and inserted Y95 into CrFDX2 through site-directed mutagenesis. We show that mutations that replace these residues with those found in CrFDX1 lower CrFDX2 0 s midpoint redox potential to values closer to that of CrFDX1, indicating that these residues contribute to functional differences between CrFDX1 and CrFDX2. The major observed difference consisted of altered midpoint redox potential of the FDX cluster, resulting in changes in its catalytic efficiency with respect to FNR-mediated NADP ? reduction, as well as alterations in the maximum rates of hydrogenase-catalyzed H 2 photo-production in vitro.

Plasmid construction
We constructed several over-expression plasmids for all the mature, codon-optimized (Mr. Gene, Germany) versions of the Chlamydomonas FDX proteins, using a modified version of the pRSETA vector (Life Technologies, USA; Michoux et al. 2010) as the backbone. For the pulldown experiments, we used N-terminal His-tagged CrFDX1 and CrFDX2 proteins that were expressed from the pRESTA His-FDX1 and pRESTA His-FDX 2 expression constructs, respectively. For this, the CrFDX1 and CrFDX2 coding sequences were amplified using FDX1 primers (Peden et al. 2013) and the following FDX2 primers: FDX2-Fw GGATCCTTCAAAGTCACCTTCAAAA CCCCAAAAGGTG-3 0 and FDX2-Rev 5 0 -ACATCGTCA TTTTAACCGATCAAGAATCAAAATTGTGAGAAT TC-3. The amplified gene sequences were cloned into the BamHI and EcoRI restriction sites on the vector. The tobacco etch virus protease (TEV)-cleaved versions of the CrFDX1 and CrFDX2 proteins were expressed as His-GST-TEVcs-FDX1 and His-GST-TEVcs-FDX2 fusion proteins and contained a linker sequence (Yacoby et al. 2012) between the His-GST tandem affinity tag and the FDX sequence. These versions were used for UV/Vis and EPR spectroscopy. The cleaved versions of the CrFDX1 and CrFDX2 proteins used for CD spectroscopy that yielded CrFDX2 crystals for X-ray crystallography were expressed as FDX1-TEVcs-GST-His and FDX2-TEVcs-GST-His fusion proteins that contained the same linker sequence (Yacoby et al. 2012) between the FDX and the GST-His tandem affinity tag. To generate the expression constructs for the point-mutated CrFDX2 s (M62F, r95Y and M62F/ r95Y), the C-terminal-encoding fragment of the CrFDX2 protein was excised from the His-GST-TEVcs-FDX2 construct using HincII and EcoRI. Subsequently, digested PCR fragments generated with the following primers replaced the excised fragment: (A) CrFDX2 M62F: Fw 5 0 -GCGGTCGA CCAATCCGACCAAAACTTTTTGGACGAAGATCAA TTG-3 0 , (B) CrFDX2 M62F Rv: 5 0 -GGAATTCTCACAAT TTTGATTCTTGATC.

Protein purification
The CrFDX1 and CrFDX2 over-expression plasmids were transformed into E. coli KRX cells (Promega, USA). For expression in Terrific Broth with 200 lg/ml Ampicillin (TB; VWR, USA), a starter culture was grown overnight at 37°C and diluted 1:100 in a 100-ml TB subculture the following morning. After the subculture had reached an OD600 of *0.7, 10 ml of it were used to inoculate 1 L of TB media supplemented with 0.4 % (w/v) glycerol. At OD600 of *0.7, IPTG and Ferric ammonium citrate were added to final concentrations of 1 mM and 0.05 % (w/v), respectively, (Peden et al. 2013). The cells were harvested and resuspended in 100 ml lysis buffer (25 mM Tris pH 7.9, 100 mM NaCl, and 1 mM DTT) for breakage. The supernatant obtained after centrifugation was incubated for 1 h at 4°C with 20 ml of glutathione affinity resin (Genscript, USA). After the incubation period, the resin was washed with 15 column volumes (CV) of lysis buffer or until the wash solution became clear and colorless. Protein elution was performed with 2 9 CV of elution buffer (25 mM Tris pH = 7.9, 100 mM NaCl, and 10 mM reduced glutathione). 20 mg of TEV-His (His-tag purified from pRK193 (Kapust et al. 2001; Addgene, USA) were added to cleave the affinity tag. After a 2-h incubation period at RT, the sample was applied to a TALON Cobalt affinity chromatography column (*20 ml resin (Clontech, USA) packed in a XK16/20 (GE Healthcare column) coupled to an Ä kta FPLC (GE Healthcare, USA), see supplemental Fig. 1. We used 25 mM Tris pH 7.0, 100 mM NaCl, and 5 % (v/v) glycerol as the running buffer at a flow rate of 5-10 ml min -1 . The flow-through was collected and loaded onto a HiLoad TM 26/60 Superdex TM 75 prep grade (GE Healthcare, USA) following the purification method developed by Peden et al. (2013). The iron content was determined using a colorimetric assay described (Winkler et al. 2009b), which uses ferrozine under reductive conditions after digestion of the protein in 4.5 % (w/v) KMnO4 and 1.2 N HCl.

Protein crystallization
CrFDX2 protein crystals were obtained by the sitting drop vapor diffusion method, using a 96-well plate with Crystal Screen HT (Hampton Research, USA). Fifty lL of well solution were added to the reservoirs and drops were made with 0.2 lL of well and 0.2 lL of protein solution using a Phoenix crystallization robot (Art Robbins Instruments, USA). Protein crystals grew at 20°C in 0.1 M HEPES pH 7.0 and 3.2 M ammonium sulfate as the well solution. The protein solution contained 17.3 mg/ml of protein in 25 mM Tris pH 7.0, 200 mM NaCl, and 5 % (v/v) glycerol.

Data collection and processing
The CrFDX2 protein crystal was flash-frozen in a nitrogen gas stream at 100 K before data collection, using an inhouse Bruker 9 8 MicroStar X-Ray generator with Helios mirrors and Bruker Platinum 135 CCD detector. A well solution containing 5 % (v/v) glycerol and 5 % (v/v) ethylene glycol was added into the drop before freezing to prevent ice formation. Data were indexed and processed with the Bruker suite of programs version 2011.2-0 (Bruker AXS, USA).

Structure solution and refinement
Intensities obtained from data processing (derived from diffraction intensities) were converted into structure factors and 5 % of the reflections were flagged for Rfree calculations using the programs SCALEPACK2MTZ, ctruncate, MTZDUMP, Unique, CAD, FREERFLAG, and MTZU-TILS from the CCP4 package ). The program MrBUMP version 0.6.1 ) automatically solved the structure using the FASTA (Pearson and Lipman 1988) and MOLREP (Vagin and Teplyakov 2010) programs for sequence searches and molecular replacement. Refinement and manual correction were performed using the REFMAC5  version 5.7.0032 and the Coot (Emsley et al. 2010) version 0.6.2 programs. The MOLPROBITY method (Chen et al. 2010a) was used to analyze the Ramachandran plot, and root mean square deviations (rmsd) of bond lengths and angles were calculated from ideal values of Engh and Huber stereochemical parameters (Engh and Huber 1991). The Wilson B-factor was calculated using the ctruncate version 1.5.1, and average B-factors were calculated using the ICM version 3.7-2a program (Molsoft LLC, USA). Coot, PyMOL (http://www.pymol.org), and ICM (http:// www.molsoft.com) were used for comparing and analyzing structures. Figure 1 was done using ICM and PyMOL was used to make Fig. 2. The data collection and refinement statistics are shown in Table 1.

Isolation of thylakoids
In order to generate the thylakoids membrane for the H 2 and NADP assays, Chlamydomonas cells were harvested from cultures grown in Tris/Acetate/Phosphate medium (TAP) (pH 7.2) (Harris 2009). Algal cultures were maintained at 25°C, vigorously bubbled with air enriched with 3 % (v/v) CO 2 , stirred using a magnetic stirrer bar, and illuminated with continuous light of 80 lmol photon m -2 s -1 . These cells were then washed with 1/5 volume of buffer 1 (containing 1 0.35 M sorbitol, 20 mM HEPES, pH 7.5, 2.0 mM MgCl2) and repellet as above. The cells were broken using a french press and the thylakoids were pelleted at 40,000 9 G (for 20 min at 4°C). The final thylakoids were resuspended in equal or greater volume of buffer 1, homogenized and spinned at 1200 9 G for 30 s to pellet unbroken cells. The supernatant was removed and the pelletted thylakoids stored at -80°C at with a final concentration of 1.0 mg Chl -1 ml -1 .

Hydrogen photo-production
A master mix was prepared for the hydrogen photo-production assay as follows (amount per assay): 900 ml buffer A (50 mM tris-HCl pH 7.4, 3.35 mg ml -1 bovine serum albumin, 10 mM MgCl 2 , and 200 mg ml -1 sucrose), 5 lL of DCPIP (0.01 mM in buffer A), 10 lL of DCMU cluster. Panel C is rotated 180°with respect to panels A and B. The orientation chosen here highlights the different-charged region of CrFDX2 with the red region being the area of CrFNR/CrHYDA1 interaction (0.3 mM in DMSO), 10 lL of sodium ascorbate (1 M), 10 lL glucose oxidase (30 mg/ml in buffer A), 50 lL glucose (1 M), 10 lL catalase (10 mg ml -1 in buffer A), 50 lL 96 % (v/v) ethanol, and 100 nM HYDA1 hydrogenase (in buffer A), as described previously (24). All solutions had been previously degassed and were mixed inside a MBRAUN glove box in a 100 % N 2 atmosphere. FDX was placed in 9-ml serum vials containing the master mix to a final concentration of 10 mM. After addition of Chlamydomonas thylakoids to a final concentration of 25 mg ml -1 chlorophyll in the dark, 1.2 ml of the master mix were transferred to serum vials. The vials were sealed with rubber septa and wrapped in aluminum foil. A zero time-point sample was taken, the vial was unwrapped, and illuminated at 400 lmol photons m -2 s -1 generated by a LED light source (2000 W Diamond Series, www.advan cedLEDlights.com). Hydrogen in the head-space was measured at different time points by a gas chromatograph (400 ll injection volume), and the resulting hydrogen production rates were calculated using data from three  Statistics for the highest resolution bin are in parenthesis a Rint ¼ P I À \I [ j j = P I j j where I is the intensity of an individual reflection and \I[ is the mean intensity of a group of equivalents, and the sums are calculated over all reflections with more than one equivalent measured b Chen et al. (2010b) c Chen et al. (2010a) Photosynth Res (2016) 128:45-57  49 replicate samples for each FDX tested (lmol H 2 lg Chl -1 h -1 ).

NADPH photo-production
Initially, three solutions were prepared: (a) a buffer master mix containing (for each FDX/concentration combination) 1 ll of DCMU (0.3 mM in DMSO), 2.5 ll of DCPIP (0.01 mM), 5 ll of sodium ascorbate (1 M), Chlamydomonas thylakoids to a Chl concentration of 50 lg ml -1 (in the final assay) and buffer A (see above) to a total volume of 268 ll per buffer master mix; (b) a protein mix (final volume of 208 ll) containing 0. Then, in the dark, 208 ll of the buffer master mix were added to each protein mix and 130 ll of this mixture were added to three separate wells (triplicate samples). The plate was kept in the dark (t = 0 was taken) and then illuminated at 300 lmol photons m -2 s -1 from an LED light source (2000 W Diamond Series, www.advancedLEDlights.com). Further absorbance measurements were taken at various time points and recorded by the Infinite M200Pro plate reader (Tecan, USA). NADPH production rates (lmol NADPH lg Chl -1 h -1 ; for FNR1 at 1 mM in a final volume of 1 ml) were calculated based on a NADPH standard curve to determine the amount of NADPH produced in the assay (assay volume is 138 ll). Subsequently, the inverse of the rates were plotted over the inverse of the FDX concentrations in a Lineweaver-Burk plot and both K m and V max values were calculated for three replicates of each FDX and concentration combinations.

EPR monitored redox titrations
Potentiometric titrations of FDX2 mutants and FDX2 for reference were carried out anaerobically in a MBraun box (N 2 atmosphere, 25°C) using an ORP triode electrode (internal Ag/AgCl reference, platinum sensor, Thermo Scientific 9678BNWP). The electrode was connected via a BNC cable to a pH meter (Oakton) operating in relative mV mode and calibrated to a standard solution (Orion 967901). All values were adjusted ?200 mV and reported versus the normal hydrogen electrode (NHE) potential. The reaction was carried out in a custom vessel (Allen Scientific Glass, Boulder CO) with magnetic stirring using a similar reductive titration procedure as described previously (Usselman et al. 2008). NaDT was used as the reductant and added in 2 uL increments (2 mM stock) with a Hamilton repeating dispenser to the protein (2-5 mg/mL) buffer solution (50 mM Tris pH 7.8, 100 mM NaCl, 20 % glycerol). The protein buffer solution was supplemented with a redox mediator cocktail (3 lM final concentration) to allow for fast equilibrium between protein and reductant (Dutton 1978). The cocktail consisted of indigo disulfonate (E m = -255 mV vs NHE), phenosafranine (E m = -255 mV vs NHE), benzyl viologen (E m = -361 mV vs NHE), and methyl viologen (E m = -440 mV vs NHE). Samples at poised redox potentials were removed from the vessel after several minutes of equilibration following NaDT addition at roughly 20 mV increments and transferred to 4 mM EPR tubes (Wilmad LabGlass). EPR tubes were sealed with septa and frozen in liquid nitrogen. EPR spectra were recorded on a Bruker ELEXSYS E500 CW X-band spectrometer system outfitted with an Oxford Instruments cryostat and temperature controller and cylindrical (SHQ) Bruker resonator. Spectra were collected at optimal power and temperature settings (1.0 mW, 23 ± 3 K) as determined from power saturation and temperature analysis of reduced wild-type samples (Supplemental Fig. 3). Other spectrometer settings were as follows: microwave frequency, 9.385 GHz; modulation frequency, 100 kHz; modulation amplitude, 10.0 G; and time constant, 327.68 ms. Simulations of the spectra were carried out in EasySpin (Stoll and Schweiger 2006).
To determine the midpoint potential (E m ) of the [2Fe2S] 2?/1? cluster, plots of signal amplitude of the reduced [2Fe2S] 1? signal (measured at the g = 2.05/2.06 peak) versus sample potential (E) were fitted a form of the n = 1 Nernst equation (Hagen 2008).
Fits were carried out using the nonlinear least-squares curve fitter in OriginPro. Errors in the E m values are estimated at ±6 mV from the standard error of the fits and voltage readouts during the experiment.

Crystal structure of CrFDX2
The structure of CrFDX2 was refined to a resolution of 1.18 Å with R and Rfree of 0.109 and 0.147, respectively (Fig. 1). There is only one molecule in the asymmetric unit in complex with the [2Fe2S] cluster (Fig. 1a), and it shows a typical ferredoxin fold with a b-sheet formed by five bstrands covered by a single a-helix (Fukuyama 2004). The [2Fe2S] cluster of CrFDX2 is coordinated by four cysteine residues: Cys38, Cys 43, Cys 46, and Cys 76 (Fig. 2a, b; the numbering differs from that in Fig. 3 by 1, due to the lack of M at the start of each recombinant protein sequence). This structure has been deposited into the protein data bank (PDB; www.rcsb.org) with entry code 4ITK.

Structural comparison with other FDXs
Pair-wise secondary structure matching by the PDB-fold program (Krissinel and Henrick 2004) found 60 unique structural matches for CrFDX2 from the protein data bank with at least 70 % secondary structure similarity. Out of these, the first 53 were [2Fe2S]-ferredoxins and the remaining 7, although having high protein fold similarity to CrFDX2, showed less than 20 % sequence similarity with it. The most similar match to CrFDX2 was the cyanobacterium Mastigocladus laminosus [2Fe2S]-FDX (PDB ID: 1RFK; (Fish et al. 2005)), with a secondary structure similarity of 100 %, sequence similarity of 66 %, and Ca root mean square deviation of 0.84 Å 2 , suggesting similar backbones between the two proteins. Further inspections of similar ferredoxin structures showed significant variability between the positions of the backbone atoms away from the iron-sulfur cluster. This was the case even when the overall structure seemed to be highly similar. To properly find similar structures that might have been missed by the structural similarity search, we searched the PDB using sequence homology with the ICM program and found 36 structures with sequence similarity above 25 %. The best hit was Chlorella fusca FDX1 (CfFDX1, PDB ID: 1AWD Bes et al. 1999) with sequence similarity of 68 % and Ca root mean square deviation of 0.79 Å 2 . Closer inspection of this structure showed that it indeed was highly similar to CrFDX2.

Algal CrFDX1/CrFDX2 binding interface with CrHYDA1/CrFNR1 and point mutations
The CrFDX1 and CrFDX2 proteins are highly homologous, with 67 % sequence identity (Fig. 3). Furthermore, CrFDX2 (a root-type ferredoxin Terauchi et al. 2009) shares a practically identical protein backbone with the leaf-type CfFDX1 (Fig. 3) indicating that both FDX types would be expected to share the same binding interface with CrHYDA1/CrFNR1 (Fig. 2a). For this interaction, we assumed that the binding interface is centered on the ironsulfur cluster, based on a thorough analysis of related binding interfaces from literature (Chang et al. 2007;Hurley et al. 1993a, b;Morales et al. 2000). The distance between the iron-sulfur clusters is crucial for electron tunneling. According to known structures and computer modeling of electron tunneling enzyme complexes, the edge-to-edge distances should not exceed 14 Å in the absence of other additional cofactors to act as electron relays. This is the maximum distance that allows for physiologically relevant electron-transfer rates (Moser et al. 2010;Page et al. 1999;Gray and Winkler 1996). Using this information, we visually identified residues in the vicinity of the iron-sulfur cluster that were conserved between CrFDX1 and CfFDX1 but not in CrFDX2. We specifically focused on two amino acids located near the electron acceptor site of FDX1, namely F62 and Y95: (Fig. 2b shows the location of these residues). In vitro hydrogen photo-production rate To evaluate the effect of the selected mutated amino acid residues on CrFDX2 biochemical properties, we measured hydrogen photo-production rates driven by CrFDX1, CrFDX2, and the point-mutated versions of CrFDX2 (M62F, r95Y, and M62F/r95Y) (see Table 2). In agreement with Peden et al. (2013), CrFDX1 promoted the highest hydrogen photo-production rate, 489 (±48) lmol H 2 lg Chl -1 h -1 , while CrFDX2 displayed a 5.6-fold lower rate (86 ± 11 lmol H 2 lg Chl -1 h -1 ). Interestingly, each of the CrFDX2 point-mutants resulted in rates that were almost twofold higher than the native CrFDX2 (M62F, 150 ± 41; and r95Y, 134 ± 42 lmol H 2 lg Chl -1 h -1 ), but lower than the CrFDX1 rates. This effect appears to be additive, as the CrFDX2 double mutant (M62F/r95Y) displayed the highest rate among the three mutants (264 ± 69 lmol H 2 lg Chl -1 h -1 ). The HYDA1 H 2 evolution rate with CrFDX2 M62F/r95Y was threefold higher than with CrFDX2 WT, and half of that with CrFDX1. These data show that F62 and Y95 each function to support productive electron-transfer complexes between CrFDX1 and CrHYDA1 and, when engineered onto CrFDX2, confer it higher catalytic rates. In other words, mutations that introduce these CrFDX1 amino acid residues onto CrFDX2 are likely to induce similar structural changes in CrFDX2 to promote higher hydrogen production rates, closer to those measured with CrHYDA1.

In vitro NADPH photo-production kinetic parameters
The kinetic parameters K m (lM), V max (lmol NADPH lg Chl -1 h -1 ) and k cat (s -1 ), of CrFDX1, CrFDX2, and the CrFDX2 point-mutants for the NADPH photo-production reaction were determined from the respective Lineweaver-Burk plots (see Table 3 and Supplemental Fig. 5). Interestingly, the V max values for NADPH photo-production were similar for CrFDX1 and CrFDX2, at 185 (±68) and 177 (±47) lmol NADPH lg Chl -1 h -1 , respectively. However, the K m for CrFDX2 (0.18 ± 0.01, lM) was significantly lower than that for CrFDX1 (0.40 ± 0.04, lM), indicating CrFDX2 0 s higher affinity for CrFNR1. On the other hand, K m values for each of the CrFDX2 mutants were higher than that for the native CrFDX2 (Table 3) (Table 3). Although CrFDX2 supported a slightly lower V max , the lower K m with CrFNR1 resulted in a twofold larger k cat /K m , indicating the formation of a more efficient catalytic complex. On the other hand, CrFDX2 M62F presented a fourfold lower k cat /K m compared to CrFDX2 and twofold lower than CrFDX1. The CrFDX2 r95Y variant led to a slightly higher V max but lower k cat /K m , a value that is more similar to that measured with CrFDX1. When the M62F mutation was paired with r95Y in the CrFDX2 double mutant, the resulting variant supported the highest k cat , but led to a large, 13-fold decrease in k cat /K m compared to CrFDX2. Overall, the kinetics suggest that CrFDX2, under growth conditions where it is present in equimolar amounts to CrFDX1, is better at electron transfer with FNR1 than CrFDX1, by virtue of forming a more efficient catalytic complex with FNR1. In contrast, CrFDX1 catalyzes higher rates of H 2 evolution, and the changes (although small) observed with the M62F and r95Y mutants of CrFDX2 are evidence for involvement of these residues in H 2 evolution. The CrFDX1, CrFDX2, and mutated CrFDX2 proteins used for the hydrogen photo-production assay were the cleaved forms purified from proteins over-expressed in E. coli from the respective FDX-TEVcs-GST-His constructs. Individual rates were calculated from four time points (approximately t = 60, 180, 300 and 420 min) and the averaged rates are shown

Midpoint redox potential determination
The midpoint redox potentials of the CrFDX2 mutants were determined to test the effect of the mutations on the electron transfer properties of the [2Fe2S] cluster. Both of the single mutations shifted the potential more negative compared to CrFDX2, with the M62F mutation giving the largest shift bringing it close to the midpoint potential of CrFDX1 (Fig. 4, Table 4). Surprisingly, the double mutant did not show an additive shift of the single mutations but rather displayed a redox potential similar to that of the r95Y alone. While the overall rhombic signal assigned to the [2Fe2S] cluster was almost identical in all cases, slight shifts in the g-values were observed for the mutants. It should be noted that appearance of other small signals were observed; however, these are likely from the redox cocktail and formation of radical species during the course of reduction with NaDT. The shifts in the rhombic [2Fe2S]cluster signal can be summarized by a subtle upfield energy shift from the g-values of CrFDX2 (g = 2.06, 1.97, and 1.88) to the g-values of CrFDX1 (g = 2.05, 1.96, and 1.88). Compared to r95Y (g = 2.063, 1.973, and 1.883) and M62F/r95Y (g = 2.060, 1.969, and 1.883), M62 showed the largest shift (g = 2.052, 1.959, and 1.880) resulting in its overall signal to more closely align with CrFDX1 (Fig. 4 inset). Interestingly, these trends match nicely to the midpoint potential shifts of the mutants and particularly for M62F may indicate an underlying role toward finely tuning the orientation and electronic properties of the cluster.

Discussion
Recently, we initiated efforts to fully characterize the Chlamydomonas FDX interaction network (Peden et al. 2013). We and others had shown previously that CrFDX1 plays a predominant role as an electron carrier in the cell, through electron-transfer and binding interactions with multiple partners (Noth et al. 2012;Peden et al. 2013;Terauchi et al. 2009;van Lis et al. 2013). Interestingly, CrFDX2 was demonstrated to be capable of binding in vitro to some of the same electron partners and promoting similar redox reactions as CrFDX1 (Noth et al. 2012;Peden et al. 2013;van Lis et al. 2013).
In order to determine and compare the characteristics and functions of the two CrFDXs in more details, and to identify and study the nature of their interaction with other enzymes, we performed additional biochemical and biophysical assays on the two-purified proteins. The spectroscopy studies confirmed that both proteins are highly similar, showing typical [2Fe2S]-ferredoxin spectra (Supplemental Figs. 2, 3, and 4). We also grew crystals and solved the CrFDX2 structure, which represents the first solved Chlamydomonas FDX structure. A structure of CrFDX1 from Chlorella fusca was previously reported (Bes et al. 1999), and it was used for the visual identification of differences between CrFDX1 and CrFDX2. Five different amino acids located in the vicinity of the [2Fe2S] cluster and at the binding interface between both CrFDXs and CrHYDA1 were identified (Winkler et al. 2009a). As such, they have the potential to affect the binding properties of either of the two CrFDXs, or possibly their electron transfer potentials to specific donors/acceptors. We mutated two residues in CrFDX2 to resemble those present in CrFDX1 and determined the effects on two reactions: NADP ? reduction and H 2 photo-production.
The catalytic properties of each of our FDX mutants are different with respect to electron transport to CrHYDA1 and CrFNR1, as shown in Tables 2 and 3, respectively. We demonstrated that introduction of Y95 and mutation of CrFDX2 M62 to phenylalanine directly affects its hydrogen and NADPH photo-production activity. The single CrFDX2 M62F and r95Y mutants showed higher hydrogen photo-production rates than the CrFDX2 WT protein, but not as high as that of CrFDX1. This demonstrates that CrFDX2 (M62F/r95Y) 3.44 ± 0.71 260 ± 33 939 ± 120 2.7 9 10 8 The FDX1, FDX2, and mutated FDX2 proteins used for the NADPH photo-production assay were prepared by TEV treatment of FDX-TEVcs-GST-His fusions. Both K m and V max values were calculated from three independent replicates using linear regression analyses of Lineweaver-Burk plots. Supplemental Fig. 5 shows the Lineweaver-Burk plots for the averaged rates. NADPH photo-production rates (V max ) are given in lmol NADPH lg Chl -1 h -1 with 1 mM FNR1 present and a 1-ml assay volume were calculated from the initial 15 min after mixing. The k cat values were calculated at FDX concentrations of 1 mM Photosynth Res (2016) 128:45-57 53 both residues in CrFDX1 promote high hydrogen production, either by affecting protein complex formation/stabilization, electron transfer, or both. This was to be expected from the structural results that showed the proximity of these residues to the CrFDX: CrHYDA1 binding interface (Fig. 2b). Interestingly, the double mutant was able to support hydrogen production at an even higher rate than either of the two single mutants and, in fact, twofold higher than the CrFDX2 WT ( Table 2). The additive effect of these mutations suggests that both residues in CrFDX1 contribute to its high H 2 production. Indeed, both mutations resulted in shifts in the midpoint redox potentials of CrFDX2 toward more negative values (Table 4), which favor electron transfer to CrHYDA1. We also tested the kinetics of WT and point-mutated CrFDX2 proteins in NADPH photo-production. We show that the presence of CrFDX1 F62 and Y95 in CrFDX2 affect NADPH production in opposite manners. The presence of F62 in CrFDX2 interferes mostly with its binding affinity to CrFNR (higher K m compared to wild-type FDX2) and leads to a decrease in the turnover of NADPH production (k cat ). This could be due to the direct interaction between residue 62 and CrFNR's FAD co-factor (Supplementary Fig. 4c, e), which may be weakened by the presence of methionine in this position. On the other hand, the deletion of Y95 from CrFDX2 affects all kinetic parameters of the FNR-mediated reaction, as shown by an increase in its K m , K cat and catalytic efficiency (levels similar to those obtained with CrFDX1) ( Table 3). These residues are therefore important for the CrFDX1: CrFNR interaction and photo-reduction of NADP ? . It must be noted that others had shown that the CrFNR has almost identical k cat s in the presence of either CrFDX1 or CrFDX2, but the K m for CrFDX2 was almost sixfold lower than that for CrFDX1, resulting in a sixfold higher catalytic efficiency for CrFDX2 (Hurley et al. 1997;Vieira and Davis 1986). However, the reported values were derived from an assay that indirectly measured electron transfer between ferredoxin and CrFNR using cytochrome c reductase activity; therefore their results are not directly comparable to ours.
The kinetics parameters observed for the CrFDX2 double mutant seem to be a combined effect of the two single mutations, yielding a protein with a much higher K m (as the single M62F mutant), but also high V max (high rates of NADP reduction, as the single r95Y mutant) and high k cat (higher catalytic efficiency, as the single r95Y mutant), although low catalytic efficiency. These differences are accompanied by structural differences at positions 62 and 95 in the CrFDXs, which could partially account for the observed differences in kinetics. These observations are also consistent with the concept that subtle changes involving this particular structural region have significant effects in electron transfer within the functional catalytic complex. Indeed, all the mutations introduced at position 62 and 95 of CrFDX2, singly or in combination led to changes in redox-midpoint potentials as compared to CrFDX2.The EPR titration data (Table 4) showed that the E m value of the r95Y CrFDX2 mutant shifted more negatively, closer to the CrFDX1 value (-350 vs -398 vs -331 mV for r95Y, CrFDX1 and CrFDX2, respectively). The double mutant M62F/r95Y showed a similar pattern with an E m of -350 mV where the single M62F mutant had midpoint potential of -400 mV almost identical to CrFDX1 (-398 mV Terauchi et al. 2009).
In previous literature, certain amino acid residues were shown to be important specifically for the CrFDX1/ CrHYDA1 interaction, and to be critical for efficient electron transfer between these two proteins. In silico docking analysis and site-directed mutagenesis, for Fig. 4 Redox titrations of the [2Fe2S]-EPR signal (inset) from reduced CrFDX2 wild type, r95Y, M62F/r95Y, and M62F mutants. Each point reflects a measured EPR amplitude of the g = 2.05/2.06 peak for individual samples poised at particular potentials (E) vs NHE. The midpoint potentials were determined by fitting the potentiometric curves to the n = 1 electron form of the Nernst Equation Table 4 The midpoint potentials of FDX2 mutants compared to CrFDX1 and CrFDX2 wild types CrFDX2 (r95Y) -350 CrFDX2 (M62F/r95Y) -356 a Ref (Galván and Márquez 1985) b This study c Ref. (Terauchi et al. 2009) instance, identified (among ten amino acid residues tested) CrHYDA1 K396, and FDX1 E122 (amino acid numbers represent the position in the protein prior to cleavage of the transit peptide) as the major contributors to the formation of the FDX-HYD1 complex (Winkler et al. 2010). Interestingly, the two CrHYDA1 and CrHYDA2 share conservation of the required lysine, and five CrFDXs (except for CrFDX3) contain the conserved glutamic acid residue. Residues D56 and F93 in FDX1 were also shown to be important for the CrFDX1-CrHYDA1 interaction; F93 together with E122 and Y126 was proposed to be involved in stabilizing the redox state of [2Fe2S] cluster of CrFDX1, suggesting their probable role in electron transfer between CrFDX1 and CrHYDA1 (Winkler et al. 2010). Indeed, when mutated to nonconserved residues, the respective recombinant proteins showed a decreased V max for H 2 photo-production of more than threefold compared to the WT value (11).
Electrostatic interactions have also been demonstrated to be crucial for the interaction between ferredoxins and all their other target enzymes, such as FNR, FD:thioredoxins reductase, nitrite reductase, glutamate synthase, and sulfite reductase (Hanke and Mulo 2013). More specifically, CrFDX1 interaction-complex studies have provided evidence for the essential role of another conserved glutamate, E91, located in the short C-terminal tail of CrFDX1, and of a negatively charged patch, located in its most N-terminal a1-helix, which includes D25, E28, and E29. All four negatively charged residues are conserved only in CrFDX1, CrFDX2, and CrFDX5, out of the six CrFDXs (Terauchi et al. 2009). E91 is involved in forming complexes with nitrite reductase (NiR), glutamate synthetase (FDX-GOGAT), and the photosystem I subunit C (PSAC). Mutations of this residue in CrFDX1 diminish its catalytic activity in the reactions involving these enzymes (Fischer et al. 1998;GarciaSanchez et al. 1997). In addition, a triple D25A/E28Q/E29Q mutant protein showed less efficient interaction with those same three interacting enzymes (GarciaSanchez et al. 1997;Jacquot et al. 1997).
Besides electrostatic interactions, the midpoint redox potential of the two CrFDXs plays an important role in their physiological functions. Normally, CrFDX2 catalyzes nitrite reduction (Terauchi et al. 2009), a reaction with a midpoint redox potential of about -300 mV; NADP ? reduction and H 2 production, on the other hand, are catalyzed by CrFDX1 in vivo, in reactions that require more negative redox potentials (-320 and -400 mV, respectively). It was shown that the presence of an aromatic residue at position 65 in Anabaena ferredoxin is essential for effective electron transfer with FNR (Hurley, cheng et al. 1993), and that natural variants of animal and bacterial FDXs in which methionine is replaced by phenylalanine show a shift in the E m to more positive values (Hurley et al. 1993a(Hurley et al. , b, 1997. We suspected that the switch from phenylalanine at residue 62 in CrFDX1 to methionine in CrFDX2 (Fig. 2b), in particular, was responsible for the significantly different CrFDX2 ability to catalyze NADPH and H 2 production. In this study, we measured and compared redox potential and kinetic parameters for the NADPH and H 2 photo-production between the CrFDX2 and the CrFDX2 M62F and r95Y mutants with those of CrFDX1.
NADPH production, which involves NADP ? ? 2H ? binding, involves a more complex reaction, since it requires the binding of both NADP ? and 2H ? to the FDX/FNR complex when compared to the FDX/ HYDA interaction which requires only the binding of 2H ? . Finally, although our CrFDX2 crystal structure shows that both residues are located near the FDX [2Fe2S] cluster and could therefore influence the catalytic activity of FDX when in complex with FNR/HYDA, it seems that this is in fact organism dependent. In Anabaena, for example, FDX undergoes a conformational change at the level of the loop that contains F65 (the equivalent of F62 in Chlamydomonas) upon binding to FNR (Morales et al. 2000). Furthermore, this amino acid is proposed to be involved in the electronic coupling between the two redox centers (Hurley et al. 1993a). On the other hand, the maize leaf FDX/FNR crystal structure complex revealed that the equivalent amino acid of F65 (here Tyr 63) is neither in close contact with the [2Fe2S] cluster of the FDX, nor is between the two prosthetic groups from FNR (FAD) and FDX ([2Fe2S]) suggesting a different role in electron transfer to FNR for that residue (Kurisu et al. 2001). Unfortunately we do not have a crystal structure of the Chlamydomonas FDX/FNR complex to verify either mechanism. An alternative possible explanation for our kinetic data is that the mutations could interfere with PSI binding, which would further affect electron transfer to the FDXs. The data could therefore indicate that the mutations perturb the conformation of the CrPSI: CrFDX2 complex solely or in addition to the CrFDX2:CrHYDA1 complex. It is known that PSI subunits interact with CrFDX, although it remains unclear how many sites are present and/or available for FDX binding on CrPSI. Furthermore, no information regarding the K m for CrPSI: Cr FDX electron transfer is available, although K d values for WT and some PSI mutants ranging from 6 to 0.12 lM have been reported (Setif 2001;Setif et al. 2002). Finally, it must be noted that the error bars reflecting the double mutant data were particularly high, possibly due to the higher instability of the double mutant protein. The actual kinetic values must thus be taken only as representing a trend, not an actual number.
In summary, despite their high sequence similarity and comparable physical characteristics, CrFDX1 and CrFDX2 also exhibit structural differences that affect their electrontransfer function. A previous structural model predicted differences in the surface charge distribution between the two proteins (Terauchi et al. 2009), and our CrFDX2 structure and biochemical results further show that the CrFDX2 M62 and CrFDX1 95Y residues make significant contributions to the binding interfaces of the respective CrFDXs with CrHYDA1 and CrFNR1, as well as affecting their redox potentials. These distinct differences must certainly contribute to the different in vivo specificities of the two proteins (Gou et al. 2006;Terauchi et al. 2009). In this report, we indirectly demonstrate that the CrFDX1 F62 and Y95 residues are important for hydrogen photo-production, as progressively increased hydrogen production rates are measured when these residues are introduced into CrFDX2. Residues F62 and Y95 also affect NADPH photo-production and have opposite impacts on the kinetic parameters of that reactions. We thus confirm that CrFDX2 can potentially replace CrFDX1 in CrFDX1-dependent reactions and that differences between the two proteins rely on differences between only a few amino acid residues.