Light-induced H₂ generation in a photosystem I-O₂-tolerant [FeFe] hydrogenase nanoconstruct

Significance

Generating dihydrogen (H₂) in an environmentally friendly way is an important challenge. A recently characterized O₂-stable [FeFe] hydrogenase presents a unique opportunity for redirecting energy produced by Photosystem I (PS I) to H₂ production in a phototrophic biological system that only requires sunlight and earth-abundant elements. Our study presents a strategy to couple such an [FeFe] hydrogenase to PS I by fusing the former with a stomal subunit of PS I (PsaE). Not only does the chimeric nanoconstruct generate reasonably high rates of H₂ when illuminated, but it also functions in the presence of O₂. By investigating catalytic properties and drawbacks of the nanoconstruct, this work sets the stage for engineering sustainable biofuel production in vivo.

Abstract

The fusion of hydrogenases and photosynthetic reaction centers (RCs) has proven to be a promising strategy for the production of sustainable biofuels. Type I (iron-sulfur-containing) RCs, acting as photosensitizers, are capable of promoting electrons to a redox state that can be exploited by hydrogenases for the reduction of protons to dihydrogen (H₂). While both [FeFe] and [NiFe] hydrogenases have been used successfully, they tend to be limited due to either O₂ sensitivity, binding specificity, or H₂ production rates. In this study, we fuse a peripheral (stromal) subunit of Photosystem I (PS I), PsaE, to an O₂-tolerant [FeFe] hydrogenase from Clostridium beijerinckii using a flexible [GGS]₄ linker group (CbHydA1-PsaE). We demonstrate that the CbHydA1 chimera can be synthetically activated in vitro to show bidirectional activity and that it can be quantitatively bound to a PS I variant lacking the PsaE subunit. When illuminated in an anaerobic environment, the nanoconstruct generates H₂ at a rate of 84.9 ± 3.1 µmol H₂ mg_chl^–1 h^–1. Further, when prepared and illuminated in the presence of O₂, the nanoconstruct retains the ability to generate H₂, though at a diminished rate of 2.2 ± 0.5 µmol H₂ mg_chl^–1 h^–1. This demonstrates not only that PsaE is a promising scaffold for PS I-based nanoconstructs, but the use of an O₂-tolerant [FeFe] hydrogenase opens the possibility for an in vivo H₂ generating system that can function in the presence of O₂.

Oxygenic photosynthesis drives a large fraction of life on earth by storing solar energy within chemical bonds through the act of CO₂ fixation and ATP formation. Central to this process are the two photosynthetic reaction centers (RCs), Photosystem I (PS I) and Photosystem II (PS II). These large pigment–protein complexes are found only in the thylakoid membranes of cyanobacteria, algae, and higher plants.

PS I is unique among Type I (iron-sulfur cluster) RCs, as its core is heterodimeric, containing a pseudosymmetric arrangement of electron transfer cofactors (Fig. 1A) (1, 2). The key electron transfer molecules bound to the PsaA/PsaB core include six pseudosymmetrical chlorophyll (Chl) a molecules consisting of a Chl dimer termed the primary donor P₇₀₀ and the primary acceptors, A_A/A_B and A_0A/A_0B (“A” and “B” denote the subunit) (3). Following the absorption of a photon of light by a core or antenna Chl pigment, the excitation energy is transferred to P_700. The excited singlet state of P₇₀₀ is followed by a donation of an electron to A_A/A_B or A_0A/A_0B, thereby initiating a charge-separated state. The electron is subsequently transferred through a phylloquinone (A_1A or A_1B), a [4Fe-4S] cluster, F_X, and finally to a pair of [4Fe-4S] clusters, F_A and F_B, located in the PsaC subunit. The P₇₀₀^∙ +/[F_A/F_B]^– charge-separated state has a sufficient lifetime to allow for a soluble [2Fe-2S] (plant-type) ferredoxin (Fd) to accept the electron from the [F_A/F_B]^– clusters. The oxidized P₇₀₀^∙ + is subsequently reduced by plastocyanin or cytochrome (Cyt) c₆, thereby recovering the P₇₀₀ ground state and resetting the system for a subsequent light-induced turnover.

Fig. 1.

(A) Electron-transfer cofactors of the CbHydA1-PsaE:PS I_ΔPsaE nanoconstruct based on the structures of the PS I from Synechococcus elongatus (PDB: 1JB0) and [FeFe] hydrogenase from Clostridium beijerinckii (PDB: 6TTL). (B) Schematic representation of the protein fusion strategy employed.

An extensive chlorophyll antenna in PS I ensures a high cross-sectional area for photon capture with a quantum efficiency approaching unity, which means that virtually every photon is transferred to the core chlorophylls and converted into a charge-separated state. PS I (and PS II) are so efficient that during periods of high light intensity, the downstream processes of CO₂ fixation and ATP production cannot keep up with the reducing equivalents produced (4, 5). The efficiency of PS I presents an opportunity to direct the excess energy to alternative (engineered) pathways to power chemical transformations, such as the reduction of protons to produce dihydrogen (H₂).

Previous attempts to generate H₂ using PS I in vitro involved deposition of Au, Ru, Ir, Os, or Pt metals onto the stromal side of thylakoid membranes (6–10), implantation of Au and Pt nanoparticles (11–15), and the binding of cobaloxime and Ni-based catalysts (16). The use of metals has proven effective for the light-induced production of H₂. However, the need for expensive metals and the difficulty of generating these nanoconstructs in vivo present a significant drawback of this approach for utilization in commercially viable H₂ production. Engineering sustainable H₂-generating microorganisms by coupling PS I to an enzymatic H₂-producing system could be a rational alternative.

Hydrogenases are enzymes catalyzing the reversible oxidation of H₂. For PS I-based nanoconstructs, two superfamilies have been considered: [FeFe] hydrogenases and [NiFe] hydrogenases, named after the metal composition of their active site. [NiFe] hydrogenases are found in a wide variety of anaerobic and aerobic bacteria and archaea (17–19). [FeFe] hydrogenases comprise a functionally and structurally diverse superfamily typically found in obligate anaerobes (20). However, some green algae have been shown to express [FeFe] hydrogenases and produce H₂ under anaerobic conditions (21, 22). The ability of [FeFe] hydrogenases to produce H₂ at a higher rate than the energy throughput of PS I sparked widespread interest in this family of enzymes for biohydrogen production engineering efforts (23, 24).

Lubner et al. assembled a highly efficient nanoconstruct by directly coupling an [FeFe] hydrogenase from Clostridium acetobutylicum (CaHydA1) to PS I using a molecular wire in vitro (23). This PS I-CaHydA1 nanoconstruct achieved high rates of light-induced H₂ generation (up to 105 e^– PS I⁻¹ s⁻¹, SI Appendix, Table S1), outperforming electron transfer rates in natural photosynthesis by a factor of two. More recent work by Gorka et al. demonstrated an alternative strategy of utilizing A_1A/A_1B-based molecular wires to afford extraction of the electron from PS I (11, 12). One primary obstacle in utilizing wire-based coupling in vivo is the establishment of a molecular wire-S-[4Fe-4S] bond. It is thus unsurprising that despite the extraordinary success of these in vitro studies, no reports exist, to our knowledge, demonstrating the implementation of molecular-wire coupling strategies in vivo.

A more promising strategy for coupling PS I complex with a hydrogenase in vivo is by engineering a genetic fusion of the latter and one of the stromal subunits of PS I. Ihara et al. reported fusing a [NiFe] hydrogenase from the β-proteobacterium Cupriavidus necator H16 (formerly known as Ralstonia eutropha H16) to PsaE from Thermosynechococcus elongatus (25). More recently, Appel et al. demonstrated coupling of the native cyanobacterial [NiFe] hydrogenase from Synechocystis sp. PCC 6803 to PS I by fusion with PsaD (26). Unfortunately, these studies demonstrated low rates of light-driven H₂ evolution, most likely due to the bias of [NiFe] hydrogenases for H₂ uptake (27–29). Kanygin et al. generated chimera proteins formed between PsaC and [FeFe] hydrogenases HydA1 and HydA2 from Chlamydomonas reinhardtii (Cr) that complexed with PS I in vivo and generated H₂ in the light (30, 31). While the constructs performed exceptionally well in an anaerobic environment (SI Appendix, Table S1), the authors reported a loss of hydrogenase activity under aerobic conditions, likely due to the O₂-induced degradation of the algal [FeFe] hydrogenases (30).

The sensitivity of [FeFe] hydrogenases to O₂ is well documented and, until recently, was thought to be a universal property of this class of enzymes that stems from the unique construction of the active site (32–35). The active site of the [FeFe] hydrogenases (H-cluster) is composed of a [4Fe-4S] subcluster bound via a bridging cysteine thiolate to a labile [2Fe] subcluster. The Fe atoms of the subcluster are ligated by CO and CN^– and a bridging azadithiolate moiety (29, 36). In the presence of O₂, the H-cluster irreversibly degrades, likely involving the generation of reactive oxygen species at the open coordination site of the [2Fe] subcluster. Therefore, the presence of O₂ under photoautotrophy would inevitably result in an irreversible loss of hydrogenase activity (27, 32, 37).

It would seem as though the choice is between O₂ tolerance at the expense of H₂ generation or high rates of H₂ production in strictly anaerobic environments. However, the emergence of a subclass of O₂ tolerant [FeFe] hydrogenases (24, 32, 38) might allow PS I-[FeFe] hydrogenase nanoconstructs to be viable for light-driven H₂ generation in phototrophs. We have shown that [FeFe] hydrogenase from C. beijerinckii (CbHydA1) enters an O₂-protected state (termed H_inact) upon exposure to air. Notably, the reactivation requires a simple one-electron reduction of H_inact state. While the details of what confers O₂ stability are still under investigation, a structural study of CbA5H, a close homolog of CbHydA1, in the O₂-protected state suggests that Cys367 can act as a labile and protective cap, shielding the subcluster from O₂ (39). In support of this hypothesis, our recent electron paramagnetic resonance (EPR) investigation provided evidence for enhanced protein mobility around the H-cluster in CbHydA1 (40). Taken in concert with the recent work by the Happe group (41), these results suggest that protein flexibility may play a vital role in the ability of CbHydA1 to access the O₂-protected H_inact state. Therefore, understanding whether a protein modification, e.g., fusion with PsaE, affects its O₂ tolerance and activity is paramount to the prospects for employing this enzyme or its homologs in phototrophic H₂ production.

In this work, we sought to investigate the feasibility of coupling CbHydA1 to PS I complex from Synechococcus sp. PCC 7002 (hereafter S. 7002) by fusing CbHydA1 with PsaE. We chose PsaE as the fusion counterpart due to the viability of PsaE-deletion variants of cyanobacteria (PS I_ΔPsaE) (42). Aside from the [NiFe] hydrogenase-PsaE fusion, PsaE has also been used successfully for engineering PS I-based chimeric nanoconstructs for nonhydrogenase systems (43, 44).

We describe the design of the construct consisting of a modified CbHydA1 connected to a 6X-His tagged PsaE protein via a flexible amino acid linker (four-times repeat of Gly-Gly-Ser), hereafter termed CbHydA1-PsaE. We show that with the addition of the linker group and PsaE moiety, the hydrogenase retains bidirectional activity, albeit with some modification of activity. We also show that the presence of CbHydA1 does not interfere with the binding of PsaE to PS I_ΔPsaE. This PS I-CbHydA1 nanoconstruct generates H₂ upon illumination in both anaerobic and aerobic conditions, opening the door to the creation of in vivo H₂-generating systems that utilize natural photosynthetic machinery.

Results

Characterization of CbHydA1-PsaE Fusion Protein.

The CbHydA1-PsaE fusion protein was designed by combining the amino acid code for CbHydA1 followed by a [GGS]₄ linker, the code for wild-type (WT) PsaE from S. 7002, and a C-terminal His₆ tag to aid in purification (Fig. 1 and SI Appendix, Fig. S1). We truncated the CbHydA1 domain at the N terminus by 21 amino acids as this portion of the sequence is designated as a signaling sequence and thus may hinder the expression. To account for the potential differences in H-cluster assembly efficiency during synthetic maturation, we estimated the concentration of the fusion protein containing fully assembled H-cluster by incubating samples under a carbon monoxide atmosphere and subsequently quantifying the amount of H_ox-CO state by continuous wave (CW) EPR (Materials and Methods and SI Appendix, Fig. S9).

Intriguingly, the hydrogenase activity analysis indicates that the enzymatic activity of the CbHydA1-PsaE fusion protein obtained using this methodology is about 3-fold lower than that of CbHydA1_WT (Fig. 2). We estimated the rate of H₂ uptake to be 524 ± 156 and 173 ± 33 µmol H₂ mg_prot^–1 min^–1 for holo-CbHydA1_WT (SI Appendix, Fig. S3) and holo-CbHydA1-PsaE (SI Appendix, Fig. S4), respectively. H₂ evolution assays provided a specific activity of 1,087 ± 92 and 534 ± 70 µmol H₂ mg_prot^–1 min^–1 for holo-CbHydA1_WT (SI Appendix, Fig. S3) and holo-CbHydA1-PsaE (SI Appendix, Fig. S4) fusion proteins, respectively.

Fig. 2.

Specific activities for (A) H₂ evolution at pH 6.9 and (B) H₂ uptake at pH 8.0 for CbHydA1_WT (orange) and CbHydA1-PsaE fusion (yellow).

To investigate whether the difference in the activity of the WT and fusion proteins is due to a difference in the electronic structure of the H-cluster, we performed a series of Fourier-transform infrared (FTIR) spectroscopy measurements, focusing on CO and CN⁻ stretching bands of the H-cluster (38). The CbHydA1-PsaE fusion protein obtained immediately after activation with the synthetic [2Fe]_H precursor (labeled “as prepared”) exhibited a mixture of IR signals characteristic of the H_ox, the H_redH⁺, and H_ox-CO states with the CO/CN⁻ band positions matching well those reported in the literature for CbHydA1_WT (Fig. 3A and SI Appendix, Fig. S5) (38). Upon exposure to air for 20 min, FTIR spectra of the fusion protein (Fig. 3B) exhibit a complete conversion of the H-cluster to the inactive state (H_inact) identical to that of CbHydA1_WT (38). We were able to reactivate the H-cluster by incubation with 1 atm H₂ after exposure to air (Fig. 3C). We also observed that the CbHydA1-PsaE fusion protein can be inhibited with CO, exhibiting a typical IR spectrum of the H_ox-CO state (Fig. 3D). Finally, as with CbHydA1_WT, the H_ox-CO state of CbHydA1-PsaE fully converts into the H_inact state (Fig. 3E) upon consequent exposure to air. Overall, this set of experiments demonstrates that the fusion of CbHydA1 with PsaE does not noticeably affect the electronic structure of the H-cluster and its interaction with diatomic gases (H₂, O₂, and CO). Furthermore, it appears that CbHydA1-PsaE retains the ability to convert into the H_inact state under aerobic conditions, either from the active or the CO-inhibited states.

Fig. 3.

FTIR spectra of the CbHydA1-PsaE fusion protein under various conditions: (A) as prepared, (B) after exposure to air for 20 min, (C) after exposure to air for 20 min followed by anaerobic reactivation with H₂, (D) after incubation with CO gas for 15 min, and (E) after incubation with CO gas for 15 min followed by exposure to air for 20 min. The shaded areas designate specific states of the H-cluster with color-coding indicated above trace (E). The y-scale is shown as a bar below trace C. Trace D was scaled by 0.5.

The catalytic ability of the CbHydA1-PsaE fusion protein was further analyzed using protein film voltammetry (PFV), in which protein is adsorbed on a rotating disk electrode. In cyclic voltammetry (CV) experiments using a protein-coated working electrode, a positive current indicates H₂ uptake, while a negative current signifies H⁺ reduction (45–47). In the case of CbHydA1_WT, a drop of current at higher potentials is expected due to the enzyme’s conversion to the H_inact state (38). We evaluated the CbHydA1-PsaE fusion protein at various pH values to assess potential differences in the pH dependence of the catalysis from the WT counterpart (Fig. 4). Overall, the pH dependence of CV traces are remarkably similar to the typical response observed for CbHydA1_WT. Importantly, the fusion protein inactivates with a forward sweep to higher potentials and reactivates with a backward sweep. Quite intriguingly, a comparison between CbHydA1_WT and CbHydA1-PsaE shows a shift of the reactivation wave to lower potentials for the latter. This discrepancy is more prominent at higher pH values, with about 95 mV downshift of the mid-point of reactivation [E_switch (48)], but it is still noticeable at lower pH values (60 to 74 mV difference). A lower apparent mid-point potential for the reactivation of CbHydA1-PsaE suggests a higher propensity of CbHydA1-PsaE to remain in the H_inact state relative to CbHydA1_WT (41, 49). This increased tendency to inactivate may contribute to the lower activity of the fusion protein in the in vitro activity assays described above, although further investigation is needed to confirm this point.

Fig. 4.

Room temperature CV measurements of CbHydA1-PsaE fusion protein adsorbed on a rotating disc graphite electrode measured at pH 6.0 (A), pH 7.0 (B), pH 8.0 (C), and pH 9.0 (D). Black dashed lines represent data for CbHydA1_WT obtained under similar conditions. Experimental conditions: scan rate, 10 mV/s; working electrode rotated at 1,000 rpm; atmosphere, 1 atm of H₂.

Binding of CbHydA1-PsaE Fusion Protein to PS I_ΔPsaE.

To assess the ability of the CbHydA1-PsaE fusion protein to bind to the ΔPsaE variant of PS I (PS I_ΔPsaE), and to assess the efficacy of forward electron transfer in this orthogonal pathway, we performed time-resolved optical experiments. The time-resolved absorption at 830 nm reports on the reduction of P₇₀₀^∙ + after a 7-ns, single-turnover actinic laser flash at 532 nm. P₇₀₀^∙ + reduction can occur from either a reduced acceptor in the electron transfer chain (termed charge recombination), or an exogenous electron donor (ascorbate, DCPIP, Cyt c₆, etc.). Charge recombination kinetics from every cofactor in the acceptor chain have been well characterized for PS I_WT and numerous PS I variants (50–53). Deviations from those values serve as a reliable diagnostic tool for understanding changes in cofactor energetics within the complex (30, 51).

The kinetics of P₇₀₀^∙ + decay after an actinic flash in the presence of ascorbate as an electron donor are shown in Fig. 5, along with the respective component analysis. The complete set of P₇₀₀^∙ + decay is shown in supporting information (SI Appendix, Fig. S6). To provide a model-free assessment of the decay traces, we employed inverse Laplace transformation (iLT) analysis by CONTIN (see materials and methods) (54). As expected, PS I_WT displays a largely monophasic decay with a lifetime of ~100 ms, closely matching the previously measured recombination between [F_A/F_B]^– and P₇₀₀^∙ + (SI Appendix, Fig. S6). The PS I_ΔPsaE variant (Fig. 5A) appears to display a largely monophasic P₇₀₀^∙ + reduction but with a noticeably slower decay with a dominant lifetime of 381 ms (58%). This change is consistent with previous reports in an in vivo PsaE deletion mutant of S. 7002 (55).

Fig. 5.

Anaerobic P₇₀₀ ^∙ + reduction kinetics after a 532 nm actinic flash. (A) PS I_ΔPsaE (red) PS I_ΔPsaE:apo-CbHydA1-PsaE (blue); (B) CONTIN iLT of A. (C) PS I_ΔPsaE:holo-CbHydA1-PsaE under 3% H, 97% N₂ (blue), and under 100% He (red). (D) CONTIN iLT of C. P₇₀₀ ^∙ + was monitored by absorbance at 830 nm. Traces in (B) and (D) are normalized to the sum of preexponential factors. Black traces in (A) and (C) are reconstructions of kinetics using (B) and (D), respectively.

Although the origin of the alterations to P₇₀₀^∙ + decay kinetics imposed by the absence of PsaE is not understood (see Discussion), the change is a reliable reporter for the binding of PsaE to PS I. We used this finding to investigate the binding of the CbHydA1-PsaE fusion protein to PS I_ΔPsaE in vitro. First, we performed binding experiments on the apo-form of the fusion protein lacking the H-cluster (apo-CbHydA1-PsaE). Upon anaerobic addition of apo-CbHydA1-PsaE fusion protein to PS I_ΔPsaE in a threefold excess, the P₇₀₀^∙ + reduction kinetics remain mostly monophasic and closely resemble those of PS I_WT (Fig. 5B and SI Appendix, Fig. S6), with lifetimes of 122 ms (~54%) and a smaller contribution from a faster decay kinetic with lifetime of 25 ms (~20%). The absence of the 381 ms reduction phase suggests a near-quantitative binding of the fusion protein to PS I.

Next, we performed similar experiments on the active fusion protein, expecting a significant elongation of the P₇₀₀^∙ + lifetime due to the loss of e^– to H₂ production by the hydrogenase domain. In this case, regeneration of the P₇₀₀ ground state is only possible via reduction by an exogenous electron donor. Slow donors, such as ascorbate, reduce P₇₀₀^∙ + in a matter of several seconds. However, the holo-CbHydA1-PsaE:PS I_ΔPsaE complex (Fig. 5 B and D) kept under a 3% H₂ + 97% N₂ atmosphere exhibited a somewhat faster P₇₀₀^∙ + decay than that of PS I_WT. As shown by iLT, the kinetic decay is heterogeneous, with a dominant process consisting of two phases at 43 ms and 96 ms (together contributing ~68% to the total decay) and a smaller contribution from a slower bi-phasic decay (~23%). This is a stark departure from the apo-CbHydA1-PsaE:PS I_ΔPsaE sample, suggesting an alteration to the charge recombination pathways in the presence of the active hydrogenase domain. Most notably, the CONTIN analysis of this dataset reveals the complete absence of a >1 s charge recombination. We hypothesized that holo-CbHydA1-PsaE acts as a reductant of the FeS clusters in PsaC in the presence of H₂ effectively shutting down forward electron transfer. Consequently, we repeated the same experiments under a pure He atmosphere, which should bias the evidently bidirectional hydrogenase domain (see PFV experiments above) toward H₂ evolution. Under this condition, the PS I-CbHydA1 complexes (Fig. 5 B and D and SI Appendix, Fig. S6) display a noticeably different multiphasic decay of P₇₀₀^∙ +. The kinetics with lifetimes of 28.5 ms (17.5%) and 128 ms (40%) are highly similar to that of apo-CbHydA1-PsaE:PS I_ΔPsaE. In addition, there is another substantial phase at >3 s (28.0%). This long-lived phase indicates P₇₀₀^∙ + reduction via ascorbate, suggesting electron loss due to forward electron transfer from PS I to the [FeFe] hydrogenase domain. Note that we have estimated a 32% yield of the fully assembled H-cluster (Materials and Methods). This fraction of active fusion protein matches well with the 28% contribution of the long-lived phase, suggesting a high efficacy of the forward electron transfer to the H-cluster in the holo-CbHydA1-PsaE:PS I_ΔPsaE nanoconstruct. However, as we will discuss below, such extrapolation must be done with care.

Light-Induced H₂ Generation by the Nanoconstruct.

We next sought to quantify the amount of H₂ generated by the PS I-CbHydA1 complex by conducting gas chromatography (GC) analysis of the sample’s headspace in a capped vial under illumination with a 100 W xenon arc lamp. Before the analysis, the headspace of each sample vial was purged in the dark with ultrapure He (UHP He) for 20 min to remove residual H₂ from the anaerobic chamber atmosphere. The rate obtained for the anaerobically prepared PS I-CbHydA1 nanoconstruct (Fig. 6A and SI Appendix, Fig. S8A) is 84.9 ± 3.1 µmol H₂ mg_chl^–1 h^–1 (4.05 e^– PS I^–1 s^–1). We detected small amounts (0.44 to 0.74 µmol H₂ mg_chl^–1 h^–1) of H₂ in control experiments, including the blank buffer sample. The detection of H₂ in controls is likely due to outgassing residual dissolved H₂ in the sample solution (prepared under 3% H₂/97% N₂ atmosphere) after purging with UHP He. The sample lacking PS I_ΔPsaE is the exception, showing an increased rate equivalent to 1.8 ± 0.4 µmol H₂ mg_chl^–1 h^–1 under identical sample conditions, indicating some residual hydrogenase activity likely due to the presence of a large excess of ascorbate in the sample.

Fig. 6.

Light-induced H₂ generation rates for the CbHydA1-PsaE:PS I_ΔPsaE nanoconstruct under (A) anaerobic and (B) aerobic conditions (dark blue), along with controls for each condition (light blue).

To assay aerobic H₂ generation by the designed nanoconstruct, we repeated the above experiments, but allowing samples to incubate on air for at least 2 h prior to capping the GC vial to accumulate H₂. The rate of light-induced H₂ production for these samples was noticeably lower than the anaerobically prepared samples, 2.2 ± 0.5 µmol H₂ mg_chl^–1 h^–1 (0.10 e^– PS I^–1 s^–1) (Fig. 6B and SI Appendix, Fig. S8B). As expected, we detected no H₂ generated in most control experiments since the sample vials were kept in the air outside an anaerobic chamber with no possibility of H₂ leaching into the sample space. However, the PS I_ΔPsaE sample (“no fusion” in Fig. 6) did show small amounts of H₂ produced when prepared at 12-fold higher concentrations than the anaerobically prepared samples. Under these conditions, the rate of H₂ production was estimated to correspond to 0.053 ± 0.005 µmol H₂ mg_chl^–1 h^–1. It should be noted that minor H₂ production by PS I_ΔPsaE alone parallels observations by Ihara et al., who observed small amounts of H₂ produced from a Cyt c₃/PS I complex (56). Also, in the high-concentration sample of CbHydA1-PsaE fusion protein (no PS I_ΔPsaE), we were able to observe detectable H₂ production that would correspond to 0.0036 ± 0.0001 µmol H₂ mg_chl^–1 h^–1 (calculated with an equimolar fusion:PS I ratio) indicating that a small fraction of hydrogenases remains active, even under aerobic conditions in the presence of ascorbate.

Discussion

Fusion of PsaE to CbHydA1 Changes Its Catalytic Activity.

Our experiments indicate a measurable impact of the fusion of PsaE on the activity of CbHydA1, as evident from two- to threefold lower H₂ oxidation and proton reduction rates than that of a WT protein. In PFV experiments, we observed an increased propensity of the CbHydA1 for inactivation in the PsaE-fusion variant compared to the WT enzyme. The shift of the reactivation CV wave to a lower potential in the CbHydA1-PsaE fusion protein can be a consequence of either a lower mid-point reduction potential of the H_inact state or faster kinetics of inactivation (39, 40). Recently, using EPR, we demonstrated the existence of two structural isoforms of CbHydA1. We proposed that the unique-to-CbHydA1 isoform 1 [EPR signals H_ox(1) and H_ox-CO(1)] is the inactivation-preceding form that relates to the aberrant position of the Cys367 facilitated by a rearrangement of the respective flexible loop (39, 40). The dominance of the isoform 1 in EPR measurements of CO-inhibited CbHydA1-PsaE variant (SI Appendix, Fig. S7) is in line with the observed downshift of the reactivation wave in PFV. Therefore, we consider it likely that adding PsaE to CbHydA1 affects the mobility of the core protein structure around the H-cluster, resulting in an increased propensity for inactivation in the fusion protein and, thus, lower activity. Furthermore, isoform 1 may naturally be less active since it proposedly relates to an off-H⁺-pathway arrangement of Cys367 (40).

We also cannot exclude the possibility that the fusion of PsaE to CbHydA1 could affect the electronic structure of the two accessory [4Fe-4S] clusters (F-clusters) present in the enzyme. Indeed, there is a slight but noticeable shift in the ratio between H₂ evolution and H₂ uptake toward the former in the fusion protein. Also, as-prepared WT and fusion enzymes obtained under identical conditions show somewhat different ratios of H_ox and H_redH⁺ states in IR (SI Appendix, Fig. S5), possibly due to a somewhat different catalytic bias causing a shift of equilibrium between the two states. However, as both catalytic rates are lower in the fusion protein, further investigation into the redox potentials of the F-clusters is needed to understand this effect fully. Such experiments are underway in our laboratory.

Electron Transfer in the Modified PS I Complexes.

Our study of the P₇₀₀^∙ + reduction kinetics after an actinic laser pulse provided invaluable insights into the electron transfer mechanisms with the modified PS I complexes. First, the experiments confirmed the early report by Yu et al. (55) that the absence of PsaE elongates the lifetime of P₇₀₀^∙ +-[F_A/F_B]⁻ charge-separated state. Charge recombination rates are inextricably tied to the redox potential of the bound cofactors, the reorganization energies associated with the site, and the distance between the cofactors. Hence, our working hypothesis is that the lack of a PsaE subunit affects the mobility of the PsaC and PsaD subunits, ultimately resulting in a shift in the redox potential of either F_A and F_B or their distance from F_x and thus causing a change in the lifetime of the charge-separated state. While understanding this effect is outside the scope of this work, it is an important line of research to address in the future, as the structural modifications imposed on PsaE by the fusion with CbHydA1 may pose additional effects on the electron transfer pathways within the PS I complex.

Incubation of PS I_ΔPsaE with apo- and holo-CbHydA1-PsaE results in a shift to PS I_WT-like P₇₀₀^∙ +-[F_A/F_B]^– recombination lifetimes, thus indicating a near-quantitative formation of the PS I_ΔPsaE:CbHydA1-PsaE complex. It is somewhat surprising that when we incubate PS I_ΔPsaE with the apo-form of the fusion protein (F-clusters present, but no active site), there is very little of the long-lived phase, implying almost no electron transfer from PS I to the hydrogenase. However, it is important to note that the resulting data are averaged over 512 traces. Without an electron acceptor such as the active center (H-cluster), the F-clusters will likely remain reduced after the first couple of flashes for the duration of the experiment, assuming a high quantum yield of electron transfer. As a result, forward electron transfer from F_A/F_B to the hydrogenase domain will be effectively blocked. The time-resolved optical experiments on PS I_ΔPsaE:holo-CbHydA1-PsaE nanoconstruct in the presence of H₂ demonstrate that the electron transfer from [F_A/F_B]^– to the F-cluster is not entirely unidirectional. Were that the case, H₂ oxidation by the H-cluster would result in a reduction of the F-cluster but not the [F_A/F_B] pair, and hence, we would expect kinetic traces to be similar to that of PS I_ΔPsaE:apo-CbHydA1-PsaE. To the contrary, our time-resolved experiments showed that under a 3% H₂ atmosphere, the charge recombination is substantially more complex. CONTIN analysis shows a complex multiphasic kinetic profile with the dominant decay phase being faster than that observed in the PS I_ΔPsaE:apo-CbHydA1-PsaE nanoconstruct. Therefore, the data indicate that the H₂-uptake by the hydrogenase domain not only saturates the F-cluster, but also partially reduces the [F_A/F_B] couple. This “backfire” effect could inhibit forward electron transfer within PS I, potentially accelerating charge recombination. In line with this suggestion, the removal of H₂ from the headspace resulted in a substantial elongation of the charge recombination lifetime. In this case, two primary decay phases are the PS I_WT-like lifetime of ~100 ms and a long-lived charge-separated state >3 s. The long-lived phase, whereby P₇₀₀^∙ + is reduced by ascorbate, increases in amplitude from ~4% of the total signal to ~30%. This increase provides compelling evidence for the forward electron transfer from PS I to hydrogenase with the consequent loss of the electron to proton reduction. The remaining 100 ms phase is likely wherein the fusion protein is bound, but forward electron transfer is not occurring in PS I. We note that in our spin-quantification experiments, only a third of CbHydA1-PsaE fusion proteins contain a fully assembled H-cluster (SI Appendix). This alone could be sufficient to explain why forward electron transfer is only 31% efficient. However, whether there is a preference for binding holo-CbHydA1-PsaE rather than apo-CbHydA1-PsaE to PS I_ΔPsaE is not known, as discussed below.

Comparison with Other Nanoconstructs.

The H₂ production rate we observe for our nanoconstruct (84.9 ± 3.1 µmol H₂ mg_chl^–1 h^–1) is similar to other PS I-nanoconstructs tested under roughly comparable conditions (SI Appendix, Table S1) (11–15, 23, 25, 56–63). Due to subpar H-cluster incorporation rate of ~31% for the CbHydA1-PsaE fusion protein, the maximum theoretical rate could be upward of ~250 µmol H₂ mg_chl^–1 h^–1 or ~12 e^–(PS I)^–1s^–1. However, we refrain from using such an extrapolation as many factors can contribute to the overall quantum yield. While data suggest a near-quantitative binding of a fusion protein to PS I_ΔPsaE, it is possible that there is a binding preference of apo- vs holo-proteins, or steric limitations on the trimer affecting the number of CbHydA1 domains able to interact with PsaC while bound. Also, the ratio between forward electron transfer and charge recombination rates may play a role, as well as the backfire effect discussed above.

When discussing related PS I-based nanoconstructs, perhaps the most appropriate comparison would be that with the PsaC-fusion variants of [FeFe] hydrogenases 1 and 2 from C. reinhardtii (30, 31). The overall electron throughput of these systems, as determined experimentally, appears to be 8 to 70 times higher than that observed in our case (SI Appendix, Table S1). However, these measurements were carried out under in vivo conditions where reduction of P₇₀₀^∙ + occurs via natural and efficient electron transport, while we utilized a soluble Cyt c₆ as an electron donor in vitro. Hence, the H₂ production rate is likely rate-limited in the latter by the recycling of P₇₀₀ rather than by the performance of the active chimeric nanoconstructs. Note that cross-linking Cyt c₆ to PS I resulted in a nearly sevenfold increase in H₂ production rate in the PS I-wire-CaHydA1 nanoconstructs reported by the Golbeck group under comparable conditions, with a maximum 70-fold increase achieved by lowering pH (up to 2,200 μmol H₂ mg_Chl^–1 h^–1, SI Appendix, Table S1) (23). It would thus be interesting to perform similar Cyt c₆ cross-linking experiments for our nanoconstructs to estimate the maximum possible H₂ production rate for the CbHydA1-PsaE:PS I_ΔPsaE nanoconstruct.

The Nanoconstruct Can Generate H₂ Under Aerobic Conditions.

Demonstrating detectable H₂ production by the nanoconstruct under aerobic conditions is highly encouraging. It is important to note that there are several avenues by which O₂ can interrupt energy and electron transfer both within PS I and between PS I and the hydrogenase. Within PS I complex, O₂ can react with Chl triplet states during light harvesting (64). O₂ can also accept low-potential electrons from [F_A/F_B]^– cluster pair and even from the A₁ phylloquinones (65–67). A gradual reduction in O₂ concentration in the headspace (SI Appendix, Fig. S8B) supports this notion. Therefore, electron transfer from [F_A/F_B]^– to CbHydA1 is in competition not only with charge recombination but also with the reduction of molecular oxygen. For CbHydA1, O₂ will also functionally inactivate the active site, requiring reducing equivalents from PS I for periodic reactivation. Given this combination of factors, it is not surprising that the rates of H₂ production by the CbHydA1-PsaE:PS I nanoconstruct are lower than for anaerobically prepared samples. It may be possible to improve the efficiency of the forward electron transfer by varying the length and rigidity of the linker group to increase both protein activation, specific activity, and electron transfer from PS I. Further investigation into the interaction of O₂ with the parts of the nanoconstructs will be required to parse out the contributing factors and devise a mitigation strategy. Nonetheless, even with reduced rates in the presence of O₂, the CbHydA1-PsaE:PS I nanoconstruct shows promise for an in vivo system that functions in the presence of constantly changing levels of O₂.

In summary, we have successfully generated a unique chimeric protein nanoconstruct through the fusion of a 7.6 kDa stromal subunit of PS I (PsaE) from the cyanobacterium Synechococcus sp. PCC 7002 onto the C terminus of an oxygen-tolerant [FeFe] hydrogenase from C. beijerinckii (CbHydA1) via a flexible [GGS]₄ linker group. We show that CbHydA1 can be synthetically activated in vitro and retain native-like bidirectional hydrogenase activity. Our transient absorption studies demonstrate that the PsaE substituent provides a mode for selective and nearly quantitative binding of the fusion protein to available PS I_ΔPsaE-cores. We were able to detect light-induced H₂ evolution even in an aerobic environment. Our findings provide confidence in PsaE as a viable scaffold for the binding of exogenous proteins to PS I cores. The detailed investigation of the H₂-producing chimeric nanoconstruct presented here establishes an important basis for future engineering in vivo H₂-generating systems that can function within O₂-evolving photosynthetic pathways.

Materials and Methods

Preparation of Protein Samples.

CbHydA1-PsaE and CbHydA1_WT were expressed and purified similarly to methods found in the literature (38). In short, the protein was expressed in Escherichia coli BL21 (DE3) ΔiscR strain (The Pennsylvania State University, Golbeck lab) (68). Luria–Bertani (LB) growth medium (3L) was supplemented with 100 mM MOPS pH 7.4, 2 mM ferric ammonium citrate, 2 mM L-cysteine, and 100 µg/mL ampicillin. Cultures were first grown aerobically at 37 °C to an O.D.₆₀₀ of 0.6 ± 0.2. Then, the cultures were placed on ice to halt growth. While chilling, the cultures were supplemented with 0.5% glucose and 25 mM sodium fumarate. After chilled, the cultures were brought inside an anaerobic chamber. Cultures were induced anaerobically with the addition of 1.5 mM IPTG and allowed to express for 20 h at 20 °C. The cells were lysed via anaerobic sonication for 1 h with continuous stirring. After centrifugation, the CbHydA1-PsaE fusion protein was purified using a Co-NTA affinity chromatography (SI Appendix, Fig. S2). The purified protein was buffer-exchanged using a PD-10 column into a buffer containing 100 mM HEPES, pH 7.5, 5 mM dithiothreitol, and 300 mM KCl. Protein yields averaged ~3.5 mg per L of cell growth. Reconstitution of the iron-sulfur clusters was performed through a slow addition of 15-fold excess FeCl₃·6H₂O and Na₂S·9H₂O, consecutively, over the course of 30 min and 60 min, respectively. Synthetic maturation using a [2Fe]_H synthon Fe₂[μ-S₂C₂H₄NH](CO)₄(CN)₂ was performed according to previously published procedures (38). PS I_WT and the PS I_ΔPsaE variant were isolated from variants of Synechococcus sp. PCC 7002 using previously reported methods (SI Appendix, Fig. S10) (55, 69), with some modifications as described in detail in SI Appendix.

Quantification of H-Cluster Incorporation.

To account for a potentially inefficient H-cluster assembly during synthetic maturation, we estimated concentrations of the fully assembled H-cluster by incubating samples with CO and subsequently quantifying the amount of the H_ox-CO state by CW EPR spectroscopy. The fusion protein was incubated under CO gas for 30 min before analysis via CW EPR (SI Appendix, Fig. S7). We used FTIR to confirm the complete conversion of the H-cluster to the H_ox-CO state. The power dependence of the double integral of the H_ox-CO EPR signals was compared to that of the standard (321 µM MbN₃, SI Appendix, Fig. S9), from which the concentration of the H-cluster in the sample was extracted. This concentration was then compared against the concentration of protein determined by the Bradford assay to calculate the percent of H-cluster incorporation. For the samples used in the experiments presented here, H-cluster synthesis efficiency was estimated to be 63% and 31% for CbHydA1_WT and CbHydA1-PsaE, respectively. These correction factors were used in all reported hydrogenase activity values.

EPR and FTIR Spectroscopy.

EPR spectra reported were obtained using the Magnettech MS5000 X-band spectrometer augmented with a liquid He flow cryostat (Oxford Instruments ESR900). For the measurements, 250 μL of samples were placed in a 4 mm OD clear fused quartz tube.

FTIR measurements were made on a Nicolet is50 FTIR (Thermos Scientific) spectrometer equipped with an LN₂-cooled MCT-A detector with a 3.6 µm long-pass filter (Edmund Optics). Experiments were conducted in a home-built cell consisting of two 32 mm OD, 2 mm thick CaF₂ windows, with a 40 µm PTFE spacer. All samples were prepared and measured at room temperature. Datasets were collected with 2,000 scans with a resolution of 2 cm^–1. All EPR and FTIR spectra were processed in MATLAB using home-build software (Kazan viewer, https://github.com/AlexeySilakov/KazanViewer/).

PFV.

PFV measurements were made in an anaerobic chamber (Coy Laboratory Products, Inc.) with a WaveNow potentiostat (Pine Research), with a variable-speed rotator (Pine Research). Measurements were performed using a home-built rotating disc electrode using a 5 mm OD edge-plane pyrolytic graphite PG-HT encased in a 12 mm OD PEEK cylinder. The surface of the electrode was first polished using 2000 grit sandpaper and then sonicated in 50% ethanol. The reference electrode used was a 3 M NaCl Ag/AgCl electrode (BASi MF-2052). The counter electrode was a Pt wire wound into a spiral (Sigma Aldrich, 0.25 mm OD). Calibration of the reference electrode was performed using a methyl viologen solution, E₀(MV²⁺/MV^∙ +) = −445 mV vs. NHE.

Time-Resolved Optical Spectroscopy.

A detailed description of the home-built set-up is provided elsewhere (52). An actinic laser flash at 532 nm with a pulse width of 7 ns (full-width half-maximum) generates a single turnover event in PS I. Changes in absorbance were monitored via an 830 nm laser diode (CrystaLaser, model number DL830-100-O) that is split into measuring and reference beams. The difference signal was amplified with a 200 MHz bandwidth differential amplifier (Thor Labs PDB460A), which was subsequently processed using a 1 GHz bandwidth, 8-bit, 2 GS/s, PCI card (National Instruments NI-5154) controlled by home-built software. Data analysis was performed using home-built software in Python. Regularized iLTs of kinetic decay traces were performed using CONTIN software (54). Regularization order (mode) was set to NORDER = 2, and the regularization parameter was set to ALPST = 2E-5; the program was restricted to finding non-negative solutions (NONNEG = 1) as no subsequent oxidation of P₇₀₀ is expected after the initial generation of the charge-separated state under conditions used. All samples contained a Chl concentration of 100 µg/mL, corresponding to a concentration of 1.16 µM PS I, Tris-HCl (pH 8.0), 500 µM sodium ascorbate, and 0.05% (w/v) β-DDM. More details on sample preparation can be found in supplemental information.

Light-Induced H₂ Generation.

CbHydA1-PsaE:PS I_ΔPsaE nanoconstructs were generated anaerobically through a 30-min incubation of PS I_ΔPsaE with the CbHydA1-PsaE fusion protein in a ratio of 1:3, PS I:CbHydA1-PsaE. Following incubation, the solution was washed five times with 50 mM Tris pH 8.0, 50 mM NaCl, and 0.05% β-DDM through a 100 kDa Centricon centrifugal filter (Amicon Ultra, 0.5 mL). H₂ production rates were determined in accordance with previously published methods (23, 57, 58). 500 μL of the assembled PS I–CbHydA1 nanoconstruct solution was placed into a 8 mL glass vial, capped with gray-butyl septa, and kept in the dark. A sacrificial electron donor (100 mM ascorbate) and 20 µM Cyt c₆ (SI Appendix, Fig. S11) were added for reduction of P₇₀₀⁺. For anaerobic experiments, the vial was purged with ultrapure He for 20 min to remove residual anaerobic chamber dihydrogen. Samples were measured immediately before and periodically during illumination. For aerobic experiments, samples were equilibrated on air before capping for at least 2 h and the headspace was not displaced. The headspace was analyzed with a Shimadzu GC-2010 Plus gas chromatograph equipped with a RT-Msieve 5A column (Restek) and a barrier ionization detector (BID 2010 Plus), using ultrapure He as the carrier gas with a linear velocity of 42.5 cm/s. The injection temperature was 150°C, the column temperature was 35°C, and the BID temperature was 280°C.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. Original datasets are available on Figshare https://doi.org/10.6084/m9.figshare.26405266) (70).