Membrane Protein Modiﬁed Electrodes in Bioelectrocatalysis

: Transmembrane proteins involved in metabolic redox reactions and photosynthesis catalyse a plethora of key energy-conversion processes and are thus of great interest for bioelectrocatalysis-based applications. The development of membrane protein modiﬁed electrodes has made it possible to e ﬃ ciently exchange electrons between proteins and electrodes, allowing mechanistic studies and potentially applications in biofuels generation and energy conversion. Here, we summarise the most common electrode modiﬁcation and their characterisation techniques for membrane proteins involved in biofuels conversion and semi-artiﬁcial photosynthesis. We discuss the challenges of applications of membrane protein modiﬁed electrodes for bioelectrocatalysis and comment on emerging methods and future directions, including recent advances in membrane protein reconstitution strategies and the development of microbial electrosynthesis and whole-cell semi-artiﬁcial photosynthesis.


Introduction
Membrane proteins constitute 20-30% of all proteins encoded by both prokaryotic and eukaryotic cells. They perform a wide variety of functions, including material transport, signal transduction, catalysis, proton and electron transport ( Figure 1) [1]. They are also key to a number of earth's most fundamental reactions, such as respiration and photosynthesis [2,3]. Redox enzymes in the respiratory chain catalyse a variety of fundamental processes for energy conversion and fuel production, including H 2 oxidation, O 2 reduction, and carbon and nitrogen cycling. Membrane proteins that are involved in the light reaction of photosynthesis harvest light and facilitate electron transfer essential for solar energy conversion. The amphiphilic nature of membrane proteins makes them difficult to isolate, study and manipulate. Despite these challenges, membrane proteins have been widely advocated and studied for applications in bioelectrocatalysis, such as biofuel cells [4] and semi-artificial photosynthesis [5].
Here, we will review electrochemical studies of membrane proteins with the view to using these systems for bioelectrocatalysis. To aid discussion later on, we will briefly introduce a small selection of membrane enzymes active in bioenergy conversion, although this is far from a comprehensive list. We will then summarise the main strategies to immobilise membrane proteins on electrodes and discuss common techniques used to characterise these electrodes, including electrochemistry, spectroscopy, spectroelectrochemistry, microscopy and quartz crystal microbalance. Finally, some critical application challenges and potential future research directions will be highlighted that might find application in bioelectrocatalysis. Specifically, we will focus on two emerging directions. One is the reconstitution of membrane proteins into hybrid vesicles to extend their functional lifetime. The other is the use of microorganisms for microbial electrosynthesis and semi-artificial photosynthesis.

Membrane Enzymes in Biofuel Conversion
Nature offers highly specialised enzyme machineries that can be exploited for (bio)fuel conversion. Among them, membrane-bound hydrogenases, found in many bacteria, archaea and lower eukaryotes, are metalloenzymes capable of catalysing the reversible oxidation of H2 to protons and electrons [6]. Depending on the metal located in the active site, three phylogenetically unrelated classes can be identified: [NiFe], [FeFe] and, less common, [Fe] hydrogenases. Most hydrogenases are rapidly and almost completely inactivated by O2 [7]. However, aerobic or facultative aerobic H2oxidising bacteria have a particular subtype of O2-tolerant [NiFe] hydrogenase, which withstands the presence of O2 and are membrane-bound hydrogenases (MBHs) [8]. The most studied O2-tolerant [NiFe] hydrogenases are found in Ralstonia eutropha, Ralstonia metallidurans, Aquifex aeolicus, Hydrogenovibrio marinus and Escherichia coli. MBHs are multimeric proteins with one large subunit and one small subunit, bound to the periplasmic side of the cytoplasmic membrane through a transmembrane protein (b-type cytochrome) [8]. The [NiFe] active site, with the same configuration as other [NiFe] hydrogenases, is deeply buried in the large subunit. MBHs couple oxidation of hydrogen in the large subunit to the reduction of either menaquinone-7 or ubiquinone-8 in the b-type cytochrome in the membrane.
In the respiratory chain, terminal oxidases catalyse the reduction of molecular oxygen to water without formation of reactive oxygen species (ROS). They oxidise quinones (ubiquinone and menaquinone oxidases) or cytochromes (cytochrome c oxidases) and can be classed into haem-copper oxidases [9], bd oxidases [10] or alternative oxidases [11]. For instance, cytochrome bo3 enzymes are haem-copper oxidases in bacteria such as E. coli and couple oxidation of ubiquinol-8 to the reduction of oxygen [12]. Cytochrome bd is a quinol-dependent terminal oxidase found exclusively in prokaryotes and is structurally unrelated to haem-copper oxidases [13].
Nitrate reductases are molybdoenzymes capable of reducing nitrate (NO3 − ) to nitrite (NO2 − ). This enzyme family can be found in eukaryotic and prokaryotic cells. Eukaryotic nitrate reductases are present in plants, algae and fungi, and are involved in the assimilation of nitrate. Prokaryotic nitrate reductases are classified into three classes: assimilatory nitrate reductases (Nas), periplasmic nitrate reductases (Nap) and respiratory nitrate reductase (Nar). The latter are transmembrane enzymes which use nitrate as the electron acceptor of an anaerobic respiratory chain [14]. Facultative anaerobic bacteria use these alternative respiratory enzymes in oxygen depleted environments, replacing oxygen with different electron acceptors [15]. For instance, under anaerobic conditions, E. coli expresses a respiratory membrane-bound NarGHI coupled with a membrane-bound formate dehydrogenases (FDH-N). These two membrane enzymes form the supermolecular formate:nitrate oxidoreductase system. FDH-N oxidises formate to CO2, after which electrons are transferred to nitrate with the formation of a proton-motive force across the cytoplasmic membrane [16].

Membrane Enzymes in Biofuel Conversion
Nature offers highly specialised enzyme machineries that can be exploited for (bio)fuel conversion. Among them, membrane-bound hydrogenases, found in many bacteria, archaea and lower eukaryotes, are metalloenzymes capable of catalysing the reversible oxidation of H 2 to protons and electrons [6]. Depending on the metal located in the active site, three phylogenetically unrelated classes can be identified: [NiFe], [FeFe] and, less common, [Fe] hydrogenases. Most hydrogenases are rapidly and almost completely inactivated by O 2 [7]. However, aerobic or facultative aerobic H 2 -oxidising bacteria have a particular subtype of O 2 -tolerant [NiFe] hydrogenase, which withstands the presence of O 2 and are membrane-bound hydrogenases (MBHs) [8]. The most studied O 2 -tolerant [NiFe] hydrogenases are found in Ralstonia eutropha, Ralstonia metallidurans, Aquifex aeolicus, Hydrogenovibrio marinus and Escherichia coli. MBHs are multimeric proteins with one large subunit and one small subunit, bound to the periplasmic side of the cytoplasmic membrane through a transmembrane protein (b-type cytochrome) [8]. The [NiFe] active site, with the same configuration as other [NiFe] hydrogenases, is deeply buried in the large subunit. MBHs couple oxidation of hydrogen in the large subunit to the reduction of either menaquinone-7 or ubiquinone-8 in the b-type cytochrome in the membrane.
In the respiratory chain, terminal oxidases catalyse the reduction of molecular oxygen to water without formation of reactive oxygen species (ROS). They oxidise quinones (ubiquinone and menaquinone oxidases) or cytochromes (cytochrome c oxidases) and can be classed into haem-copper oxidases [9], bd oxidases [10] or alternative oxidases [11]. For instance, cytochrome bo 3 enzymes are haem-copper oxidases in bacteria such as E. coli and couple oxidation of ubiquinol-8 to the reduction of oxygen [12]. Cytochrome bd is a quinol-dependent terminal oxidase found exclusively in prokaryotes and is structurally unrelated to haem-copper oxidases [13].
This enzyme family can be found in eukaryotic and prokaryotic cells. Eukaryotic nitrate reductases are present in plants, algae and fungi, and are involved in the assimilation of nitrate. Prokaryotic nitrate reductases are classified into three classes: assimilatory nitrate reductases (Nas), periplasmic nitrate reductases (Nap) and respiratory nitrate reductase (Nar). The latter are transmembrane enzymes which use nitrate as the electron acceptor of an anaerobic respiratory chain [14]. Facultative anaerobic bacteria use these alternative respiratory enzymes in oxygen depleted environments, replacing oxygen with different electron acceptors [15]. For instance, under anaerobic conditions, E. coli expresses a respiratory membrane-bound NarGHI coupled with a membrane-bound formate dehydrogenases (FDH-N). These two membrane enzymes form the supermolecular formate:nitrate oxidoreductase system. FDH-N oxidises formate to CO 2 , after which electrons are transferred to nitrate with the formation of a proton-motive force across the cytoplasmic membrane [16].

Membrane Proteins in Photosynthesis
Photosynthesis is the natural process by which photosynthetic organisms convert light energy into chemical forms. The key steps, which include light harvesting, charge separation and electron transport, occur in the membrane. For oxygenic photosynthesis in algae, higher plants and cyanobacteria, two photosynthetic complexes are involved: photosystem I (PSI) and photosystem II (PSII) [17], each with a peripheral antenna system: light harvesting complex I (LH I) for PSI and light harvesting complex II (LH II) for PSII [18]. Anoxygenic photosynthesis in bacteria, such as purple bacteria, is conducted by just one type of photosystem [19]. There are two types of light-harvesting complexes in most purple bacteria. The light-harvesting complex I and reaction centre forms the RC-LH I complex which is surrounded by multiple light-harvesting complex II [20]. In both oxygenic and anoxygenic photosynthesis, light-harvesting complexes exert the vital function to effectively absorb light and transfer the energy to the reaction centres [21]. Light-harvesting complexes absorb a limited spectral range depending on their natural pigment. The light-induced charge separation occurs at the reaction centres: PSII catalyses the light-driven water oxidation to reduce quinones, while the reaction centre of purple bacteria only catalyses the reduction of quinones (part of the cyclic electron transfer process). PSI is not catalytically active but instead serves as an "electron pump" which can be potentially coupled to the other redox catalysts. Electrodes interfaced with photosynthetic proteins have found broad applications in biosensors, biophotovoltaic cells and solar fuels generation [5,22,23].

Membrane Protein Electrode Design
Redox enzymes are molecular electrocatalysts that can function either in solution or while immobilised on an electrode surface. Immobilising enzymes on the electrodes has several advantages for electrocatalysis and this review will consider only immobilised systems. It is important that the enzymes retain their structural integrity and catalytic activity upon immobilisation [24]. Unlike soluble proteins, transmembrane proteins exist within lipid membranes and, consequently, are less stable in an aqueous environment where the amphiphilic properties of membrane proteins can lead to aggregation and denaturation. Therefore, it is challenging to retain the stability and function of membrane proteins in in vitro studies [25]. Suitable detergents or mixed lipid/detergent systems are needed to maintain an amphiphilic environment surrounding the membrane protein and mimic the original membrane conditions [26,27]. Alternatively, for some multisubunit, heteromeric membrane proteins the solubility problem can be circumvented by either purifying only the soluble subunits or engineering the protein to express only the soluble subunits [28]. For instance, the large and small subunits of membrane-bound hydrogenases can be separated from the transmembrane "anchor" subunit (b-type cytochrome). More sophisticated approaches have been developed for the mammalian membrane-bound cytochrome P450 that has been bioengineered without the hydrophobic membrane anchor domain [29]. However, similar methods are not always applicable for other proteins, either because the catalytic centre is located in a polytopic membrane subunit or because the water-soluble subunits on their own are not stable. It is worth mentioning that the activity of certain membrane proteins can be dependent on the presence of particular annular lipids [30] and therefore the strategies that improve the solubility by removing the need of the membrane environment can have some potential downsides.
Various immobilisation methods have been developed to achieve efficient electron transfer between a membrane protein and an electrode. There are several aspects to consider when designing and assembling (membrane) protein modified electrodes for bioelectrocatalysis: (1) orientation of the protein on the electrode surface, (2) preservation of the protein structural integrity and functionality, (3) low overpotential to minimise the energy loss, (4) protein loading of the electrode [24,31,32]. Some methods developed for electrocatalysis using soluble proteins can be adapted to detergent solubilised membrane proteins. An alternative strategy to the use of detergent is represented by reconstitution of membrane proteins within a lipid membrane on the electrode surface and several strategies have been developed to achieve this. We will discuss the most commonly used immobilisation methods.

Unmodified Electrode
Detergent solubilised membrane proteins can be absorbed directly on carbon, metal or semi-conductor electrodes in an approach known as protein film electrochemistry ( Figure 2a) [33]. A frequently used electrode material is pyrolytic graphite "edge" (PGE) which has a rough, negatively charged surface, which can be tailored with polycations such as polymixin B or poly-l-lysine if required. Respiratory nitrate reductase (NarGHI) isolated from Paracoccus pantotrophus in n-dodecyl β-D-maltoside (DDM) buffer was adsorbed onto PGE electrodes and showed direct electron transfer (DET) with the electrode [34]. The catalytic activity of NarGHI from E. coli was studied over a wide pH range (5 < pH < 9), nitrate concentrations and in the presence of inhibitor [35]. Similarly, purple bacteria RC-LH I was solubilised in 0.1% lauryldimethylamine N-oxide (LDAO) buffer and immobilised directly on a bare Au electrode [36]. RC-LH I did not exhibit DET, but showed photocurrents (~64 nA/cm 2 ) when ubiquinone-0 and cytochrome c were present in solution, indicating that electron transfer needs to be mediated by small redox compounds in this system (mediated electron transfer, MET). Like water-soluble proteins, membrane proteins on bare electrodes often lack a specific orientation and this can impede efficient electron transfer between electrode and proteins. Making use of the amphiphilic nature of membrane proteins, a densely packed protein monolayer with defined orientation can be formed on electrode surfaces by Langmuir−Blodgett deposition and this was shown to increase the electron transfer efficiency [37][38][39][40][41]. An example of this strategy can be found in Kamran et al. where detergent solubilised-RC-LH I was first pre-assembled in a monolayer at a water-air interface and then transferred onto a gold-coated electrode, reaching photocurrent values of~45 µA/cm 2 [37].

SAM Modified Electrode
An electrode surface can be modified to promote protein immobilisation and control the orientation of the protein on the surface. A common method to functionalise metallic electrode surfaces, in particular gold electrodes, is to form a self-assembly monolayer (SAM) of thiols (e.g., alkanethiols) ( Figure 2b). By changing the length of the thiol compound (e.g., the alkyl chain), the distance between the protein and electrode can be controlled. More importantly, by changing the terminal group of the SAM (e.g., n-hydroxy-alkanethiol or n-amino-alkanethiol), the surface chemistry of the electrode can be controlled. As protein binding to surfaces is typically governed by van-der-Waals and electrostatic interactions, the surface chemistry strongly influences the orientation of proteins on the surface [42,43].
Besides tuning the chemistry of the electrode surface, proteins can be genetically engineered to control their orientation on the electrode through affinity interactions. RCs from Rhodobacter sphaeroides in LDAO detergent have been immobilised on SAM-modified electrodes terminated by Ni-NTA groups. By genetically engineering a poly-histidine tag (His7) at the C-terminus of the M-subunit of the RC, the primary donor of RC was positioned to face the electrode [44]. The histidine tag can also be engineered on H subunit of RCs to achieve the opposite orientation [45]. The Ni-NTA modified gold electrodes have also been used to immobilise His-tagged PSII solubilised in DDM buffer [46]. A different strategy that has been explored is to modify the gold electrode surface with an amine terminated SAM for further reaction with terephthaldialdehyde (TPDA). The TPDA modified SAM reacts with lysine residues from PSI to form covalent imine bonds [47]. Such an approach does not create the same orientational control compared to the His-tag/NTA coupling. Irrespective of selective orientation on the surface, mediators are often required for efficient electron exchange with membrane proteins. Small proteins like cytochrome c are widely used as mediators for PSI and purple bacteria RC [48]. Water-soluble redox mediators such as 2,6-dichloro-1,4-benzoquinone (PSII), 2,6-dichlorophenolindophenol (PSI), sodium ascorbate (PSI) and ferricyanide (PSI) are also commonly used to facilitate efficient electron transfer between the protein and electrode [49]. electrode and proteins. Making use of the amphiphilic nature of membrane proteins, a densely packed protein monolayer with defined orientation can be formed on electrode surfaces by Langmuir−Blodgett deposition and this was shown to increase the electron transfer efficiency [37][38][39][40][41]. An example of this strategy can be found in Kamran et al. where detergent solubilised-RC-LH I was first pre-assembled in a monolayer at a water-air interface and then transferred onto a goldcoated electrode, reaching photocurrent values of ~45 μA/cm 2 [37].

Nanoparticle Modified Electrode
Nanoparticles (NPs) have been shown to enhance interfacial electron transfer between a variety of electrodes and proteins [50] and gold NP-modified electrodes have been successfully used to study membrane proteins (Figure 2c), such as terminal oxidases: E. coli cytochrome bd (a quinol oxidase) [51], Paracoccus denitrificans cytochrome aa 3 [52] and Thermus thermophilus cytochrome ba 3 [53] (two cytochrome c oxidases). The use of NPs facilitated DET between the electrode and protein, enabling more in-depth studies on enzyme activity. The size of the NPs significantly affects the electron transfer rates and smaller particles reduced the requirement of overpotential for O 2 reduction activity by E. coli cytochrome bo 3 [54]. The group of Hellwig has recently studied the role of the surface charge of thiol-modified gold NPs, the length of the thiols and the effect of phospholipid composition on the interaction and DET between NP and the membrane enzyme, cytochrome bd [55]. Both cytochromes bo 3 and bd used in the aforementioned studies were isolated and stabilised in DDM buffer. Besides gold NP, various other particles have been used for protein immobilisation [56]. The Armstrong group used graphite microparticles to pair electron donor and acceptor membrane enzymes [57].
[NiFe] membrane-bound hydrogenase was reconstituted with either E. coli nitrate reductase (NarGHI) or E. coli fumarate reductase (FrdAB; not a membrane protein) on single microparticles; both systems catalysed the reductions of nitrate or fumarate, respectively, by hydrogen. In 2015, Duca et al. [58] demonstrated a cascade electrochemical reduction of nitrate to ammonia by immobilising E. coli respiratory nitrate reductase (NarGHI) on an electrode (in polyoxyethylene 9-dodecyl ether detergent) with Pt or Rh nanoparticles.

Redox Polymers
Redox polymers are widely used for wiring redox proteins on electrode surface ( Figure 2d) [59]. Redox polymers act simultaneously as immobilisation matrix and as redox mediators. High coverages of proteins can be achieved and the ubiquitous localisation of redox mediators bound to the polymer matrix negates the need to control protein orientation. These polymer-bound redox mediators overcome mass transport limitations, typically observed with freely diffusing redox mediators. The chemical and physical properties of the redox polymers can be tailored by tuning the polymer backbone and the redox mediators. The Schuhmann group has extensively explored Os redox polymers to "wire" DDM-solubilised PSI [60] and PSII [61] on electrode surfaces for photocurrent generation. A two-compartment cell with a PSII photoanode and a PSI photocathode was constructed to mimic the Z-scheme of natural photosynthesis and an open-circuit voltage (OCV) of 90 mV was achieved [62]. By tuning the redox potential of Os complexes in the redox polymers to match the redox sites of the proteins (PSII and PSI), the OCV could be increased from 90 mV to 372 mV [63]. Interestingly, by exploiting the pH-dependent properties of Os-modified polymer, the group of Plumeré improved the interfacial electron transfer rates for the PSI photocathode, even exceeding rates observed in natural photosynthesis [64].

Membrane Modified Electrode
Because of the nature and functions of the membrane proteins, it is important that their assembly on electrodes preserves their structural integrity and functionality. In the examples provided so far, detergents are used to retain stability of the immobilised membrane enzymes. Although detergents are required to accommodate the amphiphilic nature of membrane proteins, they also have adverse effects to membrane protein stability and are likely to influence the electrode-protein interaction. To mimic the native environment of membrane proteins, electrodes have been modified with model membranes, as reviewed previously [65][66][67]. These membrane-modified electrodes can be categorised into hybrid bilayer lipid membrane (hBLM) system ( In hBLM systems a phospholipid layer is absorbed onto a self-assembled monolayer (SAM) of alkylthiols ( Figure 2e). This strategy was used by the Hawkridge group to study cytochrome c oxidase immobilised on gold electrode [68,69]. In sBLMs, a lipid bilayer is non-covalently bound to the electrode surface ( Figure 2f). Photosynthetic reaction centres including Rhodobacter sphaeroides RC [70,71], spinach photosystem I [72] and photosystem II [73] have been integrated into sBLMs on pyrolytic graphite. Electron transfer between electrode and the aforementioned proteins was achieved and showed well-defined peaks in voltammetry which corresponded to the redox sites of the proteins. Noji et al. incorporated Rhodopseudomonas palustris RC-LH I into sBLMs on an ITO electrode. Anionic phospholipid like phosphatidylglycerol (PG) were shown to stabilise the charge-separated state of RC-LH I and enhance the photocurrent [74]. In tBLM systems, the membrane is 'tethered' to the lipid-modified electrode surface via a linker ( Figure 2g). Cytochrome bo 3 has been incorporated into tBLMs and was shown to retain its catalytic activity [75]. We previously included the membrane-bound [NiFe]-hydrogenases (MBH) from R. eutropha into a tBLM approach and used electrochemistry to study the activity of the entire heterotrimeric membrane-bound form of the enzyme [76]. In 2016, Pelster and Minteer isolated mitochondrial electron transport chain (ECT) enzymes, reconstituted them into liposomes and immobilised them onto a gold electrode in a tBLM. The authors used this reconstructed mitochondrial inner membrane biomimic to show the interdependence of the different complexes on bioelectrocatalytic activity [77]. In ptBLMs, the membrane protein is first anchored to an electrode via a His-tag/NTA interaction and subsequently reconstituted into a lipid membrane (Figure 2h). Ataka et al. have immobilised the Rhodobacter sphaeroides cytochrome aa 3 (a cytochrome c oxidase) by using a ptBLM system [78][79][80]. Two opposite protein orientations were compared, by varying the position of the His-tag. Cytochrome c bound and exchanged electrons with cytochrome c oxidase only when Catalysts 2020, 10, 1427 7 of 29 the latter was orientated with its subunit II (the binding domain for cytochrome c) facing the bulk aqueous medium.

Multi-Layer Assembly, Multilayered Lipid Membrane Stacks, 3D Structure Electrode
An important aspect to improve the performance of bioelectrocatalysis is catalyst loading on the electrode surface. For membrane proteins, this can be achieved via two alternative strategies: multi-layer (membrane) assembly (Figure 2d,i,j) and/or 3D electrode structure (Figure 2k). For a more detailed discussion on layer-by-layer assembly approach for redox proteins, we also refer to [65,81]. Multilayer immobilisation within a redox polymer matrix was already discussed in Section 3.4. As an example of a layer-by-layer approach (Figure 2i), PSI has been co-assembled with cytochrome c as mediator using DNA as an anionic polymer [82]. The purple bacteria RC was also co-assembled into a multilayer architecture by alternating layers of RC with a cationic polymer poly(dimethyldiallylammonium chloride) (PDDA) [83,84]. Our group used poly-l-lysine (PLL) as an electrostatic polymer to construct multilayered lipid membrane stacks ( Figure 2j) [85,86]. Two membrane proteins, E. coli cytochrome bo 3 and R. eutropha MBH, were incorporated into these lipid membranes stacks. Lipophilic quinones, the natural substrates of cytochrome bo 3 and MBH, diffuse freely between the multilayered membranes and mediated electron transfer between the proteins and the electrode. Catalytic activity was shown to increase linearly with the number of membrane layers for at least up to 5 layers [86].
The high surface area of so-called 3D electrodes can be used to immobilise proteins at high loading [87]. Mesoporous metal electrodes have been shown to increase protein loading compared to planar electrodes, e.g., rough silver electrode for bacteria RC [88] and nanoporous gold electrode for PSI [89]. A mesoporous WO 3 -TiO 2 film electrode has been reported for the entrapment for bacterial RCs [90] and mesoporous indium tin oxide (ITO) electrodes have been used to immobilise PSII [91]. The ITO electrode can be modified with SAM to covalently bind and orientate PSII with the electron acceptor side of towards the electrode surface, enhancing electron transfer kinetics and electrode stability [92]. A hierarchically structured, inverse opal, mesoporous (IO-meso) ITO electrode was later developed to provide even larger surface areas [93]. These electrodes were combined with redox polymers to electrically wire PSII with the 3D structure of IO-meso ITO, yielding photocurrent densities of up to~410 µA cm −2 [94]. Mesoporous ITO electrodes were also applied to co-immobilise cytochrome c and PSI and photocurrent densities >150 µA·cm −2 were achieved [95]. The effect of pore size was studied by comparing photocurrents between mesoporous (20-100 nm) and macroporous (5 µm) electrodes using 2,6-dichlorophenolindophenol (DCPIP) and ascorbate as redox mediator for PSI. The macroporous electrode showed three times higher photocurrent than the mesoporous electrode [96]. The authors observed that the macroporous electrode increased the active surface area twice compared to the mesoporous electrode with the same PSI mass loading. They concluded that the increase in photocurrent was due to multilayers of PSI deposited along pore walls and the macropores enhanced the MET within a single pore.

Methods to Characterise Membrane Protein Modified Electrode
The characterisation of electrodes and protein films is very often achieved through a combination of powerful and advanced techniques and their combination with electrochemical tools. In this section, we will focus on the most commonly used techniques to study membrane proteins on electrode surfaces and we will report some representative works.

Electrochemical Methods
Protein film electrochemistry (PFE) is a well-established and important technique to study protein modified electrodes and has proven powerful to probe the thermodynamic, kinetic and catalytic properties of redox proteins, typically with voltammetry [97][98][99]. PFE is also compatible with membrane-modified electrode and allows to probe the membrane proteins within detergent solutions, Catalysts 2020, 10, 1427 8 of 29 polymer matrices or model-membrane environment [100,101]. We refer to the following reviews for a detailed description of PFE methods [102][103][104][105][106].
Besides voltammetry, impedance spectroscopy has been used to study the electrode-membrane protein interface. The quality and the structure of electrode surfaces modified with planar lipid membranes (Figure 2e-h) are particularly suitable for investigation by electrochemical impedance spectroscopy (EIS). EIS can monitor the resistance and the capacitance of a planar membrane covering the electrode. For ideal planar lipid bilayers on the electrode, the capacitance value should be in the range of~0.5 µF·cm −2 with a resistance of >MΩ cm 2 [107]. Disorders and defects of these bilayers will result in lower resistance and/or higher capacitance values. For instance, we monitored the formation of multilayer membrane stacks (Figure 2j) by EIS [86] and observed only small reductions in capacitance upon the formation of each additional bilayer, indicating that the additional lipid bilayers permeable to ions and thus contain large or many defects. EIS can also provide information on whether protein incorporation in the membranes affects the electrode structure. For instance, incorporation of cytochrome bo 3 into tethered membranes (tBLM, Figure 2g) had almost no effect on the capacitance, indicating that cytochrome bo 3 did not induce large defects in the tBLM [75]. Similarly, only small changes in capacitance were observed of a hBLM (Figure 2e) after integrating human enzyme cytochrome P450, indicating that the hBLM retains its integrity upon protein adsorption [108]. Besides the membrane, the quality of a SAM on metal electrodes is also often evaluated with EIS. Compact, well-formed SAMs act as ideal insulating layers with high resistance values. The compactness of the SAM will influence subsequent protein immobilisation steps and can determine the thickness of the SAM, impacting on interfacial electron transfer kinetics.
Protein-film photoelectrochemistry (PF-PEC) is a PFE technique which has been developed for photosynthetic proteins [22,[109][110][111]. PF-PEC combines illumination with PFE to investigate photoactive proteins. Voltammetry and chronoamperometry can be used to illustrate the charge transfer processes and the kinetics of the light driven reactions. Recently, other electrochemical setups have been used to investigate photosynthetic proteins, including rotating ring disk electrode (RRDE) and scanning electrochemical microscopy (SECM). The set-up of RRDE includes a central rotating disc electrode and a ring electrode surrounding it. The potential on these two electrodes can be controlled independently and products generated by a protein film at the central disk electrode are transferred to the ring electrode for electrochemical analysis. For instance, oxygenic photoreactivity of PSII was studied with RRDE [112] and products generated by PSII at the central disc, e.g., oxygen and radical species, were detected by the ring electrode. This study revealed ET pathways that generate reactive oxygen species and O 2 by PSII. SECM employs a microelectrode as a tip above the protein modified electrode to detect the products generated locally. It was used to monitor H 2 evolution of a PSI-Pt complex within redox polymer under illumination [113]. SECM has also been used to quantify the reduction of charge carriers (methyl viologen) by PSI and compared this to the photocurrent [114]. Methyl viologen is often used as a charge carrier to collect electron from PSI in biophotovoltaic systems, but reoxidation on the electrode leads to charge recombination. The authors compared PSI on gold and silicon surfaces, both with Os-based redox polymers. SECM showed that gold and silicon exhibits different photocurrents due to different charge recombination kinetics, i.e., electron transfer kinetics for the reoxidation of the methyl viologen radical cation. Finally, by monitoring the photocurrent and H 2 O 2 generation of a PSI photocathode with SECM, it was shown that light-induced formation of reactive oxygen species caused degradation of PSI modified electrodes [115].

Spectroscopic Methods
While electrochemical methods provide information about redox reactions, in situ spectroscopic methods can be applied to characterise the structure of the protein and the electrode assembly. Surface-enhanced Raman spectroscopy (SERS) is a surface-sensitive technique for molecules and proteins immobilised on roughened metallic surfaces, in particular Ag [116]. The SERS effect is due to a large increase in the Raman cross-section of molecules in contact with the metal surface, which is enhanced at metallic nanostructured surfaces due to local surface plasmon resonance (LSPR) effects. SERS has been used to successfully characterise oxidised and reduced forms of adsorbed cytochromes (cytochrome c, haemoglobin and myoglobin) under precise potential control with low laser power [117]. In 2005, Hrabakova et al. showed that no structural changes occurred to the haem sites (e.g., a and a 3 ) of cytochrome c oxidase embedded in a phospholipid bilayer tethered to a functionalised silver electrode via a histidine-tag [118]. In 2011, Weidinger group used SERS to study the electron transfer of MBH from R. eutropha H16 [119]. Interestingly, the authors compared the behaviour of the entire enzyme with just the transmembrane "anchor" subunit (b-type cytochrome) and determined two independent pathways for the electrons from the active site of the enzyme to the electrode. A slow rate pathway crosses all the three subunits of the enzyme, whilst a faster pathway only crosses two subunits and leaves out the transmembrane anchor subunit which, however, contributes to the stabilisation of the enzyme on the electrode.
A different variation of conventional infrared spectroscopy is represented by surface-enhanced infrared absorption spectroscopy (SEIRAS) in which the signal enhancement is due to plasmon resonance from a nanostructured metal thin film [120]. The Hellwig group used SEIRAS to characterise the deposition of gold NPs on gold electrodes, the modification of gold NPs with thiols and, finally, the absorption of cytochrome bo 3 on these modified electrodes [51,54]. Wiebalck et al. characterised the formation of a tBLM and incorporation of functional cytochrome bo 3 from E. coli by SEIRA spectroscopy [121].

Spectroelectrochemistry
Spectroelectrochemistry combines electrochemistry and spectroscopic approaches to monitor a specific molecular response while varying the electrode potential [122,123]. This can be done with proteins in solution [124] or immobilised on the electrode surface [125,126]. Insights into reaction mechanisms of membrane enzymes could be gained by methods such as UV-vis spectroelectrochemistry and infrared spectroelectrochemistry. UV-vis spectra of a protein are often specific for the redox state of a cofactor, e.g., haems, FeS clusters or flavins. A common application is a spectroelectrochemical titration to determine the reduction potential of a cofactor in an enzyme in solution, where results are usually evaluated by fitting the titration data to the Nernst equation. An example is a study of the cytochrome bc 1 complex where the midpoint potentials of the cofactors were determined by UV-vis and IR spectroelectrochemical titrations [127,128]. The redox potentials of the primary electron acceptor pheophytin a in photosystem II [129] and the primary electron donor P700 in photosystem I [130] were also revealed by UV-vis spectroelectrochemistry with proteins in solution. UV-vis spectroelectrochemistry can also be used to study proteins on electrode surfaces. For instance, Haas et al. used UV-vis spectroelectrochemical titrations to determine the midpoint potentials for cytochrome c oxidase either in solution or in a Langmuir-Blodgett monolayer film [131]. A multi-protein system was studied with UV-vis spectroelectrochemistry in which an outer-membrane cytochrome, MtrC from Shewanella oneidensis MR-1 (a peripheral membrane protein), was used as an electron conduit between an electrode and other redox enzymes. MtrC was shown to be an effectively transfer-electron conduit by monitoring the absorbance of reduced Fe II -haems in MtrC [132].
Infrared (IR) spectroscopy is a powerful technique to detect structural changes of proteins during or in response to an electrochemical reaction although few studies have been reported for membrane proteins. Reactions can be triggered either by light or redox potential and infrared difference spectroscopy of proteins is typically measured to investigate the reaction mechanisms. This approach has been developed for membrane proteins involved in photosynthesis, respiration and metabolic pathways [100]. Electrochemical SEIRA was used to study the conformational changes of the R. sphaeroides cytochrome c oxidase induced by direct electron transfer in ptBLM system [133]. The cytochrome c oxidase was shown to change from a non-activated to an activated state after it involving enzymatic reaction. By applying periodic potential pulses switching between −800 mV and open circuit potential to control the state of cytochrome c oxidase, the kinetics of the conformational changes was monitored by time-resolved SEIRA spectroelectrochemistry [134]. The bacterial respiratory ubiquinol/cytochrome bo 3 couple was incorporated into a tethered bilayer lipid membrane (tBLM) on SAM modified electrode. The transmembrane proton gradient was successfully monitored by spectroelectrochemical SEIRA [121].

Electron Microscopy
Electron microscopy (EM) techniques are available to investigate membrane proteins (typically by cryo transmission electron microscopy, TEM or cryo-EM) and electrode structures (typically by scanning electron microscopy, SEM) at various scales. However, EM has not been widely used to study membrane proteins immobilised on electrodes as TEM and SEM are based on different modes of observation [135]. In 2015, Monsalve et al. [136] used SEM and TEM to characterise the morphology and size distribution of gold nanoparticles on a gold electrode used for direct absorption of Aquifex aeolicus [NiFe] membrane-bound hydrogenase surrounded by DDM detergent.

Atomic Force Microscopy
Atomic force microscopy (AFM) is a scanning probe microscopic technique based on a nanoprobe, a tip placed at the end of a long and narrow cantilever, that interacts with the sample surface to measure the topography of a surface [137]. Although AFM has been widely used to study lipid membranes [138][139][140], membrane proteins [141][142][143][144] and electrochemical systems with soluble proteins [145][146][147], AFM is less often used to study membrane proteins on electrode surfaces. In 2006, we used tapping-mode AFM to analyse the distribution of cytochrome bo 3 on a tethered bilayer lipid membrane (tBLM) on stripped gold electrode [75]. AFM combined with Polarisation Modulation Infrared Reflection-Absorption Spectroscopy (PM-IRRAS) was used to study the different orientation of membrane-bound Aquifex aeolicus (Aa) [NiFe] hydrogenase immobilised on hydrophilic and hydrophobic SAM on gold electrodes. This work highlighted that on charged or hydrophilic interfaces, H 2 oxidation proceeds through both direct and mediated electron transfer processes, while on hydrophobic surfaces, a mediator is required [148]. In 2014, Gutiérrez-Sanz et al. [149] characterised the functional reconstitution of respiratory complex I on SAM on gold electrode. Their AFM study showed the formation of a phospholipid bilayer on SAM modified gold electrode and protrusions of 6-8 nm height were observed which were ascribed to the hydrophilic arm of complex I as this arm extends outside the membrane.

Quartz Crystal Microbalance
Quartz crystal microbalance (QCM) is a sensitive mass sensor which utilises acoustic waves generated by oscillating a piezoelectric, single crystal quartz plate to measure mass changes in the order of nanograms. The association of QCM with dissipation monitoring (QCM-D) allows to also measure the energy loss or dissipation (∆D) of the system [150]. QCM was used to monitor the different stages of the ptBLM formation with cytochrome c oxidase from Rhodobacter sphaeroides as model protein [151]. The adsorption of cytochrome c oxidase at the surface decreased the resonance frequency while increased the dissipation. The Hawkridge group investigated the electrostatic association between cytochrome c and cytochrome c oxidase immobilised in hBLM with QCM [68,69]. The QCM data revealed the binding of cytochrome c and cytochrome c oxidase at different ionic strength which was related to the mediated electron transfer. QCM-D was also used to characterise the formation of the cystamine-pyrroloquinoline quinone-thylakoids layers onto SAMs [152].

Membrane Protein Modified Electrodes for Bioelectrocatalysis
The increased demand to produce energy and value-added chemicals from cheap and environmentally friendly renewable resources has driven the recent advances in bioelectrocatalysis research towards the development of alternative systems. The applications of bioelectrocatalysis range Catalysts 2020, 10, 1427 11 of 29 from biosensors, energy conversion devices and bioelectrosynthesis [153][154][155][156]. The majority of the studies employ soluble proteins; however, some membrane proteins with suitable catalytic or electron transfer properties also find applications in bioelectrocatalysis. Some of them show advantages over soluble proteins or have unique functions. Here, we will discuss the membrane protein electrode applications in the field of bioelectrocatalysis.

Enzymatic Biofuel Cells (EBCs)
One of the most studied energy conversion devices is the enzymatic biofuel cell (EBC) which uses oxidoreductase enzymes as biocatalysts to convert chemical energy into electrical energy [157][158][159]. A typical EBC consists of a two-electrode cell in which the biofuels (such as H 2 , formate) are oxidised at the bioanode and the oxidants are reduced at the biocathode (usually O 2 is reduced to water) [160]. Among these, H 2 /O 2 biofuel cells are one of the most investigated enzymatic systems [161]. Hydrogenases are promising biocatalysts to fabricate high performance H 2 -oxidation bioanodes. However, the extreme oxygen sensitivity of highly active hydrogenases is one of the main limitations of hydrogenase based H 2 bioanodes [162]. The O 2 sensitive hydrogenase can be protected by a low-potential viologen redox polymer matrix [163] or enzymatic O 2 scavenger [164]. However, these protection mechanisms either consume the electron from H 2 oxidation or add chemicals for protection. O 2 -tolerant MBH therefore shows great advantage for H 2 /O 2 biofuel cell. The first membrane-less H 2 /O 2 cell was assembled by Armstrong group. Two pyrolytic graphite electrodes, coated respectively with MBH from R. eutropha and laccase from Trametes versicolor (Tv), were immersed in a H 2 /air flushed solution. The system reached an OCV of~970 mV and a maximum power output of~5 µW [165]. High OCVs were achieved because MBH directly exchanged electrons with the electrode and no mediators were required. The Re MBH was solubilised from the membrane extracts by 2% Triton-X 114 and isolated via a Strep-tag sequence on the small subunit. It is possible that the MBH was devoid of the transmembrane cytochrome b anchor and this could explain why there was no need to use mediators. The Re MBH also showed CO-tolerance [165]. The same group further improved the performance of the H 2 /O 2 cell by replacing the Re MBH with MBH (similarly Strep-tagged on the small subunit) from R. metallidurans, which is more active and stable to O 2 exposure (Figure 3a). The authors emphasised that such a system would have the advantage to perform H 2 conversion even in H 2 -poor mixtures. They also showed that three cells in series provided a total OCV of 2.7 V which was sufficient to power a wristwatch for 24 h [166].
As described earlier, we have presented a strategy for using the full heterotrimeric MBH as a biocatalyst (including the cytochrome b anchor) and have used multi-layer membrane stacks on gold electrode to increase MBH loading [86]. In our approach, we used cytoplasmic membrane extracts of R. eutropha and created interconnected layers of membranes with each layer containing anchored MBHs. Lipophilic quinones were used as mediator, shuttling electron between electrode and protein. However, the irreversible electrochemical behaviour of the quinone redox reaction increases the required overpotential for H 2 oxidation, which will limit the power output of the devices and more research is needed to resolve this.
To optimise EBCs, gas diffusion electrodes have been investigated and enzyme coverage optimised. The porous structure of the gas diffusion electrode can increase the mass loading of the biocatalyst and overcome the mass transport limitation of gases. In 2016, Kano group used an O 2 -tolerant MBH from Hydrogenovibrio marinus and an O 2 -sensitive [NiFe]-hydrogenase from Desulfovibrio vulgaris "Miyazaki F" to create DET-type gas diffusion electrodes [167]. The authors did not specifically comment on whether their enzyme purification methods might have affected the presence of the cytochrome b anchor subunits of MBH. The MBHs were isolated through two different procedures. The large and small subunits of the O 2 -tolerant MBH were isolated through detergent solubilisation and maintained in 0.025% Triton-X. The O 2 -sensitive MBH went through a trypsinisation process which could lead to the separation of the transmembrane cytochrome b anchor. These H 2 oxidation electrodes generated a current density of 10 mA·cm −2 in the half-cell configuration. Contrary to the O 2 -tolerant MBH, the O 2 -sensitive MBH did not show overpotential for H 2 oxidation and, on this basis, was selected by the authors as bioanode of a H 2 /O 2 EBC for a further study. Coupling this O 2 sensitive MBH anode with bilirubin oxidase (BOD) from Myrothecium verrucaria immobilised on Ketjen black-modified waterproof carbon papers (KB/WPCC) electrode, a dual gas-diffusion membrane-and mediator-less H 2 /air-breathing biofuel cell was constructed (Figure 3b) which showed maximum power density in the range of 6.1 mW·cm −2 at 0.72 V [168]. The O2-reduction biocathodes for EBC are usually based on multi-copper oxidases like bilirubin oxidase or laccase which can reduce O2 almost without overpotential [169]. Among the membrane proteins, cytochrome c oxidase has been studied as O2 reduction catalyst. Katz and Willner assembled a membrane-less glucose/O2 biofuel cell with cytochrome c/cytochrome c oxidase as O2 reducing cathode [170]. The cytochrome c was assembled on a maleimide modified gold electrode through a cysteine residue to link cytochrome c oxidase. In a follow-up work, the authors developed an electroswitchable and tunable biofuel cell. In this case, the cathode was modified with poly(acrylic acid) loaded with Cu 2+ to covalently attach cytochrome c and link the latter to cytochrome c oxidase [171]. Although cytochrome c was able mediated electron transfer to cytochrome c oxidase for O2 reduction, only an OCV of 0.12 V was obtained, likely limited by the redox potential of cytochrome c (~0.25 V vs. SHE), which is much lower than that of H2O/O2 (0.82 V vs. SHE at pH 7). Although cytochrome c oxidase is not commonly used for O2 reducing cathode in biofuel cell, the recent work with gold NPs by the group of Hellwig has shown renewed possibilities for cytochrome c oxidase as catalyst (see Section 3.3). Recently, Wang et al. [172] reported that cytochrome c oxidase from acidophilic bacterium Acidithiobacillus ferrooxidans can reduce O2 at exceptionally high electrode potentials (+700 to +540 mV vs. NHE). The low overpotential for O2 reduction of this cytochrome c oxidase makes it an attractive biocatalyst as cathode of biofuel cells in the future.
Besides the use of H2 as fuel for EBC, formate is also a valuable feedstock for biofuel cells because its redox potential is similar to H2/H + . The Kano group reported a mediated electron transfer type formate/O2 biofuel cell by coupling formate dehydrogenase modified bioanode with BOD modified biocathode [173]. In nature, some formate dehydrogenases can catalyse the inverse reaction to reduce CO2 to formate [174]. However, to the best of our knowledge, the research in this field has been conducted only on soluble formate dehydrogenases. The ability of CO2 reduction and the applications of membrane-bound formate dehydrogenase need to be further explored.
Respiratory nitrate reductase (Nar) catalyses NO3 − reduction to NO2 − , this is an essential step to produce NH3. Today there is no enzyme known to reduce NO3 − to NH3 directly [175]. DET with nitrate reductase has been shown to support the electrochemical reduction of NO3 − to NO2 − [34,35]. A full reduction of NO3 − to NH3 was demonstrated by a cascade electrocatalysis process which combined nitrate reductase and noble metal catalyst to reduce NO2 − to NH3 [58].

Biophotoelectrocatalysis (PEC)
Biophotoelectrodes fabricated on planar carbon or SAM-modified metal electrode usually show low photocurrent density which limits their applications in bioelectrocatalysis. However, the The O 2 -reduction biocathodes for EBC are usually based on multi-copper oxidases like bilirubin oxidase or laccase which can reduce O 2 almost without overpotential [169]. Among the membrane proteins, cytochrome c oxidase has been studied as O 2 reduction catalyst. Katz and Willner assembled a membrane-less glucose/O 2 biofuel cell with cytochrome c/cytochrome c oxidase as O 2 reducing cathode [170]. The cytochrome c was assembled on a maleimide modified gold electrode through a cysteine residue to link cytochrome c oxidase. In a follow-up work, the authors developed an electroswitchable and tunable biofuel cell. In this case, the cathode was modified with poly(acrylic acid) loaded with Cu 2+ to covalently attach cytochrome c and link the latter to cytochrome c oxidase [171]. Although cytochrome c was able mediated electron transfer to cytochrome c oxidase for O 2 reduction, only an OCV of 0.12 V was obtained, likely limited by the redox potential of cytochrome c (~0.25 V vs. SHE), which is much lower than that of H 2 O/O 2 (0.82 V vs. SHE at pH 7). Although cytochrome c oxidase is not commonly used for O 2 reducing cathode in biofuel cell, the recent work with gold NPs by the group of Hellwig has shown renewed possibilities for cytochrome c oxidase as catalyst (see Section 3.3). Recently, Wang et al. [172] reported that cytochrome c oxidase from acidophilic bacterium Acidithiobacillus ferrooxidans can reduce O 2 at exceptionally high electrode potentials (+700 to +540 mV vs. NHE). The low overpotential for O 2 reduction of this cytochrome c oxidase makes it an attractive biocatalyst as cathode of biofuel cells in the future.
Besides the use of H 2 as fuel for EBC, formate is also a valuable feedstock for biofuel cells because its redox potential is similar to H 2 /H + . The Kano group reported a mediated electron transfer type formate/O 2 biofuel cell by coupling formate dehydrogenase modified bioanode with BOD modified biocathode [173]. In nature, some formate dehydrogenases can catalyse the inverse reaction to reduce CO 2 to formate [174]. However, to the best of our knowledge, the research in this field has been conducted only on soluble formate dehydrogenases. The ability of CO 2 reduction and the applications of membrane-bound formate dehydrogenase need to be further explored.
Respiratory nitrate reductase (Nar) catalyses NO 3 − reduction to NO 2 − , this is an essential step to produce NH 3 . Today there is no enzyme known to reduce NO 3 − to NH 3 directly [175]. DET with nitrate reductase has been shown to support the electrochemical reduction of NO 3 − to NO 2 − [34,35]. A full reduction of NO 3 − to NH 3 was demonstrated by a cascade electrocatalysis process which combined nitrate reductase and noble metal catalyst to reduce NO 2 − to NH 3 [58].

Biophotoelectrocatalysis (PEC)
Biophotoelectrodes fabricated on planar carbon or SAM-modified metal electrode usually show low photocurrent density which limits their applications in bioelectrocatalysis. However, the development of redox polymer electrodes, layer-by-layer assembly and 3D architectures has enhanced the performance of biophotoelectrodes [176,177]. PSII is the only natural protein able to catalyse the photooxidation of water. The electron flow can be blocked by herbicide compounds since they bind to the terminal plastoquinone Q B of PSII. This inhibition effect can be exploited for designing PSII light-driven biosensor to detect herbicides [178]. A similar approach to detect herbicides was taken with purple bacteria RCs [179].
As water oxidation biophotocatalyst, PSII attracts more attention for solar energy conversion to generate electricity or fuels. As mentioned above, PSII photoanodes have been connected to PSI photocathodes (both in Os redox polymers) to mimic natural photosynthesis Z-scheme for solar-to-electricity generation [62,63]. In a similar approach using a benzoquinone redox polymer, a PSII photoanode (O 2 evolution) has been connected to a bilirubin oxidase cathode for O 2 reduction [180]. Recently, PSII was integrated together with PbS quantum dots within a TiO 2 inverse opal electrode to perform H 2 O photooxidation. This was combined with an inverse opal antimony-doped tin oxide (ATO) cathode modified with bilirubin oxidase to catalyse oxygen reduction. This system achieved a high open-circuit voltage of about 1V under illumination (Figure 4a,b) [181].
A full water splitting process can be realised by combining a PSII photoanode with a cathode modified with hydrogenases. In contrast to EBCs (Section 5.1), water-soluble hydrogenases (typically [FeFe] hydrogenases, but also [NiFe] hydrogenases; H2ases) are most commonly used for water splitting systems. When combining PSII and H2ases, there is an energy gap between the terminal electron acceptor Q B within PSII and one of the FeS cluster within H2ase. A biophotoelectrochemical cell with a PSII photoanode and a H2ase cathode thus requires an applied bias voltage of 0.8 V to drive H 2 O splitting [93]. This was improved by wiring H2ase and p-Si on an inverse opal TiO 2 photocathode, lowering the required applied bias voltage to 0.4 V for H 2 O splitting [182]. Finally, by integrating PSII on a diketopyrrolopyrrole dye-sensitised TiO 2 photoanode and connecting it with a H2ase cathode, a bias-free photoelectrochemical cell for H 2 O splitting was developed by the group of Reisner [183]. Coupling a dye-sensitised PSII photoanode with a W-dependent formate dehydrogenase (FDH) cathode, a biophotoelectrochemical cell was constructed for CO 2 reduction at a small bias voltage of 0.3 V (Figure 4c,d) [184]. The latter study showed the possibility for rational design of biophotoelectrochemical cells for value-added chemicals generation beyond H 2 .
Unlike PSII, PSI does not directly catalyse technologically valuable reactions such as water oxidation. However, photoexcitation of PSI provides the reductive potential to drive reactions with other catalysts, such as Pt and H2ase [185]. Recent studies show that it is possible to drive H 2 production from light with PSI and H2ase by electrode design. Photoelectrodes were manufactured in a 'layered' fashion using an Os redox polymer, PSI and, finally, a polymer/H2ase mix [186]. Photoelectrochemical H 2 production is achieved at an onset potential of +0.38 V vs. SHE. In this study, the PSI was randomly orientated and did not form a compact layer, likely limiting efficiency by charge recombination between the carrier and mediator or electrode. In a more recent study, an anisotropically oriented PSI monolayer was formed using Langmuir-Blodgett deposition [187]. A compact and oriented PSI layer minimises charge recombination and enables unidirectional electron transfer to H2ase for H 2 evolution. Combining this PSI/H2ase photocathode with a PSII photoanode created a system able of bias-free light-driven water splitting [187]. Langmuir-Blodgett deposition transfers only a monolayer of PSI and this might limit the performance of the biophotoelectrode as this limits the loading or coverage of PSI.
One of the limitations with biophotoelectrodes is the limited lifetime of the isolated proteins, especially PSII [154]. Light-induced formation of reactive oxygen species can further limit the lifespan of proteins [115]. Another limitation of biophotoelectrodes is that photosynthetic proteins only use a limited range of the solar spectrum which reduces solar conversion efficiency. The absorption spectral range can be enhanced by attaching complementary chromophores to light-harvesting complex proteins [188]. It also can be improved by integrating biological light-harvesting antenna complexes or organic dyes/synthetic compounds to the RCs [189,190]. electron acceptor QB within PSII and one of the FeS cluster within H2ase. A biophotoelectrochemical cell with a PSII photoanode and a H2ase cathode thus requires an applied bias voltage of 0.8 V to drive H2O splitting [93]. This was improved by wiring H2ase and p-Si on an inverse opal TiO2 photocathode, lowering the required applied bias voltage to 0.4 V for H2O splitting [182]. Finally, by integrating PSII on a diketopyrrolopyrrole dye-sensitised TiO2 photoanode and connecting it with a H2ase cathode, a bias-free photoelectrochemical cell for H2O splitting was developed by the group of Reisner [183]. Coupling a dye-sensitised PSII photoanode with a W-dependent formate dehydrogenase (FDH) cathode, a biophotoelectrochemical cell was constructed for CO2 reduction at a small bias voltage of 0.3 V (Figure 4c,d) [184]. The latter study showed the possibility for rational design of biophotoelectrochemical cells for value-added chemicals generation beyond H2.

Extending the Lifetime of Membrane Enzymes
A major drawback of enzymes in EBC is their limited active lifetime, which usually ranges from few hours to several days [191]. This applies particularly when membrane enzymes in detergent solutions are used, and functional reconstitution of membrane enzymes into an amphiphilic bilayer, such as liposome or polymersome vesicles, could represent a strategy to extend the enzyme lifetime. Liposomes offer great biocompatibility because they mimic the natural environment of membrane proteins, but they lack chemical and physical long-term stability [192]. Polymersomes offer a more robust amphiphilic polymer environment with increased chemical and physical stability [193,194]. However, this non-native polymeric environment might be limiting the functional incorporation of a wider range of membrane proteins [195]. Recently, hybrid vesicle systems composed of a mixture of lipids and block copolymers, have been developed with the rationale to provide a compromise between the biocompatibility of liposomes and the stability and robustness of polymersomes. In 2016 we showed that hybrid vesicles, composed of biocompatible lipids and stable PBd-PEO copolymer, supported higher activity of reconstituted cytochrome bo 3 than the proteopolymersomes and significantly extended the functional lifetime of the membrane enzyme when compared to standard proteoliposomes [196] ( Figure 5). In 2018, we achieved increased stability of cytochrome bo 3 reconstituted in hybrid vesicles up to 500 days [197]. Similarly, recent work from Dimova group showed that functional integration of cytochrome bo 3 oxidase in synthetic membranes made of PDMS-g-PEO was capable of lumen acidification and such reconstituted system showed to increase the active lifetime and resistance to free radicals [198]. In 2017, Otrin et al. [199] demonstrated a similar ability to store gradients by reconstituting cytochrome bo 3 together with an ATP synthase in hybrid vesicles constituted of the same copolymer, PDMS-g-PEO. In 2018, Smirnova et al. presented a method that allowed transfer of a functional membrane protein, cytochrome c oxidase (cytochrome aa 3 or yeast Complex IV), with a disc of its native lipids into pre-formed liposomes of well-defined lipid composition and size using amphipathic styrene maleic acid (SMA) copolymer [200]. This recent advance in using SMA copolymer for membrane enzymes isolation and reconstitution offers the advantage to maintain the native phospholipids environment surrounding the proteins and, moreover, could reduce time and cost for enzyme isolation and reconstitution processes by avoiding detergent mediated extraction [201]. Further research into affordable purification strategies and extending the stability of commercially-relevant membrane enzymes is required for membrane enzymes to find applications in bioelectrocatalysis. An alternative approach would be to omit purification altogether and exploit the regenerative capacity of micro-organisms in microbial electrosynthesis.
Catalysts 2020, 10, x FOR PEER REVIEW 15 of 28 advance in using SMA copolymer for membrane enzymes isolation and reconstitution offers the advantage to maintain the native phospholipids environment surrounding the proteins and, moreover, could reduce time and cost for enzyme isolation and reconstitution processes by avoiding detergent mediated extraction [201]. Further research into affordable purification strategies and extending the stability of commercially-relevant membrane enzymes is required for membrane enzymes to find applications in bioelectrocatalysis. An alternative approach would be to omit purification altogether and exploit the regenerative capacity of micro-organisms in microbial electrosynthesis. Reprinted with the permission from ref [196]. Copyright (2016), The Royal Society of Chemistry.

Microbial Electrosynthesis and Whole-Cell Semi-Artificial Photosynthesis
Microorganisms are of great interest for electrochemical applications because of their selfreproduce nature and diverse metabolic processes. The possibility to use directly the microorganisms overcomes the need to purify and manipulate the proteins. Many microorganisms are exoelectrogens and can communicate with electrodes through extracellular electron transfer (EET), either (1) directly through membrane-bound cytochromes and conductive filaments (nanowires) or (2) via a soluble redox species, such as flavins, either natively secreted by the microorganism or added as a mediator [202].

Microbial Electrosynthesis
Microbial electrosynthesis is an electricity-driven process which generates chemicals using microorganisms as catalysts [203]. To date, the most well-studied exoelectrogens are Geobacter spp. and S. oneidensis [204]. They are dissimilatory metal-reducing bacteria that can reduce extracellular metal ions (and bulk electrodes) as part of their anaerobic respiration. These electron transfer steps can be reversed to gain electron from electrode for reductive synthesis, some catalysed by membrane enzymes in the microorganisms ( Figure 6) [205,206]. It has been shown that Geobacter spp. can reduce nitrate to nitrite [207], fumarate to succinate [207] and proton to hydrogen [208] using electrode as the electron donor. S. oneidensis was also shown to electrocatalytically reduce fumarate to succinate [209]. It also has been reported that S. oneidensis can catalyse CO2 reduction into formic acid with electron input from the cathode [210]. The cathodic CO2 reduction in S. oneidensis is associated with the electron uptake through outer-membrane c-type cytochromes [211]. The efficiency of the microbial electrosynthesis is mainly limited by the EET. With the genetic engineering of the  [196]. Copyright (2016), The Royal Society of Chemistry.

Microbial Electrosynthesis and Whole-Cell Semi-Artificial Photosynthesis
Microorganisms are of great interest for electrochemical applications because of their self-reproduce nature and diverse metabolic processes. The possibility to use directly the microorganisms overcomes the need to purify and manipulate the proteins. Many microorganisms are exoelectrogens and can communicate with electrodes through extracellular electron transfer (EET), either (1) directly through membrane-bound cytochromes and conductive filaments (nanowires) or (2) via a soluble redox species, such as flavins, either natively secreted by the microorganism or added as a mediator [202].

Microbial Electrosynthesis
Microbial electrosynthesis is an electricity-driven process which generates chemicals using microorganisms as catalysts [203]. To date, the most well-studied exoelectrogens are Geobacter spp. and S. oneidensis [204]. They are dissimilatory metal-reducing bacteria that can reduce extracellular metal ions (and bulk electrodes) as part of their anaerobic respiration. These electron transfer steps can be reversed to gain electron from electrode for reductive synthesis, some catalysed by membrane enzymes in the microorganisms ( Figure 6) [205,206]. It has been shown that Geobacter spp. can reduce nitrate to nitrite [207], fumarate to succinate [207] and proton to hydrogen [208] using electrode as the electron donor. S. oneidensis was also shown to electrocatalytically reduce fumarate to succinate [209]. It also has been reported that S. oneidensis can catalyse CO 2 reduction into formic acid with electron input from the cathode [210]. The cathodic CO 2 reduction in S. oneidensis is associated with the electron uptake through outer-membrane c-type cytochromes [211]. The efficiency of the microbial electrosynthesis is mainly limited by the EET. With the genetic engineering of the microorganisms, the EET efficiency can be improved [212][213][214]. Advances in synthetic biology allow the rational design of non-natural functions in order to increase the diversity of products obtainable from microbial electrosynthesis [215]. For instance, by engineering genes for an ATP-dependent citrate lyase into Geobacter sulfurreducens, the microorganism is able to fix CO 2 through a reverse TCA cycle using an electrode as electron donor [216]. A genetically engineered S. oneidensis with heterologous Ehrlich pathway genes was shown to produce isobutanol by supplying electricity [217]. It has been demonstrated that S. oneidensis can use electrons supplied by an electrode to reduce O 2 . This study also showed that the cathodic reaction can reduce NAD + via proton-pumping NADH oxidase complex I [218]. With addition of a light-driven proton pump (proteorhodopsin) in the S. oneidensis to generate proton-motive force, the electron transferred from cathode to quinone pool can be used to reduce NAD + to NADH by native NADH dehydrogenases. This was demonstrated by the reduction of acetoin to 2,3-butanediol via a heterologous butanediol dehydrogenase (Bdh) which is a NADH-dependent enzyme [219].
Catalysts 2020, 10, x FOR PEER REVIEW 16 of 28 citrate lyase into Geobacter sulfurreducens, the microorganism is able to fix CO2 through a reverse TCA cycle using an electrode as electron donor [216]. A genetically engineered S. oneidensis with heterologous Ehrlich pathway genes was shown to produce isobutanol by supplying electricity [217]. It has been demonstrated that S. oneidensis can use electrons supplied by an electrode to reduce O2. This study also showed that the cathodic reaction can reduce NAD + via proton-pumping NADH oxidase complex I [218]. With addition of a light-driven proton pump (proteorhodopsin) in the S. oneidensis to generate proton-motive force, the electron transferred from cathode to quinone pool can be used to reduce NAD + to NADH by native NADH dehydrogenases. This was demonstrated by the reduction of acetoin to 2,3-butanediol via a heterologous butanediol dehydrogenase (Bdh) which is a NADH-dependent enzyme [219].

Whole-Cell Based Semi-Artificial Photosynthesis
Semi-artificial photosynthesis has attracted great attention for harvesting solar energy for electricity or chemical generation [5]. Instead of isolated photosynthetic proteins (e.g., PSII), photosynthetic microorganisms such as algae [221], cyanobacteria [222][223][224][225] and purple bacteria [226], or extracted organelles such as thylakoids [227] have been studied for biophotoelectrochemical systems. The development of 3D structured electrodes increased the loading capacity which enhances photocurrents [228]. However, low EET rates still limit the efficiency of these photoelectrodes. It has been shown that redox polymers can improve electron transfer kinetics between microorganisms and electrode [221,226]. Furthermore improvement in EET kinetics and new electrode designs for accommodating microorganisms are required to boost the performance of photoelectrochemical

Whole-Cell Based Semi-Artificial Photosynthesis
Semi-artificial photosynthesis has attracted great attention for harvesting solar energy for electricity or chemical generation [5]. Instead of isolated photosynthetic proteins (e.g., PSII), photosynthetic microorganisms such as algae [221], cyanobacteria [222][223][224][225] and purple bacteria [226], or extracted organelles such as thylakoids [227] have been studied for biophotoelectrochemical systems. The development of 3D structured electrodes increased the loading capacity which enhances photocurrents [228]. However, low EET rates still limit the efficiency of these photoelectrodes. It has been shown that redox polymers can improve electron transfer kinetics between microorganisms and electrode [221,226]. Furthermore improvement in EET kinetics and new electrode designs for accommodating microorganisms are required to boost the performance of photoelectrochemical devices that are based on photosynthetic microorganisms [111].
An emerging direction in the semi-artificial photosynthesis is to interface synthetic light-harvesting materials with non-photosynthetic microorganisms for value-added chemicals [229,230]. As a model microorganism, S. oneidensis has been widely studied for integration with artificial light-harvesting materials for solar-driven microbial synthesis. The transmembrane cytochrome MtrCAB, helped by OmcA located on the outside of the membrane, are known to transfer electrons between the interior of the cell and extracellular materials ( Figure 6) [231,232]. In the presence of a sacrificial electron donor, MtrC and OmcA can be photoreduced by water-soluble photosensitisers including eosin Y, fluorescein, proflavine, flavin, and adenine dinucleotide, riboflavin and flavin mononucleotide [233]. In an in vitro approach, it was showed that dye-sensitised TiO 2 nanoparticles can photoreduce MtrC or OmcA, either in solution or one an electrode surface [233][234][235] and it might be possible to extend this process to reductive photosynthesis in S. oneidensis [236]. To provide a proof-of-concept, we recently reconstituted MtrCAB into proteoliposomes encapsulating a redox dye, Reactive Red 120 (RR120), which can be reductively bleached. Using light-harvesting nanoparticles including dye-sensitised TiO 2 nanoparticles, amorphous carbon dots and nitrogen doped graphitic carbon dots, we observed RR120 reductive decomposition inside the lumen of the MtrCAB proteoliposomes, confirming that electrons were transferred from the nanoparticles via transmembrane MtrCAB complex into the liposome lumen [237]. These results show that the rational design of light-harvesting nanoparticles and protein hybrids can lead to the development of semi-artificial photosynthetic systems for solar fuels synthesis.

Conclusions
Membrane proteins play an important role for biofuel conversion and photosynthesis, and the catalytic and electron transfer ability of the membrane proteins makes them attractive for applications in bioelectrocatalysis. Recent developments in bioelectrochemistry have provided the means to control and improve electron transfer rate between immobilised proteins and electrodes, stabilising the activity of immobilised enzymes. High surface areas of so-called 3D electrodes are able to increase the protein loading, which is desired for bioelectrocatalysis. Despite these enhanced performances, there is still the need to further optimise this technology for practical use in bioelectrocatalysis devices, especially for membrane enzymes. Combining several immobilisation strategies would provide possible solutions. Advances in characterisation techniques enable detailed characterisation of the structure of the electrode, the electron transfer process, catalytic activity and protein structure, which in turn provides an in-depth understanding of the catalytic reaction mechanism, aiding the rational design of the electrode interface for bioelectrocatalysis. Membrane proteins are known to be difficult to study and manipulate due to their amphiphilic nature. Preserving their stability and functionality is crucial for successful application of this technology. Recently developed redox polymers and hybrid vesicles, composed of lipids and block copolymers, are promising platforms to stabilise membrane proteins on the electrode surface. Using exoeletrogenic microorganisms for electrosynthesis and semi-artificial photosynthesis is an encouraging research direction, which will simplify time and cost related with protein purification processes. The ability of microorganisms to regenerate and self-replicate (at an energy cost), also removes the need to stabilise the biocatalyst. With synthetic biology, the microorganisms can be further engineered for diverse biofuel and chemical synthesis. However, the extracellular electron transfer kinetics might still be limiting the efficiency of the whole-cell based systems, especially when organisms are used that have not been optimised by evolution for extracellular electron transfer.