Engineering of bespoke photosensitiser–microbe interfaces for enhanced semi-artificial photosynthesis

Biohybrid systems for solar fuel production integrate artificial light-harvesting materials with biological catalysts such as microbes. In this perspective, we discuss the rational design of the abiotic–biotic interface in biohybrid systems by reviewing microbes and synthetic light-harvesting materials, as well as presenting various approaches to coupling these two components together. To maximise performance and scalability of such semi-artificial systems, we emphasise that the interfacial design requires consideration of two important aspects: attachment and electron transfer. It is our perspective that rational design of this photosensitiser–microbe interface is required for scalable solar fuel production. The design and assembly of a biohybrid with a well-defined electron transfer pathway allows mechanistic characterisation and optimisation for maximum efficiency. Introduction of additional catalysts to the system can close the redox cycle, omitting the need for sacrificial electron donors. Studies that electronically couple light-harvesters to well-defined biological entities, such as emerging photosensitiser–enzyme hybrids, provide valuable knowledge for the strategic design of whole-cell biohybrids. Exploring the interactions between light-harvesters and redox proteins can guide coupling strategies when translated into larger, more complex microbial systems.


Introduction
Generating fuels from renewable energy, such as sunlight, is of paramount importance in our urgent transition away from fossil fuels.In its most basic form, three components are required for photoredox catalysis: a catalyst for chemical reduction, a catalyst for chemical oxidation, and a light absorber to harvest solar energy.These components can be synthetic, natural, or a combination of both.Semi-articial photosynthesis has been previously dened as interfacing biological catalysts, which provide unparalleled catalytic specicity for the production of desirable chemicals, with synthetic lightharvesting materials to create biohybrid systems. 1 Here, we focus specically on the integration of microbial catalysts directly with synthetic nanoparticle light-harvesters.We refer the reader to previous reviews on other systems of semi-articial photocatalysis, including photoelectrochemical cells (PECs), which are not discussed here. 2,3While the use of microbes as biological catalysts in solution with light-harvesting nanoparticles has been investigated for decades, 4 a landmark study in 2016 initiated a renaissance of research into these biotechnologies by demonstrating that nanoparticles could be synthesised and then localised at the surface of the microbe. 5A long-lived physical interface between microbe and photosensitiser eliminates diffusion as a limiting factor.Synthetic materials can have advantages over natural photosystems in terms of resistance to photodamage; some subunits of natural photosystems require repair as oen as every 30 minutes. 6urther, microbial approaches exploit the capability of live cells to self-replicate and repair, reducing the input of materials such as puried proteins that require costly and time-consuming production.The ability to genetically program cells also provides opportunities to engineer them as factories for a specic chemical product. 7,8hile light-harvesters and microbes have been previously reviewed in depth, and whole-cell biohybrids discussed and analysed, 2,9-16 the challenge of creating a dened electron transfer interface between photosensitiser and microbe has only been highlighted. 17The nature of this abiotic-biotic interface in existing biohybrids is oen unclear and difficult to characterise retrospectively. 18Thus, the question we aim to address here is: how can we establish controllable and characterisable pathways for electron transfer, from photosensitiser to cytoplasmic enzymes, that can be built-in by design?What research should we turn to for creating these biohybrid blueprints, and where are remaining gaps of knowledge that prevent progression?In this perspective, we propose that the rational design of these systems should exploit microbes with natural mechanisms of electron uptake, and involve engineered attachment of the photosensitiser to the biological catalyst.We support this view with a detailed discussion of research that we hope can benet the ongoing optimisation of whole-cell biohybrids in the future.

Components for semi-artificial photosynthesis
Construction of a controllable, optimisable and scalable biohybrid requires consideration of three important areas: selection of an effective light absorber for photoelectron generation (together with a catalyst for regeneration and preferably the synthesis of an added-value chemical); selection of a suitable microbe for catalysis; and nally, electronic coupling of the two via a dened electron transfer pathway, from (photo)electron generation through to intracellular catalysis (see Fig. 1).Most research has focused on using the microbe as a site of fuel production through reductive catalysis, although some recent studies reversed this concept and instead use the microbe as an electron donor. 19The components discussed in this review are broadly applicable to both designs.
A wide range of articial light-harvesters have been applied, including organic and inorganic dyes, nanomaterials, and organic semiconductors. 12,20,21It is essential to consider the compatibility of these materials with the microbes involved in each system, along with characteristics such as band gap, redox potential, excited state lifetime, stability, and ease of chemical modication.Designing an interaction between the lightharvester and microbe will rely heavily on the mode by which electrons are able to be internalised.Several successful biohybrid systems for fuel production have not fully characterised their photosensitiser-microbe interface for electron transfer, limiting both controllability and optimisability.3][24][25][26] Mediators are membrane soluble electron carrier molecules that can shuttle electrons between the extracellular environment and the cytoplasm.While this concept works in systems that involve a sacricial electron donor (SED) to regenerate the lightharvesting material and is valuable at the research stage, systems addressing the ultimate vision of catalysing an oxidative half reaction with a reactant other than a SED could be short-circuited in the presence of redox mediator.Separating the oxidative and reductive half-reaction in different compartments with macroscopic (photo)electrodes would present a potential solution to this short-circuiting.However, there are several disadvantages to these assemblies of macroscopic wiring: they are technically more complicated, and rely on electrodes and ion-exchange membranes which can be both toxic and expensive.Maintaining the neutral pH required by biohybrids is difficult in a set-up where mass transport can create large pH gradients at the electrode interface.Moreover, the single compartment aqueous systems discussed in this perspective are intrinsically compatible with tubular photobioreactors, 27 making scalable solar light harvesting relatively easy.For a self-generating system, it is also important to Fig. 1 Illustration of the biohybrid concept (in this case using the microbe as a site of reductive catalysis and fuel production): an illuminated (Gram-negative) bacterium, with an inset showing a light-harvesting nanoparticle located on the outer membrane.(1) Artificial light-harvesting material (grey dashed circle) transitions to excited state on illumination: schematically indicated is an electron (e − ) that moves into a higher energy orbital (in a molecular photosensitiser) or from the valence band to the conduction band (in a nanoparticle).( 2) Electron transfer between the light-harvester and the site of catalysis: photoelectrons must be generated in close proximity to the microbe before passing through the microbial membrane to drive intracellular catalysis.(3) Microbial catalysis: the organism is selected or engineered to express enzymes that catalyse production of a desired chemical product.Finally, the light-harvester must be replenished by an oxidation reaction to complete the redox cycle, here indicated by catalytic water oxidation.
consider that growing microorganisms is more controllable in solution than on electrodes.
The case will be presented for exploiting the dened, naturally evolved, characterisable pathways of transmembrane electron transfer in exoelectrogenic bacteria as a simple method for electronically coupling extracellular photoelectrons with intracellular enzymes.Emerging molecular details of these pathways can provide context for informed advances.We argue here that rational design of a specic, engineered interaction at the abiotic-biotic interface is not only superior to interaction by diffusion but also fundamental for systems to be controllable, optimisable and scalable.This perspective presents a range of strategies used to couple photosensitisers with proteins, a discussion that aims to assist in designing the conjugation of photosensitisers to biological components at the surface of microbes.

Current approaches in biohybrid research
Current biohybrid systems can be classied broadly into two themes: (1) synthesised nanomaterials can be directly added to the growth-media of the bacteria and (2) nanoparticles can be bio-precipitated by the bacteria itself upon addition of starting materials.Furthermore and as mentioned above, a distinction can be made on whether or not redox mediators are used to shuttle electrons from the nanomaterial to the intracellular environment of the microbe.Each of these approaches has a distinct effect on the photocatalytic performance (refer to Table 1 for a summary of key features of the systems discussed, including performance parameters).
Due to a lack of electronic interaction between photosensitisers and microbes, some systems require a redox mediator such as methyl viologen (MV) for electron transfer into bacteria.For example, a biohybrid system has been constructed in which the phototropic bacterium Rhodopseudomonas palustris is further photosensitised with conjugated polymers, oli-gouorene (OF) and polythiophene (PTP). 22The natural H 2 generation by these bacteria was increased when MV was added to mediate electron transfer from the conjugated polymers to the intracellular enzymes.MV has also been used to shuttle electrons between a TiO 2 nanoparticle photosensitiser suspended in bacteria culture media with a recombinant strain of Escherichia coli.Within the E. coli microbe, electrons were transmitted to an exogenously expressed [FeFe]-hydrogenase derived from Clostridium acetobutylicum NBRC 13948 to produce H 2 . 23,24Photosensitisers Eosin Y and ruthenium-tris-2-2 0 -bipyridine (Ru(bpy) 3 2+ ) have been coupled to Shewanella oneidensis MR-1 as a biocatalyst using MV, driving H 2 generation along with the reduction of fumarate, pyruvate and CO 2 to, respectively, succinate, lactate and formate. 25An E. coli system, expressing [FeFe]-hydrogenases for H 2 production and using Eosin Y and Ru(bpy) 3 2+ as photosensitisers, was enhanced by redox mediators MV and diquat derivative 1,1 0 -(1,3-propylene)-5,5 0 -dimethyl-2,2 0 -bipyridinium (DQ). 26Both mediators signicantly improved the performance of the system although, due to its more negative reduction potential, a faster reaction rate was achieved using DQ.Use of mediators is not only unsustainable due to potential toxicity, but also the lack of a dened interface between components.Moreover, as mentioned, when an oxidative half reaction is coupled to the biohybrid system to close the redox cycle in place of an SED, the free redox mediator can cause back reactions and short-circuit the electron for the reductive reaction.Therefore our view is that redox mediators, although a useful research tool, limit optimisation of electronic coupling and thus the longevity of the system.
Biohybrids based on the self-assembly of CdS nanoparticles precipate Cd 2+ and S 2− at the surface of bacteria.The rst system to pioneer this method used self-photosensitisation of Moorella thermoacetica to drive solar-powered reduction of CO 2 to acetic acid. 5Similar systems constructed with Desulfovibrio desulfuricans, Citrobacter freundii and S. oneidensis MR-1 were used for light-driven H 2 generation, with D. desulfuricans-CdS displaying the highest activity. 28Quantum yields of 23% and 4% were obtained for this biohybrid, with and without the addition of MV respectively.The photoheterotrophic bacterium R. palustris was coated with CdS nanoparticles for N 2 xation. 29In this example, CdS has been shown to form clusters that are able to cross the cell membrane, leading to the assumption that this provides a feasible electron transfer pathway within the cell.In the E. coli-CdS biohybrid for H 2 generation, 30 nanoparticles are localised across the periplasmic space of the E. coli membrane.It was proposed that the intracellular surface of the CdS particles could interact with bioactive components and integrate into the hydrogen production pathway.Methanosarcina barkeri-CdS biohybrids were able to reduce CO 2 to methane. 31The CH 4 yield increased by approximately 250% when CdS were doped with Ni. 32 The introduction of Ni dopant was proposed to suppress the recombination of electron-hole pairs in the biohybrid, leading to enhanced electron transfer.Additionally, Ni induces alterations in the metabolic state of M. barkeri, leading to increased expression of various proteins involved in electron transfer, energy conversion, and CO 2 xation.It was proposed in this study that photoelectron transfer for CO 2 reduction occurred through membrane-bound proteins.However, the type of membrane protein was not identied.In a recent study involving S. oneidensis MR-1-CdS biohybrids for H 2 generation, four possible electron transfer pathways were proposed. 33ydrogen could be produced directly by the light-harvesting nanoparticles themselves, by hydrogenase enzymes, by nanoparticle-hydrogenase enzyme hybrids, or by electron transfer from the nanoparticles to intracellular hydrogenases through the outer membrane proteins.The complexity of these systems, and lack of clarity due to multiple possible routes, makes it difficult to ascertain precise electron transfer pathways.
Instead of nanoparticles self-assembled by the bacteria upon addition of starting materials, pre-synthesised nanoparticles (or other nano-scale materials) can be added directly to the bacteria media.For example, CuInS 2 /ZnS quantum dots (QDs) have been added to a S. oneidensis strain expressing periplasmic hydrogenases. 34The QDs can be taken up into the bacterial periplasm where they interact directly with the hydrogenases for H 2 production.A Sporomusa ovata-CdS hybrid was used to convert CO 2 to acetic acid.An electron transfer pathway has been proposed 35 where photoelectrons are simultaneously accepted by the membrane-associated avoprotein and ferrodoxin before being shuttled to the intracellular space via menaquinone.In a Clostridium autoethanogenum-CdS hybrid for CO 2 reduction, it has been proposed that electrons generated from nanoparticles are largely transported to the intracellular matrix via metal ion or avin-binding proteins. 36CdS nanorods 37 and graphiticcarbon nitride (g-C 3 N 4 ) 38 were combined with Cupriavidus necator for the production of the bioplastic polyhydroxybutyrate (PHB) from CO 2 or fructose.It was found that synthesised CdS nanorods with optimised shape and photoactivity improved the performance of the biohybrid system.However, the electron transfer pathway from photosensitiser to bacterium remains unclear.For other systems, more information about transmembrane electron transport is known.In a Cu 2 O/reduced graphene oxide (RGO)-S.oneidensis MR-1 biohybrid for H 2 generation, Cu 2 O acted as the semiconductor for harvesting solar energy. 39RGO nanosheets provided a surface to enable efficient photoelectron collection from Cu 2 O and efficient electron distribution to the cells.When metal reducing proteins such as MtrA, MtrB, and MtrC/OmcA (described in detail below) were not expressed, H 2 generation was signicantly inhibited.This suggested that, in this case, these specic outer membrane proteins provided the main electron transfer pathway between the semiconductor and bacterium.Most existing biohybrid systems are designed for reductive reactions that require SEDs to regenerate the photosensitiser.Systems capable of catalysing an oxidative reaction that do not require a costly SED are rare (see Fig. 1).A pioneering biohybrid system was recently developed which performs H 2 O oxidation and CO 2 reduction simultaneously, without sacricial reagents or organic mediators. 40In this system, the CO 2 -xing acetogenic bacterium S. ovata is combined with a photocatalytic sheet of nanoparticle clusters (Cr 2 O 3 /Ru-SrTiO 3 :La,RhjITOjRuO 2 -BiVO 4 :Mo).Acetate and oxygen are produced by the biohybrid system using only sunlight, CO 2 and H 2 O.A solar-to-acetate conversion efficiency of 0.7% was achieved at ambient conditions.S. ovata used both H 2 and electrons derived directly from the photocatalytic sheet to reduce CO 2 to acetate.
The formation of biohybrids using a variety of lightharvesters and many different microbes reects positively on the robustness of this concept.However, a fundamental disadvantage of some biohybrid systems discussed in this section is the challenge to explain non-specic binding and unclear electron transfer pathways across microbial membranes.Systems for which these pathways are not built in by design require extensive characterisation to understand and improve function, and are at risk of operating in ways other than the desired solar-driven pathway.Our perspective is that this can be mitigated by rationally designing a known electron transfer route to reach the site of catalysis.Microorganisms that naturally contain transmembrane electron transfer machineries provide an elegant design solution for cellular internalisation of electrons, advancing biohybrid systems for which the electron transfer pathway can be more easily dened, understood, and optimised.

Natural electron transport pathways across the microbial membrane
In nature, exoelectrogenic microbes can respire using extracellular material as a terminal electron acceptor, while electrotrophs accept electrons from the environment for reductive metabolism. 41,42This highly specic functionality has allowed exoelectrogenic bacteria, archaea, microalgae and fungi to occupy extreme environmental niches, for example those rich in minerals and subject to anoxic conditions or uctuating availability of terminal electron acceptors. 42Species across these different kingdoms of life have independently evolved mechanisms and machineries for the exchange of electrons between the intracellular and extracellular environments, commonly involving cross-membrane transport by redox mediators (Fig. 2a) and/or a chain of conductive cytochromes that act as wires through the membrane(s) and peptidoglycan cell wall (Fig. 2b and c).This natural function provides an excellent opportunity to exploit transmembrane electron transfer by coupling exoelectrogens to extracellular electron donors such as electrodes or photosensitisers.Exoelectrogens have been used extensively in systems of microbial electrosynthesis to power reductive metabolism for chemical production, and also within microbial fuel cells (MFCs), which harness electrons released from oxidative catabolism to produce an electrical current. 43,44lectron transfer mechanisms in bioelectrochemical systems have been comprehensively reviewed elsewhere. 45,46Here, we focus on the mechanistic details of electron transmission within examples of representative species, with an outlook towards integration into biohybrid solar fuel production systems.
8][49] With a fully sequenced and annotated genome, 50 it has been used extensively in MFC studies, leading to a detailed characterisation of the protein machinery between the inner membrane quinone pool and the extracellular space (see Fig. 2c). 43The outer membrane spanning Mtr complex, made up of protein components MtrA, MtrB and MtrC, is integral within this pathway.Electrons from the quinone pool are rst passed to inner membrane c-type cytochrome CymA.They are then shuttled by periplasmic electron carriers to cytochrome MtrA, a decaheme chain encased within the insulating pore of outer membrane spanning b-barrel MtrB, before emerging on the outside of the cell via lipid-anchored decaheme cytochrome MtrC. 51MtrC interacts with extracellular electron acceptors and passes electrons laterally to both other MtrC proteins and OmcA, a similar cytochrome in terms of structure and function.A specic arrangement of heme groups was revealed by the crystal structure of MtrC: the shape of a 'staggered cross'. 52Eight heme groups form a roughly linear chain perpendicular to the cell membrane, with the two central hemes also part of an orthogonal tetraheme chain.Rapid electron exchange between these heme groups enables the "nanowire" function. 53,54The junction where the heme chains intersect, comprising co-planar and T-shaped heme stacking motifs, allows electrons derived from inside the cell to travel in three possible directions to surface cytochromes or extracellular acceptors with approximately equal efficiency. 55Controlling expression levels of MtrCAB has been shown to signicantly alter extracellular electron transfer efficiency. 56f key importance for biohybrid development is the possibility to reverse the direction of transmembrane electron ow in S. oneidensis; the membrane machinery can be induced to pass electrons from the extracellular space into the cytoplasm. 52The energetics of this reversed electron transfer through the MtrCAB  58,63,85 pathway were analysed by measuring the electron ux between a counter electrode and a bacteria-coated working electrode, using the cathodic current caused by reduction of fumarate to succinate as a read-out. 57Electrons were successfully transferred to periplasmic fumarate reductase FccA.Situated at the surface, MtrC provides a potential entry point for injection of extracellular photoelectrons into the Mtr pathway within a biohybrid system.
In another well studied exoelectrogen Geobacter sulfurreducens, 58 the protein complexes and transmembrane electron transport mechanisms are less well characterised, with two tandem four-gene clusters coding for an assortment of periplasmic and outer-membrane cytochromes thought to be similar in function to the Mtr proteins of S. oneidensis. 59,602][63][64][65] Similarly to MtrC and OmcA of S. oneidensis, these Geobacter proteins have hemes packed in approximately linear chains.OmcE forms thinner lament structures than OmcS and OmcZ, and undergoes a greater number of post-translational modications, indicating a variation in properties between lament types. 65The T-stacked hemes of OmcZ have been shown to pack at a closer distance than those of other Omc proteins in G. sulfurreducens, a feature hypothesised to be responsible for the exceptionally high electronic conductivity of this protein (OmcZ laments have 1000fold higher conductivity than OmcS).Additionally, a branched arrangement of heme groups is likely integral to the crosslinking function of OmcZ laments in biolm formation. 61,62,66espite contrasting amino acid sequences, the physical arrangement of heme groups in an unbranched structure is highly conserved between OmcS and OmcE. 65Similarly, conductive lament structures recently noted in two major orders of archaea also exhibit non-homologous protein sequences yet comparable heme arrangements, indicating the convergent evolution of heme structures both within and across multiple species for optimised electron transfer. 67In the context of biohybrid design, it could be possible to wire photosensitisers directly to these laments.
In contrast to S. oneidensis, for which only inner membrane cytochrome CymA is linked to extracellular electron transport, G. sulfurreducens expresses various different inner membrane cytochromes depending on the types and redox states of available electron acceptors. 68ImcH, for example, is required for electron transfer to high redox potential acceptors, but not those below −0.1 V versus standard hydrogen electrode (SHE).The opposite response pattern was found for CbcL, 69 and others such as CbcBA are required for very specic redox potential ranges. 70Multiple available electron transfer pathways allow rapid switching in a uctuating environment.Recently, the formation of intracytoplasmic membrane structures has been discovered. 71It is proposed that these extra inner membrane folds provide a larger surface area for electron transfer proteins under low energy conditions, in order to increase the quantity of electrons able to be transmitted from the cytoplasm.This native versatility makes G. sulfurreducens a highly promising candidate for scalable biohybrid systems.3][74][75] Many exoelectrogenic archaea are also reliant on cytochromes for transmembrane electron transfer although, intriguingly, there are examples of archaea that are able to pass electrons to other microbes but do not express multi-heme proteins, indicating alternative electron transfer pathways in nature. 42Despite the gaps in knowledge that remain for many exoelectrogenic species, the certainty that there exists a natural transmembrane electron transfer pathway to be harnessed makes this class of microbes a promising choice for biohybrid systems.
Though the mechanistic details of electron transfer in exoelectrogens are continually emerging, injecting electrons into the cell is not so simple as directly reversing the natural pathway; some research has shown that there may be routes which bypass the Mtr pathway. 86Engineering exoelectrogens to enhance transmembrane electron transport could improve future systems. 87As well as further characterisation of these machineries, 57,88 the interface between photosensitiser and microbe should become a major complimentary research focus.Direct electron injection is a crucial and under-researched stage of electron transfer in biohybrid design.Rational design of this interface is required, taking into account both physical attachment and electron transfer between components.

Rational design of the photosensitiserprotein interface
Many current biohybrid systems are limited by a lack of molecular design and control at the photosensitiser-microbe interface. 17Here, we highlight a range of research that would be valuable to explore when designing this essential interface: hybrid proteins.These are systems in which a photosensitiser has been specically bound to a protein, oen an isolated redox enzyme, to create a direct and dened path of electron transfer. 16,89Fast electron transfer is necessary for the optimisation of biohybrid systems.This needs precise engineering at the molecular level; according to the Moser-Dutton ruler, a distance of #14 Å is required between the photosensitiser and the distal redox co-factor of the enzyme. 90NiFeSe]-hydrogenase enzymes 91,92 from Desulfomicrobium baculatum, among others, have been photosensitised using ruthenium-dye-sensitised titanium oxide nanoparticles. 93ttachment of TiO 2 nanoparticles to hydrogenase enzymes was inspired by the stable electrochemistry observed when the enzyme was adsorbed onto a TiO 2 -ITO electrode in comparison to graphite electrodes. 94Coupling TiO 2 nanoparticles to a ruthenium photosensitiser expanded the range of photoabsorption into the visible range, with TiO 2 extending the lifetime of the charge separated state and acting as an attachment surface for the hydrogenase.There are several systems which use the ability of TiO 2 to adsorb enzymes, including a CO 2reducing enzyme (CODH) from Carboxydothermus hydrogenoformans 95 and formate dehydrogenase (FDH) from Desulfovibrio vulgaris. 96TiO 2 -adsorbed CODH has been coupled to light absorbing Ag nanoclusters stabilised with polymethacrylic acid. 97In addition to Ru-sensitised TiO 2 , biohybrid systems are established for polymeric carbon nitride (CN X ), CN X -TiO 2 and Eosin Y as photosensitisers. 98-100A [NiFeSe]hydrogenase system photosensitised with CN X , despite a weaker photosensitiser-enzyme interaction than for Rusensitised TiO 2 , demonstrated much higher stability and remained active for 2 days compared to 8 hours. 94,98The interaction was enhanced by adding TiO 2 as a surface for adsorption (CN X -TiO 2 ). 99However, TiO 2 systems have the disadvantage of generating toxic reactive oxygen species (ROS) in the presence of O 2 .To overcome this limitation, Eosin Y was introduced as a photosensitiser.This organic dye actively protects the enzyme by reacting with O 2 to create a pseudo-anaerobic environment under air.Even in atmospheric conditions, the light-driven hydrogenase was able to maintain ∼11% of the photoactivity measured under anaerobic conditions. 100In an alternative approach, CN X has been recently demonstrated to exhibit a strong electrostatic interaction with an [FeFe]-hydrogenase from Clostridium pasteurianum 101 without the need for TiO 2 .
Light-harvesting nanoparticles such as carbon dots (CDs) have been functionalised with surface groups to engineer the nature of their interaction with an enzyme.Likely involving hydrogen bonding, FDH was found to bind CDs functionalised with either positive (-NHMe 2 + ) or negative (-COO − ) groups in equivalent amounts. 102However, the CD surface chemistry affects their localisation on the protein surface and thereby (presumably) the kinetics of photoelectron transfer: CDs with positive surface groups were able to drive CO 2 reduction to formate in the presence of a sacricial electron donor, while CDs with negative surface groups were not.This demonstrates how integral the positioning of the photosensitiser component is for successful electron transfer in biohybrid systems.Another study investigated the interaction of functionalised CDs with FccA and [NiFeSe]-hydrogenase enzymes and drew a similar conclusion. 103In a different example, the electrostatic interaction between CDs and an [FeFe]-hydrogenase enzyme was disrupted by the addition of ethylenediaminetetraacetic acid (EDTA) as an SED, with hydrogen evolution aer 1 hour around 4.4-fold higher when replaced by a different SED: triethanolamine (TEOA). 104he nature of electrostatic self-assembly between QDs and enzymes has been analysed in detail for CdS and CdTe nanoparticles coupled to hydrogenase enzymes. 105Functionalisation of these nanoparticles with a self-assembled monolayer of 3mercaptopropionic acid (3-MPA) allowed solvation in an aqueous environment and, due to the negative charge of exposed carboxylate groups, facilitated interaction with positively charged areas of the protein surface.A C. acetobutylicum [Fe-Fe]-hydrogenase, with surface-localised [FeS] clusters accessible for direct electron transfer, was adsorbed to 3-MPAfunctionalised CdS nanocrystals. 106The rate of catalysis was shown to increase with hydrogenase only up to a certain enzyme:CdS ratio before declining, an effect attributed to back electron transfer due to steric hindrance of the hydrogenase crowding.An optimised ratio of 0.67 : 1 for hydrogenase : CdS achieved ∼20% quantum yield for H 2 production.This value remained constant over a range of light intensities, indicating that the limiting step was not the rate of electron transfer or catalysis, but rather the rate of photoelectron generation.The quantum yield could thus be improved in this example by increasing the absorbance range of the photosensitiser component, but the electron transfer interface itself is considered highly effective.3-MPA-functionalised CdS nanorods/ nanocrystals and other QDs have also been adsorbed to an Azotobacter vinelandii nitrogenase [Mo-Fe] catalytic centre. 107,108he production yield when using CdS nanorods as a photosensitiser far outcompeted that of Ru-based compounds for this example, indicating the extent to which the photosensitiser component can affect the functionality of a catalyst. 109As described for hydrogenase systems, the nature of physical attachment between functionalised CdS photosensitisers and the protein is attributed to self-assembly via surface electrostatics. 106,110Physical binding of CdS to the enzyme in solution was demonstrated using dynamic light scattering (DLS), with the hybrid exhibiting a signicantly higher average hydrodynamic diameter than either individual component.2][113] 3-MPAcoated CdS nanorods have also been coupled to an oxidoreductase enzyme from Magnetococcus marinus MC-1. 114H 2 production was achieved in a similar electrostatically conjugated hydrogenase system, with 3-MPA-functionalised CdTe nanocrystals as the light-harvesting component. 110The CdTe-hydrogenase hybrid was conrmed to form a stable single complex using Native Polyacrylamide Gel Electrophoresis (PAGE).With this technique, CdTe nanocrystals were clearly seen to form a uorescent hybrid species with hydrogenase on the gel.The kinetics of adsorption were analysed: despite being weaker than a true covalent bond, the interaction was concluded to form a stable complex.Electron transfer between components was also investigated.Contrary to expectations, it was found that mixing at a lower hydrogenase : CdTe ratio resulted in much higher electron transfer efficiency and lightdriven H 2 production rate.These observations indicate that in this case, a lower ratio of enzyme per nanocrystal made for a more optimal system, a result attributed to high enzyme coverage forcing hydrogenases into an orientation less effective for electron transfer. 110A study involving uncoated CdS nanomaterials coupled to CODH additionally highlights the effect nanoparticle shape and size can have on catalytic activity. 115Due to the importance of this interface for electron transfer efficiency, the position, size and quantity of light-harvesters per microbe are important factors to optimise.
Closing the redox cycle is an important challenge to address.Very few examples of this have been achieved across any system of biohybrid photocatalysis, with water oxidation an obvious target when considering fuel production on a global scale. 40owever, other substrate candidates aside from water are feasible for oxidation in a scaled-up system, for instance biomass (lignocellulose), 116 which is already widely available and circular, or plastic 117 (which will likewise become circular). 118,119Recently, the reductive chemistry of a nanoparticle-enzyme biohybrid was successfully coupled to a valueadding oxidative reaction: TiO 2 -adsorbed FDH enzymes catalysed reduction of CO 2 to formate, while the TiO 2 photosensitiser was regenerated through oxidation of glucose to arabinose and formate. 116This example neatly demonstrates the ability to design a system which couples two redox reactions and generates multiple desirable products, eliminating the need for a costly and thus uneconomical SED.Compartmentalised Fig. 3 Photosensitiser-protein hybrid interfaces: (a) non-specific electrostatic interaction between a light-harvesting material and an enzyme (based on [NiFeSe]-hydrogenase): 103 distance dependency of electron transfer to the catalytic centre makes catalytic activity directly dependent on the location of the nanoparticle, which can be investigated but not controlled using this binding method.(b) Non-specific covalent binding of a light-harvesting material to lysine residues at an enzyme surface (based on [NiFe]-hydrogenase): 122 proximity to the active site can again be investigated but not controlled via this method.(c) Specific covalent conjugation between a light-harvesting material and a transmembrane cytochrome (based on MtrCAB from S. oneidensis): 126 binding directly via an engineered cysteine residue allows targeting to an optimal site for electron injection into the redox protein.(d) Summary of the physical photosensitiser-protein binding methods discussed in this perspective.Note: some studies [91][92][93][95][96][97][98][99]115,116,127,128,130 do not distinguish between weak intermolecular forces (such as van der Waals and hydrogen bonding) and stronger electrostatic interactions and these have not been included in the figure . photocatalysis becomes hihly important in these systems to eliminate back-reactions.
While the examples above are successful electron transfer interfaces, coupling methods that rely on non-specic adsorption are not inherently transferable to whole-cell biohybrids due to the complexity of the microbial surface. 120,121An approach where the photosensitiser is covalently tethered to the protein might solve this issue, and has two major advantages: (a) the method can be more easily applied to any system with chemically adjustable components (e.g. chemical dyes and functionalised nanoparticles), and (b) it is possible to target photosensitisers to a specic site, thus allowing optimisation of the electron transfer interface.CdS/CdSe quantum rods coated with carboxylic acid-containing ligands have been covalently bound to a [NiFe]-hydrogenase from Aquifex aeolicus via amide bond formation with lysine residues. 122Biohybrid complex formation was veried by both colocalisation of nanoparticle and enzyme on native PAGE and peptide mass ngerprint.Expanding this method to bind a targeted site on the enzyme would allow further optimisation of the electron transfer interface (see Fig. 3).4][125] Targeting the Ru electron donor site to an optimised location for electron transfer required comparing cysteine mutants at several positions on the enzyme, each of which signicantly affected catalytic rate.This highlights the sensitivity of the abiotic-biotic interface on electron transfer, and reiterates the importance of designing and investigating an optimal physical interaction for efficient electron transfer.

Coupling to membrane cytochromes of exoelectrogenic microbes
Building on the concept of interfacing redox enzymes with photosensitisers, a small number of studies have demonstrated binding light-harvesting materials directly to the membrane cytochromes of exoelectrogenic bacteria.While these membrane proteins are not enzymes themselves, the concept of creating an electron transfer interface from light absorber to redox protein for downstream catalysis is translatable between the two research areas.Coupling transmembrane cytochromes to light-harvesters and achieving electron transfer provides a convincing proof-of-principle for the ultimate application of photosensitising microbes.
As discussed, exoelectrogenic bacteria have the ability to transmit electrons between intracellular and extracellular environments. 48The transmembrane protein complex MtrCAB from S. oneidensis MR-1 provides the point of entry for extracellular electrons to access intracellular electron transfer pathways. 43,52A selection of light-harvesting nanoparticles were tested for adsorption with MtrC, the outward-facing protein subunit, assembled as a monolayer on electrode surfaces: RuP-TiO 2 -COO − (negatively charged), RuP-TiO 2 -NH 3 + (positively charged), CdS-COO -and CdS-N(CH 3 ) 2 H + . 127,128It was found that binding of MtrC was highest and most rigid for RuP-TiO 2 -COO − .The photocurrent was linearly dependent on light intensity, indicating that the electron transfer interface was not limiting.However, variation in binding properties between these different nanoparticles demonstrates the difficulty in predicting interactions between components.Non-specic adsorption (see Fig. 3) is inherently difficult to predict or control.These results reinforce the perspective that binding methods with built-in specicity, such as targeted covalent binding, are essential for guaranteeing an efficient electron transfer interface. Previously, we coupled MtrC to Ru-based photosensitisers (a) adsorbed on TiO 2 nanoparticles 130 and (b) via site-specic covalent labelling. 131The latter achieved ultrafast electron transfer on the ps timescale.CDs were introduced to this biohybrid design as a comparably sustainable source of light-harvester.The surface groups of CDs are amenable to chemical modication, and they have been previously demonstrated to pass electrons to the MtrCAB complex when mixed free in solution. 132Covalent binding was achieved through functionalisation of carboxylic acid residues at the CD surface with maleimide groups, followed by reaction with an engineered cysteine residue of MtrC.The covalent conjugation of photosensitiser-protein components was veried on native PAGE, where the conjugated protein was visualised to carry CD uorescence and migrate as a wider, more diffuse band (indicating addition of the heterogeneously sized CD particles).Electron (or energy) transfer was shown to occur between components: the uorescence lifetime of the CD became shorter when interfaced with MtrC.This chemical modication of MtrC demonstrates the use of sustainably produced, low-cost components to create a system that could in theory be translated into a whole-cell biohybrid system using S. oneidensis MR-1 for solar fuel production. 132pcA, a periplasmic cytochrome from exoelectrogen G. sulfurreducens, 133 has been conjugated to cysteine-binding Rubased photosensitisers for the same purpose.In this study, site-directed mutagenesis was used to introduce non-native cysteine residues to PpcA as a model for targeted conjugation. 134The aim was to form chiral linkages with photosensitiser components, allowing sterically specic conformations that mimic the physically constrained photosystem structures in nature.Structural aspects of the conjugates were analysed alongside rates of photo-induced electron transfer (PET).With PET rates of 6 ps, 130 ps and 35 ns for 3 different mutants, a clear preference for location of electron injection was demonstrated.It was also shown that certain mutants favoured one enantiomer of the chiral photosensitisers, indicating different environmental constraints at each position.A further study validated the conclusion that the tertiary structure of the conjugate directly controls efficiency of electron transfer, this time focussing on comparing cysteine positions introduced directly into heme binding motifs. 135Similar conclusions were reached in a study of Ru-dye sensitised STC, a periplasmic cytochrome from S. oneidensis MR-1. 136These detailed analyses of geometries, interactions and exibility at the site of conjugation are invaluable for rational design of biohybrids such as these going forwards.While PpcA and STC are less likely to be candidates for translating into a microbial biohybrid system, being located in the periplasm, these proof-of-principle studies lay the foundations for fundamental design principles: precise structural control of the point of electron injection from lightharvesting materials to the heme chains of cytochrome proteins.
The examples discussed above have been chosen to highlight advantages and disadvantages of various protein conjugation methods for translation into whole-cell biohybrid systems.Interfacing photosensitisers with redox proteins requires nuance and rational design.An optimum conjugation method will depend greatly on the nature of the components and the desired interaction.To achieve competitive electron transfer efficiencies in a biohybrid system, an evidence-based approach to directing the site of electron exchange between components is required.While it could be argued that there currently exists too little characterisation for strategic design in this manner, we propose that research should continue to pursue a deeper understanding of prospective biohybrid proteins and their interactions with photosensitisers to accelerate the innovation of successful system designs in the future.

Toxicity of non-biological components
The toxicity of synthetic light-harvesters on bacteria, and their effect on the environment when applied at a large scale, both play a crucial role in ensuring longevity, sustainability and feasibility of biohybrid systems.This aspect has received limited attention in other reviews on biohybrid systems and is not consistently mentioned in studies.As discussed previously, many different light-harvesters have been employed in biohybrid systems (Table 1).Toxicity is dependent on the type of light-harvester, the concentration, the species of microorganism, and conditions such as light intensity.In this perspective, we aim to provide a more thorough discussion of this critical factor (see Table 1 for a summary of known photosensitiser toxicities under specic conditions, including light/ dark, for the examples of solar-driven biohybrids discussed).
][25]100 MV has been identied as toxic to microorganisms.Alongside other disadvantages previously discussed, this factor makes it unsuitable for long term use.Additionally, organic dye Eosin Y has been found to be harmful for both microorganisms and human health, exhibiting increased production of ROS under visible light illumination. 137,138Ruthenium photosensitisers are specically used for anticancer therapy and as antimicrobial agents due to their cytotoxicity. 139,140Metal-based photosensitisers in general can oen result in the incidental dissolution and accumulation of toxic metal ions.The toxicity of these components in the context of specic microbial biohybrid systems, however, is not well researched.
QDs are semiconducting nanoparticles that have unique photophysical properties.The nature of the primary material, the size and shape of the nanoparticles and their specic surface functionalisation are all factors that play an essential role in their potential toxicity. 141However, the precise toxicity of nanomaterials to microorganisms and the natural environment is a serious concern that remains largely undetermined, despite extensive research in this area. 142,143Cadmium-containing QDs such as CdS and CdTe are most commonly used and have been applied in biohybrid systems involving both redox enzymes and bacteria. 5,33,106,107,110Cd 2+ exhibits high toxicity to a range of microorganisms including bacteria, algae and fungi. 144CdS nanoparticles can either be directly added to the bacteria, or formed via bio-precipitation on the bacteria in the presence of cysteine.Cell proliferation for E. coli was shown to be heavily inhibited with a Cd 2+ concentration higher than 0.1 mM in the absence of cysteine. 30On addition of cysteine, cell growth significantly increased for Cd 2+ concentrations up to 0.4 mM.However, despite this reduction in toxicity, cysteine addition to biohybrid systems has disadvantages: due to relying on excess thiol for stabilisation, there is no obvious way to couple bio-precipitated CdS systems to a meaningful oxidative reaction, and additionally cysteine can itself be metabolised by some microbial systems. 145arious QDs added directly in solution (CdTe/CdS/TGA, CdTe/CdS/cysteamine, CdTe and CuInS2/ZnS/PMAL) have been evaluated for their nanotoxicology with S. oneidensis MR-1, using a colony-forming units method and a uorescence viability assay. 146CdTe and CuInS 2 /ZnS/PMAL QDs displayed no toxicity, but CdTe/CdS/TGA, CdTe/CdS/cysteamine QDs signicantly inhibited the bacteria growth.In another study, seven different core-shell QDs were tested for toxicity to nonphotosynthetic bacteria (A.vinelandii and C. necator) for the purpose of solar fuels and chemicals generation. 147The surfaces of the QDs were each capped with ligands that affected surface charge: negatively charged MPA, positively charged cysteamine (CA), and zwitterionic cysteine.The Cys-capped QDs showed no inhibition on bacteria growth.However, MPA-capped QDs impaired growth moderately, and CA-capped QDs strongly, clearly illustrating how surface chemistry affects the toxicity of nanoparticles.It has also been demonstrated that activity in D. desulfuricans-CdS biohybrids decreased when cadmium concentrations surpassed 3 mM, likely due to toxicity. 28In the S. ovata-CdS hybrid system, a colony-forming unit assay showed that cell viability in the CdS-containing biohybrid was decreased compared to that of cells without CdS, further indicating toxicity of Cd-based photosensitisers. 35Toxicity of CdS was tested on the microorganisms C. necator, Saccharomyces cerevisiae, Gluconacetobacter xylinus, and Anabaena sp.PCC7120 using different concentrations of CdS nanorods, with toxicity observed for all except C. necator. 37Nanoparticles can also have toxic effects on the metabolism of different microbes, including protein expression.An example is Xanthobacter autotrophicus in which 727 proteins were upregulated and 53 downregulated upon addition of CdTe nanoparticles. 148Proteomic and metabolic analyses show that signicant alterations occurred for electron transport chain and redox signalling pathways.Expression of N 2 -xation related proteins such as nitrogenases was increased, while expression of ATP synthase and overall ATP concentration was lowered.Interaction between S. oneidensis MR-1 and several nanoparticles including CdS has been shown to increase biolm formation. 149Addition of TiO 2 has been shown to increase riboavin secretion. 150In conclusion, it is difficult to predict the toxic or metabolic effects of different light-harvesting materials on different organisms.
Gold nanoclusters (AuNCs), considered to be a less toxic material than cadmium-containing QDs, were used as photosensitisers (accumulating intracellularly) for the acetogenic bacterium M. thermoacetica. 151Both with and without the presence of AuNC, the M. thermoacetica system could maintain a relatively high proliferation rate and cell viability until the fourth day, demonstrating biocompatibility of these AuNCs.Recently, carbon-based nanomaterials such as CN X and CDs have been applied to biohybrid systems for photobiocatalysis. 98,99,103,104,152CDs are generally considered to be environmentally benign.Biocompatibility with S. oneidensis MR-1 was evaluated by assessing the viability of bacteria incubated with CDs.Cell viability was found to be maintained >96% aer 94 h, conrming the high compatibility and low cytotoxicity of the CDs for this bacterium.They have, however, been implicated in increased intracellular signaling. 153Organic semiconductors such as perylene diimide derivative (PDI) and poly(uorene-cophenylene) (PFP) are also considered non-toxic as a photosensitiser for biohybrid semi-articial photosystems. 154n comparison to organic dyes, nanoparticles offer greater potential for adjustment of both their band gap and surface chemistry.Their high stability and durability are advantageous.However, the impact of nanoparticles both on the environment and the bacteria remains uncertain, with toxicity varying depending on the species of microbe involved.To create a more sustainable system, it is imperative to focus on development of benign photosensitisers such as gold nanoclusters, organic semiconductors, and carbon-based nanomaterials which exhibit biocompatibility and lower toxicity.

Outlook
The aim of this perspective is to pitch a forward-looking "roadmap" for the future of whole-cell solar biohybrids, highlighting a range of research areas that can be explored for solutions to limitations of current systems.Exploiting microbes with natural pathways of transmembrane electron transfer provides a characterisable route for internalisation of photoelectrons, and strategies for conjugating photosensitisers to enzymes can be explored for transferable electron transfer pathways that facilitate a controllable and optimisable interface.Rather than providing a conclusive set of studies, we propose that the eld continues to examine research across these areas for inspiration, harnessing new materials, components and conjugation methods from a wide range of studies for application in the whole-cell biohybrid vision.
The main focus has been to highlight methodology used to anchor two redox components together in a such a way that electron transfer is achieved.However, it is also possible to examine protein conjugation methods for purposes such as energy transfer.While electron transfer requires a distance of #14 Å between redox centres (according to the Moser-Dutton ruler), energy transfer requires a distance of a 100-500 Å. Physical conjugation methods can nonetheless be explored.6][157] CdTe QDs were bound with high affinity to an engineered His-tag with a modiable linker sequence at the RC surface. 156,158Methods to conjugate lightabsorbing organic dyes and uorophores have been established that exploit the presence of available cysteine and lysine residues on the RC.A synthetic aryleneethynylene uorophore, designed to act as an antenna pigment, was covalently bound to lysine residues via a succinimidyl ester group at a ratio of 4.1 ± 0.3 uorophores per protein. 159In another study, uorescent dyes functionalised with maleimide groups were targeted to engineered cysteine residues. 160These types of studies highlight further translatable conjugation methods for targeting lightharvesters to specic locations on a protein.
In a different approach, Texas Red (TR)-modied phospholipid has been used for energy transfer to light-harvesting complex II (LHCII) from spinach. 161,162In this system, the light-havester is not coupled to a protein, but to the membrane.Further studies expand the concept to include dye-conjugated phospholipids ATTO647N-DOPE and Cyanine-7-DOPE, and lipophilic dialkylcarbocyanine dyes DiI and DiR. 163This type of interaction demonstrates the possibility of designing an energy (or electron) transfer interface between membrane-anchored components via proximity rather than direct physical conjugation.However, organic dyes do not absorb a broad spectrum of light, and are less durable than alternative photosensitisers such as semi-conducting nanoparticles.COOH-coated semiconducting graphitic CDs have been modied with a phospholipid. 164Chemically functionalising membrane components with nanoparticles such as this could be highly applicable to the overall biohybrid concept.
An alternative to labelling natural membrane components is to create synthetic membrane-insertable photosensitisers such as amphipathic uorescent nanoparticles.Lipophilic CDs have been developed that undergo self-assembly into structures similar to liposomes (CDsomes). 165They have been used for selective labelling of membrane structures in live mammalian cells, demonstrating high biocompatibility and negligible cytotoxicity, along with the key ability to insert into natural cellular membrane environments.Due to their ability to photoluminesce, these CDs have additionally been investigated for use in photocatalysis. 166At 365 nm, CDsomes exhibit a uorescent on-state, and at 530 nm a non-uorescent off-state.Using this photo-switchability, it was found that in the on-state catalysis of O 2 to H 2 O 2 was driven, and in the off-state ROS were generated.The principle of using a synthetic membraneembedded CD for photochemistry is fascinating for further research.In another study, a synthetic transmembrane photosystem I mimic has been developed that, on excitation by visible light, is able to pass electrons to an organic dye. 167The oligo-uorene chromophore structure spans the width of an average phospholipid bilayer, situating itself in a perpendicular orientation that mimics natural transmembrane structures.Photons are channelled to an energy acceptor, generating photoelectrons that can be transferred to acceptors at the membrane surface.Energy transfer between the synthetic light-harvester and organic dye Eosin Y was demonstrated in liposomes by measuring uorescence emission quenching.Another system uses a perylene diamine chromophore that imbeds in liposomal membranes. 168Use of liposomes as a platform for energy transfer, electron transfer and chemical conversion is a branch of its own within the eld of developing bio-based systems for solar fuel generation. 169,170Within the context of designing whole-cell systems, the relevance of synthetic membrane components such as these can only be judged by their ability to insert into the biological membranes of microbes and interact with native components.For many of these synthetic chromophores, this remains to be seen.
In conclusion, we propose that a broad range of research be used to inform the strategic design of biohybrid systems.The route of a photoelectron, from generation to intracellular catalysis, can be broken down into two main optimisable stages: the interface between photosensitiser and microbe, and internalisation of the electron by the microbe.Delving into the fundamental interactions between photosensitiser and redox protein components will reduce the unknowns when translated into larger, more complex systems.Using known methods for directly tethering key biohybrid components can allow control and optimisation of their interaction.In many cases, there is not yet enough understanding of how the building blocks interact: it is therefore of high importance to continue investigating and engineering these materials and interfaces, and to ultimately use this accumulated knowledge to establish blueprints for the future rational design of solar-driven biohybrids for fuel production.

Table 1
Properties of selected biohybrid systems