Chapter 1: Making Maquette Models of Bioenergetic Structures

Extensive work developing X-ray crystal structures of core natural bioenergetic proteins exposed the arrangement of hundreds to thousands of amino acids that surround and bind a range of different cofactors that support bioenergetic light absorption and respiratory and photosynthetic electron and proton transfer.^1–12  Such atomic resolution offers the promise of a detailed understanding of how these naturally evolved systems function to direct electron and proton transfer to support life, or how they may dysfunction in disease.

Indeed, such structures have provided a good appreciation of how electrons tunnel from one redox centre to another across the intervening electrically insulating protein medium to form chains of redox centres that connect catalytic centres of bond breaking and making. A survey of dozens of individual bioenergetic electron-transfer reactions, both productive and unproductive, using both natural and unnatural redox centres, has shown that a relatively simple formula depending on only three parameters captures the quantum mechanics adequately to provide estimates of electron-tunnelling rates within an order of magnitude.^13,14  By constructing a matrix of electron-tunnelling rates between all redox centres in a protein, we can model the progress of electron transfer to reveal the relative weaknesses and strengths in natural electron-transfer chain design.^15–17 

However, factors other than electron-tunnelling distance can dominate performance, especially at the catalytic centres terminating electron-transfer chains.¹⁸  Here, the complexity of natural bioenergetic protein structures obscures which engineering elements are critical for function and which elements may be incidental remnants of an evolutionary legacy. While mutation of individual residues can verify their functional importance, mutation imperfectly clarifies the residue's role. This is because natural selection has an inherent tendency to create systems in which component parts become progressively interdependent^19,20 —the protein-level equivalent to Muller's genetic ratchet.^21,22  Previously independent amino acid residues came to depend upon one another for continued function. Thus mutation of a residue with only indirect support of catalytic function may devastate function and seem to be of central importance. In addition, any one amino acid in natural proteins tends to support many functions, a molecular expression of Darwin's principle of multiple utility;²³  for example, a particular amino acid may support not only chemical activity, but also protein folding, stability and dynamics.^24,25 

The need to unravel this complexity inspired design and construction of simplified maquettes of natural proteins aiming for intentional simplicity, robustness and independent function. De novo designed proteins use first principles of protein folding and are untouched by natural selection. These constructs can be exploited to reverse-engineer natural protein function and rigorously test hypotheses. The basic functional principles resolved in this way may then be used to construct novel protein systems that introduce new functions or surpass natural proteins in robustness and performance.

De novo design and construction of cofactor binding protein maquettes exposes the default properties of cofactors in a protein matrix independent of natural selection and gain insight into how redox proteins may have operated early in the evolution of life.^26,27  Such designs help us to isolate and understand the means by which protein can be engineered to manipulate cofactor properties towards desired function.²⁸  Furthermore, by independently recapitulating critical properties of natural bioenergetic proteins in designed constructs, we gain confidence in protein engineering principles—the maquette approach is the redox-protein counterpart of Feynman's principle “What I cannot create, I do not understand”.²⁹  This chapter surveys the present state of cofactor functionalized bioenergetic maquette design.

De novo protein design is highly diverse, including architectures quite unlike natural proteins. For example, some early maquette designs sought to guide protein assembly through a template assisted synthetic protein (TASP) strategy whereby separate helical peptides were site specifically attached to a synthetic cyclic peptide template.^30–32  In this chapter, we will focus on designs that can be expressed in a cell; such designs can benefit from integration with natural cellular biochemistry and offer the possibility of conferring new functions to the organism.

De novo protein designs commonly exploit the fundamental principle of binary patterning of amino acid heptad repeats to promote spontaneous helical bundle association.^33,34  Hydrogen bonding between alpha-helical amide backbone oxygens and nitrogens folds two helical turns for every seven amino acids. Selecting a sequence in which residues on one face of the helix are non-polar, while the remaining are polar, drives helical self-association to bury hydrophobic residues in the bundle core. The structurally resolved two-helix coiled-coil protein tropomyosin³⁵  was used as an early model, in which heptad positions were labelled sequentially from a through g, with the a and d positions associated with core hydrophobic residues. By shifting the hydrophobic : hydrophilic balance from two to three non-polar residues per heptad at positions a, d and e, four-helix bundles form spontaneously.

For example, the heptad repeat LQQLLQX where L is Leu, Q is Gln and X is E (Glu) for one sequential pair of helices and K (Lys) for the other pair, forms a fully antiparallel helical bundle arrangement common in natural helical coiled-coil proteins. High-affinity wrapping of the helices in a left handed coiling is driven by the “knob-into-hole” interactions first predicted by Francis Crick, in which amino acid side chains (knobs) pack into spaces between the side chains on adjacent helices (holes).³⁶  Connecting the four helices into a single chain with Gly rich loops facilitates independent adjustment of the sequences of each helix, assisting in site-specific binding of redox cofactors and permitting surface residue salt-bridges between helices that stabilize the structure. For example, in the sequence just described, negatively charged E residues in helices one and two salt-bridge with positively charged K residues in helices three and four when helices are threaded in a counter-clockwise pattern, as shown in Figure 1.1a. It is not uncommon to generate helical bundle maquettes that resist thermal unfolding even at 100 °C.^25,37,38 

A single chain also facilitates the use of negative design, in which plausible alternative assemblies of helices into a bundle are destabilized. Negative design was effectively demonstrated by the DeGrado group with single-helix peptides of different charge patterns that assembled into heterotetrameric four-helix bundles only when component peptides with stabilizing charge pairing interactions were mixed with the proper stoichiometry; when component peptides were kept separate, the peptides were unstable.^39,40  In the single chain example of Figure 1.1, the alternative helical threading, in which the positions of helices two and four are swapped (Figure 1.1c) buries hydrophobic residues just as well; however this threading will be electrostatically disfavoured as it pairs the negatively charged E residues at one helical interface and the positively charged K residues at another interface.⁴¹  Thus surface charge patterning provides both positive and negative design to reliably force a particular helical bundle threading, as verified by X-ray crystallography (Figure 1.1b).

A great advantage of helical coiled-coil designs is adaptable modularity. Helices can be lengthened or shortened by extending or truncating the helical heptad pattern without disturbing the local protein folding. This frees the designer of a coiled-coil maquette to employ electron-tunnelling theory to select the optimal number of cofactors and inter-cofactor distances to support desired electron-transfer function on predetermined timescales. It also promotes compact designs lightening the expression load on the organism. Currently, maquette helices range from five to 13 turns for a length of 26 to 62 Å. The individual helices of a binary patterned four-helix bundle assembly can be extended to include regions of predominantly non-polar residues to facilitate transmembrane insertion into lipid bilayers. In this way a redox cofactor electron-transfer chain can extend from the aqueous phase to cross the low dielectric region of the membrane,^42,43  much like cytochrome bc₁.

Of the 20 natural amino acids commonly coded into proteins, the nitrogen of His and the sulfur of Cys are frequently employed to ligate the metals found in a range of biologically common redox cofactors including hemes, iron–sulfur clusters and light-activated tetrapyrroles. Inserting these residues into the binary patterned sequences anchors cofactors at locations prescribed by electron-transfer theory. For the ligation of metal cations in redox clusters, Cys and His can be used alone or in combination with the carboxylic acids of Asp and Glu.

Taking inspiration from heme binding natural four helix bundle proteins cytochrome b₅₆₂ and cyt bc₁, early heme maquette designs bound four hemes per helical bundle through the elementary strategy of placing a pair of His residues at “a” positions separated by four turns on each helix⁴⁴  while placing bulkier core Phe residues midway between to separate the helices for heme binding. As the helices self-associate, His residues in the helical core form two clusters at opposite ends of the bundle.⁴⁵  Added heme B initially partitions into the hydrophobic bundle core, and then recruits a bis–His ligation of the heme iron.⁴⁶  Relatively close packing of hemes in the bundle core leads to conspicuous heme–heme electrostatic interactions—reduction of each heme makes the reduction of the next heme less favourable. This leads to a wide split in heme redox midpoint potentials (E_m values) within the same heme maquette, ranging from −80 to −230 mV. Heme redox potentials are also shifted electrostatically by modulating the charge patterning on the maquette surface.⁴⁷ 

Heme A and Heme C bind to bis–His sites in maquettes,^48,49  as do many iron-porphyrin variants.⁴⁶  For four-helix bundle maquettes with bis–His sites in the core, strong Fe porphyrin binding requires an amphipathic character for the tetrapyrrole peripheral substituents. Polar groups (such as propionates on protoporphyrin IX) should be placed on a single or at least adjacent edges of the macrocycles and non-polar groups on remote edges. This provides a balance between tetrapyrrole water solubility with less aggregation in the aqueous phase, and maquette hydrophobic core solubility for heme partitioning into maquettes before bis–His ligation.⁴⁶ 

Early heme maquette designs were structurally characterized only in the apo-form by X-ray crystallography.⁵⁰  Recently, crystal structures of heme containing maquettes of the coiled-coil design (Figure 1.2) have been achieved to reveal details of the heme structural environment. Despite the intentional absence of sequence similarity between maquettes and natural heme proteins, the maquette heme environment resembles cyt bc₁. This suggests that structure of natural proteins at heme sites may be largely a consequence of the helical bundle fold. One difference between heme in the larger, membrane embedded cyt bc₁ and the smaller and water-soluble coiled-coil maquette is that heme propionates in the maquette are stabilized by access to the aqueous phase towards the end of the bundle. The ∼10 nM bis–His heme binding affinity in these coiled-coil designs results in spontaneous incorporation of heme during in vivo expression in E. coli.

Heme binds to these designs in many different ligation motifs besides bis–His, replicating many of the heme binding geometries seen in natural proteins (Table 1.1).⁴¹  Such ligation alternatives modulate electron transfer properties to support engineering functions beyond electron-transfer chain action. Notably, a myoglobin-like His–Ala version of this maquette binds oxygen to the ferrous state stably for days without auto-oxidation.

As heme binds to bis–His sites stronger than to single His sites, mixing bis–His and mono-His sites within a maquette enables site specific loading with different tetrapyrroles. Light activatable, Zn metallated tetrapyrroles, including porphyrins, chlorins and bacteriochlorins, bind readily to mono-His sites.^51,52  Zn tetrapyrroles are attractive for light-activated electron transfer designs because long-lived triplet excited states form with high yields to facilitate longer-range electron transfer with other redox cofactors. Light-activatable Mg chlorins bind to maquettes as well.⁵³  Indeed, when expressed in chlorophyll producing algae, natural Mg chlorins can spontaneously insert into maquette His sites. Although not a covalently bound cofactor, carotenoids can also spontaneously insert in the hydrophobic core during maquette expression in algae. The cobalt analogue of heme B (Co protoporphyrin IX) readily binds to maquette potentially providing a means of catalytic reduction of protons to hydrogen.⁵⁴ 

Cys groups inserted near His in a CXXCH motif attack heme B vinyl groups to form thio-ether linked heme C analogous to the natural cytochromes c.^49,55,56  Although this reaction takes place in vitro, heme ligation is improved through interaction with natural heme lyases during in vivo expression. Even though de novo maquette sequences are entirely different from the natural substrates of heme lyases, the coupling interaction is efficient.

Similarly, Cys reacts with the vinyls of open chain tetrapyrrole bilins, components of the natural light-harvesting phycobilisome system. As with heme, maquettes interact with natural lyases to enhance binding of a range of bilin types and colours including phycocyanobilin and phycoerythrobilin. Bilin binding efficiency depends upon the positioning of the Cys residue in the maquette sequence. However, when present in sufficient concentrations in the cell, bilins such as biliverdin will also spontaneously associate with maquette hydrophobic cores, even without anchoring Cys residues. Biliverdin ligates to maquettes without the assistance of lyases during expression in mammalian cells, forming an analogue to natural phytochromes.

When exposed to iron sulphide under reducing conditions, Cys binds directly to Fe₄S₄ clusters, analogous to the ubiquitous and primeval iron–sulfur redox proteins.⁵⁷  A 16 amino acid loop sequence is sufficient to create a ferredoxin analogue.⁵⁸  Inserting this sequence as a pair of loops into heme binding maquettes creates a compact six-redox centre ferredoxin-heme maquette.⁵⁹  Complementing the Cys loop with three nearby His anchors a Ni(ii)–Fe₄S₄ bridged cluster, an analogue to natural carbon monoxide dehydrogenase.⁶⁰  Fe₄S₄ clusters also form with Cys residues in the helical core.⁶¹  A maquette beta sheet fold has been used to form a single iron, tetra-thiolate centre as a rubredoxin analogue.⁶² 

Sulfur chemistry is also readily exploited for site-specific cofactor anchoring in maquettes through means not found in natural proteins. Single Cys residues anchor suitably activated cofactors such as halogen-activated flavin^63,64  or quinone,⁶⁵  or maleimide-activated ferrocene⁶⁶  or bacteriochlorin.⁶⁷ 

Maquette His and Cys residues have been used to ligate individual metal ions. The Pecoraro group has engineered three-helix bundle maquettes that tris-thiolate bind Hg(ii)⁶⁸  or As(iii),^69,70  and tris–His bind catalytically active Cu(i/ii)⁷¹  or Zn(ii).⁷²  Site specific metal ligation is seen as Hg(ii) ions bound preferentially to a tris-thiolate site while Zn(ii) ions bound to a tris–histidine site at the other end of the maquette.⁷² 

Clusters of acidic carboxylate residues Asp and Glu together with His, often bind redox active and inactive metal cations in natural proteins. The four-helix bundle frame of natural bacterio-ferritin inspired the design of a series of metal binding maquettes, some of which are catalytically active.⁷³  These maquettes bind bivalent metal ions with relative affinities of Mn<Fe<Co<Zn>Cd,^73–75  consistent with the Irving–Williams series.⁷⁶  These de novo designs have recently been adapted to bind combinations of redox active tetrapyrroles and metal clusters for extended chains of light and electron-transfer active cofactors. X-ray crystal structures closely reproduce the metal ligand arrangement of natural bacterio-ferritin even in the extended designs (Figure 1.6).

The amino acids Tyr and Trp function as high potential redox centres in natural electron-transfer chains and show similar redox properties in maquette environments.^65,77–82  Maquettes form light-activated flavin-Trp radical pairs that show conspicuous magnetic field sensitivity, reproducing essential elements of cryptochrome.⁶⁴  For these non-metallic amino acid redox centres, proton-coupled electron transfer rates may be slower than electron tunnelling and reflect protein dynamics.

Artificial redox cofactor amino acids have been inserted in maquette sequences using a TAG Amber codon and engineered mutant t-RNA synthetase to insert propargyloxyphenylalanine; this group reacts with azido-functionalized redox centres using click chemistry.⁶⁶  Artificial redox active amino acids have also been inserted into maquettes through intein semi-synthesis.^83,84 

Many natural bioenergetic proteins are composed of catalytic sites, where oxidative and reductive bond making and breaking take place, connected by chains of redox cofactors that relay electrons over distances of four to 14 Å by electron tunnelling through the intervening insulating protein medium.¹³  Although the effective height of the electron-tunnelling barrier between cofactors can be altered by making the medium more or less densely packed with single and double bonds, in practice natural selection has not used this means to speed productive electron transfers and slow unproductive, energy wasting electron-transfer reactions. Instead, Nature relies on the exponential dependence of the electron-tunnelling rate with the edge-to-edge distance between cofactors to guide electron transfer by means of proximity: at any step, productive electron transfer is generally shorter than unproductive electron transfer.

The driving force dependence of electron-tunnelling rates is approximately Gaussian.⁸⁵  As the driving force (−ΔG) increases from zero, the electron-transfer rate increases. However, it eventually reaches a maximum rate at a driving force defining the reorganization energy (λ). Increasing the driving force still further, into what is called the Marcus inverted region, causes electron-transfer rates to drop. This Marcus rate dependence is captured in eqn (1.1), where the log rate is in units of s⁻¹ and ΔG and λ are in units of electron volts (eV).

In this classical Marcus picture, electron-tunnelling rates show conspicuous temperature dependences when driving force does not match reorganization energy. However, working with an extensive collection of quinone electron-transfer reactions in photosynthetic reaction centres over a very wide temperature and free energy range, Gunner and Woodbury demonstrated that quantized nuclear vibrations enforce a more modest temperature dependence.^86,87  The energy of the typical characteristic frequency of vibration coupled to intraprotein electron tunnelling in natural proteins is ∼0.07 eV,¹³  larger than the 0.025 eV Boltzmann thermal energy near room temperature.

A survey of many intraprotein electron-tunnelling rates for systems with known distances, driving forces and reorganization energies led to the development of a practical electron-tunnelling rate expression¹³  that is used not only to gain insight into natural protein design, but also to design novel electron-transfer proteins.

Eqn (1.2) provides an electron-tunnelling rate for an energetically favourable reaction near room temperature in units of s⁻¹ when the R is in units of Å, and ΔG and λ are in units of eV. To estimate the rate of an uphill, endergonic reaction, the rate of the reverse exergonic reaction is first calculated using eqn (1.2), and then a rate penalty of one order of magnitude for every 0.06 eV of driving force is imposed. This is consistent with the exergonic–endergonic behaviour described by Woodbury and Gunner.⁸⁷ 

We combine the ability to estimate electron-tunnelling rates between cofactors with the facility to move maquette cofactors by adjusting the position of cofactor anchoring amino acids in maquette sequences. Designs are vetted by simulating electron-transfer performance, given the constraints of cofactor redox potentials, excited state lifetimes, etc., by solving a system of differential rate equations. Electron-transfer rates are expected to be limited by electron-tunnelling provided there is no diffusive movement or bond making/breaking coupled to the electron transfer, such a proton or hydride transfer.

Figure 1.3 shows an example of an elementary maquette design including a light excitable tetrapyrrole pigment and a tetrapyrrole electron acceptor anchored to histidines in the interior of a four-helix bundle, comparing the differences in optimal design for long-lived charge separation yield originating from the excited singlet or triplet state. The vertical axis adjusts the driving force of electron transfer from the excited pigment state P* to an electron acceptor A. This might be selected by changing the peripheral groups or protein surroundings of the P and A cofactors. The horizontal axis varies the edge-to-edge distance between P and A. This is adjusted by selecting the spacing along a helix of the cofactor anchoring amino acids, typically in increments of one helical turn, or about 5.2 Å. Optimal designs in simple cofactor dyads avoid losses that return the system to the ground state by radiative decay or rapid charge recombination. The figure illustrates that singlet systems should place histidines near three helical turns apart, while triplet systems with longer lifetimes are better if anchoring His are nearly five helical turns apart.

In constructing electron-transfer chains, it is not generally necessary that each successive electron transfer be energetically downhill (exergonic) provided the cofactors are placed close enough to one another that even uphill (endergonic) electron-tunnelling reactions are relatively fast. Indeed, many natural electron-transfer chains show an up and down energy profile.¹⁴  Catalytic bond-making and breaking generally involves pairs of electron transfers. This means that catalytic sites should be designed to accumulate multiple oxidizing or reducing equivalents, for example by placing multiple redox centres in close proximity, so that a relatively slow uphill electron transfer to a catalytic intermediate can be followed by a fast downhill electron transfer to a more stable redox state.

In natural proteins, the 6th ligation position of heme iron may bind small ligands, such as molecular oxygen, to either promote electron transfer (such as P450 or cytochrome oxidase) or suppress electron transfer, as in oxygen transport. Natural O₂ transport proteins myoglobin and haemoglobin are ligated by a single His residue with an open 6th coordination site that may be occupied by a water molecule easily displaced by ligands such as O₂, NO or CO. In natural transport proteins such as neuroglobin and cytoglobin, heme is bis–His ligated. One His residue must be displaced to bind O₂, providing a means for these globins to modulate O₂ affinity.⁸⁸ 

An early single-chain, two heme B bis–His maquette demonstrated the ability to bind both CO and O₂,⁸⁹  much like natural neuroglobin. Surprisingly, heme C bis–His^49,56  and His–Cys maquette designs also bind O₂. However, auto-oxidative electron transfer between the heme and O₂ in these initial designs on the tens of seconds timescale is much faster than the hours to days timescale typical in natural globins. The latest generation of heme maquette designs appear to have less mobility and less water access around the heme iron. The net result is oxy-ferrous lifetimes of 1–2 days, much like natural globins. Remarkably, maquette designs generally bind O₂ better than CO, in stark contrast to natural O₂ transport globins for which CO is effective poison. These properties, plus the high thermal stability of maquette designs, make them an attractive starting point for the development of an emergency oxygen transport blood supplement during treatment of trauma.

Maquettes bind a wide variety of light activated Zn metallated tetrapyrroles, including porphyrins, chlorins and bacteriochlorins (Figure 1.4).⁵¹  Assembling tetrapyrroles with different spectral properties within the maquettes leads to intraprotein excitation energy transfer (EET) and emission from the lowest energy chromophore.^51,67  Biliverdin, PEB, PCB and other fluorescing bilin chromophores bind to maquettes to support light-harvesting-antennae-like energy transfer to circular tetrapyrroles. In a design intended to couple maquette EET with natural phycobilisomes, the phycobilisome protein subunit CpcA has been fused to a maquette binding two different tetrapyrroles for three-step energy transfer (Figure 1.5).⁶⁷ 

Maquettes interact with a number of different natural proteins while being expressed in vivo, notably with heme and bilin lyase as well as membrane transporters that facilitate heme C formation. Maquettes also participate in productive electron transfer with natural proteins in vivo. For example, single-electron reduction of a maquette heme A by NADPH is moderated by the two-electron flavo-protein ferredoxin-NADPH reductase,⁴⁸  the natural redox partner of PSI. Transmembrane heme B maquette⁴³  interacts with decyl-ubiquinol and water soluble cytochrome c to reproduce quinol-cytochrome c oxidoreductase activity analogous to natural cytochrome bc₁.⁹⁰ 

Stopped flow studies show that interprotein electron-transfer between water soluble heme maquettes and natural protein cytochrome c is electrostatically modulated by ionic strength and the charge pattern of surface residues,⁹¹  analogous to the modulation observed in natural electron-transfer proteins. Details of faster intraprotein electron transfer are resolved by CO-photolysis. Just as O₂ binds to some heme maquettes, when ferrous bis–His ligated maquette hemes are exposed to sufficient concentrations of CO, one histidine is typically displaced by CO. A laser pulse releases CO reforming the bis–His state with a relatively low redox midpoint potential. This bis–His heme is now competent for rapid reduction of higher potential cofactors such as heme C in natural cytochrome c.

In multi-cofactor maquettes, both Zn tetrapyrroles and flavins engage in kinetically resolvable light-activated intraprotein electron transfer.

Zn tetrapyrroles are attractive chromophores for initiating electron transfer because intersystem crossing from excited singlet to triplet states is relatively efficient (30–70%) with triplet excited state lifetimes nearly six orders of magnitude longer than singlet lifetimes. According to eqn (1.2) and Figure 1.3, this means redox partners can be placed up to 10 Å further away and still achieve comparable yield of light-activated charge separation. The typical ∼7 ms lifetime of excited triplet Zn porphyrin is shortened by placing a heme within electron-transfer distance.⁹²  With mono-His and bis–His sites on different pairs of helices at opposite ends of the bundle with a separation of four helical turns, the edge-to-edge electron-tunnelling distance is ∼20 Å. Light-activation reveals ∼0.8 ms heme B photo-reduction at a driving force of 0.52 eV. Eqn (1.2) suggests a plausible reorganization energy for this intraprotein electron transfer of ∼0.9 eV, which would place the ∼1.2 eV charge recombination to ground state in the Marcus inverted region but closer to the estimated reorganization energy. Thus heme is transiently photo-reduced, with charge recombination about three times faster than the charge separation.

A transmembrane maquette has been engineered with three tetrapyrrole cofactor binding sites.⁴³  The ligating histidines of the middle site lie three and four helical turns from the two terminal binding sites. Binding both Zn porphyrin and heme B cofactors leads to 20 μs light induced electron transfer from the pigment to the heme acceptor, a rate consistent with the shorter and faster three helical turn tetrapyrrole spacing.⁴³ 

In natural bioenergetic proteins, flavins typically interface n=1 and n=2 electron redox centres in natural electron-transfer chains such as in respiratory Complex I and Complex II, exploiting the relative stability of the half-reduced flavin semiquinone state as an intermediate. However, flavins are light-activated as photochemical sensitizers in photo-lyase or blue light and magnetic field sensors as in cryptochromes. While most natural flavoproteins bind flavin non-covalently, in nearly 10% of cases, flavins are covalently linked. We use the covalent linkage strategy in flavo-maquette construction to hold flavins at appropriate distances for intraprotein electron transfer.^63,64  Light activation of the maquette flavin initiates photo-oxidation of sufficiently close tryptophan residues, analogous to the action of natural cryptochromes.⁶⁴  Similar to natural cryptochromes, reduced flavin-oxidized Trp radical pairs in these maquettes with lifetimes in the microsecond domain are conspicuously sensitive to applied magnetic fields. Alternatively, in the presence of another electron donor, photo-reduced flavin can reduce nearby heme, analogous to the action of flavo-cytochrome b₂.⁶⁴ 

It is difficult to achieve light-induced charge separation lifetimes long enough to support millisecond or slower redox catalysis in two-cofactor photochemical dyads. As Figure 1.3 shows, competition between productive charge separation by electron tunnelling and the typically few nanoseconds or faster decay of the chromophore excited singlet state, leads to charge separation lifetimes of hundreds of nanseconds at best. Speeding charge separation by increasing the driving force tends to speed charge recombination with the loss by P+A- short-circuiting. Lengthening excited-state lifetime by intersystem crossing to a triplet state enables longer distance electron transfer and longer charge separated lifetimes, albeit with some loss of electrochemical energy in singlet to triplet conversion.

Stabilization of charge separation for timescales long enough to support millisecond or slower catalysis requires the addition of at least one more cofactor to permit multistep, long-distance charge separation. The number of cofactors required will depend upon the energetics of the system, the required lifetime for catalysis and whether charge separation is initiated from the excited singlet or triplet state. An example of the expected performance of triplet-borne charge separating cofactor triad is shown in Figure 1.7. As in the cofactor dyads of Figure 1.3, the expected charge-separation yield is shown as dependent on two parameters that can be adjusted by design: the driving force for P to A electron-tunnelling, and the P to A edge-to-edge distance. For this figure, we included a triad donor D at 4.6 Å from P with a driving force of 0.19 eV, in the range expected for a tyrosinate donor to a metallated tetrapyrrole cation. The charge separation yield at 0.1 ms in the dyad (Figure 1.3B) is much smaller than in the triad (Figure 1.7A). Indeed, electron-tunnelling analysis indicates that charge separation should persist in cofactor triads out to tens of milliseconds (Figure 1.7B). In our experience, maquette triads can achieve charge separation lifetimes of hundreds of milliseconds.

Singlet-borne charge separation is more demanding because the initial charge separation must compete with typically nanosecond or shorter excited state decay lifetimes, which requires a shorter distance for the first step of charge separation. Under these circumstances, a cofactor tetrad will be required for millisecond or longer charge-separation lifetimes.

Not all biological electron-transfer reactions in native electron-transfer chains are rate limited by electron tunnelling. A classic example is proton-coupled electron transfer during the oxidation of tyrosine, an essential step in water oxidation catalysed by PSII. The redox potential of deprotonated tyrosinate in maquettes is close to 0.74 V,⁷⁹  which provides a favourable driving force for oxidation by a photo-oxidized Zn porphyrin cation with an E_m value near 0.91 V. At pH 9.5, approaching the reported pK_a of tyrosine of 10 to 11,⁹³  we observe tyrosinate oxidation in maquettes using a Zn porphyrin excitable pigment with an E_m of about 0.91 V together with a photo-reduced heme acceptor. The charge separation lifetime is more than 100 ms.

At pH 7.5, where the majority of Tyr is protonated, the appropriate E_m value for estimating the relevant driving force of this reaction is not the equilibrium value near 0.87 V,⁷⁹  but that of the protonated Tyr.⁹⁴  This E_m is likely closer to 1.1 to 1.4 V, based on reported operating values for Yz of PSII⁹⁵  and the pH dependence of the formal redox midpoint potential in a helical bundle.⁷⁹  Thus photo-oxidation of Tyr near neutral pH is likely endergonic in this triad. Photochemical triad maquettes do photo-oxidize Tyr at pH 7.5, albeit with smaller yield. Charge-separation lifetimes are modulated by inserting nearby protonatable residues such as a His; crystal structures show this His hydrogen bonds to Tyr much as in natural PSII.

The oxygen evolving Mn₄CaO₄ cluster in natural PSII is ligated by a cluster of aspartates, glutamates and histidines. This cluster assembles by photo-oxidizing Mn(ii) in a series of light and dark steps.^96,97  Photo-assembly is required each time PSII is replaced after photo-damage, which can be as short as 30 minutes.⁹⁸  To better understand the engineering of Mn cluster photo-assembly, two groups have modified natural proteins to couple light-activatable centres with metal-ligating amino acids. Allen and co-workers modified natural purple bacterial reaction centres by inserting ligating carboxylates near the light-activated bacteriochlorophyll dimer pigment.⁹⁹  Additionally, the E_m of the bacteriochlorophyll dimer was raised from 0.51 V to 0.80 V by adjacent mutations, enough to photo-oxidize Mn(ii) with an E_m of 0.63 V.¹⁰⁰  Although this single Mn site cannot oxidize water, it does perform light-driven oxygen production from superoxide.¹⁰¹  In collaboration with T. Wydrzynski,¹⁰²  we re-engineered the natural iron storage protein bacterioferritin for light activated electron transfer by substituting the heme iron tetrapyrrole with Zn-chlorin-e6. Natural bacterioferritin employs a cluster of two His and 4 Glu to ligate Fe(ii) and a nearby Tyr. Exposing the metal binding site to Mn(ii) instead of Fe(ii) created an artificial reaction centre proto-type that photo-oxidized tyrosine only in the presence of di-nuclear Mn(ii).

When we installed a bacterioferritin-like carboxylate/histidine ligation motif in photochemical triad maquettes adjacent to the photo-oxidizable Tyr we created a photo-oxidizable metal binding site analogous to PSII. Crystal structures reveal that metals bind in the site as designed (Figure 1.6c) while light activation shows that metal oxidation is effective. As in PSII, Mn(ii) oxidation in maquettes is a relatively slow, adiabatic electron-transfer reaction compared to a simple electron tunnelling reaction. Also as in natural photosystems, a redox pool supports repeated light-activated turnover to accumulate oxidizing equivalents. We have observed interprotein electron transfer from photo-reduced heme in maquettes to added cytochrome c. Under continuous illumination in the absence of Mn(ii) little photo-reduction of cyt c is seen. In the presence of Mn(ii), cyt c acts as an oxidizing redox pool that permits quantitation of the turnover number of reaction centre maquettes. Under continuous illumination, reduction levels off after up to four oxidizing equivalents are accumulated, indicating oxidation of multiple Mn(ii) to Mn(iii) and possibly Mn(iv).

The maquette approach allows an iterative, modular approach to reproduction of bioenergetic function that clarifies what is and is not important in engineering redox protein function. Such an approach is a valuable complement to mutagenesis of natural proteins, which often shows intolerance to progressive changes of multiple amino acids due to marginally stable folding.¹⁰³  Natural proteins are generally only as stable as they need to be under the forces of natural selection.¹⁰⁴  The broad tolerance of maquettes to insertion of a wide range of natural and unnatural redox cofactors, combined with the ability to control cofactor placement, enables ready design of a wide range of electron-tunnelling cofactor chains for long-distance electron transfer and charge separation. The most critical factors are setting appropriate edge-to-edge distances between cofactors and driving forces guided by eqn (1.2). Such designs support hundreds of millisecond charge separated lifetimes that are longer than presently reported lifetimes in other natural or synthetic cofactor triads. The design challenge now for maquettes and artificial proteins in general, is to exploit long-lived charge separated states to engineer the bond making and breaking reactions to support efficient catalysis.¹⁰⁵  In this way we can hope not only to match, but to go beyond natural design, as predicted by Emil Fisher in his Nobel lecture in 1902.¹⁰⁶  The photochemical maquettes described here are aiming for designs capable of splitting water into hydrogen and oxygen using earth abundant and economical materials as a means of addressing human sustainable energy needs.

This research was carried out as part of the Photosynthetic Antenna Research Center (PARC), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award DESC0001035 supporting, CCM, NME, JAM and PLD in maquette construction and characterization.