High Throughput Engineering to Revitalize a Vestigial Electron Transfer Pathway in Bacterial Photosynthetic Reaction Centers*

Background: Bacterial reaction centers catalyze light-induced transmembrane electron transport using only one of two chemically equivalent pathways. Results: Multiplexed screening uncovers semirandom mutations that unexpectedly activate the unused pathway by employing ionizable residues. Conclusion: High throughput mutagenesis approaches reveal structure/function relationships that govern electron transfer efficiency. Significance: Directed molecular evolution can reveal principles that enable efficient, unidirectional, transmembrane electron transfer for the design of de novo pathways or biomimetic devices. Photosynthetic reaction centers convert light energy into chemical energy in a series of transmembrane electron transfer reactions, each with near 100% yield. The structures of reaction centers reveal two symmetry-related branches of cofactors (denoted A and B) that are functionally asymmetric; purple bacterial reaction centers use the A pathway exclusively. Previously, site-specific mutagenesis has yielded reaction centers capable of transmembrane charge separation solely via the B branch cofactors, but the best overall electron transfer yields are still low. In an attempt to better realize the architectural and energetic factors that underlie the directionality and yields of electron transfer, sites within the protein-cofactor complex were targeted in a directed molecular evolution strategy that implements streamlined mutagenesis and high throughput spectroscopic screening. The polycistronic approach enables efficient construction and expression of a large number of variants of a heteroligomeric complex that has two intimately regulated subunits with high sequence similarity, common features of many prokaryotic and eukaryotic transmembrane protein assemblies. The strategy has succeeded in the discovery of several mutant reaction centers with increased efficiency of the B pathway; they carry multiple substitutions that have not been explored or linked using traditional approaches. This work expands our understanding of the structure-function relationships that dictate the efficiency of biological energy-conversion reactions, concepts that will aid the design of bio-inspired assemblies capable of both efficient charge separation and charge stabilization.

Photosynthetic reaction centers convert light energy into chemical energy in a series of transmembrane electron transfer reactions, each with near 100% yield. The structures of reaction centers reveal two symmetry-related branches of cofactors (denoted A and B) that are functionally asymmetric; purple bacterial reaction centers use the A pathway exclusively. Previously, site-specific mutagenesis has yielded reaction centers capable of transmembrane charge separation solely via the B branch cofactors, but the best overall electron transfer yields are still low. In an attempt to better realize the architectural and energetic factors that underlie the directionality and yields of electron transfer, sites within the protein-cofactor complex were targeted in a directed molecular evolution strategy that implements streamlined mutagenesis and high throughput spectroscopic screening. The polycistronic approach enables efficient construction and expression of a large number of variants of a heteroligomeric complex that has two intimately regulated subunits with high sequence similarity, common features of many prokaryotic and eukaryotic transmembrane protein assemblies. The strategy has succeeded in the discovery of several mutant reaction centers with increased efficiency of the B pathway; they carry multiple substitutions that have not been explored or linked using traditional approaches. This work expands our understanding of the structure-function relationships that dictate the efficiency of biological energy-conversion reactions, concepts that will aid the design of bio-inspired assemblies capable of both efficient charge separation and charge stabilization.
Light energy is converted to chemical energy in a series of extremely efficient electron transfer (ET) 4 reactions performed by photosynthetic reaction centers (RCs). The structures of these transmembrane protein-cofactor complexes from photosynthetic bacteria and higher plants reveal an axis of approximate 2-fold symmetry that relates both core protein subunits and functionally asymmetric sets of cofactors ( Fig. 1) (1)(2)(3)(4). In many photosynthetic bacteria, these complexes are composed of three subunits: the homologous, integral L and M polypeptides and the membrane-tethered H polypeptide. Bacterial cofactors include a bacteriochlorophyll dimer (P) that serves as the primary electron donor, two bacteriochlorophyll monomers (B A and B B ), two bacteriopheophytins (H A and H B ), and two ubiquinone 10 molecules (Q A and Q B ). In type I RCs (e.g. Photosystem I), spectroscopic evidence suggests that both pathways are active in producing reduced quinone via lightinduced ET (see e.g. Refs. [5][6][7][8]. Yet, in type II RCs (e.g. bacterial RCs and Photosystem II), the B pathway is silent, and possibly vestigial; only the A branch cofactors are active.
The relative free energies of the charge-separated states are critical in determining the directionality, rates, and yields of ET in the RC. Some strategies to manipulate these factors genetically over the last 20 years (for review, see Ref. 9) include: addition or removal of hydrogen bonds to cofactors (10 -15); addi-tion or removal of axial ligands with attendant changes of bacteriopheophytin to bacteriochlorophyll or vice versa (16 -27); modification of conserved residues or helical segments that are related by the pseudo-2-fold symmetry axis of the complex (28,29); and introduction or removal of polar or ionizable residues near the cofactors (12, 13, 16, 17, 30 -32). To disable ET from H A to Q A or from Q A to Q B specifically, the occupancy and environments of the quinone binding pockets have been manipulated by mutation (16,(23)(24)(25)(33)(34)(35)(36)(37).
The highest yields of ET from the excited dimer (P*) to H B (thus producing P ϩ H B Ϫ ) have been achieved in mutant Rhodobacter (R.) capsulatus RCs (17,31,32,35,38). In mutant RCs lacking Q A , reduction of Q B via sole activity of the B branch cofactors has been observed unambiguously (16, 24, 35, 37, 39 -41). No cases yet reproduce any of the (essentially) 100% yields characteristic of the native A side ET reactions. Rather, mutant RCs have demonstrated the magnitude of the B pathway inefficiency. The combination of a few site-specific mutations has not managed to modulate the energetics (or electronic factors) of the B branch cofactors to provide for rapid electron transfer. At present, we lack the understanding needed to use rational design to accomplish the drastic leap required to redesign the RC to use the B pathway exclusively and efficiently.
Construction of such a re-engineered RC requires sampling of a much larger number of variants than has been examined previously. Thus, an approach has been adopted that combines semirandom mutagenesis and rapid screening techniques that require minimal sample volumes. Several methods have been streamlined by implementation of an automated liquid handler, and purification of mutant RCs has been semiautomated. A high throughput spectroscopic assay evaluates the competency of candidate RCs for Q B reduction via the B side cofactors. This approach has been made feasible by advances in synthetic biology; favorable mutations that have been identified can be fed into directed molecular evolution experiments for additional optimization of the ET pathway. Techniques that have been developed here are likely to be applicable to other multisubunit protein complexes with similar gene architectures.

EXPERIMENTAL PROCEDURES
Growth-R. capsulatus strains were cultured on Super RCVPY (20) medium under semiaerobic, chemoheterotrophic conditions in the dark (165 rpm, silicone sponge closures, 33°C). Escherichia coli strains were propagated on LB or 2ϫTY medium (42). Plasmids were selected with kanamycin (30 g/ml for both organisms). Strains and plasmids used in this study are listed in supplemental Table S1.
Molecular Biology-The expression vector used in this study, pBBRKW2HTsLsM ( Fig. 2; GenBank Accession Number JN565028), was constructed from puf operon genes derived as an EcoRI-SacI fragment from plasmid pUHTMluBgl:␣ Ϫ (24) that was inserted into a derivative of broad host range cloning vector pBBR1MCS-2 (43), generating pBBRKW2HTBALMX. The expression plasmid was then modified extensively to facilitate the approach described herein. Portions of both the pufL and pufM genes were synthesized (GenScript) to encode silent changes that introduced restriction sites flanking each of the segments chosen for mutagenesis (Fig. 2). To eliminate duplicate restriction sites elsewhere, 11 restriction sites in pBBR1MCS-2 and 4 sites in the puf operon were removed (supplemental Table S2) using either site-directed mutagenesis (QuikChange; Stratagene) or mung bean nuclease. The synthetic segment of pufL was then cloned into pBBRKW2HTBALMX (Table S2) as a PstI-KpnI fragment, and the synthetic pufM segment was cloned using BglII and BamHI, thus creating pBBRKW2HTsLsM. A His 7 tag is encoded at the C terminus of pufM. The H32R mutation in the pufA gene prevents assembly of the light-harvesting (LH) I antenna complex (44). Thus, genes expressed from this plasmid in R. capsulatus strain U43 result in the antennaeless, "RC-only" photosynthetic membrane needed for quantitative extraction of RCs by mild detergents (24). A summary of substitutions carried by mutant RCs and the nomenclature used to refer to them is contained in supplemental Table S3.
Construction of the F(L181)XϩFHV, F(L216)XϩYFHV and E(L212)A-D(L213)A-F(L216)XϩYFHV sets of mutant strains was expedited with cassette-based mutagenesis. The W(M250)V control strain was constructed by site-directed mutagenesis of the synthetic BglII-BamHI fragment of the pufM segment described above. Protocols for phosphorylation and annealing of oligonucleotides, ligations, transformations, and plasmid screening were adapted for use with a liquid handling system (Biomek FX; Beckman Coulter). For each mutant set, a library of oligonucleotides was designed to encode all 20 amino acids at residue X. Oligonucleotides for the F(L181)XϩFHV set were flanked by HindIII and BbvCI restriction sites, and the oligonucleotides for the F(L216)XϩYFHV and E(L212)A-D(L213)A-F(L216)XϩYFHV cassettes were flanked by AccIII and AclI restriction sites (Table 1 and Fig. 2). Oligonucleotides were phosphorylated, annealed, and ligated to pBBRKW2HTsLsM digested with the appropriate restriction enzymes. Candidate plasmids from DH5␣ transformants were screened using restriction fragment analysis and sequenced. Correct plasmids were used to transform mobilizing donor strain S17-1 for conjugal transfer to R. capsulatus deletion strain U43 (24). To facilitate rapid recovery of expression strains, a 100-l aliquot of a 1:25,000 dilution in RCV medium (45) was plated directly to Super RCVPY agar containing kanamycin. After 2 days, R. capsulatus colonies were cultured on Super RCVPY containing kanamycin for the preparation of archival stocks.
Expression Screening-To survey expression levels, crude RCs were purified in tandem according to standard methods (as detailed in supplemental Methods) with the following modifications. Cells from small cultures (80 ml) were lysed using a microfluidizer (Microfluidics, Inc.) and mutant RCs were solubilized with Deriphat 160-C. Following overnight binding to 200 l of nickel-nitrilotriacetic acid resin (Qiagen), chromatography proceeded in 2-ml, 96-well filter plates. UV-visible spectra of dark-adapted (10 min) samples were recorded (Spectra-Max M5 e ; Molecular Devices). Based on A 865 nm , mutant RCs were classified into three expression categories relative to yields of WT RCs: abundant (Ͼ25%), moderate (2.5-25%), and poor (Ͻ2.5%).
Semiautomated RC Purification-Mutant strains that expressed RCs abundantly or moderately were grown in 2-6-liter volumes depending on the expression level. RCs were purified via semiautomated methods using ÄKTA-FPLCs (46).
RC Functionality Screening-Assaying mutant RCs for formation of P ϩ Q B Ϫ via B branch ET was accomplished with a home-built, time-resolved spectrometer designed for 100 -200-l samples arrayed in clear-bottom 96-well plates (BD Falcon). Rapid data collection and reproducibility were facilitated via use of a computer-controlled translation stage (Aerotech) to position the sample well under collinear, vertical excitation and probing (measuring) beams. RCs were excited with a single subsaturating, 5-ns, 532-nm flash from a Nd:YAG laser. The magnitude and decay of bleaching of the ground state absorbance of P were probed by a stable, continuous-wave, 850-nm laser diode (Melles-Griot) attenuated with neutral density filters so that no discernible actinic effects were observed. For added sample protection, the probe light was incident on the sample only during data collection and was otherwise blocked with a computer-controlled shutter. Following passage through the sample, the probe light was isolated from the collinearized pump-probe beams by dichroic and 850-nm interference filters. The isolated probe light was collected onto a near-infrared-sensitive photosensor module (Hamamatsu H-5784-20), whose output was recorded on a digital oscilloscope. The apparatus is capable of measuring transient spectroscopic signals  Table 1) are flanked by unique restriction sites that facilitate the insertions of oligonucleotide cassettes encoding multiple substitutions. Fifteen restriction sites (parentheses) in the plasmid were removed to enable the cassette-based mutagenesis approach.
XmnI BlpI spanning 50 s to several minutes. For 100-l volume, the effective measuring pathlength was 2 mm; sample concentrations were adjusted such that A 865 nm was 0.45-0.55. Signal intensities were normalized to the same ground state absorbance of the P band during data analysis. Because RCs are Q Aless (except for wild type) due to the W(M250)V mutation described below, the initial amplitude of each kinetic trace is proportional to the yield of ET to Q B solely via the B branch, as assayed relative to control RCs on the same plate. Further characterization of primary electron transfer events in select mutant RCs utilized a transient absorption apparatus having 130-fs excitation and white light probe flashes operating at 10 Hz (described previously in Refs. 31,38). RCs were held at ϳ10°C in an ice-cooled reservoir and flowed through a 2-mm pathlength cell.

RESULTS
A system to construct and express mutants of the L and M genes of the polycistronic, oxygen-regulated puf operon of R. capsulatus was developed previously (47)(48)(49). The expression host strain, U43, does not synthesize any light-harvesting antennae or RC complexes of the photosynthetic apparatus (i.e. LHI Ϫ LHII Ϫ RC Ϫ ). The original broad host range expression plasmid cannot be used for this type of study because its complete sequence has not been determined and its large size (32 kb) hinders its application to an approach based upon restriction enzyme cloning of mutated cassettes. Thus, pBBR1MCS-2 (5.1 kb), whose sequence is known (43), was selected as the vector for the expression plasmid and modified as described above. Yields of RCs purified from strains carrying the newly engineered expression plasmid were found to be 2-fold greater than those obtained from plasmid pUHTMluBgl:␣ Ϫ .
The initial mutagenesis template was FHV. The F mutation substitutes Phe for the native Tyr at M208 near B A and diminishes ET to the A side. The H mutation substitutes His for Leu at M212 near H A and results in incorporation of a bacteriochlorophyll in the H A site. Because this cofactor is more difficult to reduce, the A pathway is disabled further. This mutation also removes H A absorption from the 510 -550 nm spectral region, allowing quantitative assessment of H B reduction via ultrafast (picosecond) spectroscopic measurements. The V mutation, which substitutes Val for the conserved Trp at M250, prevents Q A binding and, thus, prevents ET to Q B via Q A . In other words, this mutation ensures that electrons can reach Q B only via the B side cofactors.
M208Tyr is related via the C 2 symmetry axis to residue L181Phe near B B . This conserved asymmetric pair was the target of some of the earliest mutagenesis studies of the purple bacterial RC. The swap of Tyr for the native Phe at L181 and Phe for Tyr at M208 reverses the asymmetry and enhances ET to H B while diminishing ET to the A side. When the F(L181)Y mutation is added to the FHV template, the result is the YFHV RC, which serves as another control (in addition to the wild type that results in quantitative reduction of Q B via A branch activity) for the ms screening assay because it displays ϳ30% yield of P ϩ H B Ϫ and ϳ15% yield of P ϩ Q B Ϫ (31,35,39,40,50). A large number of residues within 5-14 amino acid segments near the cofactors have been identified as mutagenesis targets (Table 1 and Fig. 2). At the outset, regions proximal to B B , H B , and Q B were targeted to further enable (compared with YFHV) the B side pathway. For all of the mutant strains reported here, protein yields from larger scale cultures and automated purification procedures correlated well with yields determined from the small scale expression screening. The use of Deriphat 160C for RC solubilization and purification ensured full occupancy of Q B in the native binding site (24).
Twenty F(L181)XϩFHV constructs comprised the initial mutant set. Residue L181 (segment D, near B B ) was chosen because of the known influence of this site on initial B side charge separation as indicated above. Eighteen of the mutant RCs were expressed moderately or abundantly (Fig. 3). None of the substitutions at L181 resulted in pronounced effects on the RC ground state absorption spectra except for the KFHV mutant, with K at L181 previously having been shown to provide a sixth ligand to B B (11).
Additional rounds of mutagenesis focused on residues near Q B in attempts to increase the low (ϳ40%) yield of ET from H B to Q B (24,35,39). Saturation mutagenesis was performed at conserved residue L216Phe (segment J) because its C 2 -symmetry-related counterpart, M250Trp, is essential for rapid ET from H A to Q A (33) and is important for Q A binding. The F(L216)X substitutions were coupled with YFHV. Additionally, because previous work showed that replacement of the native Glu and Asp at L212 and L213, respectively, with alanines stabilizes P ϩ Q B Ϫ (51), the F(L216)X variants were coupled with the E(L212)A-D(L213)AϩYFHV (AAϩYFHV) substitutions in another set of mutants. These two rounds of mutagenesis yielded 40 additional strains with 36 expressing RCs at abundant or moderate levels, 19 F(L216)XϩYFHV mutants and 17  Table S4). Only the D, H, and K substitutions failed to produce RCs in the AAϩF(L216)XϩYFHV background.

AAϩF(L216)XϩYFHV mutants (supplemental
All 54 of the mutant RCs studied here gave light-induced formation of P ϩ Q B Ϫ via the B side cofactors in the ms transient absorption assay. The relative amounts of P ϩ Q B Ϫ formed are plotted in Fig. 4. Of particular note are the F(L181)HϩFHV (HFHV), F(L181)NϩFHV (NFHV), and F(L181)DϩFHV (DFHV) mutant RCs, which produce P ϩ Q B Ϫ signals that are 20 -30% larger than that of the YFHV control. The A, G, K, and W mutations produce RCs with the poorest yields of P ϩ Q B Ϫ in the F(L181)XϩFHV series. Fig. 5A shows kinetic traces derived from the HFHV, NFHV, DFHV, and WFHV RCs along with the YFHV and W(M250)V-only control RCs from the ms screening assay. The profiles, displaying decay time constants Ն10 s, are characteristic of P ϩ Q B Ϫ charge recombination when this state is formed via activity of the B branch cofactors; in these RCs, decay of this state via routes involving Q A is prohibited by the W(M250)V mutation (24). Thus, the initial amplitudes of the kinetic traces in Fig. 5A are measures of the (relative) yields of P ϩ Q B Ϫ formed solely via B branch activity; this value is plotted for all of the mutant RCs in Fig. 4.
The results for the HFHV, NFHV, DFHV, and WFHV mutant RCs were confirmed by ultrafast transient absorption measurements (Fig. 5B). These spectra show the magnitudes of bleaching of H B at ϳ528 nm, reflecting the relative yields of ET from P* to H B and confirm an ordering of HϾNϾDϾYϾ ϾW, the same ordering as determined in the P ϩ Q B Ϫ yield assay. Additional ultrafast measurements are under way and will be reported elsewhere.
Expression yields for the F(L216)XϩYFHV and AAϩF(L216)XϩYFHV sets of mutant RCs were quite high, especially for the latter set given the number of mutations introduced into the RC. However, yields of P ϩ Q B Ϫ in these mutant RCs did not exceed that of the template YFHV. In general, there is less ET to Q B in RCs containing the E(L212)A-D(L213)A substitutions than in mutant RCs carrying the native residues at these sites. The relative rank order of the residues in each set is different (see, e.g. the effect of a Val substitution in each set), suggesting complex interplay of effects introduced by substitution of the small, nonpolar alanines for the larger, ionizable Glu and Asp.

DISCUSSION
Directionality of ET in type II RCs is a conundrum that has existed ever since the 1985 structure of the bacterial RC (3) revealed two branches of cofactors related by an axis of approximate C 2 symmetry whereas spectroscopic measurements demonstrated activity of only one of them. The free energies of the charge-separated states on the A pathway (P ϩ B A Ϫ , P ϩ H A Ϫ , and P ϩ Q A Ϫ ) are successively downhill from each other and from P*, facilitating rapid forward ET. In contrast, B branch inactivity presumably stems in large measure from P ϩ B B Ϫ being somewhat higher in free energy than P*, resulting in much slower, noncompetitive, B side charge separation. Differences between the electronic couplings of P* with B A and B B may also contribute to unidirectional A side ET. The rate constant for P* 3 P ϩ H B Ϫ on the native B pathway has been measured to be ϳ(100 ps) Ϫ1 (17,31), approximately 30-fold slower than charge separation on the A side (3-4 ps for P* 3 P ϩ H A Ϫ ), suggesting that yields of B side ET in wild-type RCs should be ϳ3%.
Although significant yields of P ϩ H B Ϫ formation have been reported in some R. capsulatus mutants (17,31,32,35,38), the largest rate constant for P* 3 P ϩ H B Ϫ is ϳ(40 ps) Ϫ1 , obtained in the YFH mutant RC (31). The best efforts by several laboratories over many years have thus achieved approximately only a 2-fold increase in the rate of ET to H B compared with the wildtype value. This enhanced rate is still a factor of 10 slower than ET to the A side. Interestingly, rapid initial charge separation to the B side has been reported in a mutant where a bacteriopheo- phytin substitutes for B B . Whereas ET from P* to this new cofactor takes place in ϳ10 ps, ET onward to H B does not occur (18), creating another, different, bottleneck in the quest for efficient use of the B side cofactors for transmembrane charge separation. Similar issues occur at the final step of ET from H B to Q B , which has a yield of only ϳ40% as mentioned above. The rate of this ET reaction is ϳ(4 ns) Ϫ1 , whereas the rate of competing charge recombination of P ϩ H B Ϫ is ϳ(3 ns) Ϫ1 (39). For comparison, the rates of the corresponding reactions on the A side are (200 ps) Ϫ1 and ϳ(20 ns) Ϫ1 .
To sample a larger number of variants and remove much of the prejudice inherent in rational design schemes, a directed molecular evolution approach has been adopted for engineering of the RC to enable efficient B branch ET. Directed molecular evolution has been applied to the discovery of new functions, increased expression, increased stability, and enhanced activity for many proteins (for review, see Refs. 52, 53) but has had relatively limited application to membrane proteins (e.g. Refs. 54 -57). The present study merges principles of both rational design and directed molecular evolution to generate versions of the heterotrimeric, protein-cofactor RC complex that have acquired increased ability to utilize a vestigial ET pathway for transmembrane charge separation.
The mutagenesis employed here is best termed "semirandom" and relies on oligonucleotide cassettes rather than errorprone PCR or fragment shuffling. The rationale for this approach is based on knowledge of the structure, awareness of the substitutions that have been tried before, the concern that massively random mutagenesis will produce RCs with assembly and/or stability problems, the difficulties imposed by the high degree of sequence similarity between the pufL and pufM genes (especially in highly conserved regions that surround the cofactors), and the fact that their coding regions overlap (Fig. 2). Because this approach has worked well here for these genes embedded in the polycistronic puf operon, it is likely that the chosen mutagenesis strategy could be applied to many gene clusters encoding transmembrane complexes where components share a high degree of sequence similarity, a common occurrence in sequence data emerging from representatives of all kingdoms.
Directional cloning of cassettes required significant downsizing and engineering of the broad host range expression vector and was enabled by synthetic biology. Surprisingly, the yield of RCs expressed from the new, partially synthetic platform vector was 2-fold greater than the yield obtained from the original expression plasmid. This result may be due to differences in plasmid copy number and/or unexpected changes in (i) transcript stability, (ii) codon preferences, or (iii) modification of regulatory elements that occurred as a result of the engineering that was required.
In the initial round of saturation mutagenesis at the L181 site, surprises were encountered when yields of B side ET were found to be highest in RCs carrying His, Asn, Asp, and Gln mutations. Numerous RCs carrying single mutations at this site have been reported previously (9,11,58), but none of these studies had ever suggested that any of the above residues would match or exceed a Tyr at L181 in enhancing B branch activity or that the functional consequences of a Glu substitution would be similar to those conferred by a Tyr substitution (Fig. 4). In fact, considering that placement of an acidic residue close to B A in the G(M201)D RC disfavors ET to the A side (12), it was surprising that an Asp substitution at L181 near B B had the opposite effect, enabling more B branch activity. Similar effects observed for the polarizable His, Asn, Asp, and Gln residues at L181 cannot all be attributable to a common fundamental mechanism, but all of these substitutions will influence, rather unpredictably, the dielectric environment of the cofactors involved in the initial electron transfer reactions. The close parallel effects on B side ET for the Asn/Asp and Gln/Glu pairs, with the smaller Asn/Asp substitutions producing RCs with larger yields of B side ET, are noteworthy. In general, an assortment of uncharged substitutions at the L181 site is tolerated with neutral functional consequences; an additional surprise is that substitutions of both small (Gly, Ala) and large (Trp) residues produce nearly equivalent yields of P ϩ Q B Ϫ . Although substitutions at the L216 site (regardless of whether they are accompanied by the Ala substitutions at L212-213) do not increase the amount of reduced quinone, trends are observed, some of which can be related to amino acid properties. In general, large residues at the L216 site result in lower yields of reduced quinone. RCs in which ionizable residues are substituted at the L216 site, when in combination with the Ala substitutions at L212-213, either fail to assemble or perform poorly in the functional assay. It is possible that substitutions within this region of the Q B binding site affect its occupancy. A disappointment is that the Trp substitution at this site does not result in increased ET from H B to Q B whereas the conserved Trp residue at the symmetry-related M250 site is critical for efficient ET from H A to Q A in the wild-type RC (33). These results suggest that targeting of residues based on their relationship to the C 2 symmetry axis in this region of the RC is less productive and emphasize that a more random, evolutionary approach will be required to enhance secondary electron transfer reactions of the B pathway.
Overall, the results reported herein validate the experimental strategy and underscore the plasticity of the RC in accommodating substitutions that result in a range of B branch activity. A subset of the new mutant RCs displays increased quinone reduction by the exclusive use of B branch cofactors. Unexpected, productive combinations of amino acid substitutions that previously have not been discovered by rational design were revealed by the approach described herein. Realization of these mutant RCs was enabled by miniaturization at two screening steps (expression testing and spectroscopic assays) and automation of protein purification and time-resolved spectroscopy. A system is now in place whereby mutant RCs with altered ET functionalities can be identified rapidly as candidates for more exhaustive functional characterization and iterative engineering to produce next generation RCs with desired ET capabilities, even those that are unimagined by Nature.