Augmenting light coverage for photosynthesis through YFP-enhanced charge separation at the Rhodobacter sphaeroides reaction centre

Photosynthesis uses a limited range of the solar spectrum, so enhancing spectral coverage could improve the efficiency of light capture. Here, we show that a hybrid reaction centre (RC)/yellow fluorescent protein (YFP) complex accelerates photosynthetic growth in the bacterium Rhodobacter sphaeroides. The structure of the RC/YFP-light-harvesting 1 (LH1) complex shows the position of YFP attachment to the RC-H subunit, on the cytoplasmic side of the RC complex. Fluorescence lifetime microscopy of whole cells and ultrafast transient absorption spectroscopy of purified RC/YFP complexes show that the YFP–RC intermolecular distance and spectral overlap between the emission of YFP and the visible-region (QX) absorption bands of the RC allow energy transfer via a Förster mechanism, with an efficiency of 40±10%. This proof-of-principle study demonstrates the feasibility of increasing spectral coverage for harvesting light using non-native genetically-encoded light-absorbers, thereby augmenting energy transfer and trapping in photosynthesis.

P hotosynthesizing organisms harvest only a small proportion of the visible and near-infrared radiation reaching the Earth's surface, and bacterial, algal and plant species have evolved to occupy particular spectral niches depending on the absorption of the particular chlorophyll, carotenoid, phycobilin and other pigments they synthesize 1,2 . Several studies have explored the possibilities of expanding the spectral coverage of light harvesting and reaction centre (RC) complexes using chlorophylls [3][4][5] or attached dye molecules [6][7][8] and the biohybrid approach, which combines parts of native antennas with synthetic chromophores, provides an emerging platform technology for the creation of artificial antennas. This flexible, versatile and tailorable strategy extends solar coverage beyond natural systems. The polypeptide scaffold often chosen for such work is based on the light-harvesting LH1 complex from the photosynthetic bacterium Rhodobacter (Rba.) sphaeroides. The LH1 antenna consists of a circular assembly of repeating a 1 b 1 BChl 2 Crt 2 units, where each a or b polypeptide is a single transmembrane helix that binds a single bacteriochlorophyll a (BChl a) and a carotenoid (Crt) molecule. This assembly surrounds, and donates excitation energy to, the RC where charge separation occurs. The discovery that an LH1 antenna can be reconstituted from purified pigments and a/b polypeptides, plus the availability of single-site alterations and truncated versions of these polypeptides, allowed the in vitro self-assembly of LH1 antenna variants [9][10][11] . Biohybrid antennas with enhanced spectral coverage can be assembled by creating attachment sites for synthetic chromophores that efficiently transfer energy to the BChl a site 5,[12][13][14][15] . The efficiencies of energy transfer to the BChl a acceptor are comparable with native antennas, and current designs can accommodate up to 60 chromophores that contribute to solar light harvesting via an energy-transfer cascade. The use of unattached accessory chromophores within detergent micelles provides another route to enhance light-harvesting capability 14,15 .
Biohybrid architectures have great potential for producing artificial light-harvesting architectures in vitro, but they utilize synthetic chemistry so they cannot replicate in vivo. To create tailor-made light harvesting antennas in vivo, we must make use of the toolbox of proteins and pigments available in nature, or create synthetic elements that can be produced by the host organism. In the purple photosynthetic bacterium Rba. sphaeroides, light is harvested by the LH2 and LH1 antenna complexes and the resulting excitation energy is used to power the reduction of quinone to quinol before the formation of a proton gradient that powers ATP synthesis 16 . Rba. sphaeroides transfers the energy from light absorbed by B875 BChls in LH1 and B850 BChls in LH2 to the RC with near 100% quantum efficiency [17][18][19] . A charge separation takes place within the RC, which converts excitation energy to biochemical energy 20 . Quinols, the eventual product of RC photochemistry, exit from the RC and traverse the LH1 ring surrounding the RC through a portal provided by the PufX polypeptide 16 .
As a pilot study to explore the possibility of creating artificial light-harvesting antennas in vivo, this work investigates the effects of incorporating the yellow fluorescent protein, YFP, as a chromophore into the Rba. sphaeroides photosystem. YFP is a well-studied and genetically programmable protein; the YFP variant used in this study, SYFP2, exhibits photostability, a highfluorescence quantum yield (B70%) and a large extinction coefficient (B100,000 M À 1 cm À 1 at 515 nm) 21 . However, YFP cannot contribute usefully to the native absorption of Rba. sphaeroides, because the carotenoid pigments of the light-harvesting complexes absorb in the 400-550 nm range. Accordingly, we constructed a carotenoid-less mutant of Rba. sphaeroides to create a baseline strain for assessing any contribution of YFP to photosynthesis. The 525 nm emission maximum of SYFP2 has some overlap with the visible-region (Q X ) absorption band of BChl a in the Rba. sphaeroides LH1 and RC complexes, and the results of this study suggest that some of the energy absorbed by YFP migrates to the RC. This work forms the basis of future studies involving the creation of novel, tailored light-harvesting antenna complexes for increased spectral coverage of in vivo photosynthesis.

Results
Spectroscopy and photosynthetic growth of RC/YFP strains. A baseline strain was constructed for the expression of YFP by deleting crtB (RSP_0270, encoding phytoene synthase) from Rba. sphaeroides 22 . This manipulation abolishes the biosynthesis of carotenoids, the 450-550 nm absorption of which would otherwise eclipse YFP absorption, as shown in the absorption spectra for WT and WT RC/YFP membranes ( Supplementary  Fig. 1). The DcrtB mutation also eradicates LH2 assembly 23,24 but does not affect the synthesis of the PufX polypeptide, which forms a portal for quinone traffic 16,25 . Although PufX is present, the carotenoid-less RC-LH1-PufX complex with YFP attached will be referred to as DcrtB RC/YFP-LH1. The gene encoding the YFP variant, SYFP2 (ref. 21), was fused to the 3 0 end of puhA, which encodes the reaction centre H subunit (RC-H). The placement of YFP on the periplasmic N terminus of RC-H was ruled out as, although it would bring the YFP closer to the RC special pair and the LH1 BChls, it would have likely blocked the cytochrome c 2docking site. The result of constructing puhA-SYFP2 is a fusion protein with YFP on the C-terminus of RC-H, located on the cytoplasmic side of the complex.
Immunoblotting with antibodies specific for the RC-H subunit or YFP showed the presence of a signal at 54.9 kDa corresponding to the expected size of a RC-H/YFP fusion (Fig. 1a). Room temperature absorption spectra of intracytoplasmic membranes (ICMs) prepared from photosynthetically grown DcrtB RC/ YFP-LH1 show the YFP peak at 517 nm, which is shifted to 519 nm at 77 K (Fig. 1b,c). Fluorescence excitation spectra of ICM prepared from DcrtB RC/YFP-LH1 recorded at 77 K exhibit a feature at 519 nm corresponding to energy-transfer from YFP (Fig. 1d).
Photosynthetic growth from absorption of light by YFP. The 400-550 nm carotenoid absorption of light-harvesting LH1 and LH2 complexes overwhelms the absorption of RC-attached YFP ( Supplementary Fig. 1), so WT and WT RC/YFP-LH1 strains grow at the same rate under illumination by white light (Supplementary Fig. 2). Nevertheless, the 525 nm emission maximum of SYFP2 has some overlap with the visible-region (Q X ) absorption band of BChl a in the Rba. sphaeroides LH1 and RC complexes, so growth tests were conducted to find out if energy absorbed by YFP could migrate to RC-LH1 complexes in vivo. The carotenoid-less DcrtB mutant of Rba. sphaeroides provides the ideal baseline strain for growth rate experiments; the DcrtB and DcrtB RC/YFP-LH1 strains were grown in white light (100 mmol photons s À 1 m À 2 ), under which conditions both strains are able to grow autotrophically. The LH2-minus positive control (carotenoids þ RC-LH1-PufX), chosen because the other two strains also lack LH2 due to the DcrtB mutation, shows strong photosynthetic growth (Fig. 2a, green). Allowing for the B30 h lag delaying the onset of photosynthetic growth, the subsequent difference between the green and blue growth curves shows that under white light illumination the DcrtB culture is light-limited by B50 %. This effect arises because the absence of carotenoids impairs the light-absorbing capacity of the LH1 antenna, thereby limiting turnover at the RCs. After B30 h the growth of DcrtB RC/YFP-LH1 cultures began to outpace the control DcrtB strain, providing evidence for in vivo energy transfer from YFP to the RC-LH1-PufX complex (Fig. 2a). In an additional growth test, an LED light source (520 nm emission maximum, 35 nm full-width-at-half-maximum (FWHM)) was used to direct excitation energy specifically to YFP. The difference between the positive control (carotenoids þ RC-LH1-PufX; green line) and DcrtB strains (no carotenoids, red/blue lines) in Fig. 2c shows that both DcrtB and DcrtB RC/YFP-LH1 are growing at sub-saturating light intensities. Illumination of carotenoid-less cultures by green LEDs barely allows native BChl a pigments to contribute towards light harvesting. Comparison of the green and blue growth curves in Fig. 2c suggests that RC turnover in the DcrtB strain runs at only B10% of its potential capacity so the addition of a 520 nm-absorbing chromophore to the DcrtB strain might be expected to be beneficial, as long as the absorbed energy can transfer to photosystem components. The emission of the green LEDs (depicted by the white region in Fig. 2b) is well matched to the absorption of YFP, and the DcrtB RC/YFP-LH1 strain (Fig. 2c, red) grows faster than DcrtB under these experimental conditions. In contrast, the DcrtB negative control absorbed enough energy to sustain only very weak photosynthetic growth (Fig. 2c, blue).
The inset in Fig. 2c shows that the growth conditions supplied sufficient oxygen for maturation of the YFP chromophore in the DcrtB RC/YFP-LH1 strain. Thus, the data presented in Fig. 2 provide further evidence for light absorption by YFP and transfer of excitation energy to the RC to drive photosynthesis and cell division. However, the absence of carotenoids in the DcrtB strain used for this work could have created a need for photoprotective pigments, and it has been suggested that fluorescent proteins can play a photoprotective role 26,27 . To confirm that the superior growth rate of the DcrtB RC/YFP-LH1 strain (versus the DcrtB negative control) was due to YFP energy transfer and not photoprotection (which could favour an increased growth rate), the expression plasmid pBBRBB-YFP, based on pBBRBB-Ppuf 843-1200 28 , was used to make cytoplasmic YFP in Rba. sphaeroides DcrtB, which is free of any attachment to membrane proteins. Photosynthetic growth curves for high light cultures show that the presence of free YFP in the cytoplasm confers no benefit on DcrtB pBBRBB-YFP ( Supplementary Fig. 3), which grows at the same slow rate as DcrtB. The effectiveness of this plasmid expression system for YFP synthesis in Rba. sphaeroides was verified by the subsequent purification of YFP from this strain, which was used as a control for the static and time-resolved fluorescence measurements.
Structural model of the RC/YFP-LH1 complex. To visualize the attachment of YFP to the RC-H subunit, the RC/YFP-LH1 complex was purified from ICM prepared from photosynthetically grown cells, as described in the Methods. Following sucrose density gradient centrifugation, ion exchange chromatography and gel filtration the final complex was reasonably pure, as judged by the A875:A280 nm absorbance ratio of 1.92 and the presence of only five main bands (the PufX polypeptide stains poorly 29 ) in a Coomassie-stained SDS-PAGE gel ( Supplementary  Fig. 4).
Purified RC/YFP-LH1 complexes were adsorbed onto the pre-coated carbon film on a copper grid and negatively stained;  an additional density, indicated by white arrows in the inset to Fig. 3a. Then 11,777 particles were selected for further analysis; Fig. 3b shows examples of averaged reference-free 2D classes and Fig. 3c,d shows the reconstructed model of the RC/YFP-LH1 complex viewed from the cytoplasmic side of the membrane and in the plane of the membrane, respectively. This model is based on half of the structure of the dimeric RC-LH1-PufX core complex 29,30 ; the extra density apparent in some of the single particles and image classes and seen in the final reconstruction arises from YFP, which clearly protrudes from the side of the complex.
YFP fluorescence lifetimes in Rba. sphaeroides cells. A homebuilt microscope 31 was used to image fluorescence from YFP, both in the free state in DcrtB pBBRBB-YFP cells, and attached to RC-H in the DcrtB RC/YFP-LH1 strain; the DcrtB strain was included as a negative control. The images in Fig. 4a,c show the YFP emission from YFP in the RC-attached and free states, respectively. Figure 4d shows fluorescence emission spectra, each recorded on a single cell; the shapes of the spectra match those for purified YFP and RC/YFP-LH1 samples in Fig. 5b. No conclusions can be drawn from the amplitudes, which reflect the relative abundance of YFP in the free and attached states in these particular cells, as well as the fluorescence yield arising from coupling to acceptors in the RC/YFP-LH1 complex. As expected, no signal was observed in cells of the control DcrtB mutant, (Fig. 4b,d, black trace). Fluorescence lifetimes of YFP in the whole cell samples were recorded to detect any quenching that would indicate Förster resonance energy transfer (FRET) from the YFP to an acceptor molecule ( Fig. 4e; Table 1). The YFP fluorescence signal from DcrtB pBBRBB-YFP, which is an average taken from eight cells, has a bi-exponential decay (Fig. 4e, black points and fitted curve) which gives an amplitude-weighted lifetime of 2.3 ns. When YFP is attached to RCs in DcrtB RC/YFP-LH1 cells the lifetimes of both the short-and long-lived components are significantly shorter (Fig. 4e, red points), with an average lifetime of 1.2 ns. These amplitude-weighted lifetimes (Table 1) and Equation 1 were used to calculate a FRET efficiency (F EET ) for energy transfer between YFP and the RC-LH1 complex of B0.48. Figure 3f shows the distances between the YFP chromophore and the RC special pair BChls, and between YFP and the nearest LH1 B875 BChl, based on the structural model of the RC/YFP-LH1 complex in Fig. 3c,d. These distances, of B85 Å and B53 Å, respectively, are both within the range for FRET, although they suggest that the most likely route for energy migration to the RC is via the LH1 BChls.
To examine this point in more detail we used Förster theory to estimate the distance between the YFP chromophore and acceptor pigments in the RC-LH1 complex, on the basis of the 0.48 value for F EET determined from the whole cell fluorescence lifetime measurements. The calculations employed Photo-chemCad 32 using three values of the refractive index (1.335, 1.45 and 1.55) to accommodate latitude in the polarity of the cofactor/protein environments ( Supplementary Fig. 5). The calculations return a YFP to acceptor distance of 43-48 Å, in good agreement with the structural model in Fig. 3f which gives a 53 Å distance between YFP and the nearest LH1 BChl.

RC/YFP fluorescence quantum yield and excited-state lifetime.
Having shown that YFP is a donor of excitation energy to RC-LH1 complexes in vivo, we set out to construct the simplest system for examining the characteristics of a coupled YFP/ energy-trapping complex. Accordingly, RC/YFP complexes with no surrounding LH1 antenna were purified from carotenoid-less ICMs. Supplementary Figure 6 shows the outcome of the   purification, which yielded complexes free of LH1; Fig. 5a shows the clearly visible YFP absorption at 514 nm (black arrow), which is not present in the RC control. YFP was purified from Rba. sphaeroides DcrtB pBBRBB-YFP ( Supplementary Fig. 7), to act as a control for the fluorescence decay measurements. The yield of excitation energy transfer (F EET ) from excited YFP (YFP*) to the RC is determined from changes in the photophysical properties of YFP À fluorescence intensity (I f ), fluorescence quantum yield (F f ), or singlet excited-state lifetime (t s ) À in the presence versus absence of the RC. The YFP*-RC energy-transfer process provides an additional decay pathway for YFP* when the RC is present, resulting in a reduction in the measured parameter w (I f , F f , t s ) so that F EET ¼ 1 À w RC/YFP /w YFP . YFP*-RC energy flow was initially measured using static and time-resolved fluorescence spectroscopy. A significant reduction in integrated YFP fluorescence intensity for RC/YFP (Fig. 5b,  As in Supplementary Fig. 5, the measured F EET ¼ 0.4 was used to estimate the effective distance between the YFP chromophore and the RC cofactors using Förster theory and the photophysical parameters for the energy donor (YFP) and acceptor (RC) given in the 'Methods' section ( Supplementary Fig. 8). The calculations return a spectral-overlap integral of J ¼ 2.96 Â 10 À 13 cm 6 , a Förster radius of R 0 ¼ 53-58 Å, and a YFP to RC distance of 56-62 Å. The discrepancy between this and the 85 Å distance in Fig. 3d are discussed later.
Ultrafast transient absorption spectroscopy of RC/YFP. Ultrafast transient absorption (TA) spectroscopy was used to measure the t s of YFP, and to demonstrate YFP*-RC energy transfer from the 'growing in' of key RC features (absorption bleaching or stimulated emission) associated with formation of the excited primary donor (P*), which initiates the charge-separation sequence in the RC. Figure 6a shows representative TA spectra in  The values of amplitudes and lifetimes presented are the mean and s.e., n ¼ 8. The fluorescence decay curves shown in Fig. 4e were analysed using OriginPro and TRI2 software packages. A1 and A2 are the amplitude contributions of the long-and short-lived components, t1 and t2 are the lifetimes of each component, and tav is the amplitude-weighted lifetime.
the near-infrared region for the isolated RC (RC-only) sample, with 515 nm excitation directed at the YFP chromophore. The spectral evolution at early times (dashed lines) arises primarily from decay of P*-stimulated emission centred at B910 nm, with little change in the bleaching of the ground state absorption band of P centred at B865 nm. After P* decay (solid lines) by charge separation, the P bleach persists because P is oxidized (P þ ) and remains so and in the same amount during the entire chargeseparation sequence, namely P*-P þ H A À -P þ Q A À -P þ Q B À . This process proceeds with essentially unity quantum yield; charge recombination P þ Q B À -P occurs on the time scale of B1-10 s (Fig. 6c). Kinetic profiles at 850 and 910 nm were fit to two exponentials plus a constant, giving average time constants of 3.9 and 306 ps, nicely matching typical wild-type RC P* (B4 ps) and P þ H A À (B250 ps) lifetimes 33 . As expected, the change in DAbs at 850 nm with time is small (nonzero because H A À in P þ H A À and the other states differ in the size of a small absorption opposing the P bleach) and remains virtually constant from B1 ns to past B60 ns, the longest time shown (Fig. 7, blue triangles).
TA spectra for the RC/YFP sample at early times (Fig. 6b, dashed lines) give spectral changes with time constants (5.3 and 331 ps) that closely resemble those for RC-only as a result of unavoidable direct RC excitation in a small fraction of the sample. However, at later times (4100 ps), the P bleaching at B865 nm triples in magnitude and P*-stimulated emission at B910 nm reappears (Fig. 6b, solid lines). These signals arise unambiguously from YFP*-RC energy transfer on the nanosecond timescale, which is in marked contrast to the data for RC-only after a few hundred picoseconds (Fig. 6a). For RC/YFP, kinetic profiles at 850 nm (Fig. 7, red circles) and 910 nm were fit to three exponentials plus a constant to account for the RC P* and P þ H A À lifetimes plus a (rise) component reflecting the YFP* lifetime, which includes YFP*-RC energy transfer. The t s for YFP* in the RC/YFP complex was determined to be 1.8 ns at 850 nm (Fig. 7, red circles)

Discussion
Photosynthetic organisms-plants, algae, cyanobacteria, purple bacteria-are demonstrably successful in occupying their particular spectral niches, but there is scope nevertheless for enhancing the spectral coverage of photosynthesis by augmenting the chlorophyll, carotenoid and phycobilin pigments normally employed for absorbing light. Future synthetic biology applications such as light-powered bioreactors could require the design and construction of organisms that can absorb light over a range of wavelengths not found for naturally-evolved antennas. Antenna modification in vitro can be achieved by attaching highly-absorbing dye molecules to purified complexes, thereby filling gaps in the native absorption spectrum. These biohybrid complexes exhibit efficient energy transfer from attached chromophores to native pigments, and by attaching several chromophores it has been possible to create energy transfer relays 5,[12][13][14][15] . Attaching up to nine Alexa Fluor 647 molecules to the LH2 complex from Rhodopseudomonas acidophila strain 10050 enhanced its light harvesting functions by occupying a vacant spectral region from 600 to 700 nm 8 . Dutta et al. 6 have shown that DNA-templated dye molecules can be attached to Rba. sphaeroides RCs in vitro; the spectral range of these dyes complemented the native absorption of the RCs, and dye-RC energy transfer was observed, with efficiencies ranging from 42 to 83%. In another study, Alexa Fluor dyes were attached directly to the periplasmic face of the RC and dye-RC energy-transfer efficiencies of 52 to 70% were obtained 7 .
The present study was performed to determine whether light harvesting and energy transfer could be augmented in vivo by genetically modifying a photosystem complex so that it carries an extra chromophore. Photosystems normally contain hundreds of pigments that feed absorbed energy into RC traps, and the attachment of YFP to the RC adds an extra chromophore to the 'photosynthetic unit', in a spectral region already occupied by carotenoids. Thus, it was necessary for this pilot study to make some prior alterations to the bacterial chassis by eliminating the LH2 antenna and carotenoids, clearing space in spectral terms to allow assessment of any contribution from YFP.
In this study we have shown that YFP, when fused to the RC of Rba. sphaeroides, is able to transfer electronic excitation energy to the native RC. This process has been observed at three levels; in bacterial cultures, in single cells, and in purified complexes (Figs 1, 2 and 4-7). Attachment of YFP to the RC increases the photosynthetic growth rate of a carotenoid-less mutant of Rba. sphaeroides (compared with a control strain with no YFP) when YFP-specific illumination is used. Although the most straightforward explanation of this result is energy transfer from YFP to LH1 and RC pigments (supported by the static and time-resolved microscopy and spectroscopy) we checked for the possibility that YFP mitigates the photosensitivity of the carotenoid-less DcrtB background strain 34,35 since it has been proposed that fluorescent proteins may have a photoprotective function 26,27 . We found that the DcrtB pBBRBB-YFP control strain with cytosolic YFP present, unattached to the RC, did not grow significantly faster than DcrtB and therefore we discount any photoprotective effect of YFP.
We attribute differences in growth rates between carotenoidcontaining and carotenoid-less strains (Fig. 2) to the lack of appropriate pigments for absorbing light at 520 nm and show that the attachment of YFP to RCs partly alleviates this problem. Other factors could potentially contribute to the relatively slow growth of carotenoid-less strains, such as destablization of the photosynthetic apparatus or susceptibility to photodamage. We have shown that only the assembly of the LH2 antenna is affected by inactivation of carotenoid biosynthesis [22][23][24][25] ; thus, we genetically removed LH2 complexes in the positive control so all strains could be compared on an equal basis in terms of antenna composition. In terms of photodamage, respiration by these photoheterotrophic bacteria during illumination alleviates phototoxic effects in carotenoid-less mutants. Thus, under white light illumination, even the DcrtB mutant achieves good rates of photosynthetic growth even with respect to the carotenoidcontaining positive control.
Direct observation of YFP-RC-LH1 energy transfer in whole cells, visualized using fluorescence lifetime microscopy (Fig. 4), is consistent with the increased photosynthetic growth rate of DcrtB RC-H-YFP compared with DcrtB (Fig. 2a-c); the in vivo efficiency of energy transfer, F EET , was calculated to be 0.48. To discover a structural basis for in vivo energy transfer, we used single particle reconstruction to obtain a structural model of the RC/YFP-LH1 complex (Fig. 3). The use of negative stain limited the resolution of the structural model to B26 Å, but the presence of YFP was evident from the outset, even in raw images, and the final structural model clearly shows a density that can be confidently assigned to YFP. The staining procedure could affect the perceived orientation of YFP with respect to the rest of the complex; we used over 11,000 particles for calculating the structural model so Fig. 3 shows an averaged orientation of YFP. The use of cryo-EM would produce a better quality structural model, which would also allow a more precise estimate of the distance between the YFP chromophore and the RC special pair. Nevertheless, the low resolution data are sufficient for a calculation of an B85 Å distance between the YFP chromophore and the RC special pair and 53 Å to the nearest LH1 B875 BChl (Fig. 3f). Our energy-transfer data provide another way to estimate distances and Förster calculations ( Supplementary  Fig. 5) show that the acceptor of excitation energy lies 43-48 Å from the YFP donor, consistent with our 53 Å estimate from the structure of the RC/YFP-LH1 complex. Thus, the most likely route for energy transfer from YFP to the RC in vivo is via the LH1 BChls.
To make a minimal energy-transfer unit free of LH1, we purified a RC/YFP complex and quantified its energy-transfer dynamics using static and time-resolved absorption and fluorescence spectroscopy. The TA measurements for the RC/YFP complexes (Figs 6 and 7) show an additionalB1.8 ns component indicating YFP*-RC energy transfer, which is not seen in the RC-only control complex. The energy-transfer yield, F EET , was calculated to be 0.35. Measurements of fluorescence yields for RC/YFP and YFP-only by two methods gave 0.47 and 0.58, and fluorescence lifetimes for RC/YFP-RC versus YFP-only using two different methods gave F EET ¼ 0.44 and 0.31. The average value for F EET is 0.4 ± 0.1. We calculated a 5 ns time constant for YFP*-RC energy transfer and a YFP to RC special pair distance of 56-62 Å ( Supplementary Fig. 8), significantly lower than the 85 Å estimate from the RC/YFP-LH1 structure (Fig. 3f). There is a likely structural basis for this apparent discrepancy; the region connecting the YFP and RC-H could accommodate some movement of the YFP in the absence of LH1, allowing YFP to move nearer to the RC 'special pair' BChls (P) and resulting in more efficient energy transfer. This possible change in YFP position, caused by the absence of LH1, is depicted in Fig. 8. The YFP-RC distance of 45 Å measured from this model is shorter than calculated from energy-transfer data, but the model represents the minimum possible YFP-RC distance and in practice the YFP likely does not approach the RC so closely, not least because some detergent molecules likely surround the central part of the RC complex.
In summary, this proof-of-principle study opens up possibilities for the creation of new energy-trapping photosynthetic pathways; we show that translational fusions of genetically-encoded light-absorbers and antenna or RC complexes can augment energy transfer and trapping in photosynthesis.

Methods
Construction of mutant Rhodobacter sphaeroides strains including the DcrtB deletion strain. A construct containing the upstream and downstream flanking regions of the gene was created to delete crtB. A 361 bp fragment upstream of crtB and the first 7 bp of the crtB gene flanked with EcoRI and XbaI restriction sites (underlined) was amplified using the primers crtBKOUF (5 0 -CCGGAATTCCA CATCACCATCACCACGGCG-3 0 ) and crtBKOUR (5 0 -GCGCTCTAGA GATCTAGGTTCTCATGAAGGTATACCG-3 0 ). A second fragment was amplified using the primers crtBKODF (5 0 -GCGCTCTAGAGGCAATCAT TCCGCGGCAAGC-3 0 ) and crtBKODR (5 0 -CCCCGCATGCGGCTGTGGCCGA GCCCTA-3 0 ) producing a downstream flanking fragment of 363 bp, which included 9 bp at the end of the crtB gene, flanked with XbaI and SphI restriction sites (underlined). Following digestion with the indicated enzymes, the fragments were sequentially cloned into the same restriction sites of the suicide vector pK18mobsacB. The resulting plasmid was introduced into Escherichia coli S17-1 competent cells and transferred to Rba. sphaeroides WT via conjugation, with first and second recombination events selected for as detailed in ref. 31. Transconjugant Rba. sphaeroides colonies were selected on M22 agar plates with 25 mg ml À 1 kanamycin. Single colonies were grown and scaled up to an 80 ml semi-aerobic culture. The cells were serially diluted 10 À 2 , 10 À 3 and 10 À 4 onto M22 agar containing 1 % (w/v) sucrose and incubated until single colonies appeared after 4-6 days. Single colonies were replica plated onto M22 sucrose plates, with and without 25 mg ml À 1 kanamycin. Colonies that grew on the antibiotic-free plate but not on the kanamycin plate were analysed by PCR to identify successful mutants.Transconjugants were confirmed by DNA sequencing; the modified DNA sequence at the location where crtB is deleted is shown in Supplementary Fig. 9.
The RC/YFP fusion strain was created using a pk18mobsacB construct designed to fuse the gene for SYFP2 (ref. 21) to the 3 0 end of the puhA gene encoding the RC-H subunit. To create a C-terminal YFP fusion protein, firstly a PCR fragment was generated using primers puhAYFPUF (5 0 -CCGGAATTCTCGGCCGGCAAG AACCCGATCGG-3 0 ) and puhAYFPUR (5 0 -GCTCCTCGCCCTTGCTCACCAT GGCGTATTCGGCCAGCATCGCCG-3 0 ). This fragment contained the last 354 bp of the puhA ORF (excluding the stop codon), with a 5 0 EcoRI restriction site (underlined) and 3 0 sequence complementary to the start of the syfp2 gene (italics). A second PCR with primers puhAYFPFor (5 0 -CGGCGATGCTGGCCGAATACGCC ATGGTGAGCAAGGGCGAGGAGCTGTTCAC-3 0 ) and puhAYFPRev (5 0 -GCGC TCTAGATCATTACTTGTACAGCTCGTCCATGCCGAGAGTGAT-3 0 ) amplified the syfp2 ORF with 5 0 sequence complementary to the end of the puhA gene (italics) and a 3 0 XbaI restriction site (underlined). The two fragments were used as a template for overlap extension (OLE)-PCR with primers puhAYFPUF and puhAYFPRev, generating a fragment in which the end of the puhA gene was joined in frame with the syfp2 gene with EcoRI and XbaI sites at the 5 0 and 3 0 ends respectively. A third fragment in which the 354 bp immediately downstream of the puhA stop codon was flanked by XbaI and HindIII restriction sites (underlined) was amplified using the primers puhAYFPDF (5 0 -GCGCTCTAGATCCCCGCAT GGCGCGGCCC-3 0 ) and puhAYFPDR (5 0 -CCCCAAGCTTTAGGGCACCGCA TAGGCCACCGC-3 0 ). Following XbaI/HindIII digestion, the downstream fragment was cloned into the suicide vector pK18mobsacB. The resulting plasmid was opened with EcoRI and XbaI and the puhA-sfYFP2 product cut with the same enzymes was inserted. This plasmid contained a total insert of 1,428 bp. The fragments were inserted in this order as the XbaI site in the OLE-PCR product is blocked by overlapping dam methylation once inserted into a plasmid. Conjugative transfer and screening of possible mutants was performed as described for the crtB deletion.
Growth of Rhodobacter sphaeroides. Rba. sphaeroides strains were grown in M22 þ medium 36 supplemented with 10,000X vitamins (0.08 M nicotinic acid, 0.01 M thiamine, 7.3 mM 4aminobenzoic acid, 0.4 mM d-biotin), to a final concentration of 1X. Liquid cultures were supplemented with 0.1% casamino acids. Single colonies were used to inoculate 10 ml starter cultures of M22 medium and grown semi-aerobically in the dark with shaking at 180 r.p.m. at 34°C for 48 h. For anaerobic photosynthetic cultures, 20 ml of M22 þ medium was used to dilute 10 ml of semi-aerobic culture in a 30 ml universal. The culture was incubated for 48 h. This culture was used to inoculate a 1 l Roux bottle filled with M22 þ medium and grown for 24 h. The culture was used to inoculate 8 l of M22 þ medium and grown to an optical density at 680 nm (OD 680 ) of 3. All cultures were gently stirred using magnetic stir bars. Light was provided by 20 W MEGAMAN CFL bulbs at 100 mmol photons s À 1 m À 2 , measured using a LI-250A Light Meter equipped with a LI-190 Quantum Sensor (LI-COR Biosciences). Where required, cells were pelleted by centrifugation and resuspended in 20 mM HEPES pH 8.
For photosynthetic growth curves, 48 h, 10 ml semi-aerobic starter cultures were diluted to 30 ml and grown photosynthetically for 72 h. Photosynthetically grown cells were used to inoculate 17 ml tubes to a starting OD 680 of B0.15. Three tubes were grown per strain and average values and standard deviations are presented (Fig. 3). Light was provided using an array of green LEDs (520 nm emission maximum, 35 nm FWHM, 170 mmol photons s À 1 m À 2 (Würth Elektronik, Germany).
Purification of RCYFP-LH1 complexes. Monomeric core complexes were purified using a method adapted from ref. 36. Lysozyme was added to the cell sample to a final concentration of 0.5 g l À 1 and incubated at 25°C for 30 min in the dark. A small spatula of deoxyribonuclease 1 from bovine pancreas (Sigma) was added to the cells. The cells were French pressed three times at 18,000 psi. The cells were kept on ice throughout. Unbroken cells were removed by centrifugation at 33,000 g at 4°C for 25 min. The supernatant was applied to the top of a 15/40% sucrose gradient followed by centrifugation for 16 h at 65,000 g. The ICM fraction was harvested from above the 40 % sucrose layer and pelleted by centrifugation at 125,171 g for 2 h and resuspended in 20 mM HEPES pH 8. The membranes were solubilised with 3% b-dodecyl maltoside (b-DDM) for 1 h at 4°C with continuous stirring in the dark. The insoluble material was removed by centrifugation at 160,000 g for 1 h. The supernatant containing solubilised membranes was layered on top of discontinuous sucrose gradients containing 20, 21.25, 22.5, 23.  After incubation with primary antibody the membrane was washed with TBS and then incubated with a 1:10,000 dilution of secondary anti-rabbit antibody conjugated with peroxidase (Sigma Aldrich) in TBS. The membrane was washed with TBS before immunodetection using the Amersham ECL Western blotting analysis system (GE Healthcare Life Sciences) according to the manufacturer's instructions. Uncropped immunoblots are displayed in Supplementary Fig. 10.
Electron microscopy. Electron microscopy and single particle reconstruction of RC/YFP-LH1 complexes was performed as described previously 37 . The detailed method was as follows. Purified RC/YFP-LH1 complexes were diluted to 0.1 mg ml À 1 . Five ml protein solution was applied to a carbon coated 400 mesh copper EM grid, which was glow discharged 30 s before use. Excess protein solution was blotted by touching the edge of the grid on the surface of Whatman filter paper. After washing twice with distilled water, grids were stained with 0.75% (w/v) uranyl formate for 30 s. The grids were imaged using a Philips CM100 electron microscope operating under 100 kV accelerating voltage at room temperature. All micrographs were recorded with a 1 K Â 1 K Gatan Multiscan 974 CCD camera at a normal magnification of Â 52,000, corresponding to 3.93 Å per pixel at specimen level with variation of under-focus value from 0.5 to 2.5 mm.
The micrographs were evaluated initially by Digitalmicrograph (Gatan). All images showing drifting, poor staining and unsuitable under-focus were discarded. Selected images were submitted to EMAN2 (ref. 38) for CTF correction and particle picking. A total of 11,777 particles, each 72 Â 72 pixels, were treated subsequently using the IMAGIC-5 software package 39 . All particles were initially bandpass filtered to suppress low spatial frequencies. A soft-edged circle mask was used to remove unwanted background of the particle. The masked data were then centered and normalized for reference-free 2D classification, from which a set of characteristic classes was selected for multi-reference alignment (MRA) and multivariant statistical analysis (MSA) classification. This procedure was iterated until a stable 2D classification was obtained. An initial model was established by assigning Euler angles to the three most different 2D classes selected from the above calculation using angular reconstitution in IMAGIC-5; this model was refined gradually by adding more 2D classes. An anchor set produced from back projection of the initial model was used to calculate the Euler angle of whole 2D class dataset, from which a new 3D model was produced. This procedure was iterated by increasing the calculation precision and decreasing the search step of the Euler angle until a stable 3D model obtained. A new reference produced from this stable 3D model was then used for a new MSA iteration. Finally, 380 2D classes were used for generation of a 3D RC/YFP-LH1 model. The final 2D class dataset was split randomly into two halves, and two 3D models were reconstituted independently. By calculation of Fourier shell correlation (FSC) between these two 3D models, a resolution of 25.9 Å for the final model ( Supplementary Fig. 11) was estimated based on 0.5 FSC criteria 40 .
Room temperature and low temperature absorbance spectra. Room temperature and 77 K absorption spectra were recorded using a Cary ultraviolet/vis spectrophotometer in the spectral range between 260 and 950 nm. For cryogenic measurements, samples suspended in a cryo-stable buffer consisting of 20 mM Tris-HCl, 80% glycerol (v/v) were cooled to 77 K in an Optistat DN-V optical cryostat manufactured by Oxford Instruments.
Low temperature spectroscopy. The absorption and excitation fluorescence spectra in Fig. 1 were recorded in a cryo-stable buffer consisting of 20 mM HEPES, 80% glycerol. Samples were cooled to 77 K as described above. Measurements were recorded on a SPEX FluoroLog spectrofluorimeter (SPEX Industries). Excitation was provided from a tungsten light source. Fluorescence excitation spectra were recorded from 400 to 910 nm with an emission of 915 nm, using an average of 10 individual scans. Excitation and emission slit widths of 5 nm were used. Fluorescence excitation spectra were normalized to 1 at 591 nm, data were subject to 2nd order smoothing with 5 neighbours.
Fluorescence imaging of YFP in Rhodobacter sphaeroides. Cells were washed three times in QH 2 O then suspended in QH 2 O. Overall, 15 ml of cell suspension was dropped on 30 ml of 1.5% agar film on a glass slide and sealed with DPX mountant (Sigma À Aldrich) between the glass slide and a coverslip. Fluorescence images were taken with an inverted fluorescence microscope (AxioObserverA1m, Zeiss) equipped with a Hal 100 halogen lamp, a high intensity HBO 100 mercury lamp and an ORCA-ER camera (Hamamatsu). Excitation light was first filtered by a 470/40 nm bandpass filter, then reflected by a 495 nm dichroic beam splitter to the sample through an objective (Plan-Apochromat Â 63/1.40 oil objective, Zeiss). Fluorescence emission was filtered by a 520/40 nm bandpass filter before detection by the camera. Each fluorescence image was taken with a 0.1 s exposure time and 50 electron multiplication gain.
Spectral and lifetime imaging of Rba. sphaeroides. The fluorescence emission properties of YFP in whole cells were measured on a home-built time-resolved fluorescence microscope. The inverted microscope is equipped with a spectrometer (Acton SP2558, Princeton Instruments), an electron-multiplying charge-coupled device (EMCCD) camera (ProEM 512, Princeton Instruments) and a Hybrid Detector (HPM-100-50, Becker & Hickl). The excitation light source was from a pulse supercontinuum white light laser (SC 480-10, Fianium) with a repetition rate of 40 MHz. The laser beam was focused on the sample surface, illuminating a diffraction limited spot using a Â 100 objective (PlaneFluorite, NA ¼ 1.4, oil immersion, Olympus). The resulting fluorescence emission was detected through the spectrometer onto the EMCCD camera and the resulting fluorescence lifetime was detected through the spectrometer onto the Hybrid Detector.
During fluorescence spectral and lifetime measurements, the excitation light was filtered by a 472/30 nm bandpass filter, then reflected by a 495 nm dichroic beamsplitter to the sample. The resulting fluorescence emission was filtered by a 496 nm long-pass filter before being detected by the cameras. The fluorescence emission was captured with a slit width of 1.5 mm and a grating of 150 line per mm working at a central wavelength of 550 nm in the spectrometer. Multiple measurements were performed on eight different cells on each sample. Each fluorescence spectrum was detected by EMCCD at an average of three frames with a 1 s exposure time and an electron multiplication gain of 80. Analysis was done with OriginPro.
For fluorescence lifetime measurements, the modulation of the laser was synchronized with a time-correlated single-photon counting (TCSPC) module (SPC-150, Becker & Hickl). Fluorescence lifetimes were recorded by parking the focused laser spot over one single cell, selecting a central wavelength of 550 nm by use of the monochromator and detected by the Hybrid Detector. SPCM software (Becker & Hickl) was used for the data acquisition. The families of decay curves were analysed with OriginPro and TRI2 software packages by fitting with a multiexponential decay function: where t i is the fluorescence lifetime, A i is the fractional amplitude contribution of the ith decay component, and B is the background. The quality of fit was judged on the basis of the reduced w 2 statistic: where I(t k ) is the data at time point k, I c (t k ) is the fit at time point k, n is the number of data points and p is the number of variable fit parameters (n-p ¼ degrees of freedom). The instrument response (IRF) of the system, measured using a mirror, was B0.18 ns, and the convolution of the decay curves with the IRF was taken into account when the fitting was performed.
Optical measurements on purified RC and RCYFP complexes. Photophysical measurements on RC, YFP or RC/YFP proteins in 0.1% LDAO, 10 mM Tris, 200 mM NaCl, pH 7.8 buffer were carried out at room temperature. Static absorption (Shimadzu UV-1800) and fluorescence measurements (Horiba Nanolog) employed B1 mM samples, which give Ar0.1 at l exc for fluorescence studies. The latter used 2 nm excitation and emission bandwidths, and spectra were corrected for IRF. Fluorescence quantum yields are the average values obtained using methods: (1) an absolute method using an integrating sphere (Horiba, Quanti-Phi), and (2) a relative method using as a standard fluorescein in 0.1 M NaOH, for which F f ¼ 0.91 was obtained from the average of two literature values 41,42 . Singlet excited-state lifetimes were determined using (1) transient absorption (TA) spectroscopy, (2) timecorrelated-single-photon-counting (TCSPC) detection of fluorescence decay (IRF function of B0.2 ns), and (3) stroboscopic measurement of fluorescence decay (IRF function of B1 ns) as described previously 43 . TA studies utilized an amplified Ti:Sapphire laser system (Spectra Physics) and Helios and Eos detection systems (Ultrafast Systems). Samples were excited with B0.5 mJ, B100 fs 515-nm excitation pulses (at 1 KHz) focused to 1 mm diameter 44 . For RC-only or RC/YFP, a spinning cell (2 mm path and 3 mm wide annulus) containing 3 ml of B10-15 mM sample was used to prevent re-excitation of the complex while in the long-lived (1-2 s) P þ Q B À state of the RC. A standard 2 mm path cuvette with stirring was used for YFP-only (600 ml at B15 mM).
Förster calculations were performed using the measured yield of excitation energy transfer (F EET ) from excited YFP to RC to obtain the donor to acceptor distance (R) between these units. The calculations employed PhotochemCAD 32 , which uses the 'R 0 method' in which F EET ¼ R 0 6 /(R 0 6 þ R 6 ), where R 0 6 ¼ (8.8 Â 10 23 ) k 2 F f J n À 4 and R 0 is the distance at which F EET ¼ 0.5. Here, the orientation factor k 2 ¼ 0.667 assuming a random orientation of transition dipoles, F f ¼ 0.70 (vide infra) is the donor (YFP) fluorescence yield, J is the spectral-overlap factor (a molar absorptivity of 288,000 mM À 1 cm À 1 at 800 nm was used for the RC 45 ), and n is the refractive index, for which values of 1.335, 1.45 and 1.55 were used to accommodate protein/cofactor environments ranging from buffer to pure hydrocarbon ( Supplementary Figs 5 and 8).
Data availability. The authors declare that the data supporting the findings of this study are available within the paper and its Supplementary Information files or are available from the corresponding author upon request.