Carbon Dots as Versatile Photosensitizers for Solar-Driven Catalysis with Redox Enzymes

Light-driven enzymatic catalysis is enabled by the productive coupling of a protein to a photosensitizer. Photosensitizers used in such hybrid systems are typically costly, toxic, and/or fragile, with limited chemical versatility. Carbon dots (CDs) are low-cost, nanosized light-harvesters that are attractive photosensitizers for biological systems as they are water-soluble, photostable, nontoxic, and their surface chemistry can be easily modified. We demonstrate here that CDs act as excellent light-absorbers in two semibiological photosynthetic systems utilizing either a fumarate reductase (FccA) for the solar-driven hydrogenation of fumarate to succinate or a hydrogenase (H2ase) for reduction of protons to H2. The tunable surface chemistry of the CDs was exploited to synthesize positively charged ammonium-terminated CDs (CD-NHMe2 ), which were capable of transferring photoexcited electrons directly to the negatively charged enzymes with high efficiency and stability. Enzyme-based turnover numbers of 6000 mol succinate (mol FccA)−1 and 43,000 mol H2 (mol H2ase) −1 were reached after 24 h. Negatively charged carboxylateterminated CDs (CD-CO2 −) displayed little or no activity, and the electrostatic interactions at the CD−enzyme interface were determined to be essential to the high photocatalytic activity observed with CD-NHMe2 . The modular surface chemistry of CDs together with their photostability and aqueous solubility make CDs versatile photosensitizers for redox enzymes with great scope for their utilization in photobiocatalysis.


INTRODUCTION
Natural photosynthesis inspires the design of synthetic systems for the conversion of solar energy into chemical fuels i.e. artificial photosynthesis. [1][2][3] This approach has spurred the development of semi-biological systems that couple a synthetic light harvesting unit to an isolated enzyme, [4][5][6][7][8] or organism. [9][10][11][12][13] Such systems combine the efficient light-harvesting properties of synthetic materials with the high selectivity and rates of biological catalysts. 9,13 Semi-biological systems require a profound understanding of the mechanisms of interfacial electron transfer between the abiotic and biotic component. Significant work has focused on integrating isolated redox enzymes into proof-of-principle photocatalytic and electrochemical systems for solar fuel synthesis. 4,6,[14][15][16] Such systems allow enzymatic redox processes to be decoupled from their biological metabolism and to be studied in isolation, providing insights into the interfacial electron transfer processes between the protein and an electrode or dye. 17,18 Hydrogenases (H 2 ases) are of particular interest for use in semi-biological solar fuels systems as they are model precious metal-free proton reduction electrocatalysts, capable of operating with turnover frequencies comparable to Pt. [19][20][21][22] The light-harvesting ability of natural photosystems has been coupled to H 2 production by H 2 ase in vitro via H 2 ase-photosystem I protein conjugates, 23,24 or by the electrochemical wiring of H 2 ase to photosystem I 25 and II. 15 Approaches employing synthetic light absorbers include immobilizing H 2 ase onto silicon photoelectrodes 6,8 as well as utilizing H 2 ase in homogeneous and semi-heterogeneous photocatalytic systems using organic and inorganic molecular dyes [26][27][28][29] or semiconductor particles such as dyesensitized TiO 2 , 30,31 graphitic carbon nitride, 32 Cd chalcogenides 33-35 and In 2 S 3 . 36 Beyond the use of enzymes as benchmark catalysts for solar fuel production, enzyme-driven catalysis is also well established for achieving complex organic redox transformations with high selectivity. 37-39 Often however, stoichiometric equivalents of expensive reducing cofactors such as NADPH are required to mediate the enzymatic reaction. 37,40,41 2 Figure 1. Representations of the two independent photocatalytic systems studied in this work: photo-excited CDs transfer electrons to a fumarate reductase (FccA; system A) or a [NiFeSe]-hydrogenase (H 2 ase; system B) for the reduction of fumarate to succinate and protons to H 2 , respectively; EDTA acts as a sacrificial hole scavenger. The location of the active site in each enzyme is labeled in addition to the Fe-heme and [Fe 4 S 4 ] cofactors for FccA and H 2 ase, respectively.
As such, there is scope for the integration of enzymes into bio-hybrid photocatalytic schemes that utilize solar energy to drive NADPH recycling, or to directly reduce the enzyme, eliminating the need for the co-factor altogether. 38,40,42,43 Only a small number of mediator-free dye-enzyme hybrids have been reported to photocatalyze organic chemical transformations however, with the photosensitizers typically being limited to (dyesensitized) TiO 2 nanoparticles, 44 organic 11 and Ru II tris(2,2'-bipyridine) dyes. 29,45 Despite the widespread use of dyes with enzymes for photocatalysis as well as in the study of electron transfer in dye-protein 46 and dye-nucleic acid 47 assemblies, there remains a limited number of suitable photosensitizers available. Photosystems 23,24 and abiotic molecular dyes 11,26 suffer from fragility. Molecular dyes can also be time-consuming to synthesize and are often expensive, particularly those containing precious-metals. 44,48 Semiconductor particles often have good stability, but common disadvantages include: poor interfacial interaction with the redox partner (graphitic carbon nitrides), 32 toxicity (Cd chalcogenides), 33-35 low elemental abundance (Cd, Te, and In) 35,36,49 and poor aqueous dispersibility (TiO 2 and carbon nitrides). 31,32,44,48 Most light absorbers are also challenging to chemically modify and can create highly oxidizing or reducing species that are detrimental to the lifetime of a bio-macromolecule. 44 Thus, there remains a need for the development of photosensitizers that are capable of interfacing efficiently with biological materials.
In this study, carbon dots (CDs) are established as versatile, low-cost, stable and efficient photosensitizers for photobiocatalysis ( Figure 1). First, a system utilizing CDs and fumarate reductase (FccA), a flavoenzyme from Shewanella oneidensis MR-1, is presented for the photoconversion of fumarate to succinate; a model sys-tem for enzyme-driven C=C bond hydrogenation, an important redox transformation in organic chemistry. 41,43,44 The study of this photocatalytic hybrid system was complemented by optical spectroscopy, monitoring the photoreduction of the four Fe 3+ -heme cofactors in FccA. 50,51 Finally, solar-H 2 generation is shown using CDs and a [NiFeSe]-H 2 ase from Desulfomicrobium baculatum (Dmb). Taking advantage of the tunable CD surface chemistry, we have synthesized negatively and positively charged CDs and demonstrate that only the cationic CDs are capable of efficient photoelectron transfer to the negatively charged proteins, due to a favorable interfacial interaction.

RESULTS AND DISCUSSION
Synthesis and Characterization of CDs. CDs are photoluminescent carbon nanomaterials that have promising light harvesting and electron transfer capabilities for applications in photocatalysis. [52][53][54] We have recently demonstrated CDs to be stable and efficient photosensitizers in a homogeneous photocatalytic H 2 generation scheme using a synthetic molecular Ni catalyst. 55,56 Although CD-protein assemblies have been studied previously 57-62 they have not been reported for the lightdriven synthesis of chemical products; i.e., artificial photosynthesis.
The solubility in water and biocompatibility of CDs has already led to significant interest in CDs for nanotherapeutic, 63,64 bioimaging 65,66 and biochemical sensing 67-69 applications. Furthermore, CD surfaces are typically carboxylic acid or amine-terminated and can be readily manipulated. This provides a versatile platform to design and tune the surface chemistry interface between the CDs and a protein, which may be crucial to optimize electron transfer efficiency. 70 Carboxylic acid and amine-terminated CDs are expected to be anionic 3 and cationic at pH neutral conditions, respectively, and are therefore well suited to probe the effect of electrostatic interactions with a redox partner.
The carboxylic acid and amine-terminated CDs used in this study were prepared as shown in Scheme 1, with experimental and characterization details described in the Experimental Section. Carboxylic-acid capped CDs (CD-CO 2 H) were synthesized by the thermal decomposition of citric acid as previously reported. 55, 71 The carboxylic acid surface groups were then selectively modified by treatment with thionyl chloride to yield acyl chloride-capped CDs (CD-COCl). 72,73 CD-COCl can be isolated by evaporation of the excess thionyl chloride and are readily soluble in polar organic solvents such as tetrahydrofuran, acetonitrile and acetone. The high reactivity of the acyl chloride group and absence of nonvolatile by-products provides an adaptable route to tunable surface chemistry without the need for elaborate purification procedures. 73 Stirring CD-COCl with the nucleophile N,N-dimethylethylenediamine (DMEN) at room temperature, followed by evaporation of the residual DMEN results in tertiary amine-capped CDs (CD-NMe 2 ). The tertiary amine was chosen to avoid crosslinking between particles or adjacent surface groups. The chemical modification of the CD-CO 2 H carboxylic acid surface groups was monitored by attenuated total reflectance Fourier transform infrared spectroscopy (ATR FT-IR; Figure 2). As expected, a clear shift in the C=O stretching frequency is observed, from = 1701 cm −1 (CD-CO 2 H), to 1760 cm −1 (CD-COCl) and then to 1654 cm −1 (CD-NMe 2 ). 73 The amide carbonyl stretch (1654 cm −1 ) in CD-NMe 2 with a corresponding amide N-H bending mode ( = 1546 cm −1 ) confirms that DMEN has reacted with CD-COCl and is covalently bound to the CD surface. Further features of the DMEN reagent are also observed on the CD-NMe 2 surface, such as the C-H stretching ( = 2948, 2865, 2827 and 2770 cm −1 ) and bending ( = 1460 cm −1 ) modes. The small peak at = 1700 cm −1 in CD-NMe 2 suggests that some acyl chloride residues from CD-COCl have been hydrolyzed possibly by residual water in the DMEN reagent. The broad signal at = 3500-3260 cm −1 in CD-NMe 2 is most likely largely due to absorbed water given the sample is very hydroscopic. The surface functionalization of CD-NMe 2 was further studied by 1 H NMR spectroscopy, shown in Figure  S1. Two sets of triplets and a singlet are observed, which we assign to the ethylene and methyl protons of the surface functional groups, respectively. No proton signals from DMEN or the protonated DMEN salt are observed in the final CD-NMe 2 product, which supports that all excess amine reagent has been removed during the work-up process.
Transmission electron microscopy (TEM) and UV-Visible (UV-Vis) absorption spectroscopy indicate that the bulk physical properties of the CDs are not significantly altered during the surface functionalization procedure. TEM analysis showed an average particle size of 6.4 nm ± 1.2 nm for CD-NMe 2 (see Figure S2), which is similar to CD-CO 2 H. 55 The UV-Vis spectrum of CD-NMe 2 in aqueous solution shows a broad absorption typical of CDs, assigned to various π-π* (C=C) and n-π* (C=O) absorptions ( Figure S3). 53 The absorption onset of CD-NMe 2 is slightly shifted to longer wavelengths compared to CD-CO 2 H, which is consistent with previous reports of amide-functionalized CDs. 72 Zeta potential (ζ) measurements were carried out to study the surface charge of the CDs ( Figure S4). The 4 isoelectric point (I p ) for CD-CO 2 H was determined to be approximately 1, whereas the I p for CD-NMe 2 was approximately 9. The surfaces of CD-CO 2 H and CD-NMe 2 are therefore negatively (ζ = −17 ± 5 mV) and positively charged (ζ = +17 ± 1 mV) under the photocatalysis conditions employed in this study (pH 6, see below). We therefore denote the CDs as CD-CO 2 − and CD-NHMe 2 + to accurately describe their ionic character in the photocatalysis sections below. Photoreduction of FccA with CDs. FccA, a flavoenzyme c 3 from Shewanella oneidensis MR-1, catalyzes the reduction of fumarate to succinate. FccA is an ideal candidate to study electron transfer from CDs due to the positive redox potential for fumarate reduction (E 0 ' = 0.03 V vs. SHE at pH 7) and low FccA overpotential (80 mV at pH 7). 44,74 Fumarate is reduced to succinate at the flavin active site by the transfer of one hydride from the flavin and one proton from a nearby arginine residue. [74][75][76] The electron transport chain from the surface to the flavin active site consists of four c-type Fe-heme groups, as shown in Figure 1. These Fe 3+ -heme cofactors are characterized by an intense Soret band at 408 nm as well as a broad feature from 500 to 600 nm. Upon reduction of the protein, new bands appear at 419, 552 and 523 nm, which can be assigned to the Fe 2+ -heme cofactors. 50,51 The characteristic absorption of the Fe 2+ -heme cofactors thus provides a useful spectroscopic handle to study the ability of CDs to photoreduce FccA using UV-Vis absorption spectroscopy. An aqueous solution containing the buffer and sacrificial electron donor ethylenediaminetetraacetic acid (EDTA; 0.1 M, pH 6) was selected as the reaction medium, as EDTA has previously been shown to be the optimized electron donor for CDs 55 under these conditions and the slightly acidic pH is within the ideal pH range for FccA activity (pH 5 -8). 50,51,74 Figure 3 shows the time-resolved UV-Vis spectra of a solution containing FccA (0.33 μM) and CD-NHMe 2 + (1 mg mL −1 ) in aqueous EDTA solution under simulated solar light irradiation (AM 1.5G, 100 mW cm -2 ) at 25 °C in the absence of fumarate (see Figure S5 for CD-CO 2 − ). Upon irradiation, the characteristic Fe 2+ -heme absorption bands are observed to appear and reaching a steady-state after approximately 8 min ( Figure S5). Although both CD-NHMe 2 + and CD-CO 2 − are capable of reducing FccA over several minutes, CD-NHMe 2 + photo-reduces FccA at a substantially higher rate, with 60% of the heme units reduced after 1 min compared to 20% when using CD-CO 2 − (see inset Figure 3). This difference in the initial rate of FccA reduction under non-turnover conditions becomes significant in the study of the CD-FccA systems during photocatalysis in the presence of substrate (see below).

Solar-driven Fumarate Reduction with CDs and
FccA. The CD-FccA system was subsequently studied under simulated solar light irradiation in the presence of fumarate and the photo-generation of succinate was monitored by 1 H NMR spectroscopy ( Figure 4, Table  S1). The photocatalytic experiments were set up by dissolving CDs (1 mg), FccA (0.22 nmol) and fumarate (10 mM) in an EDTA solution prepared in D 2 O (0.1 M, pD 6.4, 77 1 mL; see Experimental Section for details). The vials were then purged with N 2 and irradiated by a solar light simulator (AM 1.5G) at 100 mW cm −2 (1 sun). Aliquots of the photocatalysis mixture (0.2 mL) were analyzed directly by 1 H NMR after 2, 4, 8 and 24 h. The turnover number (TON) for succinate production per FccA (TON FccA ) was determined by the integral ratio of the 1 H NMR peaks corresponding to fumarate (6.44 ppm) and succinate (2.33 ppm). 44 Representative 1 H NMR spectra are shown in Figure S6. Figure 4 shows that while both CDs are capable of direct electron transfer to FccA, the photoreduction of fumarate is significantly enhanced with CD-NHMe 2 + compared to CD-CO 2 − , consistent with the higher rate of Fe 3+ -heme reduction observed spectroscopically. Control experiments in the absence of CDs, FccA or EDTA showed no succinate production. The similar bulk physical properties of the CDs suggest that the markedly different performance in photocatalytic succinate production is due to the different surface charges of the CDs,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 5 resulting in different interfacial interactions with FccA (see discussion in CD-Enzyme Interface section below).
The CD-NHMe 2 + -FccA system operated at an initial enzyme-based turnover frequency (TOF FccA ) of 1.7 ± 0.2 × 10 3 mol succinate (mol FccA) −1 h −1 and remained photoactive for at least 24 h after which time a TON FccA of 6.0 ± 0.6 × 10 3 mol succinate (mol FccA) −1 had been reached ( Figure 4, Table S1). The decrease in rate observed after 2 h is attributed to enzyme degradation rather than photoinstability of the CDs (see CD-Enzyme Interface section below). Previously FccA has only been studied under photocatalysis conditions using Ru dye-sensitized TiO 2 nanoparticles as the photosensitizer, achieving a comparable TON FccA of 5.8 × 10 3 with activity ceasing after 4. 44

Solar-driven H 2 Production with CDs and H 2 ase.
Having established that the CDs are capable of direct electron transfer to FccA in an aqueous sacrificial photocatalysis system, we investigated the CDs in a solar H 2 production system with Dmb [NiFeSe]-H 2 ase (Figure 1). Dmb [NiFeSe]-H 2 ase was chosen as it is well established for use in homogeneous and semi-heterogeneous photocatalytic H 2 generation systems. 78 [NiFeSe]-H 2 ase is a sub-class of [NiFe]-H 2 ases with a selenocysteine residue coordinated to Ni at the active site instead of a cysteine residue. 79 Three [Fe 4 S 4 ]-clusters form the electron transport chain to the active site, as shown in Figure 1.
[NiFeSe]-H 2 ases are known for their good H 2 evolution activity (E 0 ' (H + /H 2 ) = −0.41 vs. SHE at pH 7), resistance to O 2 and low product inhibition. 79,80 For the CD-H 2 ase systems studied here, EDTA was again chosen as sacrificial electron donor and pH 6 as the optimal pH for H 2 ase activity. 48 A ratio of H 2 ase (50 pmol) to CD (10 mg) was selected after optimization of the CD loading for the highest TON H2ase after 24 h; the H 2 ase loading was kept at 50 pmol to aid comparison to comparable systems with different light absorbers (see Figure S7). 31,32 The photocatalytic systems were typically assembled by dissolving CDs (10 mg) and H 2 ase (50 pmol) in a degassed aqueous solution of EDTA (0.1 M, pH 6, 3 mL) in a borosilicate glass vessel of total volume 7.74 mL. The photoreactor was purged with N 2 containing 2% CH 4 as an internal gas chromatography (GC) standard and tightly sealed. The vials were then irradiated (AM 1.5G, 100 mW cm -2 ) and the headspace gas was periodically monitored by GC. Figure 5 shows the solar H 2 production activity of H 2 ase using CD-NHMe 2 + and CD-CO 2 − as photosensitizers. Similar to their behavior with FccA, the CD surface has a significant impact on the photo-H 2 production (see discussion below). Negligible H 2 generation was observed with CD-CO 2 − , whereas CD-NHMe 2 + showed high photoactivity with a TOF H2ase of 3.9 ± 0.9 × 10 3 mol H 2 (mol H 2 ase) −1 h −1 and long-term H 2 production, reaching a TON H2ase of 52 ± 8 × 10 3 mol H 2 (mol H 2 ase) −1 after 48 h ( Figure 5, Table S2). The performance of this CD-NHMe 2 + -H 2 ase system is comparable to the previously reported system using Dmb [NiFeSe]-H 2 ase and graphitic carbon nitride as a photosensitizer where a TOF H2ase of 5.5 × 10 3 and TON H2ase of 50 × 10 3 after 48 h were reported. 32 Negligible H 2 production was observed in the dark or in the absence of CDs, H 2 ase or EDTA (Table S1, Figures 5 and S8). Photo-H 2 production activity ceased when the sample was irradiated at 25 °C after heating to 100 °C for 15 min, i.e. conditions which denature the enzyme ( Figure S9). These control experiments together indicate that H 2 production indeed arises from H 2 ase following photoexcited electron transfer from CD-NHMe 2 + . Addition of free DMEN to the CD-CO 2 − -H 2 ase system does not give rise to H 2 production activity ( Figure S9), demonstrating that the surface-bound amine is required for the high activity in the system (see below).
The light intensity was varied to gain further insight into the performance limiting factors of the CD-NHMe 2 + -H 2 ase system. Cutting the light intensity by 50% and 80% using neutral density filters resulted in a linear decrease in the TOF H2ase indicating that the system is limited by light intensity ( Figure S10, Table S3). The external quantum efficiency (EQE) was measured by irradiation of the samples using a monochromatic light 6 source at an excitation wavelength of 365 nm with an intensity of 1.18 mW cm −2 . The resulting EQE of 0.36 ± 0.02 % is an order of magnitude higher than a carbon nitride-H 2 ase system (0.07 %; λ = 360 nm) 32 and is comparable to a carbon nitride-sensitized TiO 2 system under visible light irradiation (0.51 %; λ = 400 nm). 31 Other visible light responsive, precious and toxic metalfree systems cannot compete with the CD-NHMe 2 + -H 2 ase system in terms of stability. 26

CD-Enzyme Interface.
Previous studies with immobilized H 2 ases have highlighted key features to achieve efficient non-mediated interfacial electron transfer. 17,48,70 Important parameters are the distance and orientation of the biocatalyst to the photosensitizer surface. 70 For example, strong electrostatic interactions were utilized to achieve high electron transfer efficiency between positively charged CdTe particles and the negatively charged carboxylate residues at the distal FeS cluster of a [NiFe]-H 2 ase from Thiocapsa roseopersicina. 34 Similarly, a positively charged [FeFe]-H 2 ase from Clostridium acetobutylicum has been integrated with negatively charged CdS and CdTe particles. 33,35 In order to establish whether the markedly different activity between the anionic and cationic CDs here can be explained by different electrostatic interactions with the enzymes, we conducted photocatalytic experiments in the presence of a soluble redox mediator, methyl viologen (MV 2+ ; E 0 ' = -0.45 V vs SHE). 81 A comparison of product formation in the presence and absence of MV 2+ provides an indication of the efficiency of the direct electron transfer between the CD and enzyme and, in turn, is a measure of the interfacial interaction between them.
Product formation with both FccA and H 2 ase is greatly enhanced for both types of CD in the presence of excess MV 2+ (Table 1, Figures S11 and S12). For the H 2 ase system, addition of MV 2+ enhances the CD-NHMe 2 + TOF H2ase by a factor of 6 ( Table 1). While this indicates that the CD-H 2 ase interaction is not fully optimized, the interaction compares well with previously reported benchmark systems. A CN x -H 2 ase 32 system displayed a 22 fold enhancement on addition of MV 2+ and CN x /TiO 2 -H 2 ase showed a 5 fold enhancement. 31 CD-NHMe 2 + also exhibits a good interfacial interaction with FccA, with only a factor of 2 enhancement in substrate turnover in the presence of MV 2+ . Control experiments irradiating CDs and MV 2+ in the absence of H 2 ase produced negligible H 2 .
The photocatalytic experiments with MV 2+ also highlight the stability of the CDs for sustaining enzymedriven catalysis for up to 3 days, which is consistent with the previously reported photostability of CDs with synthetic catalysts. 56 In the presence of MV 2+ solar H 2 production was sustained from the CD-H 2 ase systems for over 72 h, reaching a TON H2ase of 4.9 x 10 5 for CD-CO 2 − and 7.2 x 10 5 for CD-NHMe 2 + , respectively (Figure S11, Table S1). This long-term activity indicates that the loss of photocatalytic activity over time observed the absence of MV 2+ (Figures 2 and 4) is due to degradation of the enzymes rather than the CD, consistent with previous photocatalytic studies utilizing H 2 ase and FccA. 32,44 An important observation is that in the presence of MV 2+ , the enzyme-based TOFs for CD-CO 2 − and CD-NHMe 2 + are almost identical (Table 1, Figures S11 and S12). This implies that both enzymes are similarly active in both CD solutions under illumination and that any electrostatic interactions with the enzymes do not affect the intrinsic enzyme activity. 57 It also demonstrates that the driving force and kinetics for MV 2+ reduction and EDTA oxidation are comparable for both types of CDs. The striking difference in performance between the two CDs in these systems in the absence of MV 2+ can thus be attributed to the difference in CD surface charge and different interfacial interaction with the enzyme. 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58 59 60    Figure 6 shows the ribbon representation based on Xray crystal coordinates of FccA 75 and Dmb [NiFeSe]-H 2 ase 82 with the surface-exposed positively and negatively charged amino acid residues highlighted. The surface-exposed electron entry sites (i.e. a Fe-heme subunit for FccA and [Fe 4 S 4 ]-cluster for H 2 ase) are surrounded by negatively charged aspartate and glutamate residues in both enzymes, as well as carboxylate groups on the surface-exposed hemes in FccA. Furthermore, the I p of Dmb [NiFeSe]-H 2 ase has been reported to be at pH 5.4 48 and the enzyme is therefore predominantly negatively charged under the experimental conditions employed in this study. The high photocatalytic performance with CD-NHMe 2 + is thus consistent with a favorable electrostatic interaction with the enzymes and in particular with the concentrated negatively charged residues at the electron entry sites, allowing for efficient mediator-free electron transfer. Similarly, a poor interaction with the enzymes can be expected with negatively charged CD-CO 2 − consistent with its poor photocatalytic performance in the absence of MV 2+ .

CONCLUSIONS
CDs are established as efficient photosensitizers for light-driven catalysis with redox enzymes, combining the benefits of molecular dyes, such as good solubility in water, with the high stability of semiconductor nanoparticles. Photocatalytic H 2 generation with H 2 ase and lightdriven C=C bond hydrogenation with FccA have been demonstrated by taking advantage of the tunable surface chemistry of the CDs. Positively charged ammoniumterminated CD-NHMe 2 + display excellent interfacial interactions and efficient direct electron transfer to FccA and H 2 ase, whereas negatively charged CDs show little or no activity. Direct photoexcited electron transfer to FccA was observed spectroscopically, as well as solardriven fumarate to succinate conversion over 24 h. CD-NHMe 2 + also displayed long-term solar H 2 production with H 2 ase and compares well with previously reported systems.
Ample opportunities for enhancement of the overall performance are available, for example, through optimization of the CD properties (e.g. via graphitization or heteroatom doping) 83,84 or by covalent immobilization of the enzyme. 27 The tunable surface chemistry of CDs along with their properties of aqueous solubility, biocompatibility and high stability render CDs as versatile photosensitizers for bio-hybrid photocatalytic schemes and will also allow their use for fundamental studies of electron transfer in CD-biomolecule assemblies. 85,86

EXPERIMENTAL SECTION
Materials. Reagents used throughout this work were obtained from commercial suppliers and used as received unless otherwise stated. Laboratory grade reagents were used in all synthetic procedures and chemicals for the analytical part of the work were of the highest available purity. Millipore water (18.2 MΩ cm @ 25 °C) was used throughout this work in both synthetic and analytical procedures. Buffer solutions were made using analytical grade reagents and titrated to the desired 8 pH, as determined by a pH electrode (Mettler Toledo; SevenEasy) using NaOH. D 2 O titrations were carried out with NaOD using an electrode calibrated in aqueous solution and a correction factor of 0.4 was applied to convert the pH meter reading to pD. 77 DMEN (Aldrich, ≥ 98%) was dried over activated molecular sieves (4 Å) prior to use. Enzymes. Dmb [NiFeSe]-H 2 ase was purified following a previously reported procedure. 87 The preparation has a specific activity of 2115 µmol H 2 min -1 (mg H 2 ase) -1 . 88 FccA from Shewanella oneidensis MR-1 was purified using a previously reported procedure. 89 The purified sample coupled oxidation of dithionite-reduced methyl viologen to fumarate reduction at a rate of 18.3 mM succinate min −1 (mg FccA) −1 .

Synthesis of CDs. CD-CO 2
− was synthesized and characterized as previously described. 55,71 Briefly, citric acid (100 g) was thermolyzed under air at 180 °C for 40 h producing carboxylic acid-capped amorphous CDs as an orange-brown high-viscosity liquid. The CDs were dissolved in water and freeze-dried to yield CD-CO 2 H as a low-density yellow-orange powder. To obtain sodium carboxylate-capped CDs (CD-CO 2 − ), the CD solution was neutralized to pH 7 using aqueous NaOH solution (5 M) prior to freeze-drying. Carboxylic-acid capped CDs (CD-CO 2 H) were used for the synthesis of CD-NMe 2 and sodium carboxylate-capped CDs (CD-CO 2 − ) were used for the photocatalysis experiments. For CD-CO 2 − Na + microanalysis found: C, 37.93%; H, 3.45%; N, 0.00%. CD-COCl was prepared by refluxing CD-CO 2 H (1.10 g) in neat thionyl chloride (50 mL) at 78 °C under a N 2 atmosphere for 1 h. Thionyl chloride was removed under reduced pressure and the resulting CD-COCl dots were dissolved in dry tetrahydrofuran (5 mL) and filtered. The solvent was evaporated under high vacuum to yield acylchloride capped CDs (CD-COCl, 1.0 g). Microanalysis found: C, 54.95%; H, 3.5%; Cl, 11.19%; N, 0.00%. CD-NMe 2 was prepared by stirring CD-COCl (800 mg) with neat and dry DMEN (25 mL) at room temperature in the dark for 3 h, whereupon DMEN was removed under high vacuum. The CD-NMe 2 dots were dissolved in acetone and filtered through a 0.2 µm syringe filter. Acetone was removed under reduced pressure and the product re-dispersed in methanol (3 × 5 mL) before drying for 12 h under high vacuum at 40 °C. The resulting CD-NMe 2 was then dissolved in water and freeze-dried to yield a light brown hydroscopic solid. , which was stored frozen in the glovebox. The total volume of the photoreactor was 7.74 mL and the solution volume 3 mL. The vessels were sealed with a septum, removed from the glovebox and purged with N 2 containing 2% CH 4 as a GC standard for at least 10 min prior to irradiation. H 2 detection was carried out by GC (Agilent 7890A) with a 5 Å molecular sieve column and a thermal conductivity detector. The GC oven was held at 45 °C and N 2 was used as the carrier gas (flow rate 3 mL min −1 ). Aliquots of headspace gas (50 µL) were removed from the reaction vessel using a gas-tight syringe (Hamilton; GASTIGHT®). H 2 produced was quantified by comparison to the CH 4 internal standard. The gas chromatograph was calibrated regularly to ensure reproducibility. Treatment of data. All analytical measurements were performed in triplicate. The data were treated as follows: for a sample of n observations x i , the unweighted mean 9 value x 0 and the standard deviation σ were calculated using the equations: where n H2 is the number of moles of photogenerated hydrogen, t irr is the irradiation time (s), A is irradiation cross-section (m 2 ), N A , h and c are Avogadro's constant (mol −1 ), the Planck constant (m 2 kg s −1 ) and the speed of light (m s −1 ), respectively.