Spectroscopic elucidation of energy transfer in hybrid inorganic–biological organisms for solar-to-chemical production

Significance

Solar-powered chemical production from CO₂ promises to alleviate petrochemical consumption. Hybrid systems of an inorganic semiconductor light harvester and a microbial catalyst offer a viable way forward. Whereas a number of such systems have been described, the semiconductor-to-bacterium electron transfer mechanism remains largely unknown, limiting rational approaches to improving their performance. In this work, we look at how a semiconductor nanoparticle-sensitized bacterium transforms CO₂ and sunlight into acetic acid, a known precursor for fuels, food, pharmaceuticals, and polymers. Using time-resolved spectroscopy and biochemical analysis, we conclude that multiple pathways facilitate electron and light energy transfer from semiconductor to bacterium. This foundational study enables future investigation, understanding, and improvement of complex biotic–abiotic hybrid systems.

Abstract

The rise of inorganic–biological hybrid organisms for solar-to-chemical production has spurred mechanistic investigations into the dynamics of the biotic–abiotic interface to drive the development of next-generation systems. The model system, Moorella thermoacetica–cadmium sulfide (CdS), combines an inorganic semiconductor nanoparticle light harvester with an acetogenic bacterium to drive the photosynthetic reduction of CO₂ to acetic acid with high efficiency. In this work, we report insights into this unique electrotrophic behavior and propose a charge-transfer mechanism from CdS to M. thermoacetica. Transient absorption (TA) spectroscopy revealed that photoexcited electron transfer rates increase with increasing hydrogenase (H₂ase) enzyme activity. On the same time scale as the TA spectroscopy, time-resolved infrared (TRIR) spectroscopy showed spectral changes in the 1,700–1,900-cm⁻¹ spectral region. The quantum efficiency of this system for photosynthetic acetic acid generation also increased with increasing H₂ase activity and shorter carrier lifetimes when averaged over the first 24 h of photosynthesis. However, within the initial 3 h of photosynthesis, the rate followed an opposite trend: The bacteria with the lowest H₂ase activity photosynthesized acetic acid the fastest. These results suggest a two-pathway mechanism: a high quantum efficiency charge-transfer pathway to H₂ase generating H₂ as a molecular intermediate that dominates at long time scales (24 h), and a direct energy-transducing enzymatic pathway responsible for acetic acid production at short time scales (3 h). This work represents a promising platform to utilize conventional spectroscopic methodology to extract insights from more complex biotic–abiotic hybrid systems.

The sluggish transduction of solar energy into chemical bonds through natural photosynthesis has strangled our efforts to harvest the full bounty of the sun’s energy (1). We have temporarily sidestepped this limitation by tapping into large reserves of carbonaceous energy to drive an exponential growth in manufacturing, agriculture, urbanization, and population. However, the growing scarcity of petrochemicals has called for a return to photosynthesis––rather a new form of photosynthesis capable of keeping pace with modern society (2, 3). As a sign of progress, inorganic semiconductor light harvesters now routinely surpass the efficiency of plants (4). In contrast, synthetic catalysts still struggle to replicate the complex C–C bond formation of biology (5, 6). Significant strides toward comprehensive solar-to-chemical production have been demonstrated through several inorganic–biological hybrid systems combining inorganic semiconductor light harvesters with microbial CO₂ reduction (7–9). Recently, we have reported the self-photosensitized hybrid bacterium, Moorella thermoacetica–cadmium sulfide (M. thermoacetica-CdS), which photosynthesizes acetic acid from CO₂ via bioprecipitated CdS nanoparticles (10).

Although e⁻ transfer from electrodes to bacteria has been demonstrated across several genera, the mechanism remains in contention (11). Spectroscopic investigations of bacterium-to-electrode anodic e⁻ transfer in electrogenic microbial fuel cells have implicated cytochrome-mediated mechanisms (12). However, analogous studies of semiconductor-to-bacterium cathodic e⁻ transfer in electrotrophic organisms have remained sparse. Electron transfer first to membrane-bound or extracellular H₂ase to generate molecular H₂ as an intermediate followed by uptake into the native acetogenic Wood–Ljungdahl pathway (WLP) has been speculated or inferred (13–15). Still, detailed spectroscopic characterization has remained elusive due to the difficulty of adapting previous techniques to solid electrode platforms. In contrast, our model system, M. thermoacetica-CdS, as a translucent colloidal suspension in which an optically addressable CdS nanoparticle generates photoelectrons, eliminates the need for opaque electrodes, thereby enabling existing transmittance-based spectroscopies in uncovering the molecular basis of this charge-transfer mechanism (Fig. 1A). Here, we present transient absorption (TA) and time-resolved infrared (TRIR) spectroscopies correlated with biochemical activity to propose a model of the dynamics of inorganic–biological charge and energy transfer.

Fig. 1.

Schematic of M. thermoacetica-CdS photosynthetic charge transfer. (A) Visible light excitation of optically addressable CdS photosensitizing nanoparticles enables photosynthetic acetic acid production from CO₂, as well as characterization by TA and TRIR spectroscopy. (B) Potential e⁻ pathways in M. thermoacetica-CdS.

Upon photoexcitation of CdS, cysteine (Cys) oxidation to cystine (CySS) and H⁺ quenches the valence band h⁺, while the conduction band e⁻ may transfer to membrane-bound proteins or soluble e⁻ acceptors. Genomic mining, enzymology, and thermodynamic comparison of protein redox potentials have proposed several viable e⁻ transfer pathways (Fig. 1B) (13, 14, 16, 17). Membrane-bound NiFe H₂ases may play a significant role as H₂ directly feeds into the WLP. Demonstrations of direct e⁻ transfer between metal chalcogenide nanoparticles and purified H₂ases in vitro lend credibility to this pathway’s existence in complex whole cells (18–20). Alternative pathways have implicated e⁻ transfer first to membrane-bound cytochromes, ferredoxin (Fd), flavoproteins (Fp), and menaquinones (MK) (13, 14). Although these pathways generate reducing equivalents, implicitly, they must also couple to the formation of a H⁺ gradient to facilitate ATP synthesis by ATPase. The generation of this proton motive force, either through the transmembrane Ech complex or simply through surface proximal Cys oxidation, may be crucial to electrotrophic behavior, as a related acetogen, Acetobacterium woodii, which instead uses a Na⁺ motive force, notably cannot engage in electrotrophy (17, 21).

Results and Discussion

Biochemical Characterization.

To investigate the possibility of H₂ase mediated e⁻ transfer, M. thermoacetica-CdS was incubated from 0 to 48 h on H₂ (H₂:CO₂, 80:20), henceforth referred to as H₂ incubated) or glucose (25 mM) to vary the expression and activity of H₂ase. Activity was assayed photometrically by standard benzyl viologen (BV) reduction in the presence of H₂ (Fig. 2A and Fig. S1) (22, 23).

Fig. 2.

Biochemical assays of M. thermoacetica-CdS. (A) H₂ase activity with varying incubation time under H₂. See Materials and Methods for details of quantification. (B) CO₂-to-acetic acid conversion rates averaged over the first 3 h of photosynthesis show a decreasing trend with increasing H₂ incubation time. For comparison, 24-h glucose grown cells had a measured rate of 0.47 ± 0.15 mM h⁻¹. (C) CO₂-to-acetic acid conversion rates averaged over the first 24 h of photosynthesis show an increasing trend with increasing H₂ incubation time. Error bars represent the SD obtained from triplicate experiments.

Fig. S1.

H₂ase activity of M. thermoacetica-CdS with varying [H₂]. H₂ase activity was determined by variation of H₂ partial pressure with a fixed [BV].

Consistent with previous characterizations of M. thermoacetica, H₂ oxidation–BV reduction activity increased under increasing H₂ incubation time, presumably through increased H₂ase expression. Comparison of M. thermoacetica-CdS incubated for 24 h under glucose and H₂ showed H₂ase activity of 6.56 ± 0.92 × 10⁻¹⁵ (atm s cell)⁻¹ and 1.85 ± 8.39 × 10⁻¹⁷⋅(atm s cell)⁻¹, respectively.

To correlate enzyme activity with photosynthetic performance, the same samples were subjected to simulated solar illumination (0.5% sun, AM1.5G) and analyzed for acetic acid production. During the initial 3 h of photosynthesis, the rate of CO₂ reduction anticorrelated with H₂ase activity, with the 0-h (previously grown on glucose only) incubated sample showing the highest activity (0.23 ± 0.02 mM h⁻¹). The 24-h glucose incubated sample produced acetic acid at 0.47 ± 0.15 mM h⁻¹, one order of magnitude faster than that of the 24-h H₂ incubated sample (0.052 ± 0.009 mM h⁻¹). However, at longer illumination times, H₂ incubated samples demonstrated the opposite trend. When averaged over 24 h of illumination, photosynthesis rates increased with increasing incubation time in H₂ from 0.010 ± 0.001 mM h⁻¹ (7.1 ± 4.5% quantum yield of photons above CdS bandgap) at 0-h H₂ incubation to 0.047 ± 0.005 mM h⁻¹ (32 ± 4% quantum yield) at 36-h H₂ incubation (Fig. 2C). We note that the rate of acetic acid generation is higher within the first 3 h than the 24-h average. The 3-h rates are also not stoichiometric (e.g., 0-h H₂ sample has an apparent quantum yield of 160 ± 10%), suggesting that despite extensive washing and removal of residual glucose and H₂, some reduced intermediate carries over from the preincubation period. The observed trend indicates that lower H₂ase activity cells more effectively transfer CdS e⁻ to terminally reduce these intermediates, perhaps through a later set of enzymes in the WLP. In contrast, the 24-h averaged data agree in quantum yield with previous reports showing no such residual intermediates (10). These contrasting results suggest two competing charge-transfer mechanisms: a non-H₂ase mediated pathway dominant at short time scales (<3 h) and a H₂ase mediated pathway dominant at long time scales (∼24 h).

TA Spectroscopy.

To delve deeper into the molecular basis of these two mechanisms, we turned to time-resolved spectroscopic techniques for a dynamic understanding of the activity trends. TA decay kinetics followed the rate of photogenerated e⁻ leaving CdS for various preparations of M. thermoacetica-CdS (Fig. 3 and Fig. S2A). A transient bleach from 440 to 490 nm matched typical spectra observed with CdS e⁻ acceptor systems and was not observed in CdS-free M. thermoacetica. The spectrum of chemically precipitated CdS alone decayed much slower than M. thermoacetica-CdS, indicating that rapid quenching may result from a proximal e⁻ acceptor. The H₂ incubated M. thermoacetica-CdS displayed even faster decay kinetics than the glucose analog, correlating well with the higher H₂ase activity. These observations point to likely either faster e⁻ transfer to an acceptor site or more e⁻ acceptors available in H₂ incubated bacteria (20). Fitting each of the TA data sets to a triexponential decay revealed three lifetimes: a fast component in the range of 2–10 ps (τ₁), a longer component in the range of 20–80 ps (τ₂), and an even longer component in the range of several hundred picoseconds (τ₃) (Table S1). A multiexponential decay unsurprisingly indicates several processes at play in the complex M. thermoacetica-CdS hybrids. Rapid picosecond decays were previously measured with colloidal CdS that featured molecular acceptors with fast e⁻ transfer behavior (24–26). As hot e⁻ relaxation in Cd-chalcogenide quantum dots occurs in the subpicosecond regime, this process does not likely contribute to the TA kinetics in the measured time scale (27). However, previous studies of FeFe H₂ase-CdS constructs reported TA lifetimes in the range of 100 ns, significantly longer than the data presented here (19, 20). The discrepancy may be attributed to the presence of surface ligands (not present in M. thermoacetica-CdS) which present a charge-transfer barrier, differences in CdS-H₂ase spatial proximity, solvent effects and reorganization energies, H-bonding networks, and the relatively impaired functionality of purified enzymes under in vitro vs. in vivo conditions (28–30). A molecular carrier may also be at play, accepting charge and subsequently transferring it to H₂ase, among other potential charge-transfer pathways.

Fig. 3.

TA spectroscopy of M. thermoacetica-CdS. (A) TA plots of CdS only (cell-free), 24-h glucose incubated, and 24-h H₂ incubated. (B) Trends of exponential τ₁ and τ₂ lifetimes with increasing incubation time under H₂. (C) Weighted averages of the normalized τ₁ and τ₂ lifetimes with increasing incubation time under H₂. Error bars represent the SE associated with the exponential fitting.

Fig. S2.

Supplementary TAS spectra of M. thermoacetica-CdS. (A) Expanded time scale of spectrum presented in Fig. 3A. (B) Comparison of H₂ incubated and CO inhibited spectra.

Table S1.

Triexponential fitting data for TA spectra

H2 incubation time, h	τ1, ps	τ2, ps	τ3, ps	A1	A2	A3
0	9.26 ± 1.41	60.33 ± 19.66	1246.90 ± 249.33	0.27 ± 0.04	0.41 ± 0.07	0.32 ± 0.07
3	8.32 ± 1.55	57.77 ± 15.75	701.67 ± 23.99	0.29 ± 0.04	0.37 ± 0.06	0.34 ± 0.09
6	6.44 ± 0.24	56.72 ± 14.79	757.67 ± 39.38	0.39 ± 0.01	0.37 ± 0.07	0.24 ± 0.04
12	5.02 ± 0.37	37.22 ± 3.32	355.70 ± 17.81	0.30 ± 0.04	0.43 ± 0.04	0.27 ± 0.05
24	2.73 ± 0.46	37.41 ± 5.17	377.10 ± 40.30	0.30 ± 0.02	0.50 ± 0.03	0.20 ± 0.04
48	3.03 ± 0.95	20.68 ± 13.85	635.07 ± 407.55	0.21 ± 0.07	0.49 ± 0.07	0.29 ± 0.10

TA spectra were fit to a triexponential function of the form: $f\left(t\right)=A_1{e}^{t/{\tau }_1}+A_2{e}^{t/{\tau }_2}+A_3{e}^{t/{\tau }_3}$. Values indicated represent the value and SE associated with the triexponential regression.

The differences in lifetimes between cell-free CdS, H₂ incubated, and glucose incubated cells suggests the importance of H₂ase expression in e⁻ transfer kinetics. To correlate with the above H₂ase activity (Fig. 2A), the TA lifetimes of the H₂ incubated time series were measured (Fig. 3 B and C). Both the fast τ₁, τ₂ and their weighted average showed decreasing lifetimes with increasing H₂ase activity, suggesting that the fast e⁻ transfer kinetics was due to an increase in H₂ase e⁻ acceptor sites or molecular carriers whose appearance is correlated with H₂ase expression. Inhibition of the H₂ase active site (H cluster) with CO did not significantly change the TA kinetics, similar to previous works, in which e⁻ transfer initially proceeds through the FeS cluster chain rather than directly to the NiFe active site (Fig. S2B) (20, 31, 32).

TRIR Spectroscopy.

Whereas kinetically efficient e⁻ transfer to a membrane-bound H₂ase may explain the increasing photosynthesis rates at long time scales (Fig. 2C), TRIR helped determine the basis of the decreasing rates seen at short time scales (Fig. 2B).

We observed changes in the 1,760–1,880-cm⁻¹ spectral window, roughly corresponding to the vibrational range of CO and CN double and triple bonds, among other IR active modes characteristic of amino acid residues (Fig. 4) (33–35). The peaks decayed on the same time scale as the TA signal (Figs. S3–S6), giving further evidence that the picosecond e⁻ transfer resulted from a molecular, rather than purely physical, process. No significant changes in the 1,900–2,100-cm⁻¹ spectral window associated with the catalytic cycle of the NiFe H₂ase H cluster were observed, indicating that a significant quantity of photogenerated e⁻ was not transferred to the active site within this timeframe (18, 36, 37). Whereas the complexity of the whole-cell M. thermoacetica-CdS system renders unambiguous peak assignment beyond the scope of these initial results, careful construction of controls yielded valuable insight into the nature of these vibrational changes.

Fig. 4.

TRIR spectra of M. thermoacetica-CdS. (A and C) TRIR of 24-h H₂ incubated M. thermoacetica-CdS showing bleaching of several peaks in the region of C, N, and O double and triple bonds. (B and D) TRIR of 24-h glucose incubated M. thermoacetica-CdS.

Fig. S3.

SE of TRIR spectra. SE associated with the 24-h H₂ incubated sample at 0.25 ps.

Fig. S6.

TRIR kinetic data (continued). Intensities vs. time for several features in the TRIR spectral traces for the glucose incubated CdS-M. thermoacetica. (A–F) The legend represents the frequency at which the kinetics trance was recorded from the TRIR spectra.

Fig. S4.

TA spectra. Typical series of TA spectra showing from 1.5 to 50 ps of a 6-h H₂ incubated sample.

Fig. S5.

TRIR kinetic data. Intensities vs. time for several features in the TRIR spectral traces for the 24-h H₂ incubated CdS-M. thermoacetica. (A–F) The legend represents the frequency at which the kinetics trance was recorded from the TRIR spectra.

H₂ and glucose incubated samples yielded different TRIR spectral responses. Whereas several bleach features in the range of 1,760–1,820 cm⁻¹ on the picosecond time regime appeared for the H₂ incubated sample (Fig. 4A and Fig. S3), in the same spectral window, glucose incubated samples showed long-lived peak growth (Fig. 4B). These features may indicate H₂ase mediated charge transfer, with the differential response seen in glucose potentially representing an alternate pathway. In the region of 1,810–1,880 cm⁻¹, signal bleaches of similar time scales appeared in both H₂ and glucose incubated samples, implicating a similar e⁻ transfer pathway in both systems (Fig. 4 C and D). However, differences in their kinetic evolution point toward different utilization of this shared mechanism. Whereas the pair of bands at 1,823 and 1,827 cm⁻¹ retained a roughly 1:1 ratio for the H₂ incubated sample, in the glucose incubated sample, the 1,823-cm⁻¹ feature did not grow in until 1–2 ps after the pump excitation, and decayed back to zero faster than the 1,827-cm⁻¹ feature. We conclude that these spectral responses point toward different predominant e⁻ transfer pathways in H₂ vs. glucose incubated samples. Further elucidation of the e⁻ transfer pathways can be accomplished through fluorescent labeling studies, 2D electron spectroscopy, and synthetic biology (38–40).

Proposed e⁻ Transfer Mechanism of M. thermoacetica-CdS.

Taken together, the biochemical, TAS, and TRIR data suggest two competing pathways for e⁻ transfer within M. thermoacetica-CdS (Fig. 5). Initially, low H₂ase expression likely favors e⁻ injection to a membrane-bound e⁻ acceptor (Fd, Fp, cytochrome, MK) that generates a proton motive force for ATP generation, as well as potentially directly reduces CH₂–THF for the final stages of the WLP, using up any accumulated intermediates. However, slower charge transfer leads to poorer quantum efficiency, likely due to e⁻–h⁺ recombination losses. Additionally, this pathway cannot generate high energy reducing equivalents [H₂, NAD(P)H, Fd] needed for the first reductive steps of the full WLP. As H₂ase expression increases, charge-transfer kinetics favors e⁻ transfer to a membrane-bound H₂ase to generate molecular H₂. The possibility also exists of electron transfer to a molecular acceptor, which subsequently transfers an electron to H₂ase. Whereas this H₂ase mediated pathway may display higher quantum efficiency, the higher photosynthetic rates only kick in once a significant concentration of extra(intra)cellular H₂ accumulates. This results in lower photosynthetic rates in the first 3 h with increasing H₂ase activity (e⁻ diverted from the non-H₂ase pathway), but eventually higher rates after 24 h. This H₂ then likely enters the normal WLP via oxidation by the HydABC complex generating both ATP and reducing equivalents (17). Interestingly, the presence of CdS on M. thermoacetica may also induce novel, yet highly functional electron transfer pathways, different from what occurs naturally.

Fig. 5.

Proposed dual pathway of charge and energy transfer in M. thermoacetica-CdS. (A) The proposed non–H₂ase-mediated pathway predominant in glucose incubated cells, transiently faster during the initial 3 h of photosynthesis. (B) Membrane-bound H₂ase mediated pathway dominant in H₂ incubated cells and photosynthetically faster at long time intervals.

Conclusions

In conclusion, the work here represents an initial experimental probe into the photosynthetic, electrotrophic behavior of M. thermoacetica-CdS. With the careful construction of controls and mild, biocompatible probing conditions, complex biological behaviors may be studied through conventional spectroscopic techniques. Increasing H₂ase activity correlated with more efficient long-term photosynthetic rates of acetic acid generation. Evidence provided by TA and TRIR likewise supported the existence of a charge-transfer pathway that correlated with results from biochemical characterization. Our proposed two-pathway mechanism bears further investigation to probe more acutely the molecular and enzymatic basis of this biotic–abiotic charge transfer. Such insights will ultimately lead to a deeper understanding of the burgeoning complex nexus of inorganic materials and biological systems, and provide a rational framework for the optimization and design of next-generation solar-to-chemical systems.

Materials and Methods

Preparation and Biochemical Characterization of M. thermoacetica-CdS.

M. thermoacetica-CdS was prepared and assayed as previously described with modifications noted below (10). After full growth, M. thermoacetica-CdS was centrifuged under N₂ and resuspended in 20% of the original volume of defined photosynthesis media (DPM) with 0.1 wt % Cys·HCl and 2.5 atm H₂:CO₂ (80:20) or 25 mM glucose.

At various time points, aliquots of M. thermoacetica-CdS were sampled and stored at 4 °C under N₂:CO₂ (80:20).

BV reduction assays were conducted in a modified procedure as previously described (22). In short, 0.1 mL of M. thermoacetica-CdS was added to 5 mL of 50 mM Pipes buffer (pH 7) with [BV dichloride] ranging from 0.5 to 8 mM under a 2.5-atm H₂:CO₂ head space in a Hungate tube (Chemglass, Inc.). For varying [H₂:CO₂], volumes of H₂:CO₂ were syringe injected into a N₂:CO₂ headspace with [BV] = 4 mM. Kinetics of BV reduction was monitored at 578 nm (Shimadzu UV3101PC UV-Vis-NIR Spectrophotometer with an integrating sphere) for the initial 30 s. Rates were cell normalized by OD₆₀₀ correlated with manual cell counting.

H₂ase activity was calculated assuming pseudo–first-order kinetics (constant [BV] or [H₂]). The activity is thus defined as mM BV formed (mM benzyl viologen)^-1⋅(atm H₂)^-1 s⁻¹ cell⁻¹.

For photosynthesis measurements, 1 mL of M. thermoacetica-CdS was diluted into 4 mL of DPM with 0.1 wt % Cys·HCl. The suspension was illuminated by a filtered 75-W xenon lamp (Newport Corp.; AM1.5G, 0.5% sun) with heating and stirring (55 °C, 150 rpm) under a 2.5-atm 80:20 N₂:CO₂ atmosphere. Time points were centrifuged to remove cells and nanoparticles and assayed by quantitative proton nuclear magnetic resonance spectroscopy.

TA Spectroscopy.

Broadband TA spectra were obtained using an Ultrafast Systems Helios TA system with a Coherent Libra amplified Ti:sapphire laser system and Coherent OPerA optical parametric amplifier (OPA) pump/probe source. Briefly, the samples were excited with ∼50-fs laser pulses generated by the OPA at a repetition rate of 1 kHz. TA spectra were obtained by time-delaying a broadband supercontinuum probe pulse that is overlapped in time and space with the femtosecond pump pulse. The supercontinuum is produced by focusing a small portion of the amplified laser fundamental into a sapphire plate. Multiwavelength TA spectra were recorded using dual spectrometers (signal and reference) equipped with fast Si array detectors. In all experiments, the fluence value was held constant at 0.6 μJ cm⁻² to rule out effects from exciton–exciton annihilation as a result of high-power excitation. TA data were fit to a multiexponential decay.

TRIR Spectroscopy.

TRIR spectroscopy was performed with a home-built setup described elsewhere (41). A 400-nm pump pulse was used and the spectral region of 1,700–2,800 cm⁻¹ was probed. The sample was circulated in a N₂ purged and sealed flow cell through 2 IR-transparent CaF₂ windows spaced 150 or 75 µm apart. To exclude the contributions of hot e⁻ in the conduction band to the IR spectrum at early time scales as a broad positive absorbance, TRIR data were baseline-normalized by using a flat, featureless area of the TRIR spectrum as the baseline.