Light-driven carbon−carbon bond formation via CO2 …

Loading...

Taking too long?

Reload document

| Open in new tab

Light-driven carbon−carbon bond formation via CO2
reduction catalyzed by complexes of CdS nanorods
and a 2-oxoacid oxidoreductase

Hayden Hambya,1, Bin Lib,2, Katherine E. Shinopoulosa,3, Helena R. Kellerc, Sean J. Elliottb, and Gordana Dukovica,4

aDepartment of Chemistry, University of Colorado Boulder, Boulder, CO 80309; bDepartment of Chemistry, Boston University, Boston, MA 02215;
and cMaterials Science and Engineering, University of Colorado Boulder, Boulder, CO 80303

Edited by Catherine J. Murphy, University of Illinois at Urbana–Champaign, Urbana, IL, and approved November 18, 2019 (received for review March 18, 2019)

Redox enzymes are capable of catalyzing a vast array of useful
reactions, but they require redox partners that donate or accept
electrons. Semiconductor nanocrystals provide a mechanism to
convert absorbed photon energy into redox equivalents for
enzyme catalysis. Here, we describe a system for photochemical
carbon−carbon bond formation to make 2-oxoglutarate by cou-
pling CO2 with a succinyl group. Photoexcited electrons from cadmium
sulfide nanorods (CdS NRs) transfer to 2-oxoglutarate:ferredoxin oxi-
doreductase from Magnetococcus marinus MC-1 (MmOGOR), which
catalyzes a carbon−carbon bond formation reaction. We thereby de-
couple MmOGOR from its native role in the reductive tricarboxylic
acid cycle and drive it directly with light. We examine the dependence
of 2-oxoglutarate formation on a variety of factors and, using ultra-
fast transient absorption spectroscopy, elucidate the critical role of
electron transfer (ET) from CdS NRs to MmOGOR. We find that the
efficiency of this ET depends strongly on whether the succinyl CoA
(SCoA) cosubstrate is bound at the MmOGOR active site. We hypoth-
esize that the conformational changes due to SCoA binding impact
the CdS NR−MmOGOR interaction in a manner that decreases ET
efficiency compared to the enzyme with no cosubstrate bound. Our
work reveals structural considerations for the nano−bio interfaces
involved in light-driven enzyme catalysis and points to the competing
factors of enzyme catalysis and ET efficiency that may arise when
complex enzyme reactions are driven by artificial light absorbers.

nanocrystal | redox enzyme | electron transfer | photochemistry

In learning how to catalyze complex chemical reactions,

chemists often turn to nature for inspiration (1–4). Redox
enzymes catalyze a vast variety of reactions in nature by moving
multiple electrons and substrates through catalytic cycles with
low energy loss and high selectivity (5, 6). As the complexity of
chemical transformations increases, so does the utility of chem-
ical pathways that have evolved in nature to target those specific
transformations (7–10). However,
in nature, redox enzymes
operate as parts of complex biochemical cycles in which the
electrons involved in redox transformations travel via indirect,
tortuous, and often diffusion-limited pathways (2, 10). Recently,
there has been a surge in research directed at activating these
biological catalysts with photosensitizers so that photon energy is
converted into redox equivalents (11, 12). Colloidal semi-
conductor nanocrystals, in particular, have been shown to be
excellent partners for redox enzymes, allowing reduction reac-
to H2, N2 to NH3, and CO2 to CO to be driven
tions such as H
directly by photoexcited electrons (13–17). Due to quantum
confinement, nanocrystals have easily tunable electronic struc-
ture, enabling control of their band gaps and redox potentials
(18–20). They are also strong light absorbers with surfaces that
can be chemically adapted to interact with enzymes without
inhibiting activity (12, 21).

While several nanocrystal−enzyme systems have demonstrated
promising light-driven catalytic activity, we do not yet have
enough understanding of how electrons move through these
systems to provide rational design principles. Insights so far

+

come primarily from complexes of cadmium sulfide (CdS)
nanocrystals with hydrogenases to drive the relatively simple and
fast 2-electron reduction of 2 protons to H2 (13, 22–26). How-
ever, because the utility of enzyme catalysis increases with in-
creasing chemical complexity that
to access by
artificial means (7–10), it is important to understand how en-
zymes that undergo complicated catalytic cycles can be driven by
photoexcited nanocrystals.

is difficult

In this work, we address such catalytic complexity by targeting
carbon−carbon (C−C) bond formation via CO2 reduction. The
enzyme we selected, a 2-oxoglutarate:ferredoxin oxidoreductase
(OGOR), is one component of the reductive tricarboxylic acid
cycle, an alternative system for autotrophic growth found in
many microorganisms (27–29). OGOR catalyzes the conversion
of CO2 and succinyl coenzyme A (SCoA) to 2-oxoglutarate and
CoA. This reaction occurs with a standard redox potential, E°′, on
the order of −500 mV vs. standard hydrogen electrode (SHE) at
pH 7 (8, 29, 30) and utilizes low potential electrons donated from
the redox partner ferredoxin (Fd) (30). Here we make use of
the recently reported OGOR from Magnetococcus marinus MC-1

Significance

Nature uses enzymes to catalyze a vast array of complex
chemical reactions. Enzymes that catalyze reduction reactions
require partners that provide the electrons ultimately used in
catalysis. In nature, the associated electron transfer sequences
can be complicated, energy-inefficient, and rate determining.
Here, we employ artificial electron donors, photoexcited
semiconductor nanocrystals, to provide the electrons for ca-
talysis. We couple these nanocrystals with an enzyme that
catalyzes the formation of carbon−carbon bonds via CO2 re-
duction. With this architecture, we demonstrate light-driven
formation of 2-oxoglutarate, storing some of the photon en-
ergy in the chemical product. We examine the role of electron
transfer from the nanocrystal to the enzyme in the light-driven
chemical conversion and learn that the enzyme modulates this
process during catalysis.

Author contributions: H.H., B.L., K.E.S., S.J.E., and G.D. designed research; H.H., K.E.S., and
H.R.K. performed research; B.L. and S.J.E. contributed new reagents/analytic tools; H.H.
and H.R.K. analyzed data; and H.H., B.L., S.J.E., and G.D. wrote the paper.

The authors declare no competing interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.

1Present address: Laboratory Services, Oregon Department of Agriculture, Portland, OR 97209.

2Present address: Department of Chemistry, University of Michigan, Ann Arbor, MI 48109.

3Present address: Dupont Electronics & Imaging, Marlborough, MA 01752.

4To whom correspondence may be addressed. Email: gordana.dukovic@colorado.edu.

This article contains supporting information online at https://www.pnas.org/lookup/suppl/
doi:10.1073/pnas.1903948116/-/DCSupplemental.

First published December 18, 2019.

www.pnas.org/cgi/doi/10.1073/pnas.1903948116

PNAS |

January 7, 2020 | vol. 117 | no. 1 | 135–140

Y
R
T
S
I
M
E
H
C

Y
R
T
S
I
M
E
H
C
O
I
B

l

1
2
0
2

,
4
1

y
u
J

n
o

t
s
e
u
g

y
b

d
e
d
a
o
n
w
o
D

l

(Mm), which uses a native Fd redox partner (MmFd1) to supply
electrons for the generation of 2-oxoglutarate (31).

We describe catalytic activity of MmOGOR driven by photo-
excited electrons from CdS nanorods (NRs) to photochemically
produce 2-oxoglutarate (Fig. 1). The NRs act as an Fd mimic
that generates reducing equivalents by light absorption (rather
than chemical reduction), injects them into MmOGOR, and
supports catalytic turnover. The use of CdS NRs to drive catal-
ysis allows us to 1) demonstrate direct photochemical activation
of a complex enzymatic catalytic cycle that involves large sub-
strates, significant conformational changes during catalysis, and
eventual formation of C−C bonds and 2) examine the relation-
ship between the electron transfer (ET) from photoexcited CdS
NRs to MmOGOR and the catalytic cycle of the enzyme. When
coupled with CdS NRs, MmOGOR catalytic activity resembles
that of MmFd1-driven MmOGOR, including a similar maximum
turnover frequency (TOFmax) for 2-oxoglutarate formation.
However, the quantum yield (QY) of product formation is only
∼1%, meaning that only 1% of photoexcited electrons generated
in CdS NRs are used for 2-oxoglutarate formation. We use
transient absorption (TA) spectroscopy of NRs to probe the
kinetics of ET from the NRs to MmOGOR and inform on the
origins of this inefficiency. We find that the efficiency of ET from
NRs to MmOGOR decreases dramatically when the cosubstrate
SCoA is bound to the enzyme active site. Dynamic light scat-
tering (DLS) measurements show that binding of SCoA to
MmOGOR weakens the binding of the enzyme to the CdS NRs.
We hypothesize that the conformation changes in MmOGOR in
the presence of SCoA modify the NR−enzyme interaction in a
way that decreases the fraction of photoexcited electrons trans-
ferred to the enzyme active site poised for catalysis. Our work
reveals previously unappreciated complexities of the nanocrystal-
driven enzyme redox catalysis, namely the impact of conforma-
tional changes during catalysis on the ability of the enzyme to
accept electrons from the nanocrystal. As such, our work illus-
trates how achieving synergistic and energy-efficient nanocrystal-

driven enzyme catalysis will require an understanding of the in-
terplay between enzyme catalysis and ET as well as its control.

Results and Discussion
Photochemical 2-Oxoglutarate Formation in a Mixture of CdS NRs and
MmOGOR. Fig. 1 shows an estimated energy level diagram for the
light-driven 2-oxoglutarate formation from the SCoA and CO2
starting materials using the combination of CdS NRs and
MmOGOR. Structural and optical characterization of CdS NRs
can be found in SI Appendix, Fig. S1. The energy used to form
2-oxoglutarate at the MmOGOR active site is supplied by a
continuous wave 405-nm laser that excites electrons above the
band gap in CdS NRs. Upon subpicosecond carrier cooling, the
electrons are at the lowest conduction band level (32, 33). Its
energy corresponds to a potential of ≤−0.7 V (all potentials are
reported vs. SHE at pH 7) (15, 34, 35), which is comparable to
the redox potential of [4Fe−4S] clusters in MmFd1, the natural
Fd partner for MmOGOR (−0.635 V and −0.485 V for the 2
[4Fe−4S] clusters) (31). The redox potential of the [4Fe−4S]
cluster in MmOGOR has been estimated as ∼−0.5 V (31). In
addition to their energy level structure, CdS NRs were chosen
−1 per
because of their high molar absorptivity (∼7 × 107 M
particle at 405 nm), which allows for a high excitation rate per
particle (∼50 to 1,000 photons absorbed per NR per second of
illumination, depending on the laser power). Photoexcited
electrons removed from the NRs are regenerated using a pH 6.8
buffer mixture containing radical scavenging buffers, including
one that has previously been used to scavenge holes from CdS
nanocrystals (14, 36). We supply CO2 by adding bicarbonate to
the buffered solution, which maintains the concentration of CO2
at ∼7.8 mM.

−1 cm

Fig. 1, Inset depicts, to scale, a complex of a CdS NR and one
MmOGOR homodimer using the recently reported crystal struc-
ture of MmOGOR (Protein Data Bank [PDB] ID code 6N2N)
(31). CdS NR surfaces are capped with 3-mercaptopropionate
(MPA), a ligand that has previously been used to interface CdS
NRs with enzymes such as hydrogenase and MoFe protein of ni-
trogenase (13, 15). At pH > 6, the carboxylate group of MPA is
deprotonated, lending an overall negative charge to the nano-
crystal surface. In this regard, the NRs resemble the redox part-
ners of MmOGOR, Fds, which are predominately negatively
charged near neutral pH (31, 37, 38). In complexes of MPA-
functionalized CdS NRs and a hydrogenase enzyme, the nega-
tive charge of the surface-capping ligands on the NRs enables a
biomimetic interaction wherein the NRs bind at the positively
charged pocket where Fd docks in nature (13). Based on this
precedent, in Fig. 1, Inset, we depict the NRs as binding near the
positively charged area on the MmOGOR surface.

Direct evidence of photochemical 2-oxoglutarate formation
using the CdS−MmOGOR system came from high-resolution
mass spectrometry (MS) measurements of the filtered reaction
mixture (SI Appendix, Fig. S2). Furthermore, when bicarbonate
isotopically labeled with 13C was used as a starting material,
we detected 13C-labeled 2-oxoglutarate (SI Appendix, Fig. S3).
While MS provides a signature of the photochemical product, it
is not well suited for quantitative rate measurements. For that,
we turned to in situ assay methods. To monitor light-driven
2-oxoglutarate formation in real time, we adapted an assay that has
been previously used for the measurement of oxoacid formation
with MmOGOR and similar enzymes (31, 39, 40). This assay
employs a second enzyme, glutamate dehydrogenase (GDH),
which consumes reduced nicotinamide adenine dinucleotide
(NADH) while catalyzing the amination of 2-oxoglutarate (Fig.
2) (39, 40). GDH turnover is ∼1,000 times faster than the
turnover of MmOGOR, so that the NADH consumption reports
on the 2-oxoglutarate formation rate. We monitor the depletion
of NADH in situ under illumination by its absorbance spectrum.
The reaction mixture for the assay exhibits a superposition of the

Fig. 1. Energy level diagram with approximate redox potentials for pho-
tochemical 2-oxoglutarate formation. VB = valence band; CB = conduction
band. Light absorption by CdS NRs is followed by ET to MmOGOR, where
SCoA and CO2 combine to make 2-oxoglutarate and CoA. Photoexcited holes
are scavenged by the buffer mixture, D. (Inset) Schematic to scale depiction
of a CdS NR bound to MmOGOR (PDB ID code 6N2N) showing the surface
residue charges as well as the [4Fe−4S] clusters and active sites in MmOGOR
and the surface-capping ligands on the NRs. Positively charged residues are
shown in blue, neutral are in white, and negative are in red. Atom colors:
C (gray), N (blue), O (red), P (orange), S (yellow), Fe (rust), Cd (off-white).

l

1
2
0
2

,
4
1

y
u
J

n
o

t
s
e
u
g

y
b

d
e
d
a
o
n
w
o
D

l

136 | www.pnas.org/cgi/doi/10.1073/pnas.1903948116

Hamby et al.

is calculated using e340 for NADH (6,400 M
enzyme concentration (39).

−1 cm

−1) and the

Characterization of Photochemical 2-Oxoglutarate Formation by the
CdS NR−MmOGOR System. To compare the photochemical
2-oxoglutarate production by CdS−MmOGOR complexes
with the previously described natural Fd-driven activity of the
same enzyme (31), we investigated the dependence of the CdS−
MmOGOR reaction on reactant concentrations and excitation
frequency. Fig. 3A shows the dependence of the light-driven en-
zyme TOF on the concentration of SCoA. The concentrations of
CdS NRs and MmOGOR were 44 nM and 100 nM, respectively.
CO2 concentration was 7.8 mM, and excitation frequency was 890
photons absorbed per NR per second. At lower SCoA concen-
trations, TOF increases linearly and saturates above 0.1 mM. The
functional
the dependence resembles that seen in
Michaelis−Menten kinetics. For the purposes of comparison with
Fd-driven OGOR systems in the literature, we analyze our data
using the Michaelis−Menten treatment.

form of

TOFmax for the data in Fig. 3A is 21 ± 0.87 moles of product
per mole of MmOGOR per minute of illumination, which is
similar to the TOFmax obtained for Fd-driven MmOGOR (27.2
moles of product per mole of enzyme per minute) (31). The
Michaelis constant KM describes the concentration of SCoA at
which the TOF is at half its maximum and informs on the
strength of substrate binding to the enzyme. For our system, KM
for SCoA is 0.050 ± 0.0086 mM (Fig. 3A), which is similar to that
obtained for Fd-driven MmOGOR at KM = 0.032 mM (31).
Approximating TOFmax as a lower limit to kcat, the ratio of these
parameters, TOFmax/KM, defines the catalytic efficiency, which in
−1 for SCoA. This is within a factor of 2
our system is 7,000 s
of TOFmax/KM for SCoA in the natural Fd-driven system, which
−1 at 30 °C. Our experiments were
has a value of ∼14,000 s
carried out at room temperature (22 °C) for compatibility with
the optical excitation apparatus. The similarity between MmOGOR
activity driven by Fd and by NRs suggests that the enzyme func-
tion is not adversely impacted by NR binding and, more specifi-
cally, that binding of SCoA to the MmOGOR active site is not
significantly perturbed by the presence of the CdS NRs.

−1 M

−1 M

The kinetics of 2-oxoglutarate production also depend on CO2
concentration (Fig. 3B). The concentrations of CdS NRs and
MmOGOR were 44 nM and 100 nM, respectively. The concen-
tration of SCoA was held constant at 100 μM, and the excitation
frequency was 890 photons absorbed per NR per second. In this
experiment, the MmOGOR TOF reaches a maximum near 39 ±
3.7 mol of product per mole of MmOGOR per minute of illu-
mination, obtained using a fit to the Michaelis–Menten model.

Y
R
T
S
I
M
E
H
C

Y
R
T
S
I
M
E
H
C
O
I
B

Fig. 2. Reaction scheme and representative absorption spectra over a total
of 10 min of illumination during the in situ product detection assay, overlaid
with the NADH spectrum. (Inset) Absorbance at 340 nm (a340nm), the maxi-
mum of NADH absorbance, as a function of time during the assay.

CdS NR and NADH absorption spectra as shown in Fig. 2. Both
components absorb much more strongly than MmOGOR, which
does not show detectable signal in the absorption spectra at these
concentrations. In the dark, we observe a negligible absorbance
change. After light is turned on, we see a decrease in NADH
absorbance while the CdS NR absorption remains constant, as
seen in the lowest-energy absorption peak at 460 nm (Fig. 2).
To monitor 2-oxoglutarate formation, we measure the absor-
bance at 340 nm during concurrent illumination with 405-nm
light (Fig. 2, Inset). Control experiments with SCoA, GDH,
MmOGOR, and CdS NRs removed (SI Appendix, Table S1)
show that all 4 components are needed to observe the NADH
consumption rate of the complete system. These control exper-
iments account for a small degree of background NADH con-
sumption due to processes other than enzymatic 2-oxoglutarate
generation. The photochemical 2-oxoglutarate generation rate of
the complete system is measured from the slope of the absor-
bance change at 340 nm during illumination, with the slope of
NADH consumption in the dark and in the absence of SCoA
during illumination subtracted. The enzyme TOF, defined as
moles of product per mole of OGOR per minute of illumination,

Fig. 3. Dependence of MmOGOR TOF on species participating in light-driven 2-oxoglutarate formation using the CdS−MmOGOR system. Solid black lines are fits
to the Michaelis−Menten model. (A) MmOGOR TOF dependence on SCoA concentration is similar to that observed in the Fd-driven system. (B) The dependence on
concentration of CO2 suggests that CO2 binding is weaker than that of SCoA. (C) The dependence of MmOGOR TOF on NR excitation frequency shows both the
linear regime where the product formation is limited by the electron flux and the saturation regime where the enzyme limits the turnover.

Hamby et al.

PNAS |

January 7, 2020 | vol. 117 | no. 1 | 137

l

1
2
0
2

,
4
1

y
u
J

n
o

t
s
e
u
g

y
b

d
e
d
a
o
n
w
o
D

l

The measured KM is larger for CO2 (3.0 ± 1.0 mM) than for
SCoA, implying that CO2 does not bind as strongly. The weaker
binding of CO2 has also been observed in an OGOR analog,
pyruvate:Fd oxidoreductase from Moorella thermoacetica, during
pyruvate synthesis, suggesting a similarity of the CO2 binding sites
between the 2 enzymes (39). This similarity illustrates the overall
biochemical challenge of binding CO2, compared to a much
larger, more chemically diverse SCoA molecule (or acetyl-CoA in
the case of pyruvate synthesis) (39).

Fig. 3C shows the dependence of light-driven MmOGOR
TOF on the excitation frequency of the NRs. The concentrations
of CdS NRs and MmOGOR were 44 nM and 100 nM, re-
spectively, while SCoA and CO2 were at 100 μM and 7.8 mM,
respectively. At lower excitation frequencies, under 200 photons
absorbed per NR per second, the TOF increases linearly, sug-
gesting a regime where the TOF is limited by electron flux from
the NRs to OGOR (13). TOF saturates at higher excitation fre-
quencies, where the TOF is limited to 39 ± 3 moles of product per
mole of MmOGOR per minute of illumination. In this saturation
regime, the product formation is limited by the enzyme turnover.
The analogous experiment in the Fd-driven system is the de-
pendence of TOF on Fd concentration, which follows the same
Michaelis−Menten functional form, and the TOFmax is 30.6 moles
of product per mole of enzyme per minute (31). While the satu-
ration due to enzyme turnover at high excitation is expected, the
excitation frequency needed to reach that saturation is surprisingly
high. The QY of 2-oxoglutarate formation (QYOG), defined as
molecules of product formed per 2 photons absorbed (because it
takes 2 electrons to make a product molecule), is ∼1% in the linear
region. This result means that only 1% of photoexcited electrons
end up in a 2-oxoglutarate molecule, while the remaining 99%
decay by unproductive pathways. This observation is in contrast to
QYs of H2 formation in systems that combine CdS NRs with hy-
drogenase, where the QYs are on the order of tens of percent (13).

Characterization of ET from CdS NRs to MmOGOR. The upper limit
on the value of QYOG is defined by the efficiency of injection of
photoexcited electrons from a CdS NR to MmOGOR. Conse-
quently, to shed light on factors that control the value of QYOG,
we examine the kinetics of that step using TA spectroscopy. This
technique monitors the decay of photoexcited electrons in the
NRs on the timescales from 100 fs to ∼1 μs. The TA spectrum of
CdS NRs exhibits a strong bleach feature around 460 nm (Fig. 4,
Inset), which corresponds to photoexcited electrons at the con-
duction band edge (32). The bleach signal decays due to relaxation
processes such as electron−hole recombination, electron trapping,

and ET (23, 24, 32, 33). Fig. 4A compares the decay of the bleach
signal for CdS NRs and a 2.5:1 molar mixture of MmOGOR
(1.7 μM) and CdS NRs (670 nM). As has been observed with hy-
drogenase, bleach decay is faster in the presence of MmOGOR due
to ET (22–24). Notably, when SCoA (1.0 mM) is added to the
mixture of CdS NRs and MmOGOR (Fig. 4B), the change in the
CdS NR bleach decay is much less pronounced, indicating that ET
from CdS NRs to MmOGOR is less efficient in the presence of
cosubstrate SCoA than when it is absent.

Quantitative modeling of the TA data provides insights into
the source of this decreased efficiency. To describe the electron
decay kinetics in the mixture of CdS NRs and MmOGOR, we
follow a model previously developed for complexes of CdS NRs
with hydrogenase (23). The decay of the CdS NR bleach is
multiexponential and can be attributed to electron−hole re-
combination with the rate constant k0 and electron trapping on
the NR surface, which occurs with the rate constant ktr. The trap
density is described with a Poisson distribution with the average
number of traps 〈Ntr〉. The additional decay due to ET occurs
with a rate constant kET. A mixture of CdS NRs with MmOGOR
leads to a distribution of population of complexes (0, 1, 2, etc.,
enzymes per NR), which can also be described with a Poisson
distribution with the average number of enzymes that can accept
electrons per NR abbreviated as 〈NOGOR〉 (22–24). The val-
ues of these parameters for the TA experiments with and without
SCoA are shown in SI Appendix, Table S2. As expected, the pa-
rameters associated with CdS NR relaxation (k0, ktr, 〈Ntr〉) are
similar in both cases (and similar to previously reported values)
(23, 24), so we focus on parameters that determine the efficiency
of ET, namely 〈NOGOR〉 and kET. For the case of the mixture of
CdS NRs with MmOGOR without SCoA added (Fig. 4A),
−1. This value
〈NOGOR〉 is 1.4 ± 0.2 and kET is 2.1 ± 0.4 × 107 s
of kET is of the same order of magnitude as that reported for the
CdS NR−hydrogenase system with the same surface-capping li-
gands (23, 24). The value of kET is also comparable to k0 and the
rate of electron trapping. In other words, ET from a CdS NR to
MmOGOR is in direct kinetic competition with the other electron
relaxation processes in the NR. When SCoA is added, 〈NOGOR〉
decreases by a factor of 4, to 0.35 ± 0.1, indicating a factor of 4
decrease in the number of adsorbed enzymes accepting electrons
from CdS NRs. The value of kET is more difficult to compare
−1) in the
due to a large error in the fit value (2.6 ± 1.3 × 107 s
presence of SCoA, but it stays within the same order of mag-
nitude in the 2 cases. Overall, the TA data suggest that binding
of SCoA to MmOGOR results in a decreased number of en-
zymes capable of accepting electrons bound per NR.

Characterization of Binding between CdS NRs and MmOGOR. Inspired
by the TA data, we characterized the binding between MmOGOR
(480 nM) and CdS NRs (250 nM) in the absence and presence of
SCoA using DLS, a technique that measures particle hydrody-
namic diameter by tracking particle diffusion (Fig. 5 and SI Ap-
pendix, Table S3). As shown in Fig. 5A, DLS measurements report
an average hydrodynamic diameter of CdS NRs of 22.1 ± 6.5 nm,
which is comparable to the NR dimensions we measured from
transmission electron microscopy (TEM) images (25.0 ± 2.4 nm in
length and 4.3 ± 0.4 nm in diameter). Due to the nonisotropic
particle shape, the relationship between hydrodynamic diameter
measured by DLS and particle dimensions is not straightforward,
but DLS is qualitatively consistent with the TEM results. The
hydrodynamic diameter of MmOGOR is 10.0 ± 2.0 nm (Fig. 5A),
which is consistent with the longest distance of 12.3 nm de-
termined from the crystal structure (31). When MmOGOR and
the NRs are mixed at a ratio of 2 enzymes to 1 NR, the average
value of hydrodynamic diameter is 61.7 ± 27.0 nm (Fig. 5B). We
attribute this larger size to complexation between the CdS NRs
and MmOGOR. The large SD is likely due to a distribution of
populations of complexes (e.g., 1 enzyme per NR, 2 enzymes per

Fig. 4. Comparison of TA decay kinetics of the CdS NR bleach at 444 nm in
the absence and presence of MmOGOR when SCoA is (A) absent and (B)
included in the sample mixture. (Inset) TA spectrum of CdS NRs at 1-ns time
delay. The TA data indicate that ET efficiency decreases when SCoA is added.
ΔA = change in probe absorbance due to the pump; λ = wavelength.

l

1
2
0
2

,
4
1

y
u
J

n
o

t
s
e
u
g

y
b

d
e
d
a
o
n
w
o
D

l

138 | www.pnas.org/cgi/doi/10.1073/pnas.1903948116

Hamby et al.

in MmOGOR upon SCoA binding impacts the strength of the NR−
enzyme interactions and impacts the ET pathway.

MmOGOR, like many other thiamine pyrophosphate-dependent
enzymes, consists of 2 active sites which are known to be structurally
and functionally asymmetrical during catalysis (43). Based on the
KM value for SCoA (Fig. 3A), at the SCoA concentration of 1 mM,
on average, one of the 2 active sites has an SCoA molecule bound in
the TA experiment. Because the interaction with the SCoA-bound
subunit is weaker, the CdS NR is more likely to be bound to the
other subunit, such that ET sends the electron to the active site that
is not poised for catalysis. This asymmetry creates a competition for
the electron in which the active site with SCoA bound is less likely
to receive it. In other words, while the overall ET efficiency de-
creases when SCoA is added to the CdS−MmOGOR mixture,
that decrease is disproportionately borne by the side of the en-
zyme that is poised for catalysis. The overall result of conforma-
tional change upon SCoA binding is decreased efficiency of ET
from an NR to the MmOGOR active site with SCoA bound,
resulting in a relatively low QYOG.

Thanks to the broad tunability of properties of semiconductor
nanocrystals, it may be possible to modify their interaction with
MmOGOR to make it more conducive to ET. As a proof of
principle, in SI Appendix, Table S4, we show that MmOGOR can
be driven by CdS and CdSe quantum dots (QDs) with diameters
of 4.3 and 3.1 nm, respectively. These particles are closer in size
to Fd from Pseudomonas aeruginosa, a homolog of MmFd1,
which has the approximate dimensions of 4 × 3 × 3 nm (44).
Because these particles are ∼15 times smaller in volume than the
NRs used above, they absorb less light and therefore make less 2-
oxoglurarate. The relative QY of product formation, which
normalizes for photon absorption, is highest for CdSe QDs with
3.1 nm diameter (SI Appendix, Table S4). Both the particle
composition and size impact the driving force for ET. Particle
size also determines how it fits in the Fd-binding pocket of the
enzyme. Further work is needed to disentangle the effects of
particle composition, size, and shape to identify the optimal
configuration for driving MmOGOR catalysis with light. Fur-
thermore, it has been shown that the ligands capping nanocrystal
surfaces play a critical role in ET to enzymes by impacting factors
such as donor−acceptor distance (and therefore electronic cou-
pling) as well as effective particle size (24). Ligands can also
change the driving force for ET (35). We expect that nanocrystal
surface functionalization strategies can be developed to control the
ET process. Finally, strategies for covalently binding hydrogenases
to CdS surfaces of nanocrystals have been reported before (26, 45)
and may be adaptable for directing nanocrystal−MmOGOR
binding and the subsequent ET. Development of new electrostatic
and covalent binding strategies should take into account the en-
zyme domain movement during catalysis. Beyond the C−C bond
formation example reported here, the domain motions that appear
critical to timing of ET steps may be a more general phenomenon
in complex redox biocatalysts, requiring multiple substrates as well
as the control of proton and ET steps. Our work described here
highlights the complexities that arise when driving enzyme catalysis
of large molecule transformations with photoexcited nanocrystals
and provides an initial set of guiding principles for the design of
the next generation of nanocrystal−enzyme architectures.

Conclusions
We reported and examined the photochemical formation of
2-oxoglutarate from CO2 and SCoA in a mixture of CdS NRs
and MmOGOR. CdS NRs serve as light absorbers that transfer
electrons to MmOGOR, where catalysis occurs. We find that
interaction with NRs does not adversely impact the catalytic
activity of the enzyme. However, the QY of 2-oxoglutarate
production is relatively low, meaning that most photoexcited
electrons are wasted rather than converted to the product. Our work
reveals that the product formation is limited by conformational

Y
R
T
S
I
M
E
H
C

Y
R
T
S
I
M
E
H
C
O
I
B

Fig. 5. DLS measurements for characterization of CdS NR−MmOGOR
binding: (A) CdS NRs and MmOGOR; (B) mixture of CdS NRs and MmOGOR in
a 1:2 molar ratio and the same mixture with SCoA added. The increase in
average particle hydrodynamic diameter in the CdS−MmOGOR sample is
due to complexation between the NRs and MmOGOR. The addition of SCoA
leads to dissociation of NR−MmOGOR complexes.

NR, etc.). We caution that the differences in refractive indices
between the 2 dissimilar components, strong dependence of
scattering intensity on particle size, and the nonisotropic shape of
the NRs preclude us from making a quantitative interpretation of
the measured hydrodynamic diameter values. Rather, we observe
a clear qualitative increase in the average particle size when
CdS NRs are mixed with MmOGOR.

When SCoA is present in the CdS NR−MmOGOR mixture,
DLS data suggest that less MmOGOR is adsorbed on CdS NRs
(Fig. 5B and SI Appendix, Table S3). Fig. 5B shows how the
hydrodynamic particle diameters change when 50 μM SCoA is
added to a mixture of 250 nM CdS NRs and 480 nM MmOGOR.
While the CdS−MmOGOR sample shows an average diameter of
61.7 ± 27 nm, after SCoA addition, 2 populations are detected,
both at smaller average diameters (33.2 ± 5.8 and 15.4 ± 2.6 nm).
A shift to smaller diameters indicates increased dissociation of
CdS−MmOGOR complexes. In combination with the TA data,
the DLS data suggest that, while the presence of CdS NRs does
not markedly impact the binding of SCoA to the MmOGOR ac-
tive site (Fig. 3A), the binding of SCoA to the MmOGOR active
site strongly impacts the interaction of MmOGOR with CdS NRs.
The decrease in the number of bound enzymes accepting elec-
trons, seen in the TA measurements, is consistent with the weaker
CdS−MmOGOR interaction seen in the DLS experiments.

Structural Basis for the Substrate