Preparation and characterization of metal-substituted carotenoid cleavage oxygenases

Article abstract of DOI:10.1007/s00775-018-1586-0

Carotenoid cleavage oxygenases (CCO) are non-heme iron enzymes that catalyze oxidative cleavage of alkene bonds in carotenoid and stilbenoid substrates. Previously, we showed that the iron cofactor of CAO1, a resveratrol-cleaving member of this family, can be substituted with cobalt to yield a catalytically inert enzyme useful for trapping active site-bound stilbenoid substrates for structural characterization. Metal substitution may provide a general method for identifying the natural substrates for CCOs in addition to facilitating structural and biophysical characterization of CCO-carotenoid complexes under normal aerobic conditions. Here, we demonstrate the general applicability of cobalt substitution in a prototypical carotenoid cleaving CCO, apocarotenoid oxygenase (ACO) from Synechocystis. Among the non-native divalent metals investigated, cobalt was uniquely able to stably occupy the ACO metal binding site and inhibit catalysis. Analysis by X-ray crystallography and X-ray absorption spectroscopy demonstrate that the Co(II) forms of both ACO and CAO1 exhibit a close structural correspondence to the native Fe(II) enzyme forms. Hence, cobalt substitution is an effective strategy for generating catalytically inert but structurally intact forms of CCOs.

Full text of DOI:10.1007/s00775-018-1586-0

JBIC Journal of Biological Inorganic Chemistry
https://doi.org/10.1007/s00775-018-1586-0
ORIGINAL PAPER
Preparation and characterization of metal‑substituted carotenoid
cleavage oxygenases
Xuewu Sui1,6 · Erik R. Farquhar2,3 · Hannah E. Hill · Johannes von Lintig · Wuxian Shi · Philip D. Kiser1,4,5
1
1
2,3
Received: 7 March 2018 / Accepted: 20 June 2018
©
The Author(s) 2018
Abstract
Carotenoid cleavage oxygenases (CCO) are non-heme iron enzymes that catalyze oxidative cleavage of alkene bonds in
carotenoid and stilbenoid substrates. Previously, we showed that the iron cofactor of CAO1, a resveratrol-cleaving member
of this family, can be substituted with cobalt to yield a catalytically inert enzyme useful for trapping active site-bound stil-
benoid substrates for structural characterization. Metal substitution may provide a general method for identifying the natural
substrates for CCOs in addition to facilitating structural and biophysical characterization of CCO-carotenoid complexes under
normal aerobic conditions. Here, we demonstrate the general applicability of cobalt substitution in a prototypical carotenoid
cleaving CCO, apocarotenoid oxygenase (ACO) from Synechocystis. Among the non-native divalent metals investigated,
cobalt was uniquely able to stably occupy the ACO metal binding site and inhibit catalysis. Analysis by X-ray crystallography
and X-ray absorption spectroscopy demonstrate that the Co(II) forms of both ACO and CAO1 exhibit a close structural cor-
respondence to the native Fe(II) enzyme forms. Hence, cobalt substitution is an eꢀective strategy for generating catalytically
inert but structurally intact forms of CCOs.
Keywords Non-heme iron enzyme · Cobalt substitution · X-ray absorption spectroscopy · Crystal structure
Abbreviations
ACO
Synechocystis sp. PCC6803 apocarotenoid
oxygenase
Electronic supplementary material The online version of this
article (https://doi.org/10.1007/s00775-018-1586-0) contains
supplementary material, which is available to authorized users.
BCO
β-Carotene oxygenase
CAO1
CCO
Neurospora crassa carotenoid oxygenase 1
Carotenoid cleavage oxygenase
Extended X-ray absorption ꢁne structure
High-performance liquid chromatography
Inductively coupled plasma-optical emission
spectroscopy
EXAFS
HPLC
ICP-OES
*
Philip D. Kiser
pdk7@case.edu
1
Department of Pharmacology, School of Medicine, Case
Western Reserve University, 10900 Euclid Ave., Cleveland,
OH 44106, USA
LB
Lysogeny broth
NSLS
RMS
National Synchrotron Light Source
Root-mean-square
2
3
National Synchrotron Light Source-II, Brookhaven National
Laboratory, Upton, NY 11973, USA
SDS-PAGE Sodium dodecyl sulfate-polyacrylamide gel
electrophoresis
Center for Proteomics and Bioinformatics, Center
for Synchrotron Biosciences, School of Medicine, Case
Western Reserve University, 10900 Euclid Ave., Cleveland,
OH 44106-4988, USA
SSRL
TCEP
XANES
XAS
Stanford Synchrotron Radiation Lightsource
Tris-carboxyethyl phosphine
4
Cleveland Center for Membrane and Structural Biology, Case
Western Reserve University, 1819 E 101st Street, Cleveland,
OH 44106, USA
X-ray absorption near edge spectroscopy
X-ray absorption spectroscopy
X-ray diꢀraction
XRD
5
6
Research Service, Louis Stokes Cleveland VA Medical
Center, 10701 East Boulevard, Cleveland, OH 44106, USA
Present Address: Department of Genetics and Complex
Diseases, Harvard T.H. Chan School of Public Health,
Boston, MA 02115, USA
Vol.:(0
1
123456789)
3

JBIC Journal of Biological Inorganic Chemistry
Introduction
([15, 16], but see [17]) the CCO iron-oxy complex remains
poorly characterized from an experimental standpoint.
Likewise, until recently, limited information was availa-
ble on the mode of substrate binding to the CCO active site
and the proximity of the scissile double bond to the iron
cofactor and its O2 binding site. Carotenoids are hydropho-
bic compounds and thus inherently poorly water soluble.
The limited concentrations of the compounds achievable
in aqueous solution thwarted attempts to form ES com-
plexes in suꢂcient yield for crystallographic studies [11,
18]. These barriers have been overcome by two advances.
The ꢁrst is the use of substrate/transition state analogs
with improved water solubility as demonstrated for the
RPE65 subgroup of CCOs [19] and later through the use
of stilbene-cleaving CCOs, whose substrates are naturally
more water soluble compared to carotenoids, as structural
targets. In the latter case, co-crystallization of a native
iron enzyme in this subfamily (Novosphingobium aromati-
civorans carotenoid oxygenase 1, NOV1) with resveratrol
led to tenuously supported active site electron density
for this compound, likely due to a mixture of un-cleaved
substrate, reaction intermediates and cleavage products
all being present within the crystal [16, 17]. The second
advance was the generation of a cobalt-substituted stil-
bene-cleaving CCO from Neurospora crassa (CAO1) that
was shown to be catalytically inert and allowed the crystal-
lographic observation of well-resolved enzyme–substrate
complexes [20].
Carotenoid cleavage oxygenases are an ancient family of
enzymes that catalyze the oxidative scission of alkene
substrates [1]. These include carotenoids and apocarot-
enoids but also metabolically unrelated compounds such
as stilbenoids and other phenylpropanoids [2]. Mamma-
lian genomes encode two such enzymes with documented
oxidative cleavage activity [3]. The ꢁrst to be identiꢁed
is an enzyme known as β-carotene oxygenase 1 (BCO1),
which cleaves β-carotene symmetrically to form retinal.
The second, BCO2, cleaves at the 9–10 (9′–10′) position of
the polyene chain in a number of related non-cyclic and bi-
cyclic carotenoids such as lutein and zeaxanthin, respec-
tively. Substrate speciꢁcity is similarly well established in
the CCOs of higher plants which comprise two broad sub-
classes catalyzing the cleavage of all-trans or 9-cis-carote-
noids/apocarotenoids to form chromophores, volatiles and
signaling molecules including abscisic acid and strigol-
actone precursors [4–7]. Microbial CCOs also can cleave
various carotenoid and apocarotenoid substrates although,
in many cases, the physiological substrates remain poorly
deꢁned [8]. Similarly, the biological roles and biologically
relevant substrate speciꢁcity of stilbenoid-cleaving CCOs,
which are conꢁned to eubacteria and fungi, remain rela-
tively poorly understood despite their potential importance
in lignin degradation and pathogen defense mechanisms
against phytoalexins such as resveratrol [9, 10].
The oxidative cleavage of alkene bonds by CCOs is
accomplished through the use of a mononuclear non-
heme iron center [11]. The iron cofactor presumably medi-
Metal substitution has previously been used as a strategy
to study the structure and function of mononuclear non-
heme iron oxygenases [21]. Manganese and cobalt have
been used most frequently in such investigations owing to
similarities in their preferred coordination geometries, ionic
radii and stable ionization states. Mn(II) and Co(II) both
have elevated reduction potentials relative to Fe(II), which
disfavors the formation of oxidized metal species that are
critically involved in O2 activation [22]. Hence, Mn(II)- and
Co(II)-substituted versions of mononuclear non-heme iron
dioxygenase are typically catalytically inactive. Important
exceptions to this generalization are some extradiol dioxy-
genases and cupin-family dioxygenases which are active in
Mn(II) and Co(II) forms [23–25].
ates reductive activation of dioxygen (O ) to overcome
2
the kinetic barrier associated with spin inversion of the
O triplet ground state during its reaction with singlet
2
organic molecules [12]. The CCO iron coordination motif
is unique among non-heme iron proteins characterized to
date. It consists of an absolutely conserved set of four His
residues that are arranged in a pseudo-octahedral geom-
etry, three of which are engaged in hydrogen bonds with
outer sphere Glu residues that are also highly conserved
within the family [13]. The iron is thus coordinatively
unsaturated by protein ligands leaving two sites localized
in cis potentially available for ligand binding. One of these
sites cannot readily accommodate ligands in most of the
CCOs characterized to date due to the close proximity of a
Thr or Val residue methyl group. The other location, which
is in direct contact with the substrate binding site, binds
an aquo or hydroxo ligand in the resting state of these
The success of metal substitution strategies in the study of
non-heme iron dioxygenase structure and reaction chemistry
suggests that this approach might also be viable for the study
of CCO-carotenoid interactions which to date have resisted
atomic level structural characterization. In this study, we
examine the susceptibility of a prototypical carotenoid-
cleaving CCO, Synechocystis apocarotenoid oxygenase
(ACO), to metal-substitution and ascertain the eꢀects of
such replacement on the catalytic properties and structure of
CCOs. The results validate the use of Co-substituted CCOs
to infer biological properties and chemical reactivity.
enzymes and is presumably also the binding site for O
2
during catalysis. The mode of O binding to the iron cofac-
2
tor in CCOs has been studied computationally [14], but
despite modeled O molecules in CCO crystal structures
2
1
3

JBIC Journal of Biological Inorganic Chemistry
Materials and methods
Reagents
ꢃash-frozen and stored in liquid nitrogen or placed on ice
for immediate use.
ACO activity assays
Except as noted below, chemical reagents were purchased
from Sigma-Aldrich (St. Louis, MO) or USB Biochemi-
cals (Cleveland, OH) in the highest purity form availa-
ble. Water from a Milli-Q puriꢁcation system (resistivity
The enzymatic activity of metal-free, native and putative
metal-substituted forms of ACO was measured by HPLC as
previously described with minor changes [18]. Speciꢁcally,
2 μg of puriꢁed ACO were added to a 200 μL reaction buꢀer
consisting of 20 mM HEPES–NaOH, pH 7 and 0.05% (w/v)
Triton X-100 (Anatrace, Maumee, OH). Reactions were
initiated by addition of an ethanolic solution of all-trans-
8′-apocarotenal to a ꢁnal concentration of 100 μM. Reac-
tion mixtures were placed in a shaker-incubator operating
at 28 °C with 500 rpm shaking for 1 min and then quenched
by addition of 300 μL of methanol. The reaction products,
all-trans-retinal and 8′-hydroxy-15′-apocarotenal, as well
as the remaining substrate were extracted with 500 μL of
hexane and analyzed directly by high-performance liquid
chromatography as previously described [18]. The amount
of generated all-trans-retinal was quantiꢁed by comparison
against a standard curve that was generated with known
amounts of all-trans-retinal (Toronto Research Chemicals,
Toronto, Canada, >95% purity).
1
8.2 mΩ cm) (ED Millipore, Billerica, MA) was used to
prepare all reagents and buꢀered solutions.
Expression and purifcation oꢀ metal‑substituted
ACO
All glassware and containers were extensively rinsed with
Milli-Q water to minimize potential metal contamination.
Expression of native, iron-containing ACO from lysog-
eny broth (LB) media was performed according to a pub-
lished method [18]. For the production of non-native metal
ACO, a pET3a plasmid containing the coding sequence of
ACO (Diox1, GenBank BAA18428.1) from Synechocystis
sp. PCC6803 was transformed into the T7 express BL21
Escherichia coli strain (New England Biolabs, Ipswich,
MA). Two milliliter of LB media containing 100 μg/mL
ampicillin were inoculated with fresh bacterial colonies
from an LB agar plate containing 100 μg/mL ampicillin
and grown for ~ 5 h at 37 °C with shaking at 235 rpm.
Then the cells were collected by centrifugation at 3220g
for 15 min at 37 °C. The cell pellets were resuspended in
ACO activity was also measured spectrophotometrically
using a Lambda Bio spectrometer (Perkin Elmer, Waltham,
MA) or a Flexstation 3 plate reader (Molecular Devices, San
Jose, CA) as previously described with minor changes [26,
27]. Six microgram of puriꢁed ACO was added to 100 μL of
reaction buꢀer with or without test compounds. Following a
10 min incubation period at 28 °C, all-trans-8′-apocarotenal
was added to the enzyme at a ꢁnal concentration of 25 μM
in a reaction volume of 200 μL. Assays were also performed
with protein omitted to determine the rate of non-enzymatic
loss of apocarotenoid substrate.
1
0
mL of M9 minimal media (1X M9 salts, 2 mM MgSO ,
4
.4% w/v glucose, 0.8% v/v glycerol and 0.1 mM CaCl )
2
and transferred into 500 mL of the same media contain-
ing 100 μg/mL ampicillin. Cells were grown at 37 °C
with 235 rpm shaking to an OD
of ~ 0.5 at which
00 nm
6
time the temperature was lowered to 28 °C and an addi-
tional 100 μg/mL of ampicillin were added to the culture.
Protein expression was induced by addition of isopro-
pyl β-d-1-thiogalactopyranoside to a ꢁnal concentration
of 100 μM. At the time of induction, 15 mg of diꢀer-
ent metal salts (CoCl ·6H O, MnCl ·4H O, CuCl ·2H O,
Analysis oꢀ metal content
The transition metal content of each ACO sample was meas-
ured by inductively coupled plasma optical emission spec-
troscopy (ICP-OES) analyses at the University of Minnesota
Soil Research Analytical Laboratory. Samples for ICP-OES
analysis were prepared by digesting each puriꢁed, putatively
metal-substituted ACO sample in 1% v/v HNO3. The pro-
teins precipitated immediately upon addition of HNO3 and
the samples were incubated for 3 h at room temperature with
gentle rocking. Protein precipitates were then removed by
ꢁltration through a 0.22 μm membrane (Millipore) and the
aqueous solutions that contained dissociated metals were
analyzed by ICP-OES. The buffer solution was treated
identically to determine the background metal content. A
summary of the metal analyses for each sample is shown in
Table 1. To accurately quantify the protein concentration
2
2
2
2
2
2
(
NH ) Fe(SO ) ·6H O) were added into the 500 mL cul-
4 2 4 2 2
ture for intended production of metal-substituted ACOs.
One culture was grown without any metal supplementa-
tion for production of metal-free ACO (apo-ACO). Cells
were harvested by centrifugation after overnight growth,
resuspended in 20 mM HEPES–HCl, pH 7 buꢀer and
stored at − 80 °C. ACO puriꢁcation was carried out as
previously described [18]. Co-CAO1 was expressed in
minimal media containing CoCl and puriꢁed as previ-
2
ously described [20]. Puriꢁed protein samples were either
1
3

JBIC Journal of Biological Inorganic Chemistry
Table 1 ICP-OES of ACO
Expression media
Protein concentra-
tion (mg/mL)
Metal concentration (ppm)
Target
metal/
ACO
preparations
Fe
Co
Cu
Mn
LB
1.097
1.847
1.234
0.382
1.508
0.722
1.522
<0.01
<0.01
1.075
<0.01
<0.01
<0.01
0.043
0.021
0.037
0.284
0.018
0.033
<0.01
<0.01
<0.01
<0.01
0.018
1.349
0.704
0.802
0.635
0.012
–
M9+Fe
M9+Co
M9+Cu
M9+Mn
M9
1.338
<0.01
0.012
0.378
0.034
<0.01
in the samples and thus the protein–metal stoichiometry,
amino acid quantiꢁcation was performed on puriꢁed ACO
with 10% glycerol and ꢃash cooled in liquid nitrogen
before X-ray exposure. Co-CAO1 was crystallized and cry-
oprotected as previously described [20]. Diꢀraction data
were collected on beamline X29 at the National Synchro-
tron Light Source (Co-ACO crystals) or NECAT beam-
line 24-ID-E of the Advanced Photon Source (Co-CAO1
crystals). For Co-ACO crystals, data were collected at the
wavelength of peak ꢃux, as well as at wavelengths above
and below the cobalt K absorption edges to allow assess-
ment of Co active site occupancy. Data sets were processed
with XDS [28]. Structures were determined by direct
refinement using previously determined isomorphous
structural models (PDB accession codes: 4OU8 for Co-
ACO and 5U8Y for Co-CAO1). Initial models were then
subjected to multiple rounds of manual model rebuilding
and updating in Coot [29] followed by restrained reꢁne-
ment in Refmac [30]. Refmac input ꢁles were prepared
with the CCP4 interface [31]. The stereochemical qual-
ity of the models was assessed with the Molprobity [32]
and wwPDB validation server [33]. Anomalous log-like-
lihood gradient maps were computed using Phaser [34].
A summary of the X-ray data and reꢁnement statistics is
shown in Table 2. All structural ꢁgures were prepared with
PyMOL (Schrödinger, New York, NY).
(
Texas A&M University, Protein Chemistry Laboratory).
The molar extinction coeꢂcient of ACO was determined to
be 75,249 M− cm at 280 nm [27].
1
−1
Optical spectroscopy
Optical spectra of ACO samples were recorded at ambient
room temperature with a Lambda Bio spectrometer (Perkin
Elmer) using micro UV-transparent cuvettes (Perkin Elmer).
Samples of ACO generated in the absence of added iron or
cobalt were used to subtract background absorbance in the
5
00–600 nm wavelength range of the Co-ACO sample to
isolate the weak Co-associated d–d absorption bands. This
was accomplished by scaling the apo-ACO 417 nm absorp-
tion peak to match that of the 429 nm Co-ACO absorption
peak and then subtracting the resulting spectra. Spectra were
scaled and plotted using Excel (Microsoft, Redmond, WA)
and Sigmaplot (Systat, San Jose, CA).
Protein crystallization, structural determination
and analysis
Prior to initiation of crystallization trials, puriꢁed Co-
ACO samples were loaded onto a 25 mL Superdex 200
gel ꢁltration column (GE Healthcare, Little Chalfont, UK)
equilibrated with buꢀer containing 20 mM HEPES–NaOH,
pH 7.0, and 0.02% (w/v) Triton X-100. The protein eluted
in a single, symmetrical peak at ~ 13 mL. Protein-con-
taining fractions within this elution peak were collected
and concentrated to 8–10 mg/mL. Crystallization was
performed by the hanging-drop, vapor-diꢀusion method
by mixing 1 μL of puriꢁed ACO with 1 μL of reservoir
buꢀer containing 0.1 mM Bis–tris propane–HCl, pH 6.0,
Sample preparation ꢀor X‑ray absorption
spectroscopy
Purified Co-ACO (> 95% purity by SDS-PAGE) was
concentrated to a cobalt content of ~ 1.28 mM in 20 mM
HEPES–NaOH, pH 7.0 containing 20% glycerol. Puriꢁed
Co-CAO1 (>95% purity by SDS-PAGE) was concentrated
to a cobalt content of 0.75 mM in 16 mM HEPES pH
7, 160 mM NaCl and 20% glycerol. Samples were ꢃash
cooled in liquid nitrogen and stored on dry ice before use.
Samples were loaded into copper sample cells (NSLS)
or Delrin cuvettes (SSRL) wrapped in Kapton tape and
ꢃash frozen in liquid nitrogen immediately prior to X-ray
absorption spectroscopy (XAS) measurements.
1
8–22% (w/v) sodium polyacrylate 2100, and 0.2 M NaCl.
Crystallization trays were prepared at room temperature
and then placed in an 8 °C incubator. Rod-shaped crystals
appeared within 1–2 weeks and grew to ꢁnal dimensions
of ~ 100 × 100 × 500 μm. Mature crystals were cryopro-
tected by soaking in the reservoir solution supplemented
1
3

JBIC Journal of Biological Inorganic Chemistry
Table 2 X-ray diꢀraction data collection and structure reꢁnement statistics
Crystal
Co-ACO
Co-ACO post edgeb
Co-ACO pre edgeb
Co-CAO1
Data collection and processinga
X-ray source
Wavelength (Å)
NSLS X29
1.07500
P2 2 2
NSLS X29
1.60210
NSLS X29
1.61460
NECAT 24IDE
0.97946
Space group
P3 21
1
1
1
2
Unit cell parameters (Å)
Resolution (Å)
Unique reꢃections
Multiplicity
118.07, 124.53, 202.61
48.4–2.21 (2.34–2.21)
145,134 (23,929)
6.3 (6.0)
118.75, 124.82, 202.66
48.51–2.82 (2.99–2.82)
138,304 (21,680)
3.8 (3.8)
118.78, 124.93, 202.84
48.54–2.81 (2.98–2.81)
139,028 (21,919)
3.8 (3.7)
100.69, 448.38
50–2.2 (2.33–2.2)
133,339 (20,826)
3.4 (3.3)
c
Completeness (%)
97 (88)
15.3 (1.2)
98.3 (95.6)
10.5 (1.22)
98.3 (95.7)
8.7 (0.8)
98 (96.1)
10.4 (1.1)
⟨
I∕ꢀI⟩
RmergeI (%)
CC1/2 (%)
10.2 (181)
99.9 (79.2)
52.6
13 (107.4)
99.2 (48.9)
52.6
17.4 (157.6)
98.6 (33.6)
54.6
7.9 (108.9)
99.8 (41.9)
52.9
2
Wilson B factor (Å )
Refnementd
Resolution (Å)
No reꢃections
Rwork/Rfree (%)
Total atoms
Protein atoms
Active site metals
48.4–2.21
135,096
20.8/22.9
15,788
15,100
4
49.1–2.2
127,272
17.9/20.5
16,883
15924
4
Water
B-factor⟩ (Å )
681
59.9
955
53.2
2
〈
Protein
Active site metals
Water
60.2
37.5
54.0
53.3
44.2
50.4
RMS deviations
Bond lengths (Å)
Bond angles (º)
0.013
1.55
0.009
1.36
Ramachandran plot (%
98.2/0
98/0
e
favored/outliers)
Molprobity score (%)
PDB accession code
100
6BIG
100
6B86
a
Data were processed with XDS [28]
Friedel pairs unmerged
Numbers in parentheses are for the highest resolution shell of data
Reꢁnement was carried out using REFMAC [30]
b
c
d
e
Ramachandran analysis was carried out using Molprobity [32]
XAS data collection
maximal photoelectron wave vector value (k ) for cobalt
max
−
1
EXAFS to ~12.0 Å . At SSRL, the storage ring operated
at 3.0 GeV and 500 mA in top-oꢀ mode. A Si(220) double
crystal monochromator provided energy selection, with an
upstream Rh-coating collimating mirror rejecting higher
harmonics prior to the monochromator and a downstream
Rh-coated toroidal mirror focusing the monochromatic
radiation at the sample. Temperature control was provided
by cryostats of either the He Displex type (NSLS, 15–20 K
sample temperature) or Oxford liquid He design (SSRL,
10 K sample temperature). XAS data were acquired in
XAS data were obtained at the National Synchrotron Light
Source (NSLS) on beamline X3B (ACO) and the Stan-
ford Synchrotron Radiation Lightsource on beamline 9-3
(
CAO1). At NSLS, the storage ring operated at 2.8 GeV and
1
80–300 mA. A Si(111) double crystal monochromator with
sagittal focusing of the second crystal provided energy selec-
tion and horizontal focusing, with a downstream Ni-coated
cylindrically bent mirror rejecting higher harmonics and
providing vertically focusing. The Ni coating restricted the
1
3

JBIC Journal of Biological Inorganic Chemistry
ꢃuorescence mode using Canberra solid state germanium
detectors (NSLS: 31-element discrete; SSRL: 100 pixel
monolithic) with XIA digital data acquisition electronics.
Data were typically acquired in 10 eV steps in the pre-edge
region (1 s acquisition time), 0.3 eV steps along the edge
ΔE was allowed to ꢃoat at a single common value for all
0
shells. Histidine multiple scattering was evaluated using a
method where the imidazole moiety is represented by four
sets of grouped paths representing the dominant scattering
3
contributions [38]. The ꢁt was evaluated in k -weighted
(
2 s acquisition time), and 0.05k steps in the EXAFS up
R-space, and ꢁt quality was judged by the reported R-factor
2
2
to 14–15k (acquisition time increasing from 2–9 s in a k
weighted fashion). Samples were monitored for evidence
of radiation damage as indicated by red-shifts in either the
pre-edge or edge energies, and new spots were exposed as
needed. A cobalt metal foil placed between ion chambers
downstream of the sample was used for energy calibration,
with the ꢁrst inꢃection point of the edge set to 7709.0 eV.
and reduced χ .
Results
ACO is robustly expressed in minimal media
containing non‑native metals
XAS data analyses
We initiated the study by testing the ability of ACO to stably
bind a series of ꢁrst row transition metals that serve as cofac-
tors in other oxygenase enzymes, namely, Mn(II), Co(II)
and Cu(II). We introduced these metals as well as the native
Fe(II) cofactor through their addition to M9 minimal growth
media used for the bacterial expression cultures. We omitted
metal from one culture to test the ability to produce metal-
free apo enzyme. SDS-PAGE analysis of solubilized protein
following ammonium sulfate precipitation of the cell lysate
supernatant demonstrated that ACO produced with diꢀer-
ent treatments was expressed at levels comparable to that
obtained from Fe(II)-supplemented cultures, with the excep-
tion of the sample obtained from manganese-containing cul-
ture, whose higher ACO expression level was attributable
to the greater cell density obtained in this media (Fig. 1a).
Notably, samples produced by minimal media expression
XAS data were processed and averaged using Athena [35]
for NSLS X3B data, whereas EXAFSPAK [36] was used
for SSRL 9-3 data. Pre-edge peak ꢁtting analysis was car-
ried out with Fityk using a previously reported protocol [20]
over the energy range of 7702–7718 eV. Extended X-ray
absorption ꢁne structure (EXAFS) analysis was performed
with Artemis [35]. Theoretical phase and amplitude param-
eters were calculated for a cobalt-substituted model of the
Fe-ACO active site using FEFF6L [37]. From this, relevant
paths were incorporated into the ꢁtting model and evalu-
ated for the signiꢁcance of their contributions. For a given
shell in all simulations, the coordination number n was ꢁxed,
2
while r and σ were allowed to ꢃoat. The amplitude reduc-
2
tion factor S was ꢁxed at 0.9, while the edge shift parameter
0
Fig. 1 Expression and puriꢁcation of ACO cultured in minimal media
with or without added metals. a ACO samples obtained after ammo-
nium sulfate fractionation. Precipitated proteins were re-dissolved in
molecular mass as described previously [18]. b Chromatograms
(A
) recorded during puriꢁcation of the ACO samples shown in a
80 nm
2
by gel ꢁltration chromatography. The monomeric ACO used for sub-
sequent studies eluted at approximately 75 mL with some variability
between samples. The larger peak at ~45 mL corresponds to the void
volume of the column
2
0 mM HEPES–NaOH, pH 7.0, separated by SDS-PAGE and visu-
alized by Coomassie Blue staining. The primary band of each sam-
ple migrated at ~50 kDa, slightly lower than the actual 54 kDa ACO
1
3

JBIC Journal of Biological Inorganic Chemistry
had improved initial ACO protein purity following ammo-
nium sulfate fractionation relative to comparable samples
obtained from LB media [18]. Further puriꢁcation by gel-
due to the signiꢁcant chemical diꢀerences between carot-
enoids and stilbenoids, as well as electrostatic diꢀerences
at the metal center observed between CAO1 and ACO by
Mössbauer spectroscopy [20]. Moreover, other redox-
active metals such as copper may aꢀord a catalytically
active enzyme. To test these hypotheses, we measured the
ability of the various metal-bound forms of ACO to cleave
one of its established substrates, all-trans-8′-apocarotenal,
leading to the formation of all-trans-retinal [8] (Fig. 2a).
As expected, ACO expressed in the presence of Fe(II)
displayed activity towards this substrate (Figs. 2b, 3a).
By contrast, the activity of Mn(II) and metal-free ACO
were drastically lower in comparison, the residual activity
being attributable to contaminating iron in the prepara-
tions (Fig. 2b and Table 1). We also found that Co(II)- and
Cu(II)-ACO preparations were virtually devoid of carot-
enoid cleavage activity, despite the target metals being
present in the samples. To test the stability of metal bind-
ing, we also carried out assays in the presence of 10 μM
Fe(II) (Figs. 2c, 3a). We observed elevated activity for
Fe(II)-ACO and metal-free ACO under these conditions
consistent with the existence of vacant metal-binding sites
in those samples that can be occupied by the added Fe(II).
Notably, incubation with Fe(II) alone had a minimal eꢀect
on the apocarotenoid absorbance spectrum ruling out non-
enzymatic chemistry as the cause of the enhanced retinal
formation (Fig. 3a). The ability of apo-ACO activity to be
stimulated by externally added ferrous iron to the maxi-
mum level of Fe-ACO demonstrated that the apo enzyme
is structurally intact with a fully accessible metal-binding
site (Fig. 2c). Notably, the activity augmentation for the
Fe-ACO sample (~ 2-fold increase) was somewhat greater
than that expected from the apparent number of vacant
active sites available in the sample (~ 30% of total ACO,
Fig. 2d and Table 1). This ꢁnding could be explained by
the presence of active site Fe(III) in a portion of the ACO
sample that can be reduced or displaced by Fe(II) to give
an active enzyme. Indeed, we observed that Fe-ACO activ-
ity could be enhanced by addition of another reducing
agent, Tris-carboxyethyl phosphine (TCEP), as previously
described for other CCOs (Fig. 3b) [20, 39]. Interestingly,
among the three samples prepared with non-native diva-
lent metals, only the Co-ACO sample remained inactive
upon addition of ferrous iron, whereas manganese- and
copper-containing samples displayed activity augmenta-
tion comparable to apo-ACO (Fig. 2b, c). To further probe
the interaction of cobalt with ACO, we tested the eꢀects
of Fe(II) and Co(II) addition on the activity of Co-ACO
and Fe-ACO to gauge the susceptibility of the metals
loaded into the protein during expression to be displaced
(Fig. 3b). We found that prolonged incubation of Fe-ACO
with 100 μM Co(II) led to a roughly 30% decrease in cata-
lytic activity, whereas incubation of Co-ACO with 100 μM
ꢁltration chromatography revealed that all samples exhibited
similar elution proꢁles but with varying amounts of pro-
tein eluting in the peak corresponding to non-aggregated
ACO (Fig. 1b). These data indicate that ACO expression is
not markedly aꢀected by depletion of native Fe(II) or sup-
plementation with non-native metals although ꢁnal protein
yields were lower for ACO expressed in the presence of cop-
per or without added metal.
Non‑native metals are incorporated into ACO
during expression
To determine whether metals added to the E. coli cul-
tures were incorporated into ACO, we analyzed each puri-
ꢁ
ed protein sample by multi-element inductively coupled
plasma-optical emission spectroscopy (ICP-OES). We used
the experimentally determined ACO molar absorptivity at
2
80 nm to calculate the sample protein concentrations and
ensure an accurate metal:protein stoichiometry [27]. Protein
puriꢁed from LB media revealed a slight excess of iron in
the sample compared to the ACO concentration suggesting
either adventitious iron binding to ACO or the presence of
iron bound to contaminating proteins (Table 1). None of
the other examined metals were present in the sample in
signiꢁcant amounts, except for a small amount of contami-
nating copper. By contrast, ACO from M9 minimal media
supplemented with Fe(II) exhibited ~70% iron occupancy.
The slightly sub-stoichiometric iron occupancy in this
sample is consistent with studies on other Fe(II) depend-
ent dioxygenases expressed in Fe-supplemented minimal
media [25]. Likewise, ACO from Co(II)- or Cu(II)-contain-
ing cultures displayed levels of target metal incorporation of
0
.8 and 0.63, respectively (Table 1), suggesting that these
metals could occupy the ACO metal-binding center. Little
or no iron was found in either the Co or Cu preparations.
By contrast, the Mn(II) preparation displayed only minimal
manganese incorporation (1.2%), suggesting that Mn(II) has
properties disfavoring its binding to the CCO metal-binding
site. Notably, the sample produced in media without added
metal contained negligible levels of all metals examined,
demonstrating the feasibility of producing metal-free apo-
ACO in a soluble form.
Non‑native metals can occupy the ACO active site
but do not support catalytic activity
Previously, we showed that the stilbene-cleaving CCO,
CAO1, is catalytically inert when prepared in Co(II) sub-
stituted form [20]. However, we considered the possibility
that metal-substituted ACO may retain catalytic activity
1
3

JBIC Journal of Biological Inorganic Chemistry
Fig. 2 Enzyme activity analysis of apo-, native- and metal-substituted
ACO by HPLC. a Reaction scheme showing the retinal-forming
activity of ACO using the established substrate, all-trans-apo-8′-
carotenal. b and c show representative HPLC analyses of the retinal
product generated by ACO samples that were produced in the pres-
ence or absence of Fe(II) or non-native metals. The reactions were
performed without (b) and with (c) Fe(II) in the assay mixture to
ascertain metal binding site accessibility. The peak representing all-
trans-retinal product is denoted by an asterisk, while the peak corre-
sponding to all-trans-apo-8′-carotenal is denoted by a double dagger.
The results are quantiꢁed in d. Data are shown as mean± SD calcu-
lated from measurements carried out in triplicate
Fig. 3 Spectrophotometric assessment of metal and reductant addi-
tion on the activity of Fe- and Co-bound forms of ACO. a Raw UV/
Vis absorbance data recorded in the presence and absence of Fe-
ACO with or without added iron. Samples containing Fe-ACO show
a reduction in substrate absorbance at 425 nm with a corresponding
increase in product absorbance at 350 nm following the 10 min incu-
bation period, whereas iron alone caused negligible loss of substrate
absorbance. Samples were pre-incubated for 10 min at room tempera-
ture and spectra were recorded following a 10 min reaction period.
b Relative activity levels of ACO samples pre-incubated with metals
or the reductant, Tris-carboxyethyl phosphine (TCEP). Samples with
only metals added were pre-incubated at room temperature for 22 h
to allow time for potential metal exchange, whereas samples contain-
ing TCEP with or without added iron and those containing iron- and
TCEP-only were pre-incubated at 28 °C for 10 min prior to initiation
of the assay. Assays were performed in triplicate at 28 °C. The results
are displayed as mean± SD
Fe(II) for an equal period of time did not enhance catalytic
activity. Taken together with the ICP-OES analyses, these
data indicate that iron, cobalt and copper all can occupy
the ACO metal binding site and suggest that of the tested
non-native metals the enzyme binds cobalt with highest
aꢂnity. However, only Fe(II) was able to support ACO
catalytic activity.
1
3

JBIC Journal of Biological Inorganic Chemistry
Characterization oꢀ Co(II)‑ACO by optical
spectroscopy suggests a fve‑coordinate metal
complex
410–429 nm absorption peaks remain obscure although the
similarity in absorbance maxima suggest a related origin
that is modulated by the metal content of the samples. The
apo-ACO 417 nm and Co-ACO 429 nm absorption bands
possessed tails that extended into the 500–600 nm range
adding background to the bands of interest. We removed
this background through scaling the apo-ACO spectrum so
that its 417 nm absorbance level matched that of the Co-
ACO 429 nm absorbance followed by subtraction from
the Co-ACO spectrum (Fig. 4). The resulting diꢀerence
spectrum was bimodal with a minor peak at 555 nm and a
major peak at 584 nm, the latter giving a molar absorptiv-
Whereas apo-ACO was pale yellow, purified Co-ACO
exhibited a bronze color that could be partially attributed
to light absorption by the bound transition metal. High spin
Co(II) d–d electronic transitions, which occur in the visible
region of the electromagnetic spectrum, are spin-allowed
but Laporte-forbidden or partially forbidden depending upon
whether or not the metal ligand environment is centrosym-
metric [40]. Owing to these selection rules, the strength of
high spin Co(II) optical absorption bands, which are typi-
cally centered in 500–600 nm wavelength range, are cor-
related with the coordination number and symmetry of
the metal ion [23, 41]. We recorded electronic absorption
spectra on concentrated samples of Co- and apo-ACO to
evaluate the absorptivity of the Co(II) ion in the CCO His-
rich ligand environment. Besides the main protein-related
peak at ~280 nm, Co-ACO exhibited additional features at
−
1
−1
ity of 134 M cm on the basis of the cobalt concentra-
tion within the sample (Fig. 4, inset). The spectral shape
closely resembled that of octahedral high-spin Co(II) model
compounds, which undergo two electronic state transitions
elicited by ~ 500–600 nm photons, whereas the extinction
coeꢂcient value was in the middle of the range expected
for a ꢁve coordinate Co(II) center [23, 41]. These ꢁndings
suggest that one or possibly two additional ligands besides
the four protein-associated imidazoles occupy the Co(II)
coordination sphere. The absence of an absorption band in
the ~ 670–833 nm region of the spectrum, corresponding
4
29, 540 nm and 575 nm, whereas apo-ACO showed only
a 417 nm peak whose height was about three times smaller
than that of the Co-ACO 429 nm peak at equal protein con-
centrations. Fe-ACO also exhibited a similar but broader
absorption peak at 410 nm comparable in height to the Co-
ACO 429 nm peak at equal protein concentrations (Fig. 4).
The identities of the chromophores giving rise to these
4
4
to a A′ → E′ electronic transition, argues against a D3 -
2
h
symmetric site, whereas three transitions in the 500–600 nm
range are theoretically expected for a high spin Co(II) in a
C -symmetric environment [41]. For the latter case, it is
4
v
conceivable that two of these transitions are close in energy
for the CCO Co(II) complex and thus not resolved leading
to the observed bimodal absorption feature.
Crystal structure oꢀ Co‑ACO reveals a structurally
preserved active site compared to Fe‑ACO
The ICP-OES, catalytic activity and optical spectroscopy
data provided strong evidence for the binding of Co(II) at
the ACO active center. We conꢁrmed this by crystallizing
ACO expressed in the presence of Co(II) and determining its
structure through X-ray diꢀraction (Table 2). Crystallization
conditions were identical to those used for native Fe-ACO
and no extra Co(II) was supplied in the crystallization cock-
tail. In this putative Co-substituted ACO structure, a strong
F –F electron density peak was present at the metal-binding
o
c
site indicative of a bound metal within the catalytic center.
To verify its identity as cobalt, we collected anomalous dif-
fraction data above and below the cobalt K absorption edge
allowing the calculation of cobalt-speciꢁc diꢀerence elec-
tron density maps (Table 2). The metal-binding site con-
tained a strong peak in the post-edge, NCS-averaged, anoma-
lous log-likelihood gradient map, whereas the corresponding
peak in the pre-edge map was much weaker (Fig. 5a). The
diꢀerence in scattering strength could not be attributed to
error diꢀerences between the two data sets as the anomalous
Fig. 4 Optical spectroscopy of Co-ACO. Spectra for apo-ACO, Co-
ACO, and Fe-ACO are shown as green, orange and slate colored
lines, respectively. The spectra have been scaled to their 410–429 nm
absorption peaks to facilitate comparison of absorption features in the
5
00–600 nm region of interest. The inset shows the (Co-ACO–apo-
ACO) diꢀerence spectrum with the molar absorptivity calculated on
the basis of the cobalt concentration in the sample
1
3

JBIC Journal of Biological Inorganic Chemistry
Fig. 5 Crystal structures of cobalt-substituted CCOs. a Conꢁrma-
tion of active-site bound cobalt in Co-ACO crystals. NCS-averaged
anomalous log-likelihood gradient maps were calculated with data
collected above (red mesh) and below (blue mesh) the cobalt K
absorption edge. Both maps are contoured at 9 RMSD. Anoma-
lous scattering in the vicinity of the sulfur atoms of Met and Cys
residues (shown as sticks) demonstrates comparable quality of the
pre- and post-edge data sets. The much stronger density peak at the
metal binding site in the post-edge map demonstrates the presence of
cobalt. b Comparison of the active site structure of Co-ACO (orange
sticks, salmon and red spheres for Co and solvent, respectively) and
Fe-ACO (grey sticks, grey spheres circumscribed with orange and red
dashed lines for Fe and solvent, respectively). c Comparison of the
active site structure of Co-CAO1 (aqua sticks, salmon and red spheres
for Co and solvent, respectively) and Fe-CAO1 (grey sticks, grey
spheres circumscribed with orange and red dashed lines for Fe and
solvent, respectively)
scattering associated with sulfur atoms of Cys and Met was
comparable between the two maps (Fig. 5a). These ꢁndings
supported the modeling of cobalt within the ACO active site.
After reꢁnement, the B-factor of the cobalt atom was similar
to the surrounding ligand atoms in all four molecules of the
asymmetric unit, suggesting near complete occupancy of
cobalt in the active site of the crystallized protein. This ꢁnd-
ing contrasts with an ~80% cobalt occupancy in the Co(II)-
ACO sample as determined by ICP-OES analysis (Table 1),
which may indicate preferential incorporation of Co-loaded
protein molecules into the crystal.
coordination geometry and conformations of ꢁrst and second
sphere iron-coordinating residues were indistinguishable
between the two metal-bound forms (Fig. 5b). The metal
center of Co-ACO is ꢁve-coordinate with distorted C sym-
4v
metry in agreement with the preceding optical absorption
data. We did observe a slight contraction of the His–metal
and solvent–metal bond lengths in Co-ACO as compared
to Fe-ACO although these diꢀerences were not signiꢁcant
within the coordinate uncertainties (Table 3).
Cobalt substitution preserves active site structure
in a stilbene‑cleaving CCO
We next analyzed the structural consequences of cobalt
substitution into the ACO active site. The structure of
Co-ACO is nearly indistinguishable from that of Fe-ACO
based on visual inspection with an overall root-mean-square
We previously showed that Co substitution in the stilbene-
cleaving CAO1 results in undetectable catalytic activity [20]
as we have observed here for ACO. Although the previously
reported crystal structure of Co-CAO1 did not show nota-
ble structural diꢀerences compared to that of Fe-CAO1,
the resolution and data quality of the former structure was
substantially inferior compared to the latter [20]. To allow
a more rigorous comparison, we collected data on a Co-
CAO1 to a higher than previously achieved resolution, with
signiꢁcantly improved quality indicators (Table 2). Similar
to ACO, we found only minor structural diꢀerences in active
site residues between Co-CAO1 and Fe-CAO1 (Fig. 5c and
Fig. S1c and d). Notably, in each monomer of the asym-
metric unit, cobalt was consistently bound deeper within the
4-His metal binding pocket, in closer proximity to His197
by ~0.4 Å relative to iron. Additionally, we observed sub-
tle variability in the positioning of the coordinated solvent
between the two metal-bound forms of the enzyme. Like
(
RMS) diꢀerence between Cα atoms of 0.4 Å, comparable
to the RMS diꢀerences between monomers within each of
the asymmetric units (Fig. S1a and b). The most variable
regions included residues 205–211 and 229–234 which
comprise the top face of propeller blade II and were previ-
ously shown to exhibit ꢃexibility [18], and residues 119–123
which are located in an α-helical segment. Both regions are
involved in ACO membrane association and do not contrib-
ute to active site structure. With regard to the active site
tunnel, only Phe236 showed a consistent conformational
diꢀerence between the two structures wherein the Cβ–Cγ
bond was rotated by ~ 30° resulting in a slightly diꢀerent
orientation of the phenyl ring. Previously, we showed that
Phe236Ala mutant ACO is catalytically active, which argues
against a subtle rotation of the Phe236 side chain leading to
an inactive enzyme [27]. Focusing on the metal center, the
1
3

JBIC Journal of Biological Inorganic Chemistry
Table 3 ACO and CAO1 metal–His bond lengths derived from XAS and XRD studies
Co-ligand bond lengths (Å)
XRD (6BIG)a
Fe-ligand bond lengths (Å)
XRD (4OU9)d
XASc
2.10
XASd
2.15
ACO
His183
His238
His304
His484
Solvent
2.13± 0.03b
2.21± 0.03
2.17± 0.01
2.23± 0.03
1.96± 0.18
2.19± 0.02
2.25± 0.04
2.19± 0.04
2.23± 0.02
2.08± 0.06
1.95
2.01
XASe
c
e
XRD (6B86)
XAS
XRD (5U8X)
CAO1
His197
His248
His313
His510
Solvent
2.04± 0.03
2.11± 0.02
2.03± 0.03
2.13± 0.01
1.96± 0.11
2.09
2.43± 0.06
2.17± 0.06
2.34± 0.03
2.25± 0.05
2.45± 0.02
2.15
1.95
1: 1.99
2
: 2.51
a
PDB accession codes are shown in parentheses for X-ray diꢀraction (XRD) structures. Maximum likelihood coordinate uncertainty estimates
are as follows: Co-ACO (6BIG), 0.177 Å; Fe-ACO (4OU9), 0.1 Å; Co-CAO1 (6B86), 0.164 Å; Fe-CAO1 (5U8X), 0.137 Å
Values are the averages and standard deviations computed from the four molecules in the asymmetric unit of each structure
From Tables 4 and 5 of the present work
From Ref. [18]
b
c
d
e
From Ref. [20]
in ACO, we observed a shortening of the metal–His bond
lengths in cobalt-substituted CAO1 relative to the native iron
form, although the bond lengths for Fe-CAO1 could be arti-
information on the metal site symmetry and coordination
number, whereas the extended x-ray absorption ꢁne struc-
ture (EXAFS) provides accurate bond lengths for the metal
center as it exists in solution. X-ray absorption near-edge
spectroscopy (XANES) of both Co-ACO and Co-CAO1
suggested that both enzymes contain Co(II) centers hav-
ing similar geometric and electronic structures (Fig. 6a).
Both Co-substituted proteins exhibit a single 1s-to-3d pre-
edge transition at ca. 7709 eV that can be ꢁt to a single
Gaussian-type function with integrated peak areas of 11.6
for Co-CAO1 and ~13 for Co-ACO. These pre-edge areas
are on the high side of reported Co(II) pre-edge peak areas,
which typically range between 3–6 for six-coordinate sites
and >10 for lower symmetries [42–44]. Our pre-edge analy-
sis, therefore, supports a ꢁve-coordinate geometry for the
cobalt site in both CAO1 and ACO. Of note, the sharper
peak shape for Co-CAO1 likely reꢃects the higher resolu-
tion Si(220) monochromator crystal set used for collection
of this data set. Both Co-ACO and Co-CAO1 also exhibit
two distinct edge inꢃections (7718.5/7722.5 eV for Co-ACO
and 7719.0/7723.5 eV for Co-CAO1), paralleling the spec-
tral shapes seen for the native Fe(II)-containing forms [20,
45]. The blue-shift of the Co-CAO1 edge implied a some-
ꢁcially elongated owing to submaximal iron occupancy [20].
By contrast, the cobalt ions in Co-CAO1 have B-factors that
ε
are comparable to the coordinating His N atoms indicat-
ing full cobalt occupancy at the metal-binding site. Taken
together with the preceding structural data on ACO, cobalt
substitution appears to preserve active site structure in CCOs
without supporting catalytic activity.
XAS confrms structural similarity oꢀ Co‑ACO
and Co‑CAO1 versus Fe counterparts
but with slightly contracted metal–ligand bond
lengths
The X-ray diꢀraction data presented in the preceding sec-
tions suggested some subtle structural diꢀerences between
the Fe- and Co-forms of ACO and CAO1, but the limited
resolution of the data precluded a deꢁnitive assessment.
Additionally, the coordination number and symmetry of
the cobalt center in Co-ACO was left somewhat ambigu-
ous by the preceding optical absorption data. We addressed
both of these remaining issues by recording cobalt K-edge
XAS spectra on concentrated Co-ACO and Co-CAO1 sam-
ples. Analysis of pre-edge absorption features provides
what higher eꢀective nuclear charge (Z ) for this enzyme’s
eꢀ
Co(II) center compared to that of ACO, with a similar
trend also observed for the native Fe(II) preparations [20].
1
3

JBIC Journal of Biological Inorganic Chemistry
reꢃected in the geometric structure seen by X-ray diꢀraction
(
XRD) or XAS (vide infra) within the resolution limitations
of available data.
EXAFS analysis conꢁrmed the structural model for the
cobalt center derived by XRD (Fig. 6b, c and Table 3). The
ꢁ
rst coordination shells of both Co-ACO and Co-CAO1
can be adequately ꢁt with a disordered single shell of 5–6
N/O scatterers at ca. 2.06 Å (Tables 4, 5). However, in both
cases splitting the ꢁrst coordination shell into a 5-coordinate
shell consisting of 4 Co–N/O scatterers at ca. 2.10 Å and 1
Co–O/N scatterer at 1.95 Å elicits a substantial improve-
ment in ꢁt quality and reduction in the magnitude of the
Debye–Waller factors. We ascribe the longer scatterers to
protein-derived His residues, while the shorter Co–O bond
likely corresponds to a solvent-derived hydroxo ligand.
Evaluation of outer shell features via multiple scattering
methods conꢁrms the presence of four His-derived imida-
zole units bound to the cobalt center in each enzyme. Our
EXAFS results conꢁrm that the crystallographically derived
structures are retained in solution, as both Co-ACO and Co-
CAO1 show nearly superimposable distorted ꢁve-coordinate
cobalt sites with a modest ~ 0.05 Å reduction in average
M(II)–N and M(II)–O
bond lengths going from
solvent
His
Fe(II) to Co(II).
Discussion
The preceding results demonstrate the feasibility of substi-
tuting non-native metals in carotenoid-cleaving members of
the CCO family as shown here for a prototypical member,
Synechocystis ACO. We also describe the production of
essentially metal-free ACO in a folded, soluble form through
expression of the protein in un-supplemented M9 minimal
media. This work builds upon our prior study on the stil-
bene-cleaving CCO, CAO1, where it was shown that cobalt
could be stably incorporated into the metal-binding site of
the enzyme [20]. In the present study we ꢁnd that cobalt
is also eꢀectively taken up into the ACO active site and
coordinates tightly to the 4-His metal-binding motif within
the active site. Copper also appears capable of occupying
the ACO metal-binding site, although we found it could be
displaced by iron in contrast to cobalt. Notably, this behavior
contrasts with a more stable binding for Cu(II) compared
to Co(II) expected from the Irving–Williams series [46].
Interestingly, we were unable to achieve eꢀective manga-
nese substitution even with high levels of this metal added
to the bacterial culture. We cannot attribute this to reduced
cell viability as manganese supplementation was found to
boost growth rate and ꢁnal cell density. Manganese binding
is favored at six-coordinate, oxygen/His sites [47], which are
provided by facial triad enzymes [25, 48, 49]. By contrast,
the canonical 4-His metal-binding motif of CCOs admits
Fig. 6 XAS analysis of Co-ACO and Co-CAO1. a Normalized
XANES spectra of as-isolated Co(II)-CAO1 (solid line) and Co(II)-
ACO (dashed line). The inset shows an expansion of the pre-edge
3
region. Representative best ꢁts (bolded entries in Tables 4, 5) to k -
weighted EXAFS data of Co(II)-ACO (b), and Co(II)-CAO1 (c).
Experimental data are shown as dotted lines, while the best ꢁts are
shown as solid lines
A more electropositive environment at the metal-binding
site for CAO1 compared to ACO was also previously noted
from Mössbauer spectroscopy isomer shifts, which reꢃect
s orbital electron density at the nucleus [20]. However, we
note that this apparent electronic structure diꢀerence is not
1
3

JBIC Journal of Biological Inorganic Chemistry
Table 4 Selected EXAFS ꢁts
χ2
Fit
Co–N/O
Co–O
Co•••His
ΔE0
R
for Co(II)-ACO
n
r
σ2
n
r
σ2
n
r
σ2
1
2
3
4
5
6
7
8
4
5
6
4
5
4
4
4
2.06
2.06
2.06
2.08
2.07
2.08
2.08
2.10
5.5
7.2
8.7
3.4
5.4
4.5
3.5
4.6
−0.53
−1.32
−2.08
−2.89
−3.91
−4.55
−2.65
0.92
129.31
89.97
92.66
0.0585
0.0407
0.0419
0.0191
0.0275
0.0263
0.1852
0.0180
1
1
2
1
1
1.92
1.90
1.93
1.92
1.95
1.2
3.9
6.8
1.2
2.2
77.54
111.61
107.08
158.77
36.51
3
3.03
6.5
3
4
4
.16
.27
.30
20.0
11.6
6.6
9
4
2.10
4.3
1
1.95
2.0
4
3.04
8.3
16.4
6.5
4.9
0.61
29.79
0.0147
3
4
4
.19
.21
.30
Fitting range was k=2.0–12.0 Å− (resolution=0.16 Å) with back transform ranges of 1–2.2 Å for ꢁts 1–6
1
2
2
−3
and 1–4.0 Å for ꢁts 7–9. r is in units of Å; σ is in units of 10 Å; ΔE is in units of eV; R represents the
0
2
fractional misꢁt of the data, while χ is the χ ꢁtting metric normalized by the number of independent data
points in a given ꢁt
The best ꢁrst-shell single scattering model is italicized, while the best multiple scattering model, corre-
sponding to the best ꢁt curve in Fig. 6b, is given in bold italics
Table 5 Selected EXAFS ꢁts
χ2
Fit
Co–N/O
Co–O
Co•••His
ΔE0
R
for Co(II)-CAO1
n
r
σ2
n
r
σ2
n
r
σ2
1
2
3
4
5
6
7
4
5
6
5
4
4
4
2.07
2.07
2.06
2.07
2.08
2.08
2.10
5.4
6.9
8.3
5.2
3.4
3.7
3.7
3.58
2.74
1.98
0.36
1.27
1.11
3.64
84.52
54.93
49.53
39.27
33.37
82.60
20.94
0.0750
0.0487
0.0440
0.0216
0.0184
0.1985
0.0277
1
1
1
1
1.91
1.93
1.93
1.95
3.3
1.5
2.0
1.8
3
3.05
5.9
3
4
4
.19
.27
.37
19.4
12.2
9.0
8
4
2.09
3.8
1
1.95
1.9
4
3.06
7.6
3.63
17.00
0.0248
3
4
4
.22
.24
.32
15.7
4.8*
4.8*
Fitting range was k=2.0–13.5 Å− (resolution=0.14 Å) with back transform ranges of 1–2.15 Å for ꢁts
1
1
–5 and 1–4.15 Å for ꢁts 6–8. Parameters marked with * were reꢁned to a common value in the ꢁt. r is
2
−3
in units of Å; σ is in units of 10 Å; ΔE is in units of eV; R represents the fractional misꢁt of the data,
0
2
2
while χ is the χ ꢁtting metric normalized by the number of independent data points in a given ꢁt
The best ꢁrst-shell single scattering model is italicized, while the best multiple scattering model, corre-
sponding to the best ꢁt curve in Fig. 6c, is given in bold italics
only a single solvent-derived oxygen into the coordination
sphere, which could disfavor manganese incorporation.
Metal substitution strategies have been employed in
a number of other mononuclear non-heme iron proteins
as a means to investigate their catalytic and structural
properties. Cobalt-substitution in particular has been suc-
cessfully employed for proteins belonging to well-charac-
terized non-heme iron enzyme families. In the case of the
Rieske enzyme, dicamba dioxygenase [21], as well as TauD
[50], a member of the α-ketoglutarate/taurine-dependent
1
3

JBIC Journal of Biological Inorganic Chemistry
dioxygenase family, cobalt substitution results in catalyti-
cally inert enzymes. In both cases, the activation of dioxygen
involves formation of high-valent iron-oxy species [51]. The
inability of Co(II) to functionally substitute in these enzymes
can thus be rationalized from the inaccessibility of cobalt to
these higher oxidation states due to its elevated reduction
potential relative to iron. Likewise, Co(II) substitution in the
intradiol-cleaving protocatechuate-3,4-dioxygenase results
in an inactive enzyme [52] owing to the requirement of tri-
valent iron for the initial step of catalysis and the reduced
ability for cobalt to cycle between the +2 and +3 oxidation
states under normal biological conditions.
observed between the Co- and Fe-bound forms of both
ACO and CAO1, indicating that changes in their substrate
binding sites due to the presence of non-native metal are
not responsible for the absence of activity. We did observe
by both X-ray crystallography and XAS a slight but con-
sistent shortening in metal–ligand bond lengths (~0.05 Å)
in the Co-bound forms of both enzymes relative to their
Fe-bound counterparts (Table 2). The shorter metal–ligand
bond lengths in Co-CCOs could be at least partially rational-
ized by the smaller ionic radius of high-spin Co(II) relative
to high-spin Fe(II) [55]. Regardless, it seems unlikely that
this small diꢀerence in metal–ligand bond length on its own
could lead to such a profound change in catalytic activity.
However, the diꢀerence in bond length may reꢃect varying
strength of metal–ligand interactions. Such a diꢀerence is
suggested by our activity assay data where addition of Fe(II)
to Co-ACO had a minimal ability to boost catalytic activ-
ity, whereas incubation of Fe-ACO with Co(II) resulted in
a substantial suppression in catalytic activity indicative of
active site Fe(II) substitution by Co(II) (Fig. 3b).
By contrast, the cobalt-substituted forms of 2,3-HPCD
[
23], a vicinal oxygen chelate enzyme and the cupin-fold
ring-cleaving oxygenase, quercetin 2,3-dioxygenase [24,
3], both exhibit robust catalytic activity in relation to their
5
Fe-bound forms. Despite their phylogenic diversity, these
enzymes share a common catalytic theme: aromatic alkene
bond cleavage of a redox-active substrate via direct substrate
coordination of the metal center through one or two phenolic
oxygen atoms. Both enzymes are thought to use their metal
cofactor to channel electrons from the organic substrate to
bound dioxygen resulting in simultaneous activation of both
substrates with only a transient change in metal redox state
Co-substituted CCOs may be valuable tools for eluci-
dating biochemical properties and biological roles of these
enzymes. We previously showed that Co-substitution in
CAO1 allows for stable trapping of active site-bound sub-
strate for structural characterization. Co-substitution in
carotenoid-cleaving CCOs, as shown to be feasible in the
present study, is likely to facilitate three-dimensional struc-
ture determination of CCO–carotenoid Michaelis complexes,
which to date have evaded characterization. This goal will,
however, likely require other advances to overcome the
problem of carotenoid aqueous insolubility. As mentioned
in the Introduction, the biological substrates of some CCOs
remain to be elucidated [8, 10]. Co-substituted CCOs, being
catalytically inert but structurally intact, may be helpful for
identiꢁcation of high aꢂnity substrates through aꢂnity
chromatography coupled with mass spectrometry. Addition-
ally, Co-substituted CCOs may also be useful for carrying
out binding assays in the absence of confounding catalysis
to assess diꢀerences in substrate aꢂnity.
[
22, 25, 54], which explains the ability of cobalt to support
catalysis in these enzymes.
CCOs have features common to both Co-active and Co-
inactive dioxygenase enzymes. On one hand, like some
Rieske dioxygenases including dicamba dioxygenase, CCOs
substrates do not directly coordinate to the metal center
when they are bound in the active site. But like the cat-
echol and quercetin dioxygenases, CCO substrates are redox
active and possibly do not require a powerful oxidizing agent
(
e.g., high valent iron-oxy species) for cleavage of the tar-
get alkene bond. Thus, it is interesting that Co-substituted
forms of both carotenoid- and stilbenoid-cleaving CCOs are
catalytically inactive [20]. While our current understand-
ing of the exact iron-oxy species involved in CCO catalysis
is limited, these results suggest that the inactivity of Co-
CCOs may relate to the necessity for a relatively long-lived
iron(III)-superoxo or conceivably a higher valent iron in the
CCO catalytic cycle. This diꢀerence between CCOs and
the Co-active dioxygenases mentioned above could relate
to the absence of direct substrate binding to the iron center,
which excludes a direct pathway for electron ꢃow, as well
as eliminates possible redox tuning of the metal by organic
substrate coordination.
Acknowledgements We thank Dr. Krzysztof Palczewski and members
of the Kiser laboratory for valuable comments on the manuscript. This
research was supported in part by grants from the National Institutes of
Health (NIH) (EY009339 to P.D.K., EY020551 to J.v.L.), the Depart-
ment of Veterans Aꢀairs (IK2BX002683 to P.D.K.) and the Burroughs
Wellcome Fund (Award #1015187 to P.D.K.). This work is based in
part upon research conducted at the APS Northeastern Collaborative
Access Team beamlines supported by grants GM103403, RR029205
and DE-AC02-06CH11357. Data for this study were measured at
beamlines X3B and X29 of the National Synchrotron Light Source.
Financial support comes principally from the Oꢂces of Biological
and Environmental Research and of Basic Energy Sciences of the U.S.
Department of Energy, and from the National Center for Research
Resources (P41RR012408), the National Institute of General Medical
Sciences (P41GM103473) of the National Institutes of Health and the
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
The absence of catalytic activity in Co-CCOs could
also result from a perturbation in the active site structure
upon replacement of the native Fe(II) cofactor. The crys-
tallographic and XAS data that we present have allowed
us to rigorously evaluate this potential explanation. Little
or no consistent diꢀerences in active site residues were
1
3

JBIC Journal of Biological Inorganic Chemistry
(
P30-EB-09998), and the U.S. Department of Energy Oꢂce of Sci-
24. Schaab MR, Barney BM, Francisco WA (2006) Biochemistry
45:1009–1016
ence under Contract DE-AC02-98CH10886 to Brookhaven National
Laboratory. Use of SSRL beamline 9-3 was made possible by the U.S.
Department of Energy Oꢂce of Science under Contract DE-AC02-
25. Emerson JP, Kovaleva EG, Farquhar ER, Lipscomb JD, Que L Jr
(2008) Proc Natl Acad Sci USA 105:7347–7352
7
6SF00515 to SLAC National Accelerator Laboratory and NIH grant
26. Choi EH, Suh S, Sander CL, Hernandez CJO, Bulman ER,
Khadka N, Dong Z, Shi W, Palczewski K, Kiser PD (2018) Hum
Mol Genet. https://doi.org/10.1093/hmg/ddy128
P41-GM-103393 (SSRL Structural Molecular Biology Program).
2
7. Sui X, Zhang J, Golczak M, Palczewski K, Kiser PD (2016) J Biol
Chem 291:19401–19412
Open Access This article is distributed under the terms of the Crea-
tive Commons Attribution 4.0 International License (http://creativeco
mmons.org/licenses/by/4.0/), which permits unrestricted use, distribu-
tion, and reproduction in any medium, provided you give appropriate
credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made.
2
2
8. Kabsch W (2010) Acta Crystallogr D 66:125–132
9. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Acta Crystal-
logr D 66:486–501
3
0. Murshudov GN, Skubak P, Lebedev AA, Pannu NS, Steiner RA,
Nicholls RA, Winn MD, Long F, Vagin AA (2011) Acta Crystal-
logr D 67:355–367
3
1. Winn MD, Ballard CC, Cowtan KD, Dodson EJ, Emsley P, Evans
PR, Keegan RM, Krissinel EB, Leslie AG, McCoy A, McNicholas
SJ, Murshudov GN, Pannu NS, Potterton EA, Powell HR, Read
RJ, Vagin A, Wilson KS (2011) Acta Crystallogr D 67:235–242
References
1
2
.
.
Wyss A (2004) J Nutr 134:246S–250S
32. Chen VB, Arendall WB 3rd, Headd JJ, Keedy DA, Immormino
RM, Kapral GJ, Murray LW, Richardson JS, Richardson DC
(2010) Acta Crystallogr D 66:12–21
Kamoda S, Saburi Y (1993) Biosci Biotechnol Biochem
5
7:926–930
3
4
5
.
.
.
von Lintig J, Vogt K (2004) J Nutr 134:251S–256S
33. Read RJ, Adams PD, Arendall WB 3rd, Brunger AT, Emsley P,
Joosten RP, Kleywegt GJ, Krissinel EB, Lutteke T, Otwinowski
Z, Perrakis A, Richardson JS, Sheꢄer WH, Smith JL, Tickle IJ,
Vriend G, Zwart PH (2011) Structure 19:1395–1412
34. Read RJ, McCoy AJ (2011) Acta Crystallogr D 67:338–344
35. Ravel B, Newville M (2005) J Synchrotron Radiat 12:537–541
36. George GN (2000) Stanford Synchrotron Radiation Lightsource.
SLAC National Accelerator Laboratory, Stanford
Kloer DP, Schulz GE (2006) Cell Mol Life Sci 63:2291–2303
Schwartz SH, Tan BC, Gage DA, Zeevaart JA, McCarty DR
(
1997) Science 276:1872–1874
6
.
Alder A, Jamil M, Marzorati M, Bruno M, Vermathen M, Bigler
P, Ghisla S, Bouwmeester H, Beyer P, Al-Babili S (2012) Science
3
35:1348–1351
7
8
9
.
.
.
Al-Babili S, Bouwmeester HJ (2015) Annu Rev Plant Biol
6:161–186
Ruch S, Beyer P, Ernst H, Al-Babili S (2005) Mol Microbiol
6
37. Rehr JJ, Deleon JM, Zabinsky SI, Albers RC (1991) J Am Chem
Soc 113:5135–5140
5
5:1015–1024
38. Costello A, Periyannan G, Yang KW, Crowder MW, Tierney DL
(2006) J Biol Inorg Chem 11:351–358
Brefort T, Scherzinger D, Limon MC, Estrada AF, Trautmann
D, Mengel C, Avalos J, Al-Babili S (2011) Fungal Genet Biol
39. von Lintig J, Vogt K (2000) J Biol Chem 275:11915–11920
40. Housecroft CE, Sharpe AG (2008) Inorganic chemistry. Pearson
Education Limited, Essex, pp 637–685
4
8:132–143
1
1
1
1
1
1
0. Diaz-Sanchez V, Estrada AF, Limon MC, Al-Babili S, Avalos J
(
2013) Eukaryot Cell 12:1305–1314
41. Bertini I, Luchinat C (1984) Adv Inorg Biochem 6:71–111
42. Zhang JH, Kurtz DM, Maroney MJ, Whitehead JP (1992) Inorg
Chem 31:1359–1366
1. Kloer DP, Ruch S, Al-Babili S, Beyer P, Schulz GE (2005) Sci-
ence 308:267–269
2. Costas M, Mehn MP, Jensen MP, Que L Jr (2004) Chem Rev
43. Padden KM, Krebs JF, MacBeth CE, Scarrow RC, Borovik AS
(2001) J Am Chem Soc 123:1072–1079
1
04:939–986
3. Sui X, Kiser PD, Lintig J, Palczewski K (2013) Arch Biochem
Biophys 539:203–213
44. Padden KM, Krebs JF, Traꢀord KT, Yap GPA, Rheingold AH,
Borovik AS, Scarrow RC (2001) Chem Mater 13:4305–4313
45. Kiser PD, Farquhar ER, Shi W, Sui X, Chance MR, Palczewski K
(2012) Proc Natl Acad Sci USA 109:E2747–E2756
46. Irving H, Williams RJP (1953) J Chem Soc 3192–3210. https://
doi.org/10.1039/JR9530003192
4. Borowski T, Blomberg MR, Siegbahn PE (2008) Chemistry
1
4:2264–2276
5. Messing SA, Gabelli SB, Echeverria I, Vogel JT, Guan JC,
Tan BC, Klee HJ, McCarty DR, Amzel LM (2010) Plant Cell
2
2:2970–2980
47. Bock CW, Katz AK, Markham GD, Glusker JP (1999) J Am Chem
Soc 121:7360–7372
1
6. McAndrew RP, Sathitsuksanoh N, Mbughuni MM, Heins RA,
Pereira JH, George A, Sale KL, Fox BG, Simmons BA, Adams
PD (2016) Proc Natl Acad Sci USA 113:14324–14329
48. Roach PL, Clifton IJ, Fulop V, Harlos K, Barton GJ, Hajdu
J, Andersson I, Schofield CJ, Baldwin JE (1995) Nature
375:700–704
1
1
7. Kiser PD (2017) Proc Natl Acad Sci USA 114:E6027–E6028
8. Sui X, Kiser PD, Che T, Carey PR, Golczak M, Shi W, von Lintig
J, Palczewski K (2014) J Biol Chem 289:12286–12299
49. Hegg EL, Que L Jr (1997) Eur J Biochem 250:625–629
50. Kalliri E, Grzyska PK, Hausinger RP (2005) Biochem Biophys
Res Commun 338:191–197
1
2
9. Kiser PD, Zhang J, Badiee M, Li Q, Shi W, Sui X, Golczak M,
Tochtrop GP, Palczewski K (2015) Nat Chem Biol 11:409–415
0. Sui X, Weitz AC, Farquhar ER, Badiee M, Banerjee S, von Lintig
J, Tochtrop GP, Palczewski K, Hendrich MP, Kiser PD (2017)
Biochemistry 56:2836–2852
51. Price JC, Barr EW, Tirupati B, Bollinger JM Jr, Krebs C (2003)
Biochemistry 42:7497–7508
52. Fujisawa H, Uyeda M (1974) Eur J Biochem 45:223–231
53. Gopal B, Madan LL, Betz SF, Kossiakoꢀ AA (2005) Biochemistry
44:193–201
2
1. D’Ordine RL, Rydel TJ, Storek MJ, Sturman EJ, Moshiri F, Bart-
lett RK, Brown GR, Eilers RJ, Dart C, Qi YL, Flasinski S, Frank-
lin SJ (2009) J Mol Biol 392:481–497
54. Gunderson WA, Zatsman AI, Emerson JP, Farquhar ER, Que
L Jr, Lipscomb JD, Hendrich MP (2008) J Am Chem Soc
130:14465–14467
2
2
2. Fetzner S (2012) Appl Environ Microbiol 78:2505–2514
3. Fielding AJ, Kovaleva EG, Farquhar ER, Lipscomb JD, Que L
55. Shannon RD (1976) Acta Crystallogr A 32:751–767
(
2011) J Biol Inorg Chem 16:341–355
1
3