Characterization of the DNA-mediated Oxidation of Dps, a
Bacterial Ferritin
Anna R. Arnold
,
Andy Zhou
, and
Jacqueline K. Barton
*
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena,
California 91125, USA
Abstract
Dps proteins are bacterial ferritins that protect DNA from oxidative stress and have been
implicated in bacterial survival and virulence. In addition to direct oxidation of the Dps iron sites
by diffusing oxidants, oxidation from a distance via DNA charge transport (CT), where electrons
and electron holes are rapidly transported through the base-pair
π
-stack, could represent an
efficient DNA protection mechanism utilized by Dps. Here, we spectroscopically characterize the
DNA-mediated oxidation of ferrous iron-loaded Dps. X-band EPR was used to monitor the
oxidation of DNA-bound Dps after DNA photooxidation using an intercalating ruthenium
photooxidant and the flash-quench technique. Upon irradiation with poly(dGdC)
2
, a signal arises
with
g
= 4.3, consistent with the formation of mononuclear high-spin Fe(III) sites of low
symmetry, the expected oxidation product of Dps with one iron bound at each ferroxidase site.
When poly(dGdC)
2
is substituted with poly(dAdT)
2
, the yield of Dps oxidation is decreased
significantly, consistent with guanine radical intermediates facilitating Dps oxidation. We have
also explored possible protein electron transfer (ET) intermediates in the DNA-mediated oxidation
of ferrous iron-loaded Dps. Dps proteins contain a conserved tryptophan residue in close
proximity to the iron-binding ferroxidase site (W52 in
E. coli
Dps). In EPR studies of the
oxidation of ferrous iron-loaded Dps following DNA photooxidation, a W52A Dps mutant was
significantly deficient compared to WT Dps in forming the characteristic EPR signal at
g
= 4.3,
consistent with W52 acting as an ET hopping intermediate. This effect is mirrored
in vivo
in
E.
coli
survival in response to hydrogen peroxide, where mutation of W52 leads to decreased survival
under oxidative stress.
Graphical Abstract
*
Corresponding Author
. jkbarton@caltech.edu.
ASSOCIATED CONTENT
Primer sequences, detailed Dps overexpression and purification protocol, and Figures S1–S7. This material is available free of charge
via the Internet at
http://pubs.acs.org
.
The authors declare no competing financial interest.
HHS Public Access
Author manuscript
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Published in final edited form as:
J Am Chem Soc
. 2016 September 7; 138(35): 11290–11298. doi:10.1021/jacs.6b06507.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
INTRODUCTION
Dps proteins are dodecameric (12-mer) bacterial ferritins that protect DNA from oxidative
stress, and have been implicated in bacterial survival and virulence.
1
This protection is
thought to derive from the ferroxidase activity of Dps, where Dps proteins simultaneously
deplete ferrous iron and hydrogen peroxide, reactive species that can otherwise form
damaging hydroxyl radicals via Fenton chemistry.
2
Like other ferritins, Dps proteins are
spherical, with a hollow core where oxidized iron is reversibly stored. Some Dps proteins
nonspecifically bind DNA, such as that from
Escherichia coli
which utilizes N-terminal
lysine residues for DNA binding.
3
Within cells, Dps is upregulated by the transcriptional
regulator OxyR in response to oxidative stress;
4
Dps is also upregulated in stationary phase,
when an additional physical component of Dps protection may be biocrystallization with
DNA.
5
DNA charge transport (CT), where electrons and electron holes are efficiently transported
through the base-pair
π
-stack, represents a powerful means to carry out redox chemistry
from a distance.
6
Moreover, DNA CT chemistry is remarkably sensitive to the integrity of
the intervening DNA. We have explored biological applications of DNA CT, where we have
seen this chemistry being utilized as a first step for DNA repair proteins containing 4Fe4S
clusters to signal one another and thus localize to the vicinity of a lesion within the vast
milieu of the genome.
7
We have also found examples where DNA CT facilitates the
selective activation of redox-active transcription factors to respond to oxidative stress from a
distance.
6
Given these applications of DNA CT, we considered whether Dps, in addition to
interacting directly with diffusing oxidants, might also utilize DNA CT to protect DNA from
a distance. Guanine is the most easily oxidized base within DNA, and the presence of
adjacent stacked guanines further lowers the 5’-guanine oxidation potential;
8
thus, radicals
are characteristically formed at guanine multiplets upon DNA photooxidation.
9
A long
distance protection mechanism via DNA CT would involve electron transfer from Dps
through the DNA
π
-stack to fill the hole on guanine radicals, restoring the integrity of the
DNA. In this way, Dps could respond to an oxidative affront to the DNA, even if the protein
is bound, at the minimum, a hundred base-pairs away.
10
Indeed, we have previously shown biochemically that
E. coli
Dps loaded with ferrous iron at
the ferroxidase sites can protect DNA from oxidative damage through DNA CT.
11
[Ru(phen)
(dppz)(bpy
′
)]
2+
, where phen is 1,10-phenanthroline, dppz is dipyrido[2,3-a:2
′
,3
′
-
c]phenazine, and bpy
′
is 4-butyric acid-4
′
-methyl-2,2
′
-bipyridine, was covalently tethered
to the 5
′
end of mixed-sequence 70-mer DNA and served as the distally tethered,
Arnold et al.
Page 2
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
intercalated photooxidant generated
in situ
by the flash-quench technique. Upon excitation
with visible light (“flash”), the ruthenium(II) excited state can be oxidatively quenched
(“quench”) by a diffusing quencher, here [Co(NH
3
)
5
Cl]
2+
, to form a highly oxidizing
intercalated Ru(III) complex (1.6 V versus NHE).
12
In the absence of protein, oxidative
damage is observed preferentially at a guanine triplet within the 70-mer DNA duplex.
Titrating in ferrous iron-loaded Dps significantly attenuates the level of oxidative damage at
the guanine triplet, while Apo-Dps and ferric iron-loaded Dps, which lack available reducing
equivalents, do not display this protection.
11
Luminescence experiments rule out direct
interaction between the ruthenium photooxidant excited state and Dps, consistent with a
long range DNA-mediated oxidation mechanism.
Long-distance, DNA-mediated oxidation of Dps could be an effective mechanism for
bacteria to protect their genomes from oxidative insults, contributing to pathogenic survival
and virulence.
Here we spectroscopically characterize the DNA-mediated oxidation of ferrous iron-loaded
Dps. Electron paramagnetic resonance (EPR) spectroscopy has previously been used to
observe oxidation of the 4Fe4S cluster of the base excision repair glycosylase MutY
following flash-quench DNA photooxidation.
12b
In this work, we use X-band EPR
spectroscopy to observe the oxidation of DNA-bound
E. coli
Dps loaded with ferrous iron at
the ferroxidase sites and to investigate the DNA-mediated characteristics of this oxidation
(Figure 1).
We also explore possible protein electron transfer intermediates in the DNA-mediated
oxidation of ferrous iron-loaded Dps. There is a highly conserved tryptophan residue in
close proximity (approximately 5 Å) to the di-iron ferroxidase site in Dps proteins, W52 in
E. coli.
13
Aromatic amino acids such as tryptophan and tyrosine can act as electron transfer
(ET) hopping intermediates in proteins, allowing for rapid ET across the protein where a
single ET process would be kinetically difficult.
14
Because of the location of this aromatic
tryptophan residue between the ferroxidase site and the outer protein shell where the DNA
must be located, it is an attractive candidate as a hopping intermediate to facilitate ET
between the ferroxidase site of Dps and DNA. Previous work has suggested an important
role for this conserved tryptophan residue in Dps proteins. Upon oxidation with hydrogen
peroxide of Dps loaded with only 6 Fe(II)/Dps, UV-visible stopped flow experiments with
E.
coli
Dps were able to observe spectra with maxima at 512 and 536 nm, consistent with a
neutral tryptophan radical.
15
By comparison with site-directed mutagenesis studies on
L.
innocua
Dps, which also contains a tyrosine residue nearby the ferroxidase site, the
tryptophan radical in
E. coli
was ascribed to W52.
15
A double mutant of
L. innocua
Dps,
where both of the aromatic residues in proximity to the ferroxidase site were mutated, was
assayed for its ability to protect plasmid DNA from degradation by ferrous iron and
hydrogen peroxide. Given that the protective capacity of the
L. innocua
Dps double mutant
was significantly attenuated, it was concluded that these conserved aromatic residues act as a
trap for electron holes generated by the oxidation of insufficient ferrous iron by hydrogen
peroxide.
15
,
16
Interestingly, 24-mer ferritins contain a conserved tyrosine residue in close
proximity to the ferroxidase site that has also been proposed to act as a molecular capacitor,
Arnold et al.
Page 3
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
although other studies contend with this hypothesis.
17
Overall, the conserved aromatic
residue in close proximity to the ferroxidase site may play an important role in ferritins.
Thus, here we investigate two
E. coli
Dps W52 mutants: W52Y, where an aromatic residue
is maintained at this position, and W52A, which abolishes the aromatic residue adjacent to
the ferroxidase site. We compare these mutants with the wild type (WT) protein in EPR
studies of the oxidation of ferrous iron-loaded Dps following DNA photooxidation. Because
the intercalating ruthenium photooxidant is a one-electron oxidant, sending one electron
hole at a time into the DNA
π
-stack, we examine the possible role of
E. coli
Dps W52 as an
electron transfer hopping intermediate rather than a molecular capacitor. In addition to EPR,
we also probe the role of W52 with respect to cellular survival in studies of
E. coli
under
oxidative stress.
EXPERIMENTAL METHODS
Materials
The alternating co-polymer DNA duplexes poly(dGdC)
2
and poly(dAdT)
2
were purchased
from Sigma. The DNA duplexes were passed through Bio-rad spin columns (6 K MWCO)
before use and quantified based on their molar absorptivity values in base-pairs,
18
(poly(dGdC)
2
:
ε
254
= 16,800 M
−1
cm
−1
, poly(dAdT)
2
:
ε
262
= 13,200 M
−1
cm
−1
). Duplexes
were then dried on a speed-vac, brought into an anaerobic chamber, and re-suspended in
deoxygenated buffer for EPR experiments. Buffers (50 mM Tris or 50 mM MOPS, pH 7.0,
150 mM NaCl, 5% glycerol) were deoxygenated in a Schlenk flask by at least 4 cycles of
freeze-pump-thaw.
[Ru(phen)(dppz)(bpy
′
)]
2+
was synthesized according to published methods,
19
purified by
reversed-phase chromatography, and characterized by NMR and ESI mass spectrometry
(expected for the +2 ion: 409.62 m/z, observed: 410.2 m/z). The ruthenium photooxidant
was brought into the anaerobic chamber as a solid powder, re-suspended with deoxygenated
buffer, and a sample removed for quantification based on UV-Vis absorption (
ε
440
= 21,000
M
−1
cm
−1
). [Co(NH
3
)
5
Cl]Cl
2
was purchased from Sigma (99.995% pure) and used as
received. The Co quencher was brought into the anaerobic chamber as a solid powder, re-
suspended with deoxygenated buffer, and a sample removed for quantification based on UV-
Vis (
ε
550
= 47.5 M
−1
cm
−1
).
W52 mutagenesis
The W52A and W52Y
E. coli
Dps mutants were made with a Quikchange II-E Site-Directed
Mutagenesis Kit (Stratagene) using a pBAD18-
dps
plasmid (containing the WT
E. coli dps
gene and an ampicillin resistance cassette) donated by Dr. Roberto Kolter as a template.
20
Primers were purchased from Integrated DNA Technologies. All mutagenized plasmids were
sequenced (Laragen) to confirm the desired sequences. After creating the mutant pBAD18-
dps
plasmids, the
E. coli
cell line ZK2471 (
dps
::
kan
Δ
recA
Δ
ara
) donated by Dr. Roberto
Kolter was made electrocompetent and the plasmid was transformed via electroporation into
these cells. Primer sequences are provided in the Supporting Information.
Arnold et al.
Page 4
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Dps overexpression, purification, and iron loading
WT and W52 mutant
E. coli
Dps proteins were overexpressed and purified according to
previous procedures,
11
with one significant modification: the addition of a HiTrap Heparin
HP affinity column to ensure removal of endogenous DNA. A detailed protocol is provided
in the Supporting information. Proteins were deoxygenated in Schlenk tubes by rapid cycles
of vacuum and argon according to previous procedures,
11
and brought into the anaerobic
chamber. Proteins were then anaerobically incubated with excess ferrous iron to load the
ferroxidase sites, with unbound iron removed by size exclusion chromatography.
11
The
number of iron ions bound per Dps dodecamer was then quantified by separately measuring
the protein and iron concentration. Protein concentrations were measured using either a
Bradford assay (Sigma) or
ε
280
values calculated using the ExPASy ProtPram tool (
http://
web.expasy.org/-protparam/
) with calculated molar absorptivity values for WT, W52A and
W52Y
E. coli
Dps dodecamers of 1.86 × 10
5
M
−1
cm
−1
, 1.20 × 10
5
M
−1
cm
−1
, and 1.37 ×
10
5
M
−1
cm
−1
, respectively. Iron concentration was quantified according to [Fe(bpy)
3
]
2+
absorbance (
ε
522
= 8790 M
−1
cm
−1
) using a denaturing method detailed elsewhere.
11
As-
purified, Dps was considered to be Apo-Dps with typically ≤1 Fe/Dps.
Circular dichroism of Dps
Protein concentrations were determined using the calculated
ε
280
values described above.
Spectra were recorded at 25°C on a Model 430 circular dichroism spectrometer (AVIV) in a
buffer consisting of 50 mM Tris, pH 7.0, 150 mM NaCl. The spectra shown are the average
of three individual scans, with a spectrum of buffer alone subtracted.
EPR sample preparation
EPR samples were prepared in an anaerobic chamber using the anaerobic materials outlined
above. Samples were loaded into EPR tubes within the anaerobic chamber, sealed with
septa, and parafilmed around the septa seal. Under the conditions used in these experiments,
all samples containing both Dps and DNA precipitated, expected behavior associated with
Dps condensing DNA. EPR tubes were then brought out of the anaerobic chamber, frozen in
liquid nitrogen, and kept in the dark until measurement. Precipitated samples were
thoroughly mixed before freezing in liquid nitrogen. For chemically oxidized samples,
ferrous iron-loaded protein (approximately 120 μL) was added to the bottom of an EPR tube.
Approximately 20 μL of ferricyanide solution was added to the top of the EPR tube, which
was then sealed. Upon removal from the anaerobic chamber, the solutions were mixed
together and immediately frozen in liquid nitrogen within approximately 5–10 seconds of
the initiation of mixing.
EPR experiments
EPR spectra were measured on an X-band Bruker EMX spectrometer equipped with an
ER4119HS resonator and an Oxford ES9000 cryostat. Instrumental settings are detailed in
figure captions, but were generally as follows: modulation amplitude = 10 G at 100 kHz,
frequency = 9.37 GHz, microwave power = 6.4 mW, temperature = 10 K. Samples in
Suprasil quartz EPR tubes were irradiated while freezing in liquid nitrogen in an unsilvered
Dewar. The excitation source was a xenon lamp equipped with a lens to focus the beam and
Arnold et al.
Page 5
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
a 320 nm long-pass filter to remove UV light. Each sample was irradiated for approximately
10 seconds. For each sample, a dark control (DC) EPR spectrum was first measured at 10 K.
The sample was then thawed, mixed, and irradiated while freezing with liquid nitrogen as
described. The EPR spectrum of the irradiated sample was then measured under identical
instrumental settings. For data analysis, the DC spectrum was smoothed and subtracted from
the irradiated sample.
Hydrogen peroxide survival experiments
This protocol is adapted from that reported by Martinez and Kolter.
20
Hydrogen peroxide
was purchased from Macron (30% solution). Lyophilized catalase from bovine liver (≥
20000 units/mg protein) was purchased from Sigma and re-suspended in buffer (50 mM
K
3
PO
4
, pH 7) to make a stock concentration of 0.4 mg/mL. Overnight cultures of the
E. coli
ZK2471 strain (
dps
::
kan
Δ
recA
Δ
ara
) containing WT, W52A, or W52Y pBAD18-
dps
plasmids were prepared by inoculating single colonies in 5 mL of LB media containing 100
μg/mL ampicillin and 50 μg/mL kanamycin. After overnight shaking at 37°C, the cultures
were diluted 1:500 into 10 mL of fresh LB media also containing antibiotics. For each WT,
W52A, and W52Y, both induced and un-induced 10 mL cultures were prepared: L-arabinose
was added to induce Dps overexpression (0.2% w/v final concentration), and an equivalent
volume of sterile water was added to un-induced cells. Cultures were then incubated at 37°C
with shaking (200 rpm) for 3 hours until OD
600
= 0.3–0.4 (exponential phase). Separately,
the growth of the induced cultures were monitored over 8 hours for growth differences. Once
reaching exponential phase, cultures were serially diluted in LB media for a total of 10,000-
fold dilution with a final 1 mL aliquot volume. Hydrogen peroxide was added to these
aliquots to attain a final concentrations of 3 or 5 mM in the cell aliquots and mixed by
pipetting. After 15 minutes at RT, catalase solution was added to each aliquot to stop the
reaction (50 μL, working concentration: 50 μg/mL culture) and mixed by pipetting. Cultures
were incubated for 15 minutes after catalase addition to ensure complete hydrogen peroxide
reaction. Finally, cultures were plated in 10 μL droplets onto LB agar plates containing
ampicillin and kanamycin and incubated at 37°C overnight. The number of colonies in each
droplet was manually counted the subsequent day after plate imaging.
RESULTS
EPR Spectroscopy of WT
E. coli
Dps
In spectroscopic studies with Dps, we expect to observe the oxidation of mononuclear iron
sites under the conditions of this study. At the inter-subunit ferroxidase sites of Dps, two
irons are bound by two conserved histidine ligands and two conserved carboxylate ligands,
glutamate and aspartate.
1
This ligand coordination sphere creates two binding sites with very
different iron affinities: while one site has a relatively high affinity, the other site binds iron
weakly.
21
After binding, ferrous iron is oxidized and shuttled to the core of the protein for
storage. Whereas 24-mer ferritins react rapidly with dioxygen as an oxidant, Dps proteins
react slowly with dioxygen and much more quickly with hydrogen peroxide.
2
We have
previously found that ferricyanide also functions well as a chemical oxidant of
E. coli
Dps in
solution.
11
,
2
Whereas full occupation of the twelve di-iron centers of the protein would
correspond to 24 Fe(II)/Dps, we have found that under the anaerobic conditions used in our
Arnold et al.
Page 6
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
experiments (i.e. in the absence of oxidants),
E. coli
Dps binds only 12 Fe(II)/Dps.
11
This
loading agrees with studies on
Bacillus anthracis
and
Listeria innocua
Dps proteins, where a
bridging oxidant seems to be required to tether the lower affinity iron and form the di-iron
site.
21
,
23
Coupled with the specificity of iron binding evidenced by its abrogation in the
E.
coli
Dps ferroxidase site double mutant H51G/H63G, the 12 Fe(II)/Dps corresponds to
binding only at the higher affinity iron site of each ferroxidase center in the dodecameric
protein.
Mononuclear high-spin Fe(III) sites of low symmetry (i.e., non-heme) typically display an
EPR signal with an apparent
g
-value of 4.3, and this is what we observe (Figure 2).
24
,
25
This
mononuclear high-spin Fe(III) signal at
g
= 4.3 has been frequently reported in the EPR
spectra of 24-merferritins.
26
The EPR spectrum of iron-bound Dps proteins is consistent but
has been reported only once previously,
27
and has not yet been reported for
E. coli
Dps.
We first used chemical oxidation to examine the Dps oxidation products we might expect in
DNA flash-quench studies. All EPR samples described in this study were prepared
anaerobically in order to prevent dioxygen oxidation of ferrous iron loaded Dps. As
expected, Apo-Dps, which has not been loaded with iron, and Dps loaded with ferrous iron
are EPR-silent (Figure 2). However, when WT
E. coli
Dps loaded with 12 Fe(II)/Dps is
mixed anaerobically with stoichiometric ferricyanide and frozen in liquid nitrogen, within 5–
10 seconds, a split signal at
g
= 4.3 is observed at low temperature (10 K). Given that ferri-
cyanide has a different
g
-value and ferrocyanide is EPR-silent, and that the steady-state UV-
Visible spectrum of ferrous iron-loaded Dps incubated with ferricyanide indicates the
formation of oxidized iron species,
11
this signal at
g
= 4.3 can be assigned to oxidized ferric
iron at the mononuclear ferroxidase site in Dps. This signal was confirmed to be neither
power saturated nor overmodulated under the conditions used in this study. No other EPR-
active species are apparent in wide spectra from 500 to 4500 gauss (data not shown). When
instead the ferrous iron-loaded Dps was incubated with ferricyanide for much longer times,
no EPR-active species were observed (data not shown), likely because the oxidized iron was
translocated to the core of the protein, forming EPR-silent polynuclear species.
Next, we investigated the oxidation of DNA-bound WT Dps following DNA photooxidation
via the flash-quench technique. We compare the yield of Dps oxidation with the alternating
copolymers poly(dGdC)
2
and poly(dAdT)
2
in order to determine if guanine radical is an
important intermediate in Dps oxidation. Here, the sample is irradiated in an EPR tube while
freezing in liquid nitrogen in a clear dewar in order to trap reactive intermediates. For each
sample, an individual dark control (DC) was measured at low temperature (10 K). The
sample was then thawed, mixed, and irradiated while freezing to generate oxidative DNA
damage via the flash-quench technique. Efforts were made to irradiate all samples under
identical conditions for 10 seconds. The EPR spectrum of the irradiated sample was then re-
measured under identical instrument settings. All spectra shown in Figure 3 have had an
individual DC subtracted; thus, all features are a function of irradiation.
In a sample containing 20 μM Dps loaded with 12 Fe(II)/Dps, poly(dGdC)
2
at a
concentration of 1 mM base-pairs, 20 μM non-covalent [Ru(phen)(dppz)(bpy
′
)]
2+
and 120
μM [Co(NH
3
)
5
Cl]
2+
in a buffer of 50 mM Tris, pH 7.0, 150 mM NaCl, 5% glycerol, a split
Arnold et al.
Page 7
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
nearly isotropic signal at
g
= 4.3 is observed upon irradiation (Figure 3). Comparison with
the chemically oxidized sample indicates that this species corresponds to oxidized ferric iron
at the mononuclear ferroxidase site in Dps that was formed after DNA oxidation with the
ruthenium photooxidant. DC subtracted spectra in this trial were quantified by double
integration over the range of 1400 to 1700 gauss. Consistent trends were seen in two
separate trials. In contrast to the full sample (containing poly(dGdC)
2
DNA, ferrous iron-
loaded Dps, Ru photooxidant, and Co quencher), an 8-fold decreased signal was observed in
a sample lacking DNA (Minus DNA). An attenuated signal was also observed in an
irradiated sample that contained ferrous iron-loaded Dps, poly(dGdC)
2
, and [Ru(phen)
(dppz)(bpy
′
)]
2+
but lacked quencher (Light control). Some signal was observed in the light
control sample (1.7-fold less than the full sample), even though steady-state room
temperature luminescence experiments with mixed-sequence 70-mer duplex DNA
containing covalently tethered [Ru(phen)(dppz)(bpy
′
)]
2+
indicated that Dps does not quench
the ruthenium(II) excited state.
11
When Apo-Dps is substituted for Dps loaded with ferrous
iron, no signal at
g
= 4.3 is observed, confirming protein-bound iron as the origin of this
signal. Importantly, when poly(dAdT)
2
is substituted for poly(dGdC)
2
, the observed signal is
significantly attenuated (3.0-fold), suggesting that guanine radicals play a role in Dps
oxidation.
The wide EPR spectrum of the sample containing poly(dGdC)
2
with ferrous iron-loaded Dps
from 500 to 4500 gauss shows that the features evident upon irradiation are the
g
= 4.3
signal previously discussed, a broad signal at lower magnetic fields due to Co
2+
produced in
the flash-quench reaction, and a small signal at
g
= 2 (Figure S1). The
g
= 2 signal is likely
an organic radical, either guanine radical or tryptophan radical; however, we were unable to
obtain reproducible results at
g
= 2. Despite investigating different temperatures and
microwave powers to prevent power saturation, signals at
g
= 2 were too small for
conclusions to be drawn. While the tryptophan radical (W52) has been previously observed
in
E. coli
Dps with UV-Visible stopped-flow experiments,
15
it is likely that we lack the time
resolution to observe this radical with EPR. In Tris buffer, there is a signal at
g
= 4.3 when a
sample of ferrous iron only is mixed with ferricyanide (Figure S2). Therefore, it is important
to control for oxidized iron remaining bound to Dps in DNA photooxidation experiments.
One approach is to compare samples prepared in a MOPS buffer. In MOPS buffer, no signal
is observed when a sample of ferrous iron only is mixed with ferricyanide (Figure S3).
However, a small, unsplit signal is apparent at
g
= 4.3 in a sample of ferrous iron-loaded
Dps, poly(dGdC)
2
DNA, ruthenium photooxidant and diffusing quencher upon irradiation.
While the significantly smaller signals in MOPS buffer makes comparisons between samples
(i.e., different DNAs) difficult, this MOPS result suggests that at least some of the free iron
in Tris samples remains bound to the protein after oxidation.
Structure and Fe binding of W52 Dps mutants
The Dps monomer is composed of a four helix bundle with two helix-turn-helix motifs.
28
The far-UV circular dichroism (CD) spectra of WT Apo-Dps is consistent with this
α
-
helical structure
29
(Figure 4). Comparison of the WT, W52A and W52Y Dps CD spectra
shows that overall protein folding is relatively unaffected by these mutations.
Arnold et al.
Page 8
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Iron binding at the ferroxidase site was also investigated for both mutations. As previously
described for WT Dps,
11
the proteins were incubated anaerobically with excess ferrous iron
and unbound iron was subsequently removed with size exclusion chromatography. The
number of irons bound per Dps dodecamer was then quantified by the formation of
[Fe(bpy)
3
]
2+
after protein denaturation and addition of reductant and 2,2
′
-bipyridine. When
Dps concentration is measured via the Bradford reagent or calculated
ε
280
values, the
Fe(II)/Dps can be quantified. In one trial, the WT protein bound 14.6 ± 0.5 Fe(II)/Dps,
whereas the W52A and W52Y mutants bound only 8.6 ± 0.4 and 10.6 ± 0.4 Fe(II)/Dps,
respectively. Equivalent results were obtained in other trials. Note that small increases from
12 Fe(II)/Dps in the WT protein are likely due to a minor degree of oxidation due to trace
oxygen, allowing for some di-iron site formation. Figure 4 shows the [Fe(bpy)
3
]
2+
spectra
normalized to protein concentration for WT, W52A, and W52Y Dps. W52A Dps binds iron
on the order of 60% of WT, while W52Y is slightly better, binding 70% of the iron of WT
Dps. Thus iron binding is somewhat attenuated, but not abrogated, for these mutations.
Furthermore, these W52 mutations are not expected to affect the DNA binding of Dps;
previous work demonstrated that iron-loaded Dps binds DNA similarly to the Apo-protein.
11
Additionally, we observed the same level of precipitation and DNA condensation with the
W52 mutants and WT.
EPR results comparing WT Dps with W52 mutants
The ability of ferrous iron-loaded W52A and W52Y Dps to be oxidized by ferricyanide, a
diffusing oxidant, was first explored using EPR spectroscopy. Figure 5A shows oxidation of
the W52 mutants compared to WT with excess ferricyanide, with the EPR intensity adjusted
for iron loading (i.e., Intensity / (Fe
W52A
/Fe
WT
)). When the intensity of the EPR signal
resulting from ferricyanide oxidation is adjusted for iron loading in this manner, the W52
mutants show similar yields of iron oxidation to WT Dps, with W52Y showing a slightly
increased signal relative to WT. This result indicates that oxidation of the mononuclear iron
site by a diffusing oxidant is not affected in W52A and W52Y Dps compared to WT.
Next, the X-band EPR spectrum of ferrous iron-loaded WT
E. coli
Dps was compared to
W52A and W52Y Dps upon DNA photooxidation using the flash-quench technique.
Samples containing ferrous iron-loaded Dps, poly(dGdC)
2
DNA, non-covalent [Ru(phen)
(dppz)(bpy
′
)]
2+
and [Co(NH
3
)
5
Cl]
2+
were irradiated for identical lengths of time. The yield
of iron oxidation at
g
= 4.3 was attenuated in the W52 mutants compared to the WT protein,
even when adjusted for iron loading (Figure 5B). Overall smaller signals in Figure 5B
compared to the WT spectra in Figure 3 are due to a lower modulation amplitude in the
former (5 G and 10 G, respectively), while the intensity difference between Figures 5A and
5B is likely due to a poor kinetic window and inefficieny in irradiation for observing DNA
photooxidation. When the DC-subtracted, adjusted spectra are quantified by double
integration from 1400 to 1600 gauss, the W52A iron signal is 3.4-fold less than WT, while
the W52Y signal is 1.8-fold less than WT. The proficiency in oxidation of the iron sites in
these W52 Dps mutants by a chemical oxidant that directly diffuses to the iron site,
combined with the deficiency in the yield of iron oxidation upon DNA photooxidation,
suggests that W52 could play a role in mediating ET from the iron site to the DNA. As with
Arnold et al.
Page 9
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
WT Dps, we do not have the time resolution to observe a tryptophan or tyrosine radical
directly using EPR spectroscopy.
It is noteworthy that the EPR spectra of DC samples (i.e. before irradiation) of the W52
mutants show evidence of Co
2+
formation, whereas the WT protein does not (Figure S4).
The cobalt quencher, [Co(NH
3
)
5
Cl]
2+
, is a low-spin Co
3+
d
6
species with
S
= 0. Upon
reduction to Co
2+
, the complex becomes labile, forming [Co(H
2
O)
6
]
2+
, a high-spin
d
7
species with
S
= 3/2 (EPR-active). Therefore, there may be some direct electron transfer
from the ferrous mononuclear iron site to the Co
3+
quencher to yield EPR-active Co
2+
in
these mutants, perhaps because the ferroxidase site is more solvent-accessible. However,
there is very little evidence of ferric iron formation in the DC spectrum of the W52A mutant,
and in W52Y, a relatively small percentage of the total amount of iron in the sample is
oxidized, allowing ample room for an increase upon irradiation. Thus the lower yield of iron
oxidation that we observe upon DNA photooxidation with the W52 mutants is significant,
supporting our EPR results that suggest W52 as an electron transfer intermediate in Dps.
Hydrogen peroxide survival assay
The biological consequence of mutating W52 was also investigated by measuring the
survival of
E. coli
upon exposure to hydrogen peroxide for cells containing WT, W52A, or
W52Y Dps. The
dps
knockout
E. coli
strain (
dps
::
kan
Δ
recA
Δ
ara
) (ZK2471) was
transformed with a pBAD18 plasmid containing either the
E. coli
WT, W52A, or W52Y
dps
gene under the control of an inducible promoter. In the absence of hydrogen peroxide at the
inducer concentrations used in this study, no growth difference was observed between the
strains (Figure S5).
Adapted from the method of Martinez and Kolter,
20
cells were grown overnight and diluted
into fresh media with the addition of either the inducer (+), L-arabinose, or sterile water (−).
Cells were then grown to exponential phase (OD600 = 0.3–0.4) and challenged with 3 or 5
mM hydrogen peroxide. After quenching the reaction with catalase, cells were diluted and
plated in multiple droplets in order to quantify colony forming units (CFU). The results from
the 0.2% w/v of L-arabinose inducer at 10,000-fold dilution are shown in Figure 6. Percent
survival was calculated as the fraction of surviving colonies over the number of seeded
CFUs as calculated by dilution-adjusted OD
600
readings. Each mutant data set was then
normalized to the WT survival percentage. The standard error of the mean was calculated by
treating each droplet as a data point (n=16). Raw data are shown in Figure S6. As was
previously observed,
20
there is also a clear difference in bacterial survival after treatment
with hydrogen peroxide between induced and un-induced cells (Figure S7) demonstrating
that Dps is needed for protection.
As is evident in Figure 6, cells containing W52Y Dps survive the hydrogen peroxide
challenge more effectively than those with W52A Dps. Additionally, the relative percent
survival varies with hydrogen peroxide concentration. Whereas at 3 mM H
2
O
2
, cells
containing W52Y and W52A Dps survive at levels 77% and 41% of WT, respectively, when
cells are further challenged by 5 mM hydrogen peroxide, survival drops to 62% of WT for
W52Y and 24% for W52A Dps cells. Thus, W52 is an essential player for cells containing
Dps to survive this hydrogen peroxide challenge. There is certainly a component of Dps Fe
Arnold et al.
Page 10
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
loading proficiency (W52Y and W52A Dps bind iron at approximately 70% and 60%,
respectively, of WT,
vide supra
) in this survival. However the difference between observed
survival and Dps Fe loading, combined with the variation of survival with H
2
O
2
concentration shows that an additional component also affects survival; we propose that this
factor may be the efficiency of DNA CT. Furthermore, this
in vivo
trend (WT > W52Y >
W52A) parallels what we observe in EPR experiments, where the largest attenuation in the
yield of iron oxidation following DNA photooxidation is seen with W52A Dps. The
correlation between our EPR and
in vivo
experiments supports our proposal that survival
depends on both Dps Fe loading proficiency and the efficiency of DNA CT.
DISCUSSION
In earlier studies, we have shown biochemically that ferrous iron-loaded
E. coli
Dps can
protect DNA from oxidative damage generated using the flash-quench technique.
11
The
absence of this protective ability in Apo-Dps and Dps loaded with ferric iron, which both
lack available reducing equivalents, suggested that ferrous-iron loaded Dps protects DNA by
becoming oxidized via DNA CT to fill guanine radical holes. Here using EPR, we show
directly the oxidation of WT ferrous iron-loaded
E. coli
Dps following DNA photooxidation
generated by the flash-quench technique. Because Dps is loaded with one ferrous iron per
ferroxidase site, this oxidation is evidenced by the appearance of mononuclear ferric iron
species of low symmetry at an apparent
g
-value of 4.3. This signal is absent in controls
without DNA and is attenuated in the irradiated control that lacks the diffusing quencher
necessary for oxidation by flash-quench.
The results described here furthermore support the idea that guanine radicals facilitate Dps
oxidation. When poly(dGdC)
2
is substituted with poly(dAdT)
2
, the yield of Dps oxidation is
decreased significantly. A similar dependence on guanine radicals was observed in the
oxidation of the 4Fe4S cluster of the base excision repair protein MutY following DNA
photooxidation by flash-quench.
12b
More efficient protein oxidation by guanine radicals is
likely a kinetic effect. Back electron transfer processes generally decrease the observed yield
of oxidized protein using the flash-quench technique. Adenine radicals would be expected to
have short lifetimes compared to the neutral guanine radical, which persists for
milliseconds.
12a
Thus, as was described for MutY,
12b
it appears likely that guanine radical
formation allows more time for the oxidation of Dps by better competing with rapid back
electron transfer to the intercalated ruthenium photooxidant, resulting in higher yields of
Dps oxidation with guanine radical as an intermediate. Additionally, poorly stacked ATAT
tracts do not conduct charge efficiently;
30
this poor CT may also be a factor in the lower
yield of protein oxidation with poly(dAdT)
2
DNA. Generally, the more favorable oxidation
of Dps by guanine radicals also supports a sequential process, using DNA-mediated CT,
where after guanine radicals are produced, Dps is oxidized to fill guanine radical holes.
Can we also obtain information concerning the path for electron transfer? In 12-mer Dps
proteins, there is a conserved tryptophan residue in close proximity to the ferroxidase site
(W52 in the
E. coli
protein), whereas 24-mer ferritins contain a conserved tyrosine residue.
In both cases, this aromatic residue has been proposed to act as a molecular capacitor,
providing an extra electron during iron oxidation in order to prevent formation of oxygen
Arnold et al.
Page 11
J Am Chem Soc
. Author manuscript; available in PMC 2016 October 07.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript