of 22
A Redox Role for the [4Fe4S] Cluster of Yeast DNA Polymerase
δ
Phillip L. Bartels
1
,
Joseph L. Stodola
2
,
Peter M.J. Burgers
2,*
, and
Jacqueline K. Barton
1,*
1
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena,
CA
2
Department of Biochemistry and Molecular Biophysics, Washington University School of
Medicine, St. Louis, MO
Abstract
A [4Fe4S]
2+
cluster in the C-terminal domain of the catalytic subunit of the eukaryotic B-family
DNA polymerases is essential for the formation of active multi-subunit complexes. Here we use a
combination of electrochemical and biochemical methods to assess the redox activity of the
[4Fe4S]
2+
cluster in
Saccharomyces cerevisiae
polymerase (Pol)
δ
, the lagging strand DNA
polymerase. We find that Pol
δ
bound to DNA is indeed redox-active at physiological potentials,
generating a DNA-mediated signal electrochemically with a midpoint potential of 113 ± 5 mV
versus NHE. Moreover, biochemical assays following electrochemical oxidation of Pol
δ
reveal a
significant slowing of DNA synthesis that can be fully reversed by reduction of the oxidized form.
A similar result is apparent with photooxidation using a DNA-tethered anthraquinone. These
results demonstrate that the [4Fe4S] cluster in Pol
δ
can act as a redox switch for activity, and we
propose that this switch can provide a rapid and reversible way to respond to replication stress.
For Table of Contents Only
*
to whom correspondence should be addressed at burgers@wustl.edu and jkbarton@caltech.edu.
Supporting Information
DNA synthesis and purification for electrochemistry, preparation of DNA-modified gold electrodes, DNA substrate design, scan rate
dependence, UV-visible spectra and CV of EndoIII in polymerase buffer, Pol
δ
SQWV +/− PCNA, bulk electrolysis and spectroscopic
characterization, full alkaline agarose gels, polyacrylamide gel analysis, AQ assay with
E. coli
Klenow fragment. This material is
available free of charge via the Internet at
http://pubs.acs.org
.
HHS Public Access
Author manuscript
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Published in final edited form as:
J Am Chem Soc
. 2017 December 20; 139(50): 18339–18348. doi:10.1021/jacs.7b10284.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
INTRODUCTION
During genomic replication, eukaryotic cells divide the task of DNA synthesis between three
B-family DNA polymerases (Pols): Pol
α
, Pol
δ
, and Pol
ε
(
1
). Each of these enzymes
forms a multi-subunit complex composed of a catalytic subunit and a B-subunit, with
additional accessory subunits present in Pol
δ
and Pol
ε
(
2
). Recent work has shown that a
[4Fe4S]
2+
cluster in the C-terminal domain (CTD) of the catalytic subunit is essential for the
formation of multi-subunit complexes in the case of Pol
δ
(
3
). Functionally, the B-family
polymerases are believed to carry out DNA synthesis in a division of labor model, with the
DNA primase-Pol
α
complex initiating 5
-3
DNA synthesis by forming an RNA-DNA
hybrid primer that is then extended by Pol
ε
on the continuously generated leading strand
and by Pol
δ
on the discontinuously formed lagging strand (
4
). Additional roles for Pol
δ
during leading strand replication have been suggested (
5
,
6
). Pol
δ
is also involved in various
DNA recombinatorial and repair processes (
7
).
While the C-terminal [4Fe4S] cluster clearly plays a role in complex formation (
3
), several
lines of evidence suggest a more direct functional role. First, a 2.5 Å X-ray crystal structure
of the yeast Pol
α
CTD in complex with its B-subunit contained zinc in place of a cluster,
demonstrating that structural integrity can be supported by simpler metals (
8
). Given the
metabolic expense of [4Fe4S] cluster biosynthesis and loading into target proteins, the strict
conservation of this cofactor in the B-family polymerases suggests that it serves an
important function (
9
). Indeed, the importance of [4Fe4S] clusters in these enzymes is
emphasized by the presence of an additional cluster in the unique Pol
ε
N-terminal domain
(
10
).
The [4Fe4S] clusters perform a wide range of roles in biology including enzymatic catalysis
and electron transfer (
11
). In the DNA polymerases, the cluster is not required for catalysis
(
3
,
12
). Many DNA-processing enzymes have now been shown to contain [4Fe4S] clusters,
and, in many cases, a DNA-bound redox activity of the cluster has been demonstrated (
13
16
). These diverse proteins include base excision repair glycosylases, repair helicases and
DNA primase. As in the Pol
δ
holoenzyme, the clusters are largely redox-inert in the
absence of DNA (
17
20
). However, when bound to DNA, these protein cofactors undergo a
significant negative shift in redox potential, activating the clusters toward oxidation (
21
23
).
Electrochemical experiments with DNA-bound proteins show a reversible redox signal with
potentials ranging from 65–95 mV versus the normal hydrogen electrode (NHE) (
13
16
).
EPR studies support the assignment of the reversible signal to the [4Fe4S]
3+/2+
couple
favored by high-potential iron proteins (HiPIP) that are electron carriers (
13
,
24
25
). In
addition to modulating redox potential, the
π
-stacked base pairs of DNA can act as a
medium for long-range charge transport between redox-active proteins (
26
). DNA-mediated
charge transport (DNA CT) is characterized by a shallow distance dependence and high
sensitivity to base pair stacking, making it an excellent reporter of DNA integrity (
26
).
Importantly, although DNA CT can be attenuated by proteins that bend the duplex or flip out
bases, DNA CT can proceed unimpeded through nucleosome-wrapped DNA (
26
).
The redox activity of the [4Fe4S] cluster appears to be utilized in many of these proteins as a
switch to regulate DNA binding and therefore activity. For the DNA repair enzyme,
Bartels et al.
Page 2
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Endonuclease III (EndoIII), the negative shift in redox potential associated with DNA
binding has been shown to lead to a 500-fold increase in DNA affinity for the oxidized
[4Fe4S]
3+
cluster versus the reduced 2+ form (
27
). In the case of human DNA primase, the
oxidation state of the [4Fe4S] cluster also controls template binding, and redox switching
through electron transfer between clusters in primase and Pol
α
has been proposed to
regulate RNA primer handoff (
16
).
Here we focus on Pol
δ
, a central B-family polymerase. We utilize a combination of
electrochemical, spectroscopic, and biochemical techniques to investigate redox activity in
this enzyme and to understand the consequences of redox switching for polymerase activity.
These studies provide a new perspective on polymerase regulation under oxidative stress.
MATERIALS and METHODS
Protein expression and purification
Yeast Pol
δ
(WT and exo
D520V), RFC, RPA, PCNA, and
E. coli
EndoIII were expressed
according to previously published protocols (
3
,
28
).
DNA preparation
The DNA substrate for electrochemistry consisted of a 49:58-mer primer-template
composed of three oligomers: a 20-mer with a 3
thiol modification, a 38-mer, and a 49-mer
complement; sequences are as follows (see also Figure 1c):
20-mer Thiol: 5
- GCT GTC GTA CAG CTC AAT GC - 3
- (CH
2
)
2
O(CH
2
)
3
SH
38-mer: 5
- TAA CAG GTT GAT GCA TCG CGC TTC GGT GCT GCG TGT CT -
3
49-mer: 5
- GCA TTG AGC TGT ACG ACA GCA GAC ACG CAG CAC CGA
AGC GCG ATG CAT C - 3
The bold G of the 49-mer was changed to an A or an abasic (AP) site for CA mismatch and
abasic site discrimination experiments.
DNA replication assays used single-stranded M13mp18 plasmid purchased from New
England Biolabs (NEB). Primers were purchased from IDT and purified by HPLC as
described above. Primed DNA was formed by heating a 1:1 plasmid/primer mix in activity
buffer (50 mM Tris-HCl, pH 7.8, 50 mM NaCl) to 90° C for 5
and cooling to RT over
several hours. The M13mp18 DNA primer had the following sequence (complementary to
positions 6265-6235):
5
- GAC TCT AGA GGA TCC CCG GGT ACC GAG CTC G - 3
Primers were radiolabeled by incubating 10 pmol of 31-mer M13mp18 primer with T4
polynucleotide kinase (PNK) and [
γ
-
32
P] ATP (Perkin Elmer) in T4 buffer (NEB) for 15
minutes at 37° C. Reactions were stopped by addition of EDTA to 10 mM and heating at 75°
C for 10 minutes. 2 log DNA ladder (NEB) was dephosphorylated by calf intestinal alkaline
phosphatase (CIAP; 60 minutes, 37° C) prior to labeling in the same manner. As an
additional size standard, duplexed M13mp18 DNA was linearized by digestion with HincII
Bartels et al.
Page 3
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
(60 minutes, 37° C) and dephosphorylated by CIAP prior to radiolabeling. Proteins and
unincorporated ATP were removed using spin columns (BioRad Microspin6) equilibrated in
Pol
δ
activity buffer (50 mM Tris-HCl, pH 7.8, 50 mM NaCl). T4 PNK, CIAP, HincII, and
dsM13mp18 DNA were purchased from NEB.
Pol
δ
Electrochemistry
All electrochemical experiments were performed using a potentiostat equipped with a
multiplexer, both from CH Instruments. Experiments used a standard 3-electrode cell
composed of an Au working electrode, a Ag/AgCl reference electrode in 3 M NaCl
(BASInc), and a 1 mm diameter Pt wire counter electrode (Lesker). Potentials were
converted from Ag/AgCl to NHE by adding 212 mV to the potential as measured by Ag/
AgCl; this correction accounted for both ambient temperature and the use of 3 M NaCl for
reference storage (
29
). Experiments with Pol
δ
were all run in polymerase storage buffer (30
mM HEPES, pH 7.4, 350 mM NaAc, 1 mM DTT, 0.1 mM EDTA, 10% glycerol v/v, and
0.01% w/v decaethylene glycol monododecyl ether).
Because PCNA can slide directly onto DNA with open ends, the clamp loader complex RFC
was excluded from these experiments (
30
). WT Pol
δ
3
-5
exonuclease activity was
prevented by excluding Mg
2+
from the buffer. In initial experiments, 3–5 μM WT Pol
δ
or
exonuclease-deficient Pol
δ
D520V (DV) (
31
) was incubated on the electrode for several
hours in the presence of 5–10 μM PCNA. To spare enzyme, later experiments used Pol
δ
DV
at 500 nM in combination with 5.0 μM PCNA, 80 μM dATP and 8 mM MgAc
2
. Cyclic
voltammetry (CV; 100 mV/s scan rate) and square wave voltammetry (SQWV; 15 Hz
frequency, 25 mV amplitude) scans were taken once per hour for several hours. Between
scans, electrodes were covered in Parafilm and stored in a humid environment to minimize
evaporation. CV scan rate dependence was assessed after 3 hours using rates of 20, 50, 80,
100, 200, 500, 750, and 1000 mV/s. In experiments with abasic and CA mismatch DNA,
signal attenuation was calculated as follows:
[1 − ((peak area on abasic or CA mismatch DNA)/(peak area on unmodified DNA))]
∗ 100%
(1)
Pol
δ
concentrations are reported as the concentration of [4Fe4S] cluster, determined by UV-
visible absorbance at 400 nm (
ε
400
= 13000 M
−1
cm
−1
) (
3
). The [4Fe4S] cluster loading was
in the range of 70–85%, determined by dividing [4Fe4S] concentration by total protein
concentration as measured by Bradford assay, Pierce BCA assay, and UV-visible absorbance
at 280 nm (
ε
280
= 194100 M
−1
cm
−1
; estimated using the EXPasy ProtParam tool). Bradford
and BCA assay standard curves were generated using a BSA standard, and both kits were
purchased from Thermo Scientific. UV-visible spectra were taken on a Cary Varian
instrument using 100 μL quartz cuvettes purchased from STARNA Cells.
When possible, diffusion coefficients were obtained from the scan rate dependence of the
CV current using the Randles-Sevcik equation (
32
):
Bartels et al.
Page 4
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
I
p
= [0.4463(F
3
/RT)
1/2
](n
3/2
)(A)(D
1/2
)(C°)ν
1/2
(2)
I
p
is the peak current in amperes, F is Faraday’s constant (96485 C·mol
−1
), R is the universal
gas constant (8.314 J·(mol·K)
−1
), T is temperature in K, n is the number of electrons
transferred per CV peak, A is electrode area in cm
2
, D is the diffusion coefficient in cm
2
·s
−1
,
C° is bulk protein concentration in mol·cm
−3
, and
ν
is the scan rate in V·s
−1
. Experimental
values of D were compared to those estimated by the Stokes-Einstein equation,
D = k
B
T/6πηR
(3)
where k
B
is Boltzmann’s constant (1.38 × 10
−23
J·K
−1
), T is the incubation temperature (293
K),
η
is the solution viscosity (estimated to be 1.38 × 10
−3
Pa·s for an aqueous solution with
10% glycerol), and R is the hydrodynamic radius. R was estimated to be ~26 Å from
dimensions obtained from X-ray crystal structures of DNA-bound Pol3, Pol1–Pol12, and
PCNA (PDB ID 3IAY, 3FLO, and 4YHR, respectively).
Electrochemical Oxidation and Spectroscopic Analysis of Pol
δ
To prevent cluster degradation in the presence of O
2
, bulk electrolysis was performed in an
anaerobic glove bag (COY) under a 95% N
2
, 5% H
2
atmosphere with an O
2
-scavenging
catalyst present. Buffers were degassed by bubbling in argon for several hours and stored
open in the glove bag overnight prior to experiments. For spectroscopic characterization, a
150 μL sample of 1–2 μM Pol
δ
was added to two identical DNA-modified electrodes. On
one electrode, a potential of 0.412 V vs NHE was applied for ~15 minutes, while no
potential was applied to the other. Oxidation yields were estimated by taking the difference
between the total charge obtained in the presence of Pol
δ
and that generated by electrolysis
with buffer alone. Following electrolysis, UV-visible and electron paramagnetic resonance
(EPR) spectroscopy were used to confirm the integrity of the cluster after electrolysis.
Samples were sealed in cuvettes for UV-visible spectroscopy, and subsequently returned to
the glove bag and added to EPR tubes. Tubes were sealed by Parafilm and frozen in liquid
nitrogen outside the bag. Continuous wave X-band EPR was performed at 10 K, and each
experiment consisted of 9 sweeps taken at 12.88 mW microwave power, 2 G modulation
amplitude, and 5.02 × 10
3
receiver gain.
Pol
δ
Activity Assays
Immediately prior to assays, Pol
δ
DV was oxidized on Au rod electrodes exactly as
described for spectroscopic characterization, but the sample was diluted to 190 nM in
degassed storage buffer in a total volume of 30–40 μL. Reduction of oxidized sample was
carried out at a potential of −0.188 V vs NHE for a similar length of time. In parallel with
electrolysis, 140 μL reaction mixes (0.1mg/mL BSA, 80 μM each dNTP, 500 μM ATP, 2.0
nM M13mp18 with a
32
P-labeled primer, 8.0 mM MgAc
2
, 500 nM RPA, 5.0 nM RFC and
5.0 nM PCNA, 50 mM NaCl, 50 mM Tris-HCl, pH 7.8) were prepared inside the glove bag.
The PCNA sliding clamp was loaded onto the primer end by incubating the reaction mix
Bartels et al.
Page 5
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
with the RFC clamp loader and ATP for 1 minute at 30 °C. After clamp loading, reactions
were initiated by the addition of 2 nM (final concentration) oxidized, untreated, or re-
reduced Pol
δ
DV. Reactions were run at 30 °C, and 20 μL aliquots were removed and
quenched at specific time points by adding 10 μL stop mix (10 mM EDTA and 0.1% v/v
SDS final concentration). The polymerase was heat-inactivated for 10–20 minutes (55 °C),
and samples were counted on a liquid scintillation counter to determine exposure time (1
hour per 300,000 counts). Samples were dried on a speed vacuum and dissolved in alkaline
gel buffer (500 mM NaOH, 10 mM EDTA) with 1× alkaline loading dye (6× stock: 300 mM
NaOH, 6.0 mM EDTA, 18% Ficoll w/v, 0.25% xylene cyanol w/v, and 0.15% bromocresol
green w/v), and equivalent amounts of radioactivity were then loaded onto a 1% alkaline
agarose gel and run at 30 V for 14–15 hours. Gels were neutralized in 7% TCA (w/v) in
water for 30 minutes at RT and dried under mild pressure for several hours, exposed on a
phosphor screen and visualized on a Typhoon phosphorimager (GE Healthcare). Products
were analyzed using ImageQuant software (GE Healthcare). The relative amounts of DNA
synthesis were determined by dividing the volume of the largest band in an oxidized sample
by the equivalent band in the appropriate untreated sample.
To limit DNA synthesis to that of a single processive cycle by the PCNA-Pol
δ
complex,
0.01 mg/mL heparin was included in reactions (
33
) that were then analyzed on 5%
polyacrylamide gels. In these instances, Pol
δ
was added after clamp loading and reactions
were started by adding a mix of dNTPs and heparin. Quenched reactions were then counted,
dried, and redissolved in 2.0 μL formamide loading dye. Immediately prior to gel loading,
samples were heated at 90 °C for 10 minutes, and gels were run at ~50 W for 5 hours in 1×
TBE buffer. Polyacrylamide gels were then exposed and imaged by phosphorimager
analysis.
Tris-HCl, NaCl, MgAc
2
, BSA, and heparin were purchased from Sigma-Aldrich, while
dNTPs and ATP were from NEB. dNTPs, ATP, and MgAc
2
were thoroughly degassed prior
to reaction, and the protein stocks were kept open during a series of vacuum/nitrogen/gas
mix purges to minimize residual oxygen.
Chemical Oxidation of Pol
δ
For photooxidation, the 31-mer M13mp18 primer was covalently modified with a 5
anthraquinone (AQ). AQ was prepared as a phosphoramidite and added to the unmodified
DNA on a DNA synthesizer according to previously reported procedures (
34
). The presence
of AQ was verified by MALDI-TOF, and the modified primer was annealed to M13mp18
DNA in Pol
δ
activity buffer (50 mM Tris-HCl, pH 7.8, 50 mM NaCl). Because the 5
AQ
modification prevented
32
P end-labeling, DNA was labeled by adding 2 μCi [
α
-
32
P] dATP
(Perkin Elmer) to the reaction, and incorporation of [
α
-
32
P] dATP was facilitated by
lowering the concentration of cold dATP from 80 μM to 10 μM.
Anaerobic reaction mixes lacking dNTPs were prepared in glass vials and incubated under a
solar simulator equipped with a UVB/C long pass filter or in the dark for 30 minutes. To
ensure complete oxidation, 2-fold molar excess of both PCNA and Pol
δ
DV were included.
As controls, reactions were also run using unmodified DNA (no AQ) and AQ reactions were
repeated with 140 nM Klenow fragment exo
(NEB). After treatment, samples were
Bartels et al.
Page 6
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
returned to the glove bag and transferred into Eppendorf tubes containing dNTPs to start the
reaction. Free dNTPs were removed using BioRad Microspin 6 columns in SCC buffer (GE
Healthcare) and sample radioactivity was quantified on a liquid scintillation counter.
Samples were then run out on a 1% alkaline agarose gel and visualized by phosphorimaging
analysis. Overall [
α
-
32
P] dATP incorporation was used to compare overall DNA synthesis
levels by dividing the total radioactivity counts in oxidized samples by those of dark
controls.
RESULTS
Electrochemical Characterization of Pol
δ
To determine whether Pol
δ
holoenzyme was redox active in the presence of DNA, we
carried out electrochemistry on DNA-modified gold electrodes. In initial experiments, 3 μM
WT Pol
δ
in storage buffer was combined with 10 μM PCNA and incubated on the electrode
for several hours. CV scans taken hourly reveal a reversible signal with a midpoint potential
of 116 ± 3 mV vs NHE (Figure 1a). This signal grows in over time to reach a maximum size
of 41 ± 4 nC and −51 ± 2 nC for the reductive and oxidative peaks at a 100 mV/s scan rate
after two hours of incubation (Figure 1). The CV current varies linearly with the square root
of the scan rate (Figure S1), as expected of a diffusive rather than adsorbed species (
32
). The
diffusive nature of the signal is in agreement with earlier studies of DNA binding proteins
(
14
). No differences were observed between aerobic and anaerobic electrochemistry carried
out in a glove bag, indicating that the cluster is relatively stable in air and consistent with the
general long-term stability of B-family DNA polymerases (
35
). The redox couple observed
was attributed to the [4Fe4S]
3+/2+
based on the fact that Pol
δ
is HiPIP-like, being EPR-
silent unless oxidized (
3
). In addition, the electrochemical signal is similar to the DNA
glycosylases EndoIII and MutY, in which the identity of the couple has been established by
EPR (
13
,
24
).
We could obtain quantifiable signals at lower concentrations by adding dNTPs and Mg
2+
to
enhance protein association with the DNA. To prevent degradation of the DNA substrate by
the 3
-5
exonuclease activity of WT Pol
δ
, we turned to the exonuclease-deficient mutant
Pol
δ
DV (D520V) for these experiments (
31
). At 113 ± 5 mV vs NHE, the midpoint
potential of Pol
δ
DV is indistinguishable from WT (Figure 1, S3). By adding 80 μM dATP
(the incoming nucleotide), 8.0 mM MgAc
2
, and 5.0 μM PCNA, we were able to see signals
with Pol
δ
concentrations as low as 500 nM (Figure 1). Under these conditions, the
maximum CV peak areas were 6.9 ± 1 nC and −7.5 ± 1 nC for the reductive and oxidative
peaks at a scan rate of 100 mV/s.
The midpoint potential of Pol
δ
is within the HiPIP potential regime, but it is slightly higher
than the 65–95 mV vs NHE reported for DNA-bound repair proteins (
13
,
28
). We used the
well-studied
E. coli
BER glycosylase EndoIII (
28
) to determine if the measured potential is
truly distinct or the result of different buffer conditions. In our standard phosphate buffer (20
mM sodium phosphate, pH 7.5, 100 mM NaCl, 1 mM EDTA, 10% glycerol v/v), the
EndoIII midpoint potential is ~80 mV versus NHE (
28
). However, when we exchanged
EndoIII into Pol
δ
storage buffer, CV and SQWV carried out with 140 μM or 1.5 μM
EndoIII result in a midpoint potential of 113 ± 3 mV versus NHE, which is indistinguishable
Bartels et al.
Page 7
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
from Pol
δ
(Figure S2b). UV-visible spectra confirm the stability of EndoIII in Pol
δ
buffer
(Figure S2a). The observed increase in EndoIII potential is most likely due to the
significantly higher ionic strength of Pol
δ
storage buffer (350 mM NaAc in Pol
δ
buffer vs
150 mM NaCl in EndoIII buffer) (
36
,
37
). In any case, the fact that the EndoIII and Pol
δ
redox potentials are indistinguishable under identical buffer conditions supports the assertion
that the [4Fe4S] cluster resides in the same narrow potential regime in both proteins.
Given that Pol
δ
ordinarily functions in complex with PCNA, we next asked what effect
PCNA might have on the electrochemical properties of Pol
δ
. In the absence of PCNA, the
midpoint potential is unaltered at 115 ± 8 mV versus NHE, but the signal was markedly
smaller, reaching a maximum CV peak area of 0.4 ± 0.1 nC for the reductive peak and −0.7
± 0.1 nC for the oxidative peak (Figure S3). The signal also decays more rapidly with PCNA
absent, suggesting lower polymerase stability in the absence of PCNA. To compare the
signals with and without PCNA more quantitatively, diffusion coefficients (D) under both
conditions were calculated using the Randles-Sevcik equation (
32
). At maximum signal size,
D was found to be 6.7 ± 3 × 10
−6
cm
2
· s
−1
with PCNA and dATP present, which is within
one order of magnitude of an estimate (6.0 × 10
−7
cm
2
· s
−1
) based on the Stokes-Einstein
equation. The difference between these values most likely reflects errors from the use of
multiple partial crystal structures to estimate the hydrodynamic radius. In the absence of
PCNA, D decreases to 1.2 ± 0.3 × 10
−7
cm
2
·s
−1
, which is on the same order of magnitude as
the value estimated from the Stokes-Einstein equation. These results are consistent with
differently shaped complexes, in agreement with the known elongate form of Pol
δ
alone in
solution and the more compact form expected when multiple subunits have engaged with
PCNA (
3
,
38
). To see if dNTPs contribute to the shape of PCNA-bound Pol
δ
, we prepared a
surface with Pol
δ
and PCNA but lacking dATP and Mg
2+
. Under these conditions, the
signal is comparable to that in the absence of PCNA, giving a D value of 2.2 ± 0.7 × 10
−7
cm
2
· s
−1
. Taken together, these results indicate that PCNA does not affect the potential of
Pol
δ
but is critical for effective DNA binding, and the entire complex is more likely to
remain DNA-bound when dNTPs are present.
To determine if the Pol
δ
signal is DNA-mediated, we carried out electrochemistry using
DNA containing either an abasic site or a CA mismatch 6 nucleotides from the thiolated end
(Figure 1c). In addition to base stack integrity, DNA-mediated signaling in proteins is
dependent on film morphology, with DNA-bound proteins capable of charge transport both
through the DNA bases and directly through the monolayer surface (Figure S4) (
28
). In
general, closely packed monolayers provide more DNA which may be less accessible to
large proteins, while the opposite is true of loosely packed monolayers. To address all of
these issues together, we prepared monolayers in the presence of 100 mM MgCl
2
to form
closely packed islands of DNA (30–50 pmol·cm
−2
) in parallel with standard loosely packed
monolayers (15–20 pmol·cm
−2
) formed without Mg
2+
(Figure S4) (
39
,
40
). For both
morphologies, half of each chip contained well-matched DNA, and the other half contained
either abasic or CA mismatch DNA (Figure S4a). In all experiments, we used 500 nM Pol
δ
DV in the presence of PCNA, dATP, and MgAc
2
.
Consistent with previous studies on EndoIII (
28
), Pol
δ
redox potentials are identical on both
film morphologies, supporting the assertion that all observed signals are from DNA-bound
Bartels et al.
Page 8
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
proteins (Figure S4). On closely packed DNA films, signal size is highly variable, but 46
± 33% signal attenuation as determined by SQWV was observed on abasic DNA (Figure
S4). No mismatch discrimination was observed, and even the abasic site discrimination
decreased over time as more protein diffused to the surface. In general, the signals on closely
packed monolayers are consistent with variable DNA accessibility and significant steric
hindrance causing protein-DNA complexes to lie flat on the surface (Figure S4). In contrast,
loosely packed monolayers show very consistent signal sizes and significant charge
attenuation on both DNAs containing abasic sites and CA-mismatches, reaching a maximum
of 44 ± 16% on abasic DNA and 46 ± 29% on CA-mismatch DNA after 2 hours of
incubation as measured by SQWV (Figure 1, Figure S4). Abasic site and mismatch
discrimination also remain stable over time, indicating that DNA accessibility and steric
hindrance are not significant problems on loosely packed DNA. Together, these results
confirm that Pol
δ
is capable of DNA-mediated signaling and emphasize the importance of
substrate accessibility when assessing CT by large protein complexes.
Activity Assays with Oxidized and Reduced Pol
δ
Having seen that DNA binding can activate Pol
δ
for redox activity, we next asked how the
cluster oxidation state might affect polymerase activity. As purified, Pol
δ
exists largely in
the [4Fe4S]
2+
state (
3
), so any assessment of activity differences would require extensive
oxidation to generate sufficient amounts of the [4Fe4S]
3+
cluster for comparison. To this
end, we turned to bulk electrolysis on DNA-modified electrodes, applying an oxidizing
potential of 0.412 V verses NHE for 15–20 minutes.
To prevent aerobic degradation of oxidized cluster, all experiments were performed in a
glove bag under a 95% N
2
/5% H
2
atmosphere.
To achieve high yields, bulk electrolysis is best done on an electrode with a relatively large
surface area. Multiplexed chips have many advantages for electrochemical characterization,
but only a single electrode can be addressed at a time and each sample in a quadrant is
distributed between 4 electrodes. To overcome these limitations, we switched to single gold
rod electrodes for these experiments. In order to characterize this system, a sample of
concentrated (150 μL of 2.74 μM) Pol
δ
was oxidized on a DNA-modified electrode in a
custom-made glass cell for several hours (Figure S5a, b). UV-visible spectra were taken
before and after electrolysis, after which the sample was then frozen for EPR in parallel with
untreated protein (Figure S5c, d). The [4Fe4S] cluster oxidation generally results in a broad
increase in UV-visible absorbance from 300–450 nm with a less distinct peak at 410 nm in
both the [4Fe4S]
3+
and [3Fe4S]
+
species (
41
43
). After bulk electrolysis, increased
absorbance from 300–400 nm, consistent with cluster oxidation, was indeed observed. No
significant increase in absorbance at 800 nm occurred after oxidation, and the 280 nm peak
associated with aromatic and thiolated amino acid residues remained distinct. From our own
observations, protein aggregation tends to generate a U-shaped curve with high absorbance
at 800 nm and a shallow, poorly defined peak at 280 nm, and the lack of these features in our
spectra indicate that oxidized Pol
δ
did not aggregate (Figure S5c). EPR signals are small as
a result of the low sample concentration, but clear signals at g = 2.08 and g = 2.02 are
present in the oxidized sample (Figure S5d). These signals are consistent with a combination
Bartels et al.
Page 9
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
of [4Fe4S]
3+
and [3Fe4S]
+
cluster oxidation products (
24
,
42
43
). A smaller signal at g =
2.02 was also present in the native sample, consistent with earlier reports of residual [3Fe4S]
+
cluster in untreated Pol
δ
(
3
). That some [3Fe4S]
+
cluster would occur upon oxidation is
not surprising, and similar results have been obtained for EndoIII and MutY (
13
,
24
). In
earlier studies, loss of iron was attributed in part to damage incurred upon freezing, which
may have also happened here. Furthermore, oxidized Pol
δ
was stored away from protective
DNA long enough to take a UV-visible spectrum prior to freezing. In any case, some
[4Fe4S]
3+
cluster was still observed, and the [3Fe4S]
+
cluster that did occur would have
formed as a degradation product of the [4Fe4S]
3+
cluster (
13
).
Because activity assays require only low nanomolar polymerase concentrations, bulk
electrolysis for these experiments was carried out with 190 nm Pol
δ
DV to minimize
sample waste. Oxidation yields under these conditions were higher, typically around 75 –
90% as determined from the total charge passed. After electrolysis, untreated or oxidized Pol
δ
DV was added directly to pre-made reaction mixes to a final concentration of 2 nM
(Figure 2a). When run out on an alkaline agarose gel, it is apparent that at early time points
less DNA synthesis was carried out by oxidized Pol
δ
(Figure 2b, S6a). DNA synthesis can
be more quantitatively compared for the oxidized versus untreated sample by dividing the
amount of frontier products (highest molecular weight major products) in the oxidized
sample by the amount present in untreated samples. Using this analysis, oxidized Pol
δ
at
60–80% yield forms only 30–50% as many large (~ 5 kb) DNA products as untreated Pol
δ
after 30 seconds (Figure 2c). Significantly, higher oxidation yields lead to lower activity
levels. In any case, this difference gradually decreases over the course of 10 minutes.
Regardless of oxidation state, no DNA synthesis occurs in samples lacking PCNA,
confirming that all observed DNA synthesis is processive (Figure S6c).
Reduction of the oxidized Pol
δ
stock by electrolysis at −0.188 V versus NHE effectively
restores DNA synthesis, reaching 90% of untreated levels at early time points (Figure 2b, c,
S6b). Critically, this result both confirms the reversibility of oxidative slowing and provides
support for the [4Fe4S]
3+
cluster as the major oxidation product. As mentioned earlier, the
reversible electrochemical signals are consistent with [4Fe4S]
3+/2+
cycling, but EPR
spectroscopy with oxidized Pol
δ
showed evidence of both [4Fe4S]
3+
and [3Fe4S]
+
products
in the sample, leading to some ambiguity. However, the nearly complete restoration of native
activity levels upon re-reduction would not be expected if most of the cluster had degraded
to the [3Fe4S]
+
state, supporting the [4Fe4S]
3+
cluster as the major oxidation product. These
combined results thus indicate that the [3Fe4S]
+
cluster seen by EPR likely forms after the
[4Fe4S]
3+
major product degrades over time in the absence of DNA; we have observed this
previously with sample freezing for
E. coli
EndoIII and MutY following chemical oxidation
(
13
,
24
).
To gain further insight into the effect of oxidation, reaction rates were estimated by
comparing frontier velocities. Velocities were calculated by dividing the amount of the
largest quantifiable band of DNA by time and the number of polymerase molecules present
(
44
). This method yields maximum rates of 118 ± 63 (SD, n = 7) nt/s per enzyme for
untreated Pol
δ
and 21 ± 27 (SD, n = 5) nt/s for oxidized Pol
δ
at 2 minutes. The rates
obtained for untreated Pol
δ
are consistent with previously published
in vitro
results (
33
),
Bartels et al.
Page 10
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
while the oxidized form is significantly slower. This calculation represents just an upper
estimate, as typical bulk electrolysis fails to oxidize around 20–30% of the enzyme; thus,
some of the DNA synthesis observed in oxidized samples can be attributed to the non-
oxidized population. Indeed, the comparable amounts of DNA synthesis observed after 5–10
minutes could have resulted from either slow oxidized polymerase catching up or
redistribution by of residual native Pol
δ
in the oxidized sample. Overall, it is clear from
these experiments that oxidation leads to a decrease in replication rate, but resolution on
alkaline agarose gels is insufficient to distinguish between complete stalling or dramatic
slowing of DNA synthesis.
To distinguish between stalling and slowing of DNA synthesis by oxidized Pol
δ
, reactions
were analyzed on 5% polyacrylamide gels to obtain increased resolution in the 30–1000
nucleotides range. In addition, DNA synthesis was limited to that of a single processive
cycle by the PCNA-Pol
δ
complex by adding heparin, which traps dissociated Pol
δ
(
33
).
Without the heparin trap, products up to 7 kb were observed, due to multiple processive
cycles of synthesis (Figure 2, S6). In order to visualize products at all sizes, reactions
containing heparin were divided in two, with half loaded onto a polyacrylamide gel and half
onto an alkaline agarose gel. With heparin present, alkaline agarose gels demonstrate a
severe limitation to DNA synthesis, with no products larger than ~1 kb observed on alkaline
agarose gels (data not shown). When these products are resolved on polyacrylamide gels, a
greater proportion of both very small (~primer length) and very large (~1 kb) products are
formed by untreated Pol
δ
, while the oxidized form generates more intermediate products
between 30 and 1000 nucleotides (Figure S7).
These results illustrate several important points about the effects of oxidation on Pol
δ
DNA
synthesis. First, they verify that the oxidized form remains active and does not completely
stall. Second, the relatively greater amounts of very small products and unextended primers
in reactions with native sample indicate greater susceptibility to dissociation from DNA and
trapping by heparin, while the native form that does associate produces longer products. In
contrast, oxidized Pol
δ
leaves fewer primers unextended or fully extended, instead making
more intermediate products between 150 bp and 1 kb. The greater proportion of extended
primers is consistent with tighter DNA binding after cluster oxidation, as has been observed
with both primase and DNA repair proteins (
16
,
27
). However, the slower procession
indicates that tighter binding impedes rapid procession, acting as a brake on PCNA-mediated
DNA synthesis. These experiments suggest that the similar activity levels observed on
alkaline agarose gels at time points beyond 5 minutes could be explained by either the
oxidized form gradually catching up or by redistribution of the residual native enzyme in the
sample. Regardless of the precise details, the overall impact of polymerase stalling, with a 6-
fold decrease in rate, would be significant on the timescale of S-phase; unperturbed yeast S-
phase lasts ~30 minutes, while using oxidized Pol
δ
moving at 20 nt/s to replicate the
lagging strand of the yeast genome would itself require 27 minutes (
45
47
).
Chemical Oxidation of Pol
δ
with Anthraquinone
Electrochemical oxidation provides clear advantages in estimating yields and re-reducing the
oxidized sample, but the use of chemical oxidants is much more common. Thus, we were
Bartels et al.
Page 11
J Am Chem Soc
. Author manuscript; available in PMC 2018 April 03.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript