Article
The Pentose Phosphate Pathway Regulates the
Circadian Clock
Graphical Abstract
Highlights
d Pentose phosphate pathway regulates circadian oscillations
through NADPH metabolism
d Inhibition of pentose phosphate pathway remodels circadian
gene expression
d NRF2 connects redox oscillations to transcriptional rhythms
d Pentose phosphate pathway modulation alters rhythmic
behavior and tissue clocks
Authors
Guillaume Rey, Utham K. Valekunja,
Kevin A. Feeney, ..., Vidya Velagapudi,
John S. O’Neill, Akhilesh B. Reddy
Correspondence
areddy@cantab.net
In Brief
Here, Rey et al. identify the pentose
phosphate pathway (PPP), which
generates NADPH, as an importantRey et al., 2016, Cell Metabolism 24, 462–473
September 13, 2016 ª 2016 The Authors. Published by Elsevier Inc.
http://dx.doi.org/10.1016/j.cmet.2016.07.024regulator of redox and transcriptional
oscillations. Inhibition of this highly
conserved metabolic pathway affects
circadian rhythms in flies, mice, and
human cells.
Accession Numbers
GSE74439Current models of circadian clock control
emphasize transcriptional networks.
t
o
is
1,3
e
0
r
phosphate pathway (PPP) oxidize glucose to produce NADHof phylogenetically disparate organisms ranging from bacteria to
humans (Edgar et al., 2012). In a simple model of non-transcrip-
tional circadian oscillations, the red blood cell, oxidation cycles
would affect redox oscillations. To disrupt NADPH production,
we used 6-aminonicotinamide (6AN). This compound is metab-
olized into an analog of NADP+, thus competitively inhibitingoccur in association with robust circadian oscillations of the
core cellular reductants NADH and NADPH (O’Neill and Reddy,
2011).
the critical NADPH-producing enzymes 6-phosphogluconate
dehydrogenase (PGD) and glucose 6-phosphate dehydroge-
nase (G6PD) (Ko¨hler et al., 1970). Consistent with our hypothesis,NRF2. Thus, the PPP regulates circadian rhythms via
NADPH metabolism, suggesting a pivotal role for
NADPH availability in circadian timekeeping.
INTRODUCTION
Mammalian models of the circadian clock center on transcrip-
tion-translation feedback loop mechanisms, involving the core
transcription factors BMAL1 and CLOCK (Bass, 2012). However,
recent evidence has uncovered the existence of transcription-in-
dependent mechanisms of circadian timekeeping (Cho et al.,
2014; Nakajima et al., 2005; O’Neill and Reddy, 2011; O’Neill
et al., 2011). These likely preceded the existence of transcrip-
tional oscillations during evolution, as highlighted by rhythms in
the oxidation and reduction of peroxiredoxin proteins in a range
RESULTS
Inhibition of the PPP Alters Circadian Redox and
Transcriptional Oscillations
In red blood cells, peroxiredoxin oxidation rhythms resonate
with NADPH oscillations (O’Neill and Reddy, 2011). NADPH
powers intracellular redox defense and is used by the peroxir-
edoxin system during its catalytic cycle to remove harmful
reactive oxygen species (Wood et al., 2003). We therefore
measured NADPH accumulation in human osteosarcoma
(U2OS) cells, an established and robust cellular clock model
(Liu et al., 2008), and found similar redox oscillations to those
seen in red blood cells previously (O’Neill and Reddy, 2011)
(Figure 1B).
Given that the PPP is a major source of NADPH in the cell (Fan
et al., 2014), we hypothesized that inhibiting its metabolic fluxThe Pentose Phosphate Pa
Regulates the Circadian Cl
Guillaume Rey,1 Utham K. Valekunja,1 Kevin A. Feeney,1,3 L
Laura Ansel-Bollepalli,1 Vidya Velagapudi,2 John S. O’Neill,
1University of Cambridge Metabolic Research Laboratories, Wellcom
Cambridge CB2 0QQ, UK
2Metabolomics Unit, Institute for Molecular Medicine Finland (FIMM),
3Present address: Cell Biology Division, MRC Laboratory of Molecula
4Lead Contact
*Correspondence: areddy@cantab.net
http://dx.doi.org/10.1016/j.cmet.2016.07.024
SUMMARY
The circadian clock is a ubiquitous timekeeping sys-
tem that organizes the behavior and physiology of or-
ganisms over the day and night. Current models rely
on transcriptional networks that coordinate circadian
gene expression of thousands of transcripts. How-
ever, recent studies have uncovered phylogeneti-
cally conserved redox rhythms that can occur inde-
pendently of transcriptional cycles. Here we identify
the pentose phosphate pathway (PPP), a critical
source of the redox cofactor NADPH, as an important
regulator of redox and transcriptional oscillations.
Our results show that genetic and pharmacological
inhibition of the PPP prolongs the period of circadian
rhythms in human cells, mouse tissues, and fruit flies.
These metabolic manipulations also cause a remod-
eling of circadian gene expression programs that in-
volves the circadian transcription factors BMAL1 and
CLOCK, and the redox-sensitive transcription factor462 Cell Metabolism 24, 462–473, September 13, 2016 ª 2016 The A
This is an open access article under the CC BY license (http://creativeand NADPH, respectively (Figure 1A). These pathways are com-
mon to most aerobic organisms and produce an important frac-
tion of the cellular pool of NAD(P)H (Fan et al., 2014). Since the
peroxiredoxin oxidation cycle is directly influenced by the avail-
ability of NADPH (Wood et al., 2003), we hypothesized that these
cellular reduction pathways might regulate redox and transcrip-
tional oscillations in nucleated cells.
Using a combination of pharmacologic and genetic ap-
proaches, we found that inhibition of the PPP altered circa-
dian rhythms in human cells. We observed similar effects in
mouse tissues, and we also found that PPP inhibition affected
the pattern of rhythmic behavior in Drosophila. Our study indi-
cates that the interplay between redox and transcriptional cy-
cles relies on the circadian transcription factors BMAL1/
CLOCK and the redox-sensitive transcription factor NRF2.
Moreover, we identify the histone acetyltransferase P300 as a
redox-dependent modulator of BMAL1/CLOCK transactivation
ability.Cell Metabolism
Article
hway
ck
a Wulund,1 Nikolay B. Milev,1 Alessandra Stangherlin,1
and Akhilesh B. Reddy1,4,*
Trust-MRC Institute of Metabolic Science, Addenbrooke’s Hospital,
0290 Helsinki, Finland
Biology, Francis Crick Avenue, Cambridge CB2 0QH, UK
In central carbon metabolism, glycolysis and the pentoseuthors. Published by Elsevier Inc.
commons.org/licenses/by/4.0/).
HNADPH
Glucose
A B
NADP6AN treatment prolonged the period of NADPH oscillation to

30 hr (Figure 1B). We next measured peroxiredoxin oxidation
in U2OS cells and found that these rhythms were similarly
affected by inhibition of the PPP (Figures 1C, 1D, S1A, and
S1B, available online), indicating that the availability of NADPH
regulates circadian redox oscillations. 6AN treatment indeed
drove the NADP+:NADPH redox poise in favor of oxidation by
decreasing NADPH by 
50%, consistent with its expected ef-
fect (Figure 1E). In contrast, NAD+:NADH ratio remained un-
Phase Response Curve
24 36 48
-12
-6
0
6
12
Ph
as
e 
di
ff e
re
nc
e 
(h
)CTRL 6AN
Time of treatment (h)
24 48 72 120 144 168 192
400
600
800
1000
6AN
CTRL
Time (h)
NADH
Glycolysis
ATP
Pyruvate
Oxidative
Phosphorylation
Pentose
Phosphate
Pathway
C
Dose Wash
22
24
26
28
30
6AN
CTRL***
*
Pe
rio
d 
(h
)
NADP+:NADPH
CTRL 6AN
0.0
0.5
1.0
1.5
2.0
R
el
at
iv
e 
le
ve
ls
***
FE
D
PRDX-SO2/3
monomer
ACTB
0 12 24 36 48 60
Time (h)
PRDX-SO2/3
monomer
ACTB
Control (DMSO)
6AN
0 12 24 36 48
-0.4
-0.2
0.0
0.2
0.4
Time (h)
R
el
at
iv
e 
Ab
un
da
nc
e 
(A
U
)
G
H I
20 
20 
J
Ti
m
e 
of
 tr
ea
tm
en
t (
h)
24 48 72
24
36
30
42
24 48 72
24
36
30
42
1
0
-1
40
40
6AN
CTRL
0.6
0.8
1.0
1.2
1.4
0 12 24 36 48 60
0.6
0.8
1.0
1.2
Time (h)
N
or
m
al
is
ed
 P
R
D
X
-S
O
2/
3 
m
on
om
er
CTRL
6AN
Time (h) Time (h)
Wash off
R
aw
 b
io
lu
m
in
es
ce
nc
e 
(A
U
)
ment, removal of the
(5 mM 6AN, 25.65 ±
We further validated t
repress PPP activity
down expression of
and PGD in Bmal1:lu
with pharmacological
in the NADP+:NADPH
period lengthening for
Cell Metaboglycolysis, the pentose phosphate pathway (PPP),
and oxidative phosphorylation in mitochondria.
(B) NADPH levels in Bmal1:luc U2OS cells treated
with 5 mM 6-aminonicotinamide (6AN) versus
control (DMSO) for 2 consecutive days (mean ±
SEM, n = 3–4).
(C) Representative immunoblots showing over-
oxidized peroxiredoxin (PRDX-SO2/3) monomers
with loading controls (b-actin, ACTB) for Bmal1:luc
U2OS cells treated with 5 mM 6AN versus control
(DMSO). Molecular weights (kDa) shown on right
side of blots.
(D) Quantification by densitometry of immunoblots
from (C). Values were normalized to the average for
each blot (mean ± SEM, n = 3).
(E) NADP+:NADPH ratio of cells treated with 6AN
(mean ± SEM, n = 3–4; two-tailed Student’s t test;
***p < 0.001)., baseline corrected
Figure 1. The PPP Regulates Redox and
Transcriptional Oscillations in Human Cells
(A) Schematic of glucose metabolism showing(F) Bioluminescence traces for Bmal1:luc U2OS
cells treated with 5 mM 6AN versus control
(DMSO), followed by wash off after 96 hr. (mean
values shown, n = 3–6).
(G) Quantifications of the period length from (F)
before and after wash off (mean ± SEM, n = 3–6;
two-tailed Student’s t test; ***p < 0.001, *p < 0.05).
(H and I) Heatmaps showing bioluminescence
traces for Per2:luc U2OS cells treated at the indi-
cated time points with 5 mM 6AN (I) or control
(DMSO) (H) until the end of the experiment. Each
row represents a different time of treatment.
(J) Phase-response curve showing the phase shifts
caused by treatment with 6AN compared to control
(DMSO) at different time of the day (mean ± SEM,
n = 3–6).
changed (Figure S1C) and treatment with
6AN did not acutely affect glycolysis or
mitochondrial respiration rates (Figures
S1D and S1E).
We then investigated the effect of
PPP inhibition on transcriptional oscilla-
tions using U2OS cells stably expressing
the Bmal1:luciferase (Bmal1:luc) reporter
construct (Liu et al., 2008). Treatment of
Bmal1:luc cells with 6AN caused a strong
and reversible effect on transcriptional os-
cillations. Oscillation period was length-
ened by 3 hr (Figures 1F, 1G, and S1F;
5 mM 6AN, 28.48 ± 0.16 hr versus control,
25.50 ± 0.05 hr), and after 96 hr of treat-
drug restored almost normal oscillations
0.10 hr versus control, 24.97 ± 0.21 hr).
his effect by using a genetic approach to
. We used RNA interference to knock
the NADPH-producing enzymes G6PD
c cells (Figures S2A–S2F). In agreement
manipulations, we observed an increase
redox ratio (Figure S2D) and a significant
both genes (Figure S2E).
lism 24, 462–473, September 13, 2016 463
+Total NAD NADPCTRL
24 48 72 120 144 168 192
400
600
800
1000
1200 DHEA
CTRL
Time (h)
A
C
B
D E
Pe
rio
d  
(h
)
NADP+:NADPH
Wash off
R
aw
 b
io
lu
m
in
es
ce
nc
e 
(A
U
)Having shown that tonic inhibition of the PPP modulated the
period of redox and transcriptional oscillations, we tested
whether such metabolic perturbation could also reset the phase
of circadian oscillations in a time-of-day-dependent manner. To
this end, we administered 6AN treatment around the clock and
assessed the phase of oscillations following treatment, gener-
ating a ‘‘phase-response curve.’’ Inhibiting the PPP had a strong
resetting effect, inducing large phase advances or delays in
rhythms depending on the time of day when the treatment
started (Figures 1H–1J). Together, these results implicate the
PPP as a regulator of two key facets of circadian pacemaker
function (period and phase of oscillation).
The PPP Affects Circadian Oscillations via NADPH
Metabolism
Since there is cellular interconversion of NAD+ and its phosphor-
ylated form (NADP+), we next investigated if the effects of PPP
perturbation could involve this pathway. This is important
because NAD+ metabolism forms a feedback loop with the
core circuitry of the circadian transcriptional network (Nakahata
et al., 2009; Ramsey et al., 2009), and therefore changes in NAD+
0.0
0.5
1.0
1.5
R
el
at
iv
e 
le
ve
ls
***
0.0
0.5
1.0
1.5
2.0
2.5
R
el
at
iv
e 
le
ve
ls
DHEA
6AN
6AN+NMN
**
***
**
**
DHEA
6AN
6AN+NMN
24 48 72 96 120
0
100
200
300
400
Time (h)
24
26
28
30
***
*** *
Pe
ri o
d 
l e
n
gt
h 
(h)
CTRL DHEA
0.0
0.5
1.0
1.5
2.0
R
el
at
iv
e 
le
ve
ls
**
F G
CTRL
6AN
6AN+NMN
B
as
el
in
e 
su
bt
ra
ct
ed
bi
ol
um
in
es
ce
nc
e 
(A
U
)
ure 2C). In contrast to
accumulation of total
on period does not d
the change in NADP+
determined conditions
in 6AN-treated cells, u
a precursor of NAD+ (F
restored close to norm
was a rescue of the a
period was not abolis
fore suggested that i
circadian oscillations
nisms, and that the
metabolism.
In addition to its re
biosynthetic function,
der to globally assess
performedmetabolom
PPP inhibitors (Figure
perturbations in the le
correlated between th
464 Cell Metabolism 24, 462–473, September 13, 2016treated cells; mean ± SEM, n = 3–4; two-tailed
Student’s t test; ***p < 0.001, **p < 0.01).
(E) Treatment with 500 mM NMN does not
restore NADP+:NADPH ratio to normal levels (two-
tailed Student’s t test, control [DMSO] versus
treated cells; mean ± SEM, n = 3–4, ***p < 0.001,
**p < 0.01).
(F) Bioluminescence traces for Bmal1:luc U2OS
cells treated with 6AN, or 6AN and NMN, versus
control (DMSO) (mean values shown, n = 8).
(G) Quantifications of the period length from (F):NADPH CTRL
Dose Wash
22
24
26
28
30
DHEA
CTRL
***
Figure 2. Manipulation of the PPP Affects
Circadian Oscillations through NADPH
(A) Bioluminescence traces for Bmal1:luc U2OS
cells treated with 50 mM dehydroepiandrosterone
(DHEA) versus control (DMSO), followed by wash
off after 96 hr. (mean values shown, n = 3–6)
(B) Quantifications of the period length from (A)
before and after wash off (mean ± SEM, n = 3–6;
two-tailed Student’s t test; ***p < 0.001).
(C) NADP+:NADPH ratio of cells treated with DHEA
(mean ± SEM, n = 3–4; two-tailed Student’s t test;
**p < 0.01).
(D) Treatment of Bmal1:luc U2OS cells with 5 mM
6AN decreases the levels of total NAD, while in-
cubation with 50 mM DHEA has no effect. NAD
levels in presence of 6AN can be restored by
addition of 500 mM NMN (control [DMSO] versus(two-tailed Student’s t test; mean ± SEM, n = 8,
***p < 0.001, *p < 0.05).
might potentially contribute to the period
phenotype seen with PPP inhibition.
Therefore, we tested the effect of
dehydroepiandrosterone (DHEA), a non-
competitive inhibitor of G6PD (Raineri
and Levy, 1970), and again found a
reversible period lengthening (Figures
2A, 2B, and S3A–S3D) and an increase
in the NADP+:NADPH redox ratio (Fig-
6AN treatment, DHEA did not affect the
NAD (Figure 2D), showing that the effect
epend on NAD levels and is specific to
:NADPH redox ratio. Importantly, we also
under whichwe could rescue NAD+ levels
sing nicotinamide mononucleotide (NMN),
igures 2D and 2E). When NAD levels were
al with NMN in the presence of 6AN, there
mplitude of oscillations, but the effect on
hed (Figures 2F, 2G, and S3E). This there-
nhibition of the PPP differentially affects
through direct and indirect redox mecha-
prolonged period is specific to NADPH
dox role, the PPP is a key contributor to
especially for nucleic acid synthesis. In or-
the effect of inhibition by 6AN or DHEA, we
ics profiling of U2OS cells treated with the
S3F; Table S3). We observed only mild
vels of 90 metabolites, and the changes
e 6AN and DHEA treatments (Figure S3G).
A B CTRL
Pearson’s rAlthough the levels of ribose-5-phosphate (R5P), which is impor-
tant for nucleotide synthesis, were slightly decreased in the both
conditions, the levels of nucleotides and nucleosides were
largely unchanged, indicating that the non-oxidative branch of
the PPP, downstream of the NADPH-producing enzymes, was
sufficient to provide substrates for synthetic pathways. Thus,
the effects of both 6AN and DHEA appear specific to the oxida-
tive (NADPH-producing) branch of the PPP, without a significant
impact on its biosynthetic functions.
NR1D1
0
10
20
30
40
50
NR1D2
24 36 48 60 72
0
10
20
30
40
PER3
24 36 48 60 72
0
5
10
15
20
25
BMAL1
0
5
10
15
DBP
24 36 48 60 72
0
10
20
30
40
TEF
0
5
10
15
20
25
6ANCTRL
Time (h) Time (h)Time (h)
C
CTRL
D
CT24
CT28
CT32
CT36
CT40
CT44
CT48
CT52
CT56
CT60
CT64
CT68
C
T2
4
C
T2
8
C
T3
2
C
T3
6
C
T4
0
C
T4
4
C
T4
8
C
T5
2
C
T5
6
C
T6
0
C
T6
4
C
T6
8
CT24
CT28
CT32
CT36
CT40
CT44
CT48
CT52
CT56
CT60
CT64
CT68
C
T2
4
C
T2
8
C
T3
2
C
T3
6
C
T4
0
C
T4
4
C
T4
8
C
T5
2
C
T5
6
C
T6
0
C
T6
4
C
T6
8
6AN
C
TR
L
6A
N
414 453
26
E
F
G
0 8 16 24
24
Time (h)
36 48 60
0
40
80
0 8 16 24
Phase (h)Phase (h)
m
R
N
A 
ex
pr
es
si
on
 (F
P
K
M
)
24
Time (h)
36 48 60
CTRL 6AN
Fr
eq
ue
nc
y
Fr
eq
ue
nc
y
p=0.002
0.95
1.00
CTRL 6AN
H
0
40
80
P
ha
se
 d
iff
er
en
ce
 6
A
N
 v
s.
 D
M
S
O
 (h
)
21−
6−
0
6
21
Circadian
genes
Clock
genes
BMAL1 DBP NR1D1 NR1D2 PER3 TEF
0.0
0.5
1.0
1.5
2.0
siCTRL
siG6PD
***
***
** ***
** ***R
el
at
iv
e 
m
R
N
A 
ex
pr
es
si
on
Pe
rio
d 
(h
)
CTRL 6AN CTRL 6AN
Circadian
genes
Clock
genes
p = 1.6*10-5
p = 2.8*10-4
20
24
28
32
**
**
treatment did not g
observed high correl
samples (Pearson co
time points; Figure 3A
bolism caused a profo
(Figures 3B–3D). Using
mark, 2014), we dete
the control and 6AN c
common transcripts (
Cell Metaboby RNA-seq in Bmal1:luc U2OS cells incubated
with 5 mM 6AN or control (DMSO). The heatmap
shows the Pearson’s correlation coefficient be-
tween time points for log-transformed fragments
per kilobase of transcript per million (FPKM) of the
14,686 expressed transcripts.Bmal1:luc cells were
synchronized with a dexamethasone shock and
total RNA was collected at the indicated time
points.
(B) Heatmap representation of the temporal accu-
mulation of mRNA for circadian transcripts in the
6AN (453) and control (414) conditions. The RAIN
algorithm (Thaben and Westermark, 2014) was6AN Figure 3. PPP Inhibition Remodels Circa-
dian Gene Expression
(A) Time course of mRNA expression determinedused to detect circadian transcripts (p % 0.01) in
each dataset.
(C) Overlap between the rhythmic transcripts de-
tected in the 6AN and control conditions (Fisher
test on contingency table, p = 0.002).
(D) Phase histogram of rhythmic transcripts shown
in (B).
(E) Boxplot representation of period length for 6AN
and control mRNA profiles for circadian transcripts
(RAIN algorithm, p % 0.01) and clock gene tran-
scripts (list of 20 well-described circadian genes;
Table S4) (Wilcoxon rank-sum test, with p values as
shown).
(F) Boxplot representation of phase differences
between 6AN and control mRNA profiles for
circadian transcripts (detected in the 6AN or con-
trol condition with RAIN algorithm, p % 0.01) and
clock gene transcripts (Kuiper’s one-sample test of
uniformity; **p < 0.01).
(G) Profiles of mRNA accumulations for the six
clock genes that are detected as circadian in both
conditions.
(H) mRNA accumulation of clock gene transcripts
in Bmal1:luc U2OS cells following siRNA knock-
down with 50 nM G6PD or control siRNA (negative
control #1) (mean ± SEM, n = 3; two-tailed Stu-
dent’s t test; ***p < 0.001, **p < 0.01).
Remodeling of Circadian Gene
Expression by NADPH Metabolism
How are perturbations in redox oscilla-
tions transduced into alterations in circa-
dian gene expression and, ultimately, to
organism behavior? To probe this, we
performed time course analyses of
U2OS cells and determined their gene
expression profiles by RNA sequencing
(RNA-seq) (Figures 3A and S4A). 6AN
lobally affect the transcriptome, as we
ation between control and 6AN-treated
rrelation coefficient >0.94 between all
). However, perturbation of NADPH meta-
und change in circadian gene expression
the RAIN algorithm (Thaben and Wester-
cted 414 and 453 circadian transcripts in
ondition, respectively (Figure 3B), with 26
Figure 3C). We validated these analyses
lism 24, 462–473, September 13, 2016 465
using two other algorithms, Fisher test (Rey et al., 2011) and
ARSER (Yang and Su, 2010), and found a considerable overlap,
as 147 and 169 genes were detected by the three methods in the
control and 6AN condition, respectively (Figure S4B). Gene
ontology (GO) analysis of rhythmic transcripts revealed that
genes involved in metabolic processes were enriched in both
conditions, while GO annotations related to circadian rhythms
were highly enriched in the 6AN condition (Figure S4C). Our
results thus indicate that perturbation of the PPP was able
to extensively remodel circadian gene expression, as high-
lighted by the altered phase distribution of mRNA expression
(Figure 3D).
This led us to investigate further how the period and phase of
circadian transcripts were changed following PPP inhibition. We
found that the median period of oscillations was increased by
treatment with 6AN (Figure 3E; control versus 6AN, 23.2 hr
versus 24.1 hr). Interestingly, the effect on clock genes was
especially pronounced (control versus 6AN, 21.8 hr versus
27.2 hr) but only marginally contributed to the shift in the period
distributions of all circadian transcripts (Figure S4D). Similarly,
we computed the distribution of phase differences between
6AN and control conditions. We observed a phase delay for
both circadian and clock gene sets (Figures 3F and S4E), again
with a stronger effect on clock genes. Accordingly, NR1D1,
NR1D2, TEF, DBP, BMAL1, and PER3, the six clock genes that
are rhythmic in both conditions, displayed prolonged periods
and phase delays in their mRNA accumulation profiles (Fig-
ure 3G). Most other clock genes (Table S4) had similar effects
on circadian gene expression, even if they were not necessarily
detected as statistically rhythmic (Figure S4F). In order to vali-
date the effect of PPP inhibition on circadian gene expression,
we silenced the expression of the enzyme G6PD by small inter-
fering RNA (siRNA) knockdown. In agreement with treatment
with 6AN, we observed a perturbation of the circadian gene
network, since the expression of clock genes was severely dis-
rupted (Figure 3H). Therefore, inhibition of the PPP remodels
circadian expression by changing the period and phase of circa-
dian transcripts, with an effect especially prominent on clock
genes.
The Circadian Transcription Factors BMAL1 and CLOCK
Are Activated by a Change in Redox Environment
Since the core circadian transcription factors BMAL1 and
CLOCK regulate the expression of most of the clock genes, we
hypothesized that perturbation in circadian gene expression
may involve a change in BMAL1/CLOCK DNA-binding activity.
This response could indeed result from an altered NADP+:
NADPH (or NAD+:NADH) ratio, since these dinucleotides have
been reported to affect the binding affinity of several circadian
PAS-domain transcription factors in vitro (Rutter et al., 2001).
We therefore performed chromatin immunoprecipitation fol-
lowed by sequencing (ChIP-seq) in Bmal1:luc U2OS cells to
delineate genome-wide binding patterns of these transcriptionfactors. We found that the number of shared BMAL1/CLOCK
genomic binding sites increased from 147 to 439 (3-fold in-
crease) following 6AN treatment (Figures 4A and 4B). CLOCK
was mostly affected, since we observed a more than 4-fold
rise in genomic binding peaks (Figure 4C). Moreover, BMAL1
and CLOCK binding strengths significantly increased at 439
466 Cell Metabolism 24, 462–473, September 13, 2016shared peaks following PPP inhibition (Figure 4D), indicating
enhanced DNA-binding activity that is consistent with elevated
expression of several BMAL1/CLOCK targets upon 6AN treat-
ment (Figures 3G and S4F).
Increased DNA-binding activity of BMAL1/CLOCK was
accompanied by changes in chromatin state at their genomic
binding sites (Figures S5A and S5B). We measured two epige-
netic marks of transcriptionally active chromatin by ChIP-seq:
histone H3 lysine 9 acetylation (H3K9ac) and histone H3 lysine 4
trimethylation (H3K4me3) (Figures 4E and 4F). While H3K4me3
profiles remained unchanged, H3K9 showed a local increase
around BMAL1/CLOCK sites (Figure 4E). This effect was not
due to a widespread increase in H3K9 acetylation near active
promoters, as H3K9ac profiles around transcription start sites
(TSSs) of expressed genes were not affected (Figure 4F). More-
over, elevated H3K9 acetylation was specific to 6AN peaks, as
we did not observe similar effects at BMAL1/CLOCK peaks
from the control condition (Figure S5C). Notably, BMAL1/
CLOCK binding and H3K9 acetylation were associated with
rhythmic expression of nearby transcripts. Indeed, we found
that the fraction of rhythmic transcripts increased with the fold
change in BMAL1/CLOCK binding and H3K9 acetylation
following 6AN treatment (Figures S5D and S5E).
We next investigated the mechanism by which redox imbal-
ance could affect chromatin states. First, we excluded the
NAD+-dependent deacetylase SIRT1 as a mechanism driving
this change, since SIRT1–/– mouse embryonic fibroblasts
exposed to 6AN still exhibited alterations in clock gene mRNA
patterning (Figure S5F). Moreover, rescue of NAD+ levels with
NMN did not restore normal DNA-binding activity of BMAL1/
CLOCK or levels of H3K9ac (Figure S5G). These results indicate
that SIRT1 and other NAD+-dependent deacetylases, including
SIRT6, are not likely to significantly contribute to the chromatin
state changeswe sawwith redox perturbation. Therefore, we hy-
pothesized that the archetypal histone acetyltransferase P300
might mediate these effects, since it is able to form disulphide
bridges with the FOXO transcription factors by a redox-depen-
dent mechanism (Dansen et al., 2009) and has been shown to
interact with clock proteins (Etchegaray et al., 2003). We
measured P300 protein accumulation in the nucleus and
observed increased levels following 6AN treatment (Figures 4G
and 4H). Furthermore, ChIP analyses revealed that clock gene
loci exhibited elevated P300 binding upon PPP inhibition (Fig-
ure 4I), strongly implicating redox-dependent acetylation by
P300 at these genomic regions. Interestingly, increased P300
binding and H3K9ac were specific to direct BMAL1/CLOCK tar-
gets—those with mRNA expression in phase with DNA-binding
activity (Rey et al., 2011) (Figures 4I and 4J; Table S4). Further-
more, we investigated ifP300 knockdown by siRNA could antag-
onize the effect of 6AN on circadian oscillations. Consistent with
its role in activating circadian transcription, P300 knockdown
caused a strong decrease of the amplitude of circadian oscilla-
tions (Figure S5H). However, at low siRNA concentrations, wefound that P300 knockdown was able to partially reverse the
strong period lengthening effect of 6AN, as it reduced the period
difference to only 1 hr compared to control (Figure S5I). These
results thus indicate that PPP inhibition leads to a redox-depen-
dent activation of BMAL1/CLOCK that is mediated by the his-
tone acetyltransferase P300.
A C ENRF2 Signaling Links Changes in Redox Balance to
Circadian Gene Expression
Overlap between circadian and BMAL1/CLOCK-bound genes
was significant for the control, but not the 6AN, condition
(Fisher test; control, p < 1 3 103; 6AN, p = 0.05; Figure S6A),
suggesting that additional transcription factors were likely to
contribute to the remodeling of circadian gene expression.
D
BMAL1
CTRL
2314
CLOCK
6AN
289
N
um
be
r o
f t
ag
s 
pe
r p
ea
k
p<2e-16
p<2e-16
CTRL
6AN
BMAL1 CLOCK
BMAL1 CLOCK
14
7 
sh
ar
ed
 s
ite
s
43
9 
sh
ar
ed
 s
ite
s
BMAL1
6AN
1369
1577
CLOCK
CTRL
B
0
40
80
12
0
CTRL 6AN
CLOCK binding
0
40
80
BMAL1 binding
I
-100
0
10
20
30
N
or
m
al
is
ed
 re
ad
 n
um
be
r
0
5
10
15
N
or
m
al
is
ed
 re
ad
 n
um
be
r
B
%
 in
pu
t
CTRL
6AN
0.8
0.9
1.0
1.1
1.2
1.3
N
uc
le
ar
 P
30
0
Nuclear
CTRL 6AN
α-P300
α-U2AF65
*
220 
60 
G H
Figure 4. BMAL1/CLOCK Are Activated by Inhibition of the PPP
(A and B) BMAL1 and CLOCK ChIP-seq binding profiles around the 147 and
conditions (B). Bmal1:luc U2OS cells were treated with 5 mM 6AN or control (DM
(C) Venn diagram showing the overlap of ChIP-seq peaks for BMAL1 and CLOC
(D) Distributions of number of tags per peak in the 439 peaks shared between
(Kolmogorov-Smirnov test performed between the indicated distributions, with p
(E and F) Genomic profiles of H3K9ac (top) and H3K4me3 (bottom) densities aro
14,686 expressed transcripts (F).
(G) Immunoblot showing P300 nuclear accumulation in cells treated with 6AN or
shown on right side of blots.
(H) Densitometric quantification of blots from (G) (two-tailed Student’s t test, *p <
(I) ChIP followed by quantitative real-time PCR of P300 following 6AN treatment
(J) Distribution of fold changes in H3K9ac density (6AN versus control) for direct an
as shown).FConsistent with this observation, we found enriched DNA
motifs for other transcription factors in the 6AN condition (Fig-
ure 5A) and, in particular, a motif corresponding to the redox-
sensitive transcription factor NRF2 (Chorley et al., 2012)
(Figure S6B). NRF2-like motifs showed a positional correlation
with the canonical BMAL1/CLOCK binding motif (E-box) when
6AN-treated cells were assessed (Figure 5B). Importantly, we
J
Direct Indirect
−1
.0
0.
0
1.
0
Lo
g2
(fo
ld
ch
an
ge
)
H3K9Ac
p<0.005
BMAL1/CLOCK
targets
H3K4me3
0 -500 0 500 1000
Distance to peak center (bp)
H3K9ac
CTRL
6AN
H3K4me3
-1000 -500 0 500 1000
0
10
20
30
40
Distance to TSS (bp)
N
or
m
al
is
ed
 re
ad
 n
um
be
r
H3K9ac
0
5
10
15 CTRL
6AN
N
or
m
al
is
ed
 re
ad
 n
um
be
r
Expressed genesMAL1/CLOCK 6AN peaks
P300 ChIP
ACTB NR1D1 DBP PER1 CRY2
0.0
0.1
0.2
0.3
0.4 6AN
CTRL
*
**
***
439 BMAL1/CLOCK peaks bound, respectively, in the control (A) and 6AN
SO) and chromatin was extracted after 24 hr of incubation.
K. The total number of peaks for each set is given.
6AN BMAL1 and CLOCK ChIP-seq for BMAL1 (top) and CLOCK (bottom)
values as shown).
und BMAL1/CLOCK 6AN peaks (E) and transcription start sites (TSSs) of the
control (DMSO). U2AF65 is shown as loading control. Molecular weights (kDa)
0.05).
or control (DMSO) (mean ± SEM, n = 3; two-tailed Student’s t test, *p < 0.05).
d indirect BMAL1/CLOCK target genes (Wilcoxon rank-sum test, with p values
Cell Metabolism 24, 462–473, September 13, 2016 467
A Bobserved a significant increase in NRF2 nuclear accumulation
following 6AN treatment (Figures 5C and 5D), indicating that
PPP inhibition leads to the activation of NRF2. H3K9ac
genomic profiles around NRF2 ChIP-seq peaks (Chorley et al.,
2012) were not altered by 6AN treatment, indicating that
NRF2 activation is not associated with H3K9 acetylation
(Figure S6C).
We next investigated whether NRF2 could mediate the inter-
action between redox balance and circadian oscillations. First,
we observed that a significant fraction of circadian transcripts
in the control and 6AN conditions were NRF2 targets (44 and
E
G
Figure 5. NRF2 Mediates the Effect of Redox Perturbation on Circadia
(A) De novo HOMER motif analysis of the indicated sets of BMAL1/CLOCK peak
(B) Histogram of E-box and NRF2-like motif positions around BMAL1/CLOCK pe
(C) Immunoblot showing NRF2 nuclear accumulation in cells treated with 6AN or
shown on right side of blots.
(D) Densitometric quantification of blots from (C) (mean ± SEM, n = 3; two-tailed
(E) ChIP followed by quantitative real-time PCR of NRF2 following 6AN treatment
HMOX1, heme oxygenase 1.
(F) mRNA accumulation of clock gene transcripts in Bmal1:luc U2OS cells follow
Bmal1:luc cells were synchronized 72 hr after transfection with a dexamethasone
control (DMSO) (mean ± SEM, n = 3; two-tailed Student’s t test; ***p < 0.001, **p
(G and H) Bioluminescence recordings of Bmal1:luc U2OS cells transfected with
treatment at 1.25 mM (G) or control (DMSO) (H) (left; mean, n = 8). Quantifications
two-tailed Student’s t test; **p < 0.01).
468 Cell Metabolism 24, 462–473, September 13, 2016C D48 genes, respectively), implicating this redox transcription fac-
tor in the control of circadian gene expression (Figures S6D and
S6E). Two important NRF2 targets, glutathione reductase (GSR)
and thioredoxin reductase 1 (TXNRD1), which both use NADPH
as reducing agent for cellular redox defense, also displayed
rhythmic mRNA accumulation, even though they were not
statistically detected as circadian (Figure S6F). Importantly, the
circadian transcriptional repressor NR1D1 was among NRF2
targets, with inducible binding sites at its promoter and in its
first intron (Figure S6G) (Chorley et al., 2012). Accordingly, we
found that perturbation of the PPP caused an increase in NRF2
F
H
n Oscillations
s.
aks bound only the 6AN condition.
control (DMSO). U2AF65 is shown as loading control. Molecular weights (kDa)
Student’s t test, **p < 0.01).
or control (DMSO) (mean ± SEM, n = 3; one-tailed Student’s t test, *p < 0.05).
ing knockdown with 20 nM NRF2 siRNA or control (non-targeting siRNA #1).
shock, and total RNA was collected after 24 hr incubation with 5 mM 6AN or
< 0.01). NS, not statistically significant by t test.
20 nM NRF2 siRNA or control (non-targeting siRNA #1) combined with 6AN
of circadian period length of bioluminescence traces (right; mean ± SEM, n = 8;
Pe
ri o
d 
(h
)
Pe
rio
d 
(h
)CA
B
24 48 72 120 144 168
0
50000
100000
Time (h)
50000
100000
SCN
Liver Wash off
D
Wash off
De
tre
nd
ed
 b
io
lu
m
in
es
ce
nc
e 
(A
U
)
bi
ol
um
in
es
ce
nc
e 
(A
U
)DNA binding toNR1D1 and its known target gene heme oxygen-
ase 1 (HMOX1) (Figure 5E), suggesting that the PPP-dependent
activation of NRF2 could relay redox signals to the circadian
network through NR1D1.
In order to functionally validate the role of NRF2 in mediating
the effect of 6AN on circadian gene expression, we silenced
NRF2 expression using siRNA and measured the mRNA expres-
sion of several clock genes (Figures 5F and S6H). We found that
both NR1D1 and PER3 lost their responsiveness to 6AN
treatment whenNRF2was silenced, indicating thatNRF2 knock-
down can reverse the effects caused by inhibition of the PPP.
Bioluminescence recordings of Bmal1:luc U2OS cells confirmed
this hypothesis, since NRF2 silencing in 6AN-treated cells
reduced the period by 2 hr compared to control siRNA (Fig-
ure 5G; siCTRL, 30.6 ± 0.5 hr; siNRF2, 28.0 ± 0.2 hr). In contrast,
NRF2 silencing in control cells did not affect the period of oscil-
lations (Figure 5H; siCTRL, 26.6 ± 0.1 hr; siNRF2, 27.1 ± 0.2 hr).
We further validated this effect using DHEA and found again that
NRF2 silencing reversed the period lengthening caused by
inhibition of the PPP (Figures S6I and S6J). Our results thus iden-
tify NRF2 as a key connection between redox and circadian
oscillations.
0
Dose Wash off
22
24 48 72 120 144 168
Time (h)
Control 6AN
Pe
rio
d 
(h
)
22
24
26
28
30
32
34
10
20
0 24 48 72
Control
10
20
0
6AN
R
aw
 b
io
lu
m
in
es
ce
nc
e 
(A
U)
Time (h)
E F
De
tre
nd
ed
 
ganotypic slice cultur
and a key metaboli
(mPer2Luc) mice (Yoo
lengthening at the hig
26.6 ± 0.2 hr; liver, 2
0.2 hr and liver, 23.4 ±
(Figures 6A–6D, S7A,
after 4 days of incub
both tissues, illustrat
perturbations.
Given that an ensem
behavior due to dispe
within the tissue, we p
SCN slices to investig
cells. Similar to the eff
hibited a period leng
(500 mM 6AN, 27.5 ± 0
6E and S7C), togethe
Notably, we obtained
cell oscillations in Per2
S7E–S7G), and we di
morphology upon exp
Cell Metabodistribution of circadian period lengths in the 6AN
and control (DMSO) conditions (n > 100; two-tailed
Student’s t test, p < 1 3 1016).
Inhibition of the PPP Modulates
Circadian Oscillations in Mouse
Tissues
Having explored how inhibition of the PPP
leads to altered circadian gene expres-
sion, we set out to study the effect of
these metabolic perturbations on circa-
dian behavior, an important output of the
clockwork. Destruction of erythrocytes
(hemolysis) when the 6AN target G6PD is
deficient (Cappellini and Fiorelli, 2008)
indicated that an in vivo approach in live***
Dose Wash off
22
24
26
28
Control
250μM
500μM
24
26
28
Control
250μM
500μM**
NS
NS
Figure 6. PPP Perturbation Disrupts Circa-
dian Oscillations in Mouse Tissues
(A and B) Bioluminescence recordings of supra-
chiasmatic nuclei (SCN) (A) and liver slices (B) from
mPer2Luciferase (mPer2Luc) mice treated with the
indicated concentration of 6AN or control (DMSO)
(left), followed by wash off (right) with control
medium.
(C and D) Quantifications of the period length from
SCN (C) and liver (D) slices treated with 6AN or
control (DMSO) before and after wash off (mean ±
SEM, n = 3–7; two-tailed Student’s t test; 500 mM
versus control; ***p < 0.001, **p < 0.01). NS, not
statistically significant by t test.
(E) Single-cell bioluminescence traces of SCN
slices from mPer2Luc mice incubated either with
500 mM 6AN or control (DMSO) (n > 100).
(F) Violin plot representing the difference in themice would not yield meaningful results.
We therefore next analyzed the effects of
PPP inhibition on circadian rhythmicity in
primary tissues ex vivo. Treatment of or-
es of the suprachiasmatic nucleus (SCN)
c tissue, the liver, from mPer2Luciferase
et al., 2004) with 6AN elicited a period
hest concentration (500 mM 6AN; SCN,
6.0 ± 0.5 hr versus control SCN, 24.9 ±
0.4 hr), similar to the effect in human cells
and S7B). Importantly, removal of 6AN
ation restored a normal period length in
ing the reversible nature of the redox
ble of oscillators may exhibit comparable
rsion in the phase of individual oscillators
erformed single-cell imaging of mPer2Luc
ate the effect on period length in individual
ects at the population level, single cells ex-
thening of >2 hr when treated with 6AN
.2 hr versus control, 25.0 ± 0.1 hr) (Figures
r with a higher damping rate (Figure S7D).
similar results when measuring single-
:luc U2OS cells treated with 6AN (Figures
d not observe deleterious changes in cell
osure to 6AN or DHEA for several days
lism 24, 462–473, September 13, 2016 469
M
ed
ia
n 
ac
tiv
ity
 p
er
 3
0 
m
in
ut
e 
bi
n 
(A
U
)
Time (days)
0 1 2 3 4 5 6
1mM 6AN
0.1% DMSO
10mM 6AN
1% DMSO
15mM 6AN
1.5% DMSO
0
1
2
0
1
2
0
1
2
0
1
2
5mM 6AN
0.5% DMSO
Hours
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
Hours
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
Hours
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
Hours
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
01
:00
02
:00
03
:00
04
:00
05
:00
06
:00
07
:00
08
:00
09
:00
10
:00
11
:00
12
:00
13
:00
14
:00
15
:00
16
:00
17
:00
18
:00
19
:00
20
:00
21
:00
22
:00
23
:00
00
:00
1
3
5
7
1
3
5
7
1
3
5
7
1
3
5
7
D
ay
s
0mM 6AN
15mM 6AN
. % DMSO
% DMSO
A B
C
D
Circadian
Gene
Expression
CLOCK/BMAL1
Ox
NRF2
P300
SH
PRDXPRDX
SO2/3
Circadian
Gene
Expression
CLOCK/BMAL1Activity
Red
Circadian
Redox
Oscillations
Ox Red
ReducingOxidising
Activity
Glycolysis
Pentose
Phosphate
Pathway
NADPH
NADPHNADPH
NADPH
NADP+
NADPH NADPH
NADPH
NADPH
NADPH
NADPH
NADPH
NADPH
NADPH
NADPH
NADPH
NADPH
A
ut
oc
or
re
la
tio
n
-0.2
0.0
0.2
0.4
-0.2
0.0
0.2
0.4
-0.2
0.0
0.2
0.4
-0.2
0.0
0.2
0.4
12 18 24 30 36
Time (h)
2.3 h
Figure 7. PPP Inhibition Affects Behavioral Rhythms in Flies
(A) Median activity plots ofDrosophila melanogaster (Canton-S strain) behavioral activity with concentrations of 6AN ranging from 1 to 15mM in their usual growth
medium. As a control, DMSO was used at the specified concentration to directly match the concentration experienced with the 6AN dose.
(legend continued on next page)
470 Cell Metabolism 24, 462–473, September 13, 2016
(Figure S7H). This highlights the significant and specific effect
that perturbing central cellular metabolism has on cell-autono-
mous circadian oscillations.
Fly Behavioral Rhythms Respond to Redox Perturbation
Similarly to Mammals
In order to circumvent the systemic effects associated with 6AN
in mice, we measured the effect of PPP inhibition on Drosophila
melanogaster locomotor activity. Flies are ideal organisms to
study the effect of metabolic inhibitors because they lack red
blood cells, which are sensitive to such treatment when admin-
istered to rodents. In addition, their transcriptional clockwork is
largely similar in architecture to mammals, with orthologs of
BMAL1 and CLOCK driving gene expression of repressors
(Young and Kay, 2001). We therefore recorded locomotor
behavior of flies using a video recording system (Figure S7I).
Behavioral recordings of flies fed 1–15 mM 6AN in their agar
growth medium revealed a dose-dependent effect on behavioral
rhythms (Figure 7A). PPP inhibition caused only a mild reduction
in the amplitude of activity rhythms at all concentrations,
enabling us to measure their behavioral rhythms over the course
of several days. Treatment with 6AN lengthened the period up to
2.3 hr at the highest dose (15mM) (Figure 7B), and this effect was
visible in the locomotor activity of individual flies (Figures 7C and
S7J). These results thus show that inhibition of the PPP not only
affects cell-autonomous circadian oscillations but also complex
behavioral rhythms controlled by the circadian clock.
DISCUSSION
Models for circadian timekeeping in all species currently incor-
porate similar transcriptional mechanisms. However, each spe-
cies’ clock relies on a different set of clock genes in its timing
system, given that these are not evolutionarily conserved be-
tween kingdoms (Young and Kay, 2001). Recently, an alternative
type of circadian oscillation, the oxidation of peroxiredoxins, has
been reported in a diverse range of species (Edgar et al., 2012),
implying that redox oscillations could be a more fundamental
timekeeping mechanism. We set out to investigate how such
non-transcriptional oscillations may be connected to circadian
transcriptional rhythms. Our results demonstrate that manipula-
tion of the PPP, a key pathway in NADPH metabolism, affects
circadian oscillations in human cells, mouse tissues, and living
flies. Identification of the PPP as a modulator of redox oscilla-
tions indicates that the overoxidation pattern of peroxiredoxin
may be a reporter of more fundamental oscillations in the
formof NADPH rhythms (Figure 7D). It also suggests that NADPH
metabolism may be an important parameter in the generation of
circadian redox oscillations, in light of previous findings showing
similar NADPH rhythmicity in non-transcriptional models (O’Neill
and Reddy, 2011). Moreover, these redox rhythms may have
physiological importance, since several studies have described
rhythms in NADP+:NADPH ratio in rodents (Reddy and Rey,2014).
(B) Mean autocorrelation of activity plots highlights the period difference between
as a shaded gray area and autocorrelation values outside these boxes are signifi
(C) Representative actograms of individual flies following treatment with the indi
(D) Schematic showing how perturbation of the PPP regulates circadian redox aWe found that redox perturbations increased the DNA-binding
activity of BMAL1/CLOCK, which in turn led to profound qualita-
tive and quantitative changes in circadian gene expression. The
effects were especially prominent for clock genes, but inhibition
of the PPP also caused a switch in the set of output circadian
genes. Indeed, the sets of genes being rhythmic with or without
PPP inhibition diverged considerably. Perturbation of NADPH
metabolism also led to an increased density of histone H3K9
acetylation near BMAL1/CLOCK sites, indicating redox-depen-
dent chromatin remodeling. We showed that the redox-sensitive
histone acetyltransferase P300 accumulated in the nucleus after
PPP inhibition and subsequently displayed increased binding at
BMAL1/CLOCK sites. Interestingly, this effect was more pro-
nounced at direct BMAL1/CLOCK target genes—those with
mRNA expression corresponding to BMAL1/CLOCK binding in
mouse liver (Rey et al., 2011)—suggesting that P300 is mainly
associated with transcriptionally active BMAL1/CLOCK com-
plexes. This is consistent with the fact that the genome-wide
bindingof P300 is inphasewithBMAL1/CLOCKbinding inmouse
liver (Koike et al., 2012). Thus, our study indicates that P300 links
redox rhythms to circadian transcription by modulating BMAL1/
CLOCK transactivation ability in a redox-dependent fashion.
Our study also revealed the important role ofNRF2 in the inter-
play between redox and circadian oscillations. Previous studies
have shown clock-controlled activity of Nrf2 in the mouse lung
(Pekovic-Vaughan et al., 2014) and proposed that Nr1d1 may
respond to oxidative stress signals through an NRF2 binding
site in its promoter (Yang et al., 2014). Here we find that NRF2
and BMAL1/CLOCK have overlapping transcriptional regulatory
programs, likely through cooperative binding to common
genomic sites, and may therefore contribute to circadian tran-
scription, as suggested by the number of NRF2 target genes
rhythmically expressed. Moreover, our data strengthen the
notion that NR1D1 could integrate circadian and redox signals,
but most importantly reveal the role ofNRF2 as an important reg-
ulatory node between redox rhythms and circadian transcrip-
tional oscillations in nucleated cells. Indeed, we found that
NRF2 is necessary for relaying redox perturbation caused by in-
hibition of the PPP to the circadian clockwork. These findings will
be of great importance in building an integrated model of the
circadian clock that encompasses its transcriptional and meta-
bolic components. In addition, these results also provide a novel
molecular mechanism by which redox imbalance, as experi-
enced in cancer, cardiovascular disease, and neurodegenera-
tive disease, could lead to circadian disruption.
In conclusion, we show that the PPP is an important regulator
of circadian redox and transcriptional oscillations. We also iden-
tify P300 and NRF2 as two parallel mechanisms that connect
redox oscillations to BMAL1/CLOCK-mediated transcriptional
oscillations in nucleated cells. In a physiological context, the
PPP is a fundamental player in anabolic cellular processes and
is emerging as a determinant in cancer because of its role in
curbing oxidative stress (Masri et al., 2015; Patra and Hay,2014; Tsouko et al., 2014). Since the circadian transcriptional
6AN and DMSO conditions. The 95%confidence interval (white noise) is shown
cant at p < 0.05 (n = 24 male flies per group).
cated concentration of 6AN or control (DMSO).
nd transcriptional oscillations.
Cell Metabolism 24, 462–473, September 13, 2016 471
network rhythmically regulates over 40% of all protein-coding
genes in the body (Zhang et al., 2014), an implication of our re-
sults is that disruption of metabolic pathways as occurs in
many metabolic disorders and cancers could impact signifi-
cantly on tissue gene expression programs and associated
organ physiology via its effect on the clockwork.
EXPERIMENTAL PROCEDURES
Cell Culture and Bioluminescence Assays
Bmal1:lucU2OS and Per2:lucU2OS cells were a gift from Dr. Andrew Liu, Uni-
versity of Memphis (Liu et al., 2008). U2OS cells were cultured in standard con-
ditions. For bioluminescence recordings, U2OS cells were synchronized by
changing medium to ‘‘Air Medium’’ (Hastings et al., 2005). Bioluminescence
assays were performed at 37C using 12-well and 96-well plates in custom-
made bioluminescence recording systems (Cairn Research Ltd) composed
of a charge-coupled device (CCD) camera (Andor iKon-M 934) mounted on
the top of an Eppendorf Galaxy 170R CO2 incubator. Bioluminescence data
traces were analyzed using a modified version of the R script ‘‘CellulaRhythm’’
(Hirota et al., 2008).
Gel Electrophoresis and Immunoblotting
Bmal1:lucU2OS cells treated with 5mM6AN or control (DMSO) were synchro-
nized with a dexamethasone shock and lysed in 13 SDS sample buffer at the
indicated time points. NuPAGE Novex 10% Bis-Tris gradient gels were run
according to the manufacturer’s protocol with a nonreducing MES SDS buffer
system. Protein transfer to nitrocellulose for blotting was performed and
membranes were incubated in anti-PRDX-SO3 (LF-PA0004, Thermo Fisher
Scientific) or anti-ACTB (sc-47778, Santa Cruz) overnight at 4C. Immunoblot
signals were first normalized with loading control (actin) and then normalized to
the average for each replicate.
siRNA Transfections
For bioluminescence experiments, 90 mL cell suspension (0.5–13 105 cells per
mL) were seeded in 96-well plates. Cells were transfected with the indicated
siRNAs (see Table S1 for details) 20–24 hr after seeding using Lipofectamine
RNAiMAX (Life Technologies) according to manufacturer’s instructions. The
mediumof transfectedcellswaschanged to ‘‘AirMedium’’ for bioluminescence
recording 72 hr after transfection. When combined with drug experiments, sol-
vent (DMSO) was kept at a concentration of 0.25% for control and treatment
conditions. For gene expression analyses after siRNA knockdown, siRNA
transfections were performed as described above, except that they were per-
formed in 12-well plates, keeping the ratio between cell number and transfec-
tion reagent constant. Cells were synchronized with dexamethasone and
cultured in DMEM supplemented with 5 mM 6AN or control (DMSO) 72 hr after
transfection. After 24 hr incubation, RNA was extracted with TRI-Reagent in
triplicate and purified with Direct-zol RNA MiniPrep kit (Zymo Research).
RNA-Seq
For mRNA expression time course, Bmal1:luc U2OS cells were synchronized
with dexamethasone (Figure S4A) and cultured in DMEM as described above,
supplementedwith 5mM6AN or amatched amount of DMSO (0.5%) as a con-
trol. At the time points indicated in the main text, RNA was extracted with TRI-
Reagent in triplicate and purified with Direct-zol RNA MiniPrep kit (Zymo
Research). RNA-seq libraries were prepared as described in the detailed pro-
tocol provided in Supplemental Experimental Procedures. Sequencing using a
HiSeq platform with single-end 50 bp reads and subsequent quality filtering of
reads was performed according to manufacturer’s instructions (Illumina).
ChIP-SeqChIP was performed on Bmal1:lucU2OS using amodified version of an estab-
lished protocol (Mortazavi et al., 2006) provided in Supplemental Experimental
Procedures. ChIP-seq libraries were prepared as described for RNA-seq sam-
ples, except that fragment size selection was performed after end repair using
AMPure XP Magnetic Beads. Sequencing using a HiSeq platform with paired-
end 101 bp reads and subsequent quality filtering of reads was performed
according to manufacturer’s instructions (Illumina).
472 Cell Metabolism 24, 462–473, September 13, 2016Nuclear Fractions
Nuclear fractions were prepared from Bmal1:luc U2OS cells treated with 6AN
or control (DMSO) for 24 hr using the NE-PER reagents (Thermo Fisher
Scientific) according to manufacturer instructions. Nuclear lysates were
diluted with denaturing LDS sample buffer (Invitrogen) with 50 mM TCEP
and heated to 70C for 10 min before loading on gels. Nuclear extracts were
analyzed by immunoblotting as described in the Supplemental Experimental
Procedures, except that NuPAGE Novex 4%–12% Bis-Tris gradient gels
were used. The following antibodies were used: anti-p300 (N-15), sc-584,
Santa Cruz; anti-NFE2L2, Antibody EP1808Y, OriGene Technologies; and
anti-U2AF65 U4758, Sigma.
Organotypic Slice Culture and Bioluminescence
All animal experimentation was licensed by the UK Home Office under the An-
imals (Scientific Procedures) Act 1986, and according to the European Parlia-
ment and Council of the European Union Directive 2010/63/EU. Local Ethical
Review was also conducted by the University of Cambridge. Prior to use in ex-
periments, animals were group housed in individually ventilated cages under a
12:12 light:dark (LD) cycle with food and water available ad libitum. SCN and
liver slices were extracted from 8- to 12-week-old adult mPer2Luc mice (Yoo
et al., 2004). Slices were cultured on a membrane (Merck Millipore,
PICM0RG50) in a sealed dish. Slices were then transferred to custom-imaging
incubators for whole-explant bioluminescence recording, or microscopes for
single-cell bioluminescence imaging. Whole-explant imaging of SCN and liver
slices was performed using an Andor iKon-M 934 cooled CCD camera
mounted CO2 incubator at 37
C. Single-cell images were recorded from
SCN slices placed into an Okolab stage-top heated chamber (37C) mounted
on an inverted Nikon Eclipse Ti-Emicroscope equippedwith an electron-multi-
plied CCD (EM-CCD) camera (Hamamatsu ImagEM 1K, C9100-14).
Fly Behavioral Assays
Wild-type Canton-S flies were bred and grown on standard yeast cornmeal
agar medium at 25C in 12 hr:12 hr LD cycles. For behavioral recording exper-
iments, individual flies were placed into wells of a 96-well plate following brief
exposure to CO2 anesthesia. Each well contained an equal volume of assay
medium (5% sucrose, 1% agar), supplemented with 6AN or DMSO (control)
at concentrations indicated in the main text. Although the concentrations of
drug were high in comparison to those used in our cell and tissues studies,
it is important to note that the Drosophila were ingesting agar dosed with the
drug and therefore received a much lower effective concentration. Using a
custom-made infrared video recording system, the locomotor activity of indi-
vidual 4- to 7-day-old flies was recorded in constant darkness (DD) following
2 days of entrainment in LD cycles (which were not recorded). The videos
were processed using Ethovision XT v10 software (Noldus) to quantify the
locomotor activity of the flies.
ACCESSION NUMBERS
The GEO (http://www.ncbi.nlm.nih.gov/geo/) accession number for the RNA-
seq and ChIP-seq data reported in this paper is GEO: GSE74439.
SUPPLEMENTAL INFORMATION
Supplemental Information includes Supplemental Experimental Procedures,
seven figures, and four tables and can be found with this article online at
http://dx.doi.org/10.1016/j.cmet.2016.07.024.
AUTHOR CONTRIBUTIONS
A.B.R. and G.R. designed and planned experiments with contributions fromJ.S.O. G.R., K.A.F., N.B.M., U.K.V., and A.S. performed cell experiments.
U.K.V. performed fruit fly experiments. G.R. and U.K.V. performed and
analyzed RNA-seq and ChIP-seq experiments. V.V. supervised the metabolo-
mics analyses and performed data analyses, and V.V. and G.R. analyzed re-
sults. L.W. and L.A.-B. performed mouse experiments. G.R., A.B.R., N.B.M.,
L.W., U.K.V., and V.V. analyzed the data. A.B.R. andG.R. wrote themanuscript
with contributions from all of the authors.
ACKNOWLEDGMENTS
A.B.R. acknowledges funding from the Wellcome Trust (100333/Z/12/Z,
100574/Z/12/Z), the European Research Council (ERC Starting Grant No.
281348, MetaCLOCK), EMBO Young Investigators Programme, the Lister
Institute of Preventive Medicine, and the Medical Research Council (MRC_
MC_UU_12012/5). J.S.O. is supported by the Medical Research Council
(MC_UP_1201/4) and the Wellcome Trust (093734/Z/10/Z). G.R. is supported
by an SNSF Postdoctoral Mobility Fellowship and an EMBO Long-Term
Fellowship. We thank A. Liu (University of Memphis) for stable Bmal:luc and
Per2:luc U2OS cells, L. Guarante (Massachusetts Institute of Technology)
for SIRT1–/– mouse embryonic fibroblasts, and N. Matthews and colleagues
(Francis Crick Institute, Advanced Sequencing Facility) for assistance with
high-throughput sequencing.
Received: February 20, 2016
Revised: May 24, 2016
Accepted: July 28, 2016
Published: August 18, 2016
REFERENCES
Bass, J. (2012). Circadian topology of metabolism. Nature 491, 348–356.
Cappellini, M.D., and Fiorelli, G. (2008). Glucose-6-phosphate dehydrogenase
deficiency. Lancet 371, 64–74.
Cho, C.-S., Yoon, H.J., Kim, J.Y., Woo, H.A., and Rhee, S.G. (2014). Circadian
rhythm of hyperoxidized peroxiredoxin II is determined by hemoglobin autox-
idation and the 20S proteasome in red blood cells. Proc. Natl. Acad. Sci. USA
111, 12043–12048.
Chorley, B.N., Campbell, M.R., Wang, X., Karaca, M., Sambandan, D.,
Bangura, F., Xue, P., Pi, J., Kleeberger, S.R., and Bell, D.A. (2012).
Identification of novel NRF2-regulated genes by ChIP-seq: influence on reti-
noid X receptor alpha. Nucleic Acids Res. 40, 7416–7429.
Dansen, T.B., Smits, L.M.M., van Triest, M.H., de Keizer, P.L.J., van Leenen,
D., Koerkamp, M.G., Szypowska, A., Meppelink, A., Brenkman, A.B., Yodoi,
J., et al. (2009). Redox-sensitive cysteines bridge p300/CBP-mediated acety-
lation and FoxO4 activity. Nat. Chem. Biol. 5, 664–672.
Edgar, R.S., Green, E.W., Zhao, Y., van Ooijen, G., Olmedo, M., Qin, X., Xu, Y.,
Pan, M., Valekunja, U.K., Feeney, K.A., et al. (2012). Peroxiredoxins are
conserved markers of circadian rhythms. Nature 485, 459–464.
Etchegaray, J.P., Lee, C., Wade, P.A., and Reppert, S.M. (2003). Rhythmic his-
tone acetylation underlies transcription in the mammalian circadian clock.
Nature 421, 177–182.
Fan, J., Ye, J., Kamphorst, J.J., Shlomi, T., Thompson, C.B., and Rabinowitz,
J.D. (2014). Quantitative flux analysis reveals folate-dependent NADPH pro-
duction. Nature 510, 298–302.
Hastings, M.H., Reddy, A.B., McMahon, D.G., and Maywood, E.S. (2005).
Analysis of circadian mechanisms in the suprachiasmatic nucleus by trans-
genesis and biolistic transfection. Methods Enzymol. 393, 579–592.
Hirota, T., Lewis,W.G., Liu, A.C., Lee, J.W., Schultz, P.G., and Kay, S.A. (2008).
A chemical biology approach reveals period shortening of the mammalian
circadian clock by specific inhibition of GSK-3beta. Proc. Natl. Acad. Sci.
USA 105, 20746–20751.
Ko¨hler, E., Barrach, H., and Neubert, D. (1970). Inhibition of NADP dependent
oxidoreductases by the 6-aminonicotinamide analogue of NADP. FEBS Lett.
6, 225–228.
Koike, N., Yoo, S.-H., Huang, H.-C., Kumar, V., Lee, C., Kim, T.-K., and
Takahashi, J.S. (2012). Transcriptional architecture and chromatin landscape
of the core circadian clock in mammals. Science 338, 349–354.Liu, A.C., Tran, H.G., Zhang, E.E., Priest, A.A., Welsh, D.K., and Kay, S.A.
(2008). Redundant function of REV-ERBalpha and beta and non-essential
role for Bmal1 cycling in transcriptional regulation of intracellular circadian
rhythms. PLoS Genet. 4, e1000023.
Masri, S., Kinouchi, K., and Sassone-Corsi, P. (2015). Circadian clocks, epige-
netics, and cancer. Curr. Opin. Oncol. 27, 50–56.Mortazavi, A., Leeper Thompson, E.C., Garcia, S.T., Myers, R.M., andWold, B.
(2006). Comparative genomics modeling of the NRSF/REST repressor
network: from single conserved sites to genome-wide repertoire. Genome
Res. 16, 1208–1221.
Nakahata, Y., Sahar, S., Astarita, G., Kaluzova, M., and Sassone-Corsi, P.
(2009). Circadian control of the NAD+ salvage pathway by CLOCK-SIRT1.
Science 324, 654–657.
Nakajima, M., Imai, K., Ito, H., Nishiwaki, T., Murayama, Y., Iwasaki, H.,
Oyama, T., and Kondo, T. (2005). Reconstitution of circadian oscillation of
cyanobacterial KaiC phosphorylation in vitro. Science 308, 414–415.
O’Neill, J.S., and Reddy, A.B. (2011). Circadian clocks in human red blood
cells. Nature 469, 498–503.
O’Neill, J.S., van Ooijen, G., Dixon, L.E., Troein, C., Corellou, F., Bouget, F.-Y.,
Reddy, A.B., and Millar, A.J. (2011). Circadian rhythms persist without tran-
scription in a eukaryote. Nature 469, 554–558.
Patra, K.C., and Hay, N. (2014). The pentose phosphate pathway and cancer.
Trends Biochem. Sci. 39, 347–354.
Pekovic-Vaughan, V., Gibbs, J., Yoshitane, H., Yang, N., Pathiranage, D., Guo,
B., Sagami, A., Taguchi, K., Bechtold, D., Loudon, A., et al. (2014). The circa-
dian clock regulates rhythmic activation of the NRF2/glutathione-mediated
antioxidant defense pathway to modulate pulmonary fibrosis. Genes Dev.
28, 548–560.
Raineri, R., and Levy, H.R. (1970). On the specificity of steroid interaction with
mammary glucose 6-phosphate dehydrogenase. Biochemistry 9, 2233–2243.
Ramsey, K.M., Yoshino, J., Brace, C.S., Abrassart, D., Kobayashi, Y.,
Marcheva, B., Hong, H.-K., Chong, J.L., Buhr, E.D., Lee, C., et al. (2009).
Circadian clock feedback cycle through NAMPT-mediated NAD+ biosyn-
thesis. Science 324, 651–654.
Reddy, A.B., and Rey, G. (2014). Metabolic and nontranscriptional circadian
clocks: eukaryotes. Annu. Rev. Biochem. 83, 165–189.
Rey, G., Cesbron, F., Rougemont, J., Reinke, H., Brunner, M., and Naef, F.
(2011). Genome-wide and phase-specific DNA-binding rhythms of BMAL1
control circadian output functions in mouse liver. PLoS Biol. 9, e1000595.
Rutter, J., Reick, M., Wu, L.C., and McKnight, S.L. (2001). Regulation of clock
and NPAS2 DNA binding by the redox state of NAD cofactors. Science 293,
510–514.
Thaben, P.F., and Westermark, P.O. (2014). Detecting rhythms in time series
with RAIN. J. Biol. Rhythms 29, 391–400.
Tsouko, E., Khan, A.S., White, M.A., Han, J.J., Shi, Y., Merchant, F.A., Sharpe,
M.A., Xin, L., and Frigo, D.E. (2014). Regulation of the pentose phosphate
pathway by an androgen receptor-mTOR-mediated mechanism and its role
in prostate cancer cell growth. Oncogenesis 3, e103.
Wood, Z.A., Schro¨der, E., Robin Harris, J., and Poole, L.B. (2003). Structure,
mechanism and regulation of peroxiredoxins. Trends Biochem. Sci. 28, 32–40.
Yang, R., and Su, Z. (2010). Analyzing circadian expression data by harmonic
regression based on autoregressive spectral estimation. Bioinformatics 26,
i168–i174.
Yang, G., Wright, C.J., Hinson, M.D., Fernando, A.P., Sengupta, S., Biswas,
C., La, P., and Dennery, P.A. (2014). Oxidative stress and inflammation modu-
late Rev-erba signaling in the neonatal lung and affect circadian rhythmicity.
Antioxid. Redox Signal. 21, 17–32.
Yoo, S.-H., Yamazaki, S., Lowrey, P.L., Shimomura, K., Ko, C.H., Buhr, E.D.,
Siepka, S.M., Hong, H.-K., Oh, W.J., Yoo, O.J., et al. (2004). PERIOD2:
LUCIFERASE real-time reporting of circadian dynamics reveals persistent
circadian oscillations in mouse peripheral tissues. Proc. Natl. Acad. Sci.
USA 101, 5339–5346.Young, M.W., and Kay, S.A. (2001). Time zones: a comparative genetics of
circadian clocks. Nat. Rev. Genet. 2, 702–715.
Zhang, R., Lahens, N.F., Ballance, H.I., Hughes, M.E., and Hogenesch, J.B.
(2014). A circadian gene expression atlas in mammals: implications for biology
and medicine. Proc. Natl. Acad. Sci. USA 111, 16219–16224.
Cell Metabolism 24, 462–473, September 13, 2016 473