TTT-Feedback Regulation of Transcriptional Termination by the Mammalian Circadian Clock PERIOD

Complex
The Inner Workings of a Clock
Eukaryotic circadian clocks are built at least in part on transcriptional feedback loops, but the
mechanisms underlying circadian feedback are poorly understood. Padmanabhan et al. (p. 599, published
online 5 July) explored the transcriptional feedback mechanism at the heart of the mammalian circadian
clock. The proteins PERIOD (PER) and CRYPTOCHROME suppress transcription of their own genes. PER
complexes do so in part by recruiting a histone deacetylase to promoters of clock genes. But PER is also
present on DNA in a complex with Senataxin, a helicase that functions in transcriptional termination.
Senataxin appears to be inhibited in the PER complex, thus inhibiting termination and further reducing the
rate of transcription.
Abstract
Eukaryotic circadian clocks are built on transcriptional feedback loops. In mammals, the PERIOD
(PER) and CRYPTOCHROME (CRY) proteins accumulate, form a large nuclear complex (PER complex),
and repress their own transcription. We found that mouse PER complexes included RNA helicases DDX5
and DHX9, active RNA polymerase II large subunit, Per and Cry pre-mRNAs, and SETX, a helicase that
promotes transcriptional termination. During circadian negative feedback, RNA polymerase II accumulated
near termination sites on Per and Cry genes but not on control genes. Recruitment of PER complexes to the
elongating polymerase at Per and Cry termination sites inhibited SETX action, impeding RNA polymerase II
release and thereby repressing transcriptional reinitiation. Circadian clock negative feedback thus includes
direct control of transcriptional termination.

Eukaryotic circadian clocks are based on transcriptional negative-feedback loops (1, 2). In mammals,
the three PER and two CRY proteins accumulate, form a large nuclear complex (3), and associate with the
dimeric transcription factor CLOCK-BMAL1 (4–7) at Per and Cry promoters (8),
repressing Per and Cry transcription (5–7, 9). PER complexes include the RNA-binding protein NONO
(non-POU domain containing, octamer binding protein) and the histone methyltransferase WDR5 (WD
repeat-containing protein 5), and they function in part by recruiting a SIN3 histone deacetylase complex to
clock gene promoters (8).
To explore the clock mechanism, we affinity-purified nuclear PER complexes (10) from livers of FH-
PER1 and PER2-FH (N-terminal Flag-hemagglutinin–PER1 and C-terminal Flag-hemagglutinin–PER2)
mice (8) obtained at circadian time 20 hours (CT20), the peak phase of PER negative feedback. Purified FH-
PER1 and PER2-FH samples contained many proteins not found in controls, and the copurifying proteins
appeared essentially identical in the two cases (Fig. 1A). In addition to known PER-associated proteins (8),
mass spectrometry revealed RNA helicases DDX5 (DEAD-box polypeptide 5) and DHX9 (DEAH-box
protein 9) in both samples. DDX5 and DHX9 are related to FRH (FREQUENCY-interacting helicase), a
component of the Neurospora clock (11, 12). DDX5 and DHX9 specifically coimmunoprecipitated with
endogenous PER2 from mouse tissue extracts (CT20) (Fig. 1B). When tissues were harvested at CT8, the
trough of PER2 abundance, little or no DDX5 or DHX9 was coimmunoprecipitated (Fig. 1C).
 Fig. 1 DDX5 and DHX9 are components of nuclear PER complexes from mouse liver (CT20). (A)
Silver-stained polyacrylamide gels showing proteins copurifying with tagged PER1 (FH-PER1) or tagged
PER2 (PER2-FH) and respective control purifications from wild-type littermates with only native PERs.
Molecular size markers (in kilodaltons) are at left. Proteins migrating as expected for some known PER-
associated proteins are indicated. (B) DDX5 and DHX9 are in a complex with endogenous PER2. Nuclear
extracts from mouse liver or lung (input) and PER2 or control IgG immunoprecipitates (IP) from the extracts
were probed with antibodies to PER2, DDX5, DHX9, and IgG light chain (IgG LC). (C)
Coimmunoprecipitation as in (B), performed on nuclear extracts of livers obtained at CT8 or CT20; IgG
heavy chain (IgG HC) served as loading control. (D) Density gradient sedimentation of affinity-purified
PER2-FH complexes from mouse liver nuclei (CT20). Top, immunoblots of gradient fractions (10 to 50%
Optiprep, left to right; larger complexes toward right) probed with antibody to FLAG (to monitor PER
complexes) and antibodies to DDX5 and DHX9. Position of marker protein is at the top. Bottom, purified
complexes identical to those above treated with ribonucleases A and T1 (+ RNAse) before centrifugation.
Expand for more
 
We depleted DDX5 or DHX9 from a circadian reporter fibroblast line (13) and
monitored Per1 transcription and clock function. Depletion of either resulted in a selective increase in
transcription of the Per1 gene and a shortening of circadian period (figs. S1 and S2). DDX5 and DHX9 are
thus important for PER feedback and clock function.
Density gradient centrifugation of highly purified PER complex (CT20) revealed that DDX5 and DHX9
are constituents of large PER complexes, which are enriched in low-mobility, modified PER2 proteins (Fig.
1D, top). Smaller complexes are likely dissociation products. Treatment of affinity-purified material with
ribonucleases before centrifugation resulted in a partial shift toward smaller molecular-size fractions of
PER2 and DHX9, particularly modified PER2 (Fig. 1D, bottom), indicating the presence of stabilizing
RNA.
These findings suggested a role for PER proteins in circadian feedback beyond orchestrating histone
deacetylation (8). DDX5 and DHX9 function in transcription and pre-mRNA processing (14). Both are
associated with elongating RNA polymerase II (15) and are components of the 3′ transcriptional termination
complex (16). Thus, circadian feedback might impinge on these processes.
We tested whether PER complex associated with RNA polymerase II and transcriptional elongation
factors during circadian feedback. Elongating RNA polymerase II large subunit (phosphorylated on C-
terminal domain serine-2, serine-5, or both), CDK9 (cyclin-dependent kinase–9), and CBP80 (nuclear cap-
binding protein, 80 kD) specifically coimmunoprecipitated with PER2 from liver nuclear extracts (CT20)
(Fig. 2, A and B, and fig. S3A). Density gradient analysis of highly purified PER2-FH complexes (as in Fig.
1A) indicated that these factors, along with known clock proteins, are components of large PER complexes
(Fig. 2C). None of the proteins were detected in gradient fractions when the affinity purification was
performed using control mice with untagged PER2 (fig. S3B), and the proteins did not cosediment when the
analysis was performed with unpurified extracts (fig. S3C), as expected if only a small fraction of these
proteins is in a PER complex.

Fig. 2 Elongating RNA polymerase II large subunit (A to C) and circadian target gene pre-mRNAs (D to
F) in PER complexes. (A) Coimmunoprecipitation of RNA polymerase II large subunit with endogenous
PER2. Nuclear extracts from liver obtained at CT20 (input) and proteins immunoprecipitated (IP) with
control IgG or antibody to PER2 were probed with antibodies to the indicated proteins. (B)
Coimmunoprecipitation of phosphorylated RNA polymerase II large subunit and associated transcriptional
elongation factors with endogenous PER2, as in (A). (C) Immunoblots of fractions from density gradient
sedimentation of affinity-purified PER2-FH complexes from mouse liver nuclei (as in Fig. 1D).
(D) Per and Cry pre-mRNAs, but not control pre-mRNAs, associated with nuclear PER complex from
mouse liver (CT20). Data show quantitative RT-PCR from RNA extracted from PER2 IP relative to control
IgG IP (mean ± SEM; n = 3 for each; cycle thresholds of 17 to 24; t test, one-
tailed: Per1, Per2, Cry2 each P < 0.01). No signals were detected when reverse transcriptase was omitted.
(E) Control identical to (D), except that tissue was obtained at CT8, when there is little or no PER complex.
(F) As in (E), except that starting material was PER2 or control IgG ChIP samples from mouse liver nuclear
extract. n.d., not detected (cycle threshold >26).
Expand for more
 
If PER proteins act on elongating polymerase, then PER complexes might contain
nascent Per and Cry pre-mRNAs but not arbitrary pre-mRNAs. We performed PER2 or immunoglobulin G
(IgG) control immunoprecipitations from liver nuclear extracts (CT20), purified RNA from the samples, and
assayed for pre-mRNAs by quantitative reverse transcription–polymerase chain reaction (RT-
PCR). Per1, Per2, and Cry2 (but not control) pre-mRNAs specifically coimmunoprecipitated with PER2
(Fig. 2D). We did not detect Per3 or Cry1 pre-mRNAs, perhaps indicating differences in PER function at
different target genes. None of the pre-mRNAs were detected in immunoprecipitations from tissue collected
at CT8, the trough of PER2 abundance (Fig. 2E). We obtained similar results from chromatin-associated
PER complexes (Fig. 2F). Thus, PER complexes selectively associate with active transcriptional machinery
on the Per1, Per2, and Cry2 genes during negative feedback.
If active RNA polymerase II is inhibited by PER action, then arrested polymerases might accumulate
within the Per1and Cry2 genes during circadian negative feedback. We prepared chromatin from mouse
livers obtained every 4 hours across a circadian cycle, performed RNA polymerase II chromatin
immunoprecipitations (ChIP), and monitored genome-wide RNA polymerase II occupancy by deep
sequencing. A circadian rhythm of the polymerase at the 5′ transcriptional start site and along the length of
clock genes was observed that was consistent with the onset of circadian transcription (Fig. 3). However,
during negative feedback, RNA polymerase II accumulated near the 5′ transcriptional start site and 3′
transcriptional termination site of the Per1 and Cry2 genes (Fig. 3, A and B, and figs. S4 and S5A), with the
3′ accumulation evident even when viewed on a mega–base pair scale (fig. S4). We did not detect a 3′
accumulation at any time on either the Bmal1 gene, which has a different mechanism of circadian regulation
(17) (Fig. 3C and fig. S5A), or on arbitrary genes scattered throughout the genome (fig. S5, B and C). The
results for Per1 were confirmed by ChIP-PCR, and we found coaccumulation of PER2 and DHX9 with RNA
polymerase II at the Per1 3′ site (Fig. 3, D and E). Because impaired transcriptional termination causes
accumulation of RNA polymerase II just downstream of 3′ termination sites and inhibits transcription (18),
these results raised the possibility that PER complexes might inhibit 3′ processing at
the Per1 and Cry2 genes.

Fig. 3 Accumulation of RNA polymerase II near 3′ termination sites of Per1 and Cry2 genes during

circadian negative feedback. (A to C) Circadian ChIP-sequencing (seq) profiles of RNA polymerase II on
the Per1, Cry2, and Bmal1 genes, respectively, in mouse liver. Top, quantitative RT-PCR pre-mRNA profiles
[relative to Hprt (hypoxanthine-guanine phosphoribosyltransferase) pre-mRNA] indicating the relative
transcription of the genes across the circadian cycle (mean ± SEM; n = 3 for each). Bottom, ChIP-seq
profiles showing RNA polymerase II occupancy (vertical traces) along the genes (5′ to 3′, left to right) at the
six indicated circadian times. DNA scale, lower right. At bottom, thick blocks represent exons; thin blocks,
untranslated regions. The 5′ transcriptional start and 3′ termination sites are centered within gray bars.
Arrows denote accumulation of RNA polymerase II near 3′ termination sites. (D and E) ChIP analysis of
RNA polymerase II, PER2, and DHX9 near Per1 5′ initiation site (D) or 3′ termination site (E) at indicated
circadian times (relative to control IgG ChIP).
Expand for more
 
Supporting this notion, we found that PER2 was associated at CT20, the trough of Per1 transcription,
with SETX (senataxin) (Fig. 4A), a helicase important for transcriptional termination of some genes (18).
After cleavage of the nascent transcript, unwinding of RNA-DNA duplexes by SETX at the 3′ termination
site permits the XRN2 nuclease to degrade the downstream 3′ RNA and thereby release the polymerase (18).
We readily detected XRN2 in a complex with SETX, but PER complex included SETX but no detectable
XRN2 (Fig. 4A), suggesting that PER complex might inhibit SETX activity, blocking subsequent processing
by XRN2. SETX bound at or near Per1 and Cry2 3′ termination sites in a temporal pattern that mirrored that
of RNA polymerase II, including an accumulation at CT22 (fig. S5D), suggesting that SETX functions in the
termination of both genes (19).

Fig. 4 PER complex associates with SETX and antagonizes its action in transcriptional termination of
the Per1gene. (A) Coimmunoprecipitations from mouse liver nuclei (CT 20) showing that a complex of
PER2 with SETX excludes XRN2. (B) Immunoblot showing depletion of SETX in a fibroblast cell line after
electroporation of SETX small interfering RNA (siRNA) compared to control siRNA. α-Tubulin, loading
control. (C) Quantitative RT-PCR assay showing abundance of 3′-read-through mRNAs of indicated genes
after depletion of SETX (normalized to control siRNA; mean ± SEM; n = 3 for each). (D) ChIP assay
(relative to control IgG ChIP) showing RNA polymerase II near 3′ termination sites of indicated genes after
depletion of SETX. (E) Circadian oscillations of bioluminescence in synchronized circadian reporter
fibroblasts after introduction of SETX siRNA (blue) or control siRNA (yellow). Traces from four (SETX
siRNA) or three (control) independent cultures are shown. (F to H) Quantitative RT-PCR assay showing
circadian profile of pre-mRNA and 3′-read-through product for Per1 or (I to K) a control gene in mouse
liver. (F, G, I, and J) relative to Hprt; mean ± SEM; n = 3 for each.
Expand for more
Inefficient termination produces increased “3′-read-through” of mRNAs (18). Depletion of SETX (Fig.
4B) caused an increase in 3′-read-through at the Per1 and Cry2 (but not Bmal1) genes (Fig. 4C) and an
accumulation of RNA polymerase II at or near the 3′ termination sites of the Per1 and Cry2 (but not Bmal1)
genes (Fig. 4D). In addition, it caused a marked loss of circadian amplitude (Fig. 4E), suggesting an
important role for SETX in the clock.
If PER complex inhibits SETX, then there should be a naturally occurring circadian rhythm of Per1 3′-
read-through in vivo, with the peak occurring at the trough of transcription. Indeed, we found
rhythmic Per1 3′-read-through in mouse liver that was antiphase to Per1 transcription (Fig. 4, F and G). At
the transcriptional trough, 3′-read-through RNAs accounted for at least a 10-fold higher proportion of
transcripts than at the peak (Fig. 4H). A control gene showed no such effect (Fig. 4, I to K).
Thus, Per1 transcriptional termination appears to be under circadian control in vivo.
 These results reveal an unsuspected mechanism of circadian clock negative feedback. Because of
coupling between initiation and termination (18, 20–22), inhibition of termination reduces the rate of
transcription (18, 20). During negative feedback, we observed accumulation of RNA polymerase II at the 5′
site as well as the 3′ site of the Per1 and Cry2genes (Fig. 4, A and B), perhaps reflecting reduced initiation
secondary to inhibited termination.
The mammalian PER complex thus has at least two actions in circadian feedback. It represses
transcription by recruiting a SIN3 histone deacetylase complex to clock gene promoters (8), and it inhibits
termination, likely by antagonizing the action of SETX at the 3′ termination site. Both processes contribute
to circadian Per1 gene repression, but one or the other could predominate at different target genes.
In Drosophila, chromatin-associated PER spreads far downstream of the transcription start site (23),
consistent with regulation of one or more postinitiation steps. By distributing circadian feedback repression
across two or more steps in transcription, the PER complex might offer the advantages of averaging out
noise and reducing the energetic costs of cyclical reversal of gene repression.
Acknowledgments
We thank M. Liu for technical assistance and J. Gray and M. Greenberg (Harvard Medical School) for
helpful comments. All data are tabulated in the paper and supplementary materials. This work was supported
by the G. Harold and Leila Y. Mathers Charitable Foundation (C.J.W.), a David Mahoney Fellowship (K.P.),
and an Edward R. and Anne G. Lefler Center fellowship (M.S.R.).