Interlocked feedback loops contribute to the
robustness of the Neurospora circadian clock
Ping Cheng, Yuhong Yang, and Yi Liu*
Department of Physiology, University of Texas Southwestern Medical Center, Dallas, TX 75390-9040
Communicated by Steven L. McKnight, University of Texas Southwestern Medical Center at Dallas, Dallas, TX, April 5, 2001 (received for review
January 15, 2001)
Interlocked feedback loops may represent a common feature
among the regulatory systems controlling circadian rhythms. The
Neurospora circadian feedback loops involve white collar-1 (wc-1),
wc-2, and frequency (frq) genes. We show that WC-1 and WC-2
proteins activate the transcription of frq gene, whereas FRQ
protein plays dual roles: repressing its own transcription, probably
by interacting with the WC-1?WC-2 complex, and activating the
expression of both WC proteins. Thus, they form two interlocked
feedback loops: one negative and one positive. We establish the
physiological significance of the interlocked positive feedback
loops by showing that the levels of WC-1 and WC-2 determine the
robustness and stability of the clock. Our data demonstrate that
oscillation and the more robust the overt rhythms. Our data also
show that, despite considerable changes in the levels of WC-1,
WC-2, and FRQ, the period of the clock has been limited to a small
also important for determining the circadian period length of the
iological, behavioral, cellular, and biochemical activities. At the
molecular level, a common theme of various circadian oscillators
is a network of positive and negative elements that form the core
of the oscillators that establish the negative feedback loops
generating the basic circadian rhythmicity (1). In a simple view,
every oscillator has both positive and negative elements to
comprise the feedback loop. The positive elements of the loop
activate the expression of the negative elements, whereas the
negative elements feedback to block their own activation by the
positive elements. The identified positive elements in Neuro-
spora, Drosophila, and mammals are all PAS domain-containing
transcription factors (2–7). These factors form heterodimeric
complexes and activate the transcription of the negative ele-
ments in each system, and the protein products of these negative
elements feedback to inhibit their own expression (8–12).
Recently, studies in Neurospora, Drosophila, and mammals
have significantly furthered our view of the negative feedback
nature of the circadian oscillator with the identification of
interlocked feedback loops (13–15). In each system, the negative
elements of the oscillator have been found to activate the
expression of one of the positive elements. Thus, the negative
and the positive elements form another positive feedback loop
interlocked with the negative feedback loop. The similarity of
such an arrangement in different clock systems suggests that it
may be a common aspect in the eukaryotic circadian oscillators.
However, evidence to support the physiological significance of
the positive feedback loops is still lacking.
In the Neurospora frq-wc based circadian feedback loops, the
two PAS domain-containing transcription factors, WHITE
COLLAR-1 (WC-1) and WC-2, form heterodimeric complexes
and function as the positive components (2, 16), whereas two
forms of the FRQ protein are the negative elements (9, 17).
WC-1 and WC-2 proteins have two different roles in Neurospora
n eukaryotic and certain prokaryotic organisms, circadian
clocks are responsible for controlling a wide variety of phys-
(2, 18, 19). First, they are both essential for the light induction
of gene expression of all known light-induced genes, including
light induction of frq and light resetting of the clock. Second, in
constant darkness, WC-1 and WC-2 are required for the acti-
vation of frq and the generation of circadian rhythms. In mutants
protein are very low, and the clock is not running at normal
The frequency (frq) gene, the first known Neurospora clock
gene, plays an essential role in the frq-wc based circadian
feedback loop (1, 9, 20, 21). Both forms of FRQ protein (large
and small FRQ forms) can negatively feedback to repress their
own transcription (9, 17). This repression is probably achieved
through protein–protein interactions between FRQ and the WC
proteins (22). Recently, Lee et al. (15) revealed that FRQ also
positively regulates the level of WC-1, leading to rhythmic
expression of WC-1 through a posttranscriptional mechanism,
and thereby they form a second positive feedback loop. In this
study, we demonstrate that both WC proteins are positively
regulated by FRQ and that the levels of WC-1 and WC-2
determine the robustness and stability of the Neurospora clock.
Thus, the positive feedback loops formed by FRQ and the WCs
are important for maintaining robustness and stability of the
Materials and Methods
Strains and Culture Conditions. The bd, a (wild-type clock) strain
was used as the wild-type strain in this study. 93-4 (his-3, bd,
frq10), 87-12 (wild-type clock, his-3, bdA), 161-8 (his-3, bd, wc-1),
and 241-23 (his-3, bd, wc-2?) were the host strains for various
his-3 targeting constructs used (2, 23, 24). In frq10, qa-FRQ;
wc-1?, qa-WC-1; wc-2?, qa-WC-2; wc-1?, qa-WC-1; and wc-2?,
qa-WC-2 strains, qa-FRQ and qa-WC constructs were trans-
formed into various host strains at the his-3 locus (9, 25). For
each transformation, the transformants were first checked by
Western blot analysis, and the positive transformants were
examined by race tube assays.
Liquid culture conditions were the same as previously de-
scribed (21), except a lower glucose concentration was used in
the media (1? Vogel’s?0.1% glucose?0.17% arginine). Such a
concentration of glucose was used because the induction of the
qa-2 promoter is significantly inhibited by the normally used 2%
glucose in the media (9). For rhythmic experiments, the Neu-
rospora cultures were moved from LL to DD at time 0 and were
harvested in constant darkness at the indicated time (hours).
Race tube assay media contains 1? Vogel’s?0% glucose?0.17%
arginine?50 ng/ml biotin?1.5% agar. Densitometric analysis of
race tubes and calculations of period length were performed as
Abbreviation: QA, quinic acid.
*To whom reprint requests should be addressed. E-mail: Yi.Liu@UTsouthwestern.edu.
The publication costs of this article were defrayed in part by page charge payment. This
article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
§1734 solely to indicate this fact.
June 19, 2001 ?
vol. 98 ?
no. 13 www.pnas.org?cgi?doi?10.1073?pnas.121170298
Plasmids. To make the his-3 targeting qa-WC constructs, a PCR
fragment containing the promoter of qa-2 (9) was inserted into
the BglII site of the pDE3dBH. The resulting plasmid
pDE3dBH-qa-2 was the parental vector for pqa-WC constructs.
pqa-WC vectors were made by inserting the PCR fragment
containing the entire WC-1 ORF (from nucleotide ?18 to the
SnaBI site of the wc-1 locus) or the StuI–SmaI fragment con-
taining the entire wc-2 ORF into the SmaI site of pDE3dBH-
qa-2 (18, 19). The resulting pqa-WC constructs were targeted by
transformation to the his-3 locus of the host strains as previously
Protein and RNA Analyses. Protein extraction, Western blot anal-
ysis, and immunoprecipitation assay are as previously described
(22, 26). Equal amounts of total protein (40–100 ?g) were
loaded in each protein lane, and after the blots were developed
by chemiluminescence (ECL, Amersham Pharmacia), they were
stained by amido black to verify equal loading of protein (17).
RNA extraction and Northern blot analysis were performed as
previously described (20). Equal amounts of total RNA (40 ?g)
were loaded onto agarose gels for electrophoresis, and the gels
were blotted and probed with RNA probe specific for frq, wc-1,
or wc-2 (18–20).
The Levels of Both WC-1 and WC-2 Proteins Are Positively Regulated
by FRQ. To investigate the role of FRQ in regulating the expres-
sion of WC-2, we first examined the level of WC-2 in frq null
strains in constant light (LL) and in constant darkness (DD). As
shown in Fig. 1 A and B, the level of WC-2 was significantly lower
under both conditions in the frq null strains than in the wild-type,
suggesting that, like WC-1, the expression of WC-2 is also
positively regulated by FRQ. The various FRQ bands seen on
Western blots are the results of protein phosphorylation and two
alternatively translated FRQ forms (26). To demonstrate the
positive role of FRQ on both WC-1 and WC-2, their protein
levels were examined in a strain (frq10, qa-FRQ) in which the
endogenous frq locus is deleted (23) and the FRQ ORF is under
the control of the quinic acid (qa-2)-inducible promoter (9). In
the presence of 1 ? 10?2M quinic acid (QA), the levels of both
WC-1 and WC-2 were significantly increased in this strain,
whereas their amounts were unchanged by the addition of QA in
the frq10strain (Fig. 1C). Together, these results demonstrate
that FRQ positively regulates the expression of both WC-1 and
To understand how FRQ regulates wc-1 and wc-2, mRNA
levels of wc-1 and wc-2 were examined in frq10, wc-1?(making a
truncated WC-1 protein), and wc-2?(wc-2 knock-out) strains
(Fig. 1D). Our data indicate that the mechanisms of FRQ
controlling wc-1 and wc-2 are different. In agreement with
previous reports, the level of frq mRNA is extremely low in the
wc mutant strains (2), and the regulation of wc-1 by frq appears
to be posttranscriptional because the level of wc-1 mRNA in frq10
is similar to that of the wild-type (15). In addition, wc-1 does not
appear to autoregulate its own mRNA level in constant condi-
tions, as indicated by the similar wc-1 level in the wc-1 mutant
indicating that the positive regulation of FRQ on WC-2 is at least
partially achieved by increasing the abundance of wc-2 mRNA.
Because FRQ levels oscillate daily, one would predict that the
because of the positive effect of FRQ (15). However, we found
no robust rhythm of wc-2 mRNA (data not shown) or WC-2
protein in constant darkness under the conditions examined.
Despite the lack of a clear rhythm in WC-2 protein level (see
Figs. 3 and 4), in some of our rhythmic experiments, we found
that WC-2 level, like that of WC-1, showed a trough around
DD14–18 (Fig. 1E). Therefore, it is also possible that WC-2
oscillates with a very low amplitude. Similar to what has been
reported before (15), the level of WC-1 was rhythmic with a
trough around DD14–18 and a peak around DD 24–29. How-
ever, the amplitude of the WC-1 rhythm has been variable in our
hands, especially with the second trough of the rhythm (compare
WC-1 rhythms in this and other figures in this study).
WC-1 and WC-2 Positively Regulate frq.In wc-1 or wc-2 mutants, the
levels of frq mRNA and FRQ protein are very low (Fig. 1D), and
the clock is not running at normal conditions (2), suggesting that
both WC-1 and WC-2 are the positive elements of the frq
of WC-1 and WC-2 on frq expression, we made strains (wc-1?,
qa-WC-1 and wc-2?, qa-WC-2) in which the ORF of either WC-1
or WC-2 is under the control of the quinic acid-inducible
promoter (qa-2) at the his-3 locus in either the wc-1?or wc-2?
strain (2, 24). In the presence of QA, the qa-2 promoter results
in the constitutive induction of the controlled gene (9, 27). At
DD14, 1 ? 10?2M QA was added, and protein expressions were
monitored afterward. As can be seen from the control lanes
(?QA) in Fig. 2 A and B, there is little WC-1 or WC-2 expressed
without the inducer (the low-level expression was because of the
basal activity of the qa-2 promoter, also see Figs. 3 and 4). The
level of FRQ in the wc-1?, qa-WC-1 strain was also very low
(only the extensive phosphorylated forms of FRQ can be seen)
may be rhythmic (note the change in FRQ phosphorylation
status). These data indicate that even a low level of WC-2 is able
to support a significant level of FRQ expression, suggesting that
WC-2 is not a limiting factor for frq expression in a wild-type
strain. After the addition of QA, the level of WC-1 or WC-2 was
increased immediately (peaking after 2 h), suggesting that there
(B). Total protein extracts were prepared from the wild-type and the frq null
strains (frq9and frq10) grown in LL or in DD (at different times). Western blot
analysis was performed by using WC-1, WC-2, or FRQ antisera (22). Represen-
tative results from three independent experiments are shown. (C) Western
blot analysis shows that FRQ positively regulated the levels of WC-1 and WC-2
10?2M QA for several hours before being transferred into constant darkness
and were harvested at DD24. (D) Northern blot analysis reveals that frq
differentially regulates wc-1 and wc-2. Cultures were harvested at DD14.
Similar results were obtained for cultures harvested in LL. (E) Western blot
analysis shows that the WC-2 level does not fluctuate much in constant
darkness in the wild-type strain.
Both WC-1 and WC-2 are positively regulated by FRQ. The WC-2 level
Cheng et al.
June 19, 2001 ?
vol. 98 ?
no. 13 ?
was almost no delay between their transcription and translation
(Fig. 2 A and B). The induction of either WC-1 or WC-2 quickly
leads to the induction of the frq mRNA and FRQ protein. Note
that the low molecular weight FRQ forms (the newly synthesized
and little phosphorylated two alternatively translated FRQ
forms) started to appear in 2–4 h, and FRQ levels reached the
peak in 8 h. Interestingly, the level of frq RNA decreased after
about 6 h of induction, suggesting that the induction of WC-1 or
WC-2 in these strains started running of the clock.
Somewhat to our surprise, we found that the levels of WC-2
were comparable when with or without QA in the wc-1?,
qa-WC-1 strain. Two lines of evidence suggest that even the
low-level expression of FRQ in wc-1?strain is able to maintain
a normal WC-2 level. First, some FRQ is still expressed, albeit
at a low level, in wc-1?strain (Fig. 2C). Second, WC-1 does not
appear to negatively regulate WC-2 because the induction of
WC-1 in a frq?strain did not affect the level of WC-2 (Fig. 2D);
therefore, the normal WC-2 level in wc-1?strain is not because
of the absence of WC-1. These data also help to explain the fact
The Levels of WC-1 and WC-2 Determine the Robustness of the
Neurospora Circadian Conidiation Rhythm. What is the biological
significance of the positive role of FRQ on the expression of the
WCs? One hypothesis is that the positive regulation of FRQ on
WCs would promote the robustness and stability of the clock, as
more WC proteins will require more FRQ protein to block their
of FRQ. To test this hypothesis, we examined the circadian
conidiation rhythms by race tube assays in the wc-1?, qa-WC-1
and wc-2?, qa-WC-2 strains. Because the level of WC-1 or WC-2
can be induced to different levels in these strains by different QA
concentrations, the clock phenotypes under these conditions
should inform us about the importance of the WC levels on the
clock. For the wild-type strain, the presence of different con-
centrations of QA (?glucose media) had no influence on the
conidiation rhythms (Fig. 3A), and it also had no effect on the
shown). The broader conidiation peaks observed at 1 ? 10?2M
QA was because of the carbon source effects of the high
concentration of QA. This phenomenon is known to occur when
the concentration of the carbon source, such as glucose, sucrose,
or QA, is high, as more conidia and aerial hyphae are made, but
the clock is not affected by these conditions. As shown in Fig. 3B
(Upper), the wc-1?, qa-WC-1 strain was arrhythmic when QA
concentration was less than 1 ? 10?7M. At 1 ? 10?6M QA, it
showed one conidiation peak in the first day in DD, but the
rhythm could not be sustained. When QA was at 1 ? 10?5M,
clear circadian conidiation rhythms could be observed, but the
very broad conidiation peaks suggest that the amplitude of
rhythm was low. As the concentration of QA further increased,
the conidiation rhythms became more robust, as indicated by
more defined conidiation peaks. These conidiation rhythm data
were in agreement with the induction of WC-1 and FRQ in the
cell (Fig. 3B, Lower). When QA was less than 1 ? 10?5M, the
levels of WC-1 and FRQ were low. At 1 ? 10?5M QA, a
significant amount of WC-1 was induced, leading to a higher
FRQ expression level. At 1 ? 10?4M QA, the levels of both
WC-1 and FRQ were comparable with those in the wild-type
strain. At higher QA concentrations, their levels were signifi-
cantly higher than those of the wild-type. In contrast to FRQ
and WC-1, the WC-2 level changed little, suggesting that
even low-level expression of FRQ can support normal WC-2
The wc-2?, qa-WC-2 strain showed results very similar to
wc-2?, qa-WC-2 (B) strains were used in these experiments. At DD14, 1 ? 10?2
M QA was added to half of the cultures. Northern blot and Western blot
analyses were performed. The representative results from several indepen-
dent experiments are shown. The weak and high molecular weight FRQ
blot analyses show that there is a low level of FRQ expression in wc-1?strain.
(D) Western blot analyses show that the level of WC-1 does not affect the
abundance of WC-2 in a frq10, qa-WC-1 strain.
conidiation rhythms of the wild-type (A), wc-1?, qa-WC-1 (B, Upper), and
wc-2?, qa-WC-2 (C, Upper) strains in the presence of different concentrations
of QA (labeled at left). The race tubes shown are representative samples from
six replicate tubes. The period lengths of the rhythm (Ave. ? SEM) are labeled
the levels of WC-1, WC-2, and FRQ in the wild-type, wc-1?, qa-WC-1 or wc-2?,
qa-WC-2 strain. Liquid cultures were grown in media with different concen-
trations (indicated above) of QA and were harvested at DD24. No QA was
www.pnas.org?cgi?doi?10.1073?pnas.121170298Cheng et al.
those of the wc-1?, qa-WC-1 strain, except that it became
rhythmic at a lower QA concentration (Fig. 3C). When QA
concentration was less than 1 ? 10?7M, one conidiation peak
could be observed in the first day, but the rhythm could not be
sustained in the later days. When QA was above 1 ? 10?6M, the
strain exhibited clear circadian rhythms of conidiation, and the
rhythms became more robust at higher QA concentrations. At
the molecular level, some FRQ and WC-1 were expressed even
when there was no QA present, which was because of the low
expression level of WC-2 resulting from the basal activity of the
qa-2 promoter (see the longer exposure of WC-2 Western blot
in Fig. 3C). At 1 ? 10?6M QA, the induction of WC-2 and FRQ
started to be seen. As more WC-2 was induced by higher
concentrations of QA, the levels of FRQ and WC-1 were higher,
but the increase of FRQ expression appeared to be limited. At
1 ? 10?3M QA, although the WC-2 level was higher than that
of the wild-type, the levels of FRQ were comparable, suggesting
that WC-2 is not the limiting factor for the WC-1?WC-2 complex
in a wild-type strain.
Interestingly, although the robustness of the conidiation
rhythms changed as the levels of the WCs and FRQ changed in
these two strains, the period of the conidiation rhythms varied
only a couple of hours. When the levels of WCs and FRQ were
much lower than those of the wild-type and the amplitude of the
overt rhythms was low, the period length of the overt rhythm was
only about 2 h longer than that of the wild-type. When the levels
of WCs and FRQ were higher than those of the wild-type due to
higher concentration of the inducer, the period of the rhythms
stayed very close to that of the wild-type (Fig. 3). The small
period changes despite dramatic variations in the levels of the
clock components suggest that the parallel changes in WCs and
FRQ levels, results of the interlocked nature of the feedback
loops, is important for determining the circadian period of the
Together, these data suggest that the main effect of high WC
levels is to increase the robustness and stability of the circadian
clock. When either WC-1 or WC-2 level is low and rate-limiting
in the mutant strains, the clock runs poorly and it is hard to
sustain, whereas the clock runs more robust when their levels are
Higher Level of WC-1 or WC-2 Leads to Higher Level of FRQ Oscillation.
Because the levels of WC-1 and WC-2 determine the level of
FRQ and the robustness of the conidiation rhythms, we would
expect to see higher WC-1 and WC-2 levels leading to more
robust FRQ oscillations. To confirm this, the levels of WC-1,
WC-2, and FRQ were monitored in constant darkness in the
wc-1?, qa-WC-1 and wc-2?, qa-WC-2 strains. In general, these
molecular data match the race tube data. When there was no
inducer, the levels of WC-1 and FRQ were very low and
arrhythmic in the dark in the wc-1?, qa-WC-1 strain (Fig. 4A).
At 1 ? 10?5M QA, when clear conidiation rhythms started to
be seen on race tubes (Fig. 3B), low-level oscillations of WC-1
and FRQ could be observed (Fig. 4 B and D). When QA was
increased to 1 ? 10?2M, there was a robust and high level
oscillation of FRQ (Fig. 4 A, B, D). Although this high concen-
tration of QA increased WC-1 significantly, the abundance of
WC-1 was still rhythmic, suggesting that the posttranscriptional
regulation of FRQ on WC-1 expression is still functioning
despite the high and constitutive expression of wc-1.
For the wc-2?, qa-WC-2 strain, although the levels of WC-2
and FRQ were low without the inducer (the low level of WC-2
was due to leaky expression from the qa-2 promoter), the level
of FRQ was cyclic with a low amplitude (note the changes in
FRQ amount and phosphorylation states) (Fig. 4 C and E).
However, this weak FRQ rhythm was only limited to the first day
expression of FRQ and WC-2 cannot sustain the rhythmicity for
long. These data corresponded well with the 1-day conidiation
rhythm we observed on race tubes (Fig. 3C). In the presence of
1 ? 10?2M QA, as in the wc-1?, qa-WC-1 strain, the higher level
of WC-2 led to a robust and a higher level of FRQ oscillation
(Fig. 4 C and E). Surprisingly, we also found that the FRQ level
was comparable in constant light with or without QA in both
strains despite the dramatic difference in WC-1 or WC-2 level
that the requirement for WC-1 and WC-2 on FRQ expression in
LL is different from that in the dark. Together, these data
required for the functioning of the clock, and higher levels of
WC-1 and WC-2 lead to more robust oscillation of FRQ.
WC-1 Is the Limiting Factor in the WC-1?WC-2 Complex.Because both
WC proteins are required for the activation of FRQ and they
interact with each other to form heterodimeric complexes (16),
we wondered which protein is the limiting factor for the complex
in a wild-type strain. As described before, the results in Figs. 3
and 4 suggest that WC-1 is the limiting factor for frq expression
in the WC-1?WC-2 complex. To directly compare the levels of
WC-1 and WC-2, we generated strains in which WC-1 or WC-2
is tagged with the identical five c-Myc epitope tags (22). Western
blot analysis by using c-Myc monoclonal antibody revealed that
the relative molar concentration of WC-1 to WC-2 is about 1 to
6 in LL (data not shown), similar to what was recently reported
(29). Although this ratio can inform us about the relative amount
of the two proteins, it does not reflect the ratio of the functional
proteins in vivo because the functional levels of the two proteins
can be affected by regulations such as nuclear import or homo-
multimer formation (16, 30). Therefore, we decided to examine
the limiting role of WC-1 by demonstrating its limiting effects on
leads to a higher level of FRQ oscillation. (A and B) wc-1?, qa-WC-1 cultures
were grown in media with or without QA (1 ? 10?5or 1 ? 10?2M) in DD. The
left lane in B shows the protein levels of the wild-type culture grown in LL. (C)
A higher level of WC-2 resulted in a higher level of FRQ oscillation. wc-2?,
qa-WC-2 cultures were grown in media with or without 1 ? 10?2M QA in DD.
(D and E) Densitometric analyses of the Western blots shown in B and C,
Cheng et al.
June 19, 2001 ?
vol. 98 ?
no. 13 ?
the formation of the WC-1?WC-2 complex and on the activation
of FRQ expression.
To show the limiting effect of WC-1 on the WC-1?WC-2
complex formation, we used the same protein extracts of the
wc-1?, qa-WC-1 strain shown in Fig. 3B (Lower) and performed
immunoprecipitation using WC-2 antiserum. Protein extracts
from three different QA concentrations were used (from 1 ?
10?5to 1 ? 10?3M) because at 1 ? 10?5and 1 ? 10?4M of QA,
the levels of WC-1 and FRQ were comparable with those in the
wild-type strain, whereas their levels were much higher at 1 ?
10?3M QA than those in the wild-type (Fig. 3B). In contrast,
there was almost no change in the WC-2 level at different QA
concentrations (Figs. 3B and 5A). If WC-1 is the limiting factor
in the WC-1?WC-2 complex, the increase of WC-1 level at 1 ?
10?3M QA should result in a similar amount of increase of the
WC-1?WC-2 complex. As predicted, immunoprecipitation using
WC-2 antiserum showed that significantly more WC-1 was
coprecipitated with WC-2 at 1 ? 10?3M QA, whereas the
amount of WC-2 precipitated was about the same at different
QA concentrations (Fig. 5A, Right). The amount of increase of
WC-1 found in the complex was similar to the amount of its
increase in the total protein extracts. Importantly, FRQ was also
found in the complex, and the amount of FRQ coprecipitated
increased as the amount of WC-1?WC-2 complex increased.
To examine the limiting role of WC-1 in elevating FRQ levels
in a wild-type situation, we introduced the qa-WC-1 construct
into a wild-type strain and examined the circadian conidiation
rhythms and protein rhythms of the transformants. As we
expected, the higher level of WC-1 rhythm led FRQ to oscillate
at a higher level, whereas there was almost no change in the level
of WC-2 (Fig. 5B). Similar to the results shown in Fig. 3, there
was only a small period shortening effect despite higher than
wild-type levels of WC-1 and FRQ (the broader conidiation
peaks at 1 ? 10?2M QA are a carbon source effect). Consistent
with these data, a higher level of WC-2 expression did not result
in a higher level of FRQ in a wild-type, qa-WC-2 strain (Fig. 5D).
Together, these data demonstrate that WC-1 is the limiting
factor for the WC-1?WC-2 complex in the wild-type strain, and
a higher level of WC-1?WC-2 complex requires a higher level of
FRQ to interact with to block its transcriptional activation of frq
and close the circadian negative feedback loop.
In this study, we show that WC-1 and WC-2 are required for the
activation of frq transcription, with WC-1 being the limiting
factor in the WC-1?WC-2 complex. The levels of WC-1 and
WC-2 are positively regulated by FRQ, thereby they form
positive feedback loops. Interestingly, the mechanisms whereby
frq regulates wc-1 and wc-2 are different; frq regulates wc-1
posttranscriptionally, and it does not influence the abundance of
wc-1 transcripts, whereas its effect on wc-2 is at least partially
achieved by increasing the steady-state level of wc-2 mRNA.
When the level of either WC-1 or WC-2 is below a certain
threshold level, the level of FRQ is low and arrhythmic, and the
circadian rhythmicity cannot be sustained under normal condi-
tions. If the levels of WC-1 and WC-2 are above certain levels,
they will increase FRQ level, resulting in its rhythm and the
running of the clock. The higher the levels of WC-1 and WC-2,
the higher the level of FRQ oscillation and the more robust and
stable the overt rhythmicity (Fig. 6). Although our experiments
did not completely dissociate the interlocked feedback loops,
which will require the elimination of the regulations of FRQ on
WCs while keeping frq intact, the positive regulations of FRQ on
the levels of both WC-1 and WC-2 suggest that these interlocked
positive feedback loops should be important for promoting the
robustness and stability of the clock, which is an integral part of
an optimally functioning circadian system.
Another important implication of this study is that, because of
the interlocked nature of the circadian feedback loops, the levels
Immunoprecipitation analysis shows that WC-1 is the rate-limiting factor in
the WC-1?WC-2 complex. The same protein extracts (1 ? 10?5, 1 ? 10?4, and
1 ? 10?5M QA samples) shown in Fig. 3B were used, and they were immu-
noprecipitated with the WC-2 antiserum. Both the total extracts and the
pellets of immunoprecipitation were subjected to Western blot analysis. (B)
Western blot analysis shows that a higher level of WC-1 leads to a higher level
of FRQ oscillation in a wild-type strain. wc-1?, qa-WC-1 liquid cultures were
grown in media with or without 1 ? 10?2M QA in DD. (Right) The densito-
the conidiation rhythms of the wc-1?, qa-WC-1 strain at different concentra-
tions of QA (labeled at left). (D) Western blot analysis shows that the increase
of WC-2 level in a wild-type strain (wc-2?, qa-WC-2) did not result in the
increase of FRQ expression.
WC-1 is the rate-limiting factor in the WC-1?WC-2 complex. (A)
oscillator. WC-1 and WC-2 form heterodimers to activate the transcription of
frq. FRQ proteins interact with the WC-1?WC-2 complex to inhibit their
transcriptional activation, forming the negative feedback loop. FRQ also
positively regulates the levels of both WC-1 and WC-2, forming the positive
oscillation. When the levels of WCs are low, there is either no FRQ oscillation
or it oscillates at a low level, whereas high levels of WCs lead to a robust and
a high-level oscillation of FRQ. The parallel changes in WCs and FRQ limit the
period of clock to a small range.
(A) Model for gene regulation within the Neurospora circadian
www.pnas.org?cgi?doi?10.1073?pnas.121170298Cheng et al.
of the clock components are not the major determining factors
in setting period length of the clock. As shown in Figs. 3–5, the
parallel changes of WC-1 and FRQ levels had limited the period
length to a small range (about 2 h), even though their levels
varied considerably from one condition to another (from sig-
nificantly lower to higher than normal). Therefore, the inter-
locked feedback loops may be important for setting the period
also made the Neurospora clock into a highly dynamic but stable
process, a fact that may also explain phenomena such as tem-
perature compensation of the clock (21). In this view, we expect
that changes in activity or stability of the clock components
should play important roles in determining the period length of
the clock (24, 28, 31, 32).
Although the expression of both WCs is positively regulated
by FRQ, the positive effect of FRQ on WC-1 plays a more
important role in the Neurospora clock. Because WC-1 is the
limiting factor in the WC-1?WC-2 complex, its rhythmic expres-
sion would ensure robust rhythmicity by allowing proper
amounts of the transcription activation complexes to be formed
at different times of the day to start or close a cycle. The long
delay between the expression of FRQ and the increase of WC-1
is important (15) because it allows the negative repression of frq
to occur at a time (subjective morning) when FRQ level is high
while the amount of the WC-1?WC-2 complex is relatively low.
On the other hand, the positive effect of FRQ on the expression
of WC-2 seems to have only a limited role in the clock of a
wild-type strain. First, WC-2 is not the rate-limiting factor of the
WC-1?WC-2 complex, and its level does not fluctuate much in
the constant darkness (15). Second, unlike the level of WC-1,
WC-2 level does not change significantly with the increase of
FRQ level under constant conditions, and even a low level of
FRQ is able to maintain normal expression level of WC-2 (Figs.
2–5). Therefore, the positive role of FRQ on WC-2 appears only
to maintain the level of WC-2 in excess.
In both Drosophila and mammals, the negative elements of the
circadian feedback loops were also found to positively regulate
the expression of one of the positive elements and lead to their
rhythmic expressions. In flies, PERIOD and TIMELESS elevate
the level of dclock, whereas mCRYs and mPER2 increase the
expression of Bmal1 mRNA in mammals (13, 14). Unlike the
posttranscriptional role of FRQ on WC-1, these regulations
appear to be achieved at least partially at the transcript level. In
factors (WC-1 in Neurospora, dCLOCK in Drosophila, and
BMAL1 in mammals) of the positive elements in the het-
erodimeric transcription activation complexes are all expressed
rhythmically, whereas the levels of the other partners of the
complexes (WC-2 in Neurospora, CYCLE in Drosophila, and
CLOCK in mammals) remain pretty much unchanged through-
out the day (11, 14, 33, 34). Based on our results in Neurospora,
we predict that the interlocked feedback loops in Drosophila and
mammals are also important for sustaining robust rhythmicity in
We thank Drs. Michael Collett, Jay Dunlap, and Jennifer Loros for
reading of the manuscript. This work was supported by National Insti-
tutes of Health Grant GM 62591 (to Y.L.). Y.L. is a Louise W. Kahn
endowed scholar in Biomedical Research at the University of Texas
Southwestern Medical Center.
1. Dunlap, J. C. (1999) Cell 96, 271–290.
2. Crosthwaite, S. K., Dunlap, J. C. & Loros, J. J. (1997) Science 276, 763–769.
3. Allada, R., White, N. E., So, W. V., Hall, J. C. & Rosbash, M. (1998) Cell 93,
4. Rutila, J. E., Suri, V., Le, M., So, W. V., Rosbash, M. & Hall, J. C. (1998) Cell
5. King, D., Zhao, Y., Sangoram, A., Wilsbacher, L., Tanaka, M., Antoch, M.,
Steeves, T., Vitaterna, M., Kornhauser, J., Lowrey, P., et al. (1997) Cell 89,
6. Gekakis, N., Stankis, D., Nguyen, H. B., Davis, F. C., Wilsbacher, L. D., King,
D. P., Takahashi, J. S. & Weitz, C. J. (1998) Science 280, 1564–1569.
7. Kay, S. A. (1997) Science 276, 1093.
8. Young, M. W. (1999) Recent Prog. Horm. Res. 54, 87–94.
9. Aronson, B., Johnson, K., Loros, J. J. & Dunlap, J. C. (1994) Science 263,
10. Darlington, T. K., Wager-Smith, K., Ceriani, M. F., Stankis, D., Gekakis, N.,
Steeves, T., Weitz, C. J., Takahashi, J. & Kay, S. A. (1998) Science 280,
11. Kume, K., Zylka, M. J., Sriram, S., Shearman, L. P., Weaver, D. R., Jin, X.,
Maywood, E. S., Hastings, M. H. & Reppert, S. M. (1999) Cell 98, 193–205.
12. Lee, C., Bae, K. & Edery, I. (1999) Mol. Cell. Biol. 19, 5316–5325.
13. Glossop, N. R., Lyons, L. C. & Hardin, P. E. (1999) Science 286, 766–768.
14. Shearman, L. P., Sriram, S., Weaver, D. R., Maywood, E. S., Chaves, I., Zheng,
B., Kume, K., Lee, C. C., van der Horst, G. T., Hastings, M. H. & Reppert, S. M.
(2000) Science 288, 1013–1019.
15. Lee, K., Loros, J. J. & Dunlap, J. C. (2000) Science 289, 107–110.
16. Talora, C., Franchi, L., Linden, H., Ballario, P. & Macino, G. (1999) EMBO J.
17. Liu, Y., Garceau, N., Loros, J. J. & Dunlap, J. C. (1997) Cell 89, 477–486.
18. Linden, H. & Macino, G. (1997) EMBO J. 16, 98–109.
19. Ballario, P., Vittorioso, P., Magrelli, A., Talora, C., Cabibbo, A. & Macino, G.
(1996) EMBO J. 15, 1650–1657.
20. Crosthwaite, S. K., Loros, J. J. & Dunlap, J. C. (1995) Cell 81, 1003–1012.
21. Liu, Y., Merrow, M. M., Loros, J. J. & Dunlap, J. C. (1998) Science 281,
22. Cheng, P., Yang, Y., Heintzen, C. & Liu, Y. (2001) EMBO J. 20, 101–108.
23. Aronson, B. D., Johnson, K. A. & Dunlap, J. C. (1994) Proc. Natl. Acad. Sci.
USA 91, 7683–7687.
25. Bell-Pedersen, D., Dunlap, J. C. & Loros, J. J. (1996) Mol. Cell. Biol. 16,
26. Garceau, N., Liu, Y., Loros, J. J. & Dunlap, J. C. (1997) Cell 89, 469–476.
27. Merrow, M., Garceau, N. & Dunlap, J. C. (1997) Proc. Natl. Acad. Sci. USA 94,
28. Liu, Y., Loros, J. & Dunlap, J. C. (2000) Proc. Natl. Acad. Sci. USA 97,
29. Denault, D. L., Loros, J. J. & Dunlap, J. C. (2001) EMBO J. 20, 109–117.
30. Ballario, P., Talora, C., Galli, D., Linden, H. & Macino, G. (1998) Mol.
Microbiol. 29, 719–729.
31. Price, J. L., Blau, J., Rothenfluh, A., Adodeely, M., Kloss, B. & Young, M. W.
(1998) Cell 94, 83–95.
32. Lowrey, P. L., Shimomura, K., Antoch, M. P., Yamazaki, S., Zemenides, P. D.,
Ralph, M. R., Menaker, M. & Takahashi, J. S. (2000) Science 288, 483–492.
33. Lee, C., Bae, K. & Edery, I. (1998) Neuron 21, 857–867.
34. King, D. P. & Takahashi, J. S. (2000) Annu. Rev. Neurosci. 23, 713–742.
Cheng et al.
June 19, 2001 ?
vol. 98 ?
no. 13 ?