Neuron, Vol. 32, 657–671, November 20, 2001, Copyright 2001 by Cell Press
Circadian Regulation of Gene Expression Systems
in the Drosophila Head
Adam Claridge-Chang,1,5 Herman Wijnen,1,5
Felix Naef,2 Catharine Boothroyd,1
Nikolaus Rajewsky,3 and Michael W. Young1,4
1
Laboratories of Genetics,
2
Mathematical Physics, and
3
Theoretical Condensed Matter Physics
The Rockefeller University
1230 York Avenue
New York, New York 10021
Summary
Mechanisms composing Drosophila’s clock are conserved within the animal kingdom. To learn how such
clocks influence behavioral and physiological rhythms,
we determined the complement of circadian transcripts in adult Drosophila heads. High-density oligonucleotide arrays were used to collect data in the form
of three 12-point time course experiments spanning
a total of 6 days. Analyses of 24 hr Fourier components
of the expression patterns revealed significant oscillations for ⵑ400 transcripts. Based on secondary filters
and experimental verifications, a subset of 158 genes
showed particularly robust cycling and many oscillatory phases. Circadian expression was associated
with genes involved in diverse biological processes,
including learning and memory/synapse function, vision, olfaction, locomotion, detoxification, and areas
of metabolism. Data collected from three different
clock mutants (per0, tim01, and ClkJrk), are consistent
with both known and novel regulatory mechanisms
controlling circadian transcription.
Introduction
Daily rhythms of behavior and physiology are established
by circadian clocks. Intracellular biochemical oscillators
have been linked to circadian rhythms in bacteria, fungi,
plants, and animals. Genetic screens have allowed components of such clocks to be identified and have provided significant insights into their mechanisms (Allada
et al., 2001; Cermakian and Sassone-Corsi, 2000; Loros
and Dunlap, 2001; Reppert and Weaver, 2001; Ripperger
and Schibler, 2001; Williams and Sehgal, 2001; Young
and Kay, 2001).
In Drosophila, several genes form an autoregulatory
network that allows self-sustained molecular rhythmicity and perception/adjustment to daily environmental cycles (Allada et al., 2001; Edery, 2000; Williams and
Sehgal, 2001; Young and Kay, 2001). The transcription
factors CLOCK (CLK) and CYCLE (CYC) bind to homologous sequences (E boxes) in the promoters of period
(per), timeless (tim), and vrille (vri), coordinately inducing
their expression. After a delay that appears to be mediated by the casein kinase I⑀ ortholog double-time (dbt),
4
Correspondence: [email protected]
5
These authors contributed equally to this work.
PER and TIM proteins physically associate to form a
repressor of CLK/CYC. VRI negatively regulates per and
tim expression through a separate autoregulatory path,
but also determines cycling expression of a key neuropeptide, PDF (Blau and Young, 1999). Cycling accumulation of this peptide hormone is required for normal patterns of rhythmic locomotor activity (Helfrich-Forster et
al., 2000; Renn et al., 1999).
VRI’s contribution to PDF expression and behavioral
signaling indicated a potential mechanism for creating
extensive pathways of circadian regulation. Since CLK/
CYC activity oscillates, patterns of diverse gene expression might be influenced by the clock through transcriptional controls using this circadian activator. Evidence
for such a strategy has also come from independent
studies of the mammalian clock, which is highly homologous to that of the fly. Vasopressin, a well-known modulator of neuroendocrine function, oscillates due to direct
regulation by orthologs of CLK/CYC in mammals (Jin et
al., 1999).
Known clock-controlled functions in Drosophila include eclosion, vision, olfaction, courtship, locomotion,
and sleep. In the face of such functional variety, a single
transcriptional strategy for generating rhythmic gene
expression might be considered restrictive. In fact, strict
adherence to the patterns of CLK/CYC control described above could predict gene expression rhythms
with a limited number of phases, even a preponderance
of correlated expression patterns overlapping that of
per, tim, and vri. To address such hypotheses, and to
identify a comprehensive set of clock-controlled effectors of rhythmic behavior, we conducted RNA assays
of adult fly heads using high-density oligonucleotide
arrays. Circadian resolution of gene expression was
achieved by sampling adult Drosophila heads over six
24 hr intervals. Our survey included profiles of gene
expression obtained in light/dark cycles and in constant
darkness. Fourier analysis of the data allowed us to
identify genes with oscillatory expression patterns. Verification of the microarray results by extensive Northern
blot analyses showed a significant enrichment of genes
displaying periodic expression in a set of about 400
candidates. We further distinguished a subset of genes
(158) that exhibited particularly robust patterns of circadian expression. This latter group contains genes with
a wide variety of oscillatory phases and molecular functions including learning and memory, vision, olfaction,
locomotion, detoxification, and areas of metabolism. In
parallel to the time-resolved experiment, we used additional microarrays to assess profiles of gene expression
in arrhythmic clock mutants. This experiment indicated
that Drosophila’s circadian clock regulates gene activity
through both known and novel transcriptional mechanisms.
Results and Discussion
Experimental Overview
We performed a genome-wide expression analysis
aimed at identifying all transcripts from the fruit fly head
Neuron
658
that exhibit circadian oscillations in their expression.
Our experiments and analyses were designed to take
full advantage of two defining properties of circadian
rhythms: (1) ⵑ24 hr periodicity and (2) self-sustained
propagation in a constant environment. By taking time
points every 4 hr, we obtained a data set that has a high
enough sampling rate to reliably extract 24 hr Fourier
components. We decided to perform time course experiments spanning a day of entrainment followed by a
day of free-running to take advantage of both the selfsustaining property of circadian patterns and the improved amplitude and synchrony of circadian patterns
found during entrainment. We analyzed 36 RNA isolates
from wild-type adult fruit fly heads, representing three
2 day time courses, on high-density oligonucleotide
arrays. Each array contained 14,010 probe sets (each
composed of 14 pairs of oligonucleotide features) including ⵑ13,600 genes annotated from complete sequence determination of the Drosophila melanogaster
genome (Adams et al., 2000). To identify different regulatory patterns underlying circadian transcript oscillations, we collected four-point time course data from
three strains of mutant flies with defects in clock genes
(per0, tim01, and ClkJrk) during a single day of entrainment.
Because all previously known clock-controlled genes
cease to oscillate in these mutants but exhibit changes
in their average absolute expression levels, we focused
our analysis of the mutant data on changes in absolute
expression levels rather than on evaluations of periodicity.
Identification of Circadian Transcripts
In a simultaneous analysis of the three separate time
courses, we extracted the 24 hr Fourier component (F24)
for all of the observed expression patterns. We used
the p value associated with F24 (p-F24) to rank the
interrogated transcripts according to the perceived robustness of their oscillation. A set of filters was applied
to select against unreliable oscillations. We demanded
that the measured range of oscillation and the measured
expression values exceeded noise thresholds and that
circadian transcripts showed a positive 24 hr autocorrelation in at least one of the experiments. According to
this procedure, 293 genes emerged as significantly
rhythmic with 24 hr periodicity (p-F24 ⬍ 0.05; see Experimental Procedures and supplemental data).
In order to further define a set of circadian transcripts
based on independent experimental data in addition to
the statistical predictions derived from our microarrays,
we performed Northern blots for a sample of genes
taken from ordered segments of the p-F24 ⬍ 0.05 set.
For the interval p-F24 ⬍ 0.02, each of 16 sampled genes,
distributed over the full range of the interval, was verified
as strongly rhythmic with a phase concordant with that
predicted from the microarray data. This cutoff defined
a set of 158 genes and included six known circadian
genes beyond the 16 sampled (cf. Figures 1–3). Given
the limited resolution of Northern blots and the extension
of the verification to greater than 10% of this p-F24
class, the results indicate that periodic gene expression
patterns are efficiently identified by our experimental
and statistical method. Northern sampling also indicated substantial enrichment for circadian gene expression in the p-F24 0.02–0.05 interval; in fact, enrichment
for cycling gene expression continued to be evident by
Northern sampling throughout the interval p-F24 0.05–
0.1 (see Conclusions).
Organization of the Subset of 158 Circadian
Transcripts by Phase of Peak Expression
Figure 1A shows an arrangement of the circadian expression patterns identified by our selection criteria ordered by peak phase. Each column represents a time
point, and each row represents a gene. The colors represent the magnitude of the log ratios for the measured
average difference values normalized per experiment,
with shades of red indicating an increase and shades
of green indicating a decrease (Eisen et al., 1998). Oscillating patterns of all possible phases are represented.
When organized into 4 hr phase groups ranging from
ZT0 (Zeitgeber Time 0; lights on) through ZT23 (just
before lights on), the 158 genes show the following distribution of peak phases: 30 genes in ZT0-4 (e.g., Clk, to,
ninaA, Pdh), 19 in ZT4-8 (e.g., e, Ugt35b), 17 in ZT8-12
(e.g., trpl, ATPCL, Zw), 34 in ZT12-16 (e.g., tim, vri, per,
Slob), 34 in ZT16-20 (e.g., Cyp4d21, Cyp18a1), and 24
in ZT20-24 (e.g., Rh5, amphiphysin).
Hierarchical Clustering
To organize our circadian transcripts in a way that was
informed by the data from the mutants and grouped,
we performed hierarchical clustering. We included both
the log ratio wild-type data (normalized per experiment)
and the log ratios for each of the three clock mutants
(normalized to the entire data set) to achieve clusters
that have both a more or less uniform phase and a
uniform pattern of responses to defects in the circadian
clock. Figure 1B shows the organization of our 158 genes
after hierarchical clustering with an equal weighting for
all included log ratio measurements (see also Supplemental Figure S1 at http://www.neuron.org/cgi/content/
full/32/4/657/DC1).
One of the most interesting clusters generated by this
organization is the per cluster (displayed in Figure 4A).
This cluster contains genes that have an expression
peak around ZT16 and a tendency to be reduced in
expression in the ClkJrk mutant. Strikingly, all genes previously known to show this pattern of oscillation (per,
tim, vri; Allada et al., 1998; Blau and Young, 1999) are
found in this cluster. In fact, the tim gene, which has
multiple representations on our oligonucleotide arrays,
has two independent representations in this cluster. Together with the novel oscillator CG5798, per, tim, and
vri form a subcluster (average phase ZT14) that shows
upregulation in both the per0 and tim01 mutants. The fact
that per, tim, and vri all function in the central circadian
clock raises the possibility that several other genes from
this cluster, including the ubiquitin thiolesterase gene
CG5798 and the gene coding for the channel modulator
Slowpoke binding protein (Slob; discussed further below), may function in the circadian clock or directly
downstream of it.
The genes in a second cluster (Clk cluster, Figure
4B) are primarily grouped together based on their peak
phase (average phase ZT2). By virtue of the mutant expression data, we can recognize several subclusters
within this phase group. The known circadian genes Clk
and takeout (to) are part of this cluster. Clk is found in
a clustered pair with the leucyl aminopeptidase gene
CG9285. In terms of chromosomal organization, to,
CG11891, and CG10513 map closely together on chro-
Circadian Gene Expression in the Drosophila Head
659
Figure 1. Organization of 158 Circadian Expression Patterns by Phase and by Hierarchical Clustering
In both diagrams (A and B), columns represent experimental time points, and rows represent genes. The colors in an ascending order from
green to black to red represent the magnitude of the normalized log ratios for the measured average difference values. The data collected
from wild-type flies in three 12-point time courses of 2 days each is normalized per experiment and shown in the phase diagram (A) and in
the first 36 columns of the clustergram (B). Data collected from three arrhythmic mutants, per0, tim01, and ClkJrk, in four-point time courses
spanning 1 day each is normalized to all 48 data points and displayed in the last 12 columns of the clustergram (B). The blocked horizontal
bars below the phase diagram and the clustergram correspond to the light-dark entrainment schedule used in the time course experiments.
White blocks indicate periods of exposure to light, whereas black blocks correspond to periods of darkness, and the gray blocks indicate
periods of subjective light under the free-running condition of constant darkness. The vertical bar on the right of the phase diagram indicates
the peak phase for the genes in each of the rows relative to the light-dark entrainment schedule. The time of lights-on is indicated by ZT0
and 24 (Zeitgeber time 0 and 24), whereas lights-off is denoted by ZT12. The clustergram in (B) was generated by hierarchical clustering using
the uncentered average linkage method from the “CLUSTER” program giving each time point equal weight. The tree on the left of the
clustergram indicates the pairwise similarity relationships between the clustered expression patterns. The italicized gene names on the right
of the clustergram (ap, apterous; per, period; e, ebony; Clk, Clock) correspond to the positions of four clusters that each contain one of these
genes.
mosome 3R. If we consider all genes with p-F24 lower
than 0.05 we find two additional circadian genes in this
chromosomal region (CG11852, CG10553). Interestingly,
the Clk cluster contains three pairs of homologous
genes with very similar expression patterns: the UDPglycosyl transferase genes Ugt35a and Ugt35b, the enteropeptidase genes CG9645 and CG9649, and the
long-chain fatty acid transporter genes CG6178 and
CG11407. In the first two cases, the homologous genes
are also directly adjacent to each other on the chromosome. An overview of the map positions of all circadian
genes in our study is available as supplemental information online (http://www.neuron.org/cgi/content/full/32/
4/657/DC1). Apart from Ugt35a and Ugt35b, several
other genes with a predicted function in detoxification are
members of the Clk cluster (CG17524, CG8993, CG3174,
Cyp6a21). It may also be noteworthy that the genes
for three oxidoreductases found in this group (Pdh,
CG15093, CG12116) have almost identical phases (ZT3).
All genes of the apterous (ap) cluster are defined by
both the oscillatory phase of their expression pattern
(average phase ZT17) and by a distinct expression profile in the three clock mutants (see Figure 4C). Although
the 6 hr sampling interval in the mutant data makes it
difficult to reliably detect oscillations, it seems that the
majority of the genes in this cluster shows some degree
of periodicity in the three mutant light-dark regime (LD)
time courses. Although we cannot rule out that there
are circadian oscillations independent from the known
clock genes, we favor the hypothesis that there may be
Neuron
660
Figure 2. Log Ratio-Plotted Expression Data for 22 Transcripts in the Drosophila Head
Plots with six peaks represent 36 time points of array data collected over 6 days; the dashed lines indicate the boundary of separate 2 day
experiments. Plots with two peaks represent 12 time points of Northern blot data collected over 2 days; these are adjacent to the corresponding
array data. In each case, the raw data are coplotted with a 24 hr guideline (red). We emphasize that the shown guidelines are not sinusoidal
fits but show the 24 hr component in the pattern. The amplitude of that component is equal to the amplitude of the data only when the latter
follow perfect sinusoidal behavior, and necessarily smaller otherwise. Estimated phases and log ratio amplitudes are noted for each gene.
Genes are grouped by biological process.
Circadian Gene Expression in the Drosophila Head
661
Figure 3. Organization of a Subset of the 158 Circadian Genes (p-F24 ⬍ 0.02) by Molecular Function Category
Green print corresponds to published oscillators, red print denotes Northern-verified oscillators, blue print denotes a Northern-verfied oscillator
(latheo) outside of the 158 (p-F24 ⵑ0.05). Peak phases in black print correspond to times of (subjective) light, whereas those in white print
correspond to times of (subjective) darkness.
a light-driven response underlying the observed mutant
expression pattern. The genes in this group may, therefore, be regulated not only by the circadian clock, but
also by a direct light-dependent mechanism. It should
be mentioned that we specifically searched for evidence
of gene expression patterns that were purely light-driven
in wild-type flies, but found little indication of such regulation. Instead, we encountered genes with both a strong
light-driven oscillation and a weak circadian component
(data not shown). apterous (ap) encodes a LIM-homeobox transcription factor, which is known to act both
in neural development and in neuropeptide expression
(Benveniste et al., 1998). The ap cluster includes the
genes for the transcription factor moira, the synaptic
regulator syndapin, two septins (Sep1, CG9699), and
two ATP binding cassette (ABC) transporters (CG6162,
CG9990). In terms of chromosomal organization, CG6166,
the gene adjacent to CG6162 on chromosome 3R is
homologous to CG9990 and coregulated with CG6162
and CG9990.
The founding member of the fourth cluster (Figure 4D;
average phase ZT9), ebony (e), encodes ␤-alanyl-dopamine synthase and has roles in both cuticle tanning and
regulating circadian locomotor behavior (Newby and
Jackson, 1991). Among the other cluster members are
six genes that function in protein cleavage (CG9377,
(A) Array data for known clock genes showing robust circadian expression, establishing the technique as a powerful way to detect circadian
expression.
(B) Circadian transcripts connected to synaptic function, including transmission and plasticity.
(C) Genes connected to amine neurotransmitter transduction and metabolism.
(D) Five genes expressed in the eye. Rh5, Rh4, trpl, and ninaA all play a role in visual transduction.
(E) Circadian proteases.
(F) Oscillating transcripts for antioxidation and detoxification proteins.
Neuron
662
Circadian Gene Expression in the Drosophila Head
663
Ser99Da, SP1029, CG7828, CG11531, BcDNA:GH02435),
three transcription factor genes (CG15632, CG17257,
CG6755), as well as two genes each that act in signal
transduction (pn, loco), the cytoskeleton (TpnC47D,
Chd64), and lipid metabolism (ATPCL, CG1583).
Regulation of Circadian Transcripts
in Arrhythmic Flies
The per0, tim01, and ClkJrk mutations affect genes that are
essential for maintaining circadian rhythms and result in
both molecular and behavioral arrhythmicity. In addition
to abrogating the oscillations of the per, tim, vri, to,
and Clk transcripts, these mutations also affect their
absolute levels of expression. per0 and tim01 flies have
intermediate or somewhat elevated levels of per, tim, vri,
and to transcript, and decreased levels of Clk transcript
whereas ClkJrk mutants have the opposite effect (Hardin
et al., 1990; Sehgal et al., 1994, 1995; Allada et al., 1998;
Blau and Young, 1999; Glossop et al., 1999; So et al.,
2000). Based on these observations, we gathered genome-wide expression data from per0, tim01, and ClkJrk
mutant flies in three separate four-point time course
experiments. We employed a rank-sum Wilcoxon test
to determine if any of the interrogated transcripts were
significantly up- or downregulated in any of the mutants when compared to our total wild-type expression
data set.
Out of the 14010 probe sets on our arrays, 4865
showed up- or downregulation in one or more of the
three mutants with a p value lower than 0.05, 2544 were
significantly changed in the tim01 flies, 1810 were significantly different in per0, and 2181 in ClkJrk. It is unclear
what proportion of these changes depends on the actual
mutations themselves, because (1) our three mutant fly
strains have different genetic backgrounds and (2) we
collected data for only one population of each mutant
strain. Although there are known examples of noncircadian genes whose expression is affected by clock mutations, we decided that it would be more informative to
consider effects of the clock mutations only with respect
to the subset of 158 strong oscillators. Among this set,
we find 72 genes with one or more significant expression
changes in the three clock mutant strains (See Figure
5; *p ⱕ 0.05, **p ⱕ 0.01).
Included in our regulated set are tim (twice independently), vri, to, and Clk, and their patterns agree with
previously published observations. The hierarchical
clustergram of Figure 5 shows four basic patterns of
regulation: (type I) increased in per0 and tim01 but decreased in ClkJrk (e.g., vri, CG5798), (type II) decreased in
per0 and tim01 but increased in ClkJrk (e.g., Clk, CG15447),
(type III) decreased in all three mutants (e.g., Ugt36Bc/
CG17932, ea), and (type IV) increased in all three mutants
(e.g., Pdh, CG11891). Type I and II match the two known
modes of regulation for circadian genes. The behavioral
and molecular phenotypes of the per0 and tim01 mutations are almost identical (Sehgal et al., 1994, 1995). It
may, therefore, be relevant that we find no circadian
genes that are significantly upregulated in per0 and significantly downregulated in tim01 or vice versa. Apart
from genes that were a priori predicted to have expression patterns of type I (vri, tim, to) or II (Clk), we find
novel genes for each of these two expression patterns.
The average phases for the type I and type II subclusters
depicted in Figure 5 are, respectively, ZT12 and ZT7,
but there is large variation in phase among the members
of each of these subclusters. to is in the type I subcluster
and has a phase peak at ZT2, whereas CG15447 is in
the type II subcluster and peaks at ZT10. This phenomenon of phase differences among transcripts with a similar response to clock defects has been described previously for type I regulation in the case of to (So et al.,
2000). Here, we detect a similar phenomenon for genes
with Clk-like type II regulation. Type III and IV predict a
novel and unexpected response to the circadian mutants.
Enrichment for Known Circadian
Regulatory Elements
We tested the promoter sequences of our set of 158
genes for the presence of known and candidate circadian enhancer motifs (see Experimental Procedures; see
also supplemental data online at http://www.neuron.
org/cgi/content/full/32/4/657/DC1). The results suggested
that in fact this set is enriched in such elements. For
example, the frequency of E boxes in our set of oscillators (42 hits in total) is significantly higher than the frequency in random selections of genes. The significance
of this result does, however, depend on the inclusion of
well-studied genes (per, tim, vri). Some novel oscillators
with remarkable frequencies of “circadian transcription
elements” are loco (1 PDP1 site; 3 W boxes; 8 CRE
elements; 3 TER elements), trpl (1 E box; 1 PDP1 site;
9 W boxes; 6 CRE elements; 12 TER elements). Rh5 (5
W boxes; 5 CRE elements; 6 TER elements), and Slob
(1 E box; 1 PERR element; 2 PDP1 sites; 6 W boxes;
15 TER elements). It is noteworthy that many robustly
cycling genes have no known circadian transcription
elements in their promoters or first introns.
Molecular Function of Genes Expressed
with a Circadian Rhythm
We have also organized our set of 158 circadian genes
according to annotated or predicted molecular function
(see Figure 3). Several of these functional classes may
Figure 4. Clustergrams for Expression Patterns Grouped by Hierarchical Clustering
Columns represent experimental time points (denoted by genetic background, yw, cn bw, tim01, per0, or ClkJrk; and time points in Zeitgeber
time, ZT, or Circadian time, CT), and rows represent genes (indicated by their symbol and peak phase). The colors in an ascending order
from green to black to red represent the magnitude of the normalized log ratios for the measured average difference values. The blocked
horizontal bars below the clustergrams correspond to the light-dark entrainment schedule used in the time course experiments. White blocks
indicate periods of exposure to light, whereas black blocks correspond to periods of darkness, the gray blocks indicate periods of subjective
light under the free-running condition of constant darkness. The trees on the left of the clustergrams indicate the pairwise similarity relationships
between the clustered expression patterns. Boxes around sets of genes denote subclusters of genes with closely related expression patterns
that are derived from the predicted hierarchy. (A) per (period) cluster; (B) Clk (Clock) cluster; (C) ap (apterous) cluster; (D) e (ebony) cluster.
Top row shows average expression pattern for each cluster.
Neuron
664
Figure 5. Clustergram for Expression Patterns Obtained in Three
Clock Mutants
Columns represent experimental time points (denoted by genetic
background, tim01, per0, or ClkJrk; and time points in Zeitgeber time,
ZT) and rows represent genes with significant regulation in these
mutants. The colors in an ascending order from green to black to
red represent the magnitude of the normalized log ratios for the
measured average difference values. The blocked horizontal bars
below the clustergrams correspond to the light-dark entrainment
schedule used in the time course experiments. White blocks indicate
periods of exposure to light, whereas black blocks correspond to
periods of darkness. The tree on the left of the clustergram indicates
the pairwise similarity relationships between the clustered expression patterns. The symbols preceding the gene names for each
row correspond to the significance of the p value of the rank-sum
Wilcoxon test for each of the three mutants (—, not significant; *p ⱕ
0.05; **p ⱕ 0.01) in the order tim01/per0/ClkJrk. The blocks with Roman
numerals indicate four types of expression patterns.
provide insights into pathways influencing rhythmic behavior.
Synaptic Transmission and Plasticity
An emerging theory of the function of sleep postulates
that it is required for neural plasticity, synaptic maintenance, and remodelling (Hendricks et al., 2000a). Behaviorally defined sleep has been identified in the fly, with
behavioral recordings in LD indicating increased rest
during the dark phase (Hendricks et al., 2000b; Shaw et
al., 2000). Our assay for circadian expression uncovered
a number of genes known to be involved in synaptic
function and synaptic plasticity.
Three oscillating trancripts encode synaptic vesicle
endocytosis factors: ␤-adaptin (Bap; p-F24 0.007, ZT13),
AP-1␥ (p-F24 0.039, ZT18), and syndapin (p-F24 0.007,
ZT17). The first two are adaptors between the budding
membrane and clathrin lattice, while syndapin is thought
to connect vesicle endocytosis to actin (Qualmann and
Kelly, 2000; Takei and Haucke, 2001). Clock modulation
of the synaptic vesicle pool is consistent with the idea
of modulated synaptic function, although it is unclear
what effect raising or lowering endocytotic factors
would have on synaptic function.
The Slob transcript peaks at ZT15 by microarray and
Northern blot verification (Figure 2B) and is downregulated in the ClkJrk mutant, suggesting that CLK acts as
an activator of Slob. There is an E box 5.4 kb upstream
of the Slob transcriptional start site raising the possibility
that CLK acts directly on the Slob promoter.
If the cycling transcript can be shown to produce
oscillating Slob protein, this could indicate a potent
mechanism for rhythmic control of synaptic function,
including synaptic plasticity: Slob protein binds the calcium-dependent potassium channel Slowpoke (Slo) and
has been shown to increase Slo activity and voltage
sensitivity (Schopperle et al., 1998). Slob is in turn bound
by a second channel regulator, Leonardo. Hypomorphic
mutations of leonardo produce defects in learning, and
electrophysiological analyses of the larval neuromuscular junction (NMJ) in these mutants show presynaptic
function and plasticity is greatly impaired in these animals (Broadie et al., 1997). In contrast to Slob, Leonardo
is a strong inhibitor of Slo, but requires Slob for this
interaction (Schopperle et al., 1998; Zhou et al., 1999).
All three proteins colocalize to the presynaptic bouton
at larval NMJs (Zhou et al., 1999). Thus, Slob may contribute to a switching mechanism that ultimately places
Slo channel activity under circadian control.
Slo channels are widely expressed in the adult fly
head, including the eye, lamina, medulla, central brain,
and mushroom bodies, but it is not known which subset
of these areas contain oscillating Slob expression
(Becker et al., 1995). We performed in situ hybridization
with larval brains to localize Slob RNA expression. Prominent staining was observed in a restricted region consistent with placement of the developing mushroom
body (Figure 6A) (Truman et al., 1993). The staining also
corresponds well with that region of the larval brain
receiving PDF-rich projections of the circadian pacemaker cells, lateral neurons (LNs [Helfrich-Forster et al.,
2000]). In future studies, it will be important to determine
whether presence of the innervating LNs is required for
cycling Slob expression.
leonardo was initially implicated in presynaptic func-
Circadian Gene Expression in the Drosophila Head
665
Figure 6. Localization of Novel Cycling Genes by In Situ Hybridization
(A) In situ hybridization of Slob antisense probe to larval brains
is consistent with mushroom body staining, and the brain region
innervated by PDF-expressing pacemaker cells (LNs).
(B) In situ hybridization of CG13848 antisense probe to adult fly
head slices shows staining specific to basal cells in the eye, which
is high at ZT2 and low or absent at ZT14.
tion by the effect of mutations on learning (Broadie et
al., 1997). Mutations of latheo also cause learning defects, and this protein is also found at larval NMJs.
Lowered latheo function has been associated with elevated synaptic transmission and reduced synaptic plasticity (Rohrbough et al., 1999). latheo showed cycling
expression with peak accumulation at ZT12-15 in our
microarray analysis (p-F24 0.05) and in Northern assays
(Figure 2B). We also detect rhythmicity in the expression
of dunce (p-F24 0.04) and Calpain-B (p-F24 0.03) (Figure
2B) genes involved in learning and synaptic long-term
potentiation, respectively (Dudai et al., 1976; Lynch,
1998).
Amine Neurotransmitter-Related Functions
Two serotonin receptor transcripts, 5-HT2 and 5-HT1A,
were found to oscillate with phases of ZT15 and ZT18,
respectively (see Figure 2C for 5-HT2 data). Serotonin
is known to be involved in a variety of neuronal processes in animals, including synaptic plasticity, clock
entrainment, and mating behavior (Ehlen et al., 2001;
Horikawa et al., 2000; Kandel et al., 2000; Lee et al.,
2001). The 104 serotonergic neurons in the adult CNS
have been mapped, but no studies have been done
of either 5-HT receptor localization or receptor mutant
phenotypes (Valles and White, 1988). Neither of these
receptors are orthologs of the mammalian 5-HT receptor
implicated in photic clock entrainment; this would be
represented by the Drosophila 5-HT7 gene (Tierney,
2001), which we find with p-F24 0.067. 5-HT1A belongs
in a class of receptors that respond to agonists by decreasing cellular cAMP, while 5-HT2 is homologous to
mammalian receptors whose main mode of action involves activation of phospholipase C, a function involved in synaptic plasticity (Tierney, 2001; Kandel et
al., 2000).
The ebony transcript was found to oscillate by microarray and Northern blot, showing a peak of expression around ZT5 (Figure 2C). ebony oscillation connects
with a body of earlier evidence linking ebony to circadian
acitivity rhythms. ebony hypomorphs show severe defects in circadian rhythm including arrhythmicity/aberrant periodicity in the free-running condition, as well as
abnormal activity patterns in LD conditions (Newby and
Jackson, 1991). Ebony is a putative ␤-alanyl dopamine
synthetase, and hypomorphs show elevated levels of
dopamine (Hovemann et al., 1998; Walter et al., 1996).
Dopamine has been implicated in control of motor behavior, as it induces reflexive locomotion in decapitated
flies, and this response is under circadian control (Andretic and Hirsh, 2000). Our results suggest that oscillations of ebony contribute to the assembly of rhythmic
locomotor behavior. Other evidence points to a role in
clock resetting. In addition to impaired entrainment in
LD, ebony flies show an abnormal ERG, and ebony is
strongly expressed in the lamina and the medulla optic
neuropile, a region associated with vision rather than
motor control (Newby and Jackson, 1991; Walter et al.,
1996).
Vision
The Drosophila eye is both a likely target of clock control
and partly responsible for photic input to the central
pacemaker (Helfrich-Forster et al., 2001; Yang et al.,
1998). Several genes found to oscillate by our microarray
assay are components of visual processes.
Photoreceptor cells contain peripheral clocks, suggesting that visual function may be regulated by the
clock. In vertebrates, the synthesis of various visual
components is known to be under circadian control
(Hartman et al., 2001; Korenbrot and Fernald, 1989). In
Drosophila, electroretinogram (ERG) measurements of
visual sensitivity revealed a 4-fold cycle in sensitivity,
with a minimum at ZT4 and a broad peak around lights
off (ZT12) (Chen et al., 1992). This suggests that some
of the fly visual components are clock controlled. However, a previous study of five major phototransduction
components found no cycling of either mRNA or protein
(Hartman et al., 2001). In our genome-wide assay, the trpl
transcript was found to oscillate with peak expression at
ZT11 (p-F24 0.0026; Figure 2D). TRPL is one of two ion
channels in the visual transduction pathway, along with
TRP, a paralog. TRPL and TRP open in response to
a G protein-coupled phosphoinositide cascade that is
initiated by the isomerization of rhodopsin by light (Hardie and Raghu, 2001). Their opening produces the lightsensitive conductance in the photoreceptors. Amorphic
mutants of each channel show visual defects, while the
Neuron
666
double null genotype results in a blind fly (Reuss et al.,
1997). A cycling trpl transcript could contribute to the
visual sensitivity cycle. First, the two phenomena are in
the same phase with both sensitivity and trpl expression
peaking around lights-off. Second, it is estimated that
TRPL contributes about half of the wild-type conductance, allowing for a substantial range of modulation by
reducing TRPL function (Reuss et al., 1997). Another
possible role for oscillating TRPL function is connected
with circadian entrainment. In addition to being blind,
trp/trpl double null mutants show reduced circadian behavioral resetting and less TIM degradation in response
to light pulses (Yang et al., 1998). We note that the
established entrainment factor CRY is known to cycle
and that interactions of CRY and TIM are essential for
light-dependent TIM degradation (Ceriani et al., 1999;
Emery et al., 1998). The oscillating, clock-related protein
VIVID has also been shown to regulate photo-entrainment in Neurospora (Heintzen et al., 2001).
Our microarray experiments show that two opsin
genes are under circadian control: Rh5 and Rh4. The
Rh5 mRNA rhythm (p-F24 0.0098) was confirmed by
Northern blot to peak at ⵑZT 21, while the Rh4 array
data show a circadian pattern (p-F24 0.037) with a peak
4 hr later, at ZT1 (Figure 2D). Rh5 is a blue-absorbing
rhodopsin expressed in a subset of R8 cells at the base
of the retina, while Rh4 is a UV-absorbing pigment expressed in apical R7 cells (Chou et al., 1996; Salcedo
et al., 1999). Rh5 is never expressed in an R8 cell underlying an Rh4-expressing R7 cell, so in this way all ommatidia would contain one cycling rhodopsin (Chou et al.,
1996). In terms of regulating sensory receptiveness to
light, it is unclear why these two opsins should be targets
for clock control. The major blue rhodopsin Rh1 does
not cycle so it seems unlikely that an oscillation in these
two minor pigments would produce overall tuning of the
sensitivity of the fly visual system (Hartman et al., 2001).
NinaA is a rhodopsin chaperone and is required to
move Rh1 from the endoplasmic reticulum (ER) to the
rhabdomeric membrane (Colley et al., 1991). ninaA mutants display aberrant accumulation of Rh1 protein in
the ER. ninaA mRNA shows cycling expression in fly
heads by both array and Northern blot, with peak expression around ZT2 (Figure 2D). It is tempting to hypothesize that early morning ninaA upregulation would
have the effect of releasing a reservoir of Rh1 from the
ER, making it available for use in visual transduction.
However, a previous assay of NinaA levels in an LD
regime revealed no oscillation in protein levels (Hartman
et al., 2001). If true, this would represent a surprising
example of a robustly oscillating mRNA producing a
constitutive cognate protein. Finally, although its function is still unknown, Photoreceptor dehydrogenase
(Pdh) is a robustly oscillating transcript (see Figure 2D
and Figure 3) with an extremely high level of expression
in the screening pigment cells of the eye (Brunel et al.,
1998).
Rhythmic Proteases and Accessory Factors
Fifteen of the identified oscillatory genes are implicated
in protein cleavage (see Figure 2E and Figure 3). CG7828
and BcDNA:GH02435 (peak phase ZT6 and ZT10, respectively) mediate ubiquitination of protein substrates, thus
targeting them for degradation. Conversely, CG5798 and
CG7288 cycle with a peak at respectively, ZT14 and
ZT16, and each produces a ubiquitin specific protease
(Ubp; cleaves ubiquitin from ubiquitin-protein conjugates) that may act to prevent the degradation of specific protein targets (Chen and Fischer, 2000; Huang et
al., 1995). Two metalloprotease genes, two aminopeptidase genes, and five serine peptidases show circadian
oscillation. Four of the five serine peptidase genes cycle
with a peak phase between ZT4-7. This profusion of
oscillating proteases suggests that circadian proteolysis may represent a broad mechanism of clock control,
both of clock components themselves, as well as output
factors.
While circadian transcriptional mechanisms are relatively well understood, less is known about posttranslational mechanisms of circadian regulation. Proteases
are known to be involved in circadian control of the
degradation of some central clock components, and the
clock proteins PER, TIM, CLK, CRY, and VRI are all
known to undergo daily cycles of protein accumulation.
Degradation of TIM is responsible for photic resetting
of the Drosophila clock. This is thought to be mediated
by interaction with CRY, followed by ubiquitination and
proteasome-dependent loss of TIM (Naidoo et al., 1999).
Nothing is known about the factors mediating TIM degradation in the dark, yet patterns of CG5798 expression
may be of interest in this regard as it encodes a cycling
Ubp whose peak expression (ZT14) immediately precedes an interval of rapid TIM accumulation in pacemaker cells (Figure 2E).
A likely clock-related target of one or more proteases
is the neuropeptide PDF, whose regulation may be crucial to linking the clock to behavior. While pdf RNA
is expressed constitutively, the peptide accumulates
rhythmically under indirect control of the clock gene
vri (Blau and Young, 1999; Park and Hall, 1998). This
mechanism has not been further explored, but a clear
possibility is that a PDF propeptide is cleaved rhythmically, allowing cyclical release of active PDF (HelfrichForster et al., 2000). Of the 15 cycling proteases suggested by our study (Figure 3), CG4723 may be of special
interest due to its inclusion in a class of proteases known
to cleave neuropeptides (Isaac et al., 2000). The phase
of CG4723 expression, ZT4, also coincides with that of
PDF immunoreactivity in the Drosophila head (Park et
al., 2000).
Detoxification and Oxidative Stress
One hypothesis of sleep characterizes it as a cellular
detoxification and repair process (Inoue et al., 1995; Yin,
2000). We find twelve genes that have a predicted role
in detoxification. Three additional redox enzymes may
also participate in this process. Toxins are initially modified into reactive species by reductases/dehydrogenases
and cytochrome P450 molecules. Then glutathione-Stransferases (GSTs) and UDP-glycosyl transferases add
polar groups to the substrates to render them hydrophilic for elimination by secretion (Salinas and Wong,
1999). Firstly, following this pathway, we find the genes
for three circadian dehydrogenases: Pdh, CG10593, and
CG12116. Secondly, we find both morning and night
cytochrome P450 genes: Cyp6a21 and Cyp305a1 peak
early in the day (ZT0 and ZT5), whereas Cyp18a1 and
Cyp4d21 (see Figure 2F and Figure 3) peak at approximately the same time late at night (ZT18 and ZT19).
Thirdly, while CG17524 is the only GST gene found in
Circadian Gene Expression in the Drosophila Head
667
our core set, we noticed that two other members of the
GST type III gene cluster on chromosome 2R show 24
hr periodicity (CG17523, p-F24 0.025; CG17527, p-F24
0.031). Fourthly, UDP-glycosyl transferase genes are
represented in our circadian set by Ugt35a, Ugt35b,
and Ugt36Bc. Of these, Ugt35b (Figure 2F) encodes
an antennal specific transcript with a potential role in
olfaction, whereas Ugt35a is expressed more uniformly
(Wang et al., 1999).
CG13848 (see Figures 2F and 6B) is an ␣-tocopherol
transfer protein (␣-TTP) that is strongly expressed in
certain basal cells of the eye, showing peak expression
around dawn. Humans with mutations in the homologous ␣-TTP show neurodegenerative ataxia that is associated with a deficiency in ␣-tocopherol (vitamin E) incorporation into lipoprotein particles (Ouahchi et al., 1995).
Vitamin E acts as an antioxidant, and it is thought that
this activity allows it to protect neurons from damage
by free radicals (Ouahchi et al., 1995). Also from human
studies, it is known that photoreceptor cells are particularly susceptible to oxidative damage due to high levels
of polyunsaturated fatty acids in the photoreceptor
membrane, and their exposure to visible light (Beatty et
al., 2000). Thus, we propose that the dawn phase of
CG13848 represents an upregulation of ␣-TTP for increased daytime transfer of photoprotective vitamin E
into the photoreceptor membrane. Also in our circadian
set, Catalase (encoded by Cat) and a thioredoxin (encoded by CG8993) are both involved in neutralizing reactive oxygen species (Gutteridge and Halliwell, 2000).
Metabolism
Different aspects of metabolism are represented among
the selected set of oscillating transcripts: lipid metabolism (five genes), amino acid metabolism (three), carbohydrate metabolism (three), and glycoprotein biosynthesis (two). Intriguingly, Zw encodes glucose-6-phosphate
1-dehydrogenase of the pentose-phosphate pathway
(PPP), while CG10611 encodes fructose-bisphosphatase in gluconeogenesis. The two pathways have antagonizing roles in glucose metabolism; both genes are
key control points in their respective pathway and are
maximally expressed in opposite phases (Voet and Voet,
1990). Zw transcripts are at their zenith just before dusk
(ZT11), whereas CG10611 transcripts peak at dawn
(ZT0). Thus, the antiphase oscillation of these two genes
may produce daily alternation between glucose anabolism and catabolism. Zw and CG10611 transcript levels
also respond in opposite fashion to clock defects: Zw
levels are significantly decreased in per0 and tim01 mutants, whereas, CG10611 is significantly upregulated in
tim01 (Figure 5).
In more direct relation to the clock itself, Zw is the
first committed step in the PPP and generally thought
to control flux through this pathway (Stryer, 1988; Voet
and Voet, 1990). The PPP is the major pathway of NAD
(or NADP) conversion to NAD(P)H (Stryer, 1988; Voet
and Voet, 1990). It was recently shown that NAD(P)H
can bind homologs of CLK and CYC, promoting their
dimerization and DNA binding (Rutter et al., 2001). Maximal Zw expression at ZT11—and therefore presumably
NAD(P)H production via the PPP—is coincident with
maximal per and tim transcription by CLK/CYC (So and
Rosbash, 1997). This information is consistent with Zw
participating in a NAD(P)H-mediated autoregulatory
loop of the clockworks (Schibler et al., 2001).
Nucleic Acid Metabolism
We find a subset of 15 genes involved in nucleic acid
metabolism (Figure 3). This includes five genes encoding
specific RNA polymerase II transcription factors. Four
of these encode parts of the circadian clock itself (per,
tim, vri, and Clk), whereas the fifth one, apterous (ap),
generates a homeobox transcription factor with a role
in neurogenesis and the expression of neuropeptides
(Benveniste et al., 1998; Hobert and Westphal, 2000).
Taf30␣2 is a subunit of the general transcription factor
TFIID, while moira (mor) is part of the SWI-SNF chromatin-remodeling complex (Aoyagi and Wassarman, 2000;
Crosby et al., 1999). One splicing factor gene (DebB)
and one DNA repair gene (CG4049) are also found
among this set of oscillating transcripts.
Cytoskeleton
Circadian genes with a role in the cytoskeleton include
those encoding two actin binding proteins (Chd64,
CG11605), a troponin C (TpnC47D), and two septins
(Sep-1, CG9699). Chd64 (phase peak ZT4) and TpnC47D
(phase peak ZT7) function specifically in muscle contraction. Another Drosophila Troponin C, TpnC73F, is
found to peak at ZT6 and has a p-F24 of 0.08.
Conclusions
We have recognized a set of 158 genes expressed with
a robust circadian rhythm in the adult Drosophila head
by microarray screening. These encompass a wide variety of molecular functions, and expression patterns represented essentially all circadian phases. A larger set
of genes was identified (393 entries, p-F24 ⬍ 0.05; 293
entries after secondary filters), and our statistical approach again indicated significant circadian rhythmicity
for these, but they were characterized by somewhat less
robust oscillations than those of p-F24 ⬍ 0.02. Independent verifications indicated substantial enrichment for
cycling gene expression in this larger set and beyond,
to ⵑp-F24 ⬍ 0.1. At a p-F24 cutoff of 0.1, 532 genes pass
our secondary filters for 24 hr autocorrelation, noise,
and the range-to-noise measure. From the frequency of
Northern-verified oscillations detected in this larger pool
of candidate genes, we believe we can usefully predict
that the total complement of circadian genes would include ⵑ400–500 in the adult head.
There are important factors that might lead us to underestimate the total complement of circadian genes.
Our approach will favor genes that are homogeneously
expressed in the head. If the same gene is expressed
with varied phases in different head tissues, this will
lessen the robustness of the apparent oscillation and
phase. Similarly, if only a restricted portion of the head
generates the cycling gene pattern, but constitutive expression is found elsewhere in the head, amplitude of
the signal will be diminished. Differences of this sort
might be expected in cases where a cycling gene product produces a limited physiological effect. Regulation
of this type might be expected in the antennae, where,
for example, electrophysiological responses to odorants
vary with a circadian rhythm (Krishnan et al., 1999). We
should also stress that we have sampled only the fully
differentiated adult head. If transient patterns of circa-
Neuron
668
dian expression occur during development, these would
go unnoticed in our experiment. Given the plethora of
tissues housing autonomous circadian clocks, an expanded list of rhythmic genes would probably be derived
from any related sampling of the body (e.g., wings, legs,
excretory, and digestive tissues), especially as this tissue autonomy may reflect a requirement for tissue-specific pathways of circadian control that lie downstream
from a largely uniform clock mechanism (Plautz et al.,
1997).
How will the many patterns of cycling gene expression
be further explored? Molecular tools that reveal the importance of oscillating gene activity have already been
applied to a study of several clock genes in Drosophila.
In these studies, oscillating patterns of a target gene’s
expression have been replaced with constitutive activity. Central questions related to vri, per, and tim function
have each been explored in this manner (Blau and
Young, 1999; Yang and Sehgal, 2001). The present study
allows an expansion of such work to address the molecular connections between individual behaviors and circadian clocks.
Experimental Procedures
Fly Strains
Wild-type strains used were yw and cn bw. Clock mutant strains
used were per0 (Konopka and Benzer, 1971), tim01 (Myers et al.,
1995), and ClkJrk (Allada et al., 1998).
Microarray Experiments
Fly strains were entrained to a 12 hr:12 hr light-dark (LD) regime at
25⬚C for 3 days and harvested onto dry ice either at 6 hr intervals
during the following day of light-dark entrainment (mutant strains)
or at 4 hr intervals during the next day of light-dark entrainment and
an additional day of free-running in constant darkness (wild-type
strains). Frozen flies were broken up and passed through wire mesh
to separate the heads from the rest of the animal. This procedure
faithfully isolated heads from other body parts and should exclude
contaminating signals from sources outside the head. The heads
were used to isolate crude RNA with either Rnazol (Tel-Test) or
Trizol (Life Technologies). The yield was approximately 1 ␮g total
RNA/fly head. This RNA was then further purified using Rneasy
columns (Qiagen) following the RNA cleanup protocol. 25 ␮g of this
material was used to make biotin-labeled cRNA probe following the
methods described in the Affymetrix GeneChip Expression Technical Manual. cDNA synthesis was carried out with T7-dT24 primer
(MWG Biotech), Superscript Choice (Life Technologies), and additional enzymes from New England Biolabs. In vitro transcription was
done using the ENZO Bioarray HighYield RNA transcript labeling kit.
Hybridization, washing, staining, and scanning of the target cRNA to
Affymetrix oligonucleotide arrays were done as per the Affymetrix
GeneChip manual. Uniformity of the hybridization response from
one array to another was verified by hybridizing RNA samples from
the same strain to duplicate oligonucleotide arrays for several time
points (see supplemental data online).
Data Analysis
The circadian tendency of a time course experiment was quantified
using the 24 hr Fourier component and its associated p value (F24
and p-F24, respectively). These quantities were determined after
transforming the average difference (AD) scores into log ratios. For
each time point, the expression level relative to the mean (taken in
log coordinates) over that experiment was computed. AD scores
below one standard deviation of residual background (Q) were set
to Q. Fourier scores were calculated after appending the three normalized time courses, so that phase coherence between the patterns in the three experiments was automatically enforced. The resulting F24 scores are independent of the order in which the time
courses were appended. p-F24 was determined from ranking the
real Fourier score with respect to scores obtained from 10,000 random permutations of the time points (time points from different time
courses were not mixed), for each gene separately. All observed
expression patterns were ranked according to ascending p-F24, so
that 188 genes passed a filter of p-F24 ⬍ 0.02. Progressively lowering the cutoff to 0.05, 0.1, and 0.15, respectively, yields 393, 817,
and 1232 entries.
To identify possible false positives that may have passed the
above criteria by accident, we applied several fairly mild secondary
filters in series. First, we considered the 24 hr autocorrelation (AC24),
which measures the difference in pattern between the LD and DD
parts of the experiment, and has a range in ⫺1:1. We required that
the highest of the three AC24 values was larger than 0. Second, to
eliminate genes with too low over all signal intensity, we required
that more than half of all measured AD values for a given gene be
larger than 4Q. Finally, we considered a signal-to-noise like indicator
RN (range-over-noise), defined for each time course as the ratio
between the amplitude of oscillation (maximum minus minimum AD
over the time course) to the sum of the noise levels Q at the extrema.
Oscillations with low RN are likely to be the consequence of noise;
we, therefore, selected for genes with average RN ⬎ 3.5.
Analysis of the Data from Circadian Mutants
The significance of differences between AD values estimated for
wild-type flies and AD values for per0, tim01, or ClkJrk mutant flies
was assessed using a rank-sum Wilcoxon test comparing the data
for all 36 wild-type time points separately to each set of four mutant
time points.
Verification of Microarray Results
A sampling of genes that were indicated by the oligonucleotide array
data to have a circadian oscillation were independently assayed by
Northern blot. Blots were made with 12 samples of RNA collected
over 2 days, one LD and one DD, in a similar configuration to the
array experiments. Probe template was sourced from the Drosophila
Gene Collection (Berkeley Drosophila Genome Project), and each
clone was sequence verified. Blots were visualized and quantified
with a Storm Phosphorimager (Molecular Dynamics). Only band signals that were measured to be at least 3-fold above the blot background were considered reliably quantifiable. A gene was considered to have a confirmed oscillation if it showed two peaks spaced
at 24 hr, with phase of peak expression that was within 4 hr (the
sampling window) of the array data. So as to have a quantification
of the oscillation that is directly comparable with the array result,
Fourier analysis was also done on the Northern data. Based on the
distribution of p-F24 values for the transcripts that were uniformly
verified to cycle by sampling with Northern blots, we decided that
a p-F24 cutoff of 0.02 would define a subset of 158 genes with
particularly robust oscillations.
Clustering
Hierarchical clustering on the subset of 158 genes was performed
using the “CLUSTER” program (Eisen et al., 1998) on the log ratio
values normalized per time course experiment for the wild-type data
and normalized to the entire data set for the mutant data, giving
each of the arrays equal weight. Before calculating the log ratios
for clustering, average difference values below 4Q were set to 2.8Q.
Clustergrams were visualized with the aid of the “TREEVIEW” program. From the set of 158, a subset of 72 genes that showed significant up- or downregulation in at least one of the three mutant data
sets was identified. For this set of 72, hierarchical clustering was
performed on the mutant data.
Promoter Site Analysis
The upstream sequences associated with predicted circadian genes
were analyzed to test if they were enriched for known circadian
transcription factor binding sites. For each transcript, an upstream
sequence with a maximum of 10 kb was defined by the translation
start of the transcript and the beginning of the next upstream coding
sequence. The following definitions of binding sites were used: E
box, CACGTGAG(C)C with a perfect match for the first six bases
and possibly one mismatch for the last three bases; W box, TTG
GCCAGCAA with two mismatches; CRE, TGACGTCA with one mis-
Circadian Gene Expression in the Drosophila Head
669
match; PERR, GTTCGCACAA with one mismatch; TER, GCACGTTG
with one mismatch; PDP1, GAATTTTGTAAC with two mismatches. A
comprehensive description of the promoter site analysis is available
online.
In Situ Hybridizations
Larval brain in situ hybridization was performed as per Price et al.
(1998). Adult fly head in situ hybridization was done as per SchaerenWiemers and Gerfin-Moser (1993). Probe templates were from the
Drosophila Gene Collection (Berkeley Drosophila Genome Project).
Acknowledgments
We thank the staff of the Gene Array facility at The Rockefeller
University, in particular Greg Khitrov and Richard Pearson for help
with the hybridization and processing of oligonucleotide microarrays; Stefan Bekiranov for initial help with data analysis; Andrew
Eisberg for preparation of Northern blot probes; Mark Schroeder
for gene annotation; members of the Young laboratory for helpful
discussion; and Fernanda Ceriani, Martin Straume, and Steve Kay
for sharing results prior to publication. This work was supported by
grants from NSF, Center for Biological Timing, and NIH (GM54339)
to M.W.Y. and NIMH postdoctoral fellowship PHS MH63579 (H.W.).
Received October 16, 2001; revised November 2, 2001.
References
Adams, M.D., Celniker, S.E., Holt, R.A., Evans, C.A., Gocayne, J.D.,
Amanatides, P.G., Scherer, S.E., Li, P.W., Hoskins, R.A., Galle, R.F.,
et al. (2000). The genome sequence of Drosophila melanogaster.
Science 287, 2185–2195.
Allada, R., White, N.E., So, W.V., Hall, J.C., and Rosbash, M. (1998). A
mutant Drosophila homolog of mammalian Clock disrupts circadian
rhythms and transcription of period and timeless. Cell 93, 791–804.
Allada, R., Emery, P., Takahashi, J.S., and Rosbash, M. (2001). Stopping time: the genetics of fly and mouse circadian clocks. Annu.
Rev. Neurosci. 24, 1091–1119.
Andretic, R., and Hirsh, J. (2000). Circadian modulation of dopamine
receptor responsiveness in Drosophila melanogaster. Proc. Natl.
Acad. Sci. USA 97, 1873–1878.
Aoyagi, N., and Wassarman, D.A. (2000). Genes encoding Drosophila
melanogaster RNA polymerase II general transcription factors: diversity in TFIIA and TFIID components contributes to gene-specific
transcriptional regulation. J. Cell Biol. 150, F45–50.
Beatty, S., Koh, H., Phil, M., Henson, D., and Boulton, M. (2000). The
role of oxidative stress in the pathogenesis of age-related macular
degeneration. Surv. Ophthalmol. 45, 115–134.
Becker, M.N., Brenner, R., and Atkinson, N.S. (1995). Tissue-specific
expression of a Drosophila calcium-activated potassium channel.
J. Neurosci. 15, 6250–6259.
Benveniste, R.J., Thor, S., Thomas, J.B., and Taghert, P.H. (1998).
Cell type-specific regulation of the Drosophila FMRF-NH2 neuropeptide gene by Apterous, a LIM homeodomain transcription factor.
Development 125, 4757–4765.
Blau, J., and Young, M.W. (1999). Cycling vrille expression is required for a functional Drosophila clock. Cell 99, 661–671.
Broadie, K., Rushton, E., Skoulakis, E.M., and Davis, R. (1997). Leonardo, a Drosophila 14-3-3 protein involved in learning, regulates
presynaptic function. Neuron 19, 391–402.
Brunel, C.A., Madigan, S.J., Cassill, J.A., Edeen, P.T., and McKeown,
M. (1998). pcdr, a novel gene with sexually dimorphic expression in
the pigment cells of the Drosophila eye. Dev. Genes Evol. 208,
327–335.
Ceriani, M.F., Darlington, T.K., Staknis, D., Mas, P., Petti, A.A., Weitz,
C.J., and Kay, S.A. (1999). Light-dependent sequestration of TIMELESS by CRYPTOCHROME. Science 285, 553–556.
Cermakian, N., and Sassone-Corsi, P. (2000). Multilevel regulation
of the circadian clock. Nat. Rev. Mol. Cell. Biol. 1, 59–67.
Chen, X., and Fischer, J.A. (2000). In vivo structure/function analysis
of the Drosophila fat facets deubiquitinating enzyme gene. Genetics
156, 1829–1836.
Chen, D.M., Christianson, J.S., Sapp, R.J., and Stark, W.S. (1992).
Visual receptor cycle in normal and period mutant Drosophila: microspectrophotometry, electrophysiology, and ultrastructural morphometry. Vis. Neurosci. 9, 125–135.
Chou, W.H., Hall, K.J., Wilson, D.B., Wideman, C.L., Townson, S.M.,
Chadwell, L.V., and Britt, S.G. (1996). Identification of a novel Drosophila opsin reveals specific patterning of the R7 and R8 photoreceptor cells. Neuron 17, 1101–1115.
Colley, N.J., Baker, E.K., Stamnes, M.A., and Zuker, C.S. (1991). The
cyclophilin homolog ninaA is required in the secretory pathway. Cell
67, 255–263.
Crosby, M.A., Miller, C., Alon, T., Watson, K.L., Verrijzer, C.P., Goldman-Levi, R., and Zak, N.B. (1999). The trithorax group gene moira
encodes a brahma-associated putative chromatin-remodeling factor in Drosophila melanogaster. Mol. Cell. Biol. 19, 1159–1170.
Dudai, Y., Jan, Y.N., Byers, D., Quinn, W.G., and Benzer, S. (1976).
dunce, a mutant of Drosophila deficient in learning. Proc. Natl. Acad.
Sci. USA 73, 1684–1688.
Edery, I. (2000). Circadian rhythms in a nutshell. Physiol. Genomics
3, 59–74.
Ehlen, J.C., Grossman, G.H., and Glass, J.D. (2001). In vivo resetting
of the hamster circadian clock by 5-HT7 receptors in the suprachiasmatic nucleus. J. Neurosci. 21, 5351–5357.
Eisen, M.B., Spellman, P.T., Brown, P.O., and Botstein, D. (1998).
Cluster analysis and display of genome-wide expression patterns.
Proc. Natl. Acad. Sci. USA 95, 14863–14868.
Emery, P., So, W.V., Kaneko, M., Hall, J.C., and Rosbash, M. (1998).
CRY, a Drosophila clock and light-regulated cryptochrome, is a
major contributor to circadian rhythm resetting and photosensitivity.
Cell 95, 669–679.
Glossop, N.R., Lyons, L.C., and Hardin, P.E. (1999). Interlocked feedback loops within the Drosophila circadian oscillator. Science 286,
766–768.
Gutteridge, J.M., and Halliwell, B. (2000). Free radicals and antioxidants in the year 2000. A historical look to the future. Ann. NY Acad.
Sci. 899, 136–147.
Hardie, R.C., and Raghu, P. (2001). Visual transduction in Drosophila.
Nature 413, 186–193.
Hardin, P.E., Hall, J.C., and Rosbash, M. (1990). Feedback of the
Drosophila period gene product on circadian cycling of its messenger RNA levels. Nature 343, 536–540.
Hartman, S.J., Menon, I., Haug-Collet, K., and Colley, N.J. (2001).
Expression of rhodopsin and arrestin during the light-dark cycle in
Drosophila. Mol. Vis. 17, 95–100.
Heintzen, C., Loros, J.J., and Dunlap, J.C. (2001). The PAS protein
VIVID defines a clock-associated feedback loop that represses light
input, modulates gating, and regulates clock resetting. Cell 104,
453–464.
Helfrich-Forster, C., Tauber, M., Park, J.H., Muhlig-Versen, M.,
Schneuwly, S., and Hofbauer, A. (2000). Ectopic expression of the
neuropeptide pigment-dispersing factor alters behavioral rhythms
in Drosophila melanogaster. J. Neurosci. 20, 3339–3353.
Helfrich-Forster, C., Winter, C., Hofbauer, A., Hall, J.C., and Stanewsky, R. (2001). The circadian clock of fruit flies is blind after elimination of all known photoreceptors. Neuron 30, 249–261.
Hendricks, J.C., Sehgal, A., and Pack, A.I. (2000a). The need for
a simple animal model to understand sleep. Prog. Neurobiol. 61,
339–351.
Hendricks, J.C., Finn, S.M., Panckeri, K.A., Chavkin, J., Williams,
J.A., Sehgal, A., and Pack, A.I. (2000b). Rest in Drosophila is a sleeplike state. Neuron 25, 129–138.
Hobert, O., and Westphal, H. (2000). Functions of LIM-homeobox
genes. Trends Genet. 16, 75–83.
Horikawa, K., Yokota, S., Fuji, K., Akiyama, M., Moriya, T., Okamura,
H., and Shibata, S. (2000). Nonphotic entrainment by 5–HT1A/7 receptor agonists accompanied by reduced Per1 and Per2 mRNA
levels in the suprachiasmatic nuclei. J. Neurosci. 20, 5867–5873.
Neuron
670
Hovemann, B.T., Ryseck, R.P., Walldorf, U., Stortkuhl, K.F., Dietzel,
I.D., and Dessen, E. (1998). The Drosophila ebony gene is closely
related to microbial peptide synthetases and shows specific cuticle
and nervous system expression. Gene 221, 1–9.
Huang, Y., Baker, R.T., and Fischer-Vize, J.A. (1995). Control of cell
fate by a deubiquitinating enzyme encoded by the fat facets gene.
Science 270, 1828–1831.
Inoue, S., Honda, K., and Komoda, Y. (1995). Sleep as neuronal
detoxification and restitution. Behav. Brain Res. 69, 91–96.
Isaac, R.E., Siviter, R.J., Stancombe, P., Coates, D., and Shirras,
A.D. (2000). Conserved roles for peptidases in the processing of
invertebrate neuropeptides. Biochem. Soc. Trans. 28, 460–464.
Jin, X., Shearman, L.P., Weaver, D.R., Zylka, M.J., de Vries, G.J., and
Reppert, S.M. (1999). A molecular mechanism regulating rhythmic
output from the suprachiasmatic circadian clock. Cell 96, 57–68.
Kandel, E.R., Schwartz, J.H., and Jessell, T.M. (2000). Principles of
Neural Science (New York: McGraw-Hill).
Konopka, R.J., and Benzer, S. (1971). Clock mutants of Drosophila
melanogaster. Proc. Natl. Acad. Sci. USA. 68, 2112–2116.
Korenbrot, J.I., and Fernald, R.D. (1989). Circadian rhythm and light
regulate opsin mRNA in rod photoreceptors. Nature 337, 454–457.
Krishnan, B., Dryer, S.E., and Hardin, P.E. (1999). Circadian rhythms
in olfactory responses of Drosophila melanogaster. Nature 400,
375–378.
Lee, G., Villella, A., Taylor, B.J., and Hall, J.C. (2001). New reproductive anomalies in fruitless-mutant Drosophila males: extreme lengthening of mating durations and infertility correlated with defective
serotonergic innervation of reproductive organs. J. Neurobiol. 47,
121–149.
Loros, J.J., and Dunlap, J.C. (2001). Genetic and molecular analysis
of circadian rhythms in Neurospora. Annu. Rev. Physiol. 63, 757–794.
Lynch, G. (1998). Memory and the brain: unexpected chemistries
and a new pharmacology. Neurobiol. Learn. Mem. 70, 82–100.
Myers, M.P., Wager-Smith, K., Wesley, C.S., Young, M.W., and Sehgal, A. (1995). Positional cloning and sequence analysis of the Drosophila clock gene, timeless. Science 270, 805–808.
Naidoo, N., Song, W., Hunter-Ensor, M., and Sehgal, A. (1999). A
role for the proteasome in the light response of the timeless clock
protein. Science 285, 1737–1741.
Newby, L.M., and Jackson, F.R. (1991). Drosophila ebony mutants
have altered circadian activity rhythms but normal eclosion rhythms.
J. Neurogenet. 7, 85–101.
Ouahchi, K., Arita, M., Kayden, H., Hentati, F., Ben Hamida, M.,
Sokol, R., Arai, H., Inoue, K., Mandel, J.L., and Koenig, M. (1995).
Ataxia with isolated vitamin E deficiency is caused by mutations in
the alpha-tocopherol transfer protein. Nat. Genet. 9, 141–145.
Park, J.H., and Hall, J.C. (1998). Isolation and chronobiological analysis of a neuropeptide pigment-dispersing factor gene in Drosophila
melanogaster. J. Biol. Rhythms. 13, 219–228.
Park, J.H., Helfrich-Forster, C., Lee, G., Liu, L., Rosbash, M., and Hall,
J.C. (2000). Differential regulation of circadian pacemaker output by
separate clock genes in Drosophila. Proc. Natl. Acad. Sci. USA 97,
3608–3613.
Plautz, J.D., Kaneko, M., Hall, J.C., and Kay, S.A. (1997). Independent
photoreceptive circadian clocks throughout Drosophila. Science
278, 1632–1635.
Price, J.L., Blau, J., Rothenfluh, A., Abodeely, M., Kloss, B., and
Young, M.W. (1998). double-time is a novel Drosophila clock gene
that regulates PERIOD protein accumulation. Cell 94, 83–95.
Qualmann, B., and Kelly, R.B. (2000). Syndapin isoforms participate
in receptor-mediated endocytosis and actin organization. J. Cell
Biol. 148, 1047–1062.
Renn, S.C., Park, J.H., Rosbash, M., Hall, J.C., and Taghert, P.H.
(1999). A pdf neuropeptide gene mutation and ablation of PDF neurons each cause severe abnormalities of behavioral circadian
rhythms in Drosophila. Cell 99, 791–802.
Reppert, S.M., and Weaver, D.R. (2001). Molecular analysis of mammalian circadian rhythms. Annu. Rev. Physiol. 63, 647–676.
Reuss, H., Mojet, M.H., Chyb, S., and Hardie, R.C. (1997). In vivo
analysis of the Drosophila light-sensitive channels, TRP and TRPL.
Neuron 19, 1249–1259.
Ripperger, J.A., and Schibler, U. (2001). Circadian regulation of gene
expression in animals. Curr. Opin. Cell Biol. 13, 357–362.
Rohrbough, J., Pinto, S., Mihalek, R.M., Tully, T., and Broadie, K.
(1999). latheo, a Drosophila gene involved in learning, regulates
functional synaptic plasticity. Neuron. 23, 55–70.
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.
Salcedo, E., Huber, A., Henrich, S., Chadwell, L.V., Chou, W.H.,
Paulsen, R., and Britt, S.G. (1999). Blue- and green-absorbing visual
pigments of Drosophila: ectopic expression and physiological characterization of the R8 photoreceptor cell-specific Rh5 and Rh6 rhodopsins. J. Neurosci. 19, 10716–10726.
Salinas, A.E., and Wong, M.G. (1999). Glutathione S-transferases—a
review. Curr. Med. Chem. 6, 279–309.
Schaeren-Wiemers, N., and Gerfin-Moser, A. (1993). A single protocol to detect transcripts of various types and expression levels in
neural tissue and cultured cells: in situ hybridization using digoxigenin-labelled cRNA probes. Histochemistry 100, 431–440.
Schibler, U., Ripperger, J.A., and Brown, S.A. (2001). Circadian
rhythms. Chronobiology—reducing time. Science 293, 437–438.
Schopperle, W.M., Holmqvist, M.H., Zhou, Y., Wang, J., Wang, Z.,
Griffith, L.C., Keselman, I., Kusinitz, F., Dagan, D., and Levitan, I.B.
(1998). Slob, a novel protein that interacts with the Slowpoke calcium-dependent potassium channel. Neuron 20, 565–573.
Sehgal, A., Price, J.L., Man, B., and Young, M.W. (1994). Loss of
circadian behavioral rhythms and per RNA oscillations in the Drosophila mutant timeless. Science 263, 1603–1606.
Sehgal, A., Rothenfluh-Hilfiker, A., Hunter-Ensor, M., Chen, Y., Myers, M.P., and Young, M.W. (1995). Rhythmic expression of timeless:
a basis for promoting circadian cycles in period gene autoregulation.
Science 270, 808–810.
Shaw, P.J., Cirelli, C., Greenspan, R.J., and Tononi, G. (2000). Correlates of sleep and waking in Drosophila melanogaster. Science 287,
1834–1837.
So, W.V., and Rosbash, M. (1997). Post-transcriptional regulation
contributes to Drosophila clock gene mRNA cycling. EMBO J. 16,
7146–7155.
So, W.V., Sarov-Blat, L., Kotarski, C.K., McDonald, M.J., Allada, R.,
and Rosbash, M. (2000). takeout, a novel Drosophila gene under
circadian clock transcriptional regulation. Mol. Cell. Biol. 20, 6935–
6944.
Stryer, L. (1988). Biochemistry (New York: W. H. Freeman).
Takei, K., and Haucke, V. (2001). Clathrin-mediated endocytosis:
membrane factors pull the trigger. Trends Cell Biol. 11, 385–391.
Tierney, A.J. (2001). Structure and function of invertebrate 5-HT
receptors: a review. Comp. Biochem. Physiol. A Mol. Integr. Physiol.
128, 791–804.
Truman, J., Taylor, B., and Awad, T. (1993). Formation of the adult
nervous system. In The Development of Drosophila melanogaster,
M. Bate and A. Martinez-Arias, eds. (Plainview: Cold Spring Harbor
Press), pp. 1245–1275.
Valles, A.M., and White, K. (1988). Serotonin-containing neurons in
Drosophila melanogaster: development and distribution. J. Comp.
Neurol. 268, 414–428.
Voet, D., and Voet, J. (1990). Biochemistry (New York: John Wiley
and Sons).
Walter, M.F., Zeineh, L.L., Black, B.C., McIvor, W.E., Wright, T.R.,
and Biessmann, H. (1996). Catecholamine metabolism and in vitro
induction of premature cuticle melanization in wild type and pigmentation mutants of Drosophila melanogaster. Arch. Insect Biochem.
Physiol. 31, 219–233.
Wang, Q., Hasan, G., and Pikielny, C.W. (1999). Preferential expression of biotransformation enzymes in the olfactory organs of Drosophila melanogaster, the antennae. J. Biol. Chem. 274, 10309–
10315.
Circadian Gene Expression in the Drosophila Head
671
Williams, J.A., and Sehgal, A. (2001). Molecular components of the
circadian system in Drosophila. Annu. Rev. Physiol. 63, 729–755.
Yang, Z., and Sehgal, A. (2001). Role of molecular oscillations in
generating behavioral rhythms in Drosophila. Neuron 29, 453–467.
Yang, Z., Emerson, M., Su, H.S., and Sehgal, A. (1998). Response
of the timeless protein to light correlates with behavioral entrainment
and suggests a nonvisual pathway for circadian photoreception.
Neuron 21, 215–223.
Yin, D. (2000). Is carbonyl detoxification an important anti-aging
process during sleep? Med. Hypotheses 54, 519–522.
Young, M.W., and Kay, S.A. (2001). Time zones: a comparative genetics of circadian clocks. Nat. Rev. Genet. 2, 702–715.
Zhou, Y., Schopperle, W.M., Murrey, H., Jaramillo, A., Dagan, D.,
Griffith, L.C., and Levitan, I.B. (1999). A dynamically regulated 14–
3-3, Slob, and Slowpoke potassium channel complex in Drosophila
presynaptic nerve terminals. Neuron 22, 809–818.