Identification of Circadian-Clock-Regulated Enhancers and Genes of Drosophila melanogaster by Transposon Mobilization and Luciferase Reporting of Cyclical Gene Expression

Corresponding author: Ralf Stanewsky, Institut für Zoologie, Lehrstuhl für Entwicklungsbiologie, Universitätsstraße 31, 93040 Regensburg, Germany., ralf.stanewsky{at}biologie.uni-regensburg.de (E-mail)

Communicating editor: J. J. LOROS

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

A new way was developed to isolate rhythmically expressed genes in Drosophila by modifying the classic enhancer-trap method. We constructed a P element containing sequences that encode firefly luciferase as a reporter for oscillating gene expression in live flies. After generation of 1176 autosomal insertion lines, bioluminescence screening revealed rhythmic reporter-gene activity in 6% of these strains. Rhythmically fluctuating reporter levels were shown to be altered by clock mutations in genes that specify various circadian transcription factors or repressors. Intriguingly, rhythmic luminescence in certain lines was affected by only a subset of the pacemaker mutations. By isolating genes near 13 of the transposon insertions and determining their temporal mRNA expression pattern, we found that four of the loci adjacent to the trapped enhancers are rhythmically expressed. Therefore, this approach is suitable for identifying genetic loci regulated by the circadian clock. One transposon insert caused a mutation in the rhythmically expressed gene numb. This novel numb allele, as well as previously described ones, was shown to affect the fly's rhythm of locomotor activity. In addition to its known role in cell fate determination, this gene and the phosphotyrosine-binding protein it encodes are likely to function in the circadian system.

RHYTHMIC gene expression is a crucial feature of all circadian clocks described so far at the molecular level (reviewed by YOUNG and KAY 2001 ). In Drosophila the rudiments of clock functioning are as follows (WILLIAMS and SEHGAL 2001 ), with emphasis on components that are conceptually and experimentally connected to the current study: Four factors cooperate to generate sustained molecular oscillations with a period of 24 hr. Two of these clock genes, period (per) and timeless (tim), begin to be transcribed in the late morning; their protein products PERIOD (PER) and TIMELESS (TIM) reach their maximum levels in the late night. During the night both proteins enter the nucleus, where they interact with their transcriptional activators CLOCK (CLK) and CYCLE (CYC). CLK and CYC bind as a dimer to E-box sequences present in the 5'-flanking regions of both per and tim. The interaction of PER and TIM with the CLK/CYC dimer leads to a repression of its activating function and results in decreasing levels of per and tim expression. Consequently, PER and TIM levels decline due to degradation of both proteins, ultimately resulting in a release of CLK and CYC inhibition. Such proteolysis, which must be associated with rather short half-lives of the proteins if they are to exhibit abundance fluctuations, is influenced by post-translational modifications affected by the DOUBLE-TIME (DBT) and SHAGGY kinases; among the substrates of these two enzymes are PER and TIM, respectively (KLOSS et al. 1998 ; MARTINEK et al. 2001 ).

Not all clock components exhibit daily fluctuations in their abundance: Products of the cyc and dbt genes are temporally flat (e.g., KLOSS et al. 1998 ; RUTILA et al. 1998B ). But the existence of pacemaking transcriptional regulators that do cycle suggests that they could control—in addition to their own rhythmic expression—oscillations of downstream factors functioning on circadian output pathways. This has already been demonstrated for cyanobacteria, fungi, and mammals (reviewed by JOHNSON and GOLDEN 1999 ; LOROS and DUNLAP 2001 ; REPPERT and WEAVER 2001 ). A few clock-output candidates in Drosophila have been identified in molecular screens for rhythmically expressed genes [e.g., Drosophila rhythmically expressed gene 5 (Dreg-5; VAN GELDER and KRASNOW 1996 ) and Circadianly regulated gene-1 (Crg-1; ROUYER et al. 1997 )]. However, mutant forms of these clock-controlled genes (ccgs) are few and far between (although see the takeout case of SAROV-BLAT et al. 2000 ). Certain output variants that correspond to genes whose mRNAs do not happen to cycle are known—via phenotypic screening (NEWBY and JACKSON 1993 ) or retrospective mutant identification (MCNEIL et al. 1998 ; PARK and HALL 1998 ; RENN et al. 1999 ; PARK et al. 2000 ). There is a flipside of this coin: Would molecular detection of a rhythmically expressed gene necessarily define a ccg or, instead, a factor that operates at least in part as a component of the clockworks? The vrille (vri) gene in Drosophila, whose chronobiological significance was initially pointed to by the fruits of a screen for cycling RNAs, may indeed play the dual roles just indicated (BLAU and YOUNG 1999 ).

Identification of an additional rhythm-related gene in Drosophila, cryptochrome (cry), showed how novel circadian factors can be discovered by combining mutagenesis with molecularly based phenotyping. Thus, a cry mutant was recovered in a screen that involved luminescence rhythms emanating from a per-luciferase (per-luc) transgene (STANEWSKY et al. 1998 ). The short half-life of this reporter enzyme (PLAUTZ et al. 1997 ; STANEWSKY et al. 1997 ) facilitated the application of per-luc for this and other chronobiological purposes. Analysis of the cryptochrome mutant, detected in a strain that exhibited no per-luc cycling, revealed that the gene is involved in daily resetting of the circadian clock mediated by natural cycles of light and darkness (reviewed by HALL 2000 ). cry turned out to be yet another gene for which abundances of the encoded mRNA and protein are subjected to circadian regulation (EMERY et al. 1998 ; EGAN et al. 1999 ; ISHIKAWA et al. 1999 ).

The rapid throughput permitted by real-time reporting suggested a further application of this technology, in which new genetic variants would themselves mediate molecular cycling. Thus, we adapted the enhancer-trap technique to identify rhythm-related factors by transposon-induced variants that would elicit daily oscillations of luciferase activity. Enhancer trapping has a long successful history in identifying new genes and their functions in Drosophila (reviewed by BELLEN 1999 ). This method is based on the genetic mobilization of transposable elements that have the tendency to insert in the 5' regulatory region of genes. The enhancers therein can positively influence reporter-gene activity from the minimal promoter present in most detector constructs, which often reflects at least aspects of the genes' spatial expression pattern (BELLEN 1999 ). Genes in the vicinity of the transposon insertion can easily be cloned by plasmid rescue, inverse PCR, or both (WILSON et al. 1989 ). Even though a small proportion of P-element insertions cause overtly defective phenotypes (only 10% lead to visible abnormalities or lethality; BELLEN 1999 ), some of them could induce rhythm-related behavioral defects: Rest-activity cycles (e.g., RENN et al. 1999 ) or periodic eclosion (e.g., NEWBY and JACKSON 1993 ) could be anomalous. Moreover, in many cases more severely mutated forms of the "trapped" genes can be induced by imprecise excision of the transposon in the presence of a stable source of transposase (ROBERTSON et al. 1988 ).

Against this background, we designed an enhancer-trap vector containing a luciferase-encoding sequence. This reporter was designed to be under the control of a weak constitutive promoter (Fig 1A), but genetic mobilization of the luc-containing transposon should allow detection of enhancers that normally control rhythmically expressed genes (Fig 1B). An approach similar to the temporal enhancer-trapping tactic we now report for Drosophila revealed that essentially all genes of the cyanobacterium Synechococcus are expressed rhythmically (LIU et al. 1995 ). Here we show that luminescence-based enhancer trapping is also an efficient method to detect circadianly regulated enhancers in a higher organism. The genes connected to these elements are candidates for factors that encode cyclically expressed input functions (such as CRY), pacemaker molecules (such as PER, TIM, CLK, and VRI), or output functions (such as DREG-5, CRG-1, TAKEOUT, and VRI). Uncovering novel factors in the latter category may be especially important, because little is known about the manner by which core molecular oscillators are linked to overt rhythmicity. In fact, there is a paucity of information about rhythmic biological processes in Drosophila. Apart from the aforementioned behavioral and eclosion cycles, the only other circadian rhythm known in Drosophila is one that involves cyclical sensitivity of the olfactory system in the fly's antennae (KRISHNAN et al. 1999 , KRISHNAN et al. 2001 ). We propose that a genetically oriented search for fluctuating output functions will simultaneously facilitate two investigatory processes, as the following questions are asked: What rhythmic biological phenomenon is suggested by a given enhancer-trapped strain by virtue of the product encoded at the locus and its spatial expression pattern? Will the periodic phenotype that is putatively predicted be abnormal under the influence of the inserted transposon or variants derived from it?

View larger version (18K): In this window In a new window Download PPT slide   Figure 1. (A) Schematic map of the luc-sniffer transposon. The firefly luc cDNA was cloned into the polylinker of a vector called the "universal-sniffer" pEG117 (GINIGER et al. 1993 ). The length of the transposable portion of luc-sniffer is 10 kb. luc is expressed under the control of the weak constitutive promoter, originating from the (natural) P element's transposase-encoding gene (arrow). The reporter gene is upstream of a hsp70 polyadenylation signal and the intronless mini-white+ gene (the latter serving as visible marker for recovery of transformed flies). The bacterial plasmid sequences are surrounded by the unique restriction enzymes indicated, which allow the recovery of genomic fly DNA sequences to both sides of the transposon (see MATERIALS AND METHODS). (B) Schematic view of a hypothetical enhancer-trap insertion. The clock-controlled enhancer acts on the promoter in the luc-sniffer, resulting in rhythmic luciferase expression. In addition, it regulates rhythmic expression of a gene upstream (but not one downstream) of the insertion site. Given the uncertainty about which of the neighboring loci is controlled by the enhancers, both candidates can be analyzed for temporal RNA expression of the native mRNAs they encode (see text).

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Generation of the luc-sniffer construct and P-element transformation:
The firefly luciferase (luc) cDNA isolated from Photinus pyralis was cloned from pJD261 (LUEHRSEN et al. 1992 ) into the SalI/KpnI sites of pBluescriptII SK (Stratagene, La Jolla, CA) from which parts of the polylinker (between and including the HindIII and XbaI sites) had been removed. The 1.8-kb luc fragment was then cloned into the NotI and KpnI sites of the P-element vector pEG117 carrying the mini-white⁺ gene as visible marker (GINIGER et al. 1993 ). In the final luc-sniffer construct luciferase expression is under the control of the weak constitutive transposase promoter (Fig 1A). Transformations of y Df(1)w embryos (hereafter referred to as y w) were performed using standard techniques (e.g., RUBIN and SPRADLING 1986 ). Newly created transgenic flies were recognized by w⁺-mediated eye color. Transposase was supplied by coinjection of the helper plasmid pUChs2-3 (LASKI et al. 1986 ). Two independent transformant lines were isolated, both located on chromosome 2.

Mutagenesis:
To obtain an X-chromosomal jump-start line suitable for generating new autosomal insertion lines, one of the two original insertions on chromosome 2 was mobilized by crossing it to a transposase-producing 2-3 strain (ROBERTSON et al. 1988 ). Thirty-five X-chromosomal insertion lines were recovered, from 4 of which we determined the frequency of transposition after crossing females of these strains again to 2-3. Next, 20–50 individual crosses were set up, in each of which one male carrying both the luc-sniffer and 2-3-encoded transposase was mated to y w females. The transposition frequency was determined by scoring the progeny for orange or red-eyed (mini-w⁺) males, which are produced only if a transposition from the X chromosome to an autosome occurred. One line (X-90) with a frequency of 93% (meaning that almost every single cross resulted in a new insertion line) was used as starter line for the mutagenesis.

The mutagenesis scheme is shown in Fig 1 of the supplementary material at http://www.genetics.org/supplemental. Males homozygous for a third chromosome carrying the transposase gene (2-3) and the additional dominant homozygous-viable marker Kinked (Ki; LINDSLEY and ZIMM 1992 ) were crossed to homozygous X-90 females. Single F₁ males from the progeny carrying the luc-sniffer and one copy of the transposase-encoding chromosome were crossed to virgins heterozygous for a dominantly marked (Bl) second chromosome and the In(2LR)O, Cy (CyO) balancer. In the F₂, crosses were screened for the presence of mini-w⁺ males, indicating that a transposition event from the X chromosome to one of the autosomes had occurred. Such individual males were again crossed to Bl/CyO virgins to generate stable stocks and to determine the chromosome of insertion. In the F₃, four males of each line (heterozygous for the luc-sniffer) were screened for rhythmic expression of luciferase (see below). In case the insertion was not on chromosome 2, rhythmically expressing lines were balanced by crossing transgenic males to females carrying third-chromosomal marker-bearing and balancer chromosomes [H and In(3LR)TM3, Sb, respectively].

Isolation of a novel tim loss-of-function allele, tim⁰³, and generation of a recombinant chromosome carrying this clock mutation and a novel enhancer-trapping transposon:
A novel timeless loss-of-function allele was isolated after ethyl methanesulfonate mutagenesis of a strain expressing a period-luciferase transgene (see STANEWSKY et al. 1998 for details of the screening procedure). The homozygous mutant strain exhibited arrhythmic luciferase expression and locomotor activity. Complementation analysis with flies carrying the tim⁰¹ mutation (SEHGAL et al. 1994 ) indicated that the tim gene is mutated in the novel strain. Behavioral arrhythmicity of this mutant could be rescued by introducing a tim rescue transgene (cf. RUTILA et al. 1998A ) into flies carrying tim⁰¹ over the newly isolated tim allele: Of 16 trans-heterozygous males without the rescue transgene, all were arrhythmic, whereas 9 out of 10 male flies harboring one copy of the rescue transgene showed robust locomotor rhythms in dark:dark (DD) cycles (average period 23.4 hr ± 0.1). This demonstrates that a novel tim allele was isolated. Moreover, Western blot analysis revealed that no TIMELESS protein is produced in this mutant. Sequencing the region mutated in the tim⁰¹ allele (MYERS et al. 1995 ) showed that in the novel tim allele this region is not affected. Hence, the isolated mutation represents a novel loss-of-function mutation in the tim gene and was therefore dubbed tim⁰³ [tim⁰¹ is an intragenic mutation; the so-called tim⁰² allele is a deletion of the entire locus along with neighboring second-chromosomal genes (MYERS et al. 1995 )]. To obtain recombinants between tim⁰³ (cytological map position 23F3-5) and the luc-sniffer insertion in one of our enhancer-trap lines (1-17), we crossed tim⁰³ males to 1-17/CyO virgins that were homozygous white. Since the tim⁰³ chromosome was marked with black (b; cytological map position 34B) and the transposon of line 1-17 mapped to 57A on chromosome 2, we selected for recombinants carrying the mini-w⁺ gene and b. This was possible only after backcrossing the recombinant chromosomes to the tim⁰³ b strain, since homozygous 1-17 flies are lethal due to the insertion of the luc-sniffer into a vital gene (see RESULTS). Individuals from recombinant b 1-17/CyO strains were then crossed to tim⁰¹ and male progeny were tested for the presence of locomotor activity rhythms to assure that they carry the tim⁰³ mutation (indicated by arrhythmic behavior of the tim⁰³/tim⁰¹ mutant flies); note that 1-17/+ flies exhibit normal locomotor rhythms (Table 5).

Generation of revertants of enhancer-trap lines 1-17 and 90-3:
All flies used for the generation of revertants carried the y w markers on their X chromosomes (see above). For the homozygous lethal line 1-17, the transposon was mobilized by crossing 1-17/CyO males to CyO/+; Ki 2-3/+ virgins. F₁ 1-17/CyO; Ki 2-3/+ males with pigmented eyes were individually mated with Bl/CyO virgins, and the F₂ was screened for white-eyed 1-17/Bl jump-out males. These males were crossed back to Bl/CyO females to generate 1-17/CyO males and females, which were then crossed together to check for homozygous viability. For line 90-3, homozygous males were mated with CyO/+; Ki 2-3 virgins. The next crosses were analogous to line 1-17. Because homozygosity for the luc-sniffer in line 90-3 does not cause a lethal phenotype, jump-out lines had to be screened molecularly for precise excisions. PCR using genomic DNA isolated from homozygous jump-out animals was performed. Primers were chosen from both sides of the transposon such that a 500-bp PCR product resulted in the y w control line; no product could be amplified in the original enhancer-trap line 90-3 due to the luc-sniffer insertion. Putative revertants were isolated on the basis of the assumption that a perfect revertant should again give a PCR product of 500 bp. Analysis of the PCR products from two putative revertants (90-3^rev1 and 90-3^rev2) showed that their sequences were identical to the relevant autosomal ones in the y w strain.

Plasmid rescue:
Genomic sequences flanking the sites of the luc-sniffer insertions were isolated by plasmid rescue (cf. WILSON et al. 1989 ), using the restriction enzymes EcoRI or SacII (for 3' sequences) and BamHI, BglII, or PstI (for 5' sequences). After digestion and ligation DNA was transformed into Escherichia coli XL1-Blue electrocompetent cells (Stratagene). For each rescue, plasmid-DNA from at least three colonies was isolated and subjected to restriction analyses. On the basis of the restriction pattern, representative clones were chosen for sequence analysis (see below). These data were used to perform BLAST and PSI BLAST searches (ALTSCHUL et al. 1990 , ALTSCHUL et al. 1997 ) against the nucleotide and protein sequence databases of the Berkeley Drosophila Genome Project and the National Center for Biotechnology Information, respectively (http://www.fruitfly.org/; http://www.ncbi.nlm.nih.gov/). Genomic location of the identified genes associated with the luc-sniffer insertions was determined by GadFly (FLYBASE 1999 ). To verify the transposon location, in situ hybridizations to polytene chromosomes were effected using linearized digoxygenin-labeled luciferase DNA as a probe (BLACKMAN 1996 ). In addition, Southern blots with digested genomic DNA of luc-sniffer insertion lines were performed to confirm the restriction pattern of the rescue constructs and to check for multiple insertions. In two cases the results of in situ hybridizations and Southern blots indicated the presence of more than one transposon in the genome. These lines were excluded from further analysis.

Nucleotide sequencing:
To determine the terminal sequences of plasmid-rescued genomic fragments the following primers were used: For EcoRI, SacII rescues 5'-AGTGGATGTC TCTTGCCGACG-3' (proximal ends) and 5'-GTGCCACCTGA CGTCTAACGAAACC-3' (distal ends); for BamHI BglII, and PstI rescues 5'-CCTCTCAACAAGCAACGTGCACT-3' (proximal ends) and 5'-TATAGTCCTGCTGGGTTTCGCCACC-3' (distal ends). Dye primer sequencing (Thermo Sequenase fluorescent-labeled primer cycle sequencing kit, Amersham Biosciences, Freiburg, Germany) was carried out on an ALFexpress DNA sequencer (Pharmacia Biotech).

RNA detection assays:
Total RNA was prepared from 50 heads (males and females) or from 10 male bodies per time point as described in ZENG et al. 1994 . RNase protection assays were performed as described in EMERY et al. 1998 with the following modifications: A total of 5 x 10⁵ cpm of the rp49 probe and 1 x 10⁶ cpm of all other probes were used in the hybridization reactions. Templates for the antisense RNA probe synthesis were generated by RT-PCR. The following primers were used to amplify the probes from total RNA extracted from y w flies: CG13432, 5'-TCCCACAAGGTGC ACTGTTC-3' and 5'-TCCGATGACGTATGCGTATC-3'; CG6 145, 5'-CACTCACATTGGCGCACACATAGG-3' and 5'-CATG TGCTTCTCCTGCACCAGCCA-3'; numb, 5'-GTCACACACG CACGAACCACTCGAGCG-3' and 5'-CAACGTCTGCTGACG GAAGGATCCGGTC-3'; twins/CT19500, 5'-CGGAATTCCAA AGTGCCTGTGCCCAAGA-3' and 5'-CGGGATCCTTGCTGG CTTCTTGGCTCCAT-3'; twins/CT36963, 5'-CGGAATTCGT AAAGTGCGAAATTTGCAAC-3' and 5'-CGGGATCCACCGA ACAGTTTGCGTCGATT-3'. A different probe was used in Northern blot experiments for detection of both the CT19500 and the CT36963 transcripts; this labeled DNA fragment stemmed from amplified cDNA isolated from Canton-S wild-type flies using the primers 5'-GTCAGTGAGCGTGACAAGTC-3' and 5'-GAATGAGGCGTGATCGTAGT-3': CG2207, 5'-GTGA ATTCTTGGTGATCCAAAAACCCTCAG-3' and 5'-CATCAAGC TTAGCCTTTTTCTCCGGCG-3'. To proceed with RNase protection assays, the PCR products were digested with HaeIII (CG13432), BamHI (CG6145), XhoI/BamHI (numb), EcoRI/BamHI (twins, both CT19500 and CT36963), or EcoRI/HindIII (CG2207) and subcloned into pBluescriptII SK (Stratagene). The antisense probes were transcribed from these constructs in the presence of [³²P]UTP using the T3, T7, or SP6 RNA polymerases. Riboprobes protect the following regions of mRNAs: CG13432, 229 bases of transcript CT32789 (the last 68 bases of exon 3 and the first 161 bases of exon 4); CG6145; 245 bases of transcript CT19307 (the last 146 bases of exon 1 and the first 99 bases of exon 2) and in addition 99 bases of a second transcript generated from the same gene (corresponding to the first 99 bases of exon 2 of expressed sequence tag LP03268); numb, bases 825–1472 (648 nucleotides) of the transcript are referred to as "zygotic" and bases 476–1076 (601 nucleotides) of the transcript are referred to as "maternal" (numbers and denotations according to UEMURA et al. 1989 ); twins/CT19500, 248 bases (complete exon 2); twins/CT36963, 189 bases (exon 2); CG2207: 568 bases of transcript CT7302. Primer sequences and other information regarding the generation of templates for riboprobe synthesis for the remaining genes listed in Table 4 are available on request. Gels were quantified with a Cyclone Storage Phosphor System phosphoimager (Packard, Meriden, CT) and OptiQuant analysis software.

Luciferase monitoring:
Bioluminescence measurements of individual live flies were similar to those described in STANEWSKY et al. 1997 . Every other well of 96-well microtiter plates (Optiplate; Packard, Meriden, CT) was filled with 100 µl food consisting of 1% agar, 5% sucrose, and 15 mM Luciferin (Biosynth, Staad, Switzerland) dissolved in H₂O. After entrainment for at least 3 days to a 12 hr:12 hr (12:12) light:dark (LD) regime, individual flies were placed in the food-containing wells and covered with a small plastic dome to reduce locomotor movements (cf. STANEWSKY et al. 1997 ). Luminescence emanating from each well was measured once per hour for 5–6 days in LD (same as the entrainment conditions) or for 3 days in LD followed by 2–3 days in constant darkness (DD). The resulting data were analyzed both by visual inspection of the plotted time series and by fast Fourier transform-nonlinear least-squares analysis (FFT-NLLS) (cf. PLAUTZ et al. 1997 ). The outcomes of applying these functions are period and phase values, whereby the latter describes the peak time of luciferase expression during a 12:12 LD cycle. A metric called "relative amplitude error" (rel-amp) for each fly is obtained by dividing the 95% confidence interval of the amplitude estimate by the amplitude estimate (ratio of amplitude error to most probable amplitude). This value ranges from 0 to 1, where 0 indicates a rhythm with infinite precision (zero error) and 1 indicates a rhythm that is not statistically significant (error exceeds amplitude estimate). The rel-amp is used to determine the significance of a given rhythm: rel-amps <0.7 indicate that the bioluminescence rhythm is due to rhythmic gene expression with 95% confidence (see STANEWSKY et al. 1997 ). Therefore all flies with rel-amps >0.7 were considered to be arrhythmic for reporter-gene expression.

Behavior:
Locomotor activity of adult males was monitored automatically as described in HAMBLEN et al. 1986 . The data were processed and analyzed as described in that report and in HAMBLEN-COYLE et al. 1992 . Generally, flies were entrained for 1 day in 12:12 LD at 25° and then assayed for locomotor activity for the next 5 days in the same LD regime, followed by 7 days in constant darkness (DD). Activity periods in DD were analytically determined by ² periodogram analysis ( = 0.05). The program also indicates the strength of the behavioral rhythm (cf. EWER et al. 1992 ) by computing "power" values (roughly the height of the periodogram peak) and the number of 0.5-hr bins crossing the significance line ("width"). Only flies showing periods in combination with powers 20 and width 2 were considered significantly rhythmic and had their period values listed in Table 5 (averages for all rhythmic flies from a given genotype).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Enhancer-trap mutagenesis:
We generated a modified version of the P-lacW enhancer-trap construct (BIER et al. 1989 ) by replacing the transposase-lacZ fusion gene with sequences encoding firefly luciferase. In the final luc-sniffer construct, designed to sniff out regulatory elements related to temporal control of gene expression, luciferase is expressed under control of the weak constitutive P-transposase promoter (Fig 1; bottom of Fig 2). After germline transformation, two autosomal transgenic lines were recovered. One line was used to mobilize the P element genetically by crossing it to a fly strain that constitutively expresses transposase (see MATERIALS AND METHODS). An X-chromosomal luc-sniffer insertion line (X-90) was isolated, found to exhibit temporally flat levels of luciferase, and used as a jump-start line for mutagenesis (Fig 1 of the supplementary material at http://www.genetics.org/supplemental). This was performed by remobilization of the luc-sniffer element and subsequent screening for novel autosomal insertion lines (Fig 1 of supplementary material at http://www.genetics.org/supplemental). A total of 1176 lines were generated, among which the luc-sniffer was then inserted in chromosome 2 (n = 550) or in chromosome 3 (n = 626). To determine whether expression of the reporter construct in a given autosomal location is under the control of an enhancer driving rhythmic gene expression, four males of each line were tested for rhythmic bioluminescence during 4–5 days in 12 hr light:12 hr dark (12:12 LD) conditions (Fig 2). The overall reduction of signal levels occurring during the course of such an experiment (Fig 2, Fig 3, and Fig 6) is due to substrate depletion, because the effect can be compensated for by supplying fresh luciferin after flies have been fed the initial substrate-containing food for several days (PLAUTZ et al. 1997 ).

View larger version (22K): In this window In a new window Download PPT slide   Figure 2. Examples of enhancer-trap lines exhibiting significant bioluminescence rhythms or arrhythmic reporter-gene activity. The 71 lines showing rhythmic luciferase expression in a circadian pattern were grouped in classes I, II, or III according to the robustness of the observed rhythms. Rhythm strength (expressed as rel-amp) and peak phase [expressed in hours relative to the Zeitgeber time (ZT)] of expression were determined analytically as described in MATERIALS AND METHODS. ZT0 is conventionally the time of lights on in a 12:12 LD cycle and ZT12 is time of lights off. The bottom graph depicts an arrhythmically expressing line, typical for the majority of lines recovered in the screen. The solid and open bars below the plots indicate when the lights were off and on, respectively, during such cycles. CPS, counts per second.

View larger version (44K): In this window In a new window Download PPT slide   Figure 3. Clock control of per-regulated and enhancer-driven bioluminescence rhythms. To determine whether rhythmic expression in a given enhancer-trapped line is affected by circadian-pacemaker mutations, the different class I enhancer-trap lines were crossed into period-altering (perT, top graphs in A–D) or rhythm-eliminating (per01, center in A–D) mutant genetic backgrounds. In addition, expression was analyzed in clock-normal backgrounds in LD conditions (top and center in A–D) and after transfer from LD to constant darkness (DD, bottom in A–D). Solid and open bars are as in Fig 2; shaded bars indicate when the lights would have been on in an LD cycle. (A) Bioluminescence rhythms of the plo transgenic type, in which luciferase is expressed under control of the per promoter (cf. BRANDES et al. 1996 ). Rhythms are phase advanced in perT, eliminated in per01, and continue (although dampened) in DD, indicative of strong clock control of per gene expression. (B) Similar results as in A were obtained for luc-sniffer line 8-35. (C) Enhancer-trap line 1 shows advanced peak expression in a perT genetic background, but weak rhythmicity is retained in per01. (D) Line 90-14 shows only mild phase advances of expression in perT and residual rhythmicity in per01 (cf. Table 2B). 90-14 expression was caused to be aperiodic by tim01 (cf. Table 2C) and continued to be rhythmic in DD in a clock-normal genetic background (cf. Table 2A).

View larger version (11K): In this window In a new window Download PPT slide   Figure 4. Overview of clock regulation in class I enhancer trap lines. Rhythmic bioluminescence of all such lines (see Table 2) was analyzed in two different clock-mutant backgrounds and in DD to determine the extent to which these rhythms are clock controlled. Luminescence cycling mediated by the per-luc-only (plo) transgenic type (in which per promoter sequences are fused directly to luc) is plotted (1) to exemplify circadian-pacemaker regulation of the period gene's 5'-flanking DNA. In addition, enhancer-trap lines in which a rhythmically expressed gene was identified in the vicinity (cf. Table 4) are indicated: (2) 1-17, (3) 7-32A, (4) 90-3, (5) 1-45, and (6) 90-14 (although cycling of the gene trapped in the last of these lines turned out to be irreproducible; see text). (A) Change of peak-phase luc expression in LD cycles in clock-normal per+ (open circles) vs. perT (solid circles) genetic backgrounds. Except for the plo transgenic type (1), all lines were sorted according to their peak phase in a clock-normal genetic background (cf. Table 2). ZT, Zeitgeber time in hours. (B, top) Average percentage of rhythmic individuals in each line in an LD cycle in clock-normal per+ (open circles) and per01 (shaded circles) genetic backgrounds. (B, bottom) Percentage of individuals showing circadian expression of luc in DD per+ (solid circles). Note that some class I lines exhibit stronger rhythmicity in DD compared with the plo transgenic type.

View larger version (48K): In this window In a new window Download PPT slide   Figure 5. Temporal RNA-expression profiles of genes in the vicinity of clock-regulated enhancers. Total head or body RNA (as indicated) from y w, Canton-S wild-type, or different clock-mutant backgrounds was isolated at the different Zeitgeber times (ZT), indicated above each gel image. The RNA amount at each ZT was determined by RPAs using gene-specific riboprobes (see MATERIALS AND METHODS). In addition, a probe detecting the constitutively expressed ribosomal rp49 gene was included to control for RNA loading. Band intensities were quantified (after standardization to the rp49 signal) using a phosphoimager. Solid and open bars are as in Fig 2. (A, top) Head RNA expression of CG13432 (identified by line 1-17) in a clock-normal background with 2-hr time resolution during a 12:12 LD cycle. To the right, expression in four different arrhythmic clock mutants is shown. (Bottom) Quantification of two—or three for ZT11 and ZT23—independent experiments with 2-hr time resolution. Right, quantification of signals obtained from three independent experiments involving the various clock-mutant backgrounds compared to wild type. Maximum expression in a clock-normal genetic background was set to 1.0. (B) Head RNA expression of CG6145a and CG6145b transcripts (identified by line 90-3), analyzed by RPA with 2-hr time resolution in a 12:12 LD cycle. In addition, in one experiment abundance levels at three different time points in per01 and ClkJrk mutant backgrounds were determined. (Top, right), quantification of three experiments for CG6145a (solid circles) and two experiments for CG6145b (open circles) in a clock-normal genetic background are shown. (Bottom) Quantification of CG6145 transcript levels in clock-mutant backgrounds. Maximum expression for each transcript in a clock-normal genetic background was set to 1.0. (C, top) numb RNA isolated from male bodies at 2-hr resolution in a 12:12 LD cycle. Arrows point to protected fragments of the two alternatively spliced transcripts described (UEMURA et al. 1989 ). Right, numb expression in four different clock-mutant backgrounds at two different ZTs. (Bottom) Quantification of two independent experiments performed with 2-hr resolution. Right, quantification of one experiment in which numb body RNA levels were determined in various clock-mutant backgrounds. Maximum expression in a clock-normal genetic background was set to 1.0.

View larger version (27K): In this window In a new window Download PPT slide   Figure 6. Effects of clock mutations on enhancers regulating circadian gene expression. To verify the results obtained by RPA, bioluminescence expression of class I lines that led to the identification of rhythmically expressed genes in various clock-mutant genetic backgrounds is shown. Expression in a clock-normal background is indicated with a stippled line in each case. Solid and open bars are as in Fig 2. (A) Bioluminescence rhythms normally observed in line 1-17 (identifying CG13432) are abolished by all four clock mutations applied (see also Table 2). Note also that expression in a tim lossof-function background became arrhythmic (cf. Fig 5A and Table 2C). (B and C) Bioluminescence rhythms occurring in the 90-3 (CG6145-identifying) and 7-32A (numb-identifying) lines were also abolished by the per01 mutation, confirming the RPA results.

A total of 71 lines (6%) reproducibly showed bioluminescence rhythms with cycle durations in the circadian range, which was determined after performing a numerical analysis of the raw expression data (Table 1). These results indicate that, in many or all of the 71 lines, luc has come under the control of an enhancer that would naturally mediate rhythmic expression of an endogenous gene in the vicinity of such a cis-acting regulatory element (cf. Fig 1B). For 20 lines the exact positions of the transposons were determined (see below), revealing that 2 lines had the luc-sniffer inserted at almost identical chromosomal locations. The rhythmically expressing lines were then grouped according to the robustness of the observed oscillations. This was not done subjectively, but by a formal analysis whose outcome is the so-called rel-amp, serving as a measure for rhythm strength (see MATERIALS AND METHODS). The 71 lines showing circadian bioluminescence rhythms with associated rel-amp errors <0.7 (cf. STANEWSKY et al. 1997 ) were considered to reflect rhythmic luc gene expression. Flies showing rhythms with rel-amp errors <0.5 were designated class I lines (total of 20, Table 1), representing the lines with the most robust rhythms; those with values between 0.5 and 0.6 were grouped in class II (n = 30); and those with rel-amps between 0.6 and 0.7 were placed in class III (n = 21), the group with the weakest, although still significant, rhythms (Table 1). Averaged bioluminescence rhythms for one representative line of each class are shown in Fig 2, along with one example of the majority of lines (94%) whose expression did not display a significant rhythm (Fig 2, bottom).

Another output from the numerical analysis is the average peak time (here called the phase) of the oscillation for each line. Interestingly, the majority of the rhythmic insertion lines (93%) show their expression maxima between midnight and midday (ZT19 to ZT7, Table 1). In those lines peak expression occurs later compared with transgenic plo flies, in which luc is driven by the promoter region of the clock gene per; peak expression of plo occurs at ca. ZT18 (STANEWSKY et al. 1997 ; Table 2; Fig 4A).

Clock control of enhancer-driven bioluminescence rhythms:
To ask whether the observed oscillations are mainly driven by light or controlled at least in part by the circadian clock, we analyzed all class I lines in per⁰¹ and per^T genetic backgrounds. The former is a per loss-of-function mutation (e.g., YU et al. 1987 ) that eliminates behavioral rhythms in constant conditions (DD; KONOPKA and BENZER 1971 ); flies carrying the latter allele exhibit a dramatically shortened free-running period of 16 hr (KONOPKA et al. 1994 ). In addition, both mutations affect molecular rhythmicity in LD: per⁰¹ results in elimination of rhythmic per RNA expression (HARDIN et al. 1990 ), whereas per^T leads to an earlier-than-normal rise of PER protein compared with the daily upswing of the wild-type protein (HAMBLEN et al. 1998 ), probably caused by an anomalously early rise time for per^T RNA. If enhancer-driven luc rhythms are regulated by per (and therefore by part of the central clock mechanism), they should be influenced by the two per mutations in the same way that per expression itself is altered by these alleles.

Fig 3 shows examples of the mutations' effects on different class I lines as well as on per-promoter-driven luc in the plo transgenic type. Fig 4 and Table 2 give a summary of the results of all class I lines analyzed in different clock-mutant genetic backgrounds. In most lines (80%) the trapped enhancer seems to be tightly controlled by the clock, indicated by a clear phase advance of expression in a per^T genetic background compared with expression in a per⁺ background and by elimination of rhythmic expression by per⁰¹ (Fig 3B and Fig 4; Table 2). In such lines the effects of the clock mutations are in fact similar or indistinguishable from those they exert on per-luc expression in the plo transgenic type (Fig 3A and Fig 4; Table 2). In the remaining 20% of all class I lines expression seemed to be under weak clock control only: For two lines (90-14 and 1-45) rhythmic phases of expression were only mildly advanced by per^T, and rhythmic expression was not thoroughly eliminated by per⁰¹ (Fig 3D; Table 2). In the example shown (line 90-14), 35% of the flies remained rhythmic in per⁰¹, compared with 91% rhythmic flies in a per⁺ genetic background (Fig 3D; Table 2). We also analyzed the effects of a different arrhythmia-inducing clock mutation, a loss-of-function allele of the timeless gene (tim⁰¹; SEHGAL et al. 1994 ). Rhythmic expression was observed in only 8% of the tim⁰¹; 90-14 individuals, indicating a potentially independent role of per and tim gene function in the regulation of this particular enhancer (Fig 3D; Table 2).

Two other lines (1 and 3-50) showed tight clock control based on the effects of per^T (i.e., substantial phase advance of peak expression; Fig 3C; Table 2), whereas the high number of rhythmically expressing flies in per⁰¹ (37 vs. 65% in per⁺ for line 1; 47 vs. 89% for line 3-50) points to weaker influence of the clock (Fig 3C; Table 2). We have no explanation for this substantive discrepancy, but it is possible that the enhancers we classified as "weakly clock controlled" are additionally influenced by light. In a clockless per⁰¹ genetic background, a light-inducible enhancer would still be rhythmically active in LD cycles even if it is simultaneously regulated by the clock.

An independent way to demonstrate clock control of gene expression is to measure molecular rhythms in constant conditions (all dark, no temperature fluctuations), which rules out effects on gene expression caused by environmental fluctuations. A problem with this kind of analysis is that even the molecular rhythms of the per and tim clock gene products dampen rapidly in DD (e.g., STANEWSKY et al. 1997 ). This is probably caused by internal desynchronization of clock gene cyclings within a given individual animal. Nevertheless, DD analysis of per-promoter-driven luciferase expression in the plo transgenic line showed that 48% of the individuals display significant circadian bioluminescence rhythms (Fig 3A and Fig 4B; Table 2; cf. STANEWSKY et al. 1997 ). Thus, we also analyzed expression of the class I insertion lines in DD. In most lines (n = 10) expression was rhythmic in 27–67% of individuals (Fig 4B; Table 2), comparable to what was observed for plo flies. Surprisingly, seven lines showed a higher percentage of rhythmic expression compared with plo. Between 80 and 97% of the individuals from those strains showed rhythmic luc expression in DD. Interestingly, line 90-14—which showed residual rhythmicity in a per⁰¹ genetic background (see above)—was most robustly rhythmic in DD (97% of all individuals tested; Fig 3D and Fig 4B; Table 2), indicating that it is regulated both circadianly and by light (see above). The remaining three lines (3-70, 6-4, and III129) showed rhythmic expression in only 11–13% of all individuals tested. Since these three lines showed tight clock control of expression in the per⁰¹ and per^T genetic backgrounds, we assume that in these cases the rhythmic enhancers are active in tissues that are more sensitive to internal desynchronization (see above). Note also that in lines 3-70 and III129 the luc-sniffer construct is inserted in the same gene, only 51 bp apart from each other (Table 3; see below).

A subset of the enhancer traps was also tested in the per^T genetic background in DD. As already observed in LD, this mutation has a negative effect on rhythmic luciferase expression of all lines (except 7-32A and 90-14; Table 2). The free-running period associated with the luminescence oscillations of the few significantly rhythmic per^T individuals was between 16 and 19 hr (see legend to Table 2).

Identification of genes adjacent to the clock-regulated enhancers:
Given that the great majority of the trapped class I enhancers is regulated by the circadian clock, we set out to isolate the genes in the vicinity of luc-sniffer insertions by plasmid rescue (WILSON et al. 1989 ). First, by in situ hybridization to polytene chromosomes we determined the location of each transposon in all class I lines (Table 3). This cytological mapping revealed that in each of the 20 lines a single insertion event had occurred. The luc-sniffer transposon was designed to allow recovery of genomic DNA sequences in the 3' direction of the insertion, using the SacII and EcoRI restriction enzymes; 5'-flanking material could be recovered after digestion of genomic DNA with BamHI, PstI, or BglII (Fig 1A). The rescued DNA fragments were partially sequenced using P-element-specific primers (see Fig 1 and MATERIALS AND METHODS), and these data were aligned to the whole Drosophila genome sequence (ADAMS et al. 2000 ). This information was then used to confirm the band location determined by in situ hybridization and to identify the nearest genes neighboring the transposon (Table 3). Except for one line (2-49) we were able to isolate genomic DNA sequences next to the luc-sniffer. In total, our insertions occurred near or in (i) 20 genes with a known function, i.e., those connected with a given protein as a whole; (ii) 16 genes that encode proteins with previously apprehended domains or that share homologous sequences from genes with known functions isolated in various organisms; and (iii) 12 genes with unknown functions, not even with appreciable motifs within the overall deduced amino acid sequence (ADAMS et al. 2000 ; Table 3).

In no case had an insertion occurred in the close vicinity of a gene known to function in the circadian system of Drosophila melanogaster (Table 3). Also, we did not identify genes that encode a PAS domain, a protein-protein interaction motif found in several clock proteins (e.g., those encoded by the per, Clk, and cyc genes in this species, as reviewed by YOUNG and KAY 2001 ). Two insertions (3-89, 7-32A) occurred in the vicinity of novel genes containing sequences homologous to those encoding a helix-loop-helix domain. This protein dimerization domain is found within members of transcription factor families, including those crucial for clock function (i.e., the CLK and CYC proteins). Other genes identified by our approach with a potential function in the circadian system include those encoding kinases, phosphatases, and proteases (Table 3). These are intriguing, because several clock proteins undergo daily changes in their phosphorylation pattern (i.e., PER, TIM, and CLK), and such catalytic events are associated with the timed disappearance of these proteins (YOUNG and KAY 2001 ).

Kinases involved in PER and TIM phosphorylation have been described: the casein-kinase I, encoded by the dbt gene (KLOSS et al. 1998 ; PRICE et al. 1998 ), and a glycogen synthase kinase-3 (GSK-3), encoded by shaggy (MARTINEK et al. 2001 ). Additional findings related to DBT and its PER substrate suggest that the former is not the only enzyme that mediates phosphorylation of the latter protein (SURI et al. 2000 ); thus additional rhythm-related kinases await identification, as may have now occurred (Table 3).

A phosphatase with potential relevance for the circadian system is PP2A, whose regulatory subunit B is encoded by the twins (tws) gene (UEMURA et al. 1993 ). tws functions in pattern formation during metamorphosis and is required for normal mitosis in neuroblasts of the larval central nervous system (e.g., MAYER-JAEKEL et al. 1993 ; SHIOMI et al. 1994 ). In line 90-14 (see above) the rhythmically expressed luc-sniffer element is inserted within the tws locus (Table 3). A chronobiological role played by this gene in adults could be related to a known PP2A function in mammals, for which this enzyme has been shown to inactivate the transcription factor cAMP response element binding protein (CREB; HUNTER 1995 ; HAFEN 1998 ). CREB is usually activated by phosphorylation via Ca²⁺- and cAMP-dependent signal transduction pathways (HUNTER 1995 ). In a chronobiologically important mammalian brain structure, the hypothalamic suprachiasmatic nucleus, CREB activation is regulated by the circadian clock (GINTY et al. 1993 ; DING et al. 1997 ). In Drosophila, a homolog of mammalian CREB, dCREB2 is also involved in circadian rhythm function, in that it is required for normal temporal expression of per and tim (BELVIN et al. 1999 ). Moreover, in the chick pineal gland, the RNA encoding the PP2A subunit that is homologous to the one encoded by tws is expressed rhythmically (J. OLCESE, personal communication). There were no obvious additional candidates among the sequenced genes (pointed to by our enhancer trapping) known or suspected to play a role in the circadian system (Table 3).

Identification of novel rhythmically expressed genes:
To determine whether the circadianly regulated enhancers controlling luc rhythms also influence circadian expression of endogenous genes (cf. Fig 1B), we analyzed temporal RNA patterns of a subset of those we identified. To this end we analyzed 12 genes by performing RNase protection assays (RPAs) with total RNA isolated from wild-type flies at at least six different times during a 12:12 LD cycle (Table 4). As protecting probes we generated radiolabeled antisense RNA fragments of the respective genes and one designed to detect the constitutively expressed rp49 gene as a control for equal loading of RNA (see MATERIALS AND METHODS). We assumed that a given gene might be expressed rhythmically in only a subset of the tissues in which its products are made. Therefore we analyzed RNA expression in heads and bodies separately, and given that bioluminescence rhythms were measured in male flies only, we restricted body-RNA analysis to this sex. Among the 12 genes analyzed, 4 showed reproducible RNA oscillations with amplitudes ranging from two- to fourfold (Table 4; Fig 5). This demonstrates that the bioluminescence-based enhancer-trap approach is a suitable way to isolate rhythmically expressed genes in Drosophila. We now describe the most heavily analyzed subset of the sniffed-out genetic loci, from the perspective of their informational contents and temporally varying expression of the gene products.

anon1A4 (CG2207):
The CG2207 DNA sequence is located 3 kb upstream of the luc-sniffer insertion in line 1-45. This gene (sharing no homologies with others) is also known as anon1A4 (an arbitrarily designated "anonymous" factor) and was identified in a screen for fast-evolving genes in Drosophila (SCHMID and TAUTZ 1997 ). Interestingly, per was found in that same screen by virtue of showing a similarly high rate of amino acid substitutions between D. yakuba and D. melanogaster as anon1A4 (this rate was used by SCHMID and TAUTZ 1997 as a measure to determine the speed of evolution for genes analyzed in this study). anon1A4 encodes the nuclear phosphoprotein Df31, which is thought to be involved in chromatin folding (CREVEL et al. 2000 ). CG2201 mRNA isolated from adult heads showed reproducible oscillations, whereas body RNA did not oscillate (Table 4 and data not shown).

CG13432:
In line 1-17 the transposon is inserted in the first intron of CG13432, a gene showing homology to nompA (KERNAN et al. 1994 ). This gene encodes a zona pellucida domain protein, which is required to connect mechanosensory dendrites to sensory structures (CHUNG et al. 2001 ). Analysis of the mRNA encoded by this gene revealed that it is rhythmically expressed in heads (Fig 5A; Table 4), whereas only weak fluctuations were observed in bodies (Table 4 and data not shown). Peak RNA levels occurred late at night (ZT19-ZT23), substantially later than the times of peak expression levels for either per or tim RNA (ca. ZT15). To compare the actual mRNA peaks with those of the luc-reported bioluminescence oscillations, one has to consider that the plo-reported expression peak occurs 3 hr later relative to the per mRNA peak (STANEWSKY et al. 1997 ). Similarly, the peak of bioluminescence in line 1-17 (ZT0.5) occurs several hours after that of the CG13432 RNA, indicating that the luciferase activity in this enhancer-trapped line reflects expression of the endogenous gene (Table 2). To determine whether CG13432 is regulated by the circadian pacemaker, we analyzed RNA levels in different clock-mutant genetic backgrounds. In particular we analyzed expression in per⁰¹, tim⁰¹, Clk^Jrk, and cyc⁰¹, each of which causes behavioral arrhythmicity and elimination of molecular rhythms of clock molecules (YOUNG and KAY 2001 ). Similarly, these mutations abolished RNA rhythms of CG13432, with the intriguing exception of tim⁰¹ (Fig 5A). In three independent tests RNA levels were found to be higher at ZT23 compared with ZT11. To confirm this result we analyzed bioluminescence rhythms of 1-17 flies in the same mutant backgrounds (Fig 6A). Since both tim and CG13432 are located on chromosome 2, a doubly variant chromosome was created by meiotic recombination between a tim mutation and 1-17. For this, we applied a novel loss-of-function tim allele, dubbed tim⁰³ (see MATERIALS AND METHODS for the origin of this mutation and the manner by which this double variant was recovered). Bioluminescence was measured from flies carrying the recombinant chromosome in heterozygous condition with tim⁰¹, which is noncomplementing with tim⁰³ (Fig 6A). In each of the four clock-mutant genetic backgrounds (including the tim-null one) bioluminescence oscillations were abolished (Fig 6A; Table 2C). Assuming that luc cycling in line 1-17 reflects that of endogenous CG13432 expression, this result in conjunction with the actual RNA data (Fig 5A) indicates that CG13432 is clock controlled. We have no explanation for the discrepancy between observed head RNA fluctuations and constitutive bioluminescence expression in a tim-null background (Fig 5A and Fig 6A; Table 2C). It could be that different tissues give rise to the bioluminescence signal compared with the rhythmic head signals stemming from tissue homogenates. This would suggest that the timeless clock gene plays only a minor role in regulation of CG13432 rhythms in the head. To examine this matter further, a more dense CG13432 RNA time course in a tim-null background needs to be conducted. Another 1-17-related issue is that levels of bioluminescence emanating from this luc-sniffer insertion were low in genetic backgrounds that included homozygosity for the Clk^Jrk or cyc⁰¹ mutations (luciferase-mediated counts per second were equivalent to trough levels observed in a Clk⁺ cyc⁺ background; Fig 6A). These findings are consistent with the roles played by these two genes as transcriptional activators (WILLIAMS and SEHGAL 2001 ; YOUNG and KAY 2001 ). Levels of 1-17-mediated luminescence in per⁰¹ or tim⁰¹/tim⁰³ mutant backgrounds were low to medium, suggesting that these genes have only minor effects on the expression level of CG13432 (Fig 6A). These differential effects of Clk and cyc mutations, on the one hand, and per and tim mutants, on the other hand, were not observed at the actual RNA level. Here CG13432 RNA abundances under the influence of either Clk^Jrk or per⁰¹ were found to be near the trough levels observed with clock-normal fly extracts (left-most pair of histogram bars in Fig 5A), whereas the cyc⁰¹ and tim⁰¹ mutations caused CG13432 RNA to stay at the normal peak level (Fig 5A). A reason for this discrepancy could be the apparently different levels of rp49 expression in the various clock-mutant backgrounds (Fig 5A, right), which here complicated an intergenotype comparison of rp49 normalized expression values. Given the low amplitude of normal CG13432 RNA cycling (Fig 5A, left), even small errors associated with such normalizations can obscure the actual RNA levels in the different genetic backgrounds. In any event, cycling of the CG13432 transcript was abolished by most arrhythmia-inducing clock mutations and in terms of reporter-gene cycling by all four clock-mutant alleles applied here, demonstrating that expression of this gene comes under the sway of the circadian pacemaker.

CG6145:
In line 90-3 the luc-sniffer got inserted at the CG6145 locus, whose conceptual protein has homology to NAD kinases (Table 3). Detailed analysis of this transcription unit revealed that two alternatively spliced transcripts are generated from this locus, each encoding a polypeptide with a different N terminus (data not shown, but see MATERIALS AND METHODS). The insertion site is situated at an intragenic site corresponding simultaneously to the first intron of the larger transcript (CG6145a) and to the upstream region of the first exon of the smaller transcript (CG6145b). The riboprobe used is able to detect both transcripts, and RPA analysis revealed that both RNAs are rhythmically expressed in fly heads with slightly different temporal profiles (Fig 5B; Table 4). The CG6145a transcript exhibited a similar temporal expression pattern as that described for per and tim RNAs (see above), with maximum levels observed at about ZT15. The shorter CG6145b transcript peaks 2–6 hr later, reaching its highest levels between ZT17 and ZT21 (a time course similar to the profile described in the previous subsection for the CG13432 transcript). The maximum bioluminescence for line 90-3 occurred 5–6 hr later compared with the plo-mediated (per-reporting) peak (Table 2), indicating that the luc rhythm in this enhancer-trapped type probably reflects expression of the shorter transcript. RNA fluctuations of both transcripts were abolished in genetic backgrounds that included Clk^Jrk or per⁰¹ mutations, demonstrating that CG6145 is a clock-controlled gene (Fig 5B). In agreement with this conclusion, 90-3-mediated luc rhythms were eliminated in a per⁰¹ background and phase advanced in per^T flies (Fig 4 and Fig 6B; Table 2), indicating that the luc-reported expression closely reflects that of CG6145 RNA.

numb (CG3779):
This gene is located immediately upstream of the transposon insertion in line 7-32A. numb is involved in cell fate decisions during embryogenesis and adult sensory bristle formation (reviewed by JAN and JAN 2000 ). A general role for mammalian numb in endocytic processes was suggested (SANTOLINI et al. 2000 ). The riboprobe we generated to detect numb RNA recognizes the two alternatively spliced transcripts, encoding proteins with different N termini (UEMURA et al. 1989 ; also see MATERIALS AND METHODS). We found that both numb transcripts are present in male bodies and that expression is robustly rhythmic with an approximately fourfold amplitude (Fig 5C). Peak expression occurred between the late nighttime and early morning (ZT21 to ZT1), and the lowest levels of numb transcripts were observed between ZT11 and ZT17. numb RNA peaks are therefore delayed by 6–10 hr compared with the per and tim RNA peaks, similar to what has been described for the Clk and cryptochrome RNA time courses (BAE et al. 1998 ; EMERY et al. 1998 ). The later-than-per and -tim peaks were nicely reflected by the 6-hr delay of the bioluminescence peak in line 7-32A compared with that of plo (Table 2). When we examined numb RNA levels in the genetic background of different arrhythmic clock mutants at peak and trough time points of expression (ZT15 and ZT21), in no instance were fluctuations comparable to those in a clock-normal background observed (Fig 5C). These (raw molecular) results are similar to the elimination of bioluminescence rhythms seen for line 7-32A as affected by per⁰¹ (Table 2; Fig 6C). Moreover, the actual RNA levels in this arrhythmia-inducing genetic background stayed at levels comparable to the normal trough abundance; the same was observed for the bioluminescence levels determined for per⁰¹; 7-32A males (Fig 5C and Fig 6C). We also observed numb RNA cycling in temporally collected RNAs isolated from heads, but here the peakto-trough amplitude was less pronounced compared to that of male-body RNA oscillations (data not shown). Taken together, the results stemming from recovery of the 7-32A line show that numb is a clock-controlled, rhythmically expressed gene and that numb RNA levels are reflected by the bioluminescence expression observed in line 7-32A (Fig 4, Fig 5C, and Fig 6C).

The remaining eight genes we analyzed did not show reproducible RNA rhythms (Table 4). In a Northern blot time course of the transcripts from the tws gene, head-mRNA oscillations were observed (data not shown). This result supported the potential function of this gene in the circadian system (see above). Interestingly, tws functions in the same cell fate decision pathway as the rhythmically expressed numb gene (SHIOMI et al. 1994 ), suggesting that both factors could also be involved in the same process in the circadian system; this might in turn be related to the circadian regulation and chronobiological significance of dCREB2 (see above). But when we repeated the temporal RNA analysis of tws expression by performing both Northern blotting and RPA experiments, the transcript rhythms originally observed were irreproducible (Table 4; data not shown). This makes a role for tws in the circadian system ambiguous for the time being (but see DISCUSSION).

Locomotor activity rhythms in flies mutant or potentially mutant for the rhythmically expressed genes:
A useful feature of the enhancer trapping is that the (modified) P element frequently induces a mutation in the targeted gene, which can lead to inferences about the nature of the function encoded at that locus. Therefore we analyzed the locomotor behavior of flies homozygous for a given luc-sniffer insert in cases for which the transposon had inserted near a gene revealed to be rhythmically expressed (i.e., by more than reporter-enzyme oscillations). If the rhythmically expressed gene functions somewhere along the clock-output pathway regulating circadian locomotor activity or upstream of it in the central clock works, one would expect alterations of behavioral rhythmicity in cases for which an insertion interferes with gene function.

In Table 5 free-running periods of the relevant strains are listed along with the percentages of flies exhibiting rhythmic behavior. In no case was drastic alteration of free-running periodicity or overall rhythmicity observed. Enhancer-trap line 1-45 is inserted 3 kb downstream of the rhythmically expressed gene CG2207 and therefore probably does not interfere with the function of that gene. In line 1-17 the element is located in the first intron of CG13432 (see above), and homozygosity for this insertion causes lethality. That developmental death is indeed caused by the luc-sniffer insertion was revealed after recovery of homozygous viable revertants of the element associated with this line (data not shown, but see MATERIALS AND METHODS). We could test only 1-17 heterozygous males for locomotor activity, and those flies turned out to be normal for this behavior (Table 5). In line 90-3 the insertion also occurred in intron 1 of the corresponding rhythmically expressed transcript (CG6145a) or upstream of its first exon (CG6145b; see above). Homozygous 90-3 flies are viable, which means that CG6145 is not a vital gene or that the transposon does not drastically interfere with gene function. Northern blot and RPA analyses suggest that the latter is the case, since both CG6145 transcripts appear to be generated in normal amounts and size in homozygous 90-3 flies (data not shown). The free-running periods of homozygous and heterozygous 90-3 flies were 0.5 hr longer compared with control flies (Table 5). Since revertants created by mobilization of the transposon show the same degree of period lengthening (Table 5), this effect is not caused by the insertion in CG6145, but rather represents a genetic background phenomenon.

In line 7-32 the P insertion occurred immediately upstream of the numb gene. As was the case for line 1-17 this insertion causes lethality in the homozygous condition. Testing flies in which the 7-32 element was placed (separately) over the lethal numb¹