Drosophila Photoreceptors Contain an Autonomous Circadian Oscillator That Can Function without period mRNA Cycling

Abstract

Circadian oscillations in period(per) mRNA and per protein (PER) constitute, in part, a feedback loop that is required for circadian pacemaker function in Drosophila melanogaster. Oscillations in PER are required for oscillations in permRNA, but the converse has not been rigorously tested because of a lack of measurable quantities of per mRNA and protein in the same cells. This circadian feedback loop operates synchronously in many neuronal and non-neuronal tissues, including a set of lateral brain neurons (LNs) that mediate rhythms in locomotor activity, but whether a hierarchy among these tissues maintains this synchrony is not known. To determine whether per mRNA cycling is necessary for PER cycling and whether cyclic per gene expression is tissue autonomous, we have generated per⁰¹flies carrying a transgene that constitutively expressesper mRNA specifically in photoreceptors, a cell type that supports feedback loop function. These transformants were tested for different aspects of feedback loop function includingper mRNA cycling, PER cycling, and PER nuclear localization. Under both light/dark (LD) cycling and constant dark (DD) conditions, PER abundance cycles in the absence of circadian cycling ofper mRNA. These results show that permRNA cycling is not required for PER cycling and indicate thatDrosophila photoreceptors R1–R6 contain a tissue autonomous circadian oscillator.

Genetic screens for mutations that affect circadian rhythms in Drosophila melanogaster have identified two genes that encode components of the circadian timekeeping apparatus, period (per) andtimeless (tim). Mutations in these genes can shorten (per^S,tim^SL), lengthen (per^L), or abolish (per⁰¹,tim⁰¹) circadian rhythms in locomotor activity and eclosion (Konopka and Benzer, 1971; Sehgal et al., 1994;Rutila et al., 1996). In parallel with these behavioral rhythms are molecular rhythms in the abundance of per and timgene products; per and tim RNAs peak early in the evening, whereas per protein (PER) and timprotein (TIM) peak several hours later (Siwicki et al., 1988; Hardin et al., 1990; Zerr et al., 1990; Edery et al., 1994b). Induced alterations in PER or TIM levels can shift the phase of behavioral rhythms (Edery et al., 1994a; Lee et al., 1996; Myers et al., 1996; Zeng et al., 1996), showing that behavioral rhythms are dependent on molecular rhythms.

To account for these molecular rhythms, a negative feedback loop model was proposed in which PER and/or TIM suppress per andtim gene transcription (Hall, 1995; Hardin and Siwicki, 1995; Sehgal, 1995). After the lights are turned off, PER and TIM accumulate in the cytoplasm as heteromeric complexes (Lee et al., 1996;Zeng et al., 1996) and enter the nucleus over the course of a few hours, starting ∼6 hr after the lights are turned off (Curtin et al., 1995). In the nucleus, PER and/or TIM feedback to repressper and (probably) tim gene transcription (Hardin et al., 1990; Zeng et al., 1994; Sehgal et al., 1995). Destruction of PER and TIM proteins, in turn, coincides with the accumulation ofper and tim transcripts during midday and the start of another circadian cycle.

Little is known about how PER–TIM complexes regulate transcription after they enter the nucleus. However, sequences that mediate PER- and TIM-dependent transcriptional cycling have been localized to a 69 bp fragment lying ∼500 bp upstream of the per transcription start site (Hao et al., 1997). Surprisingly, per genomic fragments lacking a promoter (Hamblen et al., 1986; Frisch et al., 1994) or driven by constitutively active promoters (Ewer et al., 1988,1990; Vosshall and Young, 1995) drive PER cycling and rescue locomotor activity rhythms in per⁰¹ flies. Although PER cycling was only measured in 12 hr light/dark (LD) cycles in these studies, such cycling presumably persists because behavioral rescue was measured in constant darkness. Behavioral rescue by these transgenes, however, raises the intriguing possibility that per RNA cycling is not necessary for PER cycling and locomotor activity rhythms.

The per gene is expressed in a variety of neuronal (i.e., photoreceptors, antennae, brain neurons and glia, and thoracic ganglion) and non-neuronal (i.e., gut, Malpighian tubules, testes, and ovary) tissues (Liu et al., 1988; Saez and Young, 1988; Siwicki et al., 1988; Ewer et al., 1992). The circadian feedback loop operates in all of these tissues except the ovary, where per is constitutively expressed (Hardin, 1994). In adults, lateral neurons (LNs) are the only cells within the per expression pattern known to control autonomously a rhythmic output, locomotor activity (Ewer et al., 1992; Frisch et al., 1994). Oscillator autonomy has also been observed in the pupal prothoracic gland (Emery et al., 1997), but whether oscillators operate autonomously in other adult tissues is unknown.

In this report we drive per from a crippled rhodopsin 1 (Rh1) promoter to determine whether permRNA cycling is required for PER cycling and whether photoreceptors contain an autonomous circadian oscillator. PER produced by this transgene cycles in abundance under LD and constant dark (DD) conditions. Because the transgene expresses constitutive levels ofper mRNA, the rhythms in PER show that an oscillator is running through some post-transcriptional mechanism. In addition, this oscillator is operating exclusively in photoreceptors R1–R6, which suggests that these cells contain an autonomous oscillator.

MATERIALS AND METHODS

Plasmid construction. TheRh1(−250)–LacZ,Rh1(−180)–LacZ, andRh1(−120)–LacZ transformation constructs were made as follows. Rh1 promoter fragments starting −250, −180, and −120 bp upstream of the Rh1 transcription start site, respectively, and extending to the Rh1 translation initiation site were generated by PCR usingRh1–per (Zeng et al., 1994) as a template. The Rh1 upstream region primers included theRh1(−250)+XhoI sense primer (5′-GCCTCGAGACTCAAGAATAATAC-3′), theRh1(−180)+XhoI sense primer (5′-GCCTCGAGCCCATTGCGATGTG-3′), and theRh1(−120)+EcoRI sense primer (5′-CCGAATTCGCGGCCGCGGTACCTGTCGACACTTT-3′). The antisense primer at the Rh1 translation initiation site wasRh1(ATG)+BamHI (5′-GCGGATCCATTGTGTTTTGGTTAC-3′). TheRh1(−250)+XhoI/Rh1(ATG)+BamHI andRh1(−180)+XhoI/Rh1(ATG)+BamHI amplification products were digested with XhoI andBamHI, and theRh1(−120)+EcoRI/Rh1(ATG)+BamHI amplification product was digested with EcoRI andBamHI and cloned into the pCaSpeR-β-gal transformation vector (Thummel et al., 1988).

The Rh1(−180)–per transformation construct was generated as follows. A 6.2 kb SalI–XbaI DNA fragment from 13.2 (HA/C) (Rutila et al., 1992), consisting of genomic sequences from the SalI site in exon 3 to an XbaI site ∼2 kb downstream of per transcribed sequences and including a C-terminal hemagglutinin tag, was cloned into Bluescript KS−. A 630 bp DNA fragment from the SalI site at −120 bp of the Rh1 promoter to the SalI site in exon 3 of the per genomic sequence was removed fromRh1–per (Zeng et al., 1994) and inserted at theSalI site upstream of the SalI to XbaIper fragment in Bluescript KS−. The orientation of theSalI fragment was checked via PCR and restriction enzyme digests, and the plasmid having the per coding region driven by the Rh1−120 promoter fragment was namedRh1(−120)–perKS. To make theRh1(−180)–perKS, we first removed anXhoI site close to the XbaI site at the 3′-end of the per genomic sequence by digestion with XhoI, then filling in with the Klenow fragment of DNA polymerase I. AnXhoI–SpeI fragment starting −180 bp upstream of the Rh1 transcription initiation site to the SalI site of per exon 3 was generated by PCR using theRh1(−180)+XhoI sense primer (see above) and the ex3SalI antisense primer (5′–GCAACGCGTTGTCGACCTTCTGGC-3′) and then was digested with XhoI and SpeI. This fragment was inserted into the Rh1(−120)–perKSplasmid after Rh1 sequences between −120 bp of the transcription start site and the SpeI site in theRh1 leader sequence were removed, thereby formingRh1(−180)–perKS. The inserts fromRh1(−180)–perKS were removed by digestion withXhoI and XbaI and were inserted into the CaSpeR4 transformation vector (Thummel and Pirotta, 1991) to form theRh1(−180)–per transformation plasmid.

Fly stocks and germ line transformation. The wild-typeD. melanogaster strain Canton-S and transgenic fly strains were raised on a cornmeal, sugar, agar yeast, and Tegosept (a mold inhibitor) medium at 25°C. P-element-mediated transformation was performed as described previously (Hao et al., 1997). Transformant lines with inserts on the second or third chromosomes were balanced with In(2LR)Cyo or In(3LR)TM2, respectively, and were crossed into a y,per⁰¹,wgenetic background.

Behavioral analysis. Locomotor activity of adult male Canton-S;y,per⁰¹,w;Rh1(−180)–per-1;y,per⁰¹,w;Rh1(−180)–per-2; and y,per⁰¹,w flies were monitored and analyzed as described by Hamblen et al. (1986). Flies were entrained in 12 hr light/dark cycles at 25°C for 72 hr; then the lights stayed off for 7 d. Locomotor activity was monitored from the first day of entrainment, and data collected during constant darkness were analyzed to determine the period and strength of the rhythm (Hamblen et al., 1986). Flies were designated rhythmic or arrhythmic based on the criteria of Ewer et al. (1992).

RNase protection assays. Flies used for time course analyses were entrained at 25°C in 12 hr light/dark cycles for at least 96 hr before collection. For each time point, RNA was extracted as described from either whole heads (Hardin et al., 1990) or eyes (Hardin et al., 1992b; Zeng et al., 1994), and 10 μg of whole-head RNA or 5 μg of eye RNA was used for RNase protection assays as described (Hardin et al., 1990). To make an RNase protection probe, we cloned aper cDNA fragment from the SpeI site in exon 2 to the SalI site in exon 3 into Bluescript KS− vector. The RNase protection probe was linearized by SpeI digestion, and an RNA probe was transcribed with T3 RNA polymerase. It protects an endogenous per fragment of 324 nucleotides (nt) and aRh1(−180)–per-derived transcript of 285 nt. Protected per bands were quantitated using a Fuji BAS50 phosphorimager. As a control for the amount of RNA in each lane, an antisense ribosomal protein 49 (RP49) probe was used in each RNase protection assay (Hardin et al., 1990).

Immunohistochemistry. For LD experiments, homozygousper⁰¹;Rh1(−180)–per-1 flies were entrained in 12 hr LD cycles for at least 4 d and collected at 4 hr intervals. For each time point, flies were immediately embedded into OCT compound (Tissue-Tek) on dry ice, and 10–12 μm horizontal cryostat sections were prepared. A rabbit polyclonal anti-PER antiserum (kindly provided by Ralf Stanewsky) preabsorbed against per⁰¹ embryos was used for immunostaining. Immunostaining detection was performed using a goat anti-rabbit secondary antibody conjugated to horseradish peroxidase as described (Siwicki et al., 1988). For DD experiments, heterozygousper⁰¹;Rh1(−180)–per-1/+ flies were entrained for 4 d in 12 hr LD cycles; the lights were then left off, and eight samples were collected every 4 hr starting at circadian time 4 (CT4) during the first and second days of constant darkness. Sample preparation and immunostaining for flies collected in DD conditions were performed as described for flies collected in LD conditions.

Western blotting and quantitation. Fly head extracts were prepared from either homozygous transgenic flies for LD experiments or from heterozygous transgenic flies for DD experiments. Flies were entrained and collected in LD and DD conditions as described above for the immunohistochemistry and then were subjected to Western blotting analyses (Edery et al., 1994b) with the following modifications: the first antibody is the same used for immunohistochemical staining and was diluted to 1:20,000, and the secondary antibody is anti-rabbit IgG horseradish peroxidase-conjugated antibody (Amersham, Arlington Heights, IL) diluted 1:5000 in blocking solution. X-ray film exposures of Western blots were scanned using OFOTO software and quantitated using National Institutes of Health Image 1.6 software. The level of PER at each time point was taken as the PER signal minus the background in each lane. In each independent time course, the highest PER signal was set to 1.0, and all other time points were normalized to this value. The normalized PER values from three independent time courses of Canton-S andper⁰¹;Rh1(−180)–per-1/+ and from two independent time courses ofper⁰¹;Rh1(−180)–per-2/+ were added together to yield pooled data for each genotype. The peak PER value from the pooled data was then set to 1.0. The normalized pooled data from each genotype were used to generate a curve based on its fit to a polynomial function using NIH Image 1.6 software.

RESULTS

A minimal rhodopsin 1 gene promoter drives constitutive expression in photoreceptors R1–R6

To test whether circadian oscillator function requiresper mRNA cycling and operates autonomously in peripheral tissue (i.e., cells that are not capable of driving locomotor activity rhythms), we had to find an appropriate promoter to driveper expression. Such a promoter should meet the following criteria: (1) it directs gene expression in a tissue that normally expresses per, (2) it expresses a constant level of mRNA, and (3) it drives expression at a level comparable with that ofper. One promoter that meets the first two criteria is theninaE (rhodopsin 1, Rh1) promoter that constitutively expresses mRNA in the six outer photoreceptor cells of each ommatidium (i.e., R1–R6) but not in the inner photoreceptors R7 and R8. However, this promoter expresses high levels of mRNA and was used to show that PER overexpression represses the endogenousper RNA cycling (Zeng et al., 1994).

To make this promoter useful for our purposes, we sought to reduce its level of expression while maintaining its tissue specificity. An earlier study had identified an Rh1 promoter fragment [Rh1(−120/+67)] with these properties (Mismer and Rubin, 1987); however, this promoter fragment was not specific to photoreceptors R1–R6 in our hands (H. Hao, Y. Cheng, and P. E. Hardin, unpublished observations). Therefore, we tested the expression level and tissue specificity of two progressively larger portions of the Rh1 promoter in transgenic flies using a lacZreporter gene (Fig. 1). Flies transformed with the Rh1(−250)–LacZ andRh1(−180)–LacZ constructs, which contain 250 or 180 bp of Rh1 upstream sequences, respectively, and leader sequences up to the initiating ATG, were sectioned and stained for β-galactosidase (β-gal) activity. Both constructs showed specific staining in the eye, but Rh1(−250)–LacZ showed considerably higher levels of expression than didRh1(−180)–LacZ (Fig. 1). Although eachRh1(−180)–LacZ line was expressed specifically in the eye, expression in two of the lines was patched (data not shown). Such a pattern may be caused by position effects because theRh1 promoter is adjacent to the 5′-end of the P-element vector. Because we wanted to minimize expression levels, we used theRh1(−180) promoter for subsequent experiments.

Fig. 1.

 Top. A crippled Rh1promoter produces low levels of β-gal in eyes. A, Schematic representation of theRh1(−250)–LacZ andRh1(−180)–LacZ DNA constructs. TheRh1(−250) and Rh1(−180) promoter fragments, which contain 250 and 180 bp of upstream sequences, respectively, and leader sequences up to the translation initiation site, are represented by black bars. These promoter fragments were fused to lacZ sequences (white bars) at the lacZ translation initiation site.B, Spatial expression and β-gal staining intensity inRh1(−250)–LacZ transgenic flies. β-Gal activity was detected via X-gal staining for 12 hr in a horizontal section through a whole head oriented with the anterior end up. β-Gal activity (blue stain) is present throughout the eyes of this and four other lines. C, Spatial expression and β-gal staining intensity inRh1(−180)–LacZ transgenic flies. β-Gal activity was detected in heads as described inB. β-Gal activity is present throughout the eyes of this and other lines. Two other lines showed eye-specific staining present in patches within the eye.

The Rh1(−180) promoter and its leader sequence were fused to per genomic sequences at the translation initiation site to drive low level per expression specifically in photoreceptors R1–R6 (Fig. 2). Eight independent transgenic lines were obtained and crossed into aper⁰¹ background to eliminate endogenous PER expression. The spatial expression pattern of PER in these transgenic strains was examined by anti-PER immunohistochemical staining. Six of the eight strains showed undesirable perexpression patterns; one expresses PER at high levels, two show PER expression in photoreceptors R1–R6 and the central brain, and three express per in only a subset of R1–R6 cells (data not shown). The other two transgenic strains have the desired PER expression pattern: photoreceptor R1–R6-specific expression at levels similar to that of PER. To confirm that PER is not expressed at levels too low to detect by immunohistochemical methods in the central brains of these two transgenic lines, we monitored the lines for locomotor activity rhythms. Both lines are arrhythmic, indicating that they do not have a functional circadian pacemaker in their brain (Table1). These two lines, designatedper⁰¹;Rh1(−180)–per-1 andper⁰¹;Rh1(−180)–per-2, were used for subsequent molecular analyses.

Fig. 2.

Bottom. Rh1(−180)–perexpresses specifically in photoreceptors R1–R6. A, Schematic representation of theRh1(−180)–per DNA construct. TheRh1(−180) promoter fragment, which contains sequences from 180 bp upstream of the transcription start site to the translation initiation site, is represented by a black bar. The Rh1(−180) promoter was fused toper genomic sequences at the pertranslation initiation site. White bars symbolizeper protein coding exons 2–8, the hatched bar denotes the per 3′-untranslated sequence,thin lines represent per intron sequences, and the dashed line denotes genomic sequences downstream from per. HA refers to the hemagglutinin tag inserted at the C terminus of PER (Rutila et al., 1992). B, Spatial expression of PER inper⁰¹;Rh1(−180)–pertransgenic flies. PER immunoreactivity was detected in flies collected at Zeitgeber time 23 (ZT23), sectioned, and incubated with anti-PER antibodies. A horizontal section of a whole head is oriented with the anterior end up. PER immunoreactivity (dark brown stain) is only present in the nuclei of photoreceptors R1–R6, which are indicated by small arrows. No immunoreactivity is seen in the area of the LNs, indicated by the large arrows. Photoreceptor specificity was obtained in a total of five transgenic strains.

Table 1.

Behavioral analysis of per⁰¹;Rhl(−180)—per transgenic flies

Genotype	No. tested	No. rhy.	No. arr.	Period (hr ± SEM)	Power (±SEM)	Width (±SEM)
Canton-S	24	22	2	23.9 ± 0.1	92.5 ± 10.9	6.0 ± 0.3
per01;Rhl(−180)—per-1	38	0	38
per01;Rhl(−180)—per-2	56	0	56
per01	57	0	57

Young male flies from wild-type Canton-S,per⁰¹; Rhl(−180)—per-1,per⁰¹; Rhl(−180)—per-2, and per⁰¹ were entrained in 12:12 hr light/dark cycle for 3 d; then their locomotor activity was monitored in constant darkness at 25°C for 7 d. Periodogram analysis was done as previously described (Hamblen et al., 1986). Power and width are defined in Frisch et al. (1994) and were used to distinguish between rhythmic and arrhythmic flies (Ewer et al., 1992). No. tested, Total number of tested flies; No. rhy., number of rhythmic flies; No. arr., number of arrhythmic flies.  

Rh1(−180) promoter activity is constitutive in DD conditions and fluctuates in LD cycles

Previous studies showed that Rh1 mRNA in wild-type flies is expressed at high levels throughout the circadian cycle (Zeng et al., 1994). To ensure that per RNA from this crippledRh1(−180) promoter does not cycle in abundance, we entrainedper⁰¹;Rh1(−180)–per-1 flies in 12 hr LD cycles for 4 d and collected flies every 4 hr during the fifth LD cycle. Total head RNA was prepared from each time point, and levels of Rh1(−180)–per mRNA were determined by RNase protection assays. Whole heads were used in these assays because the Rh1(−180)–per transgene is specifically expressed in eyes (Fig. 2; Table 1), and the transgene-derived transcript can be monitored independently of the endogenous per⁰¹ transcript. In contrast to wild-type Rh1 mRNA levels,Rh1(−180)–per mRNA levels are ∼2.5-fold higher during the day than during the night (Fig.3). To determine whether these mRNA fluctuations are attributable to the Rh1(−180) promoter or clock regulatory sequences within, or downstream of, the percoding region, we examined mRNA levels fromRh1(−180)–LacZ transgenic flies under LD conditions. As seen with Rh1(−180)–per mRNA,Rh1(−180)–LacZ mRNA was found to be ∼2.5-fold higher during the day than during the night (data not shown), indicating that the Rh1(−180) promoter, and not PER coding or downstream sequences, mediates these mRNA fluctuations.

Fig. 3.

Rh1(−180) promoter activity is constitutive in DD conditions and fluctuates in LD cycles. RNase protection assays were performed on total head RNA samples prepared fromper⁰¹;Rh1(−180)–perflies. A, Endogenous and transgene-drivenper RNA levels measured in samples collected during LD cycles at the time points indicated.Rh1(−180)–per represents the transgene-driven per-protected fragment,per⁰¹ represents the endogenousper-protected fragment, and RP49represents the ribosomal protein 49 mRNA-protected fragment. Thewhite and black boxes represent times when lights were on or off, respectively. B, Quantitation of the data in A. Relative RNA abundance refers to the values of endogenousper/RP49 (filled squares) or transgene-drivenper/RP49 (open squares), when the peak value of endogenous per mRNA was adjusted to 1.0. The white and black boxesrepresent times when lights were on or off, respectively. Similar results were obtained from six independent time courses.C, Endogenous and transgene-driven perRNA levels measured in samples collected during the first day of DD at the time points indicated. The protected fragments are as designated inA. The hatched and black boxes represent times when lights would have been on or off, respectively, if the light cycle were continued. D, Quantitation of the data in C. Relative RNA abundance is described in B. The hatched andblack boxes represent times when lights would have been on or off, respectively, if the light cycle were continued. Similar results were obtained from three independent time courses.

To determine whether the Rh1(−180) promoter is sensitive to light or is under clock control, we measuredRh1(−180)–per mRNA under constant dark conditions. Flies were entrained in 12 hr LD cycles for 4 d, transferred to constant darkness, and collected every 4 hr for one circadian cycle. Head RNA extracted from each time point was analyzed by RNase protection assays, and unlike the previous results in LD conditions, Rh1(−180)–per mRNA levels do not fluctuate under constant dark conditions (Fig. 3). Likewise, the levels of per⁰¹ transcripts are constant (Fig.3), because of the lack of oscillation in noneye head tissue where moreper⁰¹ mRNA is expressed. Identical results were obtained when Rh1(−180)–LacZ mRNA levels were measured under constant dark conditions (data not shown).

Rh1(−180)–per mRNA levels are approximately fivefold higher than endogenous per⁰¹transcript levels in these transgenic lines (Fig. 3). This difference is mainly caused by the Rh1(−180) promoter becauseRh1(−180)–LacZ mRNA levels are two- to threefold higher than wild-type per mRNA levels (data not shown). In addition, per⁰¹ mRNA levels are two- to threefold lower than that of wild-type per mRNA (Hardin et al., 1990), which would further magnify per mRNA levels for the Rh1(−180)–per transgenics. These measurements were made on homozygous transgenic lines; however, the levels of Rh1(−180)–per mRNA may be lower in heterozygotes because one of the dosage compensation elements (i.e., intron 1) is missing in these constructs (Cooper et al., 1994). Consequently, heterozygotes were used in experiments performed under constant conditions (see below).

PER abundance and nuclear localization cycle during LD conditions

Because TIM is light sensitive and is required for PER accumulation and nuclear localization (Vosshall et al., 1994; Price et al., 1995; Hunter-Ensor et al., 1996; Myers et al., 1996; Zeng et al., 1996), we expected that both PER abundance and nuclear localization would cycle under LD conditions in these transgenic lines. To determine whether PER levels cycle, we entrainedper⁰¹;Rh1(−180)–per-1 flies to 12 hr LD cycles and collected flies at 4 hr intervals during the final cycle. Protein extracts were prepared from fly heads and analyzed on Western blots with anti-PER antiserum (see Materials and Methods). The results show that PER cycles under these conditions with a lower level during the day than during the night (Fig.4A). The cycling profile under these conditions is different from that of wild-type flies in two respects; it peaks at ZT15 rather than ZT21, and PER is never completely absent during the day. These differences may be caused by either the relatively high levels of per RNA from the transgene or a premature (light-driven) rise in mRNA levels from the transgene.

Fig. 4.

PER abundance and nuclear localization cycles inRh1(−180)–per flies under LD conditions.A, Head protein extracts were prepared fromper⁰¹;Rh1(−180)–per-1 transgenic flies collected at the time points indicated, fromper⁰¹ flies collected at ZT23, and from wild-type Canton-S (CS) flies collected at ZT23. These protein extracts were used to prepare Western blots, which were probed with anti-PER antibodies. The gel strip shows the level of PER immunoreactivity for each sample. The whiteand black boxes represent times when lights were on or off, respectively. Similar results were obtained from five independent time courses. B,per⁰¹;Rh1(−180)–perflies were entrained for 4 d in 12 hr LD cycles and collected at the time points indicated during the fifth LD cycle. Horizontal sections were prepared and probed with anti-PER antibodies. One hemisphere of a head is shown for each time point and is oriented with the anterior end up. PER immunoreactivity is seen as a dark brown stain. Two independent time courses showed similar results.

Another parameter that can be used to measure oscillator function is the PER nuclear localization cycle (Curtin et al., 1995). To determine whether PER derived from the Rh1(−180)–pertransgene undergoes nuclear cycling under LD conditions, we measured PER immunohistochemically on sections of flies collected every 4 hr (Fig. 4B). Little if any nuclear PER is detected until ZT15, a time point ∼2–3 hr earlier than that when PER begins to enter the nucleus in wild-type flies. Nuclear PER peaks between ZT19 and ZT23 and then begins to decline during the light phase until it becomes undetectable at ZT11. These results show that PER nuclear localization cycles under LD conditions. However, inRh1(−180)–per flies, PER accumulation peaks at ZT15, but the peak level of nuclear PER occurs at ZT23. This is different than the situation in wild-type flies in which nuclear localization occurs between ZT18 and ZT20 (Curtin et al., 1995), ∼4–5 hr before the peak in protein accumulation (Edery et al., 1994a). This uncoupling of PER accumulation and nuclear localization suggests that these two aspects of the circadian cycle are controlled independently because nuclear localization that is solely dependent on PER concentration would result in a premature movement of PER into the nucleus.

Endogenous per⁰¹ mRNA cycling is rescued in the eyes ofper⁰¹;Rh1(−180)–perflies

In the circadian feedback loop, nuclear PER suppresses the transcription of the PER gene, thereby producing fluctuations in mRNA abundance. From this, we would predict that nuclear PER cycling seen inper⁰¹;Rh1(−180)–perflies should rescue endogenous per⁰¹ RNA cycling. To test this prediction, we performed RNase protection assays on total RNA isolated from the eyes of homozygousper⁰¹;Rh1(−180)–per-1 flies collected every 4 hr during LD cycles (see Materials and Methods). The results show that endogenousper⁰¹ mRNA cycling is rescued in the eyes of these transgenic flies and has an amplitude of approximately fourfold (Fig. 5). The peak inper⁰¹ mRNA abundance occurs at ZT11, which is 4–6 hr earlier than the per mRNA peak in wild-type flies (Hardin et al., 1990; Brandes et al., 1996). This earlyper⁰¹ mRNA peak could be accounted for by the premature accumulation of nuclear PER (Fig. 4B), which acts to decrease the levels of per mRNA. Endogenoustim mRNA also exhibited low amplitude cycling, consistent with the notion that per and tim transcription are coregulated. As expected, the levels ofRh1(−180)–per mRNA were higher during the light phase because of the sensitivity of the Rh1(−180) promoter to light (Figs. 3, 5).

Fig. 5.

Endogenousper⁰¹ and tim mRNA cycling is rescued in the eyes of Rh1(−180)–perflies. RNase protection assays were performed on total eye RNA samples (5 μg) prepared fromper⁰¹;Rh1(−180)–perflies. A, Endogenousper⁰¹, endogenous tim, and transgene-driven per RNA levels measured in samples collected during LD cycles at the time points indicated.Rh1(−180)–per represents the transgene-driven per-protected fragment,per⁰¹ represents the endogenousper-protected fragment, tim represents the endogenous tim-protected fragment, andRP49 represents the RP49-protected fragment. Thewhite and black boxes represent times when lights were on or off, respectively. B, Quantitation of the data in A. Relative RNA abundance refers to the values of endogenousper/RP49 (open squares), endogenous tim/RP49 (filled squares), or transgene-drivenper/RP49 (filled diamonds), with each peak adjusted to 1.0. Thewhite and black boxes represent times when lights were on or off, respectively. Similar results were obtained from three independent time courses.

PER abundance cycles under constant dark conditions

To determine whether PER cycling in photoreceptors R1–R6 is a circadian rhythm, we examined both PER levels and nuclear localization in constant darkness.per⁰¹;Rh1(−180)–per-1/+ flies were entrained in an LD cycle for 4 d and collected during the first 2 d of DD at 4 hr intervals. Head protein extracts were prepared and subjected to Western blotting in parallel with those of wild-type controls. The results show that PER levels in these transgenic flies fluctuate over two circadian cycles. Compared with PER cycling in wild-type flies, in which low levels are present at CT11 and high levels are present at CT23 (Fig.6A), the overall PER levels inper⁰¹;Rh1(−180)–per-1/+ flies cycle with a different phase and/or period; PER decreases during the first subjective day, reaches its trough level at CT15–19, increases from CT23 to the next CT15, and then decreases (Fig.6A). This phase and/or period difference in PER cycling is also seen inper⁰¹;Rh1(−180)–per-2/+ flies, although the delay in PER accumulation (vs wild-type PER) is greatly reduced in comparison with that ofper⁰¹;Rh1(−180)–per-1/+ flies (Fig. 6A).

Fig. 6.

Overall PER abundance cycles under DD conditions. A,per⁰¹;Rh1(−180)–per-1/+ andper⁰¹;Rh1(−180)–per-2/+ flies were entrained in LD cycles for 4 d before being transferred to constant darkness and collected at the circadian time points indicated. Head protein extracts were made and subjected to Western blotting as described in Figure 4. The gel strips show the level of PER immunoreactivity for each sample. Head extracts fromCS and per⁰¹ flies collected at ZT23 were used as positive and negative controls for PER immunoreactivity, respectively (right two lanes). Thehatched and black boxes represent times when lights would have been on or off, respectively, if the light cycle were continued. Similar results were obtained from three independent time courses for CS, three time courses forper⁰¹;Rh1(−180)–per-1, and two time courses forper⁰¹;Rh1(−180)–per-2.B, Quantitation of PER cycling under DD conditions is shown. Data from three independent time courses of CS(open squares; solid line), three independent time courses ofper⁰¹;Rh1(−180)–per-1/+ (open circles; dashed line), and two independent time courses ofper⁰¹;Rh1(−180)–per-2/+ (filled triangles; line withshort and long dashes) are plotted. The peak time point for each time course was set to 1.0, and all other time points are relative to this value. The curves were generated by fitting the normalized pooled data for each genotype to polynomial functions (see Materials and Methods). The circadian time points and thehatched and black boxes are described inA.

Quantification of these data shows that PER from both transgenes cycles over the course of 2 d with a 1.5–2-fold amplitude forper⁰¹;Rh1(−180)–per-1/+ flies and an approximately twofold amplitude forper⁰¹;Rh1(−180)–per-2/+ flies (Fig. 6B). The cycling amplitudes of these transgenic lines approach that of the 2–2.5-fold cycling seen in wild-type flies collected and tested in parallel with the transgenic lines. This dampened cycling amplitude of wild-type flies in DD conditions is similar to that seen in previous studies (Edery et al., 1994b; Zeng et al., 1996). Because PER abundance could only be measured for a limited number of cycles under DD conditions, it is not possible to determine whether the differences in peaks and troughs in PER levels are caused by period differences, phase differences, or a combination of both. However, it would not be surprising if the higherper mRNA levels and the lack of per mRNA cycling in the transgenic lines has a negative effect on PER cycling amplitude.

To determine whether nuclear PER cycles under constant dark conditions, we entrained and collectedper⁰¹;Rh1(−180)–per-1/+ flies as described above. Sections were prepared for each time point and stained with anti-PER antibodies. The nuclear PER staining signal is intense as flies enter DD conditions, declines to a minimum value between CT16 and CT20, and then increases during the next circadian day (Fig. 7). Unlike the case in LD cycles, PER staining inper⁰¹;Rh1(−180)–per-1/+ flies is never completely absent in the nucleus. The fluctuations in PER intensity on sections fit well with overall PER abundance measurements on Western blots (Fig. 6).

Fig. 7.

Top. Levels of nuclear PER cycle under DD conditions.per⁰¹;Rh1(−180)–per-1/+ flies were entrained in LD cycles for 4 d before being transferred to constant darkness and collected at the indicated time points. Horizontal sections were prepared and probed with anti-PER antibodies as described in Figure 4B. One hemisphere of a head is shown for each time point and is oriented with the anterior end up. PER immunoreactivity is seen as a dark brown stain. Two independent time courses showed similar results. Each time point represents similar results from at least four flies.

DISCUSSION

In this study, the Rh1 promoter was modified so that it would produce low levels of mRNA in photoreceptors R1–R6. This modified promoter, Rh1(−180), is >20-fold less active than the native Rh1 promoter and produces transcripts specifically in photoreceptors R1–R6 at levels approximately fourfold higher than the wild-type per mRNA peak (Figs. 1, 2, 3, 5). TheRh1(−180) promoter is constitutively active in constant darkness and is light inducible, producing 2.5-fold more mRNA in the light phase than in the dark phase (Fig. 3). This light inducibility was not seen in earlier studies with the complete Rh1promoter (i.e., having 3 kb of upstream sequences), where high levels of expression may mask differences in mRNA abundance (Zeng et al., 1994). Such diurnal fluctuations in opsin mRNA abundance have been seen in certain species of toads and fish; opsin mRNA levels are 4–10-fold higher during the light phase than in the dark phase and can be induced by light in the dark phase (Korenbrot and Fernald, 1989). However, unlike the case in Drosophila, opsin mRNA levels in these species continue to fluctuate under constant darkness, apparently under the control of a circadian oscillator (Korenbrot and Fernald, 1989).

In wild-type flies, PER cycling is mediated, in part, via interactions with the TIM protein. TIM forms heteromeric complexes with PER and is required for both PER accumulation and nuclear localization (Vosshall et al., 1994; Gekakis et al., 1995; Price et al., 1995; Lee et al., 1996; Zeng et al., 1996). Light-induced degradation of TIM, for example, in natural environmental cycling conditions, seems very likely to mediate daily resetting of the Drosophila circadian clock such that the oscillator is in synchrony with the environment (Lee et al., 1996; Myers et al., 1996; Zeng et al., 1996). InRh1(−180)–per flies, per mRNA levels are constantly higher than is the wild-type per mRNA peak. These high levels of per mRNA nevertheless give rise to fluctuations in PER abundance and nuclear localization under LD conditions (Fig. 4). This high amplitude PER cycling is apparently mediated by TIM; whenper⁰¹;Rh1(−180)–pertransgenic flies are in LD cycles, light drives TIM cycling in R1–R6 that, in turn, drives both overall PER abundance and nuclear PER to cycle. This light-mediated cycling of TIM and its effect on PER cycling have been seen in previous studies (Myers et al., 1996; Zeng et al., 1996; Dembinska et al., 1997; Stanewsky et al., 1997). However, when extremely high levels of PER are present, as is the case with theRh1–per transgenes, TIM cannot maintain PER cycling (Zeng et al., 1996).

The phase of PER cycling inper⁰¹;Rh1(−180)–perflies is earlier than that of PER in wild-type flies under LD conditions. This early peak in PER abundance (∼6–8 hr early) is not followed by an equally early movement into the nucleus, although nuclear localization is ∼3 hr early (Fig. 4). This result shows that the PER abundance and nuclear localization cycles can be disconnected, indicating that high levels of PER accumulation alone are not sufficient for nuclear localization. Because PER moves into the nucleus as a complex with TIM (Lee et al., 1996; Zeng et al., 1996), the high levels of PER seen early in the dark phase [inper⁰¹;Rh1(−180)–perflies] may not be complexed with TIM and are therefore unable to move into the nucleus.

In LD cycles, endogenous per⁰¹ mRNA cycling is rescued in photoreceptors R1–R6 ofper⁰¹;Rh1(−180)–perflies. However, this rescued per⁰¹ mRNA cycling is different than the cycling seen for wild-type perRNA (Hardin et al., 1992b; Zeng et al., 1994). First, the rescuedper⁰¹ mRNA cycling amplitude is low. This may be because PER never drops to zero as it does in wild-type flies (Fig. 4A), and constant levels ofper⁰¹ mRNA are seen in 25% of the photoreceptor cells that lack transgene expression. Second,per⁰¹ mRNA peaks at ZT11, several hours earlier than the wild-type per mRNA peak (Fig. 5). This early mRNA peak probably results from higher than normal PER levels, which would lead to premature PER nuclear translocation and a decrease in per gene transcription. Nuclear PER in our transgenic flies was detected 3 hr earlier than it was in wild-type flies (ZT15 vs ZT18) (Fig. 4B), indicating that the period of time between the per mRNA peak and the appearance of nuclear PER in these transgenic flies is similar to that seen in wild-type flies (Hardin et al., 1990; Zerr et al., 1990; Hardin and Siwicki, 1995).

Cycling of PER abundance inper⁰¹;Rh1(−180)–perflies persists in constant darkness, showing that per RNA cycling is not required for PER cycling (Fig. 6). This result is compelling because expression is restricted to photoreceptors, a cell type in which expression levels are high enough to be quantitatively measured. Previous studies in which mRNA expression levels in certain tissues are inferred, but not directly measured, are consistent with this finding. The glass promoter, which is constitutively active in wild-type flies, has been used to drive perexpression in the LNs and photoreceptors ofper⁰¹ flies. Thisglass–per transgene rescues long-period (∼28–34 hr) locomotor activity rhythms to varying degrees (25–69% rescue) in each independent line and gives rise to cycling in PER abundance and nuclear localization under LD conditions in photoreceptors (Vosshall and Young, 1995). Even though perRNA cycling was not tested in these transgenic lines, constitutive expression from the endogenous glass promoter suggested that PER cycling can occur without per mRNA cycling (Vosshall and Young, 1995).

Constructs containing a 7.2 kb fragment of per genomic DNA lacking the promoter, first exon, and most of the first intron were also used to rescue rhythms in per⁰¹flies. One of the two lines that rescue locomotor activity rhythms (7.2:2) expresses PER exclusively in LNs (Hamblen et al., 1986; Frisch et al., 1994). In this line, PER abundance and nuclear localization cycle under LD conditions (Frisch et al., 1994). This cycling is not simply driven by LD cycles, however, because it persists under DD conditions (Fig. 8). As seen in wild-type flies, the PER cycling amplitude in 7.2:2 flies dampens under DD conditions because of PER remaining in the nucleus at times when it would be undetectable in LD cycles (Fig. 8; Zerr et al., 1990). The behavioral rescue seen in the two 7.2 lines was attributed to internal (i.e., within or downstream of the transcribed region) cis-acting transcriptional cycling elements that are active in a permissive genomic environment (Frisch et al., 1994). Our results argue against the existence of internal cycling elements because we see no mRNA cycling in any of theper⁰¹;Rh1(−180)–perlines, which only differ from 7.2 lines in that they lack the last 343 bp of the first intron and 46 bp of untranslated second exon sequences. Of these differences, the first intron sequence was incapable of driving cyclic transcription (H. Hao, Y. Cheng, and P. E. Hardin, unpublished observations). Our results, therefore, suggest that rescue of PER cycling and behavioral rhythms in 7.2:2 flies occurs without underlying cycles in per mRNA.

Fig. 8.

Bottom. Nuclear PER level cycles in lateral neurons of 7.2:2 flies under DD conditions. Homozygous per⁰¹7.2:2 flies were entrained in LD cycles for 4 d before being transferred to constant darkness and collected at CT3, CT9, CT15, and CT21. Immunohistochemical staining of PER was done as described in Figure4B. One hemisphere of a head is shown for each time point and is oriented with the anterior end up. Thearrowheads denote anti-PER antibody staining in LNs. Three independent time courses showed high intensity staining at CT3 (n = 5 flies) and CT9 (n = 6 flies), low intensity staining at CT15 (n = 6 flies), and medium intensity staining at CT21 (n = 9 flies).

If per mRNA cycling is not required for PER cycling in general, then what function does transcriptional feedback serve in wild-type flies? We believe that feedback regulation of mRNA cycling is important for several aspects of circadian clock function. For example, transgenic strains assumed to express constitutive levels ofper mRNA (i.e., glass–per; 7.2:2;hsp70-7.2) exhibit period-altered and/or weak locomotor activity rhythms. Likewise,per⁰¹;Rh1(−180)–perflies, which are known to express constant levels of permRNA, exhibit an altered phase and/or period of PER cycling compared with wild type. These examples suggest that per mRNA cycling affects PER cycling and behavior; however, inappropriate permRNA or protein expression (i.e., levels that are too high or too low; altered or incomplete spatial expression) could also account for these effects on PER cycling and behavior. Another aspect of clock function that may require per and/or tim mRNA cycling is phase resetting. Because light acts to decrease TIM abundance, high levels of tim mRNA early in the dark phase were proposed to replenish TIM levels, leading to a phase delay, whereas low levels oftim mRNA late in the dark phase could not replenish TIM levels, leading to a phase advance (Hunter-Ensor et al., 1996; Myers et al., 1996; Zeng et al., 1996). Finally, circadian feedback is believed to control the expression levels of clock output genes. A group of 20Drosophila rhythmically expressed genes (Dregs) have been identified (Van Gelder et al., 1995). Rhythmic expression of several of these transcripts (i.e., Dregs 5, 6, 9, 10, and 15), as measured by mRNA cycling, is dependent on per gene function (Van Gelder et al., 1995; Van Gelder and Krasnow, 1996), indicating that circadian feedback loop function is important for rhythmic Dreg expression. Likewise, another clock-regulated gene, Crg-1, has been identified that exhibits PER-dependent mRNA cycling (Rouyer et al., 1997).

When per is specifically expressed in the photoreceptors R1–R6, PER levels show circadian cycling under both LD and DD conditions. These results demonstrate that Drosophila eyes contain an autonomous circadian oscillator (i.e., one that operates in the absence of per expression and circadian feedback loop function in other tissues). This finding is consistent with earlier studies on disconnected (disco) flies, which show that per mRNA and protein cycle in heads under both LD and DD conditions even though disco flies are behaviorally arrhythmic and lack per-expressing lateral neurons (Zerr et al., 1990; Dushay et al., 1992; Hardin et al., 1992a). These results suggest that other circadian oscillators operate in the absence of the pacemaker for locomotor activity.

Like Drosophila photoreceptors, hamster suprachiasmatic nucleus (SCN) and retina (Ralph et al., 1990; Tosini and Menaker, 1996), Xenopus eyes (Besharse and Iuvone, 1983),Aplysia eyes (Jacklet, 1969), gypsy moth testes (Giebultowicz et al., 1989), and chicken pineal glands (Takahashi et al., 1980) contain tissue autonomous circadian oscillators. This autonomy extends down to individual isolated basal retinal neurons inBulla (Michel et al., 1993), and individual (but not isolated) rat SCN neurons (Welsh et al., 1995) show circadian electrical activity. In contrast, the circadian rhythm of lateral eye sensitivity in horseshoe crab (Limulus polyphemus) is a slave oscillator controlled by the brain circadian clock (Barlow et al., 1980; Kaplan and Barlow, 1980; Barlow, 1983; Kass and Barlow, 1992). Thus, depending on the tissue and the organism, circadian oscillators can operate tissue/cell autonomously or can be dependent on oscillators in other tissues. In Drosophila, we have shown that the adult eye contains an autonomous circadian oscillator. This, along with clock autonomy in the prothoracic gland (Emery et al., 1997) and lateral neurons (Frisch et al., 1994) and lateral neuron independent oscillators in Malpighian tubules (Giebultowicz and Hege, 1997; Hege et al., 1997), suggests that the Drosophilacircadian system consists of a collection of autonomous oscillators. Synchrony among these oscillators could be mediated by the light sensitivity of TIM because a hierarchical organization is not apparent.