Light signaling to the zebrafish circadian clockby Cryptochrome 1aT. Katherine Tamai*, Lucy C. Young, and David Whitmore*

Department of Anatomy and Developmental Biology, Centre for Cellular and Molecular Dynamics, University College London,21 University Street, London WC1E 6DE, United Kingdom

Edited by Jeffrey C. Hall, Brandeis University, Waltham, MA, and approved July 30, 2007 (received for review May 17, 2007)

Zebrafish tissues and cells have the unusual feature of not onlycontaining a circadian clock, but also being directly light-responsive.Several zebrafish genes are induced by light, but little is known abouttheir role in clock resetting or the mechanism by which this mightoccur. Here we show that Cryptochrome 1a (Cry1a) plays a key role inlight entrainment of the zebrafish clock. Intensity and phase responsecurves reveal a strong correlation between light induction of Cry1aand clock resetting. Overexpression studies show that Cry1a acts asa potent repressor of clock function and mimics the effect of constantlight to ‘‘stop’’ the circadian oscillator. Yeast two-hybrid analysisdemonstrates that the Cry1a protein interacts directly with specificregions of core clock components, CLOCK and BMAL, blocking theirability to fully dimerize and transactivate downstream targets, pro-viding a likely mechanism for clock resetting. A comparison of en-trainment of zebrafish cells to complete versus skeleton photoperiodsreveals that clock phase is identical under these two conditions.However, the amplitude of the core clock oscillation is much higher ona complete photoperiod, as are the levels of light-induced Cry1a. Webelieve that Cry1a acts on the core clock machinery in both acontinuous and discrete fashion, leading not only to entrainment, butalso to the establishment of a high-amplitude rhythm and evenstopping of the clock under long photoperiods.

entrainment � oscillator � phase shift � photoperiod

An essential and defining feature of a circadian clock is that itcan be set or entrained to the local light–dark (LD) cycle.

Because light is the most typically used environmental cue, mostplants and animals have evolved circadian light detection mecha-nisms and signaling pathways that convey this light information tothe core clock machinery. Zebrafish represent an alternative, if asyet relatively unexplored, vertebrate model system for the study ofthe circadian clock. Their circadian system has some similarities tothat of Drosophila, particularly in regard to peripheral clock en-trainment (1, 2). Zebrafish tissues contain endogenous circadianoscillators that are directly light-responsive and entrainable to LDcycles in vitro (3). This direct light sensitivity extends to the earlieststages of development, as well as to embryonic cell lines, makingzebrafish cells distinct from their mammalian counterparts (3, 4).This unusual feature means that the clock mechanism and entirephotoentrainment pathway are contained within a single cell.Consequently, by transfecting zebrafish cell lines with luminescentreporters driven by the promoters of rhythmic and light-responsivegenes, we have generated a model system where we can followcircadian oscillations and their direct entrainment by light at thecellular level (5, 6).

Recently, by employing single-cell luminescent imaging, we haveshown that an asynchronous population of zebrafish cells can bestrongly reset by a single light pulse to a common phase of thecircadian cycle (6). This is the consequence of a high-amplitude,Type 0 phase response curve (PRC) that zebrafish clocks possess(5). The cellular and molecular events involved in entrainment ofthe zebrafish clock, however, have not been extensively studied. Weand others have identified several acutely light-responsive genes,including Cryptochrome 1a (Cry1a) and Period 2 (Per2) (4, 7–9).These molecules have been proposed to play key roles in entrain-

ment, although evidence for their mode of action is lacking. The aimof this study, therefore, is to explore the molecular changes thatoccur in a zebrafish clock-containing cell in response to light, withparticular emphasis on the role of Cry1a. Here we examine theeffect of light on Cry1a induction and phase shifting and establishhow the Cry1a protein interacts with the core clock machinery tobring about light-dependent changes in clock function. Results fromintensity and PRCs demonstrate a strong correlation between lightinduction of the Cry1a gene and clock resetting. Overexpressionanalysis reveals that Cry1a is a potent transcriptional repressor andappears to mimic the effect of sustained light to ‘‘stop’’ the circadianoscillator. Biochemical studies show that the Cry1a protein binds tomultiple domains of central clock components, CLOCK andBMAL. This binding then blocks transactivation, providing a likelymechanism for clock resetting and establishment of high-amplituderhythms on a LD cycle. Thus, Cry1a appears to play a critical rolein the response of the zebrafish clock to light but, interestingly, aspart of a signaling pathway to the circadian pacemaker. Zebrafishpossess a number of cryptochromes (10), and it appears that a‘‘division of labor’’ has occurred in this circadian system, with Cry1aacting as a light-signaling molecule and other cryptochromes,potentially taking on the role of circadian photopigment or coreclock protein (7, 10).

ResultsA 1-h Light Pulse at Circadian Time (CT) 16 Leads to Large Phase Shiftsand Induction of Cry1a and Per2. To investigate how light phase shiftsthe zebrafish clock, we monitored clock function by using a Per1-luciferase zebrafish cell line (5) (Per1 is also called Per4). After 3 daysof entrainment, cells were given a ‘‘white’’ light pulse (400–700 nmat 2,500 �W/cm2) for 1 h at CT16 and returned to constant darkness(DD) for a further 3 days. Bioluminescent traces show that this 1-hlight pulse leads to a dramatic phase shift of �12 h (Fig. 1A). Toidentify potential molecules involved in phase shifting, we com-pared the expression of a number of Cryptochrome (Cry1a, Cry1b,Cry2a, Cry2b, Cry3, and Cry4) and Period (Per1, Per2, and Per3)genes in light-pulsed versus dark control cells by RNase protectionanalysis. From these two gene families, only Cry1a and Per2 werestrongly induced by light (Fig. 1B and data not shown). Shorter lightpulses of 15 min also led to a significant increase in Cry1a and Per2mRNA levels (data not shown) and were therefore used for allsubsequent experiments.

Author contributions: T.K.T. and D.W. designed research; T.K.T., L.C.Y., and D.W. per-formed research; T.K.T. and D.W. analyzed data; and T.K.T. and D.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: LD, light–dark; DD, constant darkness; CT, circadian time; bHLH, basichelix–loop–helix; PRC, phase response curve; PAS, PER-ARNT-SIM.

*To whom correspondence may be addressed. E-mail: ktamai@yahoo.com andd.whitmore@ucl.ac.uk.

This article contains supporting information online at www.pnas.org/cgi/content/full/0704588104/DC1.

© 2007 by The National Academy of Sciences of the USA

14712–14717 � PNAS � September 11, 2007 � vol. 104 � no. 37 www.pnas.org�cgi�doi�10.1073�pnas.0704588104

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 16

, 202

0

Light Induction of Cry1a Correlates with the Size of Phase Shift.Zebrafish cells were pulsed with a range of light intensities toexamine whether there is a correlation between the magnitude ofCry1a induction and the size of the resulting phase shift. Tofacilitate this analysis, we generated a Cry1a-luciferase reporter cellline by stably transfecting zebrafish PAC2 cells with 4.7 kb of thezebrafish Cry1a promoter fused to luciferase. After entrainment,Cry1a- and Per1-luciferase cells were exposed to light (400–700 nm)at different intensities (0, 100, 1,000, 5,000, and 10,000 �W/cm2) for15 min at CT16 and then returned to DD for 3 days. Theseexperiments show a clear correlation between light intensity, Cry1ainduction, and the size of the Per1 phase shift (Fig. 2A). Althoughthese light intensities may appear ‘‘unnaturally’’ high, they aresimilar to the levels of sunlight we experience here in Londonduring winter (3,000–16,000 �W/cm2). Thus, for a tropical, diurnalanimal like zebrafish, these intensities are well within the averagelevels experienced in nature.

Is Cry1a induction phase-dependent, and, if so, does it correlatewith the shape of the PRC? After entrainment, both reporter celllines were pulsed with white light (400–700 nm at 5,000 �W/cm2)for 15 min at different CTs (CT0, CT4, CT8, CT12, CT16, andCT20). These results demonstrate that, although Cry1a is inducedthroughout the circadian cycle, the magnitude of its induction variessignificantly according to circadian phase, from 1.59 � 0.11-fold atCT8 up to 4.38 � 0.20-fold at CT20 (Fig. 2B). Moreover, there isa strong correlation between the magnitude of Cry1a induction andthe size of the resulting phase shift, from a 2.10 � 0.46-h phase delayat CT8 to a 15.68 � 0.67-h phase delay at CT20 (Fig. 2C).

Overexpression of Cry1a Represses Circadian Rhythms and Mimics theEffect of Sustained Light. To further analyze the role of Cry1a, wetransfected Per1-luciferase cells with HA-tagged Cry1a by usingretroviruses. Western blot and immunocytochemical analyses dem-onstrate that this tagged protein is expressed and localized to thenucleus [supporting information (SI) Fig. 7]. Bioluminescent tracesreveal that Cry1a overexpression abolishes rhythmic expression of

Per1 and significantly reduces basal levels in a dose-dependentmanner (Fig. 3A and data not shown). These results indicate thatCry1a is a potent repressor of circadian clock function and, inter-estingly, appears to mimic the effect of constant light. Indeed,sustained light treatment at 2,500 �W/cm2 appears to stop ordramatically reduce the amplitude of the circadian oscillator (Fig.3B). The oscillation then resumes from this point upon entry intoDD (Fig. 3B). This is most clearly demonstrated in experimentswith variable, long-duration light pulses starting at the same phase(‘‘forward wedge’’) (SI Fig. 8A). Under such conditions, the ze-brafish oscillator appears to stop during the light treatment and thenrestarts at a predictable, fixed point immediately after the light-to-dark transition. By ending the variable-duration light pulses at afixed time (‘‘reverse wedge’’), the clock can be seen to restart froma common phase, equivalent to CT12 (SI Fig. 8B). These obser-vations fit well with the ‘‘clock stopping’’ action of light previouslydescribed in classic studies, where circadian rhythms in eclosion of

Fig. 1. One-hour light pulse at CT16 leads to large phase shifts and inductionof Cry1a and Per2 mRNA. (A) Bioluminescent traces of Per1-luciferase cellslight pulsed for 1 h (light gray squares) or maintained in DD (dark graydiamonds). The average (� SEM) of quadruplicate wells is presented. (B) RNaseprotection assay of Cry1a and Per2. RNA was prepared from light-pulsed anddark control cells harvested at 0, 1, 2, 4, and 6 h after the light treatment. Tim2served as an internal RNA control.

Fig. 2. Light induction of Cry1a correlates with size of Per1 phase shift. (A)Cry1a- and Per1-luciferase cells were exposed to light (400–700 nm) at CT16 atthe indicated intensities for 15 min. The average (� SEM) from quadruplicatewells is presented. The fold induction of Cry1a is plotted as a bar graph, andthe size of Per1 phase shift in hours is plotted as a line graph. Cry1a-luciferase(B) and Per1-luciferase (C) cells were light-pulsed (400–700 nm at 5,000�W/cm2) at the indicated CT. The histograms represent the average (� SEM)fold induction of Cry1a (B) and size of Per1 phase shift in hours (C) from threeindependent experiments.

Tamai et al. PNAS � September 11, 2007 � vol. 104 � no. 37 � 14713

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 16

, 202

0

fly pupae and flight activity of mosquitoes have been examined (11,12). We believe that this represents a molecular correlate of thatresponse and, in the case of zebrafish, is mediated through asustained increase in Cry1a levels. What is the mechanism of Cry1aaction?

Cry1a Interacts Directly with the PER-ARNT-SIM (PAS) B Domain ofCLOCK, As Well As Multiple Regions of BMAL. In vitro studies haveshown that the Cry1a protein binds directly to core clock compo-nents, CLOCK and BMAL (13). Using the yeast two-hybrid system,we confirm most of these interactions and extend this analysis toidentify specific domains of CLOCK and BMAL involved inprotein binding. These experiments demonstrate that CLOCK1and BMAL1 interact strongly at the basic helix–loop–helix (bHLH)and PAS B domains, with little or no binding between the two PASA domains (Fig. 4C and SI Fig. 9A). Interaction assays with Cry1areveal that Cry1a binds strongly to the PAS B domain of CLOCK1and to multiple regions of BMAL1, including the bHLH, PAS B,and C-terminal transactivation domains (Fig. 4 and SI Fig. 9).Additional analysis demonstrates that Cry1a interacts withCLOCK3 and BMAL3, but not CLOCK2 or BMAL2 (SI Fig. 9B).Because all forms of CLOCK and BMAL are able to dimerize andtranscriptionally activate target genes (13, 14), these results indicatethat Cry1a can potentially block all but the CLOCK2:BMAL2heterodimer (i.e., eight of the nine potential CLOCK:BMALcombinations). Based on our Cry1a overexpression data above,however, the CLOCK2:BMAL2 heterodimer alone does not ap-pear sufficient to drive rhythmic expression of Per1. From theseresults, we believe that Cry1a interferes with CLOCK:BMALfunction in two ways: first, by inhibiting transactivation directly bybinding to the C-terminal domain of BMAL, and, second, byphysically blocking the ability of new CLOCK and BMAL proteinto form active dimers by competing for the bHLH and PAS Bdomains, regions where CLOCK and BMAL themselves directlyinteract (Fig. 5).

Clock Entrainment to Skeleton Photoperiods Reveals the ContinuousInfluence of Light on the Core Clock Mechanism. We propose thatlight can influence the motion of the circadian oscillator by inter-

fering with the ability of CLOCK and BMAL proteins to dimerizeand transactivate downstream genes, such as Per1, through anincrease in Cry1a levels. Short light pulses cause a transient increasein Cry1a transcript levels, with a resultant strong repression andphase shift of the Per1 rhythm. Moreover, constant light or con-stitutive overexpression of Cry1a appears to stop the circadianoscillator, showing a sustained action of light on the clock. Tocompare directly the continuous versus discrete action of light onthe clock, we examined entrainment to both complete and com-plementary skeleton photoperiods (‘‘two-pulse’’ entrainment).Cells were taken from an unentrained state and placed on either acomplete 12:12 LD cycle or a skeleton light cycle with 15 min oflight, corresponding to dawn and dusk, of matching intensity (2,500�W/cm2). An examination of the Per1 oscillation on a completephotoperiod shows the rapid establishment of a high-amplituderhythm from the starting DD condition (Fig. 6A). Similarly, cells onthe skeleton photoperiod show rapid entrainment and achieve astable phase relationship within 24 h, with one peak of Per1expression showing a phase angle identical to that seen on acomplete photoperiod. However, there are clear differences in thePer1 rhythm on the two lighting regimes. On the skeleton light cycle,the ‘‘second’’ of the 15-min light pulses induces an additionaltransient peak in Per1 expression at a phase equivalent to dusk,before light-dependent repression (see Discussion). Entering DDfrom this photoperiod reveals, however, that the first peak in Per1

Fig. 3. Cry1a overexpression abolishes rhythmic expression of Per1 and mimicsconstant light. (A) Bioluminescent traces of Per1-luciferase cells transfected withHA-tagged Cry1a (black squares) or pCLNCX empty vector control (gray dia-monds). After 3 days of entrainment, cells were transferred into DD for 3 days. (B)Bioluminescent traces of Per1-luciferase cells entrained for 3 days, exposed toconstant light at 2,500 �W/cm2 for 60 h, and transferred into DD.

Fig. 4. Yeast two-hybrid analysis of Cry1a binding to CLOCK1 and BMAL1. (A)CLOCK1 deletion constructs were fused to the GAL4 DNA binding domain (BD)and tested for interactions against full-length Cry1a or BMAL1 fused to theGAL4 activation domain (AD). Colored rectangles represent functional do-mains of CLOCK1:bHLH (yellow), PAS A (green), and PAS B (blue). (B) Four yeasttransformants from each plasmid combination were patched onto agar platescontaining synthetic media without leucine or tryptophan (�L�W) to selectfor plasmids. Interactions were assayed by monitoring growth on mediawithout adenine (�Ade). �� represents a strong interaction, and � indicatesno interaction. (C) BMAL1 deletion constructs were fused to the GAL4 activa-tion domain (AD) and tested for interactions against full-length Cry1a orCLOCK1 (amino acids 1–390) fused to the GAL4 DNA binding domain (BD).Colored rectangles represent functional domains of BMAL1:bHLH (yellow),PAS A (red), and PAS B (pink). Interactions were assayed as above. �/� and�/� represent weak interactions. NT, not tested.

14714 � www.pnas.org�cgi�doi�10.1073�pnas.0704588104 Tamai et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 16

, 202

0

is sustained, with the oscillator using this light pulse as the dawnsignal. The second ‘‘dusk’’ light pulse appears to generate a phasedelay in the rhythm, as predicted by the shape of the PRC. This thenleads to precise and stable timing of the Per1 peak at zeitgeber time3, as would be expected from a nonparametric entrainment modelwhere each light pulse causes an equal and opposite phase shift (15).Another, perhaps more dramatic, difference between the twolighting regimes relates to the amplitude of the Per1 rhythm, whichis significantly reduced in cells on a skeleton photoperiod. Inparticular, the trough in Per1 expression is much greater in the lateafternoon on a complete photoperiod, suggesting a much higherlevel of repression by sustained light.

A comparison of Cry1a expression in parallel shows that, underthe skeleton photoperiod, each of the entraining 15-min light pulsesinduces a short peak in Cry1a levels (Fig. 6B). Therefore, a peak inCry1a occurs to mark both dawn and dusk, with each light pulsephase shifting the oscillator to produce steady-state entrainment.After the first entraining cycle, the level of Cry1a induction is verysimilar, as would be predicted for light pulses producing advancesand delays of equal but opposite magnitude. On a completephotoperiod, only a single peak in Cry1a is observed, but theduration and amplitude of this induction are much greater than ona skeleton photoperiod. The timing of this broader peak in Cry1acorresponds perfectly to the enhanced level of repression we see inthe Per1 oscillation. Consequently, we believe that the enhancedlevels of Cry1a during the light phase of the complete photoperiodact to increase Per1 repression and, subsequently, to increase theamplitude of the circadian oscillation. So, although the phase of therhythm is set accurately by the skeleton photoperiod (demonstrat-ing nonparametric entrainment), there is a clear consequence ofsustained light on the clock, which is to establish a high-amplitudemolecular oscillation.

We then examined entrainment to a single 15-min light pulse andcompared this to a complete photoperiod. Cells exposed to the‘‘one-pulse’’ regime show rapid synchronization, with a peak in Per1expression occurring at a phase identical to cells on a complete or

two-pulse skeleton photoperiod (Fig. 6C). However, both theamplitude of the Per1 rhythm and the induction of Cry1a are higheron a one-pulse than a two-pulse lighting regime but still fail to reachthe levels seen on a complete photoperiod (Fig. 6D). Moreover,entrainment to the one-pulse skeleton is not complete, because theoscillator appears to free-run between single light pulses, with therising phase of Per1 expression clearly advancing each cycle. Thus,to fully entrain the Per1 oscillation, either the continuous presenceof light during the day or an acute light pulse in the evening togenerate a critical phase delay is required. These results are verysimilar to the relative coordination or ‘‘bouncing’’ phenomenondescribed by Pittendrigh and Daan (16) in wheel running behaviorof rodents exposed to short, one-pulse lighting regimes. In ourexperiments, the single light pulse at dawn will strike the phaseadvance region of the zebrafish PRC, which may generate ‘‘after-effects’’ in the clock mechanism and lead to a shortening of thefree-running period (17). This ‘‘history dependence’’ or aftereffectcan be clearly seen when the cells are placed into DD, because cellsexposed to the one-pulse regime show a much shorter free-runningperiod than those exposed to the complete photoperiod (24.48 �0.39 versus 27.31 � 0 h, respectively) (Fig. 6C). This historydependence, as well as partial entrainment of the Per1 waveform,may suggest multioscillator complexity even in cultured cells, eitherbetween cells in the population or possibly within a single cell.

DiscussionThe circadian pacemakers in zebrafish cell lines show high-amplitude phase shifts in response to short, 15-min pulses of whitelight (400–700 nm). An examination of genes believed to beinvolved in circadian clock function reveals that Cry1a, amongothers, is strongly induced by such light pulses. Several lines ofevidence support the idea that Cry1a is critical for light-inducedphase shifts. We have shown a strong correlation between lightintensity, induction of Cry1a expression, and size of the resultingphase shift. Furthermore, light induction of Cry1a shows a keyphase dependency, with modest increases during the day and strong

Fig. 5. Model for light-induced interactions of Cry1a with CLOCK1 and BMAL1. (A) Yeast two-hybrid assays indicate strong binding between the bHLH andPAS B domains of CLOCK1 and BMAL1. (B) Cry1a interacts directly with the PAS B domain of CLOCK1 and the bHLH, PAS B, and C-terminal domains of BMAL1.These data support the hypothesis that light-induced Cry1a inhibits CLOCK:BMAL function by binding directly to the transactivation domain of BMAL and tocritical regions where CLOCK and BMAL themselves directly interact to form an active dimer.

Tamai et al. PNAS � September 11, 2007 � vol. 104 � no. 37 � 14715

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 16

, 202

0

inductions in the late night. This is of particular interest, because wehave shown that a single light pulse can shift an asynchronouspopulation of clock cells to a common phase of the circadian cycle,equivalent to the early day or zeitgeber time 4 (6). One wouldtherefore predict that, if cells were close to zeitgeber time 4, only asmall phase shift and modest increase in Cry1a would be necessaryfor entrainment. In the late night, however, a much larger phaseshift and higher level of Cry1a induction would be required, whichis precisely what we observe experimentally.

Because such correlations are not proof of a functional role forCry1a in phase shifting, we explored this issue further by overex-pressing Cry1a in our Per1 luminescent cell line. The consequencesof this are quite dramatic, in that the clock oscillation is completelyabolished. Thus, Cry1a can clearly act as a strong transcriptionalrepressor but, more importantly, as a light-induced repressor ofclock function. If Cry1a is a key element in the input pathway, thenincreasing its levels independent of light (i.e., in the dark) shouldmimic the natural action of light. This is true in the case of Cry1aoverexpression, which mimics remarkably well the consequences ofconstant light. Both bright light and tonically overexpressed Cry1aappear to stop the circadian clock. As described in Drosophila (12),the core molecular oscillator of zebrafish seems to be held relativelymotionless for light exposures �12 h and begins to oscillate fromthat phase onwards when released back into darkness. This hasinteresting implications for the zebrafish clock on photoperiods�12 h of light, where it appears likely that this system may act morelike an ‘‘hourglass’’ than an oscillator.

Our yeast two-hybrid assays show that Cry1a can clearly interactwith the core clock transcriptional activators, CLOCK and BMAL.The consequence of this binding is to strongly disrupt transactiva-tion by the CLOCK:BMAL heterodimer, and, as such, Cry1astrongly represses the positive limb of the circadian clock mecha-nism. Upon light exposure, Cry1a is strongly induced. Per1 expres-sion levels fall rapidly and are tonically suppressed under constant

light conditions. Phase shifts and clock entrainment appear, at leastin part, to be a result of this light-dependent transcriptionalrepression by Cry1a. As shown above, the circadian oscillatorappears to be held ‘‘motionless’’ when the day length begins toexceed 12 h at about CT12. What is special about this circadianphase? It in fact corresponds to the time when expression of bothCLOCK and BMAL is becoming strongly activated (ref. 3 and datanot shown). We believe that newly synthesized CLOCK and BMALproteins may be targets for Cry1a, which binds to these keytranscriptional activators, prevents their downstream action, and,consequently, stops the clock at this phase.

It is apparent from both the literature and a closer examinationof our Per1-luciferase rhythms under various lighting conditions thatCry1a is not the only ‘‘player’’ in the light input pathway to thezebrafish clock. Light pulses also induce Per2 expression, andinjection of a Per2 antisense morpholino into early-stage zebrafishembryos has been shown to block subsequent synchronized rhythmsin zfaanat2 expression in the pineal gland (8). However, themechanism by which Per2 may be involved in clock resetting is notyet known. Curiously, light pulses also lead to the acute inductionof Per1, which occurs before the increase in Cry1a levels andsubsequent repression in Per1 expression. It takes �3 h for Cry1ato reach peak transcript levels after light exposure, during whichtime a transient increase in Per1 is observed. The mechanism andpotential role of this transient Per1 increase are not yet understood.

We have demonstrated that zebrafish cells entrain to bothcomplete and skeleton (one- and two-pulse) photoperiods. Al-though the phase angle of the Per1 peak is identical under theseconditions, the amplitude of the rhythm is very different. Moreover,using single-pulse treatments, entrainment does not appear to becomplete across the entire cycle, and there are clear aftereffects onthe resulting free-running period in DD. The duration and ampli-tude of Cry1a induction are also considerably greater under com-plete versus skeleton photoperiods. Nevertheless, we believe that

Fig. 6. Zebrafish cells entrain to complete and skeleton photoperiods. Shown are bioluminescent traces of Per1-luciferase (A and C) and Cry1a-luciferase (Band D) cells exposed to a complete versus two-pulse skeletal photoperiod (A and B) or a complete versus one-pulse skeletal photoperiod (C and D). The average(� SEM) of quadruplicate wells is presented. The rectangular bars above (complete photoperiod) and below (skeleton photoperiod) represent light (white) anddark (black) periods.

14716 � www.pnas.org�cgi�doi�10.1073�pnas.0704588104 Tamai et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 16

, 202

0

Cry1a is playing a key role under both conditions. Cry1a can act asa nonparametric or discrete entraining cue under skeletal photo-periods, causing an acute repression and phase shift of the Per1rhythm. In addition, it can act in a parametric or continuous mannerunder full photoperiods, where the sustained induction of Cry1aacts to dramatically increase the amplitude of the circadian oscil-lator. These observations, in some ways, blur the distinction ofparametric and nonparametric entrainment as representing twoseparate mechanisms. Taking this one step further, these resultssuggest that, through its light induction and subsequent inhibitoryaction on CLOCK:BMAL, Cry1a may actually stop the clockcompletely.

Materials and MethodsZebrafish Cell Lines. The generation of a Per1-luciferase zebrafish cellline has been described (5). To establish a Cry1a-luciferase cell line,a 4.7-kb fragment of the Cry1a promoter was amplified by PCRfrom bacterial artificial chromosome clone HUKGB735F08222Q(German Resource Center for Genome Research, Berlin, Ger-many) and subcloned into pGL3-Basic (Promega, Madison, WI).PAC2 cells were electroporated with 5 �g each of linearizedCry1a-luciferase DNA and pcDNA3.1/myc-His A (Invitrogen,Carlsbad, CA). After neomycin selection, bioluminescence wasmonitored on a Packard TopCount scintillation counter (28°C), andcells from one strongly luminescent well were selected for thesestudies. RNase protection assays showed that 4.7 kb of the Cry1apromoter regulated expression of luciferase, which matched that ofendogenous Cry1a (data not shown). For all experiments, cells wereplated at 2.5–5.0 � 105 cells per milliliter.

RNA Analysis. Cells were maintained on a 12:12 LD cycle for 3 daysand then transferred into DD. At CT16 of the following day, cellswere exposed to white light (250 �W/cm2) for 1 h or kept in DD ascontrols. Samples were harvested at the times indicated in the figurelegends. Total RNA was extracted in TRIzol (Invitrogen) followingthe manufacturer’s instructions. The full-length coding regions ofthe Cryptochrome (Cry1a, Cry1b, Cry2a, Cry2b, Cry3, and Cry4) andPeriod (Per1, Per2, and Per3) genes were amplified by PCR fromexisting plasmids or from PAC2 cells by RT-PCR and subclonedinto pGEM-Teasy (Promega). These plasmids were linearized andused as templates for riboprobe synthesis (Promega). RNase pro-tection assays were carried out as previously described (3).

Bioluminescence Assays. Per1- and Cry1a-luciferase cells were platedin quadruplicate wells of a 96-well plate in media containing 0.5 mMbeetle luciferin (Promega). Unless otherwise indicated, cells wereplaced on a 12:12 LD cycle for 3 days and transferred into DD. Onthe following day, samples were light-pulsed at the time, intensity,and duration indicated in Results. Bioluminescence was monitoredon a Packard TopCount NXT scintillation counter (28°C).

Retroviral Constructs and Transfections. A DNA fragment encodingN-terminally tagged HA-Cry1a was amplified by PCR and sub-cloned into the retroviral vector pCLNCX (Imgenex, San Diego,CA). Expression of cDNAs subcloned into pCLNCX is driven bythe CMV promoter. The packaging cell line GP2–293 (Clontech,Palo Alto, CA) was transfected as described (6). Retrovirus was

collected 2 days later, and Per1-luciferase cells were infected twicea day for 2 days. Transfection efficiencies of 60–90% were typicallyobtained. Single cells from a transfected population were sorted byFACS into individual wells of a 96-well plate. Several clones wereexpanded and examined by Western blot and immunocytochemis-try (Fig. 3 and data not shown).

Western Blots. Cells were harvested in 300 �l of cracking buffer (8M urea/5% SDS/40 mM Tris, pH 6.8/0.1 mM EDTA/0.4 mg/mlbromophenol blue/147 mM 2-mercaptoethanol) plus protease andphosphatase (type I and type II) inhibitors (Sigma, St. Louis, MO)and boiled for 10 min. Proteins were separated by SDS/PAGE andtransferred to nitrocellulose (Schleicher and Schuell, Dassel, Ger-many). Blots were probed with rat anti-HA antibody high-affinity3F10 (Roche, Basel, Switzerland) diluted 1:500 and developed byusing the ECL Plus Western Blotting Detection System (GEHealthcare, Chalfont, St. Giles, U.K.).

Immunocytochemistry. Cells were fixed on day 3 at zeitgeber time 3in 4% paraformaldehyde in PBS for 20 min at room temperature,washed with PBS, and permeabilized in 0.2% Triton X-100 in PBSfor 5 min at room temperature. Samples were washed with PBS,blocked in 5% albumin, and then incubated at 4°C overnight in ratanti-HA antibody high-affinity 3F10 (Roche) diluted 1:500 in 1%albumin in PBS. Cells were washed with PBS and incubated at roomtemperature for 1 h in Alexa Fluor 568 goat anti-rat (MolecularProbes, Eugene, OR) diluted 1:1,000 in 1% albumin in PBS.Samples were washed with PBS, incubated in DAPI (1:50,000) for5 min at room temperature, washed again, and mounted.

Yeast Two-Hybrid Assays. The full-length coding regions of zebrafishCry1a, CLOCK (1, 2, and 3) and BMAL (1, 2, and 3) were fusedin-frame to the GAL4 DNA-binding domain of pGBKT7 or theGAL4 activation domain of pGADT7 (Clontech). Deletion con-structs encoding the amino acids indicated in the figures were alsogenerated. pGBKT7 and GAL4 DNA binding domain fusions weretransformed into yeast strain AH109 (MATa), and pGADT7 andGAL4 activation domain fusions were transformed into strain Y187(MAT�). The two strains were mated on rich media YPDA, anddiploids were selected on synthetic dropout media minus leucineand tryptophan (Clontech). Protein interactions were routinelyassayed by monitoring growth on synthetic media minus adenine.Pairings that were negative were tested on media minus histidineplus different concentrations of 3-aminotriazole.

We thank Nick Foulkes (Institute of Toxicology and Genetics, Forschun-gszentrum Karlsruhe, Karlsruhe, Germany), Daniela Vallone (Institute ofToxicology and Genetics, Forschungszentrum Karlsruhe), Matt Pando(Exon Hit Therapeutics, Gaithersburg, MD), Amanda Carr (Institute ofOphthalmology, University College London), and Veronica Ferrer (De-partment of Anatomy and Developmental Biology, University CollegeLondon) for plasmids and reagents; Kirsty Allen and Derek Davies atCancer Research UK for expert cell sorting; and members of the D.W.laboratory for valuable comments. We are especially grateful to CarlJohnson and Terry Page (who bear no responsibility for any circadian‘‘errors’’ in the text) for many helpful comments and suggestions. This workwas supported through funds from The Wellcome Trust and the Biotech-nology and Biological Sciences Research Council.

1. Plautz JD, Kaneko M, Hall JC, Kay SA (1997) Science 278:1632–1635.2. Tamai TK, Vardhanabhuti V, Arthur S, Foulkes NS, Whitmore D (2003) J Neuroendo-

crinol 15:344–349.3. Whitmore D, Foulkes NS, Sassone-Corsi P (2000) Nature 404:87–91.4. Tamai TK, Vardhanabhuti V, Foulkes NS, Whitmore D (2004) Curr Biol 14:R104–

R105.5. Vallone D, Gondi SB, Whitmore D, Foulkes NS (2004) Proc Natl Acad Sci USA

101:4106–4111.6. Carr AJF, Whitmore D (2005) Nat Cell Biol 7:319–321.7. Cermakian N, Pando MP, Thompson CL, Pinchak AB, Selby CP, Gutierrez L, Wells DE,

Cahill GM, Sancar A, Sassone-Corsi P (2002) Curr Biol 12:844–848.8. Ziv L, Levkovitz S, Toyama R, Falcon J, Gothilf Y (2005) J Neuroendocrinol 17:314–320.

9. Hirayama J, Cardone L, Doi M, Sassone-Corsi P (2005) Proc Natl Acad Sci USA102:10194–10199.

10. Kobayashi Y, Ishikawa T, Hirayama J, Daiyasu H, Kanai S, Toh H, Fukuda I, TsujimuraT, Terada N, Kamei Y, et al. (2000) Genes Cells 5:725–738.

11. Peterson EL (1980) J Theor Biol 84:281–310.12. Pittendrigh CS (1966) Z Pflanzenphysiol 54:275–307.13. Ishikawa T, Hirayama J, Kobayashi Y, Todo T (2002) Genes Cells 7:1073–1086.14. Lahiri K, Vallone D, Gondi SB, Santoriello C, Dickmeis T, Foulkes NS (2005) PLoS Biol

3:e351.15. Pittendrigh CS, Minis DH (1964) Am Nat 98:261–294.16. Pittendrigh CS, Daan S (1976) J Comp Physiol 106:223–252.17. Pittendrigh CS (1960) Cold Spring Harbor Symp Quant Biol 25:159–184.

Tamai et al. PNAS � September 11, 2007 � vol. 104 � no. 37 � 14717

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 16

, 202

0