The circadian oscillator gene GIGANTEA mediatesa long-term response of the Arabidopsis thalianacircadian clock to sucroseNeil Dalchaua,b,1, Seong J. Baeka, Helen M. Briggsa, Fiona C. Robertsona,2, Antony N. Dodda,3, Michael J. Gardnera,4,Matthew A. Stancombea, Michael J. Haydona, Guy-Bart Stanc, Jorge M. Gonçalvesd, and Alex A. R. Webba,1

aDepartment of Plant Sciences, University of Cambridge, Cambridge CB2 3EA, United Kingdom; bMicrosoft Research, Cambridge CB3 0FB, United Kingdom;cDepartment of Bioengineering, Imperial College London, Centre for Synthetic Biology and Innovation, London SW7 2AZ, United Kingdom; and dDepartmentof Engineering, University of Cambridge, Cambridge CB2 1PZ, United Kingdom

Edited by Steve A. Kay, University of California at San Diego, La Jolla, CA, and approved January 26, 2011 (received for review October 16, 2010).

Circadian clocks are 24-h timing devices that phase cellular re-sponses; coordinate growth, physiology, and metabolism; andanticipate the day–night cycle. Here we report sensitivity of theArabidopsis thaliana circadian oscillator to sucrose, providing evi-dence that plant metabolism can regulate circadian function. Wefound that the Arabidopsis circadian system is particularly sensitiveto sucrose in the dark. These data suggest that there is a feedbackbetween the molecular components that comprise the circadian os-cillator and plant metabolism, with the circadian clock both regulat-ing and being regulated by metabolism. We used also simulationswithin a three-loop mathematical model of the Arabidopsis circa-dian oscillator to identify components of the circadian clock sensi-tive to sucrose. The mathematical studies identified GIGANTEA (GI)as being associatedwith sucrose sensing. Experimental validation ofthis prediction demonstrated that GI is required for the full responseof the circadian clock to sucrose. We demonstrate that GI acts aspart of the sucrose-signaling network and propose this role permitsmetabolic input into circadian timing in Arabidopsis.

mathematical modeling | optimization | photosynthesis | oscillations |carbohydrates

The Arabidopsis thaliana circadian clock confers growth andcompetitive advantage (1). The phase of circadian rhythms in

plants is adjusted by light signals to entrain the clock to dawn anddusk (2). Additionally, it has been proposed that the Arabidopsiscircadian clock is sensitive to nitrogen acting as a nutritional cue(3) and phytohormones, possibly as an input from stress signalingand growth pathways (4). In Arabidopsis, circadian oscillationsare generated and maintained by interlocking transcriptional–translational feedback loops and posttranslational regulation (1,5–8). In the morning, light activates the expression of two Myb-like transcription factors, CIRCADIAN CLOCK ASSOCIATED1 (CCA1) and LATE ELONGATED HYPOCOTYL (LHY),leading to expression of PSEUDO RESPONSE REGULATORS(PRR) 7 and 9, which in turn repress CCA1 and LHY expression.This “morning” loop of the oscillator is connected to “evening”expressed genes through direct repression of the expression ofTIMING OF CAB EXPRESSION 1 (TOC1/PRR1) by binding ofLHY/CCA1 to the TOC1 promoter. TOC1 is expressed in theevening and feeds back to activate LHY/CCA1 through an un-known pathway. TOC1 physically interacts with and antagonizesCCA1 HIKING EXPEDITION (CHE), a transcriptional re-pressor of CCA1 (9). CCA1 and LHY are also coupled withTOC1 through GIGANTEA (GI). LHY/CCA1 repress GI and GIforms a loop with TOC1, GI activating TOC1 and TOC1repressing GI. GI encodes a protein of unknown biochemicalfunction but in blue light, GI physically interacts with and sta-bilizes ZEITLUPE (10), an F-box protein that targets TOC1 fordegradation (11). To entrain the circadian oscillator, signals areincorporated from light, second messengers, and metabolites to

alter circadian clock gene expression and modulate circadianfunction (7, 12–15).Carbohydrates are a major energy store in most forms of life

and in plants are synthesized as a result of photosynthesis.Sugars, such as sucrose, are also important signaling moleculesregulating growth, development, gene expression, and metabo-lism. Supplying high exogenous concentrations of sucrose (e.g.,300 mM, 10% wt/vol) severely inhibits early development ofArabidopsis seedlings, resulting in small white or purple cotyle-dons (16). Lower concentrations of sucrose (30–90 mM) do notinhibit shoot development although can interfere with a range ofsignal transduction pathways. Exogenous sucrose abolishes cir-cadian oscillations in the concentration of cytosolic-free Ca2+

([Ca2+]cyt) (17) and differentially perturbs daily oscillations ofgene expression in roots compared with aerial tissue (18). Ex-ogenous sucrose also increases expression of core central oscil-lator genes CCA1, GI, and TOC1 and reduces period in constantlight (LL), although these behaviors are absent in sensitive tofreezing 6 (sfr6) mutants (19). A large number of mutants havebeen identified that are insensitive to sucrose, although it is notclear how the circadian oscillator perceives sugar signals.

ResultsTo investigate the effect of sucrose on the Arabidopsis circadianclock network, we compared rhythmic behaviors in the presenceor absence of 90 mM (3% wt/vol) exogenous sucrose. Moststudies of circadian behavior in Arabidopsis have been performedusing seedlings grown in agar media containing 90 mM exoge-nous sucrose, and the mathematical models describing theArabidopsis circadian clock are based on experimental data ac-cordingly (20–24). Experiments from our laboratory and othersinvestigating circadian rhythms of [Ca2+]cyt demonstrate that theArabidopsis circadian clock also functions in the absence of ex-ogenous sucrose (7, 17). We first compared rhythms of CCA1::luc, TOC1::luc, and GI::luc in LL with or without 90 mM sucrose

Author contributions: N.D., G.-B.S., J.M.G., and A.A.W. designed research; N.D., S.J.B.,H.M.B., F.C.R., A.N.D., M.J.G., M.A.S., and M.J.H. performed research; N.D., S.J.B.,H.M.B., and F.C.R. analyzed data; and N.D., G.-B.S., J.M.G., and A.A.R.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: ndalchau@microsoft.com or alex.webb@plantsci.cam.ac.uk.

2Present address: Plant Functional Genomics Research Group, Plant Genomic NetworkResearch Team, RIKEN, Tsurumi-ku, Yokohama City, Kanagawa 230-0045, Japan.

3Present address: Department of Biology, University of York, York YO10 5DD, UnitedKingdom.

4Present address: European Bioinformatics Institute, Wellcome Trust Genome Campus,Hinxton, Cambridge CB10 1SD, United Kingdom.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1015452108/-/DCSupplemental.

5104–5109 | PNAS | March 22, 2011 | vol. 108 | no. 12 www.pnas.org/cgi/doi/10.1073/pnas.1015452108

added to the growth media (Fig. 1 A and C). In LL, sucrosecaused a small but statistically significantly period increase inTOC1::luc rhythms (period estimates: 90 mM sucrose, 23.83 h;90 mM mannitol, 23.36 h, P = 0.0023; H2O, 23.23 h, P = 0.0043)and period decrease in GI::luc rhythms (period estimates: 90mM sucrose, 24.75 h; 90 mM mannitol, 25.78 h, P = 0.0099;dH2O, 25.63 h, P = 0.0041; Fig. 1C). In constant dark (DD),CCA1::luc, TOC1::luc, and GI::luc were rhythmic in the presenceof 90 mM sucrose (period estimates: CCA1::luc, 23.76 h; TOC1::luc, 26.50 h; GI::luc, 27.39 h) but the rhythms were either verylow amplitude or undetectable in the absence of exogenous su-crose (Fig. 1 B and D). Without exogenous sucrose, CCA1::lucluminescence oscillated in 12 h light, 12 h dark (LD) cycles, in-creased 12 h after transfer to DD in anticipation of subjectivedawn, and then decreased to a constant low level (Fig. 1B).TOC1::luc and GI::luc luminescence also oscillated in LD beforeoscillating with decreasing amplitude in DD in the absence ofexogenous sucrose (Fig. 1B). The luc reporter was active in DDin the absence of exogenous sucrose because we were able todetect low temperature induction of CCA1::luc after 60 h in DD,and the presence or absence of 90 mM sucrose did not increasethe degree of induction compared with H2O controls (Fig. S1).Our data demonstrate that the amplitude of circadian oscil-

lations in DD was increased by the continual supply of exoge-nous sucrose in the growth media. To understand sucrose

sensitivity in a more physiological setting, we were interested totest whether the circadian clock also responds to step changes insucrose concentration, such as might occur following a dark tolight transition as plants start photosynthesizing and conse-quently produce sucrose. A total of 90 mM sucrose was appliedto entrained seedlings that had previously been exposed toprolonged darkness (60–72 h). Before sucrose treatment, oscil-lations of CCA1::luc, TOC1::luc and CHLOROPHYLL A/B-BINDING PROTEIN 2 (CAB2)::luc were barely detectable andoscillations of GI::luc were low amplitude (Fig. 2). Sucrosetreatment led to a large induction of all circadian markers,resulting in renewed circadian oscillations (Fig. 2). Therefore,sucrose has both a long-term effect on circadian rhythms, po-tentially incorporating metabolic status, and a short term effectthat may be related to intercellular signaling, as has been spec-ulated to exist between roots and aerial tissue (18).To determine whether sucrose was amplifying low amplitude

oscillations, such as observed in GI::luc, or was resetting thephase of the oscillator, sucrose treatment was administered atdifferent times through the circadian cycle. When sucrosetreatment coincided with subjective dawn (CT0), CAB2::luctransiently increased before continuing with a large amplitudecircadian rhythm that peaked 26.75 h after sucrose addition (Fig.2B). When sucrose treatment coincided instead with subjectivedusk (CT12), an initial transient oscillation of CAB2::luc lumi-nescence was not observed (Fig. 2C). However, similar to theCT0 treatment, transient dynamics were followed by a circadianrhythm that peaked 25.25 h after sucrose addition (Fig. 2C),qualitatively similar to the gated response of CAB2 induction totransient pulses of red light (25). Therefore, although the timingof the sucrose treatment led to a differential transient responseof CAB2::luc expression, the resulting circadian oscillationpeaked ∼24 h after sucrose addition irrespective of the subjectivetime of application, suggesting that sucrose, like light, has thepotential to act as a zeitgeber (“time giver”), a signal that is

Time in constant light (h)-24

0

1

2

3

4

Nor

mal

ized

lum

ines

cenc

e

0123456

dH2O 90 mM Man90 mM Suc

0

2

4 dH2O 90 mM Man90 mM Suc

B

0

1

2

368

1012A

Nor

mal

ized

lum

ines

cenc

e

0.0

0.5

1.0

1.5

2.0

2.5

Period in constant light (h)23 24 25 26

Mea

n re

lativ

e am

plitu

de e

rror

0.0

0.2

0.4

0.6

0.8

1.0C

Period in constant dark (h)20 22 24 26 28 30 32 34 36

Mea

n re

lativ

e am

plitu

de e

rror

0.0

0.2

0.4

0.6

0.8

1.0D

CCA1::luc

TOC1::luc

Time in constant dark (h)

0123

10

20dH2O 90 mM Man90 mM Suc

GI::luc

24 48 72 960 24 48 72 960

Fig. 1. Exogenous supply of sucrose influences period and amplitude ofcircadian oscillations in clock gene expression. (A and B) CCA1::luc (circles),TOC1::luc (triangles) and GI::luc (squares) luminescence measured for seed-lings entrained to LD cycles and transferred to (A) LL at t = 0 h or (B) DD at t =0 h. Data were normalized by dividing each point by the mean over thewhole trace. Bars on the abscissa are open for light, solid for darkness, andhatched for (A) subjective night or (B) subjective day. (C and D) Circadianperiod estimates of CCA1::luc, TOC1::luc, and GI::luc in (C) LL and (D) DD.Relative amplitude error quantifies the rhythmic robustness (0 is sine wave, 1indicates no oscillation). Agar media supplements: solid is H2O control, openis 90 mM mannitol control, and shaded is 90 mM sucrose.

Lum

ines

cenc

e

0

100000

200000

300000

Time since treatment (h)-48 -24 48

0

25000

50000

75000

100000

0

100000

200000

300000

90 mM Mannitol 90 mM Sucrose

CCA1::luc

TOC1::luc

GI::luc

Time since treatment (h)

Lum

ines

cenc

e

0

2500

5000

7500

10000

Time since treatment (h)

Lum

ines

cenc

e

0

20000

40000

60000

CAB2::luc

CAB2::luc

CT0

CT12

A B

C

CT0

CT0

CT0

0 24

-48 -24 480 24

-48 -24 480 24

Time since treatment (h)

Fig. 2. Sucrose entrains circadian oscillations of CCA1, TOC1, GI, and CAB2promoter activity in DD. Seedlings were grown in LD12/12 cycles beforetransfer to constant dark (DD). After (A and B) 60 h (CT0) or (C) 72 h (CT12) inDD, 90 mM sucrose (open circles) or 90 mM mannitol (solid circles) was ap-plied to aerial tissue. Luminescence of (A) CCA1::luc, TOC1::luc, and GI::luc or(B and C) CAB2::luc was measured for 4–12 clusters of two to three seedlings.Bars on the abscissa are solid for the subjective night and hatched for thesubjective day.

Dalchau et al. PNAS | March 22, 2011 | vol. 108 | no. 12 | 5105

PLANTBIOLO

GY

capable of reestablishing rhythms and setting the phase of thecircadian oscillator.To investigate these experimental findings in a quantitative

framework, we considered mathematical models of the Arabi-dopsis circadian clock. Currently published models simulate thebehavior of the circadian oscillator in seedlings grown with ex-ogenous sucrose in the growth media (20–24). A three-loopmodel of the central oscillator (22), from now on referred to asthe nominal model, yields stable oscillations with a period of23.81 h in LL and a period of 23.75 h in DD, reflecting thewealth of LL and DD data available for seedlings grown withexogenous sucrose (note the most recent Arabidopsis clockmodel, presented in ref. 24, was not published at the time of thisresearch). To identify potential targets for the long-term sucroseeffect on the circadian system we exploited the differences insystem behavior observed in DD. In the absence of sucrose thesystem tended toward a steady state, whereas in the presence ofsucrose there were stable oscillations (Fig. 1 B and D), which arealso predicted by the nominal model (22).We investigated whether changes in the kinetic parameters of

the nominal model could account for the dynamical differencesof central oscillator gene expression in the absence of exogenoussucrose. The kinetic parameters describe rates of transcription,translation, light activation, degradation, and nuclear transport(22) (Table S1). The model parameters were first varied singly,implicitly assuming that sucrose targets only a single mechanism.Over a wide range of values, the oscillation period and peak/trough fold ratio were computed for simulated [LHY/CCA1mRNA] (Fig. S2A). Twenty-six of the 71 model parameters hadregions of values that led to a loss of rhythms (peak/trough foldratio <1.1) in DD, while retaining oscillations in LL with a pe-riod between 20 and 30 h (Fig. S2B), reminiscent of the behaviorof the circadian oscillator in the absence of exogenous sucrose.Simulations using the 26 parameter alterations were examined inmore detail to identify those parameter alterations that couldalso describe the dynamics seen in DD in the absence of sucrose,including anticipation of subjective dawn in the first cycle aftertransfer to DD and the fast tendency to a low steady state ofCCA1 expression (Fig. 3A). Only a reduction in n5 (maximumlight-independent rate of Y transcription; reduction to 25% ofthe nominal value) could account for both the anticipation ofsubjective dawn after 12 h of darkness and the fast tendency toa low steady state of CCA1 expression (Fig. 3A, S2C). In thisreduced-n5 model (from now on referred to as the N5 model),simulated [TOC1 mRNA] was also constantly low after transferto DD, agreeing with measured TOC1::luc in the absence ofexogenous sucrose (Fig. 3A). There were stable oscillations ofboth [LHY/CCA1 mRNA] and [TOC1 mRNA] in LL, with anearly phase compared with the nominal model (Fig. S3A). Therelative phase advance in the N5model resulted in a poorer fit ofsimulated [LHY/CCA1 mRNA] and [TOC1 mRNA] to the cor-responding experimental data (Fig. S3A). Although significantquantitatively, this difference is very minor qualitatively com-pared with the requirement that parameter alterations shouldabolish rhythms in DD. Also, if sucrose targets more than oneprocess, there may be compensatory effects that maintain therelative timing of gene expression.In a contrasting analysis, all model parameters were allowed to

vary. Enabling multiple-parameter changes had the advantage ofconsidering potential interactions in model parameters that aresucrose dependent. Modified parameter sets were generated byoptimizing a cost function that penalizes deviations in simulated[mRNA] for LHY/CCA1 and TOC1 to CCA1::luc and TOC1::lucluminescence data, respectively, in both LL and DD aftertransfer from LD cycles. Approximate solutions to the optimi-zation problem were obtained using a modified Metropolis–Hastings simulated annealing algorithm, which was run in-dependently 400 times for 1,000 iterations each (SI Materials and

Methods). For each trial run, the lowest cost parameter set(annealed solution) was retained for further analysis (for exam-ple models, see Fig. S4 and Table S2). To establish whichparameters should be modified to account for sucrose absence,the mean of each parameter was computed over the best 100annealed solutions and analyzed for their deviation from thenominal parameter value (Fig. 3B, Fig. S5, Table S2, and TableS3). The strategy of averaging over the trial optimal parametersets was chosen for two reasons. First, many parameter setsreturned cost function values that were close to the best observedset; the distribution was bimodal, so we retained only the betterhalf of the low cost mode (Fig. S5A). Second, some parametervalues were not constrained by the experimental data, as theirdistributions across low cost parameter sets were broad (Fig.S5B). Therefore, taking the average of this distribution dis-tinguishes between parameters that are required to be differentfrom their nominal value and parameters that are poorly con-strained by the data and on average will not deviate from thenominal value.The nominal model considers Y transcription to be constitu-

tive with maximum rate n4 + n5 nM/h during light and rate n5nM/h during darkness (n4 is the light-dependent rate of Y tran-scription, and n5 is the light-independent rate) and inhibited byTOC1 and LHY/CCA1 (22). The parameter that was decreasedthe most on average to account for the removal of sucrose was n5(down 39%), and the parameter that on average was increasedthe most was n4 (up 93%). As n5 > n4, this equated to an increaseof n4 + n5 (total rate of Y transcription in light) of only 6% toaccount for the absence of sucrose. Therefore, the mathematicalanalysis suggested that Y transcription is less light dependentwhen sucrose is supplied exogenously and more light dependentin the absence of exogenous sucrose. The concentration of in-tracellular sucrose ([Suc]i) is high during the day and low at nightin the absence of exogenous sucrose (26). Also, supplying su-crose exogenously increases [Suc]i (27). Therefore, our mathe-matical modeling predicts the supply of exogenous sucrose willincrease the rate of Y transcription in the dark (when endoge-nous sucrose is low), but will have little effect on Y transcriptionin the light (when endogenous sucrose is high). Consequently,the rate of Y transcription varies coincident to the endogenousintracellular concentration of sucrose.Considering parameter alterations to the nominal model of

the Arabidopsis circadian oscillator led to the hypothesis thatsucrose up-regulates Y transcription. At least some of the bi-ological function described by the hypothetical component Y isfulfilled by GIGANTEA (GI) (22). Therefore, we wanted toconsider in more detail how perturbations to the expression ofGI could be reconciled by corresponding simulations of pertur-bations to Y. In particular, we wanted to investigate whetherlight/dark-dependent posttranslational modification of Y couldinstead be a target mechanism of sucrose, as the protein turnoverof GI is light dependent (28) and our hypothesis that sucrose up-regulates Y depends upon the light/dark dependency of Y ex-pression. This is important because the nominal model does nothave specific terms to describe the regulation of Y protein deg-radation. To address this alternative hypothesis, the nominalmodel equations governing Y protein dynamics were modified toinclude a dark-dependent degradation term (from now on, thePostGI model; SI Materials and Methods). As the rate of dark-dependent degradation was increased (without changing theexisting parameters), circadian oscillations of simulated [LHY/CCA1 mRNA] and [TOC1 mRNA] were suppressed in DD (Fig.3C), similar to the luc luminescence data without exogenoussucrose. A single oscillation in simulated [LHY/CCA1 mRNA]peaking 24 h after transfer from LD cycles was retained, as ob-served for CCA1::luc luminescence (Figs. 1B and 3C), and stableoscillations were retained in LL (Fig. S3B). This suggested a hy-pothesis whereby GI degradation is high when [Suc]i is low and

5106 | www.pnas.org/cgi/doi/10.1073/pnas.1015452108 Dalchau et al.

invariant to light availability when [Suc]i is buffered with exog-enous sucrose.Our mathematical modeling suggested two competing hy-

potheses for the long-term sucrose regulation of Arabidopsiscircadian rhythms. On the basis of a parameter perturbation tothe nominal model it was hypothesized that sucrose increases Ytranscription (N5 model). Alternatively, incorporation of light/dark-dependent posttranslational modification terms into thenominal model suggested that Y protein degradation is invariantwhen [Suc]i is buffered by exogenous sucrose supply (PostGImodel). To distinguish between the behaviors of the N5 andPostGI model hypotheses, we simulated topical sucrose treat-ment after a period of prolonged darkness. Both models weresimulated in entrained LD cycles before transfer to DD for 60 or72 h. Sucrose treatment was then simulated by switching to thenominal (sucrose supplied) model (22). In the N5 model simu-lated sucrose treatment caused a large induction of both [TOC1mRNA] and [Y mRNA] that preceded reestablished oscillationsin all model components (Fig. S6A). In contrast, simulated su-crose treatment in the PostGI model led to an initial decrease in[Y mRNA], followed by a reestablishment of circadian oscil-lations (Fig. S6B). Sucrose addition in DD caused an immediateincrease in GI::luc (Fig. 2) as predicted for Y/GI by the N5 modeland the initial decrease in Y/GI predicted by the PostGI modelwas not observed. These data experimentally validate the N5model on the basis of the assumption that the behavior of thehypothetical component Y describes the behavior of GI and thatGI fulfills at least some of the functions of Y (22). However, theassumption that GI fulfills the function of Y has been questioned(29). To address this question and determine the functionalsignificance for the circadian system of our findings, we furtherexamined the model hypotheses experimentally by testing the

effects of sucrose on GI promoter activity and the response of gimutants to sucrose.We experimentally tested the hypothesis that GI acts as part of

a long term sucrose-sensing pathway for the circadian clock bycomparing circadian behavior in gi-11 null mutants (30) grown inthe presence and absence of exogenous sucrose. Luminescenceof CAB2::luc was measured in DD after transfer from LD cycles.As expected from earlier experiments measuring CCA1::luc (Fig.1B), wild-type plants had circadian oscillations of CAB2::lucexpression in DD when 90 mM sucrose was supplied in themedia and rhythms were undetectable in media supplied withwater or mannitol as a control (Fig. 4A). However, in gi-11mutants addition of sucrose to the growth media did not permitoscillations of CAB2::luc in DD (Fig. 4A). Similarly, CCR2 ex-pression was arrhythmic in the gi-3 point mutant in the presenceof sucrose in DD (31). These observations demonstrate thatfunctional GI is required for the circadian system to integratelong-term sucrose signals and implicate GI in sucrose sensing inDD. As we had observed a GI–sucrose interaction in DD, wewere interested to see whether this interaction was also observ-able in LL. We compared two seed lines of gi-11 expressingCAB2::luc with corresponding wild-type WS and found statisticallysignificant long circadian periods in both mutant lines (−Suc: WS,23.50 h; gi-11/1, 26.37 h, P = 0.006; gi-11/2, 25.35 h, P = 0.002;+Suc: WS, 23.91 h; gi-11/1, 26.21 h, P = 4.5 × 10−5; gi-11/2,24.88 h, P= 0.03l; Fig. 4 B and C). However, sucrose had no effecton the period of CAB2::luc rhythms in the wild-type and both gi-11mutant lines (Fig. 4C). Therefore, we conclude that the putativelong term GI–sucrose interaction is not active in LL.Finally, to ascertain whether GI is involved in the short-term

response of the circadian system to sucrose, we investigated theresponse of wild-type and gi-11 plants expressing CAB2::luc to

Parameter

0

log(

Ave

rage

d re

lativ

epa

ram

eter

val

ue)

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

Nor

mal

ized

lum

ines

cenc

e

0

5

10

15

0.0

0.2

0.4

0.6

0.8

1.0

1.2CCA1:luc dataN5 model

A

Time (h)0 24 48 72 96 120

0

1

2

3

4

5

Sim

ulat

ed [m

RN

A] (

nM)

0.0

0.5

1.0

1.5

2.0TOC1:luc dataN5 model

B

n5

n4

Nor

mal

ized

lum

ines

cenc

e

0

5

10

15

0.0

0.5

1.0

1.5CCA1:luc dataNominal modelPostGI model

C

Time (h)0 24 48 72 96 120

0

1

2

3

4

5

Sim

ulat

ed [m

RN

A] (

nM)

0

1

2

3TOC1:luc data Nominal modelPostGI model

10 20 4030 6050

Fig. 3. Mathematical modeling suggests thehypothetical component Y mediates sucrosemodulation of the clock. (A and C) Normal-ized CCA1::luc (solid circles, Upper) andTOC1::luc (solid triangles, Lower) lumines-cence in LD and DD compared againstequivalent simulated [LHY/CCA1 mRNA] and[TOC1 mRNA] (solid lines) from variants ofthe three-loop model (20). (A) N5 is thethree-loop model with the maximum rate oflight-independent Y transcription (n5) re-duced to 0.25 of its nominal value. (B) Aver-age alterations to each parameter of thethree-loop model after parameter reoptim-ization to account for wild-type lumines-cence data from seedlings withoutexogenous sucrose (SI Materials and Meth-ods). A simulated annealing algorithm wasrun 400 times, producing 400 parameter sets(annealed solutions). On the graph, the barsindicate the logarithm of the average of thetop 100 annealed parameter sets. Largestincrease (n4) and largest decrease (n5) areindicated on the graph. (C) PostGI is thethree-loop model with increased maximumdegradation of cytoplasmic and nuclear Y(possibly GI) protein in the absence of lightto five times the nominal rate. Bars on theabscissa are open for light, solid for darkness,and hatched for dark in a subjective day.

Dalchau et al. PNAS | March 22, 2011 | vol. 108 | no. 12 | 5107

PLANTBIOLO

GY

a step change in exogenous sucrose in DD (Fig. 4D). In both wildtype and gi-11 the response to topical addition of sucrose werealmost identical, indicating GI is not involved in short-term su-crose signaling and that there is a functional difference betweenthe experimentally observed short- and long-term responses ofthe circadian network to sucrose.

DiscussionWe have demonstrated that the primary metabolite sucrosemodulates the functioning of the Arabidopsis circadian oscillator,the effects of which are most pronounced when endogenoussucrose is low, in DD. Continual supply of exogenous sucrosecontributes to sustained oscillations in DD (Fig. 1), indicatingthe oscillator integrates long-term metabolic signals. The ad-vantage of incorporating long-term metabolic status could berelated to why circadian clocks enhance growth (1, 32). Topicalsucrose supply reestablished rhythms in plants grown withoutexogenous sucrose (Fig. 2), indicating the oscillator also inte-grates short-term metabolic signals.The suggested regulation of plant clock function by metabolic

status is reminiscent of the situation in mammals where it is

proposed that circadian oscillations of NAD levels feed back toregulate circadian function and provide input concerning themetabolic status of the cell (33). NAD has profound effects onthe circadian systems of plants and animals. Treatment withnicotinamide, which is a product and inhibitor of NAD con-sumption, increases the period of both the plant and the mam-malian circadian clock (7, 34). The effects of NAD on circadiansystems have been proposed to be mediated by deacetylases,poly-ADP ribose, and cyclic ADP ribose and it is likely that allcontribute to modulate circadian function in both kingdoms (7,33, 35, 36).Short-term sucrose signals might confer intercellular commu-

nication; for example, the coupling of oscillators in the shootsand roots of plants is perturbed by addition of sucrose (18).Another possibility is that sucrose feedback into the oscillatorprovides a further example of circadian loop formation, becausethe circadian clock regulates photosynthetic capacity (1) and wedemonstrate that sucrose affects clock function (Fig. 1). It hasbeen proposed that forming circadian clocks of multiple feed-back loops increases robustness (37). Our data show that sucrosetreatment in LL has a relatively small effect on the circadianoscillator (Fig. 1A), which is consistent with a previous study(19). In LL, the small effects of sucrose included a statisticallysignificant, but slight decrease in the period of GI:luc rhythmsand also a small, but significant increase in the period of TOC1::luc (Fig. 1). This result could be interpreted as evidence foruncoupled cell-specific oscillators (18) responding differently tothe sucrose signal. Alternatively, these apparently opposite re-sponses to sucrose in LL might be a consequence of measuringsmall responses close to the limit of resolution of the data cap-ture and the statistical analysis packages and so might not be ofphysiological consequence. We found the effect of sucrose wasmuch greater in DD. It is possible that this effect represents anexample of circadian gating of input sensitivity, similar to thatseen with the circadian gating of light input (38). However,gating is more often associated with short-term signals, and wehave identified a mechanism that perceives long-term signals.Therefore, we favor the interpretation that the circadian systemis insensitive to sucrose in the light because the effects of theexogenous sucrose are buffered by the high intracellular sucroseproduced as a result of photosynthesis. The buffering of theeffects of exogenous sucrose by intracellular sucrose in the lightmight explain why corresponding observations have been elusiveuntil now and why relatively high concentrations of sugars havebeen required to investigate sugar signaling in plants.Significant mechanistic understanding of the interactions be-

tween central oscillator components has been realized in the pasttwo decades. Whereas cross-talk with second messengers such ascyclic ADP ribose (7) and metabolites such as sucrose (19) hasbeen demonstrated, mechanistic understandings of the links be-tween clocks and signaling remain few. We exploited the quali-tative differences between plants grown in the presence andabsence of exogenous sucrose to explore with mathematicalmodeling how the long-term sucrose signals are integrated into theclock. Mathematical modeling allowed us to identify the hypo-thetical component Y as being sensitive to sucrose perturbations.Our simulations and modeling predictions suggested that sucroseincreases the rate of Y transcription and consequently we hy-pothesized that Y is required for the response of the oscillator tosucrose. In the nominal model, it is thought that at least 70% ofthe function of Y can be ascribed to GI (22). In a more recentmodel of the Arabidopsis central oscillator (24), the functions ofGI have been separated from the component Y to reflect addi-tional knowledge, although both Y and GI are modeled as beingtranscribed in response to the same transcription factors (TOC1and LHY/CCA1), consistent with the previous model (22). Incontrast, however, GI is transcribed only during light (24).Therefore, a reduction in the rate of Y but not GI transcription in

Time (h)120 144

0

10000

20000

30000

40000 WS +Sucgi-11/1 +Sucgi-11/2 +Suc

0

5

101520 WS

WS +Suc

Time (h)120

Nor

mal

ized

CA

B2:

:luc

Lum

ines

cenc

e

0

5

101520 gi-11

gi-11 +Suc

A B

CA

B2:

:luc

Lum

ines

cenc

e

020000400006000080000

100000120000140000

WSgi-11/1 gi-11/2

Time since treatment (h)

CA

B2:

:luc

Lum

ines

cenc

e

0

2000

4000

6000

D

0

2000

4000

6000

8000

1000090 mM Man90 mM Suc

gi-11

WS

Period (h)

23 24 25 26 27 28

Rel

ativ

e am

plitu

de e

rror

0.0

0.1

0.2

0.3

0.4

0.5

WS -Sucgi-11/1 -Sucgi-11/2 -Suc

WS +Sucgi-11/1 +Sucgi-11/2 +Suc

C

72 96480 2472 96480 24

4824 3612-12 0

Fig. 4. GI is required for long-term but not short-term modulation by su-crose. (A) Normalized CAB2::luc luminescence in LD and DD for wild-type WS(Upper) and gi-11 (Lower) seedlings, grown on agar media containing 90mM sucrose (open circles) or 90 mM mannitol (solid circles). (B) CAB2::lucluminescence in LD and LL comparing wild-type WS (solid symbols) and twoprogreny of a single gi-11 transformant (gi-11/1 and gi-11/2; open symbols),grown on agar media with (Upper) or without (Lower) 90 mM sucrose. (C)Circadian period estimates of CAB2::luc in LL for wild-type (filled symbols)and gi-11 (open symbols) seedlings, grown on agar media containing 90 mMsucrose (triangles) or 90 mM mannitol (circles). Relative amplitude errorquantifies the rhythmic robustness (0 is sine wave, 1 indicates no oscillation).(D) Wild-type (Upper) or gi-11 mutant seedlings were grown in entrainedLD12/12 cycles before transfer to constant dark (DD). After 60 h (CT0), 90 mMsucrose (open circles) or 90 mM mannitol (solid circles) was applied to aerialtissue. Bars on the abscissa are open for light, solid for darkness, and hatchedfor dark in the subjective day.

5108 | www.pnas.org/cgi/doi/10.1073/pnas.1015452108 Dalchau et al.

the new model (24) might cause suppressed rhythms, as observedin the absence of exogenous sucrose (Fig. 1B).On the basis of our analysis using the three-loop model, we

hypothesized that if GI does fulfill the role of Y, then sucrosesensing by the circadian oscillator should be compromised in ginulls. This proved to be the case for long-term responses to su-crose (Fig. 4), demonstrating GI is required for metabolic sensingby the plant circadian clock as suggested by overaccumulation ofstarch in gi nulls (39). However, transcriptional responses of theoscillator to sucrose were unaffected by gi-11 mutations (Fig. 4),which was not consistent with our predictions based on the N5model. The difference in response of the gi-11 nulls to long-termand transient changes in sucrose is consistent with there beinga functional difference between the biochemical networks thatperceive the fast and slow changes in intracellular sucrose imposedin this study. The difference in the response of the gi-11 nulls totransient changes in sucrose compared with that predicted by theN5 model might also reflect that GI does not fulfill all of thefunctions of the hypothetical model component Y (24, 29). Byrecognizing GI as a necessary regulator and demonstrating assaysthat admit significant sucrose sensitivity, it should now be possibleto characterize more completely the link(s) between metabolismand circadian rhythms in plants.

Materials and MethodsImaging of Luciferase Activity for Sucrose-Supplemented Growth Media Assay.Circadian rhythms of luciferase bioluminescence were measured using aPhotek IFS32 photon counting camera as described previously (40). Circadian

period estimates and relative amplitude errors (RAE) were obtained usingfast Fourier transform (nonlinear least-squares method) conducted with theBiological Rhythms Analysis Software System (BRASS) (41).

Imaging of Luciferase Activity for Topical Sucrose Treatment. Circadian rhythmsof luciferase luminescenceweremeasuredbefore andafter treatmentof90mMsucrose (Fisher Scientific) or 90 mM mannitol (BDH Laboratory Supplies). Ara-bidopsis WS seeds were surface sterilized, stratified at 4 °C for 2 d in darkness,and germinated on half-strength Murashige and Skoog medium (DuchefaBiochemie), 0.8% (wt/vol) agar, and no sucrose. Sixteen clusters of 3–5 seedlingswere grown in 12 h light/12 h dark (LD12/12) for 10 d at constant 19 °C and thentreated with 100 μL of 1 mM luciferin and placed in darkness for 60 h or 72 h.Photon counting was conducted as above for 48 h before sucrose/mannitoltreatment. A total of 100 μL of 90 mM sucrose/mannitol was then applied toseedling clusters in low intensity green light to prevent light entrainment (12clusters of sucrose treatment, 4 clusters with mannitol treatment). Photoncounting was then resumed immediately and up to 48 h following.

ACKNOWLEDGMENTS. We thank Prof. G. Coupland for donation of theWassilewskija ecotype gi-11 CAB2::luc seed line. We thank the Biotechnol-ogy and Biological Sciences Research Council, United Kingdom, for grants toA.A.R.W. (supporting N.D., A.N.D., H.M.B., F.C.R., M.A.S., and M.J.H.) and theEngineering and Physical Sciences Research Council, United Kingdom, fora grant to A.A.R.W. and J.M.G. (supporting G.-B.S.). G.-B.S. also acknowl-edges the support of the Engineering and Physical Sciences Research Coun-cil-funded Centre for Synthetic Biology and Innovation at Imperial CollegeLondon where he did part of this work. S.J.B. is grateful for funding from theKorean Science Foundation. M.J.G. was a Corpus Christi College, CambridgeJunior Research Fellow and is also grateful to the Royal Society of Londonfor a grant supporting his research. A.A.R.W. acknowledges the financialsupport of the Isaac Newton Trust, Cambridge, allowing the purchase ofa photon-counting camera used in this study.

1. Dodd AN, et al. (2005) Plant circadian clocks increase photosynthesis, growth, survival,and competitive advantage. Science 309:630–633.

2. Devlin PF (2002) Signs of the time: Environmental input to the circadian clock. J ExpBot 53:1535–1550.

3. Gutiérrez RA, et al. (2008) Systems approach identifies an organic nitrogen-responsivegene network that is regulated by the master clock control gene CCA1. Proc Natl AcadSci USA 105:4939–4944.

4. Hanano S, Domagalska MA, Nagy F, Davis SJ (2006) Multiple phytohormonesinfluence distinct parameters of the plant circadian clock. Genes Cells 11:1381–1392.

5. Turek FW, et al. (2005) Obesity and metabolic syndrome in circadian clock mutantmice. Science 308:1043–1045.

6. Gardner MJ, Hubbard KE, Hotta CT, Dodd AN, Webb AAR (2006) How plants tell thetime. Biochem J 397:15–24.

7. Dodd AN, et al. (2007) The Arabidopsis circadian clock incorporates a cADPR-basedfeedback loop. Science 318:1789–1792.

8. Alabadí D, et al. (2001) Reciprocal regulation between TOC1 and LHY/CCA1within theArabidopsis circadian clock. Science 293:880–883.

9. Pruneda-Paz JL, Breton G, Para A, Kay SA (2009) A functional genomics approach revealsCHE as a component of the Arabidopsis circadian clock. Science 323:1481–1485.

10. Kim W-Y, et al. (2007) ZEITLUPE is a circadian photoreceptor stabilized by GIGANTEAin blue light. Nature 449:356–360.

11. Más P, Kim W-Y, Somers DE, Kay SA (2003) Targeted degradation of TOC1 by ZTLmodulates circadian function in Arabidopsis thaliana. Nature 426:567–570.

12. Harmer SL, et al. (2000) Orchestrated transcription of key pathways in Arabidopsis bythe circadian clock. Science 290:2110–2113.

13. Schultz TF, Kay SA (2003) Circadian clocks in daily and seasonal control ofdevelopment. Science 301:326–328.

14. Yin L, et al. (2007) Rev-erbalpha, a heme sensor that coordinates metabolic andcircadian pathways. Science 318:1786–1789.

15. Dalchau N, et al. (2010) Correct biological timing in Arabidopsis requires multiplelight-signaling pathways. Proc Natl Acad Sci USA 107:13171–13176.

16. Huang Y, et al. (2010) SUGAR-INSENSITIVE3, a RING E3 ligase, is a new player in plantsugar response. Plant Physiol 152:1889–1900.

17. Johnson CH, et al. (1995) Circadian oscillations of cytosolic and chloroplastic freecalcium in plants. Science 269:1863–1865.

18. James AB, et al. (2008) The circadian clock in Arabidopsis roots is a simplified slaveversion of the clock in shoots. Science 322:1832–1835.

19. Knight H, Thomson AJW, McWatters HG (2008) Sensitive to freezing6 integratescellular and environmental inputs to the plant circadian clock. Plant Physiol 148:293–303.

20. Locke JCW, Millar AJ, Turner MS (2005) Modelling genetic networks with noisy andvaried experimental data: The circadian clock in Arabidopsis thaliana. J Theor Biol234:383–393.

21. Locke JCW, et al. (2005) Extension of a genetic network model by iterativeexperimentation and mathematical analysis. Mol Syst Biol 1:2005.0013.

22. Locke JCW, et al. (2006) Experimental validation of a predicted feedback loop in themulti-oscillator clock of Arabidopsis thaliana. Mol Syst Biol 2:59.

23. Zeilinger MN, Farré EM, Taylor SR, Kay SA, Doyle FJ, 3rd (2006) A novel computationalmodel of the circadian clock in Arabidopsis that incorporates PRR7 and PRR9.Mol SystBiol 2:58.

24. Pokhilko A, et al. (2010) Data assimilation constrains new connections and com-ponents in a complex, eukaryotic circadian clock model. Mol Syst Biol 6:416.

25. Millar AJ, Kay SA (1996) Integration of circadian and phototransduction pathways inthe network controlling CAB gene transcription in Arabidopsis. Proc Natl Acad SciUSA 93:15491–15496.

26. Lu Y, Gehan JP, Sharkey TD (2005) Daylength and circadian effects on starchdegradation and maltose metabolism. Plant Physiol 138:2280–2291.

27. Kurepa J, Smalle J, Montagu MV, Inzé D (1998) Effects of sucrose supply on growthand paraquat tolerance of the late-flowering gi-3 mutant. Plant Growth Regul 26:91–96.

28. David KM, Armbruster U, Tama N, Putterill J (2006) Arabidopsis GIGANTEA protein ispost-transcriptionally regulated by light and dark. FEBS Lett 580:1193–1197.

29. Ito S, et al. (2009) A genetic study of the Arabidopsis circadian clock with reference tothe TIMING OF CAB EXPRESSION 1 (TOC1) gene. Plant Cell Physiol 50:290–303.

30. Richardson K, et al. (1998) T-DNA tagging of a flowering-time gene and improvedgene transfer by in planta transformation of Arabidopsis. Aust J Plant Physiol 25:125–130.

31. Mizoguchi T, et al. (2005) Distinct roles of GIGANTEA in promoting flowering andregulating circadian rhythms in Arabidopsis. Plant Cell 17:2255–2270.

32. Ni Z, et al. (2009) Altered circadian rhythms regulate growth vigour in hybrids andallopolyploids. Nature 457:327–331.

33. Nakahata Y, Sahar S, Astarita G, Kaluzova M, Sassone-Corsi P (2009) Circadian controlof the NAD+ salvage pathway by CLOCK-SIRT1. Science 324:654–657.

34. Nakahata Y, et al. (2008) The NAD+-dependent deacetylase SIRT1 modulates CLOCK-mediated chromatin remodeling and circadian control. Cell 134:329–340.

35. Ikeda M, et al. (2003) Circadian dynamics of cytosolic and nuclear Ca2+ in singlesuprachiasmatic nucleus neurons. Neuron 38:253–263.

36. Panda S, Poirier GG, Kay SA (2002) tej defines a role for poly(ADP-ribosyl)ation inestablishing period length of the Arabidopsis circadian oscillator. Dev Cell 3:51–61.

37. Stelling J, Gilles ED, Doyle FJ, 3rd (2004) Robustness properties of circadian clockarchitectures. Proc Natl Acad Sci USA 101:13210–13215.

38. Hubbard KE, Robertson FC, Dalchau N, Webb AAR (2009) Systems analyses ofcircadian networks. Mol Biosyst 5:1502–1511.

39. Eimert K, Wang SM, Lue WI, Chen J (1995) Monogenic recessive mutations causingboth late floral initiation and excess starch accumulation in Arabidopsis. Plant Cell7:1703–1712.

40. Love J, Dodd AN, Webb AAR (2004) Circadian and diurnal calcium oscillations encodephotoperiodic information in Arabidopsis. Plant Cell 16:956–966.

41. Plautz JD, et al. (1997) Quantitative analysis of Drosophila period gene transcriptionin living animals. J Biol Rhythms 12:204–217.

Dalchau et al. PNAS | March 22, 2011 | vol. 108 | no. 12 | 5109

PLANTBIOLO

GY