- American Society of Plant Physiologists
Abstract
The Arabidopsis early flowering 3 (elf3) mutation causes arrhythmic circadian output in continuous light, but there is some evidence of clock function in darkness. Here, we show conclusively that normal circadian function occurs with no alteration of period length in elf3 mutants in dark conditions and that the light-dependent arrhythmia observed in elf3 mutants is pleiotropic on multiple outputs normally expressed at different times of day. Plants overexpressing ELF3 have an increased period length in both constant blue and red light; furthermore, etiolated ELF3-overexpressing seedlings exhibit a decreased acute CAB2 response after a red light pulse, whereas the null mutant is hypersensitive to acute induction. This finding suggests that ELF3 negatively regulates light input to both the clock and its outputs. To determine whether ELF3's action is phase dependent, we examined clock resetting by using light pulses and constructed phase response curves. Absence of ELF3 activity causes a significant alteration of the phase response curve during the subjective night, and constitutive overexpression of ELF3 results in decreased sensitivity to the resetting stimulus, suggesting that ELF3 antagonizes light input to the clock during the night. The phase of ELF3 function correlates with its peak expression levels in the subjective night. ELF3 action, therefore, represents a mechanism by which the oscillator modulates light resetting.
INTRODUCTION
Organisms exhibit daily biological rhythms in many physiological processes. These overt rhythms proceed even in the absence of external environmental cues and are thought to be regulated by an internal biochemical oscillator, the circadian clock. The circadian clock allows organisms to anticipate changes in their environment, such as the onset of dawn and dusk, thereby coordinating temporal phases of physiology with the external environment. In plants, the circadian clock organizes a wide array of daily rhythms, including the expression of genes involved in photosynthesis, sugar metabolism, nutrient assimilation, stomatal opening, and leaf movements. The oscillator possesses a resetting mechanism to receive timing cues from the environment. Resetting of the clock by light enables the oscillator to maintain a stable phase relationship with the external photoperiod and to detect gradual shifts in the onset of dawn as seasons change. The oscillator and the light-resetting mechanism work in concert to measure photoperiod and thereby coordinate photoperiod-dependent events such as floral induction.
The early flowering3-1 (elf3-1) mutation in Arabidopsis was identified in a screen for mutants with altered flowering time phenotypes (Zagotta et al., 1992). The mutant flowers early in both short and long photoperiods compared with the wild type (Zagotta et al., 1996). Previous analyses of overt circadian rhythms in the mutant suggested a gross defect in oscillator function and/or regulation. Clock-regulated rhythms of leaf movement, hypocotyl cell elongation, and transcription from the chlorophyll A/B–binding protein 2 gene (CAB2) promoter become arrhythmic in elf3-1 mutant seedlings under free-running conditions of constant light (LL; Hicks et al., 1996; Reed et al., 2000). A weaker allele, elf3-7, which has a less severe flowering time phenotype than does the elf3-1 allele, also exhibits similar arrhythmia in clock outputs under LL (Reed et al., 2000). In constant dark (DD), however, elf3 mutants exhibit some signs of clock function (Hicks et al., 1996). Therefore, it was proposed that ELF3 mediates an interaction between light and the circadian clock and that defects in ELF3 function lead to light-dependent arrhythmia. Recent reports support this hypothesis, suggesting that ELF3 interacts physically with a photoreceptor (Liu et al., 2001) and acts as a zeitnehmer for light signaling to the circadian clock (McWatters et al., 2000).
In Arabidopsis, the interaction between light and the circadian clock is dependent on both the fluence rate and the quality of light. For diurnal organisms with a functional clock in LL, there exists an inverse relationship between light intensity and free-running period length: increasing light intensity causes a shortening of period length of the overt rhythm (Aschoff's rule; see Aschoff, 1979), resulting in a classic fluence rate response curve (FRC). Such FRCs can be constructed for different wavelengths of light. In Arabidopsis, two major classes of photoreceptors, phytochromes and cryptochromes, mediate red and blue light photoreception, respectively (Somers et al., 1998a). Blue light and red light FRCs, along with genetic analyses, have suggested the involvement of different photoreceptors under different light quality and fluence rates in signaling to the clock (Millar et al., 1995; Devlin and Kay, 2000). For example, mutants lacking phytochrome B (phyB) exhibit a specific deficiency in response to high fluence rates of red light for shortening of period length, and transgenic plants overexpressing phyB have period lengths shorter than those of wild-type plants (Millar et al., 1995). Such FRC analyses have been used to place period-affecting mutants in either the light input pathway or the oscillator (Somers et al., 1998b, 2000).
Another variable for the interaction between light input and the circadian clock occurs in a phase-dependent manner. Although light is a strong resetting stimulus for the circadian oscillator, the same amount of light administered as a bright pulse at different phases of the circadian cycle may cause phase resetting of different magnitudes and/or directions (Johnson, 1992). Such phase shifts can be plotted as a function of the circadian phase of the stimuli to make a phase response curve (PRC). Typically, PRCs of circadian oscillators for light pulses exhibit phase delays in the early subjective night and phase advances in the late subjective night, with little phase resetting occurring during the subjective day (Johnson, 1998). Regulation of the magnitude and direction of phase resetting allows the organism to maintain a stable phase relationship between the circadian oscillator and the external light environment. The shape of the PRC reflects the strength of the resetting stimulus. For example, lower doses of light pulses during subjective night produce smaller phase shifts than do those caused by stronger light pulses (Johnson, 1992). Therefore, the shape of the PRC can be used to measure the sensitivity of the oscillator to light input at different phases.
However, the fundamental requirement for using PRCs as a tool to determine light inputs is an assay of robust circadian rhythmicity that persists in DD, which allows the effects of light pulses on phase resetting to be measured. In Arabidopsis, most of the overt rhythms used to study clock function for an extended time, such as rhythmic transcription from the CAB2 promoter, do not persist in DD (Johnson, 1998). This dilemma was overcome recently with the use of transgenic plants containing a luciferase transcriptional reporter driven by the CCR2 (for COLD CIRCADIAN REGULATED 2) promoter, which exhibits robust bioluminescence cycling under free-running conditions of both LL and DD (Strayer et al., 2000).
An additional interaction between the oscillator and light signaling in Arabidopsis occurs at the output levels. Some clock-regulated genes respond to light pulses by means of an acute phytochrome-mediated induction of transcription (Karlin-Neumann et al., 1988). This acute induction is followed by circadian cycling that persists for one or two cycles. The magnitude of acute induction varies with the time of day, suggesting possible gating by the circadian clock (Millar and Kay, 1996). It is not known conclusively whether the light input pathway to the clock participates in the gating of the acute response. Mutants that affect light input, such as elf3, should prove useful in characterizing this pathway.
The ELF3 gene was cloned recently and found to code for a novel protein with no significant sequence similarity to characterized proteins in the existing public databases. The elf3-1 allele has a nonsense mutation in the third of four exons (Hicks et al., 2001). This mutant accumulates neither ELF3 transcript nor protein product; therefore, it is likely a null allele (Hicks et al., 2001; Liu et al., 2001). To determine the role of ELF3 in the Arabidopsis circadian system, we characterized the expression pattern of ELF3 and analyzed the effects of the misexpression of ELF3 on circadian rhythms and light responses. Furthermore, we used the ccr2::luc reporter to determine conclusively the effect of elf3 mutations on oscillator function in the absence of light.
RESULTS
ELF3 Expression Is Regulated by the Circadian Clock
To determine ELF3's role in the circadian system and to gauge the influence of the clock and light input on ELF3, we defined ELF3's transcriptional expression by using noninvasive imaging of plants containing a fusion construct of the ELF3 promoter and the firefly luciferase coding region, as pioneered with cab2::luc (Millar et al., 1992).
Bioluminescence from elf3::luc in the wild-type background maintained a robust circadian rhythm after transfer to LL conditions (Figure 1A) . In LL, elf3::luc expression peaked at ∼16 hr circadian time and had an antiphasic relationship with the expression of the morning-phased reporter, cab2::luc. Under free-running conditions of constant darkness (DD), cycling of elf3::luc persisted with a similar phase (Figure 1D). Although elf3::luc expression was less robust after a few days in DD than in LL, a weak elf3::luc rhythm was detectable in these dark-adapted plants up to 6 days after transfer to dark. The cycling exhibited by elf3::luc was consistent with that observed in RNase protection assays (Hicks et al., 2001) and protein gel blot analyses of nuclear proteins (Liu et al., 2001). Also, the phase of elf3::luc expression is roughly equivalent to the predicted phase of ELF3 function on the basis of the phase at which the oscillator arrests in the elf3 mutants in LL (McWatters et al., 2000).
ELF3 Transcription Is Regulated by the Circadian Clock.
Seedlings were grown under 12-hr-light (white fluorescent; 50 to 60 μmol·m−2·sec−1)/12-hr-dark cycles for 6 days before transfer to LL (time 0) or DD (time 12). Shaded bars indicate subjective day and night. WT, wild type.
(A) elf3::luc expression is rhythmic in continuous white light. Mean bioluminescence traces (±se) from wild-type elf3::luc (closed circles; n = 5) and cab2::luc (gray squares; n = 6) seedlings are plotted.
(B) elf3 mutations cause arrhythmic expression of elf3::luc in LL. Mean elf3::luc bioluminescence traces (±se) from wild-type (closed circles; n = 7), elf3-1 (open triangles; n = 12), and elf3-7 (closed gray triangles; n = 9) seedlings are plotted.
(C) elf3 mutations lead to arrhythmic ccr2::luc expression in LL. Mean ccr2::luc bioluminescence traces (±se) from wild-type (closed circles; n = 12), elf3-1 (open triangles; n = 12), and elf3-7 (closed gray triangles; n = 12) are plotted.
(D) elf3::luc expression is rhythmic in DD. Representative bioluminescence traces from individual wild-type elf3::luc (closed circles) and ccr2::luc (open diamonds) seedlings are plotted. Data from elf3::luc plants are plotted on the left y axis, and ccr2::luc data are plotted on the right y axis.
(E) ccr2::luc rhythms are robust in DD, and period length is unaffected by ELF3 levels. Mean ccr2::luc bioluminescence traces from wild-type (closed circles; n = 27), elf3-1 (open triangles; n = 27), and ELF3-OX (open circles; n = 27) are plotted. Data from wild-type and ELF3-OX plants are plotted on the left y axis, and elf3-1 data are plotted on the right y axis.
elf3 Mutations Cause Conditional Arrhythmia of Both Day- and Night-Phased Circadian Markers
Circadian expression of ELF3 may define a clock output that regulates a subset of clock-controlled genes with a similar phase of expression. Lesions in ELF3 have been shown to cause arrhythmic expression of the morning-specific circadian clock–regulated genes CAB2 and LHY (Hicks et al., 1996; Schaffer et al., 1998). To determine whether ELF3 acts directly upstream of genes expressed in the morning in a promoter-specific manner, we tested the effect of elf3 mutations on clock-controlled genes that were expressed at different phases. Expression of both elf3::luc (Figure 1B) and ccr2::luc (Figure 1C; peaks at ∼12 hr circadian time) became arrhythmic when the elf3 mutants were placed in LL.
ELF3 Is Not a Classic Component of the Central Oscillator
Both the circadian rhythm of ELF3 expression and the global effects of mutations at the ELF3 locus on multiple circadian outputs suggest a role for ELF3 in the central oscillator or in the light input pathway to the clock. If it is a component of the oscillator, ELF3's action should persist in the absence of light, as has been shown for the oscillator mutant toc1 (Somers et al., 1998b). However, if it is an input component, the phenotypes of the elf3 mutation should be restricted to LL conditions only. Although the expression of cab2::luc in etiolated seedlings after a red light pulse showed some sign of clock function, it was difficult to assess the pace of the oscillator from the low amplitude and the rapidly damping cab2::luc rhythm under the free-running conditions of DD. However, using the ccr2::luc reporter, we were able to show that circadian regulation in elf3-1 mutants persisted for several days under DD conditions (Figure 1E). Although the amplitude of ccr2::luc rhythm in DD was affected by the loss of ELF3, neither the phase nor the period length of the rhythm in the mutant (period length [τ] = 25.41 ± 0.28 hr [variance-weighted mean ± variance-weighted se] as estimated by fast Fourier transformation–non-linear least squares [FFT-NLLS] analysis; n = 17) was statistically different (using Student's two-tailed heteroscedastic t test) from that of the wild-type rhythm (τ = 25.07 ± 0.09 hr; n = 23).
The persistence of rhythmic expression of a clock-controlled gene in DD suggests the presence of a functional clock in the elf3-1 mutant in DD. However, it is possible that cycling ELF3 action is essential for functioning of the clock under LL; therefore, it may act as an essential clock component under LL conditions. If this were true, constitutive expression of ELF3 would disrupt clock function in LL, leading to arrhythmia. To test this hypothesis, we made use of plants constitutively overexpressing a genomic copy of ELF3 driven by a strong viral promoter (the 35S promoter from cauliflower mosaic virus). Overexpression of ELF3 protein was verified by protein gel blot analysis (Liu et al., 2001). Several circadian markers were used to assay the state of the circadian system in ELF3-overexpressing (ELF3-OX) plants. Transcription of ELF3 and CCA1 (for CIRCADIAN CLOCK ASSOCIATED 1, a myb-related factor; Wang and Tobin, 1998), which encode gene products believed to be closely associated with clock function, continued to cycle with a circadian rhythm in ELF3-OX plants under LL (Figure 2A) . Also unaffected by ELF3 overexpression was the rhythmic transcription of two circadian clock–regulated output genes, CAB2 and CCR2 (Carpenter et al., 1994; Kreps and Simon, 1997) (Figure 2B). We checked for cycling expression of clock outputs in ELF3-OX seedlings under DD and found that all of the circadian reporters assayed continued to cycle: ccr2::luc (Figure 1E), cab2::luc (data not shown), and elf3::luc (data not shown). The persistence of overt rhythms in plants constitutively expressing ELF3 excluded the possibility that ELF3 cycling is required for oscillator function. It appears that ELF3 expression alone is sufficient for a functional clock in both LL and DD.
The Circadian Clock Remains Intact in Seedlings Overexpressing ELF3.
Luminescence in seedlings constitutively overexpressing a genomic copy of ELF3 under the control of the cauliflower mosaic virus 35S promoter (ELF3-OX) was determined as described in Figure 1 and Methods. After entrainment in 12-hr-light/12-hr-dark cycles, plants were transferred to continuous red light (∼200 μmol·m−2·sec−1) (time 0).
(A) Expression of genes closely associated with circadian function, elf3::luc (closed circles; n = 36) and cca1::luc (open squares; n = 9), is rhythmic in ELF3-OX seedlings. Data from elf3::luc plants are plotted on the left y axis, and cca1::luc data are plotted on the right y axis.
(B) Expression of the circadian output genes cab2::luc (closed gray squares; n = 14) and ccr2::luc (open diamonds; n = 10) is rhythmic in ELF3-OX seedlings. Data from cab2::luc plants are plotted on the left y axis, and ccr2::luc data are plotted on the right y axis.
Error bars indicate ±se.
Overexpression of ELF3 Lengthens Circadian Period Length in a Light-Dependent Manner
Although circadian rhythms were robust in ELF3-OX plants, we observed that in constant white light, the period length of rhythms in ELF3-OX plants (26.41 ± 0.18 hr for cab2::luc in ∼70 μmol·m−2·sec−1) was actually longer than it was in wild-type plants (24.28 ± 0.15 hr), suggesting a change in the pace of the central oscillator. However, under DD, in which robust ccr2::luc rhythms persisted, overexpression of ELF3 did not alter period length (τ = 25.10 ± 0.10 hr in ELF3-OX [n = 27]; τ = 25.07 ± 0.09 hr in the wild type [n = 23]). Rather, the phase of ccr2::luc expression after transfer to DD was delayed by ∼2 hr. This delay likely was an artifact of the ELF3-OX clock running slowly during the last light period before the seedlings were transferred from light/dark entrainment to DD.
To evaluate ELF3's role as an input to the clock, we took advantage of Aschoff's rule and constructed FRCs by measuring the period length of the clock in wild-type and ELF3-OX plants over a range of red light and blue light fluence rates. At medium to high fluence rates of red light, overexpression of ELF3 extended the period length of cab2::luc (Figure 3A) . In continuous blue light, however, the period length of cab2::luc (Figure 3B) was increased in ELF3-OX plants regardless of fluence rate. Similar effects were seen on the period length of elf3::luc in ELF3-OX plants (data not shown). These effects on period length indicate that the ELF3 gene product is involved in the repression of light input to the circadian oscillator.
ELF3 Overexpression Lengthens the Period of the Circadian Clock in Red Light and Blue Light.
After entrainment, wild-type (WT) and ELF3-OX seedlings were transferred to continuous red light or blue light at specific fluence rates, and luminescence was determined as described in Figure 1 and Methods. Variance-weighted means (± variance-weighted se) after >112 hr in free-running conditions are presented in FRCs. Asterisks, P < 0.01 (Student's two-tailed heteroscedastic t test).
(A) Effect of red light fluence rate on free-running period of cab2::luc expression in wild-type (closed circles; n = 18, 18, 18, 18, 17, 16, and 18) and ELF3-OX (open circles; n = 16, 18, 18, 18, 13, 11, and 36) seedlings.
(B) Effect of blue light fluence rate on free-running period of cab2::luc expression in wild-type (closed circles; n = 14, 13, 12, 18, 22, 18, 18, and 17) and ELF3-OX (open circles; n = 6, 17, 17, 18, 18, 18, and 18) seedlings.
ELF3 Attenuates Light Induction of a Circadian Output
elf3 null mutants are hypersensitive to the phytochrome-mediated acute induction of cab2::luc (Anderson et al., 1997), suggesting that ELF3 also may antagonize the light induction of light-responsive clock outputs. ELF3-OX seedlings, therefore, should have decreased induction and/or increased decay kinetics of cab2::luc expression after a light pulse. Etiolated seedlings were given a 2-min red light pulse, and bioluminescence rhythms from cab2::luc were followed for several hours (Figure 4A) . Consistent with previous reports, cab2::luc expression in elf3-1 seedlings increased nearly 12-fold compared with that of the preflash level, whereas the wild-type seedlings showed only an approximately sixfold increase. Although the acute induction was decreased in ELF3-OX seedlings (approximately fourfold induction), the difference in induction levels between the wild type and ELF3-OX was not statistically significant. Interestingly, the decay kinetics of the acute induction also differed among the three genotypes. In both elf3-1 and wild-type seedlings, cab2::luc expression continued to increase for ∼3 hr after light treatment, whereas in ELF3-OX seedlings, expression of cab2::luc peaked at least 1 hr earlier. Furthermore, in the ELF3-OX seedlings, the decrease after the acute induction was faster and to a more reduced level than was that in the wild type. elf3-1 exhibited the slowest decay kinetics.
ELF3 Attenuates Phytochrome-Mediated Induction of Circadian Outputs.
Luminescence from clusters of ∼50 etiolated cab2::luc seedlings was measured before and after a 2-min red light (RL) pulse (∼25 μmol·m−2·sec−1 at time 0; red arrow). Preflash levels were normalized to 1 so that induction levels could be evaluated and compared.
(A) ELF3 levels affect the magnitude of cab2::luc acute induction. Normalized mean cab2::luc bioluminescence traces (±se) from the wild type (WT; blue circles; n = 3), elf3-1 (open purple triangles; n = 4), and ELF3-OX (open green circles; n = 4) are plotted.
(B) Complete phytochrome response is present in the elf3-7 weak allele. Representative individual traces from the wild type (blue circles), elf3-1 (open purple triangles), and elf3-7 (closed orange triangles) are plotted.
The acute response addressed above was followed several hours later by long-period oscillations in cab2::luc expression, a response typically attributed to regulation by the circadian clock. elf3-1 null mutants display the acute, phytochrome-mediated response but are defective in the subsequent circadian regulation (Anderson et al., 1997). Although the weak allele, elf3-7, like the null mutant, was arrhythmic in LL, we wanted to determine whether enough ELF3 activity was present in elf3-7 to allow cycling of cab2::luc after a light pulse. Indeed, induction in etiolated elf3-7 seedlings was similar to that in the wild type for both the acute and circadian responses (Figure 4B).
ELF3 Regulates Light Resetting of the Clock
ELF3's involvement in both blue light and red light signaling to the clock and in the induction of outputs suggests that its action may lie very close to the oscillator rather than in directly antagonizing a photoreceptor. Because ELF3 appears to be an integral component of light input to the clock and its own expression cycles, it likely gates both light input to the clock and acute induction of the circadian outputs by restricting the light sensitivity of these pathways in the evening. Therefore, it may be expected that ELF3 is involved in light resetting of the clock.
We probed the role of ELF3 in light resetting of the clock by measuring the phase changes of ccr2::luc expression after light pulses were given to wild-type, elf3-1 null, and ELF3-OX plants in DD. Red light and blue light PRCs (Figures 5A and 5B , respectively) showed that wild-type plants exhibited little resetting in the subjective morning, when light pulses resulted in small phase advances. Later in the subjective day, the light stimuli elicited delays that continued to increase in magnitude. The greatest phase shift was during the subjective night, when large delays became large advances at what is termed the breakpoint. Near dawn, the small phase advances observed in the early morning returned. Although the overall shapes of both red light and blue light PRCs were similar, they differed in the magnitude of their largest phase resetting. Red light produced up to ∼7-hr phase delays and ∼10-hr phase advances, whereas blue light produced up to ∼9-hr phase delays and ∼6-hr phase advances. ELF3-OX plants experienced smaller phase shifts than did wild-type plants. The effects of ELF3 overexpression on light pulse entrainment of the clock were most pronounced in the red light PRC; at most, ELF3-OX plants experienced ∼6 hr of phase delays or advances. Blue light pulses given to ELF3-OX plants also elicited smaller delays than they did in wild-type plants.
ELF3 Levels Affect Sensitivity of the Clock to Resetting by Light.
PRCs for the wild type (WT; blue circles), elf3-1 (open purple triangles), and ELF3-OX (open green circles) constructed by plotting phase shifts of ccr2::luc expression (± pooled se; circadian hour) elicited by hour-long light pulses are plotted against the circadian times at which the light pulses were administered. Phase advances are plotted as positive values, and delays are plotted as negative values. Circadian times at which a light pulse induces arrhythmicity in elf3-1 seedlings are indicated by purple ×'s above the x axis. PRCs have been double plotted to aid in identification of the circadian features.
(A) Red light PRC.
(B) Blue light PRC.
In response to both red light and blue light pulses, the breakpoint observed in wild-type PRCs was abolished by ELF3-OX: the transition from delay to advance was continuous and more gradual (Figures 5A and 5B). elf3-1 seedlings, which lack ELF3 protein, showed comparable phase delays and advances during the subjective day. However, during subjective night, when ELF3 expression would normally peak, light-pulsed elf3-1 null mutants either experienced larger than wild-type phase shifts or became arrhythmic in ccr2::luc expression. This effect was more pronounced with red light pulses than with blue light pulses. Likewise, the conditional arrhythmia of elf3 mutants was more prevalent in continuous red light than blue light (Hicks et al., 1996). When administered in early to mid subjective night, red light pulses induced arrhythmia in most of the mutant seedlings; circadian rhythms were detected in only 9% compared with 69% in the wild type during the same phase and 47% in elf3-1 during other phases. Blue light pulses, however, produced increasingly larger phase delays, up to 11 hr, in subjective night before a light pulse caused arrhythmia (none of 21 elf3-1 plants were rhythmic, whereas four of 21 wild-type plants and 18 of 21 ELF3-OX plants were rhythmic during this phase).
DISCUSSION
ELF3 Is a Circadian Clock–Regulated Negative Component of Light Signaling
Expression of ELF3 is clock regulated, and ELF3 feeds back to antagonize both light signaling to the clock and light-mediated acute induction of circadian outputs (see the model in Figure 6) . By altering the levels of ELF3 so that expression is greatest at night, the plant is able to restrict the sensitivity of its clock to light signals during the night. This gating of light input to the oscillator results in an efficient clock that is resistant to transitory fluctuations in light levels at night—levels that are present in nature but do not correspond with dawn/dusk (i.e., change in cloud cover or stellar/lunar illumination).
Model of the Role of ELF3 in the Plant Circadian System.
ELF3 has a unique role within the circadian system of Arabidopsis. Circadian regulation of ELF3 expression enables the oscillator to control the light sensitivity of both clock resetting and the induction of circadian outputs in a phase-specific manner. This is a prime example of a gating mechanism in which a circadian output feeds back to regulate input pathways.
ELF3 is not the first example of a circadian output involved in light signaling. White collar 1 (WC-1) is regulated by the Neurospora clock in a post-transcriptional manner. WC-1 is believed to feed back as an activator of frequency (frq; a Neurospora oscillator component) and clock-controlled genes as well as to mediate the light induction of frq, clock-controlled genes, and itself (Lee et al., 2000). ELF3 and WC-1 are negative and positive regulators of light signaling in their respective circadian systems. The opposite functions of these cycling proteins may indicate either that they accomplish a similar goal in distinct manners or that in each of these organisms there is a pair of proteins with opposing functions that work together to properly regulate light signaling. Indeed, the circadian regulation of phyB (Bognar et al., 1999) and other photoreceptors, including cryptochromes (Harmer et al., 2000), has been demonstrated in Arabidopsis, supporting the possibility of rhythmically expressed positive and negative regulators of light signaling in this system. Although protein levels of phyB appear not to cycle (Bognar et al., 1999), post-translational circadian regulation of photoreceptors may occur. For example, transcription of SPA1, a negative regulator of phyA signaling (Hoeker et al., 1999), is regulated by the circadian clock (Harmer et al., 2000). Proteins such as ELF3 and SPA1 may be responsible for the circadian regulation of photoreceptor activity. It is hypothesized that such feedback loops contribute to robust oscillations and stable output (Lee et al., 2000).
Although most of the circadian phenotypes linked to misexpression of ELF3 are light dependent, a dark phenotype was observed consistently. The amplitude of circadian rhythms was decreased in the absence of ELF3. Therefore, ELF3 may function independent of light signaling to promote more robust oscillations by maintaining normal oscillator amplitude.
The conditional arrhythmia in LL and the decreased amplitude in DD of circadian outputs in elf3 mutants could be accounted for if ELF3 were part of the clock. However, three lines of evidence suggest that ELF3 does not function as a central component of the molecular oscillator: (1) circadian output from elf3 mutants is rhythmic in DD; (2) overexpression of ELF3 does not abolish rhythmicity of any outputs assayed in free-running conditions; and (3) in DD, period length is independent of ELF3 levels. Furthermore, the persistence of elf3::luc cycling in ELF3-OX seedlings suggests that ELF3 is not a component of a slave oscillator responsible for its own rhythmic expression.
ELF3 Acts in the Light Input Pathway to the Clock
Both red light and blue light FRCs of ELF3-OX plants are different from that of the wild type, suggesting a role of ELF3 in general light input to the clock. Under blue light, overexpression of ELF3 caused period lengthening over a wide range of fluence rates, similar to blue light FRCs of plants deficient in both CRY1 and CRY2 function. However, the hypocotyl elongation response of ELF3-OX plants under blue light was similar to that of the wild-type plants (X. Liu and R. Wagner, unpublished results), suggesting that ELF3 is not involved in the cryptochrome-mediated blue light inhibition of hypocotyl elongation. Under continuous red light of high fluence rate, ELF3-OX lengthens the free-running period length in a manner similar to the loss of functional phyB (Somers et al., 1998a). Again, the ELF3-OX plants do not exhibit the hypocotyl length response typical of phyB mutants; rather, hypocotyl elongation is hypersensitive to red light in ELF3-OX seedlings (Liu et al., 2001). Under low fluence rates of red light, ELF3-OX did not affect the period length of cab2::luc expression. Either ELF3 acts independent of the phyA signaling pathway to the clock or the fluence rate is too small to be affected by overexpression of ELF3. In general, overexpression of ELF3 restricted light signaling to the clock and resulted in longer period length under continuous light. The predicted hypocotyl phenotypes based on FRC comparisons, however, were not observed for ELF3-OX seedlings. ELF3's role in the regulation of hypocotyl elongation remains unclear.
Phytochrome-Mediated Acute Induction of Circadian Outputs
The circadian clock represses the phytochrome-mediated response to discrete light pulses in a phase-specific manner, resulting in the gating of induction such that outputs (e.g., cab2::luc) are less sensitive to light during the subjective night (Millar and Kay, 1996). For this type of gating to exist, a circadian system must generate rhythmic expression of gene products that allow and/or inhibit induction. In Arabidopsis, phytochromes are the major positive factors, whereas ELF3 is a negative regulator of this type of induction. Altering phytochrome or ELF3 levels changes the magnitude of light induction of cab2::luc in etiolated seedlings: the more phytochrome or the less ELF3, the greater the induction after a light pulse.
Compared with the wild type, acute induction of cab2::luc was diminished in ELF3-OX (Figure 4A), phyA, and phyB etiolated seedlings (Anderson et al., 1997). The second expression peak, ∼20 hr after the light pulse, was normal in phyA but remained low in phyB and ELF3-OX seedlings (Anderson et al., 1997; see also Figure 4A). These findings suggest that ELF3 may attenuate the phyB-mediated induction of circadian outputs, but they do not rule out involvement in the phyA pathway. If high levels of ELF3 are capable of completely blocking signaling through the phytochromes that ELF3 antagonizes, induction in seedlings deficient in those phytochromes should resemble that in ELF3-OX seedlings. The phyA phyB double mutant exhibits extremely low levels of induction after a light pulse (Anderson et al., 1997). This suggests that ELF3 may not act in the phyA pathway unless the inhibition by ELF3 is not absolute. In etiolated PHYA-OX seedlings, the active form of phyA, PfrA, is induced fourfold at 2 hr after a light pulse and maintains a twofold induction for up to 50 hr compared with the preflash levels (Anderson et al., 1997). Although PfrA levels are increased and fairly constant, cab2::luc expression cycles with increased amplitude, indicating that a rhythmic, negative regulator is acting on phyA induction of circadian outputs. This factor is either ELF3 or a protein with a similar function, perhaps SPA1. A light pulse experiment with elf3-1 seedlings deficient in one or more phytochromes could specify in which pathways ELF3 acts. elf3-1 seedlings would lose their hypersensitivity to light if the phytochrome(s) antagonized by ELF3 was absent.
This hypersensitivity of etiolated elf3 null seedlings is accompanied by the absence of successive peaks in cab2::luc expression after a light pulse (Anderson et al., 1997; see also Figures 4A and 4B). The presence of increased levels of the active form of phyA after a light pulse in etiolated PHYA-OX seedlings and the inherent stability of the active form of phyB suggest that light-pulsed etiolated seedlings experience the equivalent of constant dim illumination because the phytochrome pathways continue to be active. This persistence of phytochrome signaling may be the cause of arrhythmic cab2::luc expression after a light pulse in elf3-1. However, elf3-7 seedlings appear to have enough ELF3 activity to repress cab2::luc expression and maintain a circadian rhythm. If this is the case, there may be a fluence threshold below which outputs in the elf3-7 mutant retain circadian regulation in LL.
ELF3 Modulates Resetting of the Clock by Light
Period length under continuous light is the net effect of advances and delays produced by light input to the clock at different phases; therefore, it does not reveal the phase-specific interactions between the oscillator and the light input. Clock-regulated expression of ELF3 and its inhibitory effects on light input suggest that ELF3's effects on light input may be dependent on the phase of the oscillator, with maximum effects at phases of peak ELF3 expression and minimum effects when ELF3 levels are low. PRCs were helpful in addressing this question.
For Drosophila, the shape of the PRC is determined mostly by the oscillator component TIMELESS (TIM). Phase shifts are caused by light-induced degradation of the TIM protein. The magnitude and direction of phase resetting by light pulses can be explained by the levels of TIM and its subcellular localization, respectively. In Neurospora, light pulses cause phase shifting by inducing transcription of FRQ rather than by accelerating its degradation (Crosthwaite et al., 1995). The overall shape of the Neurospora PRC is determined by FRQ transcript level, with the greatest phase shifts occurring when FRQ transcript levels are minimal (Crosthwaite et al., 1995). These examples present two potentially different modes of resetting by light: acceleration of protein degradation and induction of transcription. Because ELF3 levels affect light resetting of the clock and ELF3 attenuates light-mediated induction of circadian outputs and, perhaps, the oscillator component(s), the Arabidopsis clock may be reset by light-mediated transcriptional induction of the oscillator component(s).
The Arabidopsis clock continues to run in DD in both elf3-1 and ELF3-OX seedlings, as indicated by ccr2::luc rhythms. Therefore, it is not surprising that PRCs from each of these genotypes have an overall shape similar to that of the wild type. The magnitude of phase shifts, however, changed in the mutant backgrounds. In response to pulses of red light or blue light during subjective night, wild-type seedlings produced large delays and advances with a clear breakpoint in between (Figure 6). Such PRCs usually are produced when the light pulses are saturating. Interestingly, the phase of the breakpoint coincided with the peak expression of ELF3. In ELF3-OX seedlings, reduced phase shifts were elicited by pulses in subjective night. Such topology of PRCs is usually associated with subsaturating light pulses (Pittendrigh, 1981; Johnson, 1992), suggesting that ELF3 antagonizes the light-resetting mechanism.
Deficiency of ELF3, as in elf3-1 plants, also had a significant effect during the subjective night. Light pulses on wild-type seedlings usually produce large phase shifts of up to 10 hr during this period. With the progression of subjective night, light pulses elicited increasingly larger phase shifts in the absence of ELF3. However, at the phase when ELF3 expression was maximal and light pulses produced the largest phase shifts in the wild type, a light pulse caused arrhythmia in most of the elf3-1 plants. As with ELF3-OX plants, the effect of red light pulses on elf3-1 plants was more severe than that of blue light pulses (see discontinuity in elf3-1 PRCs in Figure 5); likewise, the conditional arrhythmia of elf3 mutants is more prevalent in continuous red light than blue light (Hicks et al., 1996). This phenomenon of the suppression of circadian rhythmicity by critical light pulses that drive the oscillator permanently or transiently to the steady state (i.e., singularity) has been described for Drosophila pseudoobscura, mammals, and plants (Leloup and Goldbeter, 2000). The inhibitory effect of ELF3 on light input to the clock seems to serve as a safeguard mechanism against such severe light-induced perturbations of the oscillator. By increasing the intensity of the light pulse stimulus, it may be possible to suppress circadian rhythmicity in wild-type Arabidopsis.
Differences in the shapes of blue light and red light PRCs in the wild type may be due to independent action of the two signaling pathways in light pulse entrainment of the oscillator. This is corroborated by differences in the PRCs of elf3-1 seedlings. Furthermore, under LL conditions, red light fluence rate showed a clear inverse relationship with free-running period length, as expected for a circadian light-signaling mechanism in a diurnal organism. However, the blue light FRC lacks a clear inverse relationship with free-running period length (Somers et al., 1998a). The two pathways may be acting on separate oscillator components or on the same oscillator component but with different kinetics of resetting.
Alterations in FRCs in ELF3-OX seedlings and altered PRCs in both ELF3-OX and elf3-1 seedlings clearly demonstrate a genetic interaction between ELF3 and both blue light and red light input to the clock. ELF3 reduces the effectiveness of light in signaling and in resetting the clock. There are two possibilities for the mechanism by which ELF3 modulates light resetting. ELF3 may act simply as a valve outside the clock to regulate the flux of light. Alternatively, it may act directly on the oscillator and regulate its light sensitivity. The second mechanism is more likely. The reduced amplitude of CCR2 cycling in elf3-1 compared with that in the wild type is indicative of ELF3's role in maintaining the amplitude of the oscillator. However, the normal period length of both elf3-1 and ELF3-OX in DD suggests that ELF3 may not be essential in setting the pace of the oscillator in the absence of light input. Unless the function of ELF3 is post-translationally circadian regulated (e.g., rhythmic modification of activity by phosphorylation), the evidence of normal clock functioning in ELF3-OX seedlings in the absence of light suggests that the mere presence of ELF3, not its cycling, is necessary for its role in maintaining a robust oscillation. However, large changes in the shape of the PRCs for both ELF3-OX and elf3-1 plants during the subjective night suggest a role in maintaining oscillator progression during the night phase of the photoperiod. Such a role is consistent with ELF3's peak phase of expression.
Summary
Negative feedback loops are a central theme in circadian biology. We have shown that ELF3 feeds back to the clock in a rhythmic manner to negatively regulate light input to the oscillator and acute induction of circadian outputs during the night. Furthermore, elf3 is the only mutation shown to make a circadian system highly susceptible to suppression of circadian rhythms by a critical light pulse.
METHODS
Mutant Alleles and Heterozygous Lines
The elf3-1 mutation is a C-to-T transition in exon 3 that results in early termination. The elf3-7 mutation is a G-to-A transition in the exon 1 donor splice site that is predicted to result in an in-frame deletion of eight amino acids (K. Hicks and R. Wagner, unpublished results). ELF3-OX plants, which constitutively overexpress ELF3 genomic sequence, are described by Liu et al. (2001). In all experiments, the wild-type control is Columbia (Col-0).
Construction of the elf3::Ω::luc Reporter Construct
The 1.9-kb HindIII restriction fragment of the ELF3 promoter, just upstream of the ELF3 gene, was inserted into pPZP221, a T-DNA transformation vector used as a promoter::luciferase reporter system. pPZP221 is composed of an omega translational enhancer followed by the coding region from the firefly luciferase gene and the nopaline synthase termination sequence (Millar and Kay, 1996).
Creation of Transgenic Lines
Arabidopsis thaliana Col-0 seedlings were transformed via the Agrobacterium tumefaciens–mediated process; vacuum infiltration was used for the overexpression lines. T1 seed were surface sterilized and sown on Petri dishes containing half-strength Murashige and Skoog (1962) medium, 0.5% (w/v) sucrose, 0.8% (w/v) agar, and antibiotic. Transgenic elf3::luc T1 seed were selected on 75 mg/mL gentamycin. T2 seedlings were evaluated in bioluminescence assays (see below). The most robust lines with a single site of insertion (as judged by Mendelian segregation) and with luciferase activity faithful to the RNase protection data were chosen for further experiments.
Bioluminescence Assays
Seed for rhythm analyses were grown on 1 × Murashige and Skoog/0.8% agar medium with 3% (w/v) sucrose and imaged as described previously (Millar et al., 1995). However, expression of ccr2::luc in constant darkness (DD) was measured and analyzed using the Night Owl imaging system with WinLight software (Perkin-Elmer), whereas elf3::luc expression in DD was measured using the Packard TopCount multiplate scintillation counter as described by Strayer et al. (2000). Transgenic cab2::luc (Millar et al., 1992) had been introgressed into Col-0 before crossing with plants of various elf3 mutant backgrounds. The other circadian markers used were constructed originally in the Columbia ecotype. Period lengths for fluence response curves were measured as described previously (Somers et al., 1998b), and period estimates were obtained according to Plautz et al. (1997) and Millar et al. (1995). Dark-grown etiolated seedlings were imaged in clusters of ∼50 seedlings as described previously (Anderson et al., 1997). To depict the level of induction after a light pulse, counts from etiolated seedlings were normalized by setting prepulse levels to 1.
Phase Response Curves
Seedlings were grown as described above. After 5 days in light/dark cycles, seedlings were transferred to DD, and imaging of bioluminescence from ccr2::luc expression was initiated with the Night Owl imaging system. Beginning after 1 full day in DD, individual plates of seedlings were exposed to 1-hr light pulses at ∼3-hr intervals for ∼24 hr. Red light pulses were 40 μmol·m−2·sec−1, and blue light pulses were 25 μmol·m−2·sec−1. The period and phase of rhythms after the light pulses were estimated (Millar et al., 1995; Plautz et al., 1997) and converted to circadian time. These phase estimates of different pulsed populations were averaged to determine the variance-weighted mean and compared with those from nonpulsed control seedlings to evaluate the magnitude and direction of the phase shifts. The pooled standard error was calculated using the following equation: 
where np and nn are the sample sizes and vwSDp and vwSDn are the variance-weighted standard deviations for pulsed and nonpulsed populations, respectively. For the elf3-1 phase response curves (PRCs), some phase shifts are not plotted due to arrhythmia after the light pulse. In these cases, circadian rhythmicity was observed after the light pulse in only 0 to four of 21 to 27 elf3-1 plants. Rhythmicity was scored for traces with FFT-NLLS phase estimates between 22 and 30 hr that were represented faithfully by their respective fits.
Acknowledgments
We thank M. Straume for advice on the statistical analysis of PRCs; D. Somers for help with imaging; E. Last and H. Johnson for technical assistance; and C. Andersson and J.A. Kreps for providing transgenic cca1::luc and ccr2::luc seed, respectively. We are grateful to D. Alabadi, S. Harmer, P. Más, T. Schultz, and M. Yanovsky for critical comments on the manuscript. This research was supported by National Science Foundation Grant MCB-9808208 and U.S.–Israel Binational Agricultural Research and Development Agency Grant US-2964-97 to D.R.W.; by National Institutes of Health Grant GM56006 to S.A.K. and S.P.; by National Institutes of Health Training Grant GM07413 to M.F.C.; and by National Science Foundation Training Grant DGE-9552837 to M.F.C. and C.A.S. This is manuscript 13821-CB of the Scripps Research Institute.
Footnotes
-
↵1 Current address: Department of Cell Biology, Scripps Research Institute, La Jolla, CA 92037.
-
↵2 Current address: Exelixis Plant Sciences, Inc., 16160 S.W. Upper Boones Ferry Rd., Portland, OR 97224.
-
↵3 Current address: Department of Biology, National Science Foundation Center for Biological Timing, University of Virginia, Charlottes-ville, VA 22903.
- Received December 29, 2000.
- Accepted April 16, 2001.
- Published June 1, 2001.
References
