The short-period mutant, toc1-1, alters circadian clock regulation of multiple outputs throughout development in Arabidopsis thaliana

Close coordination of developmental and physiological processes with the environment is essential to successful plant growth. Although many of the key events in plant development depends on receiving the appropriate environmental signals at the right time, plants also possess internal timekeeping mechanisms that allow them to keep pace with and anticipate cyclic events in their environment. By alternating between restrictive and permissive phases, such an internal timekeeper can act to both gate the effect of an environmental stimulus, and to modulate the extent to which physiological processes occur. The best understood of these endogenous clocks is the one based on a 24 hour (circadian) pacemaker.

From sleep/wake cycles in mammals (Moore, 1997) to the transcriptional control of gene expression in cyanobacteria (Golden et al., 1997), a circadian clock facilitates the coordination of behavioral, developmental, and physiological events with the daily cycles of light and dark. A genetic approach towards understanding the clock was initiated more than 25 years ago in Drosophila, through a screen for circadian clock mutants (Konopka and Benzer, 1971), and has since been extended to algae (Bruce, 1972), fungi (Dunlap, 1996), mouse (King et al., 1997), cyanobacteria (Golden et al., 1997) and higher plants (Millar et al., 1995a). All have been successful in recovering single gene mutations that disrupt the function of the circadian system (Dunlap, 1996).

The goal of most studies has been to identify the primary components of the central oscillator of the clockwork. The conceptual framework guiding this pursuit has simplified the system into three domains (Block et al., 1993). The input, or entrainment pathway, is defined by the series of processes by which environmental signals are transduced to synchronize the clock with the outside world. These signals terminate at the central oscillator, which consists of a series of processes that generate an oscillatory feedback loop that sets the period length and phase of the overt rhythms. Finally, the output pathway constitutes a second signaling cascade that couples the activity of the central oscillator to the observable rhythms (i.e. the ‘hands of the clock’).

To qualify as a bona fide component (state variable) of the central oscillator certain criteria must be met that will distinguish an oscillator element from input or output components. First, the activity or amount of the component should itself oscillate. Second, constitutive expression (or complete loss) of the component should stop all oscillations (arrythmicity). Third, a transient change in the level of the component should phase shift the clock. Rigorous application of these criteria requires close control of expression level, most easily obtained by placing a cloned gene under the control of an inducible promoter. The product of the frequency locus in Neurospora meets these criteria (Aronson et al., 1994), and the Drosophila period protein has been tested similarly (Roshbash et al., 1996; Sehgal et al., 1996). However, no single criterion alone is sufficient to identify a pacemaker element; outputs cycle by definition and disruption or constitutive expression of an output element may mask the activity of the oscillator, giving the impression of arrythmicity. Similarly, changes in the level or pulse applications of an input element, such as light, can affect period length and phase of the clock.

The semi-dominant toc1-1 mutant was recovered in a screen for period length mutants in a line of Arabidopsis expressing a cab2::luciferase (cab2::luc) reporter gene (Millar et al., 1995a). Initial characterization of toc1-1 showed that the period length in two clock-controlled outputs, cycling of cab2:: luciferase (cab2::luc) activity and the rhythmic movements of expanding cotyledons and leaves, is 2-3 hours shorter than wild-type (WT) (Millar et al., 1995a). The mRNA abundance of two closely related RNA-binding proteins has also been shown to cycle with a shorter period in toc1-1 (Kreps and Simon, 1997). Both reports were limited to describing the effects of toc1-1 in young, light grown seedlings. Demonstration of aberrant clock function throughout development in a toc1-1 background would have two important implications. First, it would support the notion of a single, central clock mechanism operating throughout the life of the plant in a diversity of tissues. Second, it would demonstrate the importance of TOC1 in controlling the activity of that clock.

We describe here a detailed characterization of both the general aspects of the circadian clock in Arabidopsis and the position of TOC1 within that system. By examining the effects of toc1-1 on a variety of clock-controlled processes throughout development, we suggest that a single system is primarily responsible for controlling most, if not all, circadian phenotypes in the plant. In addition, we present evidence that TOC1 is a principle component of the central oscillator in Arabidopsis.

toc1-1 was originally isolated in the C24 ecotype as described in Millar et al. (1995a). F₃ populations backcrossed twice (toc1-1 F3-25, F3-59, F3-79) or 3 times (toc1-1 F3-73, F3-74) were selected for shortened period length using an in vivo luminescence assay (Millar et al., 1992). Lines introgressed twice (toc1-1/Laer F3-64, F3-6, F3-11) or 3 times (toc1-1/Laer F3-32, F3-13, F3-8, F3-10) into Landsberg erecta (Laer) were similarly selected for use in flowering time experiments. cab2::luciferase luminescence assays confirmed that toc1-1 acts semi-dominantly in the Laer ecotype (data not shown). Unless otherwise noted, seeds were surface sterilized and sown onto MS (Sigma) agar plates + 3% sucrose as previously described (Anderson et al., 1997).

Light grown seedlings used in luminescence assays were entrained under 50-60 μmol m⁻² second⁻¹ white fluorescent light (WL) in 12 hour light /12 hour dark (L/D) cycles at 22°C, or under 12 hour 24°C/ 12 hour 20°C temperature cycles, constant white fluorescent light (50-60 μmol m⁻² second⁻¹), for 5 days prior to commencement of imaging. For fluence rate response tests, plants were placed below the appropriate layers of neutral density filter (Roscolux #3402 or #3404) or wire screen prior to transfer to constant red light (RL) to obtain the desired fluence rate. Luminescence assays commenced with the beginning of continuous RL and plants were imaged for 25 minutes every 2 hours as previously described (Anderson et al., 1997). In temperature entrainment experiments, plants were either continued under temperature cycles throughout data collection (driven experiments) or held at 24°C (free-run experiments). Sampling time and frequency was as in fluence rate response tests. For temperature compensation tests, the plants were entrained as above and transferred to the appropriate constant temperature prior to the final dark period. Period estimates were obtained according to Plautz et al. (1997) and Millar et al. (1995b).

Dark-grown seedlings were imaged in 25-35 seedling clusters as previously described (Anderson et al., 1997) and the time between the 2^nd and 3^rd peaks after the RL pulse was determined manually.

After stratification of seeds at 4°C for 6 days, the plates were transferred into the appropriate continuous white (50-60 μmol m⁻² second⁻¹), red (20-25 μmol m⁻² second⁻¹) or blue (40 μmol m⁻² second⁻¹) light treatment for 8 days, when hypocotyl measurements were taken with a ruler. Plants grown under saturating far-red light (FRL) were dark germinated for 24 hours before being placed under continuous illumination.

Seed were stratified on agar plates and maintained in fluorescent light under long days (LD) (16 hours light :8 hours dark) (90 μmol m⁻² second⁻¹) or short days (SD) (8 hours light:16 hours dark) (180 μmol m⁻² second⁻¹) for 7-10 days after which they were transplanted to soil. Time to flowering was taken as the number of rosette leaves at time of 1 cm high flower bolt.

Surface sterilized seeds of C24 wild-type and toc1-1 F3-59 were sown onto 0.6% agar (0.5 MS, 1% sucrose, 50 μg ml⁻¹ kanamycin), stratified at 4°C for 4 days and maintained at 20±1°C; 150 μmol m⁻² second⁻¹ PAR (10 hour daylength) for 1 week. Seedlings were transferred to sterile 3:1 Levington M3 Compost (Levington Horticulture, Ipswich, UK):sand for 1.5-2 weeks, repotted into 5 cm diameter pots and transferred to a controlled environment chamber. Plants were maintained on a 12-hour photoperiod (112 μmol m⁻² second⁻¹) under constant temperature (20±0.1°C) and relative humidity (70±3%). Chamber CO₂ concentration was monitored using an infra-red gas analyser (WMA-2, PP Systems, Hitchin, UK) and maintained at 355-360 μl l⁻¹ by controlled injection of either pure CO₂ or CO₂-free air.

Viscous flow porometer cups similar to those described by Allaway and Mansfield (1969) and Weyers and Meidner (1990) remained attached to the leaves of 5- to 6-week-old vegetative plants (one cup per leaf per plant) throughout the experiment. Viscous conductance of each leaf was recorded for 2.5 minutes every 30 minutes for 2 day/night cycles and during continuous light.

The long hypocotyl hy1 mutation severely reduces the amount of spectrally active phytochrome and also increases the period length of the circadian clock in RL, relative to the wild type (Millar et al., 1995b). To determine whether TOC1 lies within the same phototransduction pathway as that which controls hypocotyl elongation, the hypocotyl length of toc1-1 was compared to wild type grown under different light qualities.

Fig. 1 shows the hypocotyl lengths of 3 independently isolated backcrossed lines of toc1-1 maintained under continuous white, red, far-red and blue light. As previously reported (Young et al., 1992), different light qualities vary in their relative effectiveness at inhibiting hypocotyl expansion in WT Arabidopsis seedlings. At the wavelengths tested here, this spectral-dependence relationship is unchanged by the toc1-1 mutation. Although there are statistically significant differences in hypocotyl length between wild type and some toc1-1 lines under the different light treatments, overall there is no consistently longer or shorter hypocotyl length among the 3 toc1-1 populations, relative to WT. The slight differences observed are presumably due to the presence of other mutations segregating in the background which have a minor effect on growth rate under some light conditions. Therefore, if toc1-1 defines a lesion in the light signaling pathway to the clock it must be independent or in a branch downstream of components involved in the light-mediated control of hypocotyl length.

To test whether the influence of toc1-1 extends beyond its effect on the clock in light-grown seedlings, we examined the effects of the mutation on the rhythm of cab2::luc expression in dark-grown, etiolated seedlings following a RL flash.

The kinetics and magnitude of the cab2::luc expression pattern during the first 8 hours after the RL flash are identical in the WT and three backcrossed lines of toc1-1 (Fig. 2). Luminescence levels are 3- to 4-fold above background within 2 hours after the RL pulse, and return to pre-flash levels 5-6 hours later. Our earlier work demonstrated that light input through the phytochrome A and phytochrome B photoreceptors accounts for the entire extent of the rapid RL flash response (Anderson et al., 1997). These results demonstrate that the toc1-1 mutation lies in a pathway independent of the one leading from the RL activation of phytochrome to the rapid induction of CAB2 expression.

The timing and magnitude of the second peak (18-21 hours post flash), marking the beginning of the clock-controlled portion of the waveform (Anderson et al., 1997), are very similar in the two backgrounds (Fig. 2). The dip in the second peak in the WT was not consistently observed and conversely it was seen occasionally in all the toc1-1 lines tested (data not shown). However, the position of the third peak (40-49 hours post flash) relative to the second peak consistently occurs approx. 3-4 hours earlier in the toc1-1 lines (23.0, 24.5, 22.5 hour after each respective second peak) than in the WT (27.0 hours after the second peak) (Fig. 2). This effect is consistent with the shorter period length of cab2::luc expression seen in light-grown seedlings of toc1-1 (Millar et al., 1995a). cab2::luc expression at later time points is too low to assess any differences among the lines reliably. Hence, TOC1 is effective in controlling period length in both achlorophyllous etiolated cotyledons and in the green, differentiated mesophyll of expanded cotyledons and leaves.

Rhythmic variation in the size of the stomatal aperture is a well-established example of how the circadian clock can control the physiology of discretely localized, highly differentiated cell types. For the first time in Arabidopsis, we have applied viscous flow porometry to monitor cyclic changes in stomatal aperture at very high resolution (Allaway and Mansfield, 1969). By using 5- to 6-week-old rosette plants we were also able to assay the role of TOC1 at a stage in development much later than previously reported.

Under L/D cycles, stomatal conductance is approx. 3-fold higher in the light than in the dark, in both the mutant and WT, and the maxima and minima of conductance are coincident in both backgrounds (Fig. 3). In contrast, under continuous light the two traces become nearly 12 hours out of phase by the end of the 6-day sampling period due to the shorter free-running period of toc1-1. Table 1 summarizes the period estimates obtained from repeated, independent measurements of stomatal conductance on individual leaves from toc1-1 and WT rosette plants. When the values obtained from the curve-fitting program are weighted in accordance with the confidence of the period estimate, the stomatal conductance rhythm in toc1-1 averages 3 hours shorter than the wild type.

These results are again consistent with the original report of a 3-3.5 hours shorter period in cab2::luc cycling in toc1-1. Since the size of guard cells is facilitated by the flux of ions across the plasma membrane and tonoplast, this assay shows that toc1-1 alters the rhythm of a physiological process very different from the transcriptional control of CAB2 expression in seedlings, and at a more advanced stage in plant development.

The involvement of the circadian clock in the determination of flowering time has been demonstrated in a wide variety of species (Vince-Prue, 1983). In Arabidopsis, a facultative long day plant, there is a 24-hour rhythm in the effectiveness that FRL pulses given over a continuous background of white light have in accelerating flowering (Deitzer, 1984). These results, together with the recent report of a circadian clock defect in the early flowering mutant elf3 (Hicks et al., 1996), provide the only evidence associating the circadian clock with the control of flowering time in Arabidopsis. It was clearly of interest then, to examine the effects of the toc1-1 mutation on the timing of the transition from vegetative to floral development.

Table 2 presents a summary of five independent experiments testing the responsiveness of toc1-1 lines to two different photoperiods. Five randomly selected toc1-1 lines (C24 ecotype) were grown under short day (8 hours light: 16 hours dark; SD) or long day (16 hours light: 8 hours dark; LD) regimes and rosette leaf number at flowering was recorded. The most consistent effect of the toc1-1 mutation in this genetic background is a diminished effect of daylength differences on flowering time, relative to WT. Within each trial the toc1-1 lines show a significantly smaller difference between the number of leaves produced in short days compared to the number produced in long days (SD-LD). This occurs through a slightly earlier flowering in SD paired with a slightly later flowering in LD (trial 1), or largely through later flowering in LD alone (trial 2).

When introgressed into the Laer background, the toc1-1 mutation has a more dramatic effect on flowering time (Table 2). In all cases, flowering in SD is significantly earlier than in the WT, with little effect on flowering time under LD. This results in a near day-neutral plant in which the SD-LD difference is reduced to 0-3 leaves, in contrast to the 5-14 leaf difference in WT (Table 2). These results are the first to suggest that a quantitative change in the pace of the clock can affect the correct processing of daylength information and result in a change in flowering time. Further experiments are in progress to conclusively eliminate the possibility that these results arise from pleiotropic effects of toc1-1 on development.

Since alterations in both the fluence rate and quality of the light input can affect the free-running period of the circadian clock (Aschoff, 1979), it is possible that toc1-1 disrupts a component of the light input pathway. This could result in an increased sensitivity to light, causing the oscillator to run faster than wild type at some or all fluence rates. We tested the effect of light intensity on the activity of the clock by assaying the period of cab2::luc cycling under different RL fluence rates after entrainment in L/D cycles. We chose RL because phytochrome has been implicated in mediating the control of period length by light (Millar et al., 1995b).

In both WT and toc1-1 backgrounds period length is inversely proportional to the logarithm of light intensity over a 500-fold range of RL fluence rates (Fig. 4). This is consistent with reports from a wide range of organisms that demonstrate a log-linear relationship between fluence rate and period length of the clock (Aschoff, 1960). At intensities below approx. 10 μmol m⁻² s⁻¹ light is clearly the limiting factor that determines period length. Higher fluence rates (10-200 μmol m⁻² s⁻¹) do not saturate the response, but the relative effectiveness of RL in shortening period is much less than at lower intensities. This change in slope suggests that with increasing fluence rates there is a shift to a different RL signaling system or that some components of the low-fluence transduction pathway have changed in abundance or activity.

Consistent with the original tests in WL (Millar et al., 1995a), at high RL fluence rates the free-running period in toc1-1 is 3-4 hours shorter than the WT (Fig. 4). The same constitutive shortening of period length (approx. 2 hours) is maintained down to the lowest fluence rates tested, as evidenced by the parallel slopes of the two linear regressions. In addition, the transition point from near saturating to limiting light (approx. 10 μmol m⁻² s⁻¹) is identical for both the WT and mutant (Fig. 4). Together these results show that the mutation has not increased the sensitivity of the input pathway to RL, but instead suggest that toc1-1 causes the central oscillator to run at a constitutively faster pace.

One hallmark of the circadian clock is the relatively slight effect temperature has on the period length of the oscillation. Whereas a 10°C increase in temperature (Q₁₀) will typically increase a biochemical reaction 2- to 3-fold (Q₁₀=2-3), the period of the clock is compensated over a wide range of temperatures and Q₁₀s are usually between 1.0-1.1 (Lakin-Thomas et al., 1990). Some long period alleles of period and frequency disrupt normal temperature compensation, and period lengths become strongly dependent on the ambient temperature (Hall, 1997). Thus, it was of interest to test the effects of toc1-1 on this aspect of clock function.

When grown at temperatures from between 12°C and 32°C, the free-running period of the WT varies no more than 2.5 hours (Q₁₀=1.0-1.1), demonstrating that temperature compensation is a property of the Arabidopsis circadian clock (Fig. 5). Throughout this range the near parallel slopes of the two best-fit lines shows that period length in toc1-1 is consistently shorter than WT by 2-2.5 hours at all temperatures tested (Fig. 5). These results demonstrate that toc1-1 causes no disruption of the normal temperature compensating properties of the clock, and conversely, growing temperature has no effect on the period-shortening effects of the mutation.

Although light cycles are the best characterized of the signaling inputs known to entrain the circadian clock, temperature cycles are equally effective in synchronizing the central oscillator with the external environment (Bünning, 1973). In the experiments described above, repeated L/D cycles or brief light pulses were the sole entraining stimuli presented to the plant. To further assess the position of TOC1 within the circadian system we tested whether the toc1-1 lesion affects the processes of temperature entrainment.

As the small s.e.m. indicates, all WT individuals (n=35) entrain to a 12 hour/12 hour 24°C/20°C thermoperiod (Fig. 6A). Peak and trough expression corresponds to the middle of the high (24°C) and low (20°C) temperature periods, indicating that the rise and fall of cab2::luc expression is not driven by the temperature changes, but controlled by a circadian clock entrained to the 24-hour thermoperiod. toc1-1 plants (n=34) also entrain, but the sawtooth waveform and the shift in peak and trough positions to near the temperature transition points clearly indicate that the mutant processes the entraining signal conveyed by these temperature cycles differently from WT plants. When both lines are subsequently maintained in constant temperature (Fig. 6B), however, the waveform of both lines are sinusoidal with toc1-1 plants running at a period approx. 2.6 hours shorter than the WT.

Further experiments suggested that the unusual waveform in toc1-1 results from entrainment to a cycle with a period (T) far removed from that of its endogenous oscillator. When both the WT and toc1-1 were entrained to a 20 hour thermoperiod (10 hour/10 hour; 24°C/20°C), only the toc1-1 lines display a sinusoidal waveform, with peaks and troughs occurring independently of the temperature transition points (Fig. 6C). In contrast, by the end of the trial, expression maxima and minima in the WT coincide closely with the temperature transitions, similar to that seen in toc1-1 under a 24-hour thermoperiod (Fig. 6A). This complementary set of experiments shows that entrainment to a cycle that differs greatly from that of the period of the endogenous clock can result in a distortion in the normal rhythmic pattern of gene expression. This also implies that in toc1-1 the clock runs with a short period under a 24-hour temperature cycle, and does not simply revert to a shorter period during free-run under constant conditions.

These data, together with the results from Fig. 5, strongly suggest that the toc1-1 mutation causes no change in the temperature entrainment pathway nor in the temperature compensation properties of the clock itself. Hence, it is unlikely that the short period phenotype of toc1-1 is due to altered temperature relations either between the environment and the clock (the input pathway) or in the interactions between components within the clock.

Circadian rhythms have been studied in a wide array of plant species, ranging from algae to higher vascular plants. The type of processes subject to clock control vary as widely as the species, including endogenous bioluminescence and phototaxis in some algae, and CO₂ assimilation, gene transcription, leaf movement and flowering in angiosperms (Sweeney, 1987). Through the use of an in vivo bioluminescence assay (Millar et al., 1992), a novel circadian phenotype was created which facilitated the isolation of circadian rhythm mutants in Arabidopsis. Characterization of the short period mutant, toc1-1 (Millar et al., 1995a), now allows both a molecular genetic and comparative approach to the analysis of the plant circadian system throughout development.

In etiolated tissue, the relative paucity of morphological and developmental events that are clearly clock-controlled and simple to assay has meant that until recently it has been difficult to determine whether a clock is running in dark-grown plants. However, frequent monitoring of the luciferase activity of a cab2::luc reporter in a high luminescence mutant background (supernova) has shown that synchronous, rhythmic expression of promoter activity occurs with a near 24 hour period in 5-day-old etiolated Arabidopsis seedlings (Hicks et al., 1996). Still unaddressed has been the question of whether the same clock is operating in dark-grown and light-grown tissues. Very early in tobacco seed germination, two overlapping and out-of-phase rhythms in CAB mRNA levels can be induced, one set by sowing time and phytochrome-independent, and the other set by a 3-hour RL treatment. After 60 hours, only one, RL-responsive oscillation is seen, suggesting that at least one oscillator is common to both phases of development (Kolar et al., 1995).

Analogously, in Arabidopsis, the results of the red pulse experiments with toc1-1 and WT (Fig. 2) now show that at least one component, TOC1, is common between the oscillator that regulates cab2::luc activity in etiolated cotyledons, and that which controls the same transgene in chlorophyllous mesophyll cells. Therefore, light-induced morphogenesis does not result in a switch to an entirely different timing system from that used in dark-grown seedlings. Although we suspect that the same clock is operating in light- and dark-grown plants, it is possible that only a subset of components are shared between two similar but different circadian clock systems, and that TOC1 is one of those elements.

Our results also show that fully developed chloroplasts are not required for the activity of the clock that controls CAB2 transcription. This eliminates the possibility that components of the photosynthetic apparatus or products of photosynthesis are directly required for the operation of a functional clock. However, a recent report demonstrating a circadian oscillation in chloroplastic free calcium suggests the possibility of a chloroplast-based clock that is separate from an oscillator in the cytoplasm (Johnson et al., 1995). The autonomy of this rhythm from the influence of the cytoplasm can now be determined in the toc1-1 background.

A diverse range of environmental and physiological factors controls stomatal aperture size (Assman, 1993; Webb and Hetherington, 1997), and the role of the circadian clock in timing the change in stomatal opening has been demonstrated in a variety of species (Gorton, 1990; Gorton et al., 1993; Hennessey et al., 1993; Webb, 1998). Since guard cells respond to a wide range of cues, cycles in [CO₂] or photosynthesis generated from within the leaf could secondarily drive a rhythm in stomatal conductance. In the few reports comparing conductance and photosynthesis rhythms in the same plant (Phaseolus vulgaris), both processes exhibited similar period lengths (Fredeen et al., 1991), even when internal [CO₂] was held constant (Hennessey et al., 1993). In Arabidopsis, we found a similar relationship between CAB2 transcription and stomatal conductance. In the wild type, the period lengths of the cab2::luc rhythm (24.7 hours; Millar et al., 1995a) and stomatal conductance rhythm (23.9 hours; Table 1, this study) were very close, and toc1-1 also exhibited rhythms in both processes that are proportionately shorter and essentially identical (20.7 hours; Millar et al., 1995a, and 20.8 hours; this study, respectively). These results are consistent with either mesophyll-to-guard cell signaling or with separate oscillators autonomously acting in each cell. However, the persistence of a circadian rhythm in guard cell aperture in Vicia epidermal peels (Gorton et al., 1989) demonstrates that autonomy from the rest of the blade is possible. Furthermore, in Phaseolus, temperature cycles entrain rhythms in stomatal conductance but not in photosynthesis, demonstrating that these processes can occur independently of each other (Fredeen et al., 1991). Although similar studies have yet to be performed in Arabidopsis, our results are consistent with the hypothesis that TOC1 acts cell autonomously in guard cells.

For both long day (LDP) and short day plants (SDP), there is evidence linking the circadian clock to the photoperiodic control of flowering time (Vince-Prue, 1994). In LDP like Arabidopsis, extension of daylength rather than interruption of the night is often the most effective way to induce flowering (Vince-Prue, 1994). In Lolium and Hordeum, however, flowering time is accelerated by far-red light (FRL) supplements added to a background of RL and the effectiveness of this treatment is under circadian clock control (Deitzer et al., 1979, 1982). Similarly, in Arabidopsis, there is a 24-hour rhythm in the acceleration of flowering by 6-hour FRL additions to a background of continuous WL (Deitzer, 1984). This finding, coupled to our results showing that toc1-1 affects a range of clock-controlled processes, clearly suggests that the dramatic acceleration of flowering time under SD in the Laer ecotype is the result of altered processing of daylength information by toc1-1, through its effect on the circadian clock. This also implies that the same oscillator mechanism controlling photoperiodism also controls rhythms in CAB2 transcription, leaf movement and stomatal conductance.

The difference in the effect of toc1-1 on flowering time in the C24 and Laer backgrounds points to significant differences in alleles at other loci in the two ecotypes that act to influence flowering through genetic interaction with TOC1. FRIGIDA (FRI) and FLC are two loci that have been shown to interact in an allele-specific way to strongly influence flowering time in Arabidopsis (Koornneef et al., 1994; Lee et al., 1994; Sanda et al., 1997). To test for potential interactions between these loci and TOC1, introgression of different FLC and FRI alleles into the toc1-1/Laer background is underway.

Because altering the level or activity of elements of the output pathway can change only the amplitude of a simple output, TOC1 must lie within either the input pathway or the central oscillator. Hence, our effort has been to try to distinguish a component of the input signaling pathway from a bona fide pacemaker element.

Light is a universally active stimulus in setting both the phase and the free-running period of the clock. Period length is inversely proportional to the log of the fluence rate (intensity), though the slope of this relationship depends on both the organism and the light quality. In many nocturnal species (e.g. mouse, hamster) period shortens with decreasing intensity, while period length in diurnal species (including many plants) tends to lengthen as light levels drop (Aschoff, 1960; Harris and Wilkins, 1976).

Within a species, the fluence rate response curve (FRC) can help elucidate some aspects of the light input pathway. In the bioluminescent dinoflagellate Gonyalaux polyedra, the slopes of the FRCs for blue and red light are of opposite sign, indicating that at least two separate photoreceptor systems are involved in setting period length (Roenneberg, 1996). In comparing the effect of different RL fluence rates on the period of toc1-1 and WT, the parallel slopes of the two traces clearly indicate that toc1-1 is constitutively acting to shorten period length over a wide fluence range (Fig. 4). Since signaling pathways involve amplification of the primary stimulus, a qualitative change in the activity of a component within the cascade will likely affect the final response by a multiplicative (or divisive) factor, rather than as a simple arithmetic addition or subtraction. Therefore, if TOC1 were a component of the light input pathway we would expect a change in the slope of the red light FRC, relative to wild type (Fig. 4). That this was not observed over a 500-fold range of light intensity is consistent with our contention that toc1-1 is a mutated clock component causing the central oscillator to run at a faster pace, independent of light input. However, it is still possible that TOC1 is a constant factor acting on the clock in parallel to the light input pathway, or that it acts as a constitutive, fluence-independent addition to light input just prior to conjunction with the oscillator itself.

Two poorly understood but characteristic features of circadian clocks are temperature compensation and temperature entrainment. The former, demonstrated here for the first time in Arabidopsis (Fig. 5), is due in part to the nature of molecular mechanisms within the central oscillator. This is clear from the study of mutants of the Drosophila period and Neurospora frequency genes, both oscillator components, in which certain alleles (per^L; frq⁹, respectively), have lost temperature compensation (Hall, 1997; Rutila et al., 1996). In contrast, a defect in fatty acid synthesis, the cel mutation, also disrupts temperature compensation in Neurospora (Lakin-Thomas et al., 1990). Such similar effects from such different mutations indicate that we cannot position TOC1 within the circadian system based solely on the absence of an effect of toc1-1 on this basic clock property.

Temperature cycles can act (like light) as an entraining stimulus to establish a stable phase relationship between peak output activity and the temperature transitions. A phase shift occurs at least once each cycle and the extent of the shift required for entrainment depends on how close the period (T) of the entraining cycle is to the endogenous period of the oscillator. The ‘sawtooth’ waveform displayed by the short-period (21 hour) toc1-1 mutant under a 24 hour T cycle demonstrates the presence of an oscillator that is being rephased each cycle differently from the wild-type (Fig. 6A). The return to a ‘wild-type’ (sinusoidal) waveform when toc1-1 is entrained to a 20 hour T cycle confirms this view (Fig. 6C). These results strongly suggest that toc1-1 does not directly affect the process of temperature entrainment, but rather the oscillator mechanism with which the entraining signal interacts.

The evidence presented here suggests that TOC1 is neither part of the light input nor the temperature input pathways. Although this leaves the central oscillator as the most likely site of TOC1 activity, there are alternative possibilities for how this oscillator fits within the entire clockwork of the circadian system. Certain aspects of phase resetting in Drosophila suggest a scheme in which a master oscillator drives a subordinate ‘slave’ oscillator (Pittendrigh and Bruce, 1959). Recent experiments with the clock-controlled gene, AtGRP7(=CCR2), raise this possibility in Arabidopsis. Out-of-phase cycling of AtGRP7 mRNA and protein abundance suggests that the protein may be involved in the regulation of its own message abundance. This notion is strengthened by the fact that normal cycling of AtGRP7 mRNA abundance is eliminated when the polypeptide, an RNA-binding protein, is overexpressed in transgenic plants (Heintzen et al., 1997). However, since the cycling of other transcripts are unaffected in this background and AtGRP7 cycles with a shorter period in toc1-1 (Kreps and Simon, 1997), these results are consistent with a model in which TOC1 is part of a master oscillator controlling one or more subordinate feedback loops, one of which includes AtGRP7 (Heintzen et al., 1997).

A second alternative to the single oscillator model considers that there are two or more independent, distinctly different clocks that share one or more components, one of which is TOC1. These may reside in different tissues or cell types and could be the reason for the 1.5-2 hour difference between the period of cab2::luc rhythm and leaf movement rhythm in Arabidopsis seedlings, and for the different periods observed for stomatal and leaf movement rhythms in bean (Hennessey and Field, 1992; Millar et al., 1995a; Millar and Kay, 1997). However, the possibility of inter-organ differences in the light input pathway, acting differently on the same clockwork, has not been addressed in either plant. Resolution of this issue will require identification of oscillator components and input coupling factors, and/or single cell resolution of clock-controlled outputs from differentiated tissue.

We have shown that a mutation in TOC1 disrupts the function of a circadian clock that controls the timing of transcriptional, physiological and developmental processes throughout the life of the plant. These results, together with a previous report (Millar et al., 1995a), constitute the most complete description, in a single species, of the pervasive effects of the circadian clock on higher plant development. They provide the best evidence to date that a single type of circadian oscillator regulates most, if not all, diurnal and photoperiodic events in the plant and that TOC1 is central to the proper function of this oscillator. Full confirmation of our conclusion awaits the cloning of TOC1 and manipulation of its expression.

We thank Carl A. Strayer for stimulating discussions and insights, and Dr Andrew J. Millar and members of the Kay lab for critical comments on the manuscript. We acknowledge support from NIH Grant GM56006 (S. A. K.) and a NSF Post-doctoral Fellowship in Plant Biology (BIR-9403981) to D. E. S. Funding was also provided by the BBSRC UK and the European Commission (A. A. R. W) and the NERC UK (M. P.). A. A. R. W. and M. P. are extremely grateful to Professor Terry Mansfield and Professor Alistair Hetherington for their encouragement and support.