*For correspondence (fax +44 1203 523 701; e-mail Andrew.Millar@warwick.ac.uk). Abbreviations: glabrous 1, gl1;timing of CAB expression 1, toc1;early flowering3, elf3;long hypocotyl 1, hy1;late elongated hypocotyl, lhy;circadian clock-associated 1, CCA1; constant light, LL; 12 h light 12 h dark cycle, LD (12,12); fast Fourier transform non-linear least squares, FFT-NLLS; relative amplitude error, RAE.
Many endogenous and environmental signals control seedling growth, including several phototransduction pathways. We demonstrate that the circadian clock controls the elongation of the Arabidopsis hypocotyl immediately upon germination. The pattern of hypocotyl elongation in constant light includes a daily growth arrest spanning subjective dawn and an interval of rapid growth at subjective dusk. Maximal hypocotyl growth coincides with the phase during which the cotyledons are raised, in the previously described rhythm of cotyledon movement. The rhythm of hypocotyl elongation was entrained by light–dark cycles applied to the imbibed seed and its period was shortened in thetoc1–1mutant, indicating that it is controlled by a similar circadian system to other rhythmic markers. The daily growth arrest is abolished by theearly flowering 3 (elf3) mutation, suggesting that this defect may cause its long-hypocotyl phenotype. Mutations that affect the circadian system can therefore cause gross morphological phenotypes, not because the wild-type gene functions pleiotropically in several signalling pathways, but rather because the circadian clock exerts widespread control over plant physiology.
The circadian system is an endogenous, biological timer that controls a wide range of rhythmic processes, all of which maintain rhythmic periods close to 24 h under constant environmental conditions ( Lumsden & Millar 1998;Sweeney 1987). Circadian rhythms in Arabidopsis thaliana include rhythmic leaf movements (nyctinasty;Engelmann et al. 1992 ), the rhythmic opening of stomata ( Somers et al. 1998 ;Webb 1998), and the transcription of a number of genes, including chlorophyll a/b-binding protein (CAB or LHC) genes (reviewed in Fejes & Nagy 1998). Light signals interact with the circadian system, such that light controls the phase and period of the rhythms in all eukaryotes and frequently affects the amplitude of circadian rhythms in higher plants ( Johnson et al. 1998 ). Many of the targets of circadian regulation in plants are also controlled by light. Stomatal opening and CAB gene expression are both rhythmic and light-regulated, for example, and the amplitude of their light regulation is modulated (gated) by the circadian clock ( Gorton et al. 1993 ;Millar & Kay 1996).
The mechanism of the circadian oscillator in other species is thought to depend on a 24 h molecular cycle in the activity of a small number of gene products (reviewed in Dunlap 1996;Young 1998). These ‘clock genes’ were identified by mutations that altered the circadian period or abolished circadian rhythms. Similar mutants in Arabidopsis include timing of CAB expression (toc1), which has a 21 h period for CAB transcription and other rhythmic markers compared to the wild-type period of 24.5 h ( Millar et al. 1995a ;Somers et al. 1998 ), and lines that overexpress transcription factors, late elongated hypocotyl (lhy;Schaffer et al. 1998 ) and circadian clockassociated 1 (CCA1;Wang & Tobin 1998). Plants that overexpress LHY and CCA1 are arhythmic under conditions of constant light and constant darkness. The early flowering 3 (elf3) mutation, in contrast, causes arhythmia only during and immediately after light exposure, suggesting that ELF3 is involved in an interaction between phototransduction and the circadian system ( Anderson et al. 1997 ;Hicks et al. 1996 ;Millar 1998). The photoperiodic control of flowering is absent in elf3 mutants, LHY- and CCA1-overexpressing plants ( Schaffer et al. 1998 ;Wang & Tobin 1998;Zagotta et al. 1992 ;Zagotta et al. 1996 ); this is expected, because the circadian system is thought to be required for the measurement of day length ( Carré 1998;Coupland 1998;Thomas & Vince-Prue 1996). These lines also have an elongated hypocotyl (hy) phenotype under some lighting conditions ( Schaffer et al. 1998 ;Wang & Tobin 1998;Zagotta et al. 1992 ;Zagotta et al. 1996 ). It was unclear whether the hy phenotype was directly related to the circadian defects of these mutants. hy phenotypes are more typically associated with mutants that are defective in phototransduction pathways, such as hy1, which lacks the phytochrome chromophore ( Parks & Quail 1991).
We have used an automated video imaging system to show that hypocotyl elongation in Arabidopsis is controlled by the circadian clock from the very earliest stages of seedling development. The circadian rhythm of hypocotyl elongation was affected by previously described clock mutations, suggesting that the hy phenotype of some clock mutants is a specific consequence of aberrant circadian regulation, rather than a pleiotropic effect of the mutations on other signalling pathways.
Video imaging of Arabidopsis seedling growth
Seed of Arabidopsis thaliana (Columbia gl1) were germinated for two days in sterile culture under 12 h light, 12 h dark cycles [LD (12,12)]. The seeds were then transferred to constant, dim white light (LL); root radicals began to emerge from the testa at approximately this time. The growth of the seedlings was recorded for several days by the Kujata digital imaging system ( Fig. 1a). Simple image processing allowed us to trace the position of the cotyledon tips and the lowest pigmented part of the seedling, at the top of the hypocotyl ( Fig. 1a). The distance between the latter marker and the agar surface represents the length of the hypocotyl.
The hypocotyls of wild-type seedlings grow for several days under these lighting conditions. Their elongation is not uniform, however: growth slows or stops for 2–6 h intervals, approximately once every 24 h close to predicted dawn ( Fig. 1b). Ultradian (short-period) rhythms of circumnutation (the corkscrew pattern of elongation) have previously been described in Arabidopsis hypocotyls ( Schuster & Engelmann 1997). This motion is largely in the horizontal plane; it was occasionally detectable in our images (small changes in cotyledon angles, around 36 h in Fig. 1a), but rarely in the vertical elongation data.
The positions of the cotyledon tips rise progressively as the seedling hypocotyls elongate, and their positions also rise and fall rhythmically with a period of approximately 24 h ( Fig. 1b). The positions of the cotyledon tips reflect the sum of the rhythms of cotyledon movement and hypocotyl elongation; cotyledon movements alone are revealed by correcting each data point for hypocotyl elongation. Elongation of the cotyledons and petioles gives the rhythm of cotyledon movement a gradual, rising trend ( Fig. 1b). The location of the cotyledons follows a circadian rhythm, as described previously; the cotyledon tips are highest (cotyledon angle closest to vertical) during the subjective night ( Engelmann et al. 1992 ;Millar et al. 1995a ;Somers et al. 1998 ). This rhythm is generally preferable as a circadian marker because it is more robust than the rhythm of hypocotyl elongation (see below).
Light-entrained, circadian regulation of hypocotyl elongation
The timing of hypocotyl elongation coincides with the cotyledons rising, suggesting that a circadian clock also controls hypocotyl elongation. The phase of maximal elongation occurs approximately at subjective dusk, with a growth arrest at subjective dawn ( Fig. 1b). The restricted phase of elongation led to an apparent circadian gating of seedling emergence because distinct sets of seedlings commenced elongation on 2 successive days ( Fig. 2). It is possible that earlier stages of germination are also gated by the circadian clock, but these could not be reliably scored in our images.
We reversed the two initial LD (12,12) cycles to DL (12,12) cycles in order to test whether the rhythm of hypocotyl elongation could be entrained by the light:dark signals, as expected of a circadian rhythm. The seedlings from both conditions were assayed in adjacent wells, as described above. The seedlings commenced elongation on either of 2 successive days. On each day, the seedlings from LD cycles elongated at the opposite phase compared to those from DL cycles. The subsequent rhythms of elongation remained in antiphase for the duration of the experiment ( Fig. 2). The circadian system that controls this rhythm can therefore be stably entrained by light–dark signals received by the imbibed seed during germination.
Dependence of the elongation rhythm on TOC1
The toc1–1 mutation shortens the period of CAB gene expression to approximately 21 h in constant light ( Millar et al. 1995a ). We measured hypocotyl elongation in toc1–1 and its cognate parent, a transgenic cab2::luc line in ecotype C24. Hypocotyl elongation was rhythmic in both lines, but the toc1–1 mutant had a shortened circadian period ( Fig. 3a). We differentiated the raw pixel data to obtain elongation rates which were better suited to biomathematical analysis ( Fig. 3b). The standard fast Fourier transform non-linear least squares method (FFT-NLLS;Plautz et al. 1997 ) was applied in order to estimate the circadian period in these lines ( Table 1). The toc1–1 mutation shortens the period of the hypocotyl elongation rhythm to approximately 21 h, as it does for cab2::luc expression and other rhythmic markers ( Millar et al. 1995a ;Somers et al. 1998 ). The circadian system controlling hypocotyl elongation therefore has a genetically related, or possibly identical, mechanism to that controlling CAB gene expression.
Table 1. Circadian period estimates and the proportion of arhythmic seedlings
Period estimates for both hypocotyl extension and cotyledon movement were derived by the FFT NLLS program, as variance-weighted means (Period) and standard deviations (SD). The number of plants that yielded a circadian period estimate (a period in the range 15 to 35 h) is listed, together with the total number of periods (5–60 h) returned by the FFT-NLLS analysis and the percentage of plants without a circadian period estimate, as an indication of the arhythmicity of an ecotype or mutant. The total number of data traces analysed and the number of independent experiments are also shown. ND, not determined.
Arhythmia in elf3
Many of the phenotypes of the elf3 mutant and LHY- and CCA1-overexpressing plants have been directly related to primary defects in the circadian system ( Schaffer et al. 1998 ;Wang & Tobin 1998;Zagotta et al. 1992 ;Zagotta et al. 1996 ), with the notable exception of their long-hypocotyl trait. Our data show that the circadian clock regulates hypocotyl elongation in Arabidopsis, so a primary defect in the circadian system could also affect hypocotyl length. In particular, a mutation that abolished the daily growth arrest might be expected to have a longer hypocotyl overall. We tested the rhythm of hypocotyl elongation in elf3 and found that elongation was uniform, with no detectable rhythmic arrests ( Fig. 4a).
Biomathematical analysis by FFT-NLLS detects tightly clustered periods with low relative amplitude error (RAE) in data from wild-type seedlings, which is indicative of clear circadian rhythms ( Fig. 4b). The more robust rhythmicity of the cotyledon tip data ( Fig. 4b, lower left) yields a closer clustering of periods and lower RAE compared to hypocotyl elongation data (upper left). FFT-NLLS can detect multiple rhythmic components in each data trace. Wild-type data yield some period estimates outside the circadian range (15–35 h) in addition to the principal, circadian period, in part due to short-period noise and long-period trends in the data. The data for elf3, in contrast, yield few periods in the circadian range and many short-period components, in the noisier, hypocotyl elongation data ( Fig. 4b and Table 1). Only 48% of elf3 plants scored give any period for hypocotyl elongation in the circadian range, compared to 96% of wild-types; cotyledon movement similarly lacked robust circadian rhythms in elf3. The long hypocotyl of elf3–1 is therefore correlated with the absence of the rhythmic growth arrest.
Rhythmic elongation in the long-hypocotyl mutant, hy1
The hy1 mutant is deficient in phytochrome chromophore biosynthesis, and the resulting deficiency in phototransduction leads to a long-hypocotyl phenotype ( Parks & Quail 1991). Hypocotyl elongation was clearly rhythmic both in hy1 and in its cognate wild-type La(er) ( Fig. 5), indicating that the uniform elongation in elf3 is not a non-specific attribute of all long-hypocotyl mutants. The period of the elongation rhythm in hy1 was not significantly different from La(er) (M.J. Doswon-Day and A.J. Millar, unpublished results), as predicted from the wild-type period of CAB gene expression in hy1 under white light ( Millar et al. 1995b ).
Many endogenous and environmental signals regulate hypocotyl elongation; we have shown that the elongation of Arabidopsis hypocotyls is controlled by the circadian clock ( Fig. 1a). The circadian rhythm of hypocotyl elongation was entrained by light–dark cycles applied to the imbibed seed ( Fig. 2) and the toc1–1 mutation shortened the rhythmic period ( Fig. 3a and Table 1). The long-hypocotyl mutant hy1 retained rhythmic elongation ( Fig. 5); the elf3 mutation, which also results in a long-hypocotyl phenotype, abolished the elongation rhythm ( Fig. 4a).
Rhythmic growth and circadian timing
Germinating Arabidopsis seedlings exhibit circadian rhythms in hypocotyl elongation and also in cotyledon position, from the earliest stages of growth ( Fig. 1a). The hypocotyl elongates in distinct steps in seedlings of strains Columbia gl1, C24 and La(er) ( Figs 1b, 2, 3a, 4a and 5). Maximal elongation rates occur at subjective dusk, coinciding with the raising of the cotyledons ( Fig. 1). The hypocotyl growth arrest occurs as the cotyledons drop to the horizontal, close to subjective dawn ( Fig. 1b). A similar circadian rhythm of elongation had been suggested during studies of circumnutation ( Schuster & Engelmann 1997), but the imaging method employed did not clearly distinguish rhythmic hypocotyl growth from cotyledon movements, and no phase information could be deduced.
Several characteristics of the hypocotyl elongation rhythm indicated that it was a typical circadian rhythm. First, the rhythm persisted under constant environmental conditions, with a period close to 24 h in wild-type seedlings ( Table 1). Second, the rhythm was entrained by the two LD (12,12) cycles that were applied to the germinating seed, before imaging under constant light ( Fig. 2). Third, the period of the hypocotyl rhythm was shortened to approximately 21 h ( Fig. 3 and Table 1) by the toc1–1 mutation. toc1–1 is the best-characterised Arabidopsis mutant to be identified by its effect on circadian rhythms; circadian rhythms of gene expression and stomatal opening also have a 21 h period in the toc1–1 mutant ( Kreps & Kay 1997;Millar et al. 1995a ;Somers et al. 1998 ). The elongation rhythms were neither driven nor entrained by artefactual stimuli in our imaging system because seedlings that were assayed simultaneously exhibited elongation maxima at different phases ( Fig. 2) or with different periods ( Fig. 3). Hypocotyl elongation is therefore controlled by an endogenous, circadian timing system that shares at least one component (TOC1) with the system controlling CAB expression.
The seedling rhythms we describe occur strikingly early in development, before any other known circadian rhythm in Arabidopsis. The rhythms can be entrained by light signals immediately upon germination and before the cotyledons emerge from the seed coat, which they normally do on the third day. Circadian-regulated CAB gene expression has also been shown to entrain to a light signal at this stage of development in dark-grown tobacco seedlings, although the resulting rhythm was assayed several days later ( Kolar et al. 1998 ). The Arabidopsis circadian system therefore controls development throughout the life of the plant, from the earliest stage of seedling growth through to the elongation of the inflorescence stem ( Agosti et al. 1997 ).
The mechanism of circadian regulation
Circadian control may be mediated either by a direct output from the circadian clock to the immediate effectors of hypocotyl elongation, or indirectly by circadian ‘gating’ of another signalling pathway, which in turn controls the hypocotyl. Conceptually, signal transduction in the gated pathway proceeds at the phase when the gate is open, but is prevented at phases when the gate is closed. Circadian gating of light responses occurs together with a ‘basal’ circadian rhythm for both CAB gene expression and stomatal opening ( Gorton et al. 1993 ;Millar & Kay 1996). Circadian gating has previously been shown to control hypocotyl elongation and hypocotyl unhooking responses to phytochromes ( Horwitz & Epel 1978;King et al. 1982 ;Wildermann et al. 1978 ). No ‘basal’ circadian rhythms were evident in these studies but, under continuous light, such circadian gating could generate a circadian rhythm ( Kay & Millar 1993): slow elongation would result when the open gate allowed phototransduction to inhibit hypocotyl elongation (near subjective dawn) and rapid elongation when the gate was closed (near subjective dusk). The phase of inhibited hypocotyl elongation (close to subjective dawn) overlaps with the phase at which CAB gene expression becomes responsive to light activation ( Millar & Kay 1996), suggesting that shared components may participate in the circadian gating of both these light responses. The mechanism of the circadian gating interaction is not yet known, nor is it clear how widespread is the circadian modulation of other signalling pathways. The circadian or diurnal rhythms in the abundance of hormones known to control hypocotyl elongation suggest that circadian regulation may function at several levels ( Finlayson et al. 1998 ;Foster & Morgan 1995). The identification of mutants with defective circadian gating may be necessary in order to address these questions.
Circadian-related phenotypes in the elf3 mutant
toc1–1 and most other toc mutants were identified by a period defect and have no gross morphological phenotypes ( Millar et al. 1995a ;Somers et al. 1998 ). The photoperiod-insensitive mutant elf3, in contrast, has a long-hypocotyl phenotype that is most pronounced under short photoperiods and is detectable in continuous red or blue light ( Zagotta et al. 1992 ; 1996). elf3 mutants were arhythmic in the light but not in darkness, suggesting that the elf3 mutation affected an interaction between phototransduction and the circadian system ( Anderson et al. 1997 ;Hicks et al. 1996 ). It was unclear whether ELF3 also affected additional pathway(s) that controlled hypocotyl elongation independently of the circadian clock, or whether arhythmia and the long hypocotyl were due to the same defect.
We reasoned that the long-hypocotyl phenotype was likely to be secondary to a lesion in the circadian system because our data showed that the circadian clock regulates hypocotyl elongation in the wild type. Specifically, any mutant that lacked the circadian growth arrest should have a longer hypocotyl than the wild type. The daily growth arrests were completely absent in elf3 seedlings ( Fig. 4), and biomathematical analysis did not show significant circadian rhythmicity in elf3, either for hypocotyl elongation or for the position of the cotyledon tips ( Table 1). Arhythmia is not an artefact of rapid hypocotyl elongation because the phytochrome-deficient hy1 mutants grew longer hypocotyls yet remained rhythmic ( Fig. 5). The long-hypocotyl phenotype of elf3 is therefore probably a secondary consequence of a defect in the circadian system.
The overexpression of the transcription factors CCA1 and LHY severely affects the circadian system in Arabidopsis and leads to arhythmia in light and darkness ( Schaffer et al. 1998 ;Wang & Tobin 1998). The link between circadian defects and the long-hypocotyl trait is supported by the fact that plants overexpressing CCA1 and LHY exhibit a hy phenotype similar to elf3 ( Schaffer et al. 1998 ;Wang & Tobin 1998), and similarly lack the circadian arrests of hypocotyl elongation (M.J. Dawson-Day and A.J. Millar, unpublished results). The elf3 mutation may disrupt the circadian system by a mechanism different from the transcription factor overexpression because elf3 is arhythmic only after light exposure, not in darkness ( Anderson et al. 1997 ;Hicks et al. 1996 ). The identities of the photoreceptor species involved in the elf3 phenotypes remain to be determined. Irrespective of the mechanism of arhythmia in elf3, our results demonstrate that the absence of circadian control in the plant regulatory network can cause gross, visible phenotypes, as well as the more subtle alterations in biological rhythms. Physiological studies have previously related circadian dysfunction to chlorosis (for example, Highkin & Hanson 1954): the pale-green phenotype of elf3 and of other clock mutants may be due to the loss of circadian control over chloroplast functions in addition to abnormal CAB gene regulation. Further analysis of clock mutants will be necessary to evaluate the full impact of circadian timing on plant physiology, and to distinguish the effects of circadian gating and basal rhythmic mechanisms.
We obtained Arabidopsis thaliana ecotype Landsberg erecta, La(er), and the mutant hy1–1 from Maarten Koornneef (Wageningen), Columbia gl1 from Lehle seeds (Arizona) and the mutant elf3–1 in the gl1 background from Ry Meeks-Wagner (Oregon). The toc1–1 mutant and its parent, the transgenic cab2::Ω::luc line in the C24 ecotype (referred to as C24), have been described previously ( Millar et al. 1992 ; 1995a).
Seed was surface-sterilised and sown in rows on 1.0% agar containing Murashige and Skoog plant salt mixture (MS, Murashige & Skoog 1962) and 3% sucrose. The seeds were incubated in a growth chamber under LD(12,12) of 88 μmol m–2 s–1 cool white fluorescent light, at 21.5°C ± 0.5°C. After two dark intervals, two 1.5 cm2 agar blocks carrying up to eight seeds were transferred to a 25-well square tissue culture dish, placed vertically before a monochrome video camera. The seedlings were imaged in a growth room under LL of 10–20 μmol m–2 s–1 cool white fluorescent light at 21–22°C.
The Kujata is a simple PC-based video imaging system with software control of up to 14 cameras (Mark Hyett, Virginia). This system stored inverted, thresholded, 1-bit images of up to eight seedlings per camera every 20 min and 8-bit greyscale images every 2 h ( Fig. 1a). The image series were analysed post hoc using rectangular regions of interest to follow the growth of individual seedlings. Seedlings that fell over or became entangled were not analysed.
The movement of one cotyledon per seedling was recorded, being reflected by the highest white pixel value within a region of interest, over the time course ( Fig. 1). A window of between 60 and 84 h was selected for each data trace, spanning the period of most rapid hypocotyl elongation. The trace within this window was then analysed by the fast Fourier transform non-linear least squares program (FFT-NLLS;Plautz et al. 1997 ). This fits sine waves to the data and estimates their period, phase and amplitude, with associated confidence intervals. Relative amplitude error (RAE = amplitude error/amplitude) is a measure of rhythmic robustness that varies between 0 (perfect sine wave) and 1 (rhythm not significant). One to three sine waves with different periods were estimated for each data trace. For estimation of the mean circadian period, results were restricted to periods falling within the circadian range, 15–35 h, and the number of plants to which the FFT program failed to fit a period in this range was also noted. Variance-weighted means and standard deviations were calculated ( Millar et al. 1995a ) because these more strongly reflect the clearest traces. In addition, in order to portray the arhythmicity of an ecotype or mutant ( Fig. 4b), all periods within the range 5–60 h were plotted against the RAE for each estimate ( Plautz et al. 1997 ).
When the testa was omitted from the region of interest, the lowest white pixel reflected the position of the hypocotyl apex and its vertical movement over the time course. Absolute hypocotyl lengths ( Fig. 1) could be calculated from the difference in pixel positions of the testa (on the agar surface) and the hypocotyl apex; lengths in each camera were calibrated by imaging a ruler. Data presented graphically are from a single seedling of each genotype, representative of independent, replicate experiments (see Table 1). The pattern of hypocotyl extension approximates less closely to a sine wave than that for cotyledon movement ( Fig. 1b), so prior to period analysis, these data were differentiated by linear regression in a running, 11-point window, centred on the time of interest ( Fig. 3b). A 60 h window, which omitted the first growth spurt on plumule emergence, was analysed as described for cotyledon movement.
We are grateful to members of the chronobiology group at Warwick for useful discussions and to Mark Hyett for Kujata software modifications. This work was supported in part by a BBSRC grant to A.J.M. The Kujata system was funded by a Royal Society grant to I. Carré. M.J.D.-D. is a Department of Biological Sciences Research Fellow.