CCA1 and ELF3 Interact in the Control of Hypocotyl Length and Flowering Time in Arabidopsis

The circadian clock is an endogenous oscillator with a period of ~24 hours that allows organisms to anticipate, and respond to, changes in the environment. In Arabidopsis ( Arabidopsis thaliana ), the circadian clock regulates a wide variety of physiological processes including hypocotyl elongation and flowering time. CIRCADIAN CLOCK ASSOCIATED 1 (CCA1) is a central clock component, and CCA1 overexpression causes circadian dysfunction, elongated hypocotyls and late flowering. EARLY FLOWERING 3 (ELF3) modulates light input to the clock and is also postulated to be part of the clock mechanism. elf3 mutations cause light-dependent arrhythmicity, elongated hypocotyls and early flowering. Although both genes affect similar processes, their relationship is not clear. Here we show that CCA1 represses ELF3 by associating with its promoter, completing a CCA1-ELF3 negative feedback loop that places ELF3 within the oscillator. We also show that ELF3 acts downstream of CCA1, mediating the repression of PHYTOCHROME-INTERACTING FACTOR s 4 and 5 in the control of hypocotyl elongation. In the regulation of flowering, our findings show that ELF3 and CCA1 either cooperate or act in parallel through the CONSTANS/FLOWERING LOCUS T ( FT ) pathway. In addition, we show that CCA1 represses GIGANTEA and SUPPRESSOR OF CONSTANS 1 by direct interaction with their promoters, revealing additional connections between the circadian clock and the flowering pathways.


ELF3 and CCA1 Form a Negative Feedback Loop
To determine how ELF3 affects CCA1 expression, CCA1 mRNA levels were examined over a time course in ELF3-OX and elf3-1 plants by qRT-PCR (Fig. 1A).
Constitutive overexpression of ELF3 resulted in slightly elevated peak levels of CCA1 mRNA, whereas the peak levels of CCA1 mRNA were reduced in elf3-1 mutant plants.
This result is consistent with previous reports that ELF3 represses PRR9 and that PRR9 represses CCA1 expression (Nakamichi et al., 2010;Dixon et al., 2011). Therefore, ELF3 positively regulates the expression of CCA1. To determine whether CCA1 forms a feedback loop with ELF3, we examined ELF3 expression in CCA1-OX and cca1-1 plants ( Fig. 1B). In CCA1-OX plants, ELF3 was repressed and its circadian rhythms abolished.
This result is consistent with previous reports showing that overexpression of CCA1 results in arrhythmicity (Wang and Tobin, 1998). In cca1-1, ELF3 expression was slightly elevated, but still exhibited circadian rhythms with evening peaks of mRNA; maintenance of circadian rhythms in cca1-1 plants has been reported previously (Green and Tobin, 1999). Taken together, these results show that CCA1 represses the expression of ELF3. Therefore, ELF3 and CCA1 form a negative feedback loop through PRR9.

CCA1 Represses ELF3 by Interacting with its Promoter
The ELF3 gene contains a CCA1-binding site (CBS) between -248 and -241 relative to its transcription start site ( Fig. 2A) (Wang et al., 1997). To examine whether the effect of CCA1 on ELF3 expression was by direct association with its promoter, chromatin immunoprecipitation (ChIP) was performed using anti-CCA1 antibody. Figure   2B demonstrates that an anti-CCA1 antibody efficiently immunoprecipitated a fragment of ELF3 that contains the CBS but not another one located in the downstream coding region, suggesting that CCA1 binds directly to the promoter of ELF3. If this association results in repression, a pulse of CCA1 using an ethanol-inducible system (Knowles et al., 2008) should result in a reduction of ELF3 mRNA shortly after the induction of CCA1 protein. Figure 2D shows that a pulse of CCA1 causes a reduction in ELF3 expression within one hour (Fig. 2C) , suggesting that CCA1 binds directly to the ELF3 promoter to repress its expression. In WT Arabidopsis, CCA1 association with the ELF3 promoter was only observed at two hours after dawn (ZT2) when CCA1 protein is most abundant, but not at two hours after dusk (ZT14) when CCA1 protein is least abundant (Fig. 2B).
These results are consistent with CCA1 repression of ELF3 during the day and rising of ELF3 expression in the evening when CCA1 is absent.

Interactions between CCA1 and ELF3 in the Control of Hypocotyl Length
To determine whether ELF3 and CCA1 interact genetically in the regulation of hypocotyl length, we compared hypocotyl lengths of elf3-1, CCA1-OX and elf3-1 CCA1-OX plants. Under both LD and SD conditions, elf3-1 and CCA1-OX have elongated hypocotyls (Fig. 3) as previously reported (Zagotta et al., 1996;Kim et al., 2005). The hypocotyl length of elf3-1 CCA1-OX plants was more similar to the hypocotyl length of CCA1-OX than the elf3-1 mutant, which has shorter hypocotyls than CCA1-OX seedlings ( Fig. 3), suggesting that additional factors downstream of CCA1 parallel to ELF3 could 1 0 be involved in hypocotyl elongation. It is known that the circadian clock can affect hypocotyl elongation through its regulation of PIF4 and PIF5 (Nozue et al., 2007;Nusinow et al., 2011). To determine the mechanism that gives rise to the observed long hypocotyls in elf3-1, CCA1-OX and elf3-1 CCA1-OX plants, we examined the expression of PIF4 and PIF5. In both LD and SD, PIF4 and PIF5 expression was higher in elf3-1, CCA1-OX and elf3-1 CCA1-OX plants than in WT in the dark (Fig. 4, A-D). However, there were no appreciable differences in PIF4 and PIF5 expression when elf3-1, CCA1-OX and elf3-1 CCA1-OX plants were compared, suggesting that other factors contribute to their differing hypocotyl lengths.

Interactions between CCA1 and ELF3 in the Control of Flowering Time
We determined whether ELF3 and CCA1 interact genetically to control photoperiodic flowering by measuring flowering time and examining the expression of the flowering genes FT, SOC1, FLC, GI, and CO. Under LD growth conditions, elf3-1 CCA1-OX plants flowered before CCA1-OX plants, but later than elf3-1 and WT plants (Fig. 5,, suggesting that ELF3 acts mainly upstream of CCA1, but can also act coordinately or in parallel with CCA1. There were no significant differences in FLC expression levels among WT, elf3-1, CCA1-OX, and elf3-1 CCA1-OX plants (Fig. 6C), suggesting that neither CCA1 nor ELF3 are involved in the autonomous and vernalization pathways. In the elf3-1 plants, FT, GI and CO were derepressed compared with WT ( Fig.   6, A, D, and E) as previously reported (Kim et al., 2005). CCA1-OX plants exhibited repressed FT, SOC1, GI, and CO expression (Fig. 6, A, B, D, and E), which is in 1 2 CCA1-OX and cca1-1 lines. Figures 8B and 9B show that at ZT2, the fragments of GI and SOC1 containing CBSs were significantly enriched in WT and CCA1-OX plants, but not in cca1-1 plants, suggesting that CCA1 binds to the GI and SOC1 promoters in Arabidopsis. In addition, an ethanol-induced pulse of CCA1 was able to reduce GI and SOC1 expression within one hour (Figs. 8D and 9D), whereas ethanol treatment of a control line produced no changes in their expression (Figs. 8C and 9C). These results establish that repression of GI and SOC1 is due to the direct association of CCA1 with their promoters.

DISCUSSION
It has been shown that ELF3 associates with the promoter of PRR9 to repress its expression and that PRR9 represses CCA1 by interaction with its promoter (Nakamichi et al., 2010;Dixon et al., 2011). Our data that CCA1 represses ELF3 expression by association with its promoter (Figs. 1 and 2) close the negative feedback loop composed of ELF3, PRR9 and CCA1 (Fig. 10). These results support the idea that ELF3 is a component of the oscillator itself (Thines and Harmon, 2010;Dixon et al., 2011) in addition to its established role as a modulator of light input to the clock (Hicks et al., shown that CCA1 represses GI and SOC1 by association with their promoters (Figs. 8 and 9), revealing additional connections between CCA1 and the flowering pathways.
Hypocotyl elongation is controlled by many factors, including the convergence of light and clock signaling (Nozue et al., 2007). A recent report has shown that ELF3 forms a complex with ELF4 and LUX to directly repress PIF4 and PIF5 expression and an elf3 pif4 pif5 triple mutant exhibits hypocotyls of similar length to WT plants (Nusinow et al., 2011). In this study, we found that ELF3 expression is repressed directly by CCA1, and thus ELF3 acts downstream of CCA1 to repress PIF4 and PIF5. However, PIF4 and PIF5 RNA levels were elevated to similar levels in elf3-1, elf3-1 CCA1-OX, and CCA1-OX plants while elf3-1 showed a weaker hypocotyl phenotype than CCA1-OX and elf3-1 CCA1-OX plants. This is consistent with CCA1 affecting additional factors besides ELF3, PIF4 and PIF5 in controlling hypocotyl elongation (Hazen et al., 2005;Li et al., 2011).
The circadian clock regulates flowering through the photoperiodic pathway that includes CO and FT. Both ELF3 and CCA1 act as negative regulators upstream of CO and FT, but their relationship in controlling the photoperiodic pathway is unclear. Under both LD and SD conditions, elf3-1 CCA1-OX flowers before CCA1-OX plants, but after elf3-1 plants (Fig. 5), suggesting that ELF3 and CCA1 do not act in a linear pathway. In LD, expression of CO and FT were high in the elf3-1 mutant, but low in elf3-1 CCA1-OX and CCA1-OX plants (Fig.6, A and D), suggesting that ELF3 could act upstream of CCA1 in the control of CO and FT. In SD, CO and FT levels are high in elf3-1, but low in CCA1-OX plants (Fig. 7, A and D). In elf3-1 CCA1-OX plants, high levels of FT message were observed, but these levels were not in accordance with the low level of CO expression (Fig. 7,A and D). This result is in agreement with a previous report showing that ELF3 1 4 can act independently of CO to control flowering time (Kim et al., 2005). It has been shown that ELF3 represses GI transcription independently of CCA1, possibly through its repression of PRR9 (Dixon et al., 2011), and ELF3 protein is involved in facilitating the interaction between CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) and GI. The COP1-GI interaction destabilizes the GI protein to ensure proper GI function in the control of photoperiodic flowering (Yu et al., 2008). In this study, we showed that CCA1 could bind directly to the GI promoter to repress its expression (Fig. 8), adding another link within the clock, and between the clock and the photoperiodic flowering pathway. In accordance with these data, gi-3 cca1-1 lhy-11 triple mutants exhibit WT flowering time, FT activation is probably more prominent in LD. We have presented a model in which ELF3 and CCA1 independently regulate the expression of GI, and GI regulates the expression of FT through both CO-dependent and -independent pathways (Fig.10).
The MADS box transcription factor SOC1 is a key floral activator integrating multiple flowering pathways. The expression of SOC1 is positively regulated by FT (Yoo et al., 2005) and negatively regulated by FLC, which forms a floral repressor complex with other proteins and directly binds to the promoter of SOC1 (Helliwell et al., 2006). In this study, we found that elf3-1 CCA1-OX plants exhibit low levels of SOC1 message, similar to CCA1-OX plants ( Fig. 6B and 7B), indicating that CCA1 acts downstream of ELF3 to negatively regulate the expression of SOC1. This result is in accordance with a previous report showing that SOC1 expression is higher in cca1-1 lhy-11 double mutants (Fujiwara et al., 2005). Furthermore, we demonstrated that CCA1 significantly represses SOC1 expression through direct interaction with its promoter (Fig. 9), illustrating a novel way that the circadian clock controls flowering, through direct regulation of a floral integrator (Fig. 10).
The examination of flowering gene expression revealed that neither ELF3 nor CCA1, participate in the regulation of FLC ( Fig. 6C and 7C). However, the elf3-8 mutant exhibits reduced FLC expression in a photoperiod-independent manner, suggesting that ELF3 can affect FLC expression (Yu et al., 2008). In elf3-8, there is a base pair change at the exon 4 splice acceptor site, resulting in a truncated protein that includes 28 amino acids from a different frame before a premature stop codon (Hicks et al., 2001). In this study, we utilized the elf3-1 mutant which produces a truncated protein resulting from a single base pair change that gives rise to an early stop codon (Hicks et al., 1996). A detailed comparison between these two mutants has not been carried out, but they do 1 6 appear to flower at the same time under LD and SD (Liu et al., 2001;Kim et al., 2005;Yu et al., 2008;Nefissi et al., 2011). elf3-7 is a weak allele which contains a single base change at the exon 1 splice donor site (Hicks et al., 2001). Through the use of cryptic splice sites, the elf3-7 mutant can give rise to a combination of truncated ELF3 proteins, or a version that is missing eight amino acids (Hicks et al., 2001). A recent report showed that elf3-7 has less severe hypocotyl and flowering phenotypes than elf3-1, but it has the same effect on CCA1 and LHY expression (Kolmos et al., 2011), showing that it is difficult to predict how a given mutant allele of ELF3 will behave. Although elf3-1 and elf3-8 mutants have similar flowering times, how they act on the various flowering genes, including FLC, may be different because a complete understanding of the functions of ELF3 protein domains is lacking.
In summary, our study of the interplay between the central clock component CCA1 and ELF3 has completed the CCA1-ELF3 negative feedback loop that places ELF3 within the oscillator. We have shown that CCA1 represses GI and SOC1 expression by direct association with their promoters, not only highlighting the multiple roles of CCA1 in the regulation of flowering time, but also revealing additional connections between the circadian clock and the flowering pathway.

Measurement of Hypocotyl Length and Flowering Time
Arabidopsis plants were grown on soil under either LD (16 h light/8 h dark) or SD (8 h light/16 h dark) conditions with 100 µmol m -2 s -1 white fluorescent light. Hypocotyl length was measured from days five to seven. Flowering time was scored by counting the number of days to, and number of rosette leaves at, flowering.

RNA Extraction and qRT-PCR
1-2 week-old seedlings were grown on Murashige and Skoog medium (Murashige and Skoog, 1962) with 1.5% agar. For the circadian experiments, samples were collected every 4 h either during the light/dark cycle or in continuous white light. Total RNA 1 8 extraction and qRT-PCR were carried out as previously described (Lu et al., 2011).
Actin2 was used as a non-cycling reference, and the expression levels were normalized to the control. The primers used for amplification are listed in Supplemental Table S1.

ChIP
ChIP was performed as previously described (Lu et al., 2011) using affinitypurified anti-CCA1 antibody (Wang and Tobin, 1998) for immunoprecipitation. The primers used for amplification are listed in Supplemental Table S1.

Ethanol Pulse
EtOH treatment was performed as previously described (Knowles et al., 2008).
Control plants contained the regulator construct 35S::AlcR:T NOS . 10 day-old seedlings were treated with 10 min of EtOH vapor of 1% (v/v) EtOH. Samples were collected at the time of induction (0 h), 1 h, 2 h, and 4 h after EtOH treatment.

Accession Numbers
Sequence data from this article can be found in the Arabidopsis Genome Initiative data library using the following accession numbers:

Supplemental Data
Supplemental Table S1. List of PCR primer sequences

ACKNOWLEDGEMENTS
We thank David Somers for the ELF3-OX seeds, and Chuah Cha and Matt Chiang for excellent technical assistance.

LITERATURE CITED
S c i e n c e 2 9 3 : P l a n t C e l l 1 3 : P l a n t J 1 7 : N a t u r e 4 1 9 : P l a n t J 4 6 :       O  n  a  i  K  ,  I  s  h  i  u  r  a  M   (  2  0  0  5  )  P  H  Y  T  O  C  L  O  C  K  1  e  n  c  o  d  i  n  g  a  n  o  v  e  l  G  A  R  P  p  r  o  t  e  i  n  e  s  s  e  n  t  i  a  l  f  o  r  t  h  e  A  r  a  b  i  d  o  p  s  i  s  c  i  r  c  a  d  i  a  n  c  l  o  c  k  .  G  e  n  e  s  C  e  l  l  s   1  0  :   9  6  3  -9  7  2   P  a  r  a  A  ,  F  a  r  r  e  E  M  ,  I  m  a  i  z  u  m  i  T  ,  P  r  u  n  e  d  a  -P  a  z  J  L  ,  H  a  r  m  o  n  F  G  ,  K  a  y  S  A   (  2  0  0  7  )  P  R  R  3  i  s  a  v  a  s  c  u  l  a  r  r  e  g  u  l  a  t  o  r  o  f  T  O  C  1  s  t  a  b  i  l  i  t  y  i  n  t  h  e  A  r  a  b  i  d  o  p  s  i  s  c  i  r  c  a  d  i  a  n  c  l  o  c  k . P l a n t C e l l 1 9 :          The data are presented as the mean of two biological replicates ± SD. All experiments were done at least twice with similar results.