Functional importance of conserved domains in the flowering-time gene CONSTANS demonstrated by analysis of mutant alleles and transgenic plants


For correspondence (fax +49 2215062207; e-mail .


CONSTANS promotes flowering of Arabidopsis in response to long-day conditions. We show that CONSTANS is a member of an Arabidopsis gene family that comprises 16 other members. The CO-Like proteins encoded by these genes contain two segments of homology: a zinc finger containing region near their amino terminus and a CCT (CO, CO-Like, TOC1) domain near their carboxy terminus. Analysis of seven classical co mutant alleles demonstrated that the mutations all occur within either the zinc finger region or the CCT domain, confirming that the two regions of homology are important for CO function. The zinc fingers are most similar to those of B-boxes, which act as protein–protein interaction domains in several transcription factors described in animals. Segments of CO protein containing the CCT domain localize GFP to the nucleus, but one mutation that affects the CCT domain delays flowering without affecting the nuclear localization function, suggesting that this domain has additional functions. All eight co alleles, including one recovered by pollen irradiation in which DNA encoding both B-boxes is deleted, are shown to be semidominant. This dominance appears to be largely due to a reduction in CO dosage in the heterozygous plants. However, some alleles may also actively delay flowering, because overexpression from the CaMV 35S promoter of the co-3 allele, that has a mutation in the second B-box, delayed flowering of wild-type plants. The significance of these observations for the role of CO in the control of flowering time is discussed.


The CONSTANS (CO) gene was originally identified because of the late-flowering phenotype of co mutant plants (Koornneef et al., 1991; Redei, 1962). The phenotype of the mutant suggested that CO protein promotes the transition from vegetative growth to flowering, and this was supported by the demonstration that plants carrying extra copies of CO (Putterill et al., 1995) or overexpressing CO from the 35S promoter (Onouchi et al., 2000) flowered earlier than wild-type. The CO gene was cloned (Putterill et al., 1995), and the predicted protein product contains two regions of 43 amino acids towards the amino terminus of the protein that are closely related in sequence. Each of these regions contains an arrangement of cysteine residues similar to that present in the zinc fingers of GATA transcription factors, but little direct homology to these proteins was detected (Putterill et al., 1995). The construction of a translational fusion of CO to the ligand binding domain of the rat glucocorticoid receptor (CO:GR; Simon et al. 1996) provided further evidence that CO acts to influence transcription. Introduction of CO:GR into co mutant plants did not correct the mutant phenotype until the plants were treated with the steroid dexamethasone (dex). A similar fusion of the plant transcription factor LEAFY to the GR domain was retained in the cytoplasm until dex was added, suggesting that the GR domain operates in plants as it does in animals by sequestering the fusion protein in the cytoplasm (Wagner et al., 1999). Furthermore, induction of CO:GR with dex is associated with the rapid transcription of likely target genes, such as SOC1 that encodes a MADS box transcription factor (Samach et al., 2000). Taken together these data support the idea that CO acts in the nucleus to promote flowering by altering the transcription of downstream target genes.

Despite the considerable genetic and molecular data available on genes that interact with CO to regulate flowering (Kardailsky et al., 1999; Kobayashi et al., 1999; Onouchi et al., 2000; Samach et al., 2000; Soppe et al., 2000), little is known of the function of the CO protein nor of the roles of the different domains of the protein. Here several molecular-genetic approaches are used to address the function of CO. The availability of the genomic sequence of Arabidopsis (Arabidopsis Genome Initiative, 2000) allowed us to identify an extensive family of related proteins and thereby to recognize domains conserved between these proteins that may be functionally important. The functional significance of these homologies is addressed by analysing the sequence of eight mutant alleles and determining the positions of the mutations within the CO protein.

The blocks of homology identified in the sequence comparisons and the positions of the mutations suggest that CO protein has a modular structure with two zinc fingers near the amino terminus and a domain of unknown function near the C-terminus. Transcription factors frequently have a modular structure in which for example the DNA-binding domain is separable from the transcriptional activation or repression domains (Hope and Struhl, 1986; Keegan et al., 1986). Mutant proteins in which one domain is inactivated but the other is intact can act as dominant negative forms that repress the function of the wild-type protein and this has been used to investigate the function of plant transcription factors (e.g. Mizukami and Ma, 1997; Unger et al., 1993). This approach was used to address the function of CO by making transgenic wild-type plants overexpressing mutant proteins in which one or other of the protein domains are affected.

The intracellular location of plant proteins can be determined by constructing translational fusions with green fluorescent protein (GFP; Haseloff et al. 1997; Grebenok et al., 1997). This approach was also used to determine whether CO is present in the nucleus, and whether mutations that affect one of the protein domains prevent nuclear localization.

The results of these approaches are used to propose models of how CO acts to regulate flowering time.


CONSTANS contains two B-boxes that are altered in five mutant alleles

Analysis of the CO protein using the SMART program (Schultz et al., 1998, 2000) identified strong similarity between the proposed zinc fingers of CO and those of the B-box (Figure 1a and b). The B-box is a class of zinc finger, usually of the type C-X2-H-X7-C-X7-C-X2-C-X5-H-X2-H, that was identified in a variety of animal proteins including several transcription factors (XNF7, RPT-1, EFP), ribonucleoproteins (SS-A/Ro, PwA33) and proto-oncogene products (RFP, PML, TIF1) (reviewed in Borden, 1998; Reddy et al., 1992). The CO protein contains two B-box motifs that show 46% identity and 86% similarity with each other (Figure 1c). There are seven conserved residues that could act as metal-binding residues within the B-box motif, and all of these are conserved in both CO B-boxes (Figure 1b). Four of these residues were shown to bind zinc in the B-box structure of the Xenopus protein XNF-7 (Borden et al., 1995), and these four residues are conserved in both of the CO B-boxes (Figure 1b).

Figure 1.

Comparison of CO with CO-Like proteins and with B-box containing proteins of animals.

(a) Comparison of the B-boxes of CO with several B-box proteins of animals. The animal B-box proteins are XNF7 (Miller et al., 1989), PwA33 (Bellini et al., 1993), RPT-1 (Patarca et al., 1988), SS-A/Ro (Chan et al., 1991), RFP (Takahashi et al., 1988), ATDC (Leonhardt et al., 1994), EFP (Inoue et al., 1993), TIF1 (Miki et al., 1991) and PML (Goddard et al., 1991). B1 and B2 are the most N-terminal or C-terminal B-box, respectively. (b) The consensus spacing of C and H residues in animal B-box proteins compared with the spacing between these residues in the CO B-boxes. X represents any amino acid. The asterisks indicate those residues predicted to bind zinc in the NMR structure of the XNF7 B-box (Borden et al., 1995). (c) Alignment of the first and second B-boxes of CO with those of the CO-Like proteins. The predicted amino acid substitutions in co-2, co-3, co-4 and co-6, as well as the predicted deletion in co-1 are indicated. Accession numbers are CO (embl X94937; At5g15840; Putterill et al. 1995), COL1(embl Y1005; At5g15850; Putterill et al., 1997), COL2(gb L81120; At3g02380; Ledger et al. 1996), COL3(gb AC006585; At2g24790), COL4(gb AF069716; At5g24930), COL5(dgb AB018118; At5g57660), COL6(gb AC011915; At1g68520), COL7(gb AC016662; At1g73870), COL8(gb AC016041; At1g49130), COL9(gb AC009176; At3g07650), COL10(dgb AB023039; At5g48520), COL11(emb Z97338; At4g15250), COL12(dgb AP000739; At3g21880), COL13(gb AC005309; At2g47890), COL14(gb AC002332; At2g33500), COL15(gbAC069471; At1g28050) and COL16(gbAC079281; At1g25440). (d) Schematic representation of the sequence motifs found within the CO and COL proteins, or within the animal B-box proteins. The CO and COL proteins contain one or two B-boxes near their N-terminus, and a C-terminal domain recently named the CCT domain (Strayer et al., 2000). The second B-box of COL9, 10, 11, 12, 13, 14 and 15 is relatively dissimilar to that of CO and is shown as an open rectangle. Moreover, the second B-Boxes of COL9, COL10, COL11, and COL12do not exactly match the B-box consensus and therefore may not be active B-Box domains. The animal proteins carry the illustrated arrangement of RING fingers, B-boxes and coiled-coil domains. (e) Alignment of the carboxy-terminal CCT domains of CO and COL1-16. The predicted amino acid substitutions in co-5 and co-7 are illustrated.

Two CO-LIKE (COL) genes were previously described in Arabidopsis (Figure 1c; Ledger et al., 1996, 2001; Putterill et al., 1997). A BLAST search of the Arabidopsis genome sequence was performed to determine whether there were other members of this gene family. By analysing the Arabidopsis genome sequence (Arabidopsis Genome Initiative, 2000) a total of 16 COL genes were identified that contained one or two B-boxes at the amino terminus of the predicted protein (Figure 1c), and another highly conserved domain at the carboxy terminus (see below). Five COL proteins have two B-box motifs closely related to those of CO, seven contain a second B-box less closely related to those of CO and four have only one B box motif.

The CO gene was amplified by PCR from each of the seven mutants and the resulting fragments sequenced to identify the mutations. Five of the seven co alleles contain mutations that affect residues in the B-boxes (Figure 1c), suggesting that these are important for CO protein function. The co-6 mutation causes substitution of an alanine for a valine in the first B-box. The co-2 mutation converts an arginine to a histidine towards the carboxy-terminal end of the first B-box. The co-1, co-3 and co-4 mutations affect the second B-box. The co-3 mutation affects a histidine residue that, based on the analysis of the XNF-7 B-box structure, is likely to be required for zinc binding (Figure 1a, b and c; Borden et al., 1995).

CONSTANS contains a carboxy-terminal domain that is conserved among related proteins and is functionally important

The two remaining mutant alleles, co-5 and co-7, do not affect the B-box structures. These mutations affect adjacent proline and arginine residues in a highly conserved basic domain of approximately 43 amino acids near the C-terminus of the protein (Figure 1e). This novel domain was previously proposed to contain a nuclear localization sequence (Robert et al., 1998), and is also found in all 16 COL proteins (Figure 1e). Homology to this domain was also recently described in proteins that do not contain B boxes (Kurup et al., 2000; Makino et al., 2000; Strayer et al., 2000).

The carboxy-terminal domain of CONSTANS is sufficient to localize GFP to the nucleus, but the co-7 allele does not affect nuclear localization

To test whether CO protein is localized to the nucleus, and whether this is conferred by the carboxy-terminal domain translational fusions were constructed between GFP and CO or CO derivatives.

A translational fusion of GFP and CO (GFP:CO) was constructed in plasmid pAVA121 (von Arnim et al., 1998; Experimental procedures). GFP was fused to the N-terminus of the CO protein and expressed from a double 35S promoter (Experimental procedures). A transient expression assay in onion bulb epidermal cells was used to assess the cellular location of the fusion protein. Cells bombarded with the control plasmid pAVA121 showed GFP localization both in the cytoplasm and in the nucleus (Figure 2a), as previously shown (Grebenok et al., 1997; Haseloff et al., 1997). However, GFP:CO fusion protein was located in the nucleus and was not detected in the cytoplasm (Figure 2b). CO can therefore localize GFP to the nucleus of these cells, suggesting that CO is a nuclear protein. To confirm that GFP:CO retained biological activity, the 35S::GFP:CO fusion was introduced into co-2 mutants. The transgenic plants were early flowering and showed a similar phenotype to 35S::CO (Onouchi et al., 2000). The cellular location of the fusion protein was analysed in root, hypocotyl and cotyledon cells and shown to be located in the nucleus (Figure 2g, h and i).

Figure 2.

Subcellular localization of CO:GFP fusion protein.

(a)– (f) Images of onion epidermal cells stained with DAPI. In each case the tissue was viewed using epifluorescence optics with blue excitation to detect GFP (left) and UV-excitation to detect nuclei. Cells bombarded with 35S::GFP. Bars = 100 μm. (a) Cells bombarded with 35S::GFP. (b) Cells bombarded with 35S::GFP:CO. (c) Cells bombarded with GFP:CtermCO. (d) Cells bombarded with GFP:NtermCO. (e) Cells bombarded with GFP:co-5. (f) Cells bombarded with GFP:co-7. All samples were stained with DAPI and viewed under epifluorescence optics with blue (left) and UV (right) excitation. (g)– (i) Images of 11-day-old-transgenic Arabidopsis plants carrying 35S::CO:GFP. (g) Junction of hypocotyl and root. Merged images of green and red channels. GFP fluorescence detected in the green channel and chlorophyll autofluorescence in the red channel. (h) Hypocotyl tissue. Merged images of green and red channels as for G. (i) Root hairs.

To determine whether the conserved domain at the C-terminus of the CO protein is responsible for nuclear localization, a series of CO deletion derivatives were fused to GFP. The plasmid containing GFP fused to the region of CO between amino acids 304 and the C-terminal amino acid 373 (GFP:CtermCO; Experimental procedures) was bombarded into onion cells. The GFP:CtermCO fusion protein was observed only in the nucleus (Figure 2c), suggesting that this C-terminal sequence is sufficient to target GFP to the nucleus. Some proteins contain multiple NLSs (Varagona et al., 1992), and therefore to test whether another NLS was present in the N-terminal portion of CO, a truncated CO-protein lacking the C-terminal residues 302–373 was produced (GFP:NtermCO). This fusion protein was tested for intracellular localization as described earlier. Fluorescence was observed both in the cytoplasm and nucleus (Figure 2d), demonstrating that this portion of CO does not localize GFP exclusively to the nucleus. Taken together these experiments suggest that the only region of CO containing an NLS is between amino acids 304 and 373.

The mutant alleles co-5 and co-7 contain mutations in the C-terminal region of the protein that was demonstrated above to be sufficient for nuclear localization. The C-terminal regions (between amino acids 304 and 373) from these mutant proteins were fused to GFP to determine whether the mutations affected nuclear localization of the protein. Cells bombarded with 35S::GFP:co-5 DNA showed GFP localization in the cytoplasm and in the nucleus (Figure 2e), suggesting that the co-5 mutation affects the subcellular localization of the CO protein. The distribution of GFP:co-5 was similar to that of the GFP control but the level of expression was lower. However, GFP:co-7 showed GFP localization only in the nucleus (Figure 2f), suggesting that this mutation does not affect nuclear import of the CO protein.

Isolation and characterization of co-8, a likely null allele

Based on their sequences none of the seven classical co mutant alleles were predicted to certainly abolish CO function. All seven caused in-frame changes, including co-1 which was induced with X-rays. This raised the possibility that complete loss of function may generate pleiotropic effects, and that the seven classical alleles were all hypomorphs identified by screening phenotypically for late-flowering plants. Therefore, pollen irradiation was used to identify loss-of function alleles without making the assumption that these would only cause late flowering. An appropriate population of plants made with γ-irradiated pollen was described previously (Vizir et al., 1994). Pollen from Landsberg erecta plants was irradiated and used to cross-fertilize plants homozygous for the genetically linked mutations lu co-1 ms1 ttg. Twenty-six late-flowering M1 plants were identified, and these potentially carried novel co mutant alleles derived from irradiation. These plants were self-fertilized, and six of them gave rise to M2 progeny that were all late flowering, although only one of these lines was fertile. The mutation in this late-flowering line was preliminarily called co-8, and was tested further at the genetic and molecular levels.

Southern analysis of co-8 using probes derived from the CO genomic region demonstrated that a deletion had removed 1.3 kb from the 5′ end of the CO gene. The deletion included the sequence encoding the two B-boxes and approximately 1 kb of the 5′ non-coding region (Figure 3a). However, the pattern of hybridizing fragments detected in the Southern analysis could not be explained by a simple deletion. For example, probes that flanked the deletion on either side did not hybridize to the same EcoRI or HindIII fragments, although they would have been predicted to do so had a simple deletion occurred. This suggested that an insertion or an inversion had probably occurred at the deletion breakpoints.

Figure 3.

Figure 3.

Structure of the co-8 allele.

(a) Proposed derivation and final structure of the co-8 allele. The first diagram illustrates the structure of the wild-type chromosome. The CO gene is illustrated as a rectangle containing the B-boxes (marked with internal horizontal lines) and the CCT domain (marked with internal vertical lines). DNA within BAC MPI10 is marked with black squares. CO and MPI10 are approximately 17 Mb apart, and the orientation of this intervening DNA is denoted by arrowheads. The positions of primers used to analyse the structure of the mutant allele are illustrated, their use is described in the text and their sequences appear in the Experimental procedures. The second diagram shows the location of two deletions that are proposed to have occurred in the generation of the co-8 allele, and their positions relative to the primer sequences. The third diagram illustrates the proposed final structure of the allele. The intervening DNA is inverted such that the MPI10 sequence containing co75 is adjacent to CO sequence containing co17 and MPI10 sequence containing co74 is adjacent to MPI10 sequence containing co74. (b) Analysis of CO RNA present in co-8 mutants by RT–PCR. Experiment I was performed using primers oli5 and co50 (Experimental procedures). These detected the 3′ end of the wild-type CO mRNA extending from the single intron over the region encoding the CCT domain. Experiment II was performed using primers oli9 and co53 (Experimental procedures). These detected the 5′ end of the wild-type CO mRNA extending from the single intron to the position at which co53 anneals within the first exon, as shown in the diagram. Experiment III shows the detection of the APETALA2 mRNA as a control for the abundance of cDNA used in each of the previous RT–PCR experiments.

Figure 3.

Figure 3.

Structure of the co-8 allele.

(a) Proposed derivation and final structure of the co-8 allele. The first diagram illustrates the structure of the wild-type chromosome. The CO gene is illustrated as a rectangle containing the B-boxes (marked with internal horizontal lines) and the CCT domain (marked with internal vertical lines). DNA within BAC MPI10 is marked with black squares. CO and MPI10 are approximately 17 Mb apart, and the orientation of this intervening DNA is denoted by arrowheads. The positions of primers used to analyse the structure of the mutant allele are illustrated, their use is described in the text and their sequences appear in the Experimental procedures. The second diagram shows the location of two deletions that are proposed to have occurred in the generation of the co-8 allele, and their positions relative to the primer sequences. The third diagram illustrates the proposed final structure of the allele. The intervening DNA is inverted such that the MPI10 sequence containing co75 is adjacent to CO sequence containing co17 and MPI10 sequence containing co74 is adjacent to MPI10 sequence containing co74. (b) Analysis of CO RNA present in co-8 mutants by RT–PCR. Experiment I was performed using primers oli5 and co50 (Experimental procedures). These detected the 3′ end of the wild-type CO mRNA extending from the single intron over the region encoding the CCT domain. Experiment II was performed using primers oli9 and co53 (Experimental procedures). These detected the 5′ end of the wild-type CO mRNA extending from the single intron to the position at which co53 anneals within the first exon, as shown in the diagram. Experiment III shows the detection of the APETALA2 mRNA as a control for the abundance of cDNA used in each of the previous RT–PCR experiments.

To analyse the structure of co-8 more carefully both junctions between CO and the presumed rearrangement were isolated by using Inverse Polymerase Chain Reaction (IPCR) and by constructing a cosmid library from DNA extracted from co-8 mutants (Robson, 1998; Experimental procedures). The exact locations of the breakpoints were defined by DNA sequencing using primers (co52 and co17 in Experimental procedures; Figure 3a) designed to the CO gene to sequence into the rearrangement on either side. This identified the breakpoints within CO to be 1025 bp upstream of the ATG and 304 bp downstream of the ATG. Adjacent to the breakpoints at both ends of the rearrangement was novel DNA not associated with the CO gene in wild-type plants. Both of these unknown sequences were used in a BLAST search against the database and both were identical to sequences within P1 clone MPI10 that contains DNA from Arabidopsis chromosome 5. The two junction sequences however, are not directly adjacent in the sequence of MPI10 but 1208 bp apart. On the wild-type chromosome, the CO gene (on BACF14F8) is approximately 17 Mb from the DNA within clone MPI10 according to the physical maps of chromosome 5 (TAIR). The co-8 allele was probably therefore derived from an irradiation-induced event in which two deletions occurred approximately 17 Mb apart. One of these was within the CO gene, and the other within the DNA cloned in BAC MPI10. These deletions were then repaired such that the intervening 17 Mb segment of chromosome 5 was inverted (Figure 3a).

In the co-8 allele 1025 bp of the 5′ untranslated region is deleted and the truncated CO ORF is fused to DNA from BAC MPI10 that is not normally associated with the CO gene. Reverse transcriptase–PCR was used to test whether CO mRNA was present in co-8 plants (Figure 3b). No transcript was amplified from cDNA made from co-8 mutants when one of the primers (co53) used for the PCR was designed to anneal to DNA removed by the deletion in co-8, although cDNA made from wild-type plants produced a transcript of the expected size. However, PCR primers (co50 and oli5) that annealed to part of the CO ORF that is retained in co-8, amplified a product of the expected size from cDNA made from both co-8 mutants and wild-type plants. Therefore, the remaining portion of the CO ORF is still transcribed in the co-8 rearrangement.

Although a novel transcript is detected in the co-8 mutant, the deletion of both B-boxes suggests that the mutant will lack any CO activity.

All eight constans alleles are semidominant

The co-2, co-3 and co-4 alleles were previously reported to be semidominant (Koornneef et al., 1991). Based on the phenotype of homozygous co-2 mutant plants carrying transgenic copies of the wild-type gene, Putterill et al. (1995) proposed that the semidominance of co-2 was due to haploinsufficiency rather than to co-2 encoding an altered product that delayed flowering.

To determine whether all seven classical alleles and the new co-8 allele all cause semidominance, they were each independently crossed to Landsberg erecta. The F1 progeny were sown in long-day conditions and their flowering time compared to that of wild type and homozygous mutant controls. All of the F1 plants showed an intermediate flowering time phenotype (Table 1), indicating that all eight alleles are semidominant. This was confirmed in the F2 generation, in which approximately 50% of plants showed intermediate flowering times (data not shown).

Table 1.  Flowering time of wild type and the constans homozygous and heterozygous mutants, measured as the total number of leaves produced before the onset of flowering. Data from 20 individuals for each genotype ± SE
(leaf number)
(leaf number)
WT7.5 ± 0.2
co-411.3 ± 0.4 9.6 ± 0.2EMS
co-514.4 ± 0.411.5 ± 0.4EMS
co-219.4 ± 0.913.3 ± 0.3EMS
co-822.5 ± 0.316.2 ± 0.7 γ– ray
co-623.7 ± 0.511.9 ± 0.3EMS
co-125.6 ± 0.613.5 ± 0.3X – ray
co-728.5 ± 0.810.2 ± 0.2EMS
co-329.1 ± 1.215.2 ± 0.6EMS

Transgenic wild-type plants over-expressing the co-3 protein are late flowering

The CO gene contains two functional domains based on homology searches and analysis of mutant alleles. Both of these domains may facilitate interactions between CO and other proteins (see Discussion). This suggested that mutant forms of CO in which one domain is altered but the other is intact might sequester interacting proteins into inactive complexes, and thereby lead to a late-flowering phenotype. Such a dominant negative function has been proposed recently to explain the effect of mutant forms of B-box proteins (Peng et al., 2000). To test the effectiveness of this approach, the co-3 allele, that carries a mutation in the second B-box of CO, and the co-7 allele, that carries a mutation in the carboxy-terminal domain, were each expressed from the CaMV 35S promoter (Experimental procedures). These transgenes were then introduced into wild-type Landsberg erecta plants.

Approximately 150 kanamycin-resistant T1 plants were identified after infiltration of Landsberg erecta plants with Agrobacterium cells carrying the 35S::co-3 construct. Around 20 of these T1 plants appeared to flower at least slightly later than wild-type plants. The T1 plants were self-fertilized and individuals homozygous for the T-DNA were identified in five of the late-flowering lines. T3 progeny of these five lines were then scored for flowering time under long-day conditions (Figure 4). All five lines flowered significantly later then wild-type plants, although none were as late flowering as the co-2 or co-3 mutant. Northern blots demonstrated that co-3 mRNA was over-expressed in the late-flowering 35S::Co-3 lines, and therefore that the late-flowering phenotype was not due to cosuppression causing a reduction in expression of the co-3 and CO mRNAs. This indicates that over-expression of the co-3 allele, that carries a mutation in one of the B-boxes, can delay flowering of wild-type plants.

Figure 4.

The effects of overexpression of the co-3 allele. (a) Flowering times of Landsberg erecta, co-2, co-3 and 35S::CO plants compared to the flowering times of five independent 35S::co-3 transformants. The total numbers of leaves formed prior to the onset of flowering are shown. Error bars represent the SE for 20 individulas for each genotype growing under long day conditions. (b) A schematic illustration of the 35S::CO and 35S::co-3 constructs.

A similar experiment was performed with the 35S::co-7 construct and a total of 29 kanamycin-resistant T1 plants were identified. However, none of these flowered later than wild-type plants. These were self-fertilized and in the T2 generation their flowering time was tested under long and short days. None of the transformants flowered late under long days, suggesting that the 35S::co-7 construct did not generate a dominant negative phenotype. Some of the 35S::co-7 transformants flowered earlier than wild-type under short days, indicating that the protein encoded by the co-7 allele may retain some residual CO activity.


CO is a member of a novel family of B-box containing proteins

The CO zinc finger regions are most similar to those of B-box proteins. The seven potential zinc-binding residues within the B-box consensus sequence are conserved in the CO and most of the COL B-boxes. Furthermore, the co-3 allele, which causes the most severe delay in flowering time, alters a histidine in the second B-box that corresponds to a histidine shown to bind zinc in the solution structure of the B-box of the Xenopus protein XNF-7 (Borden et al., 1995). Although the zinc fingers of CO were originally compared with those of GATA transcription factors (Putterill et al., 1995), based on the spacing of four of the cysteine residues within the CO fingers, the similarity to the more recently described B-boxes is much stronger. Putterill et al. also pointed out the lack of direct homology with GATA transcription factors (Putterill et al., 1995).

The function of B-box proteins in plants has not previously been discussed. In animal proteins, the B-box domain is usually part of a tripartite motif comprising a zinc-binding RING finger and a B-box domain followed closely (5–8 amino acids) by a predicted α-helical coiled-coil domain (RBCC family; Figure 1). The spacing between the three elements is highly conserved suggesting that the relative position of the domains is of functional importance. Proteins in a subfamily of this group, defined by the gene for ataxia–telangiectasia group D (ATDC) (Leonhardt et al., 1994), have one or two B-boxes and a coiled coil domain but no RING finger. Another variation in the arrangement of RBCC domains is found in the protein kinase C-interacting protein (RBCK1) which has two coiled coil domains followed by a RING finger, a B-box and a B-box-like domain. This is the only published example of a protein that does not contain the coiled-coil domain after the B-box motif (Tokunaga et al., 1998).

CO and the other COL proteins are unusual in containing one or two B-box domains with no coiled-coil domain or RING finger. There are several plant proteins containing RING fingers (COP1, Deng et al., 1992; A-RZF, Zou and Taylor, 1997; PRT1 Potuschak et al., 1998) and in COP1 this is followed by a coiled-coil domain (Deng et al., 1992). However none of these plant RING finger proteins contain B-boxes.

The RBCC motif is believed to mediate protein–protein interactions (Borden, 1998; El-Husseini et al., 2000; Peng et al., 2000; Tsuzuki et al., 2000). However, the two B-boxes in CO may not function in a similar way to those in the RBCC motif. For example, the RBCC domain of the transcriptional corepressor KAP-1 appears to function as an integrated structural unit in which the RING finger, the B-box and the coiled-coil region are all required for interaction with the transcriptional repression module KRAB (Peng et al., 2000). However, in other cases the B-boxes appear to function autonomously. For example, the transcription factor GATA-2 interacts specifically with the B-box region of promyelocytic leukaemia protein (Tsuzuki et al., 2000).

The role of the conserved carboxy-terminal domain of CO

The carboxy-terminal region of CO was sufficient to direct GFP to the nucleus, suggesting that nuclear localization is one function of this region. Such a function for this region was originally proposed based on the similarity of a portion of it to the consensus sequence for an NLS (Robert et al., 1998). More recently, a related region in the TOC1 protein was shown to direct TOC1 to the nucleus in transient expression assays, and was termed the CCT (CO, COL, TOC1) domain (Kurup et al., 2000; Makino et al., 2000; Strayer et al., 2000). Nevertheless the carboxy-terminal domain of TOC1 shows only 51% identity to that of CO whereas the least closely related COL protein, COL14, shows 60.5% identity. The experiments described here establish that the CCT domain in CO shares the nuclear localization function with the related domain in TOC1, and the early flowering of the 35S::CO:GFP plants confirmed that CO:GFP located in the nucleus retains biological activity.

However, in addition to nuclear localization, the CCT domain probably has other functions. This was originally suggested because the conserved region is 43 amino acids long, which is a longer stretch of contiguous homology than is shown by nuclear localization sequences (Raikhel, 1992). The demonstration that the co-7 allele has a severe effect on flowering time but does not affect the nuclear localization function of this domain further suggests that the domain has an additional role in CO activity.

A CCT domain is present in at least 18 proteins, including TOC1, that do not contain B-boxes (Strayer et al., 2000 and data not shown). There are also a further 13 proteins that contain one or two B-boxes but do not have the CCT domain. These include the salt-tolerance protein STO (Lippuner et al., 1996). The existence of proteins containing only one of these domains, either B-box or CCT-domain, suggests that these domains act independently of one another. This is supported by the observation of Kurup et al. (2000) who showed that the CCT domains of CO and TOC1 (also called ABI3 Interacting Protein 1) interact in yeast cells with the Arabidopsis transcription factor ABI3. This interaction was reduced approximately two-fold by both the co-5 and co-7 mutations. Therefore, the carboxy-terminal region probably has a role in protein–protein interaction as well as in nuclear localization.

The dominance of the co mutations

Three co mutant alleles were previously shown to be semidominant with the heterozygotes showing a phenotype intermediate between the homozygous mutants and wild-type (Koornneef et al., 1991; Redei, 1962). Putterill et al. (1995) proposed that this was likely to be caused by haploinsufficiency, in which the heterozygotes did not produce enough CO protein to promote early flowering, rather than the mutant allele encoding an altered gain of function protein. This was proposed because transgenic mutants homozygous for the co-2 allele and carrying wild-type CO as a transgene flowered as early as wild-type plants. We have now shown that all eight mutant alleles are semidominant. The novel co-8 allele, which we isolated, may be a null allele because the DNA encoding the translational start site and both B-boxes is deleted, although the remaining portion of the co-8 mRNA may still be translated to produce a truncated protein. This truncated protein would carry the CCT domain (Figure 3) and may actively delay flowering, as was shown for overexpression in wild-type plants of the co-3 allele, which also carries an intact CCT domain and impaired B-box domain. In contrast, overexpression in wild-type plants of the co-7 allele, which encodes intact B-boxes and an impaired CCT domain, did not delay flowering, although this allele was semidominant when tested in heterozygous plants (Table 1). Therefore, the observation that co-8 and co-7 alleles are semidominant is consistent with the proposal that the dominance of co mutations is caused by haploinsufficiency.

Nevertheless, at least for some alleles the semidominance may be caused by a combination of haploinsufficiency and the mutant allele encoding an altered product that actively delays flowering. The late-flowering phenotype of Landsberg erecta plants carrying the 35S::co-3 transgene clearly indicates that at least when overexpressed this allele can actively delay flowering. The co-3 mutation affects a histidine residue that is predicted to be essential for zinc-binding within the second B-box. The active delay in flowering time caused by overexpression of this protein may be a consequence of the co-3 protein sequestering wild-type CO protein or proteins required for CO function into inactive complexes. The sequestration may occur by proteins binding to the first B-box of co-3 or to the CCT domain, neither of which is affected by the co-3 mutation.

Implications for the roles of CO and CO-Like proteins in regulating flowering time

CO is a nuclear protein (Figure 2) that acts to promote flowering time by rapidly inducing the expression of downstream flowering-time genes such as SOC1 and FT (Samach et al., 2000). The zinc fingers of CO are required for CO function and are most similar to B-box motifs, which are predicted to mediate protein–protein interaction rather than DNA binding. This suggests that to activate transcription of downstream genes, CO may be recruited to promoters by DNA binding proteins. Such a role for B-box proteins has been described in animals. For example, the transcription factor GATA-2 recruits the B-box protein promyelocytic leukemia (PML) to DNA, and PML enhances the ability of GATA-2 to activate transcription (Tsuzuki et al., 2000). Similarly, the Krüppel associated box (KRAB) that acts as a transcriptional repression module must interact with the RBCC protein KAP-1 in order to cause gene silencing (Peng et al., 2000). The KAP-1 protein is recruited to DNA by zinc-finger DNA binding proteins that carry the KRAB domain. These examples may describe a paradigm for CO function, and suggest that it may interact with specific DNA binding proteins that enable its recruitment to DNA. The observation that the carboxy-terminal regions of CO and TOC1 will interact with the DNA-binding protein ABI3 (Kurup et al., 2000), suggests that ABI3 or transcription factors of the same class may be responsible for the recruitment of CO and COL proteins to DNA.

The evolution of the family of 16 COL proteins that contain the B-boxes and the carboxy-terminal domain was recently discussed (Lagercrantz and Axelsson, 2000), but their function is unknown. Overexpression of COL1 shortened the period length of circadian clock regulation, but did not affect flowering time (Ledger et al., 2001). In some cases the B-boxes are closely related in sequence to those of CO (Figure 1), however, so far there is no evidence that they regulate flowering time, and they may interact with transcription factors that do not associate with CO, and thereby regulate a different set of target genes. Closely related RBCC proteins were previously shown to interact with specific protein partners (Cainarca et al., 1999).

Further understanding of the function of the CO and COL family is likely to come from identifying interacting proteins, some of which may recruit the B-box proteins to specific sets of target genes.

Experimental procedures

Plant material and growth conditions

Seeds from Landsberg erecta– Ler-0 (NW20), tt4–1 (N85) and EMS mutants co-2 (N175), co-3 (N176), co-4 (N177), co-5 (N178), co-6 (N179) and co-7 (N180) were obtained from M. Koornneef. These mutants are all in a Landsberg erecta background. Seeds from Redei's X-ray mutant co-1 (N3122) were also provided by M. Koornneef. This mutant is in Landsberg – La-0 (N1298). co-1 is also available in Landsberg erecta – co-1 er-1 (N3135). Seeds from γ-irradiated lu-1 co-1 ms1–1 ttg-1 (N240) were provided by I. Vizir.

In the summer plants in the glasshouses were grown in natural daylight. In the winter supplementary light was provided so that the minimum day length was 16 h. Flowering time was measured under defined conditions by growing plants in Sanyo Gallenkamp as described by Putterill et al. (1995) and Robson (1998).

DNA and RNA extraction

Plant genomic DNA was extracted as described by Dean et al. (1992). To make the co-8 cosmid library the DNA was further purified on a caesium chloride gradient prior to digestion (Sambrook et al., 1989). RNA for analysis of co-8 by RT–PCR was extracted as described by Putterill et al. (1995).

Cloning and sequencing of the co mutant alleles

DNA was extracted from seedlings as described above. A pair of primers designed to amplify the CO gene had previously been designed (Putterill et al., 1995); co41 (5′-GGTCCCAACGA AGAAGTGC-3′) and co42 (5′-CAGGGAGGCGTGAAAGTGT-3′). These were used to amplify a 1.95-kb fragment from wild-type and co mutants co-1 to co-7, in duplicate PCR reactions. The PCR products were blunt-ended using T4 DNA polymerase and cloned into the Eco RV site of pBluescript (SK +).

Library construction and screening

DNA from the co-8 mutant was extracted and purified as described above. The library was constructed as described in Schaffer et al. (1998) by ligating plant DNA partially digested with Sau 3 A into the Bam HI site of cosmid vector c04541 (Jones et al., 1992). The recombinant cosmids were packaged in Gigapack II Gold packaging extracts (Stratagene, La Jolla, CA, USA) according to the manufacturer's instructions and plated using Escherichia coli XL1 Blue MR.

Analysis of expression by RT–PCR

RNA was extracted from duplicate samples of tissue from 10-day-old-seedlings as described above and cDNA prepared as described by Putterill et al. (1995). Primer pairs used were as follows: to amplify the 5′ end of the CO transcript primers co53 (5′-ACGCCATCAGCGAGTTCC-3′) and oli9 (5′-AAATGTATGC-GTTATGGTTAATGG-3′) were used. To amplify the 3′ end of the CO transcript primers co50 (5′-CTCCTCGGCTTCGATTTCTC-3′) and oli5 (5′-CATTAACCATAAC-GCATACATTTC-3′) were used. Oli5 and Oli 9 were designed to anneal to the exon sequence either side of the single CO intron to prevent the amplification of contaminating DNA (Simon et al., 1996). The position of the intron is marked in the primer sequence by a hyphen. To amplify the APETALA2 cDNA as a control, primers AP2 Oli3 (5′-CTCAATGCCG-AGTCATCAGG-3′) and AP2 Oli4 (5′-CATG AGAGGAGGTTGGAAGC-3′) were used. The resulting PCR products were fractionated on an agarose gel, Southern blotted onto Hybond N+ (Amersham, Little Chalfont, UK) according to the manufacturer's instructions and probed with the CO cDNA.

Primers used to analyse the co-8 rearrangement

The following primers were used to identify and characterize the co-8 rearrangement:

  • co25 (5′-TACTGTTGTGCAAATGG-3′)and co52 (5′-GGAACAGCC ACGAAGCAAC-3′) were used in the IPCR experiment to amplify the DNA adjacent to the deletion in co-8.
  • co17 (5′-ATGGATCATGTGGACTAG-3′)anneals in the CO promoter and was used to first identify the inversion in co-8.
  • co74 (5′-GATGGGCTACGTATGCGGC-3′)and co75 (5′-GGACTA AGCATATACGACACATCTC-3′) were designed to anneal to DNA brought adjacent to each side of the deletion in co-8 by the inversion. In wild-type they anneal to DNA within P1 clone MPI10.

Construction of transformation vectors

p35S::co3 was constructed by first isolating the CaMV 35S promoter as a 350 bp Cla I –Hind III fragment from pJIT62 (Guerineau et al., 1992) and cloning it into these sites in pBluescript. A 1.7 kb Hind III fragment containing the co-3 genomic region, including the native CO polyadenylation sequences, was isolated from pco-3. Hind III cuts in this clone in the polylinker near the 3′ end of the gene and also in the plant DNA 70 bp upstream of the ATG. This was cloned into the Hind III site adjacent to the 35S promoter and orientated correctly by restriction mapping. 35S:co3 was moved as a Cla I –Bam HI fragment into the binary vector pSLJ1711 (Jones et al., 1992). p35S::CO was constructed in essentially the same way, as described by Onouchi et al. (2000).

Transformation of Arabidopsis

Landsberg erecta plants were transformed with the 35S::co-3 construct by floral dipping (Bechtold et al., 1993; Clough and Bent, 1998). The Agrobacterium strain used was C58C1 pGV101 pMP90. Kanamycin-resistant transformants (T1 generation) were selected on 1/2 × Murashige and Skooge (MS) agar. Flowering time was measured in the T3 generation using lines homozygous for the T-DNA from several independent transformants.

Bombardment of onion bulb epidermal cells

From the inner layer of onion bulb, a peel of epidermis was taken and placed inside up on top of a 50 μl drop of liquid MS on a plate containing solid MS medium (Varagona et al., 1992). The medium contained per litre, 4.3 g MS, 1 mg thiamine, 10 mg myo-inositol, 180 mg KH2PO4 and 30 g sucrose, the pH was adjusted to 5.7 with KOH. After autoclaving, 2.5 mg of amphotericin (in DMSO) was added to the medium. The onion epidermal layers were prepared just before bombardment. 20 μg of plasmid GFP(S65T):CO was precipitated onto gold particles and bombardment performed as described by McCabe and Christou (1993). After bombardment, onion cell layers were incubated at 20 °C for 5 h in complete darkness. To visualize the distribution of cellular DNA the onion peels were immersed in a solution of 0.1% (v/v) DAPI (Sigma-Aldrich, Dorset, UK) for 5 min Subsequently, they were mounted in water and examined by epifluorescence microscopy (Nikon E-800, Nikon, Melville, NY, USA).

Plasmid construction

The GFP-vector pAVA121 was provided by Dr A.G. von Arnim (von Arnim et al., 1998). This plasmid is based on the expression cassette of pRTL2 (Restrepo et al., 1990) that contains a double 35S promoter from CaMV, the translational leader sequence from tobacco etch virus (TEV), and the 35S polyadenylation signal from CaMV. The GFP cDNA is a modified version of mGFP4 (Haseloff et al., 1997) (GFP(S65T), in which the serine 65 residue is substituted by a threonine, resulting in increased absorbance of blue light and reduced absorbance of UV light (Heim et al., 1995). The CO cDNA was inserted in frame in the BglII restriction site of the C-terminus of GFP(S65T). The region corresponding to the CO C-terminus (Met304-Phe373) was amplified by PCR using the primers 5′TERCO (5′-CAA CTC GGA TCC ATG GAG AGA GAA GCC-3′) and 3′TERCO (5′- AAT CAG ATC TTT CTT TTT GCC ACA GGA G-3′). The 5′TERCO primer introduces a BamHI site before the first codon of the sequence (methionine 304) and the 3′TERCO primer introduces a BglII restriction site after the CO coding sequence. The PCR fragment was digested with BamHI and BglII and cloned into the BglII restriction site of the vector pAVA121 resulting in an in-frame translational fusion at the C-terminus of GFP. This fusion was called GFP(S65T):Cterm. The region corresponding to the first 303 amino acid residues was amplified by PCR using primers OLIGO2 (5′-TGA GGA TCC ATG TTG AAA CAA GAG AGT A-3′) and C′-STOP (5′-CT GAG ATC TCA ACT GAG TTG TGT TAC T G-3′). Oligo2 maintains the BamHI restriction site before the start codon of the CO gene and the reverse primer transforms the proline 302 codon (CCA) into a stop codon (TGA) as well as introducing a BglII restriction site after the stop codon. The amplified fragment was inserted at the 3′ end of the GFP gene, as described previously for the CO C-terminus. This fusion was called (GFP(S65T):Nterm. The primers 5′TERCO and 3′TERCO were also used to amplify the DNA encoding the C-terminal region (Met304-Phe373) of CO from the mutants co-5 and co-7. The PCR fragments were digested with BamHI and BglII and cloned into the BglII restriction site of the vector pAVA121 (creating fusion proteins GFP(S65T):Co-5 and GFP(S65T):Co-7). All the PCR fragments were sequenced to check for PCR errors.


This work was funded by grants from the BBSRC and EC to G.C. S.R.H. was supported by an EMBO long-term fellowship, M.M.R.C. by a Ph.D. studentship (PRAXIS XXI/BD/3781/94) from Fundação para a Ciência e a Tecnologia, Portugal, and P.H.R. by the EC through the REGIA project.