An allelic series reveals essential roles for FY in plant development in addition to flowering-time control

The switch to flowering is a key developmental transition in the plant life cycle. Entry into reproductive development is modulated with respect to both environmental and endogenous cues (Simpson and Dean, 2002). Genetic analysis in the model plant Arabidopsis has identified several pathways that regulate flowering time (Simpson and Dean, 2002). These pathways quantitatively regulate a key set of floral-integrator genes,the activity of which are important for flowering(Simpson and Dean, 2002). Wild accessions of Arabidopsis either overwinter vegetatively (winter annuals) or show a rapid-cycling habit, with flowering of winter annuals being promoted by the extended cold of winter, a process known as vernalization. Vernalization requirement is conferred by two genes, FRIGIDA(FRI) and FLOWERING LOCUS C (FLC)(Johanson et al., 2000; Michaels and Amasino, 1999; Sheldon et al., 1999). Active FRI alleles increase the accumulation of FLC mRNA, which encodes a MADS-box transcription factor that is a potent repressor of flowering (Michaels and Amasino,1999; Sheldon et al.,1999). FLC functions to repress flowering by antagonising the activation of floral-integrator genes such as AGAMOUS-LIKE 20/SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1(AGL20/SOC1) and FT (Hepworth et al.,2002; Michaels and Amasino,2001; Moon et al.,2003; Samach et al.,2000). The vernalization pathway acts in opposition to FRI by repressing FLC expression in response to the cold(Michaels and Amasino, 1999; Sheldon et al., 1999).

In parallel to vernalization, the autonomous pathway also acts to promote flowering by repressing expression of FLC(Michaels and Amasino, 1999; Sheldon et al., 1999). Loss of autonomous function confers a recessive vernalization requirement(Koornneef et al., 1991). The autonomous pathway currently comprises seven genes; FCA, FY, FPA, FVE,FLD, FLK and LUMINIDEPENDENS (LD), which encode putative transcriptional (FLD, LD, FVE) and post-transcriptional(FCA, FY, FPA, FLK) regulators of gene expression(Ausin et al., 2004; He et al., 2003; Lee et al., 1994; Lim et al., 2004; Macknight et al., 1997; Mockler et al., 2004; Schomburg et al., 2001; Simpson et al., 2003). Epistasis analysis indicates that these genes repress FLC in a partially non-redundant manner (Koornneef et al., 1998). FCA and FY make up one epistasis group and this reflects functional interaction of their gene products(Simpson et al., 2003). FCA is a plant-specific, nuclear RNA binding protein with a C-terminal WW protein interaction domain (Macknight et al.,1997). FCA interacts with FY, which is homologous to a highly conserved polyadenylation factor, Pfs2p(Ohnacker et al., 2000; Simpson et al., 2003).

In eukaryotes, the 3′ ends of RNA polymerase-II-generated transcripts are cleaved and polyadenylated and this is an essential step for transcript stability (Zhao et al., 1999). Genetic and biochemical approaches in Saccharomyces cerevisiae have defined a large number of conserved proteins required for RNA 3′-end processing, including the polyadenylation factor Pfs2p(Zhao et al., 1999). Pfs2p contains seven WD repeats and acts as an interaction surface within the cleavage and polyadenylation factor (CPF) 3′-end processing complex(Ohnacker et al., 2000). CPF acts with the cleavage factor I (CFI) complex to direct 3′-end processing of pre-mRNA transcripts (Zhao et al., 1999). The Arabidopsis homologue of Pfs2p is FY and this was revealed to play a role in RNA 3′-end processing through analysis of FCA gene regulation(Quesada et al., 2003; Simpson et al., 2003). FCA negatively autoregulates its expression by promoting premature cleavage and polyadenylation within intron 3, to generate the non-functional FCA-β transcript (Macknight et al., 2002; Quesada et al.,2003). This feedback mechanism requires an interaction between FCA and FY (Quesada et al., 2003; Simpson et al., 2003). The mechanism by which FCA and FY regulate FLC is unknown, but this may also involve regulated 3′-end processing. In plants, FY possesses an extended C-terminal domain in addition to seven conserved WD repeats(Fig. 1). This C terminus carries sequences predicted to interact with the FCA WW domain(Sudol and Hunter, 2000). Hence, the novel RNA-binding protein FCA is recruited to FY, a conserved polyadenylation factor, to mediate regulated 3′-processing.

Mutations in the 3′-end processing machinery are generally lethal and this is true for null pfs2 mutations in yeast(Ohnacker et al., 2000; Wang et al., 2005). However,hypomorphic mutations that are viable can also be recovered(Ohnacker et al., 2000; Zhao et al., 1999). Viable mutations in the Drosophila polyadenylation factor, suppressor of forked (su(f)) were identified as modifiers of retrotransposon insertions in the Forked gene(Parkhurst and Corces, 1985; Parkhurst and Corces, 1986). The su(f) mutation modified usage of premature polyadenylation sites within the retrotransposon long terminal repeats, restoring forkedgene expression (Parkhurst and Corces,1985; Parkhurst and Corces,1986). However, strong or null su(f) alleles are cell-autonomous, lethal mutations that prevent mitotic proliferation(Audibert et al., 1998; Audibert and Simonelig, 1999). To investigate whether FY is also an essential gene in Arabidopsis, an allelic series of fy mutations was characterised. A null fy mutation combined with conditional silencing of FY in Nicotiana demonstrate that FY is required for growth and development in plants. In addition, the different fyalleles indicate a requirement for both the conserved FY WD repeats and the C terminus in repression of FLC.

Growth conditions used for flowering-time analyses were as described previously (Macknight et al.,2002). The fy-1 and fca-1 mutants were provided by M. Koornneef (Wageningen University, The Netherlands)(Koornneef et al., 1991). The fca-9 mutant was provided by Chang-Hsien Yang (National Chung Hsing University, Taiwan). The fy-2 mutant was provided by Syngenta(Sessions et al., 2002). The fy-3 and fy-4 mutants were provided by the Arabidopsis TILLING project (Targeted Induced Local Lesions IN Genomes; ATP) (McCallum et al.,2000).

Plants carrying fy mutations were genotyped using PCR markers. The fy-3 mutation was genotyped by amplifying with the FY3F(5′-ACCACACCTTCAGGAAGACGTCTTATCTAG-3′) and FY3XBA(5′-TCACCAGAAACCATATAATTTTCATTGTGG-3′) primers, the PCR product was cut with XbaI in fy-3 mutants. The fy-4mutation was genotyped by amplifying with the FY4F(5′-CCCAAAGTGGGGAGTTTACTCTCT-3′) and FY4XBA(5′-CCTCCATCATCACCAGAAACC-3′) primers, the PCR product was cut by XbaI in fy-4 mutants. The dCAPs marker used to identify fy-1 was as described previously(Simpson et al., 2003). Plants carrying fy-2 were identified by herbicide (BASTA) resistance conferred by the T-DNA insertion or by PCR. Amplifying with a T-DNA left border primer LB (5′-TGGTTCACGTAGTGGGCCATCG-3′) and FY4 (5′-CTGTTGGAAAGGGTTGTTGTAGCCTGGAATC-3′) produces a product in the presence of the fy-2 T-DNA insertion.

The FY::GUS transgene was constructed by PCR, amplifying the FY promoter from Arabidopsis genomic DNA, made from the Columbia accession, using the primers PFYF(5′-CGAGCTCGGTGTGTTTTTGGG-3′) and PFYR(5′-CATGCCATGGTTGCCCACGGAGAAC-3′). The 1.3 kb PCR product was cloned and sequenced. The P_FY sequence was then introduced upstream of β-glucuronidase (GUS) as a NcoI-SacI fragment in the pGreen0029 binary vector(Hellens et al., 2000). The FY::GUS transgene was introduced into Arabidopsis by Agrobacterium-mediated floral dip transformation(Bechtold et al., 1993). Transformed lines were selected using kanamycin resistance.

The complementation transgene was constructed using the FYopen-reading frame amplified by RT-PCR using RNA extracted from Col seedlings with the primers FYBstEIIF (GGGTCTAGAGGTAACCTAAATTCGAACACTTTCGCAG)and F20SalIR (CCCGTCGACCTACTGATGTTGCTGATTGTT). The FY cDNA was cloned into pCAMBIA1300 as a XbaI/SalI fragment. The FY promoter was amplified using Pfu (Stratagene) from Columbia genomic DNA with the primers Pfy5 (GAGGGATCCACTATAGGTGTGGCAAAGCTCAT)and Pfy3 (GTTGCCCACGGAGAACAGT) and cloned as a BamHI/BstEII fragment upstream of the FY cDNA in pCAMBIA1300. The FY 3′-UTR was amplified from Col genomic DNA using the primers 3UTR-1 (GTAGGTCGACGTTGTATTAGTACATTAGTTT) and 3UTR-2 (CTCCGTCGACGTCTGCTGTGGTGGCTTGGGTCTT) and cloned as a SalI fragment downstream of the FY cDNA. The pCAMBIA-pFY::FY plasmid was transformed into Agrobacteriumstrain GV3101 and used to transform fy-4/FY heterozygote plants. Inheritance of fy-4 in T₁ progeny was analysed by genotyping with a dCAPs marker amplified with FY4F and FY4R(TTTAAACAGTCAATACCAGGAGCAG) and digested with XbaI.

To analyse seed development whole siliques were fixed for 1 hour at room temperature in Cornoy's solution (acetic acid:ethanol, 1:9) and then washed for 1 hour in 80% ethanol followed by 70% ethanol. The ethanol was then replaced with fresh clearing solution (chloral hydrate:H₂O:glycerol, 8:2:1) and left overnight at room temperature. Cleared seeds were dissected from siliques using 0.2 μm needles and mounted on a slide. Seeds were viewed using differential interference contrast (DIC) microscopy with a Nikon Microphot microscope(×20 or ×40 objectives). Pictures were taken using a Nikon digital camera.

Plants were stained for GUS expression as described previously(Jefferson, 1987). To analyse GUS expression during seed development, material was first fixed in Cornoy's solution for 1 hour at room temperature. After fixation the seed was extensively washed in GUS staining buffer and then GUS stained as described previously (Jefferson, 1987). After staining, seed was cleared as described above and analysed with DIC microscopy.

mRNA in situ hybridization was performed using a published protocol(Coen et al., 1990). 8 μm cross sections of Ler 10-day shoot meristems and longitudinal sections through siliques at several stages of development were used. Antisense and sense probes were constructed using the FY-CT construct as a template. The insert was amplified using M13 forward and reverse primers and transcribed with T3 RNA polymerase (antisense) and T7 RNA polymerase(sense).

LeFCA sequence from tomato (Lycopersicon esculentum) was provided by Dr R. Macknight (University of Otago, New Zealand). LeFCAsequence was PCR amplified using FCAVBAM(5′-CGGGATCCTTGTTGGATCTGTTCCTAGAAC-3′) and FCAVHIND(5′-TTCATCGATTCAGCAAATCTAACAATCAGAGG-3′) primers and cloned into the TRV-00 vector as a BamHI-HindIII fragment(Ratcliff et al., 2001). The potato (Solanum tuberosum) EST BI176637 provided StFYSolanaceous sequence. StFY sequence was PCR amplified using FYVBAM (5′-CGGGATCCAGGACAGTGTTACAACCTAGC-3′) and FYVHIND (5′-TTCATCGATTCTCGTATTGATTCTTTATGTGC-3′) primers and cloned into the TRV-00 vector as a BamHI-HindIII fragment. The TRV vectors were transformed into Agrobacterium and used to inoculate young tobacco plants as described previously(Ratcliff et al., 2001).

To analyse gene expression, RNA was extracted from leaves systemically infected with either TRV-00 (empty vector), TRV-FY or TRV-FCA and used to generate cDNA. The expression of FY, FCA and ACTIN was analysed by PCR amplification using the following primers: FYF(5′-ATGATGCGGCAGCCATCTGCATCC-3′), FYR(5′-ACCAGTGACCATCCAGTTATC-3′), FCAF(5′-ATTTGTTGGATCTGTTCCTAG-3′), FCAR(5′-TCTCGTATTGATTCTTTATGTGC-3′), ACTINF(5′-ATGGCAGACGGTGAGGATATTCA-3′), ACTINR(5′-GCCTTTGCAATCCACATCTGTTG-3′). PCR reactions were amplified for 27, 30, 32, 35 and 40 cycles and analysed using agarose gel electrophoresis with ethidium bromide staining. Gels were visualised using a fluorescence scanner (Amersham-Pharmacia).

FY (residues 416-646) sequence was amplified by PCR and cloned into the pET19b vector (Novagen) using the primers FYNDEI(5′-AATCCCAATGTTCTTATGCAGAACC-3′) and FYR(5′-CCGGTATACCTACTGCTGTTGCTGATTGTT-3′). This allowed inducible expression and purification of the FY-CT protein with a 6× histidine tag. After elution from the nickel affinity resin, the FY-CT protein was further purified using SDS-PAGE (8% acrylamide) and electroelution (BioRad). The FY-CT antigen was then concentrated using Centricon spin columns (Amicon)and dialysed against PBS. FY-CT was used to immunise two female New Zealand White rabbits according to standard procedures(Harlow and Lane, 1988). Specific FY antibodies were purified from FY cross-reactive serum using affinity purification techniques (AminoLink, Pierce). Purified antibodies were concentrated, dialysed against PBS, supplemented with 10 mg/ml BSA and 0.1%sodium azide as a preservative and stored at –80°C. The FY antibody was used at a concentration of 1:1000 according to standard procedures.

Interaction of FCA and FY proteins was tested using the in vitro GST pull-down assay previously described(Simpson et al., 2003). FY was subcloned as either WD (residues 1-415) or CT domains (residues 416-646). The FY-WD subclone was generated by PCR amplification using the WDF (5′-GGAATTCAATAAACCATGTACGCCGGCGGCGATATG-3′) and WDR (5′-CGGGATCCCTAATCTCGGGGATTATCTGC-3′) primers and cloning the PCR product under the T7 promoter in pBLUESCRIPT IISK–. The FY-CT subclone was generated by PCR amplification using the CTF (5′-GGAATTCAATAAACCATGGTTCTTATGCAGAACCAAGGC-3′) and CTR (5′-CGGGATCCCTACTGATGTTGCTGATTGTTG-3′) primers and cloning the PCR product under the T7 promoter in pBLUESCRIPTIISK-. The FY-CT subclone was then mutagenised to generate the PPLPP→AAAAA mutants using the Quikchange method (Stratagene). PPLPP-1 was mutagenised in two steps using the PA1A1(5′-CCATGGCACTGGGGGCTGCTGCTGCGGCACCTGGTCCCCACCCATCG-3′) and PA1A2(5′-CGATGGGTGGGGACCAGGTGCCGCAGCAGCAGCCCCCAGTGCCATGG-3′) oligos first, followed by the PA1B1(5′-GGGGCTGCTGCTGCGGCAGCTGCTGCCCACGCATCGCTTCTTGGAAGTGGC-3′) and PA1B2(5′-GCCACTTCCAAGAAGCGATGCGTGGGCAGCAGCTGCCGCAGCAGCAGCCCC-3′)oligos. PPLPP-2 was mutagenised using the PA21(5′-AACAACCCTTTCCAACAGCAGGCAGCTGCAGCTGCTGGCGCTGCACCAAACAACAATCAGCAAC-3′)and PA22(5′-GTTGCTGATTGTTGTTTGGTGCAGCGCCAGCAGCTGCAGCTGCCTGCTGTTGGAAAGGGTTGTT-3′)oligos.

Proteins with strong similarity to FY can be identified in many eukaryotes(Fig. 1). However, the similarity with FY is restricted to the seven N-terminal WD repeats(Fig. 1). In addition to WD repeats many FY-like proteins have extended C-terminal domains that are not highly conserved in sequence between different eukaryotes(Fig. 1). These regions typically consist of low complexity sequence but in some cases carry defined motifs. For example, the human and mouse WDC146 (HsWDC146 and MmWDC146) C termini display similarity to collagen(Ito et al., 2001)(Fig. 1). The FY C terminus carries two PPLPP motifs, the proline-rich consensus ligand for group-II WW domains such as that present in FCA (Sudol and Hunter, 2000) (Fig. 1). Binding assays were utilised in order to test whether this domain mediates recruitment of FCA to FY polyadenylation complexes directly in vitro.

The C terminus of FCA containing the WW domain was fused to glutathione S-transferase (GST) and purified from E. coli as a recombinant protein (Simpson et al.,2003). FY translated in vitro in the presence of[³⁵S]methionine was then tested in a pull-down assay for interaction with GST-FCA. Both the seven N-terminal WD repeats (FY-WD) that mediate protein-protein interactions(Smith et al., 1999)(Fig. 2B) and the C terminus of FY (FY-CT) were tested. FY-CT, but not FY-WD, was found to have strong interaction with GST-FCA (Fig. 2A). We previously demonstrated that FY fails to interact in vitro with mutated GST-FCA-WF protein, and this mutation was also found to block interaction with FY-CT (Fig. 2A) (Simpson et al.,2003). The WF mutation results in the substitution of the second conserved tryptophan of the FCA-WW domain by a phenylalanine and abolishes FCA function in vivo (Simpson et al.,2003). Hence, the FY-CT domain interacts with FCA in a WW domain-dependent manner. To investigate a binding requirement for the FY-PPLPP repeats, site-directed mutagenesis was performed. Structural work indicates that a proline-rich backbone is important for group-II WW domain binding and both FY-PPLPP repeats occur embedded in polyproline tracts(Macias et al., 2002). The PPLPP repeats and flanking prolines were mutated to alanines independently and together to generate single and double PPLPP→AAAAA mutants(Fig. 2D). Interaction assays revealed that mutation of either PPLPP repeat alone did not prevent FY-CT binding to GST-FCA in vitro (Fig. 2C). However, the double PPLPP→AAAAA mutant FY-CT showed no interaction (Fig. 2C). This demonstrates that both PPLPP within the FY-CT are capable of interaction with the FCA WW domain in vitro. Hence, the FY C-terminal domain mediates recruitment of the RNA-binding protein FCA.

The FY WD repeats and C terminus are distinct in their degree of conservation throughout eukaryotes (Fig. 1). To investigate the requirement of these domains for FY function in vivo, an allelic series of fy mutations was characterised. Forward and reverse genetics provided four fy alleles. The fy-1 mutation was isolated from an EMS screen for late-flowering mutants (Koornneef et al.,1991). Sequencing revealed a splice-acceptor mutation at exon 16 in fy-1 (Simpson et al.,2003) (Fig. 3A). Mutations caused by insertion of T-DNA can be isolated from the Syngenta SAIL(Syngenta Arabidopsis Insertion Library) collection(Sessions et al., 2002). This collection provided the fy-2 allele, which carries a T-DNA insertion within exon 16 (Fig. 3A). The fy-3 and fy-4 mutations were isolated by the Arabidopsis TILLING project(McCallum et al., 2000). TILLING allows reverse genetic isolation of EMS-induced mutations in a gene of interest. These fy alleles affect the first FY-WD repeat and introduce glycine to serine (G141S) and tryptophan to stop-codon (W150*)substitutions, respectively (Fig. 3A). Together these alleles provide mutations in both the conserved FY WD repeats and the C-terminal domain.

The fy-1, fy-2 and fy-3 mutations are viable and confer late flowering (Fig. 3B). Northern blotting and hybridization reveal that these alleles misregulate FLC, the degree of which correlates well with the severity of the late-flowering phenotype (Fig. 3C and Table 1). The fy-2 mutation has the strongest delay in flowering and the highest level of FLC expression(Fig. 3C and Table 1). FLCexpression levels in fy-3 and fy-1 are both lower(Fig. 3C and Table 1). Although fy-2 and fy-3 flower later than fy-1, a direct comparison of their flowering phenotypes is complicated by differences in genetic backgrounds. The Ler (fy-1, fca-1) accession has a much weaker FLC allele relative to Col (fy-2, fy-3, fca-9),caused by the insertion of a Mutator-like element within FLC intron 1(Gazzani et al., 2003; Michaels et al., 2003). The Ler and Col genotypes are rapid-cycling accessions in which the fy and fca mutations were isolated. Within the Col background fy-2 and fca-9 showed a similar delay in flowering time and an increase in FLC expression(Fig. 3C and Table 1). In contrast, fy-1 shows a much weaker delay in flowering than fca-1,which is also in the Ler background(Fig. 3C and Table 1). Hence, strong fy flowering-time alleles are comparable with fca in their effects on FLC expression, and fy-1 is likely to be a weak allele.

In order to determine whether the late-flowering phenotype of fyoccurs through a specific effect on FLC regulation or misregulation of multiple transcripts, an fy-2 flc-3 double mutant was generated. The flc-3 mutation is a null allele and can suppress the late flowering of mutants in the autonomous pathway and FRI-containing lines (Michaels and Amasino,2001). The flc-3 mutation was found to suppress fy-2 late flowering (Fig. 4 and Table 1). A one-way ANOVA shows these differences to be significant (P<0.001). This demonstrates that the effects of fy-2 on flowering time are a specific effect on FLC regulation and not a pleiotropic consequence of defects in RNA metabolism. Furthermore, after a 6-week vernalization treatment to repress FLC expression, the fy mutants flowered early and closely resembled wild type (Fig. 3B and Table 1).

To investigate whether the late-flowering fy alleles were null mutations, the expression of FY mRNA and FY protein was analysed. Northern blot analysis showed reduced abundance of FY mRNA in fy-1 and the presence of transcripts varying in size(Fig. 3C). This size variation is likely to be due to the utilization of multiple, cryptic splice-acceptor sites and potentially exon skipping(Simpson et al., 1998). Sequencing revealed the presence of premature stop codons in mutant fy-1 mRNAs, which would disrupt expression of the C-terminal domain(Simpson et al., 2003). No FY protein was found to accumulate in fy-1(Fig. 3D), however, the FY antibody was raised using the C-terminal domain as an antigen. A translation product of mutant fy-1 mRNA would only overlap by approximately seven amino acids with the peptide used to raise antibodies. Hence, although no FY protein is detectable in fy-1, this may be due to a lack of expressed epitopes rather than it being a null allele. The possibility of fy-1being a hypomorphic allele is also supported by its weak effect on FLC expression and flowering time(Fig. 3C and Table 1). The T-DNA in fy-2 is inserted within exon 16, 320 nucleotides downstream of the fy-1 mutation. Although FY mRNA is reduced in fy-2(Fig. 3C), the 3′ T-DNA insertion site means a truncated protein of 62.4 kDa could be produced,encoding the conserved WD repeats but lacking an intact C-terminal domain. Western blot analysis revealed the presence of this truncated FY protein in fy-2 (Fig. 3D). Hence,it appears that in both fy-1 and fy-2 the FY C-terminal domain is disrupted and this results in FLC misexpression. Disruption of this domain is likely to impair recruitment of FCA to FY complexes in vivo because of loss of the PPLPP repeats.

Genetic evidence supporting the function of the FY WD repeats in flowering time is provided by the fy-3 (G141S) substitution allele. Western blotting reveals that mutant fy-3 protein accumulates normally(Fig. 3D). Indeed, serine is a relatively common alternative residue at this position within WD repeats and would not be expected to cause severe misfolding(Smith et al., 1999). Although the C-terminal domain is severely disrupted in fy-1 and fy-2, the fy-3 allele introduces only a minor change to the WD repeats.

Screening performed by the Arabidopsis TILLING project isolated the null fy-4 mutation (McCallum et al., 2000). This allele carries a premature stop codon at the end of the first FY-WD repeat (Fig. 3A). The tryptophan affected is the signature W residue at the end of the first WD repeat and would lead to expression of a severely truncated protein. A dCAPs (derived Cleaved Amplified Polymorphic Sequence) marker designed against this mutation identified fy-4 heterozygote individuals (Fig. 5J) (Michaels and Amasino, 1998; Neff et al., 1998). However, self-fertilization of fy-4 heterozygotes yielded neither late-flowering plants nor fy-4 homozygotes, raising the possibility that this allele is lethal. Indeed, the siliques of fy-4/FY plants displayed high levels of seed abortion (Fig. 5A). To analyse the contribution of gametophytic defects to this abortion, fy-4/FY plants were crossed reciprocally to Col and the F₁ progeny genotyped for the presence of fy-4heterozygotes. In the absence of any gametophytic defect, 50% of the F₁ should be fy-4 heterozygous. If either fy-4gametophyte is lethal, then no transmission will be observed. Crosses using fy-4/FY as either pollen donor or acceptor demonstrated normal transmission of fy-4 through both gametophytes(Table 2). We then analysed seed development in fy-4/FY siliques. Plants were fixed, cleared and analysed by differential interference contrast (DIC) microscopy. Siliques were harvested when wild-type seed contained embryos at globular or heart stage(Fig. 5B,D). Alongside their healthy siblings, mutant individuals had aborted development at a very early stage post-fertilization, with the embryo only undergoing a few mitotic divisions (Fig. 5C,E). Hence,disruption of the conserved FY-WD repeats causes a failure in embryogenesis and reveals that FY is an essential gene in Arabidopsis,consistent with a conserved function in constitutive RNA 3′-end processing.

An fy-4 heterozygote was also crossed to fy-2 mutant plants. The F₁ from this cross was genotyped to find individuals carrying both the fy-2 and fy-4 alleles and their flowering time analysed. Consistent with fy-4 being a null allele and fy-2 providing a partially functional, truncated gene product these fy-4/fy-2 plants were viable yet late flowering(Table 1). To demonstrate that the lethality observed in fy-4 was due to disruption of FYwe performed a complementation experiment. A transgene was constructed to express the FY open reading frame (cDNA) under control of the endogenous FY promoter and 3′-UTR (pFY::FY). This transgene was transformed into fy-4/FY heterozygotes and the T₁ progeny were genotyped for the inheritance of fy-4. Among 133 T₁ transformants 44 were FY/FY, 78 were FY/fy-4 and 11 were fy-4/fy-4 homozygotes(Fig. 5J). As self-fertilization of fy-4/FY heterozygotes has never produced fy-4/fy-4 homozygotes we conclude that complementation was observed.

Pleiotropic functions for FY in development were suggested previously by genetic analysis within the autonomous pathway(Koornneef et al., 1998). An fy fpa double mutant was never recovered, suggestive of lethality. In contrast, both fca fy and fca fpa double mutants are viable and late flowering (Koornneef et al.,1998). To provide further insight into these interactions,reproductive defects in plants heterozygous for fy-1 and homozygous for fpa-1 mutations were analysed. The siliques of fy-1/FY,fpa-1/fpa-1 plants displayed a high incidence of aborted seed(Fig. 5F). To analyse possible gametophytic defects mutant plants were crossed reciprocally to Lerplants. Significant transmission of fpa-1 fy-1 gametes was observed through both crosses, indicating that gametophytic defects are unlikely to be the major cause of fy fpa lethality(Table 2). However, less than 50% of the F₁ progeny were fy-1 heterozygous when plants were used as either pollen-donor or acceptor(Table 2), meaning that although fy-1 fpa-1 gametes participate in fertilization, they have reduced vigour or viability. Analysis of seed development in fy-1/FY,fpa-1/fpa-1 plants revealed that mutant seed had ceased dividing and aborted very early after fertilization, with a similar phenotype to that observed in fy-4 (Fig. 5H). Furthermore, examination of fy-1, fpa-1 and fpa-2 single mutants revealed elevated levels of seed abortion relative to Ler (Fig. 5I). In contrast, fca-1 siliques were indistinguishable from wild type and fca-1 did not enhance the defect observed in fy-1 in the fca-1 fy-1 double mutant(Fig. 5I). A one-way ANOVA shows these differences to be significant (P<0.001). Hence, fy-1 and fpa-1 have weakly penetrant defects in seed development, which combine to cause synergistic lethality. These interactions are likely to reflect essential FY functions. The fact that fca-1lacks these phenotypes indicates that FCA acts more specifically in development than FY and FPA.

The embryo lethality of null fy-4 mutations precludes analysis of loss of FY function later in development. Virus-induced gene silencing (VIGS) allows conditional silencing of target genes in Nicotiana benthamiana (Lu et al.,2003). Infection of N. benthamiana with a tobacco rattle virus (TRV) vector carrying target gene sequences leads to gene silencing of the endogenous target mRNA (Ratcliff et al., 2001). This system was utilised to assay the effect of silencing FCA and FY expression during post-embryonic development. FCA and FY are well conserved throughout higher plants and BLAST searches identified several potato and tomato EST sequences with strong nucleotide similarity to AtFCA and AtFY(Macknight et al., 1997; Simpson et al., 2003). Within the Solanaceae the sequences homologous to FCA and FY were>90% identical, making them suitable for use in TRV VIGS in N. benthamiana (Lu et al.,2003). Potato (StFCA) and tomato (LeFCA) FCA sequences were cloned into a TRV vector and called TRV-FCA and TRV-FY respectively. The LeFCA sequence was very similar to AtFCA sequence from exons 3 and 4, which spans the alternatively processed FCA intron 3. The four alternative FCA mRNAs contain exon 3 sequence and hence VIGS against this sequence should target all FCA transcripts (Macknight et al., 1997; Macknight et al.,2002).

Young leaves of N. benthamiana plants were innoculated with Agrobacterium carrying the TRV binary vectors. Transformation and expression of the inserts leads to systemic infection of the plant with TRV,and VIGS-derived phenotypes in newly emerging leaves(Ratcliff et al., 2001). Plants were infected with TRV-00 empty vector as a negative control(Ratcliff et al., 2001). The TRV-00-infected plants showed no obvious phenotypes and infection was largely asymptomatic. The TRV-FCA infected plants were similar to the TRV-00 plants and lacked any obvious phenotype (Fig. 6A,B). The TRV-FY-infected plants exhibited strong phenotypes around 14 days post inoculation (dpi). The systemically infected, emerging leaves displayed yellowing and were distorted(Fig. 6C,D). After a further week of growth, the stature of TRV-FY-infected plants was reduced. These phenotypes are consistent with a reduction in cell viability in the TRV-FY-infected tissues. This supports the conclusion that FY, but not FCA, is required generally for plant growth and development.

Although silencing of FY in Nicotiana induced deleterious phenotypes, the infected plants continued to grow and produce leaves and flowers. However, VIGS does not completely eliminate target mRNAs and residual levels of FY mRNA may be sufficient for continued growth(Lu et al., 2003; Ratcliff et al., 2001). To investigate the extent of silencing induced by TRV infection, total RNA was extracted from systemically infected tissue and analysed for target gene expression using RT-PCR (Fig. 6E). Expression was assayed relative to TRV-00 infected plants and ACTIN mRNA levels, used as an internal control. Both the TRV-FY and TRV-FCA viruses induced silencing of their target endogenous gene, though in both cases residual levels of mRNA remained(Fig. 6E). The remaining FY mRNA may provide sufficient FY expression to support the continued growth.

The failure of seed development in fy-4 mutants reveals that FY is required for embryogenesis. To determine whether this reflects expression of FY during seed development, a reporter transgene was constructed. 1.3 kb of genomic promoter sequence was used to generate a FY::GUS transgene. Transformation with FY::GUS was performed in the Col background and homozygous lines were analysed for GUS expression. FY::GUS siliques were harvested at progressive stages of maturity and seeds histochemically stained for GUS expression(Fig. 7A-C). At globular stage,strong GUS staining was evident throughout the embryo, endosperm and surrounding maternal seed tissues (Fig. 7A). From heart-stage of embryogenesis onwards GUS expression was more restricted to the embryo (Fig. 7B,C), Additionally, the funiculus connecting the seed to the silique showed strong staining, which extended into the chalazal base of the seed (Fig. 7B). To confirm the FY::GUS expression pattern we also performed in situ hybridization experiments. Sectioning of embryos and hybridization with an antisense FY probe revealed expression in globular, heart and torpedo stage embryos (Fig. 7K-M). No signal was detected when a sense probe was used for hybridization(Fig. 7N). Therefore,consistent with the failure of embryo and endosperm development in fy-4 mutants, the FY promoter is active in these tissues.

The function of FY in regulating flowering time also leads to an expectation of post-embryonic expression. FY::GUS expression was analysed in 12-day-old seedlings. Strong GUS expression was evident in the meristematic regions of the shoot and the root(Fig. 7D-7F). High expression was clear in the shoot meristem and the adjacent young leaves(Fig. 7D). In cotyledons and older leaves GUS expression was only evident in the vasculature(Fig. 7D). In situ hybridization using an antisense FY probe confirmed expression in shoot apical meristems, young leaves and vasculature(Fig. 7O-Q), while no signal was detected when a sense FY probe was used(Fig. 7R). In the root, both the apical and emergent lateral meristems showed strong expression(Fig. 7E,F). The FY::GUS line with strongest levels of expression also showed staining in differentiated, non-proliferating tissues (data not shown). However,expression was still much weaker relative to meristematic and vascular regions. This expression pattern is closely related to the majority of genes defined to regulate vernalization requirement and response, including FLC (Macknight et al.,2002; Schomburg et al.,2001; Sheldon et al., 2002; He et al., 2003) (C. Lister and C. D., unpublished data).

FY expression in meristematic regions may reflect general functions in proliferating tissue or developmental control of its target gene, FLC. In addition to regulating FLC, FCA and FY also mediate FCA negative autoregulation(Quesada et al., 2003;Simpson, 2003). The P_FCA–FCA_{to exon5}:GUStransgene consists of the FCA gene with GUS inserted downstream of intron 3 (Macknight et al.,2002). FCA-mediated polyadenylation within P_FCA–FCA_{to exon5}:GUS intron 3 leads to an absence of GUS expression. Thus, the expression of this transgene can be used as an in vivo reporter of FCA autoregulation and FCA/FY activity (Macknight et al.,2002; Quesada et al.,2003). In a Ler background P_FCA–FCA_{to exon5}:GUS expression is restricted to the meristematic regions of the shoot and root and is upregulated 6 days after germination(Macknight et al., 2002; Quesada et al., 2003)(Fig. 7G). In fca-1,P_FCA–FCA_{to exon5}:GUS expression is earlier and in a broader pattern (Fig. 7H), which led to the conclusion that FCA normally limits its own expression through intron 3 autoregulation, but this mechanism is inefficient in meristematic regions (Quesada et al.,2003). Crossing P_FCA–FCA_{to exon5}:GUS into fy mutant backgrounds should also lead to spatial and temporal upregulation of transgene GUS expression. P_FCA–FCA_{to exon5}:GUS was crossed into the fy-1 and fy-2 mutants and GUS expression analysed. The P_FCA–FCA_{to exon5}:GUS fy-2 line showed pronounced GUS staining in a pattern similar to P_FCA–FCA_{to exon5}:GUS fca-1(Fig. 7I). The P_FCA–FCA_{to exon5}:GUS fy-1 line also showed increased GUS staining relative to P_FCA–FCA_{to exon5}:GUS Ler, though to a lesser extent than that seen in fy-2 and fca-1 backgrounds(Fig. 7J). The weak effect of fy-1 on P_FCA–FCA_{to exon5}:GUSexpression reflects its effects on FLC and flowering time(Fig. 3C and Table 1). Hence, FY functions in differentiated cells to promote FCA alternative polyadenylation. However, in meristematic regions this regulation is inhibited, allowing higher levels of FCA expression.

FY was originally identified during a screen for late-flowering mutants (Koornneef et al.,1991). We demonstrate here that FY also plays a role in cell viability and these functions are performed differentially by distinct domains of FY. The N terminus of FY is highly conserved with the essential yeast polyadenylation factor Pfs2p and disruption of this domain in fy-4 causes embryo lethality. In contrast, the plant-specific FY C-terminal domain appears to be required for more limited functions and from a phenotypic analysis just for control of flowering time. The C terminus contains the binding sites for FCA and thus its disruption probably prevents FCA/FY interaction in vivo. The fy-3 substitution allele demonstrates that intact WD repeats are also required for repression of FLC. WD repeats consist of four β-sheets connected by loop regions, with each WD repeat forming a propeller blade (Smith et al., 1999). The glycine (G141) residue affected in fy-3is predicted to occur in a structural residue of the B-β-strand in the first propeller blade (Smith et al.,1999). However, the stability of mutant fy-3 protein makes it unlikely that its effect is via misfolding and destabilisation(Holm et al., 2001; McNellis et al., 1994). Therefore, the G141S substitution may have a specific effect on FY-WD interactions, which reduces FLC repression. A direct test of this hypothesis awaits the identification of FY-WD interacting proteins.

The homology of FY to the Pfs2p polyadenylation factor presents a hypothesis for the cause of fy-4 lethality. Polyadenylation factors are generally essential proteins in eukaryotes and the lethality of such mutations in yeast reflects an inability to correctly process 3′ ends of transcripts (Zhao et al.,1999; Dheur et al.,2003; Ohnacker et al.,2000; Wang et al.,2005). Additionally, loss of the Drosophila Su(f)polyadenylation factor leads to cell-autonomous defects in proliferation and viability (Audibert and Simonelig,1999). FY essential functions may reflect a conserved role in general RNA 3′-end processing. However, although the WD repeats of FY and Pfs2p are highly homologous, FY fails to complement PFS2function in S. cerevisae (I.R.H. and C.D., unpublished data). Hence,some aspects of their function appear to have diverged. An alternative possibility is that FY could function to regulate a subset of RNAs, one or several of which are essential.

Although FY may function generally in RNA processing, it is possible that it performs a more specialised role in regulated 3′-end processing. In vitro study of polyadenylation has commonly used constitutively utilised 3′-end processing signals (Zhao et al., 1999). However, there is abundant evidence that regulated 3′-end processing occurs in vivo(Beaudoing and Gautheret, 2001; Edwalds-Gilbert et al., 1997). The cis signals and trans factors mediating alternative polyadenylation are poorly understood, though polyadenylation site choice during FCAautoregulation represents one instance(Macknight et al., 2002; Quesada et al., 2003). With respect to FY function in constitutive versus regulated polyadenylation it is important to consider potential redundancy within 3′-end processing complexes (Keller and Minvielle-Sebastia,1997). Pfs2p and a second polyadenylation factor, CstF-50, are proposed functional orthologues based on their domain organisation and similar protein interactions (Ohnacker et al.,2000; Takagaki and Manley,1992). S. cerevisae is unusual relative to other eukaryotes in encoding only a Pfs2p-like protein. The presence of both CstF-50-like and Pfs2p-like proteins in other eukaryotes may have facilitated functional divergence of Pfs2p-like proteins. Indeed, Pfs2p-like proteins sequenced from eukaryotes other than S. cerevisae display unusual features. The acquisition of distinct C-terminal domains is evident in many FY/Pfs2p homologues (Fig. 1). By analogy, these domains may function as binding sites for trans-regulators of 3′-end processing, similar to FCA. A developmental function for the mammalian FY/Pfs2p homologue, WDC146, is probably the result of its specific expression pattern during spermatogenesis and its absence from constitutive polyadenylation complexes (Ito et al.,2001; Zhao et al.,1999). The functions of FY/Pfs2p proteins in regulated polyadenylation may be addressed by searching for proteins interacting with these C-terminal domains. It will be interesting to determine whether distinct FY polyadenylation complexes mediate flowering time and essential functions,and furthermore whether this reflects different roles in regulated versus constitutive RNA 3′-end processing.

We thank the Arabidopsis TILLING Project, David Baulcombe, Maarten Koornneef, Richard Macknight and Syngenta for providing materials and helpful advice. We thank Victor Quesada for providing the P_FCA–FCA_{to exon5}:GUS line in fca-1 and fy-1 mutant backgrounds and John Doonan for helpful discussions on FY expression. This work was supported by the Gatsby Charitable Foundation (I.R.H.) and the Biotechnology and Biological Sciences Research Council (BBSRC) Core grant to the John Innes Centre (supporting G.G.S. and C.D.) and BBSRC grant 208/G17507 (supporting F.L.).