The developmental roles of AGL15 and AGL18, members of the AGL15-like clade of MADS domain regulatory factors, have not been defined previously. Analysis of transgenic Arabidopsis plants showed that overexpression of AGL18 produces the same phenotypic changes as overexpression of AGL15, and the two genes have partially overlapping expression patterns. Functional redundancy was confirmed through analysis of loss-of-function mutants. agl15 agl18 double mutants, but not single mutants, flower early under non-inductive conditions, indicating that AGL15 and AGL18 act in a redundant fashion as repressors of the floral transition. Further genetic analyses and expression studies were used to examine the relationship between AGL15 and AGL18 activity and other regulators of the floral transition. AGL15 and AGL18 act upstream of the floral integrator FT, and a combination of agl15 and agl18 mutations partially suppresses defects in the photoperiod pathway. agl15 agl18 mutations show an additive relationship with mutations in genes encoding other MADS domain floral repressors, and further acceleration of flowering is seen in triple and quadruple mutants under both inductive and non-inductive conditions. Thus, flowering time is determined by the additive effect of multiple MADS domain floral repressors, with important contributions from AGL15 and AGL18.
We are interested in the role that MADS domain factors play during early stages of the life cycle. We performed a family-wide RT-PCR survey and identified several clades of MADS box genes of which every member is expressed in embryos. These include the SEPALLATA genes, FLOWERING LOCUS C (FLC)-like genes, AGAMOUS-like 15 (AGL15)-like genes, and the MIKC* genes (Lehti-Shiu et al., 2005).
Members of the FLC-like group, including FLC, FLM/MAF1 and MAF2-5, play important roles as regulators of the floral transition. There are three major pathways in Arabidopsis that control this transition. The photoperiod, vernalization and autonomous pathways control flowering time by measuring daylength, exposure to cold, and endogenous factors, respectively (reviewed by Simpson et al., 1999; Amasino, 2004). Arabidopsis is a facultative long-day plant, and light-mediated expression changes allow it to measure daylength. Under inductive (LD, long-day) conditions, photoperiod pathway components act to promote flowering by inducing CONSTANS (CO) and downstream genes. Winter annual varieties of Arabidopsis are very late-flowering unless they are exposed to extended cold periods, due to operation of the vernalization pathway. The MADS domain factor FLC is upregulated in these varieties; however, after exposure to a sufficiently long cold period, FLC is repressed through histone de-acetylation and methylation (Bastow et al., 2004; Sung and Amasino, 2004). The autonomous pathway allows constitutive repression of FLC (Michaels and Amasino, 2001). Each of these pathways ultimately has an impact on the expression of downstream genes that determine meristem identity, including FLOWERING LOCUS T (FT), LEAFY (LFY) and two other MADS box genes, SOC1 and APETALA1.
Genetic analyses indicate that the MADS domain factors in the FLC-like group act as repressors of the floral transition. Mutants carrying loss-of-function flc alleles are early-flowering (Koornneef et al., 1994; Lee et al., 1994; Michaels and Amasino, 1999). The effect is particularly striking in winter-annual varieties, where flc mutations eliminate the vernalization requirement. The effects of flm/maf1 mutations are most easily discerned under non-inductive photoperiod (SD, short-day) conditions. FLM/MAF1 is thought to act as a co-regulator with the MADS domain factor SVP, and both proteins may function in the same ternary complex (Scortecci et al., 2003). flm/maf1 and svp mutants are early-flowering (Hartmann et al., 2000; Scortecci et al., 2001). The four remaining MAF (MADS affecting flowering) factors are also thought to act as floral repressors, and their activity may become important for flowering-time control under particular environmental conditions (Ratcliffe et al., 2003).
We have shown previously that AGL15 and AGL18, which constitute the AGL15-like group, are expressed both in embryos and developing endosperm (Lehti-Shiu et al., 2005). Constitutive expression of AGL15 supports the maintenance of embryonic identity and enhances the formation of somatic embryos from the shoot apical meristem in culture (Harding et al., 2003). Plants that constitutively overexpress AGL15 exhibit a whole host of phenotypic changes, including altered leaf morphology, reduced fertility, delayed flowering, and delayed floral organ abscission and senescence (Fernandez et al., 2000). Based on these phenotypic changes, AGL15 appears to act as a repressor to slow the progression of a specific subset of age-related developmental programs. However, loss-of-function mutants show no phenotypic changes, either in embryos or after germination (Lehti-Shiu et al., 2005). One possible explanation is that loss-of-function effects are masked through genetic and/or functional redundancy.
In the first part of this work, we used three approaches to look for evidence of overlap in regulatory function between AGL15 and its closest relative AGL18. First, we generated AGL18 overexpressors and compared their phenotype to that of AGL15 overexpressors. Secondly, we compared the expression patterns of AGL15 and AGL18. Third, we tested for genetic redundancy by generating and analyzing phenotypic changes in agl15 agl18 double mutants. Through these experiments, we have uncovered a role for AGL15 and AGL18 in repression of the floral transition.
In the second part of this work, we used genetic analyses and expression studies to examine the relationship between AGL15 and AGL18 and other regulators of the floral transition. We found that AGL15 and AGL18 act upstream of the floral integrators and in a partially redundant fashion with other MADS domain factors previously identified as floral repressors. Flowering time is determined by the additive effect of multiple sets of MADS domain floral repressors, including AGL15 and AGL18.
Effects of AGL18 overexpression
If AGL15 and AGL18 perform similar functions, plants that constitutively overexpress AGL18 might be expected to show some of the same phenotypic changes as plants that overexpress AGL15. To investigate this, we generated 34 lines of transgenic plants that express genomic AGL18 sequence under the control of the CaMV 35S gene promoter. Eleven of the lines showed pronounced phenotypic changes. As with AGL15 overexpressors (Fernandez et al., 2000), the rosette leaves of AGL18 overexpressors had shortened petioles, and were often mildly serrated and wavy or curled (Figure 1a). Under our inductive LD growth conditions, the transgenic plants produced an average of 9.5 rosette leaves (means for three individual lines: 7.0 ± 0.9, 9.8 ± 1.3, 11.7 ± 2.0) before flowering, while wild-type plants (Ws background) produced an average of 6.3 ± 0.5 rosette leaves before flowering. This represents an increase of approximately 50% in the length of the vegetative phase. Many of the AGL18 overexpressors showed reduced fertility, and, as is common when fertility is affected, many more flowers and secondary inflorescences were produced than in wild-type (not shown). AGL18 overexpressors also showed prolonged longevity and retention of perianth organs (Figure 1c–e). Thus, like AGL15, AGL18 has the capacity to delay floral organ senescence and abscission as well as flowering time when expressed at elevated levels.
AGL15 and AGL18 could achieve similar overexpression effects because of biochemical similarities; alternatively, their expression could be coordinated such that overexpression of one factor leads to overexpression of the other. To distinguish between these two possibilities, we analyzed selected lines by RT-PCR. We found that AGL15 transcript levels are not elevated in response to increased expression of AGL18, and AGL18 transcript levels are not elevated in response to increased expression of AGL15 (Figure 1f). We conclude that biochemical similarity is more likely to explain the similar overexpression phenotypes than regulatory interactions.
AGL15 and AGL18 expression patterns
If AGL18 and AGL15 act redundantly in planta, they should have at least partially overlapping expression patterns. We generated 16 lines of transgenic plants expressing an AGL18–GUS translational fusion and nine lines expressing an AGL15–GUS translational fusion under the control of their respective native promoters. We have used these lines previously to show that both AGL15 and AGL18 accumulate in embryos and endosperm following fertilization (Lehti-Shiu et al., 2005). In young seedlings, AGL15–GUS was expressed in the shoot and root apices, lateral root primordia and throughout the vascular system (Figure 2a,c) in multiple lines. AGL15 promoter activity and accumulation of AGL15 protein in the shoot apex of young seedlings has been reported previously (Fernandez et al., 2000). Expression in the vascular system was not observed previously for a ProAGL15:GUS reporter construct (Fernandez et al., 2000), suggesting that the coding sequence and/or introns contain important regulatory elements for expression in the vascular system. The AGL18–GUS fusion protein was expressed in a more constitutive pattern. In young seedlings, AGL18–GUS was expressed everywhere except in a portion of the hypocotyl and in newly emerging leaves (Figure 2b,d). During the reproductive phase, AGL18–GUS, like AGL15–GUS, was expressed in immature buds and at the base of the floral organs, and in the receptacle, ovules, anther filaments and stigma–style of open flowers, as shown in Figure 2(e–h). However, unlike AGL15–GUS (Figure 2i), AGL18–GUS also accumulates in pollen (Figure 2j) and seed coats (not shown). Therefore, the overlap between the expression patterns of the two proteins is extensive, with AGL15 expressed in a subset of the tissues that express AGL18. Cells that contain AGL15 are also likely to contain AGL18, which contributes to the potential for redundant action.
Roles in regulation of flowering time
To identify possible regulatory roles, we grew agl15 and agl18 single mutants and agl15 agl18 double mutants under a variety of conditions. To determine whether AGL15 and AGL18 factors might be involved in the regulation of flowering time, we grew the mutants under both LD (16 h light/8 h dark) and SD (8 h light/16 h dark) conditions. Multiple allele combinations were tested, using the loss-of-function T-DNA insertion alleles (Col background) previously identified and described by Lehti-Shiu et al. (2005).
Under LD conditions, no differences in flowering time relative to wild-type (Col) could be detected for agl15 or agl18 single mutants or agl15 agl18 double mutants (Figure 3). Under SD conditions, no differences could be detected for agl15 or agl18-1 single mutants. agl18-2 mutants are likely to produce AGL18 factors with altered I-domains (Lehti-Shiu et al., 2005), which may have a dominant negative effect, and these mutants showed accelerated flowering relative to wild-type. All agl15 agl18 double mutant combinations flowered significantly earlier than wild-type. Apart from the differences in flowering time, agl15 agl18 double mutants were indistinguishable from wild-type. No changes in leaf morphology, perianth organ longevity, fertility, fruit and seed maturation, or other developmental parameters were observed (data not shown).
To confirm that the early flowering under SD was the result of loss of AGL15 and AGL18 function, agl15-3 agl18-1 double mutants were transformed with either a genomic AGL15 construct or a genomic AGL18 construct, each with its native promoter. In these plants, flowering time was similar or even slightly delayed relative to wild-type, as shown in Table 1. We conclude that the early-flowering phenotype is a result of disruption of AGL15 and AGL18 and does not reflect another genetic change in this background.
Table 1. Complementation of the early-flowering phenotype of agl15 agl18 mutants with AGL15 and AGL18 genomic sequences
Wild-type (Col) and transgenic lines were grown under SD conditions. The mean flowering time ±1 standard deviation is shown for each line.
67.63 ± 4.98
53.13 ± 8.99
agl15 agl18 + gAGL15 line 1
72.25 ± 6.90
agl15 agl18 + gAGL15 line 2
67.71 ± 5.21
agl15 agl18 + gAGL15 line 3
65.47 ± 2.92
agl15 agl18 + gAGL18 line 1
74.56 ± 3.61
agl15 agl18 + gAGL18 line 2
75.17 ± 2.40
agl15 agl18 + gAGL18 line 3
69.88 ± 2.80
agl15 agl18 + gAGL18 line 4
74.25 ± 3.66
Interactions with flowering-time regulatory pathways
The analysis of loss-of-function effects indicates that AGL15 and AGL18 act in a redundant fashion in floral repression. To place AGL15/18 activity into the larger context of regulation of flowering time, we combined the agl15-3 and agl18-1 alleles with mutations that affect the autonomous and photoperiod pathways. For the autonomous pathway, we combined luminidependens (ld) mutations, which result in increased FLC expression and delayed flowering under LD conditions (Michaels and Amasino, 2001), with agl15 and agl18 mutations. The agl15 agl18 mutant combination did not suppress late flowering or otherwise alter flowering in the ld background (Figure 4a). Therefore, AGL15 and AGL18 not appear to act downstream of LUMINIDEPENDENS or FLC in the autonomous pathway.
For the photoperiod pathway, we combined agl15 agl18 mutations with gi and co mutations. GI (GIGANTEA) is broadly involved in circadian rhythms, while CO (CONSTANS) is a flowering-specific output of the circadian clock and a direct activator of the floral integrator FT under LD conditions (Samach et al., 2000). gi and co mutants flower late under LD conditions, and produce approximately the same number of leaves as under SD conditions (Koornneef et al., 1991). When agl15 agl18 mutations are introduced into either the gi or co backgrounds, the effect of the defects in the photoperiod pathway is partially suppressed. In co agl15 agl18 triple mutants, flowering time is reduced by approximately 50% relative to co plants, and in gi agl15 agl18 triple mutants, flowering time is reduced by approximately one-third relative to gi plants (Figure 4a). In both cases, the triple mutant plants still flower later than wild-type under LD conditions; therefore, loss of the AGL15 and AGL18 floral repressors cannot completely compensate for loss of inductive light cues. However, the requirement for these inductive cues is decreased, suggesting that there is an interaction between AGL15 and AGL18 and the activity of components of the photoperiod pathway.
co mutants show reduced expression of FT (Kobayashi et al., 1999). To determine whether agl15 and agl18 suppress co through an impact on FT expression, RT-PCR was performed with FT-specific primers and RNA isolated from 6- and 10-day-old seedlings of early-flowering Columbia wild-type, late-flowering co mutants, and intermediate-flowering co agl15 agl18 triple mutants, grown under LD conditions. In each background, accumulation of FT transcript increased in an age-dependent fashion. In the triple mutant, FT transcripts accumulate at elevated levels relative to the co mutant, and approximately wild-type levels of transcript accumulation are found at both 6 and 10 days (Figure 4b). This suggests that the interaction between AGL15 and AGL18 and the photoperiod pathway occurs at or upstream of FT induction.
To determine whether AGL15 and AGL18 are likely to be regulatory targets of CO, RT-PCR was performed on the same RNA samples with AGL15- and AGL18-specific primers (Figure 4b). Expression of AGL15 and AGL18 is unaltered in co mutants; therefore, these plants are not late-flowering because of upregulation of AGL15 or AGL18. We conclude that AGL15 and AGL18 are not components or targets of the photoperiod pathway, but act in a separate pathway that may intersect with the photoperiod pathway at the level of the floral integrator FT. In the absence of the floral repressors AGL15 and AGL18, there is a reduced requirement for CO activity for FT induction.
Genetic interactions with other MADS domain floral repressors
Other MADS domain factors are known to function as floral repressors, including FLC, SVP, FLM/MAF1 and MAF2-5. Like AGL15 and AGL18, the effects of SVP and FLM/MAF1 loss of function are most apparent under SD conditions, and they show interactions with the photoperiod pathway (Scortecci et al., 2003). To determine whether AGL15 and AGL18 act in the same pathway as SVP and FLM, we created double, triple and quadruple mutant combinations. Under LD conditions, svp and flm mutants flower slightly earlier than wild-type. However, because the change is one leaf or less, a significant difference is difficult to detect (Figure 5). Under SD conditions, the transition to flowering is slower and the differences are more easily detected. flm mutants flower early, with approximately two-thirds the number of leaves produced by wild-type (Figure 5). svp mutants flower even earlier, with approximately one-third of the number of leaves produced by wild-type (Figure 5). As reported previously, combining svp and flm mutations does not further accelerate flowering time, which is consistent with possible action as co-regulators (Scortecci et al., 2003).
When agl15 and agl18 mutations were added to backgrounds containing either svp or flm mutations, or the svp flm double mutant combination, flowering time under non-inductive conditions was further accelerated (Figure 5). Plants carrying the triple or quadruple mutant combinations flowered very early, with 11–14 total leaves. The additive interaction suggests that both sets of floral repressors (AGL15 and AGL18, SVP and FLM) make independent contributions to regulation of flowering time.
The MADS domain factor FLC is the major contributor to floral repression in late-flowering winter-annual backgrounds and is the main regulatory target of both the autonomous and vernalization pathways under LD conditions (reviewed by Amasino, 2005). Previous genetic analyses suggested that it is not a major contributor to floral repression under SD conditions (Michaels and Amasino, 2001). We created higher-order mutant combinations to examine genetic interactions between AGL15, AGL18 and FLC. flc mutants flower with approximately the same number of leaves as wild-type under both LD and SD conditions (Figure 6). When agl15 and agl18 mutations are combined with flc mutations, we see an acceleration of the transition to flowering. While the effect is most apparent under SD conditions, the agl15 agl18 flc triple mutant flowers significantly earlier than either set of parents under LD conditions as well. Under both sets of conditions, the interaction is additive; thus, as with SVP and FLM, AGL15 and AGL18 provide an activity that is separate from FLC repressor activity. In the absence of AGL15 and AGL18 activity, the contributions (or potential contributions) of FLC to floral repression under SD conditions become more apparent.
To investigate the relationship between all three sets of floral repressors (AGL15 and AGL18, SVP and FLM, and FLC), we generated an agl15 agl18 svp flc quadruple mutant. Under both LD and SD conditions, the flc svp double mutant flowers with approximately the same number of leaves as the svp single mutant. Thus, svp mutations appear to be largely epistatic to flc mutations, just as they are with flm mutations. When agl15 agl18 mutations are introduced, flowering time is further accelerated under SD conditions. The quadruple mutant flowers earlier than the flc svp double mutant or the agl15 agl18 svp and agl15 agl18 flc triple mutants (Figure 6). The interactions are additive in this case. Thus, the successive removal of floral repressors results in earlier and earlier transitions to flowering, and the requirement for inductive photoperiods is reduced. In the absence of AGL15, AGL18, SVP and FLC, there is less than a four-leaf difference between flowering time under LD and SD conditions.
Analysis of expression of floral integrators
MADS domain factors can form regulatory networks and/or interact as co-regulators in multi-protein complexes. To further position AGL15/AGL18 activity within the context of known floral repression activities, we examined the effect of the loss-of-function mutations on expression of various flowering-time genes. We first confirmed that AGL15 and AGL18 work upstream of the floral integrator FT under non-inductive conditions (Figure 7). We used RT-PCR to examine the effects of agl15 agl18 mutations on the accumulation of transcripts for FT and the MADS domain factor SOC1, which are both known direct targets of FLC (Helliwell et al., 2006; Hepworth et al., 2002). No significant changes in SOC1 levels were detected between 6 and 10 days or in either the single mutant or double mutant seedlings relative to wild-type (Figure 7). FT transcripts could be readily detected only at 10 days, and levels were elevated in the agl15 agl18 double mutants relative to wild-type and the single mutants (Figure 7). This result is consistent both with the early-flowering phenotype of the double mutant and elevated FT expression under LD conditions in the co agl15 ag18 mutant relative to the co mutant.
Analysis of expression of MADS domain floral repressors
To determine whether AGL15 and AGL18 regulate the expression of the other MADS domain floral repressors, we compared the accumulation of various transcripts in wild-type and single and double mutant plants grown under SD conditions. RNA was isolated from 7-day-old seedlings, and semi-quantitative RT-PCR was performed using gene-specific primers. If AGL15 and AGL18 act as positive regulators of floral repressors such as FLC, SVP, FLM/MAF1 or MAF2-5, we expect transcript levels to decrease in agl15 agl18 double mutants. However, no changes could be detected in either single or double mutants relative to wild-type (Figure 8).
To determine whether AGL15 and AGL18 are regulatory targets of the other MADS domain floral repressors, we examined AGL15 and AGL18 transcript levels in svp, flm and flc mutants grown for 6 or 10 days under SD conditions. If these genes were positive regulators of AGL15 or AGL18, we would expect transcript levels to decrease in the mutants relative to wild-type. However, no major changes in AGL15 or AGL18 transcript levels could be detected in these mutant backgrounds (Figure 9), and thus AGL15 and AGL18 do not appear to be regulatory targets or regulators of other MADS domain floral repressors.
AGL15 and AGL18 are structurally related and show overlapping expression patterns
AGL15 and AGL18 are structurally related MADS domain factors that may perform similar regulatory roles in tissues where they are both expressed. Based on the degree of sequence similarity and partial overlap in expression patterns, we might expect both overlapping and unique roles. AGL15 and AGL18 are each other’s closest relatives in the 109-member MADS box family in Arabidopsis, but are only 62% similar overall. Thus, they do not represent a recent duplication, as with the SHP genes, which are 90% similar. In the 57 amino acid MADS domain, AGL15 and AGL18 are 96% similar. Thirteen of the 15 amino acid differences represent conservative changes and may have little impact on target choice or binding specificity. AGL15 and AGL18 also show sequence similarities in the K-domain and C-domain, which have been shown to play important roles in protein–protein interactions in other MADS domain proteins. In particular, both AGL15 and AGL18 C-domains contain a motif of nine amino acids (SDT(T/S)LQLGL) that is conserved across species in AGL15-like clade members. A very similar motif (SDTSLKLGL) is seen at the same position in the STMADS11-like or JOINTLESS-like MADS domain proteins (Becker and Theißen, 2003; Johansen et al., 2002). In Arabidopsis, this clade has two members: AGL24, which acts as a promoter of inflorescence fate and repressor of floral fate (Yu et al., 2004), and SVP, which acts as a floral repressor (Hartmann et al., 2000). Both motifs contain an amino acid sequence (LXLXL) that has been shown to be important for transcriptional repression in auxin response factors (Tiwari et al., 2004). A similar sequence ((L/F)DLN(L/F)XP) is found in EAR (ERF-associated amphiphilic repression) motifs, which have been shown to be essential for repression in a number of zinc finger proteins and are capable of suppressing transcription activation in both inter- and intra-molecular fashion (Ohta et al., 2001). The extensive sequence similarities may form the basis of functional redundancy between AGL15 and AGL18. If they are expressed within the same cell, they may have the capacity to act interchangeably in regulatory complexes. If the conserved LXLXL motif acts in a fashion similar to EAR motifs, then association of either AGL15 or AGL18 with a complex might enhance its ability to repress or convert an activation complex into a repression complex.
The similarity in the phenotypic changes resulting from overexpression is consistent with functional redundancy. Every aspect of the AGL15 overexpression phenotype was seen in plants overexpressing AGL18, including enhanced maintenance of embryonic potential in culture (D.E. Fernandez, unpublished results). Because of their structural similarities, it is possible that AGL15 and AGL18 share a common set of target genes and/or protein interactors when they are expressed ectopically and at elevated levels. The overexpressors show very specific phenotypic changes, largely involving slowed developmental transitions. If there are many components to the transition, as in the floral transition or exit from embryogenesis or floral senescence, then effects may be easier to see in a gain-of-function situation than a loss-of-function situation. The programs that are altered in overexpressors are likely places to look for the ‘normal’ regulatory targets of AGL15 and AGL18.
AGL15 and AGL18 expression patterns are not identical, but show a considerable degree of overlap. AGL15 and AGL18 are both expressed at all stages of the life cycle. AGL15 is upregulated during the embryonic phase and is expressed in all tissues of the embryo (Heck et al., 1995). During the later stages of seed development, AGL15 protein levels decline (Perry et al., 1996). AGL15 has been shown to auto-regulate in a positive fashion (Zhu and Perry, 2005); therefore, the decline during later stages probably contributes to downregulation of AGL15 during post-germinative development. AGL18, on the other hand, is present at lower levels in embryos and is upregulated in endosperm and young seedlings (Alvarez-Buylla et al., 2000; Lehti-Shiu et al., 2005). It is expressed in most tissues in seedlings, just as AGL15 is in embryos. We speculate that an ancestral AGL15-like gene may have duplicated and then undergone spatial and temporal sub-functionalization, such that AGL15 is predominant in embryos and AGL18 is predominant in seedlings.
AGL18 may have some unique functions, because it is expressed at times and in places where AGL15 is not. For example, AGL18 shows strong expression throughout the root and in the blades of cotyledons and leaves. AGL18 is also strongly expressed in pollen (Kofuji et al., 2003). We have been unable to detect any defects in root development or pollen function in agl18 mutants.
In young seedlings, both AGL15 and AGL18 are found in the vascular system and shoot apex. Both are important tissues for the regulation of flowering time. The expression of AGL15 is largely restricted to these tissues. This distribution is coincident with the expression domain of the floral integrator FT. AGL18 is expressed throughout the seedling, which is similar to the expression pattern of other floral regulators, such as the floral integrator SOC1.
AGL15 and AGL18 act as floral repressors under both inductive and non-inductive conditions
Through analysis of loss-of-function mutations, we have established a developmental role for AGL15 and AGL18 as redundantly acting floral repressors. In agl15 agl18 double mutants, the transition to flowering occurs earlier than in wild-type. This effect is difficult to detect in summer-annual backgrounds (Col, for example) under LD conditions, where flowering is early. The effect of the mutations becomes more apparent if the strength of the photoperiodic inductive signal is reduced and the floral transition is delayed. The photoperiod pathway can be eliminated either by growing the plants under SD conditions or genetically, by introducing co mutations. In both situations, plants carrying agl15 agl18 double mutant combinations flower significantly earlier than plants that do not carry the double mutation.
AGL15 and AGL18 interact with but do not appear to be targets of the photoperiod pathway. The agl15 agl18 mutant combination only partially suppresses photoperiod pathway defects. agl15 agl18 mutants still show a strong response to photoperiod, and mutations in CO do not result in changes in AGL15 and AGL18 transcript accumulation. The data are most consistent with a model where AGL15 and AGL18 operate in a parallel pathway. Partial suppression would be seen if the two pathways converge on a common target or targets at the level of the floral integrators.
Mutations in AGL15 and AGL18 result in de-repression of the floral integrator FT, which is consistent with the early-flowering phenotype. De-repression of the floral integrator SOC1 was not observed in the agl15 agl18 mutants. The effect on FT expression may be direct or indirect. As one of the main floral integrators, FT receives input from multiple regulatory pathways, including the photoperiod pathway through CO and the autonomous and vernalization pathways through FLC. Inducible expression studies have shown that FT is a direct target of CO (Samach et al., 2000), and FLC has been shown to bind to CArG motifs in the first intron of FT (Helliwell et al., 2006). If the effect of AGL15 and AGL18 on FT expression is also direct, which is feasible because of the overlap between AGL15, AGL18 and FT expression patterns, several possibilities exist. AGL15 or AGL18 could associate with FLC-containing complexes. Alternatively, AGL15 or AGL18 may bind independently in another region of the FT locus. The FT locus contains many potential binding sites for regulatory factors, including additional CArG motifs of the type that AGL15 has been shown to prefer (Tang and Perry, 2003), just upstream of the ATG. Further analyses of protein–DNA and protein–protein interactions are required to distinguish between direct and indirect effects on FT transcript accumulation.
AGL15 and AGL18 show additive genetic interactions with SVP and members of the FLC-like clade
We used genetic and expression analyses to examine the relationship between AGL15 and AGL18 and other MADS domain factors that function as floral repressors, including SVP and members of the FLC-like clade, such as FLC and FLM/MAF1. The phenotype of agl15 agl18 double mutants most closely resembles that of flm mutants. The effects of flm/maf1 mutations are easiest to detect under SD conditions, as is the case with agl15 agl18 double mutants, and flm/maf1mutations also partially suppress photoperiod pathway mutations (Scortecci et al., 2003). These similarities suggest that AGL15, AGL18 and FLM/MAF1 might all work in the same pathway. Our expression data are not consistent with a linear pathway, i.e. one set is not the regulatory target of the other set. If all three factors are involved in the same pathway, our data would be most consistent with action as co-regulators.
Analysis of multiple genetic combinations confirmed the epistatic relationship between SVP and FLM observed in previous studies (Scortecci et al., 2003). When svp mutations are combined with flm mutations, the double mutants resemble the svp mutants with regard to flowering time under both LD and SD conditions and in the presence or absence of agl15 agl18 mutations. This is consistent with the model of SVP and FLM as co-regulators (Scortecci et al., 2003). We note, however, that, in Col, flm mutations are not as severe as svp mutations under SD conditions. This suggests that other factors may be partially redundant and may also serve as binding partners for SVP. Other members of the FLC-like clade that show epistatic relationships with svp mutations, including FLC, would be strong candidates for such binding partners.
When svp, flm or svp flm double mutants are combined with the agl15 agl18 double mutant, the effect on flowering time is additive rather than epistatic. Based on this genetic relationship, AGL15 and AGL18 may have some redundancy with SVP. AGL15- and AGL18-containing complexes could act separately and/or enhance the repression activity of the SVP-containing complexes. AGL15 can bind to DNA as a homodimer (Perry et al., 1996), and also has been shown to interact with SVP in a yeast two-hybrid assay (de Folter et al., 2005). AGL18 interactors have not been identified to date (de Folter et al., 2005). Ternary complexes that contain a combination of SVP and AGL15 or AGL18 would contain multiple copies of the LXLXL motif and might be more effective repressors as a result. Alternatively, each complex could bind to different targets and the effect on flowering time could be the collective result of separate regulatory events. Recent work on FLM is intriguing in this regard. flm mutations lead to upregulation of TSF and AGL24 transcripts, but, unlike agl15 agl18, flm mutations have little effect on FT expression under SD conditions (Sung et al., 2006).
As with svp and flm, flc mutations show an additive interaction with agl15 agl18 mutations. The effect is statistically significant under both LD and SD conditions, but is most apparent under SD conditions. This was somewhat surprising given that FLC is not thought to play a major role in floral repression under SD conditions. These results indicate that, in the absence of AGL15 and AGL18 activity, the contributions of FLC to floral repression assume a greater significance. We can reconcile this observation with the previous work by considering that both FLC and AGL15 are particularly abundant during the embryonic phase (Lehti-Shiu et al., 2005). If all of the MADS domain repressors contribute in a quantitative fashion to repression, then the flc agl15 agl18 triple mutant would exit embryogenesis with a smaller pool of functional repressors than either the flc mutant or agl15 agl18 double mutant. Alternatively, the pool of functional repressors available during embryogenesis may determine whether target genes enter the seedling development phase in either a relatively strong or relatively weak repressed state. This would probably have some impact on the level of inductive activity needed to drive the floral transition in young seedlings.
The general picture that emerges from this and other analyses is one of multiple MADS domain repressors contributing in an additive way to control of the floral transition. Individual repressors may assume a greater or lesser significance under particular environmental conditions, and collectively they represent a highly modular, redundant system. Based on our genetic data, we propose the model shown in Figure 10. Most of the floral repression activity can be attributed to complexes that include SVP and one or more of the members of the FLC-like clade. These show epistatic interactions in genetic tests, suggesting that FLC-like factors and SVP form heterodimers or associate in a ternary complex in vivo. The other component, which may or may not be an independent complex, consists of either AGL15 or AGL18. We do not expect AGL15 and AGL18 to function as heterodimers, because then each single mutant should have a phenotype that resembles that of the double mutant. All of our genetic tests to date indicate that AGL15 and AGL18 act in a fully redundant manner and show an additive relationship with complexes containing the other MADS domain floral repressors. This may involve physical interactions in a ternary complex or action in separate pathways that integrate through FT and other floral regulators. We do not yet know whether AGL15 and AGL18 act in leaf tissue or in the shoot apex or both. Future studies will be aimed at answering this question and further probing the molecular basis of AGL15 and AGL18 regulation of flowering time.
Experiments were conducted using either the Wassilewskija (Ws) or Columbia (Col) ecotypes, as indicated. Plants were grown as described previously (Lehti-Shiu et al., 2005) in growth chambers operating under either LD (16 h light/8 h dark) or SD (8 h light/16 h dark) conditions. Light was maintained at approximately 125 μE m−2 sec−1 with a mixture of cool fluorescent and incandescent bulbs.
For RNA isolations, tissue samples were frozen in liquid nitrogen immediately after harvesting and stored at −80°C. Seedlings were grown on plates containing germination medium (GM) for either 6 or 10 days under the appropriate light conditions. Mutant and wild-type plants were grown and processed in parallel. 35S:AGL15 and 35S:AGL18 seedlings were grown under LD conditions for 4 days before collection.
Generation of transgenic plants
Two AGL18 overexpression constructs were generated in a pPZP221 transformation vector (Hajdukiewicz et al., 1994). Pro35S:gAGL18 consists of the entire AGL18 coding sequence, including introns, placed downstream of the CaMV 35S constitutive promoter and upstream of the NOS terminator. The second construct, Pro35S:gAGL18-T7, is identical except for the addition of a C-terminal T7 epitope tag. The AGL18 genomic coding sequence was PCR-amplified from Ws genomic DNA with Pfu polymerase using oligos 173 (5′-ATGGGGAGAGGAAGGATTGAGATTAAGAA-3′) and 174 (5′-TCAATCAGAAGCCACTTGACTCCCAGAGT-3′) or oligos 173 and 200 (5′-CTATCCCATCTGCTGACCTCCAGTCATAGAAGCCATATCAGAAGCCACTTGACTCCCAGA-3′). Oligo 200 incorporates the 11 amino acid sequence from the N-terminus of the T7 virus capsid protein tag (MASMTGGQQMG; Novagen, http://www.emdbiosciences.com). The PCR products were cloned into the SmaI site of a modified pPZP221 vector (Hajdukiewicz et al., 1994) that contains the 35S CaMV promoter and NOS terminator sequences (DF264).
The Pro35S:AGL15 construct and the AGL15–GUS and AGL18–GUS translational fusion constructs were as described previously (Fernandez et al., 2000; Lehti-Shiu et al., 2005). The AGL15–GUS translational fusion construct includes 2.5 kb of AGL15 5’ regulatory sequence, the entire AGL15 coding sequence plus introns fused in-frame with GUS, and 1 kb of AGL15 3’ regulatory sequence. The AGL18–GUS translational fusion includes 2.2 kb of AGL18 5’ regulatory sequence, the entire AGL18 coding sequence plus introns fused in-frame with GUS, and 1.8 kb of AGL18 3’ regulatory sequence. GUS activity was visualized as described previously (Fernandez et al., 2000).
The AGL15 genomic construct (gAGL15) for complementation consisted of a 7 kb genomic fragment including approximately 2.5 kb of upstream sequence, the full coding sequence plus introns, and 2.7 kb of downstream sequence. The sequence was originally isolated as an EcoRI/BamHI genomic fragment following partial digestion. It was inserted as a SalI/BamHI fragment into the multi-cloning site of pPZP221.
The AGL18 genomic construct (gAGL18) for complementation consisted of 1.7 kb of AGL18 upstream sequence, and the full coding sequence including introns. This was amplified from the AGL18–GUS construct using oligo 588 (5′-TGGATCCAGCATGTTACAGAAGTGTCTGG-3′), which introduced a BamHI site and oligo 589 (5′-TTGGTACCTCAATCAGAAGCCACTTGACTCCCAGAC-3′), which re-introduced the stop codon along with a KpnI site. The PCR fragment was inserted into a modified pPZP221 vector (DF289) that contained the NOS terminator.
Constructs were introduced into Arabidopsis using the floral dip method with Agrobacterium tumefaciens strain GV3101 (Clough and Bent, 1998). The GUS constructs and overexpression constructs were introduced into wild-type plants (Ws ecotype). The genomic rescue constructs were introduced into the agl15-3 agl18-1 double mutant (Col ecotype). Transformants were selected on GM supplemented with 50 μg ml−1 kanamycin (AGL18–GUS) or 100 μg ml−1 gentamycin (AGL15–GUS, Pro35S:gAGL18, gAGL15 and gAGL18 constructs).
RNA isolation and RT-PCR analysis
RNA isolation and RT-PCR analysis were preformed as described previously (Lehti-Shiu et al., 2005). RNA samples were treated with RQ1 DNase (Promega; http://www.promega.com/) and various numbers of cycles were used, as specified in the results, depending on the transcript abundance.
Gene-specific primers for EF1α and individual MADS box genes have been described previously (Lehti-Shiu et al., 2005). Whenever possible, primers were designed to flank introns. The following two primers were used for analysis of FT transcripts: oligo 708 (5′-TGTTGGAGACGTTCTTGATCC-3′) and oligo 709 (5′- AGCCACTCTCCCTCTGACAA-3′).
Plants carrying various combinations of mutant alleles were identified through PCR-based genotyping. The agl15-3, agl15-4, agl18-1 and agl18-2 alleles have been described previously (Lehti-Shiu et al., 2005). agl15-3 agl18-1 double mutants were crossed to co-9 (Balasubramanian et al., 2006), gi-2, (Redei, 1962) ld-1, flc-3 (Michaels and Amasino, 1999), svp-2 (Scortecci et al., 2003) and flm-3 (Sung et al., 2006) mutants in the Col background, obtained from Dr R. Amasino (University of Wisconsin- Madison). To generate triple mutant combinations with ld and gi, late-flowering individuals in the F2 generation were screened for the presence of agl15 or agl18. To identify recombinant chromosomes carrying both co and agl15, late-flowering individuals in the F2 generation of a cross between co and agl15 were screened for the agl15 allele. To identify recombinant chromosomes carrying both flc-3 and agl15, early-flowering individuals in the F2 generation of a cross between flc-3 (FRI) and agl15 were screened for agl15.
DNA was isolated by grinding leaf tissue in an extraction buffer (50 mM Tris, pH 7.5, 250 mm NaCl, 20 mm EDTA, 0.5% SDS), centrifuging at >10 000 g for 10 min, precipitating the supernatant in 100% isopropanol for 5 min, centrifuging again for 10 min at >10 000 g, rinsing with 70% ethanol, and resuspending in 250 μl 10 mm Tris-HCl, pH 7.0. Procedures for genotyping agl15 and agl18 alleles have been described previously (Lehti-Shiu et al., 2005). The following oligonucleotide combinations were used for genotyping flc, svp, flm and co alleles: FLC, 5′-GTATCGTAGGGGAGGAAAGATAG-3′ and 5′-CTCATGTATCTATCATGGTCGCAG-3′ for both alleles; SVP, 5′-GAAGGAAGTCCTAGAGAGGCATAAC-3′ combined with 5′-CGTTAGTAATAGACTCCGACGACTG-3′ for the WT allele, and combined with 5′-GCGTGGACCGCTTGCTGCAACT-3′ for the svp allele; FLM, 5′-GAAACATTCCTCTCTCATCATCTGT-3′ combined with 5′-CCGTTATTGTGTCTACTGGAAAAT-3′ for the WT allele and combined with 5′-GCGTGGACCGCTTGCTGCAACT-3′ for the flm allele; CO, 5′-AGCTCCCACACCATCAAACTTACTACATC-3′ combined with 5′-AGTCCATACTCGAGTTGTAATCCAC-3′ for the WT allele and combined with 5′-TAGCATCTGAATTTCATAACCAATCTCGATA-3′ for the co allele.
For each data point in the flowering-time experiments, the total numbers of leaves produced at nodes on the main axis were counted. To test for differences between the means, a two-tailed t-test was used. Mean values were considered different when P < 0.05.
This work was supported by the University of Wisconsin-Madison Graduate School and by the United States Department of Agriculture (grant number 2001-35304-10887). The authors thank Dr Richard Amasino for seed stocks, comments on the manuscript, and generous sharing of insights on flowering-time regulation, and Ann Zumhagen-Krause and Espen Hesselberg for help with flowering-time analyses. Seed stocks containing agl15 and agl18 alleles were obtained from the Arabidopsis Biological Resource Center (Ohio State University, Columbus).