Introduction

The appropriate timing of flowering is tightly linked to the success of reproduction inhigher plants. Intrinsic genetic programs and various environmental factors, mainly daylength and temperature, determine the transition from vegetative to reproductivedevelopment. In particular, photoperiod provides a major cue for controlling floweringtime, as perception of light enables plants to synchronize initiation of flowering withseasonal changes in photoperiod1.

In Arabidopsis thaliana, several signaling components participate in theregulatory circuit promoting photoperiodic flowering, including GIGANTEA(GI), CONSTANS (CO) and FLOWERING LOCUS T(FT)2,3,4. FT integrates multiple flowering pathwaysand FT protein is an essential component of florigen, which moves from the induced leafto the shoot apex2,5. CO directly regulates expression of FT mRNAand CO mediates between the circadian clock and the control of flowering. CO is stablein the light, but is degraded in the dark by ubiquitin-mediated proteolysis4,6. GI and FLAVIN-BINDING, KELCH REPEAT, F-BOX PROTEIN 1 (FKF1) form acomplex and regulate the timing of CO expression. The diurnal expression ofGI and FKF1 has little overlap in SD, leading to minimal formation ofthe GI-FKF1 complex7. By contrast, in LD, the more extensive overlap ofGI and FKF1 diurnal expression leads to formation of more GI-FKF1complex. Thus, GI acts as a flowering inducer with FKF1 in the CO-FT pathwaymainly in LD. In a CO-independent flowering pathway, GI can also directlyactivate FT expression by binding to its promoter region8,indicating that GI can directly or indirectly induce FT transcription in thephotoperiod pathway.

In addition to regulation by the photoperiod pathway, genes involved in the autonomousand vernalization pathways also control FT expression. FLOWERING LOCUS C(FLC) has a central place in those two pathways and directly regulatesFT and SOC1 expression by binding to their promoters9,10,11. Chromatin remodeling also affects FLC expression. Forexample, MULTICOPY SUPPRESSOR OF IRA1 4 (MSI4)/FVE, in the autonomous pathway,negatively regulates FLC expression via histone deacetylation of the FLClocus12. Furthermore, MSI4/FVE interacts with DDB1 and HDA6, andmediates transcriptional silencing by histone modification of H3K4me313and H3K27me314. This indicates that MSI4/FVE plays a significant role inFLC expression by making a complex with various chromatin remodelingfactors.

DET1, a repressor of photomorphogenesis, was first identified as a member of theCONSTITUTIVE PHOTOMORPHOGENIC/DE-ETIOLATED/FUSCA (COP/DET/FUS) genefamily15. DET1 forms a complex with COP10 and DAMAGED DNA BINDINGPROTEIN 1 (DDB1) to promote the activity of ubiquitin-conjugating enzymes (E2) forrepression of photomorphogenesis in the ubiquitination pathway16,17.DET1 also acts as a pacemaker to adjust the period length of the circadian rhythm18, possibly through interaction with LHY and CCA119. DET1acts as a flowering repressor; det1-1 mutants flower slightly early in LD andextremely early in SD20. Despite recent advances in the understanding ofDET1 function, the molecular mechanism causing early flowering in det1-1 mutantsremains unknown.

Here we demonstrate that DET1 delays flowering time in SD, mainly by reducing theaffinity of GI binding to the FT promoter in the photoperiod pathway. DET1also contributes to upregulating FLC expression in the autonomous pathway,possibly by weakening the activity of MSI4/FVE in histone modification of the FLClocus. These effects, in turn, lead to reduced expression of FT and SOC1.These findings provide new insights into how DET1 dynamically suppresses flowering in SDand thus plays an important role in maintaining photoperiod sensitivity inArabidopsis.

Results

The det1 mutation alters the expression of flowering-timegenes

The det1 null mutants are lethal; to study the molecular mechanism bywhich DET1 functions in floral repression, we therefore used a weakallele, det1-1 and counted the rosette leaf number at bolting to measureflowering time (Fig. 1a, b). We found that det1-1mutants flower early under LD and extremely early under SD, which shows thatflowering in det1-1 mutants is photoperiod-insensitive. These resultsindicate that DET1 acts as a strong floral repressor in SD and has a key role inmaintaining the photoperiod sensitivity of the regulation of flowering time inArabidopsis.

Figure 1
figure 1

Flowering-time phenotypes of det1-1 mutants.

(a) Phenotypes of wild-type (WT, Col-0 ecotype) and det1-1 mutantplants. Plants were grown at 22°C under cool-white fluorescentlight (90–100 μmolm−2s−1) in LD(16-h light:8-h dark) or SD (10-h light:14-h dark) and photographed at 2 to4 days after bolting. Scale bars = 2 cm. (b–c) Genetic analysisto show epistasis between det1-1 and flowering mutants using double(b) and triple mutants (c). The number of rosette leaves of WT (Col-0) andflowering-time mutants grown under LD (16-h light:8-h dark) and SD (10-hlight:14-h dark) in (b) and LD (16-h light:8-h dark) and SD (8-h light:16-hdark) conditions in (c) (see Table S1). Flowering timewas measured as the number of rosette leaves at bolting. Means and standarddeviations were obtained from more than 20 plants.

Full size image

The det1-1 mutation causes period-shortening of clock-regulated geneexpression; the internal circadian periods of CAB2:LUC (encoding aluciferase) expression in det1-1 mutants were approximately 18 h incontinuous darkness and 21 h in continuous light conditions18. Toinvestigate whether the circadian defect in det1-1 mutants causesextremely early flowering under SD (Fig. 1 and Table S1), we analyzed the expression modes of floralinducers by measuring the phases and amplitudes of GI, FKF1,CO, FT and SOC1 mRNA abundance, in WT anddet1-1 mutants grown in SD (Fig. 2). In WT,GI expression peaked at ZT6 (zeitgeber time; 6 h after dawn) duringdaytime, but the peaks of FKF1 and CO expression occurred at ZT9and ZT12 during nighttime, respectively, resulting in no FTexpression21. In det1-1 mutants, GI,FKF1, CO, FT and SOC1 also showed rhythmicexpression (Fig. 2a–e) and GI expressiondid not significantly differ compared with WT (Fig. 2a).However, the peaks of FKF1 and CO expression shifted 3 h and 6 hearlier than those in WT, respectively (Fig. 2b, c).Accordingly, the peaks of GI, FKF1 and CO expressionoccurred at ZT6 during daytime in det1-1 mutants under SD. Thus, itappears that the daytime expression of CO and light-stabilized CO (Fig. 2c) can activate FT expression in det1-1mutants under SD (Fig. 2d). The waveform and peak time ofSOC1 expression did not change in det1-1 mutants, butSOC1 mRNA abundance increased (Fig. 2e),possibly due to daytime expression of CO and/or increased expression ofFT (Fig. 2c, d)9,22,23. Thus,we first speculated that circadian dysfunction might cause the early floweringin det1-1 mutants, as previously reported19.

Figure 2
figure 2

Effect of det1-1 mutation on GI, FKF1, CO,FT, SOC1 and FLC expression under SD.

The expression of GI (a), FKF1 (b), CO (c), FT(d), SOC1 (e) and FLC (f) was analyzed in Col-0 anddet1-1 mutants by real-time PCR using 3-week-old plants. Plantswere grown at 22°C under SD (8-h light:16-h dark) conditions,and plant tissues were harvested every 3 h. ACT2 expression was usedfor normalization. Means and standard deviations were obtained from threebiological replicates.

Full size image

To test whether circadian-period shortening causes the extremely early floweringof det1-1 mutants in SD (Fig. 1 and Table S1), we examined whether the flowering-time defect can berecovered when det1-1 mutants were entrained in SD (light:dark = 1:2)under reduced diurnal cycles, i.e. environmental time periods (T) of 24 T (8-hlight:16-h dark), 21 T (7-h light:14-h dark) and 18 T (6-h light:12-h dark).Although reduced diurnal cycles of 21 T and 18 T slightly delayed floweringcompared to normal cycles of 24 T, the det1-1 mutants still flowered muchearlier than WT under SD of 24 T (Fig. 3). To investigatethe cause of early flowering in det1-1 mutants under reduced T cycles, weanalyzed the phases and amplitudes of GI, FKF1, CO,FT and SOC1 mRNA abundance in det1-1 mutants grownunder SD of 18 T (Fig. S1). Unlike the SD of 24 T, the waveforms and peaks ofGI, FKF1 and CO expression in det1-1 mutantswere very similar to those of WT. However, FT and SOC1 expressionwas still upregulated in det1-1 mutants, suggesting that the internalperiod-shortening defect in det1-1 mutants cannot fully explain theextremely early flowering under SD of 24 T. The FKF1 and CO peakshifts likely produce a small effect on early flowering in det1-1mutants, because fkf1-t and co-101 mutations delayed flowering indet1-1 mutants under SD whereas they were almost ineffective inWT21 (Fig. 1b and TableS1). Thus, these results strongly suggest that other defects inmechanisms of floral repression lead to photoperiod-insensitive early floweringin det1-1 mutants, rather than the circadian dysfunction in theFKF1-CO-FT pathway.

Figure 3
figure 3

Flowering time of det1-1 mutants under reduced diurnal cycles.

(a) Effect of reduced diurnal cycles on the flowering time of det1-1mutants. Plants were entrained in SD (light [L]:dark[D] = 1:2) of 24 h (24 T = 8 L:16 D), 21 h (21 T = 7L:14 D) and 18 h (18 T = 6 L:12 D). T represents environmental time period.Means and standard deviations were obtained from more than 20 plants. Col-0means Columbia-0 ecotype (wild type). (b) Phenotypes of det1-1mutants after bolting under SD of 24 T, 21 T and 18 T. Plants were grown at22–24°C under cool-white fluorescent light(90–100 μmol m−2s−1). Scale bars = 2 cm.

Full size image

DET1 mainly functions in the photoperiod and autonomouspathways

To test which genetic pathways of floral induction are responsible for the earlyflowering phenotype of det1-1 mutants, we examined the flowering-timephenotypes of double mutants of det1-1 and mutations with late-floweringphenotypes, specifically cry2-1, fkf1-t, gi-1,co-101, ft-1 and soc1-2 (Fig. 1band Table S1). The cry2-1det1-1 double mutants flowered much earlier than the cry2-1 singlemutants in both LD and SD, suggesting that DET1 acts downstream ofCRY2. The fkf1-t det1-1 and co-101 det1-1 doublemutants exhibited intermediate flowering times compared with fkf1-t,co-101 and det1-1 single mutants in both LD and SD,suggesting that although daytime expression of FKF1 and COcontributes to early flowering in SD, det1-1 mutants can flower early inthe absence of FKF1 and CO activity in both photoperiod conditions. Ingi-1det1-1 and ft-1 det1-1 mutants, the early-flowering effect ofdet1-1 was almost abolished by gi-1 or ft-1 in both LDand SD (Fig. 1b and Table S1),indicating that GI and FT play major roles in theDET1-mediated flowering pathway.

As both the photoperiod and autonomous pathways regulate SOC1expression10, we further tested whether DET1 alsoparticipates in the autonomous pathway. We found that soc1-2det1-1 double mutants showed intermediate flowering times in both LD andSD. Also, in ft-1 soc1-2det1-1 triple mutants, the early flowering effect of det1-1completely disappeared (Fig. 1b, c and Table S1). These results indicate that the regulation of floweringtime by DET1 does not entirely depend on the FT-mediatedphotoperiod pathway, but also depends on the SOC1-mediated autonomouspathway. Thus, we further examined the expression of FLC, a major gene inthe autonomous pathway, in det1-1 mutants. We found that thedet1-1 mutants under SD had very low levels of FLC mRNA (Fig. 2f), suggesting that DET1 induces FLC expressionto repress FT and SOC1. Taking these results together, weconcluded that DET1 mainly acts in the photoperiod and autonomous pathways as astrong floral repressor.

DET1 interacts with GI in vivo

GI functions in the photoperiod pathway and det1-1 mutants did not showsignificant alterations in GI mRNA levels (Fig.2a), but the gi-1 mutation nearly abolished the early floweringeffect of det1-1 in gi-1 det1-1 double mutants (Table S1). Based on these observations, we postulated that DET1mainly regulates GI at the post-translational level. Thus, we used transgenicplants expressing a tagged GI protein (pGI:GI-HA gi-2 and pGI:GI-HAgi-2 det1-1) to examine whether DET1 negatively regulates GI stability.We found that det1-1 mutants showed no significant alteration in therhythmic accumulation of GI protein in SD (Fig. 4a). Thisindicates that the det1-1 mutation does not affect GI proteinstability.

Figure 4
figure 4

DET1 directly interacts with GI.

(a) Comparison of GI protein stability between pGI:GI-HA andpGI:GI-HAdet1-1 plants under SD conditions. The plant tissues were collectedevery 2 h during the daytime and every 4 h during the nighttime, using3-week-old seedlings. GI protein was detected with an anti-HA antibody. RFT5expression was used for normalization. Means and standard deviations wereobtained from three biological replicates. (b) Interaction of DET1-GI wastested by yeast 2-hybrid assay. The bait was full-length DET1. For prey, GIwas divided into three pieces: N-terminal (N; 1–507), middle (M;401–907) and C-terminal (C; 801–1173). Gal4 indicatesa positive control. Empty pGBKT7 (BD) and pGADT7 (AD) vectors were used asthe negative control. SD medium (-LWHA; lacking tryptophan, leucine,histidine and adenine) was used to select for the interaction between baitand prey proteins. β-galactosidase (β-Gal) activityassays were performed according to the manufacturer's protocol.Means and standard deviations were obtained from three biologicalreplicates. (c) BiFC analysis of the interaction of between DET1 and GI inthe nucleus of an onion epidermal cell. nYFP-ELF3 and cYFP-ELF4 plasmidsserved as a positive control. For the negative control, empty nYFP/GI-cYFPand nYFP-DET1/cYFP were used. Scale bar = 50 μm. (d)Coimmunoprecipitation of DET1 and GI. Total protein was extracted from2-week-old seedlings of p35S:TAP-DET1 pGI:GI-HA gi-2 andp35S:TAP-GFP pGI:GI-HA gi-2. IgG beads were used for thepull-down. An anti-HA antibody was used for GI-HA protein band.p35S:TAP-GFP pGI:GI-HA gi-2 plants served as a negative control.The upper panel is a coimmunoprecipitated sample and the middle panel isthe input sample for GI-HA protein. The lower panel shows input samples ofp35S:TAP-GFP and p35S:TAP-DET1.

Full size image

DET1 interacts with LHY and CCA1, which regulate the circadian rhythms ofexpression of clock-regulated genes19. This raises thepossibility that DET1 could negatively regulate GI activity by protein-proteininteraction. To examine this, we performed yeast 2-hybrid assays and found thatDET1 interacts with the N-terminal region of GI (amino acids[aa] 1-507) (Fig. 4b). To test thein vivo interaction of DET1 and GI, we performed bimolecularfluorescence complementation (BiFC) assays. In the onion epidermal cells, wedetected reconstituted YFP fluorescence in the nucleus when nYFP-DET1 andGI-cYFP plasmids were co-transformed (Fig. 4c). To furtherconfirm their interaction, we tested whether GI and DET1 co-immunoprecipitatefrom transgenic plants expressing tagged proteins. To that end, we sampled thep35S:TAP-DET1 pGI:GI-HA gi-2 and p35S:TAP-GFP pGI:GI-HA gi-2(a negative control) transgenic plants at ZT8 in SD and used antibodies for theTAP tag to immunoprecipitate DET1. We found that HA-GI co-immunoprecipitatedwith TAP-DET1, but not with TAP-GFP (Fig. 4d). Theseresults indicate that DET1 interacts directly with GI in the nucleus.

DET1 negatively regulates GI binding to the FT promoter

The det1-1 mutation does not alter GI mRNA expression (Fig. 2a) or GI protein levels (Fig.4a) but gi-1 shows nearly complete epistasis to det1-1 inflowering time (Fig. 1b and Table S1).Based on this observation, we hypothesized that in the photoperiod pathway, DET1negatively regulates the activity of GI, which directly upregulates FTexpression through a CO-independent pathway8. To testwhether det1-1 mutation affects the GI-FT module, we performedchromatin immunoprecipitation (ChIP) assays, using pGI:GI-HAgi-2 and pGI:GI-HAgi-2det1-1 seedlings entrained in SD, to test whether det1-1 affectsthe ability of GI to bind to the FT promoter. We collected tissues from10-day-old seedlings at ZT8 and detected relative enrichment of the promoterregions by PCR with primers for six regions of the FT promoter, asdescribed previously8. When we compared GI binding affinity tothe FT promoter regions, the amplicons close to the 5′untranslated region (UTR) were significantly more enriched in ChIP fromdet1-1 mutants (Fig. 5b). This result stronglysupports the notion that DET1 plays an important role in the suppression ofFT transcription by preventing GI binding to the FT promoter,and thus contributing to late flowering in SD conditions.

Figure 5
figure 5

DET1 affects GI binding to the FT promoter.

(a) Gene structure of FT and the amplicon regions for the ChIP assay.Six amplicon locations (I, II, III, IV, V and VI) are shown. (b) FTpromoter binding affinity of GI in the det1-1 mutant, relative to thewild type. All samples were harvested at ZT8 under SD (8-h light:16-h dark)conditions. Chromatin isolated from these samples was immunoprecipitatedwith anti-HA. Relative enrichment in Col-0, pGI:GI-HA gi-2, andpGI:GI-HA gi-2 det1-1 are shown. Means and standard deviationswere obtained from three biological replicates. This experiment wasreplicated at least three times with similar results. UBIQUITIN 10(UBI10) was used as a negative control. Black, gray and whiteboxes represent Col-0, pGI:GI-HA gi-2 and pGI:GI-HA gi-2det1-1, respectively. Asterisks indicate statistically significantdifferences compared to pGI:GI-HA as determined byStudent's t-test (*P < 0.05 and**P < 0.01, respectively).

Full size image

DET1 positively regulates FLC expression to delay flowering time inSD

In the autonomous pathway, FLC functions as a key floral repressor anddownregulates the transcription of FT and SOC124,25,26. As the transcript levels of FT and SOC1were upregulated in det1-1 mutants under SD (Fig. 2d,e) and FLC expression was almost absent in det1-1mutants entrained in SD (Fig. 2f), we reasoned that DET1also functions to delay flowering in the autonomous pathway by upregulatingexpression of FLC. A previous report showed that the COP10-DET1-DDB1complex interacts with CUL427 and the DDB1-CUL4 complex interactswith MSI4/FVE to induce FLC transcription14. Thus, weasked if DET1 interacts with MSI4 to form a DET1-MSI4 complex to regulateFLC mRNA levels. To test this, we examined the in vivointeraction of MSI4-DET1 by BiFC assays (Fig. 6a). Wedetected strong YFP fluorescence in the nuclei of cells co-transformed withplasmids expressing DET1-nYFP and cYFP-MSI4, indicating that DET1 interacts withMSI4, which directly binds to the FLC promoter to repress FLCtranscription.

Figure 6
figure 6

DET1 interacts with MSI4 and regulates histone methylation of the FLClocus.

(a) BiFC analysis of the interaction between MSI4 and DET1 in onion epidermalcells. For negative controls, nYFP/cYFP-MSI4 and DET1-nYFP/cYFP were used.Scale bar = 50 μm. (b) Relative levels of histone modificationson the FLC locus were examined by ChIP analysis using H3K4me3 andH3K27me3 antibodies in Col-0 and det1-1 plants. The top of the panelrepresents the FLC gene structure and the region used for primers (I,II and III) in the ChIP-quantitative PCR analyses. Chromatin was preparedfrom 14-day-old seedlings grown under SD (8-h light:16-h dark). FUSCA3 (FUS3) was used for the normalization of the quantitativePCR analysis. Means and standard deviations were obtained from threebiological replicates. This experiment was replicated at least three timeswith similar results. Asterisks indicate statistically significantdifference compared to Col-0 as determined by Student'st-test (*P < 0.05).

Full size image

Since MSI4 binds to the FLC promoter and alters histone modification,specifically H3K27me3 and H3K4me3, at the FLC locus13,14,we further examined the histone methylation levels of the FLC locus,using anti-H3K27me3 and anti-H3K4me3 antibodies in WT and det1-1 mutants.The ChIP analysis revealed that det1-1 mutants maintained higher levelsof H3K27me3 and lower levels of H3K4me3 at the FLC locus than did WT(Fig. 6b), consistent with the histone modificationstates observed in the early-flowering hos1-3 mutants28.Taking these results together, we suggest that the DET1-MSI4/FVE complex likelycontributes to late flowering in SD by altering histone modification of theFLC locus in the autonomous pathway.

Discussion

DET1 is involved in repression of photomorphogenesis in the ubiquitinationpathway16,17,29, light-response regulatory pathway20 and circadian period18,19. However, the function ofDET1 in the regulation of flowering time remains unclear. In this study, we provideevidence showing how DET1 regulates photoperiod sensitivity by delaying floweringtime in SD. For example, det1-1 mutants showed increased GI activity (Fig. 5) and epigenetic silencing of FLC expression (Fig. 6), resulting in upregulation of FT and SOC1.Thus, we propose a model for the regulatory role of DET1 in both photoperiod andautonomous pathways (Fig. 7).

Figure 7
figure 7

Working model of DET1 function in floral repression inArabidopsis.

DET1 suppresses FT and SOC1 expression through the photoperiodand autonomous pathways of flowering. In the photoperiod pathway, DET1mainly represses flowering by modulating GI-mediated floral induction at thetranscriptional and post-translational levels during daytime under SD. DET1represses the function of daytime-expressed GI by preventing GI from bindingto the FT promoter in a CO-independent pathway. In theautonomous pathway, DET1 interacts with MSI4/FVE and possibly modulatestrimethylation of FLC chromatin to epigenetically induce FLCexpression. Genes and proteins are represented as rectangles and ovals,respectively.

Full size image

In this study, we showed that gi-1 and ft-1 nearly completelysuppressed the early flowering of det1-1 mutants and that DET1 directlyinteracts with GI in vitro and in vivo (Fig. 4).However, DET1 does not interact with the light-input components PHYA, PHYB, CRY1C-terminus (CCT1), or CRY2 C-terminus (CCT2), or the floral inducers CO or FKF1(Fig. S3), indicating that DET1 has a unique role in the posttranslationalregulation of GI activity in the photoperiod pathway. A previous study revealed thatEARLY FLOWERING4 (ELF4), one of the circadian-clock components30,acts upstream of GI31. ELF4 represses GI binding to the COpromoter to control flowering32. Our results revealed that co-101det1-1 mutants showed intermediate flowering-time phenotypes, but in ft-1det1-1 mutants, the early flowering phenotype of det1-1 almostcompletely disappeared under LD (Fig. 1b), indicating thatDET1 function in the regulation of photoperiodic flowering mainly depends onFT expression. Thus, we hypothesized that DET1 regulates GI binding tothe FT promoter to delay flowering time and showed that GI binding to theFT promoter significantly increased in the det1-1 mutantbackground (Fig. 5). This result indicates that DET1 repressesFT expression via direct regulation of GI binding to the FTpromoter.

DET1 functions as a repressor of photomorphogenesis in darkness by forming a complexwith COP10 and DDB1 and promoting the activity of ubiquitin-conjugating E2 enzymesin the ubiquitination pathway16,17. The RING-type E3 ubiquitinligase COP1, a member of the COP/DET/FUS family15, also repressesphotomorphogenesis in darkness; cop1-4 mutants display very similarphenotypes to det1-1 mutants, such as short hypocotyls and openedcotyledons34. This implies a potential functional connectionbetween DET1 and COP1. Indeed, COP1 interacts with COP10, but not with DET116, suggesting that COP1 could interact with the COP10-DET1-DDB1 (CDD)complex to repress photomorphogenesis. In addition, cop1-4 mutants flowerextremely early under SD, similar to det1-1 mutants33. Thus,the CDD complex may function with COP1 in regulation of flowering time, although wehave no direct evidence because the det1-1 cop1-4 double mutant islethal34. COP1 directly controls GI stability by interacting withGI in the presence of ELF3 for photoperiodic flowering33. However,DET1 does not regulate GI stability but does negatively affect GI binding to theFT promoter (Fig. 4a). Therefore, although DET1 andCOP1 have very similar mutant phenotypes and post-translational behavior, they seemto regulate GI function independently through distinct molecular mechanisms.

Other negative regulators of FT transcription, including FLC, SVP, TEM1, andTEM2, bind to the regions near the 5′UTR of FT. In single mutantsof these regulatory genes, FT mRNA expression increases to levels similar tothose seen in det1-1 mutants11,35,36. Notably, SVP, TEM1,and TEM2 interact with GI to regulate FT expression, although the regulatoryfunction of their interaction is not clearly understood8. Therefore,DET1 could be involved in the function of these FT repressors. To investigatethis possibility, we examined the interaction of DET1 with these four FTrepressors by yeast 2-hybrid assays, which revealed that DET1 does not interact withFLC, SVP, TEM1, or TEM2 (Fig. S4). This result strongly suggests that DET1 mayregulate the GI-FT module independent of these known FTrepressors.

In addition, we revealed that DET1 regulates the expression of FLC, a keycomponent in the autonomous pathway. We found that the det1-1 mutants showeda remarkable decrease in FLC mRNA levels and had altered levels of H3K4me3and H3K27me3 (Figs. 2f and 6b), asobserved in the early-flowering hos1-3 mutants28. Furthermore,our examination of the components of the CDD complex showed that in addition tointeracting with DDB1, DET1 also interacts with MSI4/FVE, which repress FLCexpression in the autonomous pathway (Fig. 6a)14. This indicates that DET1 represses FLC expression possibly through directinteraction with MSI4/FVE. Meanwhile, FLC negatively regulates not only FTbut also the downstream factor SOC1, which encodes a MADS box transcriptionfactor37. In genetic analysis, ft-1 was completelyepistatic to det1-1 in LD, but in SD the ft-1det1-1 double mutants showed an intermediate phenotype, indicating incompleteepistasis. Consistent with this, SOC1 expression was upregulated indet1-1 mutants (Fig. 2e), but soc1-2 did notrescue the early flowering of det1-1 (Fig. 1b and Table S1). Notably, the ft-1 soc1-2 det1-1 triple mutantsshowed complete suppression of the early flowering of det1-1 in bothphotoperiods. This supports the idea that DET1 suppresses both FT andSOC1 via promoting FLC expression in the autonomous pathway.

DET1 interacts with LHY/CCA1 and is required for transcriptional repression ofCCA1/LHY target genes such as TOC119. These observationsindicate that DET1 functions with LHY/CCA1 to regulate the circadian rhythms ofevening genes. Moreover, DET1 could act with LHY/CCA1 to negatively regulate GIbinding to the FT promoter mainly in SD, because lhy cca1 doublemutants also exhibit photoperiod-insensitive early flowering38. Toprove this hypothesis will require further analysis, such as examination of thein vivo interaction of CCA1-GI or LHY-GI and GI binding activity to theFT promoter in either lhy cca1 double mutants or LHY orCCA1 overexpressors.

Based on these data, we propose a model for the molecular mechanism by which DET1represses flowering in non-inductive SD conditions (Fig. 7).In WT plants, the absence of FT expression under SD can be explained by theincongruity of peak expression of FKF1 and GI; GI peaks in thelate afternoon but FKF1 peaks at night, leading to reduced expression ofCO and FT during daytime21. As GI also directlyinduces FT expression in a CO-independent pathway8, wewondered why GI, which is expressed in the afternoon21 (Fig. 2a), is not capable of inducing FT expression underSD (Fig. 2a, d). In this study, we found that DET1 suppressesFT transcription by repressing GI binding activity to the FTpromoter (Fig. 5b). This model is further supported by geneticanalysis showing that gi-1 and ft-1 are almost completely epistatic todet1-1 (Fig. 1b and Table S1),indicating that DET1 mainly regulates flowering via GI.

In conclusion, we propose that DET1 functions as a strong repressor of flowering,acting in both photoperiodic and autonomous pathways (Fig. 7);DET1 suppresses flowering mainly by decreasing GI binding activity to the FTpromoter in the photoperiod pathway and epigenetically upregulating FLCexpression in the autonomous pathway. Whether DET1 acts in the CDD complex17 to delay flowering time under SD in Arabidopsis remains to beelucidated.

Methods

Plant materials and growth conditions

All the Arabidopsis thaliana lines used in this study are in the Columbia(Col-0) genetic background. Flowering-time mutants were obtained from theArabidopsis Biological Resource Center (USA), except for det1-1 which waskindly provided by Joanne Chory. cry2-1 (CS3732), gi-1 (CS3123),soc1-2 and ft-139, fkf1-t40 and co-10141 were used for genetic analysis. To createdouble and triple mutants, F1 heterozygotes were obtained by crossingthe det1-1 mutant as the female plant with other flowering-time mutantsas pollen donors. To select correct transformants, the plants showing thedet1-1 morphological phenotype were first isolated from F3plants and flowering-time mutations were finally confirmed by PCR-basedgenotyping. Plants were grown on soil at a constant 22°C under whitefluorescent light (90-100 μmolm−2s−1) in LD (16 hlight:8 h dark) and SD (10 h light:14 h dark) or SD (8 h light:16 h dark).

Analysis of flowering time

The bolting date was measured as the number of days from seed sowing to openingof the first flower and as the total number of rosette leaves at bolting. Datawere obtained from three experimental replications (20 to 60 plants perreplication).

RNA preparation and quantitative real-time PCR analysis

Tissue samples were collected every 3 h from 3-week-old seedlings. Total RNA wasextracted with the plant RNA extraction kit (Macrogen). For each sample, 2μg of total mRNA was reverse transcribed using M-MLV reversetranscriptase (Promega). The level of the transcripts was measured by real-timePCR, using GoTaq qPCR Master Mix (Promega) and the Light Cycler 2.0 instrument(Roche). Each PCR was repeated at least three times using biologicallyindependent samples. The amount of each RNA level was determined using specificprimers. The primers used for real-time PCR are listed in TableS2.

Yeast 2-hybrid assays

The full-length cDNAs of DET1, GI, PHYA, PHYB, CCT1, CCT2, CO, FKF1, FLC, SVP,TEM1, and TEM2 were amplified from wild-type total RNA usingRT-PCR. GI was divided into three parts: GI N-terminal (aa 1-507), GI middle (aa401-907) and GI C-terminal (aa 801-1173) regions. The PCR products were clonedinto pGBKT7 and pGADT7 vectors (MATCHMAKER GAL4 TWO-hybrid system 3, Clontech)to get the bait and prey clones. For the interaction study, plasmids containingfusion proteins were transformed into Saccharomyces cerevisiae AH109 andgrown on media lacking adenine, leucine, histidine and tryptophan.Galactosidase activity assays were performed according to themanufacturer's protocol.

In vivo pull-down assays

TAP-DET1 and TAP-GFP were from Xing Wang Deng. pGI:GI-HA gi-2det1-1 was obtained by crossing pGI:GI-HA gi-2 anddet1-1. For DET1-GI binding assays, TAP-DET pGI:GI-HA gi-2 andTAP-GFP pGI:GI-HA gi-2 plants were grown on MS medium in SD (8 hlight:16 h dark) for 10 days and then vacuum infiltrated for 7 ~ 10 min in 1X MS(Duchefa) liquid medium supplemented with 50 mM MG132 (Sigma) for proteasomeinhibitor treatment. After that, plants were incubated for 10 h under lightconditions. These plants were homogenized and total proteins were extracted intotal protein extract buffer [50 mM Tris-HCl (pH 7.5), 100 mM NaCl,10 mM MgCl2, 1 mM EDTA (pH 8.0), 10% glycerol, 1 mM PMSF, 1 mMDTT]. These experiments were performed with IgG beads for TAP-IP.After washing, the immunoprecipitated fractions were analyzed by immunoblotting.The TAP-DET1 and GI fusion proteins were detected by using anti-HA antibody.

Bimolecular fluorescence complementation assays

Each cDNA of GI, ELF3, DET1 and MSI4 was cloned into the BiFC gatewayvectors42 to examine their in vivo interactions. Forpartial YFP-tagged DET1 and MSI4 constructs, the cDNA of the gene was obtainedby RT-PCR from wild-type (WT, Col-0) plants and fused into four BiFC plasmidsets, pSAT5-DEST-cEYFP(175-end)-C1(B) (pE3130), pSAT5(A)-DEST-cEYFP(175-end)-N1(pE3132), pSAT4(A)-DEST-nEYFP(1-174)-N1(pE3134) and pSAT4-DEST-nEYFP(1-174)-C1(pE3136). Partial YFP-tagged ELF3 and GI constructs were previouslydescribed33. Each pair of recombinant plasmids encoding nEYFPand cEYFP fusions was mixed 1:1 (w/w), co-bombarded into onion epidermal layersusing a DNA particle delivery system (Biolistic PDS-1000/He, BioRad), andincubated on MS solid media with MG132 (50 mM) for 16–24 h at22°C under light or dark incubation, followed by observation andimage analysis using a confocal laser scanning microscope (Carl ZeissLSM710).

Chromatin immunoprecipitation assay

For the ChIP assay, Col-0, pGI:GI-HA gi-2 and pGI:GI-HA gi-2det1-1 plants were grown for 10 days under SD (8 h light:16 h dark)conditions and collected at ZT8. The samples were cross-linked with 1%formaldehyde, ground to powder in liquid nitrogen and then sonicated43. The sonicated chromatin complexes were bound with anti-HAantibody (ab9110, Abcam) for immunoprecipitation. The amount of DNA fragment wasanalyzed by quantitative real-time PCR (qPCR) using specific primers.UBI10 was used as an internal standard for normalization. The primersused for qPCR are listed in Table S2. For another ChIPassay, Col-0 and det1-1 plants were grown for 14 days under SD (8 hlight/16 h dark) conditions and collected at ZT8. For immunoprecipitation, weused the anti-trimethyl H3K4 (07-473, Millipore) and anti-trimethyl H3K27(07-449, Millipore). FUS3 was used as an internal standard fornormalization14. Experiments were performed with threebiological repeats.