Effect of photoperiod on the regulation of wheat vernalization genes VRN1 and VRN2

Abstract

Wheat is usually classified as a long day (LD) plant because most varieties flower earlier when exposed to longer days. In addition to LD, winter wheats require a long exposure to low temperatures (vernalization) to become competent for flowering. Here we show that in some genotypes this vernalization requirement can be replaced by interrupting the LD treatment by six weeks of short day (SD), and that this replacement is associated with the SD down-regulation of the VRN2 flowering repressor. In addition, we found that SD down-regulation of VRN2 at room temperature is not followed by the up-regulation of the meristem identity gene VRN1 until plants are transferred to LD. This result contrasts with the VRN1 up-regulation observed after the VRN2 down-regulation by vernalization, suggesting the existence of a second VRN1 repressor. Analysis of natural VRN1 mutants indicated that a CArG-box located in the VRN1 promoter is the most likely regulatory site for the interaction with this second repressor. Up-regulation of VRN1 under SD in accessions carrying mutations in the CArG-box resulted in an earlier initiation of spike development, compared to other genotypes. However, even the genotypes with CArG box mutations required LD for a normal and timely spike development. The SD acceleration of flowering was observed in photoperiod sensitive winter varieties. Since vernalization requirement and photoperiod sensitivity are ancestral traits in Triticeae species we suggest that wheat was initially a SD-LD plant and that strong selection pressures during domestication and breeding resulted in the modification of this dual regulation. The down-regulation of the VRN2 repressor by SD is likely part of the mechanism associated with the SD-LD regulation of flowering in photoperiod sensitive winter wheat.

Introduction

One critical trait in the adaptation of temperate grasses to cold winters is the requirement of long exposures to low temperatures (vernalization) to accelerate flowering. The vernalization requirement in temperate grasses is mainly controlled by allelic variation at the vernalization genes VRN1 and VRN2 (Jensen et al., 2005; vonZitzewitz et al., 2005; Yan et al., 2004b; Yan et al., 2003), which are different from the Arabidopsis genes with the same name (Gendall et al., 2001; Levy et al., 2002). The wheat and barley vernalization gene VRN1 is similar to the related meristem identity genes AP1, FUL, and CAL in Arabidopsis (Yan et al., 2003), which are responsible for the initiation of the transition between the vegetative and reproductive apices. Although the function of this gene is conserved, its regulation seems to differ between wheat and Arabidopsis. Whereas most of the natural allelic variation in vernalization requirement in wheat and barely is associated with mutations in the VRN1 promoter and first intron (Fu et al., 2005; vonZitzewitz et al., 2005; Yan et al., 2004a; Yan et al., 2003), in Arabidopsis this natural variability, is associated to nonfunctional or weak FRI and FLC alleles rather than to mutations in AP1, FUL, or CAL (Gazzani et al., 2003).

A more significant difference between the vernalization pathways in Arabidopsis and the temperate cereals was found between the vernalization genes involved in flowering repression. The central repressor in the vernalization pathway of Arabidopsis is the MADS-box gene FLC (Michaels and Amasino, 1999a), whereas in the temperate cereals is the Zinc finger - CCT domain transcription factor (ZCCT1) VRN2 (Yan et al., 2004b). The wheat VRN2 gene has no clear homologues in Arabidopsis, and the same is true from the Arabidopsis FLC in wheat (Yan et al., 2004b). These results suggest that the vernalization pathways of these two species (or at least the last regulatory steps of these pathways) evolved independently (Yan et al., 2004b).

Mutations in the regulatory regions of VRN1 are associated with a dominant spring growth habit (Vrn1), whereas loss-of function point mutations at the VRN2 locus (vrn2a allele) or complete gene deletions (vrn2b allele) are associated with a recessive spring growth habit in diploid wheat (Triticum monococcum L.) and barley (Hordeum vulgare L.) (Dubcovsky et al., 2005; Yan et al., 2004b). Down-regulation of VRN2 by vernalization is followed by the up-regulation of VRN1 in winter varieties from both species. These two vernalization genes show strong epistatic interactions suggesting that they are part of the same regulatory pathway (Dubcovsky et al., 2005; Takahashi and Yasuda, 1971; Tranquilli and Dubcovsky, 2000). In spring lines homozygous for the nonfunctional vrn2 alleles, allelic differences in VRN1 have no effect on flowering time. On the other hand, mutations in the promoter (Yan et al., 2004a; Yan et al., 2003) or the first intron (Fu et al., 2005) in VRN1 eliminate or reduce the effect of VRN2 and result in a spring growth habit (Dubcovsky et al., 2005; Tranquilli and Dubcovsky, 2000). These dominance and epistatic interactions suggested a model in which VRN2 is a direct or indirect repressor of VRN1 (Yan et al., 2004a; Yan et al., 2003). This model was also supported by the observation that the reduction of the VRN2 transcript levels in transgenic winter plants of hexaploid wheat ‘Jagger’ (T. aestivum L.) by RNA interference (RNAi::VRN2) (Yan et al., 2004b) results in the up-regulation of VRN1 and the elimination or reduction of the vernalization requirement.

In some photoperiod-sensitive winter varieties of hexaploid wheat the vernalization requirement can be eliminated or greatly reduced by exposing the plants to short days (SD) for several weeks and then transferring them back to long days (LD) (Evans, 1987; Krekule, 1964; McKinney and Sando, 1935). This phenomenon has not been described in Arabidopsis but is common in winter grasses from the subfamily Festucoideae (Heide, 1994). We show here that the SD replacement of vernalization in wheat is associated to the down-regulation of the vernalization gene VRN2 by SD. In addition, we identify a putative regulatory site in the VRN1 promoter responsible for differences in the regulation of this gene by SD.

Materials and Methods

Plant materials

The T. monococcum lines used in this study are listed in Table 1 with their growth habit and genotypes. The transgenic RNAi::VRN2 lines of hexaploid wheat Jagger were described before (Yan et al., 2004b).

Table 1.

Genotypes of T. monococcum lines used in this study.

VRN1 allele	VRN2 allele	Habit	Accession
Vrn1f CArG 1-bp deletion & insertion in intron 1	Vrn2	Spring	PI 10474, PI 191096, PI 191097, PI 192063, PI 266844, PI 393496, PI 503874*
Vrn1g CArG 34-bp deletion	vrn2b deletion	Spring	PI 326317*, PI 349049, PI 418582
Vrn1h Insertion in intron 1	Vrn2	Spring	PI 306540*, PI 94743
vrn1	vrn2a CCT mutation	Spring	DV92*, PI 277140
vrn1	vrn2b deletion	Spring	PI 286068, PI 323437
vrn1	Vrn2	Winter	G3116**, PI 355516, PI 355534.

^*GenBank accession numbers: DV92 (AY188331), G3116 (AY485969), PI 306540 (DQ146423), PI 326317 (DQ146422), and PI 503874 (DQ146421).

^**G3116 is the only accession of wild T. monococcum ssp. aegilopoides. All other accessions are cultivated T. monococcum ssp. monococcum.

We produced three different mapping populations using G3116 as the winter parent to confirm that the spring growth habit in T. monococcum accessions PI 503874 (Vrn1f), PI 326317 (Vrn1g), and PI 306540 (Vrn1h) was associated with variation at the VRN1 locus. Two additional F2 populations were developed from the crosses between spring line DV92 (vrn1 vrn2a) with spring lines PI 266844 (Vrn1f, Vrn2) and PI 326317 (Vrn1g vrn2b) to study the effect of the Vrn1f and Vrn1g mutations on the transcription of VRN1 under SD. In the PI 266844 x DV92 population, we first selected homozygous vrn2a plants (spring) using molecular markers, to eliminate differences due to growth habit.

Molecular markers used in the genetic studies

The polymorphic 1-bp deletion (Vrn1f) was converted into a degenerate CAPS marker (dCAPS, Michaels and Amasino, 1999b) using primers VRN1F_F (5’-ACAGCGGCTATGCTCCAGAC-3’) and VRN1F_R (5’-GGAGGATGGCCAGGCCAAATC -3’). The second nucleotide at the 3’ end of the reverse primer (underlined T) was mutated to generate a TaqI restriction site for the vrn1 allele that was absent in the Vrn1f allele. The polymorphic 34-bp deletion (Vrn1g) was tested using primers VRN1G_F (5’-ACAGCGGCTATGCTCCAG-3’) and VRN1G_R (5’-TATCAGGTGGTTGGGTGAGG-3’) flanking this deletion. Linkage between spring growth habit and the 493-bp insertion in the first intron of PI 306540 (Vrn1h Vrn2) was tested using primers VRN1INT1_F (5’-ATGAAACTCCGGCATGAAGA -3’) and VRN1INT1_R (5’-CAAAATAAAGCCGGCAAGC -3’) flanking the repetitive element insertion. VRN2 alleles were determined with molecular markers developed before (Yan et al., 2004b).

Photoperiod and vernalization conditions

LD conditions were 16 h of light and SD conditions were 8 h of light in all experiments. Unvernalized plants were grown at 18–25 °C at Photosynthetically Active Radiations (PAR) of 200–270 μE m⁻²s⁻¹. Vernalization experiments were performed for six weeks at 4 ºC under both SD and LD (PAR 150–190 μE m⁻²s⁻¹).

Transcription profiles

RNA samples were extracted using the TRIZOL method (INVITROGEN) from leaves as described before (Yan et al., 2003). Transcript levels of VRN1 and VRN2 in T. monococcum were determined using TaqMan systems (Yan et al., 2004b; Yan et al., 2003). ACTIN and UBIQUITIN TaqMan systems were used as endogenous controls (Yan et al., 2003).

VRN2 transcript levels in hexaploid wheat were determined using a SYBR GREEN® system designed on the basis of the A genome sequence of VRN2, using primers TaVRN2-F (5’-AGCCCACATCGTGCCATTT-3’) and TaVRN2-R (5’-ACCTCATCACCTTCGCTGCT-3’). The primers for the ACTIN SYBR GREEN® system were SYBR-Actin-F (5’-ACCTTCAGTTGCCCAGCAAT-3’) and SYBR-Actin-R (5’-CAGAGTCGAGCACAATACCAGTTG-3’). The difference in slopes between VRN2 and ACTIN systems was -0.01 passing the efficiency test (difference < 0.1). The efficiency of the VRN-A2 SYBR GREEN and ACTIN primers was higher than 99% based on six 2-fold dilutions tested in triplicate.

All quantitative PCR experiments were performed in an ABI7000. The 2^{−Δ ΔC_T} method (Livak and Schmittgen, 2001) was used to normalize and calibrate transcript values relative to the endogenous controls.

Results

Regulation of VRN2 transcript levels and flowering time by photoperiod

Exposure of plants of photoperiod sensitive winter accessions of diploid wheat G3116 (Fig. 1A) and hexaploid wheat Jagger (Fig. 1B) to six weeks of SD at non-vernalizing temperatures (20ºC) followed by the transfer of the plants to LD (referred hereafter as SD-LD) resulted in a significant acceleration of flowering in both species. The SD-LD experiment in hexaploid wheat also included transgenic RNAi::VRN2 Jagger plants with reduced levels of VRN2 (Yan et al., 2004b). The differences in flowering time observed between transgenic and non-transgenic Jagger plants under LD disappeared after the SD-LD treatment (Fig. 1B) suggesting that VRN2 was involved in the SD replacement of vernalization.

Figure 1.

A) Days to heading in winter T. monococcum accession G3116 under continuous LD or six weeks SD followed by LD. “Ver.” indicates plants vernalized for 6 weeks at 4 ºC whereas “Not ver.” indicates non-vernalized plants. Values are averages of 5 plants ± SE of the means. LD non-vernalized plants were discarded after 120 days without flowering. B) Days to heading in RNAi::VRN2 transgenic plants of hexaploid wheat ‘Jagger’ (TR) and the non-transgenic controls (NT). LD and SD indicate similar conditions to those described in ‘A’. C) VRN2 transcript levels in the same plants described in ‘B’. D) Comparison of VRN1 transcript levels in vernalized and unvernalized plants of T. monococcum G3116 grown under SD. E) Time course transcription profiles of VRN1 (red line, filled diamonds) and VRN2 (blue line, open diamonds). Initial samples (0w) were collected immediately before transferring unvernalized G3116 plants grown under LD for 7 weeks to SD (20°C). Plants were grown under SD for 6 weeks and samples were collected every week (1w to 6w). Finally, plants were transferred to LD for five weeks (1w to 5w). Note the delay in VRN1 transcription until the transfer to LD. C–G) Units for the Y-axis are linearized values using the 2^{(Δ ΔC_T)} method, where C is the threshold cycle (Livak and Schmittgen, 2001). ACTIN was used as endogenous control in C (SYBR GREEN) and UBIQUITIN in D and E (TaqMan).

The characterization of VRN2 transcript levels at the end of the SD treatment supported the previous hypothesis. In both transgenic and non-transgenic Jagger plants the VRN2 repressor was down-regulated by SD (Fig 1C), eliminating the effect of the down-regulation of VRN2 by RNAi. A time course of transcript levels of VRN2 during a SD-LD experiment revealed a rapid down-regulation of this gene by SD in diploid winter wheat G3116 (Fig. 1E). The VRN2 gene is also down-regulated by vernalization under LD (Yan et al., 2004b) suggesting that this gene might play an important role in the integration of the SD and vernalization signals.

In spite of the rapid down-regulation of VRN2 by SD observed in G3116 plants grown at room temperature, up-regulation of VRN1 transcripts was not observed until the plants were transferred to LD (Fig. 1E). This result differed from the up-regulation of VRN1 transcripts in the same genotype after vernalization, both under SD (Fig. 1D) or LD (Yan et al., 2004b). The lack of VRN1 up-regulation in the absence of VRN2 suggests the existence of a second VRN1 repressor, different from VRN2. VRN1 transcript levels were up-regulated when G3116 plants were transferred to LD after 6 weeks at SD (Fig. 1E), but not when the plants with the same genotype were maintained at continuous LD (Yan et al., 2003), suggesting the possibility that the regulation of this second repressor by photoperiod requires the previous down-regulation of VRN2.

To identify the VRN1 regulatory sites that interact with this second repressor we studied the transcript levels of VRN1 under SD in T. monococcum lines carrying different combinations of natural VRN1 and VRN2 mutants (Table 1). To facilitate the discussion of the differences in transcript levels observed among these lines, we first present a detailed characterization of the sequence differences among these lines and the genetic evidence confirming the presence of dominant Vrn1 alleles in these lines.

Sequence characterization of different dominant Vrn-1 alleles

The comparison of the complete sequence of three dominant Vrn1 alleles with the recessive vrn1 allele from accession DV92 (AY188331) and with partial sequences from other cultivated winter accessions of diploid wheat (Table 2) showed a limited number of polymorphisms, as expected from the recent origin of cultivated T. monococcum (Heun et al., 1997). The 15.7-kb region studied here for each allele includes the complete gene, the promoter region (2.3-kb) from the end of a 67-kb block of repetitive elements (AY188331) to the start codon and a 718-bp region after the stop codon.

Table 2.

Haplotype variation at the VRN1 gene in cultivated T. monococcum (15.7-kb). Sequence coordinates are based on the DV92 sequence.

Acc. No.	DV92*	PI 355516	PI 355534	PI 503874*	PI 326317*	PI 306540*
GeneBank	AY18833	-	-	DQ146421	DQ146422	DQ146423
Growth habit	Spring	Winter	Winter	Spring	Spring	Spring
SD induc. VRN1	No	No	No	Yes	Yes	No
VRN1 allele	vrn1	vrn1	vrn1	Vrn1f	Vrn1g	Vrn1h
VRN2 allele	vrn2a	Vrn2	Vrn2	Vrn2	vrn2b	Vrn2
Prom. -1377 1	G	G	G	G	G	A
Prom. -978	wild	wild	wild	wild	wild	9-bp del.
CArG-box	wild	wild	wild	1-bp del.	34-bp del.	wild
Int. 1 +950	wild	wild	wild	502-bp ins.	wild	493-bp ins.
Int. 1 +6447	T	G	T	G	G	G
Int. 2 +192	T	C	C	C	C	C

¹Within a foldback repetitive element located -1,026 to -1,424-bp upstream of start codon.

Figure 2 shows the location of the 1-bp deletion in T. monococcum accession PI 503874 (Vrn1f, GenBank DQ146421) and the 34-bp deletion in accession PI 326317 (Vrn1g, GenBank DQ146422) relative to the putative CArG-box (bold letters). The VRN1 promoter of accession PI 306540 (Vrn1h, GenBank DQ146423) exhibits two unique polymorphisms, but both of them far from the start codon (Table 2). The Vrn1h and Vrn1f alleles have a 493-bp insertion of a repetitive element differing only in one 9-bp indel (Fig. 2 underlined). PCR analysis with primers VRN1INT1_F and _R confirmed that the seven accessions carrying the CArG-box 1-bp deletion all have an intron insertion similar to PI 503874 (Table 1).

Fig. 2.

Sequence differences among VRN1 alleles. Top: 74-bp segment of the VRN1 promoter showing the location of the putative CArG box in blue and the CAP signal for transcriptional initiation underlined. Middle: schematic representation of the VRN1 gene in T. monococcum. Bottom: repetitive element inserted in VRN1 first intron in PI 306540 (Vrn1h) and PI 503874 (Vrn1f).

The most likely location of the regulatory region within the VRN1 intron 1 was previously delimited to a 4-kb region by overlapping several independent deletions (Fu et al. 2005). The repetitive element found in the intron of the Vrn1f and Vrn1h alleles is inserted 200-bp upstream from this critical region. However, it has been shown in Arabidopsis that the effect of the insertion of a repetitive element in the first intron of FLC has significant effects on transcription in regions beyond the insertion site (Liu et al., 2004). Therefore, this intron insertion is still a candidate for the determination of the spring growth habit associated with the Vrn1h allele. However, we cannot rule out the other two exclusive mutations in the promoter of Vrn1h (Table 2) as a cause of the spring growth habit.

Genetic analyses of the dominant Vrn1 alleles

The presence of dominant Vrn1 alleles in spring T. monococcum accessions PI 326317 (Vrn1g), PI 503874 (Vrn1f), and PI 306540 (Vrn1h) was confirmed in three different mapping populations including G3116 as the winter parent. All F1 hybrids had spring growth habit confirming that the alleles present in these three lines were dominant for spring growth habit. In each population we tested the linkage between growth habit and polymorphisms at the VRN1 locus in greenhouse experiments (20–25 ºC) under LD conditions.

Linkage between spring growth habit and the 1-bp deletion / intron one insertion found in PI 503874 (Vrn1f Vrn2) was confirmed in a population of 144 F2 individuals from the cross PI 503874 x G3116. We observed 110 spring and 34 winter plants, fitting a 3:1 segregation for a single dominant gene (P= 0.70). The dCAPS marker for the 1-bp deletion showed that all the spring plants were homozygous or heterozygous for the Vrn1f allele whereas all winter plants were homozygous for the recessive vrn1 allele, confirming that the PI 503874 has a dominant Vrn1 allele.

Linkage between spring growth habit and the 34-bp deletion found in PI 326317 was investigated in a population of 71 F2 individuals (PI 326317 x G3116). In addition to the 34-bp deletion in the CArG box, PI 326317 has a recessive vrn2b allele and therefore this population was expected to segregate for both VRN1 and VRN2. The observed segregation of 55 spring and 16 winter plants agreed with the expected 13:3 segregation for one dominant (Vrn1g) and one recessive (vrn2b) allele for spring growth habit (P= 0.41). All 16 winter plants were homozygous for the recessive vrn1 allele (absence of the 34-deletion) confirming the presence of a dominant Vrn1 allele in PI 326317.

Linkage between spring growth habit and the 493-bp insertion in the first intron of PI 306540 (Vrn1h Vrn2) was confirmed in a population of 150 F2 individuals from the cross PI 326317 x G3116. The observed 115 spring and 35 winter plants agreed with the expected 3:1 segregation for one dominant gene for spring growth habit (P= 0.64). All the winter plants were homozygous for the absence of the intron one insertion (vrn1 allele) confirming the presence of a dominant Vrn1 allele in PI 306540. Twenty one out of the 25 earliest flowering lines were homozygous for the Vrn1h allele (the other four were heterozygous) suggesting the existence of some dosage effect for the Vrn1h allele.

Regulation of VRN1 transcript levels under SD in lines carrying different vernalization alleles

Analysis of the T. monococcum lines carrying mutations in the different VRN1 regulatory regions (Tables 1 and 2) under SD pointed to the promoter CArG-box as the most likely interaction site of the second VRN1 repressor. After six weeks at SD, lines carrying a mutated CArG-box (1-bp deletion: Vrn-1f allele, or 34-bp deletion: Vrn-1g allele) exhibited significantly higher VRN1 transcript levels than the spring lines carrying the dominant Vrn-1h allele or recessive vrn-1 alleles in combination with different nonfunctional vrn2 alleles (Fig. 3A). A similar result was observed in an independent experiment including additional T. monococcum lines for the different allelic combinations (Fig. 3C).

Fig. 3.

A) VRN1 transcript levels in T. monococcum spring lines carrying Vrn1f (PI 503874, 1-bp deletion), Vrn1g (PI 349049, 34-bp deletion), Vrn1h (PI 306540, intron one insertion), and recessive vrn1 alleles combined with recessive vrn-2b (PI 323437) and vrn2a (DV92) alleles. B) Comparison of homozygous vrn1 and Vrn1 F2 plants from the cross between DV92 (vrn1 vrn2a) with PI 266844 (Vrn1f, Vrn2) or PI 326317 (Vrn1g vrn2b). We first selected plants homozygous for the recessive vrn2 allele using molecular markers to avoid differences in growth habit. C) VRN1 transcript levels in additional T. monococcum lines carrying different vernalization alleles. A–C) Units for the Y-axis are linearized values using the 2^{(−Δ ΔC_T)}method, where C is the threshold cycle (Livak and Schmittgen, 2001). TaqMan system for ACTIN (A) and UBIQUITIN (B and C) were used as endogenous controls. D–E) Apex development in plants of T. monococcum F2 plants from the cross between PI 266844 (Vrn1f) x DV92 (vrn1). Five homozygous lines with and without the 1-bp deletion in the CArG-box were selected within the homozygous vrn2a lines (spring growth habit). Apexes were observed D) after six weeks under SD (8 h of light) and E) two weeks after transferring the plants back to LD.

The association between the Vrn1 CArG-box mutations and the SD induction of VRN1 at room temperature was confirmed in two F2 populations (Fig. 3B). F2 plants carrying the mutated CArG-box (Vrn1f or Vrn1g) showed significantly higher VRN1 transcript levels than the F2 plants homozygous for the recessive vrn1 allele (Fig. 3B). Comparison of the complete VRN1 gene and promoter sequence among these alleles confirmed that except for the CArG-box, none of the other polymorphisms were associated with the acceleration of VRN1 transcripts under SD (Table 2).

The earlier up-regulation of VRN1 in plants carrying CArG-box mutations was also reflected in faster apex development. After six weeks under SD, the apices of the T. monococcum F2 plants without the CArG-box mutations were still in the vegetative stage whereas the apices of the plants with the CArG-box mutations were at the double-ridge stage (Fig. 3D). Two weeks after transferring the plants to LD, the apices of the plants without the CArG-box mutation were just initiating their differentiation whereas those with the CArG-box mutations were already at the terminal spikelet stage (Fig. 3E).

The impact of the different alleles on the apices response to photoperiod was confirmed in an independent experiment, in which lines with the vrn1 (G3116), Vrn1f (PI 503874) and Vrn1h (PI 306540) alleles were grown for 10 weeks under SD and then divided in two groups and exposed to different photoperiod treatments. The first group was exposed to SD for additional 4 weeks whereas the second group was transferred to LD for 4 weeks (Fig. 4). Control winter plants (G3116) maintained for 14 weeks at LD showed no sign of apex differentiation (Fig. 4A), whereas spring plants (Vrn1f and Vrn1h) were already heading after 14 weeks at LD (not shown). The plants with the recessive vrn1 allele (Fig. 4B) or the dominant Vrn1h allele (Fig. 4C) showed limited induction of apex differentiation after 14 weeks under SD. On the contrary, the plants carrying the Vrn1f allele showed a clear apex differentiation into the early stages of spike development. Plants with the vrn1 (Fig. 4E) and Vrn1h alleles (Fig. 4F) moved for 4 weeks to LD have started their spike development, but were significantly behind the plants with the Vrn1f allele (Fig. 4G). The Vrn1f plants under SD-LD were also significantly more advanced in their reproductive development than the Vrn1f plants of the same age kept under continuous SD (Fig. 4D).

Fig. 4.

Comparison of apex development in T. monococcum plants. A–B and E) Winter G3116 (vrn1); C and F) PI 306540 (Vrn1h); D & G) PI 266844 (Vrn1f). A) Continuous LD, vegetative stage. Under the same conditions lines with the dominant Vrn1h and Vrn1f alleles had already headed (not shown). B, C, D) Continuous SD. E, F, G) 10 weeks at SD and 4 weeks at LD (SD-LD). Note the more advanced development of the inflorescence in Vrn1f relative to vrn1 and Vrn1h in both SD and SD-LD. White bars represent 0.5 mm.

Discussion

Effect of mutations in the VRN1 promoter on its regulation by photoperiod

The absence of up-regulation of VRN1 after the down regulation of VRN2 by SD at room temperature indicated the existence of a second VRN1 repressor, different from VRN2. The VRN1 haplotype analysis, including the complete promoter region, suggests that the mutations affecting the CArG-box are the most likely regions associated with the interactions between VRN1 and this second repressor SD (Table 2). The lines carrying the 1-bp mutation in the CArG-box also have a repetitive element inserted in the first intron providing an alternative putative regulatory site. However, the line with the Vrn1h allele that has the same intron insertion but no CArG-box mutation, showed no increase of VRN1 transcripts under SD. In addition, the lines with the Vrn1g allele (34-bp deletion) that have no insertion in the first intron show increased levels of VRN1 transcripts under SD (Fig. 3A). These results indicate that the intron one insertion does not affect the site regulating this response to SD. On the contrary, all the genotypes showing the 1-bp or 34- bp in the VRN1 promoter CArG-box (Fig. 3C), and the F2 lines carrying either of these alleles showed a significant increase of VRN1 transcripts under SD (Fig. 3B). Taken together, these results suggest that the CArG-box is a critical regulatory site for the response of VRN1 to SD. CArG-boxes are generally recognized by MADS-box genes, so it is possible that the second repressor is a MADS-box gene.

Allelic variation for the insertion within the VRN1 first intron does not seem to be important in the up-regulation of VRN1 under SD. The intron one site seems to be more important in the regulation of VRN1 by VRN2. A comparison of two different VRN1- VRN2 epistatic interaction studies suggested that a large Vrn-H1 intron one deletion in barley has larger epistatic effects on VRN2 (Dubcovsky et al., 2005) than a 20-bp deletion adjacent to the VRN-A^m1 CArG-box in T. monococcum (Tranquilli and Dubcovsky, 2000). However, the fact that mutations in either of these sites are sufficient to accelerate flowering under LD suggests the existence of interactions between these two mechanisms in the regulation of VRN1 (Dubcovsky et al., 2005; Fu et al., 2005; Yan et al., 2004a). The presence of two interacting regulatory sites has been also described in the MADS-box FLC gene in Arabidopsis (Sheldon et al., 2002). Mutations in the regulatory regions located in FLC promoter or first intron are sufficient to induce a spring growth habit in Arabidopsis, a similar result to that reported for the wheat VRN1 MADS-box gene.

Effect of CArG-box mutations on apex development

The earlier induction of the meristem identity gene VRN1 under SD in the spring lines carrying the CArG-box mutations was associated with an earlier differentiation of the reproductive apices compared to spring plants with the dominant Vrn1h allele (intron insertion). These differences in apex development can affect the adaptability of wheat to different environments. Plants showing SD induction of VRN1 and early apex differentiation might be favorable in regions with mild winters that do not pose a significant risk for the delicate floral meristems. In agreement with this hypothesis most of the plants carrying these mutations were collected from temperate regions such as Spain, Portugal, Israel, U.K. (accession PI 10474 was, however, from Germany), whereas those carrying the Vrn1h allele were collected in regions subjected to severe winters such as Romania and Russia.

Although the acceleration of VRN1 transcription under SD in genotypes with CArG-box mutations resulted in an earlier differentiation of the apices, the spike development under continuous SD (Fig. 4D) proceed slowly compared with that in plants transferred to LD (Fig. 4G). This result suggests that that there is an additional regulatory point after the induction of VRN1, which requires LD to proceed with a normal and timely reproductive development. The faster apex development under LD than under SD (in the absence of VRN2) provides a possible explanation for the slightly earlier flowering of the G3116 plants after LD vernalization than after SD vernalization (Fig. 1A). It is important to point out here, that the T. monococcum accessions used in this study are photoperiod sensitive and that the responses described here might be different in photoperiod insensitive plants.

SD replacement of vernalization

Using a positional cloning approach we demonstrated before that the ZCCT1 gene is VRN2 and showed that deletions or mutations in this gene were responsible for natural differences in growth habit (Yan et al., 2004b, Dubcovsky et al. 2005). We also demonstrated that the down-regulation of VRN2 by RNAi produced a significant acceleration of flowering in transgenic Jagger plants relative to the non-transgenic controls (Yan et al., 2004b). Therefore, the down-regulation of VRN2 by SD described here is also expected to result in the acceleration of flowering and the elimination of the vernalization requirement. The convergence of photoperiod and vernalization signals at the VRN2 gene, provides a possible explanation to the interchangeability of SD and vernalization treatments in winter wheats, and likely in other temperate cereals. Krekule (1964) analyzed the responses of different wheat varieties to the SD replacement of vernalization and concluded that this response was restricted to photoperiod sensitive winter varieties.

Since photoperiod sensitivity and vernalization requirement are ancestral traits in wheat and barley (Evans and Blundell, 1994; Halloran, 1967; Karsai et al., 2004; Kihara, 1958; Yan et al., 2004b; Yan et al., 2003), the dual SD-LD induction of flowering is likely an ancestral trait in these crop species. This dual flowering induction requirement is also characteristic of most winter grasses of the subfamily Festucoideae (Heide, 1994). Therefore, it is likely that the strong selective pressures applied to the Triticeae crops during domestication and modern breeding accelerated the loss of the dual SD-LD flowering induction in a large number of varieties (Dubcovsky et al., 2005; Fu et al., 2005; Karsai et al., 2004; vonZitzewitz et al., 2005; Yan et al., 2004b; Yan et al., 2003). The absence of SD flowering induction in numerous modern wheat varieties has led to the general classification of wheat as LD plant. However, it would be more correct to refer to this species as SD-LD plant as suggested before by other researchers (Evans, 1987; McKinney and Sando, 1935). The down-regulation of the VRN2 repressor by SD reported here is likely part of the mechanism associated with this SD-LD classification.

Abbreviations

SD

short day (8h light)

LD

long day (16 h light)

PAR

Photosynthetically Active Radiation

RNAi

RNA interference