Functional characterisation of HvCO1, the barley (Hordeum vulgare) flowering time ortholog of CONSTANS

Authors


(fax +49 221 506 2207; e-mail korff@mpipz.mpg.de).

Summary

Variation in photoperiod response is a major factor determining plant development and the agronomic performance of crops. The genetic control of photoperiodic flowering has been elucidated in the model plant Arabidopsis, and many of the identified genes are structurally conserved in the grasses. In this study, HvCO1, the closest barley ortholog of the key photoperiod response gene CONSTANS in Arabidopsis, was over-expressed in the spring barley Golden Promise. Over-expression of HvCO1 accelerated time to flowering in long- and short-day conditions and caused up-regulation of HvFT1 mRNA under long-day conditions. However, the transgenic plants retained a response to photoperiod, suggesting the presence of photoperiod response factors acting downstream of HvCO1 transcription. Analysis of a population segregating for HvCO1 over-expression and natural genetic variation at Ppd-H1 demonstrated that Ppd-H1 acts downstream of HvCO1 transcription on HvFT1 expression and flowering. Furthermore, variation at Ppd-H1 did not affect diurnal expression of HvCO1 or HvCO2. Over-expression of HvCO1 increased transcription of the spring allele of Vrn-H1 in long- and short-day conditions, while genetic variation at Ppd-H1 did not affect Vrn-H1 expression. Over-expression of HvCO1 and natural genetic variation at Ppd-H1 accelerated inflorescence development and stem elongation. Thus, HvCO1 probably induces flowering by activating HvFT1 whilst Ppd-H1 regulates HvFT1 independently of HvCO1 mRNA, and all three genes also appear to have a strong effect in promoting inflorescence development.

Introduction

The time of flowering is crucial for the adaptation of plants to a given environment and has a major impact on grain yield in crop species (Cockram et al., 2007a). Seasonal changes in photoperiod are major cues controlling development in many plant species. The photoperiod flowering pathway channels inputs from light, day length and the circadian clock to promote the floral transition. In Arabidopsis, CONSTANS (CO) is a central regulator of this pathway, triggering the transcription of the gene encoding the mobile florigen hormone FLOWERING LOCUS T (FT). The FT protein moves from the leaves through the phloem to the shoot apical meristem where it induces the switch from vegetative to reproductive growth (Corbesier et al., 2007; Tamaki et al., 2007; Mathieu et al., 2007; Jaeger and Wigge, 2007). CO is regulated at the transcriptional level by several genes that are part of the circadian clock or are under circadian clock control, so that CO mRNA accumulates at the end of a long summer day. At the protein level CO is regulated by photoreceptors and the ubiquitin ligase CONSTITUTIVE PHOTOMORPHOGENIC 1 (COP1) that respectively stabilise CO in light or de-stabilize CO in the dark (Jang et al., 2008). As CO transcription occurs before dusk on long days (LD) but after dusk on short days (SD), CO protein only accumulates and mediates transcription of FT under LD (Turck et al., 2008).

Orthologous genes of CO have been identified in many species, suggesting conservation of the components of the Arabidopsis photoperiod pathway. In both monocot and dicot species, CO activates FT or related genes under inductive day lengths to promote flowering (Suárez-López et al., 2001; Hayama et al., 2003). Heading date 1 (Hd1), the rice ortholog of CO, promotes heading under SD conditions through the induction of Hd3a, which encodes a protein closely related to Arabidopsis FT (Yano et al., 2000; Kojima et al., 2002; Hayama et al., 2003). Griffiths et al. (2003) identified nine CO orthologs in barley (Hordeum vulgare), of which HvCO1 shows the closest orthology to CO and Hd1 based on the conserved CCT (CO, CO-like, TOC1) domain near the carboxy terminus and the two zinc finger B-boxes near the amino terminus. However, barley was previously shown to differ from rice in having two paralogous CO-like genes (HvCO1 on chromosome 7H and HvCO2 on chromosome 6H), of which only HvCO1 is collinear with the rice CO ortholog Hd1 (Higgins et al., 2010).

The role of CO orthologs in regulating flowering time in temperate cereals has not been elucidated, as no genetic variants in these genes or transgenic plants affecting their expression have been described. Griffiths et al. (2003) found that none of the CO orthologs in barley coincided with known flowering time quantitative trait loci (QTLs). On the other hand, Turner et al. (2005) suggested that variation in the major barley photoperiod response gene Ppd-H1 affected flowering time through shifting the diurnal expression peaks of HvCO1 and HvCO2 mRNA into the dark phase. Ppd-H1 is orthologous to the Arabidopsis clock gene PRR7, and a recessive mutation in the CCT domain of Ppd-H1 causes photoperiod insensitivity and late flowering in spring barley (Turner et al., 2005). Wild and cultivated winter barley genotypes in contrast, carry the photoperiod responsive Ppd-H1 allele, which induces early flowering in long photoperiods. Winter and spring barley genotypes are also distinguished by allelic variation at Vrn-H1 and Vrn-H2, where deletions of the entire Vrn-H2 gene and in the first intron of Vrn-H1 cause vernalisation insensitivity in spring barley (Yan et al., 2004; Hemming et al., 2008).

In temperate cereals, the photoperiod and vernalisation pathways are closely intertwined. In wheat, the TaFT protein regulates transcription of the vernalisation-responsive MADS box transcription factor VRN1 in the leaf (Li and Dubcovsky, 2008). Expression of VRN1 and TaFT is necessary to induce flowering under LD conditions (Shimada et al., 2009). In addition, Kane et al. (2007) have proposed that VRN1 expression in wheat is also controlled by the floral repressor TaVRT2, which is another MADS box transcription factor and the cereal ortholog of SHORT VEGETATIVE PHASE (SVP) in Arabidopsis (Hartmann et al., 2000). VRT2 was shown to bind to the CArG-box in the VRN1 promoter and was proposed as a vernalisation-regulated repressor of VRN1 (Kane et al., 2005, 2007). However, Trevaskis et al. (2007) have shown that SVP orthologs HvVRT2, HvBM1 and HvBM10 did not affect Vrn-H1 expression levels and primarily delayed development after floral transition. These examples, suggest a close connection between the photoperiod and vernalisation pathways; however, the genetic interactions between photoperiod response and vernalisation genes are not yet clear.

In the present study, we functionally characterise HvCO1 as the closest barley CO ortholog and its interaction with genetic variation at Ppd-H1 and Vrn-H1, to further unravel the photoperiod response pathway in barley. We demonstrate that over-expression of HvCO1 induces HvFT1 expression and accelerates time to flowering. In addition, our data suggest that variation at Ppd-H1 acts downstream of HvCO1 mRNA on expression of HvFT1 and flowering. Over-expression of HvCO1 also up-regulated the spring allele of Vrn-H1, which suggested a photoperiod control of this vernalisation-responsive gene. Out data thus demonstrate that HvCO1 functions in the control of flowering time in barley; however, the strong effect of Ppd-H1 and control of Vrn-H1 expression indicate modifications of the photoperiod response pathway in barley as compared to Arabidopsis.

Results

Over-expression of HvCO1 promotes flowering in barley but not in Arabidopsis

We tested whether HvCO1 promotes flowering in barley by over-expressing HvCO1 under the control of the maize ubiquitin promoter in the spring barley Golden Promise. Golden Promise carries the mutated, photoperiod-insensitive ppd-H1 allele and the spring alleles at Vrn-H1 and vrn-H2. Ten independent transgenic lines (4–12 plants per line) in T2 were tested for flowering time measured as days to heading under LD and SD photoperiods (Figure 1). Transgenic lines flowered on average 42 days after sowing and non-transgenic controls as well as the Golden Promise progenitor 62 days after sowing under LD conditions. Under SD conditions, transgenic lines flowered on average after 85 days, whereas non-transgenic control as well as Golden Promise plants had not flowered by 130 days after sowing, when the experiment was stopped. Over-expression of HvCO1 thus promoted flowering under LD and SD conditions, but flowering time under SD was significantly delayed compared to LD.

Figure 1.

 Flowering time of Ubi::HVCO1 lines under long-day (LD) and short-day (SD) conditions.
Ten independent Ubi::HVCO1 lines (L19 to L 31), the null segregant control (null) and the wild type (GP, Golden Promise) were grown under LD (grey bars; 16-h light) and SD (black bars; 8-h light). Flowering time is measured in days to heading; the null segregants and the wild type under SD did not flower until the experiment was stopped (NF, 130 days). Bars represent the average of 4–10 plants ± standard error.

To test whether HvCO1 also promotes flowering in Arabidopsis thaliana, we expressed HvCO1 under the CaMV35S promoter in the Arabidopsis co-2 mutant. Two independent transgenic families carrying 35S::HvCO1, the wild-type accession Landsberg erecta (Ler), 35S::HA-COco-2 and the Arabidopsis co-2 mutant were scored for the number of leaves at bolting under LD and SD (10–12 plants per genotype/condition). The transgenic lines over-expressing HvCO1 did not differ in development from the Arabidopsis co-2 mutant, while Arabidopsis lines over-expressing CO in the co-2 mutant background were early flowering (Figure S1a in Supporting Information). Reverse transcriptase-PCR analysis confirmed that the mRNAs of HvCO1 and CO were over-expressed, but only over-expression of CO induced over-expression of FT in Arabidopsis (Figure S1b–d). Therefore, over-expressing HvCO1, a CO-like gene predicted to contain only one functional B-box (Griffiths et al., 2003), did not promote FT expression and flowering in Arabidopsis.

Candidate gene expression in independent Ubi::HvCO1 plants

Gene expression of candidate genes was analysed in the independent Ubi::HvCO1 lines at the end of the day under LD and SD, when HvCO1 and HvFT1 mRNAs peak in expression. Expression analysis confirmed that HvCO1 was over-expressed under LD and SD. HvCO1 mRNA was three to ten times more abundant in transgenic plants compared with the non-transgenic controls (Figure 2a). Over-expression of HvCO1 in the transgenic lines did show significant differences between LD and SD; however, these only explained 8% of the overall expression variance (Table S1). HvFT1 was up-regulated at least five-fold in the transgenic lines under LD. Under SD, HvFT1 mRNA levels were also up-regulated in two lines but to a much smaller extent than under LD (Figure 2c).

Figure 2.

 Expression of flowering time genes in Ubi::HVCO1 lines under long-day (LD) and short-day (SD) conditions.
Ten independent Ubi::HVCO1 lines (L19 to L31), the null segregant control (null) and the wild type (GP, Golden Promise) were grown under LD (grey bars; 16-h light) and SD (black bars; 8-h light). Bars represent the average expression of the target gene normalised to HvActin of two technical replicates ± standard error.

In addition, Vrn-H1 was significantly up-regulated in transgenic lines compared with the non-transgenic control and Golden Promise under both LD and SD. However, there was a large variation in expression between transgenic lines. Transgenic line 30, for example, exhibited expression levels lower than the non-transgenic control (Figure 2d). HvVRT2 was significantly down-regulated in the transgenic lines as compared with the null segregants and Golden Promise under LD, but not under SD (Figure 2e). HvBM1 expression was significantly lower in the wild type and null segregants than transgenics under LD, but not under SD (Figure 2f). HvCO2 and HvBM10 expression levels were not significantly different between transgenic and wild-type lines and were characterized by a large variation in expression between transgenic lines (Figure 2b,g, Table S1). Expression of HvFT2, HvFT3 and HvFT4 was not detected in any genotype.

Over-expression of HvCO1 and variation at Ppd-H1 accelerate inflorescence development

The phenotypic effects of HvCO1 over-expression on meristem development were scored according to the Waddington scale in two independent Ubi::HvCO1 lines (19, 30) and Golden Promise. Plants were grown under SD for 21 days to allow for establishment of all transgenic plants, and then either transferred to LD or kept under SD. After transfer, replicate primary shoots were dissected every 2–3 days under LD and every 4–5 days under SD for all genotypes.

Transition to the reproductive stage as indicated by the formation of a ‘double ridge’ on the spike primordium (Waddington stage 2) occurred 8 days after and the beginning of stem elongation (Waddington stage 3) approximately 11 days after transfer to LD in transgenic and wild-type plants (Figures 3a and S2). However, stem elongation was significantly faster in transgenic lines than in Golden Promise. Ubi::HvCO1 lines flowered 29 days after and Golden Promise more than 43 days after transfer to LD. Under SD conditions, the Ubi::HvCO1 lines and Golden Promise showed similar patterns of meristem development as observed under LDs, so that genetic differences in meristem development were observed after the beginning of stem elongation (Waddington scale 3, Figure 3b).

Figure 3.

 Effect of HvCO1 over-expression and the Ppd-H1 mutation on meristem development.
Meristem development (Waddington scale) of two independent Ubi::HVCO1 lines (dotted and dashed lines) and Golden Promise (solid line) under long days (LD; a, 16-h light) and short days (SD; b, 10-h light) and of the spring barley cultivar Scarlett (ppd-H1, solid line) and a derived near isogenic line S42IL-107 (Ppd-H1, dashed line) under LD (c) and SD (d). The dashed lines represent the transition to the reproductive meristem (Waddington scale 2) and the beginning of stem elongation (Waddington scale 3). Values represent the average of three plants ± standard error.

Phenotypic effects of variation at Ppd-H1 on meristem development were also analysed between the spring barley Scarlett (ppd-H1) and derived introgression lines (Ppd-H1). Like in Ubi::HvCO1 lines, transition to a reproductive meristem occurred 8 days after and stem elongation (Waddington stage 3) began 11 days after transfer to LD in Scarlett and S42-IL107 (Figures 3c and S2). However, stem elongation and inflorescence development were accelerated in S42-IL107 as compared with Scarlett, flowering occurred after 29 days in S42-IL107 and after 43 days in Scarlett. In contrast to Ubi::HvCO1 and Golden Promise, meristem development of Scarlett and S42IL-107 did not differ under SD, and neither line flowered under SD (Figure 3d). Over-expression of HvCO1 and natural variation at Ppd-H1 thus had a major effect on stem elongation and flowering under LD, and over-expression of HvCO1 had similar effects under SD.

Over-expression of HvCO1 and the active Ppd-H1 allele up-regulate HvFT1 and Vrn-H1 expression

We tested the effects of HvCO1 over-expression on the diurnal expression profiles of barley flowering time genes in leaf material of Ubi::HvCO1 line 19 and Golden Promise under LD and SD. HvCO1 was over-expressed in the Ubi::HvCO1 line as compared with Golden Promise at all time points under LD and SD (Figure 4a). HvCO2 expression was not significantly different between transgenic and wild-type barley lines under LD and SD (Figure 4b). HvFT1 was up-regulated in the Ubi::HvCO1 line as compared with Golden Promise under LD, while HvFT1 mRNA levels were very low in both lines under SD (Figure 4c). Vrn-H1 expression levels were higher in Ubi::HvCO1 than Golden Promise under LD and SD (Figure 4d). HvVRT2, HvBM1 and HvBM10 exhibited a higher expression in Golden Promise than in Ubi::HvCO1 (Figure 4e–g). Consequently, over-expression of HvCO1 up-regulated HvFT1 mRNA levels only under LD, while it increased Vrn-H1 expression and down-regulated HvVRT2, HvBM1 and HvBM10 under LD and SD.

Figure 4.

 Diurnal expression of flowering time genes in Ubi::HVCO1 lines under long-day (LD) and short-day (SD) conditions.
One Ubi::HVCO1 lines (triangles) and the wild type (Golden Promise, squares) were grown under LD (solid lines; 16-h light) and SD (dashed lines; 8-h light). Transcript accumulation was measured at 2 to 4-h intervals by quantitative real-time RT-PCR analysis of specific genes and normalised to HvActin. Values represent the average of two technical replicates ± standard error.

The effects of variation at Ppd-H1 on diurnal expression patterns of flowering time genes were tested in Scarlett (ppd-H1) and S42-IL107 (Ppd-H1) under LD, as these lines only showed developmental differences under LD. The mutation in Ppd-H1 did not affect the diurnal timing of HvCO1 expression, but HvCO1 mRNA levels were reduced in S42-IL107 compared with Scarlett with the mutated ppd-H1 allele (Figure 5a). Similarly, diurnal expression patterns of HvCO2 did not show clear diurnal differences between the two different Ppd-H1 genotypes (Figure 5b). HvFT1 and Vrn-H1 mRNA levels were up-regulated in S42IL-107 compared with Scarlett (Figure 5c,d). HvFT2 was expressed at low levels in S42IL-107, but not in Scarlett (data not shown). HvVRT2, HvBM1 and HvBM10 exhibited a higher expression in Scarlett than in S42IL-107 (Figure 5e–g). Consequently, substitution of the mutated ppd-H1 allele by the photoperiod-sensitive Ppd-H1 allele and over-expression of HvCO1 had similar effects on downstream genes under LD, the up-regulation of HvFT1 and Vrn-H1 and down-regulation of HvVRT2, HvBM1 and HvBM10. In addition, natural genetic variation at Ppd-H1 did not shift diurnal expression peaks of HvCO1 and HvCO2 mRNAs.

Figure 5.

 Diurnal expression of flowering time genes in the spring cultivar Scarlett (ppd-H1) and the introgression line S42IL-107 (Ppd-H1) under long days (LD).
Scarlett (solid line) and the derived near isogenic line S42IL-107 (dashed line) were grown under LD (16-h light). Transcript accumulation was measured at 2-h intervals by quantitative real-time RT-PCR analysis of specific genes and normalised to HvActin. Values represent the average of two technical replicates ± standard error.

Genetic analysis of flowering time in a population segregating for Ubi::HvCO1, Ppd-H1 and Vrn-H1

Inheritance of phenotypic variation.  We tested for interactions between HvCO1 over-expression and variation at Ppd-H1 and Vrn-H1 in an F2 population. This population was generated by crossing Ubi::HvCO1 in the background of Golden Promise with an introgression line derived from the winter barley Igri. This line carries the winter alleles at all flowering time loci, including Ppd-H1 and Vrn-H1, but the spring allele at the vrn-H2 locus as detected in the cross Triumph × Igri (Laurie et al., 1995). An analysis of variance testing for the effects of the three known segregating genes under LD conditions, showed that variation at Ppd-H1 explained most (32%) of the phenotypic variance in days to flowering observed in the population (Table 1). Presence or absence of the transgene only explained 21%, and genetic variation at Vrn-H1 (spring/winter allele) 20% of the variation in flowering time. The interaction between Ppd-H1 and HvCO1 over-expression was significant, where the delay in flowering time in lines homozygous for the insensitive ppd-H1 was counteracted by HvCO1 over-expression (Figure 6a). The appearance of the first node on the stem which marks the beginning of stem elongation was affected by variation at Vrn-H1 (45%) and at Ppd-H1 (35%). F2 lines with the sensitive Ppd-H1 or spring Vrn-H1 allele formed a first node significantly earlier than lines with the mutated ppd-H1 or winter vrn-H1 allele, while over-expression of HvCO1 did not have a significant effect on the timing of this trait (Figure 6a,b). Ppd-H1 and Vrn-H1 had additive effects on flowering time, while variation at Vrn-H1 only affected formation of the first node in the presence of a sensitive Ppd-H1 allele (Figure 6c).

Table 1. anova– association of days to heading (DH), appearance of the first node on the stem (Node) and gene expression of HvCO1, HvFT1, Vrn-H1, HvVRT2, HvBM1 and HvBM10 with genetic variation at HvCO1 (Ubi::HvCO1), Ppd-H1 and Vrn-H1 and respective interactions under long-day (LD) and short-day (SD) conditions
 DHNodeHvCO1HvFT1Vrn-H1HvVRT2HvBM1HvBM10
F-statR2F-statR2F-statR2F-statR2F-statR2F-statR2F-statR2F-statR2
  1. F-stat, F statistics as calculated in the anova; R2, percentage of phenotypic variance explained by the respective factor.

  2. Significant effects: *< 0.5; **< 0.01; ***< 0.001.

LD
 HvCO143***0.2110.0528***0.5721***0.1410**0.139**0.190.30.010.20.01
 Ppd-H166***0.327**0.3500.0145***0.3010.015*0.110.30.011.20.04
 Vrn-H120***0.204**0.4510.0417***0.2321***0.524*0.180.80.071.30.08
 HvCO1 × Ppd-H127***0.1310.0430.0640.0210.0230.061.10.050.50.02
 HvCO1 × Vrn-H130.0300.0300.0150.065**0.1110.041.20.103.8*0.25
 Ppd-H1 × Vrn-H130.0310.1400.0110**0.1310.0230.121.80.151.80.12
SD
 HvCO179***0.87  41***0.750.10.0100.0100.016.6*0.281.10.07
 Ppd-H100.00  10.020.10.0100.0110.033.20.140.40.03
 Vrn-H110.01  10.020.90.153*0.4030.330.20.020.20.03
 HvCO1 × Ppd-H100.00  10.020.00.0000.0100.001.00.040.10.01
 HvCO1 × Vrn-H100.01  00.010.10.0200.0210.101.60.131.00.14
 Ppd-H1 × Vrn-H110.02  00.000.40.0700.0200.010.20.010.90.12
Figure 6.

 Interaction effects between HvCO1 over-expression and natural genetic variation at Ppd-H1 and Vrn-H1 on flowering time (light grey bars) and stem elongation (dark grey bars).
The F2 population HvCO1-ox × Igri (vrn-H2) derived by the cross between an introgression line in the background of the winter barley Igri (Ppd-H1, vrn-H1, vrn-H2) and two Ubi::HVCO1 lines (ppd-H1, Vrn-H1, vrn-H2) was grown for 21 days under SD and then moved to LD and flowering time was recorded. The bars represent the average flowering time (grey bars) or time to appearance of first node (dark bars) for the different allelic combinations of HvCO1, Ppd-H1 and Vrn-H1 (±standard error). Letters indicate significant differences between phenotypic values of allelic classes.

Under SD, flowering time was controlled by over-expression of HvCO1, but not by variation at Ppd-H1 and Vrn-H1.

Inheritance of expression variation.  Under LD, HvFT1 expression was primarily controlled by Ppd-H1 (30%), while genetic variation at Vrn-H1 and the interaction between Ppd-H1 and Vrn-H1 explained 23 and 13% of the expression variance, respectively (Table 1). Over-expression of HvCO1 also had a significant, but smaller (14%), effect on HvFT1 expression. Expression of HvFT1 revealed the highest correlations with flowering time and the beginning of stem elongation (appearance of first node) in the F2 population (Table S2). Vrn-H1 expression was mainly controlled by genetic variation at Vrn-H1 (52%) and by over-expression of HvCO1 under LD (13%). Interestingly, the interaction between Vrn-H1 and the transgene was significant, where over-expression of HvCO1, but not variation at Ppd-H1, induced expression of the spring allele at Vrn-H1 (Figure 7). The expression of HvVRT2 was primarily controlled by over-expression of HvCO1 (19%), but also by variation in Ppd-H1 (11%) and Vrn-H1 (18%). HvVRT2 showed the highest negative correlation to HvFT1 (−0.72) and Vrn-H1 (−0.55) expression and was positively correlated with flowering time (0.48) and the beginning of stem elongation (0.67; Table S2). HvBM1 and HvBM10 expression was not significantly affected by HvCO1 over-expression, or variation at Ppd-H1 and Vrn-H1.

Figure 7.

 Effect of variation at Ppd-H1 (light grey bars) and HvCO1 over-expression (dark grey bars) on expression levels of winter and spring Vrn-H1 alleles.
The F2 population HvCO1-ox × Igri (vrn-H2) derived by the cross between an introgression line in the background of the winter barley Igri (Ppd-H1, vrn-H1, vrn-H2) and two Ubi::HVCO1 lines (ppd-H1, Vrn-H1, vrn-H2) was grown for 21 days under short-day conditions (SD) and then moved to long days (LD). Bars represent Vrn-H1 expression normalised to HvActin of the average number of lines having the indicated allelic combination ± standard error. Letters indicate significant differences between Vrn-H1 expression of allelic classes separately for lines varying for HvCO1 and Ppd-H1.

Under SD, over-expression of HvCO1 had a significant effect on flowering time, and on expression of HvCO1 and HvBM1. In addition, variation at Vrn-H1 affected expression of Vrn-H1, but variation at Ppd-H1 did not have any effect on gene expression under SD (Table 1).

Discussion

HvCO1 over-expression causes early flowering in barley, but not in Arabidopsis

In A. thaliana, the CO protein comprises two B-Box-type zinc fingers proposed to be involved in protein–protein interactions (Khanna et al., 2009). However, the CO orthologs, Hd1 in rice and HvCO in barley, lack the highly conserved residues in the B-Box2 domain which are predicted to be required for B-box2 function (Griffiths et al., 2003). Consistent with the requirement of B-box2 in A. thaliana, over-expression of HvCO1 in the co-2 mutant did not induce FT expression and rescue the late flowering phenotype in this study (Figure S1). In contrast, Martin et al. (2004) showed that constitutive expression of LpCO, the CO ortholog in Lolium perenne, also characterized by a non-conserved B-Box2, could complement the Arabidopsis co-2 mutant. However, mutations in B-box2 of HvCO1 correspond to those found in late-flowering co mutant alleles in Arabidopsis (Robson et al., 2001) and our data suggest that two functional B-boxes are required for flower induction in Arabidopsis.

Onouchi et al. (2000) demonstrated that constitutive over-expression of CO from the CaMv35S promoter caused early flowering and almost complete insensitivity to day length in Arabidopsis. Similar to Arabidopsis, constitutive up-regulation of HvCO1 in barley induced early flowering under LD and SD in the present study. However, flowering was delayed by on average 43 days in the transgenic lines grown under SD as compared to LD. Constitutive expression of HvCO1 induced HvFT1, but the induction was significantly lower under SD as compared to LD and thus correlated with the differences in flowering between LD and SD (Figure 2c). These results suggest that photoperiod response factors in barley strongly control flowering time and HvFT1 expression downstream of HvCO1 mRNA.

Turner et al. (2005) suggested that Ppd-H1 activates HvFT1 expression by controlling the diurnal transcription of HvCO1 and HvCO2 and thus acts upstream of HvCO1. Over-expression of HvCO1 and allelic substitution of Ppd-H1 in a spring barley background both caused induction of HvFT1 and Vrn-H1 expression under LD (Figures 4c,d and 5c,d). However, variation at Ppd-H1 did not affect diurnal expression patterns of HvCO1 and HvCO2 mRNAs (Figure 5a,b). Up-regulation of HvFT1 by HvCO1 over-expression and the dominant Ppd-H1 allele demonstrated that both genes converge on HvFT1. These results suggested that Ppd-H1 may act as a photoperiod response factor downstream of HvCO1 transcription. The lower expression of HvCO1 in the introgression line (Ppd-H1) than in Scarlett (ppd-H1) may reflect the more advanced developmental stage of the introgression line, and may not be a direct effect of variation at Ppd-H1.

Genetic variation at Ppd-H1 acts downstream of HvCO1 expression on HvFT1

In the present study, analysis of the barley F2 population showed that variation at Ppd-H1 orthologous to PRR7 had a stronger effect on flowering time than over-expression of HvCO1. Variation at Ppd-H1 as identified by Turner et al. (2005) thus affected flowering time in barley independent of HvCO1 transcript levels and can explain the differences in flowering time between LD and SD observed in the transgenic lines (Figure 1). Differences in flowering time between LD and SD were also observed in the transgenic lines with the mutated ppd-H1 allele from Golden Promise (Figure 1). Consequently, the mutated ppd-H1 seems to be a hypomorphic allele with a reduced, but measurable, photoperiod response as compared with the wild-type dominant allele. The expression of HvFT1 with the highest correlation to flowering time was primarily controlled by Ppd-H1 (Table S2 and Table 1). Kikuchi et al. (2009) have already demonstrated that over-expression of HvFT1 in rice induced flowering and concluded that HvFT1 is the key gene responsible for flowering in the barley FT-like gene family (Faure et al., 2007). Variation at Ppd-H1 acted independently of HvCO1 mRNA on HvFT expression, but could be a factor modifying the HvCO1 protein, for example by stabilizing or activating the protein. Valverde et al. (2004) and Jang et al. (2008) have shown in Arabidopsis that post-translational regulation of the CO protein is crucial for its response, by stabilizing the protein at the end of a LD, and promoting its degradation throughout the night under LD and SD. In addition, Ishikawa et al. (2011) showed that in rice PhyB mediated light-induced suppression of Hd3a through post-translational modification of the Hd1 protein. On the other hand, Fujiwara et al. (2008) had already demonstrated for Arabidopsis that mutations in two other circadian clock genes LHY and CCA1 delayed flowering and suppressed FT expression independent of the CO protein, but through the stabilization of the MADS box transcription factor SVP. Likewise, Ppd-H1, orthologous to the Arabidopsis clock gene PRR7, may also have affected HvFT1 independently of HvCO1 through the SVP-like genes HvVRT2, HvBM1 or HvBM10. In the present study, barley lines carrying the transgene or the dominant Ppd-H1 allele showed a lower level of HvVRT2, HvBM1 and HvBM10 expression; however, there was no clear difference between the Ubi::HvCO1 and S42IL-107 lines, indicating that these genes were down-regulated during barley reproductive development rather than being directly affected by variation at Ppd-H1 (Figures 4 and 5).

HvCO1 up-regulates the spring allele of Vrn-H1

Flowering time and HvFT1 expression in the F2 population were also affected by the allelic status at Vrn-H1 even in the absence of Vrn-H2. Trevaskis et al. (2006) and Casao et al. (2011) have shown that in winter barley, Vrn-H1 only indirectly controls HvFT levels through the down-regulation of Vrn-H2, which is a repressor of HvFT1. However, in the present study, genetic variation at Vrn-H1 explained 17% of the variance in HvFT1 expression, where the winter allele at Vrn-H1 delayed time to flowering and reduced HvFT1 expression levels as compared to the spring allele (Table 1). Variation at Vrn-H1 thus controlled HvFT1 expression, suggesting that even in the absence of Vrn-H2, Vrn-H1 directly or indirectly controls HvFT1.

Interestingly, HvCO1 over-expression, but not variation at Ppd-H1, induced the expression of the spring Vrn-H1 allele under LD (Figure 7). These results indicate that up-regulation of the spring Vrn-H1 allele was directly induced by HvCO1 over-expression and was not an indirect effect of the more advanced development of F2-lines over-expressing HvCO1. Hemming et al. (2009) demonstrated that a region in the first intron of the Vrn-H1 gene was required to maintain low levels of Vrn-H1 expression prior to winter. Golden Promise carries a deletion of 5.2 kb in the first intron which is associated with increased Vrn-H1 expression independent of vernalisation (Hemming et al., 2009). Our results indicated that the photoperiod response pathway may be involved in the up-regulation of the spring allele of Vrn-H1, but fails to induce the winter allele prior to vernalisation. Kane et al. (2005, 2007) have suggested that VRT2 represses VRN1 and that down-regulation of VRT2 by cold allows expression of VRN1. In contrast, our results suggest that HvVRT2 acts downstream of Vrn-H1 expression. High correlations of HvVRT2 expression with time of stem elongation and HvFT1 expression indicated that HvVRT2 down-regulation occurred during development and was not a direct effect of HvCO1 over-expression. Trevaskis et al. (2007) have already shown that SVP-like genes including HvVRT2 regulate meristem identity, but are unlikely to mediate de-repression of Vrn-H1 during vernalisation. Interestingly, TaVRT2 maps to the short arms of group 7 chromosomes in a region associated with QTLs involved in heading date in wheat and barley (Kane et al., 2005; von Korff et al., 2006, 2010; Wang et al., 2009) and may thus play a role in the integration of photoperiod and vernalisation signals.

HvCO1 over-expression and variation at Ppd-H1 control late reproductive meristem development in barley

Three major phases in cereal meristem development can be distinguished: (i) the vegetative phase until formation of the double ridge in the apex, (ii) spikelet growth until formation of stamen initials and (iii) spike growth until cessation of peduncle elongation (Slafer, 2003; Nicholls and May, 1963). It has been shown that the timing and duration of the different developmental phases vary independently and are determined genetically in response to the environment (Gonzalez et al., 2005; Whitechurch et al., 2007). Analyses of development in wheat grown under artificially manipulated photoperiods have already shown that the stem elongation phase was the most sensitive to changes in photoperiod (Slafer et al., 2001). Accordingly, over-expression of HvCO1 and natural variation at Ppd-H1, both genes implicated in the photoperiod control of flowering in barley, primarily affected stem elongation and late reproductive meristem development in the present study (Figure 3). Hemming et al. (2008) also showed that natural variation at Ppd-H1 had little influence on the time to double ridge, but mainly accelerated development after the initiation of inflorescence in lines with active VRN1 alleles grown in LD. In the present study, differences in stem elongation were affected by variation at Ppd-H1 and Vrn-H1, and correlated with differences in the expression of HvFT1. Vrn1 has already been proposed to be implicated in stem elongation in wheat (Chen et al., 2009). Furthermore, studies in rice and tomato have also suggested that FT may have effects beyond floral induction on the development and architecture of the inflorescences (Lifschitz et al., 2006). In rice, high levels of Ghd7, encoding a CCT domain protein, and reduced expression of Hd3a correlated with late flowering, but also with differences in panicle size, shape and seed number (Xue et al., 2008). Similarly, heterozygosity of the loss-of-function allele of SINGLE FLOWER TRUSS (SFT), the tomato ortholog of FT, caused a shift in the balance of vegetative to reproductive growth and increased the total number of inflorescences and flowers per plant (Krieger et al., 2010). Although the genetic pathways underlying vegetative phase change are now well understood in model plants (Amasino, 2010), the source and identity of signals that initiate the transition between the juvenile to adult stages are relatively unknown (Chuck et al., 2007; Yang et al., 2011). Flowering in barley occurs several weeks after formation of the double-ridge SAM morphology which marks the transition to the reproductive meristem. These morphological changes may in fact indicate transition to an adult vegetative phase and reproductive competence which enables plant to respond to long photoperiods.

Conclusion

Functional analysis of HvCO1 in a spring and a winter barley background allowed further characterisation of the flowering time pathway in barley. Earlier work had already demonstrated that despite structural conservation of flowering time genes across species, functional modifications are common. For example, vernalisation response is mediated by FLC in Arabidopsis, VRN1 in winter cereals (Trevaskis et al., 2006) and BvFT1 in sugar beet (Pin et al., 2010). In Arabidopsis, CO functions as a central gene in the photoperiodic induction of flowering by linking the circadian clock mechanism to the genes for meristem identity. In barley over-expression of HvCO1 accelerated flowering; however, natural genetic variation in Ppd-H1 controlled flowering independent of HvCO1 mRNA. Ppd-H1 may thus bypass the established CO-FT interaction in Arabidopsis to induce flowering under long days. In addition, the morphology of flower induction differs between cereals and Arabidopsis which may have necessitated the functional modifications of flowering time orthologs, illustrated by the prominent role of FT in the late reproductive development of barley versus its function in the transition to the reproductive meristem in Arabidopsis.

Experimental procedures

Plant material and cultivation

Transgenic 35S::HvCO1 Arabidopsis lines.  The constans-2 (co-2) mutant line in the Ler background (Putterill et al., 1995) was transformed with cDNA of HvCO1 cloned by recombination using LR clonase II (Invitrogen, http://www.invitrogen.com/) into p35S::GATEWAY destination vectors under the control of the CaMV35S promoter (An et al., 2004). In addition, the co-2 mutant was transformed with 3× hemagglutinin (HA) tagged CO under the control of CaMV35S promoter in the pGreen binary vector (GC and W. Soppe, Max Planck Institute for Plant Breeding Research, Cologne, Germany, unpublished data). The binary vectors were transformed into Agrobacterium tumefaciens strain GV3101 (pMP90) or GV3101 (pMP90RK) and transformed into Arabidopsis Ler co-2 by the floral-dip method (Clough and Bent, 1998).

For flowering-time measurements in the co-2 mutant, Ler, 35S::HA-CO co-2 and 35S::HvCO1 co-2 Arabidopsis lines, plants were sown on soil and grown in controlled environment chambers under LD (16-h light/8-h dark) or SD (8-h light/16-h dark) at a constant temperature of 21°C, and numbers of leaves at bolting were counted. Ten-day-old seedlings were harvested at zeitgeber time, T12, for RNA extraction and expression analysis. The SD samples were harvested in the dark.

Transgenic Ubi::HvCO1 lines.  Golden Promise plants were transformed with an over-expression construct made by ligating cDNA clones of HvCO1 (AF490468) to the maize ubiquitin promoter (Christensen et al., 1992), and the resulting over-expression cassette was inserted into the pWBVEC8 binary vector (Wang et al., 1998). Barley plants were transformed using Agrobacterium transformation of excised embryos (Tingay et al., 1997; Matthews et al., 2001). An average of 30 independent transformants were produced, and T1, T2 and T3 plants were screened for segregation of the transgene using primers that amplify the hygromycin selectable marker gene and the HvCO1 cDNA sequence (Table S3). Ten independent transgenic T2-families designated Ubi::HvCO1 lines, the null segregant controls for the above transgenic lines and the corresponding wild type Golden Promise were sown in soil and grown under LD (16-h light/8-h dark) and SD (8-h light/16-h dark) in the greenhouse (temperature 20°C/16°C days/nights), and flowering time (measured in days to heading) was scored. Leaf material from three plants for each T2 family was harvested 14 days after sowing at the end of the day (T14 under LD and T6 under SD) for RNA extraction and expression analysis. Ubi::HvCO1 line 19 was additionally analysed for diurnal gene expression, and Ubi::HvCO1 lines 19 and 30 for meristem development (see below).

F2 population segregating for Ubi::HvCO1, Ppd-H1 and Vrn-H1

Two randomly selected Ubi::HvCO1 lines (lines number 19 and 30 in T3) were crossed with an introgression line derived from the winter barley Igri. Golden Promise carries the photoperiod-insensitive ppd-H1 allele and spring alleles at Vrn-H1 and vrn-H2. The introgression line is a BC4F2 selection from the DH population Igri × Triumph and carries the Igri alleles at all flowering time QTLs, the photoperiod-sensitive Ppd-H1 allele, the winter allele at vrn-H1, but the spring allele vrn-H2 from Triumph (Laurie et al., 1995). The derived F2 population was grown in soil in the greenhouse for 21 days under SD and then either transferred to LD or kept under SD (20°C/16°C days/nights). Plants were analysed for flowering time under LD (46 lines) and SD (42 lines) and leaf material was harvested to extract DNA for genotyping to test for the presence of the transgene and functional polymorphisms at Ppd-H1 (Turner et al., 2005) and Vrn-H1 (Cockram et al., 2007b; Table S3). In addition, leaf material was harvested from 25 F2 lines 7 days after exposure to LD (T14) and from 20 F2 lines under SD 14 days after separation of plants between LD and SD (T8) for RNA extraction and analysis of gene expression. F2 lines were selected to allow for a balanced allelic representation at HvCO1, Ppd-H1 and Vrn-H1.

Natural variation for Ppd-H1/ppd-H1

The German spring barley cultivar Scarlett and an introgression line S42IL-107 derived by crossing Scarlett with a wild barley accession ISR42-8 from Israel were used to test for the effects of natural variation at Ppd-H1 on flowering and expression of putative downstream genes (von Korff et al., 2004; Schmalenbach et al., 2011). The spring cultivar Scarlett carries the insensitive ppd-H1 allele and flowers late under long photoperiods, while the introgression line S42IL-107 carries the photoperiod-responsive Ppd-H1 allele introgressed from wild barley and flowers early under long photoperiods. Scarlett and S42IL-107 were analysed for diurnal gene expression, meristem development and flowering time (see below).

Plant development and expression analysis

Meristem dissection and diurnal expression analysis.  The two Ubi::HvCO1 lines 19 and 30 (selected for crossing to the mutant winter barley line), Scarlett and S42IL-107 were grown for 21 days under SD and then either transferred to LD or kept under SD. After transfer, three replicate primary shoots per genotype were dissected every 2–3 days under LD and every 4–5 days under SD, and the development of the meristem was scored until flowering according to the Waddington scale (Waddington et al., 1983). In addition, diurnal gene expression was tested in Ubi::HvCO1 line 19 and Golden Promise under LD and SD, and in Scarlett and S42IL-107 only under LD as the latter did not show any developmental differences under SD. At the transition to the reproductive meristem (Waddington scale 2), leaf samples were harvested every 2 h for a total of 24 h starting in the morning at lights on (T0) under LD and SD simultaneously. For each sample, two pooled plants, and two biological replicates were collected. Samples were immediately frozen in liquid nitrogen and stored at −80°C until RNA extraction.

The F2 lines under LD were scored for the formation of the first node on the main stem which marks the beginning of stem elongation (Chen et al., 2009).

RNA extraction, cDNA synthesis and real time quantitative RT-PCR

Total RNA was extracted from 100 mg of tissue using TRIzol® reagent (Invitrogen) following the manufacturer’s instructions, except for the use of RNaseH, followed by a DNase treatment (final volume 100 μl). First-strand cDNA synthesis was performed on 4 μl of total RNA using 100 U of SuperScriptTM II RT (Invitrogen) and 500 ng of poly-T primer and following the manufacturer’s recommendations (final volume 40 μl). All samples within an experiment were reverse transcribed at the same time and the resulting cDNA was diluted 1:4 in nuclease-free water and stored in aliquots at −20°C.

Real-time quantitative (q) RT-PCRs were performed on cDNA samples using gene-specific primers (Table S3). Amplifications were performed using 1 μl of cDNA, 1 U of GoTaq Flexi DNA polymerase (Promega, http://www.promega.com/), 0.2 mm deoxynucleotide triphosphate (dNTP), 2.5 mm MgCl2, 0.2 μm of each primer and 1 μl of EvaGreen (Biotium, http://www.biotium.com/). Reactions were performed on a LightCycler480 (Roche, http://www.roche.com/) with the following amplification conditions: 95°C for 5 min, 40 cycles of 95°C (10 sec), 60°C (10 sec) and 72°C (10 sec). Appropriate non-template controls were included in each 384-well PCR reaction, and dissociation analysis was performed at the end of each run to confirm the specificity of the reaction. Starting amounts for each data point were calculated based on the titration curve for each target gene and the reference (HvActin) gene using the LightCycler 480 Software (Roche; version 1.5).

Statistical analysis

Significant differences in flowering time and gene expression between Ubi::HvCO1 lines and wild type grown under LD and SD were calculated using a mixed model analysis of variance (anova). The response (Y) was modeled according to the following three-factorial layout (SAS version 9.1; SAS Institute 2009):

image

where μ is the general mean, Pk is the fixed effect of the kth treatment (LD or SD), Gi is the fixed effect of the i-th genotype (Ubi::HvCO1 or WT), Tj is the random effect of the j-th transformant within the i-th genotype. The remaining terms describe the first order interaction effect. Eijkm is the error of Yijkm.

Significant effects of the transgene and variation at the candidate genes Ppd-H1 and Vrn-H1 on flowering time and gene expression in the F2 population under LD and SD were calculated using a fixed model anova and a three-factorial layout:

image

where μ is the general mean and Eijkm is a modified residual error term. Ci, Pj and Vk are the fixed effects for the genotype levels for HvCO1 (Ubi::HvCO1 or wild type), Ppd-H1 (Ppd-H1 or ppd-H1) and Vrn-H1 (Vrn-H1, heterozygote, vrn-H1), respectively. The remaining terms describe first-order interaction effects. Three-way interactions were subsumed in the residual error, as for some allele combination just one line replicate (m = 1) was tested.

Acknowledgements

We thank Professor Dr K. Pillen, University of Halle-Wittenberg for genetic material of Scarlett and the introgression lines S42IL-107. We thank Dr Wim Soppe for seeds of the 35S::3 × HA CO co-2 Arabidopsis line. We are grateful for excellent technical assistance by K. Luxa and E. Luley. This work was supported by the Max Planck Society and by a grant from the Deutsche Forschungsgemeinschaft (DFG) SPP 1530/1 to MvK.

Ancillary