Genetic and Molecular Characterization of the VRN2 loci in Tetraploid Wheat

Winter wheat varieties require long exposures to low temperatures to flower, a process called vernalization. The VRN2 locus includes two completely linked zinc finger-CCT domain genes ( ZCCT1 and ZCCT2 ) that act as flowering repressors down-regulated during vernalization. Deletions or mutations in these two genes result in the elimination of the vernalization requirement in diploid wheat. However, natural allelic variation in these genes has not been described so far in polyploid wheat. A tetraploid wheat population segregating for both VRN-A2 and VRN-B2 loci facilitated the characterization of different alleles. Comparisons between functional and non-functional alleles revealed that both ZCCT1 and ZCCT2 genes are able to confer vernalization requirement and that different ZCCT genes are functional in different genomes. ZCCT1 and ZCCT2 proteins from non-functional vrn2 alleles have mutations at arginine amino acids at positions 16, 35 or 39 of the CCT domain. These positions are conserved between CCT and HAP2 domains supporting a model in which the action of CCT domains is mediated by their interactions with HAP2/HAP3/HAP5 complexes. This study also revealed natural variation in gene copy number, including a duplication of the functional ZCCT-B2 gene and deletions or duplications of the complete VRN-B2 locus. Allelic variation at the VRN-B2 locus was associated with a partially dominant effect, which suggests that variation in the number of functional ZCCT genes can be used to expand allelic diversity for heading time in polyploid wheat and, hopefully, improve its adaptation to different environments. CONSTANS and the CCAAT box binding complex share a functionally important domain and interact to regulate flowering of Arabidopsis. The Plant


INTRODUCTION
Wheat is one of the major crop species and occupies a wide range of environments from 65° N to 45° S (Lantican et al., 2005). This wide adaptability is favored by diverse growth habits, which include winter and spring forms. Winter wheats are sown in autumn and require long exposures to cold temperatures (vernalization) to accelerate flowering. The vernalization requirement prevents flower development during winter, protecting sensitive floral organs from freezing temperatures. Spring wheats are planted in the spring or in the fall in regions with mild winters, and do not have a vernalization requirement.
The three major genes responsible for natural variation in vernalization requirement in wheat (and also in barley) are VRN1, VRN2 and VRN3. VRN1 is a homologue of the Arabidopsis meristem identity gene AP1, which determines the transition between the production of leaves and flowers at the shoot apical meristem (Danyluk et al., 2003;Trevaskis et al., 2003;Yan et al., 2003). Mutagenized plants of diploid wheat Triticum monococcum L. (2n=14, A m genome similar to A genome of polyploid wheat) with complete deletions of the VRN1 gene fail to flower (Shitsukawa et al., 2007), indicating that VRN1 is essential for the initiation of the reproductive phase in this species. Several natural mutations have been identified in regulatory regions of the VRN1 promoter or first intron, which are associated with the elimination or reduction of the vernalization requirement and consequently with spring growth habit (Yan et al., 2003;Yan et al., 2004a;Fu et al., 2005;vonZitzewitz et al., 2005).
VRN3 is a homologue of Arabidopsis photoperiod gene FT, and in both species this gene up-regulates VRN1 transcription under long days (Yan et al., 2006;Hemming et al., 2008) through interactions with its promoter (Wigge et al., 2005;Li et al. 2008). Before vernalization, VRN3 is down-regulated by VRN2 preventing winter wheats to flower during the fall. Vernalization results in the induction of VRN1 and the down-regulation of VRN2 (Loukoianov et al., 2005;Trevaskis et al., 2006), thereby releasing VRN3 to further induce VRN1 and initiate the reproductive phase during the long days of spring (reviewed by Trevaskis et al., 2007 andDistelfeld et al., 2009). The focus of this paper is the natural variation in VRN2.
In diploid wheat, the VRN2 locus includes two tandemly duplicated genes designated as ZCCT1 and ZCCT2 (Yan et al., 2004b). These genes code for proteins that are 76% identical, each including a putative zinc finger and a CCT domain (for CONSTANS (CO), CONSTANS-LIKE (CO-like), and TIMING OF CAB EXPRESSION1 (TOC1)).
The 43-amino acid CCT domain is present in proteins involved in photoperiod, light signaling, and circadian rhythms and is well conserved among different plant species (Griffiths et al., 2003;Yan et al., 2004b). Mutations within the CCT domain are known to alter the function of proteins CO, TOC1, VRN2, and PPD-H1 (Wenkel et al., 2006). It has been shown recently that the CCT domain has similarities to a region of the yeast HEME ACTIVATOR PROTEIN2 (HAP2), a subunit of the HAP2/HAP3/HAP5 complex that binds to CCAAT boxes in the promoters of many eukaryotic genes and regulates their expression (Wenkel et al., 2006).
In wheat and barley accessions with winter growth habit, ZCCT transcripts show a progressive decrease during vernalization (under long days), which is not observed in control plants kept at non-vernalizing temperatures (Yan et al., 2004b, Trevaskis et al., 2006. Wheat and barley ZCCT genes are also down-regulated by short days (Dubcovsky et al., 2006;Trevaskis et al., 2006;2007). GHD7 (=OsI), the closest homologous gene in rice (Xue et al., 2008) is also a long day repressor of flowering down-regulated by short days.
All diploid wheat and barley accessions with winter growth habit studied so far have at least one functional ZCCT gene, whereas those with spring growth habit associated with recessive vrn2 alleles have deletions encompassing all ZCCT genes or carry mutations in conserved amino acids of the CCT domains (Yan et al., 2004b;Karsai et al., 2005;Cockram et al., 2007;Szücs et al., 2007). The presence of a single functional Vrn2 allele in heterozygous plants is sufficient to confer some vernalization requirement, so only the homozygous recessive vrn2 allele results in spring growth habit (Takahashi and Yasuda,8 Additional ZCCT genes were sequenced from winter accession of diploid Triticum and Aegilops species with genomes similar to the A, B and D genomes of hexaploid wheat. These species included T. urartu Tumanian ex Gandilyan, the donor of the A genome (Dvorak et al., 1988); Ae. speltoides Tausch (S genome), the closest extant diploid species to the B genome of tetraploid and hexaploid wheat (Dvorak and Zhang, 1990); and Ae. tauschii, the donor of the D genome of hexaploid wheat (Kihara, 1944).
The sequences for ZCCT-D1 (FJ173818) and ZCCT-D2 (FJ173822) were obtained from Ae. tauschii BAC clones 2H24, 14E16, and 78I09 (Akhunov et al., 2005), whereas the ZCCT genes from the other diploid species were obtained directly by PCR from genomic DNA. Aegilops tauschii ZCCT-D1 and -D2 sequences, together with T. urartu sequences for ZCCT-A1 (FJ173816) and ZCCT-A2 (FJ173820), and Ae. speltoides sequences for ZCCT-S1 (FJ173817) and ZCCT-S2 (FJ173821) were deposited in GenBank. The phylogenetic analysis grouped the predicted ZCCT protein sequences into two distinct clades ( Fig. 1), one including the ZCCT1 proteins from all species and the other one the ZCCT2 proteins from all species. This suggests that the duplication that originated these two genes preceded the divergence of the diploid Triticum and Aegilops species.

RFLP germplasm screen
The hybridization of Southern blots including DraI digested DNAs from wild and cultivated tetraploid and hexaploid Triticum accessions (see Material and Methods) revealed contrasting patterns of allelic diversity for different ZCCT genes. The shortest overlapping DraI restriction fragments (767-bp from ZCCT-A1 and 771-bp from ZCCT-B1) showed limited variation among accessions. Limited variation was also found for the 1,420-bp fragment corresponding to the ZCCT-A2 gene (Fig. 2). On the contrary, the restriction fragments within the region corresponding to the ZCCT-B2a and ZCCT-B2b genes were variable in size (≈3-5 kb) generating multiple haplotypes (Fig. 2, lanes 1-5).
Some cultivated durum lines showed two fragments in this region and others only one (Fig. 2).
The largest RFLP fragment corresponds to a third and more divergent ZCCT copy (ZCCT-A3). The ZCCT-A3 putative coding region is only 81-82% identical to the other two ZCCT genes and has a shorter first exon that does not include the predicted zinc-finger characteristic of other ZCCT proteins. So far, ZCCT-A3 has been found only in the A genome, 16.2 kb upstream of ZCCT-A2 (Dubcovsky and Dvorak, 2007). It is not yet known if ZCCT-A3 is translated into a functional protein and therefore, it was not included in the allelic diversity study.
In addition to the variation in restriction fragment size, the ZCCT-B1 and ZCCT-B2 genes showed polymorphisms in copy number. Six T. turgidum ssp. dicoccoides accessions from Rosh Pinna, Israel showed unusually strong hybridization signals at the restriction fragments corresponding to ZCCT-B1 and ZCCT-B2 (one accession is shown in Fig. 2, lane 8). The VRN-A2 fragments from the same accessions showed no increase in hybridization intensity, confirming equal loading of DNA in the Southern blots. Based on this result we concluded that the copy number of the ZCCT-B1 and ZCCT-B2 genes was amplified in the Rosh Pinna accessions.

Molecular characterization of the 5A-5A m translocation in BC3F2-521
The effect of the VRN2 loci on flowering time in tetraploid wheat was studied using a plant segregating simultaneously for the vrn-B2 deletion from PI470739 and the nonfunctional vrn-A m 2 from T. monococcum accession DV92. The development of this plant is described in Materials and Methods and in Fig. 3.
A total of 42 plants with winter growth habit were selected from the progeny of a BC 3 F 1 line heterozygous for different vernalization genes (Fig. 3G). The winter growth habit indicates that these plants are homozygous for the recessive vrn-A1 and vrn-B1 alleles.
Using molecular markers for VRN-A2 (Fig. 3C) and VRN-B2 (Fig. 3H) we selected plant 521 (BC 3 F 2 -521 hereafter), which was heterozygous for both VRN-A2 and VRN-B2 and homozygous for the recessive vrn-A1 and vrn-B1 alleles (Fig. 3). The progeny of this plant were used for the effect of the different VRN2 alleles on flowering time.
Several plants from the progeny of BC 3 F 2 -521 were analyzed for chromosome number and all showed 28 chromosomes indicating that the vrn-A m 2 gene was incorporated either as a complete chromosome substitution line or as a translocation line. To differentiate between these two possibilities, this line was analyzed with two molecular markers for the short and long arm of homoeologous group 5. The marker for the PINA gene (Bonafede et al., 2007), located in the distal region of the short arm of chromosome 5A m and deleted in tetraploid wheat, was not present in BC 3 F 2 -521 confirming that vrn-A m 2 was transferred to a translocated chromosome. To determine the location of the 5A -5A m translocation we developed a marker for the VRN-A1 gene, which is located on the middle of the long arm. The absence of the T. monococcum vrn-A m 1 allele indicated that the recombination event occurred between the VRN-A1 (5AL) and VRN-A m 2 (5A m L) loci.
A marker for the deletion in the VRN-A1 intron (Fu et al., 2005) was used to confirm that the allele present in BC 3 F 2 -521 was the recessive vrn-A1 allele from Durelle, as expected from its winter growth habit. A representation of the recombined chromosome is shown in Fig. 3J.

Effect of the allelic differences in VRN-A2 and VRN-B2 on flowering time
The non-vernalized progeny of line BC 3 F 2 -521 segregated into two non-overlapping groups for flowering time. The first group included 13 early-flowering plants that headed in less than 60 days (average 53.0 ± 0.4 days) and were classified as spring, whereas the second group included 41 late-flowering plants that took more than 90 days for heading (average 139.1 ± 4.6 days) and were classified as winter (Fig. 4A). The group with spring growth habit was less variable than the group with winter growth habit, which showed two peaks, likely associated to the presence of homozygous and heterozygous lines (Fig.   4A). The observed ratio between spring and winter plants differed significantly from a 1:15 ratio segregation (two dominant genes, indicating that both T. turgidum (PI470739) and T. monococcum DV92 have recessive vrn-A2 alleles. This indicates that none of the ZCCT genes present at the VRN-A2 locus is able to confer a vernalization requirement.
Genotyping with the codominant SNF-B2 marker tightly linked to the VRN-B2 locus (supporting online materials) showed that all 13 spring plants were homozygous for the recessive vrn-B2 allele from PI470739 (Fig. 5B), a result that was confirmed using PCR primers specific for the ZCCT-B1 and ZCCT-B2 genes ( The ZCCT1 and ZCCT2 genes present in the functional VRN-B2 and non-functional VRN-A2 loci from BC 3 F 2 -521 were sequenced to determine if the differences in functionality were associated with specific mutations in the conserved CCT domain. For comparison, a consensus CCT sequence was generated from different classes of CO-like proteins found in plants (Griffiths et al., 2003;Yan et al., 2004b) and was represented using a WebLogo (Crooks et al., 2004). The consensus sequence was aligned with the CCT domain sequences from winter accessions of wild diploid progenitors of cultivated wheat (T. urartu, Ae. speltoides and Ae. tauschii) (Fig. 6). The main differences among species are summarized in Table I.

ZCCT-A1:
The predicted ZCCT-A1 protein corresponding to the non-functional VRN-A2 locus from BC 3 F 2 -521 has a mutation from R to C at position 39 of the CCT domain (designated R39C hereafter, Fig. 6). This position of the CCT domain is well conserved among CCT domains from other CO-like proteins and HAP2 domains (Fig. 6).
The R39C mutation was detected in all 37 cultivated T. turgidum ssp. durum accessions analyzed in this study, but was polymorphic in cultivated T. turgidum ssp. dicoccon (present in 11 out of 22 accessions) and wild T. turgidum ssp. dicoccoides (present in 10 out of 19 accessions, Table II). One accession from Asia Minor (PI355454) showed an additional R35Q mutation. The eleven accessions of T. turgidum ssp. dicoccon that lack the R39C mutation all have a R35W mutation. The R35W mutation was also found in 8 out of the 9 accessions of T. turgidum ssp. dicoccoides that do not have the R39C mutation. T. turgidum ssp. dicoccoides accession 10-85 collected at Ammiad in Israel was the only one with no mutations in the CCT domain from ZCCT-A1 (Table II).
In addition to the R39C mutation, the ZCCT-A1 protein has a deletion of seven amino acids relative to the ZCCT-A m 1 protein from T. monococcum. These 7 amino acids are located immediately downstream of the putative zinc finger domain from amino acids 49 to 55 (numbers are relative to the initial methionine in ZCCT-A m 1). A screening using primers VRN2/22F + R (Table SI, supporting online materials) showed that the same deletion was present in all 78 tetraploid wheats tested in this study (Table II)

ZCCT-A2:
The predicted ZCCT-A2 protein corresponding to the non-functional VRN-A2 locus from BC 3 F 2 -521 has a mutation from R to C at position 16 of the CCT domain (R16C). This position is well conserved (R or K) among CCT domains from other COlike genes (except for CO-like Group II) and HAP2 domains (Fig. 6).
All 48 accessions of cultivated tetraploid wheat sequenced for this gene have the R16C mutation in the CCT domain. This mutation was also found in the 15 accessions of T.
urartu and four accessions of T. monococcum (ZCCT-A m 2) but was not detected in the predicted ZCCT2 proteins from Ae. tauschii or Ae. speltoides (Table I, Fig. 6). Three out of the four T. monococcum accessions (including DV92) have an R39C mutation in addition to the R16C mutation in ZCCT-A m 2.

ZCCT-B1:
The predicted ZCCT-B1 protein corresponding to the functional VRN-B2 locus has the same R39C mutation as the ZCCT-A1 protein coded by the non-functional VRN-A2 locus (Table I, Fig. 6). The R39C mutation was conserved in the predicted ZCCT-B1 proteins from the 37 accessions of cultivated durum wheat (Table II) (Table II).

ZCCT-B2:
The predicted ZCCT-B2a and ZCCT-B2b proteins corresponding to the functional VRN-B2 locus from tetraploid variety Langdon have no mutations in any of the conserved amino acids of the CCT domain (Table I, Fig. 6). This was also the case for the other 37 cultivated durum accessions (Fig. 6).
A screening for the 1-bp indel characteristic of the ZCCT-B2 gene duplication using PCR primers VRN2/B2/F2+R5 (Table SI,  provides additional indirect evidence for the absence of the duplication in most wild accessions. In T. turgidum ssp. dicoccon, the presence of the ZCCT-B2 duplication was confirmed in the four accessions that carry the R39C mutation at the ZCCT-B1 gene (PI319868, PI319869, PI355454, and PI352347, Table II). The presence of the ZCCT-B2 duplication was also confirmed among most of the modern T. turgidum ssp. durum varieties (36 out of 37), with 'Messapia' as the only exception (Table II). Many of the cultivated durum varieties showed two fragments in the RFLP screening ( Fig. 2)

Expression of ZCCT1 and ZCCT2 genes in tetraploid wheat
Quantification of transcript levels of ZCCT1 and ZCCT2 in tetraploid wheat leaves collected from three, four and five-week old plants showed that the average transcript levels of ZCCT2 were significantly higher than those of ZCCT1 for all three time points ( Fig. 7). Since quantitative RTPCR primers (supporting online materials) were designed to differentiate ZCCT1 from ZCCT2 but not A from B genome copies of the same gene, the transcript levels presented in Fig. 7 include both A and B homoeologues for each gene.

DISCUSSION
The results presented here indicate that the differences among ZCCT proteins coded by genes corresponding to functional and non-functional VRN2 alleles are concentrated in the CCT domain. This 43-amino acid domain is well conserved in CO and CO-like proteins (defined as being more similar to CO than to other Arabidopsis proteins like TOC1) from mosses, gymnosperms and angiosperms indicating an ancient origin (Griffiths et al., 2003). The CCT domains are involved in the nuclear localization of CO and CO-like proteins but also have additional roles. In Arabidopsis, the co-7 CCT mutation does not alter the nuclear localization of the CO protein but delays flowering significantly (Robson et al., 2001). It is possible that some mutations in the CCT domain may limit its ability to interact with other proteins (Kurup et al., 2000). It has been recently shown that the CCT domains from Arabidopsis CO and COL15 can interact with several AtHAP3 and AtHAP5 proteins in yeast, and this interaction was confirmed in plant cells and in vitro (Ben-Naim et al., 2006;Wenkel et al., 2006). Wenkel et al. (2006) proposed that CCT proteins act by replacing the HAP2 subunit of the HAP2/HAP3/HAP5 complex, altering the ability of this complex to bind to the CCAAT boxes in the promoters of target genes. Overexpression of AtHAP3b was shown to promote early flowering probably through an interaction with CO or COL proteins while hap3b, a null mutant of HAP3b, delayed flowering under long-days but not under short days (Cai et al., 2007).
CCT domains and HAP2 proteins have similar amino acids at 18 positions, which are also well conserved within each group of proteins from mosses to vascular plants (Wenkel et al., 2006). Conservation of these amino acids for more than 400 million years suggests that they play critical roles in the proper function of these proteins. Six subdomain of the HAP2 protein (Fig. 6). This HAP2 subdomain has been modeled and predicted to interact with the DNA of the CCAAT box (Romier et al., 2003). In addition to the NF-YA2 subdomain, the HAP2 protein has another NF-YA1 subdomain proposed to interact with the HAP3/HAP5 dimer, and a linker region between these two subdomains (Romier et al., 2003). It is tempting to speculate that the ZCCT proteins may also regulate flowering time through interactions with HAP proteins, and that mutations in the CCT domain may affect the ability of the ZCCT proteins to interact with DNA or with HAP3/HAP5 dimers. We have recently confirmed that ZCCT proteins can interact with several wheat HAP3 and HAP5 proteins in yeast-two-hybrid systems (C. Li, A. Distelfeld, and J. Dubcovsky unpublished) providing additional support to this hypothesis.
Interestingly, the three CCT mutations identified here in ZCCT proteins coded by genes located in non-functional VRN2 loci are located at positions 16, 35 and 39, which are conserved both between and within the CCT and HAP2 domains (Fig. 6)

The "two-ZCCT" hypothesis
Assuming that the mutations at CCT positions 16, 35 and 39 can disrupt the function of the ZCCT proteins, the following model can explain the complex results presented here.
We propose that both ZCCT1 and ZCCT2 have the ability to delay flowering and confer a vernalization requirement. We will refer hereafter to this model as the "two-ZCCT" hypothesis to facilitate the discussion. The first corollary of this hypothesis is that the presence of a functional copy of at least one of these two genes would be sufficient to confer a vernalization requirement. The second corollary of this hypothesis is that mutations in both genes are required to completely disrupt the function of a particular VRN2 locus. The following arguments are presented to support this hypothesis.

1.-Similarity of ZCCT1 and ZCCT2 CCT domains:
The CCT domains from ZCCT1 and ZCCT2 are almost identical among functional alleles from different species (Fig. 6).
The only difference between them is found at the second amino acid, which is fixed for A in the ZCCT1 proteins and varies between E, H, and Q in the ZCCT2 wheat proteins (and barley ZCCT proteins). The second amino acid of the CCT domain is also variable among CO-like proteins and is not conserved with the HAP2 protein (Fig. 6), suggesting that it may not be a critical position for the function of the CCT domain. Therefore, it is reasonable to assume that ZCCT1 and ZCCT2 may have the ability to perform similar functions.

2.-Non-functional vrn2 alleles:
The two-ZCCT hypothesis predicts that all recessive vrn2 alleles would have non-functional mutations at both ZCCT1 and ZCCT2 proteins. In agreement with this prediction the recessive vrn-A2 allele from BC 3 F 2 -521 has the R39C mutation in ZCCT-A1 and the R16C mutation in ZCCT-A2 (Table I). The deletion of 7 amino acids found downstream of the putative zinc finger in the ZCCT-A1 protein in tetraploid variety Langdon (Dubcovsky and Dvorak, 2007) and BC 3 F 2 -521 (from PI470739) does not seem to be critical for the function of the ZCCT-A1 protein, since a similar deletion was observed in T. urartu accession PI428180, which has a winter growth habit (functional Vrn-A2 allele) and a likely non-functional ZCCT-A2 protein ( Table I).
The available information from T. monococcum also supports the two-ZCCT hypothesis. Although the published conclusion is valid for T. monococcum, the current results indicate that it cannot be generalized to all Triticeae species.  (Table I). This suggests that the winter growth habit in T. urartu is also conferred by ZCCT-A1.

3.-Functional
The molecular characterization of the functional Vrn-B2 allele provided the strongest support to the two-ZCCT hypothesis. The ZCCT-B1 protein found in the parental lines of BC 3 F 2 -521 (Langdon / Durelle) has an R39C mutation identical to the one found in the ZCCT-A1 protein from the non-functional VRN-A2 allele ( Table I). The low BLOSUM62 score (-3) and the fact that this mutation alters a conserved position across HAP2 and CCT domains (Fig. 6) suggest that this ZCCT-B1 protein is non-functional. In contrast, the ZCCT-B2 protein has no mutations in the conserved amino acids of the CCT domain.
The Q mutation found in ZCCT-B2 is associated with a positive BLOSUM62 score (+2) indicative of similar biochemical properties. In addition, CCT position 2 is variable among the CCT domains of ZCCT2 and CO-like proteins, and is not conserved with the HAP2 proteins (Fig. 6). These observations suggest that this mutation may not have a negative impact on the structure or function of ZCCT-B2, and that this protein rather than ZCCT-B1 is the one conferring the strong vernalization requirement observed in the late flowering lines from the BC 3 F 2 -521 progeny. domain. However, since this position is not conserved this mutation has a small probability of disrupting the function of the Ae. speltoides ZCCT2 protein.
The lack of mutations in ZCCT1 and ZCCT2 in the functional VRN2 alleles from these two diploid species is consistent with the two-ZCCT hypothesis, but does not provide new information about the relative importance of these genes for the establishment of the vernalization requirement. The absence of mutations in the CCT domain of the ZCCT-D1 and ZCCT-D2 genes in diploid Ae. tauschii (Table I) suggests that the D genome has the potential to contribute two functional ZCCT copies to common wheat.
In summary, the hypothesis that both ZCCT1 and ZCCT2 genes can confer vernalization requirement explains well the different results on VRN2 allelic variation described in this and previous studies.

Allelic diversity in VRN2 alleles in tetraploid wheat
The R16C mutation in the ZCCT-A2 protein seems to be fixed in the A genome of tetraploid wheat, since it is present in all the A and A m diploid species sequences so far.
However, the R39C mutation in the ZCCT-A1 protein is still polymorphic among the wild and cultivated T. turgidum ssp. dicoccoides. Approximately half of the accessions of these two subspecies have the R39C mutation whereas the others do not. The R39C mutation was present in all 37 T. turgidum ssp. durum varieties analyzed in this study (Table II), suggesting that this mutation was fixed during the domestication of the modern free threshing tetraploid wheats.
Eight of the nine T. turgidum ssp. dicoccoides accessions and all the T. turgidum ssp.
dicoccon that lack the R39C mutation in ZCCT-A1 carry a R35W mutation identical to the one detected in T. monococcum accession DV92 (Table II) (Table II). We plan to cross T. turgidum ssp. dicoccoides accession 10-85 with a line homozygous for recessive vrn-A2 and vrn-B2 alleles to test the effect of the 10-85 VRN-A2 allele on flowering time.
The R39C mutation in the ZCCT-B1 gene was also polymorphic among the T. turgidum ssp. dicoccon and T. turgidum ssp. dicoccoides accessions but was fixed in all the T. turgidum ssp. durum varieties analyzed here (Table II). On the contrary, none of the ZCCT-B2 proteins from these 37 accessions of cultivated durum has mutations in the CCT domain. This result suggests that winter growth habit in cultivated tetraploid wheat is conferred mainly by the ZCCT-B2 gene(s) and that in some T. turgidum ssp. dicoccon and T. turgidum ssp. dicoccoides accessions both the ZCCT-B1 and ZCCT-B2 genes can delay flowering under long days.

VRN-B2 dosage effect
The analysis of the progeny of BC 3 F 2 -521 showed that the effect of the functional VRN-B2 locus on heading time was partially dominant (degree of dominance= 0.26), which agrees with previous results reported in barley (Szücs et al., 2007). This partial dominant effect indicates that allelic variation in the number of functional copies of ZCCT1 and ZCCT2 can affect heading time in tetraploid wheat. Therefore, the duplication of the functional ZCCT-B2 gene found in most cultivated durum and in some T. turgidum ssp.
dicoccon accessions may have contributed to the variation in heading time in tetraploid wheat. The high sequence identity between the two copies (99.7 % identity) suggests that this duplication was originated recently.
The duplication of the functional ZCCT-B2 locus provides a simple explanation for the higher transcript levels of ZCCT2 relative to ZCCT1 in tetraploid wheat (Fig. 7, A  dicoccoides from Rosh Pinna). The copy number of ZCCT-B1 and ZCCT-B2 in these accessions is currently unknown but the intensity of the hybridization signal suggests the presence of several copies (Fig. 2). We have initiated the crosses required to study the effect of this duplication on flowering time.

CONCLUSIONS
Accessions with a spring growth habit determined only by deletions or mutations in the VRN2 locus are frequent in cultivated barley (Dubcovsky et al., 2005;Szücs et al., 2007) and diploid wheat (Yan et al., 2004b). These VRN2 mutations are also found in combination with dominant Vrn1 alleles. These results suggest that vrn2 mutations alone or in combination with some dominant Vrn1 alleles might confer different responses to environmental cues from those conferred by those VRN1 mutations alone.
The discovery that durum wheat varieties have non-functional vrn-A2 alleles and the development of a codominant marker tightly linked to the vrn-B2 deletion (PI470739) will facilitate the development of spring durum wheat varieties with no functional VRN2 loci. These non-functional VRN2 alleles can then be used alone or in combination with different dominant Vrn1 alleles to develop spring durum wheat varieties with new allelic diversity in heading time.
Allelic variation for VRN2 can be widened also in the opposite direction by adding more copies of functional ZCCT genes to cultivated durum wheat. This is expected to increase vernalization requirement and/or delay flowering, although its final effect will depend on  Seeds were imbibed for 24h at 4°C to promote synchronized germination. Seedlings were transferred to pots and watered with nutrition solution. Unvernalized plants were grown in a greenhouse at room temperature (20-25°C) and long-day photoperiod (8 h dark / 16 h light). For the vernalization experiments, plant were first grown for three weeks at the same conditions described above, transferred to a cold room at 4°C and long-day photoperiod for 4 weeks, and then transferred back to the greenhouse to score heading date. Heading date was recorded at complete spike emergence. 1 The author(s) responsible for distribution of materials integral to the findings presented in this article in accordance with the Journal policy described in the Instructions for Authors ( http://www.plantphysiol.org) are: J. Dubcovsky (jdubcovsky@ucdavis.edu) and A. Distelfeld (adistel@ucdavis.edu).
Methods used for sequencing BAC clone 738D05 (VRN-B2 locus), hybridization, polymerase chain reaction (PCR) and quantitative reverse transcription PCR (Q-RTPCR); together with the markers for VRN-A m 2, VRN-B2, VRN-A1 and PINA loci are described in the supporting online materials. Primers for all the experiments are described in Table SI, also available on the supporting online materials.

Development of a tetraploid wheat line segregating for VRN-A2 and VRN-B2
The following crosses and selections were performed to introduce the non-functional vrn- Plant BC 1 #2 was crossed with the tetraploid winter wheat Durelle to incorporate the recessive vrn-A m 2 allele into a winter background (Fig. 3D). The BC 2 F 1 plant from this cross was self-pollinated and a population of 80 BC 2 F 2 plants was generated and grown in a greenhouse without vernalization (Fig. 3E). This population showed a 3:1 (62:18) segregation between winter and spring growth habit, as expected for a population segregating only for VRN-A1. Winter BC 2 F 2 lines (homozygous for recessive vrn-A1 and vrn-B1 alleles) were screened with the VRN-A m 2 CAPS marker and three lines homozygous for the recessive vrn-A m 2 allele were selected (Fig. 3F).
The selected BC 2 F 2 lines were crossed with T. turgidum ssp. dicoccon accession PI470739 (Fig. 3F), which is homozygous for a deletion encompassing both ZCCT-B1 and ZCCT-B2 genes (recessive vrn-B2 allele). Three BC 3 F 1 were self-pollinated and the resulting BC 3 F 2 seeds were grown in a greenhouse without vernalization to select winter BC 3 F 2 plants (Fig. 3G). The winter lines (homozygous vrn-A1 vrn-B1) were then screened with the VRN-A m 2 CAPS marker and with a codominant marker for SNF-B2 Since the R16C mutation is fixed in ZCCT-A2 and ZCCT-A m 2 we checked only one accession from each group for this mutation. 2 The "1" indicates that only one of the two polymorphic sites was detected but does not completely rule out the existence of the duplication.