Copy Number Variation Affecting the Photoperiod-B1 and Vernalization-A1 Genes Is Associated with Altered Flowering Time in Wheat (Triticum aestivum)

The timing of flowering during the year is an important adaptive character affecting reproductive success in plants and is critical to crop yield. Flowering time has been extensively manipulated in crops such as wheat (Triticum aestivum L.) during domestication, and this enables them to grow productively in a wide range of environments. Several major genes controlling flowering time have been identified in wheat with mutant alleles having sequence changes such as insertions, deletions or point mutations. We investigated genetic variants in commercial varieties of wheat that regulate flowering by altering photoperiod response (Ppd-B1 alleles) or vernalization requirement (Vrn-A1 alleles) and for which no candidate mutation was found within the gene sequence. Genetic and genomic approaches showed that in both cases alleles conferring altered flowering time had an increased copy number of the gene and altered gene expression. Alleles with an increased copy number of Ppd-B1 confer an early flowering day neutral phenotype and have arisen independently at least twice. Plants with an increased copy number of Vrn-A1 have an increased requirement for vernalization so that longer periods of cold are required to potentiate flowering. The results suggest that copy number variation (CNV) plays a significant role in wheat adaptation.


Introduction
Cultivated bread wheat (Triticum aestivum) is an allohexaploid combining A, B and D genomes from three diploid ancestors [1]. The wild diploids and wheat in its original domestic form are winter annual long day plants [2], germinating in autumn and requiring an extended period of cold (vernalization) before they can respond to the flowering stimulus of increasing day length (photoperiod) in the following spring. These requirements have been altered during domestication to adjust the life cycle to local conditions, enabling wheat to be highly productive in a wide range of environments.
Wheat's ancestors and many modern varieties are described as photoperiod sensitive. That is, they flower rapidly in long days but are late flowering in short days. Day neutral (photoperiod insensitive) varieties flower rapidly in short or long days. This allows production in environments where appropriate temperature and rainfall coincide with short day conditions or where early flowering avoids high summer temperatures, as in Southern Europe [3]. The day neutral phenotype was an important component of the ''Green Revolution'' and continues to be widely used globally. It results from semi-dominant mutations at one or more of the collinear (homoeoallelic) Photoperiod-1 (Ppd-1) loci on chromosomes 2A, 2B and 2D. The Ppd-1 genes are members of the pseudo-response regulator (PRR) family orthologous to the Ppd-H1 gene of barley [4,5].
Four day neutral alleles have been characterized in wheat previously. They are given an a suffix (Ppd-A1a, Ppd-B1a and Ppd-D1a) while wild type alleles have a b suffix [6]. Three Ppd-A1a alleles and the single Ppd-D1a allele are deletions that remove a shared region of the promoter likely to be important for regulation [5,7].
Ppd-1b (wild type) alleles have a marked diurnal fluctuation in expression, with very low transcript levels at dawn, a peak 3-6 h after dawn and a subsequent drop to very low levels in the dark. Ppd-1a alleles that have been studied show elevated expression throughout the 24 h period but particularly at dawn and during the dark period. This is associated with expression of TaFT1 (the wheat orthologue of the flowering inducer FLOWERING LOCUS T) in short days [5,7]. This suggests that day neutral mutations share the property of mis-expression of Ppd-1 and the induction of TaFT1 without the normal long day cue.
Vernalization is another important flowering time control. The ancestors of wheat and many modern varieties are winter types that require vernalization, but during domestication there has been selection for spring types that do not. Spring types result from mutation of the Vernalization-1 (Vrn-1), Vrn-2 or TaFT1 (also called Vrn-3) genes [13][14][15][16]. Vrn-1 encodes a MADS-box transcription factor whose expression increases quantitatively during vernalization. This potentiates flowering by inhibiting expression of the flowering repressor Vrn-2 [17,18]. Semi-dominant mutations at one or more of the homoeoallelic Vrn-A1, Vrn-B1 or Vrn-D1 genes (chromosomes 5A, 5B and 5D, respectively) are the predominant cause of spring types in wheat [16].
In addition to the winter/spring difference, winter types vary in the duration of cold necessary to complete vernalization. We refer to this as variation in vernalization requirement. This aspect of vernalization is much less studied, but recently it was proposed that Vrn-A1 was the cause; specifically a single base difference in exon 4 [19]. We investigated three UK winter wheat varieties with short, medium and long vernalization requirements, respectively. Genetic mapping, sequencing and analysis of gene expression also identified Vrn-A1 as the cause but in contrast to [19] the exon 4 difference did not distinguish the three types.
For the Ppd-B1a and Vrn-A1 alleles we investigated, the changes in flowering time were associated with increased gene copy number and not with specific sequence polymorphisms. Copy number variation (CNV) is a type of mutation that has been widely studied in human genetics [20,21] and is well recognized in studies at the whole genome level in plants [22,23] although those studies include changes arising from segemental duplications, differences between homoeologous chromosomes and variation within gene families. Recent whole genome sequence and gene expression analysis of Arabidopsis thaliana accessions [24,25] where it was shown that 9% of genes showed CNV between at least two of the 19 accessions studied but many of these are probably pseudogenes as only 388 were expressed [25]. Of these, 237 (61%) were differentially expressed and for 54 (14%) this was attributed to CNV, leading to the conclusion that CNV has a modest effect on gene expression and by extension would have little impact on phenotype. In contrast, work on CBF genes in barley in relation to abiotic stress [26] and our studies on flowering time started with phenotypic variation and identified independent CNV alleles. This suggests that CNV may be the cause of important adaptive variation.

Results
Flowering time phenotypes of Ppd-B1a alleles Firstly, we describe our analysis of Ppd-B1. The photoperiod sensitive hexaploid spring wheat variety 'Paragon' was the recurrent parent for introgression lines containing Ppd-1a alleles from 'Chinese Spring', 'Sonora64' or 'Récital'. Two independent lines of each allele were developed and from each of these ten plants at BC 3 F 4 ('Récital' Ppd-B1a allele) or BC 4 F 4 (other alleles) were grown in short days (10 h natural light) in a photoperiod glasshouse. Days to ear emergence on the main stem were recorded for each plant (Figure 1). Plants with the 'Sonora64' Ppd-B1a allele flowered earlier than those with the 'Chinese Spring' or 'Récital' Ppd-B1a alleles. Plants with Ppd-A1a or Ppd-D1a alleles flowered earlier than those with a Ppd-B1a allele, consistent with [27], showing that these Ppd-B1a alleles had comparatively weak phenotypes. This contrasts with the Ppd-B1a allele described in [28] which has a phenotype comparable to Ppd-D1a.
Copy Number Variation (CNV) at the Ppd-B1 locus It was previously shown that 'Chinese Spring' has a truncated Ppd-B1 gene in addition to an intact version [5]. These are in close proximity because both are present on individual BAC clones. To determine copy number accurately a Ppd-B1 specific quantitative TaqManH assay was developed using a region present in the intact and truncated copies. Part of the wheat CONSTANS2 gene (TaCO2, also called TaHd1 [29]) from the group 6 chromosomes was used as an internal control. Nullisomic-tetrasomic lines of 'Chinese Spring' verified the assay.
Ppd-B1b genotypes had a haploid copy number of one (i.e. one copy per 2B chromosome), while 'Chinese Spring' had four copies. Furthermore, the other genotypes with a day neutral Ppd-B1a allele also had elevated copy numbers; two in 'Récital' and three in 'Sonora64', 'Timstein' and 'C591' (Figure 2). A 'Mercia(Chinese Spring 2B)' single chromosome substitution line in which the 'Chinese Spring' 2B (carrying Ppd-B1a) replaced the 'Mercia' 2B had the same increased copy number as 'Chinese Spring' itself. Conversely, a line in which the 2B chromosome from the photoperiod sensitive variety 'Marquis' (Ppd-1Bb) replaced the 'Chinese Spring' chromosome had a low copy number, showing that all the additional gene copies were on chromosome 2B. These data suggested that increased copy number could be the basis of the Ppd-B1a alleles.
The 'Chinese Spring' type Ppd-B1a allele A BAC library [34] allowed the 'Chinese Spring' allele to be fully described. Three overlapping B genome BAC clones (97J10, 170M01 and 792K15) were sequenced and three additional BACs (56G10, 883C12 and 1630O19) were analyzed by end sequencing, restriction digestion and pulse-field gel electrophoresis (PFGE) ( Figure 3A). These BACs were originally selected by filter hybridization [5]. They were further classified by restriction mapping and PCR to identify BACs containing the truncated copy.
Sequence assembly from 792K15 was straightforward and showed a single intact copy of the Ppd-B1 gene flanked by retrotransposons. 97J10 and 170M01 did not assemble readily because of sequence differences at the ends of approximately 35 kb regions containing either an intact or truncated gene. Digestion with NotI allowed BAC clone inserts to be sized. 170M01, which did not contain the truncated copy, would contain three repeat units of approximately 35 kb. PFGE of 97J10 using AarI, which does not cut within the truncated copy, placed the truncated copy at the left end of the repeating structure (copy_1), showing that the Ppd-B1a locus of 'Chinese Spring' comprised one truncated copy and three intact copies of the Ppd-B1 gene, all in the same orientation, within a 185 kb region ( Figure 3A; JF946485).
An A/G SNP in exon 3 identified by [5] distinguished the Ppd-B1a allele of 'Chinese Spring' (A form) from the other varieties in Figure 2. Sequence trace files from the different BAC clones and a Tsp509I CAPS assay of amplicons from 'Chinese Spring' and individual BAC clones showed that all copies had the A form. Although exon and intron sequences were identical apart from the truncation in copy_1, different haplotypes were found upstream as shown by six nucleotides below each gene copy ( Figure 3A). The haplotype found in 97J10 but not in 170M01 or 792K15 was assigned to the truncated copy and 792K15 gave the haplotype of copy_4. Restriction digestion and religation of 170M01 with BsiWI recovered subclones containing the leftmost intact copy from which the haplotype of copy_2 was obtained. The remaining haplotype was assigned to copy_3. Haplotypes, BAC sizes and TaqManH assays were all consistent with a copy number of four. A model for the origin of the 'Chinese Spring' allele is shown in Figure S1. We suggest that breakage and repair generated a chromosome with two copies. Unequal crossing-over then generated three and four copy versions and partial deletion of one copy at an intermediate or final stage gave the current structure.
The junction between intact gene copies was verified by PCR amplification and sequencing and used to design a diagnostic PCR assay ( Figure S2). This produced an amplicon from 'Chinese Spring' and lines into which the 'Chinese Spring' Ppd-B1a allele was introgressed. No amplicon was produced in other Ppd-B1a or photoperiod sensitive genotypes. A PCR assay for the gene/ transposon breakpoint in the 'Chinese Spring' truncated copy gave the same result ( Figure S2). The copy number of four, the intercopy break point, the exon 3 SNP and the partial deletion of one copy were therefore specific to the 'Chinese Spring' allele.
The 'Sonora64' type Ppd-B1a allele 'Sonora64' and 'Timstein' had a haploid copy number of three ( Figure 2). Partial sequence from the 'Timstein' allele (DQ885765) was previously reported [5]. The 'Chinese Spring' sequence was used to design additional PCR amplicons to extent the sequence further upstream and downstream and to sequence the 'Sonora64' allele (JF946486). The sequences were identical, gave no indication of variation between copies and confirmed that there was no mutation such as a promoter insertion or deletion likely to cause the day neutral phenotype. However, SNPs distinguishing the 'Sonora64'/'Timstein' allele from 'Chinese Spring' were found ( Figure 3B).
Long range PCR using primers at the ends of the gene sequence amplified an intercopy fragment of approximately 9 kb from 'Sonora64' (Figure 3B; JF946487). An identical sequence was obtained from 'Timstein'. The fragment was composed of transposable elements previously found flanking the Ppd-B1 gene in 'Chinese Spring'. However, the breakpoint was different and the intercopy fragment lacked the WIS element of the 'Chinese Spring' intercopy sequence. A PCR assay for the 'Sonora64' breakpoint produced an amplicon from 'Sonora64', 'Timstein' and 'C591', the other variety with a copy number of three ( Figure 2). No PCR product was produced from 'Chinese Spring' or 'Récital' ( Figure S2). The different intercopy breakpoints and haplotypes show that the 'Sonora64'/'Timstein'/'C591' and 'Chinese Spring' alleles must have derived from independent amplification events. . shows 59 to 39 orientation) are large dark green rectangles and the solid black line shows upstream and downstream regions with no homology to known repeated sequences. Small coloured rectangles show transposable elements annotated using the TREP database (http://wheat.pw.usda.gov/ITMI/Repeats/). Vertical arrows show the junction between intact copies (black) or between the gene and transposon in the truncated copy (red). B) Haplotypes of Ppd-B1 alleles. Intervarietal SNPs are in red. These were invariant between copies in 'Chinese Spring' and for clarity only the positions in copy_2 are shown. Bases distinguishing the copies in 'Chinese Spring' are in black (IUPAC codes). Coloured rectangles in the 'Sonora64'/'Timstein' intercopy fragment are transposable elements as in (a). Vertical blue arrow shows the junction between copies. Junction sequences are given in Figure S2. doi:10.1371/journal.pone.0033234.g003 The 'Récital' type Ppd-B1a allele 'Récital' had a haploid copy number of two ( Figure 2) and the same haplotype as 'Sonora64' and 'Timstein' (Figure 3B; DQ885763). However, long range PCR using a range of primers failed to amplify an intercopy fragment. Analysis of 143 BC 3 F 2 plants in which the 'Récital' allele was introgressed into the photoperiod sensitive 'Paragon' background showed a 3:1 segregation for early flowering in short days (10 h light) which was correlated with copy number, showing the copies to be genetically linked. This suggests that the 'Récital' allele is derived from a third independent amplification event or from the 'Sonora64'/'Timstein' type by further mutation.
The 'Paragon', 'Cheyenne' and 'Renan' Ppd-B1b alleles PCR amplicons based on the 'Chinese Spring' sequence were also used to obtain extended sequence from 'Paragon' (DQ885762), 'Cheyenne' and 'Renan' which have wild type Ppd-B1b alleles ( Figure 2, S2). This showed that no SNPs were consistently associated with the Ppd-B1a alleles ( Figure 3B). This suggests that increased copy number may be the cause of the day neutral phenotype.

Ppd-B1 expression
Expression was analyzed using a Ppd-B1 specific quantitative RT-PCR assay. First, 'Mercia' (Ppd-B1b) was compared with a 'Mercia(Chinese Spring 2B)' (Ppd-B1a) single chromosome substitution line. Plants were grown in short days (9 h light; 16uC) for 21 days and RNA samples were taken from three biological replicates, each of four plants, at 3 h intervals over a 24 h period. Expression of the 'Mercia' allele was similar to that of Ppd-1b alleles in previous studies [5,7] with very low expression at dawn and a peak of expression 3-6 h after dawn. The 'Chinese Spring' Ppd-B1a allele had significantly higher levels at dawn, a higher expression peak, and elevated expression throughout the dark period ( Figure 4A). This change was similar to that found for Ppd-A1a and Ppd-D1a promoter deletion alleles [5,7] but was less extreme, consistent with the weaker phenotype of the 'Chinese Spring' Ppd-B1a allele ( Figure 1).
There was no sequence polymorphism between the expressed regions of the intact gene copies so it was not possible to determine if all were expressed. However, RT-PCR using a forward primer in exon 6 and a reverse primer in the WIS retrotransposon produced a product with intron 6 correctly spliced, showing that the truncated copy was expressed. The transcript had a stop codon 2 bp beyond the junction which would generate a protein lacking the conserved CCT domain and which is likely to be nonfunctional.
Subsequent development of 'Paragon' introgression lines with different Ppd-B1a alleles allowed an additional comparison. Plants of 'Paragon' and 'Paragon(Chinese Spring Ppd-B1a)', 'Paragon(So-nora64 Ppd-B1a)' and 'Paragon(Récital Ppd-B1a)' introgression lines were grown in short days and sampled at the end of the dark period and at three time points in the lit period as described above ( Figure 4B). This showed that increased copy number of Ppd-B1 was always associated with significantly elevated gene expression, particularly at dawn when expression in the wild type was very low.
Having identified copy number variation at the Ppd-B1 locus and described its relationship to photoperiod response, we now consider variation in vernalization requirement, which is another important flowering time control.

Flowering time after different vernalization treatments
The winter wheat varieties 'Claire', 'Malacca' and 'Hereward' require short, medium or long low temperature treatments, respectively, to achieve full vernalization. To study this quantitatively, plants were vernalized at 7uC in 8 h days in increments of 1 week for 0 to 10 weeks ('Hereward') or 0 to 6 weeks ('Claire' and 'Malacca') after which they were transferred to a heated lit glasshouse providing long days (18 h light). Day length was sufficient to satisfy photoperiod response, and genotyping showed that no Ppd-1a allele was present.
Days to ear emergence on the main stem were recorded for five plants of each genotype per time point. Unvernalized 'Claire' plants flowered about 40 days before 'Malacca' but the acceleration of flowering per week of vernalization was greater in 'Malacca' so that flowering times were similar after 4 or more weeks of cold. Unvernalized plants of 'Hereward' and 'Malacca' flowered at a similar time but the 1 to 4 week treatments in 'Hereward' had less effect on flowering, after which the response was similar to 'Malacca' ( Figure 5).

Genetic mapping of vernalization requirement
To investigate the genetic basis of long vernalization requirement in 'Hereward' we selected a 4 week vernalization treatment (7uC, 8 h light) to study a population of 77 doubled haploid (DH) lines from a 'Malacca'6'Hereward' cross. Three plants of each DH line and four plants of each parent were vernalized and then grown in a heated glasshouse with supplementary lighting providing long days (18 h light). Days to ear emergence on the main stem were recorded and averaged for each genotype. QTL mapping showed one significant peak (LOD.22.0; Figure S3) associated with barc151, an SSR marker linked to Vrn-A1 on chromosome 5A [35].

Copy Number Variation at the Vrn-A1 locus
Approximately 12 kb of Vrn-A1 sequence including all exons and introns was obtained from 'Malacca' (JF965396) and 'Hereward' (JF965397) by direct sequencing of overlapping PCR amplicons generated from gene specific primers based on the 'Triple Dirk C' winter allele (AY747601 [16]). No differences previously associated with spring alleles were found and no sequence difference was found between 'Malacca' and 'Hereward'. However, both consistently gave an anomalous C/T double peak in sequence trace files from the amplicon covering exon 4 ( Figure  S4). Cloning and sequencing of the amplicon gave individual clones with either a C or a T, showing that two versions of exon 4 were present. Chi 2 analyses showed that the number of C and T clones for 'Malacca' (63:73) only fitted a 1:1 ratio while for 'Hereward' (51:87) they only fitted a 1:2 ratio, suggesting two and three copies of Vrn-A1, respectively. Full length cDNA clones from RNA of vernalized 'Malacca' and 'Hereward' plants also had the C and T forms showing that two intact variants of the gene were present and expressed. The C/T difference in exon 4 is likely to be that previously reported as distinguishing 'Jagger' from '2174' [19]. Our results differ in that both sequences were found within 'Malacca' and 'Hereward'. The presence of C and T sequences was not due to heterozygosity or mixed seed stocks as it was consistent and did not segregate in progeny plants.
A TaqManH assay specific for Vrn-A1 (using the same internal control employed for Ppd-B1) confirmed a haploid copy number of two for 'Malacca' and three for 'Hereward' (Figure 6A). The 'Malacca'6'Hereward' DH population showed a 1:1 ratio for copy number showing that the copies were linked in both varieties and behaved as single genetic entities. When Vrn-A1 was scored qualitatively by copy number the QTL peak coincided with Vrn-A1 ( Figure S3) and higher copy number was associated with late flowering after 4 weeks vernalization ( Figure 6B). These results show that 'Malacca' has a duplication of the Vrn-A1 gene with one copy having a C in exon 4 and one copy a T. 'Hereward' has one copy of the C form and two copies of the T form which could be the result of unequal crossing over or an additional breakage and rejoining event. Long range PCR was attempted between copies but no intercopy fragment was obtained.
'Malacca' and 'Hereward' had the same haplotype as 'IL369' (AY747599). The latter has a spring allele of Vrn-A1 [36] with a deletion in intron 1 but also an intact intron 1 sequence, suggesting a duplication [16]. We found that 'IL369' had a copy number of two, suggesting it derived from the 'Malacca' type winter allele, and copy specific PCR amplicons showed that the intact and deletion copies had a C and T in exon 4, respectively ( Figure S5B). As an intron 1 deletion is considered sufficient to induce expression and confer a spring growth habit it is likely that the T form is functional even though it changes a conserved aminoacid in the K domain of the VRN-A1 protein ( Figure S5B [37]).
'Claire' had a haploid copy number of one ( Figure 6A) and its sequence (JF965395) had a different haplotype to 'Malacca' and 'Hereward'. Exon 4 in 'Claire' was the C type and there was an additional exon 7 SNP which was C in 'Claire' and T in 'Malacca' and 'Hereward' (i.e. a T was present in all the gene copies). This enabled a 'Claire'6'Hereward' F 2 population to be used for an additional test of the genetic behaviour of the 'Hereward' Vrn-A1 copies ( Figure 6C).
Ninety-six F 2 plants and four plants of each parent were given 4 weeks vernalization and grown as described above. Days to ear emergence on the main stem were recorded. Although there was a deficiency of 'Claire' type homozygotes there was a clear association of the exon 7 SNP with flowering time.  population was analyzed using the TaqManH assay ( Figure 6C). This is consistent with linkage of the 'Hereward' copies and, as with the 'Malacca'6'Hereward' data, increased copy number of Vrn-A1 was strongly correlated with later flowering.

Vrn-A1 expression
Vrn-1 expression increases quantitatively during vernalization [13,17,18]. To compare 'Claire', 'Malacca' and 'Hereward', seedlings were harvested at the end of each week in the same experiment used to measure flowering time. Vrn-A1 expression was measured by gene specific quantitative RT-PCR.
Total Vrn-A1 expression (combining the exon 4 C and T forms) increased most rapidly in 'Claire' and was slowest in 'Hereward', while 'Malacca' was intermediate ( Figure 7A). The C and T forms were also analyzed separately. The C form, which encodes the typical form of the protein, gave the clearest difference between the varieties and was most rapidly induced in 'Claire' and most slowly in 'Hereward', with 'Malacca' intermediate ( Figure 7B). Similarly, the T form was more rapidly induced in 'Malacca' than in 'Hereward' ( Figure 7C). Irrespective of whether the T form affects flowering, all the assays showed a trend that increased copy number was associated with slower induction of expression, consistent with the observed differences in flowering time.

Discussion
Changes in flowering time in wheat have previously been shown to be caused by sequence variation such as insertions, deletions and point mutations in the Ppd-1 and Vrn-1 genes. We show that flowering time variation is also associated with these genes when they show CNV but no sequence changes likely to affect function. CNV was consistently associated with altered gene expression and it is highly unlikely that CNV in independent alleles of Ppd-B1 and in Vrn-A1 would occur by chance. Taken together the observations suggest that CNV was a plausible cause of the observed flowering time variation.
CNV is a significant source of polymorphism in the human genome [20,21] but is less studied in plants, particularly in crops. However, a barley variety with increased boron toxicity tolerance had an increased copy number of a boron transporter [38] and  increased frost tolerance in barley is associated with increased CBF gene copy number at the Fr-2 locus [26]. Variation has also been reported for the Vrn-2 locus in tetraploid wheat but it is not clear if this affects phenotype [39]. In the case of CBF the variation is found within a large cluster of around ten related genes [36], while Vrn-2 is normally found in tandem with one or two closely related genes [14,39]. In these cases the presence of related sequences may favour CNV by unequal crossing over. In contrast, CNV for Ppd-B1 and Vrn-A1 affects genes that are single copy in wild type plants. It is likely that CNV alleles first arose by double-strand breakage and repair. Beneficial phenotypes resulting from altered flowering time would have been selected by farmers and phenotypes could have been enhanced by selecting higher copy numbers that resulted from unequal crossing-over.
For both genes we examined increased copy number was always associated with altered gene expression. Vrn-A1 expression is very low in germinating seedlings and increases quantitatively during vernalization [13]. Vrn-A1 was studied over 6 or 10 weeks and this showed that increasing copy number was found in alleles that required progressively more exposure to cold to become active ( Figures 5, 6, 7). This was strongly correlated to differences in flowering time.
For Ppd-B1 such a clear correlation was not observed ( Figure 4B) but all alleles with increased copy number showed a deregulated pattern of expression and significantly increased expression at times when the transcript level of wild type alleles was very low. This was similar to changes in Ppd-A1a and Ppd-D1a alleles that confer a day neutral phenotype as a result of promoter mutations [5,7] although the level of change in Ppd-B1a alleles was less, consistent with their weaker phenotypes (Figure 1). A tandem organization of CNV alleles predicts that they will not be completely stable but will give recombinants with higher or lower copy number. These will be very valuable for further study of the interaction between copy number and expression and for exploring the mechanism by which CNV might act. Possibilities include the titration of regulating proteins and, or, changes in chromatin organization. To resolve the mechanisms it is desirable to identify or create alleles in which the different gene copies can be distinguished.
Results from Ppd-B1 prompted us to re-examine Ppd-A1a and Ppd-D1a alleles using TaqManH assays, but we found no evidence for CNV. We therefore conclude that promoter mutation and CNV are two distinct mutational mechanisms leading to day neutrality. It is not clear why CNV only affects the Ppd-B1 locus. This may be chance and CNV for the A and D genomes might be found in more extensive screens. Similarly, CNV may exist for Vrn-B1 or Vrn-D1 and it would therefore be of great interest to examine a wider range of germplasm.
We conclude that CNV is a significant source of variation for flowering time in wheat and may be an important but currently under recognized component of variation in general in crops. It will be important to consider CNV in germplasm screens and genetic mapping studies and to develop efficient markers for this type of polymorphism.

Plant materials
'Mercia' and 'Mercia(Chinese Spring 2B)' lines were from the John Innes Centre Germplasm Resources Unit. 'Paragon(Récital Ppd-B1a)' lines were developed as follows. 'Récital' carries Ppd-B1a and Ppd-D1a alleles [12]. Sixty-four 'Paragon'6'Récital' BC 1 F 2 plants were assayed for Ppd-D1a as described in [5] and Ppd-D1a carriers discarded. The remainder were grown in short days (10 h natural light). Early flowering plants would have Ppd-B1a and these were backcrossed to 'Paragon'. BC 3 F 2 families in short days showed 3:1 segregation for flowering time and early flowering plants had higher Ppd-B1 copy number, showing that Ppd-B1a was the major determinant of flowering time. BC 3 F 3 families that were all early flowering in short days (10 h natural light) were selected. BC 3 F 4 plants were compared to 'Paragon' and previously described 'Paragon' BC 4 F 4 introgression lines homozygous for the 'GS-100' Ppd-A1a, 'Chinese Spring' Ppd-B1a, 'Sonora64' Ppd-B1a or 'Sonora64' Ppd-D1a alleles [27]. Plants were grown under summer conditions using a glasshouse with moving benches that travelled into a closed (dark) room after 10 h in the light.

Selection and sequencing of Ppd-B1 BAC clones
Filter hybridization was used to select clones containing Ppd-1 genes from a 'Chinese Spring' BAC library [5,34]. B genome BAC clones selected by PCR using gene specific primers were sequenced by the commercial service of the Washington University in St. Louis Genome Sequencing Centre, 4444 Forest Park Boulevard, St. Louis MO 63108, USA. BAC sequence provided restriction endonuclease site positions. Restriction using Bsu36I or BsiW1 followed by re-ligation and selection with chloramphenicol allowed recovery of the rightmost and leftmost Ppd-B1 copies, respectively, in resulting subclones. The haplotypes of individual copies was determined by standard PCR (above) and Sanger sequencing (see sequencing of Vrn-A1).

Ppd-B1 intercopy sequencing
Long range PCR using primers facing outwards from the ends of the gene sequence (catttctcccgcttgtaagg and caacgttacagcttcggtca, respectively) was conducted using the Roche Expand Long template kit (Roche GMBH, 66298 Mannheim, Germany). This produced an approximately 9 kb fragment in 'Sonora64' and 'Timstein', which was analyzed by Sanger sequencing (below) using internal primers in a stepwise manner.
PCR assays for junction sequences in Ppd-B1a alleles PCR with ''Hot Start'' polymerase and buffer (Qiagen) and an initial denaturation step of 95uC for 5 minutes was used to detect; (a) the truncated Ppd-B1 gene in the 'Chinese Spring' allele. Primers 219H05F2 (taactgctcctcacaagtgc) and 97J10R2 (ccggaacctgaggatcatc) gave a 425 bp product; (b) the junction between intact Ppd-B1 copies in the 'Chinese Spring' allele. Primers PpdB1_F25 (aaaacattatgcatatagcttgtgtc) and PpdB1_R70 (cagacatggactcggaacac) gave a 994 bp product; (c) the junction between intact Ppd-B1 copies in the 'Sonora64'/'Timstein' allele. Primers PpdB1_F31 (ccaggcgagtgatttacaca) and PpdB1_R36 (gggcacgttaacacaccttt) gave a 223 bp product. Junction sequences and primer positions for assays are given in Figure S2.

Sequencing of Vrn-A1
'Triple Dirk D' vrn-A1 sequence (AY747601) was used to design a series overlapping PCR amplicons spanning the gene. Amplicons were obtained from genomic DNA using the standard PCR protocol and were directly sequenced using ABI Big Dye Mix v3.1 (Applied Biosystems Inc) under the manufacturer's conditions, with products resolved on an ABI 3730 capillary electrophoresis instrument. Segments of Vrn-A1 from 'IL369' were amplified using primers IL369_F1 (tgaattcatgatcggagcag) and IL369_R1 (tgctgaacttctctgcaagtg) or IL396_F2 (tgctgaacttctctgcaagtg) and IL369_R1 ( Figure S5) and directly sequenced.

Time course expression of Ppd-B1
RNA of 'Mercia' and 'Mercia (Chinese Spring 2B)' was extracted from seedlings grown for 20 days after germination in a controlled environment room with short days (9 h light; 16uC). Three replicate samples from each genotype were harvested into liquid nitrogen at each three-hourly time point over 24 h. RNA was sampled from 20 day old seedlings of 'Paragon' and 'Paragon(Chinese Spring Ppd-B1a)', 'Paragon(Sonora64 Ppd-B1a)' and 'Paragon(Récital Ppd-B1a)' introgression lines in the same way.
RNA was extracted using Trizol reagent (Invitrogen; http:// www.invitrogen.com). DNA was removed by DNAseI digestion. cDNA was synthesized with Superscript II (Invitrogen) using the manufacturer's protocols with 5 mg of total RNA as template and a mixture of OligodT (12-18) (250 ng) and random hexamers (150 ng) as primers. 1/40 by volume of the final cDNA aliquot was used for real-time PCR as described below. In the case of 18S rRNA analysis, cDNA samples were diluted 1:100 and 1/40 of this dilution was used as template.
Primers for Ppd-B1 expression were agacgattcattccgctcc and agcagcaccatttgacagg (N.B these will detect transcripts from the intact and truncated copies in 'Chinese Spring'). The forward primer was specific for Ppd-B1, and genome-specificity was confirmed using 'Chinese Spring' nullisomic-tetrasomic lines. The primers generate a 549 bp amplicon from cDNA. Similarly, primers taactgctcctcacaagtgc and ccggaacctgaggatcatc were used to detect expression of the truncated copy in 'Chinese Spring' ( Figure S2). Quantitative RT-PCR was carried out essentially as in [7] (File S1).

Quantification of Vrn-A1 expression
Samples of leaf tissue from plants of each genotype/vernalisation time combination (three biological replicates per sample) were harvested in the vernalisation chamber at the end of each treatment 3-4 h after dawn and frozen in liquid nitrogen. RNA and cDNA were prepared as above, and Vrn-A1 cDNA abundance was estimated using Real-Time RT-PCR (Supporting information). Primer pairs were: VrnA1-all: cttgaacggtatgagcgctat and gcatgaaggaagaagatgaag, product size 396 bp.
VrnA1-exon 4 T form: gcatctcatgggagaggatT and gaatagtacgcctgtatgggctggat, product size 436 bp. This assay produced a low level signal in 'Claire', which does not contain the T form ( Figure 7C). We attribute this to mis-priming in the absence of a perfectly matching template. Primer positions for this assay are underlined in the sequence below. Ppd-B1 gene sequence is to the left of the arrow (exons in uppercase, intron in lowercase) and the TREP 3161 WIS element is to the right of the arrow. A 425 bp band is produced when the junction is present. …TAACTGCTCCTCACAAGTGCCGG AAGGGAAAGACGCCGACCGTGAGAACGCCATGCCATA TCTTGAGCTGAGCCTAAAGAGGTCGAGATCGACCACG GAGGGTGCGGATGCGATCCAGGAGGAACAGAGGAACG TCGTGAGACGATCAGACCTCTCGGCATTCACGAGgtgcaa agcataatatcagtgtcctttgtgaatccttaaatcatccatatgttgcatactaaccgttttcattctt tgcaagGTACAATACGTGCTCGTTCTCCAATCAAGGCGGG GCAGGGTTCGTCGGGAGCTGTTCGCCCAqcgtgactgccaag cgttcataacgtcttggttctatgggatgggtgcttcacctagcggtccttctaggacatatgctttct tggcagctatgaggatgatcctcaggttccgg. The same primers were used for RT-PCR where a 343 bp product was produced from cDNA. The product was cloned and sequenced, confirming that intron 6 was correctly spliced. The transcript has a stop codon (double underline) close to the break point which gives a predicted protein lacking a CCT domain. (B) Assay for the junction between intact Ppd-B1 gene copies in the 'Chinese Spring' allele. Primer positions for this assay are underlined in the sequence below. The left primer spans the junction between Ppd-B1 gene sequence and the TREP 3457 Danae element. The junction (vertical arrow) marks the start of the TREP 3161 WIS element sequence which contains the right primer. A 994 bp band is generated when the junction is present. …cacacaagcattatacctggaaccgtagcataattttcctcaacattttgagctacgtgtatggagtttgcattgttatcttcaacatttctcccgcttgtaaggcctgcaaaacattat gcatatagcttgtgtcggtgtacaaaagtaggagctctgcttttgacccctttacttgtgcacgag cagtcagagccacccgccacggccacgcataacagggcagggaagggaagccggagaga agccgaaggcacaagacaaacaaagcaacgacaaggaccaaggctacaaagtgcagatg ggcgaagcaggtttcccctgcaagacccttgccgggggcagcctccgcagccccggcaagcccctt tgccggggcaactcgcccacaccaatggagagagccacccttgaacccacggcctccaatgtcaaccaccacgttggaccagggctcgagaggcacctccatggtggcatgcagatctttgtgaagacatagaatgctcaagaacatgtgatgattggagggtaacaatcctcaccaagatcctcaccgaataaacccacaagacccccccggcaagatccttgccgaggacggcaagcgccacggcaa gacccttgccgggccacccaacgagaccctcgccagaggcaaccacgagcccactgccagg cccgcaccaaccaactccccgccgccgttcgcatgcagctgccaacccaaccagctgagcag gcacctgcgtggcaacatgcagcttctaggcccactcagcacacacctgtgtggcggcatgcagatcttcgtggaggctccaccaccgcaccacctcagctgcttgcctgcctacatggcaccgcat gcatcgctggccagggcgcgtgtcgaagcaaggaggagcggcgacggatgggacgggcgtc gctcccgtccccgataaagtgagggacacctaagccatgcatqgttcatcatgtgagacgga ctagtcatcatcggtgaacatctccatgttgatcgtatcttccatacgactcatgttcgaactttcg gtcccttgtgttccgagtccatgtctgtacatgctaggctcgtcaagttaacataagtgttttgcat gtgtaaatctgtcttacacccattgtatgtggacgttggaatctatcacacccgatcatcacgtgg tgcttcgaaacaatgaactt… (C) Assay for the junction between intact Ppd-B1 gene copies in the 'Sonora64'/'Timstein'/'C591' allele. Primer positions for this assay are underlined in the sequence below. The junction is shown by a vertical arrow. The TREP 3457 Danae element is to the left of the arrow, the TREP 3173 Derami element is to the right. A 223 bp band is generated when the junction is present. …tcatggttgatgcacttgccggtggattccatctctt cttgggcggatctgtcccatgagtttgttggcgcctttaccggaggtcaccaagctcatggcca ggcgagtgatttacacatcacccctcagaaggaatgagaaaacttgcgcaagtacatccaaag gttcagccgggtgcagtacaacatccccgatgttcatcccgccgtcgtgatcagtgtgttccatt agaatgtgcgcaaccgcaagatgcgtgaagagctagcgatgaacaaqaagttaaattttca tgaagtgaaataaaaggtgtgttaacgtgcccatgccatctattttcaaaaataaataaataat ctttaagaaaagttaaatgcataaaaaagtccaatagtttacatatacattggctcaatatagtg gtgaggcatatttatttatattactcatggtcttt… Descriptions of repeat sequences and TREP numbers in (A) to (C) follow Figure 3 in the main text. (TIF) Figure S3 QTL Cartographer plot of days to ear emergence in 'Malacca'6'Hereward' DH lines. Plants were vernalized for 4 weeks (5uC; 8 h light) and grown in lit glasshouse (18 h light). For each DH line the score was the mean of three plants. One region exceeded the significance threshold (black line at LOD 2.5) and this was on a 20 cM linkage group containing marker barc151 previously shown to be linked to Vrn-A1 [35]. When Vrn-A1 was scored qualitatively based on copy number scores from the TaqManH assay ( Figure 6B in the main text) the QTL peak coincided with Vrn-A1 position. (TIF) Figure S4 Example trace files of directly sequenced PCR amplicons from the region of Vrn-A1 exon 4 containing the C/T variants. 'Hereward' and 'Malacca' had a C/T double peak (arrowed) while 'Claire' and 'Chinese Spring' had only the C form.  [40] plus the exon 4 C and T forms of VRN-A1. The T form changes a conserved leucine to a phenylalanine (arrowed position). (B) Analysis of the spring allele of Vrn-A1 from 'IL369'. Intr1/C/F, Intr1/AB/R, Intr1/A/F2 and Intr1/A/R3 are previously published primers for PCR assays [16].

Supporting Information
Additional primer positions from this paper are shown as red arrows with the primer sequences underneath. IL369_F1 is specific to the intact copy and an example trace file of the IL369_F1/IL369_R1 product is shown to the right. IL369_F2/ IL369_R1 could amplify from the intact and deleted copies but the much larger amplicon from the former was not observed. An example trace file from the IL369_F2/IL369_R1 product is shown to the right. The C/T variant is highlighted. (TIF) File S1 Quantification of Ppd-B1 expression. (DOC)