Introduction

Wheat, Triticum aestivum, is an allohexaploid comprising three diploid genomes identified as A, B, and D (Gill et al. 2004). Extant diploid relatives of hexaploid wheat include T. urartu (AA), a relative of Ae. speltoides (BB), and Ae. tauschii (DD), all of which are wild species endogenous to either the Fertile Crescent or Central Asia (Giles and Brown 2006; Zohary and Hopf 2000). Approximately 0.5 million years ago, hybridization between T. urartu and the Ae. speltoides relative occurred, giving rise to tetraploid emmer wheat, T. turgidum ssp. dicoccoides (AABB) (Luo et al. 2007; Ozkan et al. 2011). Then, approximately 10,000 years ago wild emmer was domesticated by humans giving rise to T. turgidum ssp. dicoccon (Luo et al. 2007; Matsuoka 2011; Ozkan et al. 2011). Subsequent interspecific hybridization events then occurred between cultivated emmer and Ae. tauschii, ultimately producing hexaploid wheat (AABBDD) (Charmet 2011; Matsuoka 2011; Zohary and Hopf 2000).

Today wheat is cultivated throughout the world. It is grown most-successfully in the temperate climate regions at latitudes 30°–60°N and 27°–40°S, but is also grown in more tropical regions (Curtis 2002). In temperate regions, it can be either autumn-sown or spring-sown. Autumn-sown genotypes are usually of a winter growth habit characterized by a requirement for vernalization, an 8–10 week exposure to low temperatures (0–10 °C) that causes the transition from the vegetative to the reproductive growth phase (Distelfeld et al. 2009; Greenup et al. 2009). In comparison, spring genotypes, which are spring-sown, are inherently reproductively competent and do not require vernalization to flower (Distelfeld et al. 2009; Greenup et al. 2009).

Autumn-sown genotypes must possess a level of freezing tolerance that enables them to survive minimum temperatures in a given region, and also be winter-hardy such that they are able to sustain freezing tolerance for an extended period of time appropriate for a particular region. However, most modern cultivars possess a level of freezing tolerance and winter hardiness that is only marginally greater than that required for the region where they are grown, and as such severe winter kill occurs about once every 10 years (Braun and Sãulescu 2002; Fowler and Gusta 1979). Spring genotypes are typically not selected for winter survival and, in general, are much less freezing tolerant than winter genotypes (Fowler and Limin 2004; Wilen et al. 1996).

Hexaploid wheat genotypes usually exhibit much greater freezing tolerance than diploid and tetraploid genotypes (Fowler et al. 1977; Limin and Fowler 1981). Within hexaploid wheat, the hard red winter (HRW) wheats grown in the North American Great Plains tend to be about 6 °C more freezing tolerant, and have a greater endurance for prolonged exposure to freezing temperatures and desiccating conditions than the soft red winter (SRW) wheats typically grown east of the Mississippi (Gusta et al. 1997, 2001).

In diploid and hexaploid wheat, genetic analyses have identified two loci that have a major effect on freezing tolerance and winter hardiness, both of which reside on the long arm of the group 5 homoeologous chromosomes (Båga et al. 2007; Galiba et al. 1995; Roberts 1990; Snape et al. 1997; Sutka and Snape 1989; Tóth et al. 2003; Vágújfalvi et al. 2003). These are identified as Frost resistance-1 (Fr-1), and Fr-2. Fr-1 and Fr-2 have also been identified in barley at genetically co-linear positions (Francia et al. 2004, 2007; Hayes et al. 1993; Knox et al. 2010; Skinner et al. 2006). Fr-1 is thought to be due to the product of the Vernalization-1 (Vrn-1) gene acting to concomitantly induce flowering and reduce freezing tolerance, the latter of which occurs in part through the downregulation of genes at Fr-2 (Dhillon et al. 2010; Stockinger et al. 2007). Fr-2 consists of a cluster of at least 11 different C-repeat binding factor (Cbf) gene coding sequences spanning approximately 1 cM genetic distance and 1 Mb physical distance (Knox et al. 2008, 2010; Miller et al. 2006; Skinner et al. 2005). CBFs are DNA-binding transcriptional activator proteins and are a major regulatory hub that affects control over the low temperature transcriptome in plants (Stockinger et al. 1997; Thomashow 2010). In barley, several 11 Cbf orthologs are duplicated, existing as identical or nearly identical paralogs in individual genomes (Knox et al. 2010). A single barley genome may harbor both Cbf2A and Cbf2B, Cbf10A and Cbf10B, Cbf12A and Cbf12B, and Cbf15A and Cbf15B (Knox et al. 2010). Additionally, the winter-hardy winter and facultative genotypes possess multiple copies of a tandemly duplicated 22 kb genomic segment encompassing Cbf2A and Cbf4B, whereas the spring genotypes possess only a single copy of Cbf2 and Cbf4 (Knox et al. 2010). Expression analyses indicate that the higher Cbf2A–Cbf4B genomic region copy numbers give rise to increased levels of CBF2 and CBF4 transcripts (Stockinger et al. 2007). As increasing CBF transcript levels increase freezing tolerance (Thomashow 1999), the increase in Cbf2A–Cbf4B genomic region copy numbers in winter barleys over spring barleys suggests that copy number variation (CNV) plays a role in underlying the effect of Fr-H2.

A CNV is defined as a DNA segment ranging in size from 1 kb to several megabases (Mb) whose copy numbers are variable between the genomes of two or more individuals within a species (Korbel et al. 2008; Stankiewicz and Lupski 2010). In mammalian genomes where CNV has been more extensively characterized, the frequency of CNVs is two-four orders of magnitude greater than that of single nucleotide polymorphisms (SNPs) (Korbel et al. 2008; Stankiewicz and Lupski 2010). Dosage-sensitive genes within copy number variable regions or genes involved in regulatory roles have pronounced phenotypic consequences when present in variable copy numbers (Korbel et al. 2008; Stankiewicz and Lupski 2010). In plants and mammals, the major classes of genes encompassed by CNV regions seem to be those involved with environmental response (Cook et al. 2012; DeBolt 2010; Díaz et al. 2012; Knox et al. 2010; Korbel et al. 2008; McHale et al. 2012).

The increase in copy numbers of the Cbf2A–Cbf4B genomic region in winter barley genotypes over their spring genotype counterparts raised the question whether Cbf gene CNV and the distinction between winter and spring genotypes were more widespread in the Triticeae. Several independent lines of investigation suggested that Cbf14 might be copy number variable between winter and spring growth habit types, and between different market classes of winter wheats. Expression analyses of Cbf2, Cbf9, and Cbf14 in hexaploid wheat indicated that these genes were expressed to higher levels in winter wheats than in spring wheats and that Cbf14 is expressed to higher levels in the HRW wheats than in the SRW wheats (Galiba et al. 2013; Stockinger et al. 2007). Additionally, single chromosome recombinants having the Fr-2 region from the HRW wheat ‘Cheyenne’ express Cbf7 (Cbf14 using revised nomenclature) to fourfold higher levels than recombinants having the Fr-2 region from T. spelta (Vágújfalvi et al. 2005). To test whether Cbf14 might be copy number variable a small collection of hexaploid spring and winter wheats was surveyed using DNA blot hybridization (Knox et al. 2010). Hybridization of a Cbf14 gene-specific probe obtained from the T. monococcum Cbf14 promoter to SacI-digested wheat DNAs indicated that Cbf14 was copy number variable; and suggested CNV occurred on all three group 5 chromosome homoeologs (Knox et al. 2010). Here, we further explore the extent of Cbf14 CNV in wheat. To determine the genome identity (A, B, and D) of the Cbf14 cross-hybridizing bands, we utilize the set of chromosome 5 nullisomic-tetrasomic hexaploid wheat lines (Sears 1966). Mapping to the individual genomes in combination with normalization of copy numbers through the use of the single copy Puroindoline b (Pinb) gene (Chantret et al. 2005) allowed us to determine the extent of Cbf14 CNV across the A, B, and D genomes of hexaploid wheat genotypes, irrespective of their growth habit and market class. We also ask what the Cbf14 copy number states were in the genomes of the diploid wheat species T. urartu, Ae. speltoides, and Ae. tauschii. In addition, we look at the inter-varietal chromosome substitution lines that have a long and established historical role in furthering our understanding of freezing tolerance and winter hardiness, as these were some of the first genetic tools used that played a key role in the current focus on chromosome 5 (reviewed in Stockinger et al. 2006).

Materials and methods

Plant material

Wheat genotypes, their ploidy level, seed source, growth habit or market class, pedigree, and location where the accession was collected or developed are provided in Table S1. Seedling tissue used for DNA extractions from all hexaploid wheats went through single seed descent. Tetraploid durum wheat seedlings used for DNA extractions utilized seed provided by the breeder. Diploid wheat DNAs were prepared from one to three plants grown from seed obtained from the source.

DNA extractions and hybridizations

For tetraploid and hexaploid wheats, about 50 seeds were sown in 9 cm × 9 cm pots (Kord Products, http://www.kord.ca) filled with Bacto high porosity soil mix, and placed in the greenhouse. At approximately 10 cm height, seedlings were cut about 2 cm above the soil. Leaf tissue was flash frozen in liquid nitrogen, lyophilized, and used for the isolation of high molecular weight (MW) DNAs as described (Stockinger et al. 1996). Diploid wheat plants were grown to larger size prior to tissue harvest.

Approximately, 10 μg DNA was digested with SacI or BglII restriction endonucleases (NEB, http://www.neb.com), ethanol precipitated, resuspended in 10 mM Tris (pH 8.5), and quantified using the Quant-iT™ PicoGreen® dsDNA reagent (Invitrogen, http://www.invitrogen.com). Equal quantities of each DNA sample were electrophoresed on 0.8 % TAE agarose gels, transferred to Hybond N membrane (GE Healthcare, http://www.gelifescience.com) and UV-crosslinked using standard procedures (Ausubel et al. 1993). Blots were hybridized with a 189 bp Cbf14 promoter fragment (subsequently referred to as the Cbf14 5′ fragment) and a 201 bp Puroindoline b (Pinb) coding sequence (CDS) fragment. To prepare hybridization probes, Cbf14 5′ fragment 368 bp distal to the Cbf14 MET initiator codon was PCR amplified from T. monococcum accession DV92 using forward primer: 5′-agcgagctgtccttgtcatt-3′ and reverse primer: 5′-gcatcttttgtggcgaaaat-3′. An alignment of the Cbf14 genomic region from T. monococcum genotypes DV92 and G3116 and two different A genome Cbf14 variants from the HRW wheat cultivar ‘Norstar’ (Knox et al. 2008; Miller et al. 2006; Ratnayaka et al. 2005) is presented in Figure S1, showing primer sites and the regions used as a probe. The Pinb CDS was PCR amplified from the T. aestivum N5A/T5D nullisomic-tetrasomic line using forward primer: 5′-tgatggagcgatgtttcaca-3′ and reverse primer: 5′-atacctcacctcgccaaatg-3′. Primers were designed using Primer3 (http://frodo.wi.mit.edu). The Tm-Cbf14 5′ and Ta-Pinb CDS templates were cloned into pGEM-T Easy (Promega, http://www.promega.com) and sequenced. Cloned Cbf14 5′ and Pinb CDS fragments were used for generating hybridization probes by PCR amplification using the vector’s M13 primer sites. Amplified fragments were then labeled with α-dCTP 32P via random hexamer labeling using the MegaPrime DNA labeling kit (GE Healthcare, http://www.gelifescience.com). Hybridization and washes were carried out as described (Knox et al. 2010). Blots were exposed to phosphorscreens and images were scanned using a Molecular Dynamics Storm840 PhosphorImager (GE Healthcare, http://www.gelifesciences.com). Following hybridization, the probe was stripped from the filter using 0.1 % SDS at 65 °C, and the filters were reused.

Estimation of Cbf14 copy numbers

Copy numbers of Cbf14 in hexaploid and diploid wheats were estimated using the grain hardness gene Pinb as a reference standard. Signal intensities of the Cbf14 5′ and Pinb CDS cross-hybridizing probes were quantified using the ImageQuant 5.0 software (Molecular Dynamics). Cbf14/Pinb ratios were then used to estimate haploid copy numbers of Cbf14. Pinb is a 447 bp open reading frame that is present in single copy on the D genome in hexaploid wheat, is deleted in tetraploid wheats, and is present in single copy in diploid wheats (Chantret et al. 2005; Gautier et al. 2000). In all of the hexaploid wheats, the Pinb cross-hybridizing pattern detected consisted of a single MW fragment. Absence of Pinb from the tetraploid wheat genomes was consistent with our hybridization data in which there was no Pinb cross-hybridizing fragment in any of the T. turgidum accessions characterized. In some of the diploid wheats, however, more than one Pinb cross-hybridizing band was produced. In all instances, the cumulative signal intensities of the Pinb cross-hybridizing bands were used to estimate Cbf14/Pinb ratios. For hexaploid wheat, CBF14 copy number differences between the A and D genomes of an individual genotype were also estimated by normalizing the signal intensities of the A genome cross-hybridizing fragment to that of the D genome cross-hybridizing fragment. For the tetraploid T. turgidum wheat accessions, only the Cbf14-A/Cbf14-B ratios were determined.

Nomenclature and gene symbols

Reference to loci and genes follows the Recommended Rules for Gene Symbolization in Wheat (http://wheat.pw.usda.gov/GG2/Triticum/wgc/2008/Catalogue2008.pdf). As this convention uses an all uppercase non-italic format for specifying protein, an all uppercase italic was also used to specify transcripts; e.g., CBF and VRN-1, when referring specifically to the mRNAs from these genes.

Results

Genome identity of the Cbf14 cross hybridizing fragments

In an earlier study, it was suggested that the Cbf14 gene might be variable in copy number across hexaploid wheat (Knox et al. 2010). Hybridization of a unique fragment representing the T. monococcum Cbf14 promoter (Cbf14 5′) region to hexaploid wheat DNAs restriction-digested with SacI produced four cross-hybridizing bands (Knox et al. 2010). We hypothesized the different bands represented the three homoeologous Cbf14 genes from the A, B, and D genomes. To test whether this was the case, the Cbf14 5′ fragment was used as a hybridization probe against six wheat lines nullisomic-tetrasomic for the group 5 homoeologous chromosomes (Sears 1966). In these lines, deletion of a chromosome pair is compensated by the addition of its homoeologous pair. For example, the line N5A/T5B is nullisomic for chromosome 5A and tetrasomic for chromosome 5B; i.e. it lacks the chromosome 5A pair, and carries an extra copy of the 5B pair.

Hybridization of the Cbf14 5′ fragment to this panel of nullisomic-tetrasomic lines produced two or three cross-hybridizing bands (Fig. 1). Using a SacI digest, lines nullisomic for 5A (N5A/T5B and N5A/T5D) and 5D (N5D/T5A and N5D/T5B) produced three cross-hybridizing bands, while lines nullisomic for 5B (N5B/T5A and N5B/T5D) produced two cross-hybridizing bands (Fig. 1a). Using a BglII digest, three cross-hybridizing bands were produced for the 5B and 5D nullisomic lines, and two cross-hybridizing bands were produced for the 5A nullisomics (Fig. 1b). With both digests, the absent band(s) corresponded to the homoeolog for which the lines were nullisomic (Fig. 1a, b). In this way, the chromosome 5 nullisomic-tetrasomic lines revealed the genome identity of the cross-hybridizing bands. Additionally, the banding pattern indicates the presence of a SacI restriction site internal to the cross hybridizing region of the B genome Cbf14 (Fig. 1a) and a BglII site in that of the A genome Cbf14 (Fig. 1b).

Fig. 1
figure 1

Cbf14 hybridization patterns in wheat lines nullisomic/tetrasomic for group 5 chromosome homoeologs. SacI (a) and BglII (b) digested DNAs were hybridized with a probe generated from the T. monococcum Cbf14 promoter. The A, B, and D, Cbf14 cross-hybridizing fragments, MWs 2.3, 4.6, and 1.65 kb, respectively, are identified and labeled according to their respective homoeolog. SacI restricts the B Cbf14 homoeolog into two cross-hybridizing bands; the lower MW of the two is indicated using an arrowhead. BglII restricts the A Cbf14 homoeolog into two cross-hybridizing bands. Insufficient quantity of DNA precluded the N5D/T5A line from the BglII digest

Full size image

The A and D genomes of hexaploid wheat have greater Cbf14 copy numbers than the B genome

Data indicating variation in Cbf14 copy numbers across different accessions of hexaploid wheat were obtained by surveying a small collection of genotypes (Knox et al. 2010). To explore the extent of Cbf14 CNV in wheat, we surveyed a panel of 50 hexaploid wheats that included 47 T. aestivum subsp. aestivum accessions and three T. aestivum subsp. sphaerococcum accessions. Of the 47 subsp. aestivum accessions, 35 are assigned to one of the five market classes—white spring (WhS), hard red spring (HRS), white winter (WhW), SRW, and HRW (Cox 1991; Zeven and van Hintum 1992); NPGS http://www.ars-grin.gov/npgs/index.html), whereas the remaining 12 T. aestivum accessions are unclassified.

In 43 of the 50 accessions, the Cbf14 cross-hybridizing bands migrated to the same MW positions as those of the nullisomic-tetrasomic lines (Fig. S2A and S3A; Fig. 1). Of the remaining seven accessions, three including ‘Blueboy’, ‘Mayview’, and ‘Odessa’ did not produce the typical D genome fragment and instead cross-hybridized to a fragment of higher MW, just below that of the B genome fragment (Fig. S3A). These three genotypes have both the A and B cross-hybridizing fragments. Loss of the fragment cross-hybridizing at the D genome MW and appearance of the novel MW support it to be from the D genome and a Restriction Length Polymorphism, but it is also possible that this is a disomic addition from another genome and the D genome compliment is absent. Four accessions, including ‘Sonora’ and ‘Mandel Gehun’ (Fig. S2A), and ‘PI 70711’, and ‘PI 40941’ (Fig. S3A), did not produce a Cbf14 cross-hybridizing B genome fragment.

To estimate Cbf14 copy numbers, we hybridized the same filters to a coding sequence probe of the Puroindoline b (Pinb) gene and quantified Cbf14/Pinb signal intensity ratios (Fig. S2 and S3; Table S2). Cbf14/Pinb ratios were determined for the A, B, and D genomes (Fig. 2, Table S2). These ratios were then used to estimate total copy numbers of Cbf14 in the hexaploid genome. Comparison of total Cbf14 copy numbers revealed the order HRW > SRW > HRS = WhW = WhS (Fig. 2a). The differences in Cbf14 copy numbers between HRW, SRW, and HRS wheats were significant (Fig. 2a). In all genotypes, regardless of their market class and growth habit, Cbf14 copy numbers were significantly higher in the A and D genomes compared to the B genome (Fig. 2b; Table S2). In SRW and HRW wheats, Cbf14 copy numbers in the A genome were significantly higher than those in the D genome (Fig. 2b). Additionally, Cbf14 copy numbers in the A and B genomes of HRW and SRW wheats were significantly higher than those of the other three classes (Table S2).

Fig. 2
figure 2

Cbf14 copy number estimates in hexaploid wheats, T. aestivum subsp. aestivum. Genotypes are grouped into five market classes—white spring (WhS), hard red spring (HRS), white winter (WhW), soft red winter (SRW), and hard red winter (HRW). a Cbf14 copy numbers relative to Pinb in the hexaploid genome; b comparisons of Cbf14/Pinb ratios across the three genomes within each class for each genome; c Cbf14 copy numbers in the A genome relative to those in the D genome (Cbf14-A/Cbf14-D). WhS, HRS, WhW, SRW, and HRW classes include 10, 6, 6, 5, and 8 genotypes, respectively (identified in Fig. S2 and Table S2). Error bars represent standard error of mean (SEM). Statistically significant differences are indicated by different letters above the error bars (P < 0.05)

Full size image

To assay relative genome copies of Cbf14 within a genotype, the signal intensity of the A genome Cbf14 cross-hybridizing fragment was normalized to that of the D genome fragment. The ratios of Cbf14-A/Cbf14-D were approximately 1:1 for WhS, WhW, and HRS wheats (Fig. 2c; Table S2). One exception was HRS wheat ‘Cadet’, which exhibited about threefold greater copy numbers in the D genome compared to A genome (Table S2). In comparison, the CBF14-A/CBF14-D ratios for SRW and HRW wheats were about 3:1 (Fig. 2c; Table S2).

Taken together, these data indicate that most HRW wheats have greater copy numbers of Cbf14 than the wheats of other market classes. Greater Cbf14 copy numbers of HRW wheats over the SRW wheats are significant for the B genome and also appear to occur for the A genome, however, differences between HRW and SRW in the A genome copies were not significant. Among red wheats, Cbf14 copy numbers are much higher in winter genotypes (hard and soft) compared to spring genotypes primarily due to increased copy numbers in the A genome. Differences in Cbf14 copy numbers between winter and spring white wheats were not detected in any genome however. Regardless of the market class and growth habit, the B genome harbored the fewest copies of Cbf14.

Cbf14 copy numbers vary between the A, B, and D genomes of diploid wheats

Differences in Cbf14 copy numbers across the A, B, and D genomes of hexaploid wheats raised the question as to what was the state of Cbf14 copy numbers in the ancestral diploid wheats. To address this question, we surveyed accessions of the diploid ancestors, T. urartu (AA), Aegilops speltoides (BB), and Ae. tauschii (DD). We also examined several accessions of einkorn wheat, T. monococcum (AmAm), a cultivated relative of T. urartu.

Using a SacI digest and the same Cbf14 promoter fragment as a probe, most of the A, B, and D genome diploid wheats produced bands of similar MWs as those of hexaploid wheats (Fig. S4A). Nonetheless, there was some variability. The MWs of the cross-hybridizing bands in T. urartu accessions, PI 428183 and PI 428316 ran lower than the A genome band of hexaploid wheats (Fig. S4A). T. urartu accession G1812 did not cross-hybridize to the Cbf14 5′ probe (Fig. S4A). Testing a Cbf2 gene-specific probe on the same filter resulted in a cross-hybridizing band with G1812 (not shown), indicating other Fr-2 Cbfs were still present in this genotype. The Ae. speltoides accessions exhibited several different hybridization patterns (Fig. S4A). The Ae. tauschii accessions exhibited two different banding patterns both of which consisted of one strong, and one weak cross-hybridizing band (Fig. S4A). As T. urartu and Ae. tauschii are self-pollinating species, the two different banding patterns in each suggest at least two allelic forms in each species, and that there is a SacI site unevenly bisecting the region cross-hybridizing to the probe in Ae. tauschii. In comparison, Ae. speltoides is an outcrosser, and as such two bands of relative equal intensity suggest that the DNA may be heterogeneous and comprising two allelic forms.

To estimate copy numbers of Cbf14, the same blots were subsequently hybridized with Pinb and the Cbf14/Pinb ratios were determined (Fig. S4B). The resulting Cbf14 copy number estimates ranged 3–11 in T. urartu (excluding G1812), 8–12 in T. monococcum, 3–7 in Ae. speltoides, and 5–20 in Ae. tauschii (Table S3). Average Cbf14 copy numbers were significantly lower in Ae. speltoides (BB) compared to the other diploid wheats (Fig. 3). No significant differences were detected in the average Cbf14 copy numbers between the T. urartu, T. monococcum, and Ae. tauschii accessions (Fig. 3). Additionally, no association was detected between growth habit and Cbf14 copy numbers in the T. monococcum accessions (Fig. S4A; Table S3).

Fig. 3
figure 3

Mean Cbf14/Pinb ratios in diploid wild (T. urartu (AA), Ae. speltoides (BB), and Ae. tauschii (DD) and cultivated (T. monococcum AmAm) wheats. Number of genotypes within each species (n) is listed on the X axis. Error bars represent SEM. Statistically significant differences are indicated by different letters above the error bars (P < 0.05)

Full size image

Overall, Cbf14 copy numbers in diploid wheats were lower in the B genome compared to the A and D genomes, a result which paralleled the findings of the hexaploid wheats.

Cbf14 CNV between the A and B genomes of tetraploid wheats

Cbf14 copy numbers were also assayed in a population of tetraploid (AABB), T. turgidum subsp. durum wheats resulting from a breeding program directed at developing winter durum wheats (Hall et al. 2011; Schilling et al. 2003). Among this group, the majority of the accessions produced three Cbf14 cross-hybridizing bands (Fig. S5), the MWs of which were similar to those of the A and B genomic fragments in hexaploid wheats. Six accessions did not produce B genome Cbf14 cross-hybridizing bands including Odessa#66, VA05WD-1, VA05WD-12, VA05WD-16, XVAD99067-24, and XVAD99147-1 (Fig. S5). Because Pinb is absent from the tetraploid wheats (Chantret et al. 2005; Gautier et al. 2000), copy numbers of Cbf14 were estimated in the A genome relative to the B genome. The Cbf14-A/Cbf14-B ratios were in the range of 4–26 (Table S4). These ratios were among the highest in XVAD99069-18 and VA05WD-39, and lowest in XVAD99068-14 (Table S4). These Cbf14-A/Cbf14-B ratios are at best only a rough estimate of the Cbf14 copy numbers in the A genome relative to the B genome. Nonetheless, the hybridization patterns together with the Cbf14-A/Cbf14-B ratios indicate lower copy numbers in the B genome compared to the A genome of tetraploid wheats.

Cbf14 copy number differences in the inter-varietal wheat chromosome substitution lines

Chromosome substitution lines have played an important role in the genetic analyses of freezing tolerance and winter hardiness in wheat and have been instrumental in the identification of Fr-2. We hypothesized that Cbf14 copy numbers may be different between the donor and recipient genomes of these substitution lines. To test this hypothesis, we examined three independent sets of disomic chromosome 5 substitution lines.

One set included the Kharkov-Rescue 5A (K-R5A) and Winalta-Rescue 5A (W-R5A) lines in which the 5A chromosome pair from the spring wheat ‘Rescue’ replaces the 5A of winter wheats ‘Kharkov MC22’ and ‘Winalta’, respectively (MacDonald 1987; Roberts and MacDonald 1988). Hybridization of the Cbf14 5′ probe to this set of lines indicated that the signal intensity of the A genome Cbf14 cross-hybridizing fragment was much greater in the two winter wheats ‘Kharkov MC22’ and ‘Winalta’ compared to that of the spring wheat ‘Rescue’ (Fig. 4a). Signal intensities of the 5A bands in the K-R5A and W-R5A substitution lines were comparable to that of ‘Rescue’ (Fig. 4a).

Fig. 4
figure 4

Cbf14 hybridization patterns in the inter-varietal chromosome 5 substitution lines of hexaploid wheat. a Substitution lines K-R5A and W-R5A have the 5A homoeolog of ‘Rescue’ in the ‘Kharkov MC22’ and ‘Winalta’ backgrounds, respectively, b CS-CNN substitution lines in which the ‘Cheyenne’ 5A, 5B, and 5D homoeologs replace the corresponding ‘Chinese Spring’ homeolog in the ‘Chinese Spring’ background, and c reciprocal 5A, 5B, and 5D substitutions between ‘Cheyenne’ and ‘Wichita’. DNA for each substitution line was isolated from seedlings produced from a single individual plant whose heads were bagged to prevent cross pollination. The MW of the A, B, and D, cross-hybridizing fragments are 2.3, 4.6, and 1.65 kb, respectively. Note: Signal intensity differences between the ‘Wichita’ accessions, WI-KS-1, WI-KS-2, and WI-NE-1 are attributed to DNA quantity loading differences. CS Chinese Spring, CNN Cheyenne, WI Wichita. The numbers and letters added to CS, CNN, and WI indicate different sources of accessions, which are identified in Table S1

Full size image

Two additional sets of examined substitution lines had the winter wheat ‘Cheyenne’ as either the donor or recipient of chromosome 5 homoeologs. One of these was the set in which the 5A, 5B, and 5D homoeologs of ‘Cheyenne’ (CNN) replace those of the spring wheat ‘Chinese Spring’ (CS), and are identified as CS-CNN-5A, CS-CNN-5B, and CS-CNN-5D, respectively (Morris et al. 1966). The other was the set in which the 5A, 5B, and 5D homoeologs of the two winter wheats ‘Cheyenne’ and ‘Wichita’ are reciprocally substituted in the background of the other (Zemetra and Morris 1988; Zemetra et al. 1986). In addition, parental lines ‘Chinese Spring’, ‘Cheyenne’, and ‘Wichita’ obtained from multiple sources were also examined (Table S1).

The Cbf14 hybridization patterns detected with the substitution lines having the ‘Cheyenne’ 5A, 5B, and 5D homoeologs in the ‘Chinese Spring’ background indicated that in each case there was greater signal intensity of the substituted chromosome from the ‘Cheyenne’ donor over that of the corresponding homoeolog of ‘Chinese Spring’ (Fig. 4b). Comparison of the CS-CNN substitution lines to a panel of ‘Cheyenne’ accessions obtained from different sources indicated that there was variability across the different ‘Cheyenne’ accessions (Fig. 4b). The most apparent difference occurred with ‘Cheyenne’ accession CNN-NE-1. Whereas the other five ‘Cheyenne’ accessions exhibited increased signal intensities of the A, B, and D genome cross-hybridizing fragments over ‘Chinese Spring’, CNN-NE-1 exhibited greater signal intensity only in the A genome cross-hybridizing fragment (Fig. 4b).

Analysis of the ‘Cheyenne’ and ‘Wichita’ parental lines, and the set of reciprocal substitution lines indicated that the 5B homoeolog from ‘Cheyenne’ had greater signal intensity of the Cbf14 cross-hybridizing band than that of ‘Wichita’ 5B (compare WI-CNN-5B and CNN-WI-5B in Fig. 4c). To independently assay differences in B genome copies, we also used a BglII digest, which produces a lower MW B genome fragment. Using the BglII digest, the increase in Cbf14 hybridization signal intensity of the ‘Cheyenne’ 5B homoeolog over the ‘Wichita’ 5B homoeolog was much more striking (Fig. S6). This observation was supported by Cbf14-B/Cbf14-D ratios of the ‘Cheyenne’ and ‘Wichita’ reciprocal substitution lines (Table S5).

Taken together these data indicate that Cbf14 copy numbers differ among genotypes used to develop substitution lines. These substitution lines were developed 30–50 years ago (MacDonald 1987; Morris et al. 1966; Zemetra et al. 1986). Despite having gone through numerous cycles of selfing since their development (PS Baenziger, UNL, personal communication), the similar signal intensities of the chromosome 5 homoeologs between the substitution lines and the parental donor lines indicate that Cbf14 copy numbers are stable over generations.

Discussion

The work presented here reveals that Cbf14 is copy number variable across wheat. Using DNA blot hybridization, Cbf14 CNV could be robustly scored with a single probe. The SacI digest in combination with the Cbf14 promoter probe produced a simple banding pattern across a wide array of diploid, tetraploid, and hexaploid wheat germplasm. The promoter fragment provided a gene-specific probe that avoided cross-hybridization to closely related Cbf gene family members, which occurs using coding sequence (CDS) probes. This strategy also avoided differential amplification of PCR products which will occur when primer sites differ in the different genomes, and the use of less than optimal primer sites for use in quantitative PCR to obtain genome-specific estimates of Cbf14 copy numbers.

In the diploid ancestors of hexaploid wheat, a higher Cbf14 copy number state exists, and in hexaploid wheat significantly greater Cbf14 copy numbers occur in the red winter wheats over the red spring wheats, and these differences occur on all three group 5 homoeologous chromosomes. These data imply that higher Cbf14 copy numbers pre-existed in the diploid ancestors, that reduction in Cbf14 copy numbers probably occurred following polyploidization, and that the ancestral state was retained in some populations and lost in others. Cbf14 copy number reductions in the hexaploid wheat D genome most likely occurred following amphiploidisation between T. turgidum and Ae. tauschii. In the case of the hexaploid wheat A and B genomes it is less clear. The B genome fragment was variable among the group of durums assayed, indicating reductions in the B genome may have already existed in tetraploid wheat prior to amphiploidisation with Ae. tauschii. As a group the hard red wheats also had greater B genome Cbf14 copy numbers than the soft red wheats; one scenario that can explain this difference is that these two winter wheat classes are derived from a common hexaploid ancestor and reductions in B genome Cbf14 copy numbers occurred in a subset of derivative populations. However, as evidence indicates two spatially and temporally separated amphiploidisation events occurred between T. turgidum and Ae. tauschii (Giles and Brown 2006), another possible scenario is that these two winter wheat classes are derived from separate amphiploidisation events between Ae. tauschii and different T. turgidum genomes that already differed in their B genome Cbf14 copy numbers.

Market classifications and growth habit categorizations are a traditional classification system of wheats grown in North America (Cox 1991; Zeven and van Hintum 1992), which is also used by the National Plant Germplasm System for classifications. Accessions that form the foundation of North American germplasm under these groupings were introduced as landraces and have known Old World origins. Many of the soft red winter wheats were introduced from Northern and Western Europe, the hard red wheats were introduced from the South-West region of the former Soviet Union, and the HRS wheats trace to the Red Fife accession from Poland (Cox 1991; Zeven and van Hintum 1992). In comparison, white wheats were introduced into North America from Old World and New World locations, which suggest that they were first taken by European immigrants to these locations prior to their North American introduction (Cox 1991; Zeven and van Hintum 1992). Most were highly genetically variable at the time they were introduced (Cox 1991; Zeven and van Hintum 1992). While knowing the site of origin of a landrace might provide clues to adaptability based on assumption of a landrace having undergone long-term selection, drawing definitive conclusions about the evolution of a landrace due to its geographic origin is complicated by other factors. For example as the hard red wheats were introduced from the South-West of the former USSR, a logical assumption might be that these accessions were adapted to that region. While this may be the case, loss of the original landrace and introduction of different landraces from outside regions have occurred through human migration and agricultural practices and thus this may blur our current view (Zeven 1986). Nonetheless introduction of the hard red wheats made possible the cultivation of winter wheat on the Great Plains where previous attempts with the other winter wheats failed due to their inability to survive the more arid and colder winters (Quisenberry and Reitz 1974; Zeven and van Hintum 1992). Whereas our data indicate that Cbf14 copy numbers of winter and spring white wheats of these early landrace types are more similar to each other and to the red spring wheats than to either class of the red winter wheats, through plant breeding hard white winter wheats are now grown on the Great Plains (Carver et al. 2003; Haley et al. 2003; Ibrahim et al. 2008; Martin et al. 2001; Pike and MacRitchie 2004). We did not assay these modern cultivars but it would be worth knowing whether they possess introgressions of the higher copy number state alleles on all three of the group 5 homoeologs.

It is also noteworthy that the B genome homoeolog has the fewest Cbf14 copy numbers relative to the A and D genome homoeologs. This trend occurs in hexaploid, tetraploid, and diploid wheats. This is again consistent with the higher Cbf14 copy number states pre-existing in the diploid ancestors. That the increased copy numbers occur in all of the diploid genomes suggests the amplifications that generated the multiple copies probably occurred in a common ancestor that predates the divergence of the different diploid wheat species. This may have occurred simultaneously with the projected expansion of the Cbf gene family in the core Pooids (Sandve and Fjellheim 2010). At a subsequent point in time, and prior to the amphiploidisation between T. urartu and the Ae. speltoides relative, reductions in Cbf14 copy numbers occurred in the genome of the B genome diploid wheats. Further reductions in Cbf14 copy numbers then also continued in the B genome following this amphiploidisation.

In a number of accessions, Cbf14 was completely absent from a genome. There was no B genome Cbf14 5′ cross-hybridizing fragment in four of the spring growth habit hexaploid wheats including ‘Sonora’, ‘Mandel Gehun’, ‘PI 70711’, and ‘PI 40941’. There was also no B genome fragment in six of the tetraploid T. turgidum subsp. durum wheat accessions including Odessa#66, four lines derived from Odessa#66 (VA05WD-1, VA05WD-12, VA05WD-16, and XVAD99067-24), and XVAD99147-1. The A genome diploid G1812 also did not cross-hybridize to the Cbf14 5′ probe. Absence of a Cbf14 5′ cross-hybridizing fragment in these accessions may be due to a different Cbf14 promoter sequence or the complete deletion of Cbf14 from their genomes. We also screened the G1812 BAC library (Akhunov et al. 2005) using a Cbf14 CDS probe and did not recover clones (TD and EJS, unpublished data), consistent with the absence of Cbf14 from this genotype.

Differences across the multiple ‘Cheyenne’ accessions in cross-hybridizing signal intensities of 5B and 5D homoeologs indicate that some of these accessions are different. ‘Cheyenne’ was released as a pure-line selection from a Crimean landrace (Clark 1931). It is possible that ‘Cheyenne’ was originally a heterogeneous population from which subsequent selection and single seed descent resulted in different homogeneous individuals. Pure-line selection as practiced in the early part of the 20th century often involved selecting individuals for similar phenotypes from a phenotypically polymorphic landrace, and did not necessarily involve selection from one single homozygous individual. Most likely a ‘Cheyenne’ accession other than CNN-NE-1 was used to develop the CS-CNN substitution lines.

While Cbf14 gene CNV occurs on all three homoeologs, the A and D homoeologs exhibit the largest differences across wheat accessions. It is possible that the larger differences in Cbf14 copy numbers of the 5A and 5D homoeologs are an underlying genetic factor leading to the more frequent identification of 5A and 5D as affecting differences in freezing tolerances (Cahalan and Law 1979; Roberts 1986, 1990; Roberts and MacDonald 1988). Nonetheless 5B is also associated with differences in freezing tolerance, and substitution of the ‘Cheyenne’ 5B into ‘Chinese Spring’ and ‘Wichita’ results in improved freezing tolerance (Tóth et al. 2003; Veisz and Sutka 1989). It must be stressed that Cbf14 is only one gene in the cluster of at least 11 different Cbf gene CDSs at Fr-2. Data obtained using CDSs as probes against a panel of nine hexaploid wheat lines suggest that Cbf2 and Cbf9 are also copy number variable and that the trend appears to be similar to that of Cbf14 (Fig. S7). Thus, multiple Fr-2 Cbf gene orthologs may be copy number variable, their numbers may act additively, and together they contribute in a quantitative manner to the pathways that they affect regulatory control over.

While contributions of the Cbf genes may act in a quantitative manner, mutations in single Cbf genes at the Fr-2 cluster are also associated with differences in freezing tolerance. In T. monococcum, genetic analyses identified Cbf12, Cbf14, and Cbf15 as candidate genes explaining a large portion of the differences in freezing tolerance between DV92 and G3116 (Knox et al. 2008). Cbf14 and Cbf15 sequences of DV92 do not notably differ from those of G3116 (Knox et al. 2008). In the instance of Cbf12, there is a 15 bp in-frame deletion in the DNA-binding domain of the DV92 protein that eliminates binding to its cognate binding site whereas the G3116 protein is able to bind (Knox et al. 2008). A smaller QTL effect also resolves just distal to the Cbf12–Cbf14–Cbf15 cluster, and includes Cbf16, Cbf13, Cbf3, and Cbf10 (Knox et al. 2008). Because transcripts for Cbf12, Cbf15, and Cbf16 accumulate to higher levels at warmer temperatures in G3116 than in DV92, this differential regulation affect could also not be ruled out as contributing toward differences in freezing tolerance (Knox et al. 2008). It is nonetheless possible that mutations in a single Cbf12 gene or in other members of this clade (e.g., Cbf15 or Cbf16) have more pronounced phenotypic consequences than copy number differences in Cbf14 or in other members of this clade (e.g., Cbf2, Cbf4, or Cbf9) as there are a number of differences that discriminate the CBF proteins in these two clades (Skinner et al. 2005; Xue 2003).

DV92 and G3116 both appear to have 10 copies of Cbf14 based on Cbf14 to Pinb ratios of 9.5 and 10.5, respectively—data consistent with Cbf14 not being a contributing factor in freezing tolerance differences between these two genotypes (Knox et al. 2008). Although the specific activity of the probes may affect hybridization signal intensities, and thus ratios, both probes were about the same length, 189 bp (Cbf14) and 201 bp (Pinb). And while ten copies might seem high, it is not unprecedented in plants. At the rhg1 locus, which confers resistance and susceptibility to soybean cyst nematode, resistance is conferred by an allele having ten copies of a 31 kb region encompassing six genes, three of which contribute toward resistance, whereas susceptibility occurs by the allele having a single copy of this 31 kb region (Cook et al. 2012). At Fr-2, it is presently unclear how the physical arrangement of multiple Cbf14 genes occurs. The Cbf14 sequence of T. monococcum DV92 is derived from a 3.3 kb subclone from BAC clone 119P22 (Miller et al. 2006). Additional identical Cbf14 paralogs may exist on that same BAC clone, each of which would run at the same MW position on an agarose gel. This is the situation in barley in which independent bacteriophage lambda genomic clones harboring Cbf2A and Cbf4B recovered from screening non-amplified primary libraries produced identical fingerprints with the restriction enzymes used for fingerprinting (Knox et al. 2010). While the Cbf2A and Cbf4B CDSs on independent phage clones were identical to the Cbf2A and Cbf4B CDSs on different phage clones, sequencing those clones revealed polymorphisms in the intergenic regions making it technically feasible to discern that the identical CDSs are embedded in different genomic regions (Knox et al. 2010). In the instance of the wheat Cbf14 gene, another possibility is that there is only one Cbf14 CDS on the 119P22 BAC clone but there are additional Cbf14 gene paralogs residing in similar, but distinct and genetically linked genomic regions. Figure S1 shows an alignment of the genomic regions encompassing Cbf14 from T. monococcum accessions DV92 and G3116, and from two BAC clones, JF758494.1 and JF758498.1, from T. aestivum cv. ‘Norstar’. These two Norstar BAC clones are derived from the A genome yet they do not encompass identical regions (Fig. S1). These data are consistent with multiple Cbf14 genes in the A genome, and a single Cbf14 per BAC insert.

Multiple copies of Cbf14 across the different wild diploid ancestors and the reductions in copy numbers in the different cultivated tetraploid and hexaploid wheats suggest both functional importance to the multiple copy numbers in these wild ancestors and that copy number reductions likely occurred as a result of changes in selection pressures, possibly in the form of cultivation practices. Diploid wheats do not exhibit greater freezing tolerance than hexaploid wheats. On the contrary, hexaploid wheats exhibit greater freezing tolerance than the diploid and tetraploid wheats (Fowler et al. 1977; Limin and Fowler 1981, 1982). However, the winter growth habit is the ancestral form in the Triticeae (Flood and Halloran 1986; Matus and Hayes 2002; von Bothmer et al. 2003; Yan et al. 2003; Zohary 1969), and thus developmental programming controlling this habit is likely critical for these cereals in their natural habitat. In barley, cultivated genotypes carrying mutations in Vrn-1 that result in the spring growth habit exhibit reduced copy numbers of the Cbf2A–Cbf4B genomic region relative to cultivated winter barleys (Knox et al. 2010). These associations between reduced copy numbers and the spring growth habit in cultivated forms suggest reductions occurred in response to selection for the spring growth habit and possibly earlier flowering, or both. While this relationship did not hold in the T. monococcum accessions examined here, three of the four spring accessions (PI 538722, PI 427927, and DV92) retain a winter vrn-1 allele at Vrn-1 and are of spring growth habit due to mutations in Vrn-2 (Yan et al. 2004). (The fourth accession (PI 355549) was not genotyped for Vrn-2 in that work and we do not know whether it too carries the same mutation.) With a winter vrn-1 allele, VRN-1 transcripts do not accumulate until cued by endogenous and exogenous signals (Distelfeld et al. 2009; Trevaskis 2010). In the absence of VRN-1 accumulation when plants are in the vegetative phase, CBF mRNAs accumulate to much higher levels than when VRN-1 transcripts accumulate and plants are in the reproductive phase (Dhillon et al. 2010; Stockinger et al. 2007). It is not clear what mechanisms control this process, but this indicates that the plant has a mechanism in place to reduce CBF transcript levels during the reproductive transition. Selection for the spring growth habit via mutations in the Vrn-1 gene which results in the constitutive expression of VRN-1 may have rendered higher Cbf copy numbers functionless, and thus they were lost. This hypothesis could be more rigorously tested through determination of Vrn-1 allelic states, Cbf14 copy numbers, and growth habit form in a more expansive survey of T. turgidum that includes the earliest cultivated T. turgidum ssp. dicoccon alongside wild T. turgidum ssp. dicoccoides exhibiting known kinship relationships to the cultivated forms (Luo et al. 2007). However, CBFs also delay flowering (Gilmour et al. 1998; Liu et al. 1998), and as such selection for earlier flowering should not be excluded as a factor actively driving reductions in Cbf gene copy number. In essence, these relationships indicate that the primary functional role of the Cbfs probably occurs during the vegetative growth phase and not during the reproductive phase. Thus, copy numbers may be acting additively and the high copy number states in the wild diploid species may be one integral component in a mechanism to maintain and possibly even promote developmental processes associated with vegetative growth for these plants during the winter and spring seasons in their natural habitat. Retention of the ancestral higher copy number state in landrace populations giving rise to the HRW wheats may be due to constancy in the selection pressures for the winter growth habit and higher yield through a more extended vegetative growth phase. A plant that is actively maintaining the vegetative growth phase is also likely to be manifest in its ability to regrow shoots and roots from a meristematic region following winter kill. Whereas the diploid wheats have a single cadre of Cbf genes, hexaploid wheats have three cadres, which may simply act additively in contributing toward this phenotype.