Introduction

Alleles at the Reduced height (Rht)-1 and Photoperiod (Ppd)-D1 loci have a major influence on the adaptability of wheat to climate and modern agricultural practices. Rht-B1 and Rht-D1 are located on chromosomes 4B and 4D, respectively, in bread wheat (Triticum aestivum) and encode copies of the DELLA protein, a growth repressor that is normally degraded in the presence of gibberellin (GA). Rht-B1b and Rht-D1b each contain a single nucleotide polymorphism that introduces a premature stop codon (Peng et al. 1999). The resulting DELLA proteins have a reduced sensitivity to GA relative to DELLA proteins encoded by the wild type Rht-B1a and Rht-D1a alleles, thereby reducing height (Peng et al. 1999). Individually, Rht-B1b and Rht-D1b are associated with height reductions of approximately 14–17 % relative to the wild type (Flintham et al. 1997). The semi-dwarf varieties have decreased lodging and improved harvest index under intensive agriculture, and Rht-B1b and Rht-D1b were thereby important components of the ‘green revolution’ (Hedden 2003). Ppd-D1 is located on chromosome 2D in bread wheat and is a member of the pseudo-response regulator (PRR) gene family (Beales et al. 2007). In Arabidopsis thaliana, the PRR gene family members are components of the circadian clock (Mizuno and Nakamichi 2005; Matsushika et al. 2007). The wheat Ppd-D1a allele contains a 2,089 bp deletion in the promoter region, which is associated with photoperiod insensitivity (Beales et al. 2007). Under field conditions in the United Kingdom, ear emergence of Ppd-D1a varieties was accelerated by 6–8 days when October-sown, relative to varieties containing the photoperiod sensitive allele Ppd-D1b (Worland et al. 1988). Early flowering associated with Ppd-D1a is beneficial in areas such as Southern Europe, allowing plants to mature prior to summer heat stress (Worland et al. 1998a).

The Rht-1b and Ppd-D1a alleles have a pronounced effect on wheat adaptation, but the ability to further modify plant height and flowering time should aid in adapting varieties to specific environments. Several Ppd-D1 haplotypes, including Ppd-D1a, were recently assessed and the associated effects on flowering date and several agronomic traits suggest several haplotypes may be useful in fine-tuning wheat to specific environmental conditions (Guo et al. 2010). Alterations in plant height should also prove useful for improving wheat yield in specific genetic backgrounds, environments, or farming systems (Law et al. 1978; Flintham et al. 1997; Addisu et al. 2010). In addition to Rht-B1b and Rht-D1b, the GA insensitive (GAI) alleles Rht-B1c, Rht-B1d, Rht-B1e, Rht-D1c, and Rht-D1d also have an affect on plant height (Borner et al. 1996). The Rht-1 open reading frames of these alleles have been sequenced and polymorphisms have been identified that extensively or partially explain the associated changes in plant height (Pearce et al. 2011; Wu et al. 2011; Li et al. 2012a, b). The greatest reductions in height and GA sensitivity are associated with Rht-B1c and Rht-D1c (Izumi et al. 1981; Gale and Youssefian 1985; Borner and Mettin 1988). Rht-B1e and Rht-D1d are associated with heights slightly shorter than Rht-B1b or Rht-D1b (Worland and Sayers 1995; Borner et al. 1996) and Rht-B1d is associated with heights slightly taller than Rht-B1b or Rht-D1b (Worland and Petrovic 1988). Rht-B1b and Rht-D1b are estimated to be in approximately 90 % of the world’s semi-dwarf varieties (Worland et al. 1998b); hence the use of other GAI alleles is restricted, although there may be regional use, such as Rht-B1d in Italy and southern Europe (Worland and Petrovic 1988). Two additional Rht-B1 alleles that potentially affect plant height and GA sensitivity were recently identified by Wilhelm et al. (2013b). Both alleles contain a large 5′ insertion (160 or 197 bp) within 600 bp of the Rht-B1 start codon and, therefore, may disrupt the promoter region and affect DELLA expression. The 160 bp insertion is of particular interest because it occurs in the middle of a 120 bp stretch of sequence that is highly conserved in all three wheat Rht-1 homoeologs and in Rht-1 Poaceae orthologs (Duan et al. 2012; Wilhelm et al. 2013a).

In this study, the frequencies of Rht-B1a, Rht-B1b, the Rht-B1 160 and 197 bp insertions, Rht-D1a, Rht-D1b, Ppd-D1a, and Ppd-D1b were determined in a worldwide core collection of 372 wheat varieties (372CC; Balfourier et al. 2007) using PCR markers. Plant height, GA sensitivity, and days to heading (relative to 1 January) were recorded for each accession and correlated with the Rht-1 and Ppd-D1 genotype scores and the effect of each allele on the selected plant traits is estimated. In addition, two statistical methods for classifying GA response were validated in relation to Rht-1 genotype (Rht-1a versus Rht-1b) of the 372CC accessions and then used to search for novel GA insensitive alleles in the collection.

Materials and methods

Outdoor experiment design

Plant height and flowering date of the 368 bread wheat accessions present in the 372CC were examined in an outdoor experiment at the National Institute of Agricultural Botany (NIAB), Cambridge, UK in which plants were grown in 4-l pots in outdoor beds. The 372CC is representative of 98 % of the genetic diversity (based on 38 simple sequence repeats) of 3,942 accessions originating from 73 nations and includes landraces and genetically fixed lines with registration years spanning 1830–1998 (Balfourier et al. 2007). Accessions were sown in one of two sowing dates: 25 November 2008 or 18 February 2009. Each pot (experimental unit) contained four plants of an accession along with coarse compost mixed with approximately 10 g of 11-11-18 (N %-P %-K %) controlled-release fertilizer.

The November sowing comprised 249 372CC accessions (primarily winter habit types) along with 16 control lines (Online Resource 1). Controls were winter or facultative habit types. Ppd-D1 controls included Ppd-D1a and Ppd-D1b lines. Among the Rht-1 controls were Rht-B1a, Rht-B1b, Rht-B1c, Rht-B1d, Rht-B1e, Rht-D1a, Rht-D1b, and Rht-D1c near isogenic lines (NILs) in the ‘Cappelle Desprez’ and/or ‘Mercia’ background. Seeds were germinated in compost in 96-well trays in the glasshouse under natural lighting (21 °C for 16 h; 17 °C for 8 h). Following emergence, seedlings were exposed to ambient air temperatures and natural lighting in the glasshouse for 4 weeks before being transplanted to 4-l pots and placed in outdoor beds. The experimental design was an incomplete block composed of two main blocks each consisting of three beds. A pot of each control was placed in each outdoor bed (six replicates per control) and the 372CC accessions were represented once in each main block (two replicates per accession). Experimental units were randomised to achieve maximum efficiency using: http://www.niab.com/dew/.

The February sowing consisted of 119 accessions (primarily spring habit types) along with 13 controls (Online Resource 1). Ppd-D1 controls included accessions containing Ppd-D1a or Ppd-D1b. Rht-1 controls included Rht-B1a, Rht-B1b, Rht-B1c, Rht-D1a, and Rht-D1b NILs in the genetic backgrounds of the spring lines ‘April Bearded’ and ‘Bersee’. Seeds were sown directly into pots and immediately positioned in the outdoor beds. The February sowing contained two main blocks, each consisting of 1.5 outdoor beds (one outdoor bed was split between two main blocks). A pot of each control was present in each outdoor bed (three replicates per control) and the 372CC accessions were represented once in each main block (two replicates per accession). Experimental units were randomized using the Microsoft Excel random number function.

In both sowings, outdoor beds contained peat and were irrigated as needed. Roots were allowed to grow into the outdoor beds through openings in the bottoms of pots. Each outdoor bed contained 161 4-l pots (7 pots wide × 23 pots long), which included 105 experimental units and 56 pots of cv. ‘Xi19’ placed along the perimeter to reduce edge effects. Twine was strung across outdoor beds to keep plants upright. To control powdery mildew, fungicides were applied that did not contain known growth regulators. Plants were sprayed to control aphids and covered with netting just prior to harvest to prevent bird damage. Mature plant height was measured as the distance from the soil surface to the tip of the terminal grain of the head (excluding awns) of the longest tiller present in each pot. Days to heading was recorded as the number of days from 1 January until the inflorescence of the primary tiller from each pot was 50 % emerged from the flag leaf.

DNA extraction and genotyping

Leaf tissue was collected from 2- to 3-week-old plants from the outdoor experiment and DNA extracted using a modification of the method described by Fulton et al. (1995). To minimize genotyping errors, genomic DNA (gDNA) was extracted from two replicates of four bulked plants of genetically fixed lines and from four individual plants of landraces, and assayed in separate PCR reactions. PCR assays for Rht-B1a, Rht-B1b, Rht-D1a, Rht-D1b, and the presence of the Rht-B1 160 bp and 197 bp insertions were performed as described by Wilhelm et al. (2013b). The PCR assay for detecting Ppd-D1a was performed as described in Beales et al. (2007).

GA sensitivity experiments

The three experiments to examine GA responsiveness of the 372CC accessions were conducted in the glasshouse at the John Innes Centre (JIC), Norwich, UK. GA experiment 1 (GA expt 1) contained all 368 bread wheat accessions from the 372CC. GA expt 2 contained 161 bread wheat accessions. GA expt 3 contained 19 bread wheat accessions. The successive experiments were required to resolve indeterminate classifications of GA responsiveness from the preceding experiment. Each GA sensitivity experiment contained the following NIL controls: ‘April Bearded’, ‘April Bearded Rht-D1b’, ‘April Bearded Rht-B1c’, ‘Bersee’, ‘Bersee Rht-B1b’, ‘Mercia’, and ‘Mercia Rht-D1b’.

GA sensitivity tests were based on those described by Gale and Gregory (1977), except a GA− treatment was included along with the GA+ treatment. The GA+ treatment contained 10 ppm GA3 (Sigma product number G-7645) dissolved in 0.1 % ethanol. The GA− treatment consisted only of 0.1 % ethanol. For each experiment, an incomplete block design was used with GA treatments randomly assigned to whole trays within main blocks. Trays consisted of 60 wells filled with a peat/sand mixture moistened with a GA+ or GA− solution and each well contained one seedling (the experimental unit). Seedlings were grown from seeds placed on moistened filter paper and stratified at 4 °C for a minimum of 2 days followed by 3 days at room temperature prior to transplanting. Seedlings were randomly assigned a tray and well located within each GA treatment using the Microsoft Excel random number function.

GA expt 1 consisted of three main blocks with seven GA+ trays and seven GA− trays per main block (42 trays total). GA expt 2 consisted of three main blocks with three GA+ trays and three GA− trays per main block (18 trays total). GA expts 1 and 2, therefore, each contained three GA+ and three GA− plants per 372CC entry. GA expt 3 consisted of six main blocks per treatment and a total of two GA+ trays and two GA− trays. Each 372CC entry in GA expt 3 was represented once in each main block (six GA+ and six GA− plants per accession) with the following exceptions that had low germination rates: INRA-03752, two plants per treatment; INRA-00347 and INRA-01065, four plants per treatment. For all three experiments, controls were represented once per tray. Trays were placed in the glasshouse under artificial lighting (300 to 400 μmol m−2 s−1) with day/night temperatures of 20 °C/15 °C, 16 h day length. Treatments were applied to trays by gently watering with a GA+ or GA− solution as plants needed moisture. Approximately 3 weeks after each experiment began, plants were uprooted and the length from the seed to first ligule (STFL) was recorded.

Statistical analyses

To classify the GA sensitivity of accessions, a probability distribution mixture (PDM) model and a least significant difference (LSD) test were used. In the PDM model, a mixture model consisting of two independent normal distributions with different means and variances (George et al. 2000) was fitted with a probability (α) of an observation belonging to one distribution and a probability (1−α) belonging to a second distribution. An observation consisted of the mean STFL length of an accession in the GA+ treatment in an experiment. All five parameters (μ 1, σ 21 , μ 2, σ 22 , and α) were fitted by maximum likelihood in Microsoft Excel using the Solver function. Posterior probabilities of any individual observation belonging to either of the two distributions were calculated from the probability density functions of each using the estimated means and variances as (p1α)(p1α + p2(1−α))−1 and (p2α)(p1α + p2(1−α))−1. A threshold posterior probability greater than or equal to 0.95 was used to assign individual observations to one or the other of the following GA classes: GAI or GA sensitive (GAS). Individuals with posterior probabilities <0.95 for either of the two distributions were classified as having an intermediate (Int) GA response. GA sensitivity classifications based on the PDM model are reported for the GA+ treatment of GA expts 1 and 2, but not for the GA+ treatment of expt 3 or the GA treatment differences (GA trt diff; the STFL length of the GA+ treatment minus the GA− treatment of an accession) of any GA expt because it was not possible to separate these distributions. In the LSD test, the GA trt diff for each accession was tested for statistical significance by t test in GA expts 1, 2, and 3. Positive differences significant at the 5 % level were classified as GAS. Variances for the GA+ and GA− treatments were treated as unequal and estimated separately for each experiment, but pooled over all accessions. Variances and standard errors were calculated with Genstat, 12th edition (VSN International). T value probabilities were calculated with the Microsoft Excel TDIST function (two-tailed).

To estimate the effects of Rht-1 and Ppd-D1 on the measured plant traits, three sets of restricted maximum likelihood mixed model (REMLMM-1; -2; -3) analyses were conducted using a subset of 352 accessions from the 372CC. This subset excluded accessions containing heterozygotes or rare Ppd-D1 or Rht-1 alleles. In the REMLMM-1 analyses for plant height and days to heading, fixed effects were growth habit, Rht-B1, Rht-D1, Ppd-D1, and all 2-way and 3-way interaction terms (Online Resource 2a-b). For REMLMM-1 GA trt diff, fixed effects were Rht-B1, Rht-D1, Ppd-D1, and all interactions (Online Resource 2c). Fixed effects were declared significant at an F probability threshold of 0.05. REMLMM-2 analyses were conducted to estimate mean allelic effects using loci that had significant main effects in the REMLMM-1 analyses in each model. Fixed effects in the REMLMM-2 analyses were as follows: Days to heading = Rht-B1, Ppd-D1, and Rht-B1*Ppd-D1; Plant height = Rht-B1, Rht-D1, Ppd-D1, and all 2-way interactions; GA trt diff = Rht-B1, Rht-D1, and Rht-B1*Rht-D1. REMLMM-3 analyses were performed to estimate the effect of accession type or time period of varietal registration on each plant trait. Year of registration was available for only 192 of the accessions. Fixed effects in the REMLMM-3 analyses (which included significant main effects from the REMLMM-1 analyses) were as follows: Days to heading = growth habit, Ppd-D1, Rht-B1, all 2-way interactions, and accession type or registration period; Height = Rht-B1, Rht-D1, Ppd-D1, all 2-way interactions, and accession type or registration period; GA trt diff = Rht-B1, Rht-D1, Rht-B1*Rht-D1, and accession type or registration period. Variety served as the random effect in all REMLMM analyses. Fitted and predicted means were compared using an LSD test with a threshold probability of 0.05. REMLMM analyses and comparison of means were performed using Genstat, 12th edition.

Pearson’s correlation coefficient (r) was determined using the PEARSON function in Microsoft Excel.

Results

Rht-1 and Ppd-D1 allelic frequencies

Alleles present at the Rht-B1, Rht-D1, and Ppd-D1 loci were determined for each of the 368 bread wheat accessions from the 372CC and in a set of control lines (Online Resource 1). Eight accessions were not homozygous at all three loci and are not included in the summary (Table 1). Among the remaining 360 accessions, the Rht-B1 160 bp insertion was present in 58 (16.1 %) of the accessions and the Rht-B1 197 bp insertion was present in 43 (11.9 %) of the accessions. The Rht-B1 insertion data were then correlated with the Rht-B1a/b data and with a single exception, the Rht-B1 insertions occurred in accessions containing the Rht-B1a allele. Herein, Rht-B1a-0, Rht-B1a-160 and Rht-B1a-197 will be used to refer to Rht-B1a genotypes with no insertion, a 160 bp insertion, and a 197 bp insertion, respectively. The single exception was INRA-06986 (cv. ‘Tom Thumb’), for which the product associated with the 197 bp insertion was amplified, but no product was amplified with the Rht-B1a or Rht-B1b primer pairs even under polymerase extension times of 3 min. ‘Tom Thumb’ carries the dwarf allele Rht-B1c, which contains an insertion of up to 2,026 bp in the region amplified by the Rht-B1a and Rht-B1b primer pairs (Wu et al. 2011), which likely explains the absence of product in this line and in the four control NILs containing Rht-B1c. Rht–B1a-0 was the most frequent Rht-B1 allele occurring in 213 (59.2 %) of the accessions. Rht-B1b was present in 45 (12.5 %) of the accessions. Geographically, Rht-B1a-160 is most prevalent in accessions originating in Europe, central Russia, and North America, but is also present in low frequency among accessions from Asia and South America (Online Resource 3). The majority (58 %) of the Rht-B1a-197 accessions have a French origin, but this allele can also be found in accessions from other European nations, the USA, Australia, Asia, Africa, and South America. Rht–D1a and Rht-D1b were present in 315 (87.5 %) and 45 (12.5 %) of the homozygous 372CC accessions, respectively.

Table 1 Summary of 372CC genotypes
Full size table

Four different alleles were discovered with the Ppd-D1a PCR assay. The Ppd-D1b allele was present in 263 (73.1 %) of the accessions and the insensitive Ppd-D1a allele was present in 90 (25.0 %) of the accessions. Accessions containing Ppd-D1a were present in most of the geographical regions represented in the collection, although no accessions originating in Northern European nations contained this allele. A third Ppd-D1 allele was amplified using gDNA of INRA-13812 (D genome derived from Aegilops tauschii) in the Ppd-D1a PCR assay. The amplified product was approximately 450 bp in length, which is larger than the Ppd-D1b (414 bp) or Ppd-D1a (297 bp) product. For six of the accessions, no product was amplified when multiple biological samples served as template in the Ppd-D1a assay, indicating the presence of a fourth Ppd-D1 allele. The accessions consisted of three landraces from nations near the Black Sea (Azerbaijan, Georgia, and Turkey) and three genetically fixed lines from Russia, Ethiopia, and France.

Plant traits

Among the 372CC bread wheat accessions, days to heading ranged from 116 to 185 days with a mean of 141 days (Online Resource 4a). Plant heights ranged from 54 to 200.5 cm with a mean of 133 cm (Online Resource 4b). For the three GA expts combined, GA trt diff of individual accessions ranged from −14.7 to 54.3 mm with a mean of 23.7 mm (Online Resource 4c). The highest correlation coefficient (r) among traits was between plant height and GA trt diff (r = 0.58). The days to heading and GA trt diff r value was 0.10 and the days to heading and plant height r value was 0.18.

Rht-1 and Ppd-D1 allelic effects on plant traits

For days to heading, Ppd-D1 accounted for the largest proportion of variation in the REMLMM-1 analysis (Online Resource 2a). The main effects of Rht-B1 and growth habit were also statistically significant whereas Rht-D1 and all interactions were not significant. In the REMLMM-2 analysis, Ppd-D1a accessions had fitted means (FMs) that were 7.6 days earlier than Ppd-D1b accessions, a statistically significant difference at p ≤ 0.05 (Table 2). The days to heading FM of Rht-B1a-0 was also significantly less than that of Rht-B1a-160. The six accessions that did not amplify a PCR product with the Ppd–D1a primers had a mean days to heading of 142.2 days. INRA-13812, which amplified a 450 bp product with the Ppd-D1a primers, had a days to heading of 142 days.

Table 2 Mean effects of Rht-1 and Ppd-D1 alleles on 372CC plant traits
Full size table

In the plant height REMLMM-1 analysis, the main effects of Rht-B1 and Rht-D1 were the greatest and statistically significant (Online Resource 2b). Ppd-D1 had the next largest effect and was statistically significant, along with six interactions that had smaller effects. In the REMLMM-2 analysis, Rht-B1b was associated with statistically significant FM height reductions of 14.8–37.0 cm relative to the other Rht-B1 alleles (Table 2). Relative to the Rht-B1a-0 allele, Rht-B1a-160 and Rht-B1a-197 both had statistically significant reductions in height of 22.2 and 15.1 cm, respectively, when comparing FMs. Rht-D1b accessions had a significant reduction in FM height of 36.7 cm relative to Rht-D1a accessions and Ppd-D1a accessions had a significant reduction in FM height of 17.2 cm relative to Ppd-D1b accessions.

For GA trt diff, the main effects of Rht-B1 and Rht-D1 were by far the greatest and the effect of Rht-B1*Rht-D1 was also significant in the REMLMM-1 analysis (Online Resource 2c). In the REMLMM-2 analysis of GA trt diff, the effect of Rht-B1b was statistically significant relative to all three Rht-B1a alleles (9.4–12.3 mm FM reductions) and the FM of Rht-D1b accessions was significantly reduced (20.7 mm) relative to Rht-D1a accessions (Table 2). The GA trt diff FMs of Rht-B1a-160 and Rht-B1a-197 accessions were slightly reduced (2.9 and 1.0 mm, respectively) relative to the FM of Rht-B1a-0 accessions, but were not statistically significant.

In the REMLMM-3 analysis, the main effects of varietal type and registration period were statistically significant for plant height, but not for days to heading or GA trt diff (Table 3). The predicted mean for plant height in genetically fixed lines was reduced by 14.1 cm relative to landraces (Table 3a). Accessions registered in the three 20-year periods that span 1940–1999 all had significant reductions in height relative to the three time periods that span the years 1800–1939, with differences in predicted means ranging from to 17.3–27.5 cm among registration periods (Table 3b).

Table 3 Predicted mean effects of accession type (a) and registration period (b) on 372CC plant traits
Full size table

GA sensitivity test validation and mutant screen

GA response classifications of the 372CC accessions based on the PDM model and LSD test were validated relative to the Rht-B1a/b and Rht-D1a/b genotype scores (Table 4) and then used to search for GAI mutants not containing Rht-B1b or Rht-D1b. The nine accessions with an Rht-B1b + Rht-D1b genotype were all classified as GAI using the PDM model and as GAI/Int using the LSD test. The lone 372CC accession containing the Rht-B1c allele (INRA-06986; cv. ‘Tom Thumb’) and the ‘April Bearded Rht-B1c’ control were also both classified as GAI and GAI/Int using the PDM model and LSD test, respectively (Online Resource 1). Of the 74 accessions containing a single semi-dwarf allele, the PDM model classified 65 as GAI, nine as Int, and none as GAS. Use of the LSD test also resulted in the classification of the majority of these accessions as GAI/Int; however, five accessions were classified as GAS (Table 4). INRA-01647 was the only 372CC accession containing a known semi-dwarfing allele that was not classified as GAI by the PDM model or the LSD test. The control ‘Bersee Rht-B1b’ was also not classified as GAI by the LSD test or PDM model while the remaining GAI controls were correctly classified by both methods. Of the 281 Rht-B1a + Rht-D1a accessions, the PDM model identified one accession (INRA-03752) as GAI, 14 accessions as Int and the remainder as GAS. The LSD test identified one Rht-B1a + Rht-D1a accession (INRA-04670) as GAI/Int and the remainder as GAS. The Rht-B1a + Rht-D1a controls were consistently identified as GAS in each experiment using either model.

Table 4 GA sensitivity classifications of the INRA 372CC accessions by Rht genotype
Full size table

Discussion

Each of the five alleles (Rht-B1a-160, Rht-B1a-197, Rht-B1b, Rht-D1b, and Ppd-D1a) was present in the 372CC at a frequency of 0.119 or higher. The appreciable percentage of lines containing Rht-B1a-160 (16.1 %) or Rht-B1a-197 (11.9 %), the wide geographic distribution of these alleles, and their presence in recently released lines indicate these alleles are not geographically isolated or rare in modern bread wheat varieties. The 160 bp insertion was present in a 372CC accession released in 1830 and in four 372CC landraces. However, the insertion was not identified in the bread wheat ancestral lines T. dicoccoides or T. dicoccum (Wilhelm et al. 2013b), suggesting the insertion is a recent event. The 197 bp insertion was not identified in a 372CC landrace; however, the 197 bases were previously identified in T. dicoccoides and T. dicoccum and in collinear regions of the bread wheat A and D genomes, indicating the presence of the insertion is the ancestral condition (Wilhelm et al. 2013b). Rht-B1b and Rht-D1b were each present in 12.5 % of the accessions. In a separate study that examined prevalence of these alleles in 67 wheat varieties from 20 countries, Rht-B1b occurred in 58 % of the accessions and Rht-D1b in 22 % of the accessions (Tosovic-Maric et al. 2008). The relatively low occurrence of Rht-B1b and Rht-D1b in the 372CC is likely due to the inclusion of a large number of landraces and accessions that pre-date the widespread use of the Rht-1 semi-dwarfing alleles, which began in the 1960s. The Rht-B1a/b and Rht-D1a/b assays are not capable of identifying alternative GAI alleles (Pestsova et al. 2008); hence, these alleles, if present, would not have been identified. However, alternative GAI alleles are generally rare (Worland et al. 1998b) and, therefore, are most likely rare in this collection.

The photoperiod insensitive Ppd-D1a allele was present in 25.0 % of the accessions and was found in geographical regions worldwide except for Northern Europe. In a collection of more than 400 European wheat varieties released between 1940 and 2000, the Ppd-D1a allele was also rare among Northern European accessions (E. Wilhelm and D. Laurie, JIC, unpublished results). Ppd-D1a is most suited to growing seasons that are shorter than those found in Northern Europe (Worland et al. 1998a), which explains its absence in this region. The approximately 450 bp Ppd-D1 allele present in the synthetic line INRA-13812 is of similar size and origin (Ae. tauschii) to a 453 bp Ppd-D1 allele amplified by Guo et al. (2010), suggesting these alleles likely contain the same polymorphism. The 453 bp allele was shown to contain 24 and 15 bp insertions relative to Ppd-D1b (Guo et al. 2010), but the effect on days to heading was not estimated. The remaining Ppd-D1 allele identified using the Ppd-D1a/b assay is characterized by the lack of an amplified product.  The occurrence of this allele in three landraces from the Black Sea region indicates this was the likely origin. Further research is required to determine the polymorphism associated with this allele and to estimate its effect on heading date.

For days to heading, the Ppd-D1 locus had the largest effect, but the effect of Rht-B1 was also significant. Ear emergence of Ppd-D1a plants was an average of 7.6 days earlier than Ppd-D1b plants. Similarly, Worland et al. (1988) reported that Ppd-D1a generally accelerates ear emergence by 6–8 days in the UK, and Bentley et al. (2013) reported a reduction in heading date of 4–9 days associated with Ppd-D1a in UK field trials. The cause of the 2.9 delay in days to heading associated with Rht-B1a-160 relative to Rht-B1a-0 is unclear and requires further investigation. Rht-B1b and Rht-D1b were not associated with changes in days to heading, which agrees with previous observations (J. Flintham, JIC, pers. comm.).

Changes in plant height were associated with all three loci that were examined and changes in GA sensitivity were associated with only Rht-B1 and Rht-D1. Among the alleles, Rht-B1b and Rht-D1b resulted in the greatest reductions in height and GA sensitivity. Based on fitted means (Table 2), Rht-D1b accessions were 70 % the height of accessions containing Rht-D1a; Rht-B1b accessions were 78 % the height of Rht-B1a accessions and 70 % the height of Rht-B1a-0 accessions; and Rht-B1b + Rht-D1b accessions were 55 % the height of Rht-B1a + Rht-D1a accessions. Flintham et al. (1997), using four NIL series, reported that plant heights of Rht-B1a + Rht-D1b, Rht-B1b + Rht-D1a, and Rht-B1b + Rht-D1b NILs were 83, 86, and 58 %, respectively, that of Rht-B1a + Rht-D1a NILs. The larger measured effect of the Rht-1b alleles in our study relative to Flintham et al. (1997) is at least partially due to the inclusion of landraces and older varieties in the 372CC. These older accessions appear to contain fewer secondary height-reducing alleles than the modern varieties in which Rht-B1b and Rht-D1b are found (discussed below) and, therefore, may artificially increase the measured effect of these alleles. Rht-B1b and Rht-D1b height reductions are attributable to a reduced sensitivity to GA as demonstrated by the large and significant reductions in GA trt diff fitted means (Table 2) and by the classification of nearly all accessions containing Rht-B1b or Rht-D1b as GAI (Table 4). This was expected, as both alleles are well known to have height reductions associated with reduced GA sensitivity (Gale and Youssefian 1985; Peng et al. 1999). GA trt diff was decreased in Rht-B1b + Rht-D1b accessions relative to accessions containing a single dwarfing allele. This dosage response was also reported by Yamada (1990) and indicates that varieties containing a single Rht-1 semi-dwarf allele retain some sensitivity to GA, perhaps owing to the remaining homoeologous wild type copies.

Fitted mean plant heights of Rht-B1a-160 and Rht–B1a-197 accessions were reduced by 18 and 12 %, respectively, relative to Rht-B1a-0 accessions under the conditions used in this study. Small reductions in GA sensitivity of 2.9 and 1.0 mm were also associated with Rht-B1a-160 and Rht-B1a-197, respectively, relative to Rht-B1a-0 accessions, but these differences were not statistically significant. In a separate study that measured transcript abundance in leaf and shoot tissue of 5-day-old seedlings using quantitative real time PCR, small reductions in Rht-B1 transcript abundance were reported in 2 of 2 lines containing Rht-B1a-160 and in 3 of 3 lines containing Rht-B1a-197 relative to lines with no insertion, but these reductions were not statistically significant with the exception of one line that contained Rht-B1a-197 (Wilhelm et al. 2013b). Rht-1 expression patterns can differ with tissue type and developmental stage (Pearce et al. 2011); hence, changes in Rht-B1 expression or GA sensitivity associated with Rht-B1a-160 or Rht-B1a-197, if present, may be greater after the seedling stage or in other tissue types. Both insertions occur within the 600 bases upstream of the Rht-B1 start codon and the 160 bp insertion occurs within non-coding sequence that is highly conserved among the Poaceae (Duan et al. 2012; Wilhelm et al. 2013a); therefore, either insertion has the potential to disrupt an Rht-1 regulatory region. Rht-1 regulatory elements have not yet been identified and the phenotypic effects of polymorphisms occurring outside of the coding region have not been characterized. The measured effects of Rht-B1a-160 and Rht-B1a-197 on plant height in the 372CC indicate these alleles may be useful for intermediate reductions in height relative to Rht-B1b and that this is a region that merits further study. To more clearly estimate the effects of these alleles, NILs containing Rht-B1a-160 or Rht-B1a-197 are being created.

Accessions containing Ppd-D1a have plant heights that are reduced by 15 % relative to Ppd-D1b accessions when the fitted mean is calculated over the four Rht-B1 alleles or over the two Rht-D1 alleles. Ppd-D1 had almost no effect on GA trt diff. Ppd-D1a has a shared ancestry and close genetic linkage with the GA sensitive semi-dwarf allele Rht8, both of which are located on chromosome 2D (Worland et al. 1998b; Gasperini et al. 2012). Gasperini et al. (2012) attributed a 13 % reduction in plant height to the semi-dwarf Rht8 allele and a 3 % reduction to Ppd-D1a in a study that assessed recombinant inbred lines segregating at both loci. Assuming tight linkage exists between the semi-dwarf Rht8 allele and Ppd-D1a in the 372CC accessions examined herein, Rht8 is likely the primarily cause of the height reduction associated with Ppd-D1a.

The analysis of registration year period revealed that predicted mean height of accessions released in the three 20-year periods from 1940 to 1999 was 82 % that of the three periods spanning 1800–1939 even after the effects of Ppd-D1, Rht-B1, and Rht-D1 were accounted for in the model. Similarly, the predicted mean height of fixed lines was 88 % that of landraces after accounting for the effects of Ppd-D1, Rht-B1, and Rht-D1. These results suggest the presence of secondary alleles in modern bread wheat that further reduce height relative to landraces and older germplasm. The absence of a significant difference in GA trt diff between the pre-1940 and post-1940 groups suggests that this secondary height reduction does not primarily result from changes in GA sensitivity.

The high correlation between plant height and GA trt diff (r = 0.58) in this study is not surprising considering reduced stature of Rht-1b accessions is due to reduced GA sensitivity (Gale and Youssefian 1985; Peng et al. 1999). Days to heading had little effect on plant height (r = 0.18). Similarly, when the 372CC was grown in France, the correlation between plant height and ear emergence was also low (r = 0.33; Bordes et al. 2008). The slight increase in r value when the collection was grown in France may relate to changes in growth conditions or experimental design. Similar to our results, Dotlacil et al. (2003) estimated an r value of 0.19 between days to heading and plant height for European winter wheat accessions grown in the Czech Republic.

In most instances, the PDM model and LSD test correctly classified GA response relative to Rht-B1b and Rht-D1b genotype scores. In the PDM model, none of the accessions containing a known GAI allele were classified as GAS. The LSD test procedure, however, classified five GAI 372CC accessions and the control ‘Bersee Rht-B1b’ as GAS. The misclassifications were due to the large GA treatment difference present in these accessions, which is likely the result of a GA response from the remaining Rht-1 homoeologs. Among the 372CC accessions that did not contain a known GAI allele, no novel GAI alleles with an effect as strong as Rht-B1b or Rht-D1b were identified. However, INRA-03752 was classified as GAI in the PDM model and INRA-04670 was classified as GAI by the LSD test. These two accessions along with fourteen others classified as intermediate by the PDM model may have a weak GA response relative to the Rht-1b alleles. There may be other accessions with a minor GA response among the remaining 372CC as well, as a weak GA response in combination with the inherent noise of genetic background may require greater replication than that used herein to positively identify such accessions. In this study, the PDM model was superior to the LSD test in regards to correct identification of GAI alleles and only required a GA+ treatment. The drawback of this method is the need for large sets of accessions that contain a sufficient number of GAI and GAS lines to allow for discrimination between the two distributions. For instance, in GA expt 3, the number of GAI lines was insufficient to allow for effective use of the PDM model. Another potential drawback of the PDM method is that accessions with intrinsically long seedling lengths would not be easily detected without a GA− treatment. However, accessions of this nature were not encountered in the diverse 372CC, suggesting this phenotype is rare in modern hexaploid wheat. The statistical methods developed herein provide researchers two options for classifying GA response, with the PDM model being best suited for larger experiments where two defined distributions are expected.

This study provides an estimate of the effects of several alleles at the Rht-1 and Ppd-D1 loci in a worldwide collection under near-optimal growing conditions in the UK. Similar to previous studies, Rht-B1b and Rht-D1b both had strong effects on plant height and GA sensitivity and Ppd-D1a had a strong effect on days to heading. The novel Rht-B1a-160 and Rht-B1a-197 insertion alleles were associated with statistically significant reductions in plant height of 12 and 18 %, respectively, relative to Rht-B1a-0. The height reductions were less than those associated with Rht-D1b or Rht-B1b and were not clearly related to changes in GA sensitivity. Further characterization of the Rht-B1 insertion alleles is required, but our results suggest these alleles may be useful for fine adjustments in plant height. In addition, the genetic and phenotypic characterization of the 372CC and the development of statistical methods for determining GA response should aid the discovery of novel GAI alleles and the breeding of varieties with height and flowering time suited to specific environments.