Agronomic assessment of the durum Rht18 dwarfing gene in bread wheat

Keywords: coleoptile, dwarf, early vigour, establishment, germplasm, harvest index, lodging, physiology.

Dwarfing genes have been a major driver of improved adaptation and performance with breeding and domestication across all of the major cereals (Hedden 2003). In wheat, the identification and deployment of genes for semi-dwarf stature has promoted the commercial release and global adoption of wheat cultivars with greater yield potential and stability (Perry and D’Antuono 1989; Mathews et al. 2006). Selection for reduced crop height remains a key objective of wheat breeding programs worldwide owing to semi-dwarfs being less prone to lodging and producing greater numbers of grain to increase harvest index and crop yields (Hedden 2003).

Despite the availability of >20 dwarfing genes (McIntosh et al.1998; Ellis et al. 2004), the deployment of the Rht-B1b and Rht-D1b dwarfing alleles from the Japanese wheat cultivar Norin 10 has been remarkably widespread in the selection of reduced plant height globally. Indeed, ∼70% of the commercial wheat cultivars grown worldwide contain either Rht-B1b or Rht-D1b (Evans 1998) and both genes were pivotal in yield increases arising out of the Green Revolution (Hedden 2003). However, their value in improving performance in water-limited environments is less clear (e.g. Butler et al. 2005; Mathews et al. 2006). The Rht-B1b and Rht-D1b alleles are the most widely deployed of a group of gibberellic acid-insensitive (GAI) dwarfing genes that are unique in reducing cell-expansion in response to endogenous gibberellins (Hoogendoorn et al. 1990; Botwright et al. 2005). Reduced cell elongation contributes to reduction in cell length and width without affecting final cell number (Keyes et al.1989; Botwright et al. 2005). Reduction in cell size contributes to reductions in internode length, including the peduncle, to affect final plant height (Hoogendoorn et al. 1990). This reduction in cell size is ubiquitous in above-ground tissues and contributes to reductions in coleoptile and sub-crown internode lengths, coleoptile tiller frequency and size, and individual leaf area to reduce overall seedling vigour (Allan 1989; Botwright et al.2005; Rebetzke et al.2014).

Establishment is a key phase in the development of high-yielding cereal crops (Rebetzke et al. 2007). Optimal plant densities and early leaf area development rely on the seedlings’ ability to elongate and emerge with deep sowing, and are important objectives of breeding programs targeting adaptation to water-limited environments (Richards 1992). Improved establishment and early vigour are likely to be more important with climate change and predicted variability with changes in the ‘break’ for sowing and increasingly greater reliance of deep soil moisture from summer rains (Flohr et al. 2021). Greater seedling vigour has also been suggested as a trait for improving weed competitiveness (Coleman et al.2001; Zerner et al.2016) and nutrient uptake (Pang et al. 2014; Ryan et al.2015). The constraining influence of GAI alleles such as Rht-B1b and Rht-D1b on early growth has limited the capacity for breeders developing more-vigorous GAI wheats (Allan 1989; Rebetzke et al. 2007, 2014). Several major, gibberellic acid-responsive (GAR) dwarfing genes have been identified with potential to reduce plant height without affecting seedling vigour (Ellis et al. 2004; Rebetzke et al. 2011, 2012). Many of these genes are available in bread wheat and include Rht5 (Daoura et al. 2014), Rht8 (Rebetzke and Richards 2000), Rht12 (Chen et al. 2013), and Rht13 (Rebetzke et al. 2011). These genes have been reported to reduce plant height and increase grain yields through greater partitioning of carbon to growing ears, and thereby increase floret fertility to increase grain number (e.g. Rebetzke et al. 2012). Reduced height mutants in durum wheat have been reported to contain three GAR dwarfing genes: Rht14, Rht16 and Rht18 (Haque et al. 2011).

The Rht18 dwarfing gene was first deployed in a semi-dwarf durum cv. Icaro as a direct selection following fast-neutron mutagenesis of the tall wheat Anhinga (Konzak 1988). The Rht18 gene codes for the GA2oxA9 protein to lower bioactive gibberellic acid content to reduce stem elongation and plant height (Ford et al.2018). There are only two reports of the value of Rht18 in improving grain yield in wheat, and these represent performance in very small plots in single environments (Yang et al.2015; Tang 2016). The aims of the studies reported herein were to report the transfer of the Rht18 dwarfing allele from durum into bread wheat and then to assess the effect of this gene on plant height and agronomic performance in backcross-derived, Rht18, Rht-B1b and tall progeny evaluated in large plots across multiple field environments.

Hybridisation was undertaken between the Rht18-containing, Italian durum wheat cv. Icaro (PI503555) and the tall, non-Rht18 Australian commercial bread wheat cv. Halberd (PI377885). The F₁ pentaploid seed was harvested, sown and then backcrossed to Halberd to generate BC₁F₁ seeds, which were then self-pollinated to produce BC₁F₂ progeny. Only fertile wheat heads were retained because these were presumed to be fully fertile and therefore genetically hexaploid. This process was repeated for three generations to produce ∼120 BC₁F₂-derived, F₅ recombinant inbred lines (RILs). Seeds from individual F_4:5 plants were sown into rows and a single head was harvested from rows homogeneous for plant height. Harvested heads were threshed and sown into rows in the field for plant height assessment in subsequent generations. About 120 BC₁F₂-derived, F_5:7 RILs varying for height were identified and bulk harvested. All lines were then sown into a summer nursery for further seed increase. Two inbred lines were discarded based on visual evidence for partial fertility in the summer nursery. For all sowings, the KCD NIL set, containing wheat height near-isogenic lines (NILs: semi-dwarf, Rht-B1b/D1b; doubled-dwarf, Rht-B1d/D1d; and tall, Rht-B1a/D1a), as described in Richards (1992) was sown to aid in height classification.

The simple sequence repeat (SSR) marker Xwms4608 tightly linked to Rht18 was used to genotype BC₁F_5:6 RILs (see Ford et al. 2018). In total, 51 RILs homozygous for a 220 bp allele of Xwms4608 (Halberd allele) averaged mature plant heights of 113 cm, compared with 57 RILs homozygous for the 239 bp Icaro allele averaging 89 cm (−27%) plant height when assessed in a favourable nursery environment.

Experiments were undertaken at Temora and Yanco, New South Wales, in 2010, and at Yanco in 2011, 2013, 2014 and 2015. In all experiments, BC₁F_5:7 RILs were sown at an optimal 3–5 cm sowing depth in 6-m-long, 0.17-m-spaced, 5- or 10-row plots at a seeding rate of ∼200 seeds/m². Lines were replicated in a row–column experimental design containing partial (p = 1.5) or full (p = 2) replication. Nutrients were supplied at sowing as Starter 15 (14:12.7:11 N:P:S; Incitec Pivot, Melbourne, Vic., Australia) applied at 103 kg/ha and then as commercial urea topdressed at 80 kg/ha at early stem elongation. A pre-sowing irrigation followed by supplemental irrigations as required were supplied to maintain potential growth, and sowing was managed to be free of diseases and weeds with the application of appropriate fungicide and herbicide control measures.

The lines sown included the 108 BC₁-derived Rht18 RILs and several control entries including: the original parents Icaro and Halberd; and a set of Rht F₅-derived sib-based NILs varying for Rht-B1b and Rht-D1b alleles in CIMMYT-based genetic backgrounds of cvv. Aconchi, Galvez, Kauz, Nesser, Pavon and Seri (Singh et al. 2001).

For each plot, phenological development was recorded using the Zadoks development scale (Zadoks et al. 1974). Lodging was scored at multiple times throughout grain-filling in each plot (1 = perpendicular to 9 = prostate to the soil surface), and plant height was determined at physiological maturity as the distance from the soil surface to the top of the ear (awns excluded) at three random positions in each plot. At harvest maturity, ∼100 culms were hand-cut at ground level using a 40-cm-long quadrat oriented across four rows. Numbers of spikes were counted and samples air-dried at 35°C for 3 days, after which they were weighed and threshed, and grain was weighed. Harvest index calculated as the ratio of grain weight to total aboveground biomass. Plots were end-trimmed to ∼5.0 m length and the outside border rows removed before machine harvesting.

Grain size was determined as 100-grain weight from a random sample of grain from the harvest index cuts. Grain number (per area) and grain number per spike were subsequently calculated from grain size and plot yields. For two environments (Yanco 2013 and 2014), percentage shrivelled grain (‘screenings’), test weight and grain protein concentration were also determined. Nitrogen (N) yield was calculated from grain protein and grain yield, and assuming a grain protein-N correction factor of 5.7.

Three separate experiments were undertaken under field and controlled environment conditions for assessment of early vigour:

1. Ground cover was estimated on the BC₁F₂-derived RILs from digital images in three environments (Temora in 2010 and Yanco in 2010 and 2011) and encompassing one to three development stages (Z16, Z22 and Z23) representing early, mid and late vegetative growth. Digital images were converted to estimates of ground cover percentage using the vegetation-cover prediction software CanopyCover with parameters iterated before setting to minimise distortion from background soil (Li et al. 2010). Normalised difference vegetative index (NDVI) was measured using a GreenSeeker (Trimble, Sunnyvale, CA, USA) reflectance unit at the same time as ground cover images were taken.

2. Seed from the Yanco 2011 field harvest was sized and sown for early vigour determinations after Rebetzke and Richards (1999). Briefly, good-quality seeds free of any visible damage (particularly shrivelling) and weighing 40–45 mg were obtained for early vigour assessment of each NIL, RIL and parent (Icaro and Halberd). These were sown into wooden seedling trays (600 by 300 by 120 mm) containing a fertile, compost-based potting mix. The experiment was a row–column, partial replicate (p = 1.5) design constraining genotypes from being sown in the same position across replicates. After sowing, trays were placed in an outdoor nursery during the winter until ∼3.5-leaf stage, whereupon seedlings were harvested and assessed for numbers of main-stem leaves, and width and length of the first three main-stem leaves. Total leaf area was estimated using a planimeter and dry weights were determined after drying in a 70°C oven for 3 days.

3. The same seed from the Yanco 2011 field harvest was sized and sown for coleoptile/shoot length determinations after Rebetzke et al. (2007, 2014). As for Expt 2, good-quality seeds weighing 40–45 mg were obtained of each RIL, parents Icaro and Halberd, and the CIMMYT height NILs, and sown into wooden trays. Trays were then placed in a darkened growth cabinet with soil temperatures set at a constant 15°C. Seedlings were harvested at 200 degree-days and assessed for length of both coleoptile and the entire shoot (as the entire length from the seed to the tip of leaf 1).

A set of NILs varying for Rht-B1b and Rht18 was developed in the Halberd background for comparing the effects of the two genes in the field. The Rht18 NILs were developed from crossing the Rht18-containing line HI25M (a selection from the Halberd × 2/Icaro population described above) and genotyping F₁ progeny with the SSR marker Xwms4608 (see above). This was repeated five times before single-seed descent to generate 15 BC₅F₆-derived Rht18-containing NILs. Halberd was also crossed to the Rht-B1b-containing cv. Lang and backcrossed five times with selection of Rht-B1b containing F₁ progeny using the KASP (Kompetitive allele specific PCR) Rht-B1b marker (after Ellis et al. 2004). Fifteen BC₅F₆-derived NILs were then developed following single-seed descent. All 30 lines and the tall recurrent parent Halberd were grown under favourable conditions in the glasshouse, and mature seed was harvested before threshing and sowing into deep wooden seedling trays for assessment of coleoptile length (as above). These same 15 Rht-B1b and Rht18 NILs, and recurrent parent Halberd, were sown in a partial-replicated (p-rep = 1.5), row–column experimental design at the Yanco Managed Environment Facility (Rebetzke et al. 2013) in 2015 only. Plots were grown in rainfed and irrigated treatments, with diseases and nutrition managed to maximise crop growth. Sowing depth was confirmed at 4 cm (data not shown). Measurements were made at Z13 and Z14 of NDVI, and at maturity of plant height, number of spikes, grain yield, total biomass, harvest index (as grain yield/total biomass), grain size, grain number, grain protein concentration and N yield.

Data were analysed statistically after first checking for normality and error variance heterogeneity across environments. For lodging, the sampling date representing the largest lodging-score mean was used in statistical analysis. All data were analysed as measured except ground cover percentage, which was arcsine transformed prior to statistical analysis. Thereafter, a two-stage, mixed model approach was employed (Cullis et al. 1996) using the spatial models procedures in Genstat 14th Edition (VSN International, Hemel Hempstead, UK). Each environment was analysed separately with the best spatial models being determined after first fitting the experimental design and then modelling the residual variation with autoregressive row and column terms. Significant spatial effects were then identified, and residuals assessed before determinations were made as to the need for fitting of other (e.g. linear) effects (Cullis et al. 1996). The best linear unbiased estimates (BLUEs) for each environment were then used as inputs in the subsequent cross-environment analyses (after McIntosh 1983).

Analysis of the effect of each dwarfing gene on agronomic performance was assessed using: (a) a t-test for testing dwarf vs tall allele(s) in the CIMMYT-derived Rht-B1b and Rht-D1b semi-dwarfs with their corresponding CIMMYT tall NILs, and Halberd-derived Rht18 semi-dwarf RILs with their corresponding tall siblings; and (b) a Dunnett’s test for comparisons of the Halberd-derived dwarfing gene NILs with tall recurrent parent M808S and Halberd, respectively. Genetic correlations were estimated for plant height and selected traits after Holland (2006). Unless otherwise stated, statistical significance under all null hypotheses of no entry or dwarfing gene difference was at α = 0.05.

Experiments experienced good in-crop rainfalls ranging between 279 and 410 mm across years. Rainfall was particularly high in 2010 and 2013, exceeding long-term averages for the Yanco site (Fig. 1). Years 2011 and 2014 experienced good rainfall before or early in crop growth but were drier from flowering onward, in turn requiring up to 90 mm irrigation during grain-filling. Temperatures were reasonably consistent across years except for above-average temperatures during flowering and grain-filling in 2015 (Fig. 1).

Fig. 1.  Mean monthly mean temperature (bars) and total monthly rainfall (line) at Yanco and Temora NSW for years 2010–15 under study.

Environments in each site × year combination were considered favourable, with grain yields exceeding an average 5 t/ha in all but one environment (Table 1). Average plant height varied significantly among environments, ranging from 90 to 108 cm, and with ranges in plant height across lines within environments of up to 80 cm (data not shown). Grain number and size were commonly large, consistent with the commonly wet and mild conditions experienced in the sampled environments. One site (Temora 2010) was sown to all lines but they were not harvested owing to prolonged rainfall during harvest damaging the mature grain. Notwithstanding, data for plant height and lodging score were collected (Table 1) and have been included in statistical analyses of these traits (Tables 1 and 2).

Table 1.  Environment means for agronomic traits measured on dwarfing gene Halberd BC₁-derived (Rht18/rht18) RILs and CIMMYT (Rht-B1b/D1b and Rht-B1a/D1a) NILs.

Table 2.  Genotype group means for plant height and agronomic performance measured on dwarfing gene Halberd BC₁-derived (Rht18/rht18) RILs and CIMMYT (Rht-B1b/D1b and Rht-B1a/D1a) NILs evaluated across multiple environments.

Genotypic variation was large and statistically significant for plant height measured across all entries (Table 2, Fig. 2). Genotype × environment interaction was small relative to among-genotype genetic variance (data not shown), contributing to high entry-mean repeatabilities for plant height across experiments. Accordingly, the average rank correlation for plant height across environments (r_g) was high at 0.89. The range in genotype means for plant height across all environments was 78–133 cm (Fig. 2). Lines containing the 6AS molecular marker linked to Rht18 were significantly shorter (−24%) than their tall rht18 counterparts, and were commonly as short as, or shorter than, Rht-B1b- and Rht-D1b-containing CIMMYT NILs grown in the same experiments (Fig. 2, Table 2). The shortest Rht18-containing lines were significantly shorter than the Rht18 donor Icaro (cf. Fig. 2 and Table 2) (the two shortest Rht18 lines were 78 and 81 cm), whereas the tallest lines were significantly taller than the standard height parent Halberd (the two tallest lines were 131 and 132 cm in height). Reduction in plant height with Rht18 was associated with a significant reduction in (stem) lodging (Table 2), with the shortest Rht18-containing lines producing among the highest standability scores in the study (data not shown).

Fig. 2.  Overall mean and distribution of maturity plant height for the Halberd × 2/Icaro Rht18 and rht18 RILs, and both Rht-B1b and Rht-D1b semi-dwarf and corresponding Rht-B1a and Rht-D1a tall CIMMYT NILs assessed across multiple experiments representing different sites and years. Dots represent extreme line values, whiskers the 5th and 95th percentiles, boxes the 25th and 75th percentiles, and solid and dashed lines the median and mean, respectively, for each population × dwarfing gene class.

Lines containing the Rht18-linked marker had on average significantly greater grain yield across all environments than their tall siblings (Table 2). This increase in yield largely reflected significant increases in harvest index with little change in total biomass. Anthesis date was similar for semi-dwarf and tall lines (<1 day difference), and grain number was larger (+21%) in lines containing the Rht18 dwarfing allele (Table 2). Number of spikes and number of grains per spike were greater for lines containing the dwarfing Rht18 allele (+7% and 12%, respectively), and grain size was significantly smaller (−4%) for semi-dwarf lines (Table 2).

Compared with the tall parent (Halberd), the significantly reduced grain yields of the tall BC₁ derivatives (Table 2) reflects their greater heights and reduced harvest index, and their reduced proportion of Halberd ‘adaptation’ alleles (∼75%). Further, unlike the substantial testing and selection involved in the development of the older but successful cultivar Halberd, these tall progeny lines were random samples from a large genetic population selected only for plant height and not for grain yield or any other agronomic attribute.

Presence of the Rht-B1b and Rht-D1b dwarfing alleles was associated with significant height reduction in the CIMMYT-based NILs (Table 2). Differences in height between Rht-B1b and Rht-D1b NILs were small and, in most cases, not statistically significant (data not shown). In the CIMMYT NILs, the GAI dwarfing alleles were associated, on average, with a 22% reduction in plant height. Grain yield was significantly greater for Rht-B1b- and Rht-D1b-containing NILs, with differences in grain yield being comparable with yield differences in the Halberd-derived Rht18 lines (Table 2). Greater yield was associated with significantly higher harvest index determined largely though increases in grain number, and particularly number of grains per spike. Total biomass was unchanged between semi-dwarf and tall CIMMYT NILs.

The Rht18 dwarfing gene was associated with a small but non-significant (P > 0.05) reduction in average grain protein concentration compared with tall, non-Rht18 RILs (Table 3). When coupled with the increased grain yield of Rht18-containing lines, N yield was significantly greater for semi-dwarf progeny. Grain protein concentration was significantly lower for the GAI, semi-dwarf CIMMYT NILs; however, their greater yield contributed to increased N yield compared with the CIMMYT tall NILs. Screenings percentage was significantly greater and test weight smaller for Rht18 than rht18 RILs (Table 3). Consistent with Rht18 RILs, test weight was significantly smaller for Rht-B1b and Rht-D1b CIMMYT-derived NILs. This small difference in screenings between classes of CIMMYT NILs may reflect the greater average grain weight of both the semi-dwarf and tall CIMMYT NILs (Table 2).

Table 3.  Genotype group means for grain quality characteristics measured on dwarfing gene Halberd BC₁-derived (Rht18/rht18) RILs and CIMMYT (Rht-B1b/D1b and Rht-B1a/D1a) NILs evaluated across multiple environments.

Repeatabilities were high for coleoptile (85 ± 7%) and shoot (78 ± 9%) lengths, respectively, for lines assessed in the controlled environment. The tall, long-coleoptile parent Halberd produced significantly (P < 0.01) longer coleoptiles and shoots than the Rht18 donor Icaro (Table 4), with Halberd’s greater length consistent with its longer coleoptile size in previous studies (e.g. Rebetzke et al. 2007) and with the size of Rht-B1a and Rht-D1a tall CIMMYT NILs in the same growth cabinet experiment (cf. Fig 3 and Table 4). In turn, the coleoptiles and shoots of Icaro were longer than those of the Rht-B1b and Rht-D1b CIMMYT NILs (cf. Table 4 and Fig. 3). The range in coleoptile and shoot lengths was similar for lines with and without the Rht18-linked marker (Fig. 3). The range was commonly large for both genotype classes with evidence for transgressive segregation of both shorter and longer shoots. Mean coleoptile and shoot lengths were similar and not statistically different for short Rht18 and tall genotypes (Table 4).

Table 4.  Genotype means for parents (P) and semi-dwarf (Rht18) and tall (rht18) BC₁-derived RILs for early vigour characteristics measured in the field, and in growth cabinets and outdoor nurseries.

Fig. 3.  Overall mean and distribution of coleoptile and shoot lengths for Halberd × 2/Icaro Rht18 and rht18 RILs, and both Rht-B1b and Rht-D1b semi-dwarf and corresponding Rht-B1a and Rht-D1a tall CIMMYT NILs assessed at 15°C soil temperature. Dots represent extreme line values, whiskers the 5th and 95th percentiles, boxes the 25th and 75th percentiles, and solid and dashed lines the median and mean, respectively, for each population × dwarfing gene class.

The influence of Rht18 on early growth was assessed in RILs phenotyped in both field and controlled nursery environments in Canberra. Parent variety Halberd produced significantly longer and wider seedling leaves than Icaro to increase total leaf area per plant at the 3.5-leaf stage (Table 4). This contrasted with their progeny, where Rht18 lines had greater mean leaf width than rht18 lines but were not statistically different for leaf length, plant leaf area or biomass. Numbers of leaves were similar for tall and Rht18-containing lines, but the latter produced significantly more tillers at sampling (1.92 and 1.78 tillers for Rht18 and rht18 lines, respectively) (data not shown).

Multiple field observations were made of ground cover percentage and NDVI representing different seasons and sites and sometimes multiple growth stages (Fig. 4). The genetic correlation (±s.e.) of ground cover and NDVI was 0.88 (±0.04) at Yanco and 0.85 (±0.04) at Temora at the 5.0-leaf stage, and the pooled genetic correlation across sites was 0.86 (±0.03). The strong correlations reflected the consistent performance of lines for both characteristics across sample dates and environments (Fig. 4). Indeed, genotype × environment and genotype × sample date interactions were small relative to the large genetic variance for ground cover and NDVI (data not shown).

Fig. 4.  Ground cover and normalised difference vegetative index (NDVI) for the Halberd × 2/Icaro Rht18 and rht18 genotyped RILs assessed in the field at Temora in 2010 and Yanco in 2010 and 2011. Values in parentheses are the days after sowing at which sampling occurred. Dots represent extreme line values, whiskers the 5th and 95th percentiles, boxes the 25th and 75th percentiles, and solid and dashed lines the median and mean, respectively, for each character.

Averaged across all environments, ground cover was not different between Rht18 and tall lines, whereas NDVI was significantly greater for tall genotypes, although the difference was small (0.321 and 0.332 for Rht18 and tall lines, respectively). Despite the differences between genotype classes, the overlap in individual genotype values was large for both ground cover and NDVI scores (Fig. 4), highlighting that more vigorous Rht18 lines could be readily identified. Indeed, the two best performing lines for ground cover and NDVI were reduced in height and contained the Rht18 gene (data not shown).

A high water-limited grain yield potential of 5.0 t/ha was observed for the field assessment of the BC₅ Rht-B1b and Rht18 Halberd NILs at Yanco in 2015 (Table 5). Grain yields for rainfed and irrigated treatments were 3.6 and 5.9 t/ha, respectively. Significant differences were observed for plant height, with the Rht18 and Rht-B1b single-dwarfs reduced by 21% and 19% of the height of the tall recurrent parent Halberd, respectively, whereas the Rht-B1b and Rht18 NILs were themselves not statistically different. Crop lodging score was significantly higher for Halberd, and the Rht-B1b NILs had a significantly higher lodging score than the Rht18 NILs. Harvest index and grain yield of the two semi-dwarf sets of NILs were significantly greater than those of Halberd but were themselves not significantly different (Table 5). The greater grain yields of the semi-dwarf NILs reflected significantly larger grain number, but they were smaller in average grain size. Number of spikes and number of grains per spike were significantly greater for both semi-dwarf NILs than for Halberd. Number of spikes was the same for Rht-B1b and Rht18 NILs, whereas number of grains per spike was larger for the Rht18 NILs (P < 0.10). Maturity biomass was the same for the Rht-B1b and Rht18 NILs. Grain protein concentration was significantly greater for the Rht18 NILs, but smaller for the Rht-B1b NILs, than for Halberd (Table 5). Grain N yield was significantly greater for both Rht18 (102 kg/ha, +20%) and Rht-B1b (94 kg/ha, +11%) than for Halberd, and significantly greater for Rht18 (+9%) than for Rht-B1b NILs.

Table 5.  Dwarfing gene means for agronomic traits in a field sowing of Halberd-derived Rht-B1b and Rht18 BC₅-derived NILs and the recurrent parent Halberd.

Number of plants per m² was not significantly different for Rht18 (138 plants/m²) and Rht-B1b (132 plants/m²) NILs, or recurrent parent Halberd (151 plants/m²). Early ground cover was assessed as NDVI at crop development stages Z13–14 (Table 5), with Rht18 NILs being similar in NDVI to recurrent parent Halberd, whereas both Rht18 lines and Halberd were significantly more vigorous than the Rht-B1b NILs. Consistent with NDVI estimates, coleoptile lengths were similar for Rht18 and Halberd, and both were significantly longer than for Rht-B1b NILs.

The Rht18 dwarfing gene had been deployed previously in commercial durum cv. Icaro where it performed well in producing higher grain yields and improved grain quality (Konzak 1988). In our own unpublished glasshouse and field studies, comparisons between Icaro (Rht18) and both cvv. Casteloporziano (Rht14) and Edmore M1 (Rht16) indicated that Icaro was more extreme in height reduction, consistent with the comparisons reported in Haque et al. (2011). Further, genetic studies at CSIRO have highlighted Rht18 as different from the majority of other dwarfing genes in that reduced plant height was inherited as a dominant trait (and thereby had dominant gene action) in F₁ and subsequent segregating generations (G. J. Rebetzke, unpubl. data). In turn, physiological mechanisms contributing to reduction in plant height might be different from other height-reducing genes. Indeed, Ford et al. (2018) has reported the Rht18 gene as encoding the GA2oxA9 protein to lower bioactive gibberellic acid content. This differs from the Rht-B1b and Rht-D1b Green Revolution dwarfing genes, which are DELLA mutants rendering plant cells insensitive to endogenous gibberellins.

Large and repeatable genotypic differences were observed for plant height, largely reflecting variation due to Rht18. The large height range in the progeny encompassed variation in the parents (the Rht18-containing Icaro and the tall parent Halberd), with a number of progeny significantly exceeding either parent. The large range reflects combinations of both the Rht18 major dwarfing gene from Icaro and minor height-reducing alleles contributed from Halberd. The average 24% height reduction in the Rht18-containing lines was comparable to the height reductions of Rht-B1b and Rht-D1b NILs in the CIMMYT NILs (22% reduction) and in the direct height comparison with the Halberd-derived BC₅ Rht-B1b NILs. The observed height reduction with Rht18 was consistent with results reported previously for both GAR (e.g. Loskutova 1998; Rebetzke et al. 2012; Daoura et al.2014) and GAI (e.g. Brandle and Knott 1986; Flintham et al. 1997; Rebetzke et al.2001; Ellis et al.2002) dwarfing genes.

Variation in plant height among Halberd-derived progeny was strongly genetically correlated with reductions in crop lodging in the RIL population (r_g = 0.63, P < 0.01). Similarly, both the Rht18- and Rht-B1b-containing Halberd NILs scored significantly lower than Halberd for lodging, whereas the Rht18-containing NILs scored significantly lower than Rht-B1b-containing Halberd NILs. The reduced lodging with both Rht-B1b and Rht18 was consistent with the influence of height reduction on lodging reported for GAI dwarfing genes elsewhere (Allan 1989). However, the even greater reduction in lodging associated with Rht18 was quite remarkable given the similarity in plant height with both Rht-B1b and Rht18 NILs (Table 5). The increased grain yield associated with Rht18 largely reflected increases in harvest index, with little change in aerial biomass. In turn, increases in harvest index and thereby grain yield were largely due to increases in grain number (r_g = 0.84, P < 0.01) through the production of significantly more spikes and greater numbers of grains per spike (i.e. fertility). Increases in grain number were partly compensated by a reduction in average grain size. Increases in harvest index and grain number for Rht18 were consistent with the Rht-B1b Halberd NILs, and consistent with reports elsewhere for Rht-B1b and Rht-D1b (Fischer and Stockman 1986; Richards 1992; Flintham et al. 1997; Butler et al. 2005) and the alternative GAR dwarfing genes (e.g. Rebetzke and Richards 2000; Rebetzke et al. 2012). In three winter biparental populations, Rht18 was associated with reductions in height of 12–25% depending on population (Yang et al. 2015). Grain yield and mature plant biomass (on a single-plant basis) were reduced in Rht18-containing progeny, whereas harvest index was increased relative to the tall parents. Similarly, the influence of Rht18 on numbers of spikelets and grains per spike and grain size also varied across populations. In turn, the results of both Tang (2015) and Yang et al. (2015) must be treated with caution owing to plot type and size (Rebetzke et al.2014).

The Rht18 dwarfing gene was associated with small but non-significant reductions in grain protein concentration and a significant increase in grain screenings and test weight compared with non-Rht18 talls in the RIL population. However, the Rht18-containing Halberd NILs produced significantly greater grain protein concentration than both the recurrent parent Halberd and the Rht-B1b-containing NILs in the 2015 field study. In all cases, the higher yields associated with Rht18 were associated with greater N yield. Reductions in grain protein concentration have been observed elsewhere for Rht-B1b and Rht-D1b (e.g. Brandle and Knott 1986; Flintham et al. 1997; McCartney et al. 2006).

The lines developed and reported herein served as key parents in genetic studies reported by Flohr et al. (2021), Ford et al. (2018) and Tang (2015), and have been used elsewhere in germplasm development for delivery to breeding programs. In the present study, the Rht18 dwarfing gene did not reduce seedling leaf area, biomass or coleoptile length compared with either the tall recurrent parent Halberd or tall RIL siblings, and Rht18 NILs had significantly greater coleoptile length than Rht-B1b NILs. Both ground cover and NDVI were marginally smaller throughout vegetative growth for Rht18 and tall Halberd BC₁-derived RIL progeny, whereas BC₅-derived Rht18 NILs had greater NDVI than Rht-B1b NILs. Vigorous growth and rapid leaf area development are important in achieving a good plant stand and large canopy early in the season. The shorter coleoptile and reduced NDVI of the Rht-B1b NILs is consistent with previous reports for GAI semi-dwarf wheats (e.g. Allan 1989; Rebetzke et al. 2007, 2014), and the neutral effect of Rht18 on coleoptile length and early biomass is consistent with previous observations for other GAR dwarfing genes (e.g. Ellis et al. 2004; Addisu et al. 2009; Rebetzke et al. 2012). In the work of Yang et al. (2015), coleoptile lengths of Rht18 dwarf sibs were, on average, similar to those of tall parents in two populations while being ∼20% shorter in a third population. The GAI Rht-B1b and Rht-D1b dwarfing genes decrease cell elongation to reduce leaf length and shoot biomass (Ellis et al. 2004; Botwright et al. 2005), whereas Rht8 and other GAR dwarfing genes do not affect cell elongation and seedling growth relative to tall, non-dwarfing-gene-containing NILs (Ellis et al. 2004; Rebetzke et al. 2012). Despite the reduction in leaf length, Rht18 was not associated with any reduction in leaf area in either the field or controlled environment studies, and maintenance of seedling leaf area and biomass was consistent across the two contrasting genetic backgrounds. In the study of Yang et al. (2015), coleoptile length was similar for Rht18 sibs and tall parents, whereas Flohr et al. (2021) reported that coleoptiles of Rht18 BC₃ NILs were ∼30% longer and emerged significantly better than their Rht-B1b near-isogenic siblings.

In conclusion, the Rht18 dwarfing allele has been stably transferred from durum to bread wheat to reduce plant height and lodging and increase grain yield across multiple genetic backgrounds. Presence of Rht18 increases harvest index and grain number to increase grain yield, consistent with the established influence of the Green Revolution Rht-B1b dwarfing gene. This increase reflects increases in spike number and increased spike fertility to increase numbers of grains produced in each spike. The Rht18 gene appears to reduce seedling leaf length but this does not compromise leaf area owing to commensurate increases in leaf width and greater tillering early in development. The chromosomal location of the Rht18 gene on the A-genome of wheat and access to linked molecular markers should facilitate ready selection of this allele in bread and durum wheat populations targeting improved adaptation to favourable environments. Further studies are needed to confirm that benefits of this allele extend to improved performance in less favourable environments than those sampled in this study and to develop greater understanding of physiological mechanisms underpinning reduction in plant height by this genetically dominant allele.

The data that support this study will be shared upon reasonable request to the corresponding author.

The authors declare no conflicts of interest.

This research did not receive specific funding that requires acknowledgement.