Introduction

Wheat breeding is a time and resource consuming process, with the period from initial crossing to variety release taking at least 10 generations. Testing for specific quality end-uses such as breads or noodles can add selection pressures as quality testing can require additional resourcing and cause delays in providing data due to the time required to carry out analyses. To carry out traditional flour quality assessment, a milling stage is required followed by rheology tests; although most of these have been scaled down to accommodate the small amounts of grain available in the early stages of breeding programs. While the milling stage is destructive of grain, it is important to know the flour yield potential of breeding lines. The flour yield (FY) from the milling process is required to be as high as possible (78–80 %). After milling, there are a number of rheology tests but one of the most important is water absorption (WA). WA provides data on the amount of water required to produce workable dough, i.e. a dough that holds together without crumbling or being too sticky in the mixing stage. Under Australian baking conditions with high protein hard wheat a WA > 64 % is required.

A highly successful complementary testing technology to the traditional rheology methods is near infrared reflectance spectroscopy (NIRS). NIRS is a rapid and non-destructive technology and has been used for over 40 years to estimate protein and moisture in ground wheat (Osborne 2006). The technology has advanced considerably in both hardware and chemometrics. The improved technology allows for routine, non-destructive estimation of wheat quality parameters, including hardness (Osborne et al. 2001), amylose (Wesley et al. 2003) and flour quality (Osborne 1981). A recent review highlighted some of the more routine applications used for wheat quality selection in breeding programs, including colour, water absorption and flour yield (Osborne 2006). Diepeveen et al. (2010) also showed the value of using only near infrared spectra to identify genetic regions in a wheat mapping population.

The use of NIRS data to predict phenotypic data for subsequent marker analysis has the capacity to increase genetic gains made for quality traits in the early stages of breeding programs. NIRS data provides acceptable phenotypic data to identify genetic regions by molecular marker analysis in barley, including husk (Fox et al. 2006), hardness (Fox et al. 2007) and protein content (Emebiri et al. 2005). For wheat, molecular makers have been reported for hardness (Campbell et al. 2001; Osborne et al. 2001; Groos et al. 2004; Kuchel et al. 2006; Kunert et al. 2007) and protein (Kuchel et al. 2006; Groos et al. 2007) from NIRS data.

For plant breeders, molecular markers provide tools to assist in elucidating the genetics of specific breeding traits. Marker technology also offers a relatively cheap and efficient process to identify breeding lines with suitable quality attributes at early stages within breeding programs, i.e. prior to any field trials. QTL for flour yield were previously identified (Campbell et al. 2001; Kuchel et al. 2006; Kunert et al. 2007; Lehmensiek et al. 2006; McCartney et al. 2006; Parker et al. 1999) and there are several reports for WA using the Farinograph method (Kuchel et al. 2006; Ma et al. 2007; Mann et al. 2009; McCartney et al. 2006; Raman et al. 2009).

Estimation of genetic effects from phenotypic data is improved through use of an appropriate experimental design of the laboratory analysis, as both laboratory and field effects from breeding trials can impact on subsequent quality measurements. Smith et al. (2001) demonstrated an improved statistical analysis of breeding data for wheat quality and malting barley quality, respectively, through processing of samples in a randomised order, as well as with a level of duplication. Panozzo et al. (2007) highlighted the improved QTL results with second phase laboratory design for malting quality analysis.

In this study, genetic loci associated with flour yield and water absorption were investigated using laboratory (FY from milling and Farinograph water absorption) and NIRS data. Results obtained with the two different methods are compared and a summary of QTL associated with flour yield and water absorption identified by both methodologies is presented.

Materials and methods

Plant material

The doubled haploid population of 162 individuals used in this study was produced at the Leslie Research Centre, Department of Agriculture, Fisheries and Forestry Queensland, and was a cross between the two hard wheats, namely, Lang (3AG3/4*Condor//Cook/Sunco) and QT8766 (Banks//2*Parula/Pavon). Both parental lines have the puroindoline alleles Pina-D1a and Pinb-D1b for hardness. Their glutenin profiles differ only by high molecular weight subunits. Lang has subunits 1 (Glu-A1) and 7 + 8 (Glu-B1) while QT8766 has 2* (Glu-A1) and 17 + 18 (Glu-B1). Both parents have the subunits 2 + 12 at the Glu-D1 locus. The semi-dwarf habit is conditioned by Rht1 on chromosome 4B in Lang and by Rht2 on chromosome 4D in QT8766.

This population was grown at Biloela in 2004 (B04) and at Wellcamp (W05) and Biloela (B05) in 2005. The field trials consisted of 162 lines grown in a two-replicate latin row–column design, where each plot was 5 m long × 1.75 m wide. Leaf samples of each line were taken in the field at seedling stage. DNA was extracted using the Wizard® Genomic DNA Purification Kit (Promega Corporation, Australia). The protocol provided by the supplier was followed.

Milling and water absorption

A standard method was used to mill the seed samples of the individual lines of the population. A partially duplicated design block on milling day was used to assign grain from field plots to a randomised milling order. Grain from a proportion of harvested field plots was split into two samples and tested as duplicates in the laboratory design. Flour yield was obtained by milling grain samples of approximately 600 g through a Bühler MLU-202 pneumatic mill by approved AACC method 26-21A (AACC 2000). The samples were conditioned to a moisture content of 15 % prior to milling. A differential conditioning regime, depending on the grain hardness, was used to determine the amount of water needed to condition the samples; the harder the grain the higher the conditioning level required. Water absorption was determined using a 50 g Farinograph (Brabender Duisberg, Germany) as outlined in Method 06-02 (Royal Australian Chemical Institute 2000). Tests were repeated if the mean consistency at maximum development was not within 500 BU ± 30 U. Flour yield and WA will be referred to as laboratory methods throughout this paper.

Near infrared reflectance spectroscopy

A NIR System 5000 (Foss) was used for NIRS estimations. Whole grain samples were scanned using the spectral range 1,100–2,500 nm at 2 nm increments. The spectra were collected in WinISI 1.5 (Foss USA). In-house calibrations (Fox unpublished) were used to estimate wheat quality traits, including FY and WA.

Microsatellite analysis

Primer sequences for the microsatellites used in this study were obtained from the GrainGenes website, http://wheat.pw.usda.gov/GG2/index.shtml. A standard PCR was performed with the reaction mixture consisting of 30 ng DNA, 5 μM of each primer, 100 μM of each dNTP, 1.5 mM MgCl2, 1× PCR buffer (Bioline Pty Ltd., Australia) and 0.05 U BIOTAQ™ Red DNA Polymerase (Bioline Pty Ltd., Australia) in a total volume of 10 μl. The following PCR cycle profile was used: 5 min at 95 °C, followed by 35 cycles of 30 s at 95 °C, 30 s at 50–60 °C (depending on annealing temperature) and 30 s at 72 °C and 10 min at 72 °C. Amplified products were visualized using a Gel-Scan 2000™ (Corbett Life Sciences, Australia).

Linkage map construction and QTL analysis

The program MapManager QTXb20 (Manly et al. 2001) was used to distribute markers into linkage groups using a P value of 0.01 to define linkage groups. RECORD (Van Os et al. 2005) was used to order the markers within linkage groups. The Kosambi function was used to calculate map distances from recombination values. The map was curated according to the methods discussed in Lehmensiek et al. (2005).

The QTL analysis was performed with Windows QTL Cartographer Version 2.5 (Wang et al. 2007). Data were averaged across yearly replicates for composite interval mapping analysis by QTL Cartographer (QTLCart) and results were produced for each year. Permutation tests were run 1,000 times at a significance level of P = 0.001 to determine the significance of linkage to the trait. Map figures were produced with MapChart Version 2.1 (Voorrips 2002). A likelihood of odds (LOD) score >2.0 was considered suggestive and >3.0 was considered significant (P < 0.05).

The naming convention for the identified QTL uses the format ‘QFY. LA/QT-2B’ indicating a QTL for FY, followed by the cross in which the QTL was identified and the chromosome on which the QTL was mapped.

Statistical analysis

The data were analysed using a mixed model including a term for genotype as the main effect of interest, and additional terms for extraneous effects due to field position and laboratory variation. Variance parameters for the random effects were estimated using residual maximum likelihood (REML) (Patterson and Thompson 1977), and best linear unbiased estimates of genotype effects were used as input into the QTL analysis. All phenotypic data were analysed in the package ASReml (Gilmour et al. 1995).

Data were combined across multiple environments and across the traits of FY and WA via a multivariate analysis in a mixed model framework. Genotype × environment effects were fitted as random and a variance structure was used to model the genetic covariance matrix between environments. In a similar way, genetic correlations between traits were estimated at each environment by modelling the trait by genotype covariance structure. All multivariate analyses were undertaken using the fully unstructured variance model in ASReml (Gilmour et al. 1995).

Results

Flour yield and water absorption

Flour yield and water absorption appeared to be normally distributed for this population in all three experiments (Fig. 1). A summary of site and parental means for FY and WA is given for the laboratory and NIRS traits in Table 1. For the 2004 experiment, the FY and WA measured by rheological methods for both parents and population means were higher than for both 2005 experiments; however, this trend was not reflected in the NIRS measurements.

Fig. 1
figure 1

Distribution of flour yields and water absorption in the Lang/QT8766 population for rheology (a) and NIRS data (b)

Full size image
Table 1 Summary of flour yield (FY) and water absorption (WA) data obtained by rheology and NIRS methods for the Lang/QT8766 population from three sites grown over 2 years
Full size table

Transgressive segregation was observed for both traits at all sites, and a high number of individuals within the population exhibited much higher WA than the parental lines (Table 1; Fig. 1).

Genetic correlations

The genetic correlations (ranking) between sites, estimated from a multi-environment analysis, ranged from 0.85 to 0.92 for FY and 0.81 to 0.92 for WA. This demonstrated there was close agreement in genotype ranking across those sites and suggested there was little genotype × environment interaction for both FY and WA within this population when grown at these three environments.

The agreement in genotype rankings between the laboratory and NIRS measurements is best summarised through the genetic correlation values. At B04, the genetic correlation values were 0.91 and 0.74 for FY and WA, respectively. At B05, the genetic correlation values were 0.87 and 0.90, respectively, and at W05, 0.78 and 0.77, respectively. These correlations were high for both FY and WA in all three experiments.

At the trait level, there were negative genetic correlations between FY and WA in the laboratory data where values were −0.46, −0.59 and −0.60 for B04, B05 and W05, respectively.

Linkage analysis

The genetic linkage map for the Lang/QT8766 population consists of 172 markers and a total map distance of 2,045 cm (Fig. 2). A lack of polymorphic markers resulted in a low number of markers mapped on some chromosomes with the lowest number being two markers on chromosome 7A. The highest number of markers (16) mapped to chromosome 1B. The average distance between markers was 11.9 cm. Eleven and eight chromosomal regions were identified for FY and WA, respectively.

Fig. 2
figure 2figure 2

Composite map (Appels 2003) showing approximate locations of flour yield (white box) and water absorption (black box) QTL identified to date

Full size image

QTL analysis

Quantitative trait loci were identified for FY and WA using data obtained both by laboratory and by NIRS analysis. Permutation tests were run to determine which of the QTL were significant. Significant QTL (LOD > 3.0 at P < 0.05) are indicated in bold in Table 2, whereas suggestive QTL (LOD > 2.0) are indicated in italic. Eleven chromosomal regions were identified for FY. Eighteen FY QTL (P < 0.05) were identified using both the laboratory and NIRS data (Table 2); however, only four were common with both the laboratory and NIRS data (2A, 2B, 4B and 4D). Only the QTL on 4B and 4D were detected across all sites. The FY QTL on chromosome 4D was significant (P < 0.05) at all three sites and for both laboratory and NIRS data. One other QTL (3D) was detected in more than 1 year with the laboratory data, whereas two and one QTL were detected across 2 and 3 years, respectively, using the NIRS data (1D, 2D and 5A). The QTL on 4B and 4D had the highest LOD scores and phenotypic variances explained and were associated with the height genes. Interestingly, the 4B QTL was not detected with the NIRS data for the B05 and W05 trials. The QTL on 2B, 4A, 4D and 6D were contributed by Lang, whereas the remaining QTL were contributed by QT8766 (Table 2).

Table 2 Summary of QTL identified for flour yield (FY) and water absorption (WA) using rheology (laboratory) and NIRS data
Full size table

For WA, eight chromosomal regions were identified for both laboratory and NIRS data. There was only one significant (P < 0.05) QTL for laboratory data, namely chromosome 4D at B04 (Table 2). The NIRS data provided six significant QTLs with the most significant being on 4D. Suggestive QTL on 4A and 5B were detected in more than one experiment using NIRS data. With the exception of the QTLs on 1B and 4D, all QTL were contributed by Lang (Table 2).

While there were a number of QTL for each trait using each method, only the regions on 4BL and 4DS could suggest some genetic × environment interaction (Table 2). However, both of these loci were associated with plant height.

Discussion

This study identified FY and WA QTL in the Lang/QT8766 population using data from laboratory methods compared to QTL obtained by the NIRS data. Similar QTL were identified by both methods, suggesting the NIRS data could provide phenotypic data for marker analysis faster than any laboratory methodology.

Eleven and eight chromosomal regions were identified for FY and WA, respectively. In other studies, between one and five QTL were identified for FY and between one and four for WA. The FY QTL on 2A, 2B, 4B, 4D and 6A identified in our study were in similar regions to FY QTL identified previously (Campbell et al. 2001; Kuchel et al. 2006; Kunert et al. 2007; Lehmensiek et al. 2006; McCartney et al. 2006; Parker et al. 1999). The WA QTL on 4D was in a similar region to a WA QTL identified in the spring wheat cross RL4452/‘AC Domain’ (McCartney et al. 2006) and the WA QTL on 1B and 4A were in the same region as WA QTL identified in a Chara/WW2449 population (Raman et al. 2009). The 4A WA QTL coincided with QTL for grain protein content and flour protein content in the Chara/WW2449 population (Raman et al. 2009).

The most significant LOD scores (P < 0.05) and highest phenotypic variances explained for both FY and WA were observed with the QTL on chromosome 4D. These QTL were in the same interval as the plant height QTL thus suggesting they were associated with Rht-D1. Huang et al. (2006), McCartney et al. (2006) also found a major cluster of quality-related QTL on chromosome 4D in the same interval as the height QTL. In the McCartney et al. (2006) study, reduced plant height was strongly associated with quality traits such as grain and flour protein content, Farinograph water absorption, reduced loaf height and reduced loaf volume. However, no association was observed between this height gene and a reduction in FY. In our study, the QTL on chromosome 4B was significant (P < 0.05) with the laboratory data in the three trials but was only significant (P < 0.05) in one trial with the NIRS data. Of the 19 QTL identified for FY and WA, only five were detected by both the laboratory and NIRS data.

Protein content is a key parameter for selection in bread wheat breeding, and has been positively associated with water absorption. In our study we did not include protein data, but the QTL on chromosome 4A for water absorption was in the same region as a QTL reported for protein content by Raman et al. (2009). McCartney et al. (2006) reported a positive association between protein content (grain and flour) and Farinograph water absorption. Ma et al. (2007) also showed an association between protein content and WA. These reports suggest that selecting for higher protein content could result in higher water absorption. As expected, the negative genetic correlations between WA and FY also suggest that high FY may not be obtained together with high protein, but to have ideal baking quality, high protein would be important. In addition, growing breeding material in higher nitrogen environments would also be beneficial for quality testing.

Flour yield is one of the most important traits in wheat breeding and from our results and others mentioned above, QTL for FY have been detected on all 21 wheat chromosomes. This suggests for FY, QTL expression is dependent upon the genetic background. The high genetic correlations for FY suggest the selection of appropriate genetic background would help in breeding material with potentially higher FY. Thus breeding programs need to base their marker selection strategies for this trait within pedigree groupings of regionally important germplasm or varieties.

The rapid through-put of NIRS could allow breeders to develop larger populations and to grow populations in additional environments. Coupled with the use of markers, early screening to identify doubled haploid lines with alleles of interest could add value to selection prior to laboratory testing, thereby saving time and resources. In wheat programs, NIRS is routinely used for estimating protein content, but if laboratories are prepared to develop calibration for flour quality traits (FY and WA), then the power of the technology will be greatly enhanced.