Economical optimization of a breeding scheme by selective phenotyping of the calibration set in a multi-trait context: application to bread making quality.

KEY MESSAGE: Trait-assisted genomic prediction approach is a way to improve genetic gain by cost unit, by reducing budget allocated to phenotyping or by increasing the program's size for the same budget. This study compares different strategies of genomic prediction to optimize resource allocation in breeding schemes by using information from cheaper correlated traits to predict a more expensive trait of interest. We used bread wheat baking score (BMS) calculated for French registration as a case study. To conduct this project, 398 lines from a public breeding program were genotyped and phenotyped for BMS and correlated traits in 11 locations in France between 2000 and 2016. Single-trait (ST), multi-trait (MT) and trait-assisted (TA) strategies were compared in terms of predictive ability and cost. In MT and TA strategies, information from dough strength (W), a cheaper trait correlated with BMS (r = 0.45), was evaluated in the training population or in both the training and the validation sets, respectively. TA models allowed to reduce the budget allocated to phenotyping by up to 65% while maintaining the predictive ability of BMS. TA models also improved the predictive ability of BMS compared to ST models for a fixed budget (maximum gain: + 0.14 in cross-validation and + 0.21 in forward prediction). We also demonstrated that the budget can be further reduced by approximately one fourth while maintaining the same predictive ability by reducing the number of phenotypic records to estimate BMS adjusted means. In addition, we showed that the choice of the lines to be phenotyped can be optimized to minimize cost or maximize predictive ability. To do so, we extended the mean of the generalized coefficient of determination (CDmean) criterion to the multi-trait context (CDmulti).

Genomic regions associated with the nitrogen limitation response revealed in a global wheat core collection.

Modern wheat (Triticum aestivum L.) varieties in Western Europe have mainly been bred, and selected in conditions where high levels of nitrogen-rich fertilizer are applied. However, high input crop management has greatly increased the risk of nitrates leaching into groundwater with negative impacts on the environment. To investigate wheat nitrogen tolerance characteristics that could be adapted to low input crop management, we supplied 196 accessions of a wheat core collection of old and modern cultivars with high or moderate amounts of nitrogen fertilizer in an experimental network consisting of three sites and 2 years. The main breeding traits were assessed including grain yield and grain protein content. The response to nitrogen level was estimated for grain yield and grain number per m(2) using both the difference and the ratio between performance at the two input levels and the slope of joint regression. A large variability was observed for all the traits studied and the response to nitrogen level. Whole genome association mapping was carried out using 899 molecular markers taking into account the five ancestral group structure of the collection. We identified 54 main regions involving almost all chromosomes that influence yield and its components, plant height, heading date and grain protein concentration. Twenty-three regions, including several genes, spread over 16 chromosomes were involved in the response to nitrogen level. These chromosomal regions may be good candidates to be used in breeding programs to improve the performance of wheat varieties at moderate nitrogen input levels.

Characterisation of polymorphic microsatellite markers from Aegilops tauschii and transferability to the D-genome of bread wheat.

Microsatellites were isolated from a Aegilops tauschii (the D-genome donor of bread wheat) library enriched for various motifs. Primers generated from the flanking region of the microsatellites were used successfully to amplify the corresponding loci in the D genome of bread wheat. Additional amplification sometimes also occurred from the A and B genomes. The majority of the microsatellites contained (GA)(n) and (GT)(n) motifs. GA and GT repeats appeared to be both more abundant in this library and more polymorphic than other types of repeats. The allele number for both types of dinucleotide repeats fitted a Poisson distribution. Deviance analysis showed that GA and GT were more polymorphic than other motifs in bread wheat. Within each motif type (di-, tri- and tetra-nucleotide repeats), repeat number has no influence on polymorphism. The microsatellites were mapped using the Triticum aestivum Courtot x Chinese Spring mapping population. A total of 100 markers was developed on this intraspecific map, mainly on the D genome. For polyploid species, isolation of microsatellites from an ancestral diploid donor seems to be an efficient way of developing markers for the corresponding genome in the polyploid plant.

Study of the relationship between pre-harvest sprouting and grain color by quantitative trait loci analysis in a whitexred grain bread-wheat cross.

In many wheat ( Triticum aestivumL.) growing areas, pre-harvest sprouting (PHS) may cause important damage, and in particular, it has deleterious effects on bread-making quality. The relationship between PHS and grain color is well known and could be due either to the pleiotropic effect of genes controlling red-testa pigmentation ( R) or to linkage between these genes and other genes affecting PHS. In the present work, we have studied a population of 194 recombinant inbred lines from the cross between two cultivars, 'Renan' and 'Récital', in order to detect QTLs for both PHS resistance and grain color. The variety 'Renan' has red kernels and is resistant to PHS, while 'Récital' has white grain and is highly susceptible to PHS. A molecular-marker linkage map of this cross was constructed using SSRs, RFLPs and AFLPs. The population was evaluated over 2 years at Clermont-Ferrand (France). PHS was evaluated on mature spikes under controlled conditions and red-grain color was measured using a chromameter. Over the 2 years, we detected four QTLs for PHS, all of them being co-localized with QTLs for grain color. Three of them were located on the long arm of chromosomes 3 A, 3B and 3D, close to the loci where the genes R and taVp1 were previously mapped. For these three QTLs, the resistance to PHS is due to the allele of the variety 'Renan'. Another co-located QTL for PHS and grain color was detected on the short arm of chromosome 5 A. The resistance for PHS for this QTL is due to the allele of 'Récital'.

Comparative mapping of the wheat 5B short chromosome arm distal region with rice, relative to a crossability locus.

Colinearity between wheat and rice genomes is quite well established at the chromosome level, but less is known at a finer level. We tried to specify these relationships for the wheat 5BS chromosome-arm distal region, where a major locus for crossability was located. By developing AFLP markers, we succeeded to locate this major QTL more precisely. One cloned AFLP fragment mapped to rice chromosome 11, which was in agreement with a rice chromosome-11 linkage block reported in this region. However a second marker, a RFLP probe, showed a break in synteny because it mapped to rice long-arm chromosomes 1 and 5, while screening a rice BAC library with the same probe identified rice chromosomes 5 and 6. Therefore, we concluded that the syntenic relationships were more complex at the fine level. The observed results might indicate the presence of a linkage block carrying a crossability gene on wheat groups 1, 5 and 7, and also on rice chromosomes 5 and 6.

Genetic analysis of bread-making quality scores in bread wheat using a recombinant inbred line population.

Bread-making quality has been evaluated in a progeny of 194 recombinant inbred lines (RILs) from the cross between the two French cultivars Récital and Renan, cultivated in three environments. These cultivars have been previously identified as having contrasting grain protein content and dough rheology properties, although they achieve similar scores for the official bread-making test used for cultivar registration in France. However the progeny displayed a wide range of variations, suggesting that favourable alleles at several loci are present in the two parental lines. Correlation analyses revealed that bread-making scores are poorly correlated among environments, as they are poorly predicted by multiple regression on dough rheology parameters and flour-protein content. However, loaf volume was the most heritable and predictable trait. A total of seven QTLs were found for bread scores, each explaining 5.9-14.6% of trait variation and six for the loaf volume (10.7-17.2%). Most bread-making QTLs, and particularly those detected in all environments, co-located with QTLs for dough rheology, protein content or flour viscosity due to soluble pentosans (Fincher and Stone 1986; Anderson et al. in J Cereal Sci 19:77-82, 1994). Some QTL regions such as those on chromosome 3A and chromosome 7A, which display stable QTLs for bread-making scores and loaf volume, were not previously known to host obvious genes for grain quality.

Doubled haploid versus S1 family recurrent selection for testcross performance in a maize population.

Theoretically, in a recurrent selection program, the use of doubled haploids (DH) can increase genetic advance per unit of time. To evaluate the efficiency expected from the use of DH for the improvement of grain yield in a maize (Zea mays L.) population, two recurrent selection programs for testcross performance were initiated using testcross progenies from DH lines and S1 families. In 4 years one selection cycle using DH and two selection cycles using S1 families were carried out with the same selection intensity for both methods. As expected, testcross genetic variance was twice as high among DH lines as among S1 families. The predicted genetic gain was 8.2% for the DH selection cycle, and 10.6% for the two S1 selection cycles, giving a per year advantage of 29% for the S1 family method over the DH method with a cycle of 4 years. With a 3-year cycle for the DH method, both methods were expected to be equivalent. Using a tester related to the one used for selection, the genetic gains obtained were equivalent for both methods: 6.6% for the DH cycle and 7.0% for the two S1 cycles. With a 3-year cycle for the DH method, the advantage would have been in favor of DH method. Furthermore, the DH method has the advantage of simultaneously producing lines that are directly usable as parents of a hybrid. Thus, if the genetic advance per unit of time is evaluated at the level of developed varieties even with the same or with a lower genetic advance in population improvement, the DH method appears to be the most efficient.

Genetic analysis of dry matter and nitrogen accumulation and protein composition in wheat kernels.

The maximum rate and duration for grain dry matter (DM) and nitrogen (N) accumulation were evaluated in 194 recombinant inbred lines (RILs) from a cross between the two French wheat cultivars Récital and Renan. These cultivars were previously identified as having contrasting kinetics of grain DM and N accumulation. Grain protein composition was analysed by capillary electrophoresis (CE), which enabled quantification of the different storage protein fractions (alphabetagamma-gliadins, omega-gliadins, LMW glutenins, HMW glutenins, and each of their subunits). Correlation analyses revealed that DM and N accumulation rates were closely correlated and repeatable over several years, which was not the case for DM and N accumulation durations, and that protein composition was primarily influenced by the N accumulation rate. This was particularly true for the LMW-glutenins and the alphabetagamma-gliadins, the most abundant protein fractions. A genetic map of 254 molecular markers covering nearly 80% of the wheat genome was used for quantitative trait loci (QTL) analysis. A total of seven QTLs were found. Five QTLs were significantly associated with the kinetics of DM and N accumulation, and two of them also influenced protein composition. Two QTLs affected only the protein composition. One major QTL explained more than 70% of the total variation in HMW-GS Glu1B-x content.

Diallel analysis of androgenetic plant production in hexaploid Triticale (X. triticosecale, Wittmack).

Studies were made on the genetic determination of androgenetic plant yield and its two components: embryo production and green plant regeneration. This involved the analysis of a complete 7×7 diallel cross of 4 androgenetic lines and 3 lines obtained by pedigree selection, one of them having the Triticum timopheevi cytoplasm. The three traits analysed are both heritable and environmentally influenced (by season and culture medium composition). The analysis of embryo production shows a mainly nuclear inheritance, with predominantly additive gene action, but also a favourable effect of Triticum timopheevi cytoplasm. Green plant regeneration has a more complex genetic determination, with additive as well as non-additive gene action and cytoplasmic influences. Hybrids appear superior to inbred lines for embryogenesis and green plant yield, but not for green plant regeneration. Androgenetic lines used as parents did not show superiority over other parents either in their own value or in the transmission of androgenetic abilities. Genetic improvement seems to be possible by recombination in progenies of hybrids between lines having complementary abilities.

A comparison of androgenetic doubled-haploid, and single seed descent lines in Triticale.

Sixty single seed descent (SSD) lines and about 25 anther-derived doubled-haploid (DH) lines were obtained from two triticale crosses. The frequency distributions of 10 quantitative agronomic traits were compared using parametric and non-parametric tests. A multivariate discriminant analysis was subsequently carried out. Gliadin patterns obtained from each line by polyacrylamide gel electrophoresis were used to calculate intra- and inter-population diversities from relative dissimilarity indices. It was found that DH and SSD lines show significant differences in frequency distributions of 1000 grain weight in both crosses, of heading date for one cross, and of lodging susceptibility for the other cross. The results of intra- and inter-population gliadin diversity indicate that although the SSD method theoretically provides more opportunity for recombination to occur than the DH method, it did not produce a greater range of recombinants. Since there is no significant difference between SSD- and DH-line distributions for grain yield, anther culture appears to be an efficient method for producing high yielding homozygous lines from F1 hybrids of triticale in a relatively short time.

Isozyme variation and species relationships in the genus Lolium L. (ryegrasses, Graminaceae).

Thirty-two natural populations belonging to the eight species of the genus Lolium (ryegrass) or to Festuca pratensis (meadow fescue) were recorded for allelic frequencies at 13 isozyme loci. Cultivated ryegrass (L. perenne and L. multiflorum), meadow fescue, and the annual L. rigidum, are true outbreeders. The other species are true inbreeders, except for L. canariense, which shows a moderate level of cross fertilisation (20%). Hierarchical clustering from Nei's unbiased distance leads to four groups. The three self-pollinating, weed species, L. temulentum, L. remotum and L. persicum, belong to the first cluster, which is the most differentiated one. The second cluster comprises L. multiflorum, L. subulatum and most populations of L. rigidum. All L. perenne populations belong to the third cluster, as do two of L. rigidum. The average genetic distance within the L. perenne group is very low. Surprisingly, the fourth cluster groups together L. canariense and Festuca pratensis. The data suggest that L. rigidum is the species with the greatest diversity, and could be a common ancestor of the genus. Knowledge of historical processes of domestication could help to calibrate the molecular clock.

Geostatistics for spatial genetic structures: study of wild populations of perennial ryegrass.

Methods based on geostatistics were applied to quantitative traits of agricultural interest measured on a collection of 547 wild populations of perennial ryegrass in France. The mathematical background of these methods, which resembles spatial autocorrelation analysis, is briefly described. When a single variable is studied, the spatial structure analysis is similar to spatial autocorrelation analysis, and a spatial prediction method, called "kriging", gives a filtered map of the spatial pattern over all the sampled area. When complex interactions of agronomic traits with different evaluation sites define a multivariate structure for the spatial analysis, geostatistical methods allow the spatial variations to be broken down into two main spatial structures with ranges of 120 km and 300 km, respectively. The predicted maps that corresponded to each range were interpreted as a result of the isolation-by-distance model and as a consequence of selection by environmental factors. Practical collecting methodology for breeders may be derived from such spatial structures.

Hierarchical clustering of perennial ryegrass populations with geographic contiguity constraint.

An algorithm of automatic classification is proposed and applied to a large collection of perennial ryegrass wild populations from France. This method is based on an ascendant hierarchical clustering using the Euclidian distance from the principal components extracted from the variance-covariance matrix between 28 agronomic traits. A contiguity constraint is imposed: only those pairs of populations which are defined as contiguous are grouped together into a cluster. The definition of contiguity is based on a geostatistical parameter: the range of the variogramme, i.e. the largest distance above which the variance between pairs of population no longer increases. This method yields clusters that are generally more compact than those obtained without constraint. In most cases the contours of these clusters fit well with known ecogeographic regions, namely, for macroclimatic homogeneous conditions. This suggests that selective factors exert a major influence in the genetic differentiation of ryegrass populations for quantitatively inherited adaptive traits. It is proposed that such a method could provide useful genetic and ecogeographic bases for sampling a core collection in widespread wild species such as forage grasses.

Genetic analysis of grain protein-content, grain yield and thousand-kernel weight in bread wheat.

Grain yield and grain protein content are two very important traits in bread wheat. They are controlled by genetic factors, but environmental conditions considerably affect their expression. The aim of this study was to determine the genetic basis of these two traits by analysis of a segregating population of 194 F(7) recombinant inbred lines derived from a cross between two wheat varieties, grown at six locations in France in 1999. A genetic map of 254 loci was constructed, covering about 75% of the bread wheat genome. QTLs were detected for grain protein-content (GPC), yield and thousand-kernel weight (TKW). 'Stable' QTLs (i.e. detected in at least four of the six locations) were identified for grain protein-content on chromosomes 2A, 3A, 4D and 7D, each explaining about 10% of the phenotypic variation of GPC. For yield, only one important QTL was found on chromosome 7D, explaining up to 15.7% of the phenotypic variation. For TKW, three QTLs were detected on chromosomes 2B, 5B and 7A for all environments. No negative relationships between QTLs for yield and GPC were observed. Factorial Regression on GxE interaction allowed determination of some genetic regions involved in the differential reaction of genotypes to specific climatic factors, such as mean temperature and the number of days with a maximum temperature above 25 degrees C during grain filling.

Genotype x environment interactions in a core collection of French perennial ryegrass populations.

A representative sample (core collection) of natural populations of perennial ryegrass (Lolium perenne L.) from France was evaluated for agronomic traits at seven locations. This sample exhibits a high level of genotype-environment interaction for most traits. The interactions for summer-growth (a key-factor of adaptation in most French regions) were studied by means of regression using climatic factors of the evaluation sites and the sites of population origin as covariates. This method succeeded in explaining most of the interaction term and also part of the main effects. It appears that populations from either warm or dry sites generally have a positive interaction when evaluated in a site with similar characteristics, as expected as a consequence of natural selection. A population component of regression on environmental covariates, however, was significant and could be exploited through breeding to improve adaptation of perennial ryegrass to either drier or warmer regions.

Mapping QTLs for grain hardness and puroindoline content in wheat ( Triticum aestivum L.).

Genes for puroindoline-a (Pin-a), puroindoline-b (Pin-b) and grain-softness proteins (GSP) have been shown to be linked to the dominant Ha locus responsible for the soft texture of the grain. Though linkage has been demonstrated of the puroindoline genes to the Ha locus, there is no clear evidence that puroindoline content is the product of the gene Ha. A segregating population of 115 recombinant inbred lines (RILs) originating from a cross between the hexaploid Synthetic wheat ( Triticum durum x Aegilops tauschii, W 7984) and the cultivar 'Opata' (M 85) was studied in two different experimental years to detect Quantitative Trait Loci (QTLs) for three traits: grain hardness (Hard), puroindoline-a (Pin-a) and puroindoline-b (Pin-b) contents. The detection of QTLs was performed using marker linear regression. Negative correlation coefficients (-0.86 and -0.80) were identified between grain hardness and puroindoline content (a and b, respectively) on data obtained in 1996. Results obtained in 1999 confirmed the negative correlation between Hard and Pin-a (-0.73); however a positive correlation coefficient was found with Pin-b content (0.41). Total phenotypic variation explained by each QTL was calculated (R2). For each of the Hard, Pin-a and Pin-b traits one major QTL was detected on the short arm of chromosome 5D, located close to the mta9 allele (puroindoline-a). For the first year (1996) the QTL in this region explained around 63% of the phenotypic variability in grain hardness, 77% in Pin-a and 45% in Pin-b contents. These values were confirmed in trials carried out in 1999 with a R2 value of 0.71, 0.72 and 0.25 for Hard, Pin-a and Pin-b, respectively. In 1996 and 1999 a second major QTL was detected for grain hardness on the long arm of the same chromosome. Present results indicate that it cannot be definitely concluded that puroindoline content represents a linear explanation for variations in grain hardness.

A high-density molecular map for ryegrass (Lolium perenne) using AFLP markers.

AFLP markers have been successfully employed for the development of a high-density linkage map of ryegrass (Lolium perenne L.) using a progeny set of 95 plants from a testcross involving a doubled-haploid tester. This genetic map covered 930 cM in seven linkage groups and was based on 463 amplified fragment length polymorphism (AFLP) markers using 17 primer pairs, three isozymes and five EST markers. The average density of markers was approximately 1 per 2.0 cM. However, strong clustering of AFLP markers was observed at putative centromeric regions. Around these regions, 272 markers covered about 137 cM whereas the remaining 199 markers covered approximately 793 cM. Most genetic distances between consecutive pairs of markers were smaller than 20 cM except for five gaps on groups A, C, D, F and G. A skeletal map with a uniform distribution of markers can be extracted from this high-density map, and can be applied to detect and map QTLs. We report here the application of AFLP markers to genome mapping, in Lolium as a prelude to quantitative trait locus (QTL) identification for diverse agronomic traits in ryegrass and for marker-assisted plant breeding.

Molecular variability of sunflower downy mildew, Plasmopara halstedii, from different continents.

Downy mildew of sunflower (Helianthus annuus L.), caused by the pathogen Plasmopara halstedii, is a potentially devastating disease. Seventy-seven isolates of P. halstedii collected in twelve countries from four continents were investigated for RAPD polymorphism with 21 primers. The study led to a binary matrix, which was subjected to various complementary analyses. This is the first report on the international genetic diversity of the pathogen. Similarity indices ranged from 89% to 100%. Neither a consensus unweighted pair group method with arithmetic means (UPGMA) tree constructed after bootstrap resampling of markers nor a principal component analysis based on distance matrix revealed very consistent clusterings of the isolates, and groups did not fit race or geographical origins. Phylogenies were probably obscured by limited diversity. Analysis of molecular variance (AMOVA) and Nei's genetic diversity statistics gave similar conclusions. Most of the genetic diversity was attributable to individual differences. The most differentiated races also had the lowest within-diversity indices, which suggest that they appeared recently with strong bottleneck effects. Our analyses suggest that this pathogen is probably homothallic or has an asexual mode of reproduction and that gene flow among countries can occur through commercial exchanges. Knowledge of the downy mildew populations' structure at the international level will help to devise strategies for controlling this potentially devastating disease.

Detection of QTLs for heading time and photoperiod response in wheat using a doubled-haploid population.

The genetic basis of heading time in wheat (Triticum aestivum L.) was investigated through the study of flowering under normal autumn sown field conditions as well as photoperiod responses under a controlled environment. Quantitative trait loci (QTLs) for these traits were mapped in a doubled-haploid (DH) population derived from a cross between the wheat cultivars 'Courtot' and 'Chinese Spring'. A molecular marker linkage map of this cross that was previously constructed based on 187 DH lines and 380 markers was used for QTL mapping. The genome was well covered (85%) except for chromosomes 1D and 4D, and a set of anchor loci regularly spaced over the genome (one marker each 15.5 cM) was chosen for marker regression analysis. The presence of a QTL was declared at a significance threshold of alpha = 0.005. The population was grown under field conditions in Clermont-Ferrand, France during two years (1994-1995), in Norwich, U.K. over one year (1998), and also under controlled environments in Norwich. For each trait, between 2 and 4 QTLs were identified with individual effects ranging between 6.3% and 44.4% of the total phenotypic variation. Two QTLs were detected that simultaneously affected heading time and photoperiod response. For heading time, these two QTLs were detected in more than one year. One QTL located on chromosome arm 2BS near the locus Xfbb121-2B, co-segregated with the gene Ppd-B1 known to be involved in photoperiod response. This chromosome region explained a large part of the variation (23.4-44.4% depending on the years or the traits). Another region located on chromosome arm 7BS between the loci Xfbb324-7B and Xfbb53-7B also had a strong effect (7.3-15.3%). This region may correspond to a QTL for earliness per se.

An update of the Courtot x Chinese Spring intervarietal molecular marker linkage map for the QTL detection of agronomic traits in wheat.

We made an update of the intervarietal molecular marker linkage map of the wheat genome developed using a doubled-haploid (DH) population derived from the cross between the cultivars "Courtot" and "Chinese Spring". This map was constructed using 187 DH lines and 659 markers. The genome was well covered (more than 95%) except for chromosomes from homoeologous group 4 and chromosomes 5D and 7D, which had gaps slightly larger than 50 cM. A core-map based on a set of 200 anchor loci (one marker each 18.4 cM) was developed. The total length of this map was 3,685 cM which is similar to the size of the international reference map of the ITMI population (3,551 cM). Map coverage was identical for the three genomes (A, B and D) and for the number of anchor loci, as well as for the size of the map. Using this map, QTLs for several agronomic traits were detected on phenotypic data from the population grown in Clermont-Ferrand (France) under natural field conditions over 6 years, and in Norwich (UK) in controlled conditions and under natural field conditions in 1 year. Almost all of the 21 chromosomes were involved in at least one trait. However, several regions seemed to contain gene clusters either for grain traits (and thus bread-making quality) or plant development traits.