Published online 27 May 2005
Published in Crop Sci 45:1427-1431 (2005)
© 2005 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
CROP BREEDING, GENETICS & CYTOLOGY
Indirect versus Direct Selection of Winter Wheat for Low-Input or High-Input Levels
M. Brancourt-Hulmela,*,
E. Heumeza,
P. Plucharda,
D. Beghina,
C. Depatureauxb,
A. Giraudc and
J. Le Gouisa
a INRA, Unité de Génétique et d'Amélioration des Plantes, BP 136, Estrées-Mons, 80203 Péronne Cédex, France
b INRA, Station de génétique végétale, Ferme du Moulon, 91190 Gif-sur-Yvette, France
c INRA Station de Génétique et d'Amélioration des Plantes, 17 rue de Sully, BP 86510, 21065 Dijon Cédex, France
* Corresponding author (brancour{at}mons.inra.fr)
 |
ABSTRACT
|
|---|
Market prices and environmental concerns favor low-input wheat (Triticum aestivum L.) production systems. This study assesses the efficiency of low-input vs. high-input selection environments to improve wheat for low-input environments. Three standard cultivars, 11 parents, and 270 lines bred in INRA Mons-Péronne were investigated for 2 yr (1998, 1999) in France at three INRA locations. Four agronomic treatments combining two levels of fungicides with two levels of nitrogen (N) were applied. Because of seed supply, only 10 year x treatment x location combinations were conducted. Broad-sense heritabilities for grain yield (GY) ranged from 0.18 at low N without fungicide to 0.90 at high N without fungicide. Heritability estimates were higher at high N than at low N level. This was due to both an increase in error variance and a decrease in genetic variance at low nitrogen level. Heritabilities in treatments without fungicide were the same or higher than heritabilities measured in the corresponding controlled treatments. Broad-sense heritabilities for grain N content (GNC) were similar between the controlled treatments with fungicide and the corresponding treatments without fungicide. They were lower at low N level and this was due to an increased error variance in both years. Genetic correlations between the 10 experiments were always positive for GY and N content: they ranged from 0.10 to 0.95 for yield and from 0.78 to 0.98 for GNC. The relative efficiency of indirect selection to direct selection for each pair of environments ranged from 0.15 to 0.99 indicating that indirect selection was never more efficient than direct selection. Therefore, breeding programs targeting low-input environments should include low-input selection environments to maximize selection gains.
Abbreviations: GNC, grain nitrogen content • GY, grain yield • N, nitrogen
|
INTRODUCTION
|
|---|
GROWING ENVIRONMENTAL CONCERNS and lower prices in Europe tend to favor agricultural systems with lower input levels. Studies on genetic gain have shown that, though genetic progress exists at all input levels, genetic gain is lower under low-input levels. Ortiz-Monasterio et al. (1997) compared CIMMYT wheat varieties released between 1950 and 1985, including semidwarf lines development. They showed that genetic gain for GY increased with the level of applied N. Brancourt-Hulmel et al. (2003) compared varieties released in France between 1946 and 1992 and found that the genetic gain was higher at high levels of N with fungicide than at high levels of N without fungicide or at lower levels of N with or without fungicide. In a study of 22 U.S. wheat cultivars released between 1874 and 1988, Shroyer and Cox (1993) found no genetic improvement in performance under low-fertility conditions during that century.
This situation may result from breeding that has been conducted, either under high- or low-input levels, including N fertilizer and fungicides. In France, wheat breeding has been mainly conducted under a high level of inputs. Genetic gain measured under low-input levels seen in the target environments was then due to indirect selection. The relative gain of indirect versus direct selection, considering equal selection intensities, depends on heritabilities at both input levels and genetic correlation between input levels (Falconer, 1974). Heritabilities are generally lower under low-input level or in stressed environments than under high-input level (Ud-Din et al., 1992; Calhoun et al., 1994; Bänziger et al., 1997; Bertin and Gallais, 2000; Sinebo et al., 2002) but contradictory results exist (Agrama et al., 1999). Atlin and Frey (1989) and Presterl et al. (2003) found no consistent relationship between heritability and mean yield of the selection environment. Lower heritabilities at low input levels have been related to lower genetic variance (Ud-Din et al., 1992; Calhoun et al., 1994; Bänziger et al., 1997) and an increased error variance (Ud-Din et al., 1992; Bertin and Gallais, 2000).
The level of genetic correlation between two environments has been shown to greatly vary, but it mainly depends on the character studied, the genetic material, and the type of stress as well as its intensity. For example, Atlin and Frey (1989) found a lower genetic correlation between phosphorus deficient and nonstress environments than between N deficient and nonstress environments for GY of 116 random oat lines. Bänziger et al. (1997) have shown that the genetic correlation between GY of maize (Zea mays L.) under low and high N levels decreased with increasing N stress intensity, which was estimated by the relative yield reduction under low N. Cooper et al. (1997) found similar results with water stress intensity in wheat.
Based on these studies, indirect selection was predicted to be either as efficient (Atlin and Frey, 1989), more efficient (Calhoun et al., 1994), or less efficient (Ceccarelli et al., 1992; Sinebo et al., 2002) than direct selection in the target environment.
Another case of indirect selection is when selection is applied to a secondary character and the correlated response assessed on a primary character in the same environment. Hill et al. (1999) found that improving wheat yield by selecting for resistance to yellow rust (caused by Puccinia striiformis Westend.) in environments infected by yellow rust was more efficient than selecting for yield itself. The secondary character, yellow rust reaction, had substantially higher narrow-sense heritability than the primary character. In fungicide-treated plots, indirect selection was only 73% as efficient as direct selection. The authors suggested that indirect selection for yield may be the more effective under higher stress levels. In forage maize, Gallais (1984) showed that selection on some combinations of morphological traits is expected to be as efficient as direct selection for dry matter. Morphological components of yield are generally well correlated with yield and have higher heritabilities.
Indirect selection may be alternated with direct selection in breeding schemes. Comparing five selection regimes in wheat, Van Ginkel et al. (2001) showed that alternating between high and low N inputs, with the first segregating generation (F2) having been grown under high N input, resulted in the highest yields at intermediate and high N levels.
Our objective was to investigate genetic parameters, genetic and error variances, heritabilities and genetic correlations for GY and GNC of winter wheat at different N and fungicide levels to assess the relative efficiency of indirect and direct selection.
|
MATERIALS AND METHODS
|
|---|
The plant material consisted of three standard cultivars (Récital, Record, Ritmo), 11 parents (Arche, Audace, Castan, Renan, Sidéral, Soissons, Thésée, Trémie, VM014, VM201, VM202), and 270 breeding lines. The standard cultivars and parents were all varieties registered in France except VM014, VM201, and VM202, which are breeding lines from INRA Mons-Péronne breeding program. The 270 breeding lines were derived from 21 simple and three-way crosses involving the 11 parents. The F2 and F3 plants were raised in INRA Mons-Péronne in nurseries receiving 50 kg ha–1 N without fungicide treatment. A visual selection for highly heritable characteristics such as earliness, plant height, and disease resistance was performed. Seeds harvested on a selected F2 plant were sown as a row. Selected F3 rows were harvested in bulk and seeds used to grow 1998 trials. Seeds from a F4 plot in Mons were used to grow 1999 trials.
Experiments were performed at three INRA stations: Dijon (47°14' N, 5°07' E), Le Moulon (48°42' N, 2°08' E), and Mons (49°53' N, 3°00' E). A total of ten experiments were conducted (Table 1) including four experiments under high N with fungicides, three under high N without fungicide, two under low N with fungicides, and one under low N without fungicide. Soil N availability varied from 54 kg ha–1 to 154 kg ha–1 and total N supply from 55 to 259 kg ha–1 (Table 1). All breeding lines were tested in each experiment except in 1998 at Le Moulon where only 225 lines were grown because of low seed supply. Four cultivars (Arche, Record, Trémie, VM202) were grown in each block while the 270 breeding lines and the remaining parents and cultivars were randomly spread into eight blocks. These eight blocks were replicated two times in Le Moulon and three times in Mons (Table 1).
Grain yield was measured on each experiment and GNC only in Mons. N concentrations were measured with a near infrared reflectance analyzer (Technicon InfraAnalyzer 400, Technicon Instruments Corporation, Tarrytown, NY) calibrated against a Dumas procedure (Dumas, 1831).
Preliminary analyses were conducted on GY of the four standard cultivars using the PROC GLM procedure from SAS software (SAS Institute Inc., 1999) with genotype (G) and block (B) as fixed effects using the model Vij = Gi + Bj + eij.
Analyses of variance were then conducted on all genotypes with the PROC MIXED procedure from SAS software (SAS Institute Inc., 1999) with genotype (G) and replicate (R) as random effects and block (B) as fixed effects using the model Vijk = Gi + Rj + B(R)jk + eijk.
Broad sense heritability is defined as the proportion of phenotypic variance that is due to all genetic effects (Holland et al., 2003). These heritabilities were not calculated to estimate the heritability of a reference population or to make conclusions regarding the inheritance of traits (Bänziger et al., 1997). Broad-sense heritabilities (h2) were calculated on a plot basis as h2 = 
2G/
and standard error estimates 
were calculated according to Holland et al. (2003). Genetic correlations were calculated as rG12 = COVG12/(G1,G2) and their standard deviations were computed according to Becker (1984). The relative efficiency under direct selection in Experiment 1 versus indirect selection in Experiment 2 was assessed as rG12h2/h1 (Falconer, 1974).
|
RESULTS AND DISCUSSION
|
|---|
Mean GY ranged from 461 g m–2 in Le Moulon under low N with fungicide to 951 g m–2 in Le Moulon under high N with fungicide (Table 2). In Mons, averaged on both years, the yield reduction was greater for the low N treatment (–35%) than for the no fungicide treatment (–14%). Mean GNC in Mons ranged from 1.5 to 2.0% and was greatly decreased at low N levels. No differences in GNC were observed on either year for fungicide treatments. It has been shown that foliar diseases, such as brown rust (caused by Puccinia triticina Eriks.), have a greater effect on thousand kernel weight than GNC (Bancal and Huet, 2000), as most grain N comes from N taken up before flowering and is later remobilized.
View this table:
[in this window]
[in a new window]
|
Table 2. Mean, genetic variance, error variance, and heritability for grain yield measured in 10 experiments and for grain N content measured in six experiments.
|
|
Significant genotype effects were detected for GY and grain protein content in all experiments (data not shown). Broad-sense heritabilities for GY ranged from 0.18 in Le Moulon at low N without fungicide to 0.90 in Mons in 1999 at high N without fungicide (Table 2). In Le Moulon in 1999 and in Mons, heritability estimates were higher at high N than at low N level. This was due to both an increase in error variance and a decrease in genetic variance at low N levels. Ud-Din et al. (1992) found similar results for GY, where heritability estimates were higher in irrigated environments than in drought-stress environments.
Heritabilities for GY in treatments without fungicide were as high as or higher than heritabilities measured in the corresponding control treatments. In Mons in 1999, this was due to an increase in genetic variance. Broad-sense heritabilities for GNC were not different between the control treatments with fungicide and the corresponding treatments without fungicide. They were lower at low N levels and this was due to an increased error variance in both years. Nevertheless, these heritability estimates may be biased through the selection of our population. The lines were indeed selected on F2 and F3 for highly heritable and important characters (plant height, diseases, etc.), as done in nurseries in most breeding programs, to decrease their impact on further F4 plot evaluation, especially at high-input level. This could narrow the genetic variance and consequently the heritability itself. Nevertheless, the decrease of genetic variance may be limited because the lines were unselected for GY and GNC, the characters for which the question of direct or indirect selection at low input level was addressed. In addition, the shrinkage of the genetic variance may be compensated by a decrease in the environmental variance due to the selection for resistance to lodging and to diseases. It can therefore be hypothesized that the heritability estimates of yield and GNC were weakly affected by the selection for resistance to diseases and lodging.
Genetic correlations for GY between the 10 experiments were always positive and ranged from 0.10 to 0.95 (Table 3). When comparing the same treatment between two successive years, the genetic correlations (in italic) were in the 0.75 to 0.95 range, except for the low fungicide treatment in Mons, where it was much lower. Genetic correlations for GNC were also always positive and varied from 0.78 to 0.98. They were higher on average than for GY, particularly when an experiment without fungicide was involved. This may be due to the lower genotype x environment interactions for GNC in comparison to GY, as observed in other studies (Le Gouis et al., 2000, 2002).
View this table:
[in this window]
[in a new window]
|
Table 3. Genetic correlation between two environments for grain yield (above diagonal) and grain N content (below diagonal) and standard deviation. Correlations of the same treatment between two successive years are in italic.
|
|
The relative efficiency of indirect selection was compared with direct selection for each pair of environments. Indirect selection was never more efficient than direct selection, as shown for Mons in Table 4. The relative efficiency of indirect selection ranged from 0.15 to 0.99. The lowest values were found in 1999 under low N input or without fungicide. This suggests that the predictions may be affected by the stress intensity.
View this table:
[in this window]
[in a new window]
|
Table 4. Relative efficiency of direct versus indirect selection for grain yield in six environments (first line in bold) and percentage of lines to retain to keep 80% of the top 20% of lines in the target environment (second line in italic).
|
|
From a practical point of view, poor efficiency of indirect selection also implies that selection intensity must be low. When selecting in one environment, the percentage of lines retained has to be high to ensure plants that will perform best in the other environment. For example, a breeder selecting at high N levels in Mons in 1998 would have to keep 33% of the lines to ensure retention of 80% of lines in the top 20% under low N levels (Table 4). Taking into account the genotype x year interaction, 50% of all lines must be retained when comparing results obtained in 1998 at high N level to results obtained in 1999 at low N level. Therefore, breeding programs targeting either low N environments only, or both low and high N environments should include low N selection environments to maximize selection gains. According to Gallais (1984), selection in stress environments will favor selection of adaptation characters while breeding in favorable environments will select for characters linked to the maximum potential.
Grain yield and grain protein content have to be simultaneously considered, as quality is increasingly important, particularly when exporting in a global market. The genetic correlation between GY and GNC was negative for the control as well as for low N treatments and ranged from –0.52 to –0.77 (Table 5). For the treatment without fungicide, it was negative in 1998 and positive, but near zero, in 1999. The genetic correlation tended to be more negative at low N than at high N. The genetic gain for GNC was then expected to be lower at low N than at high N when selecting for yield. The relative efficiency of direct selection for GNC compared with indirect selection for GY (Table 5) showed this was the case in 1998 (–0.70 compared with –0.83 for high and low N levels respectively). The result was different in 1999 (–0.54 compared with –0.50) because of the low heritability for GY. In a recent review, the negative correlation between GY and grain protein content in bread wheat appeared highly variable (Oury et al., 2003). The correlation may be masked by environmental effects as shown in their study based on 14 yr of multisite trials conducted between 1977 and 1999 from a French wheat-breeding program. They found a better estimate of the genotypic value for these traits when reducing the environmental effects by considering means from trials conducted over diverse environments. In most cases, correlation between GY and grain protein content were lower than –0.7 (Oury et al., 2003).
View this table:
[in this window]
[in a new window]
|
Table 5. Genetic correlation and standard deviation between grain yield (GY) and grain N content (GNC) and relative efficiency of direct selection for GNC versus indirect selection for GY measured on 284 wheat genotypes grown in six environments.
|
|
The type of crosses and the parents used in this study are a good sample of the variability used in France and Northern Europe. The population of F4 lines was large in size (270 progenies). The F4 lines were considered as a reference population of our breeding material or equivalent breeding programs which could be tested at all levels of inputs. The lines were unselected for GY and GNC, the two characters under investigation, but selected on F2 and F3 for other highly heritable characters such as resistance to diseases and to lodging. Our results illustrate that a population that has been under selection for resistance to diseases and lodging still has genetic variation for other traits such as yield or GNC.
Félix et al. (2002) have recently shown that, in France, the best economic margin can often be achieved with a low level of inputs, including one application or no fungicide application instead of the usual three, a reduction of N by 30 or 60 kg ha–1, elimination of the growth regulator, and varieties adapted to this system. This situation is even likely to be more frequent if the wheat price continues to decline. Plant breeders must be able to propose varieties adapted to low-input systems. We conclude that breeding programs targeting low-input environments should include low-input selection environments to maximize selection gains. Resistance to lodging and diseases are common objectives as the French variety registration system involves trials without growth regulators and fungicides. Performance under low N input regime, however, is seldom considered. On the basis of these results, selection under low N input is advised.
|
ACKNOWLEDGMENTS
|
|---|
We thank Dominique Bouthors, Dominique Brasseur, Benoît Dugué, Jean-Pierre Noclercq, and Michèle Van Elslande for their valuable technical assistance, as well as Sonia Debomy, Jérémie Blas, and François Clay for the preliminary analyses. We also thank Dr Eric Hanocq and Dr Pascal Bertin for their helpful comments and advice in the preparation of this paper.
Received for publication July 28, 2003.
|
REFERENCES
|
|---|
- Agrama, H.A.S., A.G. Zakaria, F.B. Said, and M. Tuinstra. 1999. Identification of quantitative trait loci for nitrogen use efficiency in maize. Mol. Breed. 5:187–195.
- Atlin, G.N., and K.J. Frey. 1989. Predicting the relative effectiveness of direct versus indirect selection for oat yield in three types of stress environments. Euphytica 44:137–142.[CrossRef][ISI]
- Bancal, M.O., and P. Huet. 2000. Functioning of a crop of winter wheat subjected to a late attack by brown rust (Puccinia recondita). p. 515–528, In P. Maillard and R. Bonhomme (ed.) Fonctionnement des peuplements végétaux sous contraintes environnementales, Vol. 93. INRA Editions, Paris.
- Bänziger, M., F.J. Betran, and H.R. Lafitte. 1997. Efficiency of high-nitrogen selection environments for improving maize for low-nitrogen target environments. Crop Sci. 37:1103–1109.[Abstract/Free Full Text]
- Becker, W.A. 1984. Manual of quantitative genetics. 4th ed. Acad. Enterprise Pullmann, WA.
- Bertin, P., and A. Gallais. 2000. Genetic variation for nitrogen use efficiency in a set of recombinant maize inbred lines. I. Agrophysiological results. Maydica 45:53–66.[ISI]
- Brancourt-Hulmel, M., G. Doussinault, C. Lecomte, P. Bérard, B. Le Buanec, and M. Trottet. 2003. Genetic improvement of agronomic traits of winter wheat cultivars released in France from 1946 to 1992. Crop Sci. 43:37–45.[Abstract/Free Full Text]
- Calhoun, D.S., G. Gebeyehou, A. Miranda, S. Rajaram, and M. van Ginkel. 1994. Choosing evaluation environments to increase wheat grain yield under drought conditions. Crop Sci. 34:673–678.[Abstract/Free Full Text]
- Ceccarelli, S., S. Grando, and J. Hamblin. 1992. Relationship between barley grain yield measured in low- and high-yielding environments. Euphytica 64:49–58.[ISI]
- Cooper, M., R.E. Stucker, I.H. DeLacy, and B.D. Harch. 1997. Wheat breeding nurseries, target environments, and indirect selection for grain yield. Crop Sci. 37:1168–1176.[Abstract/Free Full Text]
- Dumas, J.B.A. 1831. Procédés de l'analyse organique. Annales de Chimie et Physique 2: 198–213.
- Falconer, D.S. 1974. Introduction to quantitative genetics [In French]. Masson, Paris.
- Félix, I., C. Loyce, C. Bouchard, J.M. Meynard, M.-H. Bernicot, B. Rolland, and H. Haslé. 2002. Associer des variétés rustiques à des niveaux d'intrants réduits. Intérêt économique et perspectives agronomiques. Perspect. Agric. 279:30–35.
- Gallais, A. 1984. Use of indirect selection in plant breeding. p. 45–60. In W. Lange, et al. (ed.) 10th Eucarpia Congress: Efficiency in plant breeding. 19–24 June 1983. Pudoc, Wageningen, the Netherlands.
- Hill, J., R. Ortiz, W.W. Wagoire, and O. Stolen. 1999. Effectiveness of indirect selection for wheat yield in a stress environment. Theor. Appl. Genet. 98:305–309.[CrossRef][ISI]
- Holland, J., W. Nyquist, and C. Cervantez-Martinez. 2003. Estimating and interpreting heritability for plant breeding: An update. Plant Breed. Rev. 22:9–112.
- Le Gouis, J., D. Béghin, E. Heumez, and P. Pluchard. 2000. Genetic differences for nitrogen uptake and nitrogen utilisation efficiencies in winter wheat. Eur. J. Agron. 12:163–173.[CrossRef]
- Le Gouis, J., D. Béghin, E. Heumez, and P. Pluchard. 2002. Diallel analysis of winter wheat at two nitrogen levels. Crop Sci. 42:1129–1134.[Abstract/Free Full Text]
- Ortiz-Monasterio, J.I., K.D. Sayre, S. Rajaram, and M. McMahon. 1997. Genetic progress in wheat and nitrogen use efficiency under four nitrogen rates. Crop Sci. 37:898–904.[Abstract/Free Full Text]
- Oury, F.X., P. Bérard, M. Brancourt-Hulmel, C. Depatureaux, G. Doussinault, N. Galic, A. Giraud, E. Heumez, C. Lecomte, P. Pluchard, B. Rolland, M. Rousset, and M. Trottet. 2003. Yield and grain protein concentration in bread wheat: A review and a study of multi-annual data from a French breeding program. J. Genet. Breed. 57:59–68.
- Presterl, T., G. Seitz, M. Landbeck, E.M. Thiemt, W. Schmidt, and H.H. Geiger. 2003. Improving nitrogen-use efficiency in European maize: Estimation of quantitative genetic parameters. Crop Sci. 43:1259–1265.[Abstract/Free Full Text]
- SAS Institute Inc. 1999. SAS/STAT User's guide. Version 8. SAS Institute Inc., Cary, NC.
- Shroyer, J.P., and T.S. Cox. 1993. Productivity and adaptive capacity of winter wheat landraces and modern cultivars grown under low-fertility conditions. Euphytica 70:27–33.
- Sinebo, W., R. Gretzmacher, and A. Edelbauer. 2002. Environment of selection for grain yield in low fertilizer input in barley. Field Crops Res. 74:151–162.[CrossRef]
- Ud-Din, N., B.F. Carver, and A.C. Clutter. 1992. Genetic analysis and selection for wheat yield in drought-stressed and irrigated environments. Euphytica 62:89–96.[CrossRef][ISI]
- Van Ginkel, M., I. Ortiz-Monasterio, R. Trethowan, and E. Hernandez. 2001. Methodology for selecting segregating populations for improved nitrogen-use efficiency in bread wheat. Euphytica 119:223–230.[CrossRef]
Related articles in Crop Science:
- THIS ISSUE IN CROP SCIENCE
Crop Science 2005 45: vii.
[Full Text]
This article has been cited by other articles:
|
|
|
|
 
R. E. Lorenzana and R. Bernardo
Genetic Correlation between Corn Performance in Organic and Conventional Production Systems
Crop Sci.,
May 1, 2008;
48(3):
903 - 910.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
B. Hirel, J. Le Gouis, B. Ney, and A. Gallais
The challenge of improving nitrogen use efficiency in crop plants: towards a more central role for genetic variability and quantitative genetics within integrated approaches
J. Exp. Bot.,
July 1, 2007;
58(9):
2369 - 2387.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
