Interrelations among vegetative,yield and bunch quality traits in short-tem oil palm progenies

Summary Breeders usually obtain information on a large number of traits in their breeding and selection programmes. However, since some or many of these traits could be related, it is desirable to reduce the number being handled to the barest minimum without sacrificing efficiency. Breeders often use correlations, stepwise multiple regressions and path coefficient analyses to determine the nature of relationships among such characteristics. The objective of this study was to use these statistical methods to determine traits that could be useful in predicting number of bunches (NB), fresh fruit bunch yield (FFB) and mean bunch weight (MBW) for 13 backcross progenies of oil palm (Elaeis sp.) grown in four replications with 12 palms per replication. Data were collected over a period of 3 or 11 years, depending on the trait. Progeny means and individual palm data, averaged across replications and years, were used for each of these analyses.Results showed that correlations involving individual palm data were similar to the progeny mean correlations. Five to seven traits accounted for about 70–90% of the variations in the dependent variables. Traits that showed the highest correlations with the dependent variables always accounted for the largest proportion of the variation (r²) in multiple regression models, but did not always have the highest direct effect (i.e. path coefficients) in path analysis of the dependent trait. Number of leaves per palm had the highest correlation (r=0.729) with, and accounted for 53.2% of the variation in NB. Path analysis however showed that percentage fruit per bunch (% F/B) was the most important determinant of NB. It exerted the highest direct effect of 0.537. The highest correlation with FFB involved number of leaves (r=0.660), which, in multiple regression models also accounted for the largest proportion (44.0%) of the variation in FFB. Path analysis showed that percentage mesocarp per fruit gave the highest direct effect (p=–0.974) for this trait. It was concluded that various combinations of number of leaves per palm, sex-ratio, percentage fruit per bunch and percentage mesocarp per fruit would be effective as indirect selection criteria for NB, FFB and MBW in this set of material.     

Estimates of within family genetic variability for clonal selection in oil palm

Within family genetic and environmental variabilities in oil palm dura (D) ×pisifera (P) hybrid progenies were estimated from a D × P progeny test trial, a clonal trial and a clonal cum D × P progeny test trial and with the objective of examining the efficiency of selecting source palms (ortets) within family for clonal propagation. The clones were derived from seedling embryos or seedlings that were progeny reproductions of families proven in the first trial. The progenies in the second trial represented the source families of two sets of clones planted in the same trial. Estimates were obtained from: between and within family variance components; between and within clone variance components; and difference between pooled within family variance and pooled within clone variance. Estimates were generally similar with the different methods used. Within family genetic variabilities were generally low in progenies derived from more inbred parents especially in bunch and oil yields. In progenies from more outbred or variable parents, within family genetic variabilities were relatively higher particularly for oil to bunch and palm height; nevertheless environmental variabilities predominated. The low efficiency of within family palm or ortet selection for clone production and the need for repeated clonal tests over time and space to identify outstanding clones are stressed and the implications to commercial clonal propagation of oil palm discussed. This revised version was published online in July 2006 with corrections to the Cover Date.     

Genetic control of fertility in the oil palm (Elaeis guineensis Jacq.)

Summary A study was conducted on the segregating populations from tenera × tenera, tenera x fertile Pisifera crosses and tenera selfings. Pisifera palms were categorised as (1) fertile Pisifera palms if producing mature ripe bunches regularly under natural conditions, and (2) partially female sterile pisifera palms if producing a few or no bunches in several years.Based on observations on the segregation patterns of the fruit forms (dura, tenera and pisifera) and bunch production patterns, a genetic model was formulated to explain the genetic basis of fertility in the oil palm. Fertility in the oil palm is shown to be controlled by a single gene which is linked to the gene controlling fruit form. Chi-square tests confirmed that the model agreed with the segregating ratios of fruit form and fertility observed.The implications of this finding with regard to oil palm breeding and improvement, and the potential of the fertile pisifera palm for increasing palm oil yields in plantations are discussed.     

Candidate traits for index selection in choice of oil palm ortets for clonal propagation

Summary The objective of the study was to examine the usefulness of: principal component scores (PC), factor analysis cum stepwise regression identified traits (FASR) and selected traits based on their higher heritabilities and genetic correlations to the objective traits (HGC); as selection traits in a desired gains selection index (Yamada et al., 1975) to improve objective traits (oil yield, kernel oil yield, height increment, bunch index and leaf area ratio) as compared to those based on all the observed traits (AO); in selecting oil palm ortets for cloning.Based on the required selection intensities (i^*) to achieve the desired gains AO indices having smaller i^*'s were most efficient followed by PC, HGC and FASR indices. Expected selection response (1/i^*), however, is expected to increase with additional selection traits. As such HGC (bunch number, kernel to fruit and mesocarp to fruit) and FASR (mesocarp to fruit, fresh fruit bunch yield, fruit to bunch and average bunch weight) indices would be useful as they achieved expected selection responses close to AO indices with a small number of traits and would also minimise the contribution of highly correlated traits to sampling errors. The results also suggested considering selection indices for only two objective traits-oid yield and kernel oil yield-instead of all five objective traits in which case screening of impracticably large populations is needed to obtain desired genetic gains.     

Assessment of yocoboue wild oil palm (Elaeis guineensis Jacq.) from Côte d'Ivoire

Following a survey involving wild oil palm groves in Côte d’Ivoire as a whole in 1968, 17 parents were chosen at Yocoboué (southernCôte d’Ivoire). Agronomic performance of the 17 parents was assessed through Deli × Yocoboué hybrids in 3 trials grown at La Mé (Côte d’Ivoire), Mondoni and La Dibamba (Cameroon).When compared to the Deli × La Mé progenies used as controls, the Deli × Yocoboué hybrids produced 91% for total bunch weight and86% for oil yields. However, the genetic variability within the Deli × Yocoboué progenies allows to select individual progenies as productive as the Deli × La Mé controls, for breeding purposes. A study of major agronomic traits led to the choice of parent YO 11 T , which combines the largest number of favourable traits. On average, Deli × Yocobouéprogenies have slower vertical growth and are susceptible to Fusarium wilt, except 2 parents (YO 3 T and YO 9 T), which seem to be resistant.     

Development of elite families in oil palm (Elaeis guineensis Jacq.)

Summary Data from 29 dura female parents, 14 pisifera male parents and their offspring were assembled to obtain quantitative genetic information on optimizing selection of elite families. Phenotypic values of the offspring were partitioned into general combining ability (GCA) of the parents and their specific combining ability (SCA). Parental GCA-values for leaf-Mg content (LMG), height increment (HT), leaf area ratio (LAR), leaf area (LA), frond production (FPR), vegetative dry matter production (VDM), the rate of increase in leaf area against age (k) and maximum leaf area (L_m) were entered in a multiple regression analysis as predictors of oil-and-kernel yield and Harvest Index (HI), i.e. the proportion of dry matter used for oil and kernel production, of the offspring. The analysis revealed that both selection objectives are positively influenced by indirect selection for high LAR and LMG, and for reduced VDM with FPR as main component. But palms with high values for LA, L_m and HT should not be considered to avoid selecting palms partly for their ability to compete with neighbouring palms for light. The contribution of SCA was relatively small for most of the characters studied, except HT, FPR and HI. From the standard deviations for GCA and SCA of the dura and pisifera parents, i.e. _GCA and _SCA, it was extrapolated that selection can potentially enhance bunch yield (by 24%), oil-to-bunch ratio (19%), kernel-to-bunch ratio (29%), oil-and-kernel yield (31%) and HI (56%) above the mean phenotypic value obtained for the offspring of the present investigation. Heritabilities, calculated as the square of the correlation of the GCA-and phenotypic values of the female parents, were low for bunch yield, oil-and-kernel yield and HI, but moderate for LAR, LMG and VDM; levels were high for oil and kernel extraction rate. These outcomes were confirmed by the correlations between values of the female parent and their offspring. Among the unbalanced incomplete mating designs studied, the rectangular lattice and the alpha design are most suitable to compare GCA-values of pairs of parents. A breeding strategy adapting the recently introduced practical possibility of clonal reproduction is discussed.     

Effects of inbreeding on growth and yield of oil palm

The effects of inbreeding were studied in three oil palm (Elaeis guineensis Jacq.) trials in the Democratic Republic of Congo (Congo). In selfings, marked inbreeding depression was observed for yield of fruit bunches, mean bunch weight and bunch number, but there was little effect on bunch composition. Most vegetative measurements were significantly affected by selfing, but leaf production rate and leaf area ratio were unaffected. Sib-crossing had less depressive effect than selfing, and in some families sib-crosses were superior to outcrosses. Where selection needs to be done within inbred families, our study suggests that leaf area ratio and bunch composition would be useful criteria.     

Breeding of high yielding and dwarf oil palm planting materials using Deli <Emphasis Type="Italic">dura</Emphasis> × Nigerian <Emphasis Type="Italic">pisifera</Emphasis> population

In practice, progeny and individual palm selection are believed to be the most suitable breeding approach for improvement of quantitative traits in oil palm because their phenotypic expressions are strongly influenced by abiotic factors. Therefore progeny selection approach was applied in this study for the selection of high fresh fruit bunch (FFB) and dwarf oil palm planting materials. Cross between Deli dura and Nigerian pisifera resulted into 34 D × P full sib progenies with 1036 seedlings. For six consecutive years, data were collected on yield and yield component traits, while vegetative traits were recorded once. Bi-parental analysis was carried out using analysis of variance, followed by progenies mean comparison, variance components, heritabilities and cluster analysis. Highly significant (P ≤ 0.01) progeny effect was recorded in this study and this had a pronounced effect on the expression of all the quantitative traits. Progenies performance of FFB varied significantly and it ranged from 166.49 to 220.06 kg/palm/year (kg/p/yr) with trial mean of 192.93 kg/p/yr. Palm height after 8 years of field planting ranged from 1.67 to 2.78 m (control cross) with trial mean of 2.12 m. Broad sense heritability (\({\text{h}}_{\text{B}}^{2}\)) was found to be very low (<17.60%) for all the yield traits, however this parameter was high for vegetative traits with palm height having \({\text{h}}_{\text{B}}^{2}\) of 90%. Cluster analysis based on all the quantitative traits grouped all the 34 DP progenies into nine distinct clusters. From this study, five progenies (DP3, DP4, DP5, DP8 and DP24) were identified to be high yielding and dwarf palms compare to trial mean. At density of 140 palm/ha, they will produce FFB of 28.63–30.81 t/ha and average of 29.69 t/ha which is about 27.15% higher in FFB when compared to the current planting material with FFB of 23.35 t/ha. In addition, the selected progenies possessed average annual palm increment of 29.82 cm/yr with range of 26 and 32.5 cm/yr which was 57.33% shorter than the current planting material with palm height increment of 45–75 cm/yr.     

Penetrance of creeping-rootedness in clonal progenies of lucerne and observations on underground morphology of plants differing for this character

Elaeis oleifera or ‘caiaué’, a close relative of oil palm (E. guineensis), has some agronomic traits of great interest for the oil palm genetic breeding such as slow growth, oil quality (mostly unsaturated) and disease resistance. An analysis of a Brazilian oil palm germplasm collection was carried out using RAPD (Random Amplified Polymorphic DNA) markers with the objective of understanding the genetic variation of ‘caiaué’ accessions collected in the Amazon Forest in the last two decades. A sample of 175 accessions obtained along the Amazon River Basin was analyzed and compared to 17 accessions of oil palm from Africa. Ninety-six RAPD markers were used in the analysis, of which fourteen were shown to be specific to oil palm, while twelve were specific to ‘caiaué’. Results showed that the Brazilian ‘caiaué’ accessions studied have moderate levels of genetic diversity as compared to oil palm accessions. The data allowed the establishment of similarity groups for ‘caiaué’ accessions, which is useful for selecting parental plants for population breeding. Cluster analysis showed that, in general, genetic similarities are not correlated to geographical distances, but consistent with geographical dispersal along the Amazon River network. AMOVA showed that most of the genetic variation is found within populations, as expected for anallogamous and long-lived perennial species. The study provides important information to define strategies for future collection expeditions, for germplasm conservation and for the use of E. oleifera in breeding programs. This revised version was published online in August 2006 with corrections to the Cover Date.     

Estimation of genetic parameters and prediction of breeding values in an autotetraploid blueberry breeding population with extensive pedigree data

The strategy for breeding blueberries has been based on phenotypic selection without accounting for pedigree correlation information. A central premise of classical quantitative genetics is that through knowledge of the relationships among individuals in a population, we can make inferences about breeding values (BVs), estimate genetic parameters, and support phenotypic selection. Genetic evaluations using best linear unbiased prediction (BLUP) and restricted maximum likelihood (REML) are founded on pedigree information and have been the standard method used in livestock and forest breeding. Despite theoretical and practical benefits, their application in blueberry breeding programs remains unexploited. The objective of this study was to evaluate the use of REML/BLUP in the University of Florida blueberry breeding program in order to estimate genetic parameters and predict BVs of primary selection traits. To do this, we collected phenotypic data for eight selectable traits (yield, flower bud density, fruit weight, fruit firmness, fruit diameter, fruit soluble solids, fruit pH, and fruit scar diameter) from 1996 individuals. The highest narrow-sense heritabilities were found for fruit weight, scar diameter, and yield. The majority of genetic correlations between traits were under 0.20. The largest genetic gain was achieved when selections from early stages were used as parents and the top 5% of the population was selected. Compared to traditional phenotypic selection, our results are evidencing of the importance of implementing REML/BLUP to estimate genetic parameters that help and support the breeding strategy and maximize genetic gains.     

Outlines of the oil palm breeding programme in Portuguese Guinea

The methods and techniques of oil palm selection and breeding are discussed in relation to the problems of the oil palm industry in Portuguese Guinea.A general survey of the oil palm populations, both cultivated and natural palm groves, was carried out in order to determine existing varieties, their distribution, range of variability, disease incidence and palm oil quality.Yield was established as the major aim in the breeding programme. Short-stem habit and adaptability to different sets of conditions are also considered.Both a long-term and a short-term project are undertaken simultaneously.The long-term programme includes selection of Deli female parents, pollen and seed introductions from pisifera and tenera palms selected in San Thome and single crosses among selected parents. Emphasis is laid on the Deli × pisifera cross to increase oil content, though Deli × tenera and Deli × Deli seed are also produced.The short-term programme involves multiple crosses. A number of mass-selected Deli palms is pollinated using pollen from pisifere or tenera palms selected in San Thome. As a result of the short-term programme over 450,000 hybrid seeds were produced up to the end of 1964.Abbreviations INEAC Institut National pour l'Étude Agronomique du Congo Belge - IRHO Institut de Recherches pour les Huiles et Oléagineux - MEAU Missão de Estudos Agronómicos do Ultramar - WAIFOR West African Institute for Oil Palm Research     

BLUP for genetic evaluation of plants in non-inbred families of annual crops

Best Linear Unbiased Prediction (BLUP) has become the most widely used method for genetic assessment of animal and perennial species, and it is potentially relevant for annual crops. The objective of this study was to assess this method for selection within non-inbred families in recurrent breeding programs. The ‘animal model’ was fitted. The data were expansion volume (EV) and grain yield of plants in recombination plots of two to three selection cycles in the popcorn population Viçosa, with half- and full-sib progenies. The ASReml program was used to perform the analyses. For both EV and yield the breeding values predicted from the additive and additive-dominant models were highly correlated. Multi-generation BLUP was, in general, more accurate than single-generation analysis. These two methods resulted in highly correlated predicted breeding values. The dominance genetic values predicted from the single- and multi-generation analysis were also highly correlated. The pedigree information reduced the percentage of coincidences among the selected individuals relative to phenotypic selection mainly in the population structured in half-sib families. Based on breeding values predicted by BLUP analysis, the most efficient selection procedure was mass selection.     

Experimental modification of reciprocal recurrent selection in oil palm breeding in Cameroon

Reciprocal recurrent selection (RRS) exploits the genetic complementarity of dura populations of palms with a small number of large bunches and tenera/pisifera group of palms with large number of small bunches. An extra phase was introduced to RRS just after the second cycle of selection where outstanding dura derived from (DA3D × DA5D) combinations of the second cycle were self-pollinated for seed production. Best palms based on the phenotype were selected and crossed with pisifera of second cycle offspring of LM2T. [(DA3D × DA5D) self-pollinated × LM2T self-pollinated] progenies were field tested at La Dibamba (Cameroun) from 1987 to 1997. They were precocious with regard to the three-first years’ bunch yields (13.16–15.77 t/ha). Homogeneity for all bunch yield components and homogeneity for vertical growth were illustrated by low values of the variation coefficient (5.75–7.75%). The oil yield was 107% of the control against 115% from [(DA3D × DA5D) × LM2T self-pollinated] progenies of the second cycle of RRS, mostly due to poor mesocarp to fruit and poor oil to mesocarp ratios which led to a low oil extraction rate. The extra selection phase led mainly to homogenisation of progenies but also reduced the fruit quality. However, three progenies whose average oil yield represented 115% of the control (4.87 t/ha) in addition to their precocity can be released to oil palm growers. The extension of RRS did not fully meet expectations and poses the problem of its reliability.     

Investigation of long-term field experiments on response of breeding lines to common scab in a potato breeding program

Common scab, caused by Streptomyces scabies, is a disease that produces scab-like surface lesions on potato tubers. Testing for susceptibility/resistance of breeding lines at the Potato Research Centre of Agriculture & Agri-Food Canada (AFC) is carried out in a scab nursery maintained at AFC for the annual scab evaluation field trials. A replicated field trial routinely consists of breeding lines from previous testing season(s) plus newly selected lines. Data of scab scores generated from long-term experiments thus formulate an incomplete 2-way table over combinations of breeding lines (genotypes) and trials (years). This requires an advanced statistical method to estimate genetic parameters for evaluation purposes. A data base with 1,435 scab index scores from 344 breeding lines were extracted from 5 years (1995–1999) of field experiments in the scab nursery maintained at AFC. The statistical method Residual Maximum Likelihood (REML) was employed to estimate variance components of the breeding population and Best Linear Unbiased Predictor (BLUP) was used to predict genetic merit of breeding lines. High heritability was obtained from variance components estimated by REML. The BLUP scores of breeding lines provided reliable evaluation of their responses to common scab. Two data base sub-sets were separately formulated from the original data base for those parents and grandparents of the breeding lines having pedigree records available. They were again subjected to REML and BLUP analyses to compare the responses to common scab and identify sources of resistance at the parental and grandparental levels. Two random data sets with equal sized samples of breeding lines were also generated from the over-all data base. The two sets of BLUP scores between corresponding breeding lines and parents showed high association which provides an assessment of the validity of the evaluation process.     

Relevant factors associated with crown expansion in oil palm (Elaeis guineensis Jacq.)

Summary Trends in crown expansion, the area of the youngest fully expanded leaf (LA) against age were studied in two oil palm progeny trials, and one density x fertilizer experiment, the latter comparing 110, 148, and 186 palms per ha each with four fertilizer levels. LA against age was best fitted by a logistic growth curve with maximum LA (L_m) and the time required to reach 0.95 L_m (t_0.95) as characteristic parameters. Progenies significantly differed in L_m and t_0.95. Increasing planting density significantly decreased L_m and t_0.95, while fertilizers increased L_m and t_0.95 only at the lowest density. Population means of the experiments markedly differed in L_m and t_0.95 values. Based on the, so called, Leaf Expansion Ratio (ER), being the ratio of LA at maximum expansion rate and LA at its final size, individual palms were selected for a rapid increase in LA, but a relatively low final size. The first characteristic means selection for rapid ground coverage, which resulted in palms with a higher Total Dry Matter Production (TDM) prior to canopy closure. The second characteristic appeared to lead to selection for high Bunch Index (the proportion of TDM used for fruit bunch production), indicating that L_m is a measure for the relative sink strength of leaves for assimilates. Selection for high ER values appeared feasible and this would increase yield during the economic life of a planting. The practical aspects of using ER values as a new selection criterion in oil palm breeding were discussed.     

Genetic diversity of wheat gene pool of recurrent selection assessed by microsatellite markers and morphological traits

Abstact An agronomic gene pool of wheat (Triticum aestivum L.) was constructed through recurrent selection. In present research, 24 wheat SSR markers determining 25 loci on 14 different chromosomes were used to evaluate the gene pool. Thirty parents used as original materials in recurrent selection were also assessed. In total, 115 alleles were detected in gene pool with an average of 4.6, ranging from 2 to 9 alleles per locus. Statistical test showed that genetic diversities had no significant difference between the gene pool and the 30 parents. Principle coordinates analysis revealed that the individuals of the gene pool were mainly divided into three groups, which was consistent with the result of cluster analysis based on genetic distance matrix of the gene pool. Cluster analysis was carried out based on Euclidian distance calculated upon five morphological trait values and the results showed that most individuals were in a group while the others scattered. Correlation analysis of genetic distance matrix and Euclidian distance matrix showed no significant correlation between two matrices. The results suggest that the gene pool is improved after several cycles of selection, while genetic variation is still maintained. Therefore, the gene pool is suitable for further breeding program.     

Analysis of growth in the oil palm II. Estimation of genetic variances of growth parameters and yield of fruit bunches

Summary Vegetative and yield measurements were made in oil palm (Elaeis guineensis Jacq.) breeding experiments leading to estimates of genetic variance for yield of fruit bunches, Crop Growth Rate, Leaf Area, Leaf Area Index, Leaf Area Ratio, Vegetative Dry Matter Production and Bunch Index (dry matter in fruit bunches/total dry matter). Results suggest the presence of genetic variability for all the traits examined, and indications are that BI, LAI and VDM have a reasonable level of genetic control. The data are discussed in relation to breeding and it is suggested that instead of straight-forward selection for high yield of individual palms, a more promising approach may be to select for palms that at high planting densities maintain reasonable BI values.     

Inheritance of some components of yield in the ‘deli dura variety’ of oil palm

Summary Genetic and environmental parameters are estimated by variance analysis of numerical data obtained from two programmes of plant breeding: biparental matings and Comstock and Robinson's North Carolina Model 1. The yield component total weight of fruit bunches and its subcomponents, average bunch weight and total bunch number are considered. Two experiments, involving different families, within the structure of each design are analysed. Good agreement is reached between the findings both within and between experiments and designs, and also from character to character. Additive genetic variation appears virtually to be absent whereas dominance effects (which as notationally defined also include epistasis) are of high magnitude even in relation to the environmental contribution. This suggests that further rapid selection progress within this variety might not be immediately likely and that past selection may have caused a fixation of additive variation. However, large errors are attached to all parameter estimates. Analysis of the data year by year rather than over years indicates that there is a fall in the relative magnitude of genetic to total phenotypic variation with increasing palm age and some explanations for this phenomenon are suggested.     

Genetic variability among elite red clover ( Trifolium pratense L.) parents used in Chile as revealed by RAPD markers

Red clover (Trifolium pratense L.) is one of the main forage species of temperate regions. Cultivars of red clover are heterogeneous which makes their genetic analysis difficult. We applied RAPDs (Random Amplifed Polymorphic DNA) in order to assess the genetic relationship and levels of genetic variability existing among a group of 16 elite red clover parents organised in four subsets of 4 parents each. Out of 55 primers 21 provided reproducible results. A total of 135 reliable and polymorphic RAPD bands were detected which were used to estimate genetic distances among pair-wise combinations of elite parents. Nei and Li's similarity values ranged from 0.60 to 0.77, with a mean of 0.66, which reflects a rather high genetic variability among the genotypes evaluated. Lower levels of genetic variability, as detected by polymorphic loci and mean heterogeneity values, were detected in a subset of parents selected for resistance to the stem nematode. Cluster analyses resolved the different sets of parents in a manner consistent with what is known from their breeding origins. An Analysis of Molecular Variance detected substantial levels of variation within subsets of parents. RAPDs represent a valuable source of genetic information for red clover breeding programmes. This revised version was published online in August 2006 with corrections to the Cover Date.     

Sustainable plant breeding

Plant breeders disrupt Hardy–Weinberg equilibrium through selection, non‐random mating, drift, migration and mutation. Sustainable plant breeding can be defined as productive and competitive breeding that is achieved without loss of genetic diversity in the elite breeding population during the professional career of the breeder. Breeding is often productive but not sustainable. From 1974 to 2000, the animal breeding programme Meatlinc in the United Kingdom had effective population size of 95, population inbreeding of 0.19% per year and generation interval of 2.15 years. Genetic progress in Meatlinc tripled in the 8 years following introduction of best linear unbiased prediction (BLUP) selection (based on the information from relatives) in 1992. Canola breeding in Australia from 1970 to 2000 had longer generation interval (6 years), smaller effective population size (<11) and higher rates of inbreeding (>0.7% per year). BLUP selection in canola was first reported in 2010. Neither programme replaced genetic diversity lost through selection and drift. Most breeding programmes violate conditions of the infinitesimal model, thereby reducing predictability of selection, but breeders can minimize these limitations to sustainable plant breeding.