CHAPTER 4
Using Molecular
Markers in
Rice Population
Improvement through
Recurrent Selection

Brigitte Courtois

Abstract

Programmes for the genetic improvement of rice populations through recurrent selection have made little use of molecular markers, even though they could be very helpful in answering certain methodological questions. This chapter describes the major types of molecular markers, and briefly reviews techniques for their use, comparing their respective advantages and disadvantages. Molecular markers can be used in genetic improvement programmes to study genetic diversity, or for mapping or marker-assisted selection. Detailed examples are given of the possible uses of markers in a recurrent selection programme for rice, such as (1) improving the selection of parents constituting the population by maximizing their genetic diversity; (2) optimizing the number of recombination cycles by assessing the evolution of deviation from panmixia in populations of successive cycles; (3) assessing the effect of population improvement on genetic diversity by comparing cycles of recurrent selection with each other or with reference sets of known diversity; (4) helping to confirm the location of the male-sterility gene used for intercrossing and assessing the extent and nature of linkage drag accompanying its use; and (5) detecting quantitative trait loci (QTLs) and manipulating the alleles at these QTLs through marker-assisted selection.

Resumen

Los programas de mejoramiento genético del arroz mediante la selección recurrente han hecho poco uso de los marcadores moleculares, que podrían ser muy útiles para contestar algunas preguntas metodológicas. Este trabajo describe los principales tipos de marcadores moleculares, presenta una breve descripción de las técnicas utilizadas con éstos y compara sus respectivas ventajas y desventajas. Los marcadores moleculares se pueden utilizar en programas de mejoramiento genético para estudiar diversidad genética, o para mapeo y selección asistida. Se incluyen ejemplos más detallados de posibles utilizaciones de marcadores en un programa de selección recurrente en arroz. Un primer aspecto podría ser para mejorar la elección de los progenitores que constituirán la población, maximizando la diversidad genética entre ellos. Otra posibilidad sería optimizar el número de ciclos de recombinación mediante la evaluación de la evolución de la desviación de la panmixia en poblaciones de ciclos sucesivas. Los marcadores podrían usarse para evaluar el efecto del mejoramiento poblacional sobre la diversidad genética comparando los varios ciclos de recurrencia entre ellos o con patrones conocidas de diversidad. Los marcadores podrían ayudar a confirmar la localización del gen de androesterilidad utilizado para intercruzar y evaluar el alcance y la naturaleza del arrastre de ligamiento ("linkage drag") que podría acompañar su uso. Los marcadores también se usan para detectar características de control genético cuantitativo (QTLs) y manipular alelos en estos QTLs mediante selección asistida por marcadores.

Introduction

Breeding programmes in Latin America have been very successful in developing productive rice lines, using the selection method known as pedigree. Good examples include the varieties CICA 8, developed in Colombia for the irrigated ecosystem, and Maravilha, developed in Brazil for the upland ecosystem. Although the pedigree method is well adapted for autogamous species, concern has recently been expressed with respect to the sustainability of genetic progress with such a breeding scheme.

The pedigree method usually involves a limited number of elite progenitors- frequently related to each other-that are crossed. Progenies are selected across generations until the material is completely fixed. Hence, only a few of these elite lines are used for the next round of crosses, thus reducing the effective population size for the next breeding cycle. Guimarães et al. (1996) presented an example of this when they analysed CIAT’s upland rice programme.

The long-term risk with pedigree selection is a reduced genetic base and small genetic gains. Cuevas-Pérez et al. (1992) and Rangel et al. (1996) showed that the genetic bases of new rice varieties were narrowing in Latin America. Santos et al. (1997), Breseghello et al. (1999) and Rangel et al. (2000) demonstrated that the genetic progress made in various rice breeding programmes was decreasing over time.

Developing populations with broad genetic bases and using improvement methods that permit the continuous accumulation of favourable alleles can counteract these disadvantages.

Population improvement through recurrent selection fits this description. In 1984, Embrapa Arroz e Feijão and CIRAD began developing rice populations with broad genetic bases. They used the recessive male-sterility gene induced in ‘IR36’ by Singh and Ikehashi (1981) to facilitate recombination. Since then, the methodology was adopted by several countries in the region (Guimarães, 2000) and elsewhere (Châtel and Guimarães, 1997; Courtois et al., 1997; Tao et al., 2000).

A population improvement programme through recurrent selection involves several steps:

• Developing a broad base population (called a ‘gene pool’)

• Extracting at random selfed progenies from the population (usually S₁ or S₂ progenies obtained after one and two generations of selfing, respectively)

• Selection from these progenies

• Intercrossing the best individuals to constitute the next population (called ‘recurrent selection population’).

Some methodological questions arise with this process: How to select parental materials to constitute the gene pool? How many crossing cycles are needed to obtain an adequate level of recombination? By how much is the population genetic diversity being reduced through recurrent selection, compared with the pedigree method, and in which direction does change occur? How can the best progenies be selected? These questions can be organized around two themes: diversity within and between populations, and selection. Until now, they have received little attention in the context of the rice crop.

We can find answers by using molecular markers-a powerful analytical tool that underwent important developments during the 1990s. Plant breeders use this readily accessible technology in a wide range of activities such as:

• Checking conformity of F₁ hybrids or varietal identity

• Understanding the genetic control of simple or quantitative traits

• Conducting marker-assisted selection to introgress genes or QTLs, or build genotypes

• Conducting diversity studies (e.g. patterns of domestication, structure of cultivated species and their pathogens, monitoring hybridization and parentages in conventional or recurrent selection programmes, and evaluations of linkage disequilibrium and gene flow).

Until recently, however, except for studies by Ferreira et al. (2000) and Filloux (2001), molecular markers were little used in the framework of recurrent selection programmes for rice.

This chapter aims to present possible uses of molecular markers within the framework of recurrent selection programmes, giving examples where possible. We will not extensively review all possibilities; instead, we will focus on a few uses of immediate interest to plant breeders, namely:

• Constituting gene pools and monitoring the evolution of diversity
• Optimizing the number of recombination cycles
• Understanding the effect of the male-sterility gene from ‘IR36’
• Exploiting the possibilities of detecting and following up QTLs across cycles in such populations.

Different types of molecular markers

Before discussing how molecular tools can be used in programmes of population improvement through recurrent selection, we will first describe the different types of markers, presenting their respective qualities and limitations. The range of molecular markers available has expanded rapidly. The commonest are restricted fragment length polymorphisms (RFLPs), amplified fragment length polymorphisms (AFLPs), random amplified polymorphic DNA (RAPD) and simple sequence repeats (SSRs), also known as microsatellites. Single-nucleotide polymorphisms (SNPs) are the newest type. See Karp et al. (1997) and De Vienne (1998) for reviews of these different technologies. No absolute hierarchy exists between the marker types-all having advantages and disadvantages (Table 1). The choice of a type of marker therefore will depend on the user’s objectives and thus be determined case by case.

Table 1. Main characteristics of major types of molecular markers.

a. Both cost and technical difficulty are highly dependent on the chosen method of visualization and, hence, on the expected throughput level

From the historical viewpoint, RFLPs were the first to be used as markers. The technique involves digesting DNA with a restriction enzyme. The fragments are separated by gel electrophoresis and transferred to a filter for hybridization with radioactively labelled probes. Given that single-copy probes are selected, RFLP markers are locus specific and co-dominant. Their polymorphism is due to variations in the restriction sites. The positions on the genetic map of the most commonly used probes are known for species of interest. RFLP markers are technologically demanding, requiring large quantities of good quality DNA. They have therefore been progressively replaced by markers based on the polymerase chain reaction (PCR-based markers).

Microsatellites consist of tandem repeats of di- to penta-nucleotide motifs. Their polymorphism lies in the variation of the number of repeats, probably because of errors during replication. At first sight, they appear not to be under selective pressure. The technique involves:

(1) amplification of the microsatellite sequence, using primers hybridizing in its flanking regions
(2) separation of the fragments on an agarose or acrylamide gel
(3) visualization of the bands, using either ethidium bromide (EtB) or silver staining for agarose gels, or radioisotopes or fluorescence (with automat sequencer) for acrylamide gels.

Microsatellite markers have multiple advantages:

• They are PCR-based markers with few requirements in terms of DNA quantity and quality
• They are co-dominant
• Because of their high rate of mutation, they are highly polymorphic, with a large number of alleles per locus
• They are abundant and well distributed in the genome.

Their main limitation is that their development, which frequently implies sequencing, is expensive. For rice, however, more than 2700 microsatellite markers, with the primers and technical conditions needed for their amplification, are available and located on genetic and physical maps (Wu and Tanksley, 1993; Akagi et al., 1996; Panaud et al., 1996; Chen et al., 1997; Temnykh et al., 2000 and 2001; McCouch et al., 2002). For each one, general indications on the allelic diversity existing within Oryza sativa are also available in databases such as Gramene www.gramene.org.

From the technological viewpoint, AFLPs as markers are intermediate between RFLPs and microsatellites. The technique combines the use of restriction enzymes and PCR amplification. The DNA is cut with two restriction enzymes, one being a frequent cutter and the other an infrequent cutter. This is followed by ligation of adapters, including restriction motifs, and a two-step PCR amplification of selected fragments. The latter procedure uses primers composed of the adapters and 1 to 3 selected nucleotides (Vos et al., 1995). It limits the number of fragments to a resolvable range. The steps of separation and visualization are identical to those for microsatellite markers. On an AFLP gel, many bands are simultaneously visualized. Each band is assimilated to an allele at a specific locus that has only two allelic forms: presence or absence of the band.

AFLP markers are mainly scored as dominant. Polymorphism is due to point mutations in the restriction sites. In contrast to the two classes of markers discussed above, their map positions are seldom known beforehand. They are probably not distributed at random throughout the genome, as they sometimes cluster in some areas. However, even with such disadvantages, they are very useful for simultaneously observing a large number of polymorphic markers.

Concerning RAPD markers, the technique consists of amplifying DNA fragments with primers that are 9 to 10 base pairs (bp) long and defined at random. Several bands are visualized simultaneously. These markers are dominant, and their polymorphism is due to sequence variations in the primer hybridization site. Although attractive for their simplicity, problems of reproducibility limit interest in their use, compared with other types of markers (Jones et al., 1997).

The newest type of markers comprises single-nucleotide polymorphisms (SNPs), which correspond to single-base substitutions in sequences. These are interesting for their extremely high density (one every 300 to 1500 bp in humans), which permits monitoring linkage over very short physical distances (see review by Brookes, 1999). However, very few have been developed for rice. Their detection and validation, demanding significant sequencing effort, are costly. Different techniques of genotyping are used (Syvänen, 2001), which can also be costly (e.g. mass spectrometry). Once these limitations are resolved, these markers will probably become widely used, because they provide detailed genomic information and have potential for automation.

Molecular markers for diversity studies

Molecular markers can be used to:

(1) estimate genetic diversity between varieties (e.g. between parents of a gene pool or between plants extracted from a population), or between populations (e.g. between recombination cycles)

(2) compare populations with reference points (e.g. groups of lines developed by classical breeding programmes, sets of local commercial varieties, or reference core collections).

This information can be used in programmes for population improvement through recurrent selection to maximize diversity within populations by crossing lines that are genetically the most distant; to manage pedigrees; or adjust the long-term strategies of improvement programmes.

Diversity studies in rice show that the various types of molecular markers show very similar structural patterns. Microsatellite markers give slightly more blurred images of the species’ structure because of their high polymorphism and number of rare alleles, but a better vision of intra-group diversity (Glaszmann et al. 2003). As a result, marker type must be chosen in function of the expected levels of polymorphism and diversity. Where the genetic base is narrow and a high resolution is needed, microsatellite markers should be preferred.

The dominance of a marker is also an important issue. Dominant markers complicate the calculation of population genetics parameters based on allelic frequencies (Lynch and Milligan, 1994). Co-dominant markers facilitate the evaluation of heterozygosity in allogamous populations, which is also a reason why microsatellites have been so widely used as markers.

Microsatellite markers, although presenting many advantages, have one disadvantage that is associated with their high rate of mutation. They do not fit well to either of the two classic models of mutation: stepwise and infinite allele mutation. Although this limitation complicates both the calculation of distances and the making of phylogenetic inferences, it is significant only in the context where phylogenetic reconstitution is the major goal.

No definitive rule exists for choosing the number of markers to use in a diversity study, but for rice, between 12 and 24 are normally used. This is a safe range according to Barry (2000), who studied the diversity of rice varieties in Guinea and showed that similar structural patterns were obtained from as many as 12 to as few as six microsatellite markers.

One marker per chromosome or chromosome arm is a reasonable marker density for ensuring the best possible coverage of the genome. Such density will also prevent bias due to the presence of genes involved in gametic or zygotic selection, or controlling traits under selection in a specific chromosome. No evidence exists that such a distribution would give better results than a random selection of markers, but it does ensure that the selected markers are not linked.

Among the possible markers for each arm, those that reveal the largest number of alleles in the studied background should be selected as priority. Those with a very high mutation rate should be avoided as they may blur results.

Molecular markers for mapping and marker-assisted selection

Molecular markers can be used to locate important genes or QTLs in the genome. Once tagged by tightly linked markers, valuable alleles of the genes can be manipulated indirectly by selecting specific alleles at the flanking markers. Marker-assisted selection can replace or, better still, be used in addition to phenotypic selection and help improve the efficiency of selection. Markers with known positions spaced at regular intervals throughout the genome are needed. For a species such as rice, whose genetic map approaches 1800 cM, between 100 and 180 markers are needed for a first approach. The markers RFLPs, microsatellites and AFLPs are interesting in this context.

For their technical simplicity, microsatellite markers are (again) currently the ones most chosen. AFLPs should be selected for quick mapping or saturating a map where microsatellites or RFLPs have left gaps. Once markers close to the gene or QTL of interest are identified, PCR-based markers are needed to manipulate them through marker-assisted selection, in which case, microsatellite markers are irreplaceable.

Possible use in population improvement programmes

Choosing parents and developing the base population

The choice of parents to constitute a gene pool depends on the objectives and strategy of each breeding programme. Both the number and nature of the parents are important. A large number of parents is not, in itself, an indicator of a broad genetic base because some may share part or most of their genomes. The genetic relatedness between parents is, as a result, an important, although complex, aspect to take into account.

So far, genetic relatedness has been evaluated on the basis of line pedigree (Graterol, 2000). Coefficients of parentage are useful, but have the limitation of assuming that traditional varieties are totally unrelated and that both parents of a single cross contribute equally to the progeny. Both assumptions are known to be wrong. Rice is strongly structured, and varieties belonging to the same varietal group are genetically closer than those varieties belonging to different groups (Glaszmann, 1987). Equal contribution from both parents is seldom true in a context of selection where plant breeders favour an ideotype.

Molecular markers are more informative. By determining the alleles present at different markers for the set of varieties to compare, we can establish a matrix of similarities that offers a truer picture of the relationships between varieties. The classic indices of similarity for comparisons between varieties are those of Dice and Jaccard for co-dominant markers, and Sokal and Michener for dominant markers (Perrier et al., 2003). After transformation of the similarity matrix to obtain distances, we can produce a geometric representation of the distance between varieties through various clustering techniques such as the unweighted pair-group method, using the arithmetic average (UPGMA), or the neighbour-joining tree method (NJ tree). Results are usually represented as dendrograms.

Lorenzen et al. (1995) provide recent examples of evaluating relatedness based on molecular markers between genotypes used in classic breeding programmes for soybean; Soleimani et al. (2002) for durum wheat; and Lima et al. (2002) for sugar cane. Information on genetic relatedness based on neutral markers can complement data on the agronomic performance and adaptation of varieties. It can also permit a sound choice of lines to be finally used as parents of a gene pool to maximize that pool’s genetic diversity.

A few examples already exist of attempts to use markers within such a context. A group of 12 isoenzymatic markers was used to evaluate distances between potential parents to create a recurrent selection population to be improved for blast resistance (Courtois et al., 1997). When closely related varieties were found, only one representative of the group was selected. Ferreira et al. (2000), using RAPD markers, compared the parents of population CNA-5 and presented data that showed the close relationships of some of the parents included in the base population.

Optimizing the number of recombination cycles

An important question that arises when constituting a gene pool or, later in the recurrent selection process, when crossing selected progenies, is how many intercrossing cycles are needed to better recombine the parents’ alleles. The objective in this case is panmixia. On a theoretical basis, Hanson (1959) concluded that three crossing cycles were needed. Marín-Garavito (1994), working on a recurrent selection population, did not find differences among the phenotypes of the first and third cycles of recombination. Nor did Ospina et al. (Chapter 17, this volume) find differences between the first and second cycles of recombination.

Such a question can also be answered by using co-dominant molecular markers. Samples of the original populations and S₀ seeds of the consecutive cycles of recombination must be available. Badan et al. (1998) and Geraldi and Souza (2000), using means and variances of several rice populations for several traits, demonstrated that a population of 200 plants was adequate. We recommend this sample size for additional evaluations.

For each marker, the allelic and genotypic frequencies in the populations and Nei’s diversity index (Nei, 1972) can be computed. These data can be used to analyse deviations from panmixia, using the Wright fixation index.

Ferreira et al. (2000), using seven microsatellite markers on a sample of 60 plants, showed that, after three recombination cycles, the CNA-5 population had lost alleles at all loci. Meanwhile, there was an increase in the frequency of alleles contributed by the male-sterility line.

In contrast, Filloux (2001) used 22 microsatellite markers on a sample of 150 individuals extracted from a population that had undergone five recombination cycles. The total absence of a deficit in heterozygosity (F_IS = 0.001) indicated that no deviation from panmixia occurred and that crossing with the male-sterility gene respected the Hardy-Weinberg laws. However, some loci had, individually, shown deviation from panmixia, whether with an excess (RM13 and RM222 on chromosomes 5 and 10) or deficit (RM261 on chromosome 4) of heterozygosity.

Both studies indicated that the chromosomal location of the markers chosen to evaluate diversity had consequences. In a cross where seeds are harvested only from male-sterile plants, selection is performed on the chromosome carrying the male-sterility gene from ‘IR36’ and, as a result, linkage drag would affect the markers linked to this gene.

To further check the efficiency of recombination in breaking unfavourable linkages, one strategy would be to select additional markers close to the markers under study. The segregation of this specific group of markers can then be compared in successive recombination cycles. The evolution of linkage disequilibrium between markers of increasing physical distances can be evaluated over several cycles. The recombinations, if they occur normally, should progressively break linkages, and linkage disequilibrium should decrease from cycle to cycle.

For studies on linkage disequilibrium, bi-allelic markers are making analyses easier but statistics for multi-allelic markers can be used when at least some alleles have high frequencies. A simulation study could be first carried out to determine how many markers-and their spacing-would be needed to achieve the most relevant results. For a given chromosome, tightly linked markers of known physical distances are available, thanks to the sequencing of the rice genome. Through sequence analysis, we can find microsatellite markers belonging to the same bacterial artificial chromosome (BAC) clone or contig, which offers a resolution that should be satisfactory for this type of study. The study can be extended to other chromosome arms to prevent bias resulting from the presence of any gene affecting recombination on the chromosome being studied.

Most research groups working with population improvement in rice use the male-sterility gene from ‘IR36’ to facilitate crossing. However, the recombination method may affect results, and it may be desirable to evaluate its effect, using a population made by manual crosses such as CIAT’s PCT-4 japonica population.

Effect of population improvement on genetic diversity

Several projects of recurrent selection, based on evaluations of S_0:2 families, are being carried out to improve various traits of agronomic importance. Examples include resistance to blast (Guimarães, 2000), tolerance of acid soils (Ospina et al., 2000) and yield. Progress made for the trait or traits under selection is usually evaluated after a few cycles of recurrent selection. This is done by comparing phenotypes of the initial population with those of the progeny from the last cycle.

Deliberate selection reduces genetic diversity to a level that depends on the initial variability and intensity of selection. Such reduction of diversity may not be harmful if the alleles eliminated are those that are unfavourable. Information on the effect of this selection strategy on total genetic diversity of populations, however, is not available. Because such populations, once improved for a trait, will be used to extract useful lines for specific conditions, it is important that the genetic diversity not targeted for selection is not too greatly reduced.

Molecular markers can be used to evaluate the degree of reduction of total diversity by comparing the diversity between successive recurrent cycles of the population. The markers can be chosen according to the allelic diversity in the subspecies used to construct the population (indica for tropical countries and japonica for temperate ones).

To conduct such a study, S₀ seeds must be available for several recurrent cycles of the population, with samples of about 200 individuals. Changes in allelic or genotypic frequencies can be monitored. These changes derive from a mixture of effects: genetic drift, unequal contributions by parents to progenies or selection. A large sample can counteract genetic drift. Unequal contribution may or may not occur in random crossing because of the male-sterility gene. Hence, it is more important to use S₀ seeds that result from intercrossing than the parents themselves or, better still, to analyse both samples to evaluate the importance of this factor. Furthermore, only samples with the same level of inbreeding should be compared (e.g. two S₀ or two S₂ samples). A final important point for a retrospective study involving parents is the availability of the same samples of seeds used for composing the population.

Results can be expected to vary according to the type of heredity of the traits under selection, which may range from oligogenic to highly multigenic. For example, a few major genes basically control tolerance of acid soils, although some minor ones can affect the trait. Blast resistance is controlled by the combined action of a series of relatively well-known major and minor genes located on all 12 rice chromosomes. Yield potential is a highly complex trait, resulting from the interaction of many minor genes. Selection will differently affect the chromosomes carrying the genes under selection and the non-carrier chromosomes.

For blast resistance, the genetics of which are more or less well understood, major and minor genes can be located relative to the chosen markers, using information available in the literature. This will allow assessing the degree to which results are affected by linkage with known resistance genes.

In recurrent selection, comparisons are made, not between varieties, but between populations. The number of alleles, and allelic and genotypic frequencies can be measured for each marker, permitting the computation of Nei’s diversity index for each marker and comparison of the various cycles. The allelic associations can be evaluated, using multivariate analyses. Factorial analysis table of distance (FATD) is a statistical tool that is commonly employed to graphically represent the relative positions of successive cycles.

Populations may be compared with each other, but also against known standards of diversity. For rice, ‘mini- GB’, a reference collection of 275 varieties, was established to represent the overall diversity of O. sativa. Isoenzymes, and hydrological and geographic criteria (Glaszmann et al. 1995), useful for identifying diversity were used. An important part of this collection was characterized, using microsatellite markers (Filloux, 2001). These data are available at CIRAD. Different subsamples of this collection, corresponding to different genetic backgrounds, can be used for specific comparisons (japonica lines for temperate-climate populations, and indica lines for tropical ones).

Filloux (2001) presents an example. He analysed the polymorphism of 22 microsatellite markers in a sample of 150 individuals derived from a restorer population for fertility. His goal was to obtain the necessary elements for optimizing a hybrid rice programme. Their diversity was compared with that of a set of male-sterile lines (which were being used in the hybrid rice programme) and with the ‘mini-GB’ indica group. Results showed an average of 4.3 alleles per locus in the restorer population, with a few loci fixed and a heterozygosity of 0.44. The population was showing diversity in terms of number of alleles, and the diversity index equalled to half of that existing in the indica core set. Results did not change if the number of microsatellite markers was reduced from 22 to 10, showing that the number of loci studied can be reduced.

Another example is the work carried out in Barbados with a recurrent population of sugar cane to improve sucrose content (Alleyne, 2004). The segregating ratios were determined for a group of markers among varieties used as parents and among progenies derived from three cycles of recurrent selection for sucrose content. The ratios were then compared to identify skewed markers, which were assumed to be linked with sucrose content. AFLPs were chosen because of the need for a large number of markers to cover the large genome of this polyploid species. Because scoring AFLP gels is time-consuming, only those markers that were visibly skewed were scored. Of 566 AFLP bands that were scored, 200 showed strong change in segregation ratios from parents to progenies of the third cycle, using Fisher’s exact test. The FATD graph confirmed the shift with most of the parents on one side of the main axis and the progenies on the other. The intermediate cycles had not been analysed but their genotyping would have been useful to determine the factors that caused the shift (selection, random drift, unequal contribution of parents to the progenies or a combination of these factors). For practical reasons related to its flowering capacity, random mating is not possible in sugar cane. Rice may be less prone to such rapid changes.

A last example: De Koeyer et al. (2001) measured the changes in allelic frequencies during seven recurrent selection cycles for yield in oat. Significant shifts were observed, but the trend was not regular for all loci and some were inconsistent across cycles. The authors assumed that these shifts were directly related to selection for yield because of the correlation between allelic frequencies at these loci and cycle averages for the agronomic traits.

Locating the male-sterility gene and evaluating linkage drag

The male-sterility gene from ‘IR36’ has been used in most populations to facilitate crossing. Molecular markers can be used to locate the gene in the rice genome. Identifying a codominant marker closely linked to the male-sterility gene would allow distinguishing heterozygotes from homozygotes for the male-sterility gene among male-fertile plants extracted from populations. Such selection would accelerate line fixation during line extraction. The easiest method for finding such markers is through bulked segregant analysis (Michelmore et al., 1991).

In this method, two separate DNA bulks are constituted: one from the pooled DNA of 10 male-sterile plants and the other from 10 male-fertile plants. The two pools are tested for allelic differences with a set of markers. Any method that enables rapid testing of a large number of markers is advisable.

AFLP markers revealed en masse are a first option, but they would be more difficult to use later, and their map locations are not directly known. Microsatellite markers can also be used to test linkage. These are not produced en masse but are easier to handle and their positions are known. They may also help in carrying out mapping more efficiently as two steps:

(1) permitting a rough map location
(2) increasing marker density in the vicinity of the markers that were first detected, thus establishing their map locations more precisely.

Preliminary results, using the approach described, showed that the male-sterility gene may be located on the short arm of chromosome 2, near RM154 (D. Filloux, 2001, unpublished data). We do not really know the effect of the male-sterility gene from ‘IR36’ on recurrent selection populations, or the extent and nature of the linkage drag associated with its use. The positions of all the QTLs identified as associated with the gene need to be confirmed that they are, in fact, within its vicinity. The literature should then be reviewed to obtain a preliminary idea of the traits that may be affected by selection favouring male-sterile plants.

To evaluate the extent of linkage drag during the phase of population improvement requires additional work. Because ‘IR36’ is an indica variety, the best background against which to study linkage drag is a japonica population. This study can be conducted in cycle 0, which represents the worst case in terms of linkage drag.

A sample of 200 plants would be taken and their genotypes determined for 10 markers that surround the male-sterility gene and which span a 50-cM interval. These markers would be chosen according to their capacity to differentiate indica alleles from japonica ones. The spacing between them can be optimized by using results from the recombination study. The size (in cM) of the indica segment can be evaluated in each of the 200 plants. During the population improvement phase, the harvest of male-sterile plants only will contribute to maintaining the ‘IR36’ alleles at these segments. In contrast, during line extraction, the elimination of the alleles for male sterility will eliminate ‘IR36’ alleles at these segments.

Marker-assisted selection for recurrent selection populations

One major use of molecular markers is to locate genes that control agronomic traits of interest on dense genetic maps and identify those markers closely linked with these genes. To facilitate mapping, specific bi-allelic populations, resulting from crosses between pure lines that correspond to the simpler genetic situation, are developed.

Mapping is more difficult with recurrent selection populations because the number of possible alleles per locus can be high. The relative value of each allele therefore needs to be evaluated. Although using the most advanced methods for QTL analysis for multi-allelic populations may not be possible, single-marker analysis based on ANOVA between the various marker classes is always possible. The recent interest in association studies involving natural populations should translate into improvements of the detection methods.

The size of the population being evaluated is important as it determines the power of detecting QTLs. Plant breeders usually evaluate 200 S_0:2 families during selection, which is probably a good compromise between power of detection and workload.

Such work was carried out for sugar cane with AFLP markers (Alleyne, 2004). The population studied comprised 141 clones resulting from three cycles of recurrent selection for sucrose content. Single-marker analysis was used to identify markers linked with juice brix and sucrose percent juice. The analyses were repeated over 3 years of data collection. Each year, QTLs were detected but, because of strong genotype-by-environment interactions, the positions of the QTLs were not consistent from year to year. However, at least one marker was significantly associated with juice brix over the 3 years. These results show the difficulty of identifying useful markers for traits of low heritability.

De Koyer et al. (2001) used RFLP markers to identify QTLs linked with yield, plant height, and flowering in a recurrent selection population of oat. The locations of the QTLs were consistent with those detected in a classic population of recombinant inbred lines.

Once markers tightly linked with traits of interest are identified, then the interest is to manipulate them. Marker-assisted selection is based on linkage disequilibrium, which is artificially increased through hybridization. Intercrossing induces recombinations and reduces linkage disequilibrium in a recurrent selection population. Thus, the positions of the QTLs need to be regularly re-evaluated, not for every cycle but at least every 2 years, as suggested by Hospital et al. (1997). Such re-evaluation may not be as tedious as it could seem. Phenotypic evaluation is carried out anyway in a classic programme of recurrent selection, and genotyping is necessary for evaluating changes in diversity. Genotyping can be focused on areas where QTLs were detected in the first round of analyses.

The problem does not occur if the polymorphic marker is directly within the gene or genes of interest. But this implies extensive work to identify such genes through positional cloning or the candidate gene approach.

The fact that phenotypic information is available means that several strategies, more efficient than selection based only on molecular data, can be used: selection based on genotype, followed by selection on phenotype to reduce costs; or combined selection on phenotype and genotype. The latter solution is the more powerful, and will produce a greater response to selection (Dekkers and Hospital, 2002).

Final comments

We have shown that molecular markers can be of real interest for evaluating and maintaining genetic diversity of populations subjected to improvement through recurrent selection. Through a more complicated procedure, these markers can also be useful in helping with selection, permitting greater genetic gains.

Moreover, markers can help answer methodological questions and increase the efficiency of recurrent selection in rice, even though the economic merit of using markers is not always obvious when they are used in addition to selection methods based on phenotype. With the availability of PCR-based markers, which are simple to use, the cost of such techniques is progressively decreasing and may become affordable in the framework of breeding programmes.

References

Akagi, H.; Yokozeki, Y.; Inagaki, A. & Fujimara, T. 1996. Microsatellite DNA markers for rice chromosomes. Theor. Appl. Genet., 94: 61-67.

Alleyne, S. 2004. Identification of molecular markers linked with sucrose production in a recurrent selection population of sugar cane. Cave Hill Campus, Barbados, University of the West Indies. 175 pp. (PhD dissertation)

Badan, A.C. de Carvalho; Geraldi, I.O.; Guimarães, E.P. & Ospina, Y. 1998. Estimativa do tamanho efetivo de amostra ideal para caracterizar uma população de arroz para as características peso de panículas e peso de 100 grãos. Genet. Mol. Biol., 21: 246.

Barry, M.B. 2000. Diversité génétique de variétés locales de riz en Guinée. Analyse de la variabilité intra-variétale au moyen de marqueurs microsatellites. Rennes, France, Université de Rennes. 30 pp. (DEA thesis)

Breseghello, F.; Rangel, P.H.N. & Morais, O.P. de. 1999. Ganho de produtividade pelo melhoramento genético do arroz irrigado no Nordeste do Brasil. Pesq. Agropecu. Brasil., 34(3): 399-407.

Brookes, A.J. 1999. The essence of SNPs. Gene, 234: 177-186.

Châtel, M. & Guimarães, E.P. 1997. Selección recurrente en arroz en África y Madagascar: Estado actual y progreso. In E.P. Guimarães, ed. Selección recurrente en arroz, pp. 151-162. Cali, Colombia, CIAT.

Chen, X.; Temnykh, S.; Xu, Y.; Cho, Y.G. & McCouch, S.R. 1997. Development of microsatellite framework map providing genome coverage in rice. Theor. Appl. Genet., 95: 553-567.

Courtois, B.; Nelson, R. & Roumen, E. 1997. Creación de un acervo genético para mejorar la resistencia parcial a Piricularia en arroz de secano, mediante selección recurrente. In E.P. Guimarães, ed. Selección recurrente en arroz, pp. 189- 202. Cali, Colombia, CIAT.

Cuevas-Pérez, F.E.; Guimarães, E.P.; Berrío, L.E. & González, D.I. 1992. Genetic base of irrigated rice in Latin America and the Caribbean, 1971 to 1989. Crop Sci., 32(4): 1054-1059.

De Koeyer, D.L.; Philips, R.L. & Stuthman, D.D. 2001. Allelic shifts and quantitative trait loci in a recurrent selection population of oat. Crop Sci., 41: 1228-1234.

De Vienne, D. 1998. Les marqueurs moléculaires en génétique et biotechnologies végétales. Paris, INRA Editions. 200 pp.

Dekkers, J.C.M. & Hospital, F. 2002. The use of molecular genetics in the improvement of agricultural populations. Nat. Rev. Genet., 3: 22-31.

Ferreira, M.E.; Pentedo, M.I. de O.; Brondani, C.; Beló, A.; Ferreira, M.A. & Rangel, P.H.N. 2000. Caracterización y uso de marcadores RAPD y microsatélites (SSR) en el monitoreo del programa de mejoramiento poblacional en arroz. In E.P. Guimarães, ed. Avances en el mejoramiento poblacional en arroz, pp. 37-62. Santo Antônio de Goiás, Brazil, Embrapa Arroz e Feijão.

Filloux, D. 2001. Analyse de la diversité génétique d’une population restauratrice de fertilité utilisée pour la création de variétés hybrides de riz. Montpellier, France, Université de Montpellier. 17 pp. (DEA thesis)

Geraldi, I.O. & Souza, Jr, C.L. de. 2000. Muestreo genético para programas de mejoramiento poblacional. In E.P. Guimarães, ed. Avances en el mejoramiento poblacional en arroz, pp. 9-20. Santo Antônio de Goiás, Brazil, Embrapa Arroz e Feijão.

Glaszmann, J.C. 1987. Isozymes and classification of Asian rice varieties. Theor. Appl. Genet., 74: 21-30.

Glaszmann, J.C.; Grivet, L.; Courtois, B.; Noyer, J.L.; Luce, C.; Jacquot, M.; Albar, L.; Ghesquière, A. & Second, G. 2003. Asian rice. In P. Hamon, M. Seguin, X. Perrier, J.C. Glaszmann, eds. Genetic diversity of cultivated tropical plants, pp. 77-98. Enfield, NH, USA, Science Publishers and CIRAD.

Glaszmann, J.C.; Mew, T.; Hibino, H.; Kim, C.K.; Vergel de Dios, T.I.; Vera Cruz, N.H.; Notteghem, J.L. & Bonman, J.M. 1995. Molecular variation as diverse source of disease resistance in cultivated rice. In Rice genetics III, pp. 460-465. Manila, Philippines, International Rice Research Institute (IRRI).

Graterol, M.E.J. 2000. Caracterización de poblaciones e introducción de variabilidad genética para iniciar un programa de mejoramiento poblacional del arroz en Venezuela. In E.P. Guimarães, ed. Avances en el mejoramiento poblacional en arroz, pp. 87-103. Santo Antônio de Goiás, Brazil, Embrapa Arroz e Feijão.

Guimarães, E.P. 2000. Mejoramiento poblacional del arroz en America Latina: Dónde estamos y para dónde vamos. In E.P. Guimarães, ed. Avances en el mejoramiento poblacional en arroz, pp. 299-311. Santo Antônio de Goiás, Brazil, Embrapa Arroz e Feijão.

Guimarães, E.P.; Borrero, J. & Ospina, Y. 1996. Genetic diversity of upland rice materials distributed in Latin America. Pesq. Agropecu. Brasil., 31(3): 187-194.

Hanson, W.D. 1959. The breakup of initial linkage blocks under selection mating systems. Genetics, 44: 857-868.

Hospital, F.; Moreau, L.; Lacoudre, F.; Charcosset, A. & Gallais, A. 1997. More on the efficiency of marker-assisted selection. Theor. Appl. Genet., 95: 1181-1189.

Jones, C.J.; Edwards, K.J.; Castiglione, S.; Winfield, M.O.; Sala, F.; Van de Wiel, C.; Bredemeijer, G.; Vosman, B.; Matthes, M.; Daly, A.; Brettschneider, R.; Bettini, P.; Buiatti, M.; Maestri, E.; Malcevschi, A.; Narmiroli, N.; Aert, R.; Volckaert, G.; Rueda, J.; Linacero, R.; Vasquez, A. & Karp, A. 1997. Reproducibility testing of RAPD, AFLP and SSR markers in plants by a network of European laboratories. Mol. Breed., 3: 381-390.

Karp, A.; Kresovich, S.; Bhat, K.V.; Ayad, W.G. & Hodgkin, T. 1997. Molecular tools in plant genetic resources conservation: a guide to the technologies. Technical Bulletin 2. Rome, International Plant Genetic Resources Institute (IPGRI). 27 pp.

Lima, M.L.A.; García, A.A.F.; Oliveira, K.M.; Matsuoka, S.; Arizono, H.; Souza, Jr, C.L. de & Souza, A.P. de. 2002. Analysis of genetic similarity detected by AFLP and coefficient of parentage among genotypes of sugar cane. Theor. Appl. Genet., 104: 30-38.

Lorenzen, L.L.; Boutin, S.; Young, N.; Specht, J.E. & Shoemaker, R.C. 1995. Soybean pedigree analysis using map-based molecular markers, I: Tracking RFLP markers in cultivars. Crop Sci., 35: 1326-1336.

Lynch, M. & Milligan, B.G. 1994. Analysis of population genetic structure using RAPD markers. Mol. Ecol., 3: 91-99.

Marín-Garavito, J.M. 1994. Efecto de número de ciclos de recombinación en la variabilidad de poblaciones de arroz (Oryza sativa L.). Palmira, Colombia, Facultad de Ciencias Agropecuarias, Universidad Nacional de Colombia. 50 pp. (BSc thesis)

McCouch, S.R.; Teytelman, L.; Xu, Y.; Lobos, K.B.; Clare, K.; Walton, M.; Fu, B.; Maghirang, R.; Li, Z.; Xing, Y.; Zhang, Q.; Kono, I.; Yano, M.; Fjellstrom, R; DeClerck, G.; Schneider, D.; Cartinhour, S.; Ware, D. & Stein, L. 2002. Development of 2240 new SSR markers for rice. DNA Res., 9: 199-220.

Michelmore, R.W.; Paran, I. & Kesseli, R.V. 1991. Identification of markers linked to disease resistance genes by bulked segregant analysis: a rapid method to detect markers in specific genomic regions using segregating populations. Proc. Natl. Acad. Sci. USA, 88: 9282-9832.

Nei, M. 1972. Genetic distance between populations. Am. Nat., 106: 283-292.

Ospina, Y.; Châtel, M. & Guimarães, E.P. 2000. Mejoramiento poblacional del arroz de sabanas. In E.P. Guimarães, ed. Avances en el mejoramiento poblacional en arroz, pp. 241-254. Santo Antônio de Goiás, Brazil, Embrapa Arroz e Feijão.

Panaud, O.; Chen, X. & McCouch, S.R. 1996. Development of microsatellite markers and characterization of simple sequence length polymorphism (SSLP) in rice (Oryza sativa L.). Mol. Gen. Genet., 252: 597-607.

Perrier, X.; Flori, A. & Bonnot, F. 2003. Les méthodes d’analyse des données. In P. Hamon, M. Seguin, X. Perrier, J.C. Glaszmann, eds. Genetic diversity of cultivated tropical plants, pp. 31-64. Enfield, NH, USA, Science Publishers and CIRAD.

Rangel, P.H.N.; Guimarães, E.P. & Neves, P. de C.F. 1996. Base genética das cultivares de arroz (Oryza sativa L.) irrigado do Brasil. Pesq. Agropecu. Brasil., 31(5): 349-357.

Rangel, P.H.N.; Pereira, J.A.; Morais, O.P. de; Guimarães, E.P. & Yokokura, T. 2000. Ganhos para produtividade de grãos pelo melhoramento genético do arroz (Oryza sativa L.) irrigado no meio norte do Brasil. Pesq. Agropecu. Brasil., 35(18): 1595-1604.

Santos, P.G.; Soares, P.C.; Soares, A.A. & Morais, O.P. de. 1997. Estimativa do progresso genético do programa de melhoramento de arroz (Oryza sativa L.) irrigado desenvolvido em Minas Gerais no período de 1974 a 1996. In Annals of the XXII Reunião da Cultura do Arroz Irrigado, 1997, Camburio, SC, Brazil, pp. 27-30. Itajai, SC, Brazil, EMPASC.

Singh, R.J. & Ikehashi, H.I. 1981. Monogenic male sterility in rice: induction, identification and inheritance. Crop Sci., 21(2): 286-289.

Soleimani, V.D.; Baum, B.R. & Johnson, D.A. 2002. AFLP and pedigree-based genetic diversity estimates in modern cultivars of durum wheat. Theor. Appl. Genet., 104: 350-357.

Syvänen, A.C. 2001. Assessing genetic variation: genotyping SNPs. Nat. Rev. Genet., 12: 930-942.

Tao, D.; Hu, F.; Yang, Y.; Xu, P. & Li, J. 2000. Yunnan, China: Mejoramiento poblacional del arroz para rendimiento de granos, resistencia a Piricularia, tolerancia al frío y calidad de grano. In E.P. Guimarães, ed. Avances en el mejoramiento poblacional en arroz, pp. 145-154. Santo Antônio de Goiás, Brazil, Embrapa Arroz e Feijão.

Temnykh, S.; DeClerck, G.; Lukashova, A.; Lipovich, L.; Cartinhour, S. & Mc- Couch, S.R. 2001. Computational and experimental analysis of microsatellites in rice: frequency, length variation, transposon association and genetic marker potential. Genome Res., 11: 1441-1452.

Temnykh, S.; Park, W.D.; Ayres, N.; Cartinhour, S.; Hauck, N.; Lipovich, L.; Cho, Y.G.; Ishii, T. & McCouch, S.R. 2000. Mapping and genome organization of microsatellite sequences in rice (Oryza sativa L.). Theor. Appl. Genet., 100: 697-712.

Vos, P.; Hogers, R.; Bleeker, M.; Reijans, M.; Van de Lee, T.; Hornes, M.; Fritjers, A.; Pot, J.; Peleman, J.; Kupier, M. & Zabeau, M. 1995. AFLP: a new technique for DNA fingerprinting. Nuclear Acids Res., 23: 4407-4414.

Wu, K.S. & Tanksley, S.D. 1993. Abundance, polymorphism and genetic mapping of microsatellites markers in rice. Mol. Gen. Genet., 241: 225-235.