Introduction

Common bean (Phaseolus vulgaris) is an important source of protein in human diets in many parts of the world. Half the grain legumes consumed worldwide are common beans (Broughton et al. 2003). Unfortunately, bean diseases considerably reduce the yields of common beans. Anthracnose, caused by the specialized hemibiotrophic fungus Colletotrichum lindemuthianum, is one of the most important diseases of common beans throughout the world, especially in tropical bean-producing regions of Latin America and Eastern Africa where climatic conditions favour disease development (Pastor-Corrales and Tu 1989). The genetics of anthracnose resistance has been studied for a long time, since this host/pathogen interaction was the first report of race-cultivar specificity (Barrus 1911, 1915; McRostie 1919). Several dominant resistance genes which condition resistance against different races of the fungus have been described, suggesting the occurrence of a gene-for-gene-type resistance (Flor 1955). These specific resistance genes were renamed Co-1 to Co-10 (McRostie 1919; Mastenbroek 1960; Bannerot 1965; Bannerot et al. 1971; Fouilloux 1976, 1979; Pastor-Corrales et al. 1994; Young and Kelly 1996; Alzate-Marin et al. 1997; Young et al. 1998; Geffroy et al. 1999; Alzate Marin et al. 2003). Use of resistant genotypes appears to be a reliable control strategy in developing countries where farmers cannot afford the purchase of pathogen-free seeds or fungicides.

At the molecular level, the recent addition of genomics to the more traditional experimental tools has strikingly increased our understanding of the molecular basis of gene-for-gene-type resistance (Michelmore 2003; Nobuta and Meyers 2005). It is now well recognized that the largest class of disease resistance (R) genes encode proteins presenting an amino-terminal nucleotide binding site plus a series of carboxy-terminal leucine-rich repeats (NBS-LRR) (Dangl and Jones 2001; Hammond Kosack and Parker 2003; McHale et al. 2006). Functional resistance genes belonging to this class have been identified in a large selection of plant species: monocots as well as dicots. They correspond to resistance genes effective against all types of pathogens and pests including fungi, bacteria, viruses, nematodes, oomycetes and insects (Dangl and Jones 2001, Hammond Kosack and Parker 2003; McHale et al. 2006). Consequently, the NBS-LRR sequences are now recognized as the “pillar” of plant disease resistance (Meyers et al. 2005; Chisholm et al. 2006).

Annotation of the Arabidopsis thaliana genomic sequence (Arabidopsis Genome Initiative 2000) has identified 149 NBS-LRR sequences (Meyers et al. 2003). The rice and Populus trichocarpa genomic sequences contain 297 and 480 NBS-LRR genes, respectively. These results confirm the abundance of this class of genes in any plant species genome (Zhou et al. 2004; Tuskan et al. 2006). However, only a small number of NBS-LRR sequences have been functionally characterized as resistance genes. For example, in A. thaliana, less than 15 NBS-LRR sequences are known to function as resistance genes (Hammond Kosack and Parker 2003; McHale et al. 2006). One of the main limitations to identify resistance genes is probably the lack of strains of a specific pathogen determining a disease resistance gene co-located with a “candidate” NBS-LRR sequence. This knowledge gap strongly reveals the necessity to increase the number of mapped disease resistance genes against specific strains of a pathogen. In common bean, an international consortium called “Phaseomics” has selected the Meso-american genotype BAT93 as a good candidate genotype to be sequenced (Broughton et al. 2003). Consequently, mapping disease resistance genes with this genotype as a parent is of great interest for future functional analyses.

In the present study, we used a strategy based on a hypothesis developed by Bennetzen and Hulbert (1992) stating that: “the number of resistance genes that can be detected is determined exclusively by the number of different isolates of the pathogen that have been identified and isolated”. Based on this hypothesis, seven previously identified strains of C. lindemuthianum (Geffroy et al. 1999), revealing differential reaction on two representative genotypes of the Andean (JaloEEP558) and Mesoamerican (BAT93) gene pools, were inoculated on a recombinant inbred lines (RILs) population derived from an intercross of these two genotypes. We mapped three additional anthracnose disease resistance genes (Co-x, Co-w, Co-u) on the integrated linkage map of common bean (Freyre et al. 1998). We also confirmed the complexity of the B4 resistance gene cluster, because 4 and 2 of the 7 tested strains revealed resistance specificities in the vicinity of two previously identified resistance genes localized at the B4 resistance gene cluster, Co-y from JaloEEP558 and Co-9 from BAT93, respectively. The Co-x resistance gene was easily mapped using traditional procedures, whereas other strains revealing digenic segregation required the implementation of a specific LOD score calculation in order to map the corresponding resistance genes (Co-u, Co-w).

Materials and methods

Plant and fungal material

An F9 recombinant inbred line (RIL) population, derived from the cross between Andean landrace JaloEEP558 and Mesoamerican breeding line BAT93 was used to map additional anthracnose disease resistance genes. These 77 RILs were used to set up an integrated linkage map of common bean (Freyre et al. 1998), and additional molecular markers, some of which correspond to genes of known function involved in resistance/defense have been subsequently added (Geffroy et al. 2000). BAT93 was developed at the Centro International de Agricultura Tropical (CIAT, Cali, Colombia) and was derived from a cross involving four Middle American genotypes (Veranic 2, PI207262, Jamapa, Great Northern Tara). JaloEEP558 is a selection from the Andean landrace Jalo obtained at the Estacao Experimental de Pato de Minas (Minas Gerais, Brazil). “La Victoire” is a French multi-susceptible common bean cultivar of Andean origin developed by the seed company “Tezier” (Valence-sur-Rhone, France).

A world collection of 177 C. lindemuthianum strains, from the Orsay (France) fungal library, was inoculated to the two parents BAT93 and JaloEEP558 (Geffroy et al. 1999). All these strains were known to attack the multi-susceptible cultivar “La Victoire”. Eight strains, from those exhibiting avirulence towards JaloEEP558 (and virulence on BAT93) or exhibiting avirulence on BAT93 (and virulence on JaloEEP558) had already been inoculated onto the RIL population, and the corresponding resistance genes were mapped (Geffroy et al. 1999). In the present study, we present results of inoculation of seven additional C. lindemuthianum strains, giving differential reactions on BAT93 and JaloEEP558. The strains were isolated from cultivated beans (the origin of the strain is indicated in parentheses): 100 (Costa Rica), E25 (Tanzania), 40 (Rwanda), 3616 (Rwanda), 82 (Costa Rica), E4 (Tanzania) and E42b (Tanzania).

Pathogenicity tests and disease scoring

Five seeds of each RIL were planted in individual pots filled with moist vermiculite. The parental lines BAT93 and JaloEEP558 and the highly susceptible cultivar « La Victoire » were included as controls. For each strain tested, three replicates of this experiment were carried, i.e. the resistance/susceptible phenotype was assessed on at least 15 seedlings for each RIL’s genotype. Seedlings were grown for 8 days at 23°C, 75% relative humidity, with 8 h dark and 16 h light photoperiods under fluorescent tubes (166μE). The seedlings, at the two leaves stage, were infected by spray-inoculation 8 days after germination and incubated at 19°C under the light conditions described above with 90% relative humidity in a control growth chamber. The inoculum was prepared by harvesting conidia by flooding C. lindemuthianum plates with distilled sterile water. The spore concentration was measured using a haematocytometer and adjusted to 106 spores/ml.

Symptoms were scored 7 days after inoculation. Plants without visible symptoms or showing very limited necrotic lesions were scored as resistant. Plants with large, sporulating lesions and dead plants were scored as susceptible. That is, plant reactions were scored qualitatively. For resistance gene nomenclature (Kelly and Young 1996), putative new specificities were labelled Co- (Co lletotrichum) plus a letter (for example Co-x, Co-w, Co-u). The mode of action of the identified R genes is not provided in the present publication.

Genetic mapping of the anthracnose resistance genes

Chi-squared (χ2) tests were used to evaluate the goodness of fit of observed and expected segregation ratios.

For the strains revealing a single resistance gene (1:1 segregation among RILs), the MAPMAKER software version 3.0 (Lander et al. 1987) was used to map the segregating resistance gene on the integrated linkage map of common bean. More precisely, we used the set of 142 makers in Geffroy et al. (2000). Linkage groups were established with an LOD threshold of 3.0 and a maximum recombination fraction of 0.3. Marker order was estimated with an LOD threshold of 2.0 based on multipoint “Compare”, “Order” and “Ripple” analyses. Map distances were estimated by the Kosambi function (Kosambi 1944).

For the strain revealing two independent resistance genes (3:1 segregation in RILs population), because the MAPMAKER software is not able to handle this case, a specific numerical code calculating the LOD score (Allard 1956) was written in FORTRAN. We developed a maximum likelihood procedure to evaluate the linkage relationships between each two independent genes (denoted G1 and G2) controlling a single phenotype and other segregating molecular markers (denoted M) in a single recombinant inbred population. Classically, the recombination rate per meiosis is denoted r, while the proportion of recombinant zygotes in RILs is denoted R. The relation between r and R for two loci M (marker) and G1 (resistance gene) in a RIL population was established in Haldane and Waddington (1931):

$$ R = \frac{{2r}} {{1 + 2r}} $$

The frequencies of the different genotypic classes are given in Table 1. The formulas of Haldane and Waddington were the basis for linkage analysis in RILs, especially in the Mapmaker software (Lander et al. 1987), where RIL data are handled as if they were backcross data, i.e. produced by a single meiosis, except that r is replaced by R to account for multiple generation effects in RILs (Martin and Hospital 2006). When a 2nd resistance gene (referred to as G2), segregating independently of G1, is also present, the frequencies of the different genotypic classes are as given in Table 2. At the phenotypic level, it was impossible to know the genotype of a resistant RIL. Therefore, the frequencies of the different phenotypic classes were obtained by summing the frequencies of the genotypic classes giving a resistant phenotype (Table 3).

Table 1 Frequencies of genotypic classes in a recombinant inbred lines (RIL) population (with an infinite number of generations) when two loci are separated by recombination rate per meiosis r
Full size table
Table 2 Frequencies of genotypic classes for the case when two independent resistance genes (G1 and G2) are present in parent A, and the recombination frequency between G1 and marker locus M is r
Full size table
Table 3 Frequencies of phenotypic classes, when two independent resistance genes are present in parent A and when there is a recombination rate r between G1 and marker M
Full size table

If the marker M and one of two independently segregating resistance genes, G1 or G2, are linked with recombination rate r, the probability of linkage is given by:

$$ P(r) = \frac{{(N)!}} {{a!b!c!d!}}x^{a} y^{b} z^{c} t^{d} $$

where a, b, c, d are the observed numbers in each phenotypic class (Table 4), N = a + b + c + d, and x, y, z, t are the expected frequencies of the different phenotypic classes under the hypothesis of linkage between the marker M and one of the two independently segregating resistance genes in the population (Table 3).

Table 4 Abbreviations used for the maximum likelihood expression
Full size table

The LOD is defined by:

$$ {\text{LOD}} \equiv {\text{Log}}_{{10}} \frac{{P(r)}} {{P{\left( {r = \frac{1} {2}} \right)}}} = a.\log {\left( {\frac{4} {3}(1 + r)} \right)} + b.\log (4r) + c.\log {\left( {\frac{2} {3}{\left( {4r + 1} \right)}} \right)} + d.\log 2 - N.\log (1 + 2r) $$
(1)

where \( P{\left( {r = \frac{1} {2}} \right)} \) is the probability of independence. The value of r maximizing the probability to have the observed numbers (a, b, c, d) is given by:

$$ \frac{{{\text{dLOD}}}} {{{\text{d}}r}} = \frac{a} {{1 + r}} + \frac{b} {r} + \frac{{4c}} {{4r + 1}} - \frac{{2N}} {{1 + 2r}} = 0 $$
(2)

This is equivalent to finding the real roots in the interval [0; 0.5], of the following third-order polynomial:

$$ - 8r^{3} d + r^{2} {\left( {6a + 14b + 12c - 10N} \right)} + r{\left( {a + 7b + 4c - 2N} \right)} + b = 0. $$

The recombination rate r is obtained numerically by a Newton Raphson method and the associated LOD value is calculated with formula (1). To declare a significant linkage between a marker and a resistance gene, we retained a minimal LOD value of 2.5. The case of two complementary resistance genes is similar and explained in Appendix.

Results

Anthracnose resistance genes in BAT93 and JaloEEP558

Resistance specificities detected in the Andean parent JaloEEP558 will be referred to as Andean specificities. In the same way, resistance specificities identified in the Mesoamerican parent BAT93 will be named Mesoamerican specificities. For strain “100” the distribution of the RILs, conformed to a ratio of one homozygous parental type resistant to one homozygous parental type susceptible (Table 5), indicating that resistance of JaloEEP558 against this strain was controlled by a single gene, temporarily designated Co-x. Co-x was mapped straightforwardly using Mapmaker 3.0 on the bean genetic map at a distal position on linkage group B1, 12.4 cM from marker PROE8b (Fig. 1). For strains E25, 40, 3616, 82 (JaloEEP558 resistant) or strains E4, E42b (BAT93 resistant) the distribution of RILs, conformed to a ratio of three homozygous parental type resistant to one homozygous parental type susceptible (Table 5), indicating that resistance was controlled by two independent genes. Segregations towards strains 40, 3616 and 82 (resistance present in JaloEEP558) were identical. A few recombinations were detected between the resistance to strain E25 and the resistance to the latter three strains. Similarly, resistances to strains E4 and E42b showed perfect co-segregation.

Table 5 Name and origin of the Colletotrichum lindemuthianum strains inoculated on the RILs population
Full size table
Fig. 1
figure 1

Distribution of genes for resistance to anthracnose (Co-) on the BAT93 × JaloEEP558 recombinant inbred lines (RILs) linkage map, based on markers used in Geffroy et al. (2000). Numbers in rectangles adjacent to a Co- resistance gene represent strains, recognized by the considered resistance gene in the present study. B1, B2 and B4 are linkage groups relevant to this study. Designations to the right represent names of markers. A scale in centimorgans is shown on left. Co-x was mapped with Mapmaker 3.0 while Co-w, Co-u were mapped using an LOD score calculation for two independent genes as described in the present study

Full size image

LOD score calculation to map two independent resistance genes (i.e. genes with duplicate effects)

Mapmaker 3.0 is not adequate to map two independent resistance genes (digenic segregation). This situation corresponds to genes with duplicate effects. A specific numerical code calculating the LOD score, was developed as described in Materials and methods section.

For each of the six strains revealing a digenic segregation, each of the 142 markers of the bean map used in Geffroy et al. (2000) with addition to Co-x identified in the present study, was tested for linkage with one of the two segregating resistance genes. For the strains E4 and E42b, two significant regions of the common bean genome were detected (Table 6; Fig. 1): one resistance gene was mapped in the vicinity of the I gene, which confers resistance against various viruses (Vallejos et al. 2006), at a distal position on linkage group B2 (r = 0.05), and a second resistance gene was mapped to the Co-9 locus on linkage group B4 (r = 0.00). This locus has been previously identified as a locus that controls resistance to C. lindemuthianum (Geffroy et al. 1999). For the latter, we assumed that it was Co-9, because it is impossible to determine if these two strains are recognized by Co-9 itself or by resistance gene(s) closely linked to Co-9. The Meso-american resistance gene located in the vicinity of the I gene is referred to as Co-u (Table 6; Fig. 1).

Table 6 Linkage between markers and resistance genes detected by the LOD score calculation adapted for two independent resistance loci segregating in a recombinant inbred line population
Full size table

For strains E25, 40, 3616 and 82, two significant regions of the common bean genome were detected: a first resistance gene was mapped in the vicinity of the Co-x anthracnose resistance gene at the end of linkage group B1, and a second gene was co-located with the Co-y anthracnose resistance gene on linkage group B4 (r = 0.00) (Geffroy et al. 1999). For strains, 40, 3616 and 82 (Table 6) one of the resistance genes co-located with the Co-x resistance gene (r = 0.00), whereas for strain E25 an r value of 0.09 separated one of the resistance genes and Co-x. For strain E25, three RILs were confirmed as susceptible to strain E25, but resistant to strain 100, which is recognized by the resistance gene Co-x. This result enabled us to infer that strain E25 is not recognized by the Andean resistance gene Co-x, but by a closely linked resistance gene. This gene was designated Co-w. We considered that the second resistance gene is Co-y (Table 6, Fig. 1) for the four strains E25, 40, 3616 and 82, because it is impossible to determine, due to the small size of the tested population, if they are recognized by Co-y, or by resistance gene(s) closely linked to Co-y.

Discussion

In the present study, three anthracnose resistance specificities were mapped on the integrated linkage map of common bean. Co-u from the Meso-american genotype BAT93 mapped in the vicinity of the I resistance gene at the end of linkage group B2. Co-x and Co-w are two closely linked genes from the Andean JaloEEP558 genotype that are mapped at the end of linkage group B1. One of them (Co-x) segregating in a 1:1 ratio in the RIL population, was easily mapped using Mapmaker software (Lander et al. 1987). Strains E4 and E42b, and E25, 40, 3616 and 82 revealed segregation of two independent resistance genes. The development of an LOD score calculation for two independent segregating genes in a RIL population was made to map the latter genes. The problem of mapping two segregating genes on a genetic map was addressed by Caranta et al. (1996) by using χ2 statistics to map corresponding resistance genes on the pepper (Capsicum annuum) map. Our method based on an LOD score calculation was the same as that used by Mapmaker (Lander et al. 1987) and has the advantage of providing the recombination value (r) between the resistance gene and the tested marker and the associated LOD score measuring its likelihood. Although, we developed this calculation for the case of two independent genes in a RIL population, it can be easily adapted to the case of two complementary genes (Appendix). Digenic segregations are often observed in disease resistance gene studies indicating that two avirulence gene products produced by a specific strain are recognized by two disease resistance genes. For fungal pathogens such as C. lindemuthianum, with no unequivocal sexual stage yet identified, it is impossible to obtain a strain with a single Avr gene by crossing.

Combination of Andean and Meso-american resistance genes was proposed as a way of achieving durable resistance against anthracnose in common bean. Indeed, adaptation of the C. lindemuthianum strains to plants of the same origin was observed in wild populations of common bean growing in the centre of diversity of Phaseolus vulgaris as well as with domesticated beans (Geffroy et al. 1999). The Andean Co-x resistance gene appears to be of particular interest for this strategy because it confers resistance to strain “100” which is virulent to G 2333, a germplasm accession of Mexican origin and resistant to a broad range of C. lindemuthianum strains (Pastor-Corrales et al. 1994). Therefore, combining the anthracnose Andean resistance gene Co-x from JaloEEP558 with the anthracnose resistance genes from G2333 of Meso-american origin has great potential. Combining resistance genes can be facilitated by the use of molecular markers linked to the relevant genes. Placing resistance genes on a map provides molecular markers that might be used directly in marker-assisted selection (MAS). In common bean, correlations have been established between several published maps (Vallejos et al. 1992; Adam-Blondon et al. 1994; Freyre et al. 1998; Blair et al. 2003), providing additional candidate markers for a target region. For example, in the case of Co-x, mapped at the end of linkage group B1, corresponding to linkage group H of the map from Vallejos et al. (1992), restriction fragment length polymorphism (RFLP) markers Bng173, Bng171, Bng122, are of particular interest. Some of these RFLP markers have been converted into PCR-based markers (Murray et al. 2002), facilitating their use in MAS strategies. Interestingly, these RFLP markers were used in soybean to increase the map density in syntenic regions of the soybean genome known to carry a soybean cyst nematode resistance (rhg1) (Concibido et al. 1996; Foster Hartnett et al. 2002). The syntenic region of M. truncatula has been sequenced but it does not seem to carry a rhg1 homolog (Mudge et al. 2005). Synteny for disease resistance genes is an open issue in plant genome. Indeed, synteny for resistance genes (macro and/or micro) has been observed between several species especially in the Solanaceae family (Lefebvre 2004), while in cereal genome rapid re-organization of R-like sequences has been reported (Leister et al. 1998).

Allelism tests involving JaloEEP558 and the Andean genotype MDRK suggested that JaloEEP558 and MDRK possess the same allele at the Co-1 locus (Kelly and Vallejo 2004). So far, the location of Co-1 resistance at the end of linkage group B1 has only been tentative, as mapping involved molecular markers linked to Co-1 (Kelly and Vallejo 2004). Consequently, assuming that the C. lindemuthianum strain mentioned in Kelly and Vallejo (2004) and strain “100” used in the present study are recognized by the same resistance specificity from JaloEEP558, our results give the first precise genetic map position of Co-1 locus (through the mapping of Co-x). Our results also provide evidence that both alleles (Melotto and Kelly 2000) and closely linked resistance genes occur at or near the Co-1 locus, as we identified two different resistance specificities (Co-w, Co-x) with different C. lindemuthianum strains in JaloEEP558. The question of allelism versus linkage is difficult to solve with small segregating populations, due to the extreme difficulty of proving that any two putative “alleles” of a cluster of disease resistance genes are truly allelic (Bennetzen and Hulbert 1992). In common bean, sequencing of the B4 resistance gene cluster, located on linkage group B4, revealed that 11 NBS-LRR sequences were located within less than 150 kb (V. Geffroy, unpublished results). Therefore, the probability of identifying at least one recombinant in small segregating population is extremely low.

In the present study, an anthracnose resistance gene Co-u was mapped in the vicinity of the I locus which controls the development of four different phenotypes in response to inoculation with Bean common mosaic virus, 10 other related potyviruses, and 1 comovirus (Vallejos et al. 2006). Therefore, in the same region of the bean genome, resistance genes conditioning resistance against viral and fungal pathogens are clustered. Comparative mapping studies within several plant genera have shown that genes for resistance are often co-localized on maps (Hulbert et al. 2001; Ashfield et al. 2003; Hayes et al. 2004; Marczewski et al. 2006). The two resistance genes (I and Co-u) might have a common origin at the molecular level; it is now well documented that most resistance genes in plants encode NBS-LRR proteins that are involved in the recognition of diverse pathogens and pests, and that these NBS-LRR sequences are often clustered (Meyers et al. 2003; McHale et al. 2006). Recently, Vallejos et al. (2006) identified a complex family of TIR-NBS-LRR genes at the I locus; one of these sequences might condition resistance against a virus and another one against C. lindemuthianum.

Our results confirm the complexity of the B4 resistance gene cluster because of seven randomly chosen strains of C. lindemuthianum from various geographical origins, two identified a resistance specificity in the vicinity of the Meso-american Co-9 gene and four identified an resistance specificity in the vicinity of the Co-y gene. The corresponding resistance genes referred to as Co-9 and Co-y, should be considered as a minimum number of functional resistance genes identified by the numerous strains. We previously identified 3 and 4 strains revealing the existence of a resistance specificity in the vicinity of Co-9 and Co-y, respectively (Geffroy et al. 1999). At the molecular level, we have confirmed the complexity of this cluster (Ferrier-Cana et al. 2003; Ferrier Cana et al. 2005), and we have now sequenced nearly 20 complete NBS-LRR sequences in the area of Co-9 and Co-y (V. Geffroy; unpublished results), again confirming that the B4 clusters in both BAT93 and JaloEEP558 are complex cluster and composed of tightly-linked specific resistance genes.