Rust of flax and linseed caused by Melampsora lini

Transcription

1 MOLECULAR PLANT PATHOLOGY (2007) 8(4), DOI: /J X Blackwell Publishing Ltd Pathogen profile Rust of flax and linseed caused by Melampsora lini GREGORY J. LAWRENCE*, PETER N. DODDS AND JEFFREY G. ELLIS CSIRO Plant Industry, GPO Box 1600, Canberra, ACT 2601, Australia SUMMARY Melampsora lini, while of economic importance as the causal agent of rust disease of flax and linseed, has for several decades been the model rust species with respect to genetic studies of avirulence/virulence. Studies by Harold Flor demonstrated that single pairs of allelic genes determine the avirulence/virulence phenotype on host lines with particular resistance genes and led him to propose his famous gene-for-gene hypothesis. Flor s inheritance studies, together with those subsequently carried out by others, also revealed that, in some cases, an inhibitor gene pair and an avirulence/virulence gene pair interact to determine the infection outcome on host lines with particular resistance genes. Recently, avirulence/virulence genes at four loci, AvrL567, AvrM, AvrP4 and AvrP/AvrP123, have been cloned. All encode novel, small, secreted proteins that are recognized inside plant cells. Yeast two-hybrid studies have shown that the AvrL567 proteins interact directly with the resistance gene protein. The molecular basis of Flor s gene-for-gene relationship has now been elucidated for six interacting gene pairs: those involving resistance genes L5, L6, L7, M, P and P2, where both the resistance gene and the corresponding avirulence gene have been cloned. In other inheritance studies it has been shown that M. lini does not possess a (+) and ( ) mating system, but may possess a two factor system. Double-stranded (ds) RNA molecules occur in many strains of M. lini: examination of the progeny of one strain that possesses 11 dsrna molecules revealed that they fall into three transmission units, designated L, A and B. The L unit consists of a single large dsrna of 5.2 kbp while the A and B units each consist of five dsrnas in the size range kbp. The three units have different sexual and asexual transmission characteristics. The L unit is encapsidated in a virus-like particle, whereas the other units are not encapsidated. The population and coevolutionary aspects of M. lini on a wild, native Australian host species, Linum marginale, have been extensively investigated. A recent molecular analysis revealed that the M. lini *Correspondence: Tel.: ; Fax: ; greg.lawrence@csiro.au isolates from L. marginale fall into two distinct lineages, one of which is apparently hybrid between two diverse genomes. Isolates in this lineage are largely fixed for heterozygosity, which suggests that sexual recombination does not occur in this lineage. INTRODUCTION Melampsora lini (Ehrenb.) Desm., the fungal pathogen responsible for rust disease on flax and linseed (Linum usitatissimum L.), is of interest for both economic and scientific reasons. It can cause severe losses in seed yield as well as reducing fibre quality in flax plants grown for linen production; consequently, breeders need to pay particular attention to ensuring that any new variety released to growers is genetically resistant to all races of the rust in the geographical area that it is intended to be grown. From the scientific viewpoint, M. lini came to prominence in 1942 when Harold Flor reported the results of an inheritance study which demonstrated that single allelic gene pairs determine the avirulence/virulence phenotype on host lines with particular resistance genes, which led him to propose his now famous gene-for-gene hypothesis. This states that pathogenicity of a physiologic race of Melampsora lini is determined by pathogenic factors specific for each resistance factor possessed by the host. Subsequent numerous inheritance studies with M. lini by Flor and others, together with host inheritance studies, provided substantial support for the gene-for-gene model, but also revealed some additional complexity in that, in some cases, an inhibitor gene pair and an avirulence/virulence gene pair were shown to interact to determine the virulence/avirulence phenotype on host lines with particular resistance genes. The numerous inheritance studies with M. lini are reviewed here: these studies are a feature of M. lini since very few inheritance studies have been undertaken with other rust species. Other features of work with M. lini include the cloning of the first avirulence genes from a rust species, investigations of the sexual and asexual transmission of encapsidated and unencapsidated double-stranded (ds) RNA molecules found in flax rust, studies of the genetic control of mating type and an extensive 2007 BLACKWELL PUBLISHING LTD 349

2 350 G. J. LAWRENCE et al. Fig. 1 Life cycle of flax rust, Melampsora lini. examination of population and coevolutionary aspects of a wild host/rust pathogen association. These are also reviewed here, along with studies of the life cycle, infection and development and axenic culture of M. lini. BIOLOGY OF MELAMPSORA LINI Melampsora lini is currently placed in the family Melampsoraceae, in the order Uredinales, in the class Urediniomycetes of the phylum Basidiomycota. It has a chromosome number of n = 18 (Boehm and Bushnell, 1992), and an estimated genome size of 170 Mb (Eilam et al., 1992; Leonard and Szabo, 2005). It has quite a wide host range in the genus Linum, being reported to occur, for example, on numerous European species (Gäumann, 1959) on several North American species (Anon., 1960), on the sole Australian species, L. marginale (McAlpine, 1906), and on the sole New Zealand species, L. monogynum (Gäumann, 1959). M. lini has also been found on 12 of the 13 species in the genus Hesperolinon in the field with the remaining species having been successfully infected in the laboratory (Springer, 2006) and an isolate from cultivated flax in the Netherlands was found to be virulent on an accession of Radiola linoides (Kowalska and Niks, 1998). The life cycle of M. lini consists of all five possible spore stages (it is therefore a long-cycle or macrocyclic rust) with the asexual cycle and all stages of the sexual cycle occurring on the same host species (autoecious). The various stages in the life cycle are shown in Fig. 1. The dikaryotic mycelium growing in susceptible MOLECULAR PLANT PATHOLOGY (2007) 8(4), BLACKWELL PUBLISHING LTD

3 Melampsora lini, rust of flax and linseed 351 flax plants produces dikaryotic urediniospores. These are the asexual repeating stage of the life cycle. Subsequently, the dikaryotic mycelium produces telia, usually on the stems of the plants, which comprise a subepidermal layer of single-celled, columnar teliospores. The thick-walled, pigmented teliospores are resistant to adverse environmental conditions and enable the fungus to survive between growing seasons. During teliospore development the two nuclei in the teliospore fuse to give a diploid nucleus which, shortly after, enters into meiosis. However, meiosis is suspended at the end of prophase 1 (diakinesis) of the first meiotic division during the period of teliospore dormancy (Boehm and Bushnell, 1992). Following the loss of dormancy meiosis is completed in the germinating teliospore and the four resulting haploid nuclei migrate to a metabasidium, a hyphal protrusion formed on the plant surface by the germinating teliospore. The metabasidium becomes separated by septa into four cells, each of which develops a conical sterigma that bears a haploid basidiospore that is discharged when mature. Freshly discharged basidiospores are uninucleate, but the nucleus undergoes mitosis almost immediately (Kapooria, 1973) to give mature single-cell basidiospores with two identical haploid nuclei, each derived from one of the four direct products of meiosis. A basidiospore, ejected from its sterigma on to a leaf, will germinate and infect the leaf to form a monokaryotic mycelium that develops, after 8 10 days, into a pycnium that possesses a specific, genetically determined, mating type. The pycnia produce haploid pycniospores (spermatia) in a liquid exudate (nectar). M. lini is heterothallic, so that transfer of nectar containing pycniospores from one pycnium to another of different mating type is required to initiate the production of dikaryotic aeciospores in an aecium. Aeciospores infect susceptible flax plants and, after 9 10 days, urediniospores are produced. In nature, nectar transfer between pycnia is accomplished by insects attracted to the nectar, by water splashing, dripping or running down plants or by the coalescing of two infections growing in close proximity. As pointed out by Flor (1956), the dikaryotic aeciospores in an aecium and the urediniospores subsequently derived from the aeciospores are genetically equivalent to the diploid progeny resulting from the union of haploid gametes in animals and plants. The ultrastructure of mycelial growth of M. lin in the host plant, and of the various spore types in its life cycle and their development, have been examined in detail (Allen, 1934; Fromme, 1912; Gold and Littlefield, 1979; Hassan and Littlefield, 1979; Littlefield and Bracker, 1972; Littlefield and Heath, 1979). Culture techniques for both the asexual and the sexual cycles have been described (Lawrence, 1988). INFECTION AND DEVELOPMENT Dikaryotic urediniospores will germinate on the surface of a leaf under conditions of high humidity or free water. Temperatures above C will inhibit germination. As detailed by Littlefield and Bracker (1972), a germ tube from a urediniospore develops a swollen tip (appressorium) over a stomate from which an infection peg penetrates between the guard cells to form a substomatal vesicle in the intercellular space beneath the guard cells. A hypha (or hyphae) from the substomatal vesicle begins to ramify intercellularly through the leaf tissue, branching as it goes, eventually forming an extensive network of hyphae. Hyphae, when in contact with a host mesophyll cell, may form a haustorial mother cell. A peg from the haustorial mother cell penetrates the cell wall of the plant mesophyll cell then elongates and expands inside the host cell to form a spheroid or lobed haustorium. The haustorium invaginates the plasma membrane of the host cell as it expands, so that the cytoplasm of the haustorium is separated from the cytoplasm of the host cell by the haustorial plasma membrane, the haustorial wall, an extrahaustorial matrix layer and the plasma membrane of the host cell. Nuclei migrate into the haustorium from its mother cell. Monoclonal antibodies specific to antigens in the cell wall of the haustorium have been isolated (Murdoch et al., 1998), a finding that emphasizes the specialized nature of this structure. The haustoria fulfil roles as feeding structures and as secretory organs. Some proteins secreted by the haustoria probably enter the plant cell (Dodds et al., 2004): these are likely to have functions that suppress the resistance responses of the host and alter host cell metabolism in ways that benefit the rust. Identifying these secreted proteins and determining their function is currently a major focus of research in plant pathology. The development of uredinia from the hyphal matrix has been examined by Hassan and Littlefield (1979). Uredinia are produced 6 10 days after inoculation with urediniospores. Development begins with a loose aggregation of undifferentiated hyphae in a substomatal cavity that orientate towards the leaf surface. The ends of the hyphae enlarge, then divide to form uredinia initial cells each of which divides again to form a basal and a terminal cell. The basal cell becomes the sporogenous cell from which a succession of urediniospores is produced. The terminal cell divides transversely to form a terminal peridial cell and an intercalary cell. The peridial cells adhere laterally with other peridial cells to form a unicellular layer while the intercalary cells below them collapse and disintegrate. From the basal, sporogenous cell urediniospore formation is initiated by the outgrowth of a spore bud that elongates and, in association with nuclei division, develops into two cells; one of the two cells (the terminal cell) develops into a urediniospore while the other develops into a pedicel that subsequently shrivels to free the urediniospore. A series of spore buds from a sporogenous cell can produce a succession of urediniospores. During urediniospore formation long cylindrical cells (paraphyses) also develop, apparently originating from some of the sporogenous cells. These paraphyses, on reaching their full length, develop enlarged, ovate heads. The role of 2007 BLACKWELL PUBLISHING LTD MOLECULAR PLANT PATHOLOGY (2007) 8(4),

4 352 G. J. LAWRENCE et al. the paraphyses seems to be to push up the layer of peridial cells and host epidermal cells resulting in their rupture, thereby allowing the urediniospores to be released. In the monokaryotic phase of the life cycle of M. lini (described in detail by Allen, 1934; Fromme, 1912; Gold and Littlefield, 1979) infection is initiated by germination of basidiospores on the leaf surface under conditions of high humidity. The germ tube, if unbranched, produces an enlarged terminal appressorium before directly penetrating the leaf epidermis via a peg-like infection hypha from the appressorium. However, germ tubes may branch, in which case the thinner branch tube will also directly penetrate the epidermis, but without appressorium formation. The initial infection hypha usually penetrates directly into an epidermal cell from which it emerges to produce a network of hyphae (whose cells contain 1 4 nuclei, Allen, 1934) that ramifies intercellularly through the leaf tissue. These monokaryotic hyphae also develop structures inside plant cells that Allen (1934) refers to as haustoria but which Gold and Littlefield (1979) call intracellular hyphae. Intracellular hyphae develop from intercellular hyphae either as lateral branches or terminal extensions or by continued growth from one host cell into another. Intercellular hyphae gain entrance to the host cell by broad penetration pegs and develop inside the living plant cell into filamentous, hyphal-like structures which may be multilobed and extensively branched and which are frequently divided into cells by the formation of septa (which contain from one to seven nuclei, Allen, 1934). The pycnial stage is initiated about 6 days after infection when hyphae from the mesophyll region of the leaf grow towards the epidermis and then aggregate to form pycnial primordia, the growth of which is focused towards the stomatal openings in the epidermis. Within the pycnial primordia specialized hyphae (pycniosporophores), which are often branched, arise, from the ends of which chains of pycniospores are produced which are subsequently exuded through the ostiole of the pycnial fruiting body on to the leaf surface in droplets of honey dew or nectar. Adjacent to the pycniosporophores, longer, more slender and less branched flexuous hyphae arise which, in maturing pycnia, extend through the stomata. Each basidiospore infection can give rise to a cluster or aggregation of pycnial fruiting bodies which, because they appear as a single infection on a leaf, are commonly described as a single pycnium. How do pycniospores transferred from one pycnium to another effect fertilization? One possibility is that the flexuous hyphae serve as receptive hyphae for the germ tubes of the transferred pycniospores. Some support for this comes from Gold and Littlefield (1979) who refer to preliminary observations (unpublished) of fusion between pycniospore germ tubes and ostiolar filaments (= flexuous hyphae?). However, Allen (1934) could find no evidence of such fusion events. Instead, Allen observed the fine germ tubes from the introduced pycniospores making intimate contacts with hyphae of the recipient pycnium and then growing on, larger in size, suggesting that the introduced pycniospores initially establish a mycelium of their own by extracting nutrients from the mycelium of the recipient pycnium. Later hyphae, which are predominantly uninucleate, grow out to the epidermis (either upper or lower), cut off 1 4 layers of buffer cells and then fuse by pairs to form two-legged cells (an observation also noted by Fromme, 1912). Allen suggests that this represents the sexual fusion event with one of the two fusion cells originating from the recipient pycnium and the other from a pycniospore derived from another pycnium. Each two-legged fusion cell forms a terminal binucleate spore initial which, by a series of further divisions, gives rise to dikaryotic aeciospores and intercalary cells alternately to form short chains. GENETIC CONTROL OF MATING TYPE Craigie (1927a,b) demonstrated that sunflower rust (Puccinia helianthi) and wheat stem rust (P. graminis) are heterothallic and proposed that the pycnia of these two rust species are of two mating types, which he designated (+) and ( ). Subsequently, it became widely accepted that a (+) and ( ) system, controlled by two alleles at a single locus, was a common feature of heterothallic rust fungi. Under such a system 50% of pair-wise crosses are expected to be compatible (produce aecia) irrespective of whether the pycnia being crossed come from the same or different strains. However, Lawrence (1980), working with flax rust, reported results inconsistent with this model. He found that nectar transfer, on a pair-wise basis, between pycnia of two unrelated strains, designated CH5 and I, resulted in aecia formation 98 out of 100 times, whereas nectar transfer between pairs of pycnia of the same strain gave aecia formation only 21 times out of 100 for strain CH5 and 35 times out of 100 for strain I. Assuming experimental error is low, the selfing data exclude a single locus model as both sets of data differ significantly from a 1 : 1 ratio; thus, the results of intercrossing strains CH5 and I could not be accounted for by postulating single-locus control with multiple alleles at that locus. The selfing data are also not consistent with a simple two-locus model, whereby an allelic difference at two unlinked loci is necessary for aeciospore formation: under such a system, 25% of selfings should produce aecia and the results of the I selfings differ significantly from this expectation. Noting that the results of selfing strain CH5 differ significantly from the results of selfing strain I (P < 0.05) and that Flor (1965) had also reported that the proportion of selfings that produce aecia can vary significantly between strains, Lawrence (1980) pointed out that any model that is proposed for the genetic control of mating type in M. lini must permit different proportions of selfings to be compatible. A model that meets this requirement is based on the incompatibility system in Schizophyllum commune, which, like M. lini, is a basidiomycete. In S. commune, the mating MOLECULAR PLANT PATHOLOGY (2007) 8(4), BLACKWELL PUBLISHING LTD

5 Melampsora lini, rust of flax and linseed 353 regardless of the second stage condition, and (iii) genetic compatibility for the first stage with genetic incompatibility for the second stage resulting in ring formation without aeciospore production. It might then be hypothesized that each stage is controlled by one of the two factors in the model described above. It is evident that further studies of the genetic control of mating type in flax rust (and other rust species?) are required. GENETICS OF THE FLAX/FLAX RUST INTERACTION The recognition of differences Fig. 2 The three outcomes observed following transfer of pycniospores from one pycnium to another (imaged 7 days after transfer). Left: the recipient pycnium continues to produce nectar. Centre: nectar production ceases, a ring of tissue develops but does not proceed to produce aeciospores. Right: a ring of aeciospores is produced. type of monokaryons is determined by two factors, both of which must differ for a mating to be compatible. Each factor is controlled by two linked loci, with an allelic difference between two monokaryons at either or both of the loci controlling each factor giving a different factor (see Kronstad and Staben, 1997). Under such a system, 25% of selfings are expected to be compatible if the parent strain is heterozygous at just one of the loci controlling each factor. However, if the parent strain in heterozygous at both loci controlling a factor, then the proportion of selfings expected to be compatible will be greater than 25%. The maximum proportion of compatible selfings will occur when the parent strain is heterozygous at all four loci controlling mating type, with the value depending on the amount of recombination between the two loci controlling each factor. While such a model can account for the data on mating type in M. lini, it must still be considered tentative, given the limited data on which it is based. With flax rust, three outcomes are observed following transfer of pycniospores from one pycnium to another: (i) no change is observed in the recipient pycnium and it continues to produce nectar, (ii) a few days after nectar transfer the recipient pycnium stops producing nectar and develops a ring or partial ring of orange tissue which subsequently breaks through the epidermis to produce aeciospores, or (iii) a few days after nectar transfer the recipient pycnium stops producing nectar and develops a ring or partial ring of tissue but this ring does not proceed to produce aeciospores. Examples of all three outcomes are shown in Fig. 2. The observation that there appear to be three outcomes from a sexual cross could indicate that there are two stages to full sexual compatibility in M. lini with (i) genetic compatibility for both stages leading to aeciospore production, (ii) genetic incompatibility for the first stage leading to continued production of nectar, Two strains of flax rust, A and B, may both be virulent on one flax line, both avirulent on a second line, have different reactions on a third line (A avirulent, B virulent) and have the reciprocal reactions (A virulent, B avirulent) on a fourth line. Reports in the early part of the twentieth century (for a review see Henry, 1926) first revealed that for flax and flax rust different stocks of one organism may react differently to a single stock of the other organism. Such observations indicated that the phenotype of growth or no growth of the rust that occurs when a particular rust strain is inoculated on to a particular host line depends not just on the genotype of the rust but also on the genotype of the host plant. A consequence of this interaction between genotypes is that (i) variation that may be present in a large collection of one of the interacting organisms will mostly go undetected if it is tested against a small collection of the other organism and (ii) a proper understanding of the genes in one organism that are involved in the interaction cannot be achieved without a knowledge of the genes that are involved in the other organism. It was an appreciation of these two points by Harold Flor that is largely responsible for our current extensive knowledge of the genetics of the interaction between flax and its rust as outlined below. Following earlier reports of variation in both flax and its rust, Flor made a systematic search for variability in both organisms (Flor, 1935, 1940, 1942a,b, 1946, 1956). He tested a large number of flax varieties with a wide range of rust cultures that were obtained initially from field collections, but later progeny from rust-breeding studies were used as well. By this means flax varieties were selected that were useful for differentiating rust cultures on the basis of their pathogenicity on these varieties. Rust isolates that differed in pathogenicity on one or more of these differential varieties were referred to as different physiologic races. By 1946, 16 varieties had been selected as differentials. Flor found, however, from host inheritance studies that a number of these varieties possessed two or more resistance genes. Thus, after 1946, he developed his differential series not only by the trial-and-error screening procedures which he had used up to this time, but also by isolating, through breeding studies, each of the resistance genes already present in his differential set into a 2007 BLACKWELL PUBLISHING LTD MOLECULAR PLANT PATHOLOGY (2007) 8(4),

6 354 G. J. LAWRENCE et al. separate line. By 1956, a series of 27 host differentials had been developed (Flor, 1954a,b, 1955, 1956) and 239 races identified. The reaction pattern and origin of each of these races, many of which were derived from inheritance studies, is given by Flor (1954a). Since 1956, five additional differentials have been identified (Flor and Comstock, 1972; Hoes and Kenaschuk, 1986; Wicks and Hammond, 1978; Zimmer and Comstock, 1973). Genetics of host reaction Henry (1930) and Myers (1937) were the first to report that rust reaction in flax is controlled by major genes, with resistance dominant to susceptibility. Two independently inherited loci were identified, which Myers named L and M. Subsequently, Flor (1947, 1955, 1956) used the varieties that he had found useful for differentiating physiological races as parents in inheritance studies. From these studies, which were very extensive, Flor obtained information about the number of resistance genes possessed by each differential variety and the linkage relationships of different resistance genes in different varieties. By 1955, 25 resistance genes had been identified in flax (see Flor, 1955) and, since then, the number has been extended to 31 (Flor and Comstock, 1972; Hoes and Kenaschuk, 1986; Islam and Mayo, 1990; Wicks and Hammond, 1978; Zimmer and Comstock, 1973). A compendium outlining the origin and history of each gene is available (Islam and Mayo, 1990). These genes occur in five series of closely linked or allelic genes at loci designated K, L, M, N and P, which contain two, 13, seven, three and six of the 31 resistance genes, respectively. The resistance determined by each of these genes has been expressed as a dominant character in all inheritance studies reported, although with a few genes dominance is incomplete against some rust strains (see Flor, 1956). Greater resistance has been consistently epistatic to lesser resistance. The five resistance loci show independent inheritance except for N and P, which are linked with about 10% recombination (Kerr, 1960; Shepherd, 1963). Many of the resistance genes have now been cloned and sequenced: the current list includes 11 genes at the L locus, one at the M locus, three at the N locus and two at the P locus (Anderson et al., 1997; Dodds et al., 2001a,b; Ellis et al., 1999; Lawrence et al., 1995). These genes all encode resistance proteins of the Toll Interleukin 1 Receptor Nucleotide Binding Site Leucine Rich Repeat (TIR-NBS-LRR) class, although the P locus proteins have an additional C-terminal domain of 150 amino acids downstream of the LRR region. Genetics of avirulence/virulence: the gene-for-gene relationship Flor (1942a) undertook the first study of the inheritance of avirulence/virulence in M. lini. He crossed race 6 race 24 to produce an F1 culture that was self-fertilized to produce a family of F2 progeny. The avirulence/virulence phenotype of each of these F2 cultures on each of the differential varieties in use at that time was then determined. The results indicated that major genes determined avirulence/virulence differences between races 6 and 24 with respect to a particular differential and that avirulence was dominant to virulence. With this finding it was evident to Flor that the phenotype of growth or no growth observed when a particular rust strain was inoculated on to a particular host line was the result of an interaction between the resistance/susceptibility genes in the host and the avirulence/ virulence genes in the pathogen. Flor (1942a) then proceeded to ask what are the relationships between the host and pathogen genes that interact. In particular he asked whether one gene in the pathogen could interact with several different resistance genes in the host or with only one gene in the host, his key question being: Is the ability of a virulent race to attack a number of flax varieties known to possess different factors for rust reaction due to one or to a number of factors each of which overcomes a specific resistance factor in the host? In the same paper Flor proposed an answer, The data... indicate that the range of pathogenicity of a physiologic race of Melampsora lini is determined by pathogenic factors specific for each resistance factor possessed by the host. This one-for-one relationship subsequently became known as the gene-for-gene relationship. The initial data that led Flor (1942a) to conclude that different resistance genes in the host interact with different avirulence/virulence factors in the pathogen came from the observation in his race 6 race 24 F2 family that single gene segregation ratios (consistent with three avirulent to one virulent) occurred on each of two differential varieties with single but different resistance genes and that the avirulence/virulence gene pairs determining the growth/no growth phenotype on each differential were different because they segregated independently of each other. Also consistent with a one-for-one relationship was the observation that two avirulence/virulence gene pairs in the race 6 race 24 hybrid controlled the growth/no growth phenotype on a variety that possessed two genes for resistance to race 6 (this variety was susceptible to race 24). Subsequently, Flor obtained more extensive data to support his gene-for-gene hypothesis. In 1946, he reported the results of testing 133 F2 progeny from the cross race 22 race 24 on 16 host differential varieties (Flor, 1946). Segregation for avirulence/ virulence occurred on 14 of the varieties: monohybrid segregation MOLECULAR PLANT PATHOLOGY (2007) 8(4), BLACKWELL PUBLISHING LTD

7 Melampsora lini, rust of flax and linseed 355 ratios (three avirulent : one virulent) occurred on each of nine host varieties possessing a single gene conferring resistance, dihybrid ratios (15 : 1) occurred on each of two host varieties possessing two genes conferring resistance and a trihybrid ratio (63 : 1) occurred on a host variety with three genes conferring resistance to the avirulent parent race. Avirulence was dominant in all cases except on the host variety Williston Brown: on this variety the F1 culture was virulent and the F2 progeny segregated 116 virulent to 17 avirulent. This result is discussed below. In a later study Flor (1955) tested 67 F2 progeny from the cross race 6 race 22 on an extended set of 32 differential varieties, most of which possessed, as far as Flor could determine, only single genes for resistance. Segregation for avirulence/virulence occurred on 24 of the 32 differentials and all segregations agreed with a monohybrid except for one dihybrid segregation on a variety known to possess two resistance genes. The majority of these segregating pairs of avirulence/virulence genes were immediately shown to be different from each other because they segregated independently of the other pairs, or with only an incomplete association. However, some groups of two or more varieties were found in which the segregation on one variety within the group was completely associated with the segregation on the other variety(ies). Flor accounted for some of these unit segregations, where no genetics of resistance in the host lines had been carried out, by assuming that the varieties within the group possessed a common resistance gene. In the remaining instances of unit segregation Flor argued that because other rust strains occur having contrasting avirulence/virulence phenotypes on the host varieties within the group, these varieties must possess different genes controlling rust reaction and there must be separate avirulence/virulence genes interacting with them in the rust. He therefore accounted for the unit segregation observed on these varieties by postulating that the avirulence/virulence genes corresponding to the resistance genes possessed by these varieties were sufficiently closely linked that no recombinant progeny occurred among the 67 that he tested. The three studies by Flor outlined above, together with subsequent studies by Flor and others (Flor, 1959, 1960, 1965; Lawrence et al., 1981a; Shepherd, 1963; Statler and Zimmer, 1976; Statler, 1979, 1990), on the inheritance of avirulence/virulence have resulted in monohybrid ratios indicating a dominance (or occasionally partial dominance) of avirulence being obtained, on at least one occasion, on differential varieties possessing 27 of the 31 characterized resistance genes and these data do not provide any evidence, with one proviso, that contradicts the assumption that the 27 pathogen genes segregating in these studies are different genes. The proviso concerns those instances where two or more avirulence specificities have shown unit inheritance on two or more differential varieties with single but different resistance genes. Here it is possible that the avirulence specificities could be determined by a single gene with multiple specificities, with varying forms of this gene determining zero, one, two or more avirulence specificities in various combinations so as to account for the different patterns of avirulence and virulence that different rust strains have on the differential varieties carrying the corresponding resistance genes. To indicate the specificity of the interaction between genes in the host and genes in the pathogen suggested by his studies, Flor (1955) proposed that the notation of a particular pathogen gene include the resistance gene with which it interacts. Thus, for example, the avirulence gene interacting with resistance gene L2 is written as A-L2 and the corresponding virulence allele as a-l2. Recently, however, the prefix Avr has been used in place of A, e.g. AvrL2. A plant may or may not possess a particular resistance gene. An attacking rust strain may or may not possess the corresponding avirulence gene. Of the four possible combinations of host and pathogen genotypes, only one gives a no-growth (plant resistant, rust avirulent) phenotype, namely plant with resistant gene and pathogen with avirulence gene. Thus, the specificity of the interaction occurs between the dominant, or co-dominant, genes in each organism, i.e. between genes for resistance in the host and genes for avirulence in the pathogen. Thus, the gene-forgene relationship can be stated, perhaps most precisely, as For each gene conferring resistance in the host there is a separate and specific gene conferring avirulence in the pathogen. This statement is not meant to imply that different genes within an organism are each located at a separate locus. Two different genes may or may not be allelic. If they are allelic then the separate and specific characteristic is conferred by different alterations within the gene. If a single pathogen gene can simultaneously confer two or more avirulence specificities, a possibility noted above, then this might be considered an exception to the gene-for-gene relationship. Linkage relationships between avirulence genes Whereas the resistance genes in flax are located at only five complex loci, the corresponding avirulence genes are more dispersed around the genome. Only four small groups of avirulence genes have been identified where the genes within each group are apparently sufficiently tightly linked to be usually inherited as a unit. These groups are A-L3, A-L4, A-L10 A-L5, A-L6, A-L7 A-M1, A-M4 A-P, A-P1, A-P2, A-P3 (Flor, 1946, 1955, 1959, 1965; Lawrence, 1988 for review; Lawrence et al., 1981a; Shepherd, 1963; Statler, 1990). The F2 and test-cross families in these studies were all quite small, being less 2007 BLACKWELL PUBLISHING LTD MOLECULAR PLANT PATHOLOGY (2007) 8(4),

8 356 G. J. LAWRENCE et al. than 100 in all cases except that of Flor (1946) where 132 F2 individuals were tested. However, there has been one detailed study. Lawrence et al. (1981b) examined 3160 progeny from a test-cross involving a parent heterozygous for A-P, A-P1, A-P2, A-P3; several individuals recombinant for avirulence specificities were obtained giving an estimated recombination frequency within the region of these genes (or within the gene) of 0.19% (about one in 500). A comment on the unit inheritance of A-L3, A-L4 and A-L10 is warranted. The cloning of the L3, L4 and L10 resistance genes (Ellis et al., 1999) revealed that the protein products of L3 and L10 are identical and differ from that of L4 by only one amino acid. Consequently it is likely that there is only a single avirulence gene determining rust reaction on the L3, L4 and L10 differential lines so that the apparent close linkage of A-L3, A- L4 and A-L10 is probably spurious. In addition to the cases of tight linkage of avirulence genes listed above, a number of cases of incomplete linkage between avirulence genes have also been reported: these are reviewed and discussed in Lawrence (1988). Inhibitor genes The gene-for-gene hypothesis, as outlined above, does not provide a complete description of the interaction between genes in the host and genes in the pathogen. Data on the inheritance of avirulence/virulence on host differentials possessing the M1, L1, L7, L8 and L10 resistance genes suggest that two allelic gene pairs, the corresponding avirulence/virulence gene pair and an inhibitor gene pair, interact to determine the growth/no growth phenotype on these differentials. Flor (1946) reported an atypical segregation for avirulence/ virulence on the variety Williston Brown, which possesses the M1 gene for resistance. He found that F2 cultures from the cross race 22 race 24 segregated in the ratio 17 avirulent to 116 virulent. Both parent races and the F1 were virulent on Williston Brown, as were nine selfed progeny of race 22 and 15 selfed progeny of race 24. An explanation that accounts for these observations (Shepherd, 1963) postulates that a dominant inhibitor gene, I-M1, present in one of the parents, interacts with a dominant gene normally controlling avirulence on Williston Brown, A-M1, to give a virulent pathogen phenotype. Under this model there are four basic genotypes, and these have the following phenotypes: Genotype Phenotype on Williston Brown I-M1/ ; A-M1/ Virulent I-M1/ ; a-m1/a-m1 Virulent i-m1/i-m1; A-M1/ Avirulent i-m1/i-m1; a-m1/a-m1 Virulent Flor s data can be accounted for if it is assumed that races 22 and 24 have genotypes I-M1/I-M1; A-M1/A-M1 and i-m1/i-m1; a-m1/a-m1, respectively; in that case these races, their selfed progeny and the F1 hybrid (I-M1/i-M1; A-M1/a-M1) would be virulent on Williston Brown as was observed. Furthermore, a 3 : 13 ratio of avirulent to virulent F2 individuals would be expected (assuming no linkage) and this was observed (P = ). This model received support from a later study (Flor, 1965) in which progeny obtained from self-fertilizing a strain avirulent on Williston Brown segregated 61 avirulent to 29 virulent. This result fits a monohybrid ratio with avirulence dominant to virulence, thus indicating that an A-M1/a-M1 gene pair does occur in the pathogen as predicted by the model. Evidence for additional inhibitor genes comes from work by Lawrence et al. (1981a) and Jones (1988a). They tested progeny obtained by self-fertilizing and intercrossing two strains of rust, designated CH5 and I. Atypical segregations for avirulence/ virulence occurred on Williston Brown (M1) in all three families consistent with both parent strains having the genotype I-M1/i-M1; A-M1/a-M1. However, similar atypical segregations for avirulence/ virulence also occurred on host differential lines possessing the L1, L7, L8 and L10 resistance genes in the CH5-selfed and/or CH5 I families. These segregations were accounted for by postulating that inhibitor genes are also involved in determining avirulence/virulence on the L1, L7, L8 and L10 resistance genes. The segregations that occurred on the M1, L1, L7, L8 and L10 differentials in the CH5-selfed and CH5 I families were not independent of each other. These significant associations are consistent with the five inhibitor genes in strain CH5 being sufficiently closely linked in coupling phase I-M1 I-L1 I-L7 I-L8 I- L10/i-M1 i-l1 i-l7 i-l8 i-l10 (order arbitrary) that no recombination had occurred between them. This suggests the possibility that all five inhibitor specificities may be determined by a single gene. This gene, or tightly linked cluster of genes, was subsequently shown to be allelic or tightly linked to the gene that inhibits avirulence only on Williston Brown (M1) (Jones, 1988b). Thus, M. lini may possess an inhibitor locus with three alleles: one that is multispecific, inhibiting A-M1, A-L1, A-L7, A-L8 and A-L10, one that is mono-specific, inhibiting only A-M1, and a null allele that inhibits no avirulence genes. The product of an inhibitor gene could function in the pathogen by, for example, inhibiting transcription of the avirulence gene, modifying the avirulence gene product or inhibiting its secretion. Alternatively, if the product of the inhibitor gene is secreted and enters the plant cell, then it could exert its effect by inhibiting the resistance mechanism in the plant. If this latter possibility were to be the case, then cloned inhibitor genes and their protein products could be useful tools for investigating the molecular basis of plant resistance. Avirulence gene expression in the monokaryon Two studies (Flor, 1959; Statler and Gold, 1980) have been undertaken to determine whether the genes that confer avirulence MOLECULAR PLANT PATHOLOGY (2007) 8(4), BLACKWELL PUBLISHING LTD

9 Melampsora lini, rust of flax and linseed 357 in the dikaryon also function to prevent growth of the monokaryon on host lines possessing the corresponding genes for resistance. Between them, these studies tested avirulence expression in the monokaryon on 25 of the 31 resistance genes. With two exceptions the tests indicated that the avirulence genes do function in the monokaryon. The exceptions involved host lines possessing resistance genes N and P4 on which it was found that basidiospores carrying the corresponding avirulence genes could grow and establish pycnial infections; in these two cases transfer of nectar between pycnia on the same host gave rise to aeciospores that were avirulent on that host. So far, no tests have been undertaken to determine whether the inhibitor genes function in the monokaryon. Another question that can be asked is whether a monokaryotic pycnium growing on a host plant with a resistance gene can be fertilized by pycniospores possessing the corresponding avirulence gene. This has been little investigated, but in one study Flor (1941) reported that a pycnium from race 22 growing on the flax variety Ottawa 770B (race 22 is virulent on Ottawa 770B) was successfully fertilized with pycniospores from race 6 (which is avirulent on Ottawa 770B). Ottawa 770B was subsequently shown to possess two resistance genes, L and P5 (see Islam and Mayo, 1990), and as an F2 family from a race 6 race 22 hybrid subsequently gave a dihybrid segregation ratio on Ottawa 770B (Flor, 1946, 1955) the A-L and A-P5 avirulence genes may both have been present in the pycniospore(s) that fertilized the pycnium on Ottawa 770B. CLONED AVIRULENCE/VIRULENCE GENES Currently, M. lini avirulence/virulence genes have been cloned from four loci. Two of these loci contain avirulence genes that interact with several different resistance genes, namely L5, L6, L7 and P, P1, P2, P3. The other two loci contain avirulence genes that interact with just one resistance gene, namely M and P4 (Catanzariti et al., 2006; Dodds et al., 2004). The approach used was initially to preselect (by one of two means, see below) a batch of M. lini genes (as represented by cdna clones) that could contain one or more avirulence genes. Each gene in the batch was then used as a probe to detect restriction fragment length polymorphisms (RFLPs) segregating in a flax rust F2 mapping family in which 16 avirulence genes segregate. Probes whose RFLPs co-segregated with an avirulence gene(s) in the mapping family became candidate avirulence genes. Genomic clones of candidate avirulence genes were then isolated and, in the absence of a method to transform M. lini, were tested for avirulence function by transient expression in flax lines with different rust resistance genotypes. Transient assays involved infiltration of flax leaves with Agrobacterium strains capable of delivering into host cells the candidate Avr gene expressed with a plant promoter. Avirulence function was demonstrated if a hypersensitive reaction developed only in a line or lines containing the resistance gene(s) corresponding to the candidate s possible avirulence specifity(ies) as identified in the mapping analysis. Initially, one cdna probe amongst a batch of cdnas preselected by suppressive subtractive hybridization to identify rust genes expressed during infection was found to co-segregate with the AvrL5, AvrL6 and AvrL7 cluster of avirulence genes. This locus was renamed AvrL567 when it was found that a single gene at this locus could interact with each of the L5, L6 and L7 resistance genes (Dodds et al., 2004). Avirulence genes corresponding to the M, P4 and P, P1, P2, P3 resistance genes were initially identified amongst a batch of cdna clones that had undergone two preselection steps (Catanzariti et al., 2006). On the expectation that avirulence genes are likely to be expressed in haustoria and their products secreted, the initial preselection step involved isolating haustoria, then making a cdna library from RNA extracted from the haustoria. Sequencing of 822 cdna clones from this library identified 429 unique sequences that were not rrna sequences. The second preselection step involved screening the sequences to identify those encoding secreted proteins based on the presence of a predicted signal peptide. Amongst 20 such candidates, one co-segregated with AvrM, another with AvrP4 and a third with the AvrP, AvrP1, AvrP2, AvrP3 cluster of avirulence genes. Transient Agrobacterium expression assays in flax lines confirmed resistance gene-specific avirulence function in all cases. With respect to avirulence on the P, P1, P2 and P3 resistance genes, one allele in strain CH5, the parent of the mapping family, conferred avirulence on P alone, whereas the other conferred avirulence on P1, P2 and P3. These alleles were therefore named AvrP and AvrP123, respectively. The specific characteristics of the cloned avirulence/virulence genes and their protein products are summarized in Table 1. Several features of these genes and their products are worthy of note. When the genes are transiently expressed without their signal peptides inside plant cells, strong resistance-gene-dependent necrotic responses are observed, implying that recognition of these avirulence proteins occurs inside plant cells. It is not known how these proteins, once secreted from the haustorium, enter the plant cells during infection, although this now appears to be a common feature of several fungal and oomycete avirulence proteins (Catanzariti et al., 2007). In the case of the AvrL567 avirulence/virulence proteins, yeast two-hybrid assays indicate that recognition is based on direct resistance avirulence protein interaction, with a close correspondence observed between the detection of a protein interaction in yeast and the induction of a hypersensitive response during transient expression in planta (Dodds et al., 2006). All the cloned avirulence/virulence genes have two features in common. Their protein products are small (see Table 1) and they each contain two or more introns: thus, AvrL567 and AvrP4 each 2007 BLACKWELL PUBLISHING LTD MOLECULAR PLANT PATHOLOGY (2007) 8(4),

10 358 G. J. LAWRENCE et al. Table 1 Characteristics of cloned avirulence/virulence genes and their protein products from flax rust, Melampsora lini. Locus Rust strains screened Gene copies/ haplotype Variant forms of avir/vir genes Predicted product sizes (no. of amino acids) Pre-protein Mature protein Features of protein product AvrL567 C, H, I, J,Bs1, Fi Various 1, 2, 3, 4 12 (7 avr, 5 vir) Novel no database matches Highly variable (25% of amino acids have one or more polymorphisms) Expressed in haustoria, but not in germ tubes Recognized inside plant cells Interact directly with R-gene protein AvrM CH5 Various 1, 5(+?) 6 (4 avr, 2 vir) (vir) Novel no database matches (vir) Expressed in germ tubes and haustoria (avir) No cysteine residues (avir) Recognized inside plant cells (avir) AvrP4 CH5, WA 1 3 (2 avr, 1 vir) Novel no database matches Recognized inside plant cells Six cysteine residues potential to form cysteine knot structure Polymorphisms occur in C-terminal 22 amino acids between conserved cysteine residues AvrP CH5 2 (1 a pseudo-gene) AvrP = Novel no database matches AvrP123 AvrP123 = cysteine residues conform to consensus spacing of Kazal family of protease inhibitors Highly polymorphic 36 amino acid differences between AvrP and AvrP123 contain two small introns and AvrP/AvrP123 three small introns in the untranslated leader region of their mrna products and AvrM contains three small introns in the coding region. This observation may indicate that introns are required for efficient gene expression in M. lini. For the AvrL567 locus, an examination of six rust strains identified haplotypes containing one, two, three or four copies of the avirulence/virulence genes, which include 12 gene variants (Dodds et al., 2006). Seven of these elicited in planta recognition specificity when tested on L5, L6 and L7 host lines, but the recognition specificity varies; one is recognized by L5 and L6 but not L7, another is recognized by L6 and L7 but not L5 while several are recognized by all three resistance genes. Thus, a single avirulence protein can be recognized by the products of three different resistance genes and variant forms of the avirulence protein can be recognized by a subset of the three resistance proteins. Similarly, a single avirulence protein from the AvrP123 locus can be recognized by the products of three different resistance genes (P1, P2 and P3). These examples could be considered exceptions to the gene-for-gene relationship. Typically, variant forms of avirulence/virulence genes at a locus have many polymorphic nucleotides which contain a significant excess of non-synonymous over synonymous changes, indicating that diversifying selection has acted on these genes (Catanzariti et al., 2006; Dodds et al., 2004). In contrast, regions flanking the genes are frequently extremely uniform, showing little polymorphism (P. N. Dodds, unpublished data). Uniform flanking regions may promote a high frequency of recombination in the vicinity of avirulence/virulence genes, which could have evolutionary benefits to the rust. If so, selection would act to promote and conserve uniformity in the flanking regions. The cloning and sequencing of M. lini avirulence/virulence genes, along with the sequencing of 822 cdna clones from a haustorial library as part of this work, represents most of the molecular work carried out with M. lini. The remaining work consists of a population study that employed nine microsatellite markers (Barrett et al., 2007; see below) and the isolation and characterization of a β-tubulin gene from M. lini (Ayliffe et al., 2001). The cloned avirulence genes interact with nine different resistance genes. Six of these resistance genes, L5, L6, L7, M, P, P2, have been cloned. Thus, the molecular basis of Flor s genefor-gene relationship has now been elucidated for six resistance gene/avirulence gene combinations. DOUBLE-STRANDED RNA MOLECULES Double-stranded (ds) RNA molecules are apparently common in rust fungi, given that they have been found in most rust species MOLECULAR PLANT PATHOLOGY (2007) 8(4), BLACKWELL PUBLISHING LTD